# BACKLOG — the single canonical open-items list > ## ▶ OPEN WORK — the actionable list (rewritten 2026-08-04) > > The table further down is now **mostly a graveyard**: of its ~19 rows, ten are struck > through and nine of those were discharged on 2026-08-03. It is kept as the construction > record. **This is the live list.** It is ordered so that the top item is the one that > unblocks the programme's single remaining scientific blocker. > > ### ▶ NEXT STEPS — recorded 2026-09-07 > > **State.** HEAD `46be6c6`, tree clean, CI green (`CI` and `Glossary` both pass at HEAD). > ⚠️ *State update 2026-09-10 (local HEAD, ten commits ahead of `46be6c6`'s successors, not pushed):* the > sentence "there is no queued Lean brick left" below is superseded — > [`generator-layer-scoping.md`](generator-layer-scoping.md) **§9** holds nine priced rows (G5, G14b, > G15–G19, W1) under the author's policy that nothing is "unscheduled"; **W1 landed the same day** > (`LF4/SectorManifold.lean`: the `ℂℙⁿ` sectors are the standard Kähler/Liouville objects, link L1 > discharged). The four lists below are otherwise as recorded. ⚠️ *2026-09-11:* the G series is > complete except G5/G17b/G18 (§9); its lay summary is [`generator-layer-scoping.md`](generator-layer-scoping.md) > §10; the doc scan of the same day (every prose surface re-read against it) is recorded in the > **▶ OUTSTANDING** block below, which also re-lists the last "not scheduled" row (`R-019`) with a price. > > ### ▶ OUTSTANDING — the priced residue, all lists (recorded 2026-09-11) > > Every open item that touches the manifold layer or was ever marked "not scheduled" or > "permanently scoped", as a numbered priced row (repository ids `Q29`–`Q33`; the scoping note's > `G` ids are given in brackets — `G` is a paper prefix, CONVENTIONS §12). The author decides; > nothing here is "unscheduled". Mathlib landings deprecate the corpus's version. > ⚠️ *2026-09-11 re-plan:* `reconstruction-status.md` §2a (i) "permanently scoped — blocked on > Mathlib" held three items. After the G series: A1 is **discharged** on `ℂℙⁿ`; A3 (smoothness of > `π`) and the Hamiltonian generation of the witnesses are **re-classified as bricks** — `Q33` and > `R-016′` here — because the API they waited for now exists in the corpus. The 48 "Mathlib has > no …" claims in module docstrings were all re-read the same day: 20 were about the manifold > API and are rewritten to point at the theorems or at `R-016′`; the rest are tagged with > pin-checked sentinels (`check-mathlib-absence`: 80 sentinels, 0 untagged). > > | # | Item | Cx | P | Value | Where / what it lands | > |---|---|---|---|---|---| > | ~~**Q29**~~ (= G5 in the scoping note) **DONE 2026-09-12** (all five milestones) — **milestone (a) DONE 2026-09-11** (12 pins; `integralFlow` with `integralFlow_add` on any compact manifold, `IsSymplectic.exists_isMIntegralCurve_hamiltonianVectorField`; `IntegralCurve/GlobalFlow.lean`); **(a′) DONE 2026-09-12** (8 pins; `continuous_integralFlow`, joint continuity of the flow, each `integralFlow hv t` a homeomorphism; `IntegralCurve/FlowContinuity.lean`); **(b′) DONE 2026-09-12** (7 pins; `Mathlib/Analysis/ODE/FlowDerivative.lean`: differentiable dependence of the local flow on the initial point with the variational equation, the Mathlib gap filled at `C¹`; flat Liouville `ContDiffAt.exists_localFlow_form_invariant`); **(c′) DONE 2026-09-12** (10 pins; `Geometry/Manifold/HamiltonianLieDerivative.lean`: Cartan's formula on a normed space, and the flat Lie derivative of a symplectic form's local representative along its local Hamiltonian vector vanishes on the chart's target); **(d′) DONE 2026-09-12** (12 pins; `Geometry/Manifold/HamiltonianFlowVolume.lean`, `Instances/ProjectiveSpaceHamiltonianFlow.lean`: ★★★ `IsSymplectic.map_hamiltonianFlow_topFormMeasure_wedgePow` — on a compact symplectic manifold the Hamiltonian flow of every `C^∞` energy preserves the measure of every power of the form, at every time; ★★★ `Projectivization.fsVolume_map_hamiltonianFlow` on `ℂℙⁿ`); **(e) DONE 2026-09-12** (8 pins; `hamiltonianFlow_schrodingerHamiltonian`, `hamiltonianFlow_torusHamiltonian`: the Schrödinger and torus flows ARE the Hamiltonian flows; `manyToOneSchrodingerSetup_flow_preserves_volume_derived`: the sector's posited `flow_preserves_volume` derived from Liouville's theorem, `LF4/SectorManifold.lean`) | Liouville at manifold level: the time-`t` map of a Hamiltonian flow preserves `ω^{∧n}`. Milestones (a) global flows on a compact manifold from Mathlib's uniform-time integral curves (M–L); (b) the Lie derivative of a manifold form + Cartan's formula — the genuine Mathlib gap (L–XL); (c) assembly with G11's `d(ι_X ω) = 0` (M). | **XL** | Low–med | Medium | `generator-layer-scoping.md` §9. Would replace `flow_preserves_volume` by a theorem for the globally Hamiltonian pieces only; Posit 3 stands regardless. | > | ~~**Q30**~~ (= G17b) **DONE 2026-09-11** (14 pins; `MetricFamily.riemannianVolume_congr_cover`, `Projectivization.riemannianVolume_fsMetric_congr_cover`) | Chart-independence of the Gram-density Riemannian volume (`√det (DφᵀGDφ) = |det Dφ| √det G` along a transition, then `chartMeasure_congr` for metrics). | **S–M** | High | Low | Makes `MetricFamily.riemannianVolume` canonical for a metric not identified chart by chart with a top form; on `ℂℙⁿ` it is inherited. | > | ~~**Q31** (= G18)~~ | ~~Darboux's theorem (Moser's trick: needs G5(a)(b) and a Poincaré lemma on a ball — the Poincaré lemma and time-dependent flows landed 2026-09-18, ▶ OPEN QUEUE #8).~~ **DONE 2026-09-21** ([`Mathlib/Geometry/Manifold/Darboux.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Darboux.lean)): Darboux in Moser's form — the constant form `ω(x₀)` in a differentiable chart with continuous inverse; residues #49 (`C¹` chart) and #50 (standard form). | ~~**XL**~~ ~~**M**~~ done (M) | Low | Low | The classical theorem a symplectic library owes; no consumer. | > | ~~**BC**~~ (the BCFJS `iff`; the scope line of `Empirical/QM/NoBroadcasting.lean`) **DONE 2026-09-13** (all six bricks; ★★★ `QuantumInfo.exists_channel_broadcasts_iff_commute` — two positive semidefinite matrices can be broadcast by a single channel iff they commute; the hard half `Channel.Broadcasts.mul_comm_of_posSemidef` by an elementary route, no fidelity, no relative entropy; **the finite-family version** `exists_channel_broadcasts_family_iff_pairwise_commute` too, 3 pins, 2026-09-13) — **BC1, BC2 DONE 2026-09-12** (18 pins; Cat-1 `Mathlib/QuantumInfo/Broadcasting.lean`: `Channel.Broadcasts`, linear in the state; ★ **support confinement** `Broadcasts.kronecker_mul_kraus_mul`, `(P ⊗ P) Kᵢ ρ = Kᵢ ρ` for a projector `P` with `P ρ = ρ`; ★★ `exists_channel_broadcasts_of_commute` — **commuting Hermitian matrices are broadcast** by the classical copier in a joint eigenbasis (Mathlib's `directSum_isInternal_of_commute`, restricted to the finitely many eigenvalue pairs); ★★ `Broadcasts.star_dotProduct_eq_zero_or_norm_eq_one` — **two broadcast pure states are orthogonal or parallel**, no-cloning at channel level, the rank-one case of the hard half); **BC3 DONE 2026-09-13** (15 pins; ★★ `Broadcasts.mul_eq_zero_of_range_disjoint` — broadcast states with disjoint supports have orthogonal supports, `B A = 0`: `suppProj` (the range projector via `starProjection`), `IsHermitian.exists_top_eigenvalue`, the tensor bound `re_star_dotProduct_kronecker_mulVec_le`, and the `μ ≤ μ √μ` squeeze; priced M–L, took M); **BC4, BC5 DONE 2026-09-13** (17 pins; ★★ `Broadcasts.block_split` — a subspace `V` with `Kᵢ V ⊆ V ⊗ V` inside the support of a broadcast `τ` gives `P_V τ P_W = 0` and `Φ` broadcasts `P_V τ P_V` and `P_W τ P_W`, `W = supp τ ⊖ V`, via `kronecker_mulVec_kraus_mulVec_sub` (the complement is confined too: the dual `Φ† (P_V ⊗ 1)` fixes `V`, its excess over `P_V` is PSD with zero trace against `τ`) and the sandwich `kraus_block_sandwich` with the partial-trace identity `Tr_B ((A ⊗ B) Y (C ⊗ D)) = A Tr_B (Y (1 ⊗ D B)) C`; ★★ `exists_boundary_point` — for PSD `ρ ≠ σ` of trace one, some `l ≥ 1` has `σ + l (ρ − σ)` PSD with a kernel vector outside `ker (ρ + σ)`: the admissible `l` form a closed bounded set, and at its supremum a missing kernel vector would let the eigenvalue-`0` eigenvectors (in `ker (ρ + σ)`, hence `ker (ρ − σ)`) and the positive eigenvalues absorb a further step; each priced M, took M); **BC6 DONE 2026-09-13** (11 pins; the rank induction: `interProj` — the projector onto the intersection of two ranges, `1 − suppProj ((1 − P₁) + (1 − P₂))` — with `interProj_kronecker_mulVec_eq_self` (intersections of tensor squares, column by column), `rank_lt_rank_of_ker` (a strict kernel inclusion is a strict rank inequality, rank–nullity), and `Matrix.PosSemidef.exists_smul_trace_one`; with `V = S₁ ∩ S₂` the boundary supports, `V = 0` is BC3 and `V ≠ 0` splits both boundary states by BC4 into block pairs whose rank of the sum is strictly smaller — witnessed by the BC5 kernel vector and by a vector of `V` — so they commute by induction, and the cross products vanish; same channel, same `N`, no compression; priced M, took M–L) | Broadcast ⇒ commute for mixed states; with BC1 the `iff`. **Landed.** | **L–XL** as re-priced on 2026-09-12; took L–XL over 2026-09-12/13 (six bricks, 79 pins) | — | — | The prize named in `lieb-dpi-scoping.md`, delivered independently of SSA/DPI and of the bridge. | > | ~~`R-016′`~~ **DONE 2026-09-14** (`LF4/ArenaSymplectic.lean` + four Category-1 modules, 19 pins; L took L) | ~~The ARENA-level `ι_X ω = dH` for the joint-arena propagators of the record layer~~ **The arena `KSigma (N+1) = ℂℙᴺ × T²` is a symplectic manifold** (★★★ `arenaForm_isSymplectic`, the form `π₁^* ω_FS + π₂^* (dθ₁ ∧ dθ₂)` over the normed model `(Fin N → ℂ) × (ℝ × ℝ)`); ★★ Liouville on it for every smooth `H` (`arenaVolume_map_hamiltonianFlow`); ★★ `ι_X ω = dH` on the arena for sector energies with `X_{H∘π₁} = (X_H, 0)`; ★★★ **the isolated ontic dynamics IS a Hamiltonian flow of the symplectic arena** (`hamiltonianFlow_sectorEnergy_schrodinger`: the Schrödinger unitary with the torus fixed is the arena's flow of `−2⟨H⟩ ∘ π₁`, by integral-curve uniqueness on the product); ★★ the torus stroke `(0, a)` is locally Hamiltonian on the arena. Category-1 pieces: `Instances/AddCircleTranslation.lean` (the translation atlas on `AddCircle T` over `ℝ`, transitions locally translations), `ProductSelfModel.lean` (a product of manifolds charted on the product NORMED space, with the transition derivative as a block map), `TranslationAtlasForm.lean` (constant forms on a translation atlas; ★★ the torus is symplectic), `ProductForm.lean` (`π₁^* α + π₂^* β`, its local representatives, `d` of it, ★★ `IsSymplectic.prodForm`). **Not delivered, re-priced:** ~~the flux obstruction at manifold level (#28)~~ (DONE same day, `LF4/ArenaStrokeFlux.lean`) and ~~the identification `arenaVolume = c · μ_FS ⊗ vol` (#29)~~ (DONE same day, `LF4/ArenaVolume.lean`; the constant's value is #32); the joint lift on the arena is `R-016` itself, paper-side. | ~~**L**~~ | — | — | The "Hamiltonian origin stated, not formalised" residue of `record-layer-plan.md` and `TOUR.md` is now a proved theorem for the isolated dynamics and a stated theorem-shape for the strokes. | > | `R-019` (existing row) | The relaxation H-theorem for the fibre (Track B / Valentini): ~~blocked on first-passage asymptotics for the hyperbolic fibre~~ **the convergence clause is proved at step `n` in `klDiv` for the hyperbolic witness and grid-coarse-grained densities (2026-10-14, ~~#118~~; first-passage asymptotics were never needed)**; `relaxation_requires_hyperbolic_fibre` (2026-09-06) is the proved precondition; **and for every `L²` initial density with no rate (2026-10-14, ~~#147~~, the cat map strongly mixing on `L²(T²)`)**; open on the map identification (Posit 1, #19) **alone**. | ~~**XL**~~ convergence **L** + **M–L** (taken) | ~~Low~~ Medium | Highest *outside* the programme | `cr-queue.md` CR-14, `residues.tsv`. The only route to new predictions past the empirically-identical ceiling. Re-listed here with a price (was "not scheduled"). | > | ~~**Q33**~~ **DONE 2026-09-11** (11 pins; `AddCircle.instIsManifold` via transport along a homeomorphism, `CSD.LF4.ksigma_isManifold`, `manyToOneSetup_pi_contMDiff`; A3 discharged) | **Smoothness of the projection `π = Prod.fst : ℂℙⁿ × T² → ℂℙⁿ`**: a `ChartedSpace` + `IsManifold` instance on `AddCircle 1` (transport along `AddCircle 1 ≃ Circle`, whose instance Mathlib has in `Instances/Sphere.lean`), hence on `KTorus` and the product arena; then `ContMDiff π` is `contMDiff_fst`. | **S–M** | High | Low–medium | Discharges the last "needs the absent API" excuse in §2a; a prerequisite of `R-016′` (the arena as a manifold). Nothing downstream consumes smoothness of `π` today. | > | ~~**Q32**~~ **DONE 2026-09-11** | Glossary staleness sweep: 11 entries flagged STALE, 1 never reviewed, `meta.sha` re-pinned. Each re-read against its module's diff since `reviewed:`; **two needed content** (`fibre`: the base measure is now derived from the volume form, W1; `schrodinger-from-records`: the continuity-only form exists), ten were docstring-only changes and got the date. Guard: 0 STALE, 0 never reviewed. | **S** | — | — | `docs/glossary.yaml` | > | ~~**Q34**~~ **DONE 2026-09-29** | **The glossary caught up with the quantum-information tranche** (author request: "lots more qit names added recently"). Three defects, all fixed: (i) `scripts/extract-eponyms.sh`'s curated `EPONYMS` list was itself stale, so the mechanical key-set check could not see the new names — thirty-one surnames added, section (A) now reports 97 eponyms present and section (B) carries no real name; (ii) **nineteen entries written** for concepts the corpus proves and the glossary never named — `knill-laflamme`, `stabiliser-code`, `transversal-gate`, `threshold-theorem`, `concatenated-code`, `magic-state-distillation`, `clifford-t`, `two-level-unitary`, `toffoli-gate`, `reversible-adder`, `helstrom-bound`, `geometric-phase`, `aharonov-bohm`, `wigner-function`, `koopman-operator`, `nelson-stochastic-mechanics`, `darboux-theorem`, `feynman-kac`, `way-theorem` — each anchored at a live theorem with the backlink added to its module (151 entries); (iii) the twenty STALE entries re-read against their modules, with **four needing content**: `steane-code` (the recovery map, the Knill–Laflamme conditions and the fault-tolerance claims it declared absent are all proved now), `magic-state` (injection, distillation and Clifford+T universality likewise), `gleason-theorem` (the projection-valued theorem of #57, which was that row's last residue), `phase-estimation` (the two-index Dirichlet bound of #12); the rest were the `fsMeasure*` rename and an index-type generalisation. Also six `eponyms:` lists that had lost a surname (Gerlach, Vaidman, Mandel, Garg, Zehnder, Jozsa), `Kraus` on `quantum-channel`, and `meta.sha` re-pinned. Guard: 0 hard findings, staleness 0, generator builds 151 pages. | **M** | — | — | `docs/glossary.yaml`, `scripts/extract-eponyms.sh`, `GLOSSARY.md` | > | ~~**Q35**~~ **DONE 2026-09-29** | **Every glossary page now has an onward path, and the new ones have faces.** (i) `site_link` added to all 127 entries that lacked one, read off the site's own sitemap and each destination opened first: `/papers` (Paper C the reconstruction, Paper D geometric QM, C1 the Bell formalisation, C2 the PBR paper), `/primer-1`–`/primer-5`, `/position` (§6 places the programme against the standard impossibility theorems), `/roadmap` (stage 3 reconstruction, stage 5 continuum limits, stage 7 field theory), `/current-claims`, `/current-limits` (§3 finite-dimensional), `/technical-notes` (TN0 notation, TN4 mixed states and POVMs, TN6 sequential measurement). **check-glossary is at 0 warnings for the first time** (152 entries, 0 hard findings, 0 warnings). (ii) Images: the fetch tool's candidate map extended to the new tranche and every candidate LOOKED AT — ten accepted (Laflamme, Dorit Aharonov, Clifford, Berry with the Lorentz medal, Yakir Aharonov, the cat-state Wigner plot with its negative fringes, Darboux, Kac, Araki, Renner), ten with no usable free image and left without one; the three the 2026-09-14 pass had deliberately rejected (Bayes, Stone's letter, Wick's diagram) stay rejected. 77 of 152 pages now carry an image. | **M** | — | — | `docs/glossary.yaml`, `docs/img/`, `scripts/lib/glossary_images.py` | > | ~~*(push)*~~ **DONE 2026-09-12** | ~~Fourteen local commits (G9+G10 … G17) are unpushed; CI has not seen the manifold layer past `46be6c6`.~~ Pushed 2026-09-12; CI green at `0dc62c1` (17 m 08 s). | — | — | — | — | > > ### ▶ FINISH LINES, THE ACTIVE PATHS AND THE SHELF (author decision, 2026-11-06) > > **Why.** In the ten days to 2026-11-06, 102 rows were opened and 95 of them were residues or > splits of the row just worked, so the open queue stayed at about 33 however fast rows closed. > Each end goal now has one theorem that finishes it, and the queue offered for picking is the > path to those theorems. Nothing is unscheduled: every row below keeps its price and its > canonical text, and the author can promote any shelved row by number. > > **Finish lines.** > > | End goal | Finish line | Active path, in order | > |---|---|---| > | QC chain (link 12) | #315 — the Steane threshold with every recovery location counted at every level; #231 closes with it | ~~#302~~ (done 2026-11-07) → ~~#312~~ (done 2026-11-07) → ~~#313~~ (done 2026-11-07) → #314 (split 2026-11-07: ~~#318~~, ~~#316~~, ~~#317~~, ~~#320~~, ~~#321~~ done; #319 (split 2026-11-08: ~~#322~~, ~~#323~~ done; #332, #324)) → #315 | > | Records→spacetime | #298 — the `d+1` capstone (Lorentz within the records' resolution in `d+1`; closes #289) | ~~#294~~ → ~~#295~~ → #296 → #297 → #298 | > | Reconstruction frontier, Statistical-mechanics layer, Reviewer-proofing, Library owed | none set yet — the author sets one | #19, #20, #21, #22, #150 · #24, #218, #224 · #211 (offered as they stand) | > > **The shelf** (open, priced, not offered in the menu unless the author asks): QC chain — #310 > (explicit constants), #239 (sharper malignant-pair constant); Records→spacetime — #254, #255, > #263, #276, #281, #282 (1+1 refinements: none moves the scaling theorem), #299, #300, #301 > (`d+1` options), #205 (the common cone, research after its first brick); #39 is the end goal's > umbrella row. > > **The residue rule (replaces "every residue becomes a row").** What a landing does not do is > recorded, with its price, as a **Remainder** line in the landing's own row. It becomes a new > numbered row only when (a) a finish line's path cannot be completed without it, or (b) the > author promotes it. A remainder is never called "unscheduled" or "too big" (the 2026-09-10 > rule stands); it is simply not a row until it is needed. > > ### ▶ OPEN QUEUE, by category (recorded 2026-09-13) — the single index > > **State.** HEAD `0dc62c1`, pushed, CI green. Row **BC** (the BCFJS `iff`, six bricks, 82 pins) and > **Q29** (= G5, manifold-level Liouville; every smooth Hamiltonian flow on `ℂℙⁿ` preserves `μ_FS`, > the Schrödinger and torus flows *are* Hamiltonian flows) closed on 2026-09-12/13; the ▶ OUTSTANDING > table above now has three live rows. Every open item of every list in this file — ▶ OUTSTANDING, > list 1 (research), the maintenance table, the expert-review table, the Q-queue, `residues.tsv` — > was already recorded where it lives; this block is the **index**, one row per item with a > **category column**, pointing at the canonical row (the canonical row keeps the full text and is > the one to edit). Prices are the canonical rows' prices; the rows marked *new* are priced here for > the first time (the records and thermodynamics next steps asked for on 2026-09-13). The author > picks by `#`; nothing is "unscheduled". Categories: **Geometry** (manifold layer), > **QIT** (quantum information / QEC), **Records** (the record layer, MD-1), **Thermo** > (`Thermo/`, `thermo-plan.md`), **Reversible** (circuits, ecdsa.fail), **Foundations** (research > frontiers with no Lean shape yet), **Positioning** (expert-review rows), **Docs** (guards, > ledgers, hardening). **End goal** (added 2026-09-21, one label per row, so the queue can be > sorted by what a row is *for*): **QC chain (link n)** — the postulates→quantum-computer > chain of [`FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md`](../docs/FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md) > (link 9 amplitude estimation, 11 arithmetic, 12 fault tolerance); **Reconstruction frontier** — > Σ+Ω → QM (Posit 1, the sector posit, ETH, the record layer); **Records→spacetime** — rows 38–39; > **One QM, many formulations** — the wider-picture rows (Feynman, Wigner–Moyal, Nelson/Bohm, > Koopman–von Neumann, Berry); **Experiment twins** — every flagship experiment gets a CSD twin; > **Reviewer-proofing** — positioning and expert rows; **Library owed / upstream** — Mathlib PRs, > pin moves, Darboux residues; **New predictions** — R-019; **Statistical-mechanics layer** — added 2026-10-12, the middle layer between records and spacetime that #100 named when it landed: typicality and concentration (`Thermo/`), dephasing and ETH, chaos and scrambling, coarse-graining and effective stochasticity. Rows #21, #101, #118, #130, #140, #141. It is where the **third** regime of “random” lives — the apparent stochasticity of macroscopic coarse variables, deterministic underneath (`CSD-CHARTER.md`, ▸ *Three regimes*) — and it is deliberately **not** a route to the Born weights, which is #100's own prohibition and `KahlerFibreMixing.lean`'s. > > > **▶ Link 12: the order** (2026-09-26). Link 12 is now the largest strand in the queue, because > its three residues were priced into rows instead of being left as prose: `R-003` became #51–#53, > `R-004` and `R-006` became #66–#79, and two of those were re-priced when their real cost showed > (#69 → #80 + #81, #72 → #83 + #84 + #85). Nineteen of them have landed, and so has #86 (#61's > second half), and so has tier-C #74 through its own #112–#117 split. **The one row still open > in the whole strand is #62.** Tier A closed the last residue in the chain that had a Lean shape; > tier B completed the fault-tolerance story; tier C is no longer deferred — only #62's parts (c) and > (d) are. > > | Order | Row | Cx | Why here | > |---|---|---|---| > | ~~A1~~ | ~~**#84**~~ | ~~M–L~~ | **DONE 2026-09-26.** The `z`-`y`-`z` Euler decomposition, the `ABC` identity and the one-control circuit. A2 and A3 are now unblocked; both are stated in its `ctrlChoice_mul` algebra. | > | ~~A2~~ | ~~**#81**~~ | ~~L~~ | **DONE 2026-09-26.** `⟨H, T⟩` dense in `U(2)` modulo phase, through a third axis built as a cross product and a general non-integrality test. **Tier A is now #73 alone.** | > | ~~A3~~ | ~~**#85**~~ | ~~L~~ | **DONE 2026-09-26.** Pattern normalisation and the ancilla-free control recursion; #72 is closed in full. What #73 consumes is `ctrlGateOf_mem_closure`. | > | ~~A4~~ | ~~**#73**~~ | ~~S–M~~ | **DONE 2026-09-26 — `R-005` CLOSED.** Clifford+T is dense in `U(2ⁿ)` modulo phase. **Tier A is complete**, and the chain's open-mathematics list is `R-016` alone. | > | ~~B1~~ | ~~**#86**~~ | ~~M–L~~ | **DONE 2026-09-27** (it took L, not M–L). The concatenated quantum recovery at level `k`: Knill–Laflamme in encoder form composes under the new tensor-over-blocks layer, and one recovery channel per level restores every code state from every pattern with at most one bad sub-block per level. **Tier B is complete.** | > | ~~B2~~ | ~~**#53**~~ | ~~M~~ | **DONE 2026-09-27.** The `Σ`-twin of the Steane recovery: the Pauli channel as the environment marginal of one flow on `ℂℙ²⁸¹⁵`, and QEC on `Σ` end to end. | > | ~~B3~~ | ~~**#79**~~ | ~~M~~ | **DONE 2026-09-27.** The decoder as sixteen `CNOT`s and no Hadamard: wire `0` carries the logical qubit uniformly in the other fourteen labels, and its amplitudes are #77's `outQubit e`. **Tier B is now #86 alone.** | > | ~~C1~~ | ~~**#82**~~ | ~~S–M~~ | **DONE 2026-09-28** (it took S). The `d(d − 1)/2` count for #70 — the phase-fixing factor the Givens construction never needed, because the pivot already holds the norm of the column. **Tier C is now #74 and #62.** | > | ~~C2~~ | ~~**#74**~~ | ~~XL~~ | **DONE 2026-10-06 — and the "not claimed" is RETIRED.** Split into #112–#117 and all six landed: Solovay–Kitaev with a gate count `K·log(1/ε)^c`, `c = log 5/log(3/2) ∈ (3, 4)`, in #117’s three-letter cost model. The constants are existential (inherited from #113’s compactness net) and the literature’s `3 + δ` stays unclaimed; **the chain now claims gate counts, not only density.** | > | C3 | **#62** | (a)(b)(e)(f) done; (c) XL; (d) XL+ | The circuit-level threshold theorem, **split 2026-09-28 with (a), (b), (e) and (f) landed the same day**: faulty gates on a transversal gadget are corrected, the **whole Clifford generating set** of the Steane code is its transversal gates (`X̄`, `Z̄`, Hadamard, and the two-block `CNOT`), and the circuit's union bound, level count and **polylogarithmic overhead** are proved. What is left: (c) fault-tolerant syndrome extraction with the extended rectangle, and (d) the simulation theorem. | > > Tier A is a chain: A1 before A2 and A3, both before A4. Tier B was three independent rows in value > order and is complete. **What is left of link 12 is #62 alone** — > its deferred parts (c) and (d), blocking nothing. > > | # | Category | End goal | Item | Cx | Value | Canonical row | > |---|---|---|---|---|---|---| > | ~~1~~ | Thermo | Experiment twins | ~~*new*~~ **TH5a DONE 2026-09-13** (`Thermo/Jarzynski.lean`, 13 pins, M took S–M: ★★ `jarzynski` / `jarzynski_freeEnergy`, the joint law `tpmLaw` is the protocol's Born statistics (`gibbsWeight_eq_re_born`), doubly stochastic transitions by unitarity, Jensen second law `mean_work_ge_freeEnergy_sub`; glossary `jarzynski-equality`) — **Jarzynski equality**, finite-dimensional two-point-measurement form: Hermitian `H₀, H₁` on `Fin d`, a unitary `U`, initial `gibbsState H₀ β`; the work variable `W = E'ₘ − Eₙ` on the outcome pair `(n, m)` with law `pₙ^{Gibbs} ‖⟨m'∣U∣n⟩‖²`, and `𝔼 e^{−βW} = Z₁/Z₀ = e^{−βΔF}` (unitarity of `U` in the eigenbases). Consumes `Thermo/FreeEnergy.lean`'s `partitionFn`/`gibbsState`/`freeEnergy`. | **M** | Medium–high (the first fluctuation theorem in the corpus; the stretch row of `thermo-plan.md` TH5) | [`thermo-plan.md`](thermo-plan.md) TH5 | > | ~~2~~ | Thermo | Experiment twins | ~~*new*~~ **TH5b DONE 2026-09-13** (`Thermo/Crooks.lean`, 11 pins, S–M took S–M: ★★ `crooks_pairwise` / `crooks_pairwise_freeEnergy` / `crooks_ratio`, the reverse transition matrix is the transpose by unitarity (`tpmTransition_star`), the work distribution `workDist` on `workValues` with ★★ `crooks` (`P_F(w) = P_R(−w)·e^{β(w−ΔF)}`) and `jarzynski_workDist`; glossary `crooks-fluctuation-theorem`) — **Crooks relation**, the pairwise form: with the reverse protocol `U†` from `gibbsState H₁ β`, `P_F(n, m) / P_R(m, n) = e^{β(W_{nm} − ΔF)}` on every outcome pair, and the histogram form by summing the pairs at a fixed `W`; TH5a is its corollary. | **S–M** (after TH5a; the pairwise ratio is two lines of `Real.exp` algebra) | Medium | [`thermo-plan.md`](thermo-plan.md) TH5 | > | ~~3~~ | Thermo | Reconstruction frontier | ~~*new*~~ **TH5c DONE 2026-09-14** (`Thermo/ETH.lean`, 11 pins, M–L took M: `SatisfiesETH hH A f ε` (diagonal form), ★★ `tendsto_birkhoffAverage_heisenbergObs` — dephasing on Mathlib's `birkhoffAverage`, the same object `Equilibration.lean` uses, with NO mixing (the mechanism is that every off-diagonal term of the sampled Heisenberg series `heisenbergObs_pow` is a unit-modulus geometric sequence with ratio ≠ 1, `tendsto_cesaro_geom`), limit = the diagonal ensemble `∑ |cₙ|² ⟨eₙ,Aeₙ⟩`; the hypothesis is `Nonresonant hH τ` on the SAMPLED phases, and `nonresonant_of_nondegenerate` derives it from a non-degenerate spectrum with `0 < |τ(Eₙ−Eₘ)| < 2π`; ★ `abs_diagonalEnsemble_sub_le_of_eth` / `_of_window`, ★ `abs_microcanonicalDiag_sub_le`, ★★ `eth_time_average` (time average → a value within `2(ε+η)` of the microcanonical average over the window). ⚠️ Discrete sampling, diagonal ETH only, the continuous-time `(1/T)∫₀ᵀ` form not stated.) — **TH5c — ETH as a hypothesis field and its consequence**: a `SatisfiesETH` predicate (diagonal matrix elements of `A` in the energy basis are within `ε` of a smooth function of the energy) and the theorem *ETH + non-degenerate spectrum ⇒ the infinite-time average of `⟨A⟩ₜ` is the diagonal-ensemble value, within `ε` of the microcanonical value*, on the time-average machinery of `Thermo/Equilibration.lean`. ⚠️ ETH *itself* (that generic Hamiltonians satisfy it) is random-matrix theory with no Lean shape; it is **not** part of this row and is listed under Foundations as #21. | **M–L** | Medium | [`thermo-plan.md`](thermo-plan.md) TH5, `Thermo/Equilibration.lean` | > | ~~4~~ | Thermo | Experiment twins | ~~*new*~~ **TH5d DONE 2026-09-14** (`RecordLayer/DrivenTwoTime.lean` + `RecordLayer/JarzynskiRecord.lean`, 19 pins, M–L took M: the generic engine is *measure, DRIVE, measure* — `driveStage` between Q25's two stages, with ★ `drivenJointRecordSector_eq` (at the level of record sectors a drive `Φ` before the second readout IS the second context pulled back along `Φ`, so `two_stage_joint` applies verbatim) and ★★ `driven_mixed_two_time_born` (`P(i,j) = Tr(ρ∣bᵢ⟩⟨bᵢ∣)·c₂.rate (Φ'[bᵢ]) j`, first readout in ANY orthonormal basis `b`, mixed preparation); then ★★ `sigma_tpm_law` — the driven two-stage preparation (Gibbs mixed preparation, first bank calibrated to the `H₀` eigenrays, `H₀`-energy context, drive `U` on the sector = the corpus's `U(N)` action on `Σ`, `H₁`-energy context) gives the joint record sector EXACTLY `tpmLaw`, no positivity hypothesis (Gibbs weights are strictly positive); ★★ `sigma_jarzynski` / `sigma_jarzynski_freeEnergy` / ★ `sigma_mean_work_ge_freeEnergy_sub` on the record probabilities; **Q29(e) consumed**: ★★ `sigma_tpm_law_hamiltonianFlow` / `sigma_jarzynski_hamiltonianFlow` with the drive written as `IsSymplectic.hamiltonianFlow` of `−2⟨H_d⟩`, identified through `hamiltonianFlow_schrodingerHamiltonian`; Gleason-free structurally (import-negative inventory); glossary `jarzynski-equality` updated. ⚠️ *Correction to this row's phrasing:* Liouville (`fsVolume_map_hamiltonianFlow`) is NOT consumed by the identity — the Gibbs preparation is the spectral mixture of Diracs and the collapsed states are Diracs, transported pointwise; only the flow's identification is used. An AC-density (`ρ_ep dμ_FS`) Gibbs preparation would consume it and is not the corpus's canonical mixed preparation.) — **TH5d — Jarzynski at the `Σ` level**, the first thermodynamic consumer of Q29: a preparation measure whose density operator is `gibbsState H₀ β` (`LF2/PreparationQdensity.lean`, W2/W3), the driving as the Schrödinger flow on the sector (`FlowChannel.lean`, W6) — which by Q29(e) is the Hamiltonian flow of `⟨H⟩` and by `fsVolume_map_hamiltonianFlow` preserves `μ_FS`, so an absolutely continuous preparation density is *transported*, `ρ ∘ Φ₋ₜ` — and the two energy readouts as the two-time record law of `RecordLayer/TwoTimeLuders.lean` (Q25); the theorem is that the record-layer work statistics are TH5a's and satisfy the equality. | **M–L** (after TH5a) | High — thermodynamics stated on `Σ`, not on bare matrices; Q29 finally consumed | [`thermo-plan.md`](thermo-plan.md) TH5; [`generator-layer-scoping.md`](generator-layer-scoping.md) §11 | > | ~~5~~ | Records | Reconstruction frontier | ~~*new*~~ **RL-1 DONE 2026-09-14** (`RecordLayer/EntangledRecord.lean`, 15 pins, M took S–M: `DensityOperatorIx.toFin` (the `LF2/ReducedDensity.lean` bridge, through a quantified labelling `e : ι ≃ Fin d`), `reducedDensity e x : DensityOperator d` with `traceForm_reducedDensity_single` = `arenaObs (leftOp ∣e⁻¹ i⟩⟨e⁻¹ i∣) x`; ★★ `entangled_local_record_born` (the mixed swap preparation at the reduced state gives the outcome sector exactly the composite arena's local observation, every composite point), ★ `entangled_local_followup`, `bell_record_weight₀/₁ = ½`; NEW generic ★★ `mixed_two_time_born` (`P(i,j) = Tr(ρΠᵢ)·c₂.rate [eᵢ] j` for every density operator — `two_stage_joint` at `mixedReadyPrep`, which had no two-time consumer) + `_first_record`/`_other_fate`; ★★ `entangled_two_time_born` = the gated item of `two-time-luders-scoping.md`, `bell_two_time_born = ½·c₂.rate`) — **RL-1 — Q27's second half**: transport the delivered `arenaObs (leftOp A) x = re tr(reducedDM x · A)` (`CV/EntangledWeights.lean`) through the `Fin`-indexed LF2 mixed tier (`DensityOperatorIx`, `MixedLuders`), and the **entangled two-time instantiation** — the swap witness over the composite arena with `reducedDM` weights (`two-time-luders-scoping.md`'s gated list). *Residue → #25.* | **M** | High — the external reader's direct question, closed at the record tier | Q27 in the Q-queue below | > | ~~6~~ | Records | Reconstruction frontier | ~~*new*~~ **RL-2 DISSOLVED 2026-09-14** — the row asked for the base-level regions `{Ωᵢ(M)}` on `ℂℙⁿ⁻¹` as FS-Voronoi cells of the context's basis, which is Paper C Appendix D's *illustrative* realisation (the paper says so; LF3 §5.12 and TN3 §5.5 say "No Use of Fubini-Study Voronoi Volumes"). Nothing in Lean uses Voronoi and nothing ever has. `record-layer-plan.md` Phase 1 (2026-07-25) tested base-only over exactly these regions at `N = 3`, found the density forced negative, and §3b drops base-sharp regions for the fibre; the landed context-fixed partition is the fibred `globalBasin` (`GlobalBasin.lean`), whose null boundaries are the register-circle seams (`NullSeamGeneralN.lean`). The 2026-09-13 pricing carried the July wording forward without noticing the plan had retired the construction. No consumer: the base-only question the geometry would feed is #22, parked. The seam-nullity fact on its own (`{‖⟨eᵢ,ψ⟩‖ = ‖⟨eⱼ,ψ⟩‖}` is `μ_FS`-null) is measure theory with no CSD consumer and is not queued. | — | — | [`record-layer-plan.md`](record-layer-plan.md) §3b, §5 step 4 | > | ~~7~~ | Records / Geometry | Reconstruction frontier | **`R-016′` DONE 2026-09-14** (`LF4/ArenaSymplectic.lean`, 19 pins, L took L): the arena `ℂℙᴺ × T²` is a symplectic manifold (★★★ `arenaForm_isSymplectic`), Liouville holds on it for every smooth `H`, `X_{H∘π₁} = (X_H, 0)` for sector energies, and ★★★ the Schrödinger unitary with the torus fixed IS the arena's Hamiltonian flow of `−2⟨H⟩ ∘ π₁` (`hamiltonianFlow_sectorEnergy_schrodinger`); the torus stroke is locally Hamiltonian on the arena. Four Category-1 modules underneath (translation atlas on `AddCircle`, products over the product normed space, constant forms, `π₁^* α + π₂^* β`). Remaining pieces re-priced as #28 and #29, both DONE the same day (`LF4/ArenaStrokeFlux.lean`, `LF4/ArenaVolume.lean`); the joint lift on the arena stays `R-016`, paper-side. | ~~**L**~~ | — | ▶ OUTSTANDING above; [`residues.tsv`](residues.tsv) R-016 | > | ~~8~~ | Geometry | Library owed / upstream | **Q31 (= G18) — DARBOUX'S THEOREM DONE 2026-09-21** ([`Mathlib/Geometry/Manifold/Darboux.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Darboux.lean), Category 1, 12 pins; M took M): the assembly of the 2026-09-18 pieces (★★ `extDeriv_radialPrimitiveForm`, the Poincaré lemma on a ball, [`Poincare.lean`](../CsdLean4/Mathlib/Analysis/Calculus/DifferentialForm/Poincare.lean); ★★ `hasFDerivAt_flow_of_variational_timeDependent`, ★ `form_invariant_of_flatLieDeriv_eq_zero_timeDependent`, ★ `exists_flow_hasFDerivAt_of_norm_fderiv_le`, [`FlowDerivative.lean`](../CsdLean4/Mathlib/Analysis/ODE/FlowDerivative.lean)). **Non-degeneracy is open** (`isOpen_nondegenerate`: it is invertibility of `curryLeft`, `ContinuousLinearEquiv.isOpen`); **Moser's field** `X_t = −(ω_t)^♭⁻¹ β` with `ω_t = ω(x₀) + t(ω − ω(x₀))` (`moserField`, `curryLeft_moserForm_moserField`) is jointly `C¹` through `contDiffAt_map_inverse` (`contDiffOn_moserFieldJoint`) and vanishes to second order at `x₀` (`hasFDerivAt_moserField_self`), so `‖DX_t‖ ≤ ¼` on a small ball by the tube lemma (`exists_moser_small_ball`) and the flow reaches time `1`; the transport `∂ₜω_t + L_{X_t}ω_t = η − dβ = 0` is the flat Cartan formula plus the Poincaré lemma; the time-`1` map approximates the identity with constant `¼e^{¼} < 1` (`approximatesLinearOn_flow`: Grönwall + the mean value inequality) and is an `OpenPartialHomeomorph` by Mathlib's `ApproximatesLinearOn.toOpenPartialHomeomorph`; its derivative is invertible because the pullback of a non-degenerate form is non-degenerate (`injective_of_compContinuousLinearMap_eq`). ★★ `exists_openPartialHomeomorph_pullback_eq` — **Darboux on a ball**: `Φ` with `x₀ ∈ Φ.source ⊆ ball x₀ R`, `Φ x₀ = x₀`, and at every point of the source an invertible derivative `D` with `Φ^*ω = ω(x₀)`, i.e. `ω(Φ x)(Du, Dv) = ω(x₀)(u, v)`; ★ `exists_openPartialHomeomorph_symm_pullback_eq` — read on the chart `Φ⁻¹`, `ω y (u, v) = ω(x₀)(D(Φ⁻¹)(y) u, D(Φ⁻¹)(y) v)`: **in the chart `Φ⁻¹` the form is constant**; ★★ `DifferentialForm.IsSymplectic.exists_openPartialHomeomorph_localRep_pullback_eq` — **Darboux on a symplectic manifold**, through `localRep`: the Darboux chart is `Φ⁻¹ ∘ chartAt E x₀`. The two residues are priced as #49 (the chart is `C¹` — continuous dependence of the variational solution on the initial point) and #50 (the standard form `∑ dpᵢ∧dqᵢ` — a symplectic basis, absent from Mathlib). | ~~**XL**~~ ~~**M** (assembly)~~ done (M) | Low — the classical theorem a symplectic library owes; no consumer | [`Poincare.lean`](../CsdLean4/Mathlib/Analysis/Calculus/DifferentialForm/Poincare.lean); [`FlowDerivative.lean`](../CsdLean4/Mathlib/Analysis/ODE/FlowDerivative.lean); [`Mathlib/Geometry/Manifold/Darboux.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Darboux.lean) | > | ~~9~~ | Geometry / Docs | Library owed / upstream | **Q8 / KG-3 reconciliation DONE 2026-09-14** (S took S): KG-3 closed in `future-work.md` as the geometric route, now in the corpus and provably meeting the Wigner one (`hamiltonianFlow_schrodingerHamiltonian`; sector and arena forms); the two Q8 rows here (triage table, Q-queue) struck — the "Mathlib wall" they named fell in-corpus 2026-09-07..12 (`fsForm_mextDeriv`, `fsForm_isSymplectic`, `fsForm_isKahler`, `fsVolume_eq_smul_fsMeasure`); A4's "manifold form: §2a wall" cell updated; `reconstruction-status.md` line on the top-power half and `connectivity-manifest.md`'s "our route is Wigner, not Ashtekar–Schilling" both dated. One priced residue: #30 (DONE 2026-09-15). | ~~**S**~~ | — | maintenance table below (Q8 / KG-1..3); [`future-work.md`](future-work.md) KG-3 | > | ~~10~~ | QIT | One QM, many formulations | ~~**Expert row G — Berry phase as fibre holonomy + Aharonov–Bohm twin**, on Mathlib's covariant-derivative API (the wall re-probe of 2026-09-07 stands).~~ **SCOPED AND THE STATEABLE HALF BUILT 2026-09-21** ([`berry-phase-scoping.md`](berry-phase-scoping.md); [`Mathlib/Analysis/InnerProductSpace/GeometricPhase.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/GeometricPhase.lean), Category 1, 7 pins; [`Empirical/QM/BerryPhase.lean`](../CsdLean4/Empirical/QM/BerryPhase.lean), Category 3, 4 pins; the row's L was the price of the bundle statement, which the pin cannot state — parallel transport, holonomy and principal bundles are 0 declarations — the stateable half took M): **BP-1** the Aharonov–Anandan geometric phase on the sphere — `connectionForm ψ t = Im⟪ψ, ψ̇⟫`, `dynamicalPhase`, `geometricPhase = φ − ∫₀ᵀ A`; ★★ `geometricPhase_rephase` (**gauge invariance**: rephasing the lift by any differentiable `θ` leaves the geometric phase unchanged — it is a function of the closed curve of rays), ★★ `horizontalLift_cyclic` (**the geometric phase is the holonomy**: the horizontal lift `e^{−i∫A}ψ` has `A = 0` and returns as `e^{iβ}`), ★ `geometricPhase_of_schrodinger` (`β = φ + T⟨H⟩`, the energy conserved); **BP-2** Berry's spin-½ cone — ★★ `geometricPhase_coneCurve_solidAngle`, the phase of `(cos θ/2, e^{it} sin θ/2)` over `[0, 2π]` is minus half the solid angle `2π(1 − cos θ)`. The CSD reading (note §5): the phase is a base quantity of the sector's closed orbits, unchanged by the fibre's torus phases; the fibre carries no holonomy of its own and none is claimed. Residues priced: #54 (curvature formula), #55 (discrete Aharonov–Bohm, the flux twin), #56 (adiabatic theorem). | ~~**L**~~ done (M) | Medium | [`berry-phase-scoping.md`](berry-phase-scoping.md); expert-review table | > | ~~11~~ | QIT / Docs | Reviewer-proofing | **BCFJS write-up DONE 2026-09-14** ([`bcfjs-note.md`](bcfjs-note.md), S–M took S): the hard half in five steps (support confinement; no-cloning as the rank-one case; disjoint supports are orthogonal supports via the top eigenvalue of `P_A P_B P_A` and the tensor bound; a cloned subspace splits a broadcast state into broadcast blocks; the boundary point of the segment) and the rank induction that assembles them, each step named by its declaration; why the route differs from BCFJS/Lindblad (no fidelity, no relative entropy, no equality case of any monotonicity); the upstream question asked at the pin with three checkable tags (`MATHLIB-ABSENT(file:Mathlib/QuantumInfo)`, `Matrix.partialTrace`, `Broadcasts`), so `check-mathlib-absence` fires the week any of them lands. | ~~**S–M**~~ | — | row BC above; `Mathlib/QuantumInfo/Broadcasting.lean` Provenance | > | ~~12~~ | Reversible | QC chain (link 9) | ~~**`R-001`** — amplitude-estimation straddling-kernel bound (BHMT Thm 11's literal `8/π²`).~~ **DONE 2026-09-23** (`PhaseEstimation.lean`: the kernel inequality `eight_mul_sq_le_sin_sq_pi_mul`, `dirichlet_two_index`, `phase_estimation_two_index`; `AmplitudeEstimation.lean`: ★★ `amplitude_estimation_bhmt` — the accepted `Finset` `{c, c+1, −c, −(c+1)}` carries `≥ 8/π²` and every index in it decodes within `2π√(a(1−a))/T + π²/T²`; 9 pins; R-001 closed; L took M) | ~~L~~ done | Low–medium | [`residues.tsv`](residues.tsv); [`amplitude-amplification-plan.md`](amplitude-amplification-plan.md) AA-5b | > | ~~13~~ | Reversible | QC chain (link 9) | ~~**`R-002`** — BHMT Thm 3 exponential-doubling QSearch schedule (expected-runtime bookkeeping).~~ **DONE 2026-09-23** (`Mathlib/QuantumInfo/QSearch.lean`: the run as independent stage variables `QSearchRun` on a probability space, the bookkeeping `meas_reach`/`lintegral_stageCost`/`lintegral_cost`, the schedule series `qsearch_partial_sum_le`, ★★ `qsearch_expected_cost`: expected applications `≤ 54/√a` — BHMT Thm 3's upper bound; consistency `exists_qsearchRun` on the infinite product of the stage laws; 18 pins; R-002 closed, so link 9 carries no open residue; M–L took M–L) | ~~M–L~~ done | Low–medium | [`residues.tsv`](residues.tsv); AA-6 | > | ~~14~~ | Reversible | QC chain (link 12) | **`R-003` DONE 2026-09-23** — active error correction, in three pieces, all landed. ~~**(a) the Knill–Laflamme conditions**~~ **(a) DONE 2026-09-21** ([`Mathlib/QuantumInfo/KnillLaflamme.lean`](../CsdLean4/Mathlib/QuantumInfo/KnillLaflamme.lean), Category 1, 10 pins; M took M): ★★ `exists_recovery_of_knillLaflamme` — from `P Eᵢᴴ Eⱼ P = cᵢⱼ P` the recovery channel is **constructed** (`c` Hermitian, diagonalised by Mathlib's `spectral_theorem`; the canonical errors `F k = ∑ᵢ U i k • E i` have orthogonal images of the code, `knillLaflamme_canonicalError`; Kraus operators `d k^{−1/2} P F kᴴ` completed by `1 − ∑ R kᴴ R k`, `recoveryChannel_apply`) with `R (Eᵢ ρ Eⱼᴴ) = c j i • ρ` on code states; ★ `exists_recovery_channel_of_knillLaflamme` — a noise channel satisfying the condition is inverted on the code, `R (𝓔 ρ) = ρ`; ★★ `knillLaflamme_of_recovery` — **the converse**: any channel correcting the errors up to scalars forces the condition (each `R k Eᵢ P` is a scalar on the code by a sandwiched rank-one identity, `exists_smul_of_sum_conj_eq`); `knillLaflamme_iff` packages the equivalence. Degenerate codes included. ~~**(b) the Steane recovery map**~~ **(b) DONE 2026-09-21** ([`Mathlib/QuantumInfo/StabilizerRecovery.lean`](../CsdLean4/Mathlib/QuantumInfo/StabilizerRecovery.lean), Category 1, 8 pins, and [`Empirical/QM/QEC/SteaneRecovery.lean`](../CsdLean4/Empirical/QM/QEC/SteaneRecovery.lean), Category 3, 6 pins; M took M): the general half — Pauli matrices on the register (`pauliMat`, group law, commutation, adjoint, all transported from `pauliOp` through `Matrix.ext_of_mulVec`), the group average as a code projector (★ `isCodeProjector_stabMat`: every signed group element is Hermitian because coherence forces `bdot (B x) (A x) = 0`), ★ `stabMat_mul_pauliMat_mul_stabMat_eq_zero` (a Pauli anticommuting with a generator is killed by the code, `P M P = P (G M) P = −P M P`), ★★ `stabMat_knillLaflamme` (a Pauli error family whose pairwise products are detected satisfies Knill–Laflamme with `c = 1`), ★★ `exists_recovery_stabMat`; the Steane instance — ★ `syndrome_pair_ne_zero` (two distinct single-qubit Paulis are told apart by the syndromes, the `22 × 22` `𝔽₂` computation by `decide`), `steane_detects_pair` (read as anticommutation with one of the six generators), ★★ `steane_knillLaflamme`, ★★★ `exists_steane_recovery` — **one channel undoes every single-qubit Pauli error on every code state** — and ★★★ `steane_recovery_logical`, the same on the density operator of an encoded qubit `a|0̄⟩ + b|1̄⟩`. The `Σ`-twin promised in this cell is NOT here: the three-qubit `registerFlow` hard-codes its environment dilation; the generic controlled-error dilation landed with #52 (`Empirical/QM/QEC/ControlledDilation.lean`), the Steane flow is priced as #53. ~~**(c) fault-tolerance claims** — see #51.~~ **(c) DONE 2026-09-23** — the code-capacity threshold (#51: [`Empirical/QM/QEC/SteaneThreshold.lean`](../CsdLean4/Empirical/QM/QEC/SteaneThreshold.lean), [`Mathlib/Probability/CodeCapacityThreshold.lean`](../CsdLean4/Mathlib/Probability/CodeCapacityThreshold.lean)); `R-003` closed in `residues.tsv`; the concatenated quantum recovery is #61 and the circuit-level theorem #62. | ~~**L–XL**~~ done (M + M + M) | Medium | [`residues.tsv`](residues.tsv); [`steane-plan.md`](steane-plan.md); [`Mathlib/QuantumInfo/KnillLaflamme.lean`](../CsdLean4/Mathlib/QuantumInfo/KnillLaflamme.lean) | > | ~~15~~ | Reversible | QC chain (link 12) | ~~**`R-004` / `R-005` / `R-006`** — magic-state distillation (15-to-1), Clifford+T density, T-gate injection.~~ **SPLIT 2026-09-23 into fourteen priced rows #66–#79** ([`magic-plan.md`](magic-plan.md) §"The split"): `R-006` T-injection = #66 (pure-state identity, S–M) → #67 (channel, closes it, M); `R-005` Clifford+T density = #68 (irrational `HT` angle, M) → #69 (`U(2)` mod phase, M–L) and #70 (two-level unitaries, M) → #71 (Gray code, M) → #72 (`C^k(U)`, M–L) → #73 (`U(2ⁿ)`, closes it, S–M), with Solovay–Kitaev efficiency #74 (XL, not claimed); `R-004` distillation = #75 (`[[15,1,3]]` combinatorics, S–M) → #76 (transversal `T`, M) → #77 (projection of `\|A⟩^{⊗15}`, M) → #78 (the `35 p³` bound and recursion, closes it, M–L), with the explicit decoder circuit #79 (M). Nothing above M–L except #74. | ~~**XL / XL / L**~~ split | Low–medium | [`residues.tsv`](residues.tsv); [`magic-plan.md`](magic-plan.md); #66–#79 | > | ~~16~~ | Reversible | QC chain (link 11) | ~~**`R-013`** — the `n`-fold hybrid amplitude equality for the measurement-gadget adder re-cost.~~ **DONE 2026-09-23** (`Mathlib/QuantumInfo/Reversible/HybridLift.lean`: the general gate lift `gateMat`, the measure-and-correct gadget `measureCorrectMat` as a *monomial* operator, ★ `hybridLin_basisState` — no tensor factor needed; `Empirical/QM/MeasurementAdderHybrid.lean`: ★★ `hybridAdd_amplitude`, `hybridAdd_sum`, 25 pins. **Finding:** #21's per-Toffoli replacement is not amplitude-exact — the carry ancillas hold majorities, not ANDs (`naive_cell_gadget_sign_flip`); the exact hybrid uses one gadget per cell with three CZ corrections, same `3n` Toffolis, `n` measurements. M took M) | ~~M~~ done | Low | [`residues.tsv`](residues.tsv) | > | ~~17~~ | Reversible / Docs | QC chain (link 11) | ~~**General Boolean→amplitude lift** (`Reversible/Lift.lean`: per-constructor lemmas for `X`/`CX`/`swap` + the gate-list fold)~~ **lift half landed 2026-09-23 with #16** (`HybridLift.lean` `gateMat` = the permutation of `denoteGate`, checked against `ccxAtMat` by `gateMat_CCX`; the fold `shadow_gate_list`) and ~~the **Ecdsafail-split doc de-application** (~70 stale ECDLP/ECDSA docstring references)~~ **DONE 2026-09-23** — re-audited: the Cat-1 `Reversible/` tree carries 26 application-flavoured lines, of which the provenance pointers (`Ecdsafail` repository, `specs/ecdsa/ecdlp-resource-plan.md`, a path that exists) and the "no ECDSA score change" disclaimers are correct as they stand (the 2026-09-07 maintenance note was right); four headers still *described* the downstream application rather than the mathematics and were rewritten (`ConstProp.lean`: constant propagation as the value-exact transform it is; `VerifiedAdderCarryClean.lean`: the inverter/harness score accounting removed; `CuccaroModAdd.lean`, `Depth.lean`: plan-stage names and the named curve replaced by "the downstream application's"). S–M took S. | ~~M~~ done + ~~S–M~~ done | Low–medium | maintenance table below | > | ~~18~~ | Foundations | New predictions | **SPLIT 2026-10-06 — its FIRST CLAUSE IS NOW A THEOREM, and the split is the finding.** `R-019` asked for a relaxation H-theorem on the fibre: *for a partition into cells of size `δ` the coarse-grained relative entropy with respect to Haar decreases, and the distribution approaches Haar.* The first clause landed in two bricks; the second is **NEW ROW 118**. **(a)** [`Mathlib/InformationTheory/CoarseGrainArrow.lean`](../CsdLean4/Mathlib/InformationTheory/CoarseGrainArrow.lean) (Category 1, 6 pins) **generalises #109's record H-theorem to any finite coarse-graining**: for a probability measure `π` on `α`, a finite `β`, measurable `q : α → β` and `π`-preserving `F`, the induced macro kernel (`coarseStep`/`coarseKernel` — condition on a cell, evolve, read the cell) is Markov, ★★★ `comp_coarseLaw` identifies one macro step with the pushforward along `q ∘ F`, ★★★ `comp_coarseLaw_of_measurePreserving` shows **invariance of the fine measure makes the coarse law stationary** (the reference law is not chosen), and ★★★ `antitone_klDiv_coarseLaw` is the H-theorem; ★★ `monotone_measureEntropy_coarseLaw` is #110's entropy form when the cells carry equal weight. #109's proof used nothing about records beyond measurability into a finite type, so this is that proof with the vocabulary removed. **(b)** [`SigmaLayer/FibreCoarseArrow.lean`](../CsdLean4/SigmaLayer/FibreCoarseArrow.lean) (Category 7, 4 pins) cuts the fibre into `(m+1)²` cells (`circCell`/`torusCell`/`fibreCell`, measurable through `AddCircle.measurableEquivIoc` and `Nat.measurable_floor`) and instantiates: ★★★ `antitone_klDiv_fibreCell_stroke` for the hyperbolic `catStroke`, ★★★ `antitone_klDiv_fibreCell_kFlow` for the translation. **THE PAIR IS THE FINDING.** `relaxation_requires_hyperbolic_fibre` proves that on this very fibre `kFlow` admits **no** summable decay envelope while `catStroke` has a finitely supported one — the corpus's reason for saying relaxation needs a hyperbolic fibre map. Yet **both** satisfy the monotonicity clause, by the same one-line instantiation. So monotone decrease is **not** the content of relaxation and cannot distinguish a relaxing fibre from a non-relaxing one: it holds for any measure-preserving map and any finite partition, cell geometry included. `R-019`'s difficulty sits entirely in the clause this does not give. ⚠️ **Why this was available and nobody saw it:** the monotonicity clause became reachable when **#109 landed on 2026-10-04**, four weeks after `R-019` was opened, and the residue text predates it. The "do not schedule this as a result" of CR-14 applied — and still applies — to the *relaxation theorem*; it never applied to the monotonicity clause. ⚠️ **Nothing here is a Track B prediction**, and the row's end goal is not met: that needs a fibre demonstrably *out of* equilibrium returning to it, which is #118. ~~**`R-019`** — the relaxation H-theorem for the fibre (Track B / Valentini); `relaxation_requires_hyperbolic_fibre` is the proved precondition; blocked on first-passage asymptotics.~~ | ~~**XL**~~ split ((a) M, (b) S–M as taken) | Highest *outside* the programme — and #118 now carries it | ▶ OUTSTANDING above; [`cr-queue.md`](cr-queue.md) CR-14; [`residues.tsv`](residues.tsv) `R-019`; #118; #109; #110 | > | ~~118~~ | Foundations | New predictions · Statistical-mechanics layer | *new (2026-10-06, out of #18's split)* **The convergence half of `R-019`: that the coarse-grained fibre divergence tends to `0`.** #18(a)(b) proved the monotone decrease and showed it is *not* the content of relaxation — it holds for the translation, which provably cannot decorrelate. What is left is the statement Track B actually needs: for the hyperbolic fibre map, `klDiv` of the cell law from the Haar cell law **tends to `0`**, with a rate in terms of the cell count `m`. The route, recorded so the row is pickable: (i) `catStroke_hasCorrelationDecay` already gives a **finitely supported** envelope — correlations are exactly zero at every nonzero lag — for the single observable `catObs`, so the first question is whether that extends to the *cell indicators*, which is a statement about the cat map's spectral gap on indicator functions rather than on a chosen observable; (ii) the finite-horizon hypothesis `MeasureTheory.HasCorrelationDecayUpTo` is the escape the residue has always named as untouched, and [`LF4/KahlerFlowFiniteHorizon.lean`](../CsdLean4/LF4/KahlerFlowFiniteHorizon.lean) already draws consequences from it for `kFlow` — the fibre analogue would be a *finite-horizon* relaxation statement, which is weaker than `R-019` and may be the honest target; (iii) the unconditional version needs **first-passage (hitting-time) asymptotics for small sets**, which is research-grade and absent from Mathlib, and that is the wall the residue recorded. ▸ **(ii) DONE 2026-10-10** ([`SigmaLayer/FibreFiniteHorizon.lean`](../CsdLean4/SigmaLayer/FibreFiniteHorizon.lean), Category 7, 7 pins; M–L as priced). ★★★ `integral_birkhoffAverage_fibreCellInd_sub_sq_le`: **at horizon `T` the mean-square deviation of a cell's time-averaged occupation from its Haar value is at most `(2/T)·Σ_{u | 19 | Foundations | Reconstruction frontier | **Posit 1's discharge** — derive `IsTorusGenerated` from the de-isolation dynamics; Q12's residue (neither landed witness is dynamical — no flow carves either cell family) folds into it. ▸ *2026-10-14, out of ~~#101~~(d):* the chaos question folds here too. The constructed de-isolation witness (`shearEvolve`, `RecordLayer/ShearWitness.lean`) translates the fibre, so it cannot decorrelate (the compact-group obstruction) and has Lyapunov exponent zero; the hyperbolic witness `catStroke` is strongly mixing (~~#147~~) with exponent `2 log φ` — both stated in Lean, ★★★ `fibre_witnesses_lyapunov` (~~#148~~, `SigmaLayer/FibreLyapunov.lean`). So ~~#148~~ gives this row a *measurable* discharge criterion — **the de-isolation flow has a positive Lyapunov exponent on the fibre** — without deciding it: which map nature uses is exactly what this row is. ▸ **BROKEN DOWN 2026-10-14 under the Σ-local reading (author's steer: Σ local, spacetime nonlocal, QM the epistemic calculation space) into #151–#153.** The preparation is `δ_p ⊗ vol`: the epistemic state is a point of the base, the ontic ignorance is the fibre. ~~#151~~ proves the preparation can reach the outcome through exactly two doors — the readout reads the base (the cell law as it stands) or the flow moves the base (back-reaction) — so a *derivation* of the rates is a theorem about the second door: ~~#152~~ (Σ-local product coupling + einselection ⇒ torus generation, and the cell law from *calibration* — DONE) and ~~#153~~ (both fibre coordinates record-bearing, I3 — **DONE 2026-10-15**: `catStroke` is two transverse coordinate-reading shears each of exponent `0`, `shearEvolve` is one along `q` with the second coordinate untouched, so a hyperbolic de-isolation must shear the second coordinate too; `SigmaLayer/FibreTwoShears.lean`). **All three of this row's 2026-10-14 pieces are landed; what remains of the row is #19 itself — which map nature uses.** ▸ *2026-10-21:* the narrowing brick ~~#175~~ (out of this row): every two-transverse-shear de-isolation with base-read offsets is translation-conjugate to the cat stroke, so the offsets are invisible to the criterion and the choice is between conjugacy classes. ▸ *2026-10-22:* ~~#182~~ carries ~~#177~~'s derivative-free return criterion to that whole class — the same return count as the cat stroke at every stroke, growth `2 log φ` and no other rate (`fibre_affine_returns`). R-015 (which coupling) and R-016 (generation on the arena) stay where the registry has them. | **XL** | Highest in the programme — the reconstruction frontier | list 1 above; [`POSITS.md`](POSITS.md) Posit 1; Q12 in the maintenance table | > | 20 | Foundations | Reconstruction frontier | **`R-012`** — the sector posit (that the typicality law is `μ_FS`; Posit 2), reduced to D1 by `IsForcedKahlerVolume` (the unique `U(N)`-invariant probability measure). | research | — | [`residues.tsv`](residues.tsv); [`POSITS.md`](POSITS.md) Posit 2 | > | 21 | Foundations | Reconstruction frontier · Statistical-mechanics layer | **ETH for generic Hamiltonians** (random-matrix theory; the hypothesis TH5c consumes) and **Q13 — trilemma exhaustiveness** over all arenas. | research | Medium | [`thermo-plan.md`](thermo-plan.md) TH5; maintenance table below (Q13) | > | 22 | Foundations | Reconstruction frontier | **General-`N` base-only A7** — ⏸ parked, not refuted; do not resume without a new idea (~~RL-2, #6, lands its geometry only~~ — #6 dissolved 2026-09-14: the base-sharp region geometry has no consumer while this row is parked). | **XL** | Low — gates nothing | list 1 above; Research table below | > | ~~23~~ | Positioning | Reviewer-proofing | ~~**Expert row D — Colbeck–Renner** in Lean (the chained-Bell family; the positioning prose is landed).~~ **DONE 2026-09-29** (`Mathlib/Probability/ChainedBell.lean` + `Empirical/QM/ColbeckRenner.lean`, 13 pins; M–L took M): the chained walk is **one observation used twice** — two marginals of one joint distribution differ by at most the disagreement probability, and add to `1` up to the agreement probability — with no-signalling the hypothesis that lets those one-pair links be composed across *different* pairs. ★★★ `abs_marginalA_sub_half_le` closes the walk with one reversed link and pins the marginal to `1/2`; ★★★ `integral_abs_marginalA_sub_half_le` transfers the bound to **every component of a mixture** (the functional is affine in the law). On the singlet the settings are `2n + 2` nodes of a great circle at `π − π/(2n+1)`, so the chain closes **exactly** (★★ `dotR_crA_zero_crB_last`, `cos 2nπ = 1`) and the bound is `π²/(8(2n+1))` → `0`: ★★★ `no_improved_predictive_power`. **The unbundling is now in the hypotheses**: parameter independence is `hA`/`hB`, measurement independence is the single shared `μ`, and ★ `exists_signalling_of_sharp` is the contrapositive. | ~~**M–L**~~ done | Low–medium | expert-review table below | > | 24 | Docs | Reviewer-proofing | **Prose audit completion** (`prose-audit.md`), **`REFERENCES.json` + line-precise citations**, the **`_of_` hypothesis pattern** (opportunistic), **TH-1 Lévy isoperimetry** (leave). | **M / M / M / L–XL** | High / medium–high / low–medium / low–medium | maintenance table below | > | ~~25~~ | Records | Reconstruction frontier | ~~*new*~~ **RL-1′ DONE 2026-09-14** (`RecordLayer/CompositeRecord.lean`, 8 pins, S–M took S: `localBlock_sum_normSq` (the join protocol's block sum at `localBlock` reindexes along `finProdFinEquiv`), `rayDensity_posSemidef` / `rayDensityIx` / `reduceBDensity` (Bob's `reduceB` as a `DensityOperator nB` through `reducedLeft` + RL-1's `toFin`), `reduceB_diag_re`; ★★ `composite_local_record_born` — on the composite arena the join-protocol weight of Bob's outcome `j` is `re (reduceB [u]) j j` for EVERY unit composite vector and every calibration, with `_trace` / `_traceForm` forms; ★★ `composite_record_eq_mixed_record` — the join-protocol record weight on the PURE composite point EQUALS the swap-protocol record weight of the MIXED preparation of Bob's reduced state, outcome by outcome: the two lines meet at a theorem) — **RL-1′ — the composite-arena seam of Q27**: the ontic form. A local block context on the composite arena (`RecordLayer/LocalBlockBridge.lean`, `localBlock`; the join protocol's `join_sector_born` at that block map gives `‖localProjB j v‖²/‖v‖²`, `norm_blockProj_localBlock`) has sector weight equal to `re tr(reduceB [v] · Πⱼ)` — the reduced-state pairing of `OnticMarginals.lean`. RL-1 delivers the same NUMBER at the local witness's mixed preparation; this row identifies it on the composite arena itself (pure composite point, apparatus coupled to one sector), so the two lines of the corpus meet at a theorem rather than at a coincidence of values. Entrywise: `‖localProjB j v‖² = ∑ₐ ‖v (a, j)‖²` is the `(j, j)` entry of `traceLeft (vecOuter v)`. | **S–M** | Medium — closes the "you modelled the local system as a classical mixture" objection at the record tier | `RecordLayer/EntangledRecord.lean` honest scope; Q27; the dynamical no-signalling row below (brick 3 is its neighbour) | > | ~~26~~ | Docs | Reviewer-proofing | ~~*new*~~ **Glossary rule-9 pass DONE 2026-09-14** (fourteen batches, commits efcbb88..831e81a): every one of the 131 entries' four registers re-read against rule 9 (subject-verb openings, a fact and its consequence joined with so/as/and/which, contractions, plain nouns, say what the reader gets, no colon reveals), technical content and named consumers unchanged, `reviewed:` bumped on every entry, one non-ASCII character removed (Klein person field). `check-glossary` PASS at every batch, staleness 0. | ~~**M–L**~~ took M | Medium — the author's stated readability bar for the public pages | `docs/glossary.yaml` header rule 9; `check-glossary.sh` | > | ~~27~~ | Docs | Library owed / upstream | **The Mathlib pin-move rename table (HY-6)** — every remaining Compat-canary warning whose replacement name is absent from the pin. ▸ **THE SWEEP RAN 2026-10-12, at the pin move** (Lean `v4.34.0`, Mathlib `5ed2965`): sections (1), (2), (3) and (6) of the inventory are **all zero** — 1,602 Lean-core sites in 207 files, the four `Mathlib.Data.*` → `Mathlib.Basic.*` import moves, `.lipschitz`, `Filter.eventuallyEq_set` and the quotient-map lemma — and the one row the 2026-10-01 measurement called out as a trap, `.lipschitz`, was indeed dot notation at every site and needed a hand patch rather than the planned sed. **What remains of this row is the `stdSimplex` restatement alone** (16 sites, 3 files, four of them axiom pins): `stdSimplex` is a `Set`, `StdSimplex` is a bundled type with `.weights`, the `Set`-level API has no `Set`-level successor, so the five theorems' *statements* change — deferred with the deprecation silenced per declaration and the reason written into `Instances/ProjectiveSpaceMomentMap.lean`. ▸ **DONE 2026-10-23 (S as priced; built first time).** The restatement: the standard simplex is Mathlib's bundled `StdSimplex ℝ (Fin (n + 1))` (namespace `Convexity`, `Mathlib/Geometry/Convex/ConvexSpace/Defs.lean`), and the `Set`-level membership's successor is `StdSimplex.range_toFun_comp_weights`. `momentMapSimplex` (the moment map's value as a simplex point, `weights_momentMapSimplex`), `momentMap_mem_range_weights`, `sqrtVec` applied to the weight vector, ★★ `momentMapSimplex_surjective` — **the moment map onto the bundled simplex is surjective** (`momentMapSimplex_mk_sqrtVec`), ★★ `range_momentMap : Set.range momentMap = Set.range StdSimplex.weights`, and `range_momentMap_eq_inter` (the non-negative sum-one vectors, through Mathlib's successor lemma). Import `Mathlib.Geometry.Convex.ConvexSpace.Defs` replaces `Mathlib.Analysis.Convex.StdSimplex`; the corpus has no `linter.deprecated` silencer left; eleven pins (five new, one renamed), all on the standard triple; `pin-move-inventory.sh` section (4) reads zero by design. **HY-6 is swept in full; the script stays for the next pin move.** The shim collapse of section (5) is **#142**. **TABLE VERIFIED AND MADE SELF-MEASURING 2026-10-01** ([`scripts/pin-move-inventory.sh`](../scripts/pin-move-inventory.sh)): ~~the sweep itself stays gated on the pin move — the replacement names do not exist here, so nothing can be applied without leaving the pin~~ — and the *table* was executable then, and it was wrong in five places after two and a half weeks of corpus growth. **FIVE DRIFTS, ALL IN HY-6 NOW:** the Lean-core count ~1,300 → **1,564**; `stdSimplex` one file → **three**, four of the new sites being **axiom pins** whose names change with the declaration; `Data.Complex.BigOperators` one importer → **two**; `Data.Complex.Basic` six → **seven**; and the `.lipschitz` row **never matched anything at all** — every site is dot notation, the qualified name `ContinuousLinearMap.lipschitz` appears nowhere, so the planned sed would have swept zero sites and reported success. The shim row was out by 13 files, and the `try rfl` line number by 516. **THE SCRIPT'S OWN LESSON:** its first draft reported three silent zeros of its own (`$` without `re.M`; `\b` after a name ending in an apostrophe), which is the same failure the `.lipschitz` row is — so every row now prints its count and a row that finds nothing is flagged loudly, because on this corpus a zero means a wrong pattern far more often than a clean sweep. | ~~**S–M**~~ done (S + S) | Medium — the next pin move is otherwise a surprise | section D below (HY-6); [`validation-hardening-plan.md`](validation-hardening-plan.md) row M canary log | > | ~~28~~ | Records / Geometry | Reconstruction frontier | **`R-016″` DONE 2026-09-14** (`LF4/ArenaStrokeFlux.lean`, 4 pins + 2 Category-1 pins, M took S–M): ★★ `not_isHamiltonianVectorField_torusStrokeField` — for `a ≠ 0` no function `H` on the arena (any function, differentiability derived) has the torus stroke `(0, a)` as its Hamiltonian vector field for `arenaForm`. Route as priced: the derivative of `H` in every torus direction `b` is `area a b` (`mfderiv_apply_torus_of_isHamiltonianVectorField`), the transverse curve `s ↦ (x, (t₁ + s b₁, t₂ + s b₂))` has manifold derivative `(0, b)` (`hasMFDerivAt_torusCurve`, from `AddCircle.hasMFDerivAt_coe` — the quotient map is a local diffeomorphism for the translation atlas — and the self-model pairing lemma), `H` changes by `area a b` over one period (`energy_torusCurve_one_sub_zero`, `constant_of_has_deriv_right_zero`) and the curve returns (`torusCurve_one_eq_zero`), so `−a₂ = a₁ = 0`. With `isLocallyHamiltonian_torusStrokeField`, "symplectic, locally Hamiltonian, not Hamiltonian" is now the strokes' classification on the symplectic manifold. | ~~**M**~~ | — | [`ArenaStrokeFlux.lean`](../CsdLean4/LF4/ArenaStrokeFlux.lean); `RecordLayer/PiecewiseHamiltonian.lean` | > | ~~29~~ | Geometry | Library owed / upstream | **`arenaVolume = c · μ_FS ⊗ vol` DONE 2026-09-14** (`LF4/ArenaVolume.lean` + two Category-1 modules, 14 pins; M–L took M): ★★★ `arenaVolume_eq_smul_kMuL` — `arenaVolume N = arenaVolume N univ • kMuL p₀` for every base point, so ★★ `kMuL_smul_map_hamiltonianFlow` is Liouville for the corpus's `μ_FS ⊗ vol_{T²}` at manifold level. Route taken: **uniqueness, not the binomial identity.** `Geometry/Manifold/FormInvariance.lean` names the chart hypothesis of `topFormMeasure_map_eq` (`PreservesLocalRep s g`, "`g^* s = s` read in charts") and closes it under iterated powers and under products of manifolds charted over `E × F` (`PreservesLocalRep.prodMap`); chart translations preserve constant forms on a translation atlas, and translation of `AddCircle T` is a chart translation. `Instances/ProjectiveSpaceFubiniStudyInvariance.lean` restates `fsVolume_map_smul`'s hypothesis for the 2-form (`preservesLocalRep_fsForm_smul`). Hence ★ `arenaVolume` is invariant under `U(N+1) × id` and under `id × T²`; its normalisation has a `U(N+1)`-invariant sector marginal (= `μ_FS` by `fsMeasure_unique`) and translation-invariant torus slices (= multiples of Haar by Mathlib's `isAddInvariant_eq_smul_of_compactSpace`), and `Measure.prod_eq` on rectangles finishes. ~~**Not delivered, re-priced as #32:** the constant's value `(N+1)·(4π)^N` and `arenaVolume ≠ 0` (the theorem holds with `c = 0` if the measure vanished).~~ **#32 DONE 2026-09-17** (`arenaVolume_univ`, `arenaVolume_ne_zero`, `arenaVolume_eq_ofReal_smul_kMuL`). | ~~**M–L**~~ | — | [`ArenaVolume.lean`](../CsdLean4/LF4/ArenaVolume.lean); [`FormInvariance.lean`](../CsdLean4/Mathlib/Geometry/Manifold/FormInvariance.lean) | > | ~~30~~ | Geometry | Library owed / upstream | **KG-3′ DONE 2026-09-15** (`Instances/ProjectiveSpaceSchrodingerHolomorphic.lean` + Cat-1 `Geometry/Manifold/HolomorphicVectorField.lean`, 11 pins; S–M took S): ★★ `isHolomorphicVectorField_schrodingerField` — `X_{−2⟨H⟩}` is a holomorphic vector field of the Kähler manifold `ℂℙⁿ` (atlas sense: in every affine chart the field's derivative commutes with `J = i·`; the field read in any chart is `schrodingerChartField` there, by uniqueness of the local Hamiltonian vector, and that is a quadratic polynomial with complex coefficients); ★★ `isHolomorphicMap_hamiltonianFlow_schrodinger` / `fsMetric_hamiltonianFlow_schrodinger` — the Hamiltonian flow of `−2⟨H⟩` is at every time a holomorphic map (`mfderiv` commutes with `fsJ`) and an isometry of the Fubini–Study metric `ω(J·,·)` (Killing), from ★ `isHolomorphicMap_smul`, ★ `fsForm_smul_mfderiv` (`U^* ω_FS = ω_FS` at bundle level) and the general ★ `IsAlmostKahler.metric_mfderiv_eq` (symplectic + holomorphic ⇒ isometry). New Cat-1 vocabulary: `IsHolomorphicVectorField J₀ X` (chart field `J₀`-holomorphic in every chart), `IsHolomorphicMap J g` (`dg ∘ J = J ∘ dg`). **Not delivered:** `L_X J` as an object, and the chart-independence of `IsHolomorphicVectorField` on a holomorphic atlas (needs the ℂ-bilinearity of a holomorphic transition's second derivative); neither is consumed. | ~~**S–M**~~ | — | [`future-work.md`](future-work.md) KG-3 | > | ~~31~~ | Docs | Reviewer-proofing | **Glossary images DONE 2026-09-14** (S + M took S–M; author request the same day, "the glossary is very dry"): 67 of the 131 pages now carry a portrait or a diagram from Wikimedia Commons, self-hosted under `docs/img/` (4.6 MB, 480px thumbnails), each with a caption and the credit line its licence asks for, shown beside the plain-terms register (landscape diagrams full width, `wide: true`), as the social-preview image, and listed on a generated `/credits/` page. Source rule: a portrait for an eponym, the standard diagram for an apparatus or construction, nothing decorative; free licences only (public domain, CC0, CC BY, CC BY-SA, Commons "Attribution"), no GFDL-only files, no local non-free uploads. `scripts/lib/glossary_images.py` is the fetch-and-review tool (Wikipedia summary API for the lead image, Commons API for licence and author); every candidate was looked at, and three were rejected (a handwritten letter for Stone, a Feynman diagram for Wick, the doubtful Bayes portrait for QBism). `check-glossary` (H) requires the file and every attribution field; `docs/img/**` triggers the Glossary workflow. 64 entries have no image because no relevant free one exists (programme-internal terms, the question pages, and eponyms without a free portrait: Duistermaat, Jarzynski, Crooks, Lüders, Naimark, Bargmann, Loewner, Lindblad, Choi, Ozawa, Vazirani, Mosca, Gottesman, Steane, Kitaev, Simon, Ghirardi, Wick, Kubo). | ~~**S + M**~~ | — | `docs/glossary.yaml` header (`image`); `scripts/lib/gen_glossary.py`; `scripts/check-glossary.sh` (H) | > | ~~32~~ | Geometry | Library owed / upstream | **The arena volume's mass and positivity DONE 2026-09-17** (`LF4/ArenaVolume.lean` + two Category-1 modules `Geometry/Manifold/WedgePowPairs.lean` and `Instances/ProjectiveSpaceTorusVolume.lean`, 25 pins; M–L took M): ★★ `arenaVolume_univ` — `arenaVolume N univ = ENNReal.ofReal ((N + 1) * (4π)^N)`; ★ `arenaVolume_ne_zero`; ★★★ `arenaVolume_eq_ofReal_smul_kMuL` — `arenaVolume N = (N+1)·(4π)^N • kMuL p₀` with the constant visible; ★★ `kMuL_map_hamiltonianFlow` — **Liouville for `kMuL` itself, unconditionally**. Route taken: **not the general binomial identity** (still unproved, `ProductForm.lean` honest scope) but a weighted pair-tuple count: `WedgeShuffle.lean`'s shuffle sum generalised to weighted pair families (`wedge_mul_apply_weightedPairs`), the general count ★★ `wedgePow_apply_of_isWeightedPairTuple` (`β^{∧k} u = k! · ∏ c j`, which now derives `wedgePow_stdForm_pairFamily`), the product basis `prodTorusBasis` as a weighted pair tuple (weights `-4` on the `N` sector pairs, `1` on the torus pair), and ★ `fsModelForm_eq_comp` (the model form at every `w` is a pullback of the form at the origin along a map of determinant `(1+‖w‖²)^{-(N+1)}`, extracted from the mass module's rotate-and-scale argument), giving the chart density `(N+1)` times the Fubini–Study density; then Tonelli, `fsVolume_univ` and the torus chart area — ★★ `topFormMeasure_prodTorus_chartAt_source` (Category 1, any periods `T, T'`: `(n+1)(4π)^n · T T'` on a product chart domain). The complement of the chart domain is null through the #29 identification (`fsMeasure_chartSource_zero`, `AddCircle.volume_compl_singleton`), so the chart value is the mass. | ~~**M–L**~~ | — | [`ArenaVolume.lean`](../CsdLean4/LF4/ArenaVolume.lean); [`ProjectiveSpaceTorusVolume.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceTorusVolume.lean); [`WedgePowPairs.lean`](../CsdLean4/Mathlib/Geometry/Manifold/WedgePowPairs.lean) | > | ~~33~~ | Positioning / Geometry | Reviewer-proofing | **Physlib Fisher–Rao bridge DONE 2026-09-16** (Nava-Hernandez, PR #1652; M took M): ★★ `Projectivization.fsMetric_eq_fisherRaoInner` — on the regular stratum, for `u` horizontal, `fsMetric x u v = fisherRaoInner (toOpenSimplex x) (momentDeriv x u) (momentDeriv x v)`, constant ONE — as a corollary of the vector-level ★ `FisherRao.fisherRaoInner_bornDeriv` (Mathlib + the mirror of Physlib's `FisherRao.lean` only) through ★ `fsMetric_eq_fsInnerHom` (the repository's metric IS the textbook formula on lifts); Braunstein–Caves projectively (`fisherInfo_bornDeriv_horizontalLift_le` / `_eq_iff`) on the way. `momentMap` moved to the Mathlib tree so the manifold closure is Category 1 (`check-import-hygiene.sh` rule 4, one dated exemption). Export package `CsdLean4/Interop/Physlib/` + `scripts/export-physlib.sh`: 44 files / 13,256 lines, nine concept slices, 1,355 declarations (every constant of the closure) on the triple, the whole tree built `--wfail` against Physlib's pin; `MANIFEST.md` is the page Nava reads. *Residues:* ~~**(a)** the hygiene doc pass — 145 docstring lines citing `specs/` / `BACKLOG` / LF consumers~~ **DONE 2026-09-16** (dead repository paths became named plans, `References:` paragraphs became `**Provenance and references.**`, consumer notes reworded; `export-physlib.sh --strict` passes at 0 hits, MANIFEST regenerated); ~~**(b)** `ProjectiveSpaceSchrodingerFlow.lean` still imports `LF4.ManyToOneSchrodingerDerived`~~ **DONE 2026-09-16** (`Mathlib/Analysis/Matrix/SchrodingerUnitary.lean`, rule 4 has no exemptions); **(d)** external review (Codex, Brief A, 2026-09-16) — two interpretation errors corrected the same day (the algebraic `4‖u‖²` bound is not the QFI bound for non-tangent directions; `fsMetric` IS the QFI in this normalisation, not a quarter of it), the moment map's factor 2 documented, injectivity on the horizontal space proved, the export tooling hardened (import grammar incl. `import all`, concept slices, hash-keyed `--wfail` builds that fail the driver, every constant swept); **(e)** external review, Brief B (the FS machinery; three scoped read-only reviewers, 2026-09-16, every concrete claim re-verified both ways) — **one BLOCKER confirmed and fixed**: `Projectivization/Topology.lean` declared its own `instMulAction : MulAction (V ≃ₗ[K] V) (ℙ K V)` under the auto-generated name Mathlib's `Projectivization/Action.lean` also uses, so the two files could not be imported together (reproduced in both import orders: `environment already contains 'Projectivization.instMulAction._proof_2'`); the local instance is gone and `mapEquiv_smul_eq` reads Mathlib's action on representatives by `rfl`. **Fixed the same day (cheap, confirmed):** the `RiemannianMetric` namespace (a Mathlib structure name) → `MetricFamily`; the orbit/field theorem `isMIntegralCurve_torusUnitary_smul` moved into the moment-map module it belongs to (the module that cited it); the local duplicate of Mathlib's `continuous_eval_const` deleted; the zero-form witness `projectiveDifferentialForm_nonempty` and the cross-slice import it forced deleted (`fsForm_ne_zero` is the non-vacuity certificate); three first-slot linearity lemmas of `HamiltonianVectorField.lean` re-proved as Mathlib's `vecCons_*`; `hamiltonian_duality` made an identification by the uniqueness lemma `eq_hamiltonianVectorFieldOf_of_forall` (non-degeneracy); eight stale scope paragraphs rewritten (closedness and the top power are theorems, not `⚠️ blocked`); the junk-value note on `IsHamiltonianVectorField`, the chart-at-only note on `IsKahler`, `bijective` for `chartMeasure_preimage_eq`, the FlowContinuity citation, a ★ mismatch; a `planning vocabulary` hygiene token (`corpus`, `milestone`, `posit`) added to the export tool and its 16 hits reworded. **Rejected (verified):** the `map_smul'` / `isProbabilityMeasure_map'` shim is the documented Mathlib-master forward-compat shim, not dead weight; `[NeZero N]` in `FubiniStudyLebesgue.lean` is needed (`volume_singleton_zero`) and already `omit`ted where it is not. **Residues after Brief B, priced; every S row done 2026-09-17 on the user's "do all the S's":** ~~**B1**~~ (S) DONE — the three conjunction capstones left `KahlerForm.lean`; the one consumer (`LF4/KahlerOnticSetup.lean`, `isFubiniStudyKahler`) builds its own bundle from the five conjuncts; ~~**B2**~~ (M took S–M) DONE 2026-09-17 — the closure has **no** line over 100 columns (was 188): `modelWithCornersSelf 𝕜 X` spelled as Mathlib's `𝓘(𝕜, X)` throughout (648 sites, terms unchanged), 87 prose lines and 56 code lines rewrapped, two over-long relative links reduced to code spans; ~~**B3**~~ (S) DONE, with a correction to the reviewer: `InnerProductSpace.complexToReal` is a `def` in Mathlib, not an instance, so there is no `inner ℝ` on a general complex space to "just use" — `metric_eq_real_inner` states the identity under the locally installed structure and the docstring says why the named `g` stays; ~~**B4**~~ (S ×3) DONE, all three confirmed and de-duplicated — `MeasurableSingletonClass (ℙ K V)` is Mathlib's `OpensMeasurableSpace.toMeasurableSingletonClass`; `compContinuousAlternatingMapL` (since 2026-09-16 Mathlib's `compContinuousAlternatingMapCLM`) was Mathlib's `compContinuousAlternatingMapCLM` (`rfl`-compatible with the coordinate change); `unitaryHaarProb` is now Mathlib's `haarMeasure ⊤` with both instances inherited (`unitaryHaar` and its four lemmas deleted); ~~**B5**~~ (M took S–M) DONE 2026-09-17 — `IsKahler`'s field `J_localTriv_symmL` constrains every chart of the atlas through its tangent trivialisation (`tangent_localTriv_symmL_eq_fderiv`); the `chartAt` form is the derived `IsKahler.J_symmL`; ℂℙⁿ's `fsForm_isKahler` proves the atlas field per affine chart (`fsJ_localTriv_symmL`); ~~**B6**~~ (M took S) RESOLVED 2026-09-17 by verification — the three flat lemmas do NOT coincide with Mathlib's `of_contDiffAt_one` / `exists_forall_mem_closedBall_*`: each adds the confinement of the solutions to a prescribed neighbourhood, which the manifold transport needs and Mathlib does not provide; uniqueness and the uniform-time principle were Mathlib's already; recorded in the module's scope note as an upstream strengthening; ~~**B7**~~ (S) DONE — the family operator is `mextDerivFamily` with `localRep_mextDerivFamily`, `contMDiff_mextDerivFamily`, `mextDerivFamily_mextDerivFamily`, `toFlat_mextDerivFamily(_zeroFormFamily)`, `mextDerivFamily_{prod,zero,const}Family`, `mextDerivFamily_fsSection`; `DifferentialForm.mextDeriv` is the bundled operator; the `fs*` prefix question is folded into B8's generality pass; ~~**B8**~~ DONE 2026-09-17 in two halves — naming: one prefix, `fs*` (`fsMeasure`, `defaultFsMeasure`, `fsMeasure_unique`, …; 1,088 renamed sites); ~~**B8′**~~ (priced M–L, took S–M) generality: slice 1 (`Projectivization/{Unitary,UnitaryTransitive,FubiniStudy,FubiniStudyUnique,TransitionProbability,MomentMap}.lean`, `Matrix/{UnitaryCompact,UnitaryHaar}.lean`) is stated over an arbitrary finite index type `{ι : Type*} [Fintype ι] [DecidableEq ι]` (`[Nonempty ι]` where `[NeZero N]` was; the reference index is `Classical.arbitrary ι`; `exists_unitary_single_eq` takes the index explicitly; the infinitude of `ℙ ℂ (EuclideanSpace ℂ ι)` is for `2 ≤ Fintype.card ι` from any two distinct indices; the moment-map barycentre is `1 / Fintype.card ι`); names unchanged, and the "233 consumer sites" of the earlier pricing were a miscount (every `(N := …)` in the repo) — the true ripple was four call sites in three files; the affine atlas of `Instances/ProjectiveSpace.lean` stays on `Fin (n + 1)`, which is genuinely index-bound (`Fin.succAbove`); ~~**B9**~~ (M–L took S–M) DONE 2026-09-17 — `Bundle.RiemannianMetric.riemannianVolume` / `Bundle.ContMDiffRiemannianMetric.riemannianVolume`: the Riemannian volume of Mathlib's bundled metrics on the tangent bundle via the bilinear family `toMetricFamily`, with cover-independence and the top-form comparison restated for them; ~~**B10**~~ (S) DONE — `_def` equations (by `rfl`) for the 20 definitions proofs unfolded most (`metric_def`, `fundamentalForm_def`, `transProbVec_def`, `insertOne_def`, `chartInv_def`, `torusChartField_def`, `chartDen_def`, `bornDeriv_def`, `pairFamily_def`, `slotPair_inl/inr`, `slotMem_inl/inr`, …); the unfolding sites themselves are untouched (the equations are definitional, so the proofs are unchanged) — 41 lesser definitions remain without one; ~~**B11**~~ (S half) DONE — `contDiff_schrodingerChartHam`, `contMDiff_schrodingerHamiltonian`, `contMDiff_torusHamiltonian` are one statement each over `m : WithTop ℕ∞` (the `ω` twins deleted); ~~**B11′**~~ (M) DONE 2026-09-17 — `contDiffAt_localRep` at every order; `contDiffAt_localHamiltonianVector_of_contMDiff` and `contMDiff_hamiltonianVectorField_of_contMDiff` are the one proof at every infinite order (`n + 1 ≤ n`), the `∞` statements are corollaries under their old names, the `ω` ones one-liners (the analytic twins deleted); on ℂℙⁿ `contMDiff_schrodingerField_of_le` / `contMDiff_torusField_of_le` with `∞`/`ω` corollaries; ~~**B12**~~ (S) DONE — ★ `topFormMeasure_addHaar_basis`: against the basis' own Haar measure `e.addHaar` the measure of a top form does not depend on the basis (`apply_basis_eq_mul_det`, `addHaar_basis_eq_smul`, `chartDensity_basis`, `chartMeasure_addHaar_basis`), so it is canonical; ~~**B12′**~~ (S) DONE 2026-09-17 — ★ `stdBasis_addHaar : (stdBasis n).addHaar = volume` (a product basis has the product parallelepiped, `parallelepiped_pi_basis`; Lebesgue measure on `ℂ` gives `basisOneI`'s unit square mass `1` through `orthonormalBasisOneI`; `Module.Basis.addHaar_eq_iff` does the rest), hence ★ `fsVolume_eq_topFormMeasure_addHaar`: `fsVolume` IS the canonical measure of `fsTopForm`, for every basis of the model; ~~**B13**~~ (M) DONE 2026-09-17 — `MeasureTheory/Measure/Haar/PiComplex.lean` (★ the Haar measure of the standard real basis of `Fin n → ℂ` is Lebesgue measure; the same family is a real orthonormal basis of `EuclideanSpace ℂ (Fin n)`; ★ `Complex.measurePreserving_ofLp`) and `Analysis/SpecialFunctions/JapaneseBracketEuclidean.lean` (★★ `lintegral_pow_inv_one_add_norm_sq`: `∫ (1 + ‖x‖²)^{-(n+1)} = πⁿ/n!` on every real inner-product space of dimension `2n`, through orthonormal-basis isometries); closure 46 modules, 1,406 declarations. Verdict: every B row is closed, B8′ included; **(f)** external review, Brief C (Codex, 2026-09-17, on `2b78ee5`) — eight findings, every one re-verified against the tree and every one real: `IsAlmostKahler` had no smoothness of `J` (fixed: the field `contMDiff`, an existential `C^∞` section of `Hom(TM, TM)` agreeing with `J`; ℂℙⁿ supplies `fsJL` / `contMDiff_fsJL`, now proved from the atlas condition alone through `contMDiff_hom_section_of_localTriv_symmL` so the predicate is not circular); the metric-volume adapter still depended on `(μ, e)` (fixed: `det_gram_basis`, `chartDensity_basis`, ★ `riemannianVolume_addHaar_basis`, and `Bundle.RiemannianMetric.canonicalVolume` with `_congr_basis` / `_congr_cover`); the almost-Kähler docstring had the sign convention of the flat file (`ω = g(·, J·)` here, `ω = g(J·, ·)` there — stated); `KahlerForm.lean`'s header quoted the unnormalised top-power formula (now `ωⁿ = (4π)ⁿ μ_FS`, `vol_g = ((4π)ⁿ/n!) μ_FS`); the hand-proved orthonormality of `Complex.euclideanBasis` duplicated Mathlib's `Pi.orthonormalBasis` (replaced); the `ContMDiffVectorBundle n` binder of the order-generic theorems was derivable from `IsManifold n` + `n + 1 ≤ n` (derived internally; `contDiffAt_localRep` keeps it, being finite-order); the transition theorem covered only `chartAt`-to-`chartAt` (added ★ `IsKahler.fderiv_atlas_transition_comm` for any two atlas charts, via `fderiv_atlas_transition_eq` and `coordChange_comp`); two scope comments were stale (fixed). Codex's Q5 route for B8′ (a generic `[Fintype ι]` implementation with the `Fin N` names as wrappers, so the 233 consumer sites need not change) is the plan when B8′ is picked; its master-drift spot check (eight files against `ce4e0229`) found no definite breakage. ~~**(c)** the Physlib PR sequence itself — `QuantumInfo/ForMathlib/FubiniStudyFisherRao.lean` after #1652 merges (one import line changes), then slices 1–6 to `PhyslibAlpha`~~ **re-scoped 2026-09-18**: Physlib gets **slice 2 only** — the vector-level file split into `FubiniStudyFisherRao.lean` (the Born map and the identity, 215 lines) and `BraunsteinCaves.lean` (homogeneous coordinates and the inequality, 319 lines); the manifold bridge and its closure stay here on the Mathlib track (consumers depend on csd-lean4 as a Lake package, same pin). The eight-point gap to a mergeable `QuantumInfo` file, checked against Physlib's linters: **1** module-doc template, **2** line length, **3** `[ref: key]` tags + two bib entries (`Braunstein:1994zz`, `Bengtsson:2006rfv`, from InspireHep), **4** repository internals in the prose, **5** length, **7** docstring register — all closed in the exporter (`scripts/export-physlib.sh` emits the two files Physlib-ready from `scripts/physlib-sidecars/`, with replicas of Physlib's module-doc, style and reference linters run at export time and `--strict` failing on any of them; nothing hand-edited); **6** namespace/names and **8** the import of Nava-Hernandez's file (his #1657 targets `PhyslibAlpha`, which `QuantumInfo` may not import) are settings of the exporter (`--physlib-dir`, `--fisher-rao-module`) and wait on him and Tooby-Smith. | **M** | High — the public commitment to Physlib, met | [`future-work.md`](future-work.md) KG-4; [`Interop/Physlib/MANIFEST.md`](../CsdLean4/Interop/Physlib/MANIFEST.md) | > | ~~34~~ | QIT | One QM, many formulations | **DONE 2026-09-19** (M–L took M; three Category 3 modules, 19 pins): [`Empirical/CSD/Algorithms/CircuitFlow.lean`](../CsdLean4/Empirical/CSD/Algorithms/CircuitFlow.lean) — a circuit is a flow on the register's sector (★ `Circuit.isUnitaryLift_flow`, ★★ `Circuit.barycenter_flow_iterate`, ★★ `Circuit.epistemicMeasure_globalBasin_flow_iterate`; `IsUnitaryLift.iterate` in LF2); [`GroverFlow.lean`](../CsdLean4/Empirical/CSD/Algorithms/GroverFlow.lean) — ★★ `grover_flow_born` (the Grover step packaged as a product of reflections, unitary); [`ShorFlow.lean`](../CsdLean4/Empirical/CSD/Algorithms/ShorFlow.lean) — ★★ `shor_flow_born_count` (oracle = permutation matrix, inverse QFT = Kronecker factor). Rule-of-two fold on the way: `Projectivization.norm_toEuclideanLin_unitary` now lives in `Projectivization/TransitionProbability.lean`, index-generic; the copies in `ProjectiveSpaceFubiniStudyMass.lean` and `Thermo/Equilibration.lean` are gone. *Was:* **Σ-twins for the algorithms** — Grover's iterate and Shor's period-finding circuit as composed `Σ`-flows on the register's joint sector, through the unconditional unitary lift (`isUnitaryLift_of_smul`, `registerFlow`) and `barycenter_flow` / `traceRight_barycenter_flow`: each algorithm's QM statement (`Empirical/QM/Algorithms/*`) restated as a property of the flowed preparation's density operator, with the readout as the de-isolation channel. The pattern is `Empirical/CSD/QEC/RegisterFlow.lean` (QEC on `Σ` end to end); the register's tensor structure enters as Posit 7 (local tomography), as everywhere. Closes the algorithm link of the postulates-to-quantum-computer chain (#35). | ~~**M–L**~~ done | Medium — the one link of that chain still stated on the QM side only, now closed | this row | > | ~~35~~ | Docs | Reviewer-proofing | **DONE 2026-09-19** ([`docs/FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md`](../docs/FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md); S–M took S): eleven links, each with its theorems and its seam, closing with the four kinds of seam and where each is tracked; 73 named declarations and 65 links checked against the tree at landing; registered with the epistemic scan and pointed at from `docs/PATHS.md`. *Was:* **The postulates-to-quantum-computer narrative** — one page per link of the chain (arena; Born; Schrödinger; observables and update; records; composites; density operators and channels; gates; algorithms; error correction; arithmetic and resource counts), each naming its theorems and stating its seam: the posits it rests on (Posits 1, 2, 4, 6, 7 of [`POSITS.md`](POSITS.md)) and the residues still open at the quantum-computer end (`R-001`–`R-006`, `R-013`). Written from [`reconstruction-status.md`](reconstruction-status.md), [`qit-chain-scoping.md`](qit-chain-scoping.md), [`POSITS.md`](POSITS.md) and [`EMPIRICAL.md`](../EMPIRICAL.md); makes no new claim, so the guards that read those ledgers read it too. | **S–M** | High — the reader's path the ledgers only imply | this row | > | ~~36~~ | QIT | One QM, many formulations | **DONE 2026-09-20 — all three rungs** (residues are #40–#44). *new (2026-09-19)* **Feynman's formulation, in three rungs** (surveyed 2026-09-19). ~~**(a) The discrete sum over paths.**~~ **(a) DONE 2026-09-19** (S–M took S; two Category 1 modules, 9 pins): [`LinearAlgebra/Matrix/PathSum.lean`](../CsdLean4/Mathlib/LinearAlgebra/Matrix/PathSum.lean) — ★ `Matrix.pow_succ_apply_eq_sum_pathWeight`, the `(i, j)` entry of `M ^ (n + 1)` as the sum over every choice of `n` intermediate states of the product of the entries along the path (`pathWeight`, `pathWeight_cons`; Mathlib had only the adjacency-matrix walk count); [`Analysis/Matrix/SumOverPaths.lean`](../CsdLean4/Mathlib/Analysis/Matrix/SumOverPaths.lean) — ★★ `Matrix.exp_add_apply_tendsto_sum_pathWeight`: for skew-Hermitian `A`, `B` the matrix element of `exp (A + B)` is the limit of the sum over paths of products of one-step amplitudes of `trotterStep A B (n + 1) = exp (A/(n+1)) * exp (B/(n+1))`, read off ★★ `trotter_skew` entry by entry (`tendsto_apply_of_tendsto`); and ★★ `Matrix.exp_neg_I_mul_smul_add_apply_tendsto_sum_pathWeight`, the same for the propagator `exp (-(i t)(H₁ + H₂))` of a Hamiltonian split into two Hermitian parts, the one-step amplitudes being the short-time propagators. Feynman's formulation as a theorem rather than a postulate, at the finite dimension the reconstruction works in; exact, no rate restated entrywise, no continuum. *Was:* for a splitting `H = A + B` the amplitude between two basis states under `exp (-i H T)` is the `n → ∞` limit of a sum over discrete paths — sequences of intermediate basis states — of products of one-step amplitudes; the interference picture the algorithm link of #35 wants. **(b) The diagrammatic rung** (CV). The lattice propagator ★★ `freeTwoPoint_eq` and Wick's four-point theorem at distinct times ★★ `timeFourPoint_wick` ([`Propagator.lean`](../CsdLean4/CV/Propagator.lean), [`Wick.lean`](../CsdLean4/CV/Wick.lean)) are landed; diagrams need three more pieces — ~~the general `2n`-point Wick pairing sum~~ **(b)(i) DONE 2026-09-19** ([`CV/WickGeneral.lean`](../CsdLean4/CV/WickGeneral.lean), 6 pins; M–L took S–M): ★★ `wordOp_vac_eq_prod_wickMoment`, Wick at **equal time for an arbitrary word** of quadratures, any length, modes and interleaving, below the ladder's threshold — the sum over pairings in multiplicity form `∏_k (count k − 1)‼ / 2^{count k / 2}`, odd counts vanishing; the matching-indexed sum and the time-separated general formula (whose pairings carry the stroboscopic kernels) are not stated — ~~the Dyson expansion of an interaction term~~ **(b)(ii) DONE 2026-09-19** ([`Analysis/Matrix/DysonSeries.lean`](../CsdLean4/Mathlib/Analysis/Matrix/DysonSeries.lean), Category 1, 10 pins; M–L took M): ★★ `Matrix.hasSum_dysonTerm` — for skew-Hermitian `A`, `B` and `0 ≤ t` the propagator `exp (t (A + B))` is the sum of the Dyson series `D₀ = exp (tA)`, `Dₙ₊₁(t) = exp (tA) ∫₀ᵗ exp (−sA) B Dₙ(s) ds`, with ★ `exp_add_sub_exp_eq` (the Duhamel identity as an integral, by the fundamental theorem of calculus on the interpolant), ★ `norm_dysonTerm_le` and ★ `norm_dysonRemainder_le` (`(‖B‖ t)ⁿ/n!`, the truncation error), and ★★ `hasSum_dysonTerm_of_isHermitian` for a Hamiltonian split into Hermitian parts — matrix-valued Bochner integration works at the pin without the operator-space instance gap of the Poincaré lemma, and the corpus's integral-free Duhamel bound stays as it was — ~~and the vertex bookkeeping~~ **(b)(iii) DONE 2026-09-20** ([`Analysis/Matrix/DysonVertex.lean`](../CsdLean4/Mathlib/Analysis/Matrix/DysonVertex.lean), Category 1, 12 pins; [`CV/FeynmanVertex.lean`](../CsdLean4/CV/FeynmanVertex.lean), CV-27, 9 pins; M–L took M): the bookkeeping at every order — the interaction picture `dysonTermI` with ★ `dysonTermI_succ` (the `n` vertices at ordered times, the time-ordered product), ★ `dysonTerm_sum_eq_sum_dysonTermLabelled` (an interaction that is a sum of vertex types expands into the sum over labellings of the `n` vertices), ★ `dysonTermI_succ_apply_diagonal` (old-fashioned perturbation theory in the eigenbasis: a free propagation between consecutive vertices and a sum over the intermediate state), ★ `dysonTerm_one_apply_diagonal_of_ne` (the first-order transition amplitude) — and the diagrams on the lattice: ★ `interactionPicture_neg_I_smul_fieldHamiltonian` (the interaction-picture vertex is the free Heisenberg vertex), ★★ `hasSum_dysonTerm_fieldHamiltonian` (the perturbed lattice propagator is the sum of its series), ★★ `dysonTerm_one_vac_wordOp` (the first-order vacuum diagram of any monomial vertex is the pairing sum of its own legs, by the general equal-time Wick), ★★ `dysonTerm_two_vac_quadratic` (the second-order vacuum diagrams of a quadratic vertex are the three-pairing kernel sum under the ordered double integral, by the four-point time-separated Wick — the connected bubble for `a ≠ b`, the disconnected double tadpole plus two bubbles for `Q_a²`). ~~**(b)(iv)** the general time-separated Wick theorem~~ **(b)(iv) DONE 2026-09-20** ([`CV/WickTime.lean`](../CsdLean4/CV/WickTime.lean), CV-28, 17 pins; M–L took M): ★★ `timeWord_vac_eq_wickSum` — **Wick's theorem at the cutoff for every word at every real time**: below `count k / 2 < N`, `⟨vac∣ Q_{k₁}(t₁) ⋯ Q_{kₘ}(tₘ) ∣vac⟩ = wickSum [(k₁, t₁), …, (kₘ, tₘ)]`, the sum over perfect matchings in first-contraction form (`contraction x y = δ_{kl} ½ e^{−it} e^{+is}`, the propagator line); proved by the first-contraction recursion ★ `timeWord_cons_vac` — the annihilator row of the leftmost quadrature commuted through the word (`mul_map_prod_sub_map_prod_mul`), the truncated CCR at each later insertion of its mode, the rank-one defect killed by the walk band (`timeWord_apply_eq_zero_of_lt`) below threshold; corollaries ★ `timeFourPoint_eq_wickSum` (CV-23b recovered at stroboscopic times), ★ `wickSum_map_zero` (CV-23d's `∏ wickMoment` recovered at equal time), `wickSum_eq_zero_of_odd`, and ★★ `dysonTerm_two_vac_wordOp` — the second-order vacuum diagrams of EVERY monomial vertex (`Q⁴`'s sunset and its disconnected partners) as the ordered double integral of the pairing sum of the word with its copy at the time difference. ~~**(b)(iv′)** the matching-indexed restatement~~ **(b)(iv′) DONE 2026-09-20** ([`Mathlib/Combinatorics/PairingSum.lean`](../CsdLean4/Mathlib/Combinatorics/PairingSum.lean), Category 1, 11 pins; `CV/WickTime.lean` + 2 pins; S–M took M): `Fin.IsPerfectMatching` (fixed-point-free involutions), `Fin.pairingSum` (the sum over perfect matchings of `∏_{i < σ i} c i (σ i)`), ★ `Fin.pairingSum_succ_succ` — the first-contraction recursion, by the gluing bijection between matchings of `Fin (n+2)` and pairs (partner of `0`, matching of the complement; `Fin.cons` + `Fin.insertNth`), ★ `Fin.card_isPerfectMatching_even` — `Fin (2n)` has `(2n−1)‼` perfect matchings; and on the lattice ★★ `wickSum_eq_pairingSum` / ★★ `timeWord_vac_eq_pairingSum` — Wick's theorem in the textbook form, the vacuum expectation of a word as the sum over the perfect matchings of its insertions of the product of the propagator lines. **(c) The continuum path integral** (EFT rung) — **SCOPED 2026-09-20**, [`feynman-continuum-scoping.md`](feynman-continuum-scoping.md). The real-time `∫𝒟x e^{iS}` is not a measure (Cameron 1960); the honest forms are Kac's Euclidean formula against Wiener measure and Nelson's time-sliced limit, and both are buildable at the pin **without unbounded-operator theory**: `e^{−tH}` is defined by the Duhamel series around the free semigroup, the free Schrödinger group by the L² Fourier isometry, and every Wiener statement is conditional on `hB : IsBrownianReal B P` (the pin has the predicates and the weak Markov property, not existence). Bricks: ~~**FC-0** three probes~~ **FC-0 DONE 2026-09-20** (all three pass at the pin, §7 of the note: the `L^∞` multiplier on `L²` is Mathlib's `holderL`, 29 lines; the two-point Wiener integral from `indepFun_shift`, 50 lines; Riemann sums of a continuous function, ~70 lines — local scratch files, they become the first lemmas of FC-2/3/4); ~~**FC-1**~~ **FC-1 DONE 2026-09-20** ([`Mathlib/Analysis/Semigroup/BoundedPerturbation.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/BoundedPerturbation.lean), Category 1, 16 pins; M–L took L): `IsContractionSemigroup` (strongly continuous, `‖S t‖ ≤ 1`, extended by the identity to `t ≤ 0`; no generator named), the vector-valued Dyson series with ★ `norm_dysonTerm_le`, the sum `dysonSum` with ★ `norm_dysonSum_le` `≤ e^{‖B‖t}‖ψ‖`, ★ `dysonSum_eq_add_integral` (the Duhamel equation), ★ `eq_dysonSum_of_duhamel` (uniqueness by the iterated Volterra bound), ★★ `dysonSum_add_time` (the semigroup law, by uniqueness), the bundled `IsContractionSemigroup.perturbed` with ★★ `perturbed_add`, and ★★ `tendsto_trotterStep_pow_apply` — **the Trotter product formula** `(S(t/n) e^{(t/n)B})ⁿ ψ → 𝒮(t) ψ` (one-step defect `2(‖B‖h)²e^{‖B‖h}‖χ‖ + h·ω`, the oscillation `ω` uniformly small on the compact orbit, telescoping against the semigroup law); `of_group` for the real-time (unitary) case. Three constructions of one operator, proved to agree. ~~Residue **FC-1′** (S–M)~~ **FC-1′ DONE 2026-09-20** (#40, [`Mathlib/Analysis/Semigroup/MatrixInstance.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/MatrixInstance.lean), Category 1, 9 pins; S–M took S–M): the engine on `ℂᵐ` with `S(t) = exp (t A)` — ★ `perturbed_eq_exp` (the engine's semigroup is `exp (t (A + B))`, by the matrix Duhamel identity and the engine's uniqueness), ★★ `hasSum_dysonTerm_of_engine` and ★★ `trotter_skew_of_engine` (the matrix Dyson series and `trotter_skew` back as corollaries, **with the skewness of `B` dropped**; strong convergence upgraded to norm convergence by `tendsto_of_forall_tendsto_apply`); the `O(1/n)` rate of `TrotterProduct.lean` is not recovered; ~~**FC-2**~~ **FC-2 DONE 2026-09-20** ([`Mathlib/Analysis/Semigroup/HeatSemigroup.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/HeatSemigroup.lean), Category 1, 17 pins; M took M): `heatSemigroup t = ∫ τ_y (·) dγ_t(y)`, the Bochner integral of translates in `L²(ℝ)` — contraction, ★ `heatSemigroup_add` by Gaussian convolution of measures, ★ `continuous_heatSemigroup_apply` by continuity of translation in `L²` (`Continuous.compMeasurePreservingLp`) and concentration of `γ_t` (scaling + dominated convergence), ★★ `isContractionSemigroup_heatSemigroup`; ★ `heatSemigroup_apply_ae_eq` the pointwise formula `(P_t f)(x) = ∫ f(x+y) dγ_t(y)` a.e. (pairing with indicators of finite-measure sets, Fubini, `ae_eq_of_forall_setIntegral_eq_of_sigmaFinite`), ★ `heatConv_eq_integral_brownian` `= E[f(x + B_t)]`; `potential V` (Mathlib's `holderL`), `perturbedHeat V t` = `e^{−t(H₀+V)}` as FC-1's Dyson series around `P_t`, ★ `perturbedHeat_eq` (Duhamel) and ★★ `tendsto_trotter_perturbedHeat` — **the Trotter product formula for the heat semigroup with a bounded potential**, the time-sliced Euclidean path integral in operator form; ~~**FC-3**~~ **FC-3 DONE 2026-09-20** ([`Mathlib/Probability/TimeSlicedWiener.lean`](../CsdLean4/Mathlib/Probability/TimeSlicedWiener.lean), Category 1, 8 pins; M took M): ★★ `pow_stepOp_apply_ae_eq_slicedWiener` — for every pre-Brownian motion, slice length `h > 0`, bounded measurable weight `g` and bounded measurable `f ∈ L²`, `((P_h ∘ M_g)ⁿ f)(x) = E[g(x+B_h) ⋯ g(x+B_{nh}) f(x+B_{nh})]` a.e. — Feynman's finite-slice formula in the Euclidean continuum, the continuum twin of rung (a); the freezing lemma `integral_prod_of_indepFun` (`E[Ψ(X,Y)] = E_ω E_ω' Ψ(Xω, Yω')` for independent `X`, `Y`), ★ `slicedWiener_succ` the Markov step through the shifted process (`IsPreBrownianReal.shift`, `indepFun_shift`), and the induction through `heatSemigroup_apply_ae_eq`; ~~★★ **FC-4** Feynman–Kac~~ **FC-4 DONE 2026-09-20** ([`Mathlib/Probability/FeynmanKac.lean`](../CsdLean4/Mathlib/Probability/FeynmanKac.lean), Category 1, 6 pins; M–L took L): ★★ `feynmanKac` — for `hB : IsBrownianReal B P` (a.s. continuous paths), bounded continuous `V` and bounded measurable `f ∈ L²`, `(e^{−t(H₀+V)} f)(x) = E[exp(−∫₀ᵗ V(x+B_s) ds) · f(x+B_t)]` a.e., the left side FC-2's `perturbedHeat` (the Dyson series around `P_t`); ★ `exp_smul_neg_potential` (the exponential of a multiplication operator is multiplication by the exponential, by the operator series applied to `f` and set-integral uniqueness), `tendsto_riemannSum` (probe P3 promoted), `tendsto_slicedWiener` (the Wiener-side limit by dominated convergence along continuous paths); the two limits of the same time-sliced expression identified on every finite-measure set. **The Euclidean path integral is a Wiener integral, as Kac stated it.** Conditional on a Brownian motion (existence is Mathlib's pending Kolmogorov–Chentsov); ~~the extension from bounded `f` to `f ∈ L²` by continuity is not stated (FC-4′, S)~~ **FC-4′ DONE 2026-09-20** (#42, [`Mathlib/Probability/FeynmanKacL2.lean`](../CsdLean4/Mathlib/Probability/FeynmanKacL2.lean), Category 1, 7 pins; S took M): ★★ `feynmanKac_Lp` for **every** `g ∈ L²` — both sides continuous in `g` on finite-measure sets, agreeing on the dense simple functions; ★ `absConv_ae_eq`, `E‖g(x+B_t)‖ = (P_t|g|)(x)` a.e., bounds the Wiener side by Cauchy–Schwarz and contraction; `ae_integrable_shift` (the functional is defined for a.e. `x`); `tendsto_slicedWiener` generalised to `f` integrable along the endpoint; ~~★★ **FC-5** Nelson's theorem in operator form~~ **FC-5 DONE 2026-09-20** ([`Mathlib/Analysis/Semigroup/SchrodingerGroup.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/SchrodingerGroup.lean), Category 1, 12 pins; L took M–L): the phase group `M_{e^{−itκ}}` of a real measurable symbol `κ`, strongly continuous through weak continuity and polarisation (no `L²` dominated convergence); the unitary group `U_κ = 𝓕⁻¹ M_{e^{−itκ}} 𝓕` of any dispersion relation through Mathlib's `Lp.fourierTransformₗᵢ` (`freeSchrodinger` is `κ = 2π²ξ²`, the symbol of `−½ d²/dx²` in Mathlib's Fourier convention); the interacting propagator `e^{−it(κ(D)+V)}` as FC-1's Dyson series with interaction `−i M_V`; ★ `exp_eq_potential` (the complex multiplier exponential — the potential kick); ★★ `nelson` — `(U_κ(t/n) · M_{e^{−i(t/n)V}})ⁿ ψ → e^{−it(κ(D)+V)} ψ`, both factors phase groups, the kinetic one in momentum space and the potential one in position space; ★ `norm_schrodinger_apply` (probability is conserved) and ★★ `exists_linearIsometryEquiv_schrodinger` (**the propagator is unitary**: the reversed dynamics `(−κ, −V)` is a two-sided inverse, because the inverse of the forward Trotter approximant `(U M)ⁿ` is the conjugate `M⁻¹(U⁻¹M⁻¹)ⁿM` of the reversed one and both converge strongly). **The real-time path integral is Nelson's strong limit, as a theorem.** ~~Not stated: the explicit action on Schwartz/Gaussian packets and the identification of `freeSchrodinger` with `−½ d²/dx²` on Schwartz functions (FC-5″, S–M)~~ **FC-5″ DONE 2026-09-20** (#43, [`Mathlib/Analysis/Semigroup/SchrodingerSchwartz.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/SchrodingerSchwartz.lean), Category 1, 8 pins; S–M took M): ★ `fourierGroup_toLp` (the group preserves Schwartz space, `U_κ(t)(f.toLp 2) = (𝓕⁻¹(e^{−itκ}·𝓕f)).toLp 2`, after ★ `hasTemperateGrowth_phaseFun`), ★ `kineticOp_eq_laplacian` (`H₀ = −½Δ`), ★★ `hasDerivAt_freeSchrodinger` (**the free Schrödinger equation** `∂_t U₀(t)f = −i H₀ U₀(t) f` in `L²` at every `t`); ~~the Gaussian packet is #48~~ **FC-5‴ DONE 2026-09-20** (#48, [`Mathlib/Analysis/Semigroup/GaussianPacket.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/GaussianPacket.lean), Category 1, 11 pins; M took M): the Gaussian `e^{−πax²}` as a Schwartz function and ★★ `freeSchrodingerS_gaussianS`, **the packet spreads** — `U₀(t) g_a = (1 + 2πiat)^{−1/2} g_{a/(1 + 2πiat)}`, the density of a real packet widening as `√(1 + (2πσt)²)`; **FC-2′** dimension `d` (S–M after FC-4); ~~**FC-6** the documents (S)~~ FC-6 done with FC-4 and FC-5. Only **FC-5′**, the kernel form with `e^{iS}` visible, is XL (oscillatory Fresnel kernel, outside the pin's L¹ Fourier theory) and is not recommended. Decisions D1–D4 in the note (one dimension first; `L²`; build now, conditional on Brownian motion; no FC-5′). | ~~**S–M** (a)~~ done; (b) ~~(i)~~–~~(iv′)~~ all done; (c) ~~FC-0~~–~~FC-6~~ all done 2026-09-20; residues ~~FC-1′ **S–M**~~ done, ~~FC-2′ **S–M**~~ done (took L), ~~FC-4′ **S**~~ done, ~~FC-5″ **S–M**~~ done, ~~Gaussian packet #48 **M**~~ done (FC-5′ **XL**, not recommended) | Medium–high for (a), delivered; low for (c) | this row | > | ~~37~~ | Docs / Positioning | Library owed / upstream | *new (2026-09-19)* **The Mathlib pull-request sequence**, one at a time, each written in this repository first and living here until it builds green. ~~Retired 2026-08-06~~ **reinstated 2026-09-19** (`CONVENTIONS.md` §7). **Pull request 1 is backported and green here** (`future-work.md` HY-7): the quotient topology on `Projectivization`, the pull-request text verbatim, plus the two continuity instances staged in [`Topology/Algebra/MulAction.lean`](../CsdLean4/Mathlib/Topology/Algebra/MulAction.lean). What remains is outside this repository and needs the author's approval at each step: push the branch to the fork, open the pull request, answer review. ▸ **DONE 2026-10-23 (author's report): pull request 1 is merged upstream** — [leanprover-community/mathlib4#44117](https://github.com/leanprover-community/mathlib4/pull/44117), *feat(Topology/Algebra): quotient topology on projective space and open quotient map*, merged by Bors. The backport stays here until a pin move carries the upstream copy, when `CONVENTIONS.md` §7's deprecation rule applies to the staged file; pull request 2 and the later candidates are opened at the author's pace, outside this table. **Pull request 2** is the rest of the same file — Hausdorffness and compactness of `ℙ K V` for `[RCLike K]` and finite-dimensional normed `V` — which stays here until 1 lands. Candidates after that are the other self-contained staged files named in the struck B6 row. | ~~**S** each~~ done (PR 1 merged) | Medium — the corpus's code in the library that will keep it, and the 461-to-82 lesson applies to every other staged file | this row; `CONVENTIONS.md` §7; [`future-work.md`](future-work.md) HY-7 | > | ~~38~~ | Foundations | Records→spacetime | **ROW CLOSED 2026-10-02 — every brick it named is built and its one open item, the `ST-3` decision, was taken by the author.** The ledger of it: the scoping note (2026-09-19), `ST-1` the causal accessibility cone ([`CV/RecordInfluence.lean`](../CsdLean4/CV/RecordInfluence.lean), 2026-10-01), `ST-2`, `ST-4`, the **event order** ([`CV/RecordCausalOrder.lean`](../CsdLean4/CV/RecordCausalOrder.lean), ★★★ `isPartialOrder_causalPrecedes` — the record events of a coupling graph are a **causal set**), and then `ST-3` in three parts once the author chose **candidate (5)**: its structure as **#99** ([`RecordLayer/MacroProjection.lean`](../CsdLean4/RecordLayer/MacroProjection.lean)), its instance as **#102** ([`RecordLayer/RecordMacrostate.lean`](../CsdLean4/RecordLayer/RecordMacrostate.lean), the record string, with no-signalling and the Born chain law both verified for it), and its stability as **#103** ([`RecordLayer/MacrostateStability.lean`](../CsdLean4/RecordLayer/MacrostateStability.lean), the macroscopic law exactly invariant under a record write and the pointwise string robust outside measure `k·N·δ`). **THE ROW'S FINDING IS CLOSED UNCHANGED, AND IT IS A NEGATIVE ONE: there is still no Lean shape for emergence proper.** What the row produced is (a) the causal shape of an **assumed** coupling graph — no metric, no signature, no continuum, no derivation of `E` — and (b) a macroscopic coordinate on `Σ` with the record statistics factoring through it. Neither is spacetime, and `M` is still a set of **labels** with no geometry. What follows the row is carried by its successors, each a numbered row and none hidden: **#39** (the `Σ`-local / spacetime-nonlocal restatement), **#98** (the two cones — reconciling this row's influence cone with `DispersionEarned`'s kinematic one, which is where causal structure would reach `M`), **#100** (effective stochasticity), **#104** (stability under motion of the base) and **#105** (where the overwhelming majority comes from). *As the row stood:* **Scoping note DONE 2026-09-19** ([`records-to-spacetime-scoping.md`](records-to-spacetime-scoping.md); M took S–M): what a record is in Lean, four candidate relations, the arena's limits, the constraining theorems, and four priced bricks `ST-1`–`ST-4`. **Finding: no Lean shape for emergence proper**; the missing piece is which coarse-graining projection defines the macroscopic coordinates (Paper D §5.2), a decision for the author, not a lemma. ~~The one Lean brick that is honest and new is `ST-1`, the influence preorder on records from the coupling-graph light cone (S–M, relative to an assumed graph);~~ **`ST-1` DONE 2026-10-01** ([`RecordInfluence.lean`](../CsdLean4/CV/RecordInfluence.lean), 254 lines, 11 pins; S–M took S–M): `Influences E S₁ S₂ n := S₂ ⊆ graphBall E S₁ n` with ★ `influences_refl`, ★ `influences_mono` and ★★ `influences_trans` — **the periods add**, which needed the two cone lemmas the CV chain had not stated (`graphBall_add`, `graphBall_subset_of_subset`). **TWO PRECISIONS ADDED 2026-10-02 after an external read of the live repository, both on wording this row and the module had loose:** (i) at a *fixed* budget the relation is **not** transitive — two reaches within `n` compose into one within `2n` — so it is a **graded reachability relation**, and the word *preorder* is now earned by the object it belongs to, `EventuallyInfluences` (`∃ n`), with ★ `isPreorder_eventuallyInfluences`; the module's title no longer says "preorder"; (ii) `Influences` is **permitted** influence, not **demonstrated** influence — every theorem built on it has the form *outside the cone ⇒ nothing happens*, and the converse (inside the cone, some influence occurs) is neither proved nor true in general, since a coupling along the path may vanish. That propagation is not fictional is the separate `spreadKick_not_supportedOn`; there is no equivalence between graph reachability and operational influenceability anywhere in the corpus, and the module now says so and calls the relation the **causal accessibility cone**; ★★ `commute_record_of_spacelike` — spacelike records are **jointly measurable**; ★★★ `arenaObs_heisenberg_kick_of_spacelike` and ★★★ `recordStroke_heisenberg_comm_kick_of_spacelike` — **a record outside the cone cannot be steered, with no error term at all** (the discrete-period companion of `record_lightcone`'s Lieb–Robinson bound: the evolved observable's support is inside the cone on the nose); and ★★★ `map_recordStroke₂_prod` — **the joint record law is the medium's product law**, for any translation-invariant channel law, so the fibre contributes neither correlation nor bias. **THE SCOPE HELD, AND THE ROW'S FINDING IS UNCHANGED:** this is the causal shape the *assumed* graph defines — no metric, no signature, no continuum, no derivation of `E` — and the factoring theorem is about the **medium**, not about the state: two spacelike records being statistically independent would also need the base readings to be independent, which needs a tensor factorisation of the field space over a mode partition that the corpus does not have. What is still missing for emergence proper is the coarse-graining projection (Paper D §5.2), a decision for the author. **TWO PIECES NAMED AND PRICED 2026-10-02, both recorded here rather than opened as rows.** **(i) THE EVENT ORDER — DONE 2026-10-02** ([`RecordCausalOrder.lean`](../CsdLean4/CV/RecordCausalOrder.lean), 7 pins; S–M took S): `RecordEvent` is a read region with the period at which its record is written, `CausalPrecedes E e₁ e₂ := e₁.time ≤ e₂.time ∧ Influences E e₁.region e₂.region (e₂.time − e₁.time)`, and ★★★ `isPartialOrder_causalPrecedes`: **the record events of a coupling graph carry a partial order — a causal set**, reflexive, transitive (★★ `causalPrecedes_trans`, the legs' periods adding exactly, `graphBall_add` with the ℕ arithmetic discharged) and **antisymmetric** (★★ `causalPrecedes_antisymm`: equal times force a zero budget, and at zero the cone is the region itself). The grading is gone because the duration is inside the relation. `EventSpacelike` is unorderedness, with ★★ `eventSpacelike_of_spacelike` — **geometry gives order** — and ★★★ `arenaObs_kick_of_eventSpacelike`: the same hypothesis that orders two events also protects the first's record from any unitary kick on the second's region, exactly. **THREE THINGS NOT CLAIMED, in the header:** the order is induced by an **assumed** `E` and is not derived from the records; it is *permitted* influence, inherited from `ST-1`; and `EventSpacelike` is **weaker** than `Spacelike` (it compares cones of different radii), which is why the dynamical theorem carries the geometric hypothesis and not the order one. *As recorded before it was built:* the cheapest brick left on this row and neutral with respect to the `ST-3` decision below. `ST-1` relates *regions* with a duration supplied separately; the object a causal structure wants is an **event**, a read region with an occurrence index, ordered by `e₁ ⪯ e₂ := e₁.t ≤ e₂.t ∧ e₂.R ⊆ graphBall E e₁.R (e₂.t − e₁.t)`. Then the duration is *inside* the relation, so this is a genuine preorder with no grading — a **causal set** on record events — and `graphBall_add` plus `graphBall_subset_of_subset` give transitivity with only the ℕ budget arithmetic to check. It would also carry the three unsteerability theorems to event pairs (spacelike = neither precedes the other). **The two caveats of `ST-1` survive unchanged**: it is *permitted* influence, and it is the shape of an **assumed** `E`. **(ii) THE `ST-3` DECISION'S OPTION SET, written down so it can be acted on.** The scoping note says `ST-3` waits on a decision but does not name the candidates; four are worth putting to the author, each with what it would buy: **(1) the record histogram** — `π'` sends a `Σ`-point to the multiset of records realised in a window (which contexts fired, how often); the record statistics then factor through it by construction, and `ST-1`'s relation descends to it; **(2) the influence quotient** — `π'` maps to the quotient of the coupling graph by the event order of (i), so the macroscopic coordinates **are** the causal order; tightest link to what is already proved; **(3) mode occupancy** — the occupancy vector of a fixed mode partition above a threshold, which connects to the CV rung; **(4) binned entanglement distance** — the finite metric structure `ST-2`'s `entDist` induces on a fixed family of sectors, with `ST-2`'s own caveat (a separation functional, not a proved metric) attached. **THE FACTORING REQUIREMENT CONSTRAINS THE FIVE, AND ROW 99 MAKES THAT PRECISE (2026-10-02).** ★ `factorsThrough_iff` says an event factors through `π'` exactly when `π'` cannot distinguish two microstates the event separates. A record basin is `globalBasin c i = {x \| x.2.1 ∈ circleCell (c.rate x.1) i}` — it depends on the **base** `x.1` through the rate field and on the **first fibre angle** `x.2.1`. So **`π'` may coarse-grain everything except the record outcomes of the family whose statistics it is supposed to carry**: it must determine, for each context in the family, which cell the fibre angle lies in. Reading the five against that: **(1)** keeps the outcomes but a histogram forgets their **order**, so the step-resolved basins of `csd_nstep_born` do *not* factor through it — only exchangeable statistics do; **(2)** as stated is not even a map out of `Σ` — the influence quotient is a coordinate on *regions and periods*, i.e. on contexts, and carries no information about which outcome occurred, so it can only be **extra structure on `M`**, not `π'` itself; **(3)** and **(4)** are honest functions of the microstate but likewise forget the outcomes (occupancy is a base reading, `entDist` is a property of the state), so neither can be the whole of `π'`; **(5)** is the only one that *defines itself* as a factoring coordinate, and row 99's `patternProjection` is already its skeleton — take the pattern of a finite family of coarse record events and the factoring plus both invariance theorems come for free, leaving **stability under the flow** as the physics. **RECOMMENDATION (2026-10-02, the author decides): (5), built as the pattern of an ORDER-RETAINING finite family of coarse record readings** — (3)'s occupancy is the natural content of those readings and (2)'s causal order is the natural extra structure on `M`, so the recommendation absorbs them rather than competing with them. **(5) THERMODYNAMIC MACROSTATES OF RECORDS** (added 2026-10-02): `π'` is the coarse-graining whose cells are *macroscopically stable equivalence classes of records* — the spacetime coordinates being thermodynamic macrostates rather than a chosen partition. This is the best-motivated candidate because it makes `π'` **selected** rather than **picked**: stability under the dynamics does the choosing. What it would need that the corpus has: the second law for a specified coarse-graining ([`Thermo/SigmaSecondLaw.lean`](../CsdLean4/Thermo/SigmaSecondLaw.lean), TH2), the Gibbs and Landauer layers (TH3, TH4), the second moments of the reduced state ([`Thermo/ReducedSecondMoment.lean`](../CsdLean4/Thermo/ReducedSecondMoment.lean)) and the conditional time-averaged equilibration of [`Thermo/Equilibration.lean`](../CsdLean4/Thermo/Equilibration.lean). **CORRECTION 2026-10-02, same day: the sentence that stood here was wrong.** It said the corpus does not have concentration. It does — ★★ `fs_chebyshev_concentration` in [`Thermo/CanonicalTypicality.lean`](../CsdLean4/Thermo/CanonicalTypicality.lean), a **polynomial-rate** concentration from the exact second moments, so the diagonal statistics of a single `μ_FS` sample do concentrate, not merely average. What is parked is the **exponential (Lévy) rate** (row 24's `TH-1`, standing decision: leave, scoped in [`th1-concentration-scoping.md`](th1-concentration-scoping.md)), which a large-system limit would want and which the polynomial tier does not block. So candidate (5)'s real missing inputs are **not** concentration: they are the macrostate equivalence itself (this candidate's own content) and **effective stochasticity**, which is now **row 100**. **A COROLLARY WORTH RECORDING: the thermodynamic *arrow* on the event order funnels into this same decision.** The causal order of [`RecordCausalOrder.lean`](../CsdLean4/CV/RecordCausalOrder.lean) says what can influence what, not which direction is future; orienting it by increasing coarse-grained record entropy is a conjecture the corpus cannot state today, because its second law is for a *specified* coarse-graining and the specification wanted is `π'` itself. **The structure half of `ST-3` needs none of these** — it is definitional (S), is now **row 99**, and can land before the choice; only the *instance* waits on the author. **`ST-2` DONE 2026-10-01** at the author's request — the note prices it as a research direction that does not feed the record layer, and that is still true — ([`EntanglementDistance.lean`](../CsdLean4/CV/EntanglementDistance.lean), 221 lines, 10 pins; M–L took M): `mutualInfo` off the two partial traces with ★★ `mutualInfo_nonneg` (subadditivity), ★★★ `mutualInfo_kronecker` — **a product state carries none, exactly** — ★★★ `mutualInfo_pureDensity` — **on a pure state `I = 2 S(ρ_A)`, the mutual information *is* the entanglement entropy doubled** (needing the new `pureDensity_mul_self`) — and the arena reading ★★★ `mutualInfo_compositeDM_join` with ★★ `entDist_mutualInfo_join`: **the two sectors of a product point are infinitely far apart**. The functional is `entDist I = −log I` with `I ≤ 0 ↦ ⊤`, ★ antitone (more correlation, less distance) and ★ zero from one nat up. ★★ `mutualInfoTri_mono` is **monotonicity under sector enlargement**, strong subadditivity rearranged — and it carries the corpus's explicit `hDPI` hypothesis rather than assuming SSA silently. **TWO THINGS NOT CLAIMED:** `entDist` is a **separation functional, not a proved metric** (no triangle inequality, here or anywhere in the corpus); and no numerical value is computed for any entangled ray — the entangled case is covered only as twice a marginal entropy. The `ℝ≥0∞` bookkeeping the note warned about was handled by the explicit `⊤` branch rather than by `Real.log 0 = 0`, which would have made unentangled sectors *coincident* instead of infinitely distant; ~~`ST-4` (the reading into `POSITS.md` and the charter as a conjecture) can land now (S)~~ **`ST-4` DONE 2026-09-19** (`POSITS.md` `C-1`, the charter's not-goal bullet, the narrative page; and the note's overstatement about finite compact arenas corrected in place at the author's word). *The row as opened:* **Records to spacetime.** The programme's next ontological step: spacetime structure arising from the record layer rather than being assumed alongside it. **Nothing of this is in Lean and nothing is claimed**; the corpus's spacetime-adjacent content today is the CV/EFT rung (locality, cones, Lieb–Robinson bounds, propagators), which is the isolated-piece image and not the record-layer story. First deliverable is a scoping note on the model of [`qit-chain-scoping.md`](qit-chain-scoping.md): what a record is, what relation between records is supposed to carry metric or causal structure, what the finite arena can and cannot support, and which existing theorems constrain the answer (`RecordLayer/`, the two-time and composite record closures, `CV/DynamicalLocality.lean`, `CV/LiebRobinson.lean`). Only after the note can a first brick be priced; it may turn out there is no Lean shape at this stage, and the note should say so if that is the finding. | ~~scoping note **M**~~ done; ~~`ST-1` **S–M**~~ done, ~~`ST-2` **M–L**~~ done (M), ~~`ST-3` gated~~ **decision taken 2026-10-02 (candidate (5)) and all three halves done — #99, #102, #103**, ~~`ST-4` **S**~~ done | Highest in the programme's own terms, and speculative — it is the reason the record layer matters beyond measurement | this row; [`records-to-spacetime-scoping.md`](records-to-spacetime-scoping.md) | > | 39 | Foundations | Records→spacetime | ▸ *2026-10-30:* **the author's fourth brief, #241–#248** (out of this row): two physical observers in relative motion recover a common 1+1-dimensional flat spacetime from their records; steps 1–2 (#241, #242) first, their outcome deciding whether the later steps run on the present dynamics. **Scoped 2026-09-19 with #38** ([`records-to-spacetime-scoping.md`](records-to-spacetime-scoping.md) §6): restated as a conjecture about one thing only, whether the tensor cut is emergent; relative to the cut, bounded influence (the light-cone and Lieb–Robinson theorems) and non-signalling Bell-violating records are already theorems, and no local explanation is possible (`bell_general_separation`, `general_ks_noNonContextualValuation`), so the emergent description must be nonlocal. "Local in `Σ`" can only mean manifold-locality of the flow, since a joint point is not a pair (`segre_not_surjective`). Not refuted, not established; its honest one-sentence form is `POSITS.md` `C-1` (`ST-4`, DONE 2026-09-19). ▸ *2026-10-22:* **the clock-and-echo programme, #183–#189** (out of this row; the author's brief split into seven priced rows): operational time as a tick count of physical records, radar coordinates from emission/return readings, the exchange's events in `CausalPrecedes`, a capstone (#187) that says what it does not claim — no metric, no volume element, no universal `c`. ~~#183~~–~~#186~~ landed the same day: the two-site hop as a `FieldStructuredFlow` with exact transfer and Rabi return (~~#183~~), radar coordinates (~~#184~~), the record clock (~~#185~~), and the common arena with one two-time measure-preserving evolution, three message-bearing records, occurrences mapped to `RecordEvent`s labelled by the physical clock, and `emission ⪯ reception ⪯ return` (~~#186~~); the capstone ~~#187~~ `operational_geometry` (`RecordLayer/OperationalGeometry.lean`) assembles them: radar coordinates from the physical clock's readings, reproduced by every cycle, with the causal chain and bounded errors — the first statement in the corpus of operational time and separation from physical records, on its own claim surface (§8.3b); ~~#188~~ (two clocks: colocated agreement under each clock's own calibration, linear drift when miscalibrated, the synchronised clock at `B` reading `A`'s radar time, and clock time `2` ticks against longest chain `≥ 3`) and ~~#189~~ (union bound on failure, the exact witness valid, no tolerance in the arc write, `τ_R ≥ 2 τ_hop` sufficient) complete the brief's seven rows; the `L`-site chain ~~#190~~ (2026-10-23) adds the first network statement: a `Schedule` of field-structured flows, exact transfer along `L` sites with phase `(−i)^{L−1}`, and **clock delay = graph distance** on the path graph (`clock_delay_eq_graphDist`), with `causalDist ≥ 2k` against `k` elapsed ticks. ▸ *2026-10-23:* **the author's second brief, split into #191–#197**: the occurrence registry with provenance and registers on the chain (~~#191~~, done 2026-10-23: `RecordOccurrence.lean`, the site events of the chain are forgetful images of actual writes), chronological against causal precedence (~~#192~~, done 2026-10-23: `chronoDist ≤ elapsed periods`, `= m` on the chain, the inflation steps are same-period steps), operational duration (~~#193~~, done 2026-10-23: clock duration = chrono distance × tick on the chain, `causalDist ≥ 2m`), **one internally activated write as the research gate (~~#194~~, done 2026-10-23: the carrier-activated write is constructed — detuned Rabi carrier, autonomous measure-preserving dynamics, `θ̇₁ = −(2g/ħ) f'(θ₂) n_B`, the write as an `ActualOccurrence`, finite retention, a positive-measure transition region for every horizon)**, its Hamiltonian generation (~~#195~~, done 2026-10-24: `activeFlow` is the Hamiltonian flow of `−(2/ħ)(J⟨X⟩ + g f(θ₂) n_B)` on `Σ`, the pointer rate forced to `−(2g/ħ) f'(θ₂) n_B`), the coupled reception (~~#196~~, done 2026-10-24: loss quadratic in `Δ`, the thresholded bit survives `|Δ| ≤ 2J`, emission ⪯ reception on the hop graph with the arena clock) and radar with a detector (~~#197~~, done 2026-10-24: one budget `T + retention + |detectorDelay| + πħΔ²/(16J³)` on radar time, separation and distance; a detector delay `d` shifts both coordinates by `d/2`) — **the second brief #191–#197 is complete**. ▸ *2026-10-24:* **the author's third brief, split into #198–#205**: the exchange with both sites detuned (~~#198~~, done 2026-10-24: the flow = sector phase × detuned exchange in `Δ_B − Δ_A`; the carrier returns to `A` as a ray at `πħ/Ω'`), the autonomous echo arena with the emission and the return as two arcs of one register at `A` (~~#199~~, done 2026-10-24: three `ActualOccurrence`s on one arena, Liouville by three skew steps, return arc after and disjoint from the emission arc under `3 r_A λ ≤ ω`), round-trip radar from `A`'s records (~~#200~~, done 2026-10-24: both timestamps `A`'s clock at `A`'s two writes, budget `T + window + emitEnd + |returnDelay| + 2 transferError` against `(0, 2τ_hop)`), the drift pointer as an autonomous hand clock (~~#201~~, done 2026-10-24: `K` arcs, `τ_tick = 2/(Kκ)` derived, resolution `τ_tick + 1/(2ω)`, each tick an `ActualOccurrence` written by the carrier), the clock-and-echo model on one arena — the brief's one exact deliverable (~~#202~~, done 2026-10-24, both halves: one autonomous arena, timestamps = the hand's readings at `A`'s two records, derived budget; the echo is the Hamiltonian flow of its energy on `ℂℙ³ × T² × T²`; ~~#212~~ couples the clock to the echo, done 2026-10-29, its generation ~~#225~~ done 2026-10-29), geometry across three stations and two routes (~~#203~~, done 2026-10-27: corrected radar separation always symmetric, triangle up to the relay delay, a metric with free relays; raw, fixed-route and uncorrected-relay separations proved not to be), detector-class stability (~~#204~~, done 2026-10-27: corrected separations of two detectors agree up to the sum of their corrected budgets; the stable regime for detuning `≤ δ`, window `≤ w`, clock `≤ T₀`), a common effective cone (#205, research; first brick done 2026-10-27: per stage the Lieb–Robinson cone is the hop-time cone widened by the universal factor `2πe`, and packaged as one flow on `n` edges it widens by `n/2` to `n` more — residue ~~#213~~, done 2026-10-27: a local Lieb–Robinson velocity `8eΔγ`, so on the chain the bound's cone is `8πe` hop-time cones wide at every length). **THE TWO CONES, NOW ROW 98 (M–L).** The corpus contains two cone structures that have never been related: the **record cone** from the coupling graph and its dynamics ([`RecordInfluence.lean`](../CsdLean4/CV/RecordInfluence.lean), [`FibredArenaBridge.lean`](../CsdLean4/CV/FibredArenaBridge.lean)'s `record_lightcone` with its Lieb–Robinson bound), and the **kinematic Lorentz cone** of [`DispersionEarned.lean`](../CsdLean4/CV/DispersionEarned.lean), where `cone_preserving_is_boost` earns the boost form and boost covariance with `m > 0` selects `ω(p) = √(p² + m²)`. That module says explicitly that it does **not** identify its light cone with the dynamical one. What is reachable is a **containment**: the record cone sits inside a kinematic cone whose slope is set by the Lieb–Robinson velocity, which is a comparison of two bounds and nothing more (M–L). What is **not** reachable at this pin is an *identification* of the two cones, which needs a continuum limit of the graph dynamics the corpus does not have — and the containment must not be read as one. *The row as opened:* **Σ-local, spacetime-nonlocal.** The reading under which entangled correlations are local in `Σ` and appear nonlocal only in the emergent spacetime of #38. **Conjecture, not a result**: the corpus proves the Bell separation, the CGLMP no-local-hidden-variable table, no-signalling and contextuality, and proves that a composite's local record weights are the reduced state's Born weights (`CV/EntangledWeights.lean`, `RecordLayer/EntangledRecord.lean`); none of that establishes the reading, and CSD is contextual and non-local by construction, so the no-go theorems constrain it rather than support it. Depends on #38 for the meaning of "spacetime" in the sentence. First deliverable is the same scoping note's second half: what would have to be true, and what would refute it. ▸ **AND IT NOW HAS A TESTABLE NARROWING, 2026-10-10: row 131** — one finite two-wing experiment required to satisfy record stability, causal separation, Bell correlations and no-signalling in the *same* dynamical model, audited for compatibility in [`two-wing-experiment-scoping.md`](two-wing-experiment-scoping.md). It does **not** establish `C-1` (the adjacency is still supplied), but it is the first thing on this row with a price rather than a conjecture. | ~~scoping note **S–M**~~ done (with #38); ~~`ST-4` **S**~~ done | High if #38 acquires a shape, otherwise positioning | this row; [`records-to-spacetime-scoping.md`](records-to-spacetime-scoping.md) §6 | > | ~~40~~ | QIT | One QM, many formulations | ~~*new (2026-09-20)*~~ **FC-1′ DONE 2026-09-20** ([`Mathlib/Analysis/Semigroup/MatrixInstance.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/MatrixInstance.lean), Category 1, 9 pins; S–M took S–M): the engine on `ℂᵐ` with `S(t) = exp (t A)`, `A` skew-Hermitian, `B` **any** matrix — `engine_dysonTerm_eq` (the engine's Dyson terms are the matrix ones applied to `ψ`), ★ `perturbed_eq_exp` (the engine's semigroup is `exp (t (A + B))`: the matrix Duhamel identity is the one analytic input, the engine's uniqueness does the rest), ★★ `hasSum_dysonTerm_of_engine` and ★★ `trotter_skew_of_engine` — `Matrix.hasSum_dysonTerm` and `TrotterProduct.trotter_skew` back as corollaries, verbatim, with the skewness of `B` no longer needed; strong convergence is norm convergence in finite dimension (`tendsto_of_forall_tendsto_apply`). Not recovered: the `O(1/n)` rate of `TrotterProduct.lean` (the engine's Trotter formula is a strong limit without a rate). **The rule of two holds: the matrix modules are the engine's instance.** | ~~**S–M**~~ done | Low–medium | #36 (c); [`feynman-continuum-scoping.md`](feynman-continuum-scoping.md) §7 | > | ~~41~~ | QIT | One QM, many formulations | ~~*new (2026-09-20)*~~ **FC-2′ — dimension `d` — DONE 2026-09-20.** **(a) The real-time half DONE 2026-09-20**: `Analysis/Semigroup/SchrodingerGroup.lean` and `SchrodingerSchwartz.lean` generalised **in place** from `ℝ` to any finite-dimensional real inner product space `E` (`ℝᵈ` in particular) — the phase group, the unitary group of a dispersion relation, `freeSymbol ξ = 2π²‖ξ‖²`, Nelson's formula, the isometry and unitarity of the propagator, the Schwartz-space action, `H₀ = −½Δ` and the Schrödinger equation now all hold on `L²(E)`; `mult` is the generic multiplier (`mult_eq_potential : mult = HeatSemigroup.potential` on `ℝ`, `rfl`), `instNontrivialL2` is the indicator of the unit ball; every pin unchanged (same names, more general statements). **(b) The Euclidean half DONE 2026-09-20** (L took L): new [`Mathlib/Probability/BrownianVec.lean`](../CsdLean4/Mathlib/Probability/BrownianVec.lean) (Category 1, 7 pins) — `IsPreBrownianVec`/`IsBrownianVec`, a Brownian motion in `ℝᵈ` as `d` jointly independent real ones; ★ `indepFun_pi_of_iIndepFun` (**independent vectors of independent pairs**, through Mathlib's `measurePreserving_arrowProdEquivProdArrow`); `Measure.pi_absolutelyContinuous_pi` (a product of absolutely continuous measures is absolutely continuous); the product Gaussian `gaussianVec`, its characteristic function and absolute continuity; `hasLaw_eval`, `shift`, ★ `indepFun_shift` (the weak Markov property in `ℝᵈ`). Then `HeatSemigroup.lean`, `TimeSlicedWiener.lean`, `FeynmanKac.lean`, `FeynmanKacL2.lean` generalised **in place** to `EuclideanSpace ℝ ι`: the Gaussian's convolution and scaling laws now come from characteristic functions (`Measure.ext_of_charFun`, `charFun_conv`, `charFun_map_smul`, `charFun_pi`) instead of the one-dimensional `gaussianReal_conv_gaussianReal`; every pin unchanged, every statement more general. **Feynman–Kac and Nelson now meet on the same `L²(ℝᵈ)`.** | (a) ~~**S–M**~~ done; (b) ~~**L**~~ done | Low–medium | #36 (c); [`feynman-continuum-scoping.md`](feynman-continuum-scoping.md) §8 D1 | > | ~~42~~ | QIT | One QM, many formulations | ~~*new (2026-09-20)*~~ **FC-4′ DONE 2026-09-20** ([`Mathlib/Probability/FeynmanKacL2.lean`](../CsdLean4/Mathlib/Probability/FeynmanKacL2.lean), Category 1, 7 pins; S took M — the continuity of the Wiener side is a real argument): ★★ `feynmanKac_Lp`, `(e^{−t(H₀+V)} g)(x) = E[exp(−∫₀ᵗ V(x+B_s) ds) g(x+B_t)]` a.e. for **every** `g ∈ L²(ℝ)`. The functional is defined for a.e. `x` (`ae_integrable_shift`: Fubini on the Gaussian shift integrability of `HeatSemigroup.lean`, pushed through the law of `B_t`); ★ `absConv_ae_eq`, `E‖g(x+B_t)‖ = (P_t |g|)(x)` a.e., gives `‖∫_A fk g‖ ≤ e^{t‖V‖} ‖1_A‖ ‖g‖` (`norm_setIntegral_fkFunctional_le`); the operator side is bounded by `‖1_A‖ ‖e^{−t(H₀+V)}‖ ‖g‖`; both agree on the dense simple functions (`Lp.simpleFunc.denseRange`), so the set integrals agree. Measurability of the functional from `tendsto_slicedWiener`, generalised in `FeynmanKac.lean` to `f` integrable along the endpoint (`tendsto_riemann_path` extracted). | ~~**S**~~ done (M) | Low | #36 (c) | > | ~~43~~ | QIT | One QM, many formulations | ~~*new (2026-09-20)*~~ **FC-5″ DONE 2026-09-20** ([`Mathlib/Analysis/Semigroup/SchrodingerSchwartz.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/SchrodingerSchwartz.lean), Category 1, 8 pins; S–M took M): the phase `e^{−itκ}` of a symbol of temperate growth has temperate growth (`s ↦ e^{is}` has all its derivatives unimodular; ★ `hasTemperateGrowth_phaseFun`), so `U_κ(t)` is Mathlib's `SchwartzMap.fourierMultiplierCLM` on Schwartz space and ★ `fourierGroup_toLp`: `U_κ(t)(f.toLp 2) = (𝓕⁻¹(e^{−itκ}·𝓕f)).toLp 2`; ★ `kineticOp_eq_laplacian`: the kinetic operator `fourierMultiplierCLM (2π²ξ²)` **is** `−½Δ` (Mathlib's `laplacian_eq_fourierMultiplierCLM`); ★★ `hasDerivAt_freeSchrodinger`: **the free Schrödinger equation** `∂_t (U₀(t) f) = −i H₀ (U₀(t) f)` in `L²` at every `t` for Schwartz `f` (dominated convergence for the difference quotient of the phase, `|(e^{−itκ}−1)/t| ≤ |κ|`, lifted through the Fourier isometry, moved by the group law); `hasDerivAt_fourierGroup_toLp` for any dispersion relation of temperate growth. ~~**Not done**: the free Gaussian packet spreading into a Gaussian of complex variance — the pin has no Gaussian bundled as a `SchwartzMap`; priced as #48.~~ The Gaussian packet landed as #48 (FC-5‴, 2026-09-20). | ~~**S–M**~~ done (M) | Medium | #36 (c); [`feynman-continuum-scoping.md`](feynman-continuum-scoping.md) §5 | > | ~~44~~ | QIT | One QM, many formulations | ~~**FC-5′** — the kernel form, `(U₀(t)ψ)(x) = (2πit)^{−1/2} ∫ e^{i(x−y)²/2t} ψ(y) dy` as an improper Fresnel integral, and the `n`-slice iterated kernel with `e^{iS_n(x₀,…,xₙ)}` visible in the integrand.~~ **DONE 2026-10-01** (`Mathlib/Analysis/Semigroup/FresnelKernel.lean`, Cat-1, 18 pins) — and **stronger than the row asked**: the formula is an **exact identity, not an improper integral**. ★★★ `freeSchrodingerS_eq_integral_fresnel`: for `t ≠ 0` and Schwartz `ψ`, `(U₀(t)ψ)(x) = (2πit)^{−1/2} ∫ e^{i(x−y)²/(2t)} ψ(y) dy` with **no regularisation left in the statement**, the integral absolutely convergent (★ `integrable_fresnelKernel_mul`) because Feynman's kernel has **constant modulus** (`norm_fresnelKernel_I_mul`) and a Schwartz `ψ` is `L¹` — the oscillation is carried by the data, not by the kernel. **THE ROUTE WAS THE ROW'S OWN:** the kernel at complex time `w`, `(4πw)^{−1/2}e^{−x²/(4w)}`, is Schwartz for `Re w > 0` with Fourier transform `e^{−4π²wξ²}` (★ `fourier_fresnelS`); convolution with it is a semigroup in `w` (★★ `fresnelOp_fresnelOp`); and the `ε ↓ 0` limit along `w = ε + it/2` is the propagator by dominated convergence on the Fourier side (★★★ `tendsto_fresnelOp`, ★★★ `tendsto_integral_fresnelKernel_freeSchrodingerS`) — the regularisation is the **proof**, and uniqueness of limits removes it from the statement. **THE ACTION IS A THEOREM:** `U₀(t) = (U₀(t/(n+1)))^{n+1}` exactly for every `n` (★★ `freeSchrodingerS_iterate`, ★★ `fresnelOp_iterate_div`), each slice one kernel integral (★ `fresnelOp_iterate_succ_apply`), and the product of the `n` kernels in that iterated integral is `((2πit/n)^{−1/2})^n e^{iS_n}` with `S_n` the discrete action `∑(q_{k+1}−q_k)²/(2Δt)` (★★★ `prod_fresnelKernel_eq_exp_discreteAction`, `discreteAction`). **WHY IT WAS NOT XL AFTER ALL — AND D4 WAS HALF RIGHT:** the pin has no oscillatory-integral theory (no `fresnel` anything, so D4's cost claim stands), but it does not need it — the pin acquired `Mathlib/Analysis/Fourier/Convolution.lean` (`SchwartzMap.convolution`, `fourier_convolution`, `convolution_mul_swap`), which with `GaussianPacket.lean`'s complex-parameter Gaussian makes the whole route M–L. **RESIDUES, recorded here and not opened as rows** (no new rows, per the author's instruction): (i) the scoping note's `ψ ∈ L¹ ∩ L²` is **not** claimed — inversion at the pin is the Schwartz theorem; (ii) for `ψ ∈ L² \ L¹` the integral is genuinely improper and does need the absent oscillatory theory; (iii) the `n`-fold integral is proved in **iterated** form (`n` nested integrals, which is what time-slicing is), not as one integral over `ℝⁿ` against the product measure; (iv) one dimension. | ~~**XL**~~ **M–L** | ~~Medium for the physics reader; the mathematics does not need it~~ **done** | #36 (c); [`feynman-continuum-scoping.md`](feynman-continuum-scoping.md) §5, §8 D4 | > | ~~45~~ | QIT | One QM, many formulations | ~~*new (2026-09-20)* **Wigner–Moyal phase space**~~ **DONE 2026-09-23** ([`Mathlib/Analysis/Fourier/Wigner.lean`](../CsdLean4/Mathlib/Analysis/Fourier/Wigner.lean), Category 1, 9 pins; M took M): the Wigner function `W_ψ(x, ξ) = 𝓕_y[ψ(x + y/2) conj ψ(x − y/2)](ξ)` on `𝓢(ℝ, ℂ)` (`wignerKernel` is Schwartz in `y`, through `compCLMOfAntilipschitz` and `bilinLeftCLM`; `wigner`); ★ real (`conj_wigner`); ★ the `ξ`-marginal is `|ψ(x)|²` (`integral_wigner_right`, Fourier inversion at the origin); ★★ `wigner_fourier` — **the momentum representation** `W_{𝓕ψ}(ξ, −x) = W_ψ(x, ξ)` by Schwartz Parseval on the modulated reparametrisations (`affineCLM`, `modCLM`; the Fourier transform of an affine reparametrisation `fourier_comp_affine` and of a modulation `fourier_char_mul`); ★ the `x`-marginal is `|𝓕ψ(ξ)|²` (`integral_wigner_left`); total mass and the position/momentum moments (`integral_integral_wigner`, `integral_integral_mul_wigner_left/right`); ★ the Gaussian `e^{−πσx²}` has the positive Wigner function `√(2/σ) e^{−2πσx²} e^{−2πξ²/σ}` (`wigner_gaussianS`, `wigner_gaussianS_pos`); ★ an odd `ψ ≠ 0` has `W_ψ(0, 0) = −2∫|ψ|² < 0` (`re_wigner_zero_zero_neg`) — negativity as the witness of non-classicality; ★★ `wigner_freeSchrodingerS` — **the free evolution is Liouville transport** `W_{U₀(t)ψ}(x, ξ) = W_ψ(x − 2πtξ, ξ)` (`p = 2πξ`; exact for the quadratic Hamiltonian). One dimension, Schwartz data. The general Weyl-quantised expectation formula is #63, the Moyal bracket's `ℏ²` expansion #64. | ~~**M**~~ done (M) | Medium — a fifth formulation, and the one the classical limit lives in | this row; [`FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md`](../docs/FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md); #63; #64 | > | ~~46~~ | Foundations / Positioning | One QM, many formulations · Reviewer-proofing | ~~*new (2026-09-20)* **Stochastic mechanics (Nelson 1966) and Bohmian trajectories** — single-trajectory formulations: Bohm's `ẋ = ∇S/m` for `ψ = R e^{iS/ℏ}` with the continuity equation `∂_t |ψ|² + ∇·(|ψ|² ∇S/m) = 0` as **equivariance** (the Born density is carried by the flow — the one theorem, S–M, from the Schrödinger equation on Schwartz functions), and Nelson's diffusion with osmotic drift whose density is `|ψ|²` for all time. What it adds for CSD: the closest existing formulation to a single-trajectory ontology, and the sharpest contrast — Bohm's trajectory lives in configuration space and is guided by `ψ`, Nelson's is stochastic, CSD's lives in Σ, is deterministic and has records do the selecting; Valentini's relaxation to `|ψ|²` is already R-019 (#18).~~ **DONE 2026-09-28** (NEW Cat-1 [`Mathlib/Analysis/Calculus/SchrodingerCurrent.lean`](../CsdLean4/Mathlib/Analysis/Calculus/SchrodingerCurrent.lean), 8 pins, and the positioning note [`specs/bohm-nelson-scoping.md`](bohm-nelson-scoping.md); **S–M / M** took S–M). **One identity carries both formulations.** For any `ψ(t, x)` satisfying `i ∂_tψ = −½ ∂²_xψ + Vψ` at the point with `V` real (`ℏ = m = 1`): ★ `hasDerivAt_probCurrent` — `∂_x J = Im(ψ̄ ∂²_xψ)`, the `Im(∂ψ̄ ∂ψ) = 0` cancellation — and ★ `hasDerivAt_probDensity_time` — `∂_t ρ = 2 Re(ψ̄ ∂_tψ)` — give ★★ `continuity_equation`: **`∂_t ρ + ∂_x J = 0`**, the Schrödinger equation's only input being `Re(−i z) = Im z` and the reality of `V`. **Bohm**: ★ `probCurrent_polar` — in polar form `ψ = R e^{iS}` the current is `J = R² ∂_x S`, the density times the guidance velocity — so ★★ `continuity_polar` is **equivariance**, `∂_t ρ + ∂_x(ρ v) = 0`: the Born density is transported by the guidance field; and ★★ `hasDerivAt_density_along_trajectory` is its **Lagrangian** form, `dρ/dt = −ρ ∂_x v` along a curve whose velocity is the guidance field (the two-variable chain rule through `HasFDerivAt.comp_hasDerivAt`). **Nelson**: the same flux split as `ρ b = ρ v + ½ ∂_xρ = ρ (v + ½ ∂_x log ρ)` (`nelsonFlux`, `nelsonFlux_eq_mul_drift`) turns the continuity equation into ★★ `nelson_fokkerPlanck`: **`∂_t ρ = −∂_x(ρ b) + ½ ∂²_x ρ`**, the Fokker–Planck equation of a diffusion with constant coefficient `½` — the osmotic term of the drift is exactly what cancels the diffusion term, which is why `|ψ|²` is the law at every time. **The M half stays unclaimed and the row says which half**: the *equation* is proved, the *process* is not — Nelson's theorem needs stochastic differential equations, absent at the pin — and Bohmian trajectories are taken as given, their global existence (singular at the nodes of `ψ`) not attempted; one dimension only, and the Schrödinger equation is a pointwise hypothesis, not a solution produced by the corpus's semigroup. **The positioning answer is now written down** ([`bohm-nelson-scoping.md`](bohm-nelson-scoping.md)): what the three single-trajectory pictures share is this identity; they differ in where the trajectory lives (configuration space guided by `ψ` / the same space with noise / `Σ`, with `ψ` only labelling an epistemic region), in what the dynamics is (`∇S/m` read off the wavefunction / the same with a drift / the Hamiltonian flow of the sector's own symplectic structure), and in what selects an outcome (an initial condition sampled from `|ψ|²` by hypothesis / the noise / **the record**, with the Born weights derived as typicality volumes). Valentini's relaxation stays `R-019` (#18). | ~~**S–M** (equivariance) / **M** (Nelson)~~ done (S–M); the diffusion *process* unclaimed (needs SDEs) | Medium–high for positioning (the reviewer's “how is this not Bohm?”) — now answered in a citable note | #18 (`R-019`); [`bohm-nelson-scoping.md`](bohm-nelson-scoping.md); expert-review table | > | ~~47~~ | QIT / Positioning | One QM, many formulations · Reviewer-proofing | ~~*new (2026-09-20)* **Koopman–von Neumann classical mechanics in Hilbert space**: for a measure-preserving flow `φ_t` the Koopman operator `U_t f = f ∘ φ_t` is a unitary group on `L²(μ)` (the corpus already has the ingredients: the arena's Liouville theorem and `birkhoffAverage`); the generator is `i{H, ·}` (the Liouvillian), `x` and `p` commute. What it adds: the exact statement of what separates quantum from classical — Hilbert space, unitarity and superposition are all available classically, the commutator `[x, p] = iℏ` is not.~~ **DONE 2026-09-28** (NEW Cat-1 [`Mathlib/Dynamics/Koopman.lean`](../CsdLean4/Mathlib/Dynamics/Koopman.lean), 5 pins, plus one theorem in [`SigmaLayer/ChartBracket.lean`](../CsdLean4/SigmaLayer/ChartBracket.lean), 1 pin; **S–M / M** took S–M). ★ `koopmanL2` — the Koopman operator `U f = f ∘ T` of a measure-preserving map as a **linear isometry** of `L²(μ)`: Mathlib has the composition as an `AddMonoidHom` (`Lp.compMeasurePreserving`) and its isometry but **not** the linear packaging, so the `•`-compatibility is proved by transporting the a.e. equality through the map (`QuasiMeasurePreserving.ae_eq_comp`). ★★ `koopmanUnitary` — **for an invertible measure-preserving map it is unitary**, a `LinearIsometryEquiv` whose inverse is the Koopman operator of the inverse map — and ★★ `koopmanL2_flow`: **the group law** `U_{s+t} = U_t ∘ U_s`, so a measure-preserving flow gives a one-parameter unitary group on `L²`. `koopmanFun_mul` is the classical side of the contrast: on observables the Koopman map is an algebra homomorphism for the **pointwise** product (and the pin records that this one is *axiom-free*, `rfl`). ★ `hasDerivAt_koopmanFun` is **the generator on differentiable observables** — `d/dt (f ∘ φ_t)(x) = Df(φ_t x)[X(φ_t x)]`, the derivative of the observable along the flow's velocity field, stated for observables valued in any normed space — and ★ `hasDerivAt_koopmanFun_poissonBracket` (ChartBracket) reads it in a Darboux chart, where the field is `hamiltonianField H` and the derivative along it is `{f, H}`: **the Koopman generator is the Liouvillian**. Honest scope: the generator is pointwise on differentiable observables along a given flow line — nothing about its `L²` domain, self-adjointness, or Stone's theorem, which is the analytic half of Koopman–von Neumann; the flow is a hypothesis (that a Hamiltonian flow preserves the measure is Liouville, `Q29`, not reproved); and the sentence the row is *for* — Hilbert space, unitarity and superposition are classically available, the commutator is not — is **positioning**, carried by the module header, since no theorem could witness it. | ~~**S–M** (Koopman unitary group) / **M** (Liouvillian)~~ done (S–M) | Medium — positioning, one paragraph and one theorem, both now landed | `LF4/ArenaSymplectic.lean`; `SigmaLayer/ChartBracket.lean`; `Thermo/ETH.lean` | > | ~~48~~ | QIT | One QM, many formulations | ~~*new (2026-09-20)*~~ **FC-5‴ DONE 2026-09-20** ([`Mathlib/Analysis/Semigroup/GaussianPacket.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/GaussianPacket.lean), Category 1, 11 pins; M took M): `gaussianS a ha : 𝓢(ℝ, ℂ)`, the Gaussian `x ↦ e^{−πax²}` (`Re a > 0`) bundled as a `SchwartzMap` — its `n`-th derivative is `P_n(x) e^{−πax²}` with `P_{n+1} = P_n' − 2πa X P_n` (`iteratedDeriv_gaussFun`) and `|x|^m e^{−cx²} ≤ 1 + m!/cᵐ` from `yᵐ/m! ≤ e^y` (`pow_mul_exp_neg_le`); its Fourier transform `a^{−1/2} e^{−πξ²/a}` as a Schwartz identity (`fourier_gaussianS`, Mathlib's `fourier_gaussian_pi`); the free phase turns it into the Gaussian of parameter `a⁻¹ + 2πit` (`smulLeft_phase_fourier_gaussianS`); ★★ `freeSchrodingerS_gaussianS` — **the packet spreads**: `U₀(t) g_a = packetAmp a t • g_{a(t)}`, `a(t) = (a⁻¹ + 2πit)⁻¹ = a/(1 + 2πiat)` (`packetParam_eq`), amplitude `(1 + 2πiat)^{−1/2}` (`packetAmp_eq`, after ★ `mul_cpow_of_re_pos`: the principal power is multiplicative on the right half-plane); ★ `freeSchrodinger_gaussianS_toLp` the same in `L²`; ★ `re_packetParam_ofReal` — for real `σ > 0` the density is `e^{−2πσ(t)x²}` with `σ(t) = σ/(1 + (2πσt)²)`, the textbook widening `√(1 + (2πσt)²)`. One dimension; the moving packet and `ℝᵈ` are not stated (the module header says so). | ~~**M**~~ done (M) | Medium | #43; [`feynman-continuum-scoping.md`](feynman-continuum-scoping.md) §5 | > | ~~49~~ | Geometry | Library owed / upstream | ~~*new (2026-09-21)* **Darboux residue (a): the chart is `C¹`.**~~ **DONE 2026-09-23** ([`FlowDerivative.lean`](../CsdLean4/Mathlib/Analysis/ODE/FlowDerivative.lean): `norm_le_exp_of_linearODE`, ★ `dist_le_of_linearODE_coeff_close` — continuous dependence of a linear ODE's solution on its coefficient, Grönwall for approximate trajectories on the operator space — and `exists_flow_hasFDerivAt_of_norm_fderiv_le` strengthened by the conjunct `x ↦ Y x t` continuous on the half-ball; [`Darboux.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Darboux.lean): the three Darboux theorems now carry `ContDiffOn ℝ 1 Φ Φ.source ∧ ContDiffOn ℝ 1 Φ.symm Φ.target` (`contDiffAt_one_iff` on the continuous derivative `x ↦ Y x 1`, Mathlib's `OpenPartialHomeomorph.contDiffAt_symm` for the inverse), and ★★ `IsSymplectic.exists_openPartialHomeomorph_localRep_pullback_eq` states that the Darboux chart `(chartAt E x₀).trans Φ.symm` contains `x₀` and belongs to `IsManifold.maximalAtlas 𝓘(ℝ, E) 1 M` (`mem_contDiffGroupoid_of_contDiffOn`, `StructureGroupoid.trans_mem_maximalAtlas`); 4 pins. The `C^k` half is #60. M took M) | ~~**M** (`C¹`)~~ done | Low — nothing consumes the chart's regularity | #8; #60 | > | ~~50~~ | Geometry | Library owed / upstream | ~~*new (2026-09-21)* **Darboux residue (b): the standard form.**~~ **DONE 2026-09-23** ([`Mathlib/LinearAlgebra/BilinearForm/SymplecticBasis.lean`](../CsdLean4/Mathlib/LinearAlgebra/BilinearForm/SymplecticBasis.lean): `IsSymplecticBasis`, ★★ `IsAlt.exists_isSymplecticBasis` — every non-degenerate alternating form on a finite-dimensional space over any field has a basis with `B pᵢ pⱼ = 0`, `B qᵢ qⱼ = 0`, `B pᵢ qⱼ = δᵢⱼ`, by induction on the dimension splitting a plane `span {p, q}` off against its `B`-orthogonal complement; ★ `IsAlt.even_finrank`; ★ `IsSymplecticBasis.apply_eq_sum`, the standard form in symplectic coordinates. [`Mathlib/Geometry/Manifold/DarbouxStandardForm.lean`](../CsdLean4/Mathlib/Geometry/Manifold/DarbouxStandardForm.lean): `ContinuousAlternatingMap.toBilinForm`, `standardSymplecticForm` (`∑ dpᵢ ∧ dqᵢ` on `ι ⊕ ι → ℝ`), ★ `exists_continuousLinearEquiv_eq_standardSymplecticForm_comp` (linear Darboux, `finrank E = 2n`), ★★ `exists_openPartialHomeomorph_pullback_standard` — a `C¹` chart `Ψ : E → ℝ^{2n}` with `C¹` inverse and `ω = Ψ^* ω_std`, ★★ `IsSymplectic.exists_openPartialHomeomorph_localRep_pullback_standard`, ★ `IsSymplectic.even_finrank`; 12 pins. `exists_openPartialHomeomorph_pullback_eq` gained the conjunct `Φ.target ⊆ ball x₀ R`. M took M) | ~~**M**~~ done | Low–medium — the even-dimension corollary is the one fact a reader expects | #8; #49; #60 | > | ~~51~~ | Reversible | QC chain (link 12) | ~~*new (2026-09-21)* **The threshold theorem, code-capacity form.**~~ **DONE 2026-09-23** ([`Mathlib/Probability/CodeCapacityThreshold.lean`](../CsdLean4/Mathlib/Probability/CodeCapacityThreshold.lean), Category 1: ★ `measure_pi_two_or_more_le` — two or more of `n` independent events of probability `≤ q` occur with probability `≤ C(n,2) q²`, the union bound over pairs through `Measure.pi_pi`; `codeCapacityBound` with closed form `(c p)^{2^k}/c`, `codeCapacityBound_le`, ★ `tendsto_codeCapacityBound` — the threshold `p < 1/c`; `ConcatPat`/`concatMeasure`/`concatBad` and ★★ `concatMeasure_concatBad_le` — the bad patterns of a `k`-fold concatenated block code obey the recursion. [`Empirical/QM/QEC/SteaneThreshold.lean`](../CsdLean4/Empirical/QM/QEC/SteaneThreshold.lean), Category 3: ★★ `steane_codeCapacity_failure_le` — under independent per-qubit noise of rate `≤ p` the Steane recovery fails to return a code state with probability `≤ 21 p²` (a pattern with at most one error is a single-qubit Pauli, `exists_singleErr_of_card_le_one`, undone by #14(b)'s recovery); ★ `steane_concatBad_le`, ★ `steane_threshold` — `(21 p)^{2^k}/21 → 0` for `p < 1/21`. 10 pins. The quantum statement is at level `1`; the concatenated quantum recovery is #61 and the circuit-level theorem #62. M–L took M) | ~~**M–L** (code-capacity recursion)~~ done (M) / circuit-level → #62 | Medium — the one sentence the document's chain still lacked, now in link 12 | #14(b); [`steane-plan.md`](steane-plan.md) | > | ~~52~~ | QIT | QC chain (link 12) | **DONE 2026-09-21** — `Empirical/QM/QEC/ControlledDilation.lean` (the generic controlled-error dilation, 9 pins: `Channel.mixedUnitaryChannel E q` IS `stinespringChannel (controlledUnitary E q e₀) e₀`; `RegisterDilation.lean` is now its `Fin 4` instance with every statement kept), `Empirical/QM/QEC/IndependentNoise.lean` (14 pins: `indepFlipChannel p`, `flipOp_eq` — every pattern is a single error times `1` or times `X̄`, ★★ `recoveryChannel_apply_indepFlipChannel_apply`: `R (N_p ρ) = (1 − p_fail) ρ + p_fail X̄ ρ X̄`, `p_fail = 3p² − 2p³`, `failProb_lt_of_lt_half`), `Empirical/CSD/QEC/IndependentNoiseFlow.lean` (11 pins: `indepFlow` on `ℂℙ⁶³`, ★★ `indepFlow_traceRight_barycenter`, ★★★ `indepFlow_recovery` — the residual failure on `Σ`). Took M. The generic dilation is Category 3 (it needs LF2's `stinespringChannel`/`embedEnv`), not the Category 1 the row-53 cell guessed. *Was:* **The `Σ` side of independent noise.** Link 10's `Σ` model ([`Empirical/CSD/QEC/ThreeQubit.lean`](../CsdLean4/Empirical/CSD/QEC/ThreeQubit.lean), [`RegisterFlow.lean`](../CsdLean4/Empirical/CSD/QEC/RegisterFlow.lean)) corrects the **single-flip** channel only; the module says the independent-noise channel's double and triple flips are neither correctable nor modelled. Brick: a `Σ`-flow whose environment marginal is the **independent bit-flip channel** (three independent kicks, the product measure on the environment), so the `Σ`-twin of QEC meets the noise model #51 uses, with the uncorrectable weight-2 and weight-3 patterns visible as the `p²` term. | **M** | Medium for the `Σ` story — the twin of #51's noise model | #51; `Empirical/CSD/QEC/` | > | ~~53~~ | QIT | QC chain (link 12) | ~~*new (2026-09-21)* **The `Σ`-twin of the Steane recovery.**~~ **DONE 2026-09-27** ([`Empirical/CSD/QEC/SteaneFlow.lean`](../CsdLean4/Empirical/CSD/QEC/SteaneFlow.lean), Category 3, 4 pins; M took M). The seven-qubit code's error correction as **one flow on `Σ`**: `steaneJointMat q` instantiates #52's generic controlled-error dilation at the Steane code's own error family — the `128`-ray register times the `22` single-error labels `Option (Fin 7 × Fin 3)` — and `steaneFlow q` is its projective action on `ℂℙ²⁸¹⁵`, a flow of the corpus's `cpSectorData` lifting the joint unitary with no hypothesis (★ `isUnitaryLift_steaneFlow`). Then ★★ `steaneFlow_traceRight_barycenter`: **the single-qubit Pauli channel is the environment marginal of the joint `Σ`-flow** (the reduced density operator of the flowed preparation is the mixed-unitary Pauli channel applied to the register's), and ★★★ `exists_steaneFlow_recovery`: **QEC on `Σ`, end to end** — one recovery channel, the code's own and **independent of both the preparation and the noise weights**, applied to that marginal, returns the register's density operator. The linearity step is `exists_steane_recovery_mixedUnitary`: the channel is a convex combination of the errors and a channel is linear, so undoing each error undoes the mixture. The CSD reading, in the header: the error is not a separate axiom but the flow's leak into the environment factor; the codespace is a region of `Σ` (a hypothesis on the preparation, not a derivation), the regions are epistemic and the measure weighting them is ontic. No Hamiltonian generates the joint unitary — the flow is the time-one map of the projective action, as for `bitFlipFlow`, `registerFlow` and `indepFlow`. | ~~M (generic dilation) +~~ ~~**M**~~ done | Medium for the `Σ` story — the twin the cell promised | #14(b); #52 | > | ~~54~~ | QIT | One QM, many formulations | ~~*new (2026-09-21)* **The curvature formula for the geometric phase**~~ **DONE 2026-09-23** — BP-3 of [`berry-phase-scoping.md`](berry-phase-scoping.md). [`Mathlib/Analysis/InnerProductSpace/GeometricPhaseCurvature.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/GeometricPhaseCurvature.lean) (Category 1, 3 pins): for a `C²` family of unit vectors `Ψ` on `[0, 1] × [0, T]` — the disc in polar form: the edge `s = 0` collapsed to a point, the edges `t = 0`, `t = T` the same curve of rays up to `e^{iφ(s)}` — the connection form has components `A_s = Im⟪Ψ, ∂_sΨ⟫`, `A_t = Im⟪Ψ, ∂_tΨ⟫`, both `C¹` with explicit derivatives through `fderivInnerCLM` (`hasFDerivAt_connectionFormS/T`); ★ `curvature_eq` — **the curvature `∂_s A_t − ∂_t A_s` is `2 Im⟪∂_sΨ, ∂_tΨ⟫`** (symmetric second derivatives, `ContDiffAt.isSymmSndFDerivAt`); ★★ `geometricPhase_eq_neg_integral_curvature` — **`β = −∫∫ curvature`**: Mathlib's divergence theorem on the rectangle (`integral2_divergence_prod_of_hasFDerivAt_off_countable`, Green) with `f = A_t`, `g = −A_s`; the loop edge gives the dynamical phase, the collapsed edge `0`, the cut `−φ` by `dynamicalPhase_rephase`. [`Empirical/QM/BerryPhaseCurvature.lean`](../CsdLean4/Empirical/QM/BerryPhaseCurvature.lean) (Category 3, 4 pins): the disc `(cos(sθ/2), e^{it} sin(sθ/2))` of the cone has curvature `(θ/2) sin(sθ)` (`curvature_coneSurface`), disc integral `π(1 − cos θ) = Ω/2` (`integral_curvature_coneSurface`); ★★ the general formula applied to the cone (`geometricPhase_coneCurve_eq_neg_integral_curvature`) and ★★ `curvature_formula_consistent`, the same equation from Berry's direct `−Ω/2` — two independent derivations agree. The identification of `curvature` with the pullback of the corpus's bundle-level `fsForm` is #65. | ~~**M–L**~~ done (M) | Medium — closes the geometric reading | #10; [`berry-phase-scoping.md`](berry-phase-scoping.md); #65 | > | ~~55~~ | QIT | Experiment twins | ~~*new (2026-09-21)* **Aharonov–Bohm, discrete** (the flux twin ER3 asked for), [`berry-phase-scoping.md`](berry-phase-scoping.md) BP-4: a ring of `N` sites with Peierls phases `e^{iΦ/N}` on the hops; the spectrum `2 cos((2πk + Φ)/N)` depends on the flux `Φ` only mod `2π` and is invariant under the gauge transformation that moves the whole phase onto one bond — the flux is observable, the vector potential is not. A circulant matrix and its eigenvectors, Category 1, with the QM-side reading in `Empirical/QM/`. The `L²(S¹)` version is CV-scale and not proposed.~~ **DONE 2026-09-28** ([`Mathlib/QuantumInfo/AharonovBohmRing.lean`](../CsdLean4/Mathlib/QuantumInfo/AharonovBohmRing.lean), Category 1, 22 pins, and [`Empirical/QM/AharonovBohm.lean`](../CsdLean4/Empirical/QM/AharonovBohm.lean), Category 3, 7 pins; **S–M** took S–M). The model is `ringHamOf N a` — a ring of `N` sites whose hop `j → j + 1` carries the Peierls phase `e^{i a j}`, Hermitian for `N ≥ 3` — with `flux a = ∑ⱼ a j`; the uniform gauge is `ringHam N Φ`. **The spectrum**: the twisted Fourier modes `ψ_k(j) = e^{−2πi jk/N}` are eigenvectors, ★★ `ringHam_mulVec_ringMode` giving `2 cos((2πm + Φ)/N)` for **any integer label** `m` of the mode (the `ZMod N` character `ZMod.stdAddChar` makes `ψ(j ± 1) = ψ(j)·χ(∓k)` a one-line rewrite, and the *integer* cast `stdAddChar_coe` keeps `ZMod.val` out of the statement entirely); ★★ `dftMatrix_conj_ringHam` — **the discrete Fourier matrix diagonalises the ring** (its unitarity is character orthogonality, `sum_stdAddChar_mul`, from `ZMod.isPrimitive_stdAddChar`) — hence ★★ `eq_ringEigval_of_mulVec`: those numbers are the *whole* spectrum, not merely some of it. **The vector potential is not observable**: ★★ `gaugeDiag_conj_ringHamOf` — conjugating by `diag(e^{i g})` shifts the bond phases by the discrete gradient of `g` — with ★ `flux_gauge` (the total is untouched: the gradient telescopes around the ring, by reindexing with `Equiv.addRight 1`, no `val` needed) and ★★★ `gaugeDiag_conj_ringHam_oneBond`: the explicit gauge `g j = (Φ/N)·j.val` pushes **the whole phase onto the single bond from `−1` to `0`** and the identity `Φ/N + g j − g (j+1) = oneBondPhase j` holds *exactly*, the wrap-around case being where `(Φ/N)·N = Φ` appears. **The flux quantum**: ★ `ringEigval_add_two_pi` (`Φ ↦ Φ + 2π` relabels `m ↦ m + 1`) and ★★ `range_ringEigval_add_two_pi` — the level *set* does not move, so a quantum is invisible. **And the flux is observable**: ★★ `energy_lt_energy_of_flux_lt` (the level of label `0` falls strictly as `Φ` grows through `[0, πN]`, for every `N ≥ 3`) and ★★★ `not_exists_unitary_conj_ringHam` — **no unitary conjugation carries the flux-free triangle to the flux-`π` triangle**: `2` is a level of the first (the `m = 0` mode) and, by the completeness theorem, of the second it is not (`ringEigval_three_pi_ne_two`, from `cos = 1` forcing `2m + 1 = 6n`). The Category-3 file reads these as the experiment, with the levels as numbers — `{2, −1, −1}` at zero flux, `{1, −2, 1}` at half a quantum — and ★★ `same_levels_of_gauge` (the one-bond ring has the same levels, the gauge-rotated modes as eigenvectors). Honest scope: a lattice ring, Peierls' substitution *as the definition* — no solenoid, no interference pattern, no continuum limit (that is the new row #90), and the general-`N` injectivity of `Φ ↦ spectrum` is the new row #91. | ~~**S–M**~~ done (S–M) | Medium — the flagship phase experiment's twin | #10; [`berry-phase-scoping.md`](berry-phase-scoping.md); #90; #91 | > | ~~56~~ | QIT | One QM, many formulations | ~~*new (2026-09-21)* **Berry's adiabatic theorem**, [`berry-phase-scoping.md`](berry-phase-scoping.md) BP-5: the eigenstate of a slowly driven `H(R(t))` follows the parameter loop up to `e^{i(φ_dyn + β)}` with error `O(1/T)` — the adiabatic theorem (Kato 1950) is an asymptotic ODE result absent from Mathlib and the corpus.~~ **DONE 2026-09-28 — BERRY'S OWN MODEL, EXACTLY, WITH THE ADIABATIC ERROR EXPLICIT** ([`Mathlib/QuantumInfo/RotatingFrame.lean`](../CsdLean4/Mathlib/QuantumInfo/RotatingFrame.lean), Category 1, 15 pins, and [`Empirical/QM/BerryAdiabatic.lean`](../CsdLean4/Empirical/QM/BerryAdiabatic.lean), Category 3, 13 pins; **XL** took M + M). **The XL and the “not recommended” were about Kato's theorem; Berry's §5 model needs none of it.** A field of fixed magnitude `B` sweeping the cone of half-angle `θ` at angular velocity `ω` is a *rigid rotation* of one Hamiltonian, `H(t) = R_z(ωt) H(0) R_z(ωt)ᴴ` (★ `drive_conj_initHam`, from the Bloch rotation ★ `axisRot_z_conj`), so the rotating frame is **exact**: the general reduction ★★ `hasDerivAt_mul_of_gen` (`u` generated by `G` on the right, `e` by `K = H₀ − G` on the left, `u t` unitary ⇒ `(ue)′ = −i H(t) (ue)` — no commutation hypothesis, no time ordering) is instantiated twice: by the matrix exponential (★★ `hasDerivAt_rotProp`, `V(t) = e^{−itG} e^{−it(H₀−G)}`, with ★ `hasDerivAt_rotState` for the state on `ℂᴺ` through ★ `inner_toEuclideanCLM_unitary`) and by the axis–angle rotations of spin-½, which mention no exponential at all: `pauliAxis a b c = aσ₁ + bσ₂ + cσ₃` with ★ `pauliAxis_mul_self`, `axisRot_eq` (`R_n(θ) = cos(θ/2)·1 − i sin(θ/2) n·σ`, a scalar combination of **two constant matrices**) and ★ `hasDerivAt_axisRot` — so the derivative is `HasDerivAt.smul_const`, with no entrywise differentiation of a matrix path and no norm-scope gymnastics. Then: ★★★ `hasDerivAt_prop` — **the time-dependent Schrödinger equation solved exactly**, `V(t) = R_z(ωt) R_k(Ωt)` with `Ω = √((B sin θ)² + (B cos θ − ω)²)` the Rabi frequency and `k` the unit axis of `K` (★ `initHam_sub_driveGen`), `V 0 = 1`, `V t` unitary, and ★★ `hasDerivAt_state` for `ψ(t) = V(t)ψ₀`; ★★ `fieldHam_mulVec_coneCurve` — **the instantaneous eigenvector of `H(t)` with eigenvalue `B/2` is exactly BP-2's cone curve** `(cos θ/2, e^{iωt} sin θ/2)`, so the ray loop whose geometric phase BP-2 computed *is* the loop this Hamiltonian traces; ★★★ `norm_inner_coneCurve_state_sq` — **adiabatic following, exactly**: the probability of still being in the instantaneous eigenstate is `1 − (ω sin θ/Ω)² sin²(Ωt/2)`, hence ★★★ `one_sub_le_norm_inner_coneCurve_state_sq` — the loss is `≤ (ω/(B − ω))²` **uniformly in time** (the rotating frame is exact here, so nothing is lost to a `T`); ★★★ `norm_inner_state_sub_exp_le` — **the phase after one turn** `T = 2π/ω` is `e^{i(−(B/2)T − π(1 − cos θ))}` up to `πω/(2(B − ω)) + (ω/(B − ω))²`: the dynamical phase `−∫₀ᵀE dt` plus **BP-2's geometric phase**, with an `O(ω)` error that is explicit, not asymptotic — and ★ `geometricPhase_cone_loop` identifies that second summand as the repo's own `GeometricPhase.geometricPhase` of the loop over `[0, T]` (the reparametrisation `t ↦ ωt` does not change it), so the statement is Berry's and not a coincidence of numbers. **One identity carries both error bounds**: ★ `one_sub_cosDelta_sq`, `1 − (k·n₀)² = (ω sin θ/Ω)²`, with `Ω ≥ B − ω` (`sub_le_rabiFreq`) doing the rest; the `2π` rotation being `−1` (`axisRot_two_pi`) is what turns the accumulated phase into the geometric one. Honest scope, in both headers: the drive is a **rigid** rotation, so the eigenvalues never move; the general adiabatic theorem (Born–Fock, Kato: an arbitrary gapped path, error `O(1/T)`) is **not proved and not claimed anywhere in this corpus** — it is the new row #88. | ~~**XL**~~ done (M + M) | Low–medium — the geometric statement is already BP-1's; only the adiabatic limit is new | #10; [`berry-phase-scoping.md`](berry-phase-scoping.md); #88 | > | ~~57~~ | QIT / Positioning | Library owed / upstream · Reviewer-proofing | ~~*new (2026-09-21)*~~ **DONE 2026-09-22 — Gleason's theorem for `ℂᴺ`, `N ≥ 3`, proved** (`Gleason/Core.lean`, ★★ `ProjectionPackage.gleason_representation`, foundational triple, no hypothesis beyond `N ≥ 3`; 137 Gleason pins). **Gleason's theorem, finite-dimensional: the core lemma on `S²`** ([`gleason-feasibility.md`](gleason-feasibility.md)). Layers A and C are proved on `main` (`Mathlib/Analysis/InnerProductSpace/Gleason/`, 38 pins): `gleason_representation_of_core` gives the theorem for `ℂᴺ`, `N ≥ 3`, from `CoreLemma` (every nonnegative frame function on `S²` is a symmetric quadratic form). What remains is Cooke–Keane–Moran §§2–7, staged so each lands green: ~~**(a)** frame-function basics on `S²` + the two warmup theorems (B0–B1, **S–M**)~~ **(a) DONE 2026-09-21** (`Gleason/Sphere.lean`, `Gleason/Warmup.lean`, 22 pins; S–M took S–M); ~~**(b)** Piron's geometric lemma by the explicit gnomonic spiral (B3, **M** — the probe and the decision point)~~ **(b) DONE 2026-09-21** (`Gleason/Piron.lean`, 19 pins, ★ `exists_descent_chain`; M took M — the decision point is passed); ~~**(c)** the basic lemma and the simple-frame-function theorem, Bell/Piron's extreme case (B2, B4, **M**)~~ **(c) DONE 2026-09-22** (`Gleason/SimpleFrame.lean`, 18 pins, ★ `eq_add_mul_latitude`; M took M); ~~**(d)** extremal values attained via Tychonoff (B5, **M–L**)~~ **(d) DONE 2026-09-22** (`Gleason/Extremal.lean`, 18 pins, ★ `exists_forall_le`; M–L took M — the second risk point is passed); ~~**(e)** the general case, `Core.lean` merges without its `sorry` (B6, **L**)~~ **(e) DONE 2026-09-22** (`Gleason/General.lean` + `Gleason/Core.lean`, 23 pins, ★★ `frameFunction_regular_sphere`, the endgame by the circle lemma `inner_sq_eq_half_of_vanish` instead of the paper's zero count; L took M–L). The claim is made now that (e) is green and the pins are clean. Residues: #58 (real spaces, A4). ~~the glossary entry `gleason-theorem` still describes only the effect version~~ **DONE 2026-09-29**: the entry now carries both forms, names `ProjectionPackage.gleason_representation` and records that neither theorem is derived from the other (they share the Jordan–von Neumann engine and the descent, nothing else). | ~~S–M~~ done + ~~M~~ done + ~~M~~ done + ~~M–L~~ done + ~~L~~ done | High — the theorem the Born-rule programme answers to; no formalisation exists in any proof assistant (1000+ list) | `Gleason/`; `LF2/EffectGleason.lean`; branch `gleason-feasibility` | > | ~~58~~ | QIT | Library owed / upstream | ~~*new (2026-09-21)* **Gleason for real Hilbert spaces (A4).**~~ **DONE 2026-09-27** ([`Mathlib/Analysis/InnerProductSpace/Gleason/Real.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/Gleason/Real.lean), Category 1, 11 pins; S–M took M). ★★★ `real_gleason_representation`: **for `N ≥ 3` every real projection package on `ℝᴺ` is `P ↦ Tr(ρ P)` for a unique real density matrix** — the corollary #57's core lemma was always going to give, and one the complex theorem does **not** contain (a real inner product space is not a complex one). The real side mirrors the complex layers, which is cheap because the Gram section of `ProjectionPackage.lean` is already generic over `RCLike`: `rankOneR` and `RealProjectionPackage` with A1 (`sum_frame_orthonormalBasis`), A2 (`isFrameFunction_restrictR` — the weight of a `3`-space is basis independent because the rank-ones of a transported orthonormal basis sum to the same matrix), then the degree-2 extension `ext` which the core lemma makes a quadratic form on the span of every triple (`ext_combR`, `exists_ext_plane`, through ★ `exists_orthonormal_tripleR`), hence ★★ `ext_parallelogramR` — **the parallelogram law on `ℝᴺ`** by Gram–Schmidt — then the real Jordan–von Neumann engine (`IsQuadraticLikeR`, `polarR`, ★ `eq_dotProduct`) giving ★★ `exists_isSymm_sphere` (**A4**), and the descent (`posSemidef_of_sphere_nonnegR`, `eq_of_sphere_quadForm_eqR`, ★ `p_eq_trace` through the real spectral resolution) giving ★★ `existsUnique_density_of_frame_quadraticR`. **Done differently from the plan's sketch** (recorded in the header): `gleason-feasibility.md` §2 priced A4 as “three vectors lie in a `3`-space, so the polarisation is bilinear without a Cauchy equation”; the proof instead reuses the engine's Cauchy step (`additive_bounded_linear`, already generic) and needs only **pairs** inside a triple — the three-vector observation is never used. Honest scope: the statement is for a *package*, as in the complex theorem; Gleason's own frame-function form for `N ≥ 4` is new row #87 (nothing needs it), `N = 2` is false as always, and the real layer duplicates the shape of the complex one rather than genericising twelve pinned modules. | ~~**S–M**~~ done | Low–medium — the library's real corollary | #57 | > | ~~59~~ | Reversible | QC chain (link 11) | ~~*new (2026-09-23)* **The Gidney-adder hybrid instance.**~~ **DONE 2026-09-23** (`Empirical/QM/MeasurementGidneyAdderHybrid.lean`: `gidneyHybridCell` — the Toffoli replaced by the single-CZ gadget, the un-shifting CNOTs kept unitary; `gidneyHybrid_invariant` — the hybrid pass against the unitary one cell by cell; ★★ `gidneyHybridAdd_amplitude`, `gidneyHybridAdd_sum`, the addends restored on every branch; 11 + 2 pins. Both measurement-based adders are now amplitude-exact on the full register at the Toffoli counts their re-costs state: link 11's measurement strand is closed. S–M took M) | ~~S–M~~ done | Low | #16; `Empirical/QM/MeasurementGidneyAdderHybrid.lean` | > | ~~60~~ | Geometry | Library owed / upstream | ~~*new (2026-09-23)* **Darboux residue (c): the `C^k` chart of a `C^k` form.**~~ **DONE 2026-09-23** — three Category-1 pieces. [`Analysis/ODE/FlowSmooth.lean`](../CsdLean4/Mathlib/Analysis/ODE/FlowSmooth.lean): ★★ `contDiffOn_flow_of_contDiffOn` — **`C^n` dependence of a flow on its initial point** (`1 ≤ n ≤ ∞`): a flow of a jointly `C^n` field, confined to a compact convex set and Lipschitz in the initial point, is `C^n` in the initial point, by induction on `n` through the pair flow `(x, Z) ↦ (α x t, Y x t ∘ Z)` of the field `(z, Z) ↦ (f t z, D(f t)(z) ∘ Z)` on `E × (E →L E)`; ★ `exists_variational_of_lipschitz` (the `C¹` case in Lipschitz form); in [`FlowDerivative.lean`](../CsdLean4/Mathlib/Analysis/ODE/FlowDerivative.lean) ★ `exists_linearODE_solution_Icc` — linear ODEs have solutions on the whole interval, no smallness condition (concatenation with the Grönwall step). [`Analysis/Calculus/ContDiffParametricIntervalIntegral.lean`](../CsdLean4/Mathlib/Analysis/Calculus/ContDiffParametricIntervalIntegral.lean): ★ `contDiffOn_intervalIntegral_of_contDiffOn` — a parametric interval integral of a jointly `C^n` integrand is `C^n` (induction through `contDiffOn_clm_apply`, scalar integrands throughout), so the Poincaré primitive (`contDiffOn_radialPrimitiveForm'`) and Moser's field (`contDiffOn_moserFieldJoint'`) are `C^n`. `Darboux.lean`: the three Darboux theorems take the order `n : ℕ∞`, `1 ≤ n`, and conclude `ContDiffOn ℝ n Φ ∧ ContDiffOn ℝ n Φ.symm`; on a symplectic manifold the chart is `C^∞` and in the `C^∞` maximal atlas; the standard form follows at order `n`. 16 pins. L took L) | ~~**L**~~ done | Low — nothing consumes the chart's regularity beyond `C¹` | #8; #49; #50 | > | ~~61~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **The concatenated Steane code: the quantum recovery at level `k`.**~~ **SPLIT 2026-09-27 into the discretization step (done) and #86** — the price was wrong, and the reason is worth recording: the level-`(k+1)` code is seven blocks each carrying a level-`k` encoded qubit, and stating *that* needs a **tensor layer over the blocks** which the repository does not have. **The discretization step is DONE 2026-09-26** ([`Empirical/QM/QEC/SteaneArbitrary.lean`](../CsdLean4/Empirical/QM/QEC/SteaneArbitrary.lean), Category 3, 4 pins; plus ★★ `recovery_apply_lin_comb` in `KnillLaflamme.lean`, 2 pins): the correction property is **bilinear in the two error slots** and a channel is linear, so a recovery corrects every linear combination of its family, up to the scalar `∑ᵢ |aᵢ|²` (for a stabiliser code the Knill–Laflamme matrix is the identity, so the cross terms vanish); whence ★★★ `steane_recovery_arbitrary` — **one channel corrects an arbitrary `2 × 2` error on any one of the seven qubits**, returning the code state scaled by `tr(MᴴM)/2` — and ★★★ `steane_recovery_unitary_qubit`, exactly for a unitary error. The bridge is ★ `gateOf_eq_sum_pauliMat`: lifting a `2 × 2` matrix to one qubit is linear and sends the four Paulis to the code's own single-qubit Pauli strings. **This closes the gap `SteaneRecovery.lean`'s honest scope named** and is exactly what #86's base case consumes. | ~~**M–L**~~ split | Medium | #14; #51; #86 | > | ~~62~~ | QIT | QC chain (link 12) | **SPLIT INTO FOUR 2026-10-02 — the remainder is rows 94–97, each pickable by number; what landed under this number is recorded below.** *new (2026-09-23)* **The circuit-level threshold theorem** (Aharonov–Ben-Or): faulty gates and measurements, error propagation through the recovery gadgets, extended rectangles, the recursive simulation. ~~Program-verification scale; the chain document keeps it **not claimed**.~~ **SPLIT INTO FIVE 2026-09-28, AND (a) IS DONE THE SAME DAY** — the row stays open at (b)–(e), each pickable by its letter. **(a) FAULTY GATES, AND THE CIRCUIT ACCOUNTING — DONE 2026-09-28** (NEW Cat-1 [`Mathlib/Probability/CircuitThreshold.lean`](../CsdLean4/Mathlib/Probability/CircuitThreshold.lean), NEW Cat-3 [`Empirical/QM/QEC/SteaneFaultyGate.lean`](../CsdLean4/Empirical/QM/QEC/SteaneFaultyGate.lean), plus ★★ `blockKron_replace_eq_gateOf_mul` in `BlockKron.lean`; 13 pins; it took M). **The noise moves off resting qubits and onto the gates.** ★★ `faultyTransversal_eq` — **error propagation**: a fault at one location of a transversal gadget, acting after the ideal single-qubit gate there, **is** the ideal gadget followed by a weight-one error (`gateOf i E * blockKron A`) — which is the reason a transversal gadget is fault-tolerant at all: one fault cannot spread inside a block. ★★★ `steane_faultyTransversal_recovery` and ★★★ `steane_faultyTransversal_recovery_unitary`: one channel returns the **ideal** output of the faulty gadget whenever that output is a code state — up to #61's scalar `tr(EᴴE)/2` in general, exactly for a unitary fault. The concrete instance needs the code's logical gate: `blockKron_pX_eq_pauliMat` — **the transversal `X` is the all-ones Pauli**, the code's `X̄` (a product of `X`s is `1` exactly on `w = z + 1⃗`) — with ★ `logicalX_density`, whence ★★★ `steane_faultyLogicalX_recovery`: **a faulty transversal logical `X̄` on an encoded qubit `a|0̄⟩ + b|1̄⟩`, one location bad and the fault an arbitrary unitary, is corrected — the recovery returns the ideally flipped `b|0̄⟩ + a|1̄⟩`**. The accounting half is what code capacity never had: ★ `card_concatPat` (a level-`k` gadget has `n^k` fault locations, its pattern space `2^{n^k}` elements — the overhead of the recursive simulation), ★★ `circuitMeasure_circuitBad_le` — **the union bound over the circuit's `N` locations** on top of #51's recursion, `N (c p)^{2^k} / c` — ★ `mul_codeCapacityBound_lt_of_log_div_lt` (any `k` with `2^k > log(cε/N)/log(cp)` does; the right-hand side is a single logarithm, so the level grows like `log log (N/ε)`), and ★★★ `exists_level_circuitMeasure_lt` with its Steane instance ★★ `steane_circuit_threshold`: **below `p < 1/21`, a circuit of `N` transversal gadgets reaches every accuracy at some concatenation level**. Honest scope, in both headers: the recovery gadget is still perfect, the quantum statement is one fault in one gadget at level one, and `Fin N` indexes locations that do nothing — there is no circuit datatype, because the union bound needs none and claiming more would be claiming (d). **(b) THE TRANSVERSAL CLIFFORD GATES OF THE STEANE CODE — DONE 2026-09-28** (NEW Cat-1 [`Mathlib/QuantumInfo/TransversalClifford.lean`](../CsdLean4/Mathlib/QuantumInfo/TransversalClifford.lean) + `SteaneFaultyGate.lean`; **M–L** took M; the two-block `CNOT` is re-lettered (f)). The general half first: ★★ `pauliMat_eq_blockKron` — **a Pauli string IS the tensor product of its single-qubit Paulis** `onePauli (aᵢ) (bᵢ)`, which is obvious on paper and has to be said once in Lean because `pauliMat` is defined by entries and `blockKron` by a product — then ★ `hGateM_conj_onePauli` (the four cases `H I H = I`, `H X H = Z`, `H Z H = X`, `H(XZ)H = −XZ`, proved with **one** `H` on each side so the `√2` is never squared) raised to the tensor power: ★★ `hadTransversal_conj_pauliMat` — `H^{⊗n} X^a Z^b H^{⊗n} = (−1)^{a·b} X^b Z^a`, **the mechanism of transversality in one equation**. The Steane half: `swapHalves` (in `Fin 6`, `k ↦ k + 3`) exchanges the family's `X`- and `Z`-labels (`steaneA_swapHalves`), the CSS condition `bdot_steaneA_steaneB` — the corpus's own `bdot_rowComb_rowComb` — kills the sign, and the swap is a bijection of the label space, so ★★ `hadTransversal_conj_steaneProj`: **the transversal Hadamard preserves the code projector**, hence ★ `hadTransversal_code_state` and ★★ `hadTransversal_conj_logicalX`/`_logicalZ` — **it exchanges `X̄` and `Z̄`**, the logical Hadamard in the Heisenberg picture. With (a)'s theorem that gives ★★★ `steane_faultyHadamard_recovery`, and the `Z` mirror of (a)'s `X̄` gives `blockKron_pZ_eq_pauliMat` + ★ `logicalZ_density` + ★★★ `steane_faultyLogicalZ_recovery`. (a)'s own `X̄` proof was **replaced** by the two-line corollary of the factorisation. Honest scope: the Hadamard's logical action is Heisenberg-picture and projector-level — `H^{⊗ 7}|0̄⟩ = (|0̄⟩+|1̄⟩)/√2`, the MacWilliams/coset computation, is not proved. **(c) FAULT-TOLERANT SYNDROME EXTRACTION AND THE EXTENDED RECTANGLE — SPLIT INTO THREE 2026-10-02, each pickable by its label.** The corpus measures the syndrome as a projective measurement (`SyndromeRecovery.lean`); the circuit-level theorem needs it as a *circuit*. **(c1)** the extraction circuit for one stabiliser generator: an ancilla block, the transversal `CNOT`s as the label permutation (f) already built (`permMat`), the syndrome read as a projector, and the no-fault theorem that it returns the syndrome and leaves a code state alone (**M–L**). **(c2)** the **verified ancilla**: the cat state, its verification measurement, and the propagation statement that one fault inside the preparation produces at most one data error or else is flagged — this is the mathematical content of *fault-tolerant measurement* and the piece the row has never had (**L**). **(c3)** the **extended rectangle**: recovery, gadget, recovery, with the goodness predicate, delivered as a `IsCorrectedStep` instance for a *faulty* recovery — which (d1)'s theorem then composes with no restatement (**M–L**). **(d) THE SIMULATION THEOREM — (d1) DONE 2026-10-02, (d2) REMAINS.** **(d1) THE COMPOSITION HALF, DONE** (NEW Cat-1 [`Mathlib/QuantumInfo/FaultTolerantComposition.lean`](../CsdLean4/Mathlib/QuantumInfo/FaultTolerantComposition.lean) + NEW Cat-3 [`Empirical/QM/QEC/SteaneCircuit.lean`](../CsdLean4/Empirical/QM/QEC/SteaneCircuit.lean); 7 pins; **M** took M): the row had per-gadget theorems and a probabilistic accounting with nothing in between — no circuit. `IsCorrectedStep P ideal step` names the **two** conditions a gadget must meet to be composable (the step returns the *ideal* output on code states, and the ideal gadget keeps code states in the code — the second is what hands the next gadget its hypothesis), and ★★★ `correctedRun_eq_idealRun` is the induction: **a circuit whose every gadget is faulty and whose every fault is corrected computes exactly what the ideal circuit computes.** `step` is an arbitrary map on density operators, so (c3)'s faulty recovery will plug into the same theorem. Instantiated for the Steane transversal Hadamard, the one gadget whose projector-level code preservation is proved: ★★ `isCorrectedStep_hadStep`, ★★★ `steane_hadamardCircuit_corrected` (a list of locations and unitary faults, one per gadget), and the concrete form ★★★ `steane_hadamardCircuit_even` — **an even-length circuit of faulty Hadamard gadgets returns the encoded state untouched**, `2j` gadgets and `2j` arbitrary unitary faults. Honest scope in both headers: the recovery gadget is still ideal (that is (c3)), only the Hadamard family is instantiated (`X̄`/`Z̄` need the all-ones Pauli's projector-level preservation, the two-block `CNOT` a register view), and there is one level. **(d2) LEVEL REDUCTION AND THE PROBABILISTIC JOIN — the row's real weight** (**L–XL**): the recursive simulation in which a level-`k` gadget *is* a level-`(k−1)` circuit, on top of #86's concatenated recovery for good patterns, and then the join with (a)'s union bound into one statement — with probability at least `1 − N(cp)^{2^k}/c` the circuit's output is exactly the ideal one. **(e) THE POLYLOG OVERHEAD — DONE 2026-09-28** (`CircuitThreshold.lean`; **S** took S): `pow_eq_rpow_logb` is the identity `L^k = (2^k)^{log₂ L}`, ★★ `exists_level_two_pow_le` takes `k = ⌈log₂ X⌉ + 1` (`X = max 1` the logarithm ratio of (a)) and shows it both meets the accuracy and keeps `2^k ≤ 4X` — the level is **doubly logarithmic** — and ★★★ `exists_level_overhead_le`: the gadget's `L^k` fault locations are at most `(4X)^{log₂ L}`, **a fixed power of a logarithm of `N/ε`**, which is what “polylogarithmic overhead” means in the threshold theorem. The honest-scope phrase (a) had left in `CircuitThreshold.lean` is gone, and so is its `check-claims` open-scope declaration. **(f) THE TWO-BLOCK TRANSVERSAL `CNOT` — DONE 2026-09-28** (NEW Cat-3 [`Empirical/QM/QEC/SteaneTransversalCNOT.lean`](../CsdLean4/Empirical/QM/QEC/SteaneTransversalCNOT.lean) + a `permMat` layer in `TransversalClifford.lean`; 14 pins; **M–L** took M–L). **The gadget is a label permutation**, `(u, v) ↦ (u, u + v)` on the two blocks' labels, so it is `permMat cnotPerm` — Hermitian, self-inverse and unitary for free (`permMat_conjTranspose`, `permMat_mul_self`, `permMat_mem_unitaryGroup`, with the row/column reindexing lemmas `permMat_mul_apply` / `mul_permMat_apply`; the same construction serves the logical `CNOT`). **The only mathematical input is that the code is LINEAR**: `steaneEnc`'s entries are coset counts (`cosetAmp`, `steaneEnc_eq_cosetAmp` — the logical states are the uniform superpositions over the row space and its coset), and ★★ `cosetAmp_shift` says that adding a vector of the coset of `x` carries the coset of `y` to the coset of `x + y` — an `Equiv.addRight` reindexing of the row space plus a 32-case `decide` on the two-element-field arithmetic — whence ★★ `steaneEnc_mul_shift` and the capstone ★★★ `cnotT_mul_pairEnc`: **`CNOT^{⊗ 7}·(V ⊗ V) = (V ⊗ V)·CNOT`, the transversal `CNOT` IS the logical `CNOT`**, in the Schrödinger picture and with no amplitudes touched (★★ `cnotT_conj_pairEnc_code` is the density-operator form). **The two-block recovery** comes from #86's encoder-form Knill–Laflamme: the family is one single-qubit Pauli per block (`pairFam`), ★★ `encoderKL_pairFam` composes the single-block condition (`steaneEnc_pauli_pair`) over the blocks in one `blockKron` rewrite, and ★★ `exists_pair_recovery` is the channel. Then ★★★ `steane_faultyCNOT_recovery`: **a fault at one of the seven `CNOT` locations — an arbitrary unitary on each of that location's two qubits — is corrected, the recovery returning the ideal output, the logical `CNOT` applied to the encoded pair**; the fault is expanded in the family by `pairFault_eq_sum` (`blockKron_sum_smul` over the two blocks with #61's `gateOf_eq_sum_pauliMat` in each) and the scalar is pinned to `1` by `smul_eq_one_of_unitary`. Honest scope: the fault is taken in **product form** `A ⊗ B` (products span the pair's operators and `recovery_apply_lin_comb` would lift it, but the general non-product fault would need a pair-register view this file does not build), and the recovery is the two-block Knill–Laflamme channel, not two independent per-block ones. **With (b), the whole Clifford generating set of the Steane code is now covered.** | ~~**XL+**~~ (a) done (M); (b) done (M); (e) done (S); (f) done (M–L); (d1) done (M); **the remainder is rows 94 (M–L), 95 (L), 96 (M–L), 97 (L–XL)** | Low–medium — the document's chain does not need it | #51; #61; #86 | > | ~~63~~ | QIT | One QM, many formulations | ~~*new (2026-09-23)* **Weyl quantisation and the phase-space expectation formula.**~~ **DONE 2026-09-29** ([`WignerWeyl.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WignerWeyl.lean), 7 pins; M–L took M). ★★★ `integral_integral_wigner_mul_conj` — **the overlap identity** `∫∫ W_φ conj (W_ψ) = |∫ φ conj ψ|²`, so phase space recovers the inner product and the Wigner map loses nothing; ★★ `integral_integral_wigner_sq` is its purity form. ★★★ `integral_conj_mul_weylOp` — **the Weyl expectation formula** `⟨ψ, Op(a) ψ⟩ = ∫∫ a W_ψ`. **THE TWO DESIGN CHOICES THAT MADE IT M INSTEAD OF L:** (i) the overlap identity needs no two-dimensional change of variables — slicewise Plancherel (`SchwartzMap.integral_inner_fourier_fourier`) turns the `ξ`-integral into the `y`-integral, the integrand factors as `G(x + y/2) conj (G(x − y/2))` for the **single Schwartz function** `G = φ conj ψ` (`mulConj`), and that is `2·(G ⋆ conj G)(2x)`, so Mathlib's `integral_convolution` finishes it; (ii) the **symbol is presented in kernel form** as a family `b : ℝ → 𝒮(ℝ, ℂ)` of Fourier slices, so the operator's `ξ`-integral converges by fiat instead of by an oscillatory-integral argument, and the proof is two one-dimensional substitutions (`integral_comp_affine_shift`) plus one Fubini swap whose integrability comes from a dominating `M u` for the slices. Residue: #92. | ~~**M–L**~~ done | Medium — makes `W` the state in phase space, not only its marginals | #45; [`Wigner.lean`](../CsdLean4/Mathlib/Analysis/Fourier/Wigner.lean) | > | ~~64~~ | QIT | One QM, many formulations | *new (2026-09-23)* **The Moyal bracket and the classical limit.** ▸ **ROW CLOSED 2026-10-21** — the interacting Wigner equation, its last deliverable, is ★★★ `hasDerivAt_wignerL2_schrodinger_eq` in [`WignerPotential.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WignerPotential.lean) (~~#167~~–~~#169~~). **THE BRACKET AND THE `ℏ²` BOUND DONE 2026-10-01** ([`MoyalBracket.lean`](../CsdLean4/Mathlib/Analysis/Fourier/MoyalBracket.lean), 385 lines, 12 pins). For `H = 2π²ξ² + V` the kinetic half is **exactly** classical and was already proved as a flow (★★ `wigner_freeSchrodingerS`); this row's content was the potential half, and it is now: ★★★ `moyalPot_eq_comm` — the potential term **is** the Wigner transform of `−i[V, ·]`, so the integral against the symmetric difference `V(x + y/2) − V(x − y/2)` is the commutator and not an analogy; ★★★ `norm_moyalPot_sub_poisson_le` — **`‖moyalPot − (1/2π)V'(x)∂_ξW‖ ≤ (sup|V'''|/24)·∫|y|³|K_ψ|`**, the textbook `1/24`, which in the scaling `ξ = p/2πℏ` is the `ℏ²` remainder; and ★★★ `moyalPot_eq_poisson_of_third_deriv_zero` — **when `V''' = 0` the two brackets are equal**, so with the free case the whole exactly-classical family (uniform force, harmonic oscillator) is covered. **WHAT IT COST, AND WHAT THE PIN LACKED:** the estimate ★★ `norm_symmDiff_sub_le` (`|V(x+h) − V(x−h) − 2hV'(x)| ≤ Ch³/3`) had to be proved by three integrations of symmetric differences, because Mathlib's Taylor machinery expands on one side of a point and the cancellation here is *between* the two sides; and the Poisson term needed ★★ `deriv_wigner_right`, the `ξ`-derivative of `W` as the factor `−2πiy` inside the integral. **THE TIME-DEPENDENT FORM, FREE PART, DONE 2026-10-01** ([`WignerEvolution.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WignerEvolution.lean), 10 pins): ★★★ `wigner_liouville` — **`∂_t W_t(x, ξ) = −2πξ ∂_x W_t(x, ξ)` for `W_t = W_{U₀(t)ψ}`**, i.e. `∂_t W + p ∂_x W = 0` in the momentum `p = 2πξ` (★★ `wigner_liouville_add_eq_zero`) — the classical Liouville equation as a **PDE satisfied by the quantum phase-space density**, both derivatives of the same function at the same point. Underneath: ★★ `hasDerivAt_wigner_left`, **`∂_x W_ψ = W(ψ', ψ) + W(ψ, ψ')`** in the cross Wigner functions of the first half, the derivative passing under the Fourier integral against the bound ★ `norm_mul_norm_le` — `1 + y² ≤ 2(1 + (x+y/2)²)(1 + (x−y/2)²)`, so the two Schwartz bounds multiply and the dominating function `4CD/(1+y²)` has **no `x` in it at all**; and ★ `hasDerivAt_wigner_freeSchrodingerS_pos`/★★ `_time`, the two derivatives of the exact transport. **MY OWN MEASUREMENT OF THIS ROW WAS WRONG AND IS CORRECTED HERE:** the row said `∂_x W` "would need differentiation under the Fourier integral in the *parameter*, absent from the pin — the same gap as #92(a)". The pin **does** have first-order differentiation under the integral sign (`Mathlib/Analysis/Calculus/ParametricIntegral.lean`, `hasDerivAt_integral_of_dominated_loc_of_deriv_le`), which is all `∂_x W` needs; what is absent is the **all-orders-with-bounds** version (that file has no `ContDiff`/`iteratedFDeriv` lemma), and that is #92(a)'s gap, not this one. So this part cost **M**, not XL. **STILL OPEN, AND THE OBSTACLE IS THE PROPAGATOR, NOT THE DERIVATIVE:** the **interacting** equation `∂_t W_{U_V(t)ψ} = −2πξ∂_xW + moyalPot`. The corpus's propagator for `H₀ + V` is a Trotter limit on `L²` (`BoundedPerturbation.lean`) with no differentiability in `t` on Schwartz functions, so it needs the unbounded-operator layer. **THAT LAYER IS NOW UNDER CONSTRUCTION AND ITS CONTINUUM FLOOR LANDED 2026-10-02** — [`MultiplicationOperator.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/MultiplicationOperator.lean) (multiplication by a real measurable function is self-adjoint on its natural domain, with a resolvent off the closure of its values and every essential value in its spectrum; the cut-offs `1_{|m| ≤ n}` exhaust the space, which replaces every limiting argument) and [`FreeHamiltonian.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/FreeHamiltonian.lean) (★★★ `isSelfAdjoint_freeSymbolOp`, ★★★ `spectrum_freeSymbolOp` = the nonnegative real axis, both inclusions — the first unbounded self-adjoint operator of the continuum here, and its spectrum). **WHAT THIS ROW NOW WAITS ON, PRECISELY, AND NONE OF IT IS A TRICK:** ~~(i) conjugating a `LinearPMap` by a unitary, so the momentum-space operator becomes `−½Δ` on position space (MATHLIB-ABSENT(LinearPMap.conjIsometry));~~ **(i) DONE 2026-10-09** ([`InnerProductSpace/LinearPMapConj.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/LinearPMapConj.lean) + the position representation in [`FreeHamiltonian.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/FreeHamiltonian.lean), 13 pins; S–M as priced). ★★★ `LinearPMap.conjIsometry` is `U T U⁻¹` with domain `U(dom T)`, and the two transfers that make it worth having are ★★★ `adjoint_conjIsometry` — **the adjoint conjugates** — with ★★★ `IsSelfAdjoint.conjIsometry`, and ★★★ `spectrum_conjIsometry` — **the spectrum is unchanged**. **WHAT MADE IT SHORT IS THE ORDER ON `LinearPMap`, NOT A DOMAIN COMPUTATION**: `conjIsometry U T†` is shown to be a *formal* adjoint of `conjIsometry U T` (four inner products, a unitary moving across each time), so Mathlib's `IsFormalAdjoint.le_adjoint` gives one inclusion, and the other is the same statement for `U⁻¹` through ★ `conjIsometry_symm_conjIsometry` (an involution) and ★ `conjIsometry_mono`, with `le_antisymm` closing it. The technical workhorse is ★ `conjIsometry_apply_of_eq`, the conjugate's value at any element whose *coordinate* is an image point: stating it that way is what avoids rewriting under a membership proof, which is where every earlier attempt died. **AND THE APPLICATION LANDED TOO, BECAUSE THE PIN HAD MORE THAN THIS ROW THOUGHT**: the Fourier transform **is** a unitary of `L²` here (`MeasureTheory.Lp.fourierTransformₗᵢ`), so `freePositionOp` is the free Hamiltonian in the **position** representation, with ★★★ `isSelfAdjoint_freePositionOp` and ★★★ `spectrum_freePositionOp` = `[0, ∞)` transferred by the conjugation lemmas and **no second computation**, plus ★★ `freePositionOp_apply` the unitary equivalence. ⚠️ **It is not yet identified with `−½ d²/dx²`** — that needs the Fourier–derivative identity at the `L²` level and the Fourier image of the momentum domain identified as `H²(ℝ)`, which is **#128**; the remaining two pieces of this row take the *operator*, so nothing here waits on it; ~~(ii) a bounded symmetric perturbation of a self-adjoint `LinearPMap` is self-adjoint on the same domain, which gives `H₀ + V` for bounded real `V`;~~ **(ii) DONE 2026-10-09** ([`InnerProductSpace/LinearPMapPerturb.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/LinearPMapPerturb.lean) + `isSymmetric_mulCLM` + the Schrödinger operator in `FreeHamiltonian.lean`, 10 pins; M as priced). ★★★ `IsSelfAdjoint.addCLM` is the bounded case of Kato–Rellich, with ★★★ `adjoint_addCLM` the general statement `(T + V)† = T† + V` (which does **not** need `T` self-adjoint) and `LinearPMap.addCLM` the sum with domain *exactly* `dom T` — Mathlib's `+` intersects domains, so `T + V.toPMap ⊤` has domain `dom T ⊓ ⊤`, equal to `dom T` but not syntactically it, and a perturbation statement wants the domain unchanged on the nose. **THE DOMAIN IS THE WHOLE CONTENT, AND IT NEEDS ONLY BOUNDEDNESS.** Symmetry of `T + V` is three lines and says nothing; with an unbounded operator the issue is always *maximality*, and ★★ `addCLM_adjointDomain` settles it — the functional `x ↦ ⟪y, V x⟫` is continuous outright, so it cannot affect whether `x ↦ ⟪y, T x⟫` is, and **no symmetry of `V` is used there at all**. That is exactly why the bounded case is elementary and Kato–Rellich proper (relative bound `< 1`, the version covering Coulomb) is not: there the domain argument needs the resolvent and a Neumann series. **AND THE HYPOTHESIS IS DISCHARGED, NOT JUST STATED** (#120's lesson): ★★ `MeasureTheory.L2.isSymmetric_mulCLM` — multiplication by a bounded **real** function is symmetric, reality being the whole of it. **THE PAYOFF IS WHY (i) AND (ii) BELONG TOGETHER**: the free Hamiltonian lives in the *momentum* representation and a potential multiplies in the *position* one, so `H₀ + V` is only a statement once both are in the same place — ~~(i)~~ moved `H₀` across the Fourier transform and (ii) adds the potential there, giving `schrodingerOp` and ★★★ `isSelfAdjoint_schrodingerOp`: **the Schrödinger operator is self-adjoint for every bounded real potential**, on the free Hamiltonian's own domain. ⚠️ **Bounded, not relatively bounded**, and nothing about semiboundedness or form sums (KLMN); ~~(iii) **Stone's theorem**, absent from the pin, to turn the self-adjoint generator into the differentiable group the equation differentiates.~~ **(iii) DONE 2026-10-09, AND THIS ROW WAS WRONG ABOUT WHAT IT NEEDED** ([`Semigroup/PhaseGenerator.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/PhaseGenerator.lean), 10 pins; M+M as re-priced). **It does not need Stone's theorem.** Stone's theorem is an *existence* statement — every self-adjoint operator generates a unitary group — and here the group was already explicit (`phaseGroup`, with its group law, unitarity and strong continuity all proved) while ~~(i)~~+~~(ii)~~ made the generator self-adjoint. What was missing is only that the two are *related*, which is a derivative computation. (The corpus **does** have Stone's theorem in finite dimensions, [`Matrix/StoneC1.lean`](../CsdLean4/Mathlib/Analysis/Matrix/StoneC1.lean), in both the `C¹` and the continuity-only forms; the infinite-dimensional existence theorem is still absent and **is not needed here**.) ★★★ `hasDerivAt_phaseGroup_apply`: on the natural domain `d/dt e^{−itκ}f = −iκ e^{−itκ}f` in `L²`. **THE DOMAIN HYPOTHESIS IS THE DOMINATING FUNCTION, not a technical convenience**: the slope's defect is multiplication by `ω_h`, so its squared `L²` norm is the *scalar* integral `∫‖f‖²‖ω_h‖²` (★ `norm_sq_eq_integral`), whose integrand tends to `0` pointwise (★ `tendsto_phaseDefect`, the scalar exponential's derivative) and is dominated by `4‖κ f‖²` — integrable exactly because `f` is in the domain, with ★ `norm_phaseDefect_le` the uniform bound riding Mathlib's global `Real.norm_exp_I_mul_ofReal_sub_one_le`. ★★★ `hasDerivAt_phaseGroup_freeSymbolOp` and ★★★ `hasDerivAt_fourierGroup_freePositionOp` are **the free Schrödinger equation in both representations**, with the generator the *self-adjoint operator* of ~~(i)~~ rather than a fresh object — which is what makes "`H₀` generates the free dynamics" a statement rather than a name; the position form applies the Fourier isometry to the momentum form through ~~(i)~~'s conjugation and recomputes nothing. ★ `phaseGroup_mem_mulDomain`: the group preserves the domain, free because a phase commutes with a multiplication. ⚠️ **The interacting generator is not here**: `schrodinger` already exists as a unitary propagator for a bounded potential (Dyson/Duhamel/Trotter in `BoundedPerturbation.lean`) and ~~(ii)~~ made `H₀ + V` self-adjoint, but differentiating the Duhamel integral — the *mild* solution being a *classical* one — needs more than the domain invariance proved here, and that is **#129**. Nothing about the classical limit waits on it: the brackets are compared at the generator, and the free evolution is now an equation. ▸ **2026-10-21: ~~#129~~ and ~~#154~~ landed, so the interacting equation is unblocked and RE-ROWED as ~~#167~~ (the Wigner function on `L²` + `∂_t W = W(ψ̇, ψ) + W(ψ, ψ̇)` along the orbit — DONE), ~~#168~~ (the free half `= −2πξ ∂_x W` on `dom H₀` — DONE) and ~~#169~~ (the potential half `= moyalPotL2` and the assembled equation `∂_t W_{ψ_t} = −2πξ ∂_x W_{ψ_t} + moyalPotL2 V ψ_t ψ_t` — DONE 2026-10-21). This row holds nothing else and is closed.** | ~~**L**~~ bracket + bound **done** (M–L); time-dependent form: **free part done** (M, the measurement that called it XL was wrong — see the row); ~~**interacting part XL** on the propagator~~ **all three of this row's named prerequisites are discharged 2026-10-09** — (i) the unitary conjugation, (ii) the bounded perturbation, (iii) the generator identification, and the third showed the row had mis-named its own blocker (it was never Stone's theorem). What remains of the *deliverable* is the interacting equation, which now sits behind **#129 (M–L)** alone — the mild-versus-classical gap for the Dyson propagator. **So the XL was wrong twice**: once on the free part and once here; **interacting equation DONE 2026-10-21** as ~~#167~~ + ~~#168~~ + ~~#169~~ (M + M + M) — **row closed** | Medium–high — the row's stated "what it adds" | #45; #63; [`SchrodingerGroup.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/SchrodingerGroup.lean) | > | ~~65~~ | Geometry | One QM, many formulations · Library owed / upstream | ~~*new (2026-09-23)* **The curvature is the Fubini–Study form.** `GeometricPhaseCurvature.lean` proves `β = −∫∫ 2 Im⟪∂_sΨ, ∂_tΨ⟫` for a family of unit vectors; the corpus's Fubini–Study form is the bundle-level `fsForm` on `ℙ ℂ (Ambient n)` ([`ProjectiveSpaceFubiniStudyForm.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFubiniStudyForm.lean), `dd^c` of the chart potential). The missing piece: for `Ψ` unit-valued, `2 Im⟪∂_sΨ, ∂_tΨ⟫` is (a fixed multiple of) `fsForm` evaluated on the pushed-forward tangent vectors of the projection `[Ψ]`, i.e. the pullback of `fsForm` along `(s, t) ↦ [Ψ(s, t)]` — the chart formula for `dd^c log(1 + ‖z‖²)` on tangent vectors and the homogeneous-coordinate lift; with it, `β = −∫∫ [Ψ]^* ω_FS` literally, and the normalisation (`Ω/2` on `ℂℙ¹`) fixes the multiple.~~ **DONE 2026-09-28** ([`Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceGeometricPhase.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceGeometricPhase.lean), Category 1, 17 pins, and three theorems appended to [`Empirical/QM/BerryPhaseCurvature.lean`](../CsdLean4/Empirical/QM/BerryPhaseCurvature.lean), Category 3, 3 pins; **M** took M). ★★★ `fsForm_eq_neg_two_mul_curvature`: **the curvature `2 Im⟪∂_sΨ, ∂_tΨ⟫` of #54 is `−(1/2)` times the Fubini–Study form of `ℂℙⁿ` evaluated on the velocities of the projected family `[Ψ]`**, hence ★★★ `geometricPhase_eq_half_integral_fsPullback`: `β = (1/2) ∫∫ [Ψ]^* ω_FS`. THE GEOMETRY IS ONE FUNCTION AND TWO OF ITS PROPERTIES: ★ `studyForm ψ u v = Im⟪u,v⟫/‖ψ‖² − Im(⟪u,ψ⟫⟪ψ,v⟫)/‖ψ‖⁴`, the Fubini–Study form on homogeneous coordinates (the `Im` twin of the corpus's own `fsInnerHom`, which is the metric in the same coordinates) — ★★ `studyForm_lift`: it is **unchanged when the lift is rescaled by any `c ≠ 0` and the velocities are shifted by any multiples of the lift**, which is exactly the freedom in lifting a curve of rays; and ★ `studyForm_of_unit`: for a *unit* lift it is plainly `Im⟪u,v⟫`, because `⟪ψ,∂ψ⟫` is purely imaginary along a family of unit vectors (★ `re_inner_deriv_eq_zero`) so the correction term is real. Between those two sits the bookkeeping: ★★ `fsModelForm_eq_studyForm` — the chart form `dd^c log(1 + ‖z‖²)` **is** `−4 studyForm` of the lifts `insertOne i w`, `insertZero i u` (the `Im` twin of `fsMetric_eq_fsInnerHom`, and where the `−4` of `fsChartForm_zero` enters); `chartVelCLM i v` — the differential of the affine chart on representatives — with ★ `hasDerivAt_coordRatio` (the chart path of a differentiable curve of representatives is differentiable: the quotient rule, coordinate by coordinate) and ★ `insertZero_chartVelCLM` (the lifted chart velocity is `(v i)⁻¹ • d` up to a multiple of `v` — precisely what `studyForm_lift` absorbs), giving ★★ `studyForm_chartVelCLM`: **the chart form on the chart velocities is the homogeneous form on the representative**, with every trace of the chart gone. ★★ `hasMFDerivAt_projCurve` makes the tangent vectors the **pushforward**: the manifold derivative of `y ↦ [γ y]` is the derivative of its chart path, so the identification is the pullback `[Ψ]^* ω_FS` and not a formula in a chart. THE NORMALISATION IS CHECKED, NOT CHOSEN: on `ℂℙ¹` the cone's disc gives ★★ `integral_fsPullback_coneSurface` — **the Fubini–Study form integrates to `−Ω` over the cap** (`Ω = 2π(1 − cos θ)` the solid angle) — so ★★ `geometricPhase_coneCurve_eq_half_integral_fsPullback` is BP-2's `−Ω/2`, the same number three ways now (Berry's direct computation, the curvature integral, the Fubini–Study integral). Honest scope: the disc is still a rectangle (Green's theorem in a chart, as in #54) and the `∫∫` is an iterated interval integral of the pullback's component — integration of a `2`-form over a `2`-chain is defined nowhere in the corpus and is not claimed; the bundled two-parameter pushforward is the new row #89, **done 2026-09-28** (`fsPullbackDifferentialForm`). | ~~**M**~~ done (M) | Medium — makes #54 the textbook statement | #54; #8; #89 | > | ~~66~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **T-injection, pure-state form (`R-006` (a)).**~~ **DONE 2026-09-23** ([`Mathlib/QuantumInfo/MagicInjection.lean`](../CsdLean4/Mathlib/QuantumInfo/MagicInjection.lean), Category 1, with #67; 6 pins; S–M took S–M): `dataAnc ψ φ : QReg 2` (data wire `0`, ancilla wire `1`), `sCorr m` = `S^m` on the data, `injectSlice φ m ψ` = `(S^m ⊗ ⟨m\|) · CNOT₀₁ · (ψ ⊗ φ)`; ★★ `injectSlice_magicState` — **the injection identity** `injectSlice magicState m ψ = (1/√2) e^{imπ/4} • T ψ` for both outcomes (four coordinate cases; `T² = S` closes the `m = 1, x = 1` case), `norm_sq_injectSlice_magicState` — each outcome has probability `½`; ★ `injectSlice_zMagicState` — with the resource `Z\|A⟩` the circuit yields `(1/√2)(−e^{iπ/4})^m • Z T ψ`: **the resource's error is transferred to the output**. | ~~**S–M**~~ done | Medium — the circuit link 12 names | `R-006`; #67; [`magic-plan.md`](magic-plan.md) | > | ~~67~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **T-injection as a channel (`R-006` (b), closes `R-006`).**~~ **DONE 2026-09-23 — `R-006` CLOSED** ([`MagicInjection.lean`](../CsdLean4/Mathlib/QuantumInfo/MagicInjection.lean)): `injectionKraus φ m` = the matrix of `ψ ↦ injectSlice φ m ψ`, i.e. `(S^m ⊗ ⟨m\|) · CNOT · (1 ⊗ \|φ⟩)`, with the closed forms `c_m • T` and `c'_m • Z T` (`injectionKraus_magicState`, `injectionKraus_zMagicState`; `\|c_m\|² = ½`); `injectionChannel : Channel` (trace-preserving, `sum_kraus_magicState`); ★★ `injectionChannel_apply` — **`Φ(ρ) = T ρ T†` for every `ρ`**: `T` enacted by Clifford gates, one measurement and a classically controlled `S`, consuming one magic state — `discharged_by` of `R-006`; `noisyInjectionChannel p` with the resource `(1 − p)\|A⟩⟨A\| + p Z\|A⟩⟨A\|Z` (Kraus `√(1−p) c_m T`, `√p c'_m ZT`) and ★ `noisyInjectionChannel_apply` — `Φ_p(ρ) = (1 − p) T ρ T† + p Z T ρ T† Z`, the error model distillation (#75–#78) consumes. M took M. | ~~**M**~~ done | Medium | `R-006`; #66; #77 | > | ~~68~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **The `HT` angle is irrational (`R-005` (a)).**~~ **DONE 2026-09-25** ([`Mathlib/QuantumInfo/CliffordTAngle.lean`](../CsdLean4/Mathlib/QuantumInfo/CliffordTAngle.lean), Category 1, 9 pins; M took M): ★ `isIntegral_two_mul_cos_of_eq_two_pi_mul` — **`2 cos(2πq)` is an algebraic integer** for every rational `q` (`e^{iθ}` is a `q.den`-th root of unity, so it is integral — witness `X^d − 1`, `monic_X_pow_sub_C` — and so is `ζ^{d−1} = ζ⁻¹`, and `ζ + ζ⁻¹ = 2 cos θ` by `Complex.two_cos`, transported to `ℝ` through `isIntegral_algebraMap_iff`); a general fact Mathlib lacks at the pin. ★ `not_isIntegral_of_sq_add_self_eq` — **no algebraic integer satisfies `x² + x = 7/4`** (it would make `7/4` integral, and `ℤ` is integrally closed in `ℚ`: `IsIntegrallyClosed.isIntegral_iff` through the `GCDMonoid` instance). The gate: `hGateM`, `tGateM`, `htht = T·H·T·H`, `trace_htht : tr = (1 + e^{iπ/4})²/2`, `det_htht : det = e^{iπ/2}`, and ★ `trace_htht_eq_cos_htAngle : tr = 2 cos(htAngle/2)·e^{iπ/4}` with `htAngle = 2 arccos(cos²(π/8))` — so `e^{−iπ/4}·T·HTH` is an `SU(2)` rotation by `htAngle`, `cos htAngle = (2√2 − 1)/4` (`cos_htAngle`), whence `(2 cos htAngle)² + 2 cos htAngle = 7/4`; ★★ `irrational_htAngle_div_two_pi` — **the angle is an irrational multiple of `2π`**; ★ `denseRange_zsmul_htAngle` (`AddCircle.denseRange_zsmul_coe_iff`) and ★★ `exists_zsmul_htAngle_approx` — **every angle is `k·htAngle` mod `2π` to any accuracy** (`dense_addSubgroupClosure_pair_iff`, `AddSubgroup.mem_closure_pair`), the form the gate-approximation argument consumes. | ~~**M**~~ done | Medium — the number-theoretic core of universality | `R-005`; #69 | > | ~~69~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **`⟨H, T⟩` is dense in `U(2)` modulo phase (`R-005` (b)).**~~ **SPLIT 2026-09-25 into #80 (done) and #81** — the price was wrong. The Euler decomposition `U = R_u(α) R_v(β) R_u(γ)` for two *arbitrary* non-parallel axes is not an M lemma: Nielsen–Chuang states it as an exercise without proof, and a real proof needs either a degree/continuity argument or a reduction to orthogonal axes. #80 builds the axis–angle layer and the one-axis density that #68's angle result was for; #81 carries the orthogonal-axis reduction, the textbook `z`-`y`-`z` decomposition and the assembly, with the route recorded. | ~~**M–L**~~ split | Medium | `R-005`; #68; #80; #81 | > | ~~70~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **Every unitary is a product of two-level unitaries (`R-005` (c)).**~~ **DONE 2026-09-25** ([`Mathlib/LinearAlgebra/Matrix/TwoLevel.lean`](../CsdLean4/Mathlib/LinearAlgebra/Matrix/TwoLevel.lean), Category 1, 9 pins; M took M): `IdOutside s U` (agrees with the identity outside the block `s × s`) with ★ `IdOutside.mul` (the product sum collapses), `.mono`, `.eq_one`, `.list_prod`, and `IsTwoLevel U = ∃ i ≠ j, IdOutside {i, j} U`; `twoLevelMat i j a b c d` with ★ `twoLevelMat_mul` — **composition is `2 × 2` matrix multiplication** (through `mul_apply_of_idOutside_pair`, which collapses the sum once and for all) — `twoLevelMat_conjTranspose`, `twoLevelMat_one` and ★ `twoLevelMat_mem_unitaryGroup` (unitary exactly when the block's columns are orthonormal); the Givens rotation `givensMat i j v` with ★ `givensMat_mulVec_apply_snd` (**it kills the `j`-component**), `givensMat_mulVec_apply_fst` (the `i`-component accumulates `√(|vᵢ|² + |vⱼ|²)`) and `givensMat_mem_unitaryGroup`; ★★ `exists_clear_column` — **a product of two-level unitaries kills any prescribed set of components of a vector**, by induction on the set; `row_eq_of_col_eq` (unitarity turns a cleared column into a cleared row); ★★ `exists_twoLevel_prod` — **every unitary is a product of two-level unitaries**, by strong induction on the support (clear the `i`-th column, fix its phase with one more two-level factor, recurse on `s.erase i`, and undo with the adjoints). The `d(d − 1)/2` count is #82, **done 2026-09-28** (`exists_twoLevel_prod_length`, in the same file). | ~~**M**~~ done | Low–medium — Mathlib-grade linear algebra | `R-005`; #71; #82 | > | ~~71~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **A two-level unitary on `n` qubits is a Gray-code sequence of CNOTs and one multiply-controlled single-qubit gate (`R-005` (d)).**~~ **DONE 2026-09-26** ([`Mathlib/QuantumInfo/MultiControlled.lean`](../CsdLean4/Mathlib/QuantumInfo/MultiControlled.lean), Category 1, 8 pins; M took M): on the qubit basis `Fin m → Fin 2`, `ctrlGate j pat a b c d` is the **multiply-controlled gate** (the `2 × 2` block on qubit `j` of the labels agreeing with `pat` away from `j`, the identity elsewhere) and ★ `ctrlGate_eq_twoLevelMat` is the exact bridge to #70: it **is** the two-level matrix on the adjacent pair `pat[j ↦ 0]`, `pat[j ↦ 1]`, so a multiply-controlled `X` is the transposition of two labels differing in one bit (★ `isMultiCtrlX_swapMat`) and ★ `exists_ctrlGate_of_adjacent` is its converse; `swapMat u v` with ★ `swapMat_apply` (its entries are the permutation's) and ★ `swapMat_conj_apply` — **conjugating by a transposition relabels**, `(S V S) k l = V (swap k) (swap l)` — whence ★ `idOutside_swapMat_conj`: the block moves to the swapped pair; ★★ `exists_multiCtrl_sandwich` — **every two-level matrix on qubits is `L · W · L`** with `L` a list of multiply-controlled `X`s (each its own inverse) and `W` one multiply-controlled gate, unitary when the matrix is, proved by recursion on `(diffSet u v).card` — **the Gray code is the recursion**: flip one differing qubit, conjugate, the differing set shrinks; ★ `exists_multiCtrl_list` — the flat product form #72 consumes. No gate counts (#82, #74). | ~~**M**~~ done | Low–medium | `R-005`; #70; #72 | > | ~~72~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **`C^k(U)` from CNOT and single-qubit gates (`R-005` (e)).**~~ **SPLIT 2026-09-26 into #83 (done), #84 and #85** — the price was wrong, and the reason is worth recording: the row offered the work-qubit Toffoli ladder as an alternative to the square-root recursion, but **the ladder is not a matrix identity** — compute/uncompute reproduces `C^k(U)` only on the sector where the work qubits are `0`, and #73 consumes an exact product. So the ancilla-free route is forced, and it needs the `z`-`y`-`z` Euler decomposition for one control (#84) and square roots of `2 × 2` unitaries for the control-count recursion (#85), each M–L on its own. #83 is the vocabulary all three are stated in. | ~~**M–L**~~ split | Low–medium | `R-005`; #71; #83; #84; #85; #73 | > | ~~73~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **Clifford+T is dense in `U(2ⁿ)` modulo phase (`R-005` (f), closes `R-005`).**~~ **DONE 2026-09-26 — `R-005` IS CLOSED** ([`Mathlib/QuantumInfo/CliffordTUniversal.lean`](../CsdLean4/Mathlib/QuantumInfo/CliffordTUniversal.lean), Category 1, 10 pins; S–M took M). ★★ `exists_phase_mem_cliffordTmLim`: **every unitary on `m ≥ 1` qubits becomes a limit of Clifford+T circuits after multiplication by one global phase.** The row assembles the chain and supplies the three joints it was missing. (i) The **exact** product, ★★ `mem_closure_elementary`: every unitary is a finite exact product of `CNOT`s and single-qubit gates — #70's two-level decomposition, #71's Gray-code sandwich, #85's control-count recursion — whose joint is ★ `block_mem_unitaryGroup_of_ctrlSetOf`, **the converse of `ctrlSet_mem_unitaryGroup`**: the earlier rows deliver a controlled gate known to be unitary while #85 wants its `2 × 2` block unitary, and the block's unitarity comes from evaluating `UᴴU = 1` at the labels that differ only on the target, where the sum collapses to two terms. (ii) The **lift**: `gateOfHom` is a monoid homomorphism and `continuous_gateOf` is continuous, so ★ `gateOf_mem_cliffordTmLim` carries #81's one-qubit closure into the `m`-qubit one — and the step that makes *modulo phase* survive is `gateOf_smul`: **a phase on the block is a global phase on the gate**, because the gate is the block tensored with the identity. Were that false, #81's phases would become relative phases and the argument would not close. (iii) The **phases add**: the gates that become limits after one global phase are themselves a submonoid (`phaseLim`), so the exact product inherits a single phase with no list bookkeeping. Honest scope in the header: density not exactness (membership in a topological closure, which is all that is true of a countable gate set), `m ≥ 1`, and **no gate counts** — Solovay–Kitaev is #74 and stays unclaimed. | ~~**S–M**~~ done | Medium — the universality statement the doc names | `R-005` (closed); #70; #71; #81; #85 | > | ~~74~~ | QIT | QC chain (link 12) | **DISCHARGED IN FULL 2026-10-06 — all five rows of its split landed the same day, and so did the sixth that came out of #115.** SOLOVAY–KITAEV IS PROVED WITH ITS GATE COUNT: #112 (commutator contraction), #113 (the net), #114 (every determinant-one unitary is a group commutator), #115 (the error at exponent 3/2), #117 (words, and the `5ⁿ` length recursion) and #116 (the pairing and the logarithms) — ★★★ `exists_word_approx_polylog` in [`SolovayKitaevCount.lean`](../CsdLean4/Mathlib/QuantumInfo/SolovayKitaevCount.lean). **The split is what moved it**: priced XL as one lump it never moved; priced as six bricks it took one day, and two of the six found that the row’s own route was wrong (#108-style) — #115 found the quoted exponent to be a cost-model choice, #116 found its "no new quantum content" estimate wrong. ~~The remainder is rows 112–116, each pickable by number.~~ The row was priced **XL** as one lump, which is why it never moved; the Dawson–Nielsen proof factors into five pieces, three of which are **M or smaller** and two of which the corpus already has the machinery for (#80's axis–angle layer, #81's density). **It is also the last unclaimed item on link 12**: `R-005` is closed by #73 and the `d(d−1)/2` count by #82, so closing #116 finishes the chain. *As opened (2026-09-23):* **Solovay–Kitaev efficiency** — approximation to `ε` with `O(log^c(1/ε))` gates. Not needed by the doc's chain (which claims density, not efficiency); a theorem of its own (N&C Appendix 3: the `ε`-net and the group-commutator shrinking lemma). | ~~**XL**~~ split | Low as correctness, **medium as completion** — it is what is left of link 12 | `R-005`; #73; #82; rows 112–116 | > | ~~83~~ | QIT | QC chain (link 12) | ~~*new (2026-09-26)* **The controlled-gate vocabulary (`R-005` (e1)).**~~ **DONE 2026-09-26** ([`Mathlib/QuantumInfo/ControlledGate.lean`](../CsdLean4/Mathlib/QuantumInfo/ControlledGate.lean), Category 1, 7 pins; M took M): one definition `ctrlSet S j pat a b c d` — the `2 × 2` block applied to qubit `j` of the labels matching `pat` on the control set `S`, the identity elsewhere — covers the whole chain's vocabulary: `S = ∅` is a single-qubit gate (`singleGate`, `xGate`), `S = {a}` with the flip block is `CNOT` (`cnotGate'`), and `S = univ.erase j` is #71's multiply-controlled gate (★ `ctrlSet_erase_eq_ctrlGate`), so the two-level theory of #70 and #71 applies to it verbatim. The algebra is blockwise: ★★ `ctrlSet_mul` — **such gates multiply by multiplying their blocks** (a copy of the `2 × 2` matrices inside the big algebra), through the two-element collapse `sum_collapse_update`/`agree_off_iff` that every computation here uses; ★ `ctrlSet_conjTranspose` transposes the block, `ctrlSet_one`, and hence ★ `ctrlSet_mem_unitaryGroup` — unitary exactly when the block is — with the unitarity of the three special cases. | ~~**M**~~ done | Low–medium | `R-005`; #71; #84; #85 | > | ~~84~~ | QIT | QC chain (link 12) | ~~*new (2026-09-26)* **The `z`-`y`-`z` Euler decomposition and one control (`R-005` (e2)).**~~ **DONE 2026-09-26** (two modules, Category 1, 16 pins; M–L took M–L). [`Mathlib/QuantumInfo/EulerDecomposition.lean`](../CsdLean4/Mathlib/QuantumInfo/EulerDecomposition.lean) proves **Nielsen–Chuang Theorem 4.1** — ★★ `exists_euler`: every `U ∈ U(2)` is `e^{iα} R_z(β) R_y(γ) R_z(δ)` — through ★★ `exists_euler_of_det_one`, and the economy that made it small is worth recording: **passing to determinant one first leaves only two entries to match**, because a determinant-one unitary is `!![a, -star b; b, star a]` (`adjugate_entries`, from `Uᴴ = adjugate U`), so the polar form of `a` and `b` gives `γ = 2 arccos ‖a‖`, `β = arg b − arg a`, `δ = −arg a − arg b` **with no case split** — where a column degenerates the matching entry is killed by `cos (γ/2) = ‖a‖ = 0` or `sin (γ/2) = ‖b‖ = 0`, and the polar identity holds at `0` too. Supporting layer: `rzMat`/`ryMat` as `SU2.axisRot` about `z` and `y`, their unitarity through the new `SU2.su2_mem_unitaryGroup` and `SU2.axisRot_mem_unitaryGroup`, ★ `rz_ry_rz_eq` (the product as an explicit matrix) and the `X`-conjugation lemmas. Then **Corollary 4.2**: ★ `abc_identity` and ★ `abc_prod_eq_one` give ★★ `exists_abc`, `U = e^{iα} A X B X C` with `A B C = 1`. [`Mathlib/QuantumInfo/ControlledSingle.lean`](../CsdLean4/Mathlib/QuantumInfo/ControlledSingle.lean) lands the circuit half, ★★ `ctrlOf_eq_circuit`: **a gate with one control is a product of two `CNOT`s and four single-qubit gates** (Fig. 4.6) — on the control bit `0` the `CNOT`s act trivially and the target sees `A B C = 1`, on the control bit `1` they insert the two `X`s and it sees `e^{iα} A X B X C = U`. The one piece of gate algebra that needs is ★★ `ctrlChoice_mul`: gates applying `M b` to the target when the control carries the bit `b` **multiply blockwise in each branch separately**, and every gate of the circuit is one of them (`gateOf_eq_ctrlChoice`, ★ `diagGate_eq_ctrlChoice` — the phase rides on the control line as `diag (1, e^{iα})`, the only place the control qubit is addressed — `cnotGate'_eq_ctrlChoice`, ★ `ctrlOf_eq_ctrlChoice`), so the whole circuit collapses to one gate and the identity is checked one control bit at a time. **#85 is stated in that algebra.** | ~~**M–L**~~ done | Medium — the single-qubit completeness fact the whole chain leans on | `R-005`; #83; feeds #81 and #85 | > | ~~85~~ | QIT | QC chain (link 12) | ~~*new (2026-09-26)* **The control-count recursion (`R-005` (e3)), closing #72.**~~ **DONE 2026-09-26** ([`Mathlib/QuantumInfo/ControlRecursion.lean`](../CsdLean4/Mathlib/QuantumInfo/ControlRecursion.lean), Category 1, 16 pins with the shared square-root layer; L took L). ★★ `ctrlSetOf_mem_closure`: **a gate with any number of controls, on any control pattern, is a product of `CNOT`s and single-qubit gates** (Nielsen–Chuang Fig. 4.10, ancilla-free), and ★★ `ctrlGateOf_mem_closure` says it for #71's `ctrlGate` — **the exact product #73 consumes**. Two ingredients, both new algebra. (i) *The pattern costs nothing*: ★ `xGate_conj_ctrlSet` — conjugating by one `X` on a control flips that control value, so only the **number** of controls matters; the mechanism is that `X` is a permutation matrix and multiplying by it relabels (`xGate_apply`, `xGate_mul_apply`, `mul_xGate_apply`). (ii) *The recursion needs two targets at once* — `C(V)` on the target and `C^{k-1}(X)` on the last control act on different qubits, which #84's one-control algebra cannot reach. So `pairSet`, the gates acting by a `4 × 4` block on a **pair** of qubits (one block where the pattern matches on the control set, another elsewhere), with ★★ `pairSet_mul`: they multiply blockwise in each branch separately, through a four-element collapse (`sum_collapse_pair`). Every gate of the recursion is one of them, so ★★ `ctrlSetOf_insert` reduces to two `4 × 4` identities that `diagBlock`/`flipBlock` settle in three rewrites: conjugating a block-diagonal gate by the flip swaps its blocks (★ `flipBlock_conj_diagBlock`), whence ★ `diagBlock_five` — `diag(1,V)·X·diag(1,V*)·X·diag(V,V) = diag(V*V, VV) = diag(1,U)` when `V² = U` — while off the pattern the same product is `diag(1,V)·diag(1,V*) = 1`. The square root comes from the new ★★ `Euler.exists_sqrt` (**every `2 × 2` unitary has a unitary square root**, via `exists_axisRot_of_det_one`: a determinant-one unitary is a rotation about some unit axis, so halve the angle). **Row #72 is now closed in full: #83 + #84 + #85.** | ~~**L**~~ done | Medium–high | `R-005`; #83; #84; feeds #73 | > | ~~82~~ | QIT | QC chain (link 12) | ~~*new (2026-09-25)* **The `d(d − 1)/2` count for the two-level decomposition.**~~ **DONE 2026-09-28** ([`Mathlib/LinearAlgebra/Matrix/TwoLevel.lean`](../CsdLean4/Mathlib/LinearAlgebra/Matrix/TwoLevel.lean), Category 1, 4 new pins; **S–M** took S). ★★★ `exists_twoLevel_prod_length`: every unitary on an index type with `d ≥ 2` elements is a product of **at most `d(d − 1)/2`** two-level unitaries. **The phase-fixing factor was never needed.** #70 spent one extra two-level factor per elimination round to turn the pivot `(G·U) i i` into `1` — but `givensMat_mulVec_apply_fst` already puts `√(|v i|² + |v j|²)` there, a *nonnegative real*. So the strengthened ★★ `exists_clear_column` — now also carrying `L.length = s.card`, `normSq ((L.prod *ᵥ v) i) = ∑_{k ∈ insert i s} normSq (v k)` and, once `s` is nonempty, the value `√(∑ …)` itself — leaves the pivot equal to **the norm of the column**, which for a unitary is exactly `1` (`sum_normSq_col`, from `(UᴴU) i i = 1` and `Complex.mul_conj`); `row_eq_of_col_eq` then clears the row, and the recursion step costs `|s| − 1` factors and no more. The count is Pascal's rule, `(m − 1) + Nat.choose (m − 1) 2 = Nat.choose m 2`, with `isTwoLevel_of_idOutside_card_le_two` as the base case — a unitary that is the identity outside at most two indices **is** two-level, so one factor — and `Nat.choose (max m 2) 2` as the carried bound, the `max` absorbing the degenerate one-index case. `exists_twoLevel_prod` is kept uncounted beside it for #71–#73. Honest scope, in the header: an upper bound, which is all Nielsen–Chuang §4.5.1 asserts — no matching lower bound is proved, and it counts two-level factors, not elementary gates (#74). **No residue: nothing new is queued by this row.** | ~~**S–M**~~ done (S) | Low | #70; #74 | > | ~~86~~ | QIT | QC chain (link 12) | ~~*new (2026-09-27)* **The concatenated Steane code: the quantum recovery at level `k`** (the second half of #61).~~ **DONE 2026-09-27** ([`Mathlib/QuantumInfo/BlockKron.lean`](../CsdLean4/Mathlib/QuantumInfo/BlockKron.lean) + the encoder form in [`KnillLaflamme.lean`](../CsdLean4/Mathlib/QuantumInfo/KnillLaflamme.lean), Category 1; [`Empirical/QM/QEC/SteaneEncoder.lean`](../CsdLean4/Empirical/QM/QEC/SteaneEncoder.lean) + [`SteaneConcat.lean`](../CsdLean4/Empirical/QM/QEC/SteaneConcat.lean), Category 3; 18 pins; L took L, on the recorded route). **The quantum statement now holds at every level.** (a) The tensor layer `blockKron M k l = ∏_b M b (k b) (l b)`: multiplicativity and multilinearity are both `Fintype.prod_sum`, and ★ `blockKron_single_eq_gateOf` ties one filled slot back to #83's `gateOf`. (b) **Knill–Laflamme in ENCODER form** `Vᴴ Eᵢᴴ Eⱼ V = cᵢⱼ • 1`, with both directions (★ `knillLaflamme_of_encoder`, ★ `encoder_of_knillLaflamme`) and a recovery (★★ `exists_recovery_of_encoderKL`) — it composes under `blockKron` in one line and needs **no rank statement** relating `steaneEnc steaneEncᴴ` to `steaneProj`. (c) The Steane encoder: isometry from the orthonormality of `|0̄⟩, |1̄⟩`, and the crux ★★ `steaneEnc_gateOf` — **an arbitrary operator on one qubit acts on the encoded qubit as the scalar `(G₀₀ + G₁₁)/2`** (#61's span step read through the encoder) — with ★★ `exists_steaneEnc_gateOf_mul_gateOf` for **two** qubits (a product of two single-qubit Paulis is `±EᵢᴴEⱼ`, so the family's identity matrix does the rest) and ★★ `exists_steaneEnc_blockKron` in the shape the induction consumes. (d) ★★ `exists_concatEnc_step`: a pair of block errors scalar on every block but at most **two** named ones is scalar one level up — two, not one, because Knill–Laflamme is a condition on *pairs* and each error may spend its arbitrary block elsewhere. (e) The tower `cLabel`/`cEnc`/`cFam`/`cErr` and the induction ★★★ `encoderKL_cErr`, hence ★★★ `exists_cErr_recovery`: **one recovery channel per level**. The pattern tie-in: `PatErr k x E` is an error of the pattern `x`, ★★ `patErr_mem_span` shows every **good** pattern's errors (at most one bad sub-block at every level, i.e. `isBad 7 k x = false`) lie in the span of the family, and ★★★ `exists_concat_recovery` / ★★★ `exists_concat_recovery_unitary` give the recovery up to a scalar and — for unitary errors, since both the error and the channel preserve the trace (`smul_eq_one_of_unitary`) — **exactly**. The patterns it says nothing about are `concatBad 7 k` (`good_eq_compl_concatBad`), whose probability is at most `(21 p)^{2^k}/21` (#51's `steane_concatBad_le`): **the encoded failure bound at every level**. Honest scope in the header: code capacity still (perfect encoder and recovery gates — faulty gates are #62); the recovery is the Knill–Laflamme channel, not a block-by-block decoder, and no optimality or efficiency is claimed; the scalar of the non-unitary statement stays existential. | ~~**L**~~ done | Medium — the quantum threshold statement the doc's link 12 stated at level one | #14; #51; #61 (its discretization step); [`Empirical/QM/QEC/SteaneArbitrary.lean`](../CsdLean4/Empirical/QM/QEC/SteaneArbitrary.lean) | > | ~~87~~ | QIT | Library owed / upstream | ~~*new (2026-09-27)* **Gleason for bare real frame functions, `N ≥ 4`.**~~ **DONE 2026-09-28** ([`Gleason/RealFrame.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/Gleason/RealFrame.lean), Category 1, 10 pins; S–M took S–M). ★★★ `existsUnique_density_of_frameFunction`: **for `N ≥ 3` a nonnegative frame function of weight `W` on `ℝᴺ` is the quadratic form of a unique PSD matrix of trace `W`** — the statement Gleason's paper is about, where #58 had the projection-package form. The gap was one fact: the core lemma lives on `ℝ³`, so the restriction of `f` to the span of an orthonormal triple must be a frame function there, i.e. **the weight of a `3`-space must not depend on the basis chosen for it**. A package reads that off `p (∑ᵢ |eᵢ⟩⟨eᵢ|)`; a bare `f` has to earn it, which this file does by completing the triple: `orthonormalBasisOfCardR` (an orthonormal family of the right cardinality **is** a basis, via `basisOfOrthonormalOfCardEqFinrank` — no span argument), `orthonormal_sum_elim` (gluing along `⊕`), ★ `sum_eq_of_basis` (the frame sum is the weight over a basis indexed by *any* finite type) and ★ `sum_elim_eq` (it splits along a completion), `exists_orthonormal_complement` (`stdOrthonormalBasis` of `Sᴿ` plus `finrank_add_finrank_orthogonal`), giving ★★ `exists_weight_restrict`. The same completion pays for two more facts the middle layer needed and a bare `f` did not have: ★ `le_weight` (`f ≤ W` on the sphere in **any** dimension — `Sphere.lean`'s `ℝ³` proof goes through the cross product) and ★ `neg_eq` (`f` is **even** on the sphere: flip one basis vector, keep the tail). #58's middle layer was **refactored** to state A4 for any function quadratic on triples (★ `exists_isSymm_sphere_of_quad`), so the package and the frame function share it; the theorem then follows after normalising by `W` (the `W = 0` case separately) and running the descent. Honest scope: `N ≥ 3`, finite dimensions, and **nothing in the corpus consumes it** — it exists because it is Gleason's own statement and an upstream PR would be asked for it. | ~~**S–M**~~ done | Low — the library's completeness | #58 | > | ~~88~~ | QIT | One QM, many formulations | *new (2026-09-28)* **The general adiabatic theorem (Born–Fock, Kato)** ▸ **SPLIT 2026-10-20 into #158–#165 (author's request), along Kato's proof.** The honest first target the row named — finite dimensions, a `C²` path, an eigenvalue of constant multiplicity isolated by a gap `g` — is taken as the whole target; each step below is a Lean brick with its inputs already in the corpus or the pin named. The chain: Riesz projector of a Hermitian matrix (#158) → its smoothness along a gapped path and `PṖP = 0` (#159) → the adiabatic intertwiner `A` and Kato's intertwining `A(s)P(0) = P(s)A(s)` (#160) → the propagators `U_T`, `U_A` at speed `T` (#161) → **Kato's comparison `‖U_T − U_A‖ ≤ C/T`** (#162, the one L) → the Born–Fock statement for states (#163) → the phase, and BP-1's geometric phase as what survives (#164) → #56's rigid rotation as an instance (#165). Prices: M, M, M, M, L, S–M, M, S — the XL was correct and is now eight rows. ▸ *Original row:* — the residue #56 does not claim: for a `C²` path `s ↦ H(s)` of Hermitian operators with a spectral gap `g > 0` around an eigenvalue, the solution of `i ψ̇ = H(t/T) ψ` started in the instantaneous eigenvector stays in it up to `O(1/T)`, uniformly on `[0, T]` — and then the geometric phase of BP-1 is what survives the limit. #56 proved the **rigidly rotated** case (a conjugation orbit of a single `H₀`) exactly, with the error explicit in `ω` and uniform in `t`; that proof does not generalise, because it uses a *constant* rotating-frame generator. The general route is Kato's: the Riesz projector `P(s)` of the isolated eigenvalue, the adiabatic intertwiner `A(s)` solving `Ȧ = K A` with `K = [Ṗ, P]`, and a Grönwall comparison of two propagators. At the pin Mathlib has none of the pieces — no Riesz projector of a parametrised family, no gap lemma, no propagator of a time-dependent generator (the corpus's own propagators are all constant-generator or rigidly rotated). Finite dimensions with a smooth path and a constant-multiplicity eigenvalue would already be the theorem, and is the honest first target. | ~~**XL**~~ split (#158–#165) | Medium — the textbook theorem, and the one thing #56 leaves unclaimed | #56; #10; [`berry-phase-scoping.md`](berry-phase-scoping.md) BP-5 | > | ~~89~~ | Geometry | Library owed / upstream | ~~*new (2026-09-28)* **The two-parameter pushforward, and the pullback of a form as a form.**~~ **DONE 2026-09-28** (the same day it was priced; extends [`Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceGeometricPhase.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceGeometricPhase.lean) and adds NEW Cat-1 [`Mathlib/Analysis/Calculus/FDeriv/Quotient.lean`](../CsdLean4/Mathlib/Analysis/Calculus/FDeriv/Quotient.lean); 13 pins; **S–M** took M). **The blocker first**: ★ `HasFDerivAt.div` — the quotient rule for the Fréchet derivative of maps into a normed field *larger than the base field*, which Mathlib has for `HasDerivAt` only (and whose `ContDiff` twin needs no new lemma, `ContDiffAt.inv` already being general in the target) — assembled from `hasFDerivAt_inv'` and `HasFDerivAt.mul` as the row predicted. With it ★ `hasFDerivAt_coordRatio` holds for a family indexed by **any** normed space `F`, and ★★ `hasMFDerivAt_projFamily` is the pushforward of `q ↦ [Ψ q]` in that generality — `F = ℝ × ℝ` **is** the two-parameter map the row asked for, and the curve lemma `hasDerivAt_coordRatio` is now proved *from* it rather than beside it. **Then the form**: `fsPullbackForm Ψ h p` = the form of `ℂℙⁿ` at `[Ψ p]` composed with `mfderiv`, so alternation is structural; ★★★ `fsPullbackForm_apply` is the **chart-free formula** `[Ψ]^* ω_FS (a, b) = −4 studyForm (Ψ p) (∂_aΨ) (∂_bΨ)` — no `idx`, no chart, valid for every nonvanishing differentiable family, unit or not — and it is what makes `fsPullbackForm_eq_chart` (the same form read in *any* chart containing the ray) and hence the smoothness possible at all: the atlas's chart choice jumps as `p` moves, the pullback does not. ★★ `contDiffAt_fsPullbackForm` (fixed chart + `contDiff_fsModelForm` + `ContDiffOn.fderivWithin` + the corpus's own jointly-smooth pullback `contDiff_uncurry_compContinuousLinearMap`) gives ★★★ `fsPullbackDifferentialForm` — **`[Ψ]^* ω_FS` as a bundled `C^∞` `2`-form**, a term of `DifferentialForm 𝓘(ℝ, F) F ∞ (Fin 2) ℝ`, which on the rectangle is exactly “a differential `2`-form on the disc's parameter domain”. The bundling cost one small instance, `instHasTranslationAtlasSelf` (a normed space charted on itself is a translation atlas, so a section's local representative is itself), added to `TranslationAtlasForm.lean` beside the circle's; and one general evaluation lemma, `ContinuousAlternatingMap.compContinuousLinearMap_apply_pair`, added to the corpus's own `Alternating/Pullback.lean`. Finally ★★ `fsPullbackForm_apply_coord` — the form on the two coordinate directions **is** #65's function `fsPullback` — so ★★★ `curvature_eq_neg_half_fsPullbackForm` and ★★★ `geometricPhase_eq_half_integral_fsPullbackForm` restate #65's theorems with the form as a form. Honest scope, in the header: this is the pullback built for *this* family from the chart-free formula, not a general “pullback of a form along a smooth map” operator (the corpus defines none), and the `∫∫` is still the iterated interval integral of the form's value on the coordinate directions — integration of a `2`-form over a `2`-chain remains undefined and unclaimed, as in #65 and #54. **No residue: nothing new is queued by this row.** | ~~**S–M**~~ done (M) | Low — presentation, not content | #65; #54 | > | ~~90~~ | QIT | One QM, many formulations | ~~*new (2026-09-28)* **Aharonov–Bohm on the circle**, the continuum residue of #55~~ **DONE 2026-09-30** ([`AharonovBohmCircle.lean`](../CsdLean4/Mathlib/QuantumInfo/AharonovBohmCircle.lean) + the Cat-3 reading in [`AharonovBohm.lean`](../CsdLean4/Empirical/QM/AharonovBohm.lean), 14 pins). ★★★ `twisted_eigen` is the eigenvalue equation `−(∂ − iΦ)²ψₙ = (n − Φ)²ψₙ`, ★★ `isLeast_range_circleEigval` the ground level `dist(Φ, ℤ)²`, ★★★ `exists_eq_of_range_circleEigval_eq` the determination of the flux modulo a quantum and up to sign (with ★★ `range_circleEigval_add_one` and ★★ `range_circleEigval_neg` the converse, so the ambiguity is exact), ★★ `gauge_periodic_iff` the gauge half — `e^{iΦx}` lives on the circle **iff** the flux is whole — and ★ `circleMode_eq_fourier` identifies the modes with Mathlib's `fourier` family. **THE ROW’S PRICE BOUGHT A ROUTE THAT WAS NOT TAKEN, AND THIS IS THE HONEST ACCOUNTING.** The **L** was for an unbounded self-adjoint operator on `L²(S¹)` with a domain; what landed states the eigenvalue equation **pointwise on the Fourier modes** and proves the determination theorem about `Set.range (circleEigval Φ)`, which is S–M and carries all three physical claims the row asked for, but is *not* a spectral statement about a self-adjoint operator. That packaging is **#93**, priced there. Two further findings: Mathlib’s `round_le` again supplied the minimality (as in #91, no `Int.fract` bookkeeping); and the twin is an **analogue, not a limit** — the flux quantum is `1` here against `2π` on the ring, the levels are `(n − Φ)²` against `2 cos((2πm + Φ)/N)`, and the *ground* level reads the flux where the ring’s *top* level does, so nothing derives one from the other and the module says so. | ~~**L**~~ done (S–M; the L’s content is #93) | Low–medium — the textbook version of a twin the corpus already has | #55; [`berry-phase-scoping.md`](berry-phase-scoping.md) BP-4 | > | ~~91~~ | QIT | Library owed / upstream | ~~*new (2026-09-28)* **The flux is determined by the spectrum, for every `N`**~~ **DONE 2026-09-30** ([`AharonovBohmRing.lean`](../CsdLean4/Mathlib/QuantumInfo/AharonovBohmRing.lean) + the Cat-3 reading in [`AharonovBohm.lean`](../CsdLean4/Empirical/QM/AharonovBohm.lean), 9 pins; M took S–M). ★★★ `exists_eq_of_range_ringEigval_eq`: **two fluxes with the same level set differ by a whole quantum or are reflections of one another**, at every ring size, with ★★★ `flux_determined_by_spectrum` the experiment-facing form. **THE ROUTE, AS THE ROW PREDICTED, AND WHAT MADE IT CHEAPER:** the top of the spectrum is `2 cos(fluxDist Φ / N)` (★★★ `isGreatest_range_ringEigval`) with `fluxDist Φ` the distance to the nearest whole quantum, and Mathlib's **`round_le`** — `|x − round x| ≤ |x − k|` for every integer `k` — gives the minimality for free, so no `Int.fract` bookkeeping was needed. The bound direction that does the work is that a label's argument is at least `fluxDist Φ / N` from a whole turn (★★ `ringEigval_le`), because `N·|x − 2πr| = |Φ − 2π(Nr − m)|` and `Nr − m` is just some integer; the maximum is attained at the label `−round(Φ/2π)` (★★ `ringEigval_neg_round`), and `Real.injOn_cos` on `[0, π]` inverts it. ★★ `range_ringEigval_neg` proves the sign really is invisible, so the statement's ambiguity is exactly the truth and not proof slack. | ~~**M**~~ done | Low — the concrete witness already carries the physics | #55 | > | ~~92~~ | QIT | One QM, many formulations | **SPLIT 2026-10-07 — (a1) LANDED, and it is the part #92's own measurement said was needed.** (b) and (c) were done 2026-09-30; (a) was priced **XL** on the measurement that `Op(a)` as a continuous operator on `𝓢(ℝ, ℂ)` is *not* a gap in the proofs but **a statement needing a different symbol class**. That diagnosis was right, and it is now acted on. **(a1) DONE** ([`Mathlib/Analysis/Fourier/WeylSymbolClass.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WeylSymbolClass.lean), Category 1, 8 pins; M as taken). `WignerWeyl.lean` had been *assuming* what the weak datum could not give — `integrable_weylPair` takes joint continuity (`hbc`) and a bound integrable in the first slot (`hM`, `hbM`) as **hypotheses** — so this file fixes the class and discharges them as theorems. ★ `exists_bound_fst`/★ `exists_bound_snd`: a Schwartz function on the plane decays like `(1 + t²)⁻¹` **in each slot separately**, uniformly in the other, from two orders of `SchwartzMap.decay` (`k = 2` and `k = 0`) with `‖p‖² ≥ pᵢ²`. `weylOpK` is the Weyl operator of a **jointly Schwartz kernel**, with ★ `weylOpK_eq_weylOp` the bridge to the old operator through any slice family. ★ `continuous_weylKernel` and ★★ `exists_integrable_bound` are the three assumed hypotheses, now theorems. And ★ `integrable_weylOpK_integrand` with ★★ `continuous_weylOpK` are **the recorded defect fixed** — on this class the Weyl operator's output *is* continuous, by `continuousAt_of_dominated` against `4C(1 + (y − x₀)²)⁻¹`, the majorant the second-slot decay supplies uniformly for parameters within `1` of the point (★ `norm_weylKernel_le`, whose arithmetic is `(d − a)² ≤ 2d² + 2a²` at `d = x − x₀`, `a = x − y`). ⚠️ **This is the symbol class, not the calculus.** `Op(a) : 𝓢(ℝ, ℂ) →L 𝓢(ℝ, ℂ)` is **not** proved: continuity of the output is the *first* Schwartz seminorm estimate, not the last. The two things still missing are exactly the two the row measured, and each is now a numbered row, with the assembly itself #122 — ~~**#120**~~ (differentiation under the integral sign to all orders with bounds, over `ℝ` — **DONE 2026-10-07**) and **#121** (the two-variable Schwartz API, i.e. slicing). ⚠️ The bridge takes the slice family as a *hypothesis*, which is why (a1)'s properties are proved for `weylOpK` directly rather than inherited. ~~**(a) REMAINS, AND THE MEASUREMENT IS THE POINT.**~~ | ~~**L**~~ (b)(c)(a1) done (M each); remainder = #120 + #121 | Medium — (b)(c) were the physics, (a1) is the class, #120 is the calculus | #63; #60; #120; #121; [`WignerWeyl.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WignerWeyl.lean) | > | ~~119~~ | Docs | Reviewer-proofing | **DONE 2026-10-07 — and the sentence the row said to re-check rather than assume survived, for a sharper reason** ([`docs/FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md`](../docs/FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md); S–M took S). A ledger-to-prose sync, no new claim: every statement added is traceable to a struck row and all ten theorem names cited were checked against the tree. **(1) Link 10.** Its table seam and its `What is not claimed` both still listed the circuit-level threshold (row 62) as open; row 62 completed **2026-10-04** through rows 94–97, so the page now records syndrome extraction *as a circuit*, the ancilla ladder's fault propagation with the cat state's justification, the extended rectangle with a *faulty* recovery, and level reduction with the probabilistic join. The error-correction seam's opening — "Three residues, all open mathematics" — was stale in its framing and now says all three are closed. **(2) Link 11.** Its seam was `—` and its body never mentioned the synthesis cost at all; a `The synthesis cost` paragraph now carries row 74's ★★★ `exists_word_approx_polylog` with `c = log 5 / log(3/2) ∈ (3, 4)`, and its non-claims carry the three the landing recorded — `K` and the accuracy existential by inheritance, the exponent a **cost model** rather than a fact about `{H, T}`, and the literature's `3 + δ` unclaimed. **(3) THE CLOSING SENTENCE, WHICH IS THE POINT.** The row flagged "what the chain does not yet contain is a fault-tolerant machine" for re-checking rather than assuming, and re-checking changed its status rather than its truth: row 97's own honest-scope note is that the threshold **statement** is a theorem in abstract form with its gadget-set premise *named and not discharged for any concrete code*. So the sentence stands, and now says *which* premise is missing instead of which theorem. The page also records that the registry is down to three open residues (`R-012`, `R-016`, `R-019`) and that every residue it was written to track at the quantum-computer end — `R-001`–`R-006`, `R-013` — is discharged, with `R-019` marked as belonging to the prediction programme and not to this chain. ⚠️ **A correction to this row as opened:** it called the page `CONVENTIONS.md` §10 territory. It is not — §10's landing surface is the README and [`docs/TOUR.md`](../docs/TOUR.md) only, and §10 itself says routine landings update `docs/`. No §10 justification was needed and the README was not touched. | ~~**S–M**~~ done (took **S**) | Medium–high — it was the reader's path, and it had been selling the corpus short | #35; #74; #62; #116; rows 94–97; `residues.tsv` | > | ~~120~~ | QIT | Library owed / upstream | **DONE 2026-10-07 — and the interface is *shown* dischargeable, not asserted to be** ([`Mathlib/Analysis/Calculus/ContDiffParametricIntegral.lean`](../CsdLean4/Mathlib/Analysis/Calculus/ContDiffParametricIntegral.lean), Category 1, 9 pins; **M–L took M**). `partialDeriv F k` is the `k`-th derivative of `F` in its **parameter**, by recursion rather than through `iteratedDeriv`, so the induction shifts the order with no API friction — ★ `partialDeriv_eq_iteratedDeriv` is the bridge for discharging bounds in either vocabulary and ★ `partialDeriv_succ_left` is the shift itself; `contDiff_partialDeriv_one` passes parameter-smoothness to the parameter derivative through `contDiff_succ_iff_deriv`. ★★★ `contDiffOn_integral_of_bound`: **if each `x ↦ F x a` is `C^n`, each `partialDeriv F k x` is measurable in `a`, and the `k`-th is bounded on an open `U` by an integrable `bound k` uniformly in the parameter, then `x ↦ ∫ a, F x a` is `C^n` on `U`** — with ★★ `contDiff_integral_of_bound` the global case and ★★ `contDiffOn_integral_of_bound_all` smoothness at every order. **The generality came out better than the row priced**: the integration variable ranges over an *arbitrary* measure space, because per-`a` smoothness in the parameter replaces the joint-smoothness hypothesis #60 needed, so no vector-space structure on the integration variable is used at all. **And the row's own warning was honoured.** It said a version with hypotheses nobody can discharge would be worse than none, so one is discharged in the file: ★★★ `contDiff_integral_schwartz_sub_mul` — **convolution with a Schwartz function is smooth to every order**, for any integrable weight — meeting all four hypotheses, via `partialDeriv_sub_mul` (the parameter derivatives are `iteratedDeriv k f (x − t)·g t`, by `deriv_comp_sub_const`) and ★ `exists_bound_iteratedDeriv` (the uniform bound from `decay 0 k`, which is exactly the "uniformly in the parameter" the bound family asks for). ⚠️ **No formula for the derivatives**: each step produces `deriv (∫ a, F · a) = ∫ a, partialDeriv F 1 · a` on `U` as a by-product, but only the smoothness is recorded — stating the identity at every order would want `iteratedDeriv` *of* the integral, which nothing needs. ⚠️ **The parameter is one-dimensional**, which is what lets the proof use `deriv` throughout and keeps the bounds scalar; the `fderiv` version over a finite-dimensional parameter is the same induction with `ContinuousLinearMap` plumbing and is not done. ⚠️ **#92(a3) is the consumer this was built for and stays open as #122** — the `U`-local bound is the shape #92 needs, but assembling `Op(a) : 𝓢 →L 𝓢` also wants the Schwartz seminorm estimates on the output, which is more than smoothness. | ~~**M–L**~~ done (took **M**) | **Medium–high — a real Mathlib gap, now closed** | #92; #60; #88; #121 | > | ~~121~~ | QIT | Library owed / upstream | *new (2026-10-07, out of #92’s split)* **The two-variable Schwartz API. ▸ (i) SLICING DONE 2026-10-07, (ii) the partial Fourier transform remains.** ▸ **ROW CLOSED 2026-10-21**: (ii)'s joint form is ★★★ `SchwartzMap.partialFourierCLM : 𝓢(ℝ × ℝ, ℂ) →L[ℂ] 𝓢(ℝ × ℝ, ℂ)` ([`PartialFourierCLM.lean`](../CsdLean4/Mathlib/Analysis/Fourier/PartialFourierCLM.lean), ~~#170~~–~~#174~~), with `wignerSchwartz ψ` closing ~~#92~~(c)'s caveat. `𝓢(ℝ × ℝ, ℂ)` exists as a type at the pin and its seminorms work, but nothing connected it to one-variable Schwartz space. **(i) DONE** ([`Mathlib/Analysis/Fourier/SchwartzSlice.lean`](../CsdLean4/Mathlib/Analysis/Fourier/SchwartzSlice.lean), Category 1, 5 pins; M–L took S–M). Mathlib had the machinery and not the instance: `SchwartzMap.compCLM` composes on the right with any map of temperate growth that does not shrink the norm, but the slice map `w ↦ (e, w)` is **affine** rather than linear and `fun_prop` has no `Prod.mk` rule for `HasTemperateGrowth`. ★ `Function.hasTemperateGrowth_prodMk_left` supplies it through `HasTemperateGrowth.of_fderiv` — the derivative is the *constant* `inr` and the value grows linearly because the product norm is a max — whence `SchwartzMap.sliceCLM`, the slice **as a continuous linear map in the kernel**, and ★ `SchwartzMap.slice_apply`. **THE PAYOFF IS THE WHOLE POINT OF (i), AND IT IS BIGGER THAN "transfers the theorems":** ★★★ `integral_conj_mul_weylOpK` — **the Weyl expectation formula on the Schwartz kernel class with no hypotheses at all**, that the expectation of the Weyl operator in a state is the phase-space average of the symbol against the state’s Wigner function. `WignerWeyl.lean` proves it for a slice family under three assumptions its datum could not supply; #92(a1) discharged two of them and the slice supplies the family, so on this class nothing is left to assume. ⚠️ **Continuity in the slice parameter is NOT part of what landed** and is now **#123**: what is continuous here is slicing *in the kernel*, which is what transfers theorems, while `e ↦ slice K e` continuous for the *Schwartz topology* is a different statement needing one more order of decay and a mean-value estimate, with no Mathlib curry to get it from. ▸ **(ii) PARTIALLY DONE 2026-10-07, AND IT TURNED UP A DEPENDENCY THE ROW DID NOT KNOW ABOUT.** ([`Mathlib/Analysis/Fourier/PartialFourier.lean`](../CsdLean4/Mathlib/Analysis/Fourier/PartialFourier.lean), Category 1, 4 pins). What landed is the **slice-wise** transform, which is the object `weylSymbol` actually is: `SchwartzMap.sliceFourierCLM` composes (i)’s `sliceCLM` with Mathlib’s `fourierTransformCLM`, so **the symbol at a fixed midpoint is a Schwartz function of the frequency, continuously and linearly in the kernel**, with ★ `weylSymbol_slice_apply` identifying it with the corpus’s `weylSymbol` and ★ `integrable_weylSymbol_slice` / ★ `exists_bound_weylSymbol_slice` the two consequences a consumer wants. ⚠️ **THE JOINT FORM IS STILL OPEN, AND IS NOT A COROLLARY.** That `(u, ξ) ↦ weylSymbol (slice K) u ξ` is Schwartz **on the plane** needs decay and smoothness *jointly*; slice-wise Schwartzness gives neither, because every constant may depend on the midpoint. ⚠️ **CORRECTED 2026-10-07: THE CURRY ROUTE IS NOT MISSING, IT IS UNSTATABLE.** Yesterday this row said the natural proof factors through the curry `𝓢(E × F, G) ≃ 𝓢(E, 𝓢(F, G))` and that Mathlib merely lacks it. It cannot have it: `SchwartzMap` requires a **normed** target and Schwartz space is Fréchet, not normed — `NormedAddCommGroup 𝓢(ℝ, ℂ)` does not synthesize, so `𝓢(E, 𝓢(F, G))` does not typecheck. **There is only one route, the bespoke one**, and #123 is not "half of a curry". The alternative is bespoke: differentiate under the integral in *both* variables to all orders, which needs the **finite-dimensional-parameter version of #120** (that row’s scope note records its one-dimensional restriction, so this is a second new row’s worth of work — ~~**#124**~~ (**DONE 2026-10-07**)), and then the seminorm estimates `(2πiξ)^c·𝓕₂K = 𝓕₂[∂_w^c K]` and `∂_ξ^b·𝓕₂K = 𝓕₂[(−2πiw)^b K]` assembled against `K`’s decay. **So the row’s `L` for (ii) was wrong**: the joint form is gated on either #123 or #124 and is **L–XL** behind them. It stays open here. ▸ **2026-10-21: (ii)'s joint form is SPLIT into #170–#174** — #170 `𝓕₂K` smooth with uniformly bounded derivatives (~~#124~~ with the Leibniz bound family, M–L); #171 the derivative identities `∂ 𝓕₂ = 𝓕₂ ∘ Dop` (M); #172 the multiplication identities `ξ 𝓕₂ = (2πi)⁻¹ 𝓕₂ ∂_w`, `u 𝓕₂ = 𝓕₂ u` (S–M); #173 the joint seminorm bounds in `mkCLM` shape (M–L); #174 the CLM, `= weylSymbol`, and `wigner ψ` Schwartz on the plane (M). The XL was the price, not a wall, and this row now holds only (i) and the slice-wise half, both done; **it closes when #174 lands** — ~~#174~~ **landed 2026-10-21, row closed.** | ~~(i) **M–L**~~ done (took **S–M**); (ii) slice-wise done, **joint form L–XL and no longer gated: ~~#124~~ landed 2026-10-07 **with the `iteratedFDeriv` bridge it needed**, so what remains is this row’s own mathematics — a Leibniz expansion of the directional derivatives of `e(−wξ)·K(u, w)` across the exponential, and the two Fourier seminorm identities `(2πiξ)^c·𝓕₂K = 𝓕₂[∂_w^c K]`, `∂_ξ^b·𝓕₂K = 𝓕₂[(−2πiw)^b K]`, assembled against `K`’s decay in each slot. **Re-priced XL**, and the pricing has now moved twice, so it is recorded as a measurement rather than an estimate** | Medium — (i) paid, and (ii) has at least stopped being mis-priced | #92; #63; #122; #123; #124 | > | ~~124~~ | QIT | Library owed / upstream | **DONE 2026-10-07 — and it refuted its own forecast about the hypotheses** ([`Mathlib/Analysis/Calculus/ContDiffParametricIntegralFDeriv.lean`](../CsdLean4/Mathlib/Analysis/Calculus/ContDiffParametricIntegralFDeriv.lean), Category 1, 6 pins; M–L took M–L). ★★★ `contDiffOn_integral_of_dirBound`: for a parameter in any **finite-dimensional** normed space, if each `x ↦ F x a` is `Cⁿ`, each directional derivative of order `≤ n` is measurable in `a`, and the one along `hs` is bounded on an open `U` by `(∏‖hᵢ‖)·bound hs.length a` with each `bound k` integrable and nonnegative, then `x ↦ ∫ a, F x a` is `Cⁿ` on `U` — with ★★ `contDiff_integral_of_dirBound` the global case and ★★ `contDiffOn_integral_of_dirBound_all` smoothness at every order. **THE ROW’S OWN FORECAST WAS WRONG, IN THE GOOD DIRECTION.** It warned that #120’s hypothesis-design question "returns in harder form: a per-order, locally-uniform bound on a *multilinear* norm". It does not have to. Bounding `iteratedFDeriv` means bounding a multilinear map, which drags currying equivalences through every step; bounding **iterated directional** derivatives keeps every hypothesis valued in `E`, and the recursion becomes literally list-append — `dirDeriv F hs` with ★ `dirDeriv_append_singleton` the shift, and `dirWeight` carrying the product of the directions’ norms so that appending one direction multiplies the weight by its norm and raises the order by one (★ `dirWeight_append_singleton`), leaving the inner call with `fun k a => ‖h‖·bound (k+1) a`, still integrable. **No multilinear norm appears in the statement at all.** ⚠️ **Two hypotheses the row did not anticipate, both in the header.** (1) One is **unavoidably operator-valued**: Mathlib’s `hasFDerivAt_integral_of_dominated_of_fderiv_le` takes the derivative as a map into `H →L[ℝ] E`, so strong measurability of `a ↦ fderiv ℝ (F · a) x` is required and cannot be reduced to directional data — it is `hmeasD`, closed under the recursion for the same list-append reason. (2) The bound family must be **nonnegative** (`hb0`), which #120 did not need: its one-dimensional engine bounds `‖F′ x a‖` directly, whereas here the directional bounds must become an *operator*-norm bound through `opNorm_le_bound`, which is false for a negative constant when `H` is trivial. ⚠️ `FiniteDimensional ℝ H` is used exactly once, in `contDiffOn_clm_apply`, to assemble the directional components back into the operator-valued derivative — the same place #60 needed it. ⚠️ **No formula for the derivatives**, as in #120. ⚠️ **#121(ii) is the consumer this was built for and stays open**: it now has one of its two routes in hand, and still needs the Fourier seminorm estimates on top. ▸ **ADDENDUM 2026-10-07 — #124 WAS UNDER-DELIVERED AS LANDED, and trying to consume it is what showed that.** `SchwartzMap.decay` gives bounds on `iteratedFDeriv`, so the directional hypotheses were not dischargeable from Schwartz data and the file was unusable by the very row it was built for. ★★ `norm_dirDeriv_le` is the bridge — an iterated directional derivative is bounded by the iterated total one times the product of the directions’ norms, by an induction that peels the **innermost** direction (peeling the head does not work: the head is the *outermost* derivative, so the result is not of the form `dirDeriv G hs` for any `G`) — and ★★ `contDiffOn_integral_of_iteratedFDeriv_bound` is the multilinear form a `decay`-armed user actually has. This also makes the module’s own claim honest: the directional interface is easier to meet *and* strictly more general, since the multilinear one follows from it. ▸ **ADDENDUM 2026-10-08 (#126's gate).** The scope note "no formula for the derivatives" is amended: ★★★ `hasFDerivAt_integral_of_bound` exports the first-order identity this induction proves and discards (weakest hypotheses: differentiability of each slice), ★★ `fderiv_integral_apply_of_bound` is its directional form, and ★★★ `norm_iteratedFDeriv_integral_le` is the **bound** it implies at every order — the identity at order `n` would need the iterated derivative of the integral as a multilinear map, which is what this file exists to avoid. The induction now **calls** the exported identity instead of rebuilding it. | ~~**M–L**~~ done (+ two addenda) | Medium–high — a real Mathlib gap, and one of the two gates on #121(ii) | #120; #121; #60; #123 | > | ~~123~~ | QIT | Library owed / upstream | *new (2026-10-07, out of #121(i))* **Slicing is continuous in the slice parameter for the Schwartz topology**, i.e. `e ↦ SchwartzMap.slice K e` is continuous into `𝓢(F, G)`. #121(i) proved slicing continuous **in the kernel** (`sliceCLM` is a continuous linear map for each fixed `e`), which is what transfers theorems; this is the other direction and it is a genuinely different statement. The route, since Mathlib has no curry `𝓢(E × F, G) ≃ 𝓢(E, 𝓢(F, G))` to get it from: for each seminorm `k, n`, bound `‖w‖ᵏ·‖∂ⁿ_w K(e, w) − ∂ⁿ_w K(e₀, w)‖` by `|e − e₀|` times a sup of `‖w‖ᵏ·‖∂_e ∂ⁿ_w K‖`, which is `K`’s own decay **at one order higher** — so the slice family is locally Lipschitz in `e`, not merely continuous. Two pieces are needed first: relating `iteratedFDeriv n` of the slice to `iteratedFDeriv n` of `K` (a precomposition-by-`inr` identity the pin may not have in usable form), and a mean-value inequality for the vector-valued `e ↦ ∂ⁿ_w K(e, ·)`. ⚠️ **Nothing in the corpus needs this** — it is recorded because #121(i) names it, and because a symbol-valued calculus would want it. ⚠️ **CORRECTED 2026-10-07: this row is NOT "the first half of a curry".** The curry `𝓢(E × F, G) ≃ 𝓢(E, 𝓢(F, G))` is not statable with Mathlib’s `SchwartzMap`, which requires a normed target — Schwartz space is Fréchet. So this row stands on its own as a continuity statement about one map, with no larger object behind it, and **nothing is gated on it**: #121(ii) now goes by #124 alone. ▸ **DONE 2026-10-09** ([`Mathlib/Analysis/Fourier/SchwartzSliceContinuous.lean`](../CsdLean4/Mathlib/Analysis/Fourier/SchwartzSliceContinuous.lean), Category 1, 6 pins; M–L as priced). ★★★ `continuous_slice`, and **the row's own route worked with both pieces it named**. (1) ★★ `iteratedFDeriv_comp_inr` — the derivatives of a slice are the kernel's precomposed with `w ↦ (e, w)`; this is the **mirror of ~~#126(ii)~~'s `iteratedFDeriv_comp_inl`**, which is why it was cheap. (2) ★★ `norm_fderiv_iteratedFDeriv_comp_inl_le` — the kernel's `n`-th derivative is differentiable in the parameter with derivative bounded by its `(n+1)`-st, i.e. `K`'s decay **at one order higher**, exactly as predicted. ★★★ `lipschitz_seminorm_slice` is the estimate: each Schwartz seminorm of `slice K e − slice K e₀` is at most `‖e − e₀‖` times **one** seminorm of `K`. **THE ROW SAID “locally Lipschitz”; IT IS GLOBALLY LIPSCHITZ** — the derivative bound is a seminorm of `K`, which does not depend on the base point, so nothing has to be localised. ⚠️ *A note on the proof*: the weight is handled by a **case split** on whether it vanishes rather than by weighting the map before the mean value step. Weighting first is the natural move (the derivative bound is then uniform in both variables) but it puts a scalar `smul` on a space of continuous multilinear maps, where the `NormSMulClass` instance the norm computation needs is **absent at this pin**; the case split costs two lines and needs no instances. ⚠️ Still **not a curry** and still **nothing is gated on it** — it landed because it was *unblocked*, not because anything waited. | **M–L** | Low — nothing is gated on it | #121; #92 | > | ~~128~~ | QIT | Library owed / upstream | *new (2026-10-09, out of ~~#64(i)~~)* **The free Hamiltonian in position space *is* `−½ d²/dx²`.** ▸ **DONE 2026-10-15 (M), on the Schwartz core.** Cat-1 [`Mathlib/Analysis/Semigroup/FreeHamiltonianSchwartz.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/FreeHamiltonianSchwartz.lean), 13 pins. For every Schwartz `φ`: `toLp_mem_freePositionOp_domain` (**Schwartz classes lie in the position-space domain**), ★★ `freePositionOp_toLp` = `freePositionOp (φ.toLp 2) = (kineticOp φ).toLp 2` (the Schwartz kinetic operator of `SchrodingerSchwartz.lean`, already `−½Δ` there by `kineticOp_eq_laplacian`), hence ★★★ `freePositionOp_toLp_eq_laplacian` = `((−½) • Δφ).toLp 2` and ★★★ `freePositionOp_toLp_ae` / `_ae_deriv` = **pointwise a.e. `(H₀φ)(x) = −½ φ''(x)`** (`iteratedDeriv 2`, `deriv (deriv φ)`, via Mathlib's `laplacian_eq_iteratedDeriv_real`). The route: `fourier_kineticOp_apply` = `𝓕(H₀φ)(ξ) = 2π²ξ²·𝓕φ(ξ)` (Fourier inversion on Schwartz space), ★ `freeSymbolOp_eq_of_coe_eq` / `freeSymbolOp_toLp_fourier` = the momentum multiplier on `(𝓕φ).toLp` is the class of `𝓕(H₀φ)`, and the `L²`/Schwartz Fourier bridges `SchwartzMap.toLp_fourier_eq` / `toLp_fourierInv_eq` carry it across `freePositionOp`'s conjugation (`fourierL2_symm_toLp_fourier`). `freeSymbol_eq_neg_half_mul_sq` records the convention this row flagged: `2π²ξ² = −½(2πiξ)²` with Mathlib's `e^{−2πixξ}`. **Not claimed:** (a) on the *whole* domain — Schwartz functions are not shown to be a core for `freePositionOp` (closure from the Schwartz restriction = the operator); (b) the identification of the domain with `H²(ℝ)` — Mathlib has no `H^s` scale on `Lp` at this pin. What stands is that the self-adjoint operator of ~~#64(i)~~ **restricts to `−½ d²/dx²` on every Schwartz function**; the two residues are the natural follow-ups if a consumer ever appears (none does). ⚠️ Lean finding: an anonymous constructor into `freeSymbolOp.domain` inside a *proof* hit a `toLp`-unfolding defeq timeout that the same constructor in a *statement* did not; stating the lemma for a variable element with `(f : Lp) = (𝓕φ).toLp 2` as hypothesis and instantiating with `rfl` avoids it. ▸ *Original row:* ~~#64(i)~~ built `freePositionOp` as the Fourier conjugate of the momentum-space multiplication operator and proved it self-adjoint with spectrum `[0, ∞)` — but it is the *conjugate*, not yet identified with a differential operator, and the row that asked for it said "`−½Δ`". Closing that gap needs two things the pin does not package: (a) **the Fourier–derivative identity at the `L²` level** — `𝓕(f'') = −4π²ξ²·𝓕f` for `f` in the domain, which Mathlib has for Schwartz functions (`SchwartzMap.fderivCLM_fourier_eq`) and not for `Lp`; the route is density of `𝓢(ℝ, ℂ)` in the domain (`denseRange_toLpCLM`, which `Lp.fourierTransformₗᵢ` is itself built from) plus closedness of the two operators; and (b) **the domain identification** — that `𝓕⁻¹{f : (1 + ξ²)f ∈ L²}` is the Sobolev space `H²(ℝ)`, which is a *definition* Mathlib has no form of at this pin (`Mathlib/Analysis/Distribution/Sobolev.lean` exists but is about tempered distributions, not an `H^s` scale on `Lp`). ⚠️ **Nothing waits on this.** ~~#64's~~ remaining two pieces (bounded symmetric perturbation, Stone's theorem) take the *operator*, and self-adjointness plus the spectrum are exactly what they need — the differential form is for a reader, not for a proof. ⚠️ **The constant is the one place a convention bites**: with `𝓕` carrying `e^{−2πix·ξ}` the symbol `2π²ξ²` *is* `−½d²/dx²`, and with the physicists' `e^{−ix·ξ}` it is not; the identification has to name its transform. | ~~**M–L**~~ done (M) | Low — explanatory; it would make a name honest that is currently only suggestive | #64; #128; [`FreeHamiltonian.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/FreeHamiltonian.lean) | > | ~~129~~ | QIT | One QM, many formulations | *new (2026-10-09, out of ~~#64(iii)~~)* **The interacting propagator's generator: the mild solution is a classical one.** ▸ **DONE 2026-10-20 (M–L), in generator form; the one remaining identification is NEW ROW #154.** Cat-1 [`Mathlib/Analysis/Semigroup/PerturbedGenerator.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/PerturbedGenerator.lean): `HasGeneratorAt T ψ v` = the orbit has right derivative `v` at `0` (`ψ ∈ D(gen)`, `gen ψ = v`; `hasGeneratorAt_iff_hasDerivWithinAt`); ★ `tendsto_inv_smul_duhamelIntegral` = the Duhamel integral is `t·Bχ + o(t)` (joint continuity at `(0,0)`, the one analytic input); ★★ `hasGeneratorAt_perturbed_iff` = **`gen 𝒮 = A + B` on `D(A)`, domains equal**; for any semigroup-like family `HasGeneratorAt.apply` (**the orbit stays in the generator's domain**, `gen (T t ψ) = T t (gen ψ)`) and ★★ `HasGeneratorAt.hasDerivAt` (**two-sided derivative at every `t > 0`** — the left quotient is `T(t−k)` applied to the right quotient at `0`, controlled by the bound on `[0,t]` and strong continuity); ★★★ `hasDerivAt_perturbed` = `d/dt 𝒮(t)ψ = 𝒮(t)(v + Bψ)`, `hasDerivWithinAt_perturbed_zero`, ★★ `hasGeneratorAt_of_perturbed` = **`𝒮(t)ψ ∈ D(A)` with `A𝒮(t)ψ = 𝒮(t)(A+B)ψ − B𝒮(t)ψ`**, ★★★ `perturbed_classical` = `d/dt 𝒮(t)ψ = (A+B)(𝒮(t)ψ)`. Cat-1 [`SchrodingerGenerator.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/SchrodingerGenerator.lean): ★ `hasGeneratorAt_freeGroup` (a vector of `dom H₀` is in the free generator's domain with value `−iH₀ψ`, from `hasDerivAt_fourierGroup_freePositionOp` at `0`), ★★★ `hasDerivAt_schrodinger` = **`d/dt U(t)ψ = −i U(t)(H₀ψ + Vψ)` at every `t > 0`** for `ψ = 𝓕⁻¹f`, `f` in the momentum domain (`_addCLM` form through `freePositionOp.addCLM (potential VL)`; `hasDerivWithinAt_schrodinger_zero` at `0`), ★★ `hasGeneratorAt_freeGroup_schrodinger` = **the orbit `U(t)ψ` stays in the free generator's domain**, ★★★ `schrodinger_classical` = `i∂ₜU(t)ψ = (H₀+V)(U(t)ψ)` with `H₀` acting on the orbit in generator form. 19 pins. **What was NOT proved here (→ ~~#154~~, DONE 2026-10-20 the same day):** that the free group's *generator* domain coincides with the concrete `freePositionOp.domain` — the converse of `phaseGroup_mem_mulDomain`, a Fatou argument; until then '`U(t)ψ ∈ dom H₀`' is the semigroup-theoretic statement, not the multiplication-domain one. Also not made: the `coeFn` identification of `potential VL` with `schrodingerOp`'s `mulCLM` for real `V`. ▸ *Original row:* ~~#64~~ is closed, and what it deliberately left is this: `SchrodingerGroup.schrodinger` is a unitary propagator for a bounded potential (Dyson series, Duhamel equation *with uniqueness*, the semigroup law and Trotter, all in [`BoundedPerturbation.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/BoundedPerturbation.lean)), and ~~#64(ii)~~ made `H₀ + V` self-adjoint on `dom H₀` — but nothing yet says `d/dt ᵎ(t)ψ = −i(H₀+V)ᵎ(t)ψ` for `ψ` in that domain. The route: the Duhamel integral `∫₀ᵗ S(t−s)B(ᵎ(s)ψ)ds` is differentiable at `t` once `s ↦ S(t−s)B(ᵎ(s)ψ)` is continuous (it is) **and** the orbit stays in the domain — ~~#64(iii)~~'s `phaseGroup_mem_mulDomain` gives domain invariance for the *free* group, and the perturbed one needs the same for `ᵎ`, which is where the work is. ⚠️ **This is the standard mild-versus-classical gap**, and it is genuinely a theorem rather than bookkeeping: a `C₀`-semigroup's mild solutions are classical only on the generator's domain, and proving the orbit stays there for a Dyson-series solution is the content. ⚠️ **Nothing in the corpus needs it**: the classical limit is compared at the generator (~~#64~~'s brackets) and the free evolution is already an equation. | ~~**M–L**~~ done (M–L) | Low–medium — it would make the interacting dynamics an equation rather than a propagator | #64; #36; [`BoundedPerturbation.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/BoundedPerturbation.lean) | > | ~~130~~ | Foundations | New predictions · Statistical-mechanics layer | *new (2026-10-10, out of ~~#118(ii)~~)* **The finite-horizon bound in relative-entropy form.** ~~#118(ii)~~ proved the `L²` statement — the mean-square deviation of a cell's time-averaged occupation from its Haar value at horizon `T`. `R-019`'s second clause is phrased in **relative entropy**, and the conversion is standard: the Haar cell law is uniform on these cells, so `klDiv ≤ χ² = (m+1)²·Σ_c (p_c − q_c)²` and ~~#118(ii)~~'s summed bound applies, at the cost of one more factor `(m+1)²`. ~~**What it needs is packaging, not mathematics**~~ ▸ **DONE 2026-10-12, and the row was right about the work but wrong about one thing** ([`SigmaLayer/FibreHorizonDivergence.lean`](../CsdLean4/SigmaLayer/FibreHorizonDivergence.lean), 7 pins, and a new Category-1 [`Mathlib/InformationTheory/ChiSquare.lean`](../CsdLean4/Mathlib/InformationTheory/ChiSquare.lean), 2 pins; S–M took S–M). **Both halves the row named:** ★★★ `InformationTheory.klDiv_uniformOn_univ_le_chiSq` — `klDiv μ uniform ≤ n·∑ (μ{x} − 1/n)²` on a finite type, from `log t ≤ t − 1` as predicted, with the two sides turning out to be the *same* sum (so the bound is tight exactly at the uniform law) — and the packaging: `horizonLaw m T p` is the occupation law as the uniform average of the `T` Dirac masses at the cells visited, with ★★ `horizonLaw_real_apply` identifying its mass at a cell with ~~#118(ii)~~'s Birkhoff average, ★ `isProbabilityMeasure_horizonLaw` and ★ `measurable_horizonLaw` (the measurability the row asked for). ★★★ `klDiv_horizonLaw_le_chiSq` is the conversion: the divergence of the occupation law from the uniform cell law is at most `(m+1)²` times the mean-square deviation of the occupations from `1/(m+1)²`, pointwise in the base point. ⚠️ **THE ONE THING THE ROW GOT WRONG, AND IT IS WORTH HAVING FOUND: “the Haar cell law is uniform on these cells” IS NOT A THEOREM OF THIS CORPUS.** The row used it to write the `(m+1)²` factor, and the chi-squared inequality is proved against `uniformOn univ`, but ~~#118(ii)~~'s deviation is measured from `coarseLaw (kMuL p₀) (fibreCell m)` — and no theorem identifies the two. It is true by the equal-width construction of `circCell`, and it is now **#144**. So the bound above is stated against the uniform law, where it needs no hypothesis, and the *combination* with ~~#118(ii)~~ waits on #144 rather than being asserted. ⚠️ The scope is unchanged, as the row said: still **time-averaged**, so still not `R-019`; ~~#118(ii)~~'s `HasCorrelationDecayUpTo` still travels; and the factor is now `(m+1)⁴` once the two are combined, so in this vocabulary “a rate set by the cell size” is *further* away than in the `L²` reading. ⚠️ The conclusion is a **lower integral** in `ℝ≥0∞` rather than a Bochner integral: `klDiv`'s measurability in its first argument is not available, and the `ℝ≥0∞` form is what the chain of estimates supports. ⚠️ **It changes nothing about the scope**: the conclusion is still *time-averaged*, so this is not `R-019` either, and the honest reason to want it is that the row's own clause is stated in `klDiv`. | **S–M** | Low — a restatement that makes ~~#118(ii)~~ answer `R-019`'s clause in its own vocabulary, and it found an unproved premise in the row's own statement of the bound | #118; #144; `R-019`; [`FibreFiniteHorizon.lean`](../CsdLean4/SigmaLayer/FibreFiniteHorizon.lean) | > | ~~131~~ | Foundations | Records→spacetime | *new (2026-10-10, at the author's instruction)* **One finite two-wing CSD experiment: record stability, causal separation, Bell correlations and no-signalling in the *same* dynamical model** — the author's own narrowing of `C-1` to something testable, far short of deriving spacetime. ▸ **THE COMPATIBILITY AUDIT IS DONE 2026-10-10** ([`two-wing-experiment-scoping.md`](two-wing-experiment-scoping.md); read-only, every claim read off a theorem or definition *type*, nothing built). **The arena is not the obstacle and Bell is not the obstacle.** `CV.RecordFibre` and `LF4.KTorus` are *the same type*; `FieldArena K N` and `CPN M` differ only by an index equivalence whose transport (`LinearIsometryEquiv.piLpCongrLeft`) **`LF6`'s own capstone already uses internally**; and for two qubits plus two pointers `FieldConfig 4 2` has cardinality 16, so one arena `KSigma 16` carries all four interfaces with `Fin K = Fin 4` *exactly* the mode index `CV`'s coupling graph wants. `LF6` and the record layer are **already connected** — `LF6`'s Born clause evaluates `epistemicMeasure` of `RecordLayer.globalBasin`. On Bell: `IsProductPartition`'s arity has no slot for the remote setting, so the collapse argument is exact, and `RemoteSettingLocalityA` escapes it because `wingA ⟨a,b⟩` keeps the full context — while `globalBasin c i = {x \| x.2.1 ∈ circleCell (c.rate x.1) i}` supplies a contextual outcome map of the admissible shape, **rates from the base and selector from the fibre**, in which a `reroute` would be *forced by* `P_st` rather than chosen to fit the marginals. ⚠️ ~~**TWO INCOMPATIBILITIES, AND THEY ARE THE ROW'S BLOCKERS.**~~ **BOTH BLOCKERS FELL, 2026-10-10 — (II) dissolved, (I) resolved by option I1 (below).** ~~**(I)** `globalBasin`'s selector is one-dimensional (`x.2.1` only; the second torus coordinate is unused), while `CV.recordStroke₂` writes *both* fibre coordinates as two channels — so the record layer models **a single joint outcome, not two independently-recorded wings**. A two-selector basin is needed, and a per-wing basin whose rates depend only on that wing's setting is *precisely* the product-partition form the no-go kills, so at least one wing's cell geometry must see the **joint** context. That is a physics choice, not a formalisation detail.~~ ~~**(II)** Two measures: `LF6`'s flow preserves `fsMeasure` on the base, the Born and record statements live in `epistemicMeasure p = δ_p ⊗ vol` which a base-moving flow does **not** preserve, and the record layer's own `sigmaShift` preserves it but moves only the fibre.~~ **(II) DISSOLVED 2026-10-10** by the transport law of obligation (1): the flow carries the prepared epistemic measure to the post-measurement one, which is where the Born clause lives, so there was never a measure to preserve. ~~**(I) stands, and is now this row's only blocker.**~~ ▸ **(I) RESOLVED 2026-10-10, at the author's decision, by option I1** ([`RecordLayer/TwoWingCoarsening.lean`](../CsdLean4/RecordLayer/TwoWingCoarsening.lean), 16 pins; S–M): **one joint ontic selector, four cells, the two wings read off by coarsening.** The tempting minimal choice — one selector per wing, each wing's geometry seeing only its own setting — fails *twice over*: it is the product-partition arity the no-go kills, **and** with independent selectors the joint law factorises, so it could not carry correlation at all. Two facts make I1 the cheapest option that is also foundationally clean: `circleCell` has **no free offset** (a cell's position is fixed by the rate vector alone), so “whose geometry sees the joint context” is really “which rate vector is the joint one” and I1 answers that there is one and it is joint; and the architecture [`sigma-fibre-contextuality.md`](sigma-fibre-contextuality.md) settles on is *probabilities Born, regions contextual*, which coarsening preserves exactly. The content is stated **once**, for an arbitrary decoding `u : Fin N → κ`, and the wings are instances: ★★★ `measure_coarseEvent` — the coarse law is the **sum of the fine Born weights** over the decoding's fibre, the shape `LF6`'s clause (3) already uses; ★★★ `measure_coarseEvent_comp` — **coarsening composes**, and ★★★ `measure_wingAEvent_eq_sum` / ★★★ `measure_wingBEvent_eq_sum` are its two instances, so **each wing's marginal is the sum over the other wing's outcomes** as a *consequence* of the fine cells partitioning `Σ`, not an assumption; ★★★ `measure_coarseEvent_eq_of_fineSum_eq` and its two wing forms — **operational no-signalling with its premise named**, A's marginal agreeing between two joint contexts exactly when their summed fine weights agree, so the construction supplies the identity while whether the premise holds stays a *checkable* property of the rate vector (for the singlet, of `P_st`) rather than something smuggled; ★★ `factorsThrough_coarseEvent` — every coarse event is **macroscopic**, a preimage of #102's record string, which is obligation (5)'s shape; ★ `wingCode_ne_of_mem_of_mem` — the **non-vacuity check**: the coarse code genuinely moves between joint contexts at one ontic point, so these outcome maps are *not* of product-partition arity (the analogue of `LF3.translation_wingA_setting_dependent`, without which the construction would be the pointwise primitive in disguise). ⚠️ **What I1 concedes — and it is a commitment about `Σ`, not a formalisation artefact — is that the two wings share ONE ontic selector.** The “two records” are two readings of `x.2.1`, so `CV.recordStroke₂`'s two channels are **not** the write mechanism for this model. That is honest for CSD (`Σ` is unified and the apparent nonlocality sits in the projection) and it is the easiest point for a critic to press; the alternatives **I2** (chained selectors) and **I3** (a non-product partition of `T²`) both repurpose `x.2.2` and are now **row 132**, which #131 does not need. ⚠️ **The measurable coordinate is still `outcomeCode`**: `fineIndex` and `wingCode` are `Option`-valued conveniences with **no** measurability claimed, which costs nothing because every probability statement here is about an *event* and `factorsThrough_coarseEvent` is what ties those events to the measurable record coordinate. ⚠️ **No `P_st` yet**: the decoding and the context are arguments, so instantiating them with `LF6`'s pointer-pair index and the joint Born rates is obligation (1)'s second half. ~~⚠️ **AND ONE PLACE THE CORPUS IS EMPTIER THAN ITS HEADLINE**: there is **no `RemoteSettingLocalityB` witness anywhere** — `translationLocality` supplies the A-wing only and its `fB` ignores `a`, so its B-wing is of product-partition arity and cannot reproduce the singlet; hence `operationalNoSignalling_of_settingLocality` has **no model at all** today.~~ ▸ **CLOSED 2026-10-11** ([`LF3/SettingLocality.lean`](../CsdLean4/LF3/SettingLocality.lean), 6 pins; S). The fix was to let **each** wing see the remote setting through its own offset: `symmTranslationMaps` reads A at `l - g C.b` and B at `l - h C.a`. The two primitives constrain *different* wings, so their relabellings are independent and both hold at once — ★★★ `symmTranslationLocalityB` is the declaration whose absence was the gap, and ★★★ `operationalNoSignalling_symmTranslation` is the payoff: **the derived predicate is inhabited, so the theorem is not vacuous.** ★ `symmTranslation_wingA_setting_dependent` and ★ `symmTranslation_wingB_setting_dependent` are what make it a *real* witness — **both** wings' readings move with the **remote** setting, so neither is of product-partition arity and the no-go does not reach them. ⚠️ **It is not shown to reproduce `P_st`**: it is a witness for the *primitives*, and its content is the non-product arity, not a Bell violation. ⚠️ **And it is not the record-layer instantiation** the note's §4 asked for — an abstract additive group, not `Σ` with the basins as readout; that is **row 137**. ▸ **The five obligations, priced in §6 of the note**: ~~(1) one physical record history **S–M** (the cheapest test of (II));~~ **(1) FIRST HALF DONE 2026-10-10, AND BLOCKER (II) IS DISSOLVED** ([`RecordLayer/FlowRecordHistory.lean`](../CsdLean4/RecordLayer/FlowRecordHistory.lean), 5 pins; S). ★★★ `map_epistemicMeasure_liftBase` is the **transport law**: the pushforward of the *prepared* epistemic measure along the lifted flow is the epistemic measure of the **image ray**. **“Preserves” was the wrong demand** — the Dirac moves with the ray and the record medium's Haar measure does not move at all, and since `LF6`'s Born clause is stated at the *post-measurement* ray that is exactly the composition the experiment needs. ★★★ `map_outcomeCode_liftBase` and ★★★ `measure_outcomeCode_liftBase` are the record history: an outcome statement proved at the image ray **is** a statement about the run that starts at the prepared one, which makes “the flow produces this record” a theorem rather than a juxtaposition of a flow clause and a volume clause; ★★★ `measure_recordString_liftBase` is the multi-context form. It needed **no** choice about blocker **(I)**, which is why it was the right first brick. ~~⚠️ **The second half remains**: tying the outcome events to `LF6`'s pointer blocks through clause (3) and the index transport of the note's §3 — this file does not import `LF6` — **S–M**;~~ ▸ **OBLIGATION (1) IS COMPLETE 2026-10-10** ([`LF6/TwoWingRecord.lean`](../CsdLean4/LF6/TwoWingRecord.lean), 14 pins; S–M, took S–M). **What was missing was not a number.** Clause (3) already said the de-isolation reproduces the singlet and `NudgeLocality.localDeisolation_A_marginal_volume_eq_half` already said each wing's marginal is `1/2` at every setting pair — but *every* one of those statements is about a **sum over an index set** (`∑ n : Fin 4, … basin volume at e (n, stIdx (s,t))`), and a sum of numbers is not an event: there was no **set** whose measure is “the two wings recorded `(s,t)`”, so there was nothing to transport along the flow, nothing to intersect, and nothing to factor through a macroscopic coordinate. I1's coarsening supplies those sets, and ★ `filter_pointerDecode` **is the index transport of the note's §3**: the fine cells the decoding sends to `(s,t)` are *precisely* `LF6`'s pointer block, so the coarsening and clause (3)'s index map describe the same basins and nothing was quietly re-indexed. Then ★★★ `toReal_measure_wingEvent_eq_P_st` — the two-wing record **event's** Born weight is `P_st a b s t` at every setting pair, no genericity hypothesis (through `localDeisolation_pointer_volume_local`); and the headline, ★★★ `toReal_measure_preimage_wingEvent_eq_P_st` — **one physical record history**: the probability *in the prepared, pre-measurement epistemic measure* that the run ends with the two wings recording `(s,t)` **is** `P_st a b s t`. The Born clause was only ever available at the *post-measurement* ray; composing it with the new ★★★ `RecordLayer.measure_preimage_liftBase` (the transport law in **preimage** form, the shape any statement about a run takes) and clause (6) (`localDeisolationFlow_realises_localNaimark`) moves it to where an experiment starts. ★★ `toReal_measure_preimage_wingAEvent_eq_of_setting` is **operational no-signalling along the dynamics**, each side in its own prepared measure after its own flow — and it **discharges, for the singlet, the premise `measure_wingAEvent_eq_of_fineSum_eq` leaves open**, which this row recorded as not discharged: `LF3.marginal_a_eq_half` supplies exactly that equality. ★★ `factorsThrough_wingEvent_pointerDecode` gives obligation (5)'s shape at this instance. ⚠️ **The marginals and no-signalling are imported arithmetic, not new content** — `localDeisolation_A_marginal_volume_eq_half` and `localDeisolation_no_signalling_A` already prove them for the index sums; the contribution is the *event* formulation and the *prepared-ray* formulation. ⚠️ **No CHSH here**, no causal structure, and the missing `RemoteSettingLocalityB` witness of §4 is still missing; (2) dynamical record stability ~~**S, conditional**~~ ▸ **DONE 2026-10-11** ([`RecordLayer/TwoWingStability.lean`](../CsdLean4/RecordLayer/TwoWingStability.lean), 12 pins; S). **The whole content is that coarsening cannot destabilise.** `wingCode` is a function of `outcomeCode` (through `fineIndex`), so ★★★ `unstable_wingCode_subset`: the coarse exceptional set sits **inside** the fine one, and every bound #103 proved is inherited with no new argument. ★★★ `measure_drift_wingCode_ne_le` is the obligation: under a drift of the record coordinate at rate `ω` the two-wing record survives all of `[0, T]` outside a set of epistemic measure `N·ω·T` — **one** joint context, so no factor of `k` — with ★★★ `measure_wingA_ne_le` / `measure_wingB_ne_le` for each wing separately and ★★ `mem_wingEvent_sigmaShift_iff` in event form. ⚠️ **The condition this was priced under still stands**: cheap for the fibre dynamics, **absent for a general `Φ_t`**, so a write that is not a translation of the record coordinate — `CV.recordStroke`'s base-dependent one included — is not covered. ⚠️ **And no sharpening**, which is worth recording: one might expect four joint outcomes to be *more* stable than `N` fine cells, and this bound does not deliver it. A write can only relabel across a **fine** cell boundary, and the coarse reading survives that exactly when the decoding gives the two *adjacent* cells the same value — a property of the decoding's adjacency, which `circleCell`'s index order and `LF6.pointerDecode`'s arbitrary index equivalence do not supply. That is **row 138**; (3) records to causal regions ~~**M**~~ ▸ **CAUSAL-SEPARATION HALF DONE 2026-10-10** ([`CV/TwoWingCone.lean`](../CsdLean4/CV/TwoWingCone.lean), 14 pins; S of the M). **One fact drives it**: `graphNeighborhood twoWingGraph wingAModes = wingAModes` — A's mode set is *closed* under the graph's one-step neighbourhood, because the only edge touching it has both endpoints inside it. So A's cone **never grows**, and ★★★ `spacelike_twoWing` holds at **every** period count: causal separation here is not a bound that degrades with time, so every theorem of `RecordInfluence.lean` applies at every `n` with no budget to track. ★ `not_eventuallyInfluences_twoWing` — neither wing *ever* influences the other; ★★★ `commute_record_twoWing` — the wings' evolved observables commute at every `n`, which is what a two-wing experiment needs in order to **have** a joint outcome; ★★★ `arenaObs_kick_twoWing` and ★★★ `recordStroke_comm_kick_twoWing` — a B-wing intervention cannot steer A's reading or A's record write, **exactly** and at every period count, with no Lieb–Robinson error term. Non-vacuous: `supportedOn_wingAModes_modeOp`, `supportedOn_wingBModes_phaseDiagU` and the witnessed corollary. ⚠️ **TWO FINDINGS, AND THEY RE-PRICE THE REST OF THIS OBLIGATION.** The audit priced the carry to `KSigma 16` as a thin transport along §3's index equivalence. **It is not.** (i) `dmVec`, `arenaObs`, `arenaKick` and `SupportedOn` are **all indexed by `FieldConfig K N` specifically**, so carrying the statement across `FieldConfig 4 2 ≃ Fin 16` needs the CV arena API generalised over its index type — which `CompositeArena.lean`'s own rule-of-two note already records as owed. That is **new row 133**. (ii) More interestingly, the record-layer counterpart is **not the same theorem**: these are statements about a *wing-local observable*, whereas under I1 the record layer's wing reading is a *coarsening of a joint code* (`momentContext`'s rate is the moment map in the configuration basis — a reading of the **whole** configuration, not of either wing), so the record-layer side of unsteerability is the *marginal* statement `toReal_measure_preimage_wingAEvent_eq_of_setting` already landed, and a genuinely wing-local `ContextField` is **new row 134** (which bears on row 132, since under I1 the two wings share one selector). The fibre needs no transport at all: `CV.RecordFibre` and `LF4.KTorus` are the same type; (4) joint Bell and no-signalling **M–L**, ~~gated on (I) and~~ **ungated 2026-10-10 by I1** and ~~still needing the missing B-wing witness~~ ▸ **the B-wing witness landed 2026-10-11** (above), so the *no-signalling* half now has a model; and ▸ **THE BELL HALF LANDED THE SAME DAY** ([`LF6/TwoWingBell.lean`](../CsdLean4/LF6/TwoWingBell.lean), 7 pins; S of the M–L). ★★★ `chsh_preparedRecord_eq_neg_two_sqrt_two`: the CHSH combination of the two-wing **record** correlations is `−2√2`, and ★★★ `bellClassicalBound_lt_abs_chsh_preparedRecord` says that exceeds the local-hidden-variable bound. ⚠️ **No Bell theorem was proved and none was needed** — `LF3.correlation_eq_neg_dot`, `Empirical.Bell.chsh_singlet_at_optimal_angles` and `Empirical.QM.E91.lhvCHSH_abs_le_two` are all **imported**. What is new is the *quantity being correlated*: until now the corpus' CHSH was a function of `P_st`, an algebraic kernel, and it is now a function of the **measures of record events on `Σ`** — sets of ontic microstates, read in `epistemicMeasure`, **after the de-isolation flow from the prepared ray** (★★★ `preparedRecordCorrelation_eq_correlation`). So Bell violation is for the first time a statement about the *same objects* as no-signalling, record stability and causal separation. ⚠️ **The four terms are four different runs**, each in its own prepared measure — which is what a Bell test is, but is **not** the single-`μ` idealisation `LF3.OperationalNoSignalling` is stated in; the two must not be run together. The obligation's remainder is the record-layer instantiation of the setting-locality primitives (row 137). ▸ **And one candidate for that instantiation is now ruled out by theorem**: ★★★ `RecordLayer.eq_of_vadd_preimage_circleCell_eq` — a **rotation** of the record circle preserves a cell's **width**, since translation preserves Haar measure and a cell's measure *is* its rate, so it can never implement a change of conditional law. The note's §4 suggested exactly a piecewise rotation; a `RemoteSettingLocalityB` reroute at the record-layer basins must instead **move mass between A's cells**, which the B-wing primitive permits (it constrains only `wingB`) but which means the two wings' relabellings cannot come from one fibrewise family; (5) the coarse projection retains them ~~**S–M**~~ ▸ **CLOSED 2026-10-11, AND DELIBERATELY WITHOUT A NEW THEOREM.** Its three parts are each already proved: the wing events are **macroscopic** (★★ `LF6.factorsThrough_wingEvent_pointerDecode`, preimages of #102's record string), they carry the **Born weights** (★★★ `LF6.toReal_measure_wingEvent_eq_P_st`, and ★★★ `toReal_measure_preimage_wingEvent_eq_P_st` along the flow), and they are **stable** (★★★ `measure_drift_wingCode_ne_le`, obligation (2)). ⚠️ **A capstone conjoining those three would be exactly the move this row was opened to avoid** — the author's instruction was that the existence claim be “supported by an explicit admissible witness, not proved by packaging the five desired conclusions as fields of a structure”. So the obligation is discharged by *naming* its three theorems here rather than by a fourth that restates them. Nothing is owed; if a capstone is ever wanted it should be the **instantiated** statement at `LF6`'s witness, and it still waits on obligation (4)'s Bell half. ⚠️ **It would not establish `C-1`**: adjacency and separation are still supplied through `E`, and nothing would show that all apparent nonlocality is a consequence of projection alone — the author's further move, replacing `E` by an effective adjacency read off the generator, is a separate row and is not priced here. | ~~**M–L** for the existence theorem, in the five obligations above~~; the audit itself is **done (S)**. ▸ **ALL FIVE OBLIGATIONS ARE NOW INDIVIDUALLY DISCHARGED (2026-10-11) — AND THE ROW IS STILL NOT DONE, THOUGH FOR A DIFFERENT REASON THAN IT WAS AN HOUR AGO.** (1) record history, (2) stability, (3)'s causal half, (4) no-signalling **and** Bell, (5) the coarse projection: each has its theorems. What the row actually asked for is the **conjunction** — all four properties “within the *same* dynamical model” — and that is **still owed**, for a reason that is structural rather than a matter of effort: the Born, record, stability, no-signalling and Bell statements live on `KSigma 16`, while causal separation (`CV.spacelike_twoWing`) lives on `FibredFieldArena 4 2`. ~~Those are the same space up to the audit's §3 index equivalence, and making them literally one arena is **row 133** — blocked on the CV arena API being index-polymorphic, which is a core refactor and not paperwork.~~ ▸ **THE ARENA GAP CLOSED 2026-10-11** (row 133's first half): `CV.twoWingArenaEquiv` identifies `FibredFieldArena 4 2` with `LF4.KSigma 16` through a linear isometry, and ★★★ `CV.sigmaRecordStroke_comm_sigmaFibredKick_twoWing` states causal separation **on `Σ`** — so all five properties are now statements about one space. ⚠️ **AND THAT EXPOSED THE REAL REMAINING GAP: ONE ARENA, TWO DYNAMICS.** Causal separation is about the CV interacting unitary `graphInteractingU 4 2 τ λ twoWingGraph gc`; the Born weights, the record history and the CHSH value are about `LF6.localDeisolationFlow`. Those are **different maps on the same space**, and *nothing in the corpus relates them* — so the experiment's causal structure comes from one dynamics and its statistics from another. That is **row 139**, and it is the last thing between #131 and its stated deliverable. ▸ **CLOSED 2026-10-12** (~~#139~~). The CSD-preferred route was taken — read the separation off the *measurement* flow rather than import a second interaction — and ★★★ `CV.reindex_localFlowUnitary_eq_joinU` makes it a theorem rather than an index convention: the de-isolation flow, read on the four-mode configuration space, **is** the wing-factorised product unitary. So ★★★ `CV.recordStroke_comm_kick_deisolation` gives causal separation **under the same interaction** the Born weights, the record history and the CHSH value are about. ▸ **WHERE THAT LEAVES THIS ROW: substantively discharged, deliberately not struck.** All five obligations have their theorems; they are about **one arena** (~~#133~~'s first half) under **one interaction** (~~#139~~). ~~Three things are still true and none is a gap in the mathematics. **(1)** One bookkeeping identity is unproved — `arenaKick (pullback) = arenaEquiv.symm ∘ localDeisolationFlow ∘ arenaEquiv` relating the CV kick to the transported projective flow; mechanical, no choices left.~~ ▸ **(1) CLOSED 2026-10-12**: ★★★ `CV.flowArenaEquiv_arenaKick` — the aligned arena identification **intertwines the CV kick with the de-isolation flow**, so every cone statement is a statement about `LF6.localDeisolationFlow` transported. The two theorem families are about the *same map*, not merely the same space. (The identification used is `CV.flowArenaEquiv`, built from the de-isolation's own wing split — **not** `ArenaIndexBridge.arenaEquiv`, whose index change is an arbitrary `Fintype.equivFinOfCardEq` and carries no wing structure; that distinction is the whole content of the step.) ▸ **SO EVERYTHING THIS ROW ASKED FOR IS PROVED, AND TWO THINGS REMAIN TRUE, NEITHER A GAP IN ITS MATHEMATICS.** **(2)** No capstone theorem conjoins the five, **and that is deliberate** — the author's instruction was that the existence claim not be “proved by packaging the five desired conclusions as fields of a structure”, so they are named here instead, all of one explicit witness: the singlet prepared as `localNudgeVec a b`, on `KSigma 16`, under `localDeisolationFlow`. **(3)** Locality and adjacency are still *assumed*, now once and physically: `localDeisolation_factorises` holds by construction, LF6-A.3 having chosen a product dilation. That is better than the graph's supplied `E`, and it is still an assumption — deriving it is `R-016`. ⚠️ **And the row never claimed `C-1`**: nothing here shows that all apparent nonlocality is a consequence of projection alone. ▸ **STRUCK 2026-10-12, AT THE AUTHOR'S INSTRUCTION.** Everything the row asked for is proved of the witness named above — one arena, one interaction, six properties. The only artefact not produced is the single conjunction, which (2) records as ruled out by the author's own instruction. The decision as it stood before the strike: The only artefact not produced is the single conjunction, which (2) records as ruled out by instruction; whether that counts as the row's “existence theorem” is the author's call, not this ledger's, so the row is left open rather than struck. Remaining price: **M–L**, almost all of it #139 | High — it tests `C-1`'s central physical claim rather than extending structures already assumed | #39; #38; `C-1`; [`two-wing-experiment-scoping.md`](two-wing-experiment-scoping.md); [`q10-no-signalling-scoping.md`](q10-no-signalling-scoping.md) | > | ~~132~~ | Foundations | Records→spacetime | *new (2026-10-10, out of #131's blocker (I))* **A two-channel record write: `x.2.2` as a second selector (options I2 and I3)** — the alternative to the I1 coarsening that #131 landed, and the thing a critic presses on. I1 gives two wings as two readings of **one** ontic selector `x.2.1`, which is honest for CSD but means `CV.recordStroke₂`'s **two** fibre channels are not the write mechanism for that model. This row is the other half of the fibre: give the second torus coordinate record content and ask whether two genuinely separate selectors can carry the singlet. **I2 — chained selectors**: a second basin family whose rates depend on the first selector's outcome, so the pair is read as (A's cell, then B's cell given A's); the joint law is then a product of a marginal and a conditional, which is *not* the Bell-local factorisation (`IsProductPartition`'s arity has no slot for the remote setting), but the conditional must be shown to come from an admissible interaction rather than chosen to fit `P_st`. ~~**I3 — a non-product partition of `T²`**: cells that are not rectangles in (`x.2.1`, `x.2.2`), so neither wing's cell geometry is a function of its own coordinate alone;~~ ▸ **CORRECTED AND RESOLVED 2026-10-10.** Two errors in that sentence. (a) The requirement is not that the *cells* be non-rectangles — it is that the **partition** not be a product of two partitions; `skewCell`'s cells are rectangles and the partition is still not a product, because the `θ₂`-arc moves with A's cell. (b) “Neither wing a function of its own coordinate alone” is **not achievable and not needed**: the fibre no-go `volume_prodPartition_eq_mul` forces only **one** wing to read both coordinates. ▸ **And I2 and I3 are the same construction**, so this row's two options are one: a correlated partition of `T²` *is* a chained one, by the chain rule. Landed under ~~#134~~ ([`RecordLayer/SkewFibrePartition.lean`](../CsdLean4/RecordLayer/SkewFibrePartition.lean)) — so what is left of this row is **only the posit question**: is `x.2.2` allowed record content at all; this needs a torus-cell API the corpus does not have ([`TorusFibre.lean`](../CsdLean4/RecordLayer/TorusFibre.lean) builds `torusCell` from `x.2.1` only and documents the second coordinate as the symplectic partner with no record content). ⚠️ **Not required by #131** — that row's existence theorem is being built on I1, and nothing in the corpus depends on this. It is worth a row because it is the only route by which the two wings could be written **independently** in `Σ`, and because repurposing `x.2.2` would have to be reconciled with the symplectic role [`TorusFibre.lean`](../CsdLean4/RecordLayer/TorusFibre.lean) and `R-016′` give it — which is a foundations question, not a refactor. ~~First deliverable either way: decide whether `x.2.2` is allowed record content at all, and say so in `POSITS.md`.~~ ▸ **DONE 2026-10-11 — `specs/POSITS.md` **Posit 10**, the record dimension of the fibre.** The entry records the decision (**I3**: `θ₂` does carry record content), what forced the shape of the question, and what the decision costs. Three things in it are worth reading before citing this row again. **(1) The question is a genuine binary, by theorem.** `SkewFibrePartition`'s ★★★ `volume_prodPartition_eq_mul` / ★★★ `not_exists_prodPartition_of_ne_mul` close the middle option — one coordinate per wing — by Fubini, so the live options are *share `θ₁`* (I1) or *one wing reads both* (I3), and there is no third. ⚠️ The no-go does **not** rule out I1: both its wings read `θ₁`, and `TwoWingCoarsening.measure_coarseEvent` carries correlation fine. **(2) The posit was never about parity.** `TorusFibre.lean`'s reason for `T²` is dimensional — an odd-dimensional fibre carries no symplectic form — and that argument delivers a second coordinate while saying nothing about whether a record may be read from it. The two are independent, and conflating them would make a modelling choice look like a structural fact; `mem_torusCell_iff` now carries a ⚠️ saying so. **(3) The decision is recorded, not implemented.** `globalBasin` and everything `LF6`'s Born clause is stated against still read `θ₁` alone, and `SkewFibrePartition` is kinematic, so the corpus *runs* on I1 while the decision is I3 — the gap is **#135**, and until it lands cite `torusCell`/`globalBasin` for what the corpus computes and Posit 10 for what it intends. ⚠️ **What the posit still does not have is a discharge**, and no cheap one exists: deriving which coordinates carry record content means showing that the record-creating interaction writes the second channel, which is the `H_int` frontier (`R-016`) again. `CV.recordStroke₂` is the field-side precedent for a two-channel write and the reason the corpus had been running two conventions at once. | ~~**M–L** (I2), **L** (I3)~~ **both landed as one under ~~#134~~**; ~~the posit question remains at **S**~~ **took S (Posit 10)** | Medium — it does not advance #131, it answers the objection #131 concedes | #131; [`two-wing-experiment-scoping.md`](two-wing-experiment-scoping.md) §5 blocker (I); [`sigma-fibre-contextuality.md`](sigma-fibre-contextuality.md); `R-016′` | > | ~~133~~ | Foundations | Records→spacetime | *new (2026-10-10, out of #131 obligation (3))* **The CV arena API over an arbitrary index type** — what makes `FieldArena 4 2` and `CPN 16` *literally* one arena rather than two spaces related by an unproved identification. `dmVec`, `arenaDM`, `arenaObs`, `arenaKick` and `SupportedOn` ([`CV/ArenaBridge.lean`](../CsdLean4/CV/ArenaBridge.lean), [`CV/ModeLocality.lean`](../CsdLean4/CV/ModeLocality.lean)) are all stated over `FieldConfig K N = Fin K → Fin N`; the mathematics uses only that the index is a finite type with decidable equality, so the generalisation is a restatement plus the transport lemmas `arenaObs_reindex` and `arenaKick_reindex` along an index equivalence (through `LinearIsometryEquiv.piLpCongrLeft` and `Projectivization.map`, with `Matrix.reindexAlgEquiv` on the operator side — `CompositeArena.lean`'s `compositeReindex` is the pattern). **The corpus already records this as owed**: `CompositeArena.lean`'s scope note says heterogeneous composites “need the arena API generalised over its index type (rule-of-two note: generalise `ArenaBridge` when next touched)”, so this discharges that note too. ⚠️ It is a refactor of a core CV module with a wide blast radius (every CV file that mentions `arenaObs`), which is why it is priced separately rather than folded into #131. ~~First deliverable: the generalised `ArenaBridge` with the old names as abbreviations, so nothing downstream breaks.~~ ▸ **THE “ONE ARENA” HALF IS DONE 2026-10-11, AND BY A DIFFERENT ROUTE THAN THIS ROW PROPOSED.** The generalisation was not needed to identify the *spaces* — only to transport the *notions*. What was missing was a plain piece of algebra the corpus lacked: Mathlib's `Projectivization.map` has **no inverse**, and this repository's `mapEquiv` is for self-equivalences, so two projectivizations that differ by an index change could not be identified at all. [`Mathlib/…/Projectivization/Congr.lean`](../CsdLean4/Mathlib/LinearAlgebra/Projectivization/Congr.lean) supplies ★★★ `Projectivization.congr` (6 pins, Cat-1, Mathlib-shaped), and [`CV/ArenaIndexBridge.lean`](../CsdLean4/CV/ArenaIndexBridge.lean) (4 pins) uses it: ★★★ `arenaEquiv`, ★★★ `fibredArenaEquiv` and `twoWingArenaEquiv : FibredFieldArena 4 2 ≃ LF4.KSigma 16`, with the index change entering as a **linear isometry** so every Born weight is untouched, and the fibre needing no transport at all (`CV.RecordFibre` and `LF4.KTorus` are the same type). ★★★ `sigmaRecordStroke_comm_sigmaFibredKick_twoWing` is the cone theorem as a statement about `Σ`. ⚠️ **Honest about what that is**: the transported operations are *defined* through the equivalence, so the theorem is `TwoWingCone`'s conjugated and contains no new mathematics — what it buys is that the properties are about **one space**. ⚠️ **And one claim was pulled before landing**: a general `congr_measurable` resolves `MeasurableSpace (ℙ K V)` by some path other than `MeasureSpace.lean`'s Borel structure when the finite-dimensionality hypotheses are absent — the unused-section-variable linter caught it — so measurability is proved at the use site instead, where the instance is unambiguous. ~~⚠️ **What is left of this row is the API generalisation itself**: `arenaObs`, `arenaDM`, `arenaKick` and `SupportedOn` over an arbitrary finite index, which is what would let the CV *notions* be read on `Σ` rather than transported wholesale, and which also discharges `CompositeArena.lean`'s rule-of-two note.~~ ▸ **THE GENERALISATION LANDED 2026-10-12** ([`CV/ArenaBridge.lean`](../CsdLean4/CV/ArenaBridge.lean), generalised in place, 1 new pin; S–M). `GenArena ι` is the generic arena and `FieldArena K N` is its instance at the configuration space, so `unitVec`, `dmVec`, `arenaDM`, `arenaObs` and `arenaKick` now read over **any** finite index — and **every existing statement elsewhere in the corpus is unchanged**: the full build has **zero** downstream errors, which is exactly what this row asked for. ★★★ `arenaObs_kick_conj` is the bridge identity stated index-generically. This also discharges `CompositeArena.lean`'s rule-of-two note. ⚠️ **Two findings about what *cannot* be generalised, and neither is a gap.** `SupportedOn` is not generalised **and cannot be**: its content is that an operator does not move the modes outside a region, which is a statement about the index *being* a function space — so the statics and light-cone sections stay `FieldConfig`-indexed. And `heisenberg` is still `FieldConfig`-indexed because its companion lemmas are about mode support, which is why the bridge identity splits in two (`arenaObs_kick_conj` generically, `arenaObs_kick` in `heisenberg` form at the configuration space). ~~⚠️ **What remains is the row's second part: the transport lemmas.**~~ ▸ **THE TRANSPORT LEMMAS LANDED 2026-10-12, AND THE ROW IS CLOSED** ([`CV/ArenaIndexBridge.lean`](../CsdLean4/CV/ArenaIndexBridge.lean), 14 new pins; [`Mathlib/LinearAlgebra/Matrix/ReindexTrace.lean`](../CsdLean4/Mathlib/LinearAlgebra/Matrix/ReindexTrace.lean), Category 1, 2 pins; S–M took S–M, and the recorded route held with no surprises). `idxSpaceEquiv`, ★★★ `idxArenaEquiv` and `idxUnitary` transport state spaces, rays and unitaries along any `e : ι ≃ κ`; ★★ `dmVec_idxSpaceEquiv` and ★★ `arenaDM_idxArenaEquiv` say the density is **reindexed, not rescaled** (an index change is an isometry, so the normalisation is the same number on both sides); and the two the row asked for by name: ★★★ `arenaObs_reindex` — **the observable is invariant**, `arenaObs (reindex e e A) (idxArenaEquiv e p) = arenaObs A p`, so a CV reading is *the same real number* on either arena — and ★★★ `arenaKick_reindex`, the same for an intervention. **THE PAYOFF IS THAT THE ROW'S TWO HALVES ARE ONE CONSTRUCTION:** ★ `arenaEquiv_eq_idxArenaEquiv` is `rfl` — the field-arena-is-`Σ` equivalence of the first half *is* the generic transport at `configEquiv K N` — so ★★★ `arenaObs_arenaEquiv` and ★★ `arenaKick_arenaEquiv` carry a CV reading and a CV intervention to `LF4.CPN (N ^ K)` with no second construction and no unproved identification. ⚠️ **Two findings, both about duplication rather than mathematics.** (i) `arenaKick_mk` **already existed** in `CompositeArena.lean`, `FieldConfig`-indexed, with its own docstring recording that it was “general-`K` API, placed here as its first consumer” — so it moved up to the generic layer and the copy is gone; the clash was found by the compiler, not by reading. (ii) The trace and multiplicativity facts about `Matrix.reindex` are **not in Mathlib** at the `v4.34.0` pin and were being re-derived; they are now Category-1 `Matrix.reindex_mul` and `Matrix.trace_reindex`, which is the second occurrence the rule of two asks for — `trace_compositeReindex` was the first. ⚠️ `SupportedOn` is still **not** transported and cannot be, for the structural reason above; nothing here is stated for it. ⚠️ The transport is **pointwise**: no measure is claimed to be carried, so a measure-theoretic statement crossing the bridge still needs `fsMeasure`-preservation, which no theorem supplies. | ~~**M–L**~~ all three parts took **S–M**: the arena half, the API generalisation, and the transport lemmas — **ROW COMPLETE** | Medium — it unblocks #131's “one arena” claim and discharges a standing rule-of-two note | #131 obligation (3); [`CV/ArenaBridge.lean`](../CsdLean4/CV/ArenaBridge.lean); [`CV/CompositeArena.lean`](../CsdLean4/CV/CompositeArena.lean); [`CV/TwoWingCone.lean`](../CsdLean4/CV/TwoWingCone.lean) | > | ~~134~~ | Foundations | Records→spacetime | *new (2026-10-10, out of #131 obligation (3))* **A wing-local `ContextField`: a record coordinate that reads one wing only** — the thing that would let [`CV/TwoWingCone.lean`](../CsdLean4/CV/TwoWingCone.lean)'s unsteerability theorems be read as statements about **records** rather than about field observables. `globalBasin c i` reads the base through `c.rate`, and every `ContextField` the corpus has is global: `momentContext`'s rate is the moment map in the configuration basis, a reading of the **whole** configuration. Under option I1 that is by design — the wings are coarsenings of one joint code, and their no-signalling statement is the *marginal* one (`LF6.toReal_measure_preimage_wingAEvent_eq_of_setting`, landed). A wing-local context would need rates built from a wing-supported observable, i.e. a per-wing POVM read off `LF6.wingDeisolationV`/`blockProj`, with the simplex conditions (`nonneg`, `sum_one`, `measurable_rate`) discharged for it. ⚠️ **This is a different model, not a missing lemma.** Two wing-local contexts each reading its own wing is exactly the two-selector shape blocker (I) ruled out for I1, so this row only makes sense alongside **row 132** (I2/I3) — or as a *second*, wing-local reading of the same joint selector, which would have to be shown consistent with the joint code rather than assumed. ~~First deliverable: decide which of those two it is, in one paragraph, before any Lean.~~ ▸ **DECIDED 2026-10-10 — the author chose I3 — AND THE FIBRE HALF IS LANDED** ([`RecordLayer/SkewFibrePartition.lean`](../CsdLean4/RecordLayer/SkewFibrePartition.lean), 13 pins; took S–M of the M). **And the choice turned out to be constrained.** ★★★ `volume_prodPartition_eq_mul`: if A's region is a `θ₁`-set and B's region a `θ₂`-set then the joint law is **exactly** the product of the marginals — Fubini, with no hypothesis on the sets — and ★★★ `not_exists_prodPartition_of_ne_mul` turns it around: **a joint law that is not the product of its own marginals has no such partition of the fibre at all.** ⚠️ **Correction to this row's first wording, 2026-10-11**: the hypothesis is that the two wings read *different* coordinates, one each, so the theorem does **not** apply to I1 — where both wings read `θ₁` and correlation is carried fine. What it rules out is the **one-coordinate-per-wing** shape, leaving exactly two live options: share a coordinate (I1) or have a wing read both (I3). It is the fibre-level analogue of `LF6.no_product_partition_realises_singlet`, and it turns `TwoWingCoarsening`'s prose claim — that independent selectors factorise the joint law — into a theorem. The construction: `skewCell` puts A's arc in `θ₁` and, in `θ₂`, the arc of the **conditional** rate `q s t / rowSum q s`. ★★★ `volume_skewCell` — **exact Born weights for an arbitrary joint law** (`rowSum · condRate = rate`), so the fibre carries *any* correlation with no error term, the singlet's included; ★★★ `volume_wingAFibre` / ★★★ `volume_wingBFibre` — both marginals come out right, and B's is the non-trivial one because the construction never mentions `colSum`, so **no-signalling for B is a theorem about the geometry**, not something arranged; ★★★ `not_exists_sndSet_of_condRate_ne` — **`x.2.1` is load-bearing for B's record**: when the conditional law genuinely depends on A's cell, *no* `θ₂`-set describes B's region. With the no-go that is one fact seen twice: correlation and the `θ₂`-nonlocality of B's region are the same thing. ⚠️ **THE FINDING — I2 AND I3 ARE THE SAME CONSTRUCTION**, so #132's two alternatives are one and the author's choice is **binary** (I1's single coarsened selector, or this), not ternary: a partition of `T²` carrying correlation must have a wing reading both coordinates, writing the joint law as marginal-times-conditional is then the chain rule, and `condRate` *is* that conditional. ⚠️ **I3 does not make the wings symmetric** — A's region *is* a `θ₁`-set; the no-go forces only **one** wing to be non-local and the chain rule picks which, so “two independent selectors” would be the wrong description: two coordinates, both load-bearing, one conditional dependence. ⚠️ **The fibre only.** The rates are an argument; no base point, no `ContextField`, no `epistemicMeasure`. The base-coupled joint context field and its skew basin — which is what would actually replace `globalBasin` — is **row 135**. | ~~**M** (and the decision first at **S**)~~ took **S–M**; residue = #135 | Medium — it is what connects the causal half of #131 to the record half; low if the author is content with the marginal statement | #131 obligation (3); #132; [`CV/TwoWingCone.lean`](../CsdLean4/CV/TwoWingCone.lean); [`RecordLayer/GlobalBasin.lean`](../CsdLean4/RecordLayer/GlobalBasin.lean) | > | ~~135~~ | Foundations | Records→spacetime | *new (2026-10-10, out of ~~#134~~)* **The base-coupled skew basin: a joint context field and its `T²` partition on `Σ`** — the step that would let I3's fibre partition actually replace `globalBasin`. [`SkewFibrePartition.lean`](../CsdLean4/RecordLayer/SkewFibrePartition.lean) takes the joint rate table as an **argument**, so it is kinematic: nothing is attached to a base point. What is owed is the analogue of `GlobalBasin.lean` — a `JointContextField N m n` (`rate : CPN N → Fin m → Fin n → ℝ` with `measurable_rate`, `nonneg`, `rowPos`, `sum_one`), the basin `skewBasin c s t = {x \| x.2 ∈ skewCell (c.rates x.1) s t}`, its measurability (the pattern is `measurableSet_globalBasin`, plus measurability of the **quotient** `rate / rowSum`, which is where the work is), and ★★★ the Born weight `epistemicMeasure p (skewBasin c s t) = ofReal (c.rate p s t)` (the pattern is `globalBasin_prob`). Then the two wing marginals transport, and the I1 coarsening of [`TwoWingCoarsening.lean`](../CsdLean4/RecordLayer/TwoWingCoarsening.lean) has a genuine alternative at the level `LF6`'s Born clause is stated at. ⚠️ It does **not** by itself give the wing-local *record* reading #131 obligation (3) wanted, because that also needs the rates themselves to be wing-local observables — under I3 the `θ₂` arc is conditioned on A's cell by construction, so “B's record reads only B” stays false and should not be claimed. ~~First deliverable: the structure and `skewBasin`'s measurability; the Born weight follows.~~ ▸ **DONE 2026-10-11** ([`RecordLayer/SkewBasin.lean`](../CsdLean4/RecordLayer/SkewBasin.lean), 16 pins; took S–M of the M). `JointContextField` is the joint context as a field over the base, and `JointContextField.rates` evaluates it to a `JointRates` at a base point — so every fibre theorem of ~~#134~~ applies unchanged and nothing had to be re-proved. ★★★ `skewBasin_prob`: **the Born weight**, `epistemicMeasure p (skewBasin c s t) = c.rate p s t`, which is the statement `LF6`'s Born clause is written in — **so I3 now reaches the level I1 was already at**. ★★★ `wingABasin_prob` / ★★★ `wingBBasin_prob` are the two wing marginals on `Σ`, and ★★★ `wingBBasin_prob_eq_of_colSum_eq` is no-signalling with its premise named in the I3 setting. The predicted cost was right: the only new regularity obligation over I1 is `measurable_condRate`, the measurability of the **quotient** `rate / rowSum`. `constJointContext` makes it non-vacuous for *every* joint table, correlated ones included. ⚠️ **It does not give a wing-local record, as this row said, and the file says so**: under I3 the `θ₂` arc is conditioned on A's cell by construction, so “B's record reads only B” is false here and is not claimed. ⚠️ **And the corpus still runs on I1** — `globalBasin`, `outcomeCode`, `recordString` and `LF6`'s capstone read `θ₁` alone; this is the I3 alternative at the same level, not a migration, and nothing here retires anything there. ⚠️ **One limitation is sharper than expected and is now row 136**: `rowPos` must hold at *every* base point, so the only joint field exhibited is the **base-independent** one, and there is no I3 analogue of `momentContext` yet. I1 has a moment-map context; I3 has a constant table. That is not parity, and it should not be described as such. | ~~**M**~~ took **S–M**; residue = #136 | Medium — it is what makes I3 a record-layer construction rather than a fibre fact | ~~#134~~; #132; #131 obligation (3); [`RecordLayer/GlobalBasin.lean`](../CsdLean4/RecordLayer/GlobalBasin.lean) | > | ~~136~~ | Foundations | Records→spacetime | *new (2026-10-11, out of ~~#135~~)* **The moment-map joint context: an I3 rate field that comes from the base** — what would put I3 at parity with I1. [`SkewBasin.lean`](../CsdLean4/RecordLayer/SkewBasin.lean) exhibits only `constJointContext`, whose table does not move with the ontic point, because `JointRates.rowPos` demands a strictly positive A-marginal at **every** base point and a `momentMap`-derived block marginal vanishes on rays whose block has zero amplitude. I1 has `momentContext`; I3 has a constant. ▸ **The route is recorded, and `rowPos` looks removable.** Lean's `x / 0 = 0` makes `condRate s t = 0` for a null row, and then `rowSum · condRate = rate` still holds there — because non-negativity forces `rate s t = 0` whenever `∑ₜ rate s t = 0` — while `sum_condRate` becomes `0` rather than `1`, which is fine for `loSum_condRate_le_one` since `loSum 0 + 0 ≤ 1` directly. So the two lemmas that use `rowPos` (`JointRates.sum_condRate`, `JointRates.rowSum_mul_condRate`) should go through on a case split, the null row contributing an **empty** `θ₂` cell of the right measure zero. ⚠️ Unverified — this is a route, not a proof, and `volume_wingAFibre`'s covering step needs re-checking in the null case. Then the field: block-coarsen `momentMap` over the system index (which is exactly what `LF6`'s pointer-block sum already does, and what `filter_pointerDecode` identifies), with measurability from `LF4.measurable_momentMap`. ~~First deliverable: drop `rowPos` from `JointRates`, or record with a counterexample why it cannot go.~~ ▸ **DONE 2026-10-11 — the route held and `rowPos` is gone.** ★ `JointRates.rate_eq_zero_of_rowSum_eq_zero` is the lemma that does it (a vanishing A-marginal forces the whole row to vanish, by non-negativity), and ★ `rowSum_mul_condRate` and ★ `loSum_condRate_le_one` now hold by a case split, the null row contributing an **empty** `θ₂` cell of measure zero. The covering step I flagged as needing re-checking turned out not to arise: `volume_wingAFibre` sums cell volumes and never uses that the conditional cells cover. Positivity survives only as an explicit hypothesis of `not_exists_sndSet_of_condRate_ne`, where it is genuinely needed — a null A-cell has no second coordinate to be non-local in. `JointContextField.rowPos` went with it, so **nothing in the structure now stands in the way of a `momentMap`-derived field**. ~~⚠️ **What is left of this row is that field itself**~~ ▸ **THE FIELD LANDED 2026-10-13 AND THE ROW IS CLOSED** ([`RecordLayer/JointMomentContext.lean`](../CsdLean4/RecordLayer/JointMomentContext.lean), 6 pins; M took S–M once the right lemma was found). `jointMomentContext b` is the moment map coarse-grained by a **degeneracy map** `b : Fin N → Fin m × Fin n`: the joint rate of `(s, t)` is the moment map summed over the directions `b` sends to that pair, with all three field conditions coming from the moment map's. ★★★ `skewBasin_prob_jointMomentContext` — the skew basins carry those weights, immediately, because `skewBasin_prob` was always generic in the field. ▸ **AND THE PARITY IS A THEOREM RATHER THAN A REMARK**, which is what the row wanted: ★★★ `jointMomentContext_rowSum` says **A's marginal IS `blockField`'s rate at A's coarsening**, with ★★ `jointMomentContext_colSum` for B, and ★ `jointMomentContext_rowSum_single` recovers `momentContext` at one B-outcome and the identity on A — so I1 is the degenerate case of this field, not a separate construction. ⚠️ **THE SINGLE-INDEX VERSION ALREADY EXISTED, AND A NEW GUARD IS WHAT FOUND IT.** `blockField` ([`OutcomeField.lean`](../CsdLean4/RecordLayer/OutcomeField.lean)) is this construction at one index — the same finite sum of moment coordinates, the same three obligations, discharged the same way — and the row's route said as much (“which is exactly what `LF6`'s pointer-block sum already does”) while pointing at the wrong file. It **cannot** be instantiated here, because `OutcomeField N K` indexes outcomes by `Fin K` and a joint table is indexed by a *product*; bridging that through `Fin m × Fin n ≃ Fin (m * n)` would make the definition unreadable for no gain. So the definition is written out and **the relationship is proved instead** — the two marginal theorems are exactly that, which is the honest substitute for literal reuse, and `check-duplicates` confirms the file adds no duplicate statement. ⚠️ **No correlation is created.** The joint table is whatever `b` and the moment map make it; a *correlated* table at a *moving* base point is this field at a `b` that entangles the two coordinates, and which `b` the two-wing experiment wants is the question `constJointContext` sidestepped by fixing the table by hand. The degeneracy map is a posit, exactly as it is for I1 (`POSITS.md` Posit 10). | ~~**S–M** for the `rowPos` removal~~ took **S**, ~~**M** for the field~~ took **S–M** — **ROW COMPLETE** | Medium — without it I3 is only realised with base-independent rates, so the I1/I3 comparison is not yet like for like | ~~#135~~; ~~#134~~; `POSITS.md` Posit 10; [`RecordLayer/GlobalBasin.lean`](../CsdLean4/RecordLayer/GlobalBasin.lean) (`momentContext`) | > | ~~137~~ | Foundations | Records→spacetime | *new (2026-10-11, out of #131 obligation (4))* **The setting-locality primitives instantiated at the record-layer basins** ▸ **DONE 2026-10-20 (L), on the fibre, with a FINDING about the primitive itself.** **The choice (the row's first deliverable): I3, and the reroute is a RECTANGLE EXCHANGE.** Under I3 the fibre is tiled exactly by the skew cells `A_s × B_{s,t}` of the context's table; A's cells do not move with B's setting when the tables are no-signalling, so **A's reroute is the identity**; for B, a change `a ↦ a'` replaces the table `q` by `q'`, and the reroute **normalises** each skew cell onto a rectangle of the horizontal band `[0,1) × C_t` (`C_t` the arc of B's marginal) by the Jacobian-one affine map stretching `θ₁` by `(rate/colSum)/rowSum` and `θ₂` by its reciprocal, then **denormalises** with the inverse exchange for `q'`: `reroute q q' = denormalize q' ∘ normalize q`. It moves mass between A's cells, as `eq_of_vadd_preimage_circleCell_eq` said it must. **Lean:** Cat-1 [`Mathlib/MeasureTheory/GluedMeasurePreserving.lean`](../CsdLean4/Mathlib/MeasureTheory/GluedMeasurePreserving.lean) (★★ `measurePreserving_glue`: pieces on an exact partition, each pushing `μ.restrict (P i)` onto `μ.restrict (Q i)`, glue to a measure-preserving map; 4 pins); [`RecordLayer/ArcAffine.lean`](../CsdLean4/RecordLayer/ArcAffine.lean) (`arcMap`, ★ `map_restrict_arcMap` = the arc transport law with the `λ⁻¹` scalar, ★★ `map_restrict_rectMap` = **Jacobian-one rectangle transport, exact**, `existsUnique_mem_circleCell` = the Born cells tile the circle EXACTLY; 4 pins); [`RecordLayer/SkewReroute.lean`](../CsdLean4/RecordLayer/SkewReroute.lean) (exact tilings `existsUnique_mem_skewCell`/`normalCell`, the pieces `toNormal`/`toSkew` with ★ exact transport, the glued `normalize`/`denormalize` ★★ measure-preserving, `reroute`, ★★ `measurePreserving_reroute`, ★★ `wingB_reroute` (needs `colSum q = colSum q'`), `wingA_eq_of_rowSum_eq`; the LF3 instance for `2×2` tables `Q : MeasurementContext → JointRates 2 2`: `skewOutcomeMaps`, ★★★ `skewLocalSettingLocalityA` (identity), ★★★ `skewLocalSettingLocalityB` (the exchange), ★★★ `operationalNoSignalling_skew` = **operational no-signalling for the I3 record basins DERIVED from the primitive**; 15 pins). **THE FINDING:** `LF3.RemoteSettingLocalityB` wants ONE `reroute a a'` for all `b`; the exchange depends on `b`, and **no `b`-blind map exists for Bell-violating tables**: a `b`-blind reroute carries `R_t(a,b) ∩ R_{t'}(a,b')` to the same intersection at `a'`, so the counterfactual joint law of B's two readings would be `a`-independent, and for the singlet at the CHSH settings it is `(1 − √2/2)/2` at one `a` and `1/2` at the other. The derivation never used blindness, so the primitive is restated with the local setting as an argument — [`LF3/SettingLocalityLocal.lean`](../CsdLean4/LF3/SettingLocalityLocal.lean) `LocalSettingLocalityA/B`, the blind form a special case (`toLocal`), ★ the same one-line derivations, ★★ `operationalNoSignalling_of_localSettingLocality` (3 pins) — and THAT is what CSD's record basins satisfy. **Row wording corrected:** under I3 (not only I1) A's half is trivial, since A's cells depend on the table only through `rowSum`. **Residues, priced:** #155 the base-coupled version over `KSigma N` via `skewBasin` (S–M), #156 the Lean proof that the blind B-primitive is unsatisfiable for the CHSH singlet tables (M), #157 the singlet's `MeasurementContext → JointRates 2 2` family as a concrete instance (S). ▸ *Original row:* — the instantiation the #131 audit (§4) actually asked for, as against the abstract witness that landed with it. `LF3.symmTranslationLocalityA`/`B` live on an arbitrary additive group with abstract wing responses; what the row wants is `Σ = KSigma N` as the state space, the record-layer basins as the readout (`globalBasin` under I1, or `skewBasin` under I3), and the `reroute` read off the cell geometry. ▸ **One thing is already settled, in the negative**: the reroute **cannot be a fibrewise rotation** of the record coordinate (★★★ `RecordLayer.eq_of_vadd_preimage_circleCell_eq` — translation preserves a cell's width, a setting change does not), so the audit's suggested “piecewise rotation / interval exchange” is not available as stated and the map must move mass **between** A's cells. It exists for measure-theoretic reasons — both of B's reading-sets have measure `colSum t`, and on a nonatomic standard space two sets of equal measure are related by a measure-preserving map — but *constructing* one, measurably and with `wingB` invariant along it, is the work, and Mathlib has no measure-space isomorphism API this reaches for directly. ⚠️ Under I1 there is a further wrinkle worth checking before any Lean: with uniform wing marginals A's region is literally setting-independent, so A's half is trivial and only B's is real. First deliverable: decide whether to instantiate under I1 or I3, and say what the reroute is, in one paragraph. | ~~**L** (and the choice first at **S**)~~ done (L) | Medium–high — it is the difference between “the primitive has a model” and “CSD's `Σ` satisfies the primitive”, which `LF3/SettingLocality.lean` itself flags as the separate open half | #131 obligation (4); ~~#135~~; [`LF3/SettingLocality.lean`](../CsdLean4/LF3/SettingLocality.lean); [`RecordLayer/SkewFibrePartition.lean`](../CsdLean4/RecordLayer/SkewFibrePartition.lean) | > | ~~138~~ | Foundations | Records→spacetime | *new (2026-10-11, out of #131 obligation (2))* **A stability bound that scales with the *coarse* boundaries** — the sharpening [`TwoWingStability.lean`](../CsdLean4/RecordLayer/TwoWingStability.lean) deliberately does not claim. ▸ **DONE 2026-10-15 (M).** [`RecordLayer/CoarseBoundaryStability.lean`](../CsdLean4/RecordLayer/CoarseBoundaryStability.lean), 20 pins. A **coarse boundary** of a decoding `u : Fin N → κ` is an index whose value differs from the *cyclically* next cell's (`IsCoarseBoundary`, `nextIdx`, `coarseBoundaryCount`). ★★★ `measure_coarseCode_ne_le`: the microstates whose coarse reading a write of size `δ ≤ 1` changes have epistemic measure at most **`#boundaries · δ`**; ★★★ `measure_exists_coarseCode_ne_le` uniformly over the horizon and ★★★ `measure_drift_coarseCode_ne_le` under a drift with `ω·T ≤ 1`. `coarseBoundaryCount_le` = at most `N`, so #103's `N · δ` is the worst case; ★★ `coarseBoundaryCount_comp_le` = **a decoding factoring as `v ∘ g` through a MONOTONE block map `g : Fin N → Fin m` has at most `m` boundaries** (the cyclic-successor image is injective on boundaries), hence ★★★ `measure_wingCode_ne_le_of_blocks` / `measure_drift_wingCode_ne_le_of_blocks`: **a contiguous two-wing decoding into the experiment's `m = 4` joint outcomes is stable outside `4 · δ`, independent of the Hilbert dimension `N`**. The route: ★ `exists_mem_globalBasin` — **every microstate lies in a basin** (the rates sum to one, so the arcs tile the whole turn; `outcomeCode_ne_zero`, `fineIndex_ne_none` — strengthens `globalBasin_ae_total`); `rep_add_coe_of_one_lt` (the wrap-around companion of `rep_add_coe`) and `rep_sigmaShift` (a write of at most one turn moves the representative by `δ`, wrapping at most once); ★★ `coarseCode_ne_subset` — a coarse relabelling puts the microstate in the **cyclic band** `(T − δ, T] ∪ (T + 1 − δ, 1]` at some coarse boundary `T` (discrete intermediate value along the cyclic path from the microstate's cell to the written point's cell, `exists_boundary_between`, with the wrap handled by the boundary at the top of the turn); ★ `measure_cyclicBand_le` — the band weighs `min T δ + (δ − T)₊ = δ`. **Two of this row's premises corrected:** 'a write moves a microstate to the next cell or nowhere' holds only when the write is shorter than the next cell, so the covering lemma replaces it; and the contiguity hypothesis is stated as the factorisation through a monotone block map — `LF6.pointerDecode` under an arbitrary `e` is not shown contiguous, and nothing here chooses `e`. Writes longer than one turn are excluded by hypothesis (`δ ≤ 1`), not covered. ▸ *Original row:* Its bound is `N·ω·T` in the **fine** cell count, inherited from #103; the two-wing record has only `m·n` values, and intuition says it should be stabler in proportion. ▸ **What it would take.** `rep_add_coe` says a write pushes a microstate towards the upper endpoint of its own arc, so the only relabelling it can cause is into the **next** cell; the coarse reading survives that whenever the decoding gives cell `i` and cell `i+1` the same value. So the coarse exceptional set is contained in the union of `edgeBand`s over the indices `i` with `w i ≠ w (i+1)`, giving a bound of `#{i \| w i ≠ w (i+1)} · ω·T`. Two pieces are missing: the “moves to the next cell” fact is not stated in `MacrostateStability.lean` (its margin lemma only says *outside* the band nothing changes), and the wrap-around case at `i = N-1` needs care. ⚠️ **And the sharpening is only worth anything for a decoding that groups adjacent cells.** `LF6.pointerDecode` groups by the pointer half of an arbitrary `e : Fin 4 × Fin 4 ≃ Fin (M+1)`, under which the four joint outcomes need not occupy contiguous arcs, so in the worst case the count is `N` again and nothing is gained. A contiguity hypothesis on `e` (or a choice of `e` that is contiguous by construction) is part of the row, not an afterthought. First deliverable: the “write moves a microstate to the next cell or nowhere” lemma in `MacrostateStability.lean`. | ~~**M**~~ done (M) | Low–medium — nothing consumes the sharper constant, but it is the difference between a bound that mentions the experiment's outcomes and one that mentions its Hilbert dimension | #131 obligation (2); #103; [`RecordLayer/MacrostateStability.lean`](../CsdLean4/RecordLayer/MacrostateStability.lean); [`RecordLayer/CircleFibre.lean`](../CsdLean4/RecordLayer/CircleFibre.lean) (`rep_add_coe`) | > | ~~139~~ | Foundations | Records→spacetime | *new (2026-10-11, out of #131 and ~~#133~~'s first half)* **One arena, two dynamics: relate the CV interacting unitary to `LF6`'s de-isolation flow** — the last thing between #131 and its stated deliverable, and it only became visible once the arena identification removed the space mismatch that was hiding it. #131's five properties now all live on `LF4.KSigma 16`, but **under two different maps**: causal separation is about `graphInteractingU 4 2 τ λ twoWingGraph gc` (the CV mode-coupling interaction, whose cone gives `CV.spacelike_twoWing`), while the Born weights, the record history and the CHSH value are about `LF6.localDeisolationFlow` (the Naimark dilation's flow). Nothing relates them, so the experiment's causal structure comes from one dynamics and its statistics from another — which is precisely *not* “within the same dynamical model”. ▸ **Three routes, in increasing strength.** (a) **Compatibility**: show the two commute, or that the de-isolation flow is `SupportedOn` the union of the wing mode sets, so the cone bound survives the measurement — cheapest, and probably what the row needs. (b) **Containment**: exhibit `localDeisolationFlow` as a `graphInteractingU` for some coupling `g` on `twoWingGraph`, which would make them one dynamics outright — note `localDeisolationFlow` is a *product* `V_A ⊗ V_B` (`localDeisolation_factorises`), and the two-wing graph has no A–B edge, so this is not obviously false. (c) **Generation**: both as flows of one Hamiltonian, which is the `H_int` frontier and is not cheap. ⚠️ Route (b) needs care about *which* unitary: `graphInteractingU` is built from phase-diagonal couplings, and a Naimark dilation isometry is not diagonal, so the honest first check is whether the two even act on the same factor of the configuration space. ~~First deliverable: decide between (a) and (b) by checking that, in one paragraph.~~ ▸ **DECIDED 2026-10-11, AND THE ANSWER IS NONE OF (a)–(c).** Asked which route CSD itself prefers, the answer is **(c) in principle** — `CSD-CHARTER.md` says “there is **one flow** and one microstate”, and the corpus gives the *same* discharge condition for Posit 1 and Posit 10: derive it from the de-isolation dynamics. But that is the `H_int` frontier, repeatedly priced as having no cheap brick. **The reachable form of the same idea is to drop the second interaction: (d) read the separation off the measurement flow itself.** ▸ **LANDED 2026-10-11** ([`CV/TwoWingFactorisation.lean`](../CsdLean4/CV/TwoWingFactorisation.lean), 7 pins; S of the S–M). `LF6.localFlowUnitary` is `reindex … (vnUnitary 2 ⊗ₖ vnUnitary 2)` and `localDeisolation_factorises` says the dilation is `V_A ⊗ V_B` — **the measurement interaction is already a product of per-wing unitaries**, and a product unitary has disjoint per-wing support. ★★★ `supportedOn_heisenberg_pow_of_factorised`: the evolved A-observable stays A-local at **every power** of a factorised flow — *there is no ball to grow*, because the B-factor acts trivially on an A-supported observable and the A-factor cannot leave A's modes. Hence ★★★ `arenaObs_kick_of_factorised` and ★★★ `recordStroke_comm_kick_of_factorised`: a B-wing intervention cannot steer A's reading or A's record write, after any number of applications of the measurement interaction, exactly — **no graph, no period count**. ★★ `leftModes_two_two` / `rightModes_two_two` (`by decide`) confirm `CompositeArena`'s mode blocks at `K₁ = K₂ = 2` **are** `TwoWingCone`'s wing sets, so nothing was re-indexed between the two routes. ▸ **What it changes for #131**: adjacency is no longer *supplied* through `E` — the assumption moves from “the graph has no A–B edge” to “the apparatus couples each wing to its own pointer”, which is about the experiment rather than an auxiliary graph. ⚠️ **Still an assumption**: `localDeisolation_factorises` holds *by construction*, LF6-A.3 having chosen a product dilation, so locality is put in somewhere more physical rather than derived — deriving it is `R-016`. ⚠️ **Weaker than a cone, deliberately**: a product unitary gives *static* locality, not a propagating bound; that is the right statement for one measurement, and `TwoWingCone` keeps the graded version for a dynamics that *is* coupled — neither supersedes the other. ⚠️ **AND THE ROW IS HALF DONE.** The theorems are stated for *any* wing-factorised flow and instantiated at `leftOp`/`rightOp`, **not at `localFlowUnitary`**. The index alignment was *checked* — `jointDilEquiv` regroups `((sᴀ,pᴀ),(s_b,p_b)) ↦ ((sᴀ,s_b),(pᴀ,p_b))`, so the dilated space is four `Fin 2` factors grouped per wing and `FieldConfig 4 2` is four `Fin 2` factors indexed by mode, with `modes 0,1 = (sysA, ptrA)` and `2,3 = (sysB, ptrB)`, which is exactly why `wingAModes = {0,1}` was right — but **carrying `localFlowUnitary` across it is bookkeeping this landing does not do**, so “the de-isolation flow satisfies the hypothesis” is a checked claim about index conventions, not a theorem. That instantiation is what would actually discharge #131's two-dynamics gap. ▸ **AND THE INSTANTIATION LANDED 2026-10-12** ([`CV/DeisolationFactorised.lean`](../CsdLean4/CV/DeisolationFactorised.lean), 10 pins; S). ★★★ `reindex_localFlowUnitary_eq_joinU`: **the de-isolation flow, read on the four-mode configuration space, IS `CompositeArena`'s product unitary** — the `jointFlowEquiv` cancels against its own inverse and what is left is the per-wing Kronecker product read through `configSplit`, which is exactly what `compositeReindex` does, so the measurement interaction and the composite arena's product dynamics are **the same matrix**. The alignment is three named steps (`wingIndexEquiv`, `wingSplitEquiv`, `configFlowEquiv`), so the hand-check of the previous landing is now a theorem. With ★★★ `joinU_wingU_eq_mul` it meets `TwoWingFactorisation`'s hypothesis, giving ★★★ `arenaObs_kick_deisolation` and ★★★ `recordStroke_comm_kick_deisolation`: **a B-wing intervention cannot steer A's reading or A's record write, at every power of the de-isolation flow**, with no graph and no period count. ⚠️ **One thread remains and it is about the projective action, not about which dynamics**: these are statements about `arenaKick` of the pulled-back unitary on `FieldArena 4 2`, while `localDeisolationFlow` is `localFlowUnitary • ·` on `CPN (4 * 4)`; they agree up to `CV.arenaEquiv`, and the identity `arenaKick (pullback) = arenaEquiv.symm ∘ localDeisolationFlow ∘ arenaEquiv` is not proved — bookkeeping with no choice left in it, recorded on #131 as the last caveat on the conjunction rather than given a row. | ~~**M–L** (and the check first at **S**)~~ the route and the theorem took **S**, the instantiation **S** | High — it is the difference between #131 satisfied and #131 nearly satisfied | #131; ~~#133~~; [`CV/ArenaIndexBridge.lean`](../CsdLean4/CV/ArenaIndexBridge.lean); [`CV/TwoWingCone.lean`](../CsdLean4/CV/TwoWingCone.lean); [`LF6/TwoWingBell.lean`](../CsdLean4/LF6/TwoWingBell.lean) | > | ~~122~~ | QIT | One QM, many formulations | *new (2026-10-07, out of #92’s split)* **The Weyl operator as a continuous map of Schwartz space: `Op(a) : 𝓢(ℝ, ℂ) →L 𝓢(ℝ, ℂ)`** — #92(a)’s actual target, which #92(a1) deliberately stopped short of. **BOTH HALVES DONE** ([`Mathlib/Analysis/Fourier/WeylSmooth.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WeylSmooth.lean), Category 1, 13 pins). ▸ **THE SMOOTHNESS HALF, 2026-10-07.** ★★★ `contDiff_weylOpK`: **the output is `C^∞`**, the second seminorm ingredient after #92(a1)’s continuity. The piece that made it reachable is ★★ `iteratedFDeriv_comp_affine`, **the iterated chain rule along an affine line** — `x` enters `K` through `((x+y)/2, x−y) = (y/2, −y) + x·(1/2, 1)`, an affine path with constant velocity, and Mathlib has the two halves (`ContinuousLinearMap.iteratedFDeriv_comp_right` for the linear part, `iteratedFDeriv_comp_add_left` for the translation) but not the composite — with ★★ `norm_iteratedFDeriv_comp_affine_le` its bound, ★ `exists_bound_snd_iteratedFDeriv` #92(a1)’s second-slot decay at every order, and ★ `inv_one_add_sq_le_of_abs_sub_le` the ball-uniform majorant. **No product rule is needed anywhere**, because `ψ` does not depend on `x` — only the chain rule does the work, which is why this half was reachable while #121(ii), which has the same kernel but an exponential that *does* depend on both variables, is not. ▸ **THE DECAY HALF, 2026-10-08 — and the obstruction this row named was the whole of it.** ★★ `iteratedDeriv_weylOpK` is ~~#125~~’s formula on this integrand, so a weight has something to move onto; ★ `abs_le_norm_weylPath` is the one inequality the half turns on — `|x| ≤ (3/2)·‖(y/2, −y) + x·(1/2, 1)‖`, so a polynomial weight in the **parameter** becomes a polynomial weight on the **kernel**, where the kernel’s own decay can eat it; ★★ `exists_bound_snd_iteratedFDeriv_weighted` does the eating, combining `K`’s decay at orders `N` and `N + 2` into a bound that carries `‖p‖ᴺ` *and* still decays in the second slot (the unweighted lemma is its `N = 0` case), which is what keeps the majorant integrable; ★★★ `exists_bound_weylOpK` integrates it — `‖x‖ᴺ·‖∂ᵏ(Op(K)ψ)(x)‖ ≤ (3/2)ᴺ·C·π·‖ψ‖₀,₀`, uniformly in `x`; and ★★★ `weylCLM` is the operator, with ★ `weylCLM_apply_eq_weylOp` **closing this row’s own gating caveat** — through #121(i)’s slice family the continuous linear map *is* `weylOp` of a slice family, so the statement holds in the vocabulary #92 stated it in and not only for `weylOpK`. **THE FINDING IS WHICH SEMINORM APPEARS.** `SchwartzMap.mkCLM` allows a finite *set* of seminorms on the right; this needs exactly one, the sup norm `‖·‖₀,₀`, because `ψ` enters the integrand undifferentiated and unweighted — the entire `x`-dependence is the kernel’s, so the weight lands wholly on `K` and nothing is ever asked of `ψ`’s own decay. The convergence is the kernel’s decay in `x − y`, not the state’s. **And the shape of the file changed the way ~~#125~~’s did**: the `have`s buried in the smoothness proof became the standalone ★★ `partialDeriv_weylIntegrand` and ★★ `norm_partialDeriv_weylIntegrand_le` (the identity and the primitive bound *both* halves run on), with `aestronglyMeasurable_partialDeriv_weylIntegrand` and `exists_weylIntegrand_ball_bound` beside them, and `contDiff_weylOpK` now **consumes** the shared ball bound instead of rebuilding it — the hypotheses exist once and both halves read them. ⚠️ **One kernel class.** This is the operator of a *jointly Schwartz* kernel; Weyl quantisation of a wider symbol class (polynomially bounded symbols, Hörmander classes) is a different theorem and is not proved. ⚠️ **The constant is not sharp** — `(3/2)ᴺ·C·π` falls out of this route, not out of an optimisation. ⚠️ **Still no pseudodifferential *calculus***: that `Op(K₁) ∘ Op(K₂)` is again a Weyl operator is **#126**, and the symbol-level star product is #64 behind #121(ii). | ~~**L**~~ done — smoothness **M** (2026-10-07), decay **S–M** (2026-10-08) once ~~#125~~ supplied the formula; the row spun out ~~#124~~ and ~~#125~~ on the way, which is where its `L` actually went | High for #92 — it is what "`Op(a)` is an operator on Schwartz space" means, and it is that row’s last piece | #92; #120; #121; #125; #126; #63 | > | ~~126~~ | QIT | One QM, many formulations | *new (2026-10-08, out of #122)* **Composition: `Op(K₁) ∘ Op(K₂)` is a Weyl operator. DONE 2026-10-08, the same day it was opened, in four slices.** ★★★ `weylCLM_comp` ([`Mathlib/Analysis/Fourier/WeylComposition.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WeylComposition.lean)): the composition of two Weyl operators of jointly Schwartz kernels **is** the Weyl operator of a single jointly Schwartz kernel, as continuous linear maps of `𝓢(ℝ, ℂ)`. It is the first thing a *calculus* needs, and the first statement in which the **operators** compose rather than the integrals. ▸ **THE GATE (S–M)** — the formula over a finite-dimensional parameter, which #125's scope note named as missing ([`ContDiffParametricIntegralFDeriv.lean`](../CsdLean4/Mathlib/Analysis/Calculus/ContDiffParametricIntegralFDeriv.lean), +7 pins): a composite kernel is a parametric integral over a **two**-dimensional parameter, so this row hit #122's wall one dimension up. ★★★ `hasFDerivAt_integral_of_bound` is the first-order identity over a parameter in any normed space, exported from inside #124's induction with the weakest hypotheses that give it, ★★ `fderiv_integral_apply_of_bound` its directional form, and ★★★ `norm_iteratedFDeriv_integral_le` **what a consumer actually needs — a bound, not an identity**: `‖iteratedFDeriv ℝ n (∫ a, F · a) x‖ ≤ ∫ a, bound n a` on `U`, since stating the identity at order `n` would need the iterated derivative of the integral as a multilinear map, the plumbing #124 exists to avoid, whereas a Schwartz estimate wants a scalar bound. The induction peels the **last** direction (★ `iteratedFDeriv_apply_succ_last`) and substitutes the first-order identity under the remaining derivatives, legitimate because it holds on all of the open `U` (★ `Filter.EventuallyEq.iteratedFDeriv_eq`, which Mathlib has for `iteratedFDerivWithin` only); **and #124's induction now calls the exported identity instead of rebuilding it**, the change ~~#125~~ made to #120. ▸ **(ii) INTEGRATING OUT ONE VARIABLE (M)** ([`SchwartzPartialIntegral.lean`](../CsdLean4/Mathlib/Analysis/Fourier/SchwartzPartialIntegral.lean), 10 pins): ★★★ `SchwartzMap.integralLastCLM : 𝓢(E × ℝ, F) →L[ℂ] 𝓢(E, F)`. Mathlib has `integralCLM`, which integrates over the *whole* domain into a scalar, and nothing that integrates out one variable. Smoothness is #124, decay is the gate, and the bridge from Schwartz data to #124's directional hypotheses is this file: ★★ `iteratedFDeriv_comp_inl` the chain rule through `y ↦ (y, t)`; ★ `contDiff_prodDirDeriv` with ★★ `dirDeriv_eq_prodDirDeriv` **discharging the measurability hypotheses nothing had discharged before** — #124 flags one as unavoidably operator-valued, and folding the inclusion into the recursion makes every directional derivative a *slice of something jointly smooth*, with the operator-valued case needing only that a normed target is metrizable because the integration domain is second countable; and ★★ `one_add_norm_pow_mul_norm_iteratedFDeriv_le` **the weight transfer**, `(1 + ‖p‖)ᴺ·‖∂ᵏG(p, t)‖ ≤ 2^(N+2)·S·(1 + t²)⁻¹` with `S` a *single* `Finset.sup` of `G`'s seminorms — #122's two-order combination with the slots swapped — whence ★★★ `norm_pow_mul_norm_iteratedFDeriv_integral_le`, where the weight rides a bound family **constant on a unit ball** and `∫ (1 + t²)⁻¹ = π` closes it. ▸ **(i) THE TENSOR PRODUCT (S–M)** ([`SchwartzTensor.lean`](../CsdLean4/Mathlib/Analysis/Fourier/SchwartzTensor.lean), 3 pins): ★★ `SchwartzMap.tensorProd` — `(p, q) ↦ f p · g q` is Schwartz on `E × F`, by `norm_iteratedFDeriv_mul_le` for the Leibniz expansion, ★ `norm_iteratedFDeriv_comp_clm_le` to return each factor's derivative to its own variable (the projections are short), and ★ `norm_le_one_add_mul_one_add` — **the weight splits multiplicatively** because the product norm is a max, so the *same* bound serves every term of the Leibniz sum. ▸ **(iii)(iv) THE COMPOSITION ITSELF (S + S–M).** **The reframing is what made it short:** a Weyl operator *is* an integral operator with a Schwartz kernel, because `x` and `y` enter `K` through the linear isomorphism `(x, y) ↦ ((x+y)/2, x − y)` — ★ `weylKernelCLM` (composition with it, a CLM by Mathlib's `compCLMOfContinuousLinearEquiv`) turns symbol-kernel into integral kernel and back, and ★ `weylOpK_eq_integral_kernel` is the operator in that form. ★ `compAffineCLM` composes with an **injective affine** map on Schwartz space, with ★ `antilipschitzWith_affine_of_leftInverse` supplying the antilipschitz constant from an explicit linear **left inverse** — what a concrete coordinate map always has. Then ★★★ `weylCompKernel` is the composite kernel: the tensor product of the two integral kernels, composed with the injective `((x, z), w) ↦ ((x, w), (w, z))`, is Schwartz on `(ℝ × ℝ) × ℝ`; (ii) integrates `w` out; the change of variables carries it back. ★★ `weylCompKernel_apply` is the formula `∫ w, k₁(x, w)·k₂(w, z)` — **the classical kernel product** — and ★★★ `weylOpK_comp` the operator identity by `integral_integral_swap`, with `integrable_weylCompIntegrand` supplying Fubini's hypothesis through the same tensor construction at a fixed output point. **THE FINDING IS THAT THE REFRAMING, NOT THE ESTIMATES, WAS THE WORK.** The row priced the kernel-level composition at `L` on the expectation of a joint seminorm estimate on `L`; in integral-kernel coordinates there is no such estimate to do — the composite is a *tensor product composed with an injective linear map and one variable integrated out*, and each of those three steps is a general Schwartz lemma. What was genuinely missing was the parametric-integral formula one dimension up, which is why the gate is the biggest of the four slices. ⚠️ **The symbol-level statement is not here.** That the composite's *symbol* is the Moyal star product `a ♯ b` needs the joint partial Fourier transform of #121(ii), and the `ℏ²` expansion is #64. ⚠️ **One kernel class**, as in #122. ⚠️ **No algebra structure is claimed** — associativity, and that these operators form an algebra, follow from this identity with the injectivity of `weylKernelCLM`, and are ~~**#127**~~ (**DONE 2026-10-09** — and it needed a theorem #126 did not have, faithfulness of `Op`). ⚠️ `Wigner.lean`'s one-dimensional `antilipschitzWith_affine` is the `L = s·id` case of the general lemma and is left as it stands, because it sits below this file in the import order. | ~~**L**~~ done the same day, **gate S–M + (ii) M + (i) S–M + (iii)(iv) S–M**; the `L` was priced on an estimate that the reframing dissolved | High — the first brick of a Weyl calculus, and it closed a Mathlib gap (partial integration of Schwartz functions) on the way | #122; #121; #124; #125; #64; #127 | > | ~~127~~ | QIT | One QM, many formulations | *new (2026-10-08, out of ~~#126~~)* **The Weyl operators of Schwartz kernels form an algebra. DONE 2026-10-09** ([`Mathlib/Analysis/Fourier/WeylAlgebra.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WeylAlgebra.lean), Category 1, 15 pins; S–M as priced). ~~#126~~ proved `Op(K₁) ∘ Op(K₂) = Op(K₁ ⋆ K₂)` and claimed nothing about the structure; this is the structure, and **the row's own route was wrong, which is the finding.** It said associativity would follow from associativity of operator composition *and* the injectivity of `weylKernelCLM`. **That does not close.** `weylKernelCLM` is the *change of coordinates* `(x, y) ↦ ((x+y)/2, x − y)`, and its injectivity says nothing about whether two different kernels can give the same operator — which is exactly what turning an operator identity into a kernel identity needs. The ingredient is **faithfulness of `Op`**, and that is a theorem rather than bookkeeping: ★★★ `weylOpK_injective` (with ★★★ `eq_zero_of_weylOpK_eq_zero` behind it) proves **the kernel is determined by the operator**, by feeding the operator the **conjugate of its own kernel slice** — `SchwartzMap.conjCLM` of #121(i)'s `slice`, conjugation being `ℝ`-linear and not `ℂ`-linear, which is why it is a map over `ℝ` — so the pairing becomes `∫ ‖k(x, y)‖² dy`, which vanishes only if the slice vanishes almost everywhere, and a continuous function vanishing almost everywhere for Lebesgue measure vanishes. ★★ `weylCLM_injective` is the same for #122's packaged operators. The algebra laws are then corollaries: ★★★ `weylCompKernel_assoc` from `ContinuousLinearMap.comp_assoc`, and ★★ `weylCompKernel_add_left`/`_right` with `_smul_left`/`_right` for bilinearity — where **the asymmetry between the slots is real, not cosmetic**: the left slot rides linearity of the integral in the kernel (★ `weylOpK_add_kernel`, ★ `weylOpK_smul_kernel`), while the right slot is the *inner* operator and has to go through the operators' action on the state. ▸ **AND THE NON-UNITALITY IS NOW A THEOREM, NOT A CAVEAT.** ★★★ `not_weylOpK_eq_id` and ★★★ `weylCLM_ne_one`: **no Schwartz kernel represents the identity.** The witness is `shrinkBump`, a bump at the origin of outer radius `2/(n+1)` as a complex Schwartz function — it keeps the value `1` at the origin for every `n` while its pairing with the kernel's slice tends to `0` by dominated convergence (dominated by the slice's own norm, the bump being bounded by `1`). A unit would have to reproduce a state's value at a point from an integral against a **bounded** kernel, and that is what fails. ⚠️ **No `Mul` instance, deliberately.** Declaring `⋆` as the `Mul` of `𝓢(ℝ × ℝ, ℂ)` would commit the type globally to the Weyl product when the pointwise product is at least as natural a choice, and a `NonUnitalAlgebra` instance would then fix which one every downstream file means. The facts are theorems; a bundled `NonUnitalAlgHom` is one `letI` away for a consumer who wants it, and nothing in the corpus does. ⚠️ **Nothing about the symbol product**: faithfulness here is of `K ↦ Op(K)` on *kernels*, and that the *symbol* composes by the Moyal star product still needs #121(ii), with the `ℏ²` expansion at #64. ⚠️ **The non-unitality is for this class only** — a wider class (distributional kernels, where the identity's kernel is a delta) is a different setting and is not formalised. | ~~**S–M**~~ done (S–M), of which the unforeseen part was faithfulness: the row had priced it as bookkeeping and it is the file's main theorem | Medium — it tidies ~~#126~~ into a structure statement, and it turned the "no unit" caveat into a proof | #126; #122; #121; #64 | > | ~~125~~ | QIT | Library owed / upstream | **DONE 2026-10-08 — and it was the restatement the row said it was** ([`Mathlib/Analysis/Calculus/ContDiffParametricIntegral.lean`](../CsdLean4/Mathlib/Analysis/Calculus/ContDiffParametricIntegral.lean), +4 pins; M took S–M). ★★★ `hasDerivAt_integral_of_bound` states the first-order identity under #120’s own hypotheses, and ★★★ `iteratedDeriv_integral_of_bound` iterates it: **on an open `U` the `n`-th derivative of the integral is the integral of the `n`-th parameter derivative**, by the same `partialDeriv_succ_left` shift the smoothness proof runs on with the identity kept instead of discarded. ★★ `iteratedDeriv_integral_of_bound_le` is the every-order-up-to-`n` form a consumer wants and `integrable_partialDeriv` the integrability the formula is false without. **The step that makes the iteration legitimate is locality**: `Filter.EventuallyEq.iteratedDeriv_eq` lets a first-order identity that holds only on `U` be substituted under the remaining derivatives at a point of the open `U` — the one thing in the row that was not already written somewhere. **And the change that matters most goes the other way**: `contDiffOn_integral_of_bound` now *calls* the exported identity instead of re-proving it inline, so the proof exists once rather than twice and the `hFdiff` it used is gone. ⚠️ ~~Still no formula over a **finite-dimensional** parameter — #124 keeps its own scope note, and #121(ii) would want that analogue.~~ **DONE 2026-10-08 as #126's gate** (`hasFDerivAt_integral_of_bound` in #124's file, plus ★★★ `norm_iteratedFDeriv_integral_le`, the bound at every order): a composite Weyl kernel is a two-dimensional parametric integral, so #126 hit this row's wall one dimension up and closed it there. ⚠️ ~~**#122’s decay half is unblocked, not done**: the estimates still have to be written.~~ **#122 landed both halves 2026-10-08 on this formula**, and what the estimates needed from here was `iteratedDeriv_integral_of_bound_le` alone — the row’s claim that they were straightforward once the formula existed held. | ~~**M**~~ done (took **S–M**) | High — it was what two rows were waiting on | #120; #124; #122; #121 | > | ~~94~~ | QIT | QC chain (link 12) | **DONE 2026-10-04** ([`Mathlib/QuantumInfo/SyndromeExtraction.lean`](../CsdLean4/Mathlib/QuantumInfo/SyndromeExtraction.lean) Cat-1 + [`Empirical/QM/QEC/SteaneSyndromeCircuit.lean`](../CsdLean4/Empirical/QM/QEC/SteaneSyndromeCircuit.lean) Cat-3, 21 pins; M–L took M) — **and the whole gadget turned out to be one observation.** A `CNOT` ladder from data to ancilla is a **permutation of basis labels**, `(z, a) ↦ (z, a + H z)`, so it is `permMat` of `TransversalClifford.lean` exactly as the row predicted, and the only arithmetic needed is `H z + H z = 0` — characteristic two, nothing else. **THE GENERIC GADGET.** `extractMat H` with ★ `extractMat_mem_unitaryGroup` (self-inverse unitary); `ancInit` the ancilla preparation (an isometry); `ancProj s` the ancilla readout and `synProj H s` the data-side syndrome projector, both Hermitian idempotents summing to `1`. ★★ `extractMat_mul_ancInit_apply` is the content: **the circuit takes the basis state `z` with a fresh ancilla to `(z, H z)`**. ★★★ `ancProj_mul_extractMat_mul_ancInit`: **reading the ancilla as `s` IS projecting the data onto the syndrome-`s` subspace** — the circuit implements the projective measurement the corpus had been assuming (`SyndromeRecovery.lean`), as an operator identity — with ★ `ancProj_mul_extractMat_mul_ancInit_of_ne` making the measurement **exhaustive** (a syndrome the data does not carry gets zero amplitude). ★★★ `extractMat_mul_ancInit_mul_synProj_zero` and ★★★ `extractMat_conj_codeState`: **on the code, extraction does nothing** — ancilla included, and as an operator identity rather than a claim about one state — with ★★ `ancProj_zero_mul_extractMat_conj_codeState` making the all-zero outcome **certain**. That is the row's "returns the syndrome and leaves a code state alone", both halves. **THE STEANE INSTANCE.** `steaneExtract` at `Steane.lean`'s Hamming check map, with ★★ `steaneExtract_tags_syndrome`, ★★★ `steaneExtract_reads_syndrome`, ★★★ `steaneExtract_code_state`, and the **distance-3 mechanism read off the ancilla**: ★★ `steane_extract_detects_single` (a single-qubit error changes what the ancilla reads, from `steane_syndrome_single_ne_zero`) and ★★ `steane_extract_names_single` (distinct single errors read differently, so **the ancilla names the faulty qubit**, from `steane_syndrome_single_injective`). **FIVE THINGS NOT CLAIMED, in both headers:** (a) **no faults, hence no fault tolerance** — this is the fault-free gadget and nothing says it is fault-*tolerant*; a fault in the ladder or the ancilla is **#95** and **#96**; (b) the ancilla is **one unverified block** — no cat state, no verification, so no bound on an ancilla fault spreading into the data, which is exactly #95; (c) **one check map applied once** — no rounds, no interleaving with gates, which is #96's extended rectangle; (d) **`X`-checks only** — the `Z`-half is the same statement in the conjugate basis and is not derived, and nothing combines the two bases for the CSS code; (e) **no decoder** — the syndrome is produced, never interpreted, and `SteaneRecovery.lean`/`SteaneArbitrary.lean` are neither re-derived nor connected to this circuit. *As opened:* **Fault-tolerant syndrome extraction, as a circuit.** The corpus measures the syndrome as a projective measurement ([`SyndromeRecovery.lean`](../CsdLean4/Empirical/QM/QEC/SyndromeRecovery.lean)); the circuit-level theorem needs it as a circuit: an ancilla block, the transversal `CNOT`s as the label permutation `permMat` that #62(f) already built, the syndrome read as a projector, and the no-fault theorem that extraction returns the syndrome and leaves a code state alone. | **M–L** | Medium — the first piece of a *faulty* recovery; rows 95 and 96 depend on it | #62; #86; [`SteaneTransversalCNOT.lean`](../CsdLean4/Empirical/QM/QEC/SteaneTransversalCNOT.lean) | > | ~~95~~ | QIT | QC chain (link 12) | **DONE 2026-10-04** ([`Mathlib/QuantumInfo/AncillaLadder.lean`](../CsdLean4/Mathlib/QuantumInfo/AncillaLadder.lean), Cat-1, 13 pins; L took M–L) — **and it priced the caveat #94 left in its own header.** `cnotGate_conj_pauliOp` propagates `X` from control to target and `Z` from **target to control**, so for a data→ancilla ladder the dangerous fault is a `Z` on the **ancilla**, and the two wirings differ exactly there. ★★★ `laddGate_conj_pauliOp` telescopes `Clifford.lean`'s one-gate theorem along the whole ladder — the `CNOT`s commute (same target, controls off it), so the ladder is an involution (★ `cnotGate_laddGate_comm`, ★ `laddGate_involutive`) and the induction closes — giving explicit `X` and `Z` label maps with **no phase**. **THE SHARED-TARGET COST.** ★★★ `zLadd_bitAt_target_apply`: a single `Z` fault on the shared ancilla comes out as a `Z` on the ancilla **and on every control in the ladder**; ★★★ `dataSupport_zLadd_bitAt_target` makes the data-side support **exactly the control list** and ★★ `card_dataSupport_zLadd_bitAt_target` gives its size as `L.length`: **one fault, `w` data errors**, which for a weight-4 check is past the distance-3 guarantee. **THE CAT-STATE COST.** ★★ `zLadd_singleton_apply` / ★★ `card_dataSupport_zLadd_singleton`: with one control per ancilla qubit the same fault reaches **one** data qubit. That contrast is the entire justification for the cat state, and it is now a theorem rather than folklore. **THE VERIFICATION HALF.** ★ `catVerify_eq_zero_iff` (verification passes exactly on the constant strings) with ★★ `not_isCat_add_bitAt` and ★★ `catVerify_ne_zero_of_single_flip`: **a single preparation bit-flip is flagged**, on any register of at least two qubits. Together that is the row's "one fault inside the preparation produces at most one data error, or else is flagged", both branches. **SIX THINGS NOT CLAIMED, in the header:** (a) **no fault model** — "one fault" means one Pauli at one place and every conclusion is a **support** statement; no fault counting, no probabilities, no threshold (that is **#96** and **#97**); (b) **the cat state is not prepared** — `IsCat` describes the labels it is supported on and `catVerify` the parity check, but no preparation circuit, no superposition, and **not** the `X`-basis measurement whose parity gives the check value, so this is the propagation-and-flagging content and not a verified-ancilla gadget; (c) **`Z` faults on the ancilla only** — `X` faults there propagate *away* from the data and other locations are not treated; (d) **weight is not correctability** — that a weight-4 error is *not corrected* by a distance-3 code is **not** proved, only that the guarantee does not apply; (e) **one register, one ladder, one check** — no rounds, no interleaving; (f) **no Steane instance** — the theorems hold for an arbitrary control list, so a Hamming row's weight-4 support is one of their instances, but the embedding of seven data qubits plus an ancilla block into a single register is not written down and no Steane-specific number is proved. *As opened:* **The verified ancilla.** The cat state, its verification measurement, and the propagation statement: **one fault inside the preparation produces at most one data error, or else is flagged**. This is the mathematical content of *fault-tolerant measurement* and the piece row 62 never had — without it a single extraction fault becomes two errors in the block and the code's distance is spent. | **L** | High within the chain: until it lands, every circuit theorem keeps an **ideal** recovery gadget | #62; row 94 | > | ~~96~~ | QIT | QC chain (link 12) | **DONE 2026-10-04** ([`Mathlib/QuantumInfo/ExtendedRectangle.lean`](../CsdLean4/Mathlib/QuantumInfo/ExtendedRectangle.lean), Cat-1, 8 pins; M–L took M) — the Aharonov–Ben-Or / Aliferis–Gottesman–Preskill rectangle argument stripped to its algebra, with a correctable set `D` as a parameter and **three properties each about a different piece**: `IsRecovery` (the recovery removes every `D`-deviation from a code state), `DeviatesBy` (the one-fault bound on one piece), `PropagatesD` (**transversality** — the gadget carries a `D`-deviation to a `D`-deviation, which is what stops a leading fault from being amplified). Hypotheses about the **pieces**, conclusions about the **composite**: nothing assumes the rectangle is correct in order to prove it. ★★★ `isCorrectedStep_exRec_leading` is **the row's instance for a FAULTY RECOVERY** — the leading recovery is the faulty piece and the rectangle is *still* a corrected step, because transversality carries its error through the gadget and the trailing recovery removes it. ★★★ `isCorrectedStep_exRec_gadget` is the same when the **gadget** is faulty, ★★★ `isCorrectedStep_exRec_of_good` packages the two as the **goodness predicate** (the single fault is anywhere but the trailing recovery), and ★★★ `isCorrectedStep_rect` is the two-piece form. ★★★ `deviatesBy_exRec_trailing` is the honest statement for the remaining case: a **trailing-recovery fault is NOT corrected here**, it is handed on as a `D`-deviation — exactly the form the next rectangle's `IsRecovery` consumes — which is the **overlapping-rectangle convention stated rather than assumed away**. ★★★ `correctedRun_eq_idealRun_of_good_exRec` is the payoff **with no restatement**: #62 (d1)'s theorem applied verbatim, which is what its abstraction in the step was for, plus ★★ `correctedRun_replicate_exRec` for a run of copies. **SIX THINGS NOT CLAIMED, in the header:** (a) **"at most one fault" is a hypothesis, not a count** — no fault locations are enumerated and no probability is assigned, which is **#97**; (b) **`D` is never computed** — no code is fixed, and in particular **the link to #95 is not made**: that row's weight-`w` propagation bound would say a shared-ancilla fault lands outside a distance-3 `D`, but connecting them needs a concrete `D` and recovery; (c) a **trailing-recovery fault is not corrected**, and that the chain of overlapping rectangles **closes** is #97; (d) **transversality is assumed** — `PropagatesD` is a hypothesis, and the corpus's concrete Steane transversal `CNOT` is not plugged in; (e) **no instance at a concrete code** — the theorems *produce* `IsCorrectedStep` instances from the piecewise hypotheses, they do not exhibit one for Steane; (f) **maps, not channels** — no positivity or trace condition, as in #62 (d1). *As opened:* **The extended rectangle.** Recovery, gadget, recovery, with its goodness predicate, delivered as an `IsCorrectedStep` instance for a **faulty** recovery — which [`correctedRun_eq_idealRun`](../CsdLean4/Mathlib/QuantumInfo/FaultTolerantComposition.lean) then composes with no restatement, because that theorem was deliberately abstract in the step. | **M–L** | High — this is what turns #62(d1) from a theorem about faulty *gates* into one about a faulty *computer* | #62; rows 94, 95 | > | ~~97~~ | QIT | QC chain (link 12) | **DONE 2026-10-04** ([`Mathlib/QuantumInfo/LevelReduction.lean`](../CsdLean4/Mathlib/QuantumInfo/LevelReduction.lean), Cat-1, 8 pins; L–XL took M–L) — **and with it #62's remainder (94–97) is complete.** The three things #94–96 deliberately left out are here: the recursion, the fault count, and the join. **LEVEL REDUCTION.** ★★★ `isCorrectedStep_correctedRun`: **a corrected circuit IS a corrected step one level up** — and the `preserves` half comes *free* from `isCodeState_idealRun`, so the only hypothesis is that the circuit's ideal run is the gadget the next level wants. `IsCorrectedAtLevel` is the recursive simulation (level `0` a corrected step; level `k+1` the corrected run of a circuit of level-`k` gadgets) and ★★★ `isCorrectedStep_of_isCorrectedAtLevel`: **correctness at every level follows from correctness at level 0**, by induction on the tower — with ★★★ `correctedRun_eq_idealRun_of_level` as its circuit form. **THE JOIN.** ★★★ `one_sub_le_measure_output_eq` needs **no measurability hypotheses at all** — subadditivity gives `1 ≤ μ s + μ sᶜ` for *arbitrary* sets, so nothing has to be shown measurable — and ★★★ `one_sub_le_measure_output_eq_concat` is **the row's statement verbatim**: for `N` gadgets, each a level-`k` concatenated block under independent noise of rate at most `p`, the output is **exactly** the ideal one with probability at least `1 − N·(C(n,2)·p)^{2^k}/C(n,2)` — the row's `1 − N(cp)^{2^k}/c` with `c = C(n,2)`. ★★ `tendsto_one_sub_codeCapacityBound` sends that to `1` below threshold. The deterministic half is #62 (d1)'s `correctedRun_eq_idealRun` and the arithmetic half is #51's `codeCapacityBound` with `concatMeasure_concatBad_le`; **neither is re-proved** — what is new is the lifting, the union bound, and the step from "the good event is likely" to "the output is exactly ideal". Supporting: ★ `measurableSet_of_concatPat` (the pattern spaces are finite with measurable singletons, so every pattern set is measurable) and ★ `circuitMeasure_coord` (the one-coordinate marginal of the product). **SIX THINGS NOT CLAIMED, in the header, and the first is the one that matters:** (a) **the deterministic premise is a NAMED HYPOTHESIS and is not discharged here** — `hgood` (a pattern with no bad gadget gives the ideal output) is what #96's good rectangles plus the level recursion supply *for a concrete gadget set*, and **no concrete gadget set is plugged in**, so nothing here is a threshold theorem *for the Steane code*; (b) **"bad" is a pattern predicate**, and the correspondence between a gadget's faults and a `ConcatPat` is not constructed; (c) **noise is independent** across gadgets and blocks — a modelling choice visible in `circuitMeasure`, with correlated and adversarial noise outside everything; (d) **no gate count and no overhead** — `N` is given, nothing bounds the circuit size needed to simulate a computation, and **no polylogarithmic-overhead claim is made or implied**; (e) the constant `C(n,2)` is **not claimed sharp**; (f) **maps, not channels** — no positivity or trace preservation, as in #62 (d1) and #96. ⚠️ So the honest status of the chain's threshold claim after this row: the **statement** is a theorem in the abstract form above, with its gadget-set premise named — **not** a threshold theorem for a concrete code, which would need that premise discharged for one. *As opened:* **Level reduction and the probabilistic join — the threshold theorem's real weight.** The recursive simulation in which a level-`k` gadget *is* a level-`(k−1)` circuit, on top of #86's concatenated recovery for good patterns, and then the join with #62(a)'s union bound into one statement: **with probability at least `1 − N(cp)^{2^k}/c` the circuit's output is exactly the ideal one**. #62(d1) is the deterministic half; this is the recursion and the probability. | **L–XL** | The row's real weight; after it the chain's threshold claim is a theorem rather than an accounting | #62; #86; #51 | > | ~~98~~ | Foundations | Records→spacetime | **DONE 2026-10-04** ([`CV/TwoCones.lean`](../CsdLean4/CV/TwoCones.lean), 14 pins; M–L took M) — **and three findings about the starting point, read off the corpus rather than assumed.** (i) **There was no cone object anywhere**: `record_lightcone` and `FieldStructuredFlow.lightcone` are *theorem* names, and `DispersionEarned`'s "cone" is the pair of light **rays** `E = ±p` entering only as the hypotheses `a + b = c + d` and `a − b = d − c`, at slope 1 with no velocity parameter — so there was nothing with a slope for the record cone to sit inside. `forwardCone V v` is that object. (ii) **There was no map from sites to space**, and `Fin K` is read as a *momentum* label in `Dispersion.lean` versus a *graph site* label in `SupportSpreading.lean`; `SiteEmbedding` **posits** the arrangement with a bound `hop` on an edge's length, and the module takes the site reading explicitly. (iii) **The factorial bound this needed was already in Mathlib** — `Stirling.le_factorial_stirling`, for all `n` with no largeness hypothesis — which is why the row's price held at M. **THE CONTAINMENT.** ★ `exists_dist_le_of_mem_graphBall` by induction on hops, hence ★★ `mem_forwardCone_of_mem_graphBall`: **with a time step `τ`, the `n`-step record cone embeds in the forward cone of slope `hop/τ`** — the statement this row asked for. Its contrapositive ★★ `disjoint_graphBall_of_notMem_forwardCone` is what makes the cone load-bearing: it discharges `record_lightcone`'s `Disjoint` hypothesis **geometrically**. **THE VELOCITY.** ★ `pow_div_factorial_le` turns Stirling into `x^d/d! ≤ (e·x/d)^d`, so ★★ `lr_bound_le_half_pow`: once `d ≥ lrVelocity·t` the Lieb–Robinson error is at most `2^{-d}`, with `lrVelocity F = 4e‖S‖` **as** the slope in hops per unit time. ★★★ `record_influence_le_of_separated` is the capstone: **a record written in `R` cannot be steered from a region whose sites are farther than `d·hop` from `R`**, to within `L_h L_g · 2 · 2^{-d} · ‖A‖`, whenever `d ≥ lrVelocity·t` — Euclidean separation gives geometric suppression. **THE KINEMATIC LINK, WHICH THE ROW DID NOT ASK FOR AND WHICH IS WORTH MORE THAN THE CONTAINMENT.** ★★ `forwardCone_eq_nonnegSpan` makes the cone the non-negative span of its two null rays, so ★★ `mapsTo_forwardCone_of_rays`: fixing both rays preserves the cone. ★★★ `forwardCone_rays_is_boost` rescales `x ↦ x/v` to reach `cone_preserving_is_boost`, so **a unimodular ray-preserving map of the slope-`v` cone IS a boost** — the two cones' symmetry groups are the same group, at theorem level rather than in prose. And ★★★ `slope_cone_symmetry_characterises_omega`: **the slope is IMMATERIAL to P4** — for *any* `v > 0`, covariance under the slope-`v` cone's unimodular symmetries plus rest energy `m` forces `ω = √(p² + m²)`, so the Lieb–Robinson velocity may be taken as the kinematic cone's slope without changing the dispersion `cone_symmetry_characterises_omega` selects. **FIVE THINGS NOT CLAIMED, in the header:** (a) **containment, never identification** — the cones are not identified and `lrVelocity` is not claimed to be a limiting speed; an identification needs a continuum limit of the graph dynamics the corpus does not have; (b) **the embedding and the time step are posited**, not derived from the dynamics, so every slope is in the units the embedding was given in; (c) `lrVelocity`'s `4e` is **not sharp** — it is what the Stirling bound gives at the chosen `2^{-d}` suppression, and `Stirling.factorial_isEquivalent_stirling` is the route to better; (d) **one spatial dimension** on the kinematic side, since the `(E,p)` plane is two-dimensional — the boost link is not extended to higher dimensions; (e) the `(E,p)`/`(t,x)` reading is a **relabelling** of one plane, exactly as `DispersionEarned`'s header says, and **no position/momentum duality** is constructed or needed. *As opened:* **The two cones.** The corpus has two cone structures that have never been related: the **record cone** of [`RecordInfluence.lean`](../CsdLean4/CV/RecordInfluence.lean) and `record_lightcone`'s Lieb–Robinson bound, and the **kinematic Lorentz cone** of [`DispersionEarned.lean`](../CsdLean4/CV/DispersionEarned.lean), where `cone_preserving_is_boost` earns the boost form and boost covariance with `m > 0` selects `ω(p) = √(p² + m²)` — a module that says explicitly it does **not** identify its light cone with the dynamical one. The brick is the **containment**: the record cone sits inside a kinematic cone whose slope is the Lieb–Robinson velocity, a comparison of two bounds and nothing more. ⚠️ An **identification** of the two cones is *not* what this is, needs a continuum limit of the graph dynamics the corpus does not have, and must not be read into the containment. | **M–L** | Medium–high — the first relation between the dynamical and the kinematic cone, and the only piece of #39 that is buildable without the `π'` decision | #39; #38; `CV/DispersionEarned.lean` | > | ~~99~~ | Foundations | Records→spacetime | **DONE 2026-10-02** ([`RecordLayer/MacroProjection.lean`](../CsdLean4/RecordLayer/MacroProjection.lean), 7 pins; S took S) — `MacroProjection X M` is a measurable map to a finite structure, and that is all `π'` is as a structure. `FactorsThrough π B` with ★ `factorsThrough_iff`: **a record event factors exactly when it cannot distinguish two microstates of the same macroscopic coordinate** — the form in which #38's candidates get checked — closed under complement, union and indexed union. ★★ `measure_eq_map_of_factorsThrough` is the engine: a factoring event has the same probability in the macroscopic law (the pushforward) as in the microscopic one, so **every identity between probabilities of factoring events transports verbatim**, which is what "the record statistics factor through `π'`" has to mean. The two theorems the note asks of any instance: ★★★ `noSignalling_map_iff` — **no-signalling survives the projection, in both directions** — and ★★★ `nstep_born_map` — **the chain law survives it**, `csd_nstep_born`'s product read with pushforward measures of coordinate sets. A by-product worth its own name: `patternMap`/`patternProjection`, the membership pattern of a finite record family, is a `MacroProjection` through which the family factors (★ `factorsThrough_patternMap`) and ★★ `patternMap_eq_of_factorsThrough` is its **universal property** — it is the **coarsest** such coordinate, so the factoring requirement always has a canonical minimal solution and every candidate of #38 is a coarsening of it. **FOUR THINGS NOT CLAIMED, in the header:** it is a frame and not a projection (nothing here names `π'`, constructs a macrostate, or shows one with any further property exists — that is #102); the two transports are **conservative on purpose** (nothing lost, nothing gained) and are not evidence for any candidate; `patternProjection` is **not** a candidate — it is tautological for events already in hand and carries no physics; and `M` has no topology, order or geometry, so a macroscopic *geometry* is further structure nothing here provides. *As opened:* **`ST-3`'s structure half.** Paper D's `π'` as a **structure**: a measurable map from `Σ` to a finite structure with the record statistics factoring through it, plus the two theorems any instance must satisfy — no-signalling and the chain law survive the projection. Definitional, buildable now, and **independent of which of the five candidates the author picks**; the *instance* stays gated on that decision in #38, which is where the five candidates are listed. | **S** | Medium now, high the moment an instance is named — it is the frame everything else in #38 hangs on | #38; [`records-to-spacetime-scoping.md`](records-to-spacetime-scoping.md) `ST-3` | > | ~~100~~ | Foundations | Records→spacetime | **DONE 2026-10-04, conditional half as priced** ([`Mathlib/Dynamics/CoarseMarkov.lean`](../CsdLean4/Mathlib/Dynamics/CoarseMarkov.lean), Cat-1, 13 pins; M–L took M–L) — **and the row's own "conditional exactly as `Thermo/ETH.lean` is" needed a NEW hypothesis object, which is the finding.** The corpus's `HasCorrelationDecay` bounds the **two-point correlation of one scalar observable**; a Markov error is about conditional **laws**, so it cannot carry this and the two are left unconnected. The hypothesis built instead is `MicroDecoupled μ C Φ ε`: conditioning on the whole coarse history, rather than on the present coarse value alone, moves the law of the **microstate** at step `n` by at most `ε`, uniformly over measurable sets. **The point of the design is that hypothesis and conclusion are about different objects** — microstate events in, coarse-process statements out — because assuming the coarse process forgets its history would be assuming the conclusion, and that circularity is the trap this row could have fallen into. **What landed.** ★★ `abs_condProb_coarseEvent_sub_le`: the **one-step Markov error is at most `ε`**, through the bridge ★ `coarseEvent_succ_eq` — a coarse future event *is* a microstate event pulled back along the flow, which is exactly the instance of the hypothesis needed. ★ `measure_historyEvent_toReal_eq_prod`: the **exact** chain rule, no hypothesis beyond non-degeneracy. ★★★ `abs_measure_historyEvent_sub_markov_le`: **the path probability factorises up to `n·ε`** — replacing every history-conditional by the one-step transition probability costs at most `n·ε` — which is the row's "approximately Markovian with an explicit error", resting on the elementary ★ `abs_prod_sub_prod_le_sum`. **TWO BOUNDARY RESULTS, and the second is why this is not vacuous.** ★★ `condProb_coarseEvent_succ_of_autonomous`: with an autonomous coarse variable (`C ∘ Φ = T ∘ C`) the transitions are `0`/`1` indicators and the error is `0` with **no** hypothesis, so the bound is attainable. ★★ `cex_not_microDecoupled`: the **four-point rotation coarse-grained into two cells** has `P(C₂ = 1 \| C₁ = 0, C₀ = 0) = 1` against `P(C₂ = 1 \| C₁ = 0) = 1/2`, so its coarse process is **not** Markov and **no `ε < 1/2` exists for it** — the hypothesis is load-bearing and the theorems are not about an empty class. **THE ROW'S THREE PROHIBITIONS HOLD, plus three more in the header:** (i) **not a derivation of the Born rule** — the orbit is deterministic throughout and what is bounded is only how far its coarse marginals sit from a Markov chain's; (ii) decorrelation is a **hypothesis** and nothing supplies it, which is forced, since finite unitary dynamics cannot supply mixing (`not_hasCorrelationDecay_blockPop_of_unitary`); (iii) the coarse-graining is a **parameter** identified with nothing — not with `π'`, not with any macroscopic projection; (iv) the error is **additive in `n`** and useless once `n ≳ 1/ε`, with no stationary, invariant-measure or infinite-horizon claim; (v) it is **not** the corpus's `HasCorrelationDecay` and neither implies the other; (vi) the non-degeneracy hypotheses are real — on a null history the conditionals are `0` by convention and the statements are about nothing. *As opened:* **Effective stochasticity: when does coarse-graining a deterministic flow give a Markov process?** The programme has two kinds of randomness and needs a third stated: the ontic flow is **deterministic**, Born randomness is **epistemic** over microstates ([`CSD-CHARTER.md`](CSD-CHARTER.md)), and the apparent stochasticity of *macroscopic* coarse variables has no statement anywhere in the corpus. The brick: for a measurable coarse-graining `C : Σ → X` and a decorrelation hypothesis on the unresolved degrees of freedom, the coarse process `C(Φ_t σ)` is approximately Markovian with an explicit error — conditional on its hypothesis exactly as [`Thermo/ETH.lean`](../CsdLean4/Thermo/ETH.lean) is conditional on ETH. ⚠️ **THREE THINGS IT MUST NOT CLAIM:** (i) it is **not** a derivation of the Born rule — chaos does not give Born, and the hitting-time route was examined and abandoned ([`q12-fibre-mechanism-scoping.md`](q12-fibre-mechanism-scoping.md), [`CSD-CHARTER.md`](CSD-CHARTER.md): the limit theorems need vanishingly rare sets and a Born partition's cells are not rare); (ii) it needs decorrelation as a **hypothesis**, because finite unitary dynamics is almost periodic and cannot supply mixing (`not_hasCorrelationDecay_blockPop_of_unitary` in [`Thermo/Equilibration.lean`](../CsdLean4/Thermo/Equilibration.lean)); (iii) the coarse-graining is a **parameter**, not the spacetime projection — that is #38's `π'`. | **M–L** conditional; unconditional **XL+** | Medium–high — the one genuinely new theorem in the statistical-mechanics middle layer, and the third level of "random" the architecture needs | #38; #99; `Thermo/Equilibration.lean`; `Thermo/ETH.lean` | > | ~~101~~ | Foundations | Statistical-mechanics layer | *new (2026-10-02)* **Generic chaos in the ontic dynamics: positive Lyapunov exponent, random-matrix level statistics, exponential OTOC.** What [`CV/ChaosBounds.lean`](../CsdLean4/CV/ChaosBounds.lean) has: ★★ `otoc_graphInteractingU_eq_zero`, **the OTOC light-cone gate** — scrambling cannot outrun causal influence, the OTOC is exactly zero while the cones are disjoint — the slow-scrambling cap (OTOC growth at most **linear** in the kicked-diagonal family), ★★ the echo price, and exact spectral revival for the free field as an integrable baseline. What it does **not** have: no positive Lyapunov exponent, no random-matrix level statistics, no spectral ramp or plateau, no generic exponential OTOC regime, and no theorem that the de-isolation dynamics is chaotic. The corpus's own family is *slow* scrambling, so this needs a different dynamical family, which is why it is XL+. ▸ **SPLIT 2026-10-14 into #148–#150; (d) folded into #19.** The "different dynamical family" now exists on the fibre: `catStroke` has decay of correlations with a rate, strong mixing and relaxation (~~#118~~, ~~#147~~). Against it the three deliverables are three different questions: the **Lyapunov exponent** (#148, M: `2 log φ` on the hyperbolic witness, zero on the constructed de-isolation witness, which is a translation) and the **classical OTOC** (#149, M: the derivative of `cat^n` has Fibonacci entries, exact exponential growth; constant on a translation) are priced; **random-matrix level statistics** (#150) stay research, for want of both the theory and an object; and "the de-isolation dynamics is chaotic" is Posit 1's discharge (#19), not a chaos question. ⚠️ **It gates nothing**: chaos is not where Born probability comes from (that route was examined and abandoned, #100), and the opposite over-reading is already refuted — [`LF4/KahlerFibreMixing.lean`](../CsdLean4/LF4/KahlerFibreMixing.lean) exhibits a **mixing** map on the fibre, so "`Σ` cannot mix" is false; it is the compact-group flows used for *isolated* QM that are not mixing. | ~~**XL+**~~ split: **M** + **M** + research | Low–medium — explanatory, not load-bearing | #100; #148–#150; #19; `CV/ChaosBounds.lean`; `LF4/KahlerFibreMixing.lean` | > | ~~102~~ | Foundations | Records→spacetime | **DONE 2026-10-02, candidate (5) at the author's decision** ([`RecordLayer/RecordMacrostate.lean`](../CsdLean4/RecordLayer/RecordMacrostate.lean), 9 pins; M took M): the coordinate is the **record string** — `outcomeCode` per context (`i.succ` on the basin of `i`, `0` on the null set where none contains the microstate) and ★★ `recordMacro`, the outcome each context of a **finite indexed** family assigns, as a `MacroProjection` onto `Fin k → Fin (N+1)`. Indexed, so the string remembers which context gave which outcome — the order-retaining reading, which is why the histogram was rejected. ★★ `globalBasin_eq_preimage`: every record event of the family factors through it. **BOTH OF #99's REQUIREMENTS ARE VERIFIED HERE, NOT MERELY TRANSPORTED:** ★★★ `noSignalling_pairFamily` — no-signalling holds **exactly**, because summing the second context's outcomes returns the first's basin whatever the second context is (any context's basins a.e. exhaust `Σ`) — and ★★★ `nstep_born_recordMacro` — **the Born chain law holds verbatim in the macroscopic law**. With ★★ `measure_map_recordMacro_singleton` (a one-context macrostate carries the Born weight) and ★ `recordString_coarsen` (dropping contexts coarsens the coordinate, so the macrostates form a lattice). **FOUR THINGS NOT CLAIMED, in the header:** (i) **this is the construction, not the physics** — stability of the cells is what makes a coordinate macroscopic and is **row 103**; (ii) **no product law for multi-context cells**, and that is contextuality showing: every context reads the *same* fibre angle through its *own* base-dependent arcs, so a joint cell is an intersection of arcs, no independence is claimed, and this is no non-contextual value assignment; (iii) the family is fixed and finite and **no canonical family is claimed** — the lattice is the space of choices, not a selection; (iv) `M` is a set of **labels**, with no geometry, order or topology — candidate (2)'s causal order would be extra structure on `M` and is not built here. *As opened:* **Macrostate construction: the instance of `ST-3`'s structure.** Row 99 builds the *frame* (`π'` as a structure plus the two invariance theorems); this row builds an **instance**, which is the first problem of the statistical-mechanics middle layer — *what counts as the same macroscopic record configuration*. **WHAT THE CORPUS HAS AND WHY IT IS NOT THIS:** `pinch` ([`Thermo/SecondLaw.lean`](../CsdLean4/Thermo/SecondLaw.lean)) is a coarse-graining the second law is stated for, but a *given* partition on density matrices rather than a construction on `Σ`; `energyWindow` ([`Thermo/EnergyWindow.lean`](../CsdLean4/Thermo/EnergyWindow.lean), with ★★ `energyWindow_ne_zero`) is a genuine microcanonical macrostate on the base, but one energy band and not an equivalence of records; `globalBasin` ([`RecordLayer/GlobalBasin.lean`](../CsdLean4/RecordLayer/GlobalBasin.lean)) partitions `Σ` by **record outcome** with the frequency theorems, but those are one context's Born cells — a *measurement* partition, not a macroscopic one. **What is absent is any equivalence or quotient on `Σ` defined by agreement of many records, and any notion of macroscopic stability.** The brick, for whichever of #38's five candidates the author picks: define the projection or equivalence, prove the record statistics factor through it, and discharge row 99's two invariance theorems for it. **M** for candidates (1) the record histogram, (2) the influence quotient and (3) mode occupancy; **M–L** for (5) thermodynamic macrostates, which additionally needs stability under the flow and would lean on `fs_chebyshev_concentration` and [`Thermo/Equilibration.lean`](../CsdLean4/Thermo/Equilibration.lean); unpriced for (4) until the `entDist` binning is fixed. ⚠️ Opening this row does **not** pick a candidate — that is #38's decision, and this row is whichever instance follows from it. | **M** – **M–L** by candidate | High — the first problem of the middle layer, and the piece that makes `π'` a construction rather than a name | #38; #99; #100 | > | ~~103~~ | Foundations | Records→spacetime | **DONE 2026-10-02** ([`RecordLayer/MacrostateStability.lean`](../CsdLean4/RecordLayer/MacrostateStability.lean), 19 pins; L took M) — **the brick picked is the measure-theoretic one, and it lands in two halves that pull against each other.** (i) ★★★ `map_recordString_sigmaShift`: a record write translates the record coordinate of the fibre, so it preserves the epistemic measure (★ `measurePreserving_sigmaShift`) and **the macroscopic law is EXACTLY invariant** — every write size, no error term, no margin hypothesis. (ii) ★★★ `measure_recordString_ne_le`: the microstates whose record string a write of size `δ` **changes** have epistemic measure at most `k·N·δ` for a family of `k` contexts over `N` outcomes, and ★ `tendsto_measure_recordString_ne` sends that to `0`. The mechanism is one lemma — ★ `rep_add_coe`, new in [`CircleFibre.lean`](../CsdLean4/RecordLayer/CircleFibre.lean) together with ★ `volume_rep_preimage_Ioc` (which `volume_circleCell` is now a corollary of): **a translation that does not wrap ADDS to the canonical representative**, so a write pushes a microstate towards the *upper* endpoint of its own arc and can only relabel what sits within `δ` of it (`robustBasin`/`edgeBand`, ★ `globalBasin_subset_robustBasin_union_edgeBand`, the margin lemma ★★ `sigmaShift_mem_globalBasin`, then ★★ `outcomeCode_sigmaShift`, ★★★ `recordString_sigmaShift` and ★★ `recordMacro_sigmaShift` at #102's macrostate level). **Three further results.** ★★ `measure_robustBasin_ge`: **robustness is controlled by the Born weight** — `rate p i − δ` bounds the robust interior below, so the heavy records are the stable ones and a cell of weight under `δ` is granted none. ★★ `measure_robustBasin_ge_one_sub`: **the majority half, conditional on concentration** — once `1 − ε ≤ rate p i` the shared-and-robust cell carries all but `ε + δ`; and since `globalBasin_prob` makes a cell's measure *exactly* its Born weight, **no cell is overwhelming unless the weights are**, so "overwhelmingly many microstates share it" is a property of the preparation and the context and **not of the coordinate** — the honest reading of this row's second half, whose hypothesis is **NEW ROW 105**. ★★★ `measure_exists_recordString_ne_le` / `measure_drift_recordString_ne_le`: **one exceptional set serves a whole horizon** (the margin bought at `δ` covers every smaller write), so under a drift of the record coordinate at rate `ω` the string survives all of `[0,T]` outside measure `k·N·ω·T` — this row's "persistence over a horizon", with no second argument and no union over times. **SIX THINGS NOT CLAIMED, in the header:** (a) **not invariance, and no mixing** — no cell is shown invariant, nothing contradicts `not_hasCorrelationDecay_blockPop_of_unitary`; exact invariance of the *law* and an `O(δ)` pointwise error are compatible precisely because one is a pushforward and the other is pointwise; (b) **only the record coordinate moves** — a flow that moves the **base** moves the arcs themselves, and a `ContextField` is only *measurable*, so that case needs a hypothesis bounding the arcs' displacement: **ROW 104, DONE THE SAME DAY**; (c) ~~the write is a **constant** translation (`CV`'s base-dependent `recordStroke` would need an a.e. identification resting on singleton measurability in `LF4.CPN N`, not established)~~ — **CORRECTED AND DISCHARGED THE SAME DAY BY #104**: `MeasurableSingletonClass (LF4.CPN N)` **is** available at the pin, so the "not established" was a bad measurement; `ae_fst_eq` puts the epistemic measure on one fibre and `map_sigmaMove_id` covers the base-dependent write, needing no measurability of the write at all; (d) **the majority hypothesis is not supplied and is not `fs_chebyshev_concentration`** — that is about base statistics, lives elsewhere in `Σ`, and no theorem joins it to the fibre bound; (e) the bound **degrades as `k·N`** and no better constant is claimed, so small-weight records are genuinely fragile; (f) **no dynamics of `Σ` is derived** — translation is the dynamics the record medium *has* (`measurePreserving_fibreShift`), not a flow from an interaction Hamiltonian. *As opened:* **Dynamical stability of record macrostates — candidate (5)'s physics.** #102 built the coordinate; what makes it *macroscopic* is that its cells are **stable**: that the same record string persists under the flow, and that overwhelmingly many microstates share it. Neither is proved. The second half has its input already — `fs_chebyshev_concentration` ([`Thermo/CanonicalTypicality.lean`](../CsdLean4/Thermo/CanonicalTypicality.lean), polynomial rate) and the conditional time-averaged equilibration of [`Thermo/Equilibration.lean`](../CsdLean4/Thermo/Equilibration.lean). ⚠️ **Stability cannot mean invariance**: finite unitary dynamics is almost periodic and supplies no mixing (`not_hasCorrelationDecay_blockPop_of_unitary`), so the statement has to be persistence over a horizon, or a measure-theoretic statement about the cell, or conditional on a decorrelation hypothesis as #100 is. Picking which of those is the brick is the first step. | **L** | High — it is the content of candidate (5), and until it lands #102 is a relabelling rather than a macrostate structure | #102; #100; #38; `Thermo/CanonicalTypicality.lean`; `Thermo/Equilibration.lean` | > | ~~104~~ | Foundations | Records→spacetime | **DONE 2026-10-02** ([`RecordLayer/BaseMotionStability.lean`](../CsdLean4/RecordLayer/BaseMotionStability.lean), 19 pins; M took M) — **and the hypothesis turned out to be cheaper than the row predicted.** The row proposed a `LipschitzWith` bound on the rate field, which would have needed a metric on `LF4.CPN N`; the margin argument actually consumes only `ArcShift c ε p q` — **every arc of the context moves by at most `ε` between the two base points** — which mentions **no metric and no continuity at all**, and ★ `arcShift_of_rate_close` reduces it to a modulus on the rates when one is available (`ArcShift c ((N+1)·η)` from rates within `η`, each endpoint being a sum of at most `N+1` rates). **What landed.** `sigmaMove f g` is a measurable motion `f` of the base **plus a base-dependent record write** `g`. ★★★ `map_sigmaMove`: **the epistemic law is carried EXACTLY to the law at the moved base**, `(epistemicMeasure p).map (sigmaMove f g) = epistemicMeasure (f p)` — the right form of #103's exact half once the base moves, and note it is deliberately *weaker*: the macroscopic law is **not** invariant, because the Born weights are now those of the moved point, which is what a base motion should do. ★★ `sigmaMove_mem_globalBasin` is the margin lemma with **both** endpoints in play, so `robustBasin₂` is **two-sided** (margin `ε` below, `δ + ε` above) exactly as the row predicted, with `robustBasin₂_zero` recovering #103's one-sided set; then ★★★ `recordString_sigmaMove` and ★★ `recordMacro_sigmaMove`. ★★★ `measure_recordString_ne_sigmaMove_le`: **the microstates a move relabels have epistemic measure at most `k·N·(δ + 2ε)`**, with ★★ `measure_recordString_ne_sigmaMove_rate_le` giving `k·N·(δ + 2(N+1)η)` in terms of the rates — the row's predicted `k·N·(δ + (N+1)ε)` was an underestimate by a factor of two on the displacement term, because the band is two-sided. ★★ `measure_robustBasin₂_ge` keeps the **Born weight** in control (`rate p i − (δ + 2ε)`), and ★★★ `measure_exists_recordString_ne_flow_le` gives **one exceptional set for a whole one-parameter family of moves over a horizon**, since the robust interior does not depend on the parameter. **A CORRECTION TO #103, MADE AT SOURCE:** that row's caveat (c) said a base-dependent write would need "singleton measurability in `LF4.CPN N`, not established". **The instance IS available at the pin** — the measurement was wrong — so ★ `ae_fst_eq` (the epistemic measure lives on the fibre over its own preparation) and ★★ `map_sigmaMove_id` **discharge the caveat**, and they need no measurability of the write at all. #103's header and this ledger are both corrected. **FIVE THINGS NOT CLAIMED, in the header:** (a) **the arc displacement is assumed, not derived** — no flow of `Σ`, least of all a Hamiltonian one, is shown to bound it, and supplying the modulus for the canonical `momentContext` is not done (`LF4.momentMap` has no recorded modulus); (b) **`f` is an arbitrary measurable map, not a flow** — no group law, no invertibility, no generator, so "flow" names the shape of the horizon statement and not a structure; (c) **still not invariance and still no mixing**; (d) the bound degrades as `k·N` and **again in `N` through the rates**, so a large alphabet needs the rate field uniformly close; (e) **the typicality half is untouched** — still #105. *As opened:* **Stability of record macrostates under motion of the BASE.** #103 proved stability under motion of the **fibre** — the macroscopic law exactly invariant, the pointwise string robust outside a set of measure `k·N·δ` — and that is the only channel a record write opens. A flow that moves the **base** moves the arcs themselves: `globalBasin c i` is cut out by `loSum (c.rate x.1) i < rep x.2.1 ≤ cellTop c i x`, so displacing `x.1` displaces *both* endpoints, and a `ContextField` is only assumed **measurable**, under which the arcs may move discontinuously and no bound exists. **The route:** give `ContextField` a modulus-of-continuity obligation (or hypothesise `LipschitzWith L fun p => c.rate p i`), then re-run #103's margin argument with margin `δ + ε` where `ε` bounds the endpoint displacement — `loSum` moves by at most `Nε` and `cellTop` by at most `(N+1)ε` — for a bound `k·N·(δ + (N+1)ε)`. Unlike #103 the band is **two-sided**, because the *lower* endpoint can move up as well, so `robustBasin` needs a margin at both ends. ⚠️ **What to check first:** whether `LF4.CPN N` carries the metric such a hypothesis needs in this corpus, and whether the canonical `momentContext` satisfies it — `LF4.momentMap` ([`Mathlib/LinearAlgebra/Projectivization/MomentMap.lean`](../CsdLean4/Mathlib/LinearAlgebra/Projectivization/MomentMap.lean)) is measurable but **no modulus is recorded**, so the hypothesis may be cheap or may be the whole row. | **M** | Medium — it is what turns #103 from stability under record *writes* into stability under a flow of `Σ`, and #100's effective stochasticity will want it | #103; #102; #100; `RecordLayer/GlobalBasin.lean`; `Mathlib/LinearAlgebra/Projectivization/MomentMap.lean` | > | ~~105~~ | Foundations | Records→spacetime | **HALF (a) DONE 2026-10-04, half (b) RE-OPENED AS ROW 106** ([`RecordLayer/PointerConcentration.lean`](../CsdLean4/RecordLayer/PointerConcentration.lean), 10 pins; M–L took M for (a)) — **the row split on inspection, and the first half had a geometric answer.** #103 needed `1 − ε ≤ c.rate p i` and recorded why it cannot be dropped. For the canonical apparatus that hypothesis **is a distance**: `offPointer i ψ` is the component of the preparation orthogonal to the `i`-th pointer ray, ★ `norm_offPointer_sq` says **the defect is exactly the missing Born weight** (`‖offPointer i ψ‖² = ‖ψ‖² − ‖ψ i‖²`), and so ★★ `one_sub_le_momentMap_of_near_pointer`: **a preparation within `√ε` of the pointer ray has rate at least `1 − ε` there**. Feeding that into #103 gives ★★★ `measure_robustBasin_ge_of_near_pointer` and its distance form ★★★ `measure_robustBasin_ge_of_dist_le`: **near a pointer ray the record macrostate is BOTH overwhelming AND robust** — the cell of outcome `i`, cut down to the microstates robust to a record write of size `δ`, carries all but `ε + δ` of the epistemic measure. ★★★ `measure_map_recordMacro_ge_of_near_pointer` says it in #102's **macroscopic law**: all but `ε` of its mass sits on one record string. ★ `offPointer_single` / ★★ `measure_robustBasin_ge_of_pointer` are the **non-vacuity** check at the pointer state itself (`ε = 0`, cell carries `1 − δ`). **FIVE THINGS NOT CLAIMED, in the header:** (a) **nothing drives the base to a pointer ray** — the concentration is where the preparation *sits*, supplied as a hypothesis with a geometric meaning, and the dynamical half is **ROW 106**; (b) it is **still not** `fs_chebyshev_concentration`, which concentrates a *base statistic* over the projective space and is a different statement about a different object; (c) **only the canonical standard-basis `momentContext`** — the unitary transport to a general apparatus is not constructed anywhere in the corpus; (d) **record macrostates are not generically overwhelming** — for a generic superposition no cell is, and the bound is vacuous once `r² + δ ≥ 1`, which is correct physics and not a gap: the content is the *rate* at which macroscopicity is bought with nearness to the pointer; (e) **one context only** — nothing about a joint cell of two contexts, where #102's contextuality warning applies. *As opened:* **Where the overwhelming majority comes from: concentration of a context's weights.** #103 proved the majority half *conditionally* — ★★ `measure_robustBasin_ge_one_sub`: given `1 − ε ≤ c.rate p i` the shared-and-robust macrostate cell carries all but `ε + δ` of the epistemic measure — and recorded **why the hypothesis cannot be removed**: `globalBasin_prob` makes a cell's measure *exactly* its Born weight, so no cell is overwhelming unless the weights are. **"Overwhelmingly many microstates share the record" is therefore a statement about the preparation and the apparatus, not about the coordinate**, and it is false for a generic superposition — which is correct physics, not a defect. ~~The brick is to supply the hypothesis from something: a pointer/decoherence model whose rate field concentrates~~ ⚠️ **CORRECTED 2026-10-04 by #106: a DECOHERENCE model cannot do it** — dephasing is modulus-preserving and a record weight is a modulus squared, so the defect is a fixed point (`not_concentrates_of_norm_coord_eq`); the honest replacement is the robust fraction *within* the realised cell, which needs no concentration at all. The route as written was (the corpus has the environment-marginal machinery of [`Empirical/CSD/QEC/SteaneFlow.lean`](../CsdLean4/Empirical/CSD/QEC/SteaneFlow.lean) and [`CV/RecordInfluence.lean`](../CsdLean4/CV/RecordInfluence.lean)'s medium, neither phrased as a rate field), or a `ContextField` built from a pointer basis with an explicit `ε`. ⚠️ **It is NOT `fs_chebyshev_concentration`** ([`Thermo/CanonicalTypicality.lean`](../CsdLean4/Thermo/CanonicalTypicality.lean)) — that concentrates a *base* statistic over the projective space and says nothing about one context's weight vector at one base point; conflating the two is the mistake this row exists to avoid. | **M–L** | Medium — it is the second half of what makes #102's coordinate *macroscopic*, and the place a decoherence story would enter the record layer | #103; #102; #100; `Thermo/CanonicalTypicality.lean`; `CV/RecordInfluence.lean` | > | ~~106~~ | Foundations | Records→spacetime | **DONE 2026-10-04 — and its first route is REFUTED, not merely unfinished** ([`RecordLayer/PointerDynamics.lean`](../CsdLean4/RecordLayer/PointerDynamics.lean), 5 pins; L took M) — **the row's premise was wrong and this corrects it.** Route (1) asked for a dephasing estimate `1 − momentMap (Φ t p) i ≤ e^{−λt}(1 − momentMap p i)`. But **dephasing acts on a pure preparation by a phase per branch — that is what destroys interference — and a phase changes no amplitude's MODULUS**, while a record weight is a modulus squared and nothing else. So ★★★ `momentMap_eq_of_norm_coord_eq`: two unit preparations with the same coordinate moduli have the **same** moment coordinates; ★★ `globalBasin_prob_eq_of_norm_coord_eq`: **no basin's probability changes**, outcome by outcome; and ★★★ `not_concentrates_of_norm_coord_eq`: **the defect `1 − momentMap p i` is a FIXED POINT of every modulus-preserving map**, so no contraction factor `q < 1` exists at any rate or any time unless the weight was already `1`. Route (1) is not an open problem — it is the wrong mechanism, and the row (which I wrote) named it as if it were reachable. **WHAT IS TRUE INSTEAD, with no concentration hypothesis at all.** ★★★ `one_sub_le_robust_fraction`: *within* the cell of outcome `i`, the fraction of microstates robust to a record write of size `δ` is at least `1 − δ / rate p i` — so **"overwhelmingly many microstates share the record" holds for EVERY outcome** once it is read as a statement about that outcome's own cell, and what it needs is `δ` small compared with the weight, **not a dominant weight**. ★★ `robust_fraction_tendsto_one` sends that fraction to `1` as the write shrinks. **SIX THINGS NOT CLAIMED, in the header:** (a) maps that genuinely **reweight** the amplitudes (amplitude damping, measurement with feedback) are **not covered and not refuted** — and physically they are *re-preparations*, not decoherence, so they cannot be what makes an existing superposition's record macroscopic; (b) **#105(a) is not withdrawn** — its statement stands, only the expectation that decoherence supplies its hypothesis is corrected; (c) **no cell is claimed large** — the replacement is conditional on the outcome, and for a genuine superposition no cell is large, which is correct physics and not a gap; (d) **route (2) is untouched** — connecting the shear de-isolation pointer to the record layer is **NEW ROW 107**, still inheriting the de-isolation obligation; (e) **no dynamics is constructed** — the no-go quantifies over preparations with equal moduli and exhibits no flow, generator, environment or time parameter; (f) the canonical standard-basis `momentContext` only. *As opened:* **What drives the base to a pointer ray — the dynamical half of the overwhelming majority.** #105(a) showed that nearness to a pointer ray buys an overwhelming *and* robust record macrostate, with the squared distance as the only input. What is missing is any dynamics that **produces** that nearness: `1 − ε ≤ momentMap p i` is a hypothesis about where the preparation sits, and nothing in the corpus moves a base point towards a coordinate ray. ⚠️ **A unitary base flow cannot do it**, which is why this is not a small row: finite unitary dynamics is almost periodic and supplies no mixing (`not_hasCorrelationDecay_blockPop_of_unitary`), so concentration needs an **environment and a reduced description** — i.e. it inherits the standing de-isolation obligation of [`RecordLayer/DeIsolationFlow.lean`](../CsdLean4/RecordLayer/DeIsolationFlow.lean) ("exhibit a pointer generated by a physical `H_int(M)`") and of [`RecordLayer/ShearDeIsolation.lean`](../CsdLean4/RecordLayer/ShearDeIsolation.lean), whose Hamiltonian-generation residue is `ShearWitness` item 1. **Two routes, both needing new material:** (1) an explicit **dephasing estimate** — a channel or environment marginal (the `controlledUnitary` dilation of `#52` and the environment marginals of [`Empirical/CSD/QEC/SteaneFlow.lean`](../CsdLean4/Empirical/CSD/QEC/SteaneFlow.lean) are the nearest tools) with a rate: `1 − momentMap (Φ t p) i ≤ e^{−λt}(1 − momentMap p i)`, which would close #105 end to end when composed with `measure_robustBasin_ge_of_dist_le`; (2) connect the shear de-isolation pointer to `momentMap` concentration directly. ⚠️ Neither route is a repackaging of existing theorems, and the row should **not** be priced as independent of the de-isolation obligation. | **L** | High — it is the one thing standing between #102/#103/#105(a) and "records are macroscopic because the dynamics makes them so"; until it lands, macroscopicity is conditional on the preparation | #105; #103; #102; `RecordLayer/DeIsolationFlow.lean`; `RecordLayer/ShearDeIsolation.lean` | > | ~~107~~ | Foundations | Records→spacetime | **DONE 2026-10-04 — the bridge is ONE condition, and it reduces to measure preservation** ([`RecordLayer/ArenaTransport.lean`](../CsdLean4/RecordLayer/ArenaTransport.lean), 8 pins; L took M). Rows #102–#106 are statements about `globalBasin`s under `epistemicMeasure`; the pointer the corpus actually *constructs* carves its sectors as **preimages** of those basins under a de-isolation propagator on a fibred arena. The bridge is `ReadsEpistemic p μ Φ`: **the arena's preparation measure pushes forward along `Φ` to the epistemic measure** — one identity, assuming nothing about the arena's type, fibre or dynamics. ★★★ `readsEpistemic_of_measurePreserving`: **the condition reduces to a preparation-preserving flow** — on a product arena `Σ × (ready register)` a flow `F` preserving the preparation makes `Prod.fst ∘ F` read the epistemic measure; and ★★★ `readsEpistemic_readyPrep_of_measurePreserving`: **the corpus's own preparation measure IS that product** — `readyPrep p = epistemicMeasure p ⊗ readyMeasure N` by definition, so for the shear arena the bridge is a single hypothesis about the propagator, `MeasurePreserving F (readyPrep p) (readyPrep p)`, and nothing else about records, cells or basins. **WHAT TRANSPORTS:** ★ `measure_preimage_eq` (every measurable record event has the same probability in the arena as in `Σ`) and ★ `isProbabilityMeasure_of_readsEpistemic` (the arena's preparation is a probability measure for free); ★★ `measure_preimage_globalBasin` — **the flow-carved sector carries the Born weight**, the shape of `shear_sector_born` from the pushforward alone; ★★ `measure_preimage_robustBasin_ge` and ★★★ `one_sub_le_robust_fraction_preimage` — **#106's robust fraction holds for the flow-carved sector**; ★★★ `measure_preimage_recordString_ne_le` — **#103's relabelling bound holds for the arena** (at most `k·N·δ`). **FOUR THINGS NOT CLAIMED, in the header:** (a) **the shear propagator is not shown to satisfy this** — the corpus has the per-sector Born identity but not the pushforward identity, and the reduction above makes the whole remaining gap `MeasurePreserving F (readyPrep p) (readyPrep p)`, which sits upstream of `ShearWitness` item 1 (the propagator's Hamiltonian generation is stated, not formalised) — that is **NEW ROW 108**; (b) **a pushforward is not a dynamics** — no generator, interaction Hamiltonian or time parameter appears, and `DeIsolationFlow.lean`'s obligation is untouched; (c) **transport preserves scope, including the limits** — #103's bound stays robustness and not invariance, #106's fraction stays conditional on the realised outcome and claims no cell is large, and the write stays a translation of the record coordinate; (d) **one preparation** — the transport is at a fixed base point `p` and nothing varies or integrates over the preparation. *As opened:* **The shear de-isolation pointer, connected to the record layer.** #106 refuted its own route (1) — dephasing cannot concentrate the Born weights, because a record weight is a modulus squared and a phase moves no modulus. Route (2) survives: [`RecordLayer/ShearDeIsolation.lean`](../CsdLean4/RecordLayer/ShearDeIsolation.lean) exhibits the **readout ∘ flow shape** — `cellPointer_outcomeSector_eq_readout` pointwise, with `basin_rate` discharged from `shear_sector_born` — and [`RecordLayer/DeIsolationFlow.lean`](../CsdLean4/RecordLayer/DeIsolationFlow.lean)'s ★★ `map_pointer_apply` makes the remaining obligation precise: supply a measurable pointer whose basins carry the Born cell measures. The brick is to connect that construction to **#102's record string and #103's robust cells** — i.e. exhibit the shear pointer's basins *as* `globalBasin`s of a `ContextField`, so that #103's stability and #106's robust-fraction bound apply to the pointer the de-isolation flow actually produces, rather than to a posited context field. ⚠️ **What it does NOT need is concentration** — that was #106's correction: the honest target is the robust fraction *within* the realised cell, which holds for every outcome. ⚠️ **It still inherits the standing de-isolation obligation**: `ShearWitness` item 1 (the propagator's Hamiltonian generation is stated, not formalised) is upstream of it, so the row should not be priced as if that were settled. | **L** | Medium–high — it is what would make the record layer's stability results about the *constructed* pointer instead of a posited one | #106; #105; #103; #102; `RecordLayer/ShearDeIsolation.lean`; `RecordLayer/DeIsolationFlow.lean` | > | ~~108~~ | Foundations | Records→spacetime | **DONE 2026-10-05 — DISCHARGED, and its recorded route is REFUTED** ([`RecordLayer/ShearArenaBridge.lean`](../CsdLean4/RecordLayer/ShearArenaBridge.lean), 10 pins; M–L took M). The row asked for route (i): build the shear propagator and prove it **preserves `readyPrep`**. ★★★ `not_measurePreserving_shearEvolve_readyPrep` shows **it cannot** — a record *is* the pointer leaving the ready arc for a pointer arc, `shearProtocol`'s own `ready_disjoint_pointer` says the arcs are disjoint, so the ready region has `readyPrep` measure `1` while its preimage has measure `0`. **Route (i) was not unfinished; it asked for the negation of the construction's purpose** — and the row (which I wrote) priced it as M–L work. (`shear_measurePreserving` is no counterexample: it preserves `μ ⊗ volume`, the *unconditioned* Haar measure on the register, and `readyPrep` conditions on the ready arc — conditioning is what the shift destroys.) **AND THE BRIDGE HOLDS ANYWAY, for less.** ★★★ `readsEpistemic_of_fst_eq`: **a propagator that does not move the system reads the epistemic measure** — no measure preservation, no hypothesis on the register; ★★★ `readsEpistemic_readyPrep_fst` and ★★★ `readsEpistemic_shearEvolve`: **the shear propagator satisfies it at every pair of times**, because the shear moves only the pointer. So #107's transports all apply to the *constructed* propagator: ★★ `readyPrep_preimage_globalBasin` (the transported basin carries the Born weight), ★★★ `readyPrep_outcomeSector_eq_preimage_globalBasin` (**and it is the same weight the flow-carved outcome sector carries**, via `shear_sector_born`, outcome by outcome), ★★★ `readyPrep_recordString_ne_le` (#103's relabelling bound on the shear arena) and ★★★ `one_sub_le_robust_fraction_readyPrep` (#106's robust fraction there). **THE PRICE, which the corpus already knows:** the property that discharges the bridge — **no back-reaction on the system factor** — is the same one `shear_base_marginal_unchanged` identifies as the witness's limitation, that the shear gives repeatability but **not** the Lüders update. So #108 is discharged *because* the witness does not collapse the state; a propagator that did disturb the system would need #107's other sufficient condition or a new argument. **THREE THINGS NOT CLAIMED, in the header:** (a) equality of **sets** — the sector identity is equality of *measures*, not of sets and not almost-everywhere agreement, which would need the injectivity of the arc labelling; (b) any progress on the de-isolation obligation — `ShearWitness` item 1 is upstream of everything here and nothing makes the shear propagator a Hamiltonian flow; (c) more than one preparation, or any selector other than the canonical `basinIndex (momentContext N)`. *As opened:* **The shear propagator preserves `readyPrep`.** #107 reduced the whole record-layer-to-arena bridge to one hypothesis: `MeasurePreserving F (readyPrep p) (readyPrep p)` for the de-isolation propagator `F` on `Σ × ready register`, after which [`RecordLayer/ArenaTransport.lean`](../CsdLean4/RecordLayer/ArenaTransport.lean)'s ★★★ `readsEpistemic_readyPrep_of_measurePreserving` carries #103's stability, #105(a)'s majority and #106's robust fraction to the sectors the flow actually carves. The brick is to **supply that hypothesis for a propagator**, not to assume it: either (i) build the shear propagator as an explicit measurable map on `LF4.KSigma N × LF4.KTorus` and prove it measure-preserving — the natural route, since the arena's measure is a product of `epistemicMeasure` with a conditioned Haar measure and `measurePreserving_add_right`-style translations on the torus fibre are already in Mathlib — or (ii) derive preservation from Liouville, via `R-016′`'s arena symplectic form, which would make it a *consequence* of the isolated dynamics rather than a construction. ⚠️ **It inherits `ShearWitness` item 1**: the propagator's Hamiltonian generation is stated, not formalised, so route (ii) is behind that and route (i) proves measure preservation of a map that is *not yet* tied to an interaction Hamiltonian. ⚠️ **What it does NOT need** is concentration (#106's correction) or any new record-layer theorem — #107 has already discharged everything downstream of the hypothesis. | **M–L** for route (i); **XL** for route (ii) | Medium — it is the last link between the constructed pointer and the record layer's stability results | #107; #106; #103; `R-016′`; `RecordLayer/ShearDeIsolation.lean`; `RecordLayer/ShearWitness.lean`; `RecordLayer/DeIsolationFlow.lean` | > | ~~109~~ | Foundations | Records→spacetime | **DONE 2026-10-04 — THE SECOND LAW AT `π′`, and Liouville forces its reference law** ([`Mathlib/InformationTheory/KlDivArrow.lean`](../CsdLean4/Mathlib/InformationTheory/KlDivArrow.lean) 4 pins + [`RecordLayer/MacrostateArrow.lean`](../CsdLean4/RecordLayer/MacrostateArrow.lean) 10 pins; M took M). *As opened (2026-10-04, out of the scoping note's `ST-3`):* **the arrow of time at the macroscopic projection.** The corpus's second law was `Thermo`'s — coarse-grained entropy monotonicity for **one specific** coarse-graining, pinching to a pointer basis (`vonNeumannEntropy_le_pinching`) — and `ST-3` said the specification wanted was `π′`. Since #102 chose `π′` and #103–#107 made it stable, the second law can now be stated **at** it. **The Cat-1 half** (absent from Mathlib, which has the three data-processing inequalities for `klDiv` but not these): ★★ `klDiv_map_measurableEquiv` — **a reversible relabelling changes no divergence** (the DPI along `e` and along `e.symm` gives equality), so an invertible step produces *nothing*; ★★★ `klDiv_comp_le_of_stationary` — **the H-theorem**: if a Markov kernel fixes `π`, every law's divergence from `π` is non-increasing, with **no** detailed balance, symmetry or double stochasticity — only stationarity; `Measure.compIterate` with ★★ `antitone_klDiv_compIterate` — monotone along the **whole trajectory**, which is what makes it an arrow rather than a one-step inequality. **The record-layer half**: `recordLaw c p` is the macroscopic law (the pushforward of the ontic measure along #102's record string); `macroKernel c p F` is **the macro dynamics a fine dynamics induces with no extra input** — the conditional law of the next macrostate given the present one, `dirac` on the null macrostates so it is Markov on the nose — and ★★★ `comp_recordLaw` shows one step of it is exactly the pushforward along `π′ ∘ F` (the chain rule for this projection, via ★★ `macroStep_apply_mul`, which clears the conditioning against the macrostate's own weight). Then ★★★ `comp_recordLaw_of_measurePreserving`: **Liouville forces the reference law** — if the fine dynamics preserves the ontic measure then the **Born record law is stationary** for the induced kernel, so what the divergence is measured from is not chosen but fixed by the invariance of the ontic measure; and ★★★ `antitone_klDiv_recordLaw`: **every macroscopic law's relative entropy from the Born record law is non-increasing, monotonically in the number of steps.** ★★ `comp_recordLaw_sigmaShift` + ★★★ `antitone_klDiv_recordLaw_sigmaShift` make **the corpus's own dynamics an instance** (a record write preserves the epistemic measure, #103). **AND THE NEGATIVE THAT LOCATES THE PRODUCTION:** ★★ `klDiv_recordLaw_sigmaShift` — #103's *exact* invariance of the macroscopic law means a record write changes the divergence from any reference **not at all, however large the write**. The production is in the conditional step — the coarse-graining — and never in the write. **FIVE THINGS NOT CLAIMED, in the headers:** (a) **the kernel is induced, not derived** — `macroKernel` exists for any measurable `F` and says nothing about why `F` is the physical dynamics; `DeIsolationFlow.lean`'s obligation and #108's `readyPrep` hypothesis are untouched; (b) **divergence, not Shannon entropy** — "entropy increases" needs `klDiv q uniform = log card − H q` *and* a uniform reference, and the Born record law is not uniform (for a genuine superposition it is the spread of Born weights); the identity needs the Radon–Nikodym derivative of one finite-type measure against another and is **new row 110**; (c) **monotone is not strict, and no rate** — a kernel can be the identity; mixing and relaxation rates are untouched; (d) one preparation, one context family; (e) **this is not an arrow on the event order** — the monotone quantity is a divergence between macroscopic laws under a step count, not a time orientation of `CV/RecordCausalOrder.lean`'s causal order, and nothing here bears on a metric, a volume element or a continuum limit. | **M** | High — it discharges `ST-3`'s standing sentence that the thermodynamic arrow waits on the `π′` decision, and it is the first ingredient any thermodynamic route to field equations needs | `ST-3`; #107; #103; #102; #100; #99; `Thermo/FreeEnergy.lean`; `Mathlib.InformationTheory.KullbackLeibler.DataProcessing` | > | ~~110~~ | Foundations | Records→spacetime | **DONE 2026-10-05 — the identity, the maximum-entropy theorem, and the H-theorem in entropy form** ([`Mathlib/InformationTheory/FiniteEntropy.lean`](../CsdLean4/Mathlib/InformationTheory/FiniteEntropy.lean) 8 pins + 1 pin in [`RecordLayer/MacrostateArrow.lean`](../CsdLean4/RecordLayer/MacrostateArrow.lean); M–L took M). **Mathlib has `klDiv` but no Shannon entropy of a measure — not for a finite type, not anywhere** — which is why #109's arrow could only be stated in divergence form. `measureEntropy μ = ∑ x, negMulLog (μ.real {x})` supplies it. The work is the Radon–Nikodym computation: ★★ `withDensity_uniformOn_univ` — **the density of a measure against the uniform law is `card · μ{x}`**, and it holds for **any** measure on a finite type, not only a probability measure (so do `absolutelyContinuous_uniformOn_univ` and ★★ `llr_uniformOn_univ_ae`, which the `unusedArguments` linter found: the hypotheses were dropped and the lemmas are strictly stronger than stated). Then ★★★ `integral_llr_uniformOn_univ` and ★★★ `klDiv_uniformOn_univ`: **the identity** `klDiv μ uniform = ofReal (log card − measureEntropy μ)`, clean because both correction terms in `klDiv`'s definition vanish for probability measures. Two classical consequences: ★★ `measureEntropy_le_log_card` — **the maximum-entropy theorem**, Gibbs' inequality read through the identity — and ★ `measureEntropy_uniformOn_univ`, the uniform law attaining it, so the bound is tight. And the row's target: ★★★ `monotone_measureEntropy_compIterate` — **the H-theorem in entropy form**, Shannon entropy non-decreasing along the whole trajectory of a Markov kernel that **fixes the uniform law**; it is #109's divergence form composed with the identity, with the maximum-entropy bound licensing the removal of `ENNReal.ofReal`. **AT `π′` the deliverable is unconditional but modest**, exactly as the row said it would be: ★★ `measureEntropy_recordLaw_le` — a `k`-context record string over `N + 1` codes carries at most `k · log (N + 1)` of Shannon entropy, whatever the preparation and whatever the context family. **THREE THINGS NOT CLAIMED, in the headers:** (a) **anything for a non-uniform reference** — entropy increases along a kernel fixing the *uniform* law, and for any other stationary law what is monotone is the divergence from *that* law (#109) and not the entropy; the two coincide only here, and **the Born record law is not uniform**, so the entropy H-theorem is **not** claimed for the record dynamics (this is the row's own caveat, confirmed); (b) anything beyond a finite type — differential entropy, countable types with infinite entropy, and the conditional and joint entropies are untouched; (c) any convergence or rate — monotone is not strictly increasing and a kernel can be the identity. *As opened (2026-10-04, out of #109):* **Shannon entropy at `π′`: the identity `klDiv q uniform = log card − H q`.** #109 proved the arrow in its divergence form, which is the general one, but the *entropy* form — "the macroscopic entropy increases" — needs two things it does not have: the identity above, and a reference law that is **uniform**. The brick is the identity: define Shannon entropy on a finite type as `∑ v, negMulLog (q {v})` (the corpus already uses `Real.negMulLog`, e.g. `LF6/Decoherence.lean`'s `sum_negMulLog_pos_of_two_nonzero`), and prove `klDiv q (uniform) = log (card) − H q` by computing the Radon–Nikodym derivative of one finite-type measure against another — which is where the work is, since Mathlib's `llr` is stated through `Measure.rnDeriv` and not through point masses. With it, #109's `antitone_klDiv_recordLaw` becomes a genuine H-theorem **whenever the stationary law is uniform**. ⚠️ **What it does NOT buy** is the Born case: the Born record law is uniform only for a flat preparation, so the entropy form is a *different* statement about `π′` and not a strengthening of #109 — #109's reference law is forced by Liouville, and this one is chosen. Worth having as Cat-1 (Mathlib has no classical Shannon entropy of a measure on a finite type). | **M–L** | Medium — the textbook form of the arrow, and Mathlib-shaped; it changes no CSD claim | #109; `Mathlib/InformationTheory/KlDivArrow.lean`; `LF6/Decoherence.lean`; `Mathlib.InformationTheory.KullbackLeibler.Basic` | > | ~~111~~ | Foundations | Records→spacetime | **DONE 2026-10-05 — the first object in the corpus that is *geometry* on records** ([`CV/CausalDistance.lean`](../CsdLean4/CV/CausalDistance.lean), 16 pins; M–L took M). `strictPrecedes E` is strict causal precedence as a relation chains run on — transitive because the order is **antisymmetric** (★ `isTrans_strictPrecedes`: a strict step cannot return) — and `causalDist E e₁ e₂` is **the length of the longest causal chain**, the supremum of `chainLengths`. It is well defined because ★★ `chainLengths_le`: a chain's events are **distinct** (a strict chain cannot repeat) and all lie between the endpoints in time, so there are at most `2 ^ K · (e₂.time + 1 − e₁.time)` of them — **the finiteness doing the work is the arena's**, not an assumption about the order, and ★ `causalDist_le` states the bound. ★★★ `causalDist_add_le` is **the reverse triangle inequality**, `causalDist e₁ e₂ + causalDist e₂ e₃ ≤ causalDist e₁ e₃` for `e₁ ⪯ e₂ ⪯ e₃`: the signature of Lorentzian geometry and the **discrete twin paradox** — a path by way of `e₂` is never longer than the longest path, so the longest chain is the *straight* one, and a metric satisfies the inequality the other way round. The proof is chain concatenation: the longest chains are **attained** (★ `causalDist_mem_chainLengths`, via `Nat.sSup_mem` on a bounded nonempty set) and `RelSeries.smash`ed. ★★ `causalDist_pos_iff`: the distance is **positive exactly on the strictly causally related pairs**, so it vanishes exactly on the spacelike-or-equal ones, with ★ `causalDist_self` and ★ `causalDist_eq_zero_of_not_causalPrecedes` the two ways of vanishing; and ★ `causalDist_le_of_causalPrecedes_right` is monotonicity into the causal future. **FOUR THINGS NOT CLAIMED, in the header:** (a) **the metric** — order plus *length* is not order plus *volume*, and the classical route from a causal order to a Lorentzian metric (Hawking–King–McCarthy and Malament for manifolds, "order plus number equals geometry" for causal sets) fixes the conformal factor from a **volume element**, which this corpus does not have on the event order — its measure lives on `Σ` and weights events by Born probability, which is not a spacetime volume; **that companion brick is M–L and still unnumbered**; (b) **any embedding** — that a discrete order with a volume is approximated by a Lorentzian manifold is a *programme* (sprinkling; the Hauptvermutung is open), so nothing says the event order is a spacetime or that `causalDist` approximates a proper time; (c) anything independent of the coupling graph — all of it is relative to a fixed `E`, exactly as `ST-1` is; (d) any continuum limit, curvature or dimension estimator. ⚠️ `causalDist` is `0` on unrelated pairs by construction (`sSup ∅ = 0` in `ℕ`), the usual convention for a Lorentzian distance, so it is **not** a metric and its vanishing does not imply equality — that is `causalDist_pos_iff`'s content, not a defect. *As opened:* **The longest chain as a Lorentzian distance on the record-event order.** `CV/RecordCausalOrder.lean` gives a partial order on record events (`isPartialOrder_causalPrecedes`) and row 98 gives a cone with a boost symmetry, but there is **no distance on the order** — no interval, no chain length, nothing (verified by inspection 2026-10-04). The brick is the causal-set distance: define the **length of the longest chain** between two events of the order, and prove the two things that make it Lorentzian rather than metric — ★★ the **reverse** triangle inequality `d(e₁,e₃) ≥ d(e₁,e₂) + d(e₂,e₃)` for `e₁ ⪯ e₂ ⪯ e₃` (the discrete twin paradox: the straight path is the *longest*), and that `d` vanishes exactly on the spacelike-or-equal pairs. Finite and discrete, so it fits the arena with no limit and no completion. ⚠️ **What it is NOT** is the metric: order plus *length* is not order plus *volume*, and the classical route from a causal order to a Lorentzian metric (Hawking–King–McCarthy, Malament for manifolds; "order plus number equals geometry" for causal sets) fixes the conformal factor from a **volume element**, which this corpus does not have on the event order — that is the companion brick (a volume/counting measure on `RecordEvent`, with interval cardinality and the compatibility theorems), **M–L, to be numbered when the author wants it**. And the embedding theorem those two would feed is a *programme* for discrete orders (the Hauptvermutung is open), not a theorem to formalise. | **M–L** | Medium–high — it is the first object in the corpus that is geometry on records rather than causal order on records, and it is provable as stated | `ST-1`; `ST-2`; #98; `CV/RecordCausalOrder.lean`; `specs/records-to-spacetime-scoping.md` §8 | > | ~~112~~ | QIT | QC chain (link 12) | **DONE 2026-10-06 — and it came with the error budget** ([`Mathlib/Analysis/CStarAlgebra/GroupCommutator.lean`](../CsdLean4/Mathlib/Analysis/CStarAlgebra/GroupCommutator.lean), 7 pins; S–M took S). ★ `norm_mul_sub_mul_comm_le` holds in **any normed ring**, not just a C\*-algebra: `‖VW − WV‖ ≤ 2‖V − 1‖·‖W − 1‖`, and the whole content is the identity `VW − WV = (V−1)(W−1) − (W−1)(V−1)` — **the commutator sees only how far the factors are from `1`**, which is why nearness to the identity is what makes it small. ★★ `CStarRing.norm_groupCommutator_sub_one_le` is the unitary form, `‖VWV⋆W⋆ − 1‖ ≤ 2‖V − 1‖·‖W − 1‖`, and **unitarity enters exactly twice**: to write `VWV⋆W⋆ − 1 = (VW − WV)(V⋆W⋆)` (`groupCommutator_sub_one_eq`), and to drop the trailing unitary from the norm (`norm_mul_mem_unitary`). ★★ `norm_groupCommutator_sub_one_le_two_mul_sq` is the form the recursion uses (`δ` in, `2δ²` out) and ★★ `norm_groupCommutator_sub_one_lt` is **the contraction**: for `0 < δ < 1/2` the commutator is *strictly* closer to `1` than its factors were. **AND THE ERROR BUDGET CAME WITH IT**, which #115 needs and the row did not price: ★ `CStarRing.norm_mul_sub_mul_le` — a product of unitaries is **stable**, `‖AB − A′B′‖ ≤ ‖A − A′‖ + ‖B − B′‖` — and ★★ `norm_groupCommutator_sub_groupCommutator_le`: **the commutator of two `ε`-approximations is a `4ε`-approximation of the commutator**, the adjoints moving by `norm_star`. Stated in a general unital C\*-ring rather than for matrices, so it specialises to any gate algebra. **THREE THINGS NOT CLAIMED, in the header:** (a) any gate count or algorithm — this is the shrinking lemma alone, with the recursion (#115), the net (#113), the decomposition (#114) and the `O(log^c(1/ε))` bound (#116) still separate rows; (b) optimality of the constants `2` and `4`, which are what the telescoping gives and which the Solovay–Kitaev exponent does not depend on; (c) anything at `δ ≥ 1/2`, where the bound is weaker than its input and says nothing — which is exactly why the algorithm needs a base accuracy from the net before it can recurse. *As opened:* **The group-commutator contraction — Solovay–Kitaev's shrinking lemma.** For unitaries `V, W`, `‖VWV⁻¹W⁻¹ − 1‖ ≤ 2‖V − 1‖·‖W − 1‖`: two gates near the identity have a commutator **quadratically** nearer it. That quadratic gain is the entire engine of the recursion. The proof is three lines of algebra — `VW − WV = (V−1)(W−1) − (W−1)(V−1)`, then unitary invariance of the operator norm to drop the trailing `V⁻¹W⁻¹` — so the only care needed is the norm scope (`Matrix.Norms.L2Operator`, as #98 needed). Cat-1, no CSD content. ⚠️ On its own it proves nothing about gate counts; it is one of the three inputs #115 consumes. | **S–M** | Medium — small, self-contained, and the lemma every later row leans on | #74; #73; `Mathlib/QuantumInfo/CliffordTUniversal.lean` | > | ~~113~~ | QIT | QC chain (link 12) | **DONE 2026-10-06 — by a shorter route than the row recorded** ([`Mathlib/QuantumInfo/CliffordTNet.lean`](../CsdLean4/Mathlib/QuantumInfo/CliffordTNet.lean), 7 pins; M took S–M). ★★ `IsCompact.exists_finite_net` is the general statement, in any metric space: **a compact set inside the closure of `D` is covered by finitely many `ε`-balls centred in `D`** — density gives one point of `D` per target, and compactness is what makes finitely many of them enough. ★ `isClosed_unitaryGroup`: the unitary group of any finite index type is closed, as the preimage of `{1}` under `U ↦ UU⋆` (Mathlib has `IsCompact.matrix` for entrywise-compact sets but nothing about the unitary group). ★★ `isCompact_su2Set`: **the determinant-one `2 × 2` unitaries are compact** — closed, and inside the closed unit ball because a unitary has operator norm `1` in the C\*-norm, the ball being compact since the space is finite-dimensional hence proper. Then ★★★ `exists_finite_cliffordT_net`, **the base case**: for every `ε > 0` there are **finitely many genuine Clifford+T words** such that every determinant-one unitary is within `ε` of one of them. **THE ROUTE IS NOT THE ONE THIS ROW RECORDED**, and the row was wrong rather than merely pessimistic: it proposed SU(2) as the continuous image of the unit sphere in `ℝ⁴` under the `su2` parametrisation, which needs that parametrisation's **surjectivity** onto the determinant-one unitaries — a statement the corpus does not have. Closed and bounded in a finite-dimensional space needs nothing new, so the sphere picture appears nowhere in the file. **FOUR THINGS NOT CLAIMED, in the header:** (a) **any construction** — the net comes from `elim_finite_subcover`, so nothing bounds its cardinality or its words' lengths, and that is not a gap to fill by the same argument but exactly **why** the recursion is needed on top of it (#115): the net is the `O(1)` base, not the algorithm; (b) any calibration of `ε₀` against #112's contraction threshold `δ < 1/2`, which is #115's business; (c) anything modulo phase — the net is for `det = 1`, as in #81; (d) `IsCompact` for the unitary group itself, which is not stated (only closedness, plus compactness of the determinant-one subset). *As opened:* **SU(2) is compact, so the Clifford+T words contain a finite `ε`-net.** Solovay–Kitaev's base case needs a *finite* word set within `ε₀` of every target. The corpus has density (#81 `exists_phase_mem_cliffordTLim`) but no net, because density alone gives one word per target. The brick: `su2 w x y z` on `w²+x²+y²+z² = 1` is the **continuous image of a sphere** in `ℝ⁴`, hence a compact set (`Metric.isCompact_sphere` + `IsCompact.image`), and a dense subset of a compact metric space has a finite `ε`-net by `IsCompact.elim_finite_subcover`. Cat-1. ⚠️ **Existence, not construction**: it gives no bound on the net's words, which is exactly why Solovay–Kitaev needs the recursion on top of it — the net is the `O(1)` base, not the algorithm. | **M** | Medium — the base case, and the first compactness statement about the gate group in the corpus | #74; #81; #80 | > | ~~114~~ | QIT | QC chain (link 12) | **DONE 2026-10-06 — IN FULL, and the two steps the row called missing both landed** ([`Mathlib/QuantumInfo/CommutatorDecomposition.lean`](../CsdLean4/Mathlib/QuantumInfo/CommutatorDecomposition.lean), 23 pins; L took L). ★★★ `exists_commutator_of_det_one`: **every determinant-one unitary is a group commutator of two unitaries within `√2·√‖U − 1‖` of the identity**, with the commutator *exact*. Read with #112 that is the Solovay–Kitaev step in both directions: a commutator of two `δ`-close unitaries is `2δ²`-close, and every `ε`-close unitary **is** the commutator of two `√(2ε)`-close ones. **THE PIECES.** ★★ `norm_su2_sub_one`: the distance to the identity is `√(2 − 2w)` with `w` the scalar part, and it needs **no eigenvalue computation** because `star M * M` is a *scalar* multiple of `1`, so the C\*-identity `‖M‖² = ‖M⋆M‖` settles it. ★★★ `skComm_eq`: **the exact commutator identity** — the commutator of the `x̂` and `ŷ` rotations by `φ` is the unit quaternion `(1 − 2s⁴, 2cs³, −2cs³, 2c²s²)`, by `su2_mul` alone, so the matrix work is quaternion algebra; the scalar part is the angle relation, **quartic in `s` hence quadratic in the angle**. ★★★ `exists_skComm_norm_eq`: every distance in `[0, 2]` is realised **exactly**, at factor angle `2 arcsin √(ε/2)`. **AND THE TWO STEPS THE ROW RECORDED AS MISSING:** ★★ `exists_su2_of_det_one` — **the `su2` parametrisation is onto the determinant-one unitaries** (the surjectivity #113 also found absent): for `det U = 1` the inverse is the adjugate and for a unitary it is the adjoint, so comparing them forces `U 1 1 = conj (U 0 0)` and `U 1 0 = −conj (U 0 1)`, exactly `su2`'s shape; and ★★ `su2_conj_pure` + ★★ `bisector_reflect` — conjugating by a `π`-rotation **reflects** the axis (`v ↦ 2(m̂·v)m̂ − v`) and the bisector reflection swaps two vectors **of equal length**, so no axis normalisation enters the assembly. **TWO SIMPLIFICATIONS FOUND:** the bisector works at any common length, and the one bad case — an axis exactly opposite the standard commutator's — is disposed of by ★ `star_groupCommutator`, because that axis belongs to the **adjoint** of the commutator, which is *the same pair in the other order*; so **no perpendicular-vector construction appears anywhere**, which was the part priced as fiddliest. **THREE THINGS NOT CLAIMED, in the header:** (a) **`V` and `W` are unitaries, not gate words** — the commutator is exact and the factors are rotations, nothing expresses them in a generating set, which is the division of labour the algorithm needs (the net #113 supplies words, the recursion #115 approximates these factors by them) and means this row alone implies **no gate count**; (b) determinant one and `2 × 2` only — a general unitary needs a phase, as in #81; (c) optimality of `√2`, lossy by design, and the factor bound's `0 ≤ φ ≤ π` range. ⚠️ Two pins are **shorter than the foundational triple** and are recorded as they are rather than padded: `conj_groupCommutator` needs `[propext, Quot.sound]` and `star_groupCommutator` needs `[propext]` alone. *As opened:* **The commutator decomposition: a near-identity rotation is a commutator of two `√ε` rotations.** | ~~**L**~~ done (L) | High within the split — nothing after it works without it | #74; #112; #113; #80 (`axisRot`, `su2_mul`); #115 | > | ~~115~~ | QIT | QC chain (link 12) | **SPLIT IN TWO 2026-10-06 — the ERROR half is DONE, the LENGTH half is NEW ROW 117** ([`Mathlib/QuantumInfo/SolovayKitaevStep.lean`](../CsdLean4/Mathlib/QuantumInfo/SolovayKitaevStep.lean), 5 pins; M–L, the error half took M). **The error half, in full.** The whole content is **why the exponent is `3/2` and not `1`**: replacing each factor of a commutator by a `δ`-approximation gives the naive bound `4δ` (#112's `norm_groupCommutator_sub_groupCommutator_le`), which **contracts nothing**, since the input error was already `δ`. The gain is that the commutator is **bilinear in the deviations from the identity**: ★ `norm_mul_sub_mul_le'` exposes the pairing (`AB − A′B′ = (A−A′)B + A′(B−B′)`), and ★★★ `norm_groupCommutator_sub_le` is **the sharp estimate**, `‖[V,W] − [Ṽ,W̃]‖ ≤ 4aδ(1 + a)` with `a` bounding all four distances to the identity — smaller than #112's by a **factor `a`**, and that factor is the whole algorithm. ★★ `norm_groupCommutator_sub_le_sqrt` is the same at #114's factor size `a = √2·√ε`, the `√ε·δ` shape, which at `δ = ε` is `ε^{3/2}`. ★★★ `skError_le_rpow` is the bookkeeping in **closed form**: `ε_{n+1} ≤ C·ε_n^{3/2}` gives `C²·ε_n ≤ (C²·ε_0)^{(3/2)^n}` by induction through `Real.rpow` — **doubly exponentially** small in the level count, which is what makes the gate count polylogarithmic, and the inequality #116 takes logarithms of. **THE LENGTH HALF IS NOT A FORMALITY**, which is why this splits rather than closes: `ℓ_{n+1} = 5ℓ_n` needs a notion of **word length** in the generating set and the corpus has none — `cliffordT` is `Submonoid.closure {H, T}`, matrices with no length function, and #113's net is a `Finset` of *matrices* rather than of words. That is **row 117**. **AND A FINDING, recorded there:** the quoted exponent `log 5 / log(3/2) ≈ 3.97` is a **cost-model choice**, not a mathematical fact about this gate set. The recursion inverts its two factors, and over the letters `{H, T}` a shortest word for `T⁻¹` is `T⁷`, so the per-level factor is `1+1+7+7+1 = 17` and the exponent is `log 17 / log(3/2) ≈ 6.99`; counting `T⁻¹` as one letter restores `5` and `3.97` — and **both generate the same submonoid** (`T⁷` is already in `closure {H, T}`), so the choice costs no density and changes only the length function. ⚠️ Also not claimed: any **constructed approximant** — `skError_le_rpow` is a statement about real sequences, saying what the recursion's error does *given* the recurrence, which the estimates above justify rather than implement. *As opened:* **The recursion: `ε ↦ Cε^{3/2}` at five times the length.** | ~~**M–L**~~ error half done (M) | Medium–high — the algorithm's error analysis | #74; #112; #113; #114; #116; #117 | > | ~~116~~ | QIT | QC chain (link 12) | **DONE 2026-10-06 — SOLOVAY–KITAEV WITH ITS GATE COUNT, AND WITH IT ROW 74** ([`Mathlib/QuantumInfo/SolovayKitaevCount.lean`](../CsdLean4/Mathlib/QuantumInfo/SolovayKitaevCount.lean), Category 1, 13 pins — 11 here and 2 in #114; **priced M, took L**). ★★★ `exists_word_approx_polylog`: **for every determinant-one `U` and every small enough `ε` there is a Clifford+T *word* within `ε` of `U` of length at most `K·log(1/ε)^c`.** **(1) Eliminating `n`**, which is what the row asked for: `skLevel b ε = ⌈log R / log(3/2)⌉` with `R = log ε / log b`, ★★ `skLevel_rpow_le` (that many levels reach `ε`) and ★★ `five_pow_skLevel_le` (five to that power is `≤ 5·R^c`), and `skExponent = log 5 / log(3/2)` bracketed by ★ `three_lt_skExponent`/★ `skExponent_lt_four` — `c ∈ (3, 4)`, the literature's `≈ 3.97`, each by **one** application of `Real.log_lt_log` to `(3/2)³ = 27/8 < 5 < 81/16 = (3/2)⁴`, so no decimal is claimed. **(2) THE PAIRING, and the row's own estimate of it was wrong.** The row said "pure real analysis on logarithms, no new quantum content"; the logarithms were indeed cheap, but the pairing #117 declined — *which* word approximates a given `U` — is the Solovay–Kitaev algorithm's correctness, and it needed two things the corpus did not have. ★★★ `exists_mem_skWords_norm_sub_le` is the recursion: approximate `U` at level `n`, take the residual `Δ = U·(ctEval u)⋆`, write it as a group commutator of two near-identity determinant-one factors (#114), approximate *those* one level down, reassemble. It lands in `skWords` — hence inherits #117's length bound untouched — because ★★ `star_ctEval` identifies #117's `ctInvWord` with the **adjoint**, so a group commutator of word values *is* the value of `v ++ w ++ ctInvWord v ++ ctInvWord w`. **The two new pieces.** ★ `ctEval_mem_unitary` (via `hGateM_mem_unitary`, `tGateM_mem_unitary` — absent from the corpus until now), which is what makes right multiplication an isometry so the residual is as close to `1` as the approximation is good. And **the determinant**, the one step that is not formal: #114 needs a determinant-one input and a Clifford+T word is not one. It is *forced* — ★ `det_ctEval_pow_eight` (the determinant of a word is an eighth root of unity, because `H² = 1` and `T⁸ = 1` and nothing else enters) with ★★ `eq_one_of_pow_eight_eq_one` (**an eighth root of unity within `1/8` of `1` is `1`**, needing no root-of-unity theory — only `(z−1)·∑_{k<8}zᵏ = z⁸−1 = 0` and `‖1−zᵏ‖ ≤ k‖1−z‖` on the unit circle, which bounds the sum by `28/8 < 8`). **#114 was strengthened in place** rather than duplicated: `axisRot_det` and `det_conj_of_mem_unitary` discharge both of its branches, so ★★★ `exists_commutator_of_det_one` now records `V.det = 1 ∧ W.det = 1` — which its construction always produced — and that is exactly what lets the recursion re-enter itself on the factors. ⚠️ **Honest scope, all in the header.** `K` and `ε₀` are **existential by inheritance**: #113's net comes from compactness with no bound on its size, and the radius forcing the residual's determinant comes from continuity of `det`. Neither is quantitative, so no constant here is — the *exponent* is. The exponent is #117's **cost model** (`H`, `T`, `T⁻¹`; over two letters a level costs `17` and the exponent is `≈ 6.99`), free by #117's ★★ `mem_cliffordT_iff_exists_word`; the literature's `3 + δ` needs a different net and **stays unclaimed**. And **nothing is computed**: the word comes from a recursion over an existence statement, not an algorithm, and `skErr`'s constant `33` is sufficient, not best. **Link 12: this discharges tier C's #74, so #62 is all that remains of the strand.** | ~~**M**~~ done (took **L**) | **High — it was the last row of link 12** | #74; #115; #117; #113; #114; #73 | > | ~~117~~ | QIT | QC chain (link 12) | **DONE 2026-10-06 — the cost model is faithful AND free** ([`Mathlib/QuantumInfo/CliffordTWords.lean`](../CsdLean4/Mathlib/QuantumInfo/CliffordTWords.lean), Category 1, 8 pins; M–L took M). **The alphabet is three letters** — `ctGen = ![H, T, T⁷]` with `ctInvLetter = ![0, 2, 1]` — and `ctEval w = (w.map ctGen).prod` evaluates a word, multiplicatively (`ctEval_append`) and into the submonoid (★ `ctEval_mem_cliffordT`). ★★ `mem_cliffordT_iff_exists_word` is **the answer to #115’s cost-model question**: the values of words are *exactly* `cliffordT`, because `T⁻¹ = T⁷` was already in `closure {H, T}` (the submonoid `ctValues` of word values contains both generators, so it contains their closure, and `ctEval` lands in it either way). **So the inverse-closed model costs no density and changes only the length function** — the `5` is bought, not assumed. `ctInvWord w = (w.map ctInvLetter).reverse` with ★★ `length_ctInvWord` (`= w.length`) and ★★ `ctEval_ctInvWord` (`ctEval (ctInvWord w) * ctEval w = 1`, by induction through `ctGen_invLetter_mul`, whose `T` cases are `T⁷·T = T⁸ = 1`): **inverting a word is free, at the same length**, which is the whole content of the per-level factor `5`. ★★★ `exists_word_net` restates #113’s net as a `Finset` of **words** with a common length bound `ℓ₀` (`Finset.image wordOf` of #113’s matrix net, `ℓ₀ = Finset.sup List.length`; `wordOf` through `mem_cliffordT_iff_exists_word`) — the base case the recursion starts from. Then `skWords F n`, the words of the Solovay–Kitaev shape `v w v⁻¹ w⁻¹ a` with all five parts at level `n`, and ★★★ `length_le_of_mem_skWords`: **length `≤ 5ⁿ·ℓ₀`.** Five parts, each inverse free, so the factor is exactly five — the row’s `ℓ_{n+1} = 5ℓ_n` in closed form. **A route finding worth recording:** the row asked for `FreeMonoid`, and that was the wrong carrier. `FreeMonoid.ofList`/`toList` are `Equiv.refl`s that survive `simp` as visible noise, and a `Fin 3 → _` defined by pattern match will not reduce against numerals, so `ctGen 0 = H` was not `rfl`. Plain `List (Fin 3)` plus `Matrix.cons` vector notation made every proof a standard `List` lemma (`prod_append`, `reverse_cons`, `length_map`); nothing was lost, since the monoid hom was never used. ⚠️ **Honest scope, all three in the header.** `skWords` is the **shape** of the recursion, not the algorithm: the theorem bounds the length of words of that form and says nothing about *which* one approximates a given `U` — that pairing is #116. The bound is an **inequality, not a count**: nothing claims `5ⁿ·ℓ₀` is attained or that these are shortest words for their values. And nothing about `ℓ₀`’s **size** — `exists_word_net` inherits #113’s existence-only net, with no bound in terms of `ε₀`. | ~~**M–L**~~ done | Medium–high — #116 now has an `ℓ` to bound | #115; #113; #116; #73; #81 | > | ~~93~~ | QIT | Library owed / upstream | **CONTINUUM COMPANION LANDED 2026-10-02:** this row's `LinearPMap.IsResolventAt`/`resolventSet`/`spectrum` layer now carries the continuum case as well — `Analysis/InnerProductSpace/MultiplicationOperator.lean` (self-adjointness and the spectrum of multiplication by a real measurable function on `L²(μ)`, where no basis is available) and `Analysis/Semigroup/FreeHamiltonian.lean` (the free Hamiltonian, spectrum `[0, ∞)`); see #64 for what the interacting case still waits on. ~~*new (2026-09-30)* **The twisted Laplacian on `L²(S¹)` as a self-adjoint operator, and its spectrum**~~ **CLOSED 2026-09-30, all three items.** (a) [`DiagonalOperator.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/DiagonalOperator.lean) (484 lines, 16 pins): the diagonal operator of a real weight family on a Hilbert basis is densely defined, ★★★ `isSelfAdjoint_diagOp`, and ★★★ `spectrum_diagOp` = the closure of the weight set — with `LinearPMap.IsResolventAt`/`resolventSet`/`spectrum` **defined** there, because Mathlib's `spectrum` is for algebra elements. (b)(c) [`CircleSobolev.lean`](../CsdLean4/Mathlib/Analysis/Fourier/CircleSobolev.lean) (325 lines, 14 pins): ★★ `fourierCoeffOn_deriv` — **`cₙ(f') = (2πin/T) cₙ(f)`** for a periodic `f`, the bridge the pin had only in boundary-term form (`fourierCoeffOn_of_hasDerivAt`, whose boundary term periodicity kills; the `n = 0` case is the fundamental theorem of calculus) — and ★★★ `fourierCoeffOn_twistedLap`: **`−(∂ − iΦ)²` is multiplication by `(n − Φ)²` in Fourier coordinates**, which is the eigenvalue equation in the form spectral theory needs. Then `twistedOp Φ = fourierBasis.diagOp (circleEigval Φ)` with ★★★ `isSelfAdjoint_twistedOp`, ★★★ `spectrum_twistedOp`, and ★★★ `exists_eq_of_spectrum_twistedOp_eq` — **the flux is determined by the spectrum of a self-adjoint operator**, which is #90's theorem with *spectrum* in place of *level set*. **`H²(S¹)` NEEDED NO CONSTRUCTION:** it is `diagDomain` — the vectors whose weighted coefficients stay square-summable — so (b)'s first half was a definition, not a build. **THE LAST MILE, WHICH THE ROW DID NOT ASK FOR AND WHICH MAKES IT HONEST:** ★★★ `twistedOp_toL2` — the operator acts on **every `C²` periodic function** exactly as the differential operator does, via `toL2`/★ `fourierCoeff_toL2` (`AddCircle.liftIoc` + `ContinuousMap.toLp`) and ★ `periodic_deriv` (the derivative of a periodic function is periodic, absent from the pin). Honest scope: the identification is on `C²` functions and on any `L²` pair with matching coefficients, **not** on every element of the domain — that is the Sobolev embedding `H²(S¹) ↪ C¹`, which the pin lacks and nothing here needs. | ~~**L**~~ done ((a) M–L, (b) M, (c) S as priced) | Medium — (a) is the reusable brick; (b)(c) buy #90 the vocabulary, and the corpus now says *spectrum* without a caveat | #90 | > | ~~80~~ | QIT | QC chain (link 12) | ~~*new (2026-09-25)* **`SU(2)` in axis–angle form, and the word's own dense circle (`R-005` (b1)).**~~ **DONE 2026-09-25** ([`Mathlib/QuantumInfo/SU2Rotation.lean`](../CsdLean4/Mathlib/QuantumInfo/SU2Rotation.lean), Category 1, 10 pins; M took M): the unit-quaternion parametrisation `su2 w x y z` as a `2 × 2` matrix with ★ `su2_mul` (the quaternion product, by entries), `su2_det = w² + x² + y² + z²`, `su2_trace = 2w`, `su2_one`; `axisRot a b c θ` with ★ `axisRot_add` (a one-parameter group for a unit axis — the scalar part needs `a² + b² + c² = 1`), `axisRot_nat_mul`, `axisRot_add_int_mul_four_pi` (period `4π`: a `2π` rotation is `−1`), `continuous_axisRot` (`continuous_matrix`); ★★ `htht_eq_su2` — **the Clifford+T word in quaternion coordinates** `e^{iπ/4}·su2(cos²(π/8), √2/4, (2−√2)/4, √2/4)`, radical-free beyond `√2`; the unit axis through the *products* `htSin·htA = √2/4`, `htSin·htB = (2−√2)/4` with `htSin² = (10 − 4√2)/16` (`Real.sin_arccos`), so ★ `htAxis_unit` and ★ `htht_eq_axisRot`: the word is `e^{iπ/4}·R_n̂(htAngle)` about `n̂ = (1, √2−1, 1)/‖·‖`, with `htht_pow` for its powers; ★ `irrational_htAngle_div_four_pi` and ★★ `mem_closure_range_axisRot` — **every rotation about the word's own axis is a limit of integer multiples of its angle**, hence (up to phase) of its powers (`image_closure_subset_closure_image` + `AddSubgroup.mem_closure_pair` + the `4π` period). | ~~**M**~~ done | Medium | `R-005`; #68; #81 | > | ~~81~~ | QIT | QC chain (link 12) | ~~*new (2026-09-25)* **From one dense circle to `SU(2)`: the orthogonal-axis reduction and the Euler decomposition (`R-005` (b2)).**~~ **DONE 2026-09-26** ([`Mathlib/QuantumInfo/CliffordTDensity.lean`](../CsdLean4/Mathlib/QuantumInfo/CliffordTDensity.lean), Category 1, 14 pins; L took L). ★★ `det_one_mem_cliffordTLim`: **every determinant-one `2 × 2` unitary is a limit of Clifford+T words**, hence ★★ `exists_phase_mem_cliffordTLim` — **`⟨H, T⟩` is dense in `U(2)` modulo phase.** The obstruction that shaped the proof: **no two axes in the Clifford orbit of the word's axis are orthogonal** (the possible dot products are `±htB²`, `±(1 − 2htB²)`, `±htA²`, `±(htA² ± 2htA·htB)`, none zero), so a third axis must be manufactured — and the mechanism is the cross product. ★ `axisRot_htAxis_pi_mul_hGateM`: the `π` rotation about the word's axis `(htA, htB, htA)` (available because #80's circle is dense and the closure is closed), followed by `H`, is a rotation about `(1, 0, −1)/√2` — the `y` components cancel because the axis has equal `x` and `z` — through `wAngle` with `2 cos wAngle = (6 + 16√2)/17`, from the closed forms `htA² = (5 + 2√2)/17` and `htB² = (7 − 4√2)/17`. **#68's cheap irrationality test cannot see that angle** — `x² + x = q` needs an integer linear coefficient and here the minimal polynomial is `X² − (12/17)X − 28/17` — so the row also proves the general form ★ `not_isIntegral_of_quadratic`: **a monic rational quadratic whose linear coefficient is not an integer has no algebraic-integer root** (the minimal polynomial of an algebraic integer has integer coefficients, and for an irrational root that quadratic *is* the minimal polynomial), whence ★★ `irrational_wAngle_div_two_pi`. Then ★ `axisRot_mem_of_irrational` (one irrationally-angled rotation fills its circle inside any closed submonoid containing it and its inverse) gives the circles about the word's axis (★★ `axisRot_htAxis_mem`, through `htht⁸` which is phase-free since `tPhase⁸ = 1`) and about `(1, 0, −1)/√2` (★★ `axisRot_wAxis_mem`, through the fourth power which is phase-free since `i⁴ = 1`); conjugating the latter by `Z = T⁴` reflects its axis to the **orthogonal** `(1, 0, 1)/√2` (★★ `axisRot_uAxis_mem`). Finally ★★ `exists_euler_two_axes` transports #84's `z`-`y`-`z` decomposition to that pair by two conjugations (`R_z(π/2)` turns `y` into `x`, `R_y(π/4)` tilts `z` and `x` onto the two axes). **Tier A of the link-12 order is complete: only #73 remains before `R-005` closes.** | ~~**L**~~ done | Medium | `R-005`; #80; #84; feeds #73 | > | ~~75~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **The combinatorics of the `[[15, 1, 3]]` code (`R-004` (a)).**~~ **DONE 2026-09-23** ([`Mathlib/QuantumInfo/ReedMuller15.lean`](../CsdLean4/Mathlib/QuantumInfo/ReedMuller15.lean), Category 1, 4 pins; S–M took S–M): `col j` = the binary digits of `j + 1` (the nonzero vectors of `𝔽₂⁴`; `col_ne_zero`, `col_injective` by `decide`), `syndrome e = H e`, `weight`, `support`, `ind`; `syndrome_ne_zero_of_weight_one/two` — no undetected pattern of weight `1` or `2`; ★ `card_undetected_weight_three` — **exactly `35` undetected patterns of weight `3`** (bijection with the 3-subsets whose columns sum to zero, `card_undetectedTriples = 35` by `decide` over the `455` triples); ★ `dotp_add_one` — **the parity rule** `e · (x + 𝟙) = e · x + |e|`: an undetected pattern acts on the logical `1̄` with the sign `(−1)^{|e|}`, so odd weight is the logical `Z̄`; ★ `card_undetected` — **the Hamming code has `2¹¹` words** (`syndrome` surjective through the explicit `preimage` on the unit columns `j = 0, 1, 3, 7`, rank–nullity over `ZMod 2`, `Module.finBasis`). | ~~**S–M**~~ done | Medium — the `35` of `35 p³` | `R-004`; #76; #78 | > | ~~76~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **Transversal `T` on the `[[15, 1, 3]]` code (`R-004` (b)).**~~ **DONE 2026-09-23** ([`Mathlib/QuantumInfo/ReedMuller15Code.lean`](../CsdLean4/Mathlib/QuantumInfo/ReedMuller15Code.lean), Category 1, 5 pins; M took S–M): the logical states `|0̄⟩ = ∑_{x ∈ C}|x⟩`, `|1̄⟩ = ∑_{x ∈ C}|x + 𝟙⟩` on `QReg 15` (`logical0`, `logical1`, `C` = the `[15, 4]` simplex code `codeword a = ∑ᵢ aᵢ · row i`); `weight_codeword` (`0` or `8`), `weight_codeword_add_one` (`7` or `15`) and `codeword_add_row` by `decide`; ★ `pauliOp_row_logical0/1` — the `X`-stabilisers `X^{row i}` fix both logical states (`X^a|x⟩ = |x + a⟩`, the codewords permuted); `tTrans` — `T^{⊗15}` as the coordinate operator `|x⟩ ↦ e^{iπ|x|/4}|x⟩`, `tTrans_eq_comp` — **it is the composite of the fifteen single-qubit `tGate`s** (`Fin.sum_univ_def` on the explicit `List.finRange 15`); ★★ `tTrans_logical0`, `tTrans_logical1` — **`T^{⊗15}` fixes `|0̄⟩` and multiplies `|1̄⟩` by `e^{−iπ/4}`: the transversal `T` is the logical `T†`** (`e^{iπ·7/4} = e^{iπ·15/4} = e^{−iπ/4}`); ★ `tTrans_logicalPlus` — the encoded magic state `T^{⊗15}|+̄⟩ = |0̄⟩ + e^{−iπ/4}|1̄⟩`. The code space is presented by its logical states; the stabiliser-average identification is not needed by #77–#78. | ~~**M**~~ done | Medium — why this code and not Steane | `R-004`; #75; #77 | > | ~~77~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23; restated 2026-09-23)* **`Z`-error patterns on the encoded magic state (`R-004` (c)).**~~ **DONE 2026-09-23** ([`Mathlib/QuantumInfo/ReedMuller15Errors.lean`](../CsdLean4/Mathlib/QuantumInfo/ReedMuller15Errors.lean), Category 1, 7 pins; M took M): `syndromeF e i = e · row i` (definitionally the `ZMod 2` syndrome of `ReedMuller15.lean`, `syndromeF_eq` is `rfl` — `ZMod 2` is `Fin 2`), `parity e = e · 𝟙` with `parity_val : (parity e).val = wt e % 2` and `parity_eq_one_iff_odd`; `pauliOp_z_comm_row` — `Z_e X^{row i} = (−1)^{(syndrome e)ᵢ} X^{row i} Z_e`; ★ `measProj_row_z` — **the `X`-check `i` on `Z_e ψ̄` is deterministic with outcome `(syndrome e)ᵢ`** (`meas_deterministic` of `Stabilizer.lean`), so `exists_reject_of_syndrome_ne_zero` (a detected pattern is rejected with certainty) and `measProj_row_z_zero` (an undetected one passes every check); `bdot_codeword`, `bdot_codeword_eq_zero` (a pattern orthogonal to the rows is orthogonal to the code); ★ `pauliOp_z_logical0/1` — an undetected pattern fixes `|0̄⟩` and multiplies `|1̄⟩` by `(−1)^{|e|}`; ★★ `pauliOp_z_encodedMagic` — **the accepted state is `|0̄⟩ + (−1)^{|e|} e^{−iπ/4}|1̄⟩`**, the encoded `T†|+⟩` for even `|e|`, its logical `Z̄`-flip for odd; `magicConj = T†|+⟩`, `sGate_magicConj : S T†|+⟩ = |A⟩`, ★ `outQubit_eq` — the logical coefficients read as one qubit are `√2 · Z^{|e| mod 2} T†|+⟩`: **the protocol distils the magic state up to the Clifford `S`, with a `Z`-error exactly when the undetected pattern has odd weight.** | ~~**M**~~ done | Medium | `R-004`; #76; #66; #78 | > | ~~78~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **The 15-to-1 distillation bound and its recursion (`R-004` (d), closes `R-004`).**~~ **DONE 2026-09-24 — `R-004` CLOSED** ([`Mathlib/QuantumInfo/ReedMuller15Distill.lean`](../CsdLean4/Mathlib/QuantumInfo/ReedMuller15Distill.lean), Category 1, 7 pins; M–L took M): the pattern measure is fifteen copies of a Bernoulli `ν` with `ν {1} = p`, `patternMeasure_singleton : μ {e} = p^{|e|}(1 − p)^{15 − |e|}` (`Measure.pi_pi` on `Set.univ_pi_singleton`, `prod_ite` + `compl_filter`); ★ `measure_acceptSet_ge` — **acceptance `≥ (1 − p)^{15}`** (the zero pattern is accepted); ★ `measure_wrongSet_le` — **accepted-and-wrong `≤ 35 p³ + 2¹¹ p⁵`**: an accepted wrong pattern has odd weight `≥ 3`, so it lies in the `35` of weight `3` (#75) or among the `≤ 2¹¹` of weight `≥ 5`, each weighted by `p^{|e|}` (`sum_measure_singleton`, `measure_union_le`); ★★ `distillation_error_le` — **the 15-to-1 bound** `P(wrong \| accept) ≤ (35 p³ + 2¹¹ p⁵)/(1 − p)^{15}` (`ProbabilityTheory.cond_apply` on the accept set), Bravyi–Kitaev's `35 p³` at leading order; ★ `distillation_error_le_ninety` — `≤ 90 p³` for `p ≤ 1/20`; the recursion `distillIter p k` (`p ↦ 90 p³`) with ★ `distillIter_le` (`p_k ≤ p (90 p²)^k`) and ★★ `tendsto_distillIter` — **below `p ≤ 1/20` the iterated error rate tends to `0`**. `discharged_by` of `R-004`. The `1/20` threshold is a sufficient condition, not Bravyi–Kitaev's exact fixed point `≈ 0.141`; the recursion assumes the output errors of a round are again independent (module header). | ~~**M–L**~~ done | Medium–high — the distillation statement the doc names | `R-004`; #75; #77 | > | ~~79~~ | QIT | QC chain (link 12) | ~~*new (2026-09-23)* **The decoder as an explicit Clifford circuit.**~~ **DONE 2026-09-27** ([`Mathlib/QuantumInfo/ReedMuller15Decoder.lean`](../CsdLean4/Mathlib/QuantumInfo/ReedMuller15Decoder.lean), Category 1, 12 pins; M took M). **Sixteen `CNOT`s, no Hadamard.** The `[15, 4]` simplex code has a *linear* logical-`Z` readout — columns `1`, `2`, `3` sum to zero, so `x₀ + x₁ + x₂` vanishes on the code and is `1` on its all-ones coset (`logicalBit_codeword`, `logicalBit_add_one`) — so `CNOT(1 → 0)` and `CNOT(2 → 0)` compute the logical bit onto wire `0` and `CNOT(0 → j)`, `j = 1 … 14`, clear it from the rest (`decoderList`, `decoderList_length`, unitary by `decoderCircuit_mem_unitaryGroup`). The general layer first: ★ `cnotGate'_apply` identifies #83's matrix as the permutation matrix of the label action `cnotAct` (the shape of `xGate_apply`), ★ `cnotGate'_mulVec_apply` reads a `CNOT` on amplitudes as a pullback, and ★★ `cnotListMat_mulVec_apply` iterates that over a gate list. Then ★ `cnotListPull_decoderList` computes this circuit's pullback as `decodeLabelInv`, whence ★★ `decoderCircuit_mulVec_apply`, **every decoded amplitude**: `(Dψ)(z) = ψ (decodeLabelInv z)`, with `decodeLabel` the inverse (★ `decoderCircuit_mulVec_basisState`, `D|x⟩ = |decodeLabel x⟩`). ★★ `decoderCircuit_logical_update`: **wire `0` carries the logical qubit** — for every label of the other fourteen wires the amplitude in the `1`-sector is `λ` times the one in the `0`-sector, the same ratio whatever they carry, which is why measuring them leaves wire `0` alone (not vacuous: `decoderCircuit_logical_apply_zero`). The mechanism is `decodeLabelInv_update_one`: the `1`-sector pulls back to the coset `C + 𝟙` where `|1̄⟩` lives, the `0`-sector into the kernel of the readout where `|1̄⟩` has no amplitude (`logical1_apply_of_logicalBit_zero`). ★★★ `decoderCircuit_encodedMagic` and `decoderCircuit_encodedMagic_apply`: **the decoded output is #77's `outQubit e`**, i.e. `√2 · Z^{|e|} T†|+⟩`, the magic state up to the Clifford `S`. Honest scope in the header: the fourteen residual wires are not exhibited as a tensor factor (the uniform ratio is that statement in coordinates), no measurement is modelled, and the corpus has no `IsClifford` predicate — the Clifford claim is carried by the explicit gate list. | ~~**M**~~ done | Low — the protocol as a circuit | `R-004`; #78 | > | ~~140~~ | Thermo / Records | Statistical-mechanics layer · Records→spacetime | *new (2026-10-12, out of the scope note #103 and #105 share)* **DONE 2026-10-12 — and the answer is negative, which is the finding** ([`RecordLayer/TypicalMajority.lean`](../CsdLean4/RecordLayer/TypicalMajority.lean), 6 pins; S–M took S–M). #103's majority half is conditional on `1 − ε ≤ c.rate p i`, and its scope note says of the Fubini–Study statistics that “those are statements about **base statistics** … and **no theorem joins them** to the fibre bound proved here”; #105 then supplied the hypothesis *geometrically* and recorded in prose that “for a generic superposition no cell is overwhelming, and that is correct physics rather than a gap”. **This row prices that prose.** ★★ `fsMeasure_momentMap_dev_le` specialises `fs_chebyshev_concentration` from a *linear* statistic to **one Born weight**, which is the form the record layer consumes since a basin is indexed by one outcome. ★★★ `fsMeasure_momentMap_ge_le` — the preparations whose outcome `i` carries weight `1 − ε` or more have Fubini–Study measure at most `(N−1)/(N²(N+1)(1−ε−1/N)²)`, so `O(1/N²)` at fixed `ε`. ★★★ `fsMeasure_exists_momentMap_ge_le` — the union bound over the `N` outcomes costs one factor `N`, so **typically no outcome at all is a majority**, at `O(1/N)`. ★★★ `fsMeasure_exists_majority_basin_le` — **the join, in #103's own vocabulary**: the preparations at which *some* basin of `momentContext` carries epistemic measure `1 − ε` or more have Fubini–Study measure at most `1/(N(1−ε−1/N)²)`, and nothing mediates, because `globalBasin_prob` makes a cell's measure *exactly* its rate. ★★ `tendsto_majorityBound` — and the price tends to `0`, so the hypothesis becomes arbitrarily untypical as the arena grows. **So base typicality does not supply #103's hypothesis — it argues against it, and that is why #105 had to be geometric.** ⚠️ **Not a refutation of anything.** #105's hypothesis is about a *particular* preparation, and a laboratory preparation is prepared rather than sampled: `measure_robustBasin_ge_of_pointer` exhibits the pointer state itself, so the exceptional set is **nonempty at every `N`**. ⚠️ **The Born weights are untouched** — `globalBasin_prob` is exact at every base point; what is atypical is *one cell dominating*, not the cell law. ⚠️ **`momentContext` only.** `CellLawFreedom.lean`'s rival rate field is not covered, since the Chebyshev input is about the moment map specifically; nor is any record *string* of several contexts, exactly as in #103 and #105. ⚠️ **Nothing about the fibre**: `measure_recordString_ne_le` is a statement about writes and is still unjoined — the half joined here is the *majority* half, the one whose hypothesis was unsupplied. ⚠️ **Polynomial because its input is**; a Lévy bound would sharpen every constant and stays inside #24. | ~~**S–M**~~ took S–M | Medium — it closes the sentence #103 and #105 both end on, and it tells the programme where the majority hypothesis can and cannot come from | #103, #105, `Thermo/CanonicalTypicality.lean` | > | ~~141~~ | Docs / Positioning | Reviewer-proofing · Statistical-mechanics layer | *new (2026-10-12)* **DONE 2026-10-12** ([`CSD-CHARTER.md`](CSD-CHARTER.md) and [`docs/glossary.yaml`](../docs/glossary.yaml); S took S) — **the corpus proves three regimes of “random” and published two.** The charter's one sentence on the subject is “all indeterminacy is epistemic — ignorance of exactly *where* in a prepared region Ω₀ the microstate is”, which covers (i) the **ontic** level, where there is none, and (ii) the **quantum operational** level, where the Born weights come from that ignorance. (iii) **Macroscopic effective stochasticity** has been a theorem since #100 (2026-10-04), and the words “coarse”, “Markov” and “macroscopic” appeared **nowhere** in the charter, nor `CoarseMarkov`/`MicroDecoupled`/effective stochasticity anywhere in the glossary's 152 entries. **What landed.** A ▸ *Three regimes of “random”* section in the charter naming all three with their anchors (`Measurement.bornMeasurement_frequency`; ★★★ `abs_measure_historyEvent_sub_markov_le` with `MicroDecoupled` as the hypothesis nothing supplies; `cex_not_microDecoupled` for its load-bearingness), plus the **negative** half: **chaos is not where the Born weights come from** — the hitting-time route was examined and abandoned as *mis-specified* rather than blocked ([`q12-fibre-mechanism-scoping.md`](q12-fibre-mechanism-scoping.md): rare sets against cells that sum to one, asymptotic against `measure_raceCell`'s exact value, one trajectory against `n` clocks) — the corpus's own flows **cannot** mix (`not_hasCorrelationDecay_of_compactGroup`, instantiated by `not_hasCorrelationDecay_blockPop_of_unitary` and `not_hasCorrelationDecay_kFlow`, so equilibration runs through dephasing and ETH), and **that is a fact about the class of maps isolated QM uses, not about `Σ`** (`torusDouble_hasCorrelationDecay` on the very same fibre), with #101 open and unclaimed. The glossary half **extends `where-do-probabilities-come-from`** rather than adding a parallel entry, because the third regime is an answer to that entry's own question and the entry already closes the time-average route. ⚠️ **(iii) adds no new kind of indeterminacy** — it is still ignorance of the microstate, read through a coarser variable, so the charter's sentence stands unqualified. ⚠️ **No Lean changed and nothing new is claimed**: every statement is an existing theorem, cited by name. | ~~**S**~~ took S | Medium — a reader of the charter could not find the third regime, and “chaos ⇏ Born” lived only inside a closed backlog cell | #100, #101, #140, `q12-fibre-mechanism-scoping.md` | > | ~~142~~ | Docs | Library owed / upstream | *new (2026-10-12, out of the `v4.34.0` pin move)* **Collapse the `MapProbability` compat shims and delete the file.** [`Mathlib/MeasureTheory/MapProbability.lean`](../CsdLean4/Mathlib/MeasureTheory/MapProbability.lean) exists because Mathlib master (2026-08-29) reworked `Measure.map`'s junk value: it removed the named theorem for “the pushforward of a probability measure is a probability measure” in favour of an unconditional **instance**, and gave `Measure.map_smul` a measurability hypothesis. The file's own note said the shims collapse “when the pin advances past the master change” — **the pin advanced on 2026-10-12**, so upstream now carries both. ~~What is left is mechanical but not trivial~~ ▸ **DONE 2026-10-13, and the file is gone** (30 files touched, 69 insertions against 154 deletions; S–M took S–M). The measured counts were **30 occurrences of `isProbabilityMeasure_map'` across 22 files** and **11 of `Measure.map_smul'` across 10** — the row's “34 in 23” and “eight” were both off, which is the third time this pin-move table has drifted and is why it measures itself now. `isProbabilityMeasure_map' h` became `inferInstance` with the measurability argument simply dropped; `Measure.map_smul' _ _ h` became `Measure.map_smul _ h.aemeasurable`, upstream taking `AEMeasurable` where the shim took `Measurable`. **The row's prediction about `show` was right, and there were five of them, not one**: the instance cannot see through a `def`, so `coarseLaw`, `gaussianVec`, `phasePrep`, `recordLaw` and `joinPrep` each need their definition unfolded first. ⚠️ **FOUR THEOREMS LOST A HYPOTHESIS, WHICH IS A STRENGTHENING AND NOT A CONVENIENCE.** `isProbabilityMeasure_coarseLaw`, `isMarkovKernel_coarseKernel`, `isMarkovKernel_macroKernel` and `isProbabilityMeasure_projectiveLaw_flow` each took a `Measurable` hypothesis *only* to feed the shim; upstream's instance is unconditional, so each now holds for an arbitrary map. The hypotheses were **removed rather than renamed to `_h`**, because a vestigial argument is a worse record than a stronger statement — every call site was in the same file, so the ripple was four lines, and the three of them that are axiom-pinned kept their names and so their pins. Where a dropped hypothesis was the only mention of an implicit `q` or `F`, the call sites name it with `(q := …)`. ⚠️ 22 now-unused imports of the shim module went with it, and `scripts/pin-move-inventory.sh` section (5) reads **zero**. ⚠️ **Nothing is wrong meanwhile**: the shims are proved from the stable `map_apply_of_aemeasurable` API and compile against both pins, so this is tidying, not a fix — which is why it was kept out of the bump rather than mixed into a 220-file commit. ⚠️ Four of the sites are in Cat-1 files that `check-mathlib-absence` scans; the deletion leaves that guard green, and `check-import-hygiene` too. | **S–M** | Low–medium — it removes a shim that now shadows upstream, and a reader of either name currently has to work out which of the two is load-bearing | #27 (HY-6 section (5)), `scripts/pin-move-inventory.sh` | > | ~~143~~ | Docs | Reviewer-proofing | *new (2026-10-12, surfaced by the `v4.34.0` bump)* ~~**31 glossary entries are flagged stale by a rename, and the staleness signal needs a decision rather than 31 re-reads.**~~ ▸ **DONE 2026-10-13 on the author's route (a), and the count in this row was wrong: 38 ENTRIES across 31 MODULES.** The row conflated the two, which is worth recording because the decision was priced on it. `check-glossary`'s staleness check asks whether the module an entry anchors to has changed since the entry was reviewed, and the bump touched 220 Lean files, so 31 entries now read as stale. **Measured rather than assumed**: of the 31 modules, **28 are byte-identical to their pre-bump content once the `if_pos`/`ite_eq_left` family rename is undone** — nothing a reader of those entries could be reading changed — and **three changed in another way**, all of it proof-side: `Mathlib/LinearAlgebra/Matrix/TwoLevel.lean` (an import move), `Mathlib/Probability/FeynmanKac.lean` (two local scalar-tower instances) and `Mathlib/Probability/IidClockRace.lean` (`eventuallyEq_set` → `eventuallyEqSet_iff`). ▸ **ROUTE (a) TAKEN, AND IT PAID FOR ITSELF.** [`scripts/glossary-staleness-triage.sh`](../scripts/glossary-staleness-triage.sh) is the measurement as a tool rather than as a claim: for each flagged entry it diffs the module **from its state at that entry's own review date** to `HEAD`, normalises away the `if_*`/`dif_*` family, the `Mathlib.Data.*` → `Mathlib.Basic.*` moves and whole import lines, and reports `rename-only` or `read`. Measured: **32 rename-only, 6 to read.** All 38 dates are now `2026-10-13` — the 32 on the measurement, the 6 on a read of the entry against its module's actual diff. ⚠️ **AND THE READ FOUND ONE GENUINE STALENESS, WHICH THE RENAME HAD NOTHING TO DO WITH**: `two-level-unitary` sent a reader to “BACKLOG row 74” for the elementary-gate count, and **#74 was discharged in full on 2026-10-06**. The sentence now answers the question instead, naming `ctrlGate_eq_twoLevelMat` and `exists_multiCtrl_sandwich`, which is where that count actually lives. Five of the six needed no change: `fubini-study-measure`, `first-passage-race`, `no-signalling`, `bell-not-join` and `feynman-kac` each describe statements their module's diff did not touch — the diffs were a docstring sentence, an alias rename, the deletion of `arenaKick_mk` (which no entry names), two local instances, and proof-side shim collapses. ⚠️ **`check-glossary` was NOT changed.** The temptation was to teach the staleness check to ignore rename-only commits; that would weaken a gate to quieten it. The triage is a separate advisory tool that partitions the gate's output, and it says in its own banner that it bumps no dates and excuses no read. ⚠️ **Nothing is broken meanwhile**: `check-glossary` reports 0 hard findings and 0 warnings and exits 0; staleness is a signal, not a gate. ⚠️ The deeper item, if the answer is (a) more than once: the check could ignore commits whose diff for a module collapses under a known rename set, which would stop a mechanical sweep from drowning the signal — that is guard engineering and would belong with #27. | ~~(a) **S** or (b) **M**~~ (a) took **S** | Medium — a staleness check that cries wolf after every pin move stops being read, which is the failure mode it was built to prevent; and the triage is reusable, so the next pin move costs a command | #27 (HY-6, the sweep that caused it), `scripts/check-glossary.sh` | > | ~~144~~ | Foundations | Statistical-mechanics layer | *new (2026-10-12, out of #130)* **The Haar cell law of the fibre grid is uniform** — `coarseLaw (kMuL p₀) (fibreCell m) {c} = 1/(m+1)²` for every cell. #130 and [`FibreFiniteHorizon.lean`](../CsdLean4/SigmaLayer/FibreFiniteHorizon.lean) both *say* the cells have equal Haar weight — they do, by construction — but **no theorem says it**, which is why #130's divergence bound is stated against `uniformOn univ` and cannot yet be combined with ~~#118(ii)~~'s `L²` bound. ▸ **DONE 2026-10-13, AND THE ENDPOINT BOOKKEEPING THE ROW PRICED NEVER HAPPENED** ([`SigmaLayer/FibreCellUniform.lean`](../CsdLean4/SigmaLayer/FibreCellUniform.lean), 15 pins; M took M, most of it the capstone rather than the uniformity). ★★★ `volume_circCell_preimage` — a circle cell has Haar measure `1/(m+1)`; ★★★ `volume_torusCell_preimage` — a fibre cell has `1/(m+1)²` by the product measure; ★★★ `coarseLaw_fibreCell_apply` and ★★ `coarseLaw_fibreCell_toReal` — **the Haar cell law of `Σ` is uniform**, which is what #130 and ~~#118(ii)~~ both assumed. ⚠️ **THE ROW'S OWN ROUTE WAS MORE WORK THAN THE PROBLEM NEEDED, which is the finding.** It laid out three cases, one per interval shape, because `circRep` lands in `(0, 1]` while a floor cell is half-open the other way. But the cell never has to be *identified* — only **bounded**: it is sandwiched between `Ioo (k/(m+1)) ((k+1)/(m+1))` and the corresponding `Icc`, uniformly in `k`, and those two have the same Lebesgue measure, so `measure_mono` in both directions pins it. `Ioo_subset_cellSet_inter` is one statement for every `k`; `cellSet_inter_subset_Icc` still splits `k < m` against `k = m`, because the clamp genuinely behaves differently there, but each half is three lines and **the `k = 0` case disappeared entirely**. No endpoint is ever decided. ⚠️ **One piece of general machinery was missing and is now extracted**: ★★ `RecordLayer.volume_rep_preimage` — the `rep`-preimage of **any** measurable set weighs what the set weighs inside one turn. The row expected to need it (`volume_rep_preimage_Ioc` “covers `k = 0` directly”), and the `Ioc` form indeed cannot see a floor cell; the general form is now the engine and the `Ioc` form is a two-line corollary of it. ▸ **AND THE PAYOFF LANDED WITH IT**: ★★★ `lintegral_klDiv_horizonLaw_le` — **`R-019`'s second clause at a finite horizon, with its rate**. In expectation over the epistemic measure the divergence of the time-averaged occupation law from the cell law is at most `(m+1)⁴·(2/T)·∑_{u | ~~145~~ | Docs | Reviewer-proofing | *new (2026-10-13, out of the guard added with #136)* **42 statements are proved twice in this corpus, and now something says so.** [`scripts/check-duplicates.sh`](../scripts/check-duplicates.sh) with [`scripts/duplicate-scan.lean`](../scripts/duplicate-scan.lean) groups every hand-written `CSD.*` theorem by its statement with **binder names erased**, so alpha-equivalent statements compare equal. Of the 6,100 theorems, 100 groups share a statement: **33 Empirical twins** (a `CSDBridge` theorem beside its `Empirical.QM` original — the point of those files, not a defect), **25 in-namespace aliases** (a headline name beside its internal one, deliberate), and **42 TO REVIEW** — two namespaces, neither twin nor alias. Examples from the scan: `CV.euclidean_norm_map_of_isom` and `LF5.toEuclideanLin_norm_map_of_isom`; `LF2.psd_smul` and `SigmaLayer.posSemidef_real_smul`; `LF6.mul_star_eq_normSq` and `Einselection.mul_star_normSq`; `LF5.single_one_ne_zero`, `LF4.cpBasisVec_ne_zero` and `RecordLayer.single_ne_zero'` — three names for one fact. ▸ **REVIEWED 2026-10-13, and the count was wrong because the GUARD was too blunt: 42 → 27.** Fifteen of the 42 were not second proofs at all but second *names*: `b92_nonorthogonal` is literally `:= bb84_states_nonorthogonal`. That is detectable structurally — **an alias cites its original** — so the scan now asks whether any member of a group appears in another member's **proof term** and reports those as `derived` rather than as findings. The classification is now 33 twins + 25 in-namespace aliases + **15 derived** + **27 to review**, and the pin is 27. ⚠️ **AND THE REVIEW FOUND A REAL DEFECT, which is the row paying for itself.** `Empirical.Hardy.HardyQMMax.exists_hardy_realisation_max` carried the heading “QM realises the Hardy constraints **at the golden-ratio maximum**”, and its statement is **identical** to `exists_hardy_realisation`'s — the maximum is in the proof's choice of witnesses (`hardyMaxVec`, `aPrimeMax`) and nowhere in the statement. The docstring now says so, and points at `hardyMax_probability_eq`, which is the theorem that does carry the maximum `(5√5 − 11)/2`. Nothing was downgraded: the Lean content was always there, in a different theorem from the one the prose implied. Note that `HardyVolume.lean`'s prose had it right all along, which is why this survived the 2026-08-08 vacuity audit (`future-work.md` HY-2) — that audit strengthened *both* Hardy existentials equally and so left the `_max` one with no maximality content. ▸ **The 27 triaged into four clusters, and the merges are #146**: (A) eight groups of one-line `√n`-power and coordinate helpers repeated across the `CSDBridge` volume files (`r2`, `r3`, `r6`, `r3_4`, `inner_three`, `normSq_coord`, `realNormSq`, `normSq_inner_single`); (B) four `phiPlus_ofLp_k` pairs between `Gates` and `SuperdenseCoding`; (C) three protocol-local helpers (`hadEntry_zero_left` × 3 — ~~there is a *fourth* copy in `MeasurementUncompute.lean` the scan does not group because its binder differs~~ **wrong, corrected under #146 (C): there are exactly three and the scan grouped all three, binder names being erased** — `inner_eq_sum`, `mul_conj_eq`); (D) eleven library-shaped pairs wanting a canonical Category-1 home (`euclidean_norm_map_of_isom`, `psd_smul`, the three `single_ne_zero` spellings, **three separate `Fact (0 < 1)` instances**, `toLin_unit_ne_zero`, `trace_outer_mul_eq_inner`, `toEuclideanLin_comp`, `vnUnitary_unitary`, `sign_sq`), plus one pair to read (`IsJointLift.protocol_outcomeSector_eq`). ▸ **(D) gained a twelfth from (A)**: the index-generic `normSq_inner_single` wants a Category-1 home that `LF4/POVMVolume.lean` can import, after which `LF4.normSq_inner_single` is a citation. ⚠️ **Not a build problem** — nothing is wrong with the corpus; this is paid in reader confusion and in the next person proving it a fourth time. ⚠️ **The guard catches exact repeats only**, which it says in its own header: it would not have caught `Matrix.reindex_mul`, because that duplication lived *inside proofs* with no declaration to compare. Near-duplicates (an `ι`-generic lemma against its `Fin N` instance) are also out of reach. Pretending otherwise would make the guard a false reassurance. | ~~**M**~~ the review took **S–M**; the merges are #146 | Medium–high — the review cut the count by a structural test and found one over-claiming theorem, which is more than the merges would have | #136; `scripts/check-duplicates.sh`; #27 | > | ~~146~~ | Docs | Library owed / upstream | *new (2026-10-13, out of ~~#145~~'s review)* **Merge the 27 genuinely duplicated statements, by cluster.** ~~#145~~ reviewed them and cut the count from 42 by a structural test; what is left is 27 groups where **neither member cites the other**, so each is a second proof of the same statement. Four clusters, each pickable on its own: ~~**(A) — S** eight one-line `√n`-power and coordinate helpers across the `CSDBridge` volume files~~ ▸ **(A) DONE 2026-10-13, pin 27 → 19** ([`Empirical/CSD/VolumeArith.lean`](../CsdLean4/Empirical/CSD/VolumeArith.lean)). One shared module in `CSD.Empirical.CSDBridge` — the namespace each volume file's own namespace sits *inside* — so **not one call site moved**: `r2`, `r3`, `r6`, `r3_4`, `normSq_coord`, `realNormSq`, `normSq_inner_single` and `inner_three` resolve exactly where they resolved before, through the enclosing namespace. 67 lines deleted from six files for 7 added (one import each), and the eight groups left the TO REVIEW list exactly as predicted. ▸ **AND ONE OF THEM GOT BETTER RATHER THAN JUST MOVED.** `normSq_inner_single` had **three** copies, not two: `Fin 4` in `BellVolume` (proved by `Fin.sum_univ_four` and a `fin_cases` over the index), `Fin 4` in `HardyVolume` and `Fin 8` in `GHZVolume` (both two lines from `EuclideanSpace.inner_single_left`). The short proof never used the dimension, so the shared version is **index-generic** and absorbs all three — including the `Fin 8` one **signal (1) could not see**, because it compares statements and a `Fin 4` statement is not a `Fin 8` statement. That is the guard's declared blind spot showing up on its first use, and the fix was free. ⚠️ **THERE IS A FOURTH COPY, AND THE GUARD'S WEAKER SIGNAL IS WHAT FOUND IT.** `LF4.normSq_inner_single` ([`LF4/POVMVolume.lean`](../CsdLean4/LF4/POVMVolume.lean)) is the same statement at the dilated index `Fin N × ι` with the *same two-line proof*. Signal (2) — same final name component, different namespace, the one the scan's own header calls weak — reported it, which is a point in that signal's favour. It is **not** retired here: `LF4` cannot import an `Empirical/` module, so the lemma has to reach a Category-1 home first, and that is cluster (D)'s shape rather than (A)'s. Making the shared version index-generic rather than `Fin n` is what leaves it one line from deletion. **Added to (D).** ~~**(B) — S** the four `phiPlus_ofLp_k` pairs between `Gates` and `SuperdenseCoding`~~ ▸ **(B) DONE 2026-10-13, pin 19 → 15** ([`SuperdenseCoding.lean`](../CsdLean4/Empirical/QM/Resources/SuperdenseCoding.lean), −16/+3). They became citations, as the row said they should: the four local `phiPlus_ofLp_*` are gone and `mulVec_phiPlus` now rewrites with `Gates.qmKetPhiPlus_ofLp_*`, which this file **already imported** ([`Gates/BellPrep.lean`](../CsdLean4/Empirical/QM/Gates/BellPrep.lean)). No new import, no new lemma, nothing else referenced the deleted names, and none was axiom-pinned. ▸ **AND THE DUPLICATION HAD A WRITTEN REASON THAT HAD EXPIRED, WHICH IS THE INTERESTING PART.** The block carried the comment “*re-derived locally; `BellPrep`'s versions are `private`*” — and `git log -S` says that was **true when it was written**: `df7efdff` (2026-05-22) added them to `BellPrep` as `private`, `aa1cc660` (2026-05-24) re-derived them here for exactly that reason, and then `ba10362d` (2026-07-22, the module-system migration) **dropped the `private`** as a side effect. From that day the copies were unnecessary; nobody noticed for nearly three months, because nothing was looking. This is the clearest case yet for the guard: the duplication was not careless, it was **correct and then quietly stopped being correct**, and no amount of care at the time of writing would have caught it. The new comment records the history rather than deleting it. ~~**(C) — S** three protocol-local helpers~~ ▸ **(C) DONE 2026-10-13, pin 15 → 12.** All three got a **canonical home that already existed**, which is the whole point of the cluster. ▸ `hadEntry_zero_left` → [`Mathlib/QuantumInfo/Hadamard.lean`](../CsdLean4/Mathlib/QuantumInfo/Hadamard.lean), **directly beside `hadEntry_zero_right`, which was already there and already `@[simp]`**. That is the sharpest instance of this row's premise yet: the Cat-1 file defining `hadEntry` held the `_right` half, and three protocol files (`Algorithms/HadamardTest`, `Algorithms/SwapTest`, `MeasurementUncompute`) each re-proved the `_left` half by the identical `simp [hadEntry]` while importing the file that should have had it. Marking it `@[simp]` for symmetry with its mirror immediately **shortened a downstream proof**: `MeasurementUncompute.corr_had_eq`'s first branch was `simpa using hadEntry_zero_left (x * y)` and is now `simp`, which `--wfail` caught as an `unnecessarySimpa` warning — the gate reporting that the merge made the corpus simpler. ▸ `inner_eq_sum` and `mul_conj_eq` → [`Mathlib/QuantumInfo/Register.lean`](../CsdLean4/Mathlib/QuantumInfo/Register.lean), whose `normSq_eq_sum_prob` is the **diagonal case of `inner_eq_sum`** — so the file had the corollary and not the lemma. Both are flagged in place as convenience specialisations rather than library additions (`inner_eq_sum` is `PiLp.inner_apply` with the `ℂ` coordinate inner product unfolded; `mul_conj_eq` is `Complex.mul_conj` carrying the norm-squared cast, which is the only reason to want it — the Born sums need that cast atom canonical). ⚠️ **AND THE CLAIM IN THIS ROW ABOUT A FOURTH COPY WAS WRONG — MINE, not the scan's.** This cluster was written as “three helpers, and note the scan *under*-reports: `hadEntry_zero_left` has a **fourth** copy in `MeasurementUncompute.lean` that is not grouped because its binder differs”. `git grep` at `1492c42f` finds **exactly three** declarations, and the scan's group named all three — including `CSD.Empirical.QM.hadEntry_zero_left`, which *is* `MeasurementUncompute`'s. Its binder genuinely differs (`(c : Fin 2)` against `(b : Fin 2)`) and that is exactly why it grouped: **`eraseBinderNames` is the feature, working**. So the guard did not under-report here; the triage mis-read its own output and invented a fourth copy. The guard's real blind spot is the one (A) hit — the same statement at a different *type* — not a different binder name. Left here as a correction rather than deleted, because it is the kind of mistake the ledger exists to catch. ~~**(D) — M** eleven library-shaped pairs wanting a canonical Category-1 home, including **three separate `Fact (0 < 1)` instances** and three spellings of “a basis vector is nonzero”; several are axiom-pinned, so a deletion moves a pin as well as its uses.~~ ▸ **(D) DONE 2026-10-14, pin 12 → 1 — ROW 146 COMPLETE.** 27 declarations deleted, 5 written once, 1 moved; and the answer to “where is the canonical home?” was **Mathlib itself** for two of the twelve pairs, an existing corpus file for four, a new Category-1 module for five, and a third file for the one to read. ▸ **Already upstream, so no lemma at all** — every copy deleted, every consumer cites Mathlib: `psd_smul` / `posSemidef_real_smul` had **three** copies, not two (`LF6/LindbladPositivity.lean` carried a third at real-scalar type, which the scan could not group), and all three are `Matrix.PosSemidef.smul`, which takes the cast scalar `(c : ℂ) • M` through `Complex.zero_le_real` and the real scalar `t • M` directly; `SigmaLayer`'s copy was *already* a one-line citation of it, the clearest sign a wrapper was never wanted. The **three `Fact ((0 : ℝ) < 1)` instances** (`LF4/ArenaSymplectic`, `LF4/KahlerInstance`, `RecordLayer/CircleFibre`), plus a fourth as a `have` in `PointerLanding`, are all `ZeroLEOneClass.factZeroLtOne` (`Mathlib/Algebra/Order/ZeroLEOne.lean`): deleted, nothing else changed. ▸ **A home that already existed** (the (C) pattern): `vnUnitary_conjTranspose_mul` (LF6) is `LF5.vnUnitary_unitary` in the defining file, verbatim; `sign_sq` (`LF3/Spinor`) is `Sign.val_sq`, `@[simp]` in `LF3/Setup.lean` six lines below `Sign.val`; `trace_outerProduct_mul_eq_inner` (`LF3/Projectors`) is `LF2.trace_outer_mul_eq_inner`, and `LF2Interface` now imports `EffectAux` for it; `toLin_unit_ne_zero`, `toEuclideanLinN_ne_zero` **and the two `toEuclideanLin_ne_zero`** (CV, LF4) are `Matrix.UnitaryGroup.toEuclideanLin_unitary_ne_zero` (Cat-1 `UnitaryTransitive.lean`), which every one of those files already imported. ▸ **A new Category-1 home** — [`Mathlib/Analysis/InnerProductSpace/EuclideanLin.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/EuclideanLin.lean), five lemmas, one an upstream candidate (`Matrix.norm_toEuclideanLin_of_isom`: Mathlib has no norm lemma for `toLpLin` at all) and four flagged in place as convenience specialisations that must not be offered upstream. They absorb the isometry-norm pair, the three spellings of “a basis vector is nonzero”, the two scalar-inner copies, the **five** copies of `toEuclideanLin (A * B) v = …` (`toEuclideanLin_comp`, `CircuitFlow`'s generic prime, `JointEig` at `Fin 2 × Fin 2`, `PauliDotRobertson` at `Fin 2`, `HamiltonianSignature` at `Fin N` and in the other orientation), and the dilated `LF4.normSq_inner_single` — absorbed by (A)'s index-generic version exactly as (A) predicted. ▸ **The pair to read**: `IsJointLift.protocol_outcomeSector_eq` and `pointerProtocol_outcomeSector` were one theorem in two files neither of which imports the other; it now lives once, in `RecordLayer/PointerProtocol.lean` beside `pointerProtocol_evolve_stroke` (the lemma it is one `show` away from), under the already-pinned name. ⚠️ **AND THE SCAN'S BLIND SPOT COST MORE THAN THE SCAN FOUND.** Eight of the 27 were invisible to signal (1): six are copies at a *specific* index type of a lemma stated for `ι` (`Fin N × ι`, a real scalar, `Fin 2 × Fin 2`, `Fin 2`, `Fin N`, `Fin 2`), and two are index-generic copies whose partner is a **Category-1** lemma — the scan reads `CSD.*` only, so `CSDBridge.Algorithms.toEuclideanLin_mul_apply'` and `CV.toEuclideanLin_ne_zero` were never compared with `Matrix.toEuclideanLin_mul_apply` and `Matrix.UnitaryGroup.toEuclideanLin_unitary_ne_zero`. All eight surfaced through signal (2) — same final name, different namespace — and only *after* the Cat-1 lemmas were named as the corpus names them. **Naming the canonical lemma the way the corpus would is therefore part of retiring its duplicates**: it is what makes the weak signal fire. ⚠️ One honest nit: the first draft of this cluster declared `EuclideanSpace.single_one_ne_zero` *inside* `namespace EuclideanSpace`, so its real name was `EuclideanSpace.EuclideanSpace.…` and three consumers failed to build — caught by the rebuild, not by any guard. **The one TO REVIEW group left is the Hardy pair ~~#145~~ already adjudicated** (same statement, different witnesses, said so in the `_max` docstring): kept by design and recorded beside the pin, so the pin is **1**, not 0, and honestly so. Closed in passing: `CircuitFlow.toEuclideanLin_pow_apply` re-derived `@[simp] toLpLin_pow` by induction and now cites it. *Gate note:* `check-axiom-imports.sh` exceeds 180 s on Windows (one `git grep` per pin over 4,967 pins); its logic re-run with a one-pass declaration index finds 0 pins outside the 856-file closure. ⚠️ **Lower the pin in `check-duplicates.sh` as each cluster lands**, which is what makes the progress measurable. ⚠️ **Not a build problem**: nothing is wrong with the corpus, and the cost is paid in reader confusion and in the next person proving it again. | ~~(A)~~ ~~(B)~~ ~~(C)~~ took **S** each; ~~(D) **M**~~ (D) took **M** | Medium — four of these were written while the author watched, which is the rate the guard exists to measure | ~~#145~~; `scripts/check-duplicates.sh`; #27 | > | ~~147~~ | Foundations | New predictions · Statistical-mechanics layer | *new (2026-10-14, out of ~~#118~~)* **Relaxation for a general `L²` initial density.** ~~#118~~ proved `R-019`'s convergence clause for densities constant on the fibre cells (and, by the Category-1 pair bound, for finite combinations of arbitrary rectangles of arcs). The general case is an `ε/3` argument: approximate `ρ ∈ L²(T²)` by a rectangle step function `ρ_ε`, apply `abs_integral_rectIndR_iterate_mul_sub_le_env` to each rectangle, and bound the remainder by `‖ρ − ρ_ε‖₂` through Cauchy–Schwarz. The missing step is the density of rectangle step functions in `L²(T²)`: simple functions on measurable sets are dense (`MeasureTheory.Lp.simpleFunc.dense`), and the passage from a measurable set to a finite union of rectangles is the regularity of Haar measure on the torus (approximate a measurable set by an open set, an open set by finitely many rectangles). No rate in general: it depends on the observable. ⚠️ The map identification (Posit 1, #19) is a separate matter and is not this row. ▸ **DONE 2026-10-14 — BY A DIFFERENT DENSE CLASS THAN THE ROW NAMED, WHICH IS THE FINDING.** The row priced the density of *rectangle step functions* (regularity of Haar, open sets by rectangles). Not needed: Mathlib already has the Fourier basis of `L²(Fin 2 → T)` (`UnitAddTorus.hasSum_mFourier_series_L2`, `Analysis/Fourier/AddCircleMulti.lean`), and `finTwoArrow` carries it to `T²` with Haar to Haar and `mFourier n ↦ chi (n 0, n 1)` — so **trigonometric polynomials** are the dense class (★★ `exists_trigPoly_integral_normSq_sub_lt`), and for them the exact statement is trivial: `∫ χ_v(cat^n p) χ_w(p) = [catZ^n v + w = 0]`, and `catZ^n v` leaves every bounded set for `v ≠ 0` (★ `tendsto_supNormZ_catZ_iterate_atTop`, from the golden-ratio growth of ~~#118~~), so the correlation of two trigonometric polynomials *equals* the product of the means from some lag on (★★ `eventually_integral_trigPoly_iterate_mul`). The `ε/3` argument then needs only Cauchy–Schwarz and measure preservation. ▸ Cat-1 [`Mathlib/Dynamics/CatMapStrongMixing.lean`](../CsdLean4/Mathlib/Dynamics/CatMapStrongMixing.lean) (nine pins, **M–L** as priced): ★★★ `tendsto_integral_comp_cat_iterate_mul` — **for every `f, g ∈ L²(T²)`, `∫ f(cat^n p) g(p) → (∫ f)(∫ g)`** — its real form, the set form `vol(cat^{-n} s ∩ t) → vol s · vol t`, and the predicate `Mixing f μ` shaped after Mathlib's `Ergodic` (with `Mixing.ergodic`): ★★★ `cat_mixing`, `cat_ergodic`. **Arnold's cat map is strongly mixing, in Lean.** ▸ On `Σ`, [`SigmaLayer/FibreRelaxation.lean`](../CsdLean4/SigmaLayer/FibreRelaxation.lean) (five pins): a fibre cell is a torus cell of the fibre coordinate (definitional), the stroke acts on the fibre with the base fixed, so ★★★ `tendsto_integral_fibreStroke_iterate_mul_snd` (the stroke is mixing on `Σ` for fibre observables), ★★★ `tendsto_fibreStepWeight` (every cell's weight after `n` strokes → `(m+1)⁻² ∫ ϱ` for every `ϱ ∈ L²`), and ★★★ `tendsto_klDiv_fibreStepLaw`: **for every nonnegative normalised `ϱ ∈ L²(T²)`, the cell law after `n` strokes converges to uniform in relative entropy.** `R-019`'s convergence clause is now unconditional in the initial density; the residue is open on the map identification alone (Posit 1, #19), its carriers updated. ⚠️ **No rate**, by design (the rate of a general density depends on its Fourier coefficients; ~~#118~~'s `klDiv_stepLaw_le` is the rate for the coarse-grained class). ⚠️ **Densities on the fibre coordinate**; a density depending on the base too reduces by integrating out the base (Jensen on a probability measure), not written. ⚠️ The map is still the witness, not a derived dynamics — nothing here is a prediction until #19. | ~~**M–L**~~ **M–L** as paid | ~~Medium — it would make `R-019`'s convergence clause unconditional in the initial density, leaving only the map identification~~ Done: it did | ~~#118~~; `R-019`; `Mathlib/Dynamics/CatMapMixing.lean`; `Mathlib/Dynamics/CatMapStrongMixing.lean`; `SigmaLayer/FibreRelaxation.lean` | > | ~~148~~ | Foundations | Statistical-mechanics layer | *new (2026-10-14, out of ~~#101~~(a))* **The Lyapunov exponent of the fibre witnesses: `2 log φ` for the hyperbolic stroke, zero for the constructed de-isolation flow.** Mathlib has no Lyapunov exponent at all (MATHLIB-ABSENT(lyapunovExponent)), so the Category-1 piece is the definition for a *linear* map `A` on a finite-dimensional normed space — `λ(A, v) = lim (1/n) log ‖Aⁿ v‖` — with the two-dimensional hyperbolic case worked out: for `[[2,1],[1,1]]` (eigenvalues `φ²`, `φ⁻²`) the exponent is `2 log φ` for every `v` off the stable line and `−2 log φ` on it. The proof is already in the corpus: `ellU_catZ_iterate` scales the unstable linear form by `φ²` per step and `abs_ellU_le`/`one_le_abs_ellU_mul_ellS` sandwich the norm (the transpose has the same spectrum). On `Σ`: `catStroke` is affine-linear on `T²`, so its derivative is that constant matrix and its exponent is `2 log φ`; `shearEvolve` (`pshift`) and `kFlow` translate the fibre, derivative the identity, exponent `0` — **the constructed de-isolation witness is not chaotic, the hyperbolic one is, and which one nature uses is #19.** ⚠️ Honest scope: the exponent of a *nonlinear* measure-preserving map (Oseledets) is not defined and not needed for either witness; no claim about the base `ℂℙ^{N−1}` flow. Shares its module with #149 (both together **M–L**). ▸ **DONE 2026-10-14, M as priced.** Cat-1 [`Mathlib/Dynamics/LyapunovExponent.lean`](../CsdLean4/Mathlib/Dynamics/LyapunovExponent.lean) (five pins): `HasLyapunovExponent A v l` for a continuous linear map (`log ‖Aⁿ v‖ / n → l`), the geometric sandwich `hasLyapunovExponent_of_geometric_bounds`, the real cat matrix `catR` as the lift of `cat` (`cat_coe`) and of `catZ` (`catR_intCast`), its two eigenforms `ellUR`/`ellSR` scaled by `φ²` and `2 − φ = φ⁻²`, and ★★ `hasLyapunovExponent_catR` (`2 log φ` off the stable line), ★ `hasLyapunovExponent_catR_stable` (`−2 log φ` on it). The one real step was the inversion `‖v‖ ≤ |ℓU v| + |ℓS v|` (`v₁ = (φ ℓU + (φ−1) ℓS)/(2φ−1)`, `v₂ = (ℓU − ℓS)/(2φ−1)`). ▸ On `Σ`, [`SigmaLayer/FibreLyapunov.lean`](../CsdLean4/SigmaLayer/FibreLyapunov.lean) (six pins): `catStroke_coe` (the stroke lifts to `catR`), `shearEvolve_snd`/`pshift_coe`/`kFlow_coe` (the constructed de-isolation witness acts on each fibre by a translation, as does the isolated fibre flow; their lifts are `v ↦ v + c`), derivatives through Mathlib's `fderiv` (`fderiv_catR`, `fderiv_transR`), and ★★★ `fibre_witnesses_lyapunov`: **the hyperbolic stroke's lift stretches every direction off the stable line at rate `2 log φ`; the translations' lifts stretch nothing.** The constructed de-isolation flow is not chaotic, the hyperbolic one is, and which map nature uses is #19, which now carries the criterion. ⚠️ Exponents of the lifts' derivatives (constant for both witnesses); no tangent bundle on `T²`, nothing about the base flow, nothing decided about #19. | ~~**M**~~ done (M) | Medium — completes the hyperbolic dossier (decay, mixing, relaxation, now stretching) and gives Posit 1's discharge a measurable form | ~~#101~~; #19; #149; `Mathlib/Dynamics/CatMapMixing.lean`; `RecordLayer/ShearWitness.lean`; `LF4/KahlerFlowNoMixing.lean`; `Mathlib/Dynamics/LyapunovExponent.lean`; `SigmaLayer/FibreLyapunov.lean` | > | ~~149~~ | Foundations | Statistical-mechanics layer | *new (2026-10-14, out of ~~#101~~(b))* **The classical OTOC of the hyperbolic stroke grows exponentially; the constructed de-isolation flow's does not.** ▸ **DONE 2026-10-15 (M).** Cat-1 [`Mathlib/Dynamics/ClassicalOtoc.lean`](../CsdLean4/Mathlib/Dynamics/ClassicalOtoc.lean): `poissonBracket f g` on `ℝ²` with canonical `(x, p)`, `classicalOtoc A n = {x ∘ Aⁿ, p}` for a linear map, constant in the phase point and equal to the `(1,1)` entry of `Aⁿ` (`classicalOtoc_eq`, `poissonBracket_clm_snd`); ★ `catR_pow_apply` = **the powers of the cat matrix are Fibonacci matrices** `[[F_{2n+1}, F_{2n}], [F_{2n}, F_{2n+1} − F_{2n}]]`; ★ `classicalOtoc_catR` = the cat map's classical OTOC is **exactly `F_{2n+1}`** (Binet form `classicalOtoc_catR_binet`); ★★ `tendsto_log_classicalOtoc_catR` = it grows at rate **`2 log φ`, the Lyapunov exponent** (read off `hasLyapunovExponent_catR` in direction `(1,0)`, whose iterates have sup norm exactly `F_{2n+1}`), ★ `tendsto_log_classicalOtoc_sq_catR` = the squared bracket, the textbook `C(t)`, at `4 log φ`; on the other side `classicalOtoc_id = 1` (every translation), `classicalOtoc_shearXR = classicalOtoc_shearYR = 1` and the vertical shear's transverse bracket is `n` (`poissonBracket_snd_shearYR`, linear) — **exponential scrambling is the composition of the two transverse shears, as ~~#153~~ said of the exponent**; ★★★ `cat_classicalOtoc`. Cat-7 [`SigmaLayer/FibreOtoc.lean`](../CsdLean4/SigmaLayer/FibreOtoc.lean): on the lifts' derivatives, ★ `classicalOtoc_catStroke_lift = F_{2n+1}`, ★★ `tendsto_log_classicalOtoc_catStroke_lift` (rate `2 log φ`), ★ `classicalOtoc_transR_lift = 1` (the `shearEvolve` fibre action and `kFlow`, rate `0`), the two shear factors `= 1`; ★★★ `fibre_witnesses_otoc`. **The three classical signatures — mixing (~~#147~~), exponent (~~#148~~), OTOC — now agree on which side each fibre map sits.** ⚠️ **Two wordings of this row corrected in the module header:** the bracket `{x ∘ Φⁿ, p}` is the `(1,1)` entry `F_{2n+1}`, not the `(1,2)` entry `F_{2n}` (which is `{p ∘ Φⁿ, p}`; every entry grows at the same rate, so nothing concluded changes), and 'twice the Lyapunov exponent' is the rate of the *squared* bracket. The quantum OTOC's exponential regime stays unpriced until a quantised fibre dynamics exists, as the row says. 27 pins. ▸ *Original row:* The corpus's OTOC is the commutator-norm diagnostic on finite unitaries (`QuantumChaos.otoc`, `Incubator/QuantumChaos/Otoc.lean`); its classical analogue for a map `Φ` of `T²` is the Poisson bracket `{x∘Φⁿ, y}`, the `(1,2)` entry of `DΦⁿ`. For the cat map `DΦⁿ = Aⁿ = [[F_{2n+1}, F_{2n}], [F_{2n}, F_{2n−1}]]` (Fibonacci, `Nat.fib`), so the classical OTOC is `F_{2n} ~ φ^{2n}/√5`: **exact exponential growth at rate `2 log φ`**, twice the Lyapunov exponent, as the textbook says. For a translation (`shearEvolve`, `kFlow`) the derivative is the identity and the bracket is constant; and the corpus's kicked-diagonal family is at most linear (`otoc_interactingU_le`). ⚠️ **Not this row:** the *quantum* OTOC's exponential regime up to the Ehrenfest time needs a quantised cat map (Hannay–Berry, dimension `N`, `ℏ = 1/2πN`); the corpus has no quantised fibre dynamics and nothing derives one — **L–XL**, unpriced until an object exists. Shares its module with ~~#148~~ (now `Mathlib/Dynamics/LyapunovExponent.lean`, which already has the eigenform scaling `ellUR_catR_pow`). | ~~**M**~~ done (M) | Medium — the explicit chaos signature on the witness, with the constructed witness on the other side of the line | ~~#101~~; #148; `CV/ChaosBounds.lean`; `Incubator/QuantumChaos/Otoc.lean` | > | 150 | Foundations | Statistical-mechanics layer | *new (2026-10-14, out of ~~#101~~(c))* **Random-matrix level statistics for a quantised fibre dynamics.** Wigner–Dyson spacing, or the ramp of the spectral form factor, need an ensemble (CUE/GUE), its joint eigenvalue density and either sine-kernel asymptotics or Weingarten calculus. Mathlib has none of it — no semicircle law (MATHLIB-ABSENT(semicircleLaw)), no joint eigenvalue density, no Weingarten calculus — and the corpus has **no object**: level statistics belong to a quantised dynamics, the fibre witness is classical, and `sff` exists only as a definition on finite unitaries with the free field's exact revival as the integrable baseline (`sff_freeFieldU`, `CV/ChaosBounds.lean`). Same wall as #21 (ETH). The one M-sized cousin — the Haar-average spectral form factor of a random unitary — is Weingarten calculus, library-grade and programme-irrelevant. ⚠️ Not blocked on effort: blocked on the theory and on the object, both. ▸ **FRAMING CORRECTED 2026-10-21 (author's steer: Σ is ontic and physical, QM is epistemic and the calculation engine — quantising the ontic fibre runs that arrow backwards and has no CSD consumer).** Two corrections, both prose, no brick: (1) for the corpus's own hyperbolic witness the literature answer is **not** random-matrix statistics — the quantised cat map is the textbook counterexample (Hannay–Berry 1980; Keating, *Nonlinearity* 4 (1991) 309: finite quantum period, degenerate eigenphases, non-Wigner–Dyson spacing), and random-matrix statistics return only for generic perturbations (Keating–Mezzadri, *Nonlinearity* 13 (2000) 747); so even the row's own question, for the row's own object, has a known negative answer. A five-brick Lean chain (finite Heisenberg translations on `ℂᴺ` and their irreducibility; the Hannay–Berry propagator as quadratic phase × DFT; exact Egorov; periodicity from the finite order of `A` in `GL₂(ZMod N)`; exact periodicity of `sff`) would prove it with no random-matrix theory at all, priced L–XL — **and is not scheduled, because it quantises the ontic fibre**, which CSD says is not what happens; it would be a Mathlib-staged quantum-chaos result with no consumer here. (2) The CSD-relevant chaos content is **classical**, on the fibre, feeding #19's criterion: that is #176 (periodic-point growth = Lyapunov exponent), out of this row. What this row still holds is the ensemble question for a *perturbed* quantisation, honestly research and off the programme's axis. | research / **XL+** | Low — explanatory, no consumer | ~~#101~~; #21; `Incubator/QuantumChaos/SpectralFormFactor.lean` | > | ~~151~~ | Foundations | Reconstruction frontier | *new (2026-10-14, out of #19 under the Σ-local reading)* **Back-reaction is necessary: the two doors through which the preparation reaches the outcome.** ▸ **DONE 2026-10-14 (M).** Cat-1 [`Mathlib/MeasureTheory/FibrePreserving.lean`](../CsdLean4/Mathlib/MeasureTheory/FibrePreserving.lean) (two pins): `FibrePreserving Φ ν` (fixes the base coordinate, acts on the fibre over each base point by a `ν`-preserving map), and ★ `FibrePreserving.measure_preimage_univ_prod`: under `δ_b ⊗ ν` such a map sends mass `ν R` into a fibre-only readout `univ ×ˢ R`, for every `b`. ▸ [`RecordLayer/BackReactionNecessity.lean`](../CsdLean4/RecordLayer/BackReactionNecessity.lean) (five pins): ★★ `no_fibrePreserving_born` — no fibre-preserving measurement with fibre-only readouts has Born statistics (two basis rays disagree); ★★★ `exists_fst_ne_of_born` — **a Σ-local measurement with Born statistics and a fibre-only readout must move the base point**: the register acts on the calculation space; ★ `globalBasin_not_fibreOnly` — the corpus's basins read the base (the other door); `shearEvolve_fibrePreserving` — the constructed witness is fibre-preserving and takes that door, its readout reading the selector. **The finding:** every landed witness, `jointLift` included (base motion only along the torus directions, rates conserved), takes the readout door; a derivation of the rates from the dynamics is a theorem about the back-reaction door, where the epistemic point is acted on by the ontic register. ⚠️ Pointwise fibre preservation (what the witnesses satisfy); readout on `Σ`'s own fibre, the register-torus version is the same argument and not repeated; nothing decides #19. | ~~**M**~~ done (M) | High — fixes the *shape* of Posit 1's discharge from the corpus's own preparation | #19; #137; `POSITS.md` Posit 1; `RecordLayer/GlobalBasin.lean`; `RecordLayer/JointLift.lean` | > | ~~152~~ | Foundations | Reconstruction frontier | *new (2026-10-14, out of #19)* **Σ-local product coupling + einselection ⇒ torus generation** — the premise conversion for Posit 1. For a de-isolation coupling of product form, (base function) × (fibre momentum), the base back-reaction is the Hamiltonian flow of the base function; if the pointer basis is the context's basis (einselection, `pointer_invariant_iff_commute`: the pointer observables are exactly those commuting with the interaction), the base function is the expectation of an operator diagonal in that basis, `∑ λᵢ Πᵢ`, so each outcome's generator is the coordinate energy — a phase-rotation Hamiltonian (`isPhaseHamiltonian_coordEnergy`) — and `IsGeneratedRateField` holds for the rates read off the coupling; `CellLawForced` then gives the moment map. **Shape:** Posit 1 ⟸ (the coupling is Σ-local of product form) ∧ (the pointer is einselected in the context's basis), exactly as WAY brick 2 converted `hgen ⟺ RecordLocallyTomographic`. What it leaves: the *choice* of coupling (R-015, boundary) and the arena generation (R-016). ⚠️ **The principle's wording is the author's**: the Lean statement is a theorem about a named class of couplings, and naming the class is a posit-level decision; the draft above is the proposal. Route: finite-dimensional linear algebra (`Commute` with all `Πᵢ` ⇒ diagonal), `IsPhaseHamiltonian` for `λᵢ ‖xᵢ‖²/2`, the degree-0 normalisation, then `generatedRateField_eq_momentMap`. ▸ **DONE 2026-10-14 (M–L as priced) — WITH ONE CORRECTION TO THE ROW'S OWN WORDING, WHICH IS THE FINDING.** [`RecordLayer/SigmaLocalCoupling.lean`](../CsdLean4/RecordLayer/SigmaLocalCoupling.lean) (seven pins). `SigmaLocalCoupling N` = a Hermitian base operator `A` einselected in the context's basis (`[|eᵢ⟩⟨eᵢ|, A] = 0`); `A_eq_diagonal`; ★★ `baseGen_eq_sum` (`½ re⟪x, Ax⟫ = ∑ λᵢ · coordEnergy i`) and ★★ `fundamentalForm_torusField` — **einselection makes the back-reaction the torus flow at the eigenvalue speeds**, `ι_{∑ λᵢ X_i} ω = dQ_A`: Posit 1's "the dynamics generates the torus action", as a theorem given einselection. ★ `drift_eq_sum_momentMap`: the register's degree-0 drift is `⟨A⟩ = ∑ λᵢ Φᵢ`. ⚠️ **The correction.** The row said `IsGeneratedRateField` holds "for the rates read off the coupling" by unit-speed normalisation. That reading is **vacuous**: the unit-speed outcome generator is `coordEnergy i` whatever `A` is, so no coupling is consulted. The non-vacuous reading is **calibration** — the rate-weighted average of the outcome labels (the eigenvalues) equals the register's drift — and it is a principle about the *mean* only, for each coupling in the context's commutant. ★★★ `isCalibrated_forall_iff`: **a rate field calibrated against every einselected coupling is the moment map, and conversely** (forward direction: the `N` projector couplings suffice); ★★★ `rate_eq_momentMap_of_calibrated`, ★★★ `isTorusGenerated_of_calibrated` (calibration ⇒ generation). **The premise conversion as landed: Posit 1 ⟸ (Σ-local product coupling `H = Q_A · p_R`) ∧ (einselected in the context's basis) ∧ (readout calibrated to the drift for every such coupling).** ⚠️ `POSITS.md` is not edited: whether calibration becomes Posit 1's live form is the author's decision. ⚠️ The fibre half (`q̇_R = Q_A`) is prose, R-016's business; which coupling is realised is R-015. The dual to `CellLawForced`: generation on the rates, calibration on the couplings, both characterise the same field. | ~~**M–L**~~ done (M–L) | Highest in the programme — it is what a discharge of Posit 1 looks like once R-015 is accepted as a boundary | #19; ~~#151~~; `RecordLayer/CellLawForced.lean`; `RecordLayer/SigmaLocalCoupling.lean`; `Empirical/CSD/PointerCommutation.lean`; `specs/way-theorem-scoping.md` (brick 2 precedent) | > | ~~153~~ | Foundations | Reconstruction frontier · Statistical-mechanics layer | *new (2026-10-14, out of #19 and ~~#148~~)* **The fibre carries both record coordinates: hyperbolicity needs the second coordinate sheared.** ▸ **DONE 2026-10-15 (S–M).** Cat-1 [`Mathlib/Dynamics/CatMapShears.lean`](../CsdLean4/Mathlib/Dynamics/CatMapShears.lean): ★ `cat_eq_shearX_comp_shearY` (`cat = shearX ∘ shearY`, the vertical shear `(x,y) ↦ (x, x+y)` then the horizontal `(x,y) ↦ (x+y, y)`), each factor a Haar-preserving bijection (`measurePreserving_shearX`/`Y`) whose lift preserves the area form (`shearXR_area`, `shearYR_area` — symplectic), `catR = shearXR ∘ shearYR`; ★★ `hasLyapunovExponent_shearXR`/`shearYR` = **each shear has exponent `0` in every direction** (unipotent: `(shearXR^n) v = (v₁ + n v₂, v₂)`, via the new general lemma `hasLyapunovExponent_zero_of_linear_growth`, the sub-exponential sibling of the geometric sandwich); ★★★ `cat_two_shears` bundles it with `2 log φ` for the composition — **the stretching is the transversality's, not either shear's.** Cat-7 [`SigmaLayer/FibreTwoShears.lean`](../CsdLean4/SigmaLayer/FibreTwoShears.lean): `qshift` (the transverse partner of `pshift`), ★★ `catStroke_eq_two_shears` = `catStroke` is `pshift (rep x.2) ∘ qshift (rep x.1)` — two transverse *coordinate-reading* shifts; ★ `catStroke_ne_pshift_type` = no single `pshift`-type map `x ↦ pshift (g x) x` is the stroke, for any `g`; ★ `shearEvolve_snd_snd` = the constructed de-isolation leaves the second fibre coordinate untouched (with `shearEvolve_snd` it is ONE `pshift`-type map, amount read off the base); ★★★ `fibre_two_shears`. **Read against #19's criterion: a de-isolation that shears along `q` alone, by any base-read amount, has exponent `0`, so a hyperbolic de-isolation must act on the second coordinate too, reading the first — Posit 10 in the I3 form.** ⚠️ **One wording of this row corrected in the module header:** the shear factors do NOT "carry the same flux obstruction as the corpus's shear" — they are linear toral automorphisms, not time-one maps of torus vector fields (the plane flow `(x,y) ↦ (x, y + tx)` descends only at integer `t`), so `not_isHamiltonianVectorField_torusStrokeField` is about a different kind of map; what is proved of the factors is symplectic + Haar-preserving. Structure, not derivation, as priced: nothing decides #19 or R-019. 24 pins. ▸ *Original row:* The cat map is the composition of two *transverse* linear shears, `(x,y) ↦ (x, x+y)` then `(x,y) ↦ (x+y, y)`, each symplectic and each carrying the same flux obstruction as the corpus's shear (`ι_X ω` closed, not exact, `PiecewiseHamiltonian.lean`); the constructed de-isolation shears along `q` only, by a base-set amount, and ~~#148~~ shows that gives exponent `0`. So a de-isolation that can discharge R-019's map identification must act on the fibre's second coordinate too — which is Posit 10 in the I3 form the author chose on 2026-10-10 (the second coordinate carries record content). **Deliverable:** `cat = shearX ∘ shearY` with each factor a measure-preserving shear of `T²` (Cat-1, S), and on `Σ` the statement that `catStroke` is two transverse `pshift`-type shears while `shearEvolve` is one (S–M). ⚠️ Structure, not derivation: it says what a hyperbolic de-isolation would have to do to the fibre, not that nature's does it. | ~~**S–M**~~ done (S–M) | Medium — the ontic fibre is where the spacetime-nonlocal correlations are Σ-local, and this is what the stretching needs of it | #19; #148; `R-019`; `POSITS.md` Posit 10; `Mathlib/Dynamics/LyapunovExponent.lean` | > | ~~154~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#129~~)* **The generator domain of the phase group is the multiplication domain.** ▸ **DONE 2026-10-20 (M).** Cat-1 [`Mathlib/Analysis/Semigroup/PhaseGeneratorDomain.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/PhaseGeneratorDomain.lean), 8 pins. ★★ `mem_mulDomain_of_hasGeneratorAt` — **the Fatou argument as planned**: along `hₙ = 1/(n+1)` the quotients `qₙ = hₙ⁻¹(e^{−ihₙκ} − 1)f` converge to `w` in `L²` so `‖qₙ‖² → ‖w‖²` (`norm_sq_eq_integral` + `ofReal_integral_eq_lintegral_ofReal`), pointwise `|qₙ|² → |κf|²` (`tendsto_phaseDefect`, a.e. for all `n` at once by `ae_all_iff`), and `lintegral_liminf_le'` gives `∫|κf|² ≤ ‖w‖² < ∞`, hence `MemLp (κ·f) 2` (`memLp_two_iff_integrable_sq_norm`, `hasFiniteIntegral_iff_ofReal`); ★ `hasGeneratorAt_phaseGroup_iff` — the two readings of the domain agree, with value `−i κ f` (`L2.mulOp`) by uniqueness of the right derivative; ★★ `mem_freePositionOp_domain_of_hasGeneratorAt` — across the Fourier transform (`hasGeneratorAt_phaseGroup_fourierL2`), a vector whose free orbit has a right derivative at `0` lies in `freePositionOp.domain`; ★★★ `schrodinger_mem_freePositionOp_domain` — **the interacting orbit `U(t)ψ` lies in `dom H₀`**, in the operator sense; ★★★ `hasDerivAt_schrodinger_orbit` — **`d/dt U(t)ψ = −i (H₀ + V)(U(t)ψ)` with `H₀ = freePositionOp` applied to the orbit point** (`freePositionOp.addCLM (potential VL)` at `⟨U(t)ψ, mem⟩`), via `hasGeneratorAt_freeGroup_of_mem` and `HasGeneratorAt.unique`. **~~#129~~'s one residue is closed: the mild solution is classical in the operator-theoretic sense.** Still not made (as before): the `coeFn` identification of `potential VL` with `schrodingerOp`'s `mulCLM` for real `V`. ⚠️ Lean finding: `HasGeneratorAt`'s real scalar action is the one `NormedSpace ℂ E` induces (`complexToReal`), while fresh `h⁻¹ • x` terms on `Lp` elaborate with `Lp`'s own instance — defeq at default transparency (`exact`/`have` work) but not for `rw`/`simp` matching; `show` the goal in your own words first, then rewrite. ▸ *Original row:* ~~#129~~ proved that the interacting orbit `U(t)ψ` stays in the domain of the *generator* of the free group (`hasGeneratorAt_freeGroup_schrodinger`: the free orbit of `U(t)ψ` has a right derivative at `0`). `PhaseGenerator.lean` has the forward half — `phaseGroup_mem_mulDomain`, and `hasDerivAt_phaseGroup_apply` (`κ·f ∈ L²` ⇒ the orbit is differentiable) — but not the **converse**: if `h⁻¹(e^{−ihκ} − 1) f` converges in `L²` as `h → 0⁺`, then `κ·f ∈ L²`. **The route is Fatou:** along a sequence `hₙ → 0⁺` the `L²` norms of the quotients converge to the limit's norm, the integrands converge pointwise to `|κ f|²` (`tendsto_phaseDefect`), so `∫ |κ f|² ≤ liminf ∫ |quotient|² < ∞` (`MeasureTheory.lintegral_liminf_le`), i.e. `f ∈ mulDomain κ`; then conjugate by `𝓕` for `freePositionOp`. **Deliverable:** `mem_mulDomain_of_hasGeneratorAt` and the corollary `schrodinger_mem_freePositionOp_domain : U(t)ψ ∈ freePositionOp.domain`, after which `schrodinger_classical` can be restated with `freePositionOp (U(t)ψ)` in place of the generator value `w`. ⚠️ Nothing consumes it; it closes the last gap between the semigroup-theoretic and the operator-theoretic readings of `dom H₀`. | ~~**M**~~ done (M) | Low — the honest last step of ~~#129~~ | ~~#129~~; [`PhaseGenerator.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/PhaseGenerator.lean); [`SchrodingerGenerator.lean`](../CsdLean4/Mathlib/Analysis/Semigroup/SchrodingerGenerator.lean) | > | ~~155~~ | Foundations | Records→spacetime | *new (2026-10-20, out of ~~#137~~)* **The skew reroute over the base: `LocalSettingLocalityB` for `Σ = KSigma N` with `epistemicMeasure p`.** ▸ **DONE 2026-10-20 (S–M), as routed.** [`RecordLayer/SkewBasinReroute.lean`](../CsdLean4/RecordLayer/SkewBasinReroute.lean), 15 pins. `rerouteBasin c c' x = (x.1, reroute (c.rates x.1) (c'.rates x.1) x.2)` — exchange along the tables at the point's OWN base point. **Measurability was the work and went as predicted:** `measurable_rectMap_param` (an affine rectangle map with six base-measurable parameters is jointly measurable on `Σ`), the column regularity (`measurable_colSum`/`colFrac`/`loSum_colFrac`/`loSum_colSum`, the pattern of `measurable_condRate`), the measurable `normalBasin`, and the glued maps over the base `normalizeB`/`denormalizeB` with ★ `measurable_normalizeB`/`_denormalizeB` — preimages split along the finitely many skew (normal) basins exactly as in the fibre proof. ★★ `measurePreserving_rerouteBasin` for `epistemicMeasure p`, every `p`: a.e. for `δ_p ⊗ vol` the base point is `p` (`ae_fst_eq`), so the reroute is a.e. `Prod.map id (reroute (c.rates p) (c'.rates p))`, which preserves `δ_p ⊗ vol` by the fibre theorem (`Measure.map_congr` + `MeasurePreserving.prod`). ★★ `wingB_rerouteBasin` / `wingA_eq_of_rowSum_eq'` pointwise from the fibre statements at `x.1`. The LF3 instance ON `Σ`: `skewBasinOutcomeMaps Q` for `Q : MeasurementContext → JointContextField N 2 2`, ★★★ `skewBasinLocalSettingLocalityA/B` for every `p`, ★★★ `operationalNoSignalling_skewBasin` = **operational no-signalling for the I3 record basins on `Σ` itself, for the epistemic measure of every preparation, derived from the primitive**; ★★★ `operationalNoSignalling_skewBasin_singlet` for the singlet as a constant field (`singletField`). The no-signalling of the tables is a hypothesis pointwise in `p`. ▸ *Original row:* ~~#137~~ instantiated the primitive on the fibre `T²` with the tables as data. ~~#135~~'s `JointContextField` attaches a table to every base point and `skewBasin c s t = {x \| x.2 ∈ skewCell (c.rates x.1) s t}`; the reroute is then `(p, θ) ↦ (p, reroute (c.rates p) (c'.rates p) θ)`, measure-preserving for `δ_p ⊗ vol` by `measurePreserving_reroute` fibrewise (`MeasurePreserving.prod` with the identity on the base), and B's reading carried by `wingB_reroute` at each `p` given `colSum` agreement pointwise in `p`. **The work is measurability of the glued map jointly in `(p, θ)`**: the piece selection `skewIdx (c.rates p) θ` depends on `p` through measurable rates (`measurable_loSum_rowSum`, `measurable_loSum_condRate` in `SkewBasin.lean`), so `{(p,θ) \| θ ∈ skewCell (c.rates p) s t}` is measurable and the glue is a finite measurable case split — the pattern of `measurableSet_skewBasin`. Deliverable: `skewBasinOutcomeMaps` on `KSigma N` and ★★★ `operationalNoSignalling_skewBasin` for `epistemicMeasure p`, every `p`. | ~~**S–M**~~ done (S–M) | Medium — it is the statement at the level `LF6`'s Born clause is stated at | ~~#137~~; ~~#135~~; [`RecordLayer/SkewBasin.lean`](../CsdLean4/RecordLayer/SkewBasin.lean); [`RecordLayer/SkewReroute.lean`](../CsdLean4/RecordLayer/SkewReroute.lean) | > | ~~156~~ | Foundations | Records→spacetime | *new (2026-10-20, out of ~~#137~~)* **The blind B-primitive is unsatisfiable at the I3 basins for the CHSH singlet — as a theorem.** ▸ **DONE 2026-10-20 (M), as routed.** [`RecordLayer/SkewRerouteObstruction.lean`](../CsdLean4/RecordLayer/SkewRerouteObstruction.lean), 8 pins. ★ `LF3.measure_counterfactual_eq_of_remoteSettingLocalityB` — **a blind B-reroute equates the counterfactual joint laws** `μ {wingB⟨a,b⟩ = t ∧ wingB⟨a,b'⟩ = t'}` at `a` and `a'` (the one relabelling carries the intersection at `a'` to the intersection at `a`; any `Σ`, any `μ`); `counterfactual_eq_iUnion` + ★ `volume_counterfactual` — at the skew cells of two tables with the same A-marginal the counterfactual set is `⋃ₛ A_s × (B_{s,t} ∩ B'_{s,t'})` with measure `∑ₛ |A_s|·|B ∩ B'|`; ★★ `volume_counterfactual_singlet` — for the singlet it is `cfJoint (a·b) (a·b') = ½min((1−d)/2,(1−d')/2) + ½min((1+d)/2,(1+d')/2)` (two initial arcs meet in the shorter, `circleCell_condRate_zero_inter`; `condRate s 0 = 2·P_st`); ★★★ `not_nonempty_remoteSettingLocalityB_singlet` — **no blind B-reroute exists for the singlet on the I3 basins**: at the circle settings `a = 0, a' = π/2, b = π/4, b' = −π/4` (`Empirical/QM/ColbeckRenner.lean`'s `circleSetting`, `dotR_circleSetting`) the law is `1/2` at `a` and `(1 − √2/2)/2` at `a'`, and `√2 ≠ 0`. The ~~#137~~ finding is now a theorem; `SettingLocalityLocal.lean` is the primitive the record layer satisfies, and this is why. Not formalised: the Fréchet-bound generalisation to arbitrary geometries with shared A-cells (prose in `SettingLocalityLocal.lean`). ▸ *Original row:* ~~#137~~ recorded the argument in prose: a `b`-blind `reroute a a'` with `wingB ⟨a',b⟩ ∘ reroute = wingB ⟨a,b⟩` for all `b` carries `R_t(a,b) ∩ R_{t'}(a,b')` onto `R_t(a',b) ∩ R_{t'}(a',b')`, so `vol (R_t(a,b) ∩ R_{t'}(a,b'))` is `a`-independent; at the I3 basins this is `∑ₛ rowSum s · |B-arc(s,t;a,b) ∩ B-arc(s,t';a,b')| = ½(1 − |c − c'|)` with `c = sin²((a−b)/2)`, `c' = sin²((a−b')/2)`, which is `(1 − √2/2)/2` at `a = 0` and `1/2` at `a = π/2` for `b, b' = π/4, 3π/4`. **Deliverable:** ★★★ `not_remoteSettingLocalityB_skew_chsh : ¬ Nonempty (RemoteSettingLocalityB volume (skewOutcomeMaps Q_singlet))` — the general lemma `measure_inter_eq_of_rerouteB` (any blind reroute equates the counterfactual intersections' measures), the arc-overlap computation `volume (circleCell r 0 ∩ circleCell r' 0) = min (r 0) (r' 0)` for two-cell partitions, and the trigonometry. ⚠️ Says what the record layer CANNOT do under the blind primitive, which is exactly why `SettingLocalityLocal.lean` exists; a reviewer will want it as a theorem, not a paragraph. | ~~**M**~~ done (M) | Medium — reviewer-proofing of the ~~#137~~ finding | ~~#137~~; [`LF3/SettingLocalityLocal.lean`](../CsdLean4/LF3/SettingLocalityLocal.lean); [`LF3/SettingLocality.lean`](../CsdLean4/LF3/SettingLocality.lean) | > | ~~157~~ | Foundations | Records→spacetime | *new (2026-10-20, out of ~~#137~~)* **The singlet's tables as a `MeasurementContext → JointRates 2 2` family, and the I3 basins' no-signalling for it.** ▸ **DONE 2026-10-20 (S).** [`RecordLayer/SingletSkewTables.lean`](../CsdLean4/RecordLayer/SingletSkewTables.lean), 8 pins: `singletJointRates a b : JointRates 2 2` from `P_st` (`P_st_nonneg`, normalisation from `marginal_a_eq_half`), `singletJointRates_rowSum`/`_colSum` = both marginals exactly `1/2`, `singletTables C := singletJointRates C.a C.b` with `singletTables_rowSum_eq`/`_colSum_eq` (no-signalling in the form the reroutes need); ★ `volume_skewCell_singlet` = **the skew cells carry `P_st` exactly**; ★★★ `skewLocalSettingLocalityBSinglet` (the rectangle exchange IS a B-reroute for the singlet), `skewLocalSettingLocalityASinglet` (identity), ★★★ `operationalNoSignalling_skew_singlet` = **the singlet on the I3 record basins is operationally no-signalling, derived from the primitive**, on the fibre partition whose cells carry the Bell-violating law. `LF3/OperationalNoSignalling.lean`'s `singlet_operational_no_signalling` was a kernel-level *verification*; this is the same statement *derived* from `LocalSettingLocality` on `Σ`'s fibre. The blind primitive is not instantiated and cannot be (~~#137~~; #156). ▸ *Original row:* ~~#137~~'s instance is for an arbitrary no-signalling family `Q`; the corpus's singlet kernel (`Singlet/Kernel`, `LF3/OperationalNoSignalling.lean`'s `P_st`) gives the concrete `Q⟨a,b⟩` with uniform marginals. Deliverable: `singletJointRates a b : JointRates 2 2` from `P_st`, `rowSum = colSum = fun _ => 1/2`, and the corollary `operationalNoSignalling_skew singletJointRates …` — so **the singlet on the I3 record basins is operationally no-signalling by the primitive**, closing the loop with `LF6`'s Bell violation on the same basins (`SkewBasin.lean`, `JointMomentContext.lean`). | ~~**S**~~ done (S) | Medium — the concrete instance the audit will ask for first | ~~#137~~; `LF3/OperationalNoSignalling.lean`; [`RecordLayer/SkewReroute.lean`](../CsdLean4/RecordLayer/SkewReroute.lean) | > | ~~158~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#88~~)* **The Riesz projector of a Hermitian matrix.** ▸ **DONE 2026-10-20 (M), as routed.** [`Mathlib/Analysis/Matrix/RieszProjector.lean`](../CsdLean4/Mathlib/Analysis/Matrix/RieszProjector.lean), 28 pins. `Matrix.rieszProj H c r := (2πi)⁻¹ • ∮_{C(c,r)} (z•1 − H)⁻¹` for ANY matrix (Mathlib's `circleIntegral`, under the `L2Operator` norm scope). **The design that made it M:** one operator `conjD d := U diag(d) U†` (`U` = `eigenvectorUnitary`), closed under `*`/`•`/`−`/`ᴴ` with the diagonals composing pointwise (`conjD_mul_conjD`, `conjD_smul`, `conjD_sub`, `conjD_conjTranspose`, `rank_conjD`, `conjD_mulVec_eigenvectorBasis`); `eq_conjD` is the spectral theorem as `H = conjD eigC`. ★ `inv_smul_one_sub_eq_conjD` — the resolvent is `conjD (fun j => (z − μⱼ)⁻¹)` off the spectrum (`inv_eq_left_inv`); `circleIntegral_sub_inv_eq_ite` — `∮ (z − μ)⁻¹ = 2πi` inside (`integral_sub_inv_of_mem_ball`) and `0` outside (Cauchy's theorem, `circleIntegral_eq_zero_of_differentiable_on_off_countable`); ★★ `rieszProj_eq_sum` / ★★ `rieszProj_eq_conjD` — **the Riesz projector is `conjD (discInd c r)`, the indicator of the disc on the eigenvalues**, by `integral_congr` on the sphere + `integral_fun_sum` + `integral_smul_const`. Corollaries in one line each: `rieszProj_mul_self`, `rieszProj_isHermitian`, `rieszProj_mul_comm` (`H P = P H`), ★ `mul_rieszProj_eq_smul` (`H P = λ • P` when the disc isolates `λ`), `rank_rieszProj` = `card {j // dist μⱼ c < r}`, `rieszProj_mulVec_eigenvectorBasis` (keep inside, kill outside). **Lesson for #159:** never `rw` the spectral theorem INTO a goal that mentions `hH` (the motive is ill-typed since `hH : H.IsHermitian`) — rewrite `← eq_conjD` from the `conjD` side, as `rieszProj_mul_comm` does. ▸ *Original row:* `rieszProj H c r := (2πi)⁻¹ ∮_{|z−c|=r} (z•1 − H)⁻¹ dz` as Mathlib's `circleIntegral` of the resolvent, for `H : Matrix (Fin N) (Fin N) ℂ` Hermitian with no eigenvalue on the circle (matrices under the `L2Operator` norm scope, a complete normed `ℂ`-space, so the integral exists). **Deliverable:** in the eigenbasis of `spectral_theorem` the resolvent is diagonal with entries `(z − μⱼ)⁻¹`, and `circleIntegral.integral_sub_inv_of_mem_ball` (and its outside-the-ball twin) evaluate each to `2πi` or `0`; hence ★★ `rieszProj_eq_sum_eigenprojections` — the projector onto the span of the eigenvectors with `|μⱼ − c| < r` — and the corollaries `rieszProj_idempotent`, `rieszProj_isHermitian`, `rieszProj_comm` (`H P = P H`), `rank_rieszProj` = the number of eigenvalues in the disc, `H * P = c-independent λ • P` when the disc isolates a single eigenvalue. Cat-1. ⚠️ **Stuck on:** nothing — Mathlib has every input | ~~**M**~~ done (M) | Medium — the object every step of Kato's proof is written in; Mathlib has `cfc` but no projector of an isolated part of the spectrum | ~~#88~~; `Mathlib/Analysis/Matrix/Spectrum.lean` (`spectral_theorem`, `mulVec_eigenvectorBasis`); `Mathlib/MeasureTheory/Integral/CircleIntegral.lean` | > | ~~159~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#88~~)* **Smoothness of the Riesz projector along a gapped path, and `PṖP = 0`.** ▸ **DONE 2026-10-20 (M), as routed.** [`Mathlib/Analysis/Matrix/RieszProjectorSmooth.lean`](../CsdLean4/Mathlib/Analysis/Matrix/RieszProjectorSmooth.lean), 28 pins. `GappedPath N` = `H`, `isHermitian`, a CONTINUOUS branch `lam`, `g > 0`, `gap : ∀ s j, μⱼ(s) = lam s ∨ g ≤ |μⱼ(s) − lam s|`; `Γ.proj s := rieszProj (H s) (lam s) (g/2)`, `Γ.mult s` = the multiplicity. ★ `HasDerivAt.mul_deriv_mul_eq_zero_of_idempotent` (any normed algebra: `P² = P` ⇒ `P Ṗ P = 0`); ★★ `contDiffOn_rieszProj` — **over a FIXED contour the Riesz projector is `C^n` in the matrix path** (the integrand `(r i e^{iθ}) • Ring.inverse (circleMap c r θ • 1 − H s)` is jointly `C^n` on the open set `W = Φ⁻¹' {units}` by `contDiffAt_ringInverse` + `Units.isOpen`, then `contDiffOn_intervalIntegral_of_contDiffOn` after rescaling `[0, 2π] → [0, 1]`); `proj_eq_near` — **the contour freezes locally**: for `|lam s − lam s₀| < g/4` the disc about `lam s₀` isolates the same eigenvalue of `H s`, so no continuity of eigenvalues is needed anywhere; ★★ `GappedPath.contDiff_proj` (`ContDiff ℝ n H → ContDiff ℝ n proj`); ★ `proj_mul_deriv_mul_proj` = **`P Ṗ P = 0`**; `H_mul_proj`/`proj_mul_H` (`H P = P H = lam • P`), `trace_proj` (`tr P = mult`), ★ `mult_eq` (**constant multiplicity**, trace continuous into `ℕ`), ★ `contDiff_lam` (**the branch is `C^n` too**, `lam = tr(H P)/mult`). **Library change:** `ContDiffParametricIntervalIntegral.lean`'s theorem generalised in place from `F : E → ℝ → ℝ` to any normed codomain `G` — the proof was unchanged, Poincaré rebuilt untouched (the row's 'repackage entrywise' was NOT needed: the matrix space under the `L2Operator` scope is a normed space, which is all the lemma ever used). Hypotheses are stated under that scope throughout. ▸ *Original row:* A **gapped path**: `H : ℝ → Matrix` Hermitian and `C^k` (entrywise, `k ≥ 1`), `lam : ℝ → ℝ` continuous, `g > 0`, with every eigenvalue of `H s` either equal to `lam s` or at distance `≥ g` from it. **Deliverable:** `P s := rieszProj (H s) (lam s) (g/2)` is `C^k` — entrywise, by `contDiffOn_intervalIntegral_of_contDiffOn` (#60's scalar-valued parametric integral; ⚠️ the operator-space instance gap means: differentiate each matrix entry of the integrand, package by linearity) with the resolvent `C^k` in `(s, z)` by `contDiffAt_ring_inverse`; `lam` is then `C^k` too (`lam s = tr (H s * P s) / tr (P s)`, so continuity of `lam` can even be weakened to 'the gap holds'); `Ṗ = P Ṗ + Ṗ P` from `P² = P`, hence ★ `P Ṗ P = 0`; constant rank by continuity. ⚠️ **Stuck on:** the parametric integral is scalar-valued at the pin: repackage entrywise, as #60 did | ~~**M**~~ done (M) | Medium — the differential geometry of the eigenbundle, in finite dimensions | ~~#88~~; #158; [`ContDiffParametricIntervalIntegral.lean`](../CsdLean4/Mathlib/Analysis/Calculus/ContDiffParametricIntervalIntegral.lean); `project_operator_space_instance_gap` | > | ~~160~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#88~~)* **The adiabatic intertwiner and Kato's intertwining identity.** ▸ **DONE 2026-10-20 (M), as routed.** [`Mathlib/Analysis/Matrix/AdiabaticIntertwiner.lean`](../CsdLean4/Mathlib/Analysis/Matrix/AdiabaticIntertwiner.lean), 18 pins (two of them `HasDerivAt.eq_mul_add_mul_of_idempotent` + `deriv_proj_eq`, exposed in #159's module). `Γ.K s := Ṗ P − P Ṗ` (`continuous_K`; ★ `conjTranspose_K` = anti-Hermitian, via `conjTransposeCLM` and `conjTranspose_deriv_proj` — the derivative of a Hermitian family is Hermitian); `Γ.IsIntertwiner T A` = `A 0 = 1 ∧ A' = K A` on `[0,T]` (one-sided at the ends). ★ `exists_isIntertwiner` by `exists_linearODE_solution_Icc` with coefficient `ContinuousLinearMap.mul ℝ _ (K s)` (bound from compactness, `opNorm_mul_apply_le`); ★ `IsIntertwiner.conjTranspose_mul_self`/`mul_conjTranspose`/`mem_unitaryGroup` (**unitary**: `(A†A)' = A†(K† + K)A = 0`, constancy by `constant_of_has_deriv_right_zero` wrapped as `eq_of_hasDerivWithinAt_zero_Icc`, then `mul_eq_one_comm`); ★★ `IsIntertwiner.conjTranspose_mul_proj_mul` (`A† P A ≡ P 0`) and ★★ **`IsIntertwiner.mul_proj_zero` = KATO'S INTERTWINING `A(s) P(0) = P(s) A(s)`**, from `deriv_proj_sub_comm_K`: `Ṗ − [K,P] = (Ṗ − ṖP − PṖ) + 2PṖP = 0` (#159's two identities, the rest is `noncomm_ring` after `P² = P`); `proj_mulVec` (vectors of the initial eigenspace land in the instantaneous one, exactly); `unique` (any two intertwiners agree: `(A†B)' = 0`). No ODE-uniqueness lemma was needed anywhere — unitarity does the work. ▸ *Original row:* `K s := [Ṗ s, P s]` is anti-Hermitian and continuous; `A : ℝ → Matrix` solves `A' = K A`, `A 0 = 1` on `[0, 1]` by `exists_linearODE_solution_Icc` (the Darboux work's linear ODE on a real normed space; matrices qualify), unique by `dist_le_of_linearODE_coeff_close` at `ε = 0`. **Deliverable:** ★ `A` is unitary (`(A†A)' = 0`), and ★★ **Kato's intertwining** `A s * P 0 = P s * A s` — `(A† P A)' = A†(Ṗ − [K,P])A = 0` because `Ṗ − [K,P] = Ṗ − (Ṗ − 2PṖP) = 2PṖP = 0` by #159. So `A` carries the initial eigenspace onto the instantaneous one, exactly. Cat-1. ⚠️ **Stuck on:** nothing — the ODE lemmas exist | ~~**M**~~ done (M) | Medium–high — the parallel transport of the eigenbundle, as an operator | ~~#88~~; #159; [`FlowDerivative.lean`](../CsdLean4/Mathlib/Analysis/ODE/FlowDerivative.lean) (`exists_linearODE_solution_Icc`, `dist_le_of_linearODE_coeff_close`) | > | ~~161~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#88~~)* **The two propagators at speed `T`.** ▸ **DONE 2026-10-20 (M), as routed, plus a library pass.** [`Mathlib/Analysis/Matrix/UnitaryPropagator.lean`](../CsdLean4/Mathlib/Analysis/Matrix/UnitaryPropagator.lean) + [`Mathlib/Analysis/Matrix/AdiabaticPropagators.lean`](../CsdLean4/Mathlib/Analysis/Matrix/AdiabaticPropagators.lean), 29 pins. **The library pass (rule of two):** #160's existence/unitarity/constancy arguments were about to recur twice, so they are now the general theory of `IsPropagator G T U` (`U 0 = 1`, `U' = G U` on `[0,T]`): ★ `exists_isPropagator` (continuous `G`), `IsPropagator.conjTranspose_mul_eq` (**Heisenberg constancy** `U† Y ≡ Y 0` for every solution `Y' = G Y`, anti-Hermitian `G`), ★ unitary, `eq_mul_init` (**uniqueness** `Y = U · Y 0`), `unique`, ★ `mul_eq_of_hasDerivWithinAt_comm` (**Heisenberg intertwining**: `Q' = [G,Q]` ⇒ `U Q(0) = Q U`); `IsIntertwiner` is now the abbreviation `IsPropagator Γ.K`, all #160 theorems are one-line instances, pins unchanged. **The row:** `physGen T s = −iT • H s`, `adGen T s = physGen + K` (both anti-Hermitian), `IsPhysPropagator`/`IsAdPropagator` on `[0,1]`; ★ both exist and are unitary; ★★ `IsAdPropagator.mul_proj_zero` = `U_A P(0) = P U_A` (because `[adGen, P] = [K, P]`, `H` commuting with `P`); `dynPhase T s = exp(−iT∫₀ˢ lam)`; ★★ **`IsAdPropagator.mul_proj_zero_eq`: `U_A(s) P(0) = e^{−iT∫₀ˢ lam} · A(s) P(0)`** — the rescaled `Z = e^{+iT∫lam} U_A P(0)` solves `Z' = K Z` (FTC for the phase, `H U_A P(0) = lam · U_A P(0)` from the intertwining), and `eq_mul_init` for the propagator `A` of `K` finishes. **FINDING: the multiplicity-one hypothesis the row planned is not needed** — the identity holds for a degenerate branch too, since `H P = lam • P` already does for any multiplicity. ▸ *Original row:* In the rescaled time `s = t/T`: `U_T` solves `U' = −iT·H(s) U`, `U 0 = 1` (the physical propagator) and `U_A` solves `U' = (−iT·H(s) + K(s)) U`, `U_A 0 = 1` (the adiabatic one). **Deliverable:** both exist on `[0,1]` (same linear ODE lemma), both are unitary, and ★ `U_A s * P 0 = P s * U_A s` by the #160 argument plus `[H s, P s] = 0`; ★ `U_A s * P 0 = e^{−iT ∫₀ˢ lam} • A s * P 0` on the eigenspace when the multiplicity is one (`H P = lam P`). The corpus's propagators were all constant-generator or rigidly rotated (`RotatingFrame.lean`'s `hasDerivAt_mul_of_gen`); this is the first time-dependent one. Cat-1. ⚠️ **Stuck on:** nothing | ~~**M**~~ done (M) | Medium — the objects the comparison is about | ~~#88~~; #160; [`RotatingFrame.lean`](../CsdLean4/Mathlib/QuantumInfo/RotatingFrame.lean) | > | ~~162~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#88~~)* **Kato's comparison: `‖U_T s − U_A s‖ ≤ C / T` uniformly on `[0, 1]`.** ▸ **DONE 2026-10-20 (L took L).** [`Mathlib/Analysis/Matrix/AdiabaticComparison.lean`](../CsdLean4/Mathlib/Analysis/Matrix/AdiabaticComparison.lean), 25 pins. ★★★ **`GappedPath.exists_norm_sub_le`: along a gapped `C²` path whose branch is an eigenvalue, `∃ C ≥ 0, ∀ T > 0, ∀ U_T U_A, ∀ s ∈ [0,1], ‖U_T s − U_A s‖ ≤ C / T`** — Kato's proof as routed: the **reduced resolvent** `redRes s = (2πi)⁻¹ ∮_{C(λ s, g/2)} (z − λ s)⁻¹ (z − H s)⁻¹ dz` (★ `IsHermitian.redResIntegral_eq_conjD` = `U diag(0 on the branch, (λ−μⱼ)⁻¹ off it) U†`, by the Cauchy evaluations `circleIntegral_sub_inv_mul_self` (double pole → 0) and `circleIntegral_sub_inv_mul_sub_inv` (partial fractions → `2πi (λ−μ)⁻¹`)), ★★ `contDiff_redRes` (the contour freezes locally exactly as for `P`; the general ★ `contDiffOn_circleIntegral` + `contDiffOn_inv_comp` were added to #159's module and `contDiffOn_rieszProj` re-based on them), the four identities `(λ−H)X = X(λ−H) = 1 − P`, `XP = PX = 0`, then `Y = XṖP + PṖX` with ★★ `Y_mul_H_sub` = **`Y H − H Y = K`** (pure algebra, no `PṖP = 0` needed) and `contDiff_Y` (`C¹`, where the `C²` of the path is spent). The estimate: `W = U_T† U_A` has `W' = U_T† K U_A`; `(U_T† Y U_A)' = −iT · U_T† K U_A + U_T†(Ẏ + YK) U_A`; so `F = W + (iT)⁻¹ U_T† Y U_A` has `‖F'‖ ≤ ME/T` (`norm_sub_le_of_hasDerivWithinAt_Icc`, the mean-value bound), and `‖W − 1‖ ≤ (2·sup‖Y‖ + sup‖Ẏ + YK‖)/T` with unitaries of operator norm `≤ 1` (`norm_le_one_of_conjTranspose_mul_self`, the C*-identity); `U_A − U_T = U_T (W − 1)`. `C = 2 sup_{[0,1]}‖Y‖ + sup_{[0,1]}‖Ẏ + YK‖` — depends on the path only (the row's `C(g, ‖Ḣ‖, ‖Ḧ‖)` is honoured as a dependence, not made a formula; that is a residue a reviewer may ask for: NEW ROW #166, S–M). ▸ *Original row:* The core. The **reduced resolvent** `X s := (2πi)⁻¹ ∮ (z − lam s)⁻¹ (z − H s)⁻¹ dz` (or algebraically `(1 − P)(H − lam)⁻¹(1 − P)`) satisfies `[H s, Y s] = K s` for `Y := X Ṗ P − P Ṗ X` (because `P` is the `lam`-eigenprojection and `K = [Ṗ, P]` maps `range P` to its complement and back); `Y` is `C¹` with `‖Y‖, ‖Ẏ‖ ≤ C(g, ‖Ḣ‖, ‖Ḧ‖)` — this is where `C²` of the path is spent. Then `U_T† U_A − 1 = ∫₀ˢ U_T† K U_A = ∫₀ˢ U_T† [iT·H… ]` — writing `K = [H, Y]` and `d/ds (U_T† Y U_A) = U_T†(iT[H,Y] + Ẏ + Y K)U_A`, integration by parts turns the integral into `(1/T)·(boundary term + ∫ U_T†(Ẏ + Y K)U_A)`, bounded by `C/T` since all propagators are unitary. **Deliverable:** ★★★ `norm_propagator_sub_adiabatic_le`. Needs operator-valued interval integration by parts (`HasDerivAt.mul` + `intervalIntegral.integral_eq_sub_of_hasDerivAt` for matrix-valued maps, fine under the `L2Operator` scope) and the bounds from #159. ⚠️ **Stuck on:** the price, not a wall — the longest single argument of the split | ~~**L**~~ done (L) | **High** — the theorem; everything else in the split is scaffolding for this row | ~~#88~~; #159; #160; #161; Kato 1950; Teufel, *Adiabatic Perturbation Theory*, Thm 2.2 | > | ~~163~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#88~~)* **The Born–Fock statement for states.** ▸ **DONE 2026-10-20 (S).** [`Mathlib/Analysis/Matrix/AdiabaticBornFock.lean`](../CsdLean4/Mathlib/Analysis/Matrix/AdiabaticBornFock.lean), 14 pins. States are points of `EuclideanSpace ℂ (Fin N)`, a matrix acts through `Matrix.toEuclideanCLM` (so the `L2Operator` norm is exactly the norm of the action, `norm_toEuclideanCLM_apply_le`). ★★★ **`GappedPath.exists_norm_one_sub_proj_le`: `∃ C ≥ 0, ∀ T > 0, ∀ U_T, ∀ s ∈ [0,1], ∀ ψ₀, P 0 ψ₀ = ψ₀ → ‖(1 − P s)(U_T s ψ₀)‖ ≤ (C/T)·‖ψ₀‖`** — the row's sentence, with the `‖ψ₀‖` factor (unit vectors give the row's form). Also ★★ `exists_norm_toEuclideanCLM_sub_le` (`‖U_T s ψ − U_A s ψ‖ ≤ (C/T)‖ψ‖` for every `ψ`, #162 on vectors), ★ `IsAdPropagator.proj_toEuclideanCLM_apply` / `toEuclideanCLM_apply_mem_range` / `one_sub_proj_toEuclideanCLM_apply` (`U_A s ψ₀ ∈ range (P s)` *exactly*, #161 on vectors), ★★ `exists_norm_proj_le` (the survival form `‖ψ₀‖ − (C/T)‖ψ₀‖ ≤ ‖P s (U_T s ψ₀)‖`, via `IsPhysPropagator.norm_toEuclideanCLM_apply` = unitaries preserve the norm of a state), and the library lemmas `norm_le_one_of_conjTranspose_mul_self_eq` (a Hermitian idempotent has norm `≤ 1`, C*-identity), `norm_proj_le`, `norm_one_sub_proj_le`, `mem_range_proj_iff`. Proof: `(1 − P s) U_T ψ₀ = (1 − P s)(U_T − U_A) ψ₀` since `(1 − P s) U_A ψ₀ = 0`, then `‖1 − P s‖ ≤ 1` and #162. `C` is #162's constant (compactness — #166 makes it a formula). ▸ *Original row:* For `ψ₀ ∈ range (P 0)`: ★★★ `‖(1 − P s) (U_T s ψ₀)‖ ≤ C / T` for all `s ∈ [0,1]` — **the solution started in the instantaneous eigenspace stays in it up to `O(1/T)`, uniformly** — and `U_T s ψ₀ = U_A s ψ₀ + O(1/T)` with `U_A s ψ₀ ∈ range (P s)` exactly. Immediate from #161 and #162. This is the sentence ~~#88~~ was named for. ⚠️ **Stuck on:** nothing once #162 lands | ~~**S–M**~~ done (S) | High — the textbook theorem in its textbook form | ~~#88~~; #162 | > | ~~164~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#88~~)* **The phase: BP-1's geometric phase is what survives the adiabatic limit.** ▸ **DONE 2026-10-20 (M took M).** [`Mathlib/Analysis/Matrix/AdiabaticPhase.lean`](../CsdLean4/Mathlib/Analysis/Matrix/AdiabaticPhase.lean), 24 pins. **The identification, as routed (uniqueness of ODE):** for a `C¹` unit section `ψ` of the eigenline (`P t ψ t = ψ t`, `‖ψ t‖ = 1`, all `t`), ★ `K_apply_section` **`K ψ = (1 − P) ψ̇`** (differentiate `P ψ = ψ`), and for a nondegenerate branch (`Γ.mult 0 = 1`; `rank_proj : rank (P s) = mult s`) ★★ `K_apply_unit_section` **`K ψ = ψ̇ − i·Im⟪ψ, ψ̇⟫·ψ`** (rank one: `P v = ⟪ψ, v⟫ ψ`, `toEuclideanCLM_apply_eq_inner_smul`; unit norm: `⟪ψ, ψ̇⟫ = i Im⟪ψ, ψ̇⟫`, `inner_deriv_eq_im_mul_I`) — Kato's generator acts on the section as covariant differentiation. So ★★ `hasDerivAt_horizontalLift_K`: **`horizontalLift ψ` solves `χ' = K χ`**, and the new vector-level uniqueness for propagators (★ `IsPropagator.conjTranspose_apply_eq`/`eq_apply_init`, Heisenberg constancy of `U† v`; derivative of `s ↦ M s (v s)` via `toEuclideanCLMᵣ` + `HasDerivWithinAt.clm_apply`) gives ★★★ **`IsIntertwiner.toEuclideanCLM_apply_eq_horizontalLift`: `A s (ψ 0) = horizontalLift ψ s` on `[0,1]`**. **The phase:** ★★ `IsIntertwiner.toEuclideanCLM_apply_one_eq` **`A 1 ψ₀ = e^{iβ} ψ₀`, `β = geometricPhase ψ 1 φ`** for a closed loop `ψ 1 = e^{iφ} ψ 0` (`horizontalLift_cyclic`); `exists_apply_one_eq` (the holonomy exists without choosing a section, from `P 1 = P 0`); `exists_cyclic_of_proj_one_eq` (rank one + `P 1 = P 0` ⇒ the section closes up to a phase), `proj_one_eq_of_closed` (`H 1 = H 0 ∧ lam 1 = lam 0 ⇒ P 1 = P 0`). ★★ `IsAdPropagator.toEuclideanCLM_apply_eq` + `exists_norm_sub_dynPhase_smul_le` (**`‖U_T s ψ₀ − e^{−iT∫₀ˢ lam} A s ψ₀‖ ≤ (C/T)‖ψ₀‖`**, any multiplicity — the adiabatic approximation with the parallel-transported eigenvector), and ★★★ **`exists_norm_sub_phase_smul_le`: `‖U_T 1 (ψ 0) − e^{−iT∫₀¹ lam} e^{iβ} (ψ 0)‖ ≤ C/T`** — dynamical phase times Berry's phase with #162's error. Hypotheses honest: the section is global (`ℝ`, the path is), `β` is section-independent by BP-1's `geometricPhase_rephase`; `C` is #162's constant (#166). ▸ *Original row:* For a nondegenerate eigenvalue (`rank P = 1`) and a closed path (`H 1 = H 0`, so `P 1 = P 0`): `s ↦ A s ψ₀` is a unit section of the eigenline with `⟪A s ψ₀, (A s ψ₀)'⟫ = ⟪ψ, K ψ⟫ = 0` — it is **BP-1's horizontal lift** (`GeometricPhase.lean`'s `horizontalLift`, `connectionForm = 0`), so `A 1 ψ₀ = e^{iβ} ψ₀` with `β` the geometric phase of the eigenvector loop, and ★★★ `U_T 1 ψ₀ = e^{−iT∫₀¹ lam} e^{iβ} ψ₀ + O(1/T)` — dynamical phase plus Berry's phase, with the error of #162. **Deliverable:** the identification of `A` with the horizontal lift, and the phase theorem. ⚠️ **Stuck on:** matching `horizontalLift`'s definition to `A s ψ₀`: a uniqueness-of-ODE step | ~~**M**~~ done (M) | High — closes the loop ~~#56~~ opened: the general theorem behind Berry's model | ~~#88~~; #163; #160; [`GeometricPhase.lean`](../CsdLean4/Mathlib/Analysis/InnerProductSpace/GeometricPhase.lean) (`horizontalLift`, `geometricPhase`, `rephase_cyclic`); [`berry-phase-scoping.md`](berry-phase-scoping.md) BP-1, BP-5 | > | ~~165~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#88~~)* **~~#56~~'s rigid rotation as an instance of the general theorem.** ▸ **DONE 2026-10-20 (S took S).** [`Empirical/QM/BerryKatoInstance.lean`](../CsdLean4/Empirical/QM/BerryKatoInstance.lean), 31 pins (4 general lemmas in `MathlibStaging`, incl. `GappedPath.proj_toEuclideanCLM_apply_of_eq` added to #163's module: an eigenvector of `H s` for `lam s` is fixed by `P s`, via `X(λ − H) = 1 − P`). **The instance:** `fieldHam_mul_self` `H(t)² = (B/2)²·1`, `fieldHam_trace = 0` ⇒ eigenvalues exactly `±B/2`, one each (`card_eigenvalues_eq_one`; general `IsHermitian.eigenvalues_sq_eq_of_mul_self`, `Fintype.card_subtype_eq_one_of_two`); ★★ **`conePath θ ω hB : GappedPath 2`** (branch `B/2`, gap `B`), `contDiff_fieldHam` (every order), `conePath_mult_eq_one`, `conePath_hex`, `conePath_proj_one_eq` (`ω = 2π`); ★★ `conePath_isPhysPropagator`: ~~#56~~'s exact `V(Ts)` is the physical propagator at speed `T` of `fieldHam θ (ωT) B`; `coneSection_*`: BP-2's cone curve is a `C¹` unit section of the eigenline, closed for `ω = 2π`, with geometric phase `−π(1 − cos θ)`. **The regression:** ★★★ `exists_norm_state_sub_phase_le` — `∃ C, ∀ ω > 0 (Ω ≠ 0), ‖ψ(2π/ω) − e^{i(−(B/2)(2π/ω) − π(1 − cos θ))} ψ₀‖ ≤ C ω/(2π)` — **the same dynamical and geometric phases as ~~#56~~'s exact `norm_inner_state_sub_exp_le`**; `exists_norm_inner_state_sub_exp_le` is the overlap form, same shape as ~~#56~~. **The honest comparison of constants:** `berry_bound_eq_kato_form`: ~~#56~~'s `πω/(2(B−ω)) + (ω/(B−ω))² = π²/((B−ω)T) + 4π²/((B−ω)²T²)` in Kato's variable `T = 2π/ω` (a `1/T` term with constant `π²/(B−ω) → π²/g`, plus `1/T²`); `berry_bound_le_kato_form`: for `ω ≤ B/2` it is `≤ (2π(π+2)/B)·ω/(2π)`, i.e. in this model the overlap statement holds with the **explicit** constant `C = 2π(π+2)/g` — a datapoint for #166 (the general `C` is still by compactness). No wrong constant caught: the general theorem specialises exactly. `BerryAdiabatic.lean`'s header no longer says the general theorem is unclaimed. ▸ *Original row:* `H t = R_z(ωt) H₀ R_z(ωt)ᴴ` with eigenvalue `B/2` and gap `B` is a gapped path; instantiate #158–#164 and compare the general constant `C/T` with ~~#56~~'s exact `πω/(2(B−ω)) + (ω/(B−ω))²`. A regression and an honest comparison, nothing more. ⚠️ **Stuck on:** nothing | ~~**S**~~ done (S) | Low — a check that the general theorem specialises correctly; its value is in catching a wrong constant | ~~#88~~; ~~#56~~; [`Empirical/QM/BerryAdiabatic.lean`](../CsdLean4/Empirical/QM/BerryAdiabatic.lean) | > | ~~166~~ | QIT | One QM, many formulations | *new (2026-10-20, out of ~~#162~~)* **An explicit adiabatic constant.** ▸ **DONE 2026-10-21 (S–M took M).** [`Mathlib/Analysis/Matrix/AdiabaticConstant.lean`](../CsdLean4/Mathlib/Analysis/Matrix/AdiabaticConstant.lean), 43 pins. ★★★ **`GappedPath.norm_sub_le_explicit`: with `‖Ḣ‖ ≤ M₁`, `‖Ḧ‖ ≤ M₂` on `[0,1]`, `‖U_T s − U_A s‖ ≤ (8M₁/g² + 4M₂/g² + 80M₁²/g³)/T`** — Teufel Thm 2.2's shape (`c₁‖Ḣ‖²/g³ + c₂‖Ḧ‖/g² + c₃‖Ḣ‖/g²`), constants from the termwise triangle inequality, not optimal. **FINDING — the row's stuck-on was a false wall:** no differentiation under `∮` is needed. Differentiating `H P = P H` and multiplying by the reduced resolvent gives ★★ `deriv_proj_eq_redRes` **`Ṗ = X Ḣ P + P Ḣ X`** (Kato's formula) by algebra from `X(λ−H) = 1−P`, `XP = PX = 0`, `PṖP = 0`; differentiating `X(λ−H) = 1−P` and `XP = 0` gives ★★ `deriv_redRes_eq` **`Ẋ = −ṖX − λ̇X² + XḢX − XṖ`** (Kato II-§2.3), with ★ `proj_mul_deriv_H_mul_proj` **`PḢP = λ̇P`** and `|λ̇| ≤ ‖Ḣ‖` (needs `‖P‖ ≥ 1`, `one_le_norm_proj`). Norms: ★ `norm_redRes_le` `‖X‖ ≤ 1/g` (`IsHermitian.norm_conjD_le`: `‖U diag(d) U†‖ ≤ sup|dⱼ|` via `l2_opNorm_diagonal`), `‖Ṗ‖ ≤ 2M₁/g`, `‖Ẋ‖ ≤ 6M₁/g²`, `‖P̈‖ ≤ 16M₁²/g² + 2M₂/g`, `‖Y‖ ≤ 4M₁/g²`, `‖K‖ ≤ 4M₁/g`, `‖Ẏ + YK‖ ≤ 80M₁²/g³ + 4M₂/g²`; #162's estimate refactored into `norm_sub_le_of_bounds` (suprema as hypotheses; `exists_norm_sub_le` is now its compactness corollary, pins unchanged). Idioms: the matrix identities are `simp only [mul_assoc, …, the right-multiplied basic relations]` into a right-associated normal form then `linear_combination (norm := module) ±h`, which handles `ℂ`-scalars and matrix atoms. **Check against ~~#165~~ (by hand, not in Lean):** for the spin-½ cone path `‖Ḣ‖ = πB sin θ`, `‖Ḧ‖ = 2π²B sin θ`, `g = B`, so `C = (8π sin θ + 8π² sin θ + 80π² sin²θ)/B ≈ 894/B` at `θ = π/2`, versus the exact model's `2π(π+2)/B ≈ 32/B` for the overlap — consistent (the general bound must be larger) and a measure of how loose the termwise constants are. ▸ *Original row:* ~~#162~~'s `C = 2 sup‖Y‖ + sup‖Ẏ + YK‖` exists by compactness; make it a formula in the gap and the derivatives of the path: `‖X‖ ≤ 1/g` (the L2 operator norm of `conjD d` is `sup|dⱼ|` — needs `Matrix.l2_opNorm` of a diagonal, which Mathlib has for `diagonal` but the corpus has not yet used), `‖Ṗ‖ ≤ (2/g)·sup‖Ḣ‖` via `Ṗ = (2πi)⁻¹ ∮ R Ḣ R dz` (differentiation under the contour integral — the derivative formula, not just smoothness), and `‖P̈‖` similarly with `Ḧ`, giving `C ≤ c₁ ‖Ḣ‖²/g³ + c₂ ‖Ḧ‖/g² + …` (Teufel Thm 2.2's form). **Deliverable:** `exists_norm_sub_le` with an explicit `C`. ⚠️ **Stuck on:** the derivative of the Riesz projector as a contour integral (differentiation under `∮`), which #159 proved smoothness of without computing — an M on its own. | ~~**S–M**~~ done (M) | Low–medium — reviewer-proofing; the theorem is already uniform in `T` | ~~#162~~; #165 (the rigid-rotation instance would check the constant); Teufel, *Adiabatic Perturbation Theory*, Thm 2.2 | > | ~~167~~ | QIT | One QM, many formulations | *new (2026-10-21, out of #64 — the interacting Wigner equation, first brick)* **The Wigner function on `L²`, and its time derivative along the interacting orbit.** ▸ **DONE 2026-10-21 (M).** [`Mathlib/Analysis/Fourier/WignerL2.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WignerL2.lean), 21 pins. The Schwartz-only `wigner` of `Wigner.lean` cannot be differentiated along `U(t)ψ ∈ L²`; the **momentum form** can: ★ `wignerL2 φ ψ x ξ = 2⟪ φ̂(ξ − ·), e(2x·) ψ̂(ξ + ·) ⟫_{L²}`, an inner product of two *continuous linear images* of the states (`wignerLeft` = reflection ∘ Plancherel, `wignerRight` = unimodular multiplier ∘ translation ∘ Plancherel, both isometries up to `‖phase‖_∞ ≤ 1`), so integrability, the bound ★ `norm_wignerL2_le` (`≤ 2‖φ‖‖ψ‖`) and the chain rule come for free. ★★ `wignerL2_toLp`: **on a Schwartz state it is `Wigner.lean`'s `wigner`** (`wigner_fourier` + Mathlib's `SchwartzMap.toLp_fourier_eq` + the substitution `y = 2u`); ★★ `hasDerivAt_wignerL2`: `∂_t W(ψ_t, ψ_t) = W(ψ̇, ψ) + W(ψ, ψ̇)` along any `L²`-differentiable path (`HasDerivAt.inner` through the two CLMs); ★★★ **`hasDerivAt_wignerL2_schrodinger`: along `ψ_t = U(t)ψ₀` (bounded real `V`, `ψ₀ ∈ dom H₀`, `t > 0`), `∂_t W_{ψ_t}(x, ξ) = W(−i(H₀+V)ψ_t, ψ_t) + W(ψ_t, −i(H₀+V)ψ_t)`** with `H₀` acting on the orbit point (~~#154~~'s `hasDerivAt_schrodinger_orbit`) — the interacting Wigner equation in operator form. Idioms: `Lp.compMeasurePreservingₗᵢ ℂ` for translations (`measurePreserving_add_left volume ξ`; the subtraction one is `Measure.measurePreserving_sub_left`), `HeatSemigroup.potential` of an `L^∞` phase class for the multiplier, `QuasiMeasurePreserving.ae_eq_comp` to move `coeFn_toLp` across a change of variable, `Measure.integral_comp_mul_left` for `y = 2u`. ⚠️ The two halves of the right-hand side are not yet identified — that is #168 and #169. | ~~**M**~~ done (M) | Medium-high — the equation ~~#64~~ named, now one identification away on each side | #64; ~~#129~~; ~~#154~~; [`Wigner.lean`](../CsdLean4/Mathlib/Analysis/Fourier/Wigner.lean) | > | ~~168~~ | QIT | One QM, many formulations | *new (2026-10-21, out of #64)* **The free half of the interacting Wigner equation on `dom H₀`: `W(−iH₀ψ, ψ) + W(ψ, −iH₀ψ) = −2πξ ∂_x W_ψ`.** ▸ **DONE 2026-10-21 (M–L took M).** [`Mathlib/Analysis/Fourier/WignerFree.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WignerFree.lean), 21 pins. **The `x`-derivative came from a theorem already in the corpus, not from a dominated-convergence argument of its own:** `e(2xu) = e^{−ix·κ(u)}` with `κ(u) = −4πu`, so the right slot `R_{x,ξ}ψ = e(2x·)ψ̂(ξ+·)` **is** `phaseGroup κ x` applied to a fixed `L²` function (★ `wignerRight_eq_phaseGroup`), and `PhaseGenerator.lean`'s `hasDerivAt_phaseGroup_apply` differentiates it in `L²` once `κ·ψ̂(ξ+·) ∈ L²` — which `ψ ∈ dom H₀` gives (`memLp_id_mul_of_mem`: `|η| ≤ 1 + 2π²η²`, so `(1+2π²η²)ψ̂ ∈ L² ⇒ ηψ̂ ∈ L²`; `memLp_shift_mul`, `memLp_kappa_shift` for the translates). ★★ `hasDerivAt_wignerRight`, ★★ `hasDerivAt_wignerL2_x` (through `HasDerivAt.inner` with the constant left slot), `deriv_wignerL2_x_eq` = `2∫ 4πiu·e(2xu) ψ̂(ξ+u) conj φ̂(ξ−u) du`. `fourier_freePositionOp` (`𝓕(H₀ψ) = 2π²η²·ψ̂` on `dom H₀`, by `conjIsometry_apply_of_eq`; `mem_freeSymbolOp_domain_of_mem` unpacks `freePositionOp_domain`), then ★★★ **`wignerL2_freePositionOp_add`: `W(−iH₀φ, ψ) + W(φ, −iH₀ψ) = −2πξ ∂_x W(φ, ψ)` for `φ, ψ ∈ dom H₀`** — pointwise algebra under the momentum integral, `(ξ−u)² − (ξ+u)² = −4ξu`, with the two `H₀` coefficients moved across the translations by `QuasiMeasurePreserving.ae_eq_comp`. Also `wignerL2_add_left/right`, `wignerL2_smul_left/right` (sesquilinearity, for #169's splitting of `H₀ + V`) and `integrable_wignerL2_integrand`. Idioms: `abs_of_nonneg (by positivity : (0:ℝ) ≤ …)` must be type-ascribed inside `rw` (the side goal is elaborated before the rewrite target is known); pull constants out of integrals with the *forward* `integral_const_mul c` on the side that has them, never `←` on a goal with two integrals. ▸ *Original row:* In the momentum form of ~~#167~~ both sides are integrals over `η` of `conj ψ̂(ξ − η/2) ψ̂(ξ + η/2) e(xη)` with weights: `H₀` multiplies by `2π²ξ²` (`freeSymbolOp`, `coeFn_freeSymbolOp`), and `(ξ+η/2)² − (ξ−η/2)² = 2ξη`, so the left side is `−2πξ · ∫ 2πiη (…)` — which is `∂_x W` once the `x`-derivative passes under the integral. **Deliverable:** `hasDerivAt_wignerL2_left` (the `x`-derivative of `wignerL2 φ ψ` for `φ, ψ ∈ dom H₀`, dominating function `|η|·|φ̂(ξ−η/2)|·|ψ̂(ξ+η/2)|`, integrable because `(1+ξ²)ψ̂ ∈ L²` gives `ηψ̂ ∈ L²`), and the identity. ⚠️ **Stuck on:** nothing — the dominating function is Cauchy–Schwarz with the extra factor `η` absorbed by the domain condition; the work is the `coeFn` plumbing of `freePositionOp` on the orbit point (`freePositionOp_apply_fourierL2`). | ~~**M–L**~~ done (M) | Medium-high — half of the equation | ~~#167~~; `FreeHamiltonian.lean`; `WignerEvolution.lean` (the Schwartz case, `hasDerivAt_wigner_left`) | > | ~~169~~ | QIT | One QM, many formulations | *new (2026-10-21, out of #64)* **The potential half and the assembled equation: `W(−iVψ, ψ) + W(ψ, −iVψ) = moyalPot V ψ`, hence `∂_t W_{U(t)ψ} = −2πξ ∂_x W + moyalPot V (U(t)ψ)`.** ▸ **DONE 2026-10-21 (M).** [`Mathlib/Analysis/Fourier/WignerPotential.lean`](../CsdLean4/Mathlib/Analysis/Fourier/WignerPotential.lean), 20 pins. **The stuck-on step — the `L²` Plancherel identity between the position and momentum forms — went by density, with no Fourier transform of any translate or modulation computed on `L²`:** the position form ★ `wignerPos φ ψ x ξ = 2⟪φ(x−·), e(−2ξ·)ψ(x+·)⟫` is, like the momentum form, an inner product of continuous linear images of the states (`reflectAt x`, `posRight x ξ = potential (phaseLinf (−ξ)) ∘ shiftAt x`), so both forms are continuous on `L² × L²` (`continuous_wignerPos`, `continuous_wignerL2`); on Schwartz states both are the cross Wigner function `wigner₂` — `wignerPos_toLp` by the substitution `y = 2v`, and `wignerL2_toLp₂` through `wignerL2_eq_wignerPos_fourier` (**the momentum form IS the position form of the transforms at `(ξ, −x)`, by unfolding**) and the cross Plancherel symmetry ★★ `wigner₂_fourier` (the two-state form of `wigner_fourier`, same proof); `SchwartzMap.denseRange_toLpCLM` + `DenseRange.induction_on₂` + `isClosed_eq` then give ★★★ `wignerPos_eq_wignerL2`. Then ★ `moyalPotL2 V φ ψ` (the position integral against `V(x+v) − V(x−v)`; ★★ `moyalPotL2_toLp` = `moyalPot` on Schwartz states), ★★★ `wignerPos_potential_add` / ★★★ `wignerL2_potential_add`: **`W(−iVφ, ψ) + W(φ, −iVψ) = moyalPotL2 V φ ψ`** for `VL =ᵐ V` real (pointwise under the integral, `conj(−iV(x−v)) − iV(x+v) = −i(V(x+v) − V(x−v))`), `wignerL2_potential_add_toLp` with `moyalPot` on the right, and ★★★ **`hasDerivAt_wignerL2_schrodinger_eq`: along `ψ_t = U(t)ψ₀`, `ψ₀ ∈ dom H₀`, `t > 0`, `∂_t W_{ψ_t}(x, ξ) = −2πξ ∂_x W_{ψ_t}(x, ξ) + moyalPotL2 V ψ_t ψ_t (x, ξ)`** — the Wigner–Moyal evolution equation for `H = 2π²ξ² + V(x)` on `L²`, assembled from ~~#167~~ (`∂_t W = W(−i(H₀+V)ψ, ψ) + W(ψ, −i(H₀+V)ψ)`), ~~#168~~ (the `H₀` half) and the potential half by sesquilinearity. **Row ~~#64~~ is closed.** Idiom: `LinearPMap.addCLM_apply` does not fire under `simp` on the orbit term — `change … (freePositionOp ⟨_, hmem⟩ + potential VL ψ_t) …` with the orbit point *spelled out* (a `_` there leaves `Submodule.subtype ⟨ψ_t, _⟩`, a different atom for `ring`, and `rw [Submodule.subtype_apply]` is rejected on motive grounds). ▸ *Original row:* `moyalPot` (`MoyalBracket.lean`) is defined for Schwartz `ψ` as the integral against the symmetric difference `V(x + y/2) − V(x − y/2)`; the identity with the `L²` form is pointwise algebra of integrands (`coeFn_potential`) plus the position-form of ~~#167~~'s `wignerL2` (the Plancherel step `wignerL2_eq_integral` in the `y`-variable, i.e. the `L²` version of `wigner_fourier`). **Deliverable:** `moyalPotL2` on `L²` for bounded `V`, `moyalPotL2_toLp = moyalPot` on Schwartz states, and ★★★ the equation `∂_t W_{U(t)ψ}(x, ξ) = −2πξ ∂_x W_{U(t)ψ}(x, ξ) + moyalPotL2 V (U(t)ψ) x ξ` for `ψ₀ ∈ dom H₀`, `t > 0`, assembled from ~~#167~~ and #168. ⚠️ **Stuck on:** the `L²` Plancherel identity between the position and momentum forms of `wignerL2` (an `L²` inner product computed two ways, `Lp.fourierTransformₗᵢ` is unitary — S–M on its own). | ~~**M**~~ done (M) | Medium-high — the row ~~#64~~ was named for | ~~#167~~; #168; `MoyalBracket.lean` (`moyalPot`, `moyalPot_eq_comm`) | > | ~~170~~ | QIT | Library owed / upstream | *new (2026-10-21, out of #121(ii) — the partial Fourier transform on the plane, brick 1 of 5)* **`𝓕₂K` is smooth on the plane, with every derivative bounded uniformly: `(u, ξ) ↦ ∫ e(−wξ) K(u, w) dw` is `C^∞` on `ℝ × ℝ` and `‖∂ⁿ 𝓕₂K(u, ξ)‖ ≤ Cₙ(K)` for all `(u, ξ)`.** ▸ **DONE 2026-10-21 (M–L took S–M).** [`Mathlib/Analysis/Fourier/PartialFourierSmooth.lean`](../CsdLean4/Mathlib/Analysis/Fourier/PartialFourierSmooth.lean), 24 pins. `SchwartzMap.partialFourier K p = ∫ e(−w p.2) K(p.1, w) dw` (`partialFourier_eq_sliceFourier`: at `(u, ξ)` it is `PartialFourier.lean`'s slice-wise object). ★★ `contDiff_partialFourier` (`C^∞` on the plane) and ★★ `norm_iteratedFDeriv_partialFourier_le`: **`‖∂ⁿ 𝓕₂K(p)‖ ≤ leibnizConst n · kSup K n · π` for every `p`**, with `kSup K n` the sup of `K`'s seminorms over `Iic (n+2, n)` — linear in one `Finset.sup`, the shape #173/#174 need. **As priced, the content was the bound family and it is one Leibniz step:** `norm_iteratedFDeriv_mul_le` on `e(−wξ) · K(u, w)`; the phase's parameter derivatives have norm exactly `(2π|w|)ⁱ` (★ `norm_iteratedFDeriv_phase_le`: the phase is `exp(c·ξ) ∘ snd` with `|c| = 2π|w|`, by `norm_iteratedFDeriv_comp_clm_le` + `norm_iteratedFDeriv_eq_norm_iteratedDeriv` + a four-line `iteratedDeriv_cexp_const_mul_ofReal`); the kernel factor's `u`-derivative is the full derivative through `fst` then `inl` (`norm_iteratedFDeriv_slice_fst_le`); and ★ `one_add_abs_pow_mul_norm_iteratedFDeriv_le` moves `(1+|w|)ⁱ` onto two extra orders of `K`'s decay, giving `(1+w²)⁻¹` uniformly in `u` — the mirror of ~~#126~~(ii)'s weight transfer with the slots swapped. The rest is ~~#124~~'s `contDiffOn_integral_of_iteratedFDeriv_bound` and `norm_iteratedFDeriv_integral_le` with ~~#126~~(ii)'s `prodDirDeriv` measurability lemmas, exactly as the row said. Snags: `fun_prop` has no rule for a constant `Mul.mul` factor in `2π · (−(q.2 q.1.2))` (write `contDiff_const.mul …`); a lambda `fun x => c ^ n * exp (c * x)` elaborates over `ℂ` unless `x : ℝ` is ascribed; `Finset.sup_mono` on seminorms needs `Seminorm.le_def.1` before it can be applied to a point. ▸ *Original row:* The integrand `F p w := e(−w p.2) · K(p.1, w)` with parameter `p = (u, ξ) ∈ ℝ × ℝ` is exactly the shape of ~~#124~~'s ★★★ `contDiffOn_integral_of_iteratedFDeriv_bound` and ★★★ `norm_iteratedFDeriv_integral_le` (`ContDiffParametricIntegralFDeriv.lean`): finite-dimensional parameter, a bound family `bound k w` integrable in `w` with `‖∂ᵏ_p F(·, w)(p)‖ ≤ bound k w` for all `p`. **The content is the bound family, and it is a Leibniz expansion across the exponential:** `∂ᵏ_p [e(−wξ) K(u, w)]` is a sum over `i ≤ k` of `(2πw)ⁱ`-sized derivatives of the phase times `∂ᵏ⁻ⁱ_u K(u, w)` (Mathlib's `norm_iteratedFDeriv_mul_le` for the product, `norm_iteratedFDeriv_comp_clm_le`-style bounds for `p ↦ e(−w p.2)`, a character composed with a linear map), so `bound k w := Cₖ · (1 + |w|)ᵏ · sup_u ‖∂^{≤k} K(u, w)‖`, which `K`'s decay at two orders apart makes `≤ C'ₖ (1 + w²)⁻¹` — the weight transfer ~~#126~~(ii) already proved as `one_add_norm_pow_mul_norm_iteratedFDeriv_le` (`SchwartzPartialIntegral.lean`), slots swapped. Measurability of the directional derivatives in `w` is `dirDeriv_eq_prodDirDeriv` (same file): `F` is jointly smooth on `(ℝ × ℝ) × ℝ`. **Deliverable:** `partialFourier K : ℝ × ℝ → ℂ` (the plain function), ★★ `contDiff_partialFourier`, ★★ `norm_iteratedFDeriv_partialFourier_le` (`≤ ∫ bound n`, uniform in `p`), and `partialFourier_apply_eq_sliceFourier` (`= sliceFourierCLM u K ξ` of `PartialFourier.lean`, so the slice-wise object is this one). ⚠️ **Stuck on:** nothing — the price. The Leibniz bound on the product with an oscillating factor is the one computation; everything else is ~~#124~~ and ~~#126~~(ii) called with the right `F`. | ~~**M–L**~~ done (S–M) | Medium — the first of the two halves #121(ii) always had (smoothness); the uniform derivative bound comes with it for free from ~~#124~~ | ~~#124~~; ~~#126~~(ii); `PartialFourier.lean` (`sliceFourierCLM`); #171 | > | ~~171~~ | QIT | Library owed / upstream | *new (2026-10-21, out of #121(ii) — the partial Fourier transform on the plane, brick 2 of 5)* **The derivative identities: `∂_u 𝓕₂K = 𝓕₂(∂_u K)` and `∂_ξ 𝓕₂K = 𝓕₂((−2πi w) K)`, hence every iterated directional derivative of `𝓕₂K` is `𝓕₂` of a Schwartz function.** ▸ **DONE 2026-10-21 (M took S–M).** [`Mathlib/Analysis/Fourier/PartialFourierDeriv.lean`](../CsdLean4/Mathlib/Analysis/Fourier/PartialFourierDeriv.lean), 24 pins. `Dop h := (h.1 : ℂ) • lineDerivOpCLM ℂ 𝓢(ℝ×ℝ, ℂ) (1, 0) + (h.2 : ℂ) • smulLeftCLM ℂ wMul` with `wMul q = −2πi q.2` (the pin's directional-derivative CLM on Schwartz space is `lineDerivOpCLM`/`∂_{m}`, not `pderivCLM`; `wMul` has temperate growth as a `ContinuousLinearMap`), `Dop_apply`, `Dop_add`, `Dop_smul`, `Dop_fst`, `Dop_snd`. ★★ **`fderiv_partialFourier_apply : fderiv ℝ (partialFourier K) p h = partialFourier (Dop h K) p`** — ~~#124~~'s `fderiv_integral_apply_of_bound` on ~~#170~~'s integrand with ~~#170~~'s bound family at orders `0` and `1`, plus the pointwise product rule ★ `fderiv_partialFourierIntegrand_apply` (phase `exp(c·ξ)` via `HasFDerivAt.cexp`, kernel via `inl ∘ fst`); `fderiv_partialFourier_fst/_snd` the named partials, ★ `hasFDerivAt_partialFourier` the CLM form (`smulRight` of the two projections). **All orders:** `dopFin n v` composes `Dop` along a tuple of directions, last first — the order `iteratedFDeriv_apply_succ_last` (~~#124~~) peels them — and ★★ **`iteratedFDeriv_partialFourier_apply : iteratedFDeriv ℝ n (partialFourier K) p v = partialFourier (dopFin n v K) p`** by induction on `n` generalising `K`, three lines per case. Also `integrable_partialFourierIntegrand`, `partialFourier_add`, `partialFourier_smul` (linearity in the kernel, for #174's `mkCLM`). Snag: `HasFDerivAt.mul` returns the Pi-product `(fun y => φ y) * (fun y => ψ y)` as its function — `rw [← hfun] at hd` cannot see it; state the `HasFDerivAt` with the intended function in a typed `have` and close it by `rw [hfun]; exact hφ.mul hK` (defeq through `Pi.mul_apply`). ▸ *Original row:* First order is Mathlib's `hasFDerivAt_integral_of_dominated_of_fderiv_le` on the integrand of #170 with #170's `bound 1`: `fderiv (𝓕₂K) p h = 𝓕₂ (Dop h K) p` where `Dop h K := h.1 • ∂_u K + h.2 • ((−2πi w) · K)` — and `Dop h` is a **continuous linear map of `𝓢(ℝ × ℝ, ℂ)` to itself** (Mathlib's `SchwartzMap.pderivCLM` along `(1, 0)` and `SchwartzMap.smulLeftCLM` of the temperate-growth function `(u, w) ↦ −2πi w`), so the identity iterates: ★★ `dirDeriv_partialFourier`: `dirDeriv (fun p => 𝓕₂K p) hs p = 𝓕₂ (Dop hs K) p` with `Dop hs` the composite over the list, by induction on `hs` through ~~#124~~'s `dirDeriv_append_singleton`. **Deliverable:** `Dop : ℝ × ℝ → (𝓢(ℝ × ℝ, ℂ) →L[ℂ] 𝓢(ℝ × ℝ, ℂ))`, ★ `hasFDerivAt_partialFourier`, ★★ `dirDeriv_partialFourier`, and the two named partials `pderiv_partialFourier_fst`/`_snd`. ⚠️ **Stuck on:** nothing. One dominated-derivative application and one list induction; the dominating function is #170's. | ~~**M**~~ done (S–M) | Medium — it is what lets #173 bound a derivative of `𝓕₂K` by an integral of a Schwartz function rather than by a derivative of one | #170; ~~#124~~ (`dirDeriv`, `dirDeriv_append_singleton`); #173 | > | ~~172~~ | QIT | Library owed / upstream | *new (2026-10-21, out of #121(ii) — the partial Fourier transform on the plane, brick 3 of 5)* **The multiplication identities: `ξ · 𝓕₂K(u, ξ) = (2πi)⁻¹ · 𝓕₂(∂_w K)(u, ξ)` and `u · 𝓕₂K(u, ξ) = 𝓕₂(u · K)(u, ξ)`, hence `(1 + ‖(u, ξ)‖)ᵏ · 𝓕₂K(u, ξ)` is a finite sum of `𝓕₂`'s of Schwartz functions.** ▸ **DONE 2026-10-21 (S–M as priced).** [`Mathlib/Analysis/Fourier/PartialFourierMul.lean`](../CsdLean4/Mathlib/Analysis/Fourier/PartialFourierMul.lean), 21 pins. `mulU = smulLeftCLM ℂ uMul` (`uMul q = q.1`), `derivW = lineDerivOpCLM ℂ 𝓢(ℝ×ℝ, ℂ) (0, 1)`; ★ `partialFourier_mulU : 𝓕₂(mulU K)(u, ξ) = u · 𝓕₂K(u, ξ)` (`u` is a parameter), `slice_derivW : slice (derivW K) u = ∂_{1} (slice K u)` (the chain rule through `w ↦ (u, w)`), ★★ `partialFourier_derivW : 𝓕₂(derivW K)(u, ξ) = 2πi · (ξ · 𝓕₂K(u, ξ))` — Mathlib's `fourier_lineDerivOp_eq` on the slice, through `partialFourier_eq_sliceFourier`, exactly as the row said. Iterates as explicit recursions `mulUPow k`, `derivWPow k` (so `_apply` is `rfl`): `partialFourier_mulUPow` (`= uᵏ 𝓕₂K`), `partialFourier_derivWPow` (`= (2πiξ)ᵏ 𝓕₂K`). ★★ **`norm_pow_mul_norm_partialFourier_le : ‖p‖ᵏ ‖𝓕₂K p‖ ≤ ‖𝓕₂(mulUPow k K) p‖ + (2π)⁻ᵏ ‖𝓕₂(derivWPow k K) p‖`** — the weight is `‖p‖ᵏ` itself (the Schwartz seminorm's weight), not the row's `(1 + ‖p‖)ᵏ`, and the family has **two** members, not `2ᵏ`: `max(|u|, |ξ|)ᵏ ≤ |u|ᵏ + |ξ|ᵏ` (`max_pow_le_add_pow`), so no binomial expansion is needed. This is the form #173 consumes with `K := dopFin n v K`. ▸ *Original row:* The `ξ` identity is **slice-wise and already in Mathlib**: `𝓕₂K(u, ·) = 𝓕 (slice K u)` (`sliceFourierCLM_apply`, `PartialFourier.lean`) and `SchwartzMap.fourier_lineDerivOp_eq` is `𝓕(∂f) = 2πiξ · 𝓕f` on `𝓢(ℝ, ℂ)`; what needs saying is that `∂_w` of the slice is the slice of `∂_w K` (`iteratedFDeriv_comp_inr` of ~~#123~~, `SchwartzSliceContinuous.lean`, at order one). The `u` identity is that `u` is a parameter of the integral. Both right-hand sides are `𝓕₂` of **continuous linear images** of `K` (`pderivCLM` along `(0, 1)`; `smulLeftCLM` of `(u, w) ↦ u`), so the weight `(1 + |u| + |ξ|)ᵏ` expands by the binomial theorem into `≤ 2ᵏ`-many terms `𝓕₂ (Pⱼ K)` with each `Pⱼ` a CLM. **Deliverable:** ★★ `mul_snd_partialFourier`, ★ `mul_fst_partialFourier`, and ★★ `norm_pow_mul_partialFourier_le`: `(1 + ‖p‖)ᵏ · ‖𝓕₂K p‖ ≤ Σⱼ ‖𝓕₂ (Pⱼ K) p‖` with the finite family `Pⱼ` of CLMs named. ⚠️ **Stuck on:** nothing; S–M because the slice-wise Fourier–derivative rule is Mathlib's and the weight is a parameter. | ~~**S–M**~~ done (S–M) | Medium — the decay half of #121(ii), reduced to a sup bound | #170; `PartialFourier.lean`; ~~#123~~ (`iteratedFDeriv_comp_inr`); #173 | > | ~~173~~ | QIT | Library owed / upstream | *new (2026-10-21, out of #121(ii) — the partial Fourier transform on the plane, brick 4 of 5)* **The joint seminorm bounds: for every `k, n` there are a finite set `s` of `K`-seminorms and a `C` with `‖p‖ᵏ · ‖∂ⁿ 𝓕₂K(p)‖ ≤ C · (s.sup (schwartzSeminorm ℂ ℂ)) K` for all `p` — the hypothesis `SchwartzMap.mkCLM` consumes.** ▸ **DONE 2026-10-21 (M–L took S–M).** [`Mathlib/Analysis/Fourier/PartialFourierBound.lean`](../CsdLean4/Mathlib/Analysis/Fourier/PartialFourierBound.lean), 5 pins. ★★★ `exists_bound_partialFourier (k n) : ∃ s C, 0 ≤ C ∧ ∀ K p, ‖p‖ᵏ * ‖iteratedFDeriv ℝ n (partialFourier K) p‖ ≤ C * (s.sup 𝓠) K`. **The row's "one real step" — linearity in `K` — is one Mathlib lemma, and it removed all the bookkeeping the row feared:** ★★ `exists_seminorm_bound_clm (T : 𝓢 →L[ℂ] 𝓢) (s) : ∃ s' C, ∀ K, (s.sup 𝓠)(T K) ≤ C * (s'.sup 𝓠) K` is `Seminorm.bound_of_continuous (schwartz_withSeminorms ℂ (ℝ×ℝ) ℂ)` applied to the continuous seminorm `(s.sup 𝓠).comp T` (a finite sup of the Schwartz seminorms is continuous because its balls are basis sets: `Seminorm.continuous_of_forall` + `hasBasis.mem_of_mem` + `SeminormFamily.basisSets_mem`). No constant is ever named; `mkCLM` asks only for existence. Then ★ `exists_seminorm_bound_Dop` (`≤ C ‖h‖ (s'.sup 𝓠) K`, from the two operators `Dop` is built from, `|h.1|, |h.2| ≤ ‖h‖`, finsets joined by `Finset.sup_mono subset_union_left/right` read through `Seminorm.le_def`), ★★ `exists_seminorm_bound_dopFin` (`≤ C (∏‖vᵢ‖) (s'.sup 𝓠) K`, induction on `n` with `Fin.prod_univ_castSucc`, `Fin.init v i = v (castSucc i)` by `rfl`), and the assembly: ~~#171~~'s `iteratedFDeriv_partialFourier_apply`, ~~#172~~'s `norm_pow_mul_norm_partialFourier_le` with `K := dopFin n v K`, ~~#170~~'s `norm_partialFourier_le` (`kSup G 0 = (Iic (2,0)).sup 𝓠 G` by `rfl`), closed by `ContinuousMultilinearMap.opNorm_le_bound` on `‖p‖ᵏ • iteratedFDeriv …` (the weight carried as a scalar so that `opNorm_le_bound` applies without dividing by `‖p‖ᵏ`). Snags: `Continuous (s.sup 𝓠)` must be written `Continuous ⇑(s.sup 𝓠 : Seminorm ℂ 𝓢(ℝ×ℝ, ℂ))` — with the bare sup, `Continuous`'s expected function type makes the elaborator look for the sup in a Pi lattice; `SeminormFamily` is an `abbrev` for a Pi type, so dot notation `𝓠.basisSets_mem` fails — write `SeminormFamily.basisSets_mem (p := 𝓠)`; `Seminorm.smul_apply`/`ContinuousMultilinearMap.smul_apply` are deprecated for `_root_.smul_apply`. ▸ *Original row:* Assembly of #170–#172: `‖∂ⁿ 𝓕₂K(p)‖` is bounded through the directional derivatives (`ContinuousMultilinearMap.opNorm_le_bound` on `iteratedFDeriv_apply_succ_last`, as ~~#124~~'s `norm_iteratedFDeriv_integral_le` does it) by `sup_{‖hᵢ‖ ≤ 1} ‖𝓕₂ (Dop hs K)(p)‖` (#171); each such term with the weight is `Σⱼ ‖𝓕₂ (Pⱼ Dop hs K)(p)‖` (#172); and `‖𝓕₂ G (p)‖ ≤ ∫ |G(u, w)| dw ≤ π · sup_{(u,w)} (1 + w²)|G(u, w)|`, which is a seminorm bound on `G` (`SchwartzMap.one_add_le_sup_seminorm_apply`). **The one real step is linearity in `K`:** `G = Pⱼ (Dop hs K)` is a continuous linear image of `K`, so its seminorms are bounded by a `Finset.sup` of `K`'s — this is the continuity of a CLM between Schwartz spaces read as a seminorm estimate (Mathlib: `Seminorm.bound_of_continuous` / `SchwartzMap.seminorm_le_bound` through `WithSeminorms`), and the constant must not depend on the direction list `hs` beyond its length and the unit bound on each `hᵢ` — `Dop h` is linear in `h`, so `‖Dop h G‖_seminorm ≤ ‖h‖ · (…)`, which is where `dirWeight` is absorbed. **Deliverable:** ★★★ `exists_bound_partialFourier (k n : ℕ) : ∃ s C, 0 ≤ C ∧ ∀ K p, ‖p‖ᵏ ‖iteratedFDeriv ℝ n (partialFourier K) p‖ ≤ C · (s.sup (schwartzSeminorm ℂ ℂ)) K`. ⚠️ **Stuck on:** the bookkeeping of `Dop` over a direction list against the Schwartz seminorm family — a quantity of plumbing, no wall. Compare ~~#122~~'s `exists_bound_weylOpK` (`WeylSmooth.lean`), the same shape of statement delivered once already. | ~~**M–L**~~ done (S–M) | Medium-high — the whole of #121(ii)'s estimate, in the exact form the packaging needs | #170; #171; #172; ~~#122~~ (`exists_bound_weylOpK`, the template); #174 | > | ~~174~~ | QIT | Library owed / upstream | *new (2026-10-21, out of #121(ii) — the partial Fourier transform on the plane, brick 5 of 5)* **The partial Fourier transform as a continuous linear map `𝓕₂ : 𝓢(ℝ × ℝ, ℂ) →L[ℂ] 𝓢(ℝ × ℝ, ℂ)`, its identification with `weylSymbol`, and the payoff: the Wigner function of a Schwartz state is Schwartz on the plane.** ▸ **DONE 2026-10-21 (M took S). #121 CLOSED.** [`Mathlib/Analysis/Fourier/PartialFourierCLM.lean`](../CsdLean4/Mathlib/Analysis/Fourier/PartialFourierCLM.lean), 13 pins. ★★★ `partialFourierCLM : 𝓢(ℝ × ℝ, ℂ) →L[ℂ] 𝓢(ℝ × ℝ, ℂ)` — `mkCLM` with ~~#170~~'s `contDiff_partialFourier`, ~~#171~~'s `partialFourier_add/smul` and ~~#173~~'s `exists_bound_partialFourier n.1 n.2` passed straight through (the bound field's shape was matched on purpose in #173); `partialFourierCLM_apply` is `rfl`; `slice_partialFourierCLM : slice (partialFourierCLM K) u = sliceFourierCLM u K` ties it to `PartialFourier.lean`; ★★ **`partialFourierCLM_apply_eq_weylSymbol : partialFourierCLM K (u, ξ) = weylSymbol (slice K) u ξ`** — the symbol of a Schwartz kernel is a Schwartz function on the plane. **The payoff:** ★ `wignerKernelS ψ : 𝓢(ℝ × ℝ, ℂ)` with `wignerKernelS_apply : = ψ(x + y/2) · conj ψ(x − y/2)` — ~~#126~~(i)'s `tensorProd ψ (conjS ψ)` composed by `compCLM` with the linear automorphism `wignerL (x, y) = (x + y/2, x − y/2)` (`norm_le_wignerL : ‖q‖ ≤ 2(1 + ‖wignerL q‖)`, the `compCLM` hypothesis; `conjS` is `bilinLeftCLM` of `(a, b) ↦ conj a · b` against the constant `1`, since the pin has no conjugation on Schwartz space) — and ★★★ **`wignerSchwartz ψ : 𝓢(ℝ × ℝ, ℂ)` with `wignerSchwartz_apply : wignerSchwartz ψ (x, ξ) = wigner ψ x ξ`: the Wigner function of a Schwartz state is a Schwartz function of `(x, ξ)`**, which is what ~~#92~~(c) asked for and `WignerCalculus.lean` recorded as not provable at the pin. Scope notes of `PartialFourier.lean`, `SchwartzSlice.lean`, `WignerCalculus.lean` and `WeylSymbolClass.lean` rewritten to say what is now proved. **The whole of #121(ii) took S–M + S–M + S–M + S–M + S against a measured XL**: every brick was priced for a wall that one existing lemma removed (~~#124~~'s all-order bound, Mathlib's slice-wise Fourier–derivative rule, `Seminorm.bound_of_continuous`). ▸ *Original row:* `SchwartzMap.mkCLM` with #170 (smoothness), #173 (the bounds) and linearity of the integral gives ★★★ `partialFourierCLM`; `partialFourierCLM_apply` is #170's function; ★★ `partialFourierCLM_apply_eq_weylSymbol`: `partialFourierCLM K (u, ξ) = weylSymbol (slice K) u ξ` (`weylSymbol_slice_apply`, `PartialFourier.lean`) — **so the symbol of a Schwartz kernel is a Schwartz function on the plane**, which is what #92(c) recorded as the reason `W` could not be shown Schwartz on the plane. Then the payoff: the Wigner kernel `(x, y) ↦ ψ(x + y/2) · conj ψ(x − y/2)` is Schwartz on the plane (★ `wignerKernel₂_prod`: `tensorProd ψ (conj ψ)` of ~~#126~~(i) composed with the linear automorphism `(x, y) ↦ (x + y/2, x − y/2)` of `ℝ × ℝ`, `SchwartzMap.compCLM` with a linear map of temperate growth — S), and `wigner ψ = partialFourierCLM (wignerKernel₂_prod ψ)` pointwise (`wigner_eq_integral`), so ★★★ `wignerSchwartz ψ : 𝓢(ℝ × ℝ, ℂ)` with `wignerSchwartz_apply : wignerSchwartz ψ (x, ξ) = wigner ψ x ξ`. **Deliverable:** `partialFourierCLM`, `partialFourierCLM_apply`, ★★ `partialFourierCLM_apply_eq_weylSymbol`, ★ `wignerKernel₂_prod`, ★★★ `wignerSchwartz`, `wignerSchwartz_apply`; and the scope notes of `PartialFourier.lean` and `WeylSymbolClass.lean` (#92's slicing caveat) rewritten to say the joint form is proved. ⚠️ **Stuck on:** nothing once #173 lands; `mkCLM` wants exactly #173's shape. | ~~**M**~~ done (S) | Medium-high — closes #121(ii) and the #92(c) caveat together; the `𝓢(ℝ × ℝ)` Wigner function is what a Moyal product on symbols would be stated on | #170; #173; ~~#126~~(i) (`tensorProd`); `WignerWeyl.lean` (`weylSymbol`); #92 | > | ~~175~~ | Foundations | Reconstruction frontier | *new (2026-10-21, out of #19 and ~~#153~~ — the narrowing brick)* **Affine hyperbolic fibre maps are translation-conjugate to the cat stroke: every two-transverse-shear de-isolation with base-read offsets is strongly mixing with Lyapunov exponent `2 log φ`.** ▸ **DONE 2026-10-21 (M took S–M).** Cat-1 [`Mathlib/Dynamics/CatMapAffine.lean`](../CsdLean4/Mathlib/Dynamics/CatMapAffine.lean) (13 pins) + SigmaLayer [`SigmaLayer/FibreAffineShears.lean`](../CsdLean4/SigmaLayer/FibreAffineShears.lean) (13 pins). ★★ `Mixing.conj_measurableEquiv : Mixing f μ → MeasurePreserving e μ μ → Mixing (e ∘ f ∘ e.symm) μ` (`Semiconj.iterate_right` for `(e ∘ f ∘ e⁻¹)^[n] = e ∘ f^[n] ∘ e⁻¹`, then one `measure_preimage_equiv` per set); `affineCat c p = cat p + c`, `affineCatShift c = (−c₂, c₂ − c₁)` with `affineCatShift_sub_cat : b − cat b = c` (one `abel` per coordinate — the `(A − I)⁻¹ = [[0, 1], [1, −1]]` of the row), ★★ `affineCat_eq_conj : affineCat c = addLeft b ∘ cat ∘ (addLeft b).symm`, ★★★ `affineCat_mixing`, `affineCat_ergodic`, `measurePreserving_affineCat`, lift `affineCatR c v = catR v + c` with `fderiv_affineCatR = catR` and ★ `affineCatR_fderiv_hasLyapunovExponent` (`2 log φ` off the stable line), `affineCat_coe`. On the fibre: `shearXOff g x = pshift (rep x.2 + g) x`, `shearYOff g x = qshift (rep x.1 + g) x` (at `g = 0` they are ~~#153~~'s `shearX`, `shearY`), ★★ **`shearXOff_comp_shearYOff : shearXOff g₁ ∘ shearYOff g₂ = affineCat (g₁ + g₂, g₂)`**, ★★★ `mixing_shearXOff_comp_shearYOff` (+ ergodic, measure-preserving) **for every pair of offsets**, `shearXOff_comp_shearYOff_coe`, ★ `lift_fderiv_hasLyapunovExponent`, ★★★ `fibre_affine_shears` the package. **Reading against #19:** the base-read offsets are invisible to the criterion *the de-isolation flow has a positive Lyapunov exponent on the fibre* — one whole conjugacy class of candidate maps (two transverse coordinate-reading shears with any offsets) is dynamically the cat stroke. The choice #19 is about is between conjugacy classes; nothing here says nature's de-isolation is in this one. Built first try on the SigmaLayer side; the only snag was a `rw` with `map_sub catHom` not matching `cat` syntactically (state the equation with `cat` in a typed `have`, the defeq is accepted there). ▸ *Original row:* ~~#153~~ says a hyperbolic de-isolation must shear both fibre coordinates, each reading the other; the constructed witness shears one coordinate by a *base-read* amount. The natural class that does both is `x ↦ A x + c` with `A` the cat matrix and `c = c(base)` an offset read off the base — two transverse coordinate-reading shears each carrying a base-read offset, `shearXOff g₁ ∘ shearYOff g₂ = affineCat (g₁ + g₂, g₂)`. **Claim:** `A − I` is invertible over `ℤ` (`det = −1`), so every such map is conjugate to the cat stroke by a translation of the fibre, `affineCat c = τ_b ∘ cat ∘ τ_{−b}` with `b = (−c₂, c₂ − c₁)`, and it inherits strong mixing (~~#147~~) and the exponent `2 log φ` (~~#148~~) unchanged — **the base-read offsets are invisible to #19's criterion; what decides it is the transverse pair, not what the base says.** **Deliverable:** Cat-1 `Mathlib/Dynamics/CatMapAffine.lean` — `Mixing.conj_measurableEquiv` (strong mixing is invariant under conjugation by a measure-preserving measurable equivalence), `affineCat`, `affineCat_eq_conj`, ★★★ `affineCat_mixing`, `affineCat_ergodic`, the lift `affineCatR` and its exponent; SigmaLayer `FibreAffineShears.lean` — `shearXOff`, `shearYOff`, ★★ `shearXOff_comp_shearYOff`, ★★★ `fibre_affine_shears` (the package read against #19). ⚠️ **Stuck on:** nothing — the conjugation is a two-line computation in `AddCircle`, the mixing transport a preimage bookkeeping lemma. ⚠️ Narrows #19, does not decide it: one whole conjugacy class of candidate maps is shown to be dynamically the cat stroke; which class nature uses stays #19. | ~~**M**~~ done (S–M) | Medium — removes a family of candidates from #19's choice by showing them equivalent; the first statement about *which* hyperbolic de-isolations there are | #19; ~~#153~~; ~~#147~~; ~~#148~~; `MovingFibreWitness.lean` (`catStroke`); `FibreTwoShears.lean` | > | ~~176~~ | Foundations | Reconstruction frontier · Statistical-mechanics layer | *new (2026-10-21, out of #19 and #150 — the classical chaos brick)* **Periodic points of the cat stroke: `#Fix(cat^[t]) = |det(Aᵗ − I)| = L₂ₜ − 2`, and its growth rate is the Lyapunov exponent `2 log φ`.** ▸ **DONE 2026-10-21 (M–L took M).** Cat-1 [`Mathlib/Dynamics/CatMapPeriodicPoints.lean`](../CsdLean4/Mathlib/Dynamics/CatMapPeriodicPoints.lean) (39 pins) + SigmaLayer [`SigmaLayer/FibrePeriodicGrowth.lean`](../CsdLean4/SigmaLayer/FibrePeriodicGrowth.lean) (3 pins). **The row's stuck-on was wrong in the good direction:** the lattice-index fact is in the pin as `Submodule.natAbs_det_basis_change` (`Mathlib/LinearAlgebra/FreeModule/Finite/CardQuotient.lean`) — the ideal version this row found is a one-line wrapper around it — so (iii) needed no Smith-form transport and no hand proof. `lucasEven n = 2F₂ₙ₊₁ − F₂ₙ` with ★ `lucasEven_eq : = φ²ⁿ + ψ²ⁿ` (Binet); `cassini : F₍ₙ₊₂₎Fₙ − F₍ₙ₊₁₎² = (−1)ⁿ⁺¹` (the pin has no Cassini; four-line induction); `catLatMat t = Aᵗ − I` as an integer matrix with ★ `det_catLatMat : = 2 − L₂ₜ`, `catLat = toLin'` of it, injective by the adjugate; `cat_iterate_coe` descends `cat^[t]` from `catR ^ t`. The parametrisation `catFixLift : ℤ² →+ T²`, `m ↦ (Aᵗ − I)⁻¹m mod ℤ²` (real adjugate over `det`): `catFixLift_mem` (every image is fixed), `catFixLift_surjective` (every fixed point is an image — lift, read `(Aᵗ − I)v ∈ ℤ²` off `AddCircle.coe_eq_zero_iff`), `catFixLift_eq_zero_iff` (kernel = the lattice `(Aᵗ − I)ℤ²`). Then ★★★ **`card_fixedPoints_cat_iterate : Nat.card {p // cat^[s+1] p = p} = (L₂₍ₛ₊₁₎ − 2).natAbs`** — `Equiv.subtypeEquivRight` + Noether (`QuotientAddGroup.quotientKerEquivRange`) + `quotientAddEquivOfEq` + the index formula with the image basis `(Pi.basisFun ℤ (Fin 2)).map (LinearEquiv.ofInjective …)` and `Basis.det_comp`/`det_self`; `card_fixedPoints_cat_iterate_real : = φ²ᵗ + φ⁻²ᵗ − 2`; ★★ **`tendsto_log_card_fixedPoints_cat : (1/t) log #Fix → 2 log φ`** by the squeeze `(1 − φ⁻²)² φ²ᵗ ≤ #Fix ≤ φ²ᵗ`. SigmaLayer ★★★ `fibre_periodic_growth`: the count, its growth rate and the Lyapunov exponent of the lift, all `2 log φ` — **#19's criterion read by counting returns agrees with its reading by stretching; both are classical statements about the ontic fibre.** Snags: `Real.log_pow` takes `(x) (n)`; `rw` sees through a `set` variable's value — `clear_value` before rewriting a `log`; `AddCircle.coe_eq_zero_iff` takes `p` explicitly; `Basis` is `Module.Basis` (open `Module`); `Matrix.dotProduct` is now root `dotProduct`; `Matrix.smul_mulVec_assoc` is `Matrix.smul_mulVec`. ▸ *Original row:* ~~#148~~ gave #19 one measurable criterion (a positive Lyapunov exponent of the de-isolation flow on the fibre); ~~#175~~ showed it is blind to base-read offsets. This row gives the criterion a second, independent reading — **a counting one**: the number of fibre configurations that return after `t` strokes grows at exactly the stretching rate. For the cat matrix, `Aᵗ = [[F₂ₜ₊₁, F₂ₜ], [F₂ₜ, F₂ₜ₋₁]]` (Fibonacci), `det(Aᵗ − I) = 2 − tr Aᵗ = 2 − L₂ₜ` (Lucas), so the fixed-point count is `L₂ₜ − 2 = φ²ᵗ + φ⁻²ᵗ − 2` and `(1/t) log #Fix(cat^[t]) → 2 log φ` — the cat-map instance of *topological entropy = Lyapunov exponent*. **Deliverable** (Cat-1 `Mathlib/Dynamics/CatMapPeriodicPoints.lean`): (i) `cat_iterate_coe`/`catR_pow` — the iterate is the Fibonacci matrix, by induction; (ii) `det_catR_pow_sub_one : det(Aᵗ − I) = 2 − L₂ₜ` with `L₂ₜ` written as `F₂ₜ₊₁ + F₂ₜ₋₁`; (iii) ★★★ `card_fixedPoints_cat_iterate : Nat.card {p // cat^[t] p = p} = L₂ₜ − 2` for `t ≥ 1`; (iv) ★★ `tendsto_log_card_fixedPoints_div : (1/t) log #Fix → 2 log φ` (Binet, `Real.coe_fib_eq`); and the SigmaLayer reading `fibre_periodic_growth` against #19. ⚠️ **Stuck on:** (iii) is the lattice-index fact `|ℤ²/(Aᵗ − I)ℤ²| = |det(Aᵗ − I)|`: the fixed points of a toral automorphism `B` with `det(B − I) ≠ 0` are `(B − I)⁻¹ℤ²/ℤ² ≅ ℤ²/(B − I)ℤ²`. The pin has it for **ideals** of a ring free over `ℤ` (`Ideal.natAbs_det_basis_change`, `cardQuot`) and the Smith normal form for submodules (`Submodule.smithNormalForm`), not the submodule-of-`ℤ²` statement itself — so either transport through the Smith form (two invariant factors, `Nat.card (ℤ ⧸ a) = |a|`) or prove the `2 × 2` case by hand. (i), (ii), (iv) have no wall. | ~~**M–L**~~ done (M) | Medium — gives #19's criterion a counting reading (returns of fibre configurations), independent of the derivative reading of ~~#148~~; the first entropy-type statement on the fibre | #19; #150; ~~#148~~; ~~#175~~; `Mathlib/Dynamics/LyapunovExponent.lean` (`catR`, `hasLyapunovExponent_catR`); `CatMapWitness.lean` (`cat`, `cat_coe`) | > | ~~177~~ | Foundations | Reconstruction frontier | *new (2026-10-21, out of #19 and ~~#176~~ — the derivative-free criterion)* ▸ **DONE 2026-10-21 (S–M as priced).** Cat-1 [`Mathlib/Dynamics/ReturnCriterion.lean`](../CsdLean4/Mathlib/Dynamics/ReturnCriterion.lean) (6 pins) + SigmaLayer [`SigmaLayer/FibreReturnCriterion.lean`](../CsdLean4/SigmaLayer/FibreReturnCriterion.lean) (5 pins). `returnCount f t := Nat.card {x // f^[t] x = x}`; `HasReturnGrowth f h` = finite ∧ nonempty return set at every `t ≥ 1` ∧ `(1/t) log returnCount → h`. `iterate_add_right : (· + c)^[t] = (· + t • c)`, `add_right_eq_self_iff`, ★★ `not_hasReturnGrowth_add_right [AddCommGroup G] [Infinite G] (c h) : ¬ HasReturnGrowth (· + c) h` — at the first stroke the return set is all of `G` (infinite, fails `finite`) or empty (fails `nonempty`); ★★★ `cat_hasReturnGrowth : HasReturnGrowth cat (2 log φ)` from ~~#176~~ (`Nat.finite_of_card_ne_zero`, `Nat.card_of_isEmpty`). Fibre: `infinite_addCircle_one` (via `AddCircle.equivIoc` + `Set.Ioc.infinite`, now an `Infinite KTorus` instance), ★★★ `catStroke_hasReturnGrowth`, `pshift_eq_add : pshift a = (· + (↑a, 0))`, ★★ `not_hasReturnGrowth_pshift (a h)`, ★★★ `fibre_return_criterion idx s t` (hyperbolic stroke has it at `2 log φ > 0`; `shearEvolve`'s fibre action is `pshift` at every base point and interval, `shearEvolve_snd`; no `pshift` has it at any rate). **Reading:** #19's criterion now has a form that mentions only which fibre configurations come back — no lift, no tangent vector, no direction — and the two corpus witnesses fall on its two sides without a derivative being taken. Which map nature uses stays #19. ▸ *Original row:* **The return criterion: a fibre map `f` *has return growth `h`* when at every stroke count `t ≥ 1` finitely many but at least one configuration returns (`f^[t] x = x`) and `(1/t) log #returns → h`; #19's criterion is `h > 0`. The hyperbolic witness has it at `2 log φ`; every translation of the fibre — hence the constructed de-isolation's fibre action at every base point — fails it at the first stroke.** ~~#148~~ stated #19's criterion through the derivative of the lift (`HasLyapunovExponent`), which needs the lift and a tangent direction. ~~#176~~ showed the hyperbolic witness's return count grows at the same rate. This row makes the counting form **the** criterion, stated for an arbitrary self-map of the fibre with no lift, no derivative and no direction — a statement about records (which configurations come back) rather than about tangent vectors. **The dichotomy that decides the witness:** a translation `x ↦ x + c` of an infinite group fixes everything (`c = 0`) or nothing (`c ≠ 0`), and its `t`-th iterate is the translation by `t • c`; so at every `t` its return set is infinite or empty and the criterion's first clause fails — no exponent, no base-read amount, no case analysis on rational versus irrational amounts. The constructed de-isolation's fibre action over any interval at any base point is `pshift a = (· + (a, 0))` (`shearEvolve_snd`), so it fails for every `a`. **Deliverable:** Cat-1 `Mathlib/Dynamics/ReturnCriterion.lean` — `returnCount f t := Nat.card {x // f^[t] x = x}`, `HasReturnGrowth f h` (finite, nonempty, growth), `iterate_add_right` (`(· + c)^[t] = (· + t • c)`), ★★ `not_hasReturnGrowth_add_right` (no translation of an infinite additive group has return growth), ★★★ `cat_hasReturnGrowth : HasReturnGrowth cat (2 log φ)` from ~~#176~~; SigmaLayer `FibreReturnCriterion.lean` — `catStroke_hasReturnGrowth`, `pshift_eq_add`, ★★ `not_hasReturnGrowth_pshift`, ★★★ `fibre_return_criterion` (the package read against #19, with `shearEvolve_snd`). ⚠️ **Stuck on:** nothing; S–M. `Infinite (AddCircle 1)` is not an instance at the pin (derive it from `AddCircle.equivIoc` and `Set.Ioc.infinite`). ⚠️ Narrows the *statement* of #19's criterion, does not decide #19. | ~~**S–M**~~ done (S–M) | Medium — the criterion in record terms, which is the form a derivation of torus generation from dynamics could consume; and the witness fails it without any derivative | #19; ~~#176~~; ~~#148~~; ~~#153~~; `RecordLayer/ShearWitness.lean` (`pshift`, `shearEvolve`); `SigmaLayer/FibreLyapunov.lean` (`shearEvolve_snd`) | > | ~~178~~ | Foundations | Reconstruction frontier | *new (2026-10-21, out of #19 and ~~#177~~ — the criterion by the trace, forward direction)* ▸ **DONE 2026-10-21 (L took M–L).** Cat-1 [`Mathlib/Dynamics/ToralReturns.lean`](../CsdLean4/Mathlib/Dynamics/ToralReturns.lean), 49 pins. `toral B p = (B₀₀ • p.1 + B₀₁ • p.2, B₁₀ • p.1 + B₁₁ • p.2)` with `toral_mul`, `toral_pow : (toral B)^[t] = toral (Bᵗ)`, `toral_fixed_iff` (fixed points = kernel of `toral (B − 1)`), `toral_coe` (descent from the real matrix). ★★★ **`card_ker_toral (hC : det C ≠ 0) : Nat.card {p // toral C p = 0} = (det C).natAbs`** — ~~#176~~'s parametrisation with the `2 × 2` adjugate written out (`adjMat`, `adjMat_realMat : adj(C)·C·v = det C • v`), the same `Submodule.natAbs_det_basis_change`, and `toLin'_injective_of_det_ne_zero` by the adjugate; `det_sub_one_fin_two : det(M − 1) = det M − tr M + 1`; ★★ `card_fixedPoints_toral_pow : #Fix(toral Bᵗ) = |2 − tr Bᵗ|` for `t ≥ 1`. `cayley_hamilton_fin_two : M² = (tr M) • M − (det M) • 1` (explicit `2 × 2`, `simp only` with `Matrix.one_fin_two` — `simp` alone turns `n • M` into `↑n * M` in the matrix ring via `zsmul_eq_mul` and loses), `trace_pow_add_two : tr Bᵗ⁺² = tr B · tr Bᵗ⁺¹ − tr Bᵗ`; `spectralRadius B = (|tr B| + √(tr B² − 4))/2`, `one_lt_spectralRadius`, `spectralRadius_add_inv : ρ + ρ⁻¹ = |tr B|`, `traceSign B = ±1`, ★★ **`trace_pow_eq : tr Bᵗ = εᵗ (ρᵗ + ρ⁻ᵗ)`** (two-step induction, one `linear_combination` with `ρ · ρ⁻¹ = 1` and `ε² = 1`), `trace_pow_ne_two` (`tr Bᵗ = 2 ⇒ (ρᵗ − 1)² = 0 ⇒ ρᵗ = 1`, impossible — no sign split), ★★★ **`hasReturnGrowth_toral (hτ : 2 < |tr B|) (hdet : det B = 1) : HasReturnGrowth (toral B) (log ρ)`** by the squeeze `(1 − ρ⁻¹)² ρᵗ ≤ #Fix ≤ 4 ρᵗ`. Instance: `cat_eq_toral : cat = toral !![2,1;1,1]`, `spectralRadius_catMat = φ²`, `log_spectralRadius_catMat = 2 log φ`, `cat_hasReturnGrowth'` — ~~#176~~/~~#177~~ recovered from the general theorem. **For #19: an integer-matrix candidate for the fibre map is tested by one integer, `|tr B| > 2`.** ▸ *Original row:* **Every hyperbolic toral automorphism has return growth: for `B ∈ SL₂(ℤ)` with `|tr B| > 2`, `HasReturnGrowth (toral B) (log μ)` where `μ > 1` is the spectral radius `(|tr B| + √(tr B² − 4))/2` — so a candidate fibre map given as an integer matrix is tested by one integer, its trace.** ~~#176~~–~~#177~~ did the cat matrix by hand (Fibonacci entries, Cassini). The general statement replaces them by three abstract facts: (i) the kernel of a toral endomorphism `toral C` with `det C ≠ 0` has `|det C|` points — ~~#176~~'s parametrisation `m ↦ (1/det C)·adj(C)·m mod ℤ²` with Mathlib's `Matrix.adjugate` in place of the explicit formulas, and the same `Submodule.natAbs_det_basis_change`; (ii) `det(M − 1) = det M − tr M + 1` for `2 × 2`, so `#Fix(toral Bᵗ) = |2 − tr Bᵗ|`; (iii) Cayley–Hamilton `B² = (tr B)·B − 1` gives the recurrence `tr Bᵗ⁺² = tr B · tr Bᵗ⁺¹ − tr Bᵗ`, solved by `λᵗ + λ⁻ᵗ` with `λ + λ⁻¹ = tr B`, `|λ| = μ > 1`; then `tr Bᵗ = 2` would force `λᵗ = 1`, impossible, so every return set is finite and nonempty, and `μᵗ − 2 ≤ #Fix ≤ μᵗ + 3` squeezes the growth to `log μ`. **Deliverable** (Cat-1 `Mathlib/Dynamics/ToralReturns.lean`): `toral B`, `toral_mul`, `toral_pow`, `toral_sub_one_fixed` (fixed points = kernel of `toral (B − 1)`), ★★★ `card_ker_toral (hC : det C ≠ 0) : Nat.card {p // toral C p = 0} = (det C).natAbs`, `det_sub_one_fin_two`, `trace_pow_eq` (the `λ` closed form), ★★★ `hasReturnGrowth_toral_of_two_lt_abs_trace`, and `cat` as the instance `B = [[2,1],[1,1]]`, `tr = 3`, `μ = φ²` (recovering ~~#176~~'s exponent `2 log φ` as `log φ²`). ⚠️ **Stuck on:** nothing new — every ingredient is ~~#176~~'s with `Matrix.adjugate` abstract; the sign of `λ` (negative trace) is handled by working with `|λ|` and `λᵗ = 1 ⇒ |λ|ᵗ = 1`, no parity split. | ~~**L**~~ done (M–L) | Medium-high — turns #19's criterion into a one-integer test on any integer-matrix candidate; the general theorem a reader expects after ~~#176~~ | #19; ~~#176~~; ~~#177~~; #179; `Mathlib/LinearAlgebra/FreeModule/Finite/CardQuotient.lean` | > | ~~179~~ | Foundations | Reconstruction frontier | *new (2026-10-21, out of #19 and ~~#177~~ — the criterion by the trace, converse)* ▸ **DONE 2026-10-21 (M took S–M; built first try).** Cat-1 [`Mathlib/Dynamics/ToralReturnsConverse.lean`](../CsdLean4/Mathlib/Dynamics/ToralReturnsConverse.lean), 6 pins. `exists_kernel_vector (hC : det C = 0) : ∃ k ≠ 0, toLin' C k = 0` (the row's explicit vector: `(1, 0)` if `C = 0`, else `(−C₀₁, C₀₀)` or `(−C₁₁, C₁₀)`); ★★ `infinite_ker_toral_of_det_eq_zero` (the line `s ↦ (s/kᵢ)·k mod ℤ²` on `[0, 1)`, injective by `AddCircle.coe_eq_coe_iff_of_mem_Ico` at the coordinate `i` with `kᵢ ≠ 0`); `exists_det_pow_sub_one_eq_zero (hdet) (|tr B| ≤ 2) : ∃ t, det(Bᵗ⁺¹ − 1) = 0` — `interval_cases` on the trace, the five Cayley–Hamilton cases exactly as the row listed them (`t + 1 = 2, 3, 4, 6, 1` for `tr = −2, −1, 0, 1, 2`; `noncomm_ring` for `(B ± 1)² = B² ± 2B + 1` in the matrix ring, `abel` after rewriting `B²`); ★★★ **`not_hasReturnGrowth_toral_of_abs_trace_le_two`** (the `finite` clause fails at that stroke) and ★★★ **`hasReturnGrowth_toral_iff (hdet : det B = 1) : (∃ h, HasReturnGrowth (toral B) h) ↔ 2 < |tr B|`** with ~~#178~~. **For #19:** on integer-matrix candidates for the fibre map, the return criterion is a decidable test on one integer. ▸ *Original row:* **No non-hyperbolic toral automorphism has return growth: for `B ∈ SL₂(ℤ)` with `|tr B| ≤ 2`, `¬ HasReturnGrowth (toral B) h` for every `h`. With #178: `(∃ h, HasReturnGrowth (toral B) h) ↔ 2 < |tr B|` — the return criterion on integer matrices is decided by the trace.** Five cases by Cayley–Hamilton `B² = (tr B)·B − 1`: `tr = 0` gives `B⁴ = 1`, `tr = 1` gives `B⁶ = 1`, `tr = −1` gives `B³ = 1` — at that stroke every point returns, the return set is all of `T²`, infinite; `tr = 2` gives `(B − 1)² = 0`, so `det(B − 1) = 0`; `tr = −2` gives `(B + 1)² = 0`, so `det(B² − 1) = 4·det(B + 1) = 0`. A singular `C` has infinite kernel on `T²` (a nonzero integer kernel vector `k` gives the infinite line `s·k mod ℤ²`, injective for `s ∈ [0, 1/‖k‖)`), so the `finite` clause fails. **Deliverable** (same module as #178 or `ToralReturnsConverse.lean`): `cayley_hamilton_fin_two`, the five power identities, ★★ `infinite_ker_toral_of_det_eq_zero`, ★★★ `not_hasReturnGrowth_toral_of_abs_trace_le_two`, ★★★ `hasReturnGrowth_toral_iff`. ⚠️ **Stuck on:** nothing; the integer kernel vector of a singular `2 × 2` integer matrix is explicit (a column of the adjugate, or `(−C 0 1, C 0 0)` / `(C 1 1, −C 1 0)` whichever is nonzero, or `C = 0`). | ~~**M**~~ done (S–M) | Medium-high — closes the characterisation; with #178 the criterion is a decidable test on the matrix | #19; #178; ~~#177~~ | > | ~~180~~ | Foundations | Reconstruction frontier | *new (2026-10-21, out of #19 and ~~#178~~/~~#179~~ — the orientation-reversing case)* ▸ **DONE 2026-10-21 (M as priced).** Cat-1 [`Mathlib/Dynamics/ToralReturnsReversing.lean`](../CsdLean4/Mathlib/Dynamics/ToralReturnsReversing.lean), 16 pins. `tendsto_log_div_of_bounds` (the squeeze `c μᵗ ≤ uₜ ≤ C μᵗ ⇒ (1/t) log uₜ → log μ`, now a lemma), `det_pow_sub_one_neg : det(Bᵗ − 1) = (−1)ᵗ − tr Bᵗ + 1`, `trace_pow_add_two_neg` (the signed recurrence), `spectralRadiusNeg B = (|tr| + √(tr² + 4))/2` with `one_lt_spectralRadiusNeg (tr ≠ 0)` and `spectralRadiusNeg_sub_inv : μ − μ⁻¹ = |tr B|`, ★★ `trace_pow_eq_neg : tr Bᵗ = εᵗ (μᵗ + (−1)ᵗ μ⁻ᵗ)` (same two-step induction as ~~#178~~, signs flipped in the `linear_combination`), `count_bounds_neg` (the parity split: odd strokes `μᵗ − μ⁻ᵗ = μᵗ(1 − μ⁻²ᵗ)`, even strokes `μᵗ + μ⁻ᵗ − 2 = μᵗ(1 − μ⁻ᵗ)²`, both in `[(1 − μ⁻²)² μᵗ, 4 μᵗ]` and nonzero), ★★ `card_fixedPoints_toral_pow_neg`, ★★★ **`hasReturnGrowth_toral_neg (det B = −1) (tr B ≠ 0) : HasReturnGrowth (toral B) (log μ)`**, ★★ `not_hasReturnGrowth_toral_of_trace_eq_zero_neg` (`B² = 1`, infinite return set at the second stroke), ★★★ `hasReturnGrowth_toral_iff_neg`, ★★★ **`hasReturnGrowth_toral_iff_of_det (det B = ±1) : (∃ h, HasReturnGrowth (toral B) h) ↔ (det B = 1 ∧ 2 < |tr B|) ∨ (det B = −1 ∧ tr B ≠ 0)`**. **For #19: every integer-matrix candidate for the fibre map, orientation-preserving or not, is tested by its determinant and trace.** ▸ *Original row:* **The return criterion for `det B = −1`: `toral B` has return growth iff `tr B ≠ 0`, with rate `log μ`, `μ = (|tr B| + √(tr B² + 4))/2`.** ~~#178~~–~~#179~~ assumed `det B = 1`. For `det B = −1` the eigenvalues are `λ, −λ⁻¹` (product `−1`), `tr B = λ − λ⁻¹`, and `det(Bᵗ − 1) = (−1)ᵗ − tr Bᵗ + 1` with `tr Bᵗ = λᵗ + (−λ⁻¹)ᵗ`; the recurrence from Cayley–Hamilton is `tr Bᵗ⁺² = tr B · tr Bᵗ⁺¹ + tr Bᵗ`. Hyperbolic iff `tr B ≠ 0` (then `|λ| > 1` and `det(Bᵗ − 1) ≠ 0` at every `t ≥ 1`, odd and even `t` separately: `λᵗ − λ⁻ᵗ ≠ 0` and `2 − λᵗ − λ⁻ᵗ ≠ 0`); `tr B = 0` gives `B² = 1`, every point returns at `t = 2`. **Deliverable** (Cat-1 `Mathlib/Dynamics/ToralReturnsReversing.lean`): `det_pow_sub_one_neg (hdet : det B = −1)`, the signed recurrence, the closed form with `μ`, ★★★ `hasReturnGrowth_toral_of_trace_ne_zero_of_det_neg_one`, ★★ `not_hasReturnGrowth_toral_of_trace_eq_zero_of_det_neg_one`, and the full statement on `GL₂(ℤ)`: **`(∃ h, HasReturnGrowth (toral B) h) ↔ (det B = 1 ∧ 2 < |tr B|) ∨ (det B = −1 ∧ tr B ≠ 0)`** (`det B = ±1` for an automorphism; other determinants are not automorphisms and are out of scope). ⚠️ **Stuck on:** nothing — a replay of ~~#178~~/~~#179~~ with a `(−1)ᵗ` carried through; the parity split in the nonvanishing step is the only new bookkeeping. | ~~**M**~~ done (M) | Medium — completes the trace test to all of `GL₂(ℤ)`, so every integer-matrix candidate for the fibre map is covered | #19; ~~#178~~; ~~#179~~; ~~#177~~ | > | ~~181~~ | Foundations | Reconstruction frontier | *new (2026-10-21, out of #19 and ~~#153~~/~~#177~~ — the parabolic witness's returns made explicit)* ▸ **DONE 2026-10-22 (S–M as priced; both modules built first try).** Cat-1 [`Mathlib/Dynamics/ShearReturns.lean`](../CsdLean4/Mathlib/Dynamics/ShearReturns.lean) (8 pins) + SigmaLayer [`SigmaLayer/FibreShearReturns.lean`](../CsdLean4/SigmaLayer/FibreShearReturns.lean) (1 pin). `shearX_iterate : shearX^[t] = fun p => (p.1 + t • p.2, p.2)`, `shearY_iterate`, ★★ `shearX_iterate_fixed_iff : shearX^[t] p = p ↔ t • p.2 = 0` (and `shearY`, on `p.1`), ★ `infinite_fixedPoints_shearX_iterate`/`_shearY_` (the circle `{(x, 0)}` resp. `{(0, y)}` lies in every return set; injection from `Set.Ico 0 1` as in ~~#179~~), ★★ `not_hasReturnGrowth_shearX`/`_shearY` (the `finite` clause fails already at the first stroke, and at every stroke). SigmaLayer ★★★ `fibre_shear_returns`: the two `iff`s, the two infinities, the two negations, `catStroke = shearX ∘ shearY`, and `HasReturnGrowth catStroke (2 log φ)` in one statement — **each factor returns a union of torsion circles at every stroke, the composite returns finitely many points: the transversality of the pair is what makes returns finite** (~~#153~~'s message in ~~#177~~'s vocabulary). ▸ *Original row:* **Which configurations return under a single coordinate-reading shear: `Fix(shearX^[t]) = {x : t • x.2 = 0}`, the `t` circles over the `t`-torsion points of the read coordinate — infinite at every stroke, so the shear fails the return criterion at every `t`, not merely at one.** ~~#153~~ factored the hyperbolic stroke into two transverse shears each of exponent `0`; ~~#179~~ shows abstractly that a shear (`tr = 2`) has no return growth. This row says *what* returns: `shearX^[t] (x, y) = (x + t • y, y)`, fixed iff `t • y = 0`, i.e. `y` is a `t`-torsion point of the circle (`AddCircle` has exactly `t` of them, Mathlib's `AddCircle.card_addOrderOf_eq_totient`-side API), and every `x` works — the return set is `T¹ × (ℤ/t)`. In record terms: a parabolic de-isolation returns exactly the configurations whose *read* coordinate is rational with denominator dividing `t`, all of them at once, independent of the *shifted* coordinate. **Deliverable** (SigmaLayer `FibreShearReturns.lean`, with a Cat-1 lemma `shearX_iterate` in `CatMapShears.lean`'s line): `shearX_iterate : shearX^[t] = fun p => (p.1 + t • p.2, p.2)`, ★★ `fixedPoints_shearX_iterate : shearX^[t] p = p ↔ t • p.2 = 0`, ★ `infinite_fixedPoints_shearX_iterate`, `not_hasReturnGrowth_shearX` (now by the explicit description, independent of ~~#179~~), and the same for `shearY`; the reading against #19: **the hyperbolic stroke's return sets are finite and grow, each shear's are infinite at every stroke — the stretching that makes returns finite is produced by the transversality of the pair** (~~#153~~'s message, now in return terms). ⚠️ **Stuck on:** nothing; S–M. The torsion-point count on `AddCircle` is in Mathlib (`AddCircle.addOrderOf_div_of_gcd_eq_one`, `card_addOrderOf_eq_totient`); only `t • y = 0 ↔ y` is `t`-torsion is needed here, which is the definition. | ~~**S–M**~~ done (S–M) | Medium — the records-side picture of the parabolic case; what a non-hyperbolic de-isolation would make return | #19; ~~#153~~; ~~#177~~; ~~#179~~; `Mathlib/Dynamics/CatMapShears.lean` (`shearX`, `shearY`) | > | ~~182~~ | Foundations | Reconstruction frontier | *new (2026-10-22, out of #19 and ~~#175~~/~~#177~~ — the return criterion carried to the whole affine class)* ▸ **DONE 2026-10-22 (S–M as priced; one tactic retry in the induction step, otherwise first try).** Cat-1 [`Mathlib/Dynamics/ReturnCriterionConj.lean`](../CsdLean4/Mathlib/Dynamics/ReturnCriterionConj.lean) (12 pins) + SigmaLayer [`SigmaLayer/FibreAffineReturns.lean`](../CsdLean4/SigmaLayer/FibreAffineReturns.lean) (3 pins). `iterate_conj (e : α ≃ β) f t : (⇑e ∘ f ∘ ⇑e.symm)^[t] = ⇑e ∘ f^[t] ∘ ⇑e.symm`, `fixedPointsEquivConj : {x // f^[t] x = x} ≃ {y // (⇑e ∘ f ∘ ⇑e.symm)^[t] y = y}`, `returnCount_conj`, ★★ `HasReturnGrowth.conj`, `hasReturnGrowth_conj_iff` (both directions, via `symm_conj_conj`), `HasReturnGrowth.unique` (`tendsto_nhds_unique`); `MeasureTheory.affineCat_eq_conj_equiv` (~~#175~~'s conjugacy through `MeasurableEquiv.toEquiv`), `returnCount_affineCat c t = returnCount cat t`, `card_fixedPoints_affineCat_iterate = (lucasEven (s+1) − 2).natAbs`, ★★★ `affineCat_hasReturnGrowth c : HasReturnGrowth (affineCat c) (2 log φ)`. SigmaLayer ★★★ `fibre_affine_returns g₁ g₂`: `HasReturnGrowth (shearXOff g₁ ∘ shearYOff g₂) (2 log φ)` ∧ `∀ t, returnCount (shearXOff g₁ ∘ shearYOff g₂) t = returnCount catStroke t` ∧ rate uniqueness — **what the base says at the base point changes which configurations return, never how many.** ▸ *Original row:* **Return growth is invariant under conjugation by a bijection, so every affine cat map — every two-transverse-shear de-isolation with base-read offsets (~~#175~~) — has return growth `2 log φ`, with the same return count as the cat stroke at every stroke.** ~~#175~~ showed the offsets are invisible to the *Lyapunov* reading of #19's criterion (the lift's derivative is `catR` whatever the offsets); ~~#177~~ replaced that reading by the derivative-free one (which configurations come back). This row closes the gap: for `e : α ≃ β` and `f : α → α`, `(e ∘ f ∘ e.symm)^[t] = e ∘ f^[t] ∘ e.symm`, so `e` restricts to a bijection of the return sets, `returnCount (e ∘ f ∘ e.symm) t = returnCount f t`, and `HasReturnGrowth` transports both ways. Applied to `affineCat_eq_conj` (translation-conjugate to `cat`): `returnCount (affineCat c) t = (L₂ₜ − 2)` (~~#176~~) and `HasReturnGrowth (affineCat c) (2 log φ)` for every `c`. The rate is also unique (`tendsto_nhds_unique`), so the class has return growth at exactly one rate. **Deliverable** (Cat-1 `Mathlib/Dynamics/ReturnCriterionConj.lean`, root namespace beside `HasReturnGrowth`): `iterate_conj (e : α ≃ β) (f) (t) : (⇑e ∘ f ∘ ⇑e.symm)^[t] = ⇑e ∘ f^[t] ∘ ⇑e.symm`, `fixedPointsEquiv_conj : {x // f^[t] x = x} ≃ {y // (e ∘ f ∘ e.symm)^[t] y = y}`, `returnCount_conj`, ★★ `HasReturnGrowth.conj`, `hasReturnGrowth_conj_iff`, `HasReturnGrowth.unique`; then in `MeasureTheory`: `card_fixedPoints_affineCat_iterate`, ★★★ `affineCat_hasReturnGrowth (c) : HasReturnGrowth (affineCat c) (2 log φ)`. SigmaLayer `FibreAffineReturns.lean`: ★★★ `fibre_affine_returns (g₁ g₂)`: `HasReturnGrowth (shearXOff g₁ ∘ shearYOff g₂) (2 log φ)` ∧ `∀ t, returnCount (shearXOff g₁ ∘ shearYOff g₂) t = returnCount catStroke t` ∧ uniqueness of the rate — **the base-read offsets are invisible to the return criterion too, stroke by stroke: what the base says at the base point changes which configurations return, never how many.** ⚠️ **Stuck on:** nothing; S–M. `affineCat_eq_conj` is stated through `MeasurableEquiv.addLeft`; pass to its `toEquiv` (`MeasurableEquiv.coe_toEquiv`, `coe_toEquiv_symm`). | ~~**S–M**~~ done (S–M) | Medium — closes ~~#175~~'s reading under ~~#177~~'s criterion; the first invariance lemma of the criterion | #19; ~~#175~~; ~~#176~~; ~~#177~~; `Mathlib/Dynamics/ReturnCriterion.lean` (`HasReturnGrowth`), `CatMapAffine.lean` (`affineCat_eq_conj`) | > | ~~183~~ | Records | Records→spacetime | *new (2026-10-22, the clock-and-echo programme — operational coordinates from physical records; split of the author's brief into seven rows, a–g, recorded together)* ▸ **DONE 2026-10-22 (M as priced).** Cat-1 [`Mathlib/Analysis/Normed/Algebra/ExpUnique.lean`](../CsdLean4/Mathlib/Analysis/Normed/Algebra/ExpUnique.lean) (3 pins) + [`CV/SignalExchange.lean`](../CsdLean4/CV/SignalExchange.lean) (53 pins). ★★ `exp_smul_eq_of_hasDerivAt (X) (h0 : U 0 = 1) (hU : ∀ a, HasDerivAt U (X * U a) a) : U a = exp (a • X)` (any complete normed algebra over `ℝ` or `ℂ`; `a ↦ exp (−a • X) * U a` has zero derivative), ★★ `exp_smul_eq_cosh_sinh (hXX : X * X = P) (hXP : X * P = X) : exp (a • X) = 1 + (cosh a − 1) • P + sinh a • X`. `hopX`, `oneExc`, `hopGen J ħ = −(i J/ħ) • hopX`, ★ `hopFlow J ħ : FieldStructuredFlow 2 2` (one edge `(0, 1)`), ★★ `hopFlow_val_eq : U(t) = 1 + (cos θ − 1) • P − i sin θ • X`, `θ = J t/ħ`; `hopTime = π ħ/(2 J)`, `roundTrip = 2 τ_hop` (`roundTrip_pos`); ★★ `transfer_at_hopTime : U(τ_hop) e_A = −i • e_B`, ★★ `return_at_roundTrip : U(2 τ_hop) e_A = −e_A`, `arenaFlow_hopTime`, `arenaFlow_roundTrip : arenaFlow (2 τ_hop) (msgPoint 1) = msgPoint 1`; `arenaObs_single_mk : arenaObs (single j j 1) (mk v) = ‖v j‖²/‖v‖²`; `receiver_readout_one = sin² θ`, `receiver_readout_zero = 0`; ★★★ `receiver_distinguishes` (`1` vs `0` at `τ_hop`), ★ `partial_transfer_before_hopTime` (`0 < readout < 1` on `(0, τ_hop)`), `influence_of_message` (+ `msgConfig_siteB`: the preparations agree at `B`); ★★★ `signal_exchange` the package. **Findings:** (i) Mathlib's `exp_add_of_commute` sits in the `NormedAlgebra ℚ` section — use `exp_add_of_commute_of_mem_ball` with `expSeries_radius_eq_top 𝕂 𝔸`; (ii) `Matrix.single_mul_single_of_ne` / `mul_single_apply_of_ne` take the entry `c` as the *first explicit* argument at this pin, and the `_apply` forms must be fed with `exact`, not `rw` (the Matrix-apply transparency trap); (iii) a message point defined through `msgConfig b` cannot be `rw`-reduced under `Projectivization.mk` (motive) — state `msgPoint_one : msgPoint 1 = mk (basisVec excA) _` by `rfl` and rewrite with that. ▸ *Original row:* **(a) The signal exchange: a two-site hop as a `FieldStructuredFlow`, exact transfer with its phase, the Rabi return, and a receiver readout that distinguishes two messages.** On `FieldConfig 2 2` (sites `A = 0`, `B = 1`, occupancy `0/1`) the hop `X = |10⟩⟨01| + |01⟩⟨10|` satisfies `X² = P` (the one-excitation projector) and `X P = X`, so `exp(a X) = 1 + (cosh a − 1) P + sinh a · X` for every `a : ℂ` — proved by **uniqueness of the matrix exponential as the solution of `U' = X U`, `U 0 = 1`** (a Cat-1 lemma for any complete normed algebra, via `hasDerivAt_exp_smul_const` and `is_const_of_deriv_eq_zero`), not by a series. With generator `−(i J/ħ) X`, `J, ħ > 0`: the flow is `1 + (cos θ − 1) P − i sin θ · X`, `θ = J t/ħ`; the excitation at `A` goes to `−i · (excitation at B)` at `τ_hop = πħ/2J` (the phase is a theorem), and to `−(excitation at A)` at `2 τ_hop` — **the echo is the Rabi return**, zero processing delay. Receiver readout `arenaObs (|01⟩⟨01|)`: `sin² θ` for message `1`, `0` for message `0`, so at `τ_hop` it is `1` vs `0` (★★★ `receiver_distinguishes`); **partial transfer before `τ_hop` is a theorem** (`0 < sin² θ < 1` on `(0, τ_hop)`), which is the brief's own warning made precise. Influence by differing preparations: the two message points differ only in the occupancy of `A`. The flow is a `FieldStructuredFlow`, so `lightcone` applies to it with no further hypothesis. **Deliverable:** Cat-1 `Mathlib/Analysis/Normed/Algebra/ExpUnique.lean` (`exp_smul_eq_of_hasDerivAt`, `exp_smul_eq_of_sq_eq`) + `CV/SignalExchange.lean`. ⚠️ **Stuck on:** nothing; M. The drift check is in the header: the hop is QM used as the arena's isolated-item flow (R-016′); the records at `A` and `B` are the ontic writes (#186). | ~~**M**~~ done (M) | High — the cheapest physics of the programme and the test of the flow infrastructure the rest leans on | #39; `CV/FieldStructuredFlow.lean`; `CV/ArenaBridge.lean` (`arenaKick_mk`, `arenaObs`); `specs/records-to-spacetime-scoping.md` §4 | > | ~~184~~ | Records | Records→spacetime | *new (2026-10-22, the clock-and-echo programme — operational coordinates from physical records; split of the author's brief into seven rows, a–g, recorded together)* ▸ **DONE 2026-10-22 (S as priced; built first try).** [`CV/RadarCoordinates.lean`](../CsdLean4/CV/RadarCoordinates.lean) (29 pins, namespace `CSD.CV.RadarReadings`). `structure RadarReadings (emit ret : ℝ) (emit_le_ret)` — the domain is a field; `radarTime = (ret + emit)/2`, `radarSep = (ret − emit)/2`, `radarDist c₀ = c₀ * radarSep`; `emit_le_radarTime`, `radarTime_le_ret`, `radarSep_nonneg`, `radarSep_eq_zero_iff`, `radarDist_nonneg`; ★ `radarTime_sub_radarSep = emit`, `radarTime_add_radarSep = ret`, `ext_of_radar` (equal coordinates ⇒ equal readings; the converse is `congrArg`); ★ `shift s` with `radarTime_shift = T + s`, `radarSep_shift = D`, `radarDist_shift`; ★ `rescale λ hλ` with `radarTime_rescale = λ T`, `radarSep_rescale = λ D`, `radarDist_rescale = λ R`, and **`radarDist_rescale_eq : radarDist c₀ (rescale λ r) = radarDist (c₀ λ) r` — the calibration constant absorbs the unit, nothing fixes `c₀`**; ★★ `abs_radarTime_sub_le`, `abs_radarSep_sub_le` (`|·| ≤ δ`), `abs_radarDist_sub_le` (`≤ c₀ δ`, `0 ≤ c₀`); `radarSep_eq_of_duration_eq`, `ofDuration τ d`, `radarSep_ofDuration = d/2`, `radarSep_roundTrip : (ofDuration τ (roundTrip J ħ)).radarSep = hopTime J ħ` (the *arithmetic* of ~~#183~~'s witness, not its record); ★★ `radar_coordinates` the package. ⚠️ The header says what the row said: algebra with a stated domain, honest only once #186 consumes it with readings that are physical records. ▸ *Original row:* **(b) Radar coordinates and their error bounds — the algebra.** From emission and return clock readings `τ₋ ≤ τ₊`: `T_A = (τ₊ + τ₋)/2`, `D_A = (τ₊ − τ₋)/2 ≥ 0`, `R_A = c₀ D_A` with `c₀ > 0` a *calibration* constant, never a derived signal speed. Theorems: `τ₋ ≤ T_A ≤ τ₊`; equal readings give equal assignments; an origin shift `τ ↦ τ + s` moves `T_A` by `s` and fixes `D_A`; a unit rescaling `τ ↦ λ τ`, `λ > 0`, scales both; with endpoint errors `≤ δτ`: `|T̂_A − T_A| ≤ δτ`, `|D̂_A − D_A| ≤ δτ`, `|R̂_A − R_A| ≤ c₀ δτ`. Deterministic timing error only; probabilistic failure is #189. **Deliverable:** a new module `CV/RadarCoordinates` (to be created) (definitions on `ℝ` with the stated domain `τ₋ ≤ τ₊`). ⚠️ **Stuck on:** nothing; S. ⚠️ Honest only once #186 consumes it with clock readings that are physical records; until then it is algebra with a stated domain, and the module says so. | ~~**S**~~ done (S) | Medium — the interface the capstone #187 states its conclusion in | #39; #183; #186 | > | ~~185~~ | Records | Records→spacetime | *new (2026-10-22, the clock-and-echo programme — operational coordinates from physical records; split of the author's brief into seven rows, a–g, recorded together)* ▸ **DONE 2026-10-22 (M–L took M; one build round of five routine fixes).** [`RecordLayer/RecordClock.lean`](../CsdLean4/RecordLayer/RecordClock.lean) (45 pins, namespace `CSD.RecordLayer`). `ClockArena K = KTorus × (Fin K → KTorus)` (oscillator + `K` slot registers; `K` the horizon); `winElapsed T n u = max 0 (min ((u − nT)/T) 1)` the window schedule — **a modelling input indexed by `t_Σ`, stated as such**; `clockEvolve T s t x = (pshift ((t−s)/T) x.1, fun n => pshift ((winElapsed T n t − winElapsed T n s) · shearGap K) (x.2 n))`; `clockEvolve_self`, ★ `clockEvolve_comp` (two-time law), `measurable_clockEvolve`, ★ `clockEvolve_measurePreserving` (`MeasurePreserving.prod` + `measurePreserving_pi` of `measurePreserving_pshift`) — **a dynamics**; `osc_period : pshift ((t + T)/T) q = pshift (t/T) q` — **the oscillator alone cannot count**; `writtenArc K = {rep ∈ (δ, δ + ε]}` (`= pointerArc K 0`), `readyArc_disjoint_writtenArc`, `gap_add_width_le_one`; `rep_pshift_ready`, ★ `written_of_ready`, ★ `written_mono`; `ClockReady K` (all slots ready, measurable); ★ `slot_ready_before` (`t ≤ nT`), ★ `slot_written_after` (`(n+1)T ≤ t`), ★ `slot_written_mono`; `tickCount x = #{n | x.2 n ∈ writtenArc}`, ★ `measurable_tickCount` (as a sum of indicators), `tickCount_le_horizon`, ★★ `tickCount_mono`, ★★ `tickCount_ge : (n+1)T ≤ t → n+1 ≤ tickCount` (`Finset.Iic n ⊆ filter`, `Fin.card_Iic`), ★★ `tickCount_le : t ≤ nT → tickCount ≤ n` (injection into `range n`; slots `k ≥ n` still ready), ★★ `tickCount_at_period : tickCount (clockEvolve T 0 (mT) x) = m` for `m ≤ K`, ★ `successive_ticks_distinguishable`; `tickProtocol T n τR : MeasurementProtocol (ClockArena K) 1` (start `nT`, readout `(n+1)T`, ready/pointer regions the slot's arcs; `Subsingleton.pairwise`, `Disjoint.preimage`), `tickProtocol_pointerInvariant` (frozen after readout), ★ `tick_persists` **through the corpus's `record_persists_on_interval`** (the outcome sector is `evolve start readout ⁻¹' pointerRegion`, no ready clause); `tickReading τ_tick x = tickCount x · τ_tick` (count and unit apart), ★★ `abs_tickReading_sub_lt : |tickReading T (clockEvolve T 0 t x) − t| < T` on `[0, KT]` — **calibrated with the period, the clock reads within one tick of `t_Σ`**; ★★★ `record_clock` the package. ⚠️ Scope in the header: the writes are scheduled, not triggered by the oscillator's phase (a ripple counter would make the trigger dynamical and is not built); in-window readouts are not specified; Hamiltonian generation not formalised (as for `shearEvolve`/`swapEvolve`). ▸ *Original row:* **(c) The record clock: a rotating pointer plus fresh bank slots, on one arena.** A periodic subsystem does not count cycles; a pointer that completes a cycle and swaps into a fresh slot does. Reuse the swap bank (`RecordLayer/SwapWitness.lean`: `SwapArena`, `swapG`, the slots `j : Fin K`) and the pointer rotation (`RecordLayer/PointerRotation.lean`): tick `n` writes slot `n`; the **tick count** is the number of written slots, read off the register; the observation horizon is the bank size `K`. Theorems: the readout is well defined and measurable; a concrete scheduled evolution writes tick `n` from the ready preparation; successive ticks occupy distinguishable register states; the count is nondecreasing over the horizon; tick records persist for the window (`record_persists_on_interval`); the calibrated reading `τ_A(n) = n · τ_tick`, `τ_tick > 0`, with the physical count and the unit kept apart. The write schedule is a **modelling input indexed by `t_Σ`**, stated as such; it is not emergent time. **Deliverable:** a new module `RecordLayer/RecordClock` (to be created). ⚠️ **Stuck on:** nothing; M–L — the common-arena gluing is #186's, but the clock must already be built so that it can be glued. | ~~**M–L**~~ done (M) | High — the first physical clock in the corpus; operational time as a record | #39; #186; `RecordLayer/SwapWitness.lean`; `RecordLayer/PointerRotation.lean`; `RecordLayer/RecordPersistence.lean` | > | ~~186~~ | Records | Records→spacetime | *new (2026-10-22, the clock-and-echo programme — operational coordinates from physical records; split of the author's brief into seven rows, a–g, recorded together)* ▸ **DONE 2026-10-22 (L as priced; three build rounds).** [`RecordLayer/RecordedExchange.lean`](../CsdLean4/RecordLayer/RecordedExchange.lean) (67 pins, namespace `CSD.RecordLayer`). `XArena K = FieldArena 2 2 × ClockArena K × (Fin 3 → KTorus)` (exchange, clock, registers `0` emission at `A`, `1` reception at `B`, `2` return at `A`); `hopOn u = max u 0` (switch-on at `0`), `arrived τ t = [τ ≤ t]`, `readTime = ![0, τ_hop, 2 τ_hop]`, `readSite = ![A, B, A]`, `readBit c p = [1/2 ≤ momentMap p c]` (measurable by `measurable_momentMap`); `arenaKick_eq_smul`, `measurable_arenaKick` (`continuous_const_smul` from `instContinuousSMul_projectivization`), `hopFlow_zero`, `arenaFlow_zero`; `readAt J ħ i s p` = the bit of the exchange state at the readout time when the propagator starts at `s`; **`xEvolve J ħ T s t`** — the joint two-time propagator, register `i` translated by `(arrived τ_i t − arrived τ_i s) · shearAmt 2 (readAt i s p)`; `xEvolve_self`, ★★ `xEvolve_comp` (the reading is cut-independent by `arenaFlow_add`: `readAt_arenaFlow`), `measurable_xEvolve`, `arenaFlow_measurePreserving` (`fsMeasure_smul_invariant`), ★ `xEvolve_measurePreserving : MeasurePreserving (xEvolve …) ((fsMeasure p₀).prod volume) …` via `MeasurePreserving.skew_product` — **a dynamics**; `XReady`, `xPrep b c r`, `momentMap_msgPoint`, `arenaFlow_msgPoint_zero` (the vacuum is fixed), `readBit_msgPoint_excA`, `readAt_xPrep_zero/one/two`, ★ `readAt_xPrep : readAt J ħ i (−1) (msgPoint b) = b` (reception by `arenaFlow_hopTime`, return by `arenaFlow_roundTrip`); `pshift_shearAmt_mem_pointerArc`, `readyArc_disjoint_pointerArc`; `register_before`, ★★ `register_written : readTime i ≤ t → (xEvolve … (−1) t (xPrep b c r)).2.2 i ∈ pointerArc 2 b`, `register_frozen`; `recordProtocol J ħ T i τR : MeasurementProtocol (XArena K) 2` (prepared at `−1`, readout at `readTime i`), `recordProtocol_pointerInvariant`, ★ `record_persists` through `record_persists_on_interval`; ★★ `reception_depends_on_message` (registers in disjoint arcs for `b = 1` vs `b = 0`, everything else equal), `return_depends_on_message`; `structure Occurrence (register site time message)`, `emission b`, `reception J ħ b`, `returnOcc J ħ b`, **`eventOf J ħ T x o : RecordEvent 2 = ⟨{o.site}, tickCount (xEvolve … (−1) o.time x).2.1⟩`** — the period label is the physical clock's tick count in the arena; `tickCount_clock_neg_one`; `eventOf_emission/reception/return = ⟨{0},0⟩, ⟨{1},1⟩, ⟨{0},2⟩` **under the stated calibration `T = τ_hop`** (`K ≥ 2`); `hopGraph = {(0,1)}`; ★★ `causal_chain` (`CausalPrecedes` ×2 by `decide` on the graph ball, `emission ⪯ return` by `causalPrecedes_trans`); `recordEvents j i = {x | x.2.2 i ∈ pointerArc 2 j}`, `recordPattern j := patternProjection …`, ★ `factorsThrough_recordEvents`; ★★★ `recorded_exchange` the package. **Findings:** (i) the two-time law on a skew product holds because the written bit is read at the *readout time* from the exchange state propagated from the cut — `arenaFlow_add` makes it cut-independent; a shear reading the *current* state would break the law (the swap witness's lesson, in a new place); (ii) `rw [hT] at h` with `hT : T = hopTime` rewrites every `T`, including the clock's period argument — rewrite the goal with `← hT` instead; (iii) `fin_cases i` leaves `![…] ⟨0, ⋯⟩` unreduced — state the per-index lemma with `(0 : Fin 3)` and `show` the reduced form, then `exact` by defeq. ▸ *Original row:* **(d) The common arena and the recorded exchange: clock × hop with one joint flow, occurrences, and the causal chain.** The price driver. Glue #185's clock arena and #183's exchange arena into one product arena with one evolution (the clock's schedule and the hop's flow commute, being supported on disjoint factors — `commute_of_disjointSupport`'s shape), and prove the compatibility maps; a tuple of unrelated witnesses is not accepted (the WAY brick 1 lesson). Define the **occurrences** emission / reception / return along the constructed evolution with their registers and message identity, prove the forgetful map to `CV.RecordEvent` (region + period), and prove `emission ⪯ reception ⪯ return` under `CausalPrecedes` **from the actual transfer under differing preparations**, not from graph reachability. The record predicates factor through the macroscopic projection (`RecordLayer/MacroProjection.lean`). Two clocks already live in the corpus and must be identified explicitly: `RecordEvent.time` is a period index of the write schedule, the clock reads a tick count; they coincide only if the tick schedule *is* the write schedule — a stated modelling choice. **Deliverable:** a new module `RecordLayer/RecordedExchange` (to be created). ⚠️ **Stuck on:** the price, not a wall; L. | ~~**L**~~ done (L) | High — the brick that turns #183–#185 into one experiment | #39; #183; #184; #185; `CV/RecordCausalOrder.lean`; `CV/RecordInfluence.lean`; `RecordLayer/MacroProjection.lean` | > | ~~187~~ | Records | Records→spacetime | *new (2026-10-22, the clock-and-echo programme — operational coordinates from physical records; split of the author's brief into seven rows, a–g, recorded together)* ▸ **DONE 2026-10-22 (M as priced; built first try).** [`RecordLayer/OperationalGeometry.lean`](../CsdLean4/RecordLayer/OperationalGeometry.lean) (14 pins, namespace `CSD.RecordLayer`). `arenaFlow_cycles : arenaFlow (m · roundTrip) (msgPoint 1) = msgPoint 1` (induction on `arenaFlow_add` + `arenaFlow_roundTrip`), `arenaFlow_cycle_reception`, `receiver_cycle : arenaObs occB (arenaFlow (m · roundTrip + τ_hop) (msgPoint 1)) = 1` — **the stationary witness repeats itself**; `xEvolve_exchange_of_nonneg`; `clockReading J ħ T τ_tick x t = tickReading τ_tick (xEvolve … (−1) t x).2.1`, `clockReading_eq`, `clockReading_mono`, `clockReading_at_period`; **`cycleReadings m : RadarReadings`** = the clock's readings at `2m τ_hop` and `(2m+2) τ_hop`, domain condition = tick-count monotonicity; ★★ `radarSep_cycle = τ_tick` (every cycle within the horizon), `radarTime_cycle = (2m+1) τ_tick`; ★★★ `operational_geometry` — **twelve conjuncts**: two-time law; Liouville preservation; clock records persist; exchange records persist; two distinguishable messages; receiver response `1` vs `0`; nonzero round trip; radar separation and time `τ_tick`; every cycle reproduces the separation, returns the exchange to the message point and is received with readout `1`; `emission ⪯ reception ⪯ return`; error bounds `≤ δ`, `≤ δ`, `≤ c₀ δ`; tick count `≤ K`. Hypotheses explicit: `J, ħ > 0`, `T = τ_hop` (the calibration), `2 ≤ K` (the horizon), the ready preparation, `τ_tick > 0`. **§8.3b decision confirmed in the header: a different claim surface from `record_influence_le_of_separated` (cone containment vs operational coordinates), so it stands alone.** No metric, no volume element, no universal `c`, schedule a modelling input, Hamiltonian generation not formalised — all in the header. ▸ *Original row:* **(e) The capstone: operational geometry of the clock-and-echo witness.** On the explicit finite arena and stated ready preparation, the constructed evolution produces persistent clock and exchange records; their readouts define radar coordinates that repeated exchanges reproduce; the represented events respect `CausalPrecedes`; bounded readout errors give bounded coordinate errors. Consumes #183–#186 and their proofs; no structure field may carry the experiment's behaviour. Requires: two distinguishable messages, a nonzero round trip (`2 τ_hop`), a demonstrated receiver response, explicit preparation / schedule / calibration hypotheses, finite memory and horizon handled. **Capstone discipline (§8.3b):** decide at landing whether this strengthens `record_influence_le_of_separated` (`CV/TwoCones.lean`, ~~#98~~) by a conjunct or stands as the records→spacetime capstone of a *different* claim surface (operational coordinates, not cone containment) — the second is the current expectation, to be confirmed against the rule rather than assumed. **Deliverable:** a new module `RecordLayer/OperationalGeometry` (to be created). ⚠️ **Stuck on:** #183–#186; M once they exist. What it does **not** claim, in the header: no metric, no volume element on the event order (§8 of the scoping note), no universal `c` — the cone is one edge per period in a preferred frame (~~#98~~), so a common slope is common relative to the assumed graph and schedule. | ~~**M**~~ done (M) | Highest of the programme — the first statement in the corpus of operational time and separation from physical records | #39; #183–#186; `CV/TwoCones.lean`; `specs/records-to-spacetime-scoping.md` §8 | > | ~~188~~ | Records | Records→spacetime | *new (2026-10-22, the clock-and-echo programme — operational coordinates from physical records; split of the author's brief into seven rows, a–g, recorded together)* ▸ **DONE 2026-10-22 (M as priced; one `rfl` fix).** [`RecordLayer/ClockCalibration.lean`](../CsdLean4/RecordLayer/ClockCalibration.lean) (26 pins, namespace `CSD.RecordLayer`). `TwoClocks K K' = ClockArena K × ClockArena K'`, `twoClockEvolve T₁ T₂ s t = Prod.map (clockEvolve T₁ s t) (clockEvolve T₂ s t)` (self, comp via `Prod.map_comp_map`, measurable, measure-preserving), `reading₁/reading₂`; ★★ `colocated_agreement : |reading₁ (unit T₁) − reading₂ (unit T₂)| < T₁ + T₂` on the common horizon (each clock calibrated to its own period — the *justified* calibration, by `abs_tickReading_sub_lt`); `equal_rate_agreement_at_period` (**equal rates as a hypothesis**, exact agreement at `m T`); `miscalibrated_drift : tickReading τ (clockEvolve T 0 (mT) c) − mT = m (τ − T)` (linear drift — calibration is necessary); ★★ `reception_reading_eq_radarTime` (a clock at `B`, period `τ_hop`, **started at switch-on with `A`'s — the synchronisation convention as a hypothesis** — reads `A`'s `radarTime` of the exchange at reception), `delay_eq_graph_separation` (`B ∉ graphBall {A} 0`, `B ∈ graphBall {A} 1`: one tick, one edge); `elapsed_ticks_emission_return = 2`, `two_le_causalDist` (the experiment's own chain; `causalDist_add_le` + `causalDist_pos_iff`), `bothSitesAtOne = ⟨{0,1}, 1⟩`, ★★ `three_le_causalDist` — **inserting an event lengthens the longest chain to `≥ 3` while the clock's elapsed time is `2`**, `elapsed_ticks_ne_causalDist`: clock time along a history is not the longest-chain separation; they agree only at the fixed resolution of the three events; ★★ `clock_calibration` the package. **Residue → #190** (the longer chain). ▸ *Original row:* **(f) The second clock and calibration consistency.** Add a second physical clock, colocated first: prove agreement after an explicitly justified calibration (equal tick units as a *hypothesis*, never assumed and then reported as dynamics); then a separated stationary arrangement; then, for a longer chain, how clock readings relate to transmission delays and graph separation in the witness. Comparison with `causalDist` (`CV/CausalDistance.lean`) must distinguish time along a clock history from longest-chain separation between endpoint events; a witness-specific equality or bound only where justified, with the event resolution fixed explicitly because inserting events changes chain length. **Deliverable:** a new module `RecordLayer/ClockCalibration` (to be created). ⚠️ **Stuck on:** #187; M. | ~~**M**~~ done (M) | Medium — the first test of whether different clock constructions agree, relative to the assumed graph | #39; #187; `CV/CausalDistance.lean` | > | ~~189~~ | Records | Records→spacetime | *new (2026-10-22, the clock-and-echo programme — operational coordinates from physical records; split of the author's brief into seven rows, a–g, recorded together)* ▸ **DONE 2026-10-22 (M as priced; one build round).** [`RecordLayer/ExchangeFailure.lean`](../CsdLean4/RecordLayer/ExchangeFailure.lean) (20 pins, namespace `CSD.RecordLayer`). `ExchangeValid J ħ T b x := ∀ i, register i displays b at its readout`, `recordFailure b i`, `notReady i`, `messageFailure b i := {readAt i (−1) x.1 ≠ b}`; `register_at_readout` (= `pshift (shearAmt 2 (readAt …)) (x.2.2 i)`); ★ `recordFailure_subset_union : recordFailure ⊆ notReady ∪ messageFailure`; `invalid_eq_iUnion`, `measurableSet_recordFailure/invalid`; ★★ `measure_invalid_le (μ) : μ {¬valid} ≤ Σ_i μ (recordFailure i)` (`measure_iUnion_fintype_le` — **a union bound, no independence claimed**), `measure_invalid_le' : ≤ Σ_i (μ (notReady i) + μ (messageFailure i))`, `prob_valid_ge : 1 − Σ … ≤ μ {valid}` (`prob_compl_eq_one_sub`); `exchangeValid_xPrep` (the exact witness never fails), `dirac_invalid_eq_zero`; **`volume_band_le`** — #103's `volume_rep_preimage_Ioc_le` on the register torus (the quantitative input for a register prepared off the ready arc); ★ **`pshift_shearAmt_notMem_pointerArc` — FINDING: the arc write has no tolerance**: a register whose `rep` exceeds the ready width `ε` misses the pointer arc by exactly its excess (the arcs of `ShearWitness.lean` are as wide as the ready arc); ★★ `retention_sufficient : roundTrip ≤ τR → every record alive on [readTime i, roundTrip]` (via `record_persists`), `records_at_return`; ★★ `exchange_failure` the package. ⚠️ `carrier_persistence_window` **not consumed, and the header says why**: it bounds a field-mode readout under a periodic unitary perturbation; this arena's records are torus arcs with no perturbation model and are frozen exactly after readout (`register_frozen`), so retention is exact. ▸ *Original row:* **(g) Probabilistic failure of the exchange and the retention window.** Separate from #184's deterministic timing error: a bound on the probability that the complete exchange is invalid (a record or message failure), by a **union bound** over the three writes unless independence is proved — consuming the robust-set measure bounds of ~~#103~~ (`k·N·δ`) and ~~#106~~'s robust fraction; and the retention condition `τ_R > τ_round` as a sufficient condition for the witness, derived from `carrier_persistence_window` (`CV/CarrierPersistence.lean`) where it applies. **Deliverable:** a new module `RecordLayer/ExchangeFailure` (to be created). ⚠️ **Stuck on:** #187; M. | ~~**M**~~ done (M) | Medium — robustness of the capstone under finite resolution | #39; #187; `CV/CarrierPersistence.lean`; `RecordLayer/MacrostateStability.lean` | > | ~~190~~ | Records | Records→spacetime | *new (2026-10-22, residue of ~~#188~~ — the longer chain)* ▸ **DONE 2026-10-23 (L as priced; one build round on the CV module, the clocks module built first time).** Two modules. [`CV/ChainExchange.lean`](../CsdLean4/CV/ChainExchange.lean) (49 pins, namespace `CSD.CV`): `swapSites k k'`, ★ `swapHop k k'` — **the local hop on an arbitrary edge**, the swap permutation matrix with its diagonal removed (`X² = P` the projector on what the edge moves, `X P = X`, Hermitian), ★ `swapHop_supportedOn : SupportedOn {k, k'}` (the two-site hop of ~~#183~~ never had to show this); `edgeFlow J ħ k k' : FieldStructuredFlow L 2`, ★★ `edgeFlow_val_eq` (`1 + (cos θ − 1) P − i sin θ X`, by `ExpUnique`), ★★ `edgeFlow_transfer` (`e_k ↦ −i e_{k'}` at `τ_hop`); **`Schedule K N`** — the scheduled generator as data (`stage : ℕ → FieldStructuredFlow K N`, `τ`), `completed m` (ordered product), `propagator t` (completed stages below `t` and the running one, from switch-on at `0`; identity before), `evolve s t = propagator t · (propagator s)⁻¹`, ★ `evolve_comp` — **a two-parameter dynamics, not a one-parameter group**; `propagator_at_stage`; `FieldStructuredFlow.off`, `flow_zero`; `chainSchedule n J ħ` (stage `m < n` = edge `(m, m+1)`, off afterwards; `chainSchedule_edges_of_lt/of_le`, `chainSchedule_edges_subset` — **the active edges are visible and lie on the path graph**); ★★★ `chain_transfer : U(m τ_hop) e_0 = (−i)^m e_m` for every `m ≤ n`, `chain_transfer_last` (the far end in `(L−1) τ_hop`, phase `(−i)^{L−1}`), `propagator_vac`; `occAt k`, `chainMsg n b`, `chainReadout_one` (`= 1` iff `k = m` at `m τ_hop`), `chainReadout_zero`, ★★★ `receiver_distinguishes_chain`, `excitation_position`; `pathGraph n`, `pathGraph_one = hopGraph`, ★ `mem_graphBall_pathGraph : j ∈ graphBall {0} d ↔ j ≤ d`, **`graphDist E R k := sInf {d | k ∈ graphBall E R d}`**, ★★ `graphDist_pathGraph = k`; ★★★ `scheduled_chain`. [`RecordLayer/ChainClocks.lean`](../CsdLean4/RecordLayer/ChainClocks.lean) (30 pins, namespace `CSD.RecordLayer`): `ChainArena n K = FieldArena (n+1) 2 × (Fin (n+1) → ClockArena K)`, `chainEvolve J ħ T s t` (★ `chainEvolve_comp`, `measurable_chainEvolve`, ★ `chainEvolve_measurePreserving` on `fsMeasure p₀ × Haar^{n+1}`), `ChainReady`, `chainPrep`, `siteReading`; `tickCount_at_reception = k`, ★★ `reception_reading = k τ_hop`, ★★★ **`clock_delay_eq_graphDist`** — the ticks at site `k` when the signal arrives equal `graphDist (pathGraph n) {0} k`, **clock delay = graph distance on this witness**, because every edge is on for the same hop time (the header says what an effective cone for the network would have to be: the same `τ_hop` on every edge, supplied by the schedule — nothing universal, no limiting speed); ★★ `reception_readout`; `eventOfReception`, `eventOfReception_eq = siteEvent k = ⟨{k}, k⟩` (labelled by the physical clock); `pairEvent k = ⟨{k, k+1}, k+1⟩`, the `CausalPrecedes` chain on the path graph; ★ `le_causalDist_siteEvents` (the site events alone give `≥ k`, matching the clock), ★★ `two_mul_le_causalDist` (**inserting the pair events gives `≥ 2k`**), ★★ `clock_time_lt_causalDist` (`k < causalDist` for every `k ≥ 1`: ~~#188~~'s inflation for every `k`); ★★★ `chain_clocks`. ⚠️ Synchronisation (common start at `0`) and calibration (`T = τ_hop`) are hypotheses; horizon `n ≤ K`. ▸ *Original row:* **Clock readings against transmission delay and graph separation on an `L`-site chain.** ~~#188~~ does the two-site case: the signal's clock delay is one tick and `B` is one edge from `A` (`delay_eq_graph_separation`). For a chain of `L` sites a *uniform* hopping Hamiltonian does not transfer an excitation perfectly end to end (dispersion), so exact transmission needs a **sequentially activated coupling schedule** — hop `(k, k+1)` switched on for one hop time each, in order — which must be exposed as a modelling input (the brief's own warning). Deliverable: an `L`-site `FieldStructuredFlow` with a scheduled generator (piecewise in time: the two-time law needs the crossing convention of ~~#186~~), exact transfer along the chain in `(L−1) τ_hop` with the accumulated phase `(−i)^{L−1}`, a clock at each site started at switch-on, and the theorem that the reception reading at site `k` is `k` ticks, equal to the graph separation `k` on the path graph — **clock delay = graph distance** on this witness, with its schedule visible; and the longest-chain inflation of ~~#188~~ generalised. ⚠️ **Stuck on:** nothing theoretical; the scheduled generator is new infrastructure (the corpus's flows are one-parameter or single-crossing); L. | ~~**L**~~ done (L) | Medium — the first network statement of the programme; what a universal effective cone would have to look like, relative to the graph | #39; ~~#183~~; ~~#186~~; ~~#188~~; `CV/FieldStructuredFlow.lean`; `CV/TwoCones.lean` | > | ~~191~~ | Records | Records→spacetime | *new (2026-10-23, the author's second operational-geometry brief — the occurrence registry, event order against duration, one internally activated write, its receiver, radar with a detector; split into #191–#197, recorded together)* **The occurrence registry with provenance, and reception registers on the chain.** ▸ **DONE 2026-10-23 (M, within the M–L price; three build rounds, all local: decidability of arc membership (`open Classical in`), the deprecated `if_pos` family, a wrong rewrite occurrence).** [`RecordLayer/RecordOccurrence.lean`](../CsdLean4/RecordLayer/RecordOccurrence.lean) (57 pins, namespace `CSD.RecordLayer`). `IsRecordWrite h t₀ τ₀ τ₁ w v` (not `v` on `[t₀, τ₀)`, `v` on `[τ₁, τ₁ + w]`); ★ `not_isRecordWrite_of_eq`, ★ `IsRecordWrite.not_within` — **no second write of a carrier inside a write's window, for any value**; `Observation X Ω` (kind 2), `ActualOccurrence X Ω K` (kind 1: two-time measurable `evolve`, `readout`, `ready`, the actual `prep`, `site`, `start`, `value`, `writeStart ≤ writeEnd`, `window`, and `isWrite` **as a proof field**), `toObservation`, `toRecordEvent clk` (kind 3), `recordSet` measurable, `recordPattern`, ★ `factorsThrough_recordSet`. Instances: `exchangeRegister i` (readout `arcReadout : KTorus → Fin 3`, `2` = no record; `toRecordEvent_exchangeRegister_zero/one/two` are ~~#186~~'s `eventOf` by `rfl`), ★ `exchange_no_second_write`; `clockSlot n` (write interval `[nT, (n+1)T]`, from `slot_ready_before` / `slot_written_after`). The chain: `receiverObservation k` is kind 2 (what ~~#190~~ had); `ChainXArena n K`, `chainReadAt`, `chainXEvolve` (★ `chainXEvolve_comp`, `measurable_chainXEvolve`, ★ `chainXEvolve_measurePreserving` — skew product over the chain arena), `chainXPrep`, ★ `chainReadAt_chainMsg` (register `k` reads the message at `k τ_hop`: `chainReadBit_one` through `momentMap_eq_arenaObs_occAt` and `chainReadout_one`, `chainReadBit_zero`), ★★ `chainRegister_written`, `chainRegister_before`, `chainRegister_frozen`, `chainRegister k` (the actual occurrence), ★★★ `siteEvent_eq_toRecordEvent` — **the site events of ~~#190~~'s causal chain are the forgetful images of actual writes on the common arena, labelled by the site's own clock, whose ticks are writes** — the §2 acceptance criterion; ★★★ `record_occurrences`. ⚠️ The chain registers are still written on the schedule's crossing of `k τ_hop` (as ~~#186~~); the carrier-activated write is #194. ▸ *Original row:* `RecordLayer/RecordOccurrence.lean`. `IsRecordWrite h t₀ τ₀ τ₁ w v` — a readout history is not `v` on `[t₀, τ₀)` and is `v` on `[τ₁, τ₁ + w]`: a write interval `[τ₀, τ₁]` and a finite window `w`; `ActualOccurrence X Ω` — the carrier's evolution (two-time, measurable), its readout, the ready set, the **actual preparation**, the value, the write interval and the window, with the write **discharged as a field proved from the dynamics**, never assumed. The three kinds kept apart: an actual occurrence; a possible observation (`Observation`, a measurable readout and nothing else); the region label `CV.RecordEvent`, reached by the forgetful map `toRecordEvent` under a clock readout. ★ `not_isRecordWrite_within` — **inside a write's window there is no second write of that carrier**: inspecting an unchanged register is not an occurrence. Instances discharged from the existing dynamics: the exchange's emission, reception and return registers (~~#186~~, sharp write at the readout time) and every clock slot (~~#185~~, write interval `[nT, (n+1)T]`); `recordSet` measurable and factoring through the pattern projection (`MacroProjection.lean`). **The chain of ~~#190~~ has clocks and receiver *readouts* at its sites but no registers**, so its receptions are observations, not writes: add a reception register at every site (`ChainXArena n K`, `chainXEvolve` — two-time, measure-preserving skew product), prove each is written with the message from the transmitted state at `k τ_hop` and frozen after, and that `siteEvent k` is the forgetful image of an actual write whose period label is the clock's own writes — **an experiment whose causal-event labels come from actual record writes on its common arena** (the brief's §2 acceptance criterion). ⚠️ **Stuck on:** nothing; a bounded extension of ~~#186~~ and ~~#190~~ (the generic-`n` registers cost what the smallest chain costs). | ~~**M–L**~~ done (M) | Medium — the brief's §2 acceptance criterion; the registry every later row writes into | #39; ~~#185~~; ~~#186~~; ~~#190~~; `RecordLayer/RecordedExchange.lean`; `RecordLayer/MacroProjection.lean` | > | ~~192~~ | Records | Records→spacetime | *new (2026-10-23, the author's second operational-geometry brief — the occurrence registry, event order against duration, one internally activated write, its receiver, radar with a detector; split into #191–#197, recorded together)* **Chronological precedence against causal precedence.** ▸ **DONE 2026-10-23 (S–M as priced; two build rounds, both local: `(a, b).1 = (a, b).2` hypotheses need a typed `have` before `rw`, and `RecordEvent` inequality is not `decide`-able — use the named `_ne_` lemmas).** Two modules. [`CV/ChronoPrecedes.lean`](../CsdLean4/CV/ChronoPrecedes.lean) (33 pins, namespace `CSD.CV`): `ChronoPrecedes E e₁ e₂ := CausalPrecedes E e₁ e₂ ∧ (e₁ = e₂ ∨ e₁.time < e₂.time)`, `.causalPrecedes`, `chronoPrecedes_iff_of_time_ne` (agreement on distinct periods), ★ `not_chronoPrecedes_of_time_eq` (**distinct simultaneous events are incomparable** — the handling of simultaneous writes), ★ `isPartialOrder_chronoPrecedes`; `strictChrono E` with ★ `time_lt_of_strictChrono` (**no step with unchanged occurrence time**), `isTrans_strictChrono`, `isIrrefl_strictChrono`, `strictChrono_le_strictPrecedes` (every chrono step is a causal step), ★ `time_add_length_le_of_chain` (the period strictly increases along a chain: acyclic); `chronoChainLengths`, `chronoDist`, ★★ `chronoChainLengths_le_elapsed`, ★★ `chronoDist_le_elapsed` (**the longest chrono chain is at most the elapsed periods**), ★ `chronoDist_le_causalDist`, `chronoDist_pos_iff`, `chronoDist_add_le` (the reverse triangle inequality survives). [`RecordLayer/ChainChrono.lean`](../CsdLean4/RecordLayer/ChainChrono.lean) (14 pins, namespace `CSD.RecordLayer`): ★ `not_chronoPrecedes_pairEvent` (**the inserted step of ~~#190~~'s inflation, `pairEvent k → siteEvent (k+1)`, is not a chrono step**: both at period `k+1`), `siteEvent_strictChrono_succ`, ★★ `chronoDist_siteEvents = m` (the clock's elapsed ticks under `T = τ_hop`), ★ `chronoDist_lt_causalDist` (`m ≥ 1`; `two_mul_le_causalDist` untouched); the exchange: `exchange_chrono_chain`, ★ `not_chronoPrecedes_bothSites_reception` (~~#188~~'s inserted step), ★★ `chronoDist_exchange = 2` against `three_le_causalDist`, `chronoDist_exchange_lt_causalDist`; ★★★ `chain_chrono`. **`CausalPrecedes` is not redefined**; the general statement is the inequality `chronoDist ≤ elapsed`, the equality is each witness's. ▸ *Original row:* `ChronoPrecedes E e₁ e₂ := CausalPrecedes E e₁ e₂ ∧ (e₁ = e₂ ∨ e₁.time < e₂.time)` — the sub-relation whose strict part demands a strictly later period. Bridges: it implies `CausalPrecedes`; it agrees with it on events of distinct periods; it is a partial order and its strict part is acyclic (the period strictly increases along a chain); every genuine transmission edge of the exchange and the chain is a chrono step. ★★ `chronoDist_le_elapsed` — **its longest chain is bounded by the elapsed periods** `e₂.time − e₁.time`, the site chain of the path graph attains the bound, and the inflation `two_mul_le_causalDist` stands untouched: the extra steps are exactly the same-period region-inclusion steps `pairEvent k → siteEvent (k+1)`, which `ChronoPrecedes` excludes (`not_chronoPrecedes_pairEvent`). Simultaneous writes: distinct events at one period are chrono-incomparable. **Not** a redefinition of `CausalPrecedes`, which keeps its name and its theorems. ⚠️ In general chrono chain length is not clock time either — a system may write several records per tick — so the equality on the chain is a property of that witness, and the general statement is the inequality. ⚠️ **Stuck on:** nothing. | ~~**S–M**~~ done (S–M) | Medium — the brief's §3 diagnosis, as theorems | #39; ~~#188~~; ~~#190~~; `CV/RecordCausalOrder.lean`; `CV/CausalDistance.lean` | > | ~~193~~ | Records | Records→spacetime | *new (2026-10-23, the author's second operational-geometry brief — the occurrence registry, event order against duration, one internally activated write, its receiver, radar with a detector; split into #191–#197, recorded together)* **Operational duration from calibrated readings.** ▸ **DONE 2026-10-23 (S as priced at S–M; built first time).** [`RecordLayer/OperationalDuration.lean`](../CsdLean4/RecordLayer/OperationalDuration.lean) (18 pins, namespace `CSD.RecordLayer`): `clockHistory T τ c t := tickReading τ (clockEvolve T 0 t c)`, `duration T τ c s t := clockHistory t − clockHistory s`; ★ `duration_nonneg` (`tickCount_mono`), `duration_add` (**telescoping, named as algebraic consistency, not proper time**), `duration_self`, `duration_neg`; `tickCount_clockEvolve_of_nonpos` (a switch-on reference anywhere before `0` gives the same readings), `offsetHistory`, ★ `duration_offset` (**clock-origin invariance**); ★ `duration_rescale` (`τ ↦ λ τ` scales every duration by `λ`); ★★ `abs_duration_sub_lt` (**resolution**: `|duration − (t − s)| < 2T` over the horizon, from `abs_tickReading_sub_lt` twice); `duration_eq_zero_of_writeEnd_eq` (two `ActualOccurrence`s completing together are at zero duration — duplicate descriptions, ~~#191~~); `edgeDuration_nonneg`, ★ `sum_edgeDuration` (`Finset.sum_range_sub`); the chain witness: ★★ `duration_reception` (`= m T` from the switch-on to the reception at site `m`), ★★★ `duration_eq_chronoDist_mul` — **clock duration = chrono distance × tick, while `causalDist ≥ 2m`**; ★★★ `operational_duration`. The deliverable in words is the header's last section: `causalDist` counts strict steps of permitted influence including same-period region steps; `chronoDist` keeps the period-advancing ones and is bounded by the elapsed periods; a clock duration is a difference of tick counts of one calibrated register; they coincide on the chain witness because every edge is on for one period, under the stated calibration and event resolution. ▸ *Original row:* a new module `OperationalDuration` under `RecordLayer/`: along one clock history, `duration x s t := tickReading τ (clock at t) − tickReading τ (clock at s)`; non-negativity for `s ≤ t` (`tickCount_mono`); additivity on consecutive intervals, **named as the algebraic consistency it is** (the telescoping identity is not a derivation of proper time); invariance under a common shift of the clock's origin; the scaling law under the unit rescaling `τ ↦ λ τ`; the resolution bound from `abs_tickReading_sub_lt`; zero duration between two descriptions of one occurrence (same write time); for an occurrence chain the edge durations read off one clock are non-negative and sum to the end-to-end duration. ★★ The deliverable, as theorems on the chain witness with `T = τ_hop`: `causalDist` counts permitted-influence steps including zero-duration region steps; clock duration counts ticks of a calibrated register; the chrono chain length (#192) and the duration in ticks coincide there; the causal longest chain exceeds both. The calibration and the event resolution are stated hypotheses. ⚠️ **Stuck on:** nothing. | ~~**S–M**~~ done (S) | Medium — the brief's §4 | #39; ~~#185~~; ~~#188~~; #192 | > | ~~194~~ | Records | Records→spacetime | *new (2026-10-23, the author's second operational-geometry brief — the occurrence registry, event order against duration, one internally activated write, its receiver, radar with a detector; split into #191–#197, recorded together)* **One internally activated write — the research gate.** ▸ **DONE 2026-10-23 (L as priced; the construction works as stated — two modules, 180 pins; four build rounds, the walls being Lean's rather than the physics': `Continuous.comp` against `arenaObs`/`arenaKick` unfolds `Projectivization.rep` down to `Classical.choice` and times out, so the arena continuity lemmas are stated in applied form and composed with `comp₂`; `NormedSpace.exp_continuous` now wants a `ℚ`-algebra, so exp continuity goes through `exp_analytic`).** [`CV/DetunedExchange.lean`](../CsdLean4/CV/DetunedExchange.lean) (104 pins, `CSD.CV`): `H(Δ) = J X + Δ n_B = (Δ/2) P + Ω Y` with `Y² = P`, `detunedFlow J ħ Δ : FieldStructuredFlow 2 2` (the hop at `Δ = 0`), ★★ `detunedFlow_val_eq` (**the flow in closed form**, `exp_add_of_commute` + `exp_smul_eq_cosh_sinh` twice — `ExpUnique`, as the row asked), ★★ `detuned_readout_one` = `(J/Ω)² sin²(Ωt/ħ)`, the **detuning loss** `one_sub_div_rabiFreq_sq` (`1 − (J/Ω)² = (Δ/2)²/Ω²`), attained at `detunedHopTime` and strictly positive for `Δ ≠ 0`; the vacuum fixed; ★ `arenaObs_detunedH_arenaFlow` (the carrier's energy conserved); `occHist`/`occAccum` (the receiver's occupancy and its time integral) with ★ `occAccum_add` (the flow's group law) and ★ `hasDerivAt_occAccum`; ★★ `occAccum_msgPoint_one` in closed form; `continuous_arenaDM`, ★ `continuous_arenaKick_prod` (joint continuity by descent through `mk'`, general `ι`). [`RecordLayer/ActivatedWrite.lean`](../CsdLean4/RecordLayer/ActivatedWrite.lean) (76 pins, `CSD.RecordLayer`): `detuningAt = g f(θ₂)`, `rateAt = −(2g/ħ) f'(θ₂)` (`Periodic.lift`, `Function.Periodic.deriv_periodic`; the factor `−2/ħ` is the arena form's normalisation — fixed 2026-10-24 by #195, the row's `g f'(θ₂)` was the naive one); ★ `activeFlow` (the ray by the detuned exchange at fixed `θ₂`, `θ₁` translated by `g f'(θ₂) ∫ n_B`); ★★ `activeFlow_add` (a one-parameter group), `activeEvolve` with the two-time law and ★ `activeEvolve_shift` (**autonomous — no test of `t_Σ`**); ★★ `activeFlow_measurePreserving` (**the skew product over the conserved `θ₂`**: Fubini–Study unitary × Haar translation); ★★ `activeEnergy_activeFlow` (`J⟨X⟩ + g f(θ₂) n_B` conserved), `activeFlow_angle` (`θ₂` conserved); ★★ `hasDerivAt_pointerAmount` (**`θ̇₁ = −(2g/ħ) f'(θ₂) n_B(p(t))`**; with `occHist_msgPoint_one_zero`/`_pos` the pointer starts at rest and is set in motion by the arriving carrier); ★★★ `pointerAmount_msgPoint_one` (**the message-dependent displacement** `κ(ωt − sin ωt cos ωt)/(2ω)`, `κ = −(2g/ħ) f'(θ₂)(J/Ω)²`, `ω = Ω/ħ`) against `activeFlow_msgPoint_zero` (the vacuum fixes the whole arena point); ★★ `message_one_not_displayed`/`message_one_displayed` (write interval `[1/(6κ), 2/(3κ) + 1/(2ω)]`, window `2/(3κ) − 1/ω`, under `0 < κ` and `3κ ≤ 2ω`); ★★★ `activatedRegister : ActualOccurrence ActiveArena Bool 2` at site `{B}` (~~#191~~); ★★ `records_distinguish`; ★★ `message_one_returns` (**finite retention**: an explicit later time at which the pointer is back where message `0` left it, by the IVT); ★★ `transition_region` (**the obstruction half**: for every horizon `T` the angles with `|g f'(θ₂)| T < 1/6` form an open set of positive Haar measure, nonempty by Rolle on the periodic `f`, from which the write is not displayed by `T` — exact only off a transition region, as the trilemma says and the brief allows); ★★★ `activated_write`. ⚠️ Hamiltonian generation on the arena is #195, as the row says: proved here are the dynamics, its two-time law, Liouville, energy conservation and both equations of motion (`hasDerivAt_detunedFlowVal`, `hasDerivAt_pointerAmount`), not the identification with the Hamiltonian flow of the arena form. The walls: `FrozenBase` is passed because the base moves (the detuning is the back-reaction); the chart witness of `UntriggeredFlow` has an arena-level counterpart; row 350's trigger is the carrier. ▸ *Original row:* On `FieldArena 2 2 × KTorus` (the two-site exchange of ~~#183~~ and one pointer at `B`), the **time-independent joint Hamiltonian** `𝓗(p, θ) = ⟨p, H_hop p⟩ + g · momentMap p excB · f(θ₂)` with `f` smooth and periodic: the pointer angle `θ₁` moves at rate `g · ⟨n_B⟩ · f'(θ₂)` — **only while the carrier occupies the receiver, so the activation is the arriving carrier and no test of `t_Σ` appears**; `θ₂` is conserved (`θ̇₂ = −∂𝓗/∂θ₁ = 0`, so the torus motion is generated by a periodic potential and no global translation claim is made); the back-reaction on the ray is the detuning `H_hop + g f(θ₂) n_B` — a state-dependent generator, so the uncoupled transfer formula is replaced, not reused. Deliverables: the generator as data; the flow in closed form (a detuned Rabi problem in the one-excitation sector by `ExpUnique`, the vacuum fixed); the two-time law; measure preservation (skew product: a unitary on the ray at fixed `θ₂`, a translation of `θ₁` by a function of `(p, θ₂)`); ★★★ message-dependent pointer displacement, the vacuum leaving the pointer fixed; distinguishable records for the two messages with an explicit bound on the detuning loss; a finite retention window (the carrier oscillates on, so the pointer keeps moving — the window is where the two pointer positions stay in disjoint arcs); the write registered as an `ActualOccurrence` (#191) with its write interval. The write is continuous in the preparation, so it is exact only off a transition region — the proved trilemma, and the brief allows it. ⚠️ **Stuck on:** whether the construction works as stated **is** the question; the failure mode is priced too — a theorem that no such `𝓗` on this arena separates the two messages' pointer positions, with its hypotheses. Hamiltonian generation on the arena is #195; without it this is still a dynamics (two-time, measure-preserving), not yet a Hamiltonian flow. Row 350's finding (*the trigger is the obstacle*), `FrozenBase`, `UntriggeredFlow` and `JointLift` are the recorded walls this must pass. | ~~**L**~~ done (L) | High — the first write in the corpus activated by a physical degree of freedom rather than a schedule; the brief's §5 | #39; ~~#183~~; ~~#186~~; #191; `SigmaLayer/UntriggeredFlow.lean`; `SigmaLayer/FrozenBase.lean`; `RecordLayer/JointLift.lean` | > | ~~195~~ | Records | Records→spacetime | *new (2026-10-23, the author's second operational-geometry brief — the occurrence registry, event order against duration, one internally activated write, its receiver, radar with a detector; split into #191–#197, recorded together)* **Hamiltonian generation of #194 on the arena.** ▸ **DONE 2026-10-24 (L–XL took L; the arena route, no chart fallback — three modules, 65 pins; five build rounds, all shape: tangent-space scalars need `(1 : ℝ →L[ℝ] ℝ) s`, `areaForm` evaluates only on plain `ℝ × ℝ` vectors (`show` first), `ContMDiff.mul`/`.add` need `Mathlib.Geometry.Manifold.Algebra.Structures`, `Continuous.subtype_mk _` on the unitary group leaves the topology stuck — `continuous_induced_rng`).** [`Mathlib/Geometry/Manifold/Instances/AddCirclePeriodicLift.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/AddCirclePeriodicLift.lean) (7 pins, Cat-1, `Function.Periodic`): the lift of a periodic `f` to the circle in the translation atlas has manifold derivative `f'` (`hasMFDerivAt_lift`, `mfderiv_lift`) and is `C^n` when `f` is (`contMDiff_lift`); `deriv_periodic`. [`LF4/ArenaCoupledFlow.lean`](../CsdLean4/LF4/ArenaCoupledFlow.lean) (31 pins, `CSD.LF4`): ★ `schrodingerField_add_real_smul` (the Schrödinger field is linear in the matrix, by uniqueness for the symplectic `fsForm`), ★ `isHamiltonianVectorField_torusPotential` (`X_{f(θ₂)} = (f'(θ₂), 0)` for `dθ₁∧dθ₂`), `coupledEnergy A B f = −2⟨A⟩ − 2⟨B⟩ f(θ₂)`, ★★ `isHamiltonianVectorField_coupled` (**the coupled product field the row was stuck on**: `ι_X ω = d𝓗` with `X = (X_A + f(θ₂) X_B, (−2⟨B⟩ f'(θ₂), 0))`, from `HasMFDerivAt.mul` and `prodSum_pair`), ★★ `isMIntegralCurve_coupledCurve`, ★★★ `hamiltonianFlow_coupledEnergy` (**the flow in closed form**: the Schrödinger unitary of `A + f(θ₂) B` on the sector, `θ₁` advanced by `f'(θ₂) ∫₀ᵗ (−2⟨B⟩)` along it, `θ₂` fixed — by uniqueness of integral curves on the compact arena), ★ `arenaVolume_map_coupledFlow`. [`RecordLayer/ActivatedHamiltonian.lean`](../CsdLean4/RecordLayer/ActivatedHamiltonian.lean) (27 pins, `CSD.RecordLayer`): `expectation_eq_arenaObs` (the two expectation functionals agree), `hamA = J X/ħ`, `hamB = g n_B/ħ` reindexed to `Fin 4`, ★ `activatedEnergy` with ★★ `activatedEnergy_eq` (**the Hamiltonian is `−(2/ħ)` times #194's `activeEnergy`**), ★★ `idxUnitary_detunedFlow` (the reindexed detuned exchange is the Schrödinger unitary of `A + f(θ₂) B`, by `ExpUnique` in the matrix algebra), ★ `pointerAmount_eq`, ★★★ `hamiltonianFlow_activeFlow` (**#194's `activeFlow` is the Hamiltonian flow of `𝓗` on `Σ = ℂℙ³ × T²`**, through `fibredArenaEquiv 2 2`), ★ `arenaVolume_map_activeFlow`, ★★★ `activated_hamiltonian`. **Finding:** the row's `θ̇₁ = g f'(θ₂) n_B` was the naive normalisation; with the corpus's Fubini–Study form the Schrödinger unitary of `H` is the flow of `−2⟨H⟩`, so the coupled term inherits the factor and the pointer rate is **forced** to `−(2g/ħ) f'(θ₂) n_B` — #194's `rateAt` now carries it (its statements are otherwise unchanged). The ray lives in `ℂℙ³`, not the row's `ℂℙ¹`: the one-excitation sector is a `ℂℙ¹` inside it. ▸ *Original row:* The flow of #194 is the Hamiltonian flow of `𝓗` for the arena symplectic form of `LF4/ArenaSymplectic.lean` (R-016′): the ray part by `schrodingerField_isHamiltonianVectorField` with the `θ₂`-dependent generator, the torus part `θ̇₁ = ∂𝓗/∂θ₂`, `θ̇₂ = 0`. Chart-level fallback through `ChartIntegralCurve` with the transport boundary stated, as `UntriggeredFlow.lean` does. ⚠️ **Stuck on:** the Hamiltonian vector field on the product `ℂℙ¹ × T²` of a function coupling both factors — the corpus has each factor's field and the arena form, not the coupled field; and R-016's chart→arena transport if the chart route is taken. | ~~**L–XL** (chart fallback **M**)~~ done (L, arena route) | Medium — turns #194's dynamical law into a Hamiltonian flow | #194; `LF4/ArenaSymplectic.lean`; `SigmaLayer/UntriggeredFlow.lean`; R-016′ | > | ~~196~~ | Records | Records→spacetime | *new (2026-10-23, the author's second operational-geometry brief — the occurrence registry, event order against duration, one internally activated write, its receiver, radar with a detector; split into #191–#197, recorded together)* **Coupled reception.** ▸ **DONE 2026-10-24 (M as priced; one build round of three shape fixes).** [`RecordLayer/CoupledReception.lean`](../CsdLean4/RecordLayer/CoupledReception.lean) (26 pins, `CSD.RecordLayer`): `detuned_readout_one_zero_detuning` (**at `Δ = 0` the coupled occupancy is #183's `sin²(Jt/ħ)`**); `abs_sin_sq_sub_sin_sq_le` (`sin²` is 1-Lipschitz), `rabiFreq_sub_le` (`Ω − J ≤ Δ²/(8J)`), `one_sub_div_rabiFreq_sq_le`, ★★ `occupancy_loss_le` (**the loss is continuous and quadratic**: `|n_B^Δ(t) − n_B^0(t)| ≤ (1/(4J²) + |t|/(8Jħ)) Δ²`, peak loss plus phase drift), `occupancy_loss_le_of_abs_le_one` (the row's `≤ C|Δ|` for `|Δ| ≤ 1`); ★★ `readBit_detuned_one`/`readBit_detuned_zero`/`receiver_bit_distinguishes` (**#186's thresholded receiver bit still separates the messages at the detuned hop time whenever `|Δ| ≤ 2J`**, i.e. peak loss ≤ the threshold `1/2`); ★ `formation_interval` (the `IsRecordWrite` of #194 on `[1/(6κ), 2/(3κ) + 1/(2ω)]` with the lifted displacement pinched `0 ≤ amount ≤ κ t` by the occupancy integral; retention by `message_one_returns`); `ClockedArena K = ActiveArena × ClockArena K`, `clockedEvolve` with the two-time law and Liouville (`MeasurePreserving.prod`), ★★ `clockedRegister : ActualOccurrence (ClockedArena K) Bool 2` (~~#191~~), `clockedReading`, `emissionEvent = ⟨{A}, 0⟩`, `receptionEvent_time`/`_time_le`, ★★ `emission_causalPrecedes_reception` (**the occurrence's event, labelled by #185's arena clock, is causally preceded by the emission on the hop graph** as soon as `T ≤ writeEnd`, via `tickCount_ge` and `graphBall_mono`); ★★★ `coupled_reception`. The clock remains #185's scheduled device (honest scope): the activation of the write is internal, the clock label is a second reading. ▸ *Original row:* The compatibility theorem between #194's coupled dynamics and the uncoupled transfer of ~~#183~~: the occupancy of `B` under detuning `Δ = g f(θ₂)` is the detuned Rabi law, exact transfer at `Δ = 0` and a continuous loss `|occupancy − uncoupled| ≤ C Δ`; the receiver record distinguishes the two preparations when the loss is below the arc separation; the formation interval `[τ₀, τ₁]` of the write follows from the occupancy integral and the preparation; retention per #194; the occurrence's event, labelled by the arena clock of ~~#185~~, respects `CausalPrecedes` on the hop graph. ⚠️ **Stuck on:** #194. | ~~**M**~~ done (M) | Medium — the brief's §6 acceptance criterion: one reception record whose activation is internally generated | #194; ~~#183~~; ~~#186~~ | > | ~~197~~ | Records | Records→spacetime | *new (2026-10-23, the author's second operational-geometry brief — the occurrence registry, event order against duration, one internally activated write, its receiver, radar with a detector; split into #191–#197, recorded together)* **Radar with a detector.** ▸ **DONE 2026-10-24 (M as priced; one build round, one unused hypothesis).** [`RecordLayer/RadarDetector.lean`](../CsdLean4/RecordLayer/RadarDetector.lean) (16 pins, `CSD.RecordLayer`): `transferError = πħΔ²/(16J³)` with ★ `abs_detunedHopTime_sub_hopTime_le` (the detuned arrival `πħ/(2Ω)` against #183's `τ_hop`, from `rabiFreq_sub_le`); `detectorDelay = writeEnd − detunedHopTime = 2/(3κ) + 1/(2ω) − π/(2ω)` (**the detector delay of the witness, explicit in the preparation**); `detectorReadings` (the arena clock at the emission and at a reading time in the retention window; domain by `tickCount_mono`), `detectorReadings_emit = 0` (exactly, `tickCount_at_period`), `idealReadings = (0, τ_hop)`, `radarBudget = T + retention + |detectorDelay| + transferError`, ★★ `abs_detectorReadings_ret_sub_le` (**each contribution derived for the witness** — clock resolution by `abs_tickReading_sub_lt`, the detector delay and the transfer error by the witness's formulas — **and the two input bounds labelled**: the reading is taken while the record displays, `t ≤ K T`); ★★★ `radar_with_detector` (**one bound on radar time, separation and distance** through `abs_radarTime_sub_le`/`abs_radarSep_sub_le`/`abs_radarDist_sub_le`, with the record displayed at the reading time); ★ `radarTime_delayed`/`radarSep_delayed` (**a detector of response time `d` shifts both coordinates by `d/2`** — the interpretation). Honest scope: one clock at `A` and a one-way flight (the detection is the return reading), no universal `c`, no Lorentz invariance; the #187 and #190 headers stand. **The second brief (#191–#197) is complete.** ▸ *Original row:* Replace the scheduled reception of ~~#187~~'s witness by #196's internally activated one and carry the detection delay (the write interval's length), the clock resolution (`abs_tickReading_sub_lt`), the transfer error (#196) and the retention condition (the ~~#189~~ pattern) into one bound on radar time and separation through `abs_radarTime_sub_le` / `abs_radarSep_sub_le`; each contribution derived for the witness, any input bound labelled as such; a detector with nonzero response time shifts the coordinate by its delay, and that shift appears in the interpretation. No universal `c` and no Lorentz invariance is read off any of it (the ~~#187~~ and ~~#190~~ headers stand). ⚠️ **Stuck on:** #194, #196. | ~~**M**~~ done (M) | Medium — the brief's §7 | #194; #196; ~~#187~~; ~~#189~~; `CV/RadarCoordinates.lean` | > | ~~198~~ | Records | Records→spacetime | *new (2026-10-24, the author's third operational-geometry brief — the complete physical echo with both timestamps at `A`, an internally driven clock, geometry across several paths, which measurements belong to geometry, a common effective cone; split into #198–#205, recorded together)* **The two-site exchange with both sites detuned.** ▸ **DONE 2026-10-24 (M as priced; three build rounds, all shape).** [`CV/TwoSiteDetuned.lean`](../CsdLean4/CV/TwoSiteDetuned.lean) (81 pins, `CSD.CV`): ★ `twoDetunedH_eq` (**`H = H(Δ_B − Δ_A) + Δ_A P`**), `twoDetunedFlow : FieldStructuredFlow 2 2`, `sectorPhase` with `sectorPhase_eq`/`sectorPhase_unitary`/`sectorPhase_commute`, ★★ `twoDetunedFlow_val_eq` (**the flow is the sector phase times the detuned exchange in the difference**, `exp_add_of_commute`); ★★ `arenaObs_twoDetunedFlow` (**every observable commuting with `P` reads the same along both flows at every arena point**), `occB_twoDetunedFlow = occHist J ħ (Δ_B − Δ_A)`, `arenaObs_oneExc_twoDetunedFlow` (the sector is conserved); ★★ `twoDetunedFlow_arenaFlow_msgPoint_one` (**on the ray of message `1` the two flows agree**), `twoDetunedFlow_msgPoint_zero`; ★★ `twoDetuned_readout_B_one`/`twoDetuned_readout_A_one` (`(J/Ω')² sin²(Ω't/ħ)` at `B`, `1 − (J/Ω')² sin²` at `A`), both zero for message `0`; `detunedRoundTrip = πħ/Ω' = 2 × detunedHopTime`, ★★★ `detunedFlow_return`/`twoDetuned_return` (**at the detuned round trip the carrier is back at `A` as a ray, exactly** — the column of `e_A` is the scalar `−e^{−iΔπ/(2Ω')}·e_A`), `twoDetuned_return_nsmul` (and at every multiple), `twoDetuned_return_occA = 1`/`_occB = 0`; `occHistA`/`occAccumA` (**the history of `A` and its integral** = sector weight − `occHist`/`occAccum` of `B`), `continuous_occHistA_prod`/`continuous_occAccumA_prod` (jointly in the ray and both detunings — what #199's skew product needs), `hasDerivAt_occAccumA`, `occAccumA_msgPoint_one = t − occAccum`; ★ `two_site_detuned` (six conjuncts). Nothing new beyond the difference, as priced. ▸ *Original row:* `H = J X + Δ_A n_A + Δ_B n_B = detunedH J (Δ_B − Δ_A) + Δ_A · P` (the one-excitation projector commutes with everything), so the flow is the detuned exchange of ~~#194~~ in the difference `Δ_B − Δ_A` times a sector phase, invisible on the rays of the two messages: `n_B(t) = (J/Ω')² sin²(Ω't/ħ)`, `n_A(t) = 1 − n_B(t)` for message `1` with `Ω' = √(J² + (Δ_B − Δ_A)²/4)`, both zero for message `0`; the **detuned round trip** `πħ/Ω'` at which the carrier is back at `A` with `n_A = 1`; a `FieldStructuredFlow 2 2` on the one edge. The carrier physics of the autonomous echo, nothing new in it beyond the difference. ⚠️ **Stuck on:** nothing. | ~~**M**~~ done (M) | Medium — the carrier every echo row below uses | #39; ~~#194~~; ~~#183~~; `CV/DetunedExchange.lean` | > | ~~199~~ | Records | Records→spacetime | *new (2026-10-24, the author's third operational-geometry brief — the complete physical echo with both timestamps at `A`, an internally driven clock, geometry across several paths, which measurements belong to geometry, a common effective cone; split into #198–#205, recorded together)* **The autonomous echo arena.** ▸ **DONE 2026-10-24 (L as priced; four build rounds, the arc arithmetic resolved by a band argument).** [`RecordLayer/EchoArena.lean`](../CsdLean4/RecordLayer/EchoArena.lean) (131 pins, `CSD.RecordLayer`; + 11 pins in `CV/TwoSiteDetuned.lean`: group laws `occAccumA_add`/`occAccum_twoDetuned_add`, `arenaObs_twoDetunedH_arenaFlow`, applied-form `Continuous.occAccumA`/`.occAccum`): `EchoArena = FieldArena 2 2 × (T² × T²)`, `PointerCoupling`, ★ `echoFlow` (ray by the two-site detuned exchange at the conserved angles; pointer `A` by `r_A ∫ n_A`, pointer `B` by `r_B ∫ n_B`), ★★ `echoFlow_add` (group), `echoEvolve_comp`/`echoEvolve_shift` (two-time, autonomous); ★★ `echoFlow_measurePreserving` (**Liouville** = `echoRayStep ∘ echoShiftB ∘ echoShiftA`, each measure-preserving; the generic `measurePreserving_cshift_mid` translates one circle coordinate by a measurable function of the rest, used once per pointer after `MeasurableEquiv.prodAssoc`); ★★ `echoEnergy_echoFlow`, `hasDerivAt_echoAmountA/B` (**each pointer moves only while the carrier is at its station**); `echoFlow_msgPoint_zero` (the vacuum fixes both pointers); ★★★ `echoAmountA_msgPoint_one` (`D_A = r_A (t − λ(ωt − sin ωt cos ωt)/(2ω))`, `λ = (J/Ω')²`) and `echoAmountB_msgPoint_one` (#194's `pointerDrift` with `κ_B = r_B λ`); ★ `echo_return_ray` (the carrier back at `A` at `πħ/Ω'`); `senderDrift_hopTime`, `sender_rate_hopTime` (**the sender's plateau**: at the detuned hop time pointer `A` moves at `r_A(1 − λ)`, the loss alone), `receiver_rate_roundTrip` (pointer `B` stationary at the round trip); `senderRead` (pointer `A` in thirds: ready / emitted / returned), `receiverRead` (#194's display arc); ★★★ `emissionRecord` (site `{A}`, value `1`, interval `[1/(6r_A), (1/3 + b)/μ]`), `receptionRecord` (site `{B}`, #194's interval with `κ_B`), `returnRecord` (site `{A}`, value `2`, interval `[1/(2r_A), (2/3 + b)/μ]`), window `(1/6 − 2b)/μ` with `μ = r_A(1 − λ/2)`, `b = r_A λ/(4ω)` — **all three `ActualOccurrence`s on the one arena from the one preparation**, under `0 < r_A`, `0 < κ_B`, `3 r_A λ ≤ ω`, `3 κ_B ≤ 2ω`, both pointers ready; ★★ `emission_return_disjoint` (**the return arc is disjoint from the emission arc and written after it**), ★★ `sender_returns`/`receiver_returns` (finite retention, IVT), ★★★ `autonomous_echo` (19 conjuncts). **The stuck-on resolved**: the row expected the arcs to be separated by plateau arithmetic; the band `|D_A − μt| ≤ b` about the mean growth does it for every `λ ∈ (0,1]` under the single slowness condition `3 r_A λ ≤ ω` (= `b ≤ 1/12`), and the plateaus are stated exactly alongside. Honest scope: Hamiltonian generation on `ℂℙ³ × T² × T²` is #202's second half; with `λ < 1` pointer `A` never fully stops (rate `r_A(1 − λ)` at the hop time). LEAN: `Continuous.comp₂` of `continuous_occAccumA_prod` against an expected type times out — the applied-form lemmas; `OpensMeasurableSpace`/`BorelSpace` of the depth-3 product need explicit instances or `synthInstance.maxSize 512`; compositions of shifted tuples are proved componentwise, never by whole-tuple `rfl` (eta on `AddCircle` addition blows up). ▸ *Original row:* `FieldArena 2 2 × T² × T²`: one pointer at `A` coupled to `n_A`, one at `B` coupled to `n_B`, both detunings the pointers' own back-reaction (#198). `A`'s pointer is set in motion by the carrier **twice** — the emission phase (the carrier leaves `A` as `cos²`) and the return phase (it comes back at the detuned round trip) — so **the emission and the return are the two displacement arcs of one register at `A`**, and the reception is `B`'s arc, exactly as in ~~#194~~. Deliverables: the two-time law and Liouville (a skew product over both conserved angles), the three writes as `ActualOccurrence`s (~~#191~~) with their intervals derived from the occupancy integrals `∫ n_A`, `∫ n_B`, the return arc provably disjoint from the emission arc for the stated preparations, the vacuum fixing both pointers, retention finite for all three. ⚠️ **Stuck on:** the arc arithmetic that separates the return arc from the emission arc — the emission phase deposits `κ_A π/(4ω')` before the carrier has left, and the return arc must sit beyond it with a margin that survives the detuning; this is a theorem to find, not a wall. | ~~**L**~~ done (L) | High — the brief's §1: a round trip whose every record is activated by the carrier | #39; #198; ~~#194~~; ~~#191~~; `RecordLayer/ActivatedWrite.lean` | > | ~~200~~ | Records | Records→spacetime | *new (2026-10-24, the author's third operational-geometry brief — the complete physical echo with both timestamps at `A`, an internally driven clock, geometry across several paths, which measurements belong to geometry, a common effective cone; split into #198–#205, recorded together)* **Round-trip radar from `A`'s records.** ▸ **DONE 2026-10-24 (M as priced; two build rounds, both shape).** [`RecordLayer/EchoRadar.lean`](../CsdLean4/RecordLayer/EchoRadar.lean) (33 pins, `CSD.RecordLayer`): `EchoClockedArena K = EchoArena × ClockArena K`, `echoClockedEvolve` (two-time, measurable, Liouville by `MeasurePreserving.prod`); ★★ `clockedEmissionRecord`/`clockedReturnRecord` (**`A`'s two writes on the clocked arena**, `isWrite` inherited from #199), `emissionEvent_time`/`returnEvent_time` (labels = the clock's tick count at each write's completion), ★ `emission_causalPrecedes_return` (same site, later label; `subset_graphBall`); ★ `abs_detunedRoundTrip_sub_roundTrip_le` (**outward + return transfer** ≤ `2 transferError`); `echoReturnDelay` with `echoReturnDelay_eq` (**the return write's detector response**, explicit: `(2/3 + r_A λ/(4ω))/(r_A(1 − λ/2)) − πħ/Ω'`); `echoRadarBudget = T + senderWindow + emitEnd + |echoReturnDelay| + 2 transferError`; `echoReadings` (**`A`'s clock at the two records**), ★★ `echoReadings_eq_labels` (**at the write completions the radar timestamps are `T` × the two event labels**), `idealEchoReadings = (0, 2τ_hop)` with `idealEchoReadings_radarSep = τ_hop`; ★★ `abs_echoReadings_emit_sub_le`/`abs_echoReadings_ret_sub_le` (each term derived; the two retention windows and the clock horizon are the labelled inputs); ★★★ `radar_from_sender_records` (**one bound on radar time, separation and distance against the round trip, both records displayed at their reading times** — a radar separation read entirely at one station). Honest: the emission timestamp is late by the emission write's completion (a derived detector response, in the budget as `emitEnd`); the clock is still #185's scheduled device (#201/#202); no universal `c`, no Lorentz invariance. ▸ *Original row:* Both radar timestamps are readings of `A`'s clock associated with `A`'s two writes (emission arc, return arc) on the arena of #199 with the record clock alongside; the ideal readings are `(0, 2τ_hop)`; the budget is **outward transfer + detector response + return transfer + clock resolution**, each derived for the witness (the two transfer errors from ~~#197~~'s `transferError` at the detuned round trip, the response from the arc-crossing times, the resolution from `abs_tickReading_sub_lt`), the retention window and the clock horizon labelled as the input bounds; then `abs_radarTime_sub_le` / `abs_radarSep_sub_le` / `abs_radarDist_sub_le`. This completes the operational interpretation of the detector construction: a radar separation read entirely at one station. ⚠️ **Stuck on:** ~~#199~~ done 2026-10-24; nothing. | ~~**M**~~ done (M) | High — the brief's §1 proof target | #39; #199; ~~#197~~; ~~#187~~; `CV/RadarCoordinates.lean` | > | ~~201~~ | Records | Records→spacetime | *new (2026-10-24, the author's third operational-geometry brief — the complete physical echo with both timestamps at `A`, an internally driven clock, geometry across several paths, which measurements belong to geometry, a common effective cone; split into #198–#205, recorded together)* **The drift pointer as an autonomous hand clock.** ▸ **DONE 2026-10-24 (M–L took M; five build rounds, all shape — ceiling API, a scoped `ω` notation clash, two name clashes with `RecordClock`).** [`RecordLayer/HandClock.lean`](../CsdLean4/RecordLayer/HandClock.lean) (44 pins, `CSD.RecordLayer`): `handCount K q = ⌈K·rep q⌉₊ − 1` with ★ `handCount_eq_iff` (**the arcs are the intervals `(n/K, (n+1)/K]`**, disjoint), `handReading`, `handReady` (first half-arc — the preparation width must be below the arc width, #194's lesson); `hand t q = (activeFlow t (msgPoint 1, q)).2` (**the controller is the carrier of message `1`, in the model with its back-reaction**), `hand_eq` (the drift staircase), ★ `pointerDrift_mono` (**the hand never turns back**, by `pointerAmount_add`); `tickTime = 2/(Kκ)` with ★ `pointerDrift_tickTime` (**the calibration is derived**: after `n τ_tick` the hand has crossed `n` arcs up to the band `κ/(4ω)`), `handHorizon = (2/κ)(1 − 1/K) − 1/(2ω)` (one turn; `handHorizon_pos` for `K ≥ 2`, `κ ≤ ω`); ★★ `handCount_mono` (**ticks persist**), ★★ `abs_handReading_sub_le` (**the resolution bound `τ_tick + 1/(2ω)`**, exactly as rowed: one arc of count uncertainty plus the staircase's band); `tickStart = (2n−1)/(2Kκ)`, `tickEnd = n τ_tick + 1/(2ω)`, `tickWindow = τ_tick/2 − 1/ω` (≥ 0 ⟺ `Kκ ≤ ω`), ★★ `hand_before_tick`/`hand_at_tick`, ★★★ `tickWrite n : ActualOccurrence ActiveArena ℕ 2` for `1 ≤ n ≤ K − 1` (site `{B}`, value `n`), `tickWrite_value_ne` (**distinguishable**), `tickEnd_succ` (spaced by `τ_tick`), ★ `handTick_persists`; ★ `handClock_generated` (= `hamiltonianFlow_activeFlow` at the clock's preparation — **no new geometry**); ★★★ `hand_clock` (7 conjuncts). Honest: a staircase, not a uniform rotation — the band `1/(2ω)` is the clock's intrinsic jitter; one turn only; the loss keeps the hand from stopping exactly while the carrier is at `A`. **Goal 2 of the third brief complete; #202 is unblocked on its dynamics half.** ▸ *Original row:* ~~#194~~'s pointer advances by exactly `κπ/(2ω)` per Rabi half-period and monotonically in between (`pointerDrift` has `0 ≤ D(t) ≤ κt` and `κ(t/2 − 1/(4ω)) ≤ D(t) ≤ κ(t/2 + 1/(4ω))`): read with `K` arcs of width `1/K` it is **the smallest finite clock whose internal state activates its tick writes** — the carrier is the controller, prepared explicitly, its resources (one carrier, one pointer) and its back-reaction (the detuning) in the model. Deliverables: the tick count as the arc index of the displacement, monotone non-decreasing (ticks persist), successive ticks distinguishable (disjoint arcs), the **derived calibration** `τ_tick = 2/(Kκ)`, the **resolution bound** `|tickReading − t| ≤ τ_tick + 1/(2ω)`, the horizon one turn, and Hamiltonian generation inherited from ~~#195~~ with no new geometry. ⚠️ **Stuck on:** nothing; a reading of what landed in ~~#194~~/~~#195~~. | ~~**M–L**~~ done (M) | High — the brief's §2: the reference clock internally driven | #39; ~~#194~~; ~~#195~~; ~~#185~~ | > | ~~202~~ | Records | Records→spacetime | *new (2026-10-24, the author's third operational-geometry brief — the complete physical echo with both timestamps at `A`, an internally driven clock, geometry across several paths, which measurements belong to geometry, a common effective cone; split into #198–#205, recorded together)* **The autonomous clock-and-echo model on one arena — the brief's immediate deliverable.** ▸ **DONE 2026-10-24, BOTH HALVES (dynamics M as priced; generation L as priced — the second torus factor was one `IsSymplectic.prodForm` more, as the row said).** **Dynamics:** [`RecordLayer/ClockAndEcho.lean`](../CsdLean4/RecordLayer/ClockAndEcho.lean) (25 pins): `ClockEchoArena = EchoArena × ActiveArena`, `clockEchoEvolve` (two-time, ★ autonomous `clockEchoEvolve_shift`, Liouville), ★★ `clockEchoEmission`/`clockEchoReception`/`clockEchoReturn`/`clockEchoTick` (**the echo's three writes and the clock's ticks as `ActualOccurrence`s on the one arena, every one activated by a carrier**), `clockEchoLabel` (the hand's count), `emissionLabel_time`/`returnLabel_time`, `clockReadingAt`, `clockEchoReadings` (**the timestamps are the hand's readings while `A`'s emission and return records display**), ★★ `clockEchoReadings_eq_labels` (at the write completions = `τ_tick` × the two hand labels), `clockEchoBudget = (τ_tick + 1/(2ω_c)) + senderWindow + emitEnd + |echoReturnDelay| + 2 transferError` (**the derived round-trip bound with the clock's resolution in place of #185's `T`**), ★★★ `autonomous_clock_and_echo` (the deliverable, 14 conjuncts). **Generation:** [`LF4/EchoCoupledFlow.lean`](../CsdLean4/LF4/EchoCoupledFlow.lean) (29 pins, generic `ℂℙᴺ × T² × T²`): `echoForm = prodForm arenaForm torusAreaForm`, ★ `echoForm_isSymplectic`, `twoCoupledEnergy`/`twoCoupledField`, ★ `isHamiltonianVectorField_fstLift` (**a Hamiltonian field on the arena lifts to the product**), ★ `isHamiltonianVectorField_outerTerm`, ★★ `isHamiltonianVectorField_twoCoupled` (`ι_X ω = d𝓗`), `twoCoupledMatrix`/`twoAccumA`/`twoAccumB`/`twoCoupledCurve`, ★★ `isMIntegralCurve_twoCoupledCurve`, ★★★ `hamiltonianFlow_twoCoupledEnergy` (**the flow in closed form, a skew product over both conserved angles**); [`RecordLayer/EchoHamiltonian.lean`](../CsdLean4/RecordLayer/EchoHamiltonian.lean) (19 pins): `hamOccA`, `echoEnergySigma`, ★★ `echoEnergySigma_eq` (**`−(2/ħ)` × #199's joint energy**), ★★ `idxUnitary_twoDetunedFlow` (ODE uniqueness), `twoAccumA_eq`/`twoAccumB_eq`/`echoAmountA_eq`/`echoAmountB_eq`, ★★★ `hamiltonianFlow_echoFlow` (**the echo is the Hamiltonian flow of its energy on `ℂℙ³ × T² × T²`**), ★ `map_hamiltonianFlow_echoFlow` (Liouville inherited), ★★★ `echo_hamiltonian`. The clock factor's generation is #195's `activated_hamiltonian`; the combined arena is the product of the two. Honest: the echo and the clock do not interact (a product, colocated at `A` by the site label); coupling them is a further row (~~#212~~, done 2026-10-29; generation ~~#225~~, done 2026-10-29). **THE THIRD BRIEF'S DELIVERABLE IS COMPLETE.** ▸ *Original row:* The echo of #199 and the hand clock of #201 on one common arena with one two-time measure-preserving evolution; the emission and return timestamps are readings of the autonomous clock at `A`'s two writes; the round-trip timing bound of #200 with the clock's derived resolution in place of ~~#185~~'s; explicit detector back-reaction throughout. Dynamics first (skew product, **M**); Hamiltonian generation of the combined arena after (**L**): the symplectic product `ℂℙ³ × T² × T²` and a second sector factor need `IsSymplectic.prodForm` run once more and sector energies on a product of sectors. ⚠️ **Stuck on:** ~~#199~~ ~~#200~~ ~~#201~~ (all done 2026-10-24) — nothing for the dynamics half; and for the generation half, the second torus factor in the arena of `LF4/ArenaSymplectic.lean` (one `prodForm` more, not a wall). | ~~**M** dynamics, **L** generation~~ done (M + L) | Highest in the brief — its one exact deliverable | #39; #199; #200; #201; `LF4/ArenaCoupledFlow.lean` | > | ~~203~~ | Records | Records→spacetime | *new (2026-10-24, the author's third operational-geometry brief — the complete physical echo with both timestamps at `A`, an internally driven clock, geometry across several paths, which measurements belong to geometry, a common effective cone; split into #198–#205, recorded together)* **Geometry across several paths: three stations, two routes.** ▸ **DONE 2026-10-27 (L as priced; the stuck-on resolved by choosing two distances and proving both sides).** [`CV/WalkSchedule.lean`](../CsdLean4/CV/WalkSchedule.lean) (43 pins): `Step` (hop or wait), `walkSchedule` (stage `m` = the `m`-th step's flow), `IsWalk`, `walkPos`, `hopCount`; ★★★ `walk_transfer` (**#190's chain transfer generalised to any legal walk with waits**: after `m` stages the excitation is `(−i)^{hops}` times the excitation at the walk's position, by induction through `Schedule.completed_succ_of_shift`); `walkReadout`, `walkReadout_arrival`; ★★ `walk_compose`/`walk_transfer_append` (**known relay delays compose**: route ++ wait^ρ ++ route has length `|w₁| + ρ + |w₂|`); ★★ `isWalk_cycles`, `take_cycles`, `walkPos_take_cycle`, `walkReadout_cycle` (**repeated exchanges are reproducible**); ★ `isWalk_map`, `walkPos_map`, `hopCount_map` (**relabelling the stations relabels the walk**). [`CV/ThreeStation.lean`](../CsdLean4/CV/ThreeStation.lean) (52 pins): `TravelData` (one-way travel, response and recording delays), `rawSep` and `correctedSep`; ★★ `correctedSep_comm` (**the corrected separation is always symmetric**, reciprocal or not — the round trip averages directions); ★★ `rawSep_comm_iff` (**the raw separation is symmetric exactly when `resp − recDelay` agrees at the two stations**); `RelayBound`, ★★ `correctedSep_triangle` (**triangle up to the relay delay**); `correctedSep_eq_travel_of_reciprocal`; ★★★ `correctedSep_metric` (no self-travel + positive travel + free relays ⇒ metric axioms); `correctedSep_comap`/`rawSep_comap` (**re-description invariance**); `ThreeStation` (direction-dependent direct waits, relay delays, detector delays), `directRoute`/`relayRoute`/`route` (faster), ★★ `route_spec`/`route_arrival` (**every route a legal walk arriving after exactly the operational travel time**), ★★★ `relayBound` (**the operational travel times satisfy the relay bound** — 27 cases), `metric_of_free_relays`; `roundTripWalk`, ★★ `roundTrip_spec`/`roundTrip_reception`/`roundTrip_return`, ★★★ `roundTrip_reproducible`; **where it is not a metric**: ★★ `rawSep_not_symmetric` (unequal net detector delays), ★★ `fixedRoute_not_triangle` (radar along a fixed route: `5 > 1 + 1`), ★★ `relayDelay_breaks_triangle` (`5 > 1 + 1`, yet `relayDelay_within_bound`); ★★★ `three_station_geometry` (11 conjuncts). Honest: travel times are whole numbers of hop times (every stage lasts `τ_hop`); "fastest route" is over the schedules offered, not a race; detector delays are inputs here (#199–#204 derive them); no universal `c`, no continuum metric. ▸ *Original row:* On the chain machinery of ~~#190~~ with three sites and two transmission routes (direct and via the relay), with explicitly modelled detector delays and direction dependence: repeated exchanges are reproducible (the same readings every cycle — the ~~#187~~ pattern), re-describing the same events leaves the results unchanged (the ~~#193~~ duplicate-description invariance), known relay delays compose, and **for a stated travel-time distance under stated stationary and reciprocal hypotheses, symmetry and the triangle inequality are derived** — radar separation is not a metric in general, and the row must prove where it is and state where it is not. ⚠️ **Stuck on:** the choice of distance; the stationary and reciprocal hypotheses have to be read off the dynamics (equal hop times on every edge give reciprocity by the schedule; with detectors the delays must be directionally symmetric), and the triangle inequality is where the relay delay enters. | ~~**L**~~ done (L) | Medium — the brief's §3 | #39; ~~#190~~; ~~#193~~; ~~#187~~; `CV/ChainExchange.lean` | > | ~~204~~ | Records | Records→spacetime | *new (2026-10-24, the author's third operational-geometry brief — the complete physical echo with both timestamps at `A`, an internally driven clock, geometry across several paths, which measurements belong to geometry, a common effective cone; split into #198–#205, recorded together)* **Which measurements are properties of geometry: detector-class stability.** ▸ **DONE 2026-10-27 (M–L took M; two build rounds, one lemma name).** [`RecordLayer/DetectorClass.lean`](../CsdLean4/RecordLayer/DetectorClass.lean) (14 pins, `CSD.RecordLayer`): `emitCorrection` (the emission write's completion) and `returnCorrection` (the return write's response, #200's `echoReturnDelay`) — **the derived responses a detector subtracts, known from its own preparation, not fitted**; `correctedSep`; `correctedBudget = T + senderWindow + 2 transferError` (**clock, retention window, the two transfers — nothing of the detector's response**); ★★ `abs_correctedEmit_le`, `abs_correctedRet_sub_roundTrip_le`, ★★ `abs_correctedSep_sub_hopTime_le` (**the corrected separation is within the corrected budget of the hop time**); `rawSep_eq_correctedSep_add` (**what the correction removes**: raw separation = corrected + half the net delay, so raw separations of two detectors differ by a detector-dependent amount on one arrangement); ★★★ `detector_class_stability` (**two detectors — two couplings, two preparations, two clocks — on one arrangement agree in corrected separation up to the sum of their corrected budgets**); `transferError_mono`, `correctedBudget_le_of_class`, ★★ `stable_regime` (**the regime**: for the class detuning `≤ δ`, window `≤ w`, clock `≤ T₀`, every corrected separation is within `T₀ + w + 2 transferError J ħ δ` of the hop time and any two within twice that — the hop time belongs to the arrangement, the detector enters only through the class). Honest: the residual is not zero — the loss shifts the arrival itself by `transferError`, removable only by knowing the detuning, and the regime is quantitative in `δ`; the clock is #185's (with #201's the resolution term is `τ_tick + 1/(2ω_c)`, as in #202). ▸ *Original row:* Two detectors (two couplings `g`, `f`, or two preparations `θ₂`) measuring the same arrangement; their corrected readings (the raw reading minus the detector's derived delay) agree up to a residual bounded by ~~#196~~'s loss and the two resolutions; the regime in which the inferred separation is stable across a specified class of clocks and detectors, as an explicit theorem with its class stated. ⚠️ **Stuck on:** ~~#200~~ done 2026-10-24; nothing. | ~~**M–L**~~ done (M) | Medium — the brief's §4: geometry shared by measuring systems | #39; #200; ~~#196~~; ~~#197~~ | > | 205 | Foundations | Records→spacetime | *new (2026-10-24, the author's third operational-geometry brief — the complete physical echo with both timestamps at `A`, an internally driven clock, geometry across several paths, which measurements belong to geometry, a common effective cone; split into #198–#205, recorded together)* **A common effective causal cone across apparatus.** ▸ **FIRST BRICK DONE 2026-10-27 (M as priced); the row stays open as research.** [`CV/ConeComparison.lean`](../CsdLean4/CV/ConeComparison.lean) (35 pins): ★ `norm_swapHop` (the hop on a genuine edge has norm exactly 1), ★★ `lrVelocity_edgeFlow` (a stage's Lieb–Robinson velocity is `4e|J/ħ|`), ★★★ `lrVelocity_mul_hopTime` / `lrVelocity_eq_coneRatio_mul` (**per stage the two cones are proportional with a universal ratio**: `v_LR · τ_hop = 2πe`, whatever `J` and `ħ`; `17 < 2πe < 18`), ★★ `arrival_mem_lrCone` / `lrTime_lt_arrival` (the arrival at `d` sites is inside the bound's cone, which opens `2πe` times earlier), ★★ `lr_bound_one_stage` (over one stage the bound suppresses only beyond 18 sites; the stage moves one), ★★★ `pathFlow_lrVelocity_mul_hopTime_bounds` (**FINDING: the same couplings packaged as one flow on `n` edges have a Lieb–Robinson slope `n/2` to `n` times a stage's** — `lrVelocity` is built from the whole generator's norm, so the corpus's bound does not survive a change of coarse description; residue rowed as ~~#213~~, done 2026-10-27: `localLRVelocity` = `8eΔγ`, `8πe` hop-time cones on the chain at every length), ★★★ `cone_comparison` (5 conjuncts). Answer to the first-brick question: on the schedule the hop-time cone and the Lieb–Robinson cone are one cone up to the constant `2πe` — the hop-time cone is the sharp one and the corrected radar readings (#203) read it; the bound is that cone widened. What stays open is the row itself: a dynamical condition under which one effective cone survives changes of apparatus and description. ▸ *Original row:* Whether the independent clock, carrier and detector constructions recover compatible propagation limits: a dynamical condition under which one effective cone survives changes of apparatus and of coarse description — stronger than the schedule's equal hop time on every edge. The corpus has two cones, the Lieb–Robinson cone with velocity `2e²‖S‖` (`CV/LiebRobinson.lean`) and the chain's hop-time cone; whether they, and the detectors' corrected readings, agree on one cone is the question. A universal `c`, Lorentz transformations and event-volume reconstruction are sharper targets only after this comparison. ⚠️ **Stuck on:** the theory — this is the Lorentz-invariance question in operational dress; what can be done first is the comparison of the two existing cones on the chain witness (M), priced here as the row's first brick. | research; first brick ~~**M**~~ done (M) | High if it acquires a shape, otherwise positioning | #39; #203; #204; `CV/LiebRobinson.lean`; `CV/ChainExchange.lean` | > | ~~212~~ | Records | Records→spacetime | *new (2026-10-24, residue of ~~#202~~)* **Couple the clock to the echo.** ▸ **DONE 2026-10-29, the dynamics (L as priced; generation split off as #225).** [`RecordLayer/ClockEchoCoupling.lean`](../CsdLean4/RecordLayer/ClockEchoCoupling.lean) (94 pins): the clock's carrier detuned at `A` by the echo's pointer is #199's echo flow run on the clock's view (`clockView`: its carrier, the echo's pointer `A`, the hand), so `coupledFlow` = clock step after echo step; ★★ `echoStep_comm_clockStep` (**the steps commute**: they share only pointer `A`), ★★ `coupledFlow_add`, `coupledFlow_angles` (**the three conserved angles stay conserved**), ★ `coupledEvolve_shift`, ★★ `coupledEvolve_measurePreserving` (**Liouville**, through `clockShuffle`), ★★ `coupledEnergy_coupledFlow` (**the joint energy, cross term included, is conserved**); generic window arithmetic for any displacement in a band (`bandEmitEnd`, `mem_emit_of_band`, …) and #201's hand for any `κ, ω` (★ `pointerDrift_monotone`, `driftHand_count_mono`, `abs_driftHand_reading_sub_le`, `driftHand_before_tick`, `driftHand_at_tick`); ★★ `coupledFlow_pointerA` (**reading the clock disturbs the echo's register**: pointer `A` turns by the echo's drift plus the clock carrier's); ★★ `coupledEmission`, `coupledReception`, `coupledReturn`, `coupledTick` (**the three writes and the ticks still `ActualOccurrence`s**, coupled windows; the reception's are #199's); `couplingTerm`, `coupledBudget` (**the budget acquires a term**: the coupling's shift of the return record), ★★ `couplingTerm_le` (`= (b_c − m_c·returnEnd)/m`, so `≤ (b_c + m_c·returnEnd)/m`), ★ `abs_clockRate_sub_handRate_le` / `abs_clockFreq_sub_handFreq_le` (**the clock's constants move by at most `|r_h||g_c f_c(θ₂ᴬ)|/(2J_c)` and `|g_c f_c(θ₂ᴬ)|/(2ħ)`**), ★★ `coupledBudget_of_g_eq_zero` (**zero coupling gives #202's budget exactly**); ★★★ `coupled_clock_and_echo` (15 conjuncts). Residue: the generation on the product arena, rowed ~~#225~~ (done 2026-10-29, `RecordLayer/ClockEchoHamiltonian.lean`). ▸ *Original row:* In ~~#202~~ the hand clock and the echo sit on a product arena and do not interact; the clock is colocated with `A` by its site label alone. Couple them: the clock's carrier detuned by the echo's pointer at `A` (one more back-reaction term `g_c n_{A,c} f_c(θ₂ᴬ)` in the joint Hamiltonian), so that reading the clock disturbs the echo's register and the budget acquires a term for it. Deliverables: the coupled dynamics as a skew product (the conserved angles still conserved), Liouville, the three echo writes and the ticks still `ActualOccurrence`s with the coupling's effect on each window bounded, and the generation on the product arena by the two-torus machinery of `LF4/EchoCoupledFlow.lean` run once more. ⚠️ **Stuck on:** nothing identified; the price is the bounds on the shifted windows. | ~~**L**~~ done (L; generation → #225) | Medium — closes the one honest gap in the deliverable | ~~#202~~; ~~#199~~; ~~#201~~; `LF4/EchoCoupledFlow.lean` | > | ~~225~~ | Records | Records→spacetime | *new (2026-10-29, residue of ~~#212~~)* **Generate the coupled clock-and-echo flow.** ▸ **DONE 2026-10-29 (M–L took M).** [`LF4/CrossCoupledFlow.lean`](../CsdLean4/LF4/CrossCoupledFlow.lean) (generic, 22 pins) + [`RecordLayer/ClockEchoHamiltonian.lean`](../CsdLean4/RecordLayer/ClockEchoHamiltonian.lean) (9 pins): `IsHamiltonianVectorField.prodForm_fst` / `prodForm_snd` (**a Hamiltonian field on either factor lifts to `π₁^* α + π₂^* β`**, the general form of `isHamiltonianVectorField_fstLift`); ★ `crossForm_isSymplectic` (**`(ℂℙᴺ × T² × T²) × (ℂℙᴹ × T²)` is symplectic**); `crossView`, `crossEnergy` (the echo's two-torus energy plus the second ray's read on the cross view), `crossCouplingTerm`, `crossEnergy_eq`; ★★ `isHamiltonianVectorField_crossCouplingTerm` (**the cross term's field**: `f_c(θ₂) X_{B_{c,A}}` on the second ray, `(−2⟨B_{c,A}⟩ f_c'(θ₂), 0)` on the first factor's first torus — the factorwise pairing three levels deep; the wall was the price only), ★★ `isHamiltonianVectorField_cross`, `crossField_eq_hamiltonianVectorField` (**`ι_X ω = d𝓗` for the sum**), ★★ `isMIntegralCurve_crossCurve`, ★★★ `hamiltonianFlow_crossEnergy` (**the flow in closed form**: `θ₁` advanced by its own ray's accumulation plus the second ray's); on the device: `coupledToSigma`, ★★ `coupledEnergySigma_eq` (**the Hamiltonian is `−(2/ħ)` times #212's joint energy, cross term included**), ★★★ `hamiltonianFlow_coupledFlow` (**#212's `coupledFlow` is the Hamiltonian flow of its energy**), ★ `map_hamiltonianFlow_coupledFlow` (Liouville inherited); ★★★ `coupled_hamiltonian` (4 conjuncts, as #202's `echo_hamiltonian`). No residue row. ▸ *Original row:* ~~#212~~ builds the dynamics the coupled energy `𝓗 = 𝓗_echo + 𝓗_clock`, cross term `−2⟨B_c⟩(x_c) f_c(θ₂ᴬ)` included, prescribes, and proves it a two-time, measure-preserving, energy-conserving flow; it does not prove that flow is the Hamiltonian flow of `𝓗` on `(ℂℙ³ × T² × T²) × (ℂℙ³ × T²)`. Deliverables: the product symplectic form (`IsSymplectic.prodForm` once more), the cross term's vector field `(f_c(θ₂ᴬ) X_{B_c}` on the clock's ray, `−2⟨B_c⟩ f_c'(θ₂ᴬ)` on `θ₁ᴬ)`, `ι_X ω = d𝓗` for the sum, and `hamiltonianFlow = coupledFlow` by uniqueness of integral curves. ⚠️ **Stuck on:** nothing identified; the price is that the cross term couples a ray in one product factor to a torus angle in the other, a placement `isHamiltonianVectorField_coupledTerm` (ray and torus adjacent) and `LF4/EchoCoupledFlow.lean` (one ray, tori appended) do not cover — the factorwise-pairing argument run on a five-factor product. | ~~**M–L**~~ done (M) | Medium — the coupled device generated, as #202 generated the uncoupled one | ~~#212~~; ~~#202~~; ~~#195~~; `LF4/EchoCoupledFlow.lean`; `LF4/ArenaCoupledFlow.lean`; `RecordLayer/ClockEchoCoupling.lean` | > | ~~213~~ | Foundations | Records→spacetime | *new (2026-10-27, residue of #205's first brick)* **A local Lieb–Robinson velocity.** ▸ **DONE 2026-10-27 (L as priced; the stuck-on was the price, not a wall).** [`CV/LocalLiebRobinson.lean`](../CsdLean4/CV/LocalLiebRobinson.lean) (23 pins): `edgesTouching`, `DegreeLE`, `card_edgesTouching_le`, `disjoint_edge_of_mem_edgesTouching`; ★★ `norm_commutator_step_le` (**one step of the iteration**: the terms away from the probe only conjugate, so the commutator obeys a derivative comparison against `2‖B‖ · #(touching) · φ`); ★★★ `norm_commutator_local_le` (**the local bound** `‖[A(t), B]‖ ≤ 2‖A‖‖B‖ max(|Y|,1) (4Δγt)^d/d!` for individually skew edge terms of norm `≤ γ` on a graph of degree `≤ Δ`, by induction over edge paths — no Taylor remainder, no whole-generator norm); `localLRVelocity Δ γ = 8eΔγ`, ★★ `localRate_pow_div_factorial_le` (`2^{-d}` outside the cone); ★ `arenaObs_kick_sub_le_commutator`, ★★★ `lightcone_local`, ★★★ `record_lightcone_local`, ★★★ `record_influence_le_of_separated_local` (the three consumers re-derived at the local rate); on the chain `degreeLE_pathGraph`, ★★★ `norm_commutator_pathFlow_le`, ★★★ `localLRVelocity_chain_mul_hopTime` (**`v_loc · τ_hop = 8πe` for every length**), ★★★ `localLRVelocity_lt_lrVelocity_pathFlow` (beyond eight edges the local cone is the narrower one); ★★★ `local_lieb_robinson` (4 conjuncts). Honest: each edge term must be skew on its own (`FieldStructuredFlow` asks only the sum; `edgeFlow` and `pathFlow` satisfy it, carried as a hypothesis); the constant `8e` is not optimised; the old whole-norm theorems are left in place, not replaced. ▸ *Original row:* `lrVelocity F = 4e‖Σ_e h_e‖` (`CV/TwoCones.lean`, from `norm_commutator_velocity_le` in `CV/LiebRobinson.lean`) uses the norm of the whole generator, and `pathFlow_lrVelocity_mul_hopTime_bounds` proves the consequence: on a chain of `n` edges the slope is between `n/2` and `n` times a single stage's, so the cone grows with the system rather than being a property of the couplings. Deliverable: a Lieb–Robinson bound whose velocity is built from `max_e ‖h_e‖` and the coupling graph's maximum degree (the standard path-counting form: nested commutators summed over edge paths), restated on `FieldStructuredFlow`, with `pathFlow`'s velocity then bounded independently of `n`; `record_lightcone` and `record_influence_le_of_separated` re-derived from it. ⚠️ **Stuck on:** the corpus's proof bounds the nested commutators through the global norm in one step; the price is redoing that induction as a sum over edge paths. | ~~**L**~~ done (L) | High — without it the corpus's causal cone depends on how the system is described | #205; ~~#98~~; `CV/LiebRobinson.lean`; `CV/TwoCones.lean`; `CV/ConeComparison.lean` | > | ~~206~~ | QIT | Reconstruction frontier · Reviewer-proofing | *new (2026-10-24, the QM-gap survey against HEAD `9ce090e8` — edges of ordinary QM with a Lean shape that no row carried)* **Identical particles at finite dimension — IP-1's first brick.** ▸ **DONE 2026-10-28 (M as priced).** Finding first: half of the two-particle core was already in the corpus as SL-T8 (`SigmaLayer/Symmetrization.lean`, 2026-07-18: `swap`, the projectors `½(1 ± swap)`, the `±1` dichotomy, `antisymProj (v ⊗ v) = 0`), and the HOM dip as `Empirical/QM/HongOuMandel.lean` — the survey that rowed this, and `future-work.md` IP-1, had missed both; the brick builds on them. [`SigmaLayer/IdenticalParticles.lean`](../CsdLean4/SigmaLayer/IdenticalParticles.lean) (51 pins): `bosonSector`/`fermionSector` as the `±1` eigenspaces of exchange, ★ `isCompl_bosonSector_fermionSector`, ★ `fermionSector_eq_orthogonal`; ★ `trace_swap` (`= N`), ★★ `finrank_bosonSector` (`(N+1).choose 2`), ★★ `finrank_fermionSector` (`N.choose 2`) by the trace of the projectors, and ★★ `finrank_bosonSector_eq_card` / ★★ `finrank_fermionSector_eq_card` (**one dimension per occupation configuration**: `P : Fin N → ℕ` with `∑ P = 2` for bosons, two occupied modes for fermions); ★★ `apply_diag_eq_zero_of_mem_fermionSector` (**Pauli exclusion for every fermionic state**, not only products), ★★ `wedge_eq_zero_iff` (**`v ⊗ w − w ⊗ v = 0` exactly when `v`, `w` are linearly dependent**), ★ `norm_sq_wedge` / ★ `norm_sq_symState` (`2(‖v‖²‖w‖² ∓ |⟨v,w⟩|²)`); ★★ `bosonic_enhancement` (same-mode probability **twice** the distinguishable one for orthonormal states), `fermionic_same_mode`; on the composite arena of `TensorSolved.lean`: `exchangeMatrix` (`toEuclideanLin_exchangeMatrix`), ★ `exchangeMatrix_mul_kronecker`, ★ `exchangeMatrix_mul_aliceOp` (exchange carries Alice's algebra onto Bob's), `IsExchangeSymmetric` with `isExchangeSymmetric_kronecker_self`, `isExchangeSymmetric_oneBody`, `IsExchangeSymmetric.exp_smul`; ★★ `IsExchangeSymmetric.mem_bosonSector` / `mem_fermionSector` (**the sectors are invariant**), ★★ `IsExchangeSymmetric.inner_eq_zero` (**superselection**), ★★ `IsExchangeSymmetric.inner_add_add` (no interference between sectors); ★★★ `identical_particles` (7 conjuncts). Honest: the symmetrisation postulate is shown **consistent** (a conserved, unobservable sector label), not derived; two particles only. Residue rowed as ~~#214~~ (`n`-particle sectors, done 2026-10-28), ~~#215~~ (second quantisation, done 2026-10-28; operator dictionary ~~#220~~ fermions, ~~#221~~ bosons, ~~#222~~ one-body, ~~#223~~ two-body done), ~~#216~~ (exchange statistics, done 2026-10-28; spin–statistics proper → #218), ~~#217~~ (Slater converse, done 2026-10-28). ▸ *Original row:* The one textbook pillar deliberately absent (`future-work.md` IP-1): the symmetric and antisymmetric sectors of `ℂᴺ ⊗ ℂᴺ` as subspaces, Pauli exclusion as a theorem about the antisymmetric sector (no two-particle antisymmetric state with both particles in one mode), bosonic enhancement for two particles, the symmetrisation postulate stated as a sector restriction on the composite arena of `SigmaLayer/TensorSolved.lean`; the finite second quantisation on the occupation lattice the CV track already uses (`FieldConfig K N`) as the `n`-particle generalisation. ⚠️ **Stuck on:** nothing for two particles; the `n`-particle sectors and spin-statistics proper are the pillar's **XL** remainder, priced when the brick lands. | ~~**M**~~ done (M) (brick); remainder #214–#216 | High — the largest named gap in "CSD covers ordinary QM" | `future-work.md` IP-1; `SigmaLayer/TensorSolved.lean`; `CV/FieldModes.lean` | > | ~~214~~ | QIT | Reconstruction frontier · Reviewer-proofing | *new (2026-10-28, residue of ~~#206~~)* **The `n`-particle sectors.** ▸ **DONE 2026-10-28 (M–L as priced; the stuck-on was not Burnside after all — the dimensions came from sorting, not from a trace).** [`SigmaLayer/ManyIdenticalParticles.lean`](../CsdLean4/SigmaLayer/ManyIdenticalParticles.lean) (69 pins): `ManyParticle n N = EuclideanSpace ℂ (Fin n → Fin N)`, `permMatrix`/`permOp` (`(permOp σ x) c = x (c ∘ σ)`, a unitary representation: `permOp_mul`, `inner_permOp`), `permSign`; `nBosonSector`, `nFermionSector` with pointwise forms; `symmetrizer`/`antisymmetrizer` and ★ `isProj_symmetrizer` / ★ `isProj_antisymmetrizer`; ★★ `apply_eq_zero_of_not_injective` (**Pauli for `n`**: zero on every configuration with a repeated mode); ★★ `nBosonEquiv` / ★★ `nFermionEquiv` (a boson is its values on the sorted configurations, a fermion on the strictly increasing ones — `Tuple.sort`, `Tuple.comp_perm_comp_sort_eq_comp_sort`), `strictConfigEquivFinset` (strictly increasing tuples = `n`-subsets, `Finset.orderEmbOfFin`), `monoStrictEquiv` (**stars and bars**: monotone `Fin n → Fin N` = strictly increasing `Fin n → Fin (N + n − 1)`), ★★★ `finrank_nBosonSector` (`(N + n − 1).choose n`), ★★★ `finrank_nFermionSector` (`N.choose n`), ★★ `finrank_nBosonSector_eq_card` / ★★ `finrank_nFermionSector_eq_card` (one dimension per occupation configuration); ★ `nBosonSector_orthogonal` (`n ≥ 2`), ★ `sectors_not_exhaustive` (`n = 3`, `N = 2`: `4 + 0 < 8` — mixed-symmetry components exist); `slater`, ★★ `slater_mem_nFermionSector`, ★★ `slater_eq_factorial_smul_antisymmetrizer` (**the Slater determinant is the antisymmetrised product**, up to `n!`), ★★ `slater_eq_zero_of_not_linearIndependent`; `IsPermSymmetric`, ★ `isPermSymmetric_tensorPow`, `IsPermSymmetric.exp_smul`, ★★ `IsPermSymmetric.mem_nBosonSector` / `mem_nFermionSector` (**sector invariance**), ★★ `IsPermSymmetric.inner_eq_zero` (**superselection**); ★★★ `many_identical_particles` (7 conjuncts). Honest: consistency, not derivation (#216); the mixed-symmetry sectors are not studied; the converse "independent orbitals give a nonzero Slater determinant" is rowed as ~~#217~~ (done 2026-10-28, `slater_eq_zero_iff`). ▸ *Original row:* ~~#206~~ did two particles. For `n`: the permutation action of `Equiv.Perm (Fin n)` on `EuclideanSpace ℂ (Fin n → Fin N)`, the bosonic sector (fixed by every permutation) and the fermionic sector (each permutation acts by its sign) as subspaces, their projectors `(1/n!) ∑ σ` and `(1/n!) ∑ sign σ · σ`, dimensions `(N + n − 1).choose n` and `N.choose n` matched to the occupation counts (`Sym.equivNatSumOfFintype`, `Fintype.card_finset_len`, as in #206), Pauli exclusion for `n` (zero amplitude on any configuration with a repeated mode), the Slater determinant as the antisymmetrised product, and the sector restriction for permutation-symmetric operators. ⚠️ **Stuck on:** nothing identified; the price is the trace of the averaged permutation action (the two-particle trace was one fixed-point count, here it is Burnside's lemma). | ~~**M–L**~~ done (M–L) | High — the pillar's general case | ~~#206~~; `SigmaLayer/IdenticalParticles.lean`; `future-work.md` IP-1 | > | ~~215~~ | QIT | One QM, many formulations | *new (2026-10-28, residue of ~~#206~~)* **Second quantisation on the occupation lattice.** ▸ **DONE 2026-10-28 (L as priced).** Two modules, 90 pins. [`SigmaLayer/FermionFock.lean`](../CsdLean4/SigmaLayer/FermionFock.lean): `supportedOn` / `supportedOnEquiv` and ★ `isInternal_supportedOn_fiber` (the fibres of any map split the space orthogonally); `jwSign` (parity of the occupied modes before `j`), `fAnnihilate` / `fCreate` on `FieldSpace N 2`, ★ `inner_fCreate` (creation is the adjoint); ★★★ **the CAR exactly**: `fAnnihilate_mul_fCreate_add` (`{a_i, a_j†} = δ_ij`), `fAnnihilate_mul_fAnnihilate_add`, `fCreate_mul_fCreate_add`; ★★ `fAnnihilate_eq_jwString_mul_qubitLower` (**Jordan–Wigner**, `a_j = Z_0 ⋯ Z_{j−1} σ⁻_j` with the qubit chain's single-site `qubitZ`, `qubitLower`); `fNumber`, `fTotalNumber`, `fockSector` (`mem_fockSector_iff`: the number eigenspace), ★★ `isInternal_fockSector` (**Fock = ⊕ n-particle sectors**), ★★★ `fockFermionEquiv` (**the `n`-particle Fock sector is `nFermionSector n N`**, canonically: `fockFermionEquiv_apply`, amplitude on a strictly increasing configuration = Fock amplitude on its occupied set), `finrank_fockSector` (`N.choose n`), `finrank_fermionFock` (`2^N`). [`SigmaLayer/BosonFock.lean`](../CsdLean4/SigmaLayer/BosonFock.lean): `modeOp A k` (a single-mode matrix on mode `k`: `modeOp_mul`, `modeOp_one`, ★ `modeOp_mul_comm` across modes, ★ `inner_modeOp_conjTranspose`), `bAnnihilate` / `bCreate` = the CV track's `annihilation` / `creation` on mode `k`; ★★ `bAnnihilate_mul_bCreate_sub` (`[b_k, b_k†] = 1 − L·|top⟩⟨top|_k`, the CV `truncated_ccr` on one mode), ★★ exact CCR across modes; `bosonFockSector`, ★★ `isInternal_bosonFockSector`, ★★★ `ccr_exact_on_bosonFockSector` (**exact CCR on every `n`-quantum sector with `n + 1 < L`**; `no_exact_finite_ccr` forbids it on the whole space); `monoOcc` (occupation numbers of a sorted configuration, injective by `List.Perm.eq_of_sortedLE`, onto by counting), ★★★ `bosonFockEquiv` (**for `n < L` the `n`-quantum sector is `nBosonSector n K`**), `finrank_bosonFockSector`; ★★★ `second_quantisation` (5 conjuncts). Honest: the identification is by the canonical map on amplitudes, not yet as operators — creation as (anti)symmetrised tensoring, the normalisations that make the maps isometric, and `dΓ(h)` ↔ `∑ h^{(k)}` are rowed as #220. ▸ *Original row:* The Fock space as the direct sum of the `n`-particle sectors of #214, identified with the occupation-number basis the CV track already uses (`FieldConfig K N`, `CV/FieldModes.lean`); for fermions, creation and annihilation operators satisfying the CAR **exactly** (the fermionic Fock space of `N` modes is finite, dimension `2^N`) and the Jordan–Wigner identification with a qubit chain; for bosons, the truncated ladder operators of the CV track restricted to fixed total number, where `no_exact_finite_ccr` (`CV/ApproxCCR.lean`) already says the CCR can only hold up to the cutoff. ⚠️ **Stuck on:** ~~#214~~ (done 2026-10-28) — nothing now; the price is the CAR algebra and the Fock-space identification. | ~~**L**~~ done (L) | Medium — first and second quantisation shown to be one theory at finite dimension | #214; ~~#206~~; `CV/FieldModes.lean`; `CV/ApproxCCR.lean` | > | ~~216~~ | QIT | Reviewer-proofing | *new (2026-10-28, residue of ~~#206~~)* **Exchange statistics: only bosons and fermions.** ▸ **DONE 2026-10-28 (S as priced).** [`SigmaLayer/ExchangeStatistics.lean`](../CsdLean4/SigmaLayer/ExchangeStatistics.lean) (14 pins): ★ `map_swap_eq_map_swap` (a character takes one value on all transpositions, `Equiv.Perm.isConj_swap`), ★★ `perm_char_eq_one_or_sign` (**the trivial character and the sign are the only characters of a symmetric group** with values in an integral domain; Mathlib had only the surjective-into-`ℤˣ` form, `eq_sign_of_surjective_hom`), `perm_char_eq_one_or_permSign`; ★★ `eq_one_or_eq_neg_one_of_permOp_swap` (**an exchange phase is `±1`**); `IsExchangeEigenstate` (every permutation keeps the ray), ★★★ `isExchangeEigenstate_iff` (**exchange eigenstates are exactly the bosons and the fermions**), ★★ `le_nBosonSector_or_le_nFermionSector` (a subspace on which permutations act as scalars lies in one sector), ★ `eq_zero_of_mem_nBosonSector_of_mem_nFermionSector` (`n ≥ 2`), ★ `exists_not_isExchangeEigenstate` (three particles in two modes: the premise is a restriction, not a theorem about every state); ★★★ `exchange_statistics` (6 conjuncts). Honest: ray-invariance under permutations is the premise (what "identical" means), not derived; without it the mixed-symmetry sectors remain (described in ~~#219~~, general `n` #224); anyons are excluded only because exchange is a permutation here (no braid-group arena). Spin–statistics proper rowed as #218. ▸ *Original row:* The Lean-shaped half: the only one-dimensional representations of `Equiv.Perm (Fin n)` are the trivial one and the sign (Mathlib's characterisation of `Equiv.Perm.sign`), so a sector of `n` identical particles on which permutations act by a phase is the bosonic or the fermionic one — no other one-dimensional exchange statistics at finite dimension. The other half, **spin–statistics proper** (which particles take which sign), needs Lorentz covariance and is not a finite-dimensional statement; it is recorded here as the boundary it is (the scope ladder's relativistic rung), not as a brick. ⚠️ **Stuck on:** nothing for the representation half; spin–statistics proper is outside the finite-dimensional theory. | ~~**S**~~ done (S) (statistics); spin–statistics → #218 | Medium — closes "why only two kinds" at the level QM can | ~~#206~~; #214; `future-work.md` IP-1 | > | ~~217~~ | QIT | Reviewer-proofing · Library owed / upstream | *new (2026-10-28, residue of ~~#214~~)* **Independent orbitals give a nonzero Slater determinant.** ▸ **DONE 2026-10-28 (M as priced; the stuck-on was the price).** Cauchy–Binet as a CSD-free staging module, [`Mathlib/LinearAlgebra/Matrix/CauchyBinet.lean`](../CsdLean4/Mathlib/LinearAlgebra/Matrix/CauchyBinet.lean) (8 pins): ★ `Matrix.det_mul_eq_sum_fun` (**`det (A * B)` for rectangular `A`, `B` as a sum over all functions**, the first step of Mathlib's `det_mul`), `Matrix.det_submatrix_id_eq_zero_of_not_injective`, `Tuple.injectiveEquivStrictMonoPerm` (an injective tuple is a strictly increasing one composed with a permutation, uniquely, by `Tuple.sort`) and `Tuple.sum_injective_eq_sum_strictMono_perm`, ★★ `Matrix.det_mul_eq_sum_strictMono` (**Cauchy–Binet**, the sum over strictly increasing `g : Fin k → m`, i.e. over `k`-subsets), ★ `Matrix.det_conjTranspose_mul_self_eq_sum`. In [`SigmaLayer/ManyIdenticalParticles.lean`](../CsdLean4/SigmaLayer/ManyIdenticalParticles.lean) (7 pins): `orbitalMatrix`, `gram_eq_conjTranspose_mul_orbitalMatrix`, ★★★ `norm_sq_slater` (**`‖slater φ‖² = n! · det (gram φ)`**, the `n`-particle form of ~~#206~~'s `norm_sq_wedge`), ★★ `slater_ne_zero_of_linearIndependent`, ★★ `slater_eq_zero_iff` (**zero exactly for dependent orbitals**, the `n`-particle `wedge_eq_zero_iff`), `norm_sq_slater_of_orthonormal` (`n!`, so the normalised Slater state is `slater φ / √n!`); the capstone `many_identical_particles` strengthened in place (conjunct 4 now carries the norm identity and the `↔`). Residue: none. ▸ *Original row:* ~~#214~~ proves `slater_eq_zero_of_not_linearIndependent`; the converse — `n` linearly independent orbitals in `ℂᴺ` give a Slater determinant that is not zero, so `slater φ = 0 ↔ ¬ LinearIndependent ℂ φ`, the `n`-particle form of ~~#206~~'s `wedge_eq_zero_iff` — needs a nonzero `n × n` minor of an `N × n` matrix of rank `n`. Cleanest route: the Cauchy–Binet formula `det (Aᴴ A) = ∑_{|S| = n} |det A_S|²` as a library lemma, which gives `‖slater φ‖² = n! · det (Gram φ)`, the `n`-particle form of `norm_sq_wedge`. ⚠️ **Stuck on:** nothing identified; the price is Cauchy–Binet for rectangular matrices. | ~~**M**~~ done (M) | Medium — completes the Pauli statement for `n` particles; Cauchy–Binet is library-grade on its own | ~~#214~~; ~~#206~~; `SigmaLayer/ManyIdenticalParticles.lean` | > | 218 | QIT | Reviewer-proofing | *new (2026-10-28, residue of ~~#216~~)* **Spin–statistics proper.** ~~#216~~ proves that a state whose ray every permutation keeps is a boson or a fermion; **which** particles take which sign — integer spin bosons, half-integer spin fermions — is the spin–statistics theorem, and it needs Lorentz covariance, locality (microcausality) and positivity of the energy: it is not a finite-dimensional statement. The Lean shape would be the Pauli–Burgess or Streater–Wightman argument for free fields on Minkowski space. ⚠️ **Stuck on:** the corpus has no Lorentz-covariant field theory (the scope ladder's relativistic rung; CV/ and the EFT layer are non-relativistic or effective). | **XL** | Medium — the last piece of "why these two, for these particles" | ~~#216~~; ~~#214~~; `future-work.md` IP-1 | > | ~~219~~ | QIT | Reviewer-proofing | *new (2026-10-28, residue of ~~#214~~ and ~~#216~~)* **The mixed-symmetry sectors (parastatistics).** ▸ **DONE 2026-10-28 (M as priced, `n = 3` plus the general-`n` frame).** [`SigmaLayer/MixedSymmetry.lean`](../CsdLean4/SigmaLayer/MixedSymmetry.lean) (45 pins): `inner_symmetrizer_left` / `inner_antisymmetrizer_left` (the projectors are self-adjoint), `ker_symmetrizer` / `ker_antisymmetrizer`; `mixedSector n N`, ★★ `mixedSector_eq_orthogonal` (**the orthogonal complement of bosons + fermions**), ★★ `finrank_mixedSector_add` (**`N^n = C(N+n−1,n) + C(N,n) + dim`**, every `n ≥ 2`), ★★ `map_mem_mixedSector` (**every permutation-symmetric operator keeps it**), `eq_zero_of_mem_mixedSector_of_isExchangeEigenstate`; for three particles ★★ `mem_mixedSector_three_iff` (**mixed ⟺ the cyclic average vanishes**), `sum_swaps_eq_zero_of_mem_mixedSector` (the `S₃` character `0` / `−1` on the sector), `mixedEven` / `mixedOdd` (even/odd under `P₀₁`, direct sum), ★★ `mixedIntertwiner_mixedIntertwiner` (`(P₀₂ − P₁₂)² = 3` there), ★★★ `mixedEvenEquivOdd` (**the two copies are isomorphic**), ★★★ `three_mul_finrank_mixedSector_three` (**`dim = 2N(N² − 1)/3`**), ★★ `three_mul_finrank_mixedEven` (each copy `N(N² − 1)/3`); ★★★ `mixed_symmetry` (5 conjuncts). Residue: the general isotypic decomposition (Young symmetrisers), rowed #224. ▸ *Original row:* ~~#214~~'s `sectors_not_exhaustive` shows that for `n ≥ 3` the boson and fermion sectors do not fill the space; ~~#216~~ shows the remainder carries no state whose ray every permutation keeps. What is left: describe the remainder. For `n = 3`: it has dimension `2N(N² − 1)/3`, two copies of `S₃`'s two-dimensional representation (`N³ = C(N+2,3) + C(N,3) + 2 · N(N² − 1)/3`), and permutation-symmetric dynamics keeps it; in general the isotypic decomposition of `(ℂᴺ)^{⊗n}` under `S_n` (Schur–Weyl). ⚠️ **Stuck on:** for `n = 3` nothing (explicit projectors); in general Mathlib has no Specht modules or Young symmetrisers. | ~~**M**~~ done (M) (`n = 3`); general → #224 | Low–medium — completes the picture of what the symmetrisation postulate excludes | ~~#214~~; ~~#216~~; `SigmaLayer/ManyIdenticalParticles.lean` | > | ~~220~~ | QIT | One QM, many formulations | *new (2026-10-28, residue of ~~#215~~)* **First and second quantisation as operators.** ▸ **SPLIT and fermion half DONE 2026-10-28.** The row's three parts had different prices: the fermion half (signs) is landed here, the boson half (multinomial counting) is #221, one-body operators `dΓ(h)` are #222. [`SigmaLayer/FermionDictionary.lean`](../CsdLean4/SigmaLayer/FermionDictionary.lean) (45 pins): `fermionFirst n N` (the normalised first-quantisation map, `1/√n!` times the Fock amplitude on the occupied set, extended antisymmetrically), ★★★ `inner_fermionFirst_of_mem` (**an isometry on each `n`-particle sector**, by `Tuple.sum_injective_eq_sum_strictMono_perm`), ★★ `fermionFirst_fockBasis` (**a Fock basis state is the normalised Slater determinant of its occupied basis orbitals**), `fCreate1` (`√(n+1)` × antisymmetrised `e_j ⊗ ·`, explicit form `fCreate1_apply`), ★★★ `fermionFirst_fCreate` (**`Φ_{n+1} a_j† = fCreate1 j Φ_n`**), `fAnnihilate1` (`√(n+1) x (j, ·)`), ★★★ `fermionFirst_fAnnihilate` (**`Φ_n a_j = fAnnihilate1 j Φ_{n+1}`**); sign bookkeeping: `splitPermEquiv` (`Fin (n+1) × Perm (Fin n) ≃ Perm (Fin (n+1))`, `σ = p.cycleRange⁻¹ · lift e`, sign `(−1)^p sign e`), ★ `jwSign_image` (the Jordan–Wigner sign of the `p`-th occupied mode of a sorted configuration is `(−1)^p`); ★★★ `fermion_dictionary` (4 conjuncts). ▸ *Original row:* ~~#215~~ identifies each `n`-particle Fock sector with the first-quantised sector of ~~#214~~ by the canonical map on amplitudes (`fockFermionEquiv`, `bosonFockEquiv`). What is left is the operator dictionary: (a) the normalisations — `√n!` for fermions, `√(n!/∏ P_m!)` for bosons — that make the maps isometric; (b) creation as tensoring: under the normalised map, `a_j†` (`fCreate`) is `√(n+1)` times antisymmetrised tensoring with the orbital `e_j`, `b_k†` the symmetrised one (so a Fock basis state is a normalised Slater determinant or permanent of basis orbitals); (c) one-body operators: `dΓ(h) = ∑ h_ij a_i† a_j` corresponds to `∑_k h^{(k)}` on every sector, the free-field Hamiltonian of `CV/FieldModes.lean` being the case `h` diagonal. ⚠️ **Stuck on:** nothing identified; the price is the sign and normalisation bookkeeping of (b). | ~~**M–L**~~ split: fermion half done (M); #221, #222 | Medium — completes "one theory, two formulations" at the level of operators | ~~#215~~; ~~#214~~; `SigmaLayer/FermionFock.lean`; `SigmaLayer/BosonFock.lean` | > | ~~221~~ | QIT | One QM, many formulations | *new (2026-10-28, split from ~~#220~~)* **The boson dictionary.** ▸ **DONE 2026-10-28 (M as priced).** [`SigmaLayer/BosonDictionary.lean`](../CsdLean4/SigmaLayer/BosonDictionary.lean) (36 pins): `occNat` (occupation numbers of a configuration: `occNat_comp_perm`, `occNat_cons`, `occNat_succAbove`, `occNat_eq_card`), `monotone_eq_of_occNat_eq`, `exists_perm_of_occNat_eq`; ★★ `card_comp_eq` (**the coset count** `#{σ | m ∘ σ = c} = ∏ P_k!` or `0`, by `DomMulAct.stabilizer_card`), ★★ `card_fibre_mul_prod_factorial` (**configurations with occupation numbers `P` number `n!/∏ P_k!`**); `bosonFirst hL` (`√(∏ P_k!/n!)` × the Fock amplitude on the occupation numbers; `bosonFirst_mem`), ★★★ `inner_bosonFirst_of_mem` (**an isometry on the `n`-quantum sector**), ★★ `bosonFirst_bosonFockBasis` (**a Fock basis state is the normalised symmetrised product of basis orbitals**), `bCreate1` (`√(n+1)` × symmetrised `e_k ⊗ ·`, `bCreate1_apply`), ★★★ `bosonFirst_bCreate` (**`Φ_{n+1} b_k† = bCreate1 k Φ_n`**), `bAnnihilate1`, ★★★ `bosonFirst_bAnnihilate` (**`Φ_n b_k = bAnnihilate1 k Φ_{n+1}`**), all below the cutoff (`n + 1 < L`); ★★★ `boson_dictionary` (5 conjuncts). ▸ *Original row:* The boson half of ~~#220~~: the first-quantisation map `bosonFirst n` (`√(∏ P_m!/n!)` times the Fock amplitude on the occupation numbers `P` of a configuration), an isometry on each `n`-quantum sector below the cutoff; creation `b_k†` as `√(n+1)` times symmetrised tensoring with `e_k`; annihilation as `√(n+1)` times evaluation at `(k, ·)`; a Fock basis state as a normalised permanent. The counting step: the number of configurations with occupation numbers `P` is `n!/∏ P_m!`, by orbit–stabiliser with Mathlib's `DomMulAct.stabilizer_card`. ⚠️ **Stuck on:** nothing identified; the price is the multinomial fibre count. | ~~**M**~~ done (M) | Medium — the boson half of "one theory, two formulations" | ~~#220~~; ~~#215~~; `SigmaLayer/BosonFock.lean`; `SigmaLayer/FermionDictionary.lean` | > | ~~222~~ | QIT | One QM, many formulations | *new (2026-10-28, split from ~~#220~~)* **One-body operators in the two pictures.** ▸ **DONE 2026-10-28 (M as priced).** [`SigmaLayer/OneBody.lean`](../CsdLean4/SigmaLayer/OneBody.lean) (15 pins): `oneBody h` (`∑_k h^{(k)}`), `cons_comp_succAbove_eq` (re-inserting a particle in slot `p` = updating slot `p`, then rotating it to the front), ★★ `sum_fCreate1_fAnnihilate1` / ★★ `sum_bCreate1_bAnnihilate1` (**the first-quantised identity** `∑_{ij} h_ij Create1_i Annihilate1_j = ∑_k h^{(k)}` on antisymmetric / symmetric states; the slot signs cancel in pairs), `dGammaF`, `dGammaB`, ★★★ `fermionFirst_dGammaF` (**`Φ_n dΓ(h) = (∑_k h^{(k)}) Φ_n`, every `n`**), ★★★ `bosonFirst_dGammaB` (the same for bosons below the cutoff), `dGammaF_one` / `dGammaB_one` (`dΓ(1)` = total number), ★★ `toEuclideanLin_fieldHamiltonian` (**the CV free-field Hamiltonian is `dΓ(1) + K/2`**); ★★★ `one_body` (4 conjuncts). Residue: two-body interactions, rowed #223. ▸ *Original row:* `dΓ(h) = ∑ h_ij a_i† a_j` on the Fock space corresponds to `∑_k h^{(k)}` (the same one-particle operator on every particle) on each `n`-particle sector, for fermions (from ~~#220~~'s `fermionFirst_fCreate` / `fermionFirst_fAnnihilate` and the first-quantised identity `∑ h_ij fCreate1 i (fAnnihilate1 j x) = ∑_k h^{(k)} x` on antisymmetric `x`) and for bosons (after #221); the free-field Hamiltonian of `CV/FieldModes.lean` is the case `h` diagonal. ⚠️ **Stuck on:** nothing identified — ~~#221~~ done 2026-10-28, both halves of the dictionary are in place. | ~~**M**~~ done (M) | Medium — the dynamics, not only the states, are one theory | ~~#220~~; #221; `SigmaLayer/FermionDictionary.lean` | > | ~~223~~ | QIT | One QM, many formulations | *new (2026-10-28, residue of ~~#222~~)* **Two-body interactions in the two pictures.** ▸ **DONE 2026-10-28 (M as priced).** [`SigmaLayer/TwoBody.lean`](../CsdLean4/SigmaLayer/TwoBody.lean) (19 pins): `twoBody V` (`½ ∑_{p≠q} V^{(pq)}`, `V : Matrix (Fin N × Fin N) (Fin N × Fin N) ℂ`), ★★ `oneBody_single_oneBody_single_apply` / ★★ `twoBody_eq_sum_oneBody` (**the pair operator is the hop polynomial `½ ∑ V_ab (E_{a₁b₁} E_{a₂b₂} − δ_{a₂b₁} E_{a₁b₂})` on every state**: the `p = q` terms of the double hop are the correction), `twoBody_one` (the unit interaction counts pairs, `n choose 2`); ★★ `fCreate_mul_fCreate_mul_fAnnihilate_mul_fAnnihilate` (**normal ordering `a_i† a_j† a_l a_k = a_i† a_k a_j† a_l − δ_jk a_i† a_l`**, from the exact CAR), ★★ `bCreate_mul_bCreate_mul_bAnnihilate_mul_bAnnihilate` (the same for truncated bosons, `δ_jk` carrying `1 − L P_k`), `dGamma2F_one` (`½ N̂(N̂ − 1)`), `bosonFirst_bCreate_topProj` (the top-level correction is invisible below the cutoff); ★★★ `fermionFirst_dGamma2F` (**`Φ_n W = (½ ∑_{p≠q} V^{(pq)}) Φ_n`, every `n`**), ★★★ `bosonFirst_dGamma2B` (the same for bosons below the cutoff); ★★★ `two_body` (4 conjuncts). **Stuck-on was the price, and the price fell:** normal ordering reduces the two-slot bookkeeping to ~~#222~~'s one-slot dictionary, no new sign work. ▸ *Original row:* ~~#222~~ matches one-body operators. The pair interaction `½ ∑ V_ijkl a_i† a_j† a_l a_k` on the Fock space should correspond to `½ ∑_{k ≠ k'} V^{(k k')}` on every `n`-particle sector, for fermions and for bosons below the cutoff: two applications of the creation/annihilation dictionary (~~#220~~, ~~#221~~) and a first-quantised identity with two slots removed and re-inserted. ⚠️ **Stuck on:** nothing identified; the price is the two-slot rotation bookkeeping. | ~~**M**~~ done (M) | Medium — interacting dynamics are one theory too | ~~#222~~; ~~#220~~; ~~#221~~; `SigmaLayer/OneBody.lean` | > | 224 | QIT | Reviewer-proofing | *new (2026-10-28, residue of ~~#219~~)* **The isotypic decomposition of `(ℂᴺ)^{⊗n}` for every `n`.** ~~#219~~ proves the mixed sector is the orthogonal complement of bosons + fermions with dimension `N^n − C(N+n−1,n) − C(N,n)` for every `n`, and splits it into two isomorphic halves for `n = 3`. For `n ≥ 4` it splits further by the irreducible representations of `S_n` (Young diagrams), with Schur–Weyl multiplicities (the dimensions of the `GL_N` Schur modules). First brick: `n = 4` by explicit Young symmetrisers for the shapes `(3,1)`, `(2,2)`, `(2,1,1)` with isotypic dimensions `3 · N(N+1)(N+2)(N−1)/8`, `2 · N²(N²−1)/12`, `3 · N(N+1)(N−1)(N−2)/8` (standard tableaux × hook-content), which with `C(N+3,4) + C(N,4)` must sum to `N⁴` (at `N = 2`: `5 + 9 + 2 + 0 + 0 = 16`). ⚠️ **Stuck on:** for `n = 4`, nothing but the bookkeeping of 24 permutations and three shapes; in general, Mathlib has no Specht modules, Young symmetrisers or Schur–Weyl duality. | **L** (`n = 4`), **XL** (general) | Low — completes the textbook picture; the symmetrisation postulate already excludes all of it | ~~#219~~; `SigmaLayer/MixedSymmetry.lean` | > | ~~207~~ | QIT | QC chain (link 12) | *new (2026-10-24, the QM-gap survey against HEAD `9ce090e8` — edges of ordinary QM with a Lean shape that no row carried)* **The gadget-set premise for a concrete code.** ▸ **DONE 2026-10-29 at level one (L took M).** [`Empirical/QM/QEC/SteaneMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteaneMachine.lean) (31 pins): ★ `logicalX_mul_steaneProj` / `logicalZ_mul_steaneProj` (**`X̄`, `Z̄` commute with the code projector** — `SteaneCircuit.lean`'s open item); `Gadget` (**the gadget set: transversal `H̄`, `X̄`, `Z̄`**), ★★ `Gadget.isCodeState_ideal`; `card_le_one_of_isBad_eq_false`, `exists_single_of_card_le_one` (**a level-one pattern not `isBad` is one fault at one location**); `weightOne`, ★ `isRecovery_weightOne`, ★★ `Gadget.propagatesD` (**transversality discharged**), ★★ `Gadget.deviatesBy_faulty` (**the one-fault bound discharged**); ★★★ `steane_gadgetSet_corrected` (**every good step is an `IsCorrectedStep`**, one recovery for all, by #96's `isCorrectedStep_rect`), `steane_gadgetSet_correctedAtLevel_zero`; ★★★ `steane_exRec` (**#96's extended rectangle at the Steane code**: faulty leading recovery or faulty gadget corrected, faulty trailing recovery handed on); ★★★ `steane_machine` (**`hgood` discharged** through `correctedRun_eq_idealRun_of_level`, and the threshold statement: output exactly ideal with probability `≥ 1 − 21 N p²`). Scope: level one, the machine's recovery ideal (a faulty recovery enters only as `DeviatesBy` in `steane_exRec`), one block and three gates, independent unitary faults. Residue rowed #226–#230. ▸ *Original row:* Link 12 has the threshold theorem in abstract form and the level reduction (~~#97~~, `LevelReduction.lean`); what the chain does not yet contain is a fault-tolerant machine, because the gadget-set premise is not discharged for any concrete code. Discharge it for the Steane code: the transversal gadgets of the stabiliser formalism already in the corpus, the recovery already landed, assembled into the `IsCorrectedStep` hypothesis the level reduction consumes. ⚠️ **Stuck on:** nothing identified; the price is assembling existing pieces into the exact premise shape. | ~~**L**~~ done (M, level one; residue #226–#230) | High — closes the postulates→machine chain | ~~#97~~; `Mathlib/QuantumInfo/LevelReduction.lean`; `docs/FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md` link 12 | > | ~~226~~ | QIT | QC chain (link 12) | *new (2026-10-29, residue of ~~#207~~)* **The two-block Steane machine: `CNOT̄` in the gadget set.** ▸ **DONE 2026-11-05 (M–L took M).** [`Empirical/QM/QEC/SteanePairMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteanePairMachine.lean) (25 pins): `pairProj` (**the two-block code projector**, `= pairEnc pairEncᴴ`); `blockFault`, ★★ `exists_pairRecovery` (**one channel undoes every unitary fault of weight at most one per block** — the Knill–Laflamme scalar fixed to `1` on the projector, a code state of nonzero trace), `pairWeightOne`, ★ `exists_isRecovery_pairWeightOne`; `PairGadget` (**the gadget set: `H̄`, `X̄`, `Z̄` on either block, and `CNOT̄`**), ★★ `PairGadget.isCodeState_ideal` (for `CNOT̄` from `cnotT_mul_pairEnc`); `faultOp` (a location's fault: a unitary on qubit `q` of each block), ★★ `PairGadget.deviatesBy_faulty` (**the one-fault bound for all four gadgets**); ★★ `PairGadget.propagatesD_single` (transversality for the single-block gadgets); ★★★ `steane_pairGadgetSet_corrected` (**every good step of the four-gadget set is an `IsCorrectedStep`**, one recovery); ★★★ `steane_pair_machine` (**the Steane machine over `H̄`, `X̄`, `Z̄`, `CNOT̄`**: no bad gadget → exactly ideal output; probability `≥ 1 − 21 N p²`). Scope: level one, ideal recovery, product-form faults at a location (for a single-block gadget the location includes the idle partner qubit, so a gate-only fault is the special case), seven locations per gadget so #207's constant is unchanged. `PropagatesD` for `CNOT̄` is **not** stated — it copies an `X` from control to target, so the one-set form is expected to fail; the `CNOT̄` extended rectangle is row #302. ▸ *Original row:* ~~#207~~'s machine runs `H̄`, `X̄`, `Z̄` on one block; the transversal `CNOT̄` and its one-fault recovery exist (`steane_faultyCNOT_recovery`, `cnotT_mul_pairEnc`) on the two-block register `Fin 2 → Fin 7 → Fin 2`. Deliverables: the single-block gadgets lifted to the two-block register (`blockKron` over the two blocks), the two-block code projector, `IsRecovery`/`PropagatesD`/`DeviatesBy` for the pair recovery and the lifted set plus `CNOT̄`, and `steane_machine` restated for circuits over the four gadgets. ⚠️ **Stuck on:** *done 2026-11-05 — was:* nothing identified; the price is the register lift and a second correctable set (weight one per block). | ~~**M–L**~~ done (M) | Medium — the Clifford generators in one machine | ~~#207~~; `Empirical/QM/QEC/SteaneMachine.lean`; `Empirical/QM/QEC/SteaneTransversalCNOT.lean` | > | ~~227~~ | QIT | QC chain (link 12) | *new (2026-10-29, residue of ~~#207~~)* **The level-`k` Steane machine.** ▸ **DONE 2026-10-29 (L took M; the stuck-on was the encoder route, as guessed).** [`Empirical/QM/QEC/SteaneMachineConcat.lean`](../CsdLean4/Empirical/QM/QEC/SteaneMachineConcat.lean) (28 pins): `trace_steaneProj`, ★★ `steaneProj_eq` (**`P = E Eᴴ`**: `P − E Eᴴ` a Hermitian idempotent of trace 0), `cProj_one`, `mul_steaneEnc_of_comm`; ★ `logicalX_mul_steaneEnc`, `logicalZ_mul_steaneEnc`; ★★ `exists_hadTransversal_mul_steaneEnc` (**the transversal Hadamard is `c H` on the encoder, Schrödinger picture, no coset sum** — `SteaneFaultyGate.lean`'s open item: `V X = Z V`, `V Z = X V` force a multiple of `H`); `cTrans`, ★★★ `Gadget.exists_cTrans_mul_cEnc` (**every gadget a logical gate at every level**), `isCodeState_cProj_conj`; `cFaulty`, `cFaultErr`, ★★ `cFaulty_eq` (faulty = ideal then the pattern's error), `patErr_cFaultErr`, `cFaultErr_unitary`; ★★ `exists_concat_recovery_exact` (#86's recovery exact on every code state, no trace hypothesis); ★★★ `steane_gadgetSet_corrected_level`, ★★★ `steane_machine_level` (**premise discharged at level `k`, bound `1 − N(21p)^{2^k}/21`**), ★★★ `steane_machine_threshold` (**below `p < 1/21` the bound tends to 1** — the threshold theorem for the Steane machine). Scope: one ideal recovery per gadget at level `k`, not the recursive simulation with interleaved lower-level recoveries (rowed #231); one block, three gates; independent unitary leaf faults. ▸ *Original row:* ~~#207~~'s pattern is `ConcatPat 7 1`, the seven locations of one transversal gadget. At level `k` the gadgets act on the concatenated code `cProj k` (`SteaneConcat.lean`, `7^k` qubits) and the pattern is `ConcatPat 7 k`. Deliverables: transversal `H̄`, `X̄`, `Z̄` on `cLabel k` preserving `cProj k`; a good level-`k` pattern's faults as a `PatErr`-correctable error after the ideal gadget; `exists_concat_recovery_unitary` as the recovery; and `steane_machine` at every `k`, giving `1 − N·codeCapacityBound 21 p k`. ⚠️ **Stuck on:** code preservation at level `k` — `cProj k` is built as an encoder product, not as a `stabMat`, so the level-one commutation argument does not transfer verbatim; the encoder form (`cEnc`) is the likely route. | ~~**L**~~ done (M; residue #231) | High — the threshold theorem's levels at a concrete code | ~~#207~~; ~~#86~~; `Empirical/QM/QEC/SteaneConcat.lean`; `Mathlib/QuantumInfo/LevelReduction.lean` | > | ~~228~~ | QIT | QC chain (link 12) | *new (2026-10-29, residue of ~~#207~~)* **The Steane recovery as a circuit.** ▸ **DONE 2026-10-29 (M–L took M).** [`Empirical/QM/QEC/SteaneCircuitRecovery.lean`](../CsdLean4/Empirical/QM/QEC/SteaneCircuitRecovery.lean) (41 pins): `ancRead`, ★★ `ancRead_mul_extractMat_mul_ancInit` (**prepare, extract, read = the syndrome projector**, generic); `bitProj`, `phaseProj` (**the `Z`-checks as the ladder conjugated by the transversal Hadamard**); ★ `bitProj_mul_pauliMat_mul_steaneProj` / `phaseProj_mul_pauliMat_mul_steaneProj` (on an errored code state the two syndromes are those of the error's `X` and `Z` parts); `decode`, ★ `decode_syndrome_unitErr` (**the decoder names the faulty qubit**), `syndromePair_injective`; `recKraus` (**one branch as gates**), ★★ `recKraus_eq`, `recKraus_tp`, ★★★ `circuitRecovery` (**the channel**); ★★★ `circuitRecovery_cross` (**Knill–Laflamme for every pair of single-qubit Paulis**), ★★★ `circuitRecovery_unitary` (every unitary single-qubit error undone exactly), ★★ `isRecovery_circuitRecovery`; ★★★ `steane_machine_circuit` (**#207's machine on the gate-built recovery**). `Gadget.step` / `machineCircuit` generalised to any Kraus index type. Scope: fault-free recovery, one unverified ancilla per syndrome, one round, level one (the level-`k` circuit recovery is what #231 builds). Unblocks #229. No new row. ▸ *Original row:* The machine's recovery is the channel of `steane_recovery_unitary_qubit`; the corpus has `X`-syndrome extraction as a circuit (`steaneExtract_reads_syndrome`, #94) but not the `Z`-checks, the decoder, or the correction. Deliverables: `Z`-check extraction in the conjugate basis, the syndrome-to-correction decoder for weight-one errors, and the composite circuit equal to the recovery channel on the relevant inputs. ⚠️ **Stuck on:** nothing identified; the two bases and the decoder are new code, the identities are the ones `SteaneSyndromeCircuit.lean` already proves for one basis. | ~~**M–L**~~ done (M) | Medium — the recovery the machine uses, built from gates | ~~#207~~; ~~#94~~; `Empirical/QM/QEC/SteaneSyndromeCircuit.lean`; `Empirical/QM/QEC/SteaneArbitrary.lean` | > | ~~229~~ | QIT | QC chain (link 12) | *new (2026-10-29, residue of ~~#207~~)* **Faulty recovery locations in the machine.** ▸ **SPLIT 2026-10-29 — the price was the blocker (four deliverables, one of them a new circuit); first brick DONE, the rest is #232–#234.** Brick: [`Empirical/QM/QEC/SteaneMethod.lean`](../CsdLean4/Empirical/QM/QEC/SteaneMethod.lean) (48 pins): `ancPrep`, ★★ `ancRead_mul_extractMat_mul_ancPrep` (**prepare `ψ`, extract, read = the diagonal `ψ(m + Hz)`**, generic), `ancRead_mul_kronecker`; `plusAnc`, `tCNOT`, ★★ `steaneRead_eq` (**Steane's encoded `|+̄⟩` with a transversal `CNOT` reads the syndrome, not the qubit**); `bitRound`, `phaseRound`, `methodKraus_eq`, `methodRecovery`, ★★★ `methodRecovery_apply` (**Steane's method = ~~#228~~'s recovery circuit, as channels**); `AtQubit`, `faultRead_mul_steaneProj`, ★★ `syndrome_of_faultCoeff_ne_zero` (**a fault at `q` moves the read word at most at `q`**), `atQubit_bitFault`, `atQubit_phaseFault`, `AtQubit.phaseRound_mul`; ★★★ `steaneMethod_single_fault` / `_apply` (**one fault at any extraction location of Steane's method leaves a mixture of errors on one data qubit**). Ancilla preparation ideal (→ #234). ▸ *Original row:* ~~#207~~'s `steane_exRec` takes a faulty recovery as #96's `DeviatesBy` hypothesis on the piece; the pattern counts only the gadgets' seven locations. Deliverables: the recovery circuit of #228 with a verified (cat) ancilla, its fault locations added to the gadget's pattern, a proof that one fault in it leaves a weight-one deviation (#95's ladder bound against the cat state), and `steane_machine` with recovery locations counted (the constant grows with the location count). ⚠️ **Stuck on:** ~~#228 first~~ (done 2026-10-29, `circuitRecovery`); the verified-ancilla fault analysis, which #95 has for the ladder and not for verification. | ~~**L**~~ split (brick M; rest #232–#234) | High — removes the machine's ideal recovery | ~~#207~~; #228; ~~#95~~; ~~#96~~; `Mathlib/QuantumInfo/AncillaLadder.lean` | > | ~~230~~ | QIT | QC chain (link 12) | *new (2026-10-29, residue of ~~#207~~)* **A logical `T̄` for the Steane machine.** ▸ **DONE 2026-11-05 at level one, in two bricks (L took M); residue #304, #305.** **Second brick — the injection** ([`Empirical/QM/QEC/SteaneInjection.lean`](../CsdLean4/Empirical/QM/QEC/SteaneInjection.lean), 41 pins): `injChannel` (**the gadget as one channel on the two-block register**: discard block `1` through its 128 labels and prepare the encoded magic state, transversal `CNOT̄`, the decoded transversal `Z` read-out `logProj m`, feed-forward `X̄^m` on the ancilla and `S̄^m` on the data); ★★ `injCore` (**the logical heart, outcome by outcome**: each outcome applies `T` with amplitude `magicL m`, `|magicL m|² = 1/2` — on outcome `1` the `CNOT` applied `e^{iπ/4} T†` and `S T† = T`); `sum_blockKron_isometry` (discarding through the physical labels is discarding the logical qubit); ★ `injKraus_mul_pairEnc` (**every physical Kraus operator is the logical one**); ★ `logicalInj_blockKron` (`ρ ⊗ τ ↦ tr τ · TρT† ⊗ |0⟩⟨0|`); ★★★ `injChannel_pairEnc` (**the injection gadget is the logical `T`, exactly, on every encoded input**); ★★ `injChannel_eq_idealInj`; ★★★ `steane_injection_corrected` (**a corrected step, with the magic state's error as the `DeviatesBy` hypothesis**), ★★ `deviatesBy_injChannel_magicEnc` (the exact magic state discharges it), ★★★ `steane_injection` (one pair recovery for both). Scope: level one; the magic state's error is a hypothesis, not discharged for any faulty preparation (#304); the injection is not in the pair machine's gadget set or its fault count (#305). **First brick DONE 2026-11-05 — the logical `S̄`** ([`Empirical/QM/QEC/SteaneLogicalS.lean`](../CsdLean4/Empirical/QM/QEC/SteaneLogicalS.lean), 22 pins with the gadget-set extension): `hamWt_coset_mod_four` (**the code's words have weight `0`/`4` on the row space, `3`/`7` on its all-ones coset**, by `decide`); ★★ `sTransversal_mul_steaneEnc` (**`S^{⊗7} E = E S†` exactly** — the transversal `S` is the logical `S†`), ★★ `sTransversal_conjTranspose_mul_steaneEnc` (the transversal `S†` is the logical `S`), ★ `sTransversal_mul_steaneProj` / `_conjTranspose_mul_steaneProj` (both keep the code); in `SteanePairMachine.lean` the gadget set gains `PairGadget.phase` (`S̄` on either block) with ★★ `PairGadget.propagatesD_phase`, so `steane_pair_machine` now runs over `H̄`, `S̄`, `X̄`, `Z̄`, `CNOT̄` — the two-qubit Clifford generators — with the same `1 − 21 N p²`. Scope: level one; `S̄` is not in the one-block `Gadget` type, because at the concatenated level the transversal gate alternates `S`/`S†` (row #303). ~~**Remaining (the injection, L−):** the encoded magic state, `CNOT̄`, the logical `Z̄` measurement and the `S̄` correction as a corrected step on the two-block register.~~ done (second brick). ▸ *Original row:* The machine's gates are Clifford; universality needs `T̄`. The corpus has `T`-injection (rows 66, 67) and 15-to-1 distillation on the `[[15,1,3]]` code (rows 76–79) at the physical level. Deliverables: injection at the logical level (encoded magic state, transversal `CNOT̄`, logical measurement, `S̄` correction) as a corrected step on the two-block register, with the magic state's error as a `DeviatesBy`-style hypothesis. ~~⚠️ **Stuck on:** nothing identified — ~~#226~~ (the two-block register) and the logical `S̄` both done 2026-11-05; the price is the measurement-and-feedforward step as a channel on the pair register, and the bridge from `MagicInjection.lean`'s coordinate-operator model to the matrix model.~~ closed; the bridge was not needed (the 4×4 logical identity was proved directly). | ~~**L**~~ done (M, two bricks; residue #304, #305) | High — universality in the machine | ~~#207~~; #226; ~~#66~~; ~~#67~~; `Mathlib/QuantumInfo/MagicInjection.lean` | > | 231 | QIT | QC chain (link 12) | *new (2026-10-29, residue of ~~#227~~)* **The recursive simulation at the Steane code.** ▸ **Brick 1 DONE 2026-11-06 (M)** — the recursion with its interleaved recoveries, in [`SteaneRecursiveSim.lean`](../CsdLean4/Empirical/QM/QEC/SteaneRecursiveSim.lean): `simKraus` (the level-`(k+1)` gadget's Kraus operators: a level-`(k+1)` recovery Kraus operator after the tensor of the seven blocks' level-`k` gadgets, each ending in its own recovery), `simApply`, ★ `simKraus_tp` (a channel), `exists_simTower` (#86's recoveries act Kraus operator by Kraus operator on the span), ★★★ `simKraus_mul_cEnc` (**on a good pattern every Kraus operator is a scalar times the ideal gate on the code**: good blocks by induction, the one bad block arbitrary, the tensor in the span by ★★ `blockKron_mem_span_cErr`), ★★★ `simApply_eq` (**exact on every code state**, the weights summing to one by trace preservation), ★★★ `steane_machine_recursive` (**the machine as the recursive simulation**: one recovery tower for all levels, the five-gate set of #303, bound `1 − N (21p)^{2^k}/21`). Extracted on the way: `exists_kraus_mul_mul_eq_smul` (`KnillLaflamme.lean`, the Kraus-level half of the converse), `blockKron_mem_span_cErr`, `exists_forall_ne_isBad_eq_false`, `cErr_cIdOne` (`SteaneConcat.lean`, `patErr_mem_span` now through them). 22 pins. The recursion is proved directly at the concatenated codes, not through `IsCorrectedAtLevel` (whose one code projector `P` does not change with the level). **Remainder split 2026-11-06 into #312–#315** (the recoveries built as level-`(k−1)` circuits of gadgets and their locations counted — the Aliferis–Gottesman–Preskill level reduction, which needs a decoder statement per rectangle and a rectangle for every gate the recovery uses, `CNOT̄` included, so it depends on #302); this row closes when #315 does. ▸ *Original row:* ~~#227~~'s level-`k` gadget is the transversal gate on `7^k` qubits followed by **one** level-`k` recovery; the threshold argument's simulation builds a level-`k` gadget as a level-`(k−1)` circuit of gadgets **with interleaved level-`(k−1)` recoveries**, which is what `IsCorrectedAtLevel` (#97) abstracts. Deliverables: the level-`(k−1)` recovery of each block as a gadget of the circuit, the identification of the level-`k` ideal gadget with that circuit's ideal run (`IsCorrectedAtLevel steaneProj`-style at `cProj k`), and the bound with the recoveries' locations counted. ⚠️ **Stuck on:** nothing identified for the counted version since ~~#238~~ (2026-10-30: `steane_machine_overlap` counts every recovery location at level one, `1 − 23871 N p²`; the price is the level-`k` bookkeeping — ~~#234~~ done 2026-10-30: `steane_verified_ancilla`, one verified ancilla; ~~#235~~ done 2026-10-30: channel faults before or after the `CNOT` at every extraction location; ~~#232~~, ~~#233~~ done 2026-10-29: `steane_machine_overlap` counts the recovery's locations at level one, through the generic `IsFaultChain`); the fault-free tower itself has no wall — the encoder intertwiners of ~~#227~~ give the ideal-run identity. | **L** | High — the simulation theorem at a concrete code | ~~#227~~; #229; ~~#97~~; `Empirical/QM/QEC/SteaneMachineConcat.lean`; `Mathlib/QuantumInfo/LevelReduction.lean` | > | ~~232~~ | QIT | QC chain (link 12) | *new (2026-10-29, split from ~~#229~~)* **The faulty recovery as a correctable deviation.** ▸ **DONE 2026-10-29 (M–L took M).** [`Empirical/QM/QEC/SteaneFaultyRecovery.lean`](../CsdLean4/Empirical/QM/QEC/SteaneFaultyRecovery.lean) (56 pins): `CssErr`, `cssPauli` (**the sixty-four errors** `X^a Z^b`, `wt a, wt b ≤ 1`), ★ `cssSyn_injective`; ★★★ `circuitRecovery_css_cross` (**Knill–Laflamme for all sixty-four, cross terms included**); ★★ `cssMix` (**the class**: Kraus operators in their span, `P (Σ KᴴK) P = P`), `id_mem_cssMix`, ★ `conj_cssPauli_mem_cssMix` (`X` and `Z` on different qubits), ★★ `weightOne_subset_cssMix`; ★★★ `isRecovery_cssMix`, ★★ `isRecovery_methodRecovery_cssMix`; ★★ `Gadget.propagatesD_cssMix` (via `propagatesD_cssMix_of_unitary`); `OnPair` (**a fault is any operator on data qubit `q` and ancilla qubit `q`**), `faultReadOp`, ★★ `atQubit_bitFaultOp` / `atQubit_phaseFaultOp` (linearity from ~~#229~~), `sum_faultReadOp_conjTranspose_mul`, `faultyKraus_tp`, ★★ `faultyMethod` (**with a unitary fault the faulty method is a channel**); ★★★ `deviatesBy_cssMix_of_atQubit`, ★★★ `deviatesBy_faultyMethod` (**#96's `DeviatesBy` for the recovery piece, discharged**), ★★★ `methodRecovery_faultyMethod` (one fault in a recovery is removed by the next); ★★★ `steane_exRec_faultyMethod` (**#96's extended rectangle with every piece a circuit**: leading-recovery fault or gadget fault corrected, trailing-recovery fault handed on as a CSS deviation). Generic: `DeviatesBy.mono`, `IsRecovery.anti`, `pauliMat_conj`, `sum_mul_conjTranspose_mul_of_tp`, `Controlled.gateOf_conjTranspose`, `sum_ancRead_conjTranspose_mul`, `ancPrep_conjTranspose_mul`. Residue: faults before the `CNOT` and channel faults → #235. ▸ *Original row:* ~~#229~~'s brick proves one extraction fault in Steane's method leaves `Σ_b M_b ρ M_bᴴ` with every `M_b` on one qubit; a prep fault (#234) leaves an `X` error and a `Z` error that may sit on **different** qubits. Deliverables: the CSS correctable set — mixtures `ρ ↦ Σ K_b ρ K_bᴴ` with each `K_b P` in the span of `X^a Z^b`, `wt a, wt b ≤ 1`, trace-preserving on the code; `circuitRecovery`'s Knill–Laflamme property for that 64-element family (the decoder already handles it: `decode` inverts both syndromes); `IsRecovery`, `PropagatesD` for the gadget set, and Steane's method with one fault as `DeviatesBy id` — so `steane_exRec` runs with a faulty recovery that is a circuit, not a hypothesis. ⚠️ **Stuck on:** nothing identified; the price is the mixture class and the 64-member Knill–Laflamme sum. | ~~**M–L**~~ done (M) | High — the recovery's own faults enter the rectangle as a theorem | ~~#229~~; ~~#228~~; ~~#96~~; `Empirical/QM/QEC/SteaneMethod.lean`; `Empirical/QM/QEC/SteaneMachine.lean` | > | ~~233~~ | QIT | QC chain (link 12) | *new (2026-10-29, split from ~~#229~~)* **The machine with the recovery's locations counted.** ▸ **DONE 2026-10-29 (L took M–L).** Generic: [`Mathlib/QuantumInfo/OverlappingRectangles.lean`](../CsdLean4/Mathlib/QuantumInfo/OverlappingRectangles.lean) (16 pins): `ovRun`, `ovIdeal`, `runMaps_ofFn_eq_ovIdeal`; ★★ `IsFaultChain` (**#96's three piecewise properties, one per piece, with a fault flag**), `lastFault`, `ExRecGood` (**at most one flagged piece in the `k`-th extended rectangle**); ★★★ `IsFaultChain.step` (**one overlapping rectangle**: leading fault carried and removed, gadget fault removed, trailing fault handed on), ★★★ `IsFaultChain.ovRun_spec` (**the invariant**: after `k` steps a `D`-deviation of the ideal run, exact unless the last recovery was flagged), ★★★ `IsFaultChain.decode_ovRun`; ★★ `measure_two_le_card_filter_le` (**two faults in a window: `≤ C(|W|,2) ν(fault)²`** by a union over the window's pairs, no marginal computed), ★★★ `one_sub_le_measure_output_eq_windows` (**the join for overlapping windows**). Steane: [`Empirical/QM/QEC/SteaneOverlapMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteaneOverlapMachine.lean) (40 pins): `roundFault`, ★★ `methodUnder` (**Steane's method with a unitary at every faulty extraction location of any pattern, a channel**), ★ `methodUnder_none` (= `methodRecovery`), ★ `methodUnder_single` (= ~~#232~~'s `faultyMethod`); `StepLoc`, `OvLoc N` (**one location space: 21 per step, gadget 7 + recovery 14**), `window`, ★ `card_window_le` (≤ 35), `WindowsGood`; `overlapRun`, ★ `ovIdeal_ovGadget` (= #207's `machineIdeal`); ★★★ `isFaultChain_overlap` (**the machine is an `IsFaultChain` for `cssMix`**), ★★ `exRecGood_of_windowsGood`; ★★★ `steane_machine_overlap` (**every overlapping exRec with ≤ 1 fault ⟹ state a CSS deviation of the ideal, exact if the last recovery is fault-free, ideal after a final ideal decoding; under independent faults of rate `p` at all `21N` locations, `P(ideal) ≥ 1 − 595 N p²`**, `595 = C(35,2)`). Residue: encoded input and final read-out as rectangles → #236; recovery fault model's last two cases stay #235; ancilla preparation #234; level `k` #231. ▸ *Original row:* The corpus composes **non-overlapping** corrected steps (`correctedRun_eq_idealRun`); with recoveries between gadgets the extended rectangles **overlap** (each recovery trails one gadget and leads the next), which is Aliferis–Gottesman–Preskill's setting. Deliverables: one location space for the circuit (every gadget's and every recovery's locations), the windows of the `N` extended rectangles, the invariant that after each rectangle the state is a correctable deviation of the ideal one (nonzero only if the trailing recovery was the faulty piece), a final ideal decoding, and the bound `1 − N·C(n,2)p²` by the union bound over overlapping windows (each window's marginal of the product measure is a product). ⚠️ **Stuck on:** the overlapping-rectangle invariant — #96 has its pieces (`isCorrectedStep_exRec_of_good`, `deviatesBy_exRec_trailing`) but nothing chains them; ~~#232~~ done 2026-10-29 (`deviatesBy_faultyMethod`, `isRecovery_methodRecovery_cssMix`, the class `cssMix` the chain carries). | ~~**L**~~ done (M–L) | High — the level-one machine with nothing ideal but ancilla preparation | ~~#229~~; #232; ~~#96~~; ~~#97~~; `Mathlib/QuantumInfo/LevelReduction.lean` | > | ~~234~~ | QIT | QC chain (link 12) | *new (2026-10-29, split from ~~#229~~)* **Verified ancilla preparation for Steane's method.** ▸ **DONE 2026-10-30 (L took L).** [`Empirical/QM/QEC/SteaneVerifiedAncilla.lean`](../CsdLean4/Empirical/QM/QEC/SteaneVerifiedAncilla.lean) (88 pins). **Encoder:** `cnotRelabel`, `plusGates` (nine `CNOT`s, information qubits `2,4,5,6` onto parity `0,1,3`), `cnotCircuit`, `plusInit`, ★★ `plusEncoder_mulVec` (**the encoder prepares `|+̄⟩`**, kernel-checked over 128 labels), ★★ `cnotCircuit_fault_mulVec` (**one fault anywhere leaves `pre F preᴴ |+̄⟩`**), `encoderFault_tp`. **Verification:** a checker `|+̄⟩` controls a transversal `CNOT` onto the ancilla and is read in the `X` basis (`hadB`, `verifyOut`), passing when its word is orthogonal to the Hamming code (`Accepts`); ★★★ `isPeriodic_verifyOut_prepA` (**a passing checker leaves the ancilla with `X` errors only, whatever error its preparation made** — the `CNOT` copies the `Z` part onto the checker), ★★ `isPeriodic_verifyOut_prepB` (faulty checker), ★★ `verifyOut_pairFault` (a fault on pair `j` leaves an error on qubit `j`), `verify_fault_before`, `readout_fault_eq`; ★ `IsPeriodic.eq_sum` (**`X` errors only = a combination of `X^{decode t}|+̄⟩`**, the `Z` error trivial modulo the stabiliser — stronger than the row's weight-`≤ 1`). **Factory:** ★★ `factoryOut` (pass → the ancilla; fail → a fresh `|+̄⟩`, the one fault spent), ★★ `factoryOut_norm`, `factoryOut_ideal`. **Method:** ★★ `ancMethod` (one round's ancilla from a family), ★★ `ancMethod_apply_of_eq_smul`, ★★ `exists_atQubit_ancBitOp` / `exists_atQubit_ancPhaseOp` (**an `X` error on the ancilla only mis-reads the syndrome; the decoder's correction has weight one**), ★★★ `deviatesBy_cssMix_of_exists_atQubit` (the qubit may depend on the branch — the row's 'X and Z on different qubits' case is subsumed), ★★★ `deviatesBy_ancMethod`; ★★★ `steane_verified_ancilla` (**encoder correct; fault-free, the verified method is `methodRecovery`; one fault in the ancilla's preparation, the checker's, or a verification pair ⟹ a CSS deviation**). Generic (`QuantumInfo`, 10): `permMat_mulVec_apply`, `kronecker_mulVec_prod`, `ancPrep_mulVec`, `sum_star_mulVec_mul_mulVec`, `hadTransversal_apply`, …. Residue: the factory's locations counted in ~~#233~~'s machine → #238. ▸ *Original row:* ~~#229~~'s brick takes `|+̄⟩` as prepared ideally. A fault in its encoding circuit can leave a `Z` error of weight two or more, which the transversal `CNOT` copies onto the data. Deliverables: the encoding circuit of `|+̄⟩`, a verification step (a second ancilla checked against it, or the standard fault-tolerant verification) with accept/reject, the theorem that one fault leaves an **accepted** ancilla whose `Z` error is weight `≤ 1` modulo the stabiliser (`X` errors arbitrary — they only mis-read the syndrome, and the decoder's correction is then weight one), and the prep-fault case of `steaneMethod_single_fault` (an `X` error and a `Z` error, each weight one, possibly on different qubits — #232's class). ⚠️ **Stuck on:** the verification circuit and its fault analysis are new; #95 has the ladder's propagation bound, not verification. The class the prep fault lands in exists (~~#232~~, `conj_cssPauli_mem_cssMix`). | ~~**L**~~ done (L) | High — removes the last ideal piece of the level-one recovery | ~~#229~~; #232; ~~#95~~; `Empirical/QM/QEC/SteaneMethod.lean`; `Mathlib/QuantumInfo/AncillaLadder.lean` | > | ~~235~~ | QIT | QC chain (link 12) | *new (2026-10-29, residue of ~~#232~~)* **The fault model of Steane's method, completed.** ▸ **DONE 2026-10-30 (S–M took S–M).** [`Empirical/QM/QEC/SteaneFaultModel.lean`](../CsdLean4/Empirical/QM/QEC/SteaneFaultModel.lean) (29 new pins; `PairOp`, `onPair_one` moved here): `OnPair.of_sum`, `OnPair.zero`, `OnPair.add`, `OnPair.smul`, ★ `OnPair.mul`, `OnPair.list_prod` (**the operators on one pair are a subalgebra**); `cnotData`, `cnotAnc`, ★★ `tCNOT_conj_kron` (**`CNOT (A ⊗ B) CNOT = Σ_{a,b} PₐAP_b ⊗ XᵃBXᵇ` at `q`, the other six `CNOT`s cancelling**), ★★ `OnPair.conj_tCNOT`; (i) `readAround`, ★★ `readAround_eq` (**a fault `Fb` before the `CNOT` and `Fa` after it read out as the single fault `Fa · CNOT Fb CNOT` after it**); (ii) `sum_faultReadOp_family` and the bit/phase forms (**the readout under a trace-preserving Kraus family of faults is complete**), `channelKraus`, ★★ `channelMethod`, ★ `channelMethod_unit` (= ~~#232~~'s `faultyMethod` for one operator), `aroundFam`, `aroundFam_tp`, ★ `onPair_aroundFam`, ★★★ `deviatesBy_channelMethod`, ★★★ `deviatesBy_channelMethod_around`, ★★ `methodRecovery_channelMethod` — **no closure lemma for `cssMix` was needed**: every Kraus operator of the faulty method acts on the code on qubit `q`, so `deviatesBy_cssMix_of_atQubit` applies directly; (iii) [`SteaneOverlapMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteaneOverlapMachine.lean) **widened in place** (17 new pins, 7 retired with `roundFault`): `QuantumInfo.sum_prod_ofFn_conjTranspose_mul` (**any number of channels in sequence**), `locFam` (**a location as a Kraus family on `Option κ`, identity when fault-free**), `roundKraus`, `roundKraus_none`, `roundKraus_none_of_ne`, `onPair_roundKraus_single`, `UnderIdx`, `underKraus_tp`, ★★ `methodUnder` now takes a channel `F j r q : κ → PairOp` at every extraction location, ★★★ `deviatesBy_methodUnder_single`; `isFaultChain_overlap` and ★★★ `steane_machine_overlap` restated with **channel faults on the pair at every extraction location, before or after the `CNOT`** — same bound `1 − 595 N p²`. Residue: channel faults at the **gadget** locations (still unitary on one qubit) → #237. ▸ *Original row:* ~~#232~~ proves `deviatesBy_faultyMethod` for a **unitary** operator on data qubit `q` and ancilla qubit `q` **after** the transversal `CNOT`. Deliverables: (i) a fault **before** the `CNOT` at `q` — `tCNOT F' = (CNOT_q F' CNOT_q) tCNOT`, the conjugate again `OnPair q` (the other pairs' `CNOT`s commute with it); (ii) a fault that is a **channel** (several Kraus operators on the pair): the faulty method is then a mixture of `faultyMethod` runs, still a CSS deviation (`cssMix` is closed under such mixtures); (iii) `deviatesBy_faultyMethod` restated over both, and ~~#233~~'s `methodUnder` / `isFaultChain_overlap` widened to take them, so the machine's location space puts any single-location fault model at an extraction location. ⚠️ **Stuck on:** nothing identified; (i) is a `CNOT_q` commutation computation on the pair register, (ii) a closure lemma for `cssMix`. | ~~**S–M**~~ done (S–M) | Medium — the recovery's fault model matches the gadgets' | ~~#232~~; ~~#233~~; `Empirical/QM/QEC/SteaneFaultModel.lean`; `Empirical/QM/QEC/SteaneOverlapMachine.lean`; `Empirical/QM/QEC/SteaneFaultyRecovery.lean`; `Empirical/QM/QEC/SteaneMethod.lean` | > | ~~236~~ | QIT | QC chain (link 12) | *new (2026-10-29, residue of ~~#233~~)* **Encoded input and logical read-out as rectangles.** ▸ **DONE 2026-10-30 (M–L took L).** [`Empirical/QM/QEC/SteaneLogicalIO.lean`](../CsdLean4/Empirical/QM/QEC/SteaneLogicalIO.lean) (62 pins) + 3 generic in `OverlappingRectangles.lean` (`lastFaultFrom`, `IsFaultChain.step_of_flag`, ★★★ `IsFaultChain.ovRun_spec_from` — **the chain from a deviated input**, the preparation counted in the first extended rectangle). **Input:** ★ `steaneProj_mulVec_plusAnc` (`|+̄⟩` is a code state, by counting over the 128 words), ★★ `steaneProj_cssPauli_conj` (**Knill–Laflamme for the sixty-four CSS errors**, `P E_eᴴ E_e' P = δ P`), ★★ `exists_steaneProj_conj_eq_smul` (on the code `Kᴴ K = ‖K|+̄⟩‖²` for every `K` in the CSS span), ★ `GoodAnc.exists_mem_cssSpan`, `prepState`, ★★ `prepState_none`, ★★★ `exists_prepState_single` (**one fault anywhere in the input's factory prepares a CSS deviation of `|+̄⟩⟨+̄|`**). `|0̄⟩ = H̄|+̄⟩` is the circuit's first gadget — no separate preparation needed. **Read-out:** `logicalBit` (decode the word's syndrome, correct, take the parity), `logProj`, ★ `logicalProb`, ★★ `logicalProb_steaneEnc` (**it measures the logical qubit in `Z̄`**: `σ v v` on `E σ Eᴴ`), ★ `logProj_mul_steaneProj`, `logProj_mul_mem_cssSpan`, ★★★ `logicalProb_cssMix` (**a CSS deviation of a code state never changes the logical read-out**); `measChannel` (a Kraus family on each qubit before it is read), `measChannel_none`, ★★ `measChannel_single_mem`. **Machine:** `IOLoc N` (**46 + 113 N + 7 locations**), `ioRun`, `ioWindow` (first rectangle with the input's factory; last window = last recovery + measurement), ★ `card_ioWindow_le` (≤ 219), `IOWindowsGood.good_from`, `.good_last`, ★★★ `logicalProb_ioRun`, ★★★ `steane_machine_io` (**the level-one Steane machine from preparation to read-out, nothing ideal but the classical decoding: the decoded logical read-out has its ideal distribution with probability ≥ `1 − 23871 (N + 1) p²`**). No new residue: the input's retry is #240, the constant #239, gadget channel faults #237. ▸ *Original row:* `steane_machine_overlap` starts from a code state and ends with an **ideal** decoding (`methodRecovery` applied fault-free, then the ideal output compared). Deliverables: a preparation gadget for `|0̄⟩` (from #234's verified `|+̄⟩` and the transversal `H̄`, or its own encoding circuit) whose single fault leaves a `cssMix` deviation; a **transversal measurement** gadget — each qubit read in the `Z` basis, the logical bit computed classically from the word (`decode` on the syndrome, then the parity) — and the theorem that one fault in the last recovery or in the measurement still gives the ideal logical outcome; the machine restated preparation → `overlapRun` → measurement with these windows counted, so that nothing is ideal but the classical decoding. ⚠️ **Stuck on:** nothing for the preparation half since ~~#234~~ (2026-10-30: the verified `|+̄⟩` factory, `steane_verified_ancilla`; `|0̄⟩ = H̄|+̄⟩` by the transversal Hadamard); the measurement half has nothing identified — the word read is a classical codeword plus one flipped bit, which `decode` already handles. | ~~**M–L**~~ done (L) | High — the level-one machine with no ideal quantum piece | ~~#233~~; #234; ~~#228~~; `Empirical/QM/QEC/SteaneOverlapMachine.lean`; `Empirical/QM/QEC/SteaneMethod.lean` | > | ~~237~~ | QIT | QC chain (link 12) | *new (2026-10-30, residue of ~~#235~~)* **Channel faults at the gadget locations.** ▸ **DONE 2026-11-02 (S–M took S–M).** [`Empirical/QM/QEC/SteaneOverlapMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteaneOverlapMachine.lean): `locFam_gateOf`, ★ `roundKraus_gateOf` (**with every location but `q₀` fault-free, each branch of the gadget's faults is `gateOf q₀` of a 2×2 operator** — the fault-free locations contribute `0` or `1`); `Gadget.faultyMap` (branch `s`: `roundKraus (gateOf ∘ U) b s` after the transversal gate), `Gadget.faultyChannel` (**trace preserving**, from `roundKraus_tp`; `gateOf_family_tp` rehomed here from `SteaneLogicalIO`), `Gadget.faultyMap_eq` (the ideal gadget followed by the faults), `Gadget.faultyMap_none`, ★★ `Gadget.deviatesBy_faultyMap` (**one faulty location, any trace-preserving family on its qubit ⇒ a `cssMix` deviation of the ideal gadget**; no re-indexing over branches — every branch is in `cssSpan` by `gateOf_mem_cssSpan`, the trace condition is `roundKraus_tp`); `ovFaultyGadget`/`overlapRun`, ★★★ `isFaultChain_overlap` and ★★★ `steane_machine_overlap` **restated in place with channel faults at every gadget location** (the hypothesis `∑ (U j q k)ᴴ U j q k = 1` replaces unitarity), and downstream `ioRun`, `logicalProb_ioRun`, `ioRun_ioPrune`, ★★★ `steane_machine_io` (`SteaneLogicalIO.lean`) with them. **Finding:** the gadget faults need no structure beyond trace preservation on one qubit; the one-fault deviation is the faults applied to the ideal output, exactly. Scope: single-qubit channels at each gadget location (a correlated fault across a gadget's qubits is not covered); one block, three gates (#226). No new rows. After ~~#235~~ an extraction location takes any trace-preserving Kraus family on its pair, but a gadget location still takes one **unitary** on one qubit (`Gadget.faulty U`, #207's model, `Gadget.deviatesBy_faulty`). Deliverables: `Gadget.faulty` under a pattern with a Kraus family `G j q : κ → Matrix (Fin 2) (Fin 2) ℂ` at each faulty location (the same `Option κ`-indexed ordered product as `roundKraus`), trace preserving; its one-fault form a `cssMix` deviation of the ideal gadget (each Kraus operator `gateOf q (G k)` times the transversal gate, in `cssSpan` by `gateOf_mem_cssSpan` and `Gadget.conj_cssPauli_mem_cssSpan`); `isFaultChain_overlap` and `steane_machine_overlap` restated with channels at every location of the machine. ⚠️ **Stuck on:** *done 2026-11-02 — was:* nothing identified; the product and trace-preservation lemmas are ~~#235~~'s (`sum_prod_ofFn_conjTranspose_mul`, `locFam_tp`), and the deviation is `deviatesBy_cssMix_of_atQubit`'s argument with the gadget's unitary in front. | ~~**S–M**~~ done (S–M) | Medium — one fault model, channels, at every location | ~~#235~~; ~~#207~~; `Empirical/QM/QEC/SteaneOverlapMachine.lean`; `Empirical/QM/QEC/SteaneMachine.lean`; `Empirical/QM/QEC/SteaneFaultyRecovery.lean` | > | ~~238~~ | QIT | QC chain (link 12) | *new (2026-10-30, residue of ~~#234~~)* **The verified ancilla's locations in the machine.** ▸ **DONE 2026-10-30 (M–L took L).** 72 new pins. **The factory under any pattern** ([`SteaneVerifiedAncilla.lean`](../CsdLean4/Empirical/QM/QEC/SteaneVerifiedAncilla.lean)): `EncLoc` (seven preparations, nine `CNOT`s), `encUnder` (a Kraus family after every encoder location), `encUnder_tp`, ★ `encUnder_none_mulVec`; `verUnder` (faults on the seven verification pairs and the seven checker readouts), `facIn`, `FacFaults`, `FacFaults.Valid`, `FacLoc` (**46 locations**), `facVec`, ★★ `facOut`, ★★ `facOut_norm`, ★★ `facOut_none` (fault-free → multiples of `|+̄⟩`), ★★★ `goodAnc_facOut_single` (**one fault anywhere in the factory → a good ancilla in every branch**, via ★★ `isPeriodic_verifyVec_A` / `_B` for **any** input on the faulty block and ★ `readVec_apply`). **The recovery** ([`SteaneOverlapMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteaneOverlapMachine.lean), widened in place): `RecLoc` (**106 per recovery**: 2 × (7 extraction + 46 factory)), `extractOp`, ★ `sum_extractOp_family`, `roundOp`, ★★ `recUnder` (both factories and both extractions under a pattern, a channel for every pattern), ★★ `recUnder_none` (= `methodRecovery`, via `sum_smul_conj`), ★★★ `exists_atQubit_recKraus'_single` / ★★★ `deviatesBy_recUnder_single` (**one fault anywhere in the recovery → a CSS deviation**). **The machine:** `StepLoc = Fin 7 ⊕ RecLoc` (**113 per step**), ★ `card_window_le` (**≤ 219**), `isFaultChain_overlap` and ★★★ `steane_machine_overlap` restated: **nothing ideal in the recovery but its classical decoding; `P(ideal) ≥ 1 − 23871 N p²`, `23871 = C(219, 2)`**. The pattern helpers (`locFam`, `roundKraus`, `sum_prod_ofFn_conjTranspose_mul`) moved to `SteaneFaultModel.lean`. Residue: a sharper constant by malignant-pair counting → #239; the rejected ancilla's retry counted → #240. ▸ *Original row:* ~~#234~~ proves one run of Steane's method with its ancilla from the verified factory is a CSS deviation under one factory fault; ~~#233~~'s machine still takes ideal ancillas. Deliverables: `StepLoc` widened by each recovery's factory locations (per round: the ancilla's encoder, the checker's encoder, seven verification pairs), `methodUnder` built on `ancMethod` with the pattern's faults in both the extraction and the factory, `card_window_le` restated (window size and the constant `C(w, 2)` re-priced), and `steane_machine_overlap` restated with **nothing ideal in the recovery but its classical decoding**. ⚠️ **Stuck on:** nothing identified; the one-fault case is `deviatesBy_ancMethod`, and a pattern with one fault in the window puts it in exactly one of the extraction or the factory, so `IsFaultChain`'s fields are unchanged. The price is the location bookkeeping and an encoder fault counted per gate (nine `CNOT`s, seven preparations) rather than as one family. | ~~**M–L**~~ done (L) | High — the level-one recovery with no ideal quantum piece | ~~#234~~; ~~#233~~; ~~#235~~; `Empirical/QM/QEC/SteaneVerifiedAncilla.lean`; `Empirical/QM/QEC/SteaneOverlapMachine.lean` | > | 239 | QIT | QC chain (link 12) | *new (2026-10-30, residue of ~~#238~~)* **Malignant pairs: a sharper constant.** ▸ **First brick DONE 2026-10-31: the retry's benign pairs, `81003 → 32335`.** [`Mathlib/QuantumInfo/OverlappingRectangles.lean`](../CsdLean4/Mathlib/QuantumInfo/OverlappingRectangles.lean): `measure_pair_eq`, ★★ `measure_exists_pair_le`, ★★★ `one_sub_le_measure_output_eq_pairs` (**the join over malignant pairs**: ideal whenever no pair of each window's family is faulty ⟹ `1 − N · m · p²`, `m` the largest family). [`SteaneVerifiedAncilla.lean`](../CsdLean4/Empirical/QM/QEC/SteaneVerifiedAncilla.lean): ★★ `retryOut_eq_of_first_clean` (**with the first attempt fault-free, the second attempt's faults change nothing**), `pruneSecond`, `retryOut_pruneSecond`. [`SteaneOverlapMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteaneOverlapMachine.lean): `pruneRetry` (drop a second-attempt fault whose first attempt is clean), ★★ `overlapRun_pruneRetry` (**pruning does not change the machine**); `primWin` (a window's `219` primary locations), ★ `malign` (two primary locations, or the two attempts of one factory), ★ `card_malign_le` (**`≤ C(219, 2) + 4 · 46² = 32335`**), ★★ `windowsGood_pruneRetry` (no malignant pair faulty ⟹ the pruned pattern has one fault per window); `steane_machine_overlap` **strengthened**: any pattern with no malignant pair faulty decodes ideally, and the bound is **`1 − 32335 N p²`**. [`SteaneLogicalIO.lean`](../CsdLean4/Empirical/QM/QEC/SteaneLogicalIO.lean): the same from preparation to read-out (`ioPrune`, `ioRun_ioPrune`, `ioMalign`, ★ `card_ioMalign_le` — the first window has only `159` primary locations, since `prevStep 0 = 0`, the last `113` — ★★ `ioWindowsGood_ioPrune`); `steane_machine_io` strengthened, **`1 − 32335 (N + 1) p²`**. **Finding:** the retry costs nothing in the constant beyond `4 · 46²` per window — its faults are harmful only with a first-attempt fault of the same factory. **Not done (the rest of this row):** every pair of the `219` primary locations is still malignant; the basis split — an `X`-type and a `Z`-type fault are corrected by different rounds — is not made. ▸ *Original row:* `steane_machine_overlap` bounds the failure by `C(219, 2) N p² = 23871 N p²`, counting every pair of faults in a window (since ~~#240~~ the windows hold `403` locations with the retry counted, `C(403, 2) = 81003`; with no first-attempt fault the second attempt carries weight zero (`facVec_none_of_not_accepts`), so a pair holding a second-attempt fault can be malignant only when its other fault is in the first attempt — most of the new pairs are benign). Aliferis–Gottesman–Preskill count only **malignant** pairs — those that actually defeat the rectangle (two faults in the same round's `X` part, say, rather than one `X`-type and one `Z`-type). Deliverables: a predicate on pairs of window locations that over-approximates malignancy (both faults can reach the same basis's syndrome), the theorem that a window whose faults form no such pair still decodes ideally, and the bound with the count of such pairs in place of `C(219, 2)`. ⚠️ **Stuck on:** for the retry's pairs, nothing — done 2026-10-31 (`32335`). For the basis split, the chain argument: `IsFaultChain` carries one deviation class `cssMix` with one flag per piece, so two faults in one window — even one `X`-type and one `Z`-type — leave its hypotheses; the price is a two-flag chain (an `X` deviation and a `Z` deviation, each corrected by its own round) and the per-location basis of each fault's effect read off the circuit. `one_sub_le_measure_output_eq_pairs` already takes any family. | **M–L** (remaining: **L**) | Medium — a threshold estimate closer to the literature | ~~#238~~; ~~#233~~; `Empirical/QM/QEC/SteaneOverlapMachine.lean`; `Mathlib/QuantumInfo/OverlappingRectangles.lean` | > | ~~240~~ | QIT | QC chain (link 12) | *new (2026-10-30, residue of ~~#238~~)* **The rejected ancilla's retry.** ▸ **DONE 2026-10-31 (M).** [`Empirical/QM/QEC/SteaneVerifiedAncilla.lean`](../CsdLean4/Empirical/QM/QEC/SteaneVerifiedAncilla.lean): ★ `hadTransversal_mulVec_plusAnc_of_not_accepts`, ★★ `facVec_none_of_not_accepts` (**a fault-free attempt is never rejected**: `H^{⊗7}|+̄⟩` vanishes off the words orthogonal to the code); ★★ `goodAnc_facVec_single` (one fault in an attempt, a passing word — `goodAnc_facOut_single` now a corollary); `RetryFaults`, `RetryLoc` (**ninety-two locations**, both attempts'), `RetryIdx`, `retryOutV`, ★★ `retryOut` (**on a pass the first attempt's ancilla; on a rejection the second attempt's, whatever its checker reads — nothing prepared fault-free**), ★★ `retryOut_norm`, ★★ `retryOut_none`, ★★★ `goodAnc_retryOut_single` (**one fault anywhere in either attempt leaves a good ancilla in every branch**: a fault in the first leaves the second fault-free; a fault in the second leaves the first fault-free and so never rejected). `SteaneOverlapMachine.lean` and `SteaneLogicalIO.lean` now count the retrying factory: `RecLoc` `198`, a step `205`, ★ `card_window_le` **`403`**, `steane_machine_overlap` **`1 − 81003 N p²`**, `steane_machine_io` **`1 − 81003 (N + 1) p²`** (`81003 = C(403, 2)`; the overlap capstone's docstring, still quoting the pre-#238 `21 N`, `595`, `35`, corrected). **Finding:** removing the last idealised step costs a factor `81003/23871 ≈ 3.4` in the union-bound constant at level one — the retry's locations are in every window whether or not it runs. Residue: the rejection rate and the expected number of attempts → #267; the constant itself is #239. ▸ *Original row:* On a failing checker word the factory (`factoryOutV`) prepares a fresh `|+̄⟩` **without locations of its own** — consistent with one fault per window, since the fault that caused the rejection is spent, but the retry's own locations are not in the window. Deliverables: a factory with a bounded number of attempts (two suffices at one fault per window), each with its forty-six locations, the last attempt's output used whatever it reads; `FacLoc` widened, `card_window_le` and the constant re-priced, `goodAnc_facOut_single` restated. ⚠️ **Stuck on:** nothing identified; a second attempt is a second copy of `facVec` whose input is the first attempt's discarded branch, and one fault across both attempts lands in exactly one. | ~~**M**~~ done | Medium — no idealised step left in the ancilla factory | ~~#238~~; ~~#234~~; `Empirical/QM/QEC/SteaneVerifiedAncilla.lean` | > | ~~241~~ | Records | Records→spacetime | *new (2026-10-30, the author's fourth operational-geometry brief — **two physical observers in relative motion recover a common 1+1-dimensional flat spacetime description from their own records; GR is the subsequent programme**; split into #241–#248, recorded together, with four amendments from the review: the transport now in the corpus is scheduled and so excluded by the brief's gate; the dispersion class is chosen in #241 because it decides #247; #246 compares the Lieb–Robinson bound with the measured group velocity; #248 split in two)* **Can the dynamics move a clock?** ▸ **DONE 2026-10-30 (M–L took M).** **Finding: the transport now in the corpus cannot** — #190's and #203's chains switch edges on in a prescribed order (an external trajectory) and a uniform hopping chain disperses; **the smallest viable extension is a degree of freedom that moves by its own momentum.** [`RecordLayer/MovingClock.lean`](../CsdLean4/RecordLayer/MovingClock.lean) builds the control (A) on it: `MovingState` (carried hand clock, position, momentum, carried memory, a detector per site), ★★ `movingFlow` (**one autonomous flow**: free motion under `p²/(2m)`, #194's activated flow for the clock, detectors driven by occupancy), ★★ `movingFlow_add`, ★ `hasDerivAt_pos_movingFlow` (Hamilton's equations), ★★ `siteOf_movingFlow_eq_iff` (**transport through the sites**: at site `k` exactly during `[(k − x)/v, (k + 1 − x)/v)`), ★ `siteOf_movingFlow_mono`, `mem_movingFlow` (memory retained), ★★ `abs_movingReading_sub_le` (**the carried clock keeps time during transport**, #201's bound, any momentum), ★★ `occTime_eq_inv_of_crossed` / ★★ `det_movingFlow_passage` (**a detector records the passage**: nothing before arrival, `g_d m / p` after), `movingFlow_instance`, ★★ `clock_movingFlow_indep` (**no time dilation: the control is Galilean**, clock and motion decoupled, a preferred time — the brief's useful negative), ★★★ `moving_clock_control`. Modelling assumptions stated in the header: position a continuous coordinate with sites as cells; test detectors without back-reaction. Residue: the test candidate (B) → #249. ▸ *Original row:* Inspect `RecordLayer/ClockEchoCoupling.lean`, `RecordLayer/ClockEchoHamiltonian.lean`, `RecordLayer/HandClock.lean`, `CV/ChainExchange.lean`, `CV/WalkSchedule.lean`, `CV/LocalLiebRobinson.lean`. Identify a candidate **local, autonomous** dynamics supporting: transport of an identifiable device through several sites; internal oscillation that continues during transport; memory retained during transport; signal interaction with the transported device. **Gate:** no externally prescribed trajectory — #190's and #203's schedules switch edges on in a prescribed order and are therefore excluded; if the present dynamics cannot do it, identify the missing interaction and implement the smallest viable extension. **Amendment:** the candidate declares its **dispersion class** (quadratic, as for uniform hopping — expected non-relativistic; linear, as near a two-band crossing or for a Dirac walk — the regime where dilation can emerge), and two candidates are compared: **(A) the control**, a rotor on an arena torus carrying #201's hand clock and a memory register (exact transport, expected to give a preferred frame); **(B) the test**, a two-band lattice Hamiltonian near its crossing, wave packets travelling with bounded spreading over a finite horizon and the band oscillation as the clock. Deliverable: a concrete candidate, every additional modelling assumption stated, and **at least one proved transport property**. ⚠️ **Stuck on:** nothing identified for (A); (B) needs a wave-packet spreading bound for a two-band chain, which the corpus does not have. | ~~**M–L**~~ done (M) | High — decides whether the programme runs on the present dynamics | ~~#201~~; ~~#212~~; ~~#225~~; ~~#190~~; ~~#203~~; ~~#213~~; #39 | > | ~~242~~ | Records | Records→spacetime | *new (2026-10-30, the author's fourth operational-geometry brief — **two physical observers in relative motion recover a common 1+1-dimensional flat spacetime description from their own records; GR is the subsequent programme**; split into #241–#248, recorded together, with four amendments from the review: the transport now in the corpus is scheduled and so excluded by the brief's gate; the dispersion class is chosen in #241 because it decides #247; #246 compares the Lieb–Robinson bound with the measured group velocity; #248 split in two)* **One transported clock.** ▸ **DONE 2026-10-30 for the control (A) (L took M).** [`RecordLayer/TransportedClock.lean`](../CsdLean4/RecordLayer/TransportedClock.lean): #241's flow as a two-time evolution of one product arena, ★ `measurable_transportFlow` (occupancy measurable, ★ `measurable_occTime`); ★★★ `movingTick` — **tick `n` as an `ActualOccurrence`** of that evolution; ★★★ `passageWrite` — **the passage through site `k` as an `ActualOccurrence`** (threshold `0 < θ ≤ g_d m/p`); ★★★ `transported_clock` — **concrete instance in motion** (mass, momentum `1`, start `1/2`): passage at site `1` written over `[1/2, 3/2]`, every tick `1 ≤ n ≤ K − 1` written, position `1/2 + t`, clock and motion independent of the registers (disturbance zero), memory kept, clock within `τ_tick + 1/(2ω)`. Residue: the detectors generated by a coupling with back-reaction and its disturbance bound → #251; the test candidate (B)'s detector readout → #252. ▸ *Original row:* Build #241's candidate on a finite network over a finite observation horizon. Prove bounds for: device localisation or transport fidelity; tick resolution during transport; memory retention; disturbance caused by transport. Device position defined through local observables and records, the site index kept explicitly as microscopic input. **Acceptance:** the same evolution transports the device and generates its distinguishable ticks (`ActualOccurrence`s, #191), with a concrete nonzero-motion instance satisfying the hypotheses. ⚠️ **Stuck on:** for the control (A), nothing — ~~#241~~ gives transport, ticking, memory and passage records; the ticks as `ActualOccurrence`s of the joint evolution and the detectors' generation with back-reaction remain; for the test (B), ~~#249~~ gives the clock and its dilation on modes and ~~#250~~ the localised device moving at the group velocity with its clock (`dirac_packet`); remaining for (B): the position and the ticks read by detectors on the sites as `ActualOccurrence`s, and a concrete packet instance with its error evaluated. | ~~**L**~~ done (M) | High — the brief's immediate success criterion: a clock physically moves, keeps ticking and retains its records under one autonomous evolution | #241; ~~#201~~; ~~#191~~; ~~#185~~ | > | ~~243~~ | Records | Records→spacetime | *new (2026-10-30, the author's fourth operational-geometry brief — **two physical observers in relative motion recover a common 1+1-dimensional flat spacetime description from their own records; GR is the subsequent programme**; split into #241–#248, recorded together, with four amendments from the review: the transport now in the corpus is scheduled and so excluded by the brief's gate; the dispersion class is chosen in #241 because it decides #247; #246 compares the Lieb–Robinson bound with the measured group velocity; #248 split in two)* **A stationary and a moving observer in one model — the first release milestone.** ▸ **DONE 2026-10-30 for the control (A) (L took M).** [`RecordLayer/TwoObservers.lean`](../CsdLean4/RecordLayer/TwoObservers.lean): two observers in one arena, each with hand clock, position, momentum, memory and an **encounter register** advanced while the other is within half a cell (`[−½, ½)` on `x₁ − x₀` = `(−½, ½]` on `x₀ − x₁`: one co-location, both detectors). ★★ `pairFlow`, `pairFlow_add`, ★ `measurable_pairFlow`; ★★ `obsOf_pairFlow` (**no mutual disturbance**: each observer evolves as alone); ★★ `reg_pairFlow_encounter`; ★★★ `relSpeed_of_register` (**both read the relative speed `g_d/Δη` off their own records, = `p₁/m − p₀/m`**); ★★★ `encounterWrite`, `observerTick` (`ActualOccurrence`s); `pairBudget` (clock errors + transport error, a sum of terms like `coupledBudget`), ★★ `abs_relPos_sub_le`, `abs_obsReading_sub_obsReading_le`, `pair_within_budget`; ★★★ `two_observers`; ★★ `two_observers_instance` (at rest at `2`, moving from `0` at momentum `1`: encounter written over `[3/2, 5/2]`, speed `1` read by both). Residue: the disturbance term is zero for test detectors (#251 adds it); relative motion from signals at a distance is #244; the test candidate (B) → #253. ▸ *Original row:* Two devices, each with its own clock and memory, in one arena, with physical emission and detection interactions. Relative motion defined operationally through a sequence of exchanges and compared with the moving device's transport observables. **Acceptance:** both observers function over a shared interval, with explicit bounds on clock error, transport error and mutual disturbance — the disturbance in #212's coupling-budget form (`coupledBudget`) so that the bounds compose. ⚠️ **Stuck on:** for (A), nothing — ~~#242~~ gives one transported clock with its ticks and passages as record writes; a stationary observer is the same arena at momentum `0`, so the work is the two devices on one network and their exchanges' readings; for (B), #252. | ~~**L**~~ done (M) | High — two functioning physical observers in relative motion | ~~#242~~; #252; ~~#212~~; ~~#202~~; ~~#188~~ | > | ~~244~~ | Records | Records→spacetime | *new (2026-10-30, the author's fourth operational-geometry brief — **two physical observers in relative motion recover a common 1+1-dimensional flat spacetime description from their own records; GR is the subsequent programme**; split into #241–#248, recorded together, with four amendments from the review: the transport now in the corpus is scheduled and so excluded by the brief's gate; the dispersion class is chosen in #241 because it decides #247; #246 compares the Lieb–Robinson bound with the measured group velocity; #248 split in two)* **Two-way exchanges with the moving device.** ▸ **DONE 2026-10-30 for the control (A) (L took L).** [`RecordLayer/MovingExchange.lean`](../CsdLean4/RecordLayer/MovingExchange.lean): messengers reflecting elastically off the moving observer, in an **unfolded coordinate** (relative position `−|ρ + w t|`, the method of images) so the flow with the reflections is one autonomous flow; ★★ `exchFlow`, `exchFlow_add`, ★ `measurable_exchFlow`; ★★ `occTime_eq_inv_of_crossed_neg`, ★ `bounceOcc_add`; ★★ the first observer's register through an exchange (`ExchGeom.bounceOcc_before/between/after`); ★★★ `emissionWrite` (**retained only until the return begins** — a finite window), ★★★ `returnWrite`, ★★★ `receptionWrite`; ★★ timestamps on both clocks; ★★ `elastic_exchange` (**back-reaction of a bounce**: observer's velocity moved by ≤ `2(μ/m)|u − v|`, rigid mirror = `μ/m → 0`); ★★ `exchange_train` (**repeated exchanges**: emissions one unit apart return two apart, the Galilean Doppler factor); ★★★ `moving_exchange`. Residue: launching by the observer → #254; recoil accumulating in the flow → #255; exchanges on the Dirac chain (B) → #256. ▸ *Original row:* Extend #212's coupled echo to exchanges involving the moving device. Prove that emission, reception and return are actual record writes; account for changing separation, detector response (~~#204~~), finite retention and back-reaction. **Acceptance:** each observer obtains a usable set of its own emission and return timestamps, with repeated exchanges demonstrated within the proved operating horizon. ⚠️ **Stuck on:** for (A), nothing — ~~#243~~ gives two observers in one arena with encounter records and a composable budget; the work is a signal carrier that travels between them (#212's echo has fixed sites); for (B), #253. | ~~**L**~~ done (L) | High — the radar data a moving observer can actually collect | ~~#243~~; #253; ~~#212~~; ~~#197~~; ~~#204~~ | > | ~~245~~ | Records | Records→spacetime | *new (2026-10-30, the author's fourth operational-geometry brief — **two physical observers in relative motion recover a common 1+1-dimensional flat spacetime description from their own records; GR is the subsequent programme**; split into #241–#248, recorded together, with four amendments from the review: the transport now in the corpus is scheduled and so excluded by the brief's gate; the dispersion class is chosen in #241 because it decides #247; #246 compares the Lieb–Robinson bound with the measured group velocity; #248 split in two)* **Independent radar coordinates and event matching.** ▸ **DONE 2026-10-30 for the control (A) (M–L took M).** [`RecordLayer/MovingRadar.lean`](../CsdLean4/RecordLayer/MovingRadar.lean): the event = the moving observer's **reception write** of messenger `k`; the observer at rest charts it through #184's `RadarReadings` from **its own** emission and return writes for that messenger (`moving_radar`: the readings are those writes' `writeEnd`s — the matching); the moving observer by its clock at its reception write. ★★ `ExchGeom.radarTime_radar` (**radar time = bounce time + `radarOffset`**), `radarSep_radar`; ★ `receptionEnd_eq` (**detector response delay `1/(2w)`**); ★★★ `radar_at_rest` (**at relative rest the charts agree exactly**); ★★★ `radarTime_ne_of_speed_ne` (**the control's negative: same event, different messenger speeds, different radar times**); ★★ `abs_clockRadar_sub_le` (propagated clock errors), coverage = horizons as hypotheses; ★★ `train_radar` (charts disagree by `35/48 + k/6`); ★★★ `moving_radar`. Residue: the moving observer's radar chart of events at the observer at rest → #257; (B)'s charts folded into #256. ▸ *Original row:* For a reflection event each observer uses its own clock: `T = (τ_emit + τ_return)/2`, `D = (τ_return − τ_emit)/2`, `D` in clock units first, a length calibration introduced separately. **Amendment:** an *event* is one physical record write (`ActualOccurrence`, #191), and matching means both observers' radar coordinates refer to that write. **Acceptance:** both coordinate assignments refer to shared physical events, with propagated error bounds and explicit coverage conditions. ⚠️ **Stuck on:** for (A), nothing — ~~#244~~ gives each observer its emission and return timestamps over repeated exchanges (`moving_exchange`, `exchange_train`); the messenger has no universal speed, so the charts depend on its preparation — the control's expected negative; for (B), #256. | ~~**M–L**~~ done (M) | High — two coordinate charts on one set of events | ~~#244~~; #256; ~~#184~~; ~~#200~~; ~~#191~~ | > | ~~246~~ | Records | Records→spacetime | *new (2026-10-30, the author's fourth operational-geometry brief — **two physical observers in relative motion recover a common 1+1-dimensional flat spacetime description from their own records; GR is the subsequent programme**; split into #241–#248, recorded together, with four amendments from the review: the transport now in the corpus is scheduled and so excluded by the brief's gate; the dispersion class is chosen in #241 because it decides #247; #246 compares the Lieb–Robinson bound with the measured group velocity; #248 split in two)* **Is there a common signal cone?** ▸ **DONE 2026-10-31 for the control (A), and the measured propagation of (B) priced against the bound (L took M).** [`RecordLayer/SignalCone.lean`](../CsdLean4/RecordLayer/SignalCone.lean): **(A) has no cone** — ★★★ `admissible_speed_unbounded` (through every event, messengers faster than any proposed limit are admissible); ★★ `radarTime_through` + `radarOffset_eq` (the chart of one event is `τ + (2Dy + w)/(2(w² − y²))`), ★★ `tendsto_radarOffset` (**the event's own time appears only as `w → ∞`: Galilean absolute simultaneity as a limit**), ★★ `radarOffset_pos`; the speeds the records measure (★★ `emissionWrite_duration`, `returnWrite_duration`, `receptionWrite_duration`: `y + w` out, `w − y` back, `w` at the moving observer), ★★★ `half_anisotropy` (**the relative velocity is read off any messenger**), ★★ `mean_speed` (Galilean addition), ★★ `outSpeed_ne_retSpeed` (no speed common to both observers). **(B)**: ★★★ `abs_groupVel_le_maxGroupVel` + `exists_groupVel_eq_maxGroupVel` (**the fastest packet of species `μ` moves at exactly `√(μ² + (2J)²) − μ`**, the band's width), ★★★ `maxGroupVel_lt` (**a massive species never reaches `c = 2|J|`**; on the lattice each species has its own cap, the preferred frame again), ★★ `exists_maxGroupVel_gt` (`c` the supremum over species), ★★★ `twentyOne_mul_maxGroupVel_lt` (**#213's local bound is `8e ≈ 21.7` times `c`**: the bound does not supply the chain's `c`, the band does); ★★★ `signal_cone`. Residue: the Dirac chain's own light-cone bound → #258; whether (B)'s `c` is shared by moving observers stays with #256. ▸ *Original row:* Compare propagation across directions, observer motions and more than one admissible signal preparation. Separate the Lieb–Robinson upper bound (`localLRVelocity = 8eΔγ`, ~~#213~~), the **measured** propagation — **amendment:** the largest group velocity over the admissible preparations (about `2J` on a hopping chain, far below the bound) — and any limiting speed shared across observers. **Acceptance:** agreement proved within a controlled regime, or the dependence on observer velocity, signal construction or microscopic network quantified. A propagation bound alone does not complete the row. ⚠️ **Stuck on:** for (A), the answer is in hand and needs packaging — ~~#245~~'s `radarOffset` quantifies the chart's dependence on the messenger's speed and the observer's velocity, and `radarTime_ne_of_speed_ne` shows no preparation-independent chart; remaining: the comparison with ~~#213~~'s Lieb–Robinson bound and the chain's measured group velocity; for (B), #256. | ~~**L**~~ done (M) | High — whether a `c` is physical or only a bound | ~~#245~~; #256; ~~#213~~; ~~#205~~ | > | ~~247~~ | Records | Records→spacetime | *new (2026-10-30, the author's fourth operational-geometry brief — **two physical observers in relative motion recover a common 1+1-dimensional flat spacetime description from their own records; GR is the subsequent programme**; split into #241–#248, recorded together, with four amendments from the review: the transport now in the corpus is scheduled and so excluded by the brief's gate; the dispersion class is chosen in #241 because it decides #247; #246 compares the Lieb–Robinson bound with the measured group velocity; #248 split in two)* **The coordinate transformation: Lorentz, Galilean or other.** ▸ **DONE 2026-10-31 for the control (A): Galilean, derived; Lorentz refuted for every `c` (L–XL took M).** [`RecordLayer/GalileanTransformation.lean`](../CsdLean4/RecordLayer/GalileanTransformation.lean): the **transit chart** from one exchange's four record times (each write's duration = one cell over the messenger's speed past that detector) — ★★★ `ExchGeom.transitChart_eq` (**the chart is the event itself, exactly**), ★★★ `transitChart_indep` (**whatever the messenger** — unlike #245's two-way radar); `y` and `d` read off the records (`relVel_of_writes`, `sep_of_chart`); ★★★ `galileo_first` / `galileo_second` (**one map `(t, x) ↦ (t, x − d − yt)` sends the observer at rest's chart of every matched event of both families to the moving observer's**); ★★★ `lorentz_refuted` (a clock within `ε` of the dynamics' time — #243 — refutes the dilation `τ√(1 − y²/c²)` once `τ(1 − √(1 − y²/c²)) > 2ε`, for every `c`); ★★★ `minkowski_ne` (**no interval shared**: they differ by `y²Δt²`); ★★ `train_galilean`; ★★★ `galilean_transformation`. **Finding:** in the control, simultaneity is not a convention — one-way speeds are measured locally — which is what (B) must fail to reproduce if it is Lorentzian. Derived and assumed kept apart in the header (assumed: the unit cell as length standard, the messenger's uniform speed between cell and mirror, matching by messenger label). Residue: (B)'s transformation → #259. ▸ *Original row:* From the matched event records determine the transformation between the two observers' coordinates; test whether a specified regime yields `T_B ≈ γ(T_A − vX_A/c²)`, `X_B ≈ γ(X_A − vT_A)`; then test actual clock readings for time dilation and the reconstructed interval for observer agreement. Derived from the device dynamics and operational definitions; an assumed symmetry kept distinct from a derived one; **a different transformation (a preferred frame) proved and reported** — the expected outcome for #241's control (A). ⚠️ **Stuck on:** for (A), nothing — ~~#246~~ settled the cone question (no cone; the relative velocity is read off any messenger, `half_anisotropy`; the event's own time is the infinite-speed limit of every chart); for (B), #256. The outcome is set by #241's dispersion class. For (A) the inputs are in: ~~#243~~ (the clocks agree whatever the motion), ~~#245~~ and ~~#257~~ (both radar charts, reciprocal, messenger-dependent in motion). | ~~**L–XL**~~ done for (A) (M) | High — the brief's central question | ~~#246~~; #241; ~~#257~~; #256; #39 | > | ~~248~~ | Records | Records→spacetime | ▸ **DONE 2026-11-03 — (a) for both chains and (b); the row closes.** (a) for (B) is ~~#280~~ (`radar_lorentz_scaling`). **(b):** [`RecordLayer/ApparatusIndependence.lean`](../CsdLean4/RecordLayer/ApparatusIndependence.lean) — the class is every apparatus (detector and clock) whose readings of an exchange's events lie within a stated error of the events' ideal times; the record layer's threshold detectors belong at any rate, chord and threshold (`abs_record_sub_le`, `encounter_record_window`). ★★★ `galilean_apparatus_indep` (**control: two apparatus within `ε₁`, `ε₂` give transit charts, velocity and separation estimates within `err(ε₁) + err(ε₂)` of each other**); ★★ `radar_velocity_exact` (**`B`'s radar reads `v` off its own records: radar distance = `v` × radar time before the meeting**), ★★★ `radar_velocity_error` (`v̂` within `3d/(T − 2d)`), ★★★ `radar_apparatus_indep` (**Dirac chain: two apparatus give velocity estimates within `3d₁/(T − 2d₁) + 3d₂/(T − 2d₂)` of each other and each its chart within its `radar_lorentz_error` bound of the Lorentz boost at the true `v`**). **Finding:** the effective geometry (velocity and map) is the events', not the apparatus' — every apparatus in the class gives it to within its own resolution, and the differences vanish with the resolution. Scope: the class is defined by reading errors (apparatus with unbounded lag or unknown rate are outside it); a clock's continuous membership is #283; the Dirac statements are kinematic. *new (2026-10-30, the author's fourth operational-geometry brief — **two physical observers in relative motion recover a common 1+1-dimensional flat spacetime description from their own records; GR is the subsequent programme**; split into #241–#248, recorded together, with four amendments from the review: the transport now in the corpus is scheduled and so excluded by the brief's gate; the dispersion class is chosen in #241 because it decides #247; #246 compares the Lieb–Robinson bound with the measured group velocity; #248 split in two)* **Scaling limit (a).** ▸ **(a) for the control (A) DONE 2026-10-31: the common chart is stable under finite clock readings.** [`RecordLayer/GalileanTransformation.lean`](../CsdLean4/RecordLayer/GalileanTransformation.lean): `ReadingsNear` (four readings within `ε` of an exchange's record times, writes lasting `≥ δ ≥ 4ε`, emission end and return start in `[−H, H]`); ★★★ `ReadingsNear.abs_chartTime_sub_le` (**transit-chart time within `ε(3 + 4H/δ)`**), ★★★ `abs_chartSep_sub_le` (**position within `(2ε/δ)(1 + 2H/δ)`**), ★★★ `abs_relVelEst_sub_le` (**`relVel_of_writes`' estimator within `4ε/δ²`**), ★★ `abs_sepEst_sub_le`; ★★★ `galileo_first_stable` / `galileo_second_stable` (**the map with the estimated `ŷ`, `d̂` sends each family's imperfect chart to within explicit bounds of the other's** — the discrepancy bounded from the exact `galileo_first`/`galileo_second`, no transformation assumed of the perturbed charts); ★★ `train_stable` (`y = 1`, `ε ≤ 1/16`: time `ε(71 + 32k)`, position `8ε(35 + 16k)`, velocity `64ε`, separation `ε(978 + 448k)`); ★★ `ExchGeom.readingsNear_hand`, ★★★ `galilean_transformation_clocks` (**the same with `ε = handClockErr` of each observer's own clock over its horizon**). **Finding:** every error is linear in the clock error at fixed `δ`, `H`; far reflections and fast messengers amplify it (`H/δ`, `H/δ²`); calibration, matching and free propagation stay exact assumptions. Residue: a concrete clock meeting `4 · handClockErr ≤ δ` → #264. ▸ *Original row:* A family of models with increasing network size and improving apparatus resolution: prove which errors decrease, which remain, and a quantitative convergence or stability theorem for the effective geometry. **(b)**, split from (a) at the review: vary the detector and clock construction within a stated class and prove the same effective geometry survives. ⚠️ **Stuck on:** *done 2026-11-03 — was:* for (A), nothing — (a) done 2026-10-31 (errors linear in `handClockErr`), (b) is `transitChart_indep` for messengers, so varying detectors and clocks is the open part; for (B), ~~#278~~ done 2026-11-03 (the Lorentz map within `≈ 23%`, `dirac_two_charts`); the scaling of its errors with resolution is ~~#280~~, done 2026-11-03 (`radar_lorentz_scaling`). (a) and (b) are separable. | ~~**XL**~~ done (across #247–#280 and this row's (b), S) | Medium — the claim that the geometry is effective, not an artefact of one apparatus | ~~#247~~; #259; #39 | > | ~~249~~ | Records | Records→spacetime | *new (2026-10-30, residue of ~~#241~~)* **The test candidate (B): a two-band chain near its crossing.** ▸ **DONE 2026-10-30 (L took L), with a correction.** ▸ *Precision 2026-10-31 (external review):* the clock statements are about the **defined** velocity `v_rel = cs/E` and bound the normalised rate within `(μ₁ − μ₀)/E₀` of `√(1 − (v_rel/c)²)`, not equal to one exact factor; against the measured drift velocity the lattice defect enters, and the controlled limit at nonzero physical momentum is ~~#260~~. ▸ The row as opened named the oscillation **between the two bands** as the clock; that beat runs at `2E/ħ`, which *grows* with momentum — the opposite of dilation. The right clock is an **internal two-level device** of rest energies `μ₀ < μ₁` carried by the mode (an atomic clock), its beat `E_{μ₁}(k) − E_{μ₀}(k)`. [`CV/DiracClockChain.lean`](../CsdLean4/CV/DiracClockChain.lean) (34 pins): ★★ `diracHam` (**the lattice Dirac Hamiltonian on a ring**, `−iJ(|j⟩⟨j+1| − |j⟩⟨j−1|) ⊗ σ_x + μσ_z`, local, ★ `diracHam_isHermitian`), `planeWave` (via `ZMod.toCircle`), ★★ `diracHam_mulVec_diracMode` (**exact eigenmodes**, `E = √(μ² + (2J sin θ)²)`), ★ `hasDerivAt_diracMode` (Schrödinger's equation solved by the phase), ★★ `clock_phase`; ★ `div_dispE_eq_sqrt` (**`μ/E = √(1 − (v/c)²)` exactly**, `c = 2J`), ★★ `clockRate_mem` (the clock's rate between `Δμ` times its two levels' factors), ★★★ `abs_clockRate_div_sub_le` (**time dilation derived from the dispersion, within `Δμ/E₀`**); ★ `hasDerivAt_dispE_lattice` (**group velocity = relativistic velocity × `cos θ`**), `abs_groupVel_le` (none above `c`), ★★ `lattice_dilation_defect` (**the lattice's preferred frame**: squared dilation factors from the two velocities differ by `(v/c)² sin² θ ≤ sin² θ`, vanishing as `θ → 0`); ★★★ `dirac_clock_chain`. **Finding:** relative to the mode's relativistic velocity the clock is dilated by exactly the relativistic factor (up to the level splitting); relative to the measured group velocity it is dilated by more, an `O(k²)` preferred-frame effect. Residue: a localised packet at the group velocity with a spreading bound and a position read from records → #250. ▸ *Original row:* ~~#241~~'s control is Galilean by construction (its clock and its motion are decoupled). The candidate in which time dilation can emerge is a two-band lattice Hamiltonian — a Dirac-type chain, `H = Σ_k (J σ⁺_k σ⁻_{k+1} + h.c.) ⊗ τ_x + μ τ_z` or its staggered form — near its band crossing, where the dispersion is linear with speed `c = 2Ja`: a wave packet built from momenta near the crossing travels at its group velocity with bounded spreading over a finite horizon, and the oscillation between the two bands (rate `∝ μ`) is a clock carried by the packet. Deliverables: the chain as a field-structured flow with local terms (so ~~#213~~'s local Lieb–Robinson bound applies); a packet whose position expectation moves at `v_g(k₀)` with a proved spreading bound `Δx(t) ≤ Δx(0) + t·Δv` over the horizon; the band oscillation's frequency along the packet, and the comparison with the packet at rest — `ω(k₀) = √(μ² + (c k₀)²)/ħ` against `μ/ħ` is the lattice's time-dilation factor `γ(v_g)` at leading order, the deviation at finite lattice spacing quantified. ⚠️ **Stuck on:** a wave-packet spreading bound for a two-band chain, which the corpus does not have; the dispersion relation itself is a `2 × 2` Fourier computation. | ~~**L**~~ done (L) | High — the candidate where the brief's relativistic question has content | ~~#241~~; #242; ~~#213~~; ~~#190~~; #247 | > | ~~250~~ | Records | Records→spacetime | *new (2026-10-30, residue of ~~#249~~)* **A localised device on the Dirac chain.** ▸ **DONE 2026-10-30 (L took L).** [`CV/DiracPacket.lean`](../CsdLean4/CV/DiracPacket.lean): a **packet** of upper-band modes on a window of integer momenta shorter than the ring, carrying the two-level clock; position read through the **translation phase** `T = ⟨ψ, e^{2πi x̂/L} ψ⟩` (circular position = `arg T · L/2π`, localisation = `|T|`). ★ `norm_sum_phase_sub_le`; ★ `abs_bandE_sub_le` (energy `2|J|`-Lipschitz), ★★ `abs_groupVel_sub_le` (**group velocity Lipschitz**, `C = 4J²/μ + 4|J|³/μ²`), ★★ `abs_bandE_step_sub_le`; ★ `sum_toCircle_mul`, ★★ `chainInner_planeWave` (**orthogonality on the ring**); ★ `hasDerivAt_packet` (Schrödinger), ★ `noWrap_distinct_of_window`; ★★ `transPhase_packet` (**`T(t)` exactly**, phases at `E_{m+1} − E_m`); ★★★ `transPhase_drift` (**the packet moves at `v_g(θ₀)`** within `|t|` × drift error), ★★ `norm_transPhase_ge` (**it stays localised**); ★★ `clockOverlap_eq`, ★★★ `clockOverlap_rate` (**the carried clock ticks at the centre's dilated rate** within `|t|` × the window's spread); ★★★ `dirac_packet`. ▸ *Precision 2026-10-31 (external review):* `clockOverlap_rate` bounds the complex coherence's rotation at the centre's **level splitting**, which is within `(μ₁ − μ₀)/E₀` of the dilation factor of `v_rel`, not of the drift velocity; it fixes a phase only while the coherence stays away from zero; `NoWrap`/`Distinct` prevent aliasing, not delocalisation; the transport mass `μ` is free of `μ₀`, `μ₁`; and the error terms are not shown to vanish under scaling (the `|J|³/μ²` term of `vgLip` grows as `a⁻³` under `J = c/(2a)` — **removed 2026-10-31**, #261's first brick: `vgLip = 4J²/μ`) — #261, #262. Not done here: the position read by **detectors on the sites** (moved to #242, which now has a localised clock-carrying device for both candidates); no explicit Gaussian instance. ▸ *Original row:* ~~#249~~ proves the internal clock's dilation on exact plane-wave modes, which spread over the ring. Deliverables: a wave packet of upper-band modes near momentum `θ₀` (a Gaussian or a finite window of modes) carrying the two-level clock; its position expectation on the ring moving at the group velocity `v_g(θ₀)` (`hasDerivAt_dispE_lattice`) up to a bound in the packet's momentum width; a spreading bound `Δx(t) ≤ Δx(0) + t · max |v_g'|` over a finite horizon; the clock's phase read on the packet, with its rate within the momentum width of `levelClockRate` at `θ₀`; the position read from local records (a detector per site, as in ~~#241~~). ⚠️ **Stuck on:** nothing identified; the packet is a finite sum of `diracMode`s, so its evolution is exact, and the bounds are Taylor estimates of `dispE` in `θ` — the price is the position operator on a ring (an angle, not a line) and the bookkeeping. | ~~**L**~~ done (L) | High — the moving device of the brief's step 2 with a relativistic dispersion | ~~#249~~; ~~#241~~; #242 | > | ~~251~~ | Records | Records→spacetime | *new (2026-10-30, residue of ~~#242~~)* **Detectors generated, with back-reaction.** ▸ **DONE 2026-11-04 (M) — the acceptance; measure preservation re-priced to #291.** [`RecordLayer/CoupledDetector.lean`](../CsdLean4/RecordLayer/CoupledDetector.lean): `DetProfile` (a `C¹` bump in `[0, 1]`, `0 ≤ χ ≤ 1`, Lipschitz with Lipschitz derivative); `detHam` (`H = p²/(2m) + g Σ_j χ(x − j) η_j`), `detField`, ★★ `hasDerivAt_detHam_mom`, `hasDerivAt_detHam_pos` (**the field is Hamilton's**), ★★ `lipschitz_detField`, `detFlow` (`LipschitzWith.odeFlow`), `measurable_detFlow`, `detFlow_add`; ★★★ `detHam_flow` (**energy conserved**); **the disturbance bound** (`S = Σ_j |η_j|`): ★★★ `abs_mom_sq_sub_le` (**`|p(t)² − p₀²| ≤ 2 m g S` at every time — no growth with time**), ★★ `mom_pos` (no turning back when `p₀² > 2mgS`), ★★★ `mom_mem_Icc` (**`√(p₀² − 2mgS) ≤ p(t) ≤ √(p₀² + 2mgS)`**), ★★ `pos_ge`, `pos_le`; the register: ★★ `reg_eq_of_before` (nothing written before the fastest allowed arrival), ★★★ `reg_ge_of_after` (**after the crossing the register has advanced by at least `g m A / p_hi`**, by comparison with `∫_k^x χ(u − k) du`); ★★★ `coupledPassageWrite` (**the passage as an `ActualOccurrence` of the generated flow**, threshold `0 < θ ≤ g m A / p_hi`, write interval `[(k − x₀)/(p_hi/m), (k + 1 − x₀)/(p_lo/m)]`). Scope: `η` (conserved) is a parameter of each preparation, the arena `(x, p, ζ)`; the hand clock and memory are dropped; `η = 0` recovers #242's undisturbed device. Replace #241's test detectors by a Hamiltonian coupling `g_d χ_k(x) η_k` with a smooth profile `χ_k` localised on site `k`, the register `η_k` conjugate to its readout. Prove: the coupled flow on the arena (device ⊕ registers) is the Hamiltonian flow of the sum, measure preserving; the register advances by `g_d ∫ χ_k(x(s)) ds`; the device's momentum is kicked by `−g_d χ_k'(x) η_k`, with a **disturbance bound** on its velocity and position over the horizon, in terms of `g_d`, the register's initial spread and `‖χ_k'‖`; the passage is still an `ActualOccurrence` under the bound. **Acceptance:** `passageWrite` restated on the generated flow with the disturbance bounded. ⚠️ **Stuck on:** *done 2026-11-04 — was:* the smooth profile and the coupled ODE. | ~~**M**~~ done (M) | Medium — closes the control's disturbance item: a detector that reads must push | ~~#242~~; ~~#241~~; ~~#195~~; ~~#196~~ | > | ~~291~~ | Records | Records→spacetime | *new (2026-11-04, residue of ~~#251~~)* **The coupled detector flow preserves Liouville measure.** ▸ **DONE 2026-11-04 (L took M).** [`RecordLayer/CoupledLiouville.lean`](../CsdLean4/RecordLayer/CoupledLiouville.lean): ★★ `detFlow_eq_skew` (**the coupled flow is a skew product**: the device by the planar flow of `(p/m, F(x))`, the registers shifted by an amount fixed by the device's start); for a `C²` profile, ★ `contDiff_planarField`, ★★★ `exists_hasFDerivAt_planarFlow` (**the planar flow is differentiable in the start with determinant one** — the variational solution from `hasFDerivAt_flow_of_variational`, its 2×2 determinant constant because the field is trace-free), ★★ `volume_image_planarFlow` (Mathlib's `lintegral_abs_det_fderiv_eq_addHaar_image`), ★★★ `measurePreserving_planarFlow`; ★★★ `measurePreserving_detFlow`, ★★★ `measurePreserving_coupledEvolve` (**the coupled detector flow preserves Lebesgue measure on `(x, p, ζ)` at every time** — `MeasurePreserving.skew_product` with the register shear). The anticipated wall (no planar area preservation in usable form) fell to the corpus's variational machinery plus a direct determinant computation; the flat Liouville theorem was not needed. Scope: `C²` profile as a hypothesis (`d2`, continuous); `η` a parameter, so Lebesgue on `(x, p, ζ)` per preparation (Fubini over `η` not written out). ~~#251~~ generated the detectors by `H = p²/(2m) + g Σ_j χ(x − j) η_j` and bounded the back-reaction, but did not prove the flow measure preserving. **Build:** on the full arena `(x, p, ζ, η)` (with `η` a coordinate, so the field is only locally Lipschitz — a flow on level sets of the conserved `η`, glued), Lebesgue measure is preserved: the `ζ` part is a shear (translation by an amount depending on `(x, p, η)`, preserved by Fubini), and the `(x, p)` part is a one-degree-of-freedom Hamiltonian flow, area preserving — through `FlowDerivative.lean`'s flat Liouville theorem for the area 2-form plus a change of variables (`det D = 1`), or a direct Jacobian argument. ⚠️ **Stuck on:** *done 2026-11-04 — was:* planar area preservation. | ~~**L**~~ done (M) | Medium — completes #251's "the coupled flow is the Hamiltonian flow of the sum, measure preserving" | ~~#251~~; ~~#195~~; `FlowDerivative.lean` | > | ~~252~~ | Records | Records→spacetime | *new (2026-10-30, residue of ~~#242~~)* **The Dirac packet read by detectors.** ▸ **DONE 2026-11-02 (closed at the repo scan: the acceptance is met by later rows).** ~~#253~~'s `dirac_two_observers` (`RecordLayer/DiracTwin.lean`) is one evolution that carries a device moving at `4/5` and writes its departure, return and ticks as actual occurrences; ~~#271~~'s `dirac_ticks` and ~~#272~~'s `dirac_device_ticks` write the ticks of a localized device on the evolution that carries it. Of the residues named below, ~~#265~~ and ~~#262~~ are done and #266 (one click) stays its own row. ▸ **Smallest part DONE 2026-10-31: one packet writes one site detector.** [`RecordLayer/DiracDetector.lean`](../CsdLean4/RecordLayer/DiracDetector.lean): `DetArena` (the window's amplitudes × a register at every site), `detFlow` (amplitudes advanced by their phases — ★ `packet_phaseAmp`, the chain's Dirac evolution on the window — and the register at `k` integrating the packet's site weight `Σ_c ‖ψ(k, c)‖²` with gain `g`), `detEvolve` two-time, ★ `measurable_detFlow`; ★ `siteWeight_packet_le` (a site holds at most the norm), ★★ `register_lt_of_lt` (**no detector is written before `θ / (g ⟨ψ, ψ⟩)`**); ★★★ `packetDetector` (**the passage through site `k` as an `ActualOccurrence`** of that evolution, written once the register reaches the threshold, held for any window); the instance `twoMode` (momenta `0`, `1` on `12` sites, `J = 4`, `μ = 3`): ★★ `siteWeight_twoMode` (weight `116 + 96 cos(πk/6 − 2t)`), ★★ `siteWeight_twoMode_succ` (one site every `π/12`, velocity `12/π < c = 2J = 8`), ★★★ `dirac_detector` (**the arena carries the packet, which solves the lattice Dirac equation; the detector at site `0`, threshold `g (29π + 48)`, is written at exactly `π/4`, while the register at the opposite site `6` is then `g (29π − 48)`, below it**). **Findings:** the candidate's packet writes records on its own arena; a packet of finite norm writes nothing before `θ / (g ⟨ψ, ψ⟩)`; where the record is written depends on where the packet is. Scope: the arena is the window (an invariant subspace), not the whole chain; the detector law is stated (test detectors, no back-reaction — #251); every register rises with its site's Born weight (an intensity, not one click); the instance is a two-mode beat spanning the ring, the written detector the one at the starting peak. Residue: the whole-chain arena and a distant detector written on arrival → ~~#265~~ (done 2026-10-31, `dirac_arrival`); one click selected by the record layer → #266; the clock readout → #262. ▸ *Original row:* For the test candidate (B): the packet of ~~#250~~ and its two-level clock read by **records** — a site detector whose register advances with the packet's weight on site `k` (or a two-outcome readout of the clock's levels) — as `ActualOccurrence`s of one measurable evolution of an arena (the projective state of the chain, or its Born-weighted record layer), with a **concrete packet** (explicit window and amplitudes) whose drift and clock errors are evaluated. **Acceptance:** #242's acceptance for (B): the same evolution transports the packet and writes its ticks, with a concrete nonzero-motion instance. ⚠️ **Stuck on:** *done 2026-11-02 — was:* for the smallest part, nothing — done 2026-10-31 (the arena is the window's amplitudes with integrating registers); what remains is #265, #266 and the clock readout (#262). Originally: the passage from amplitudes to records — the readout of a quantum device is a record write only through the record layer (#185's ticks are on the ontic torus, not on the chain), so the arena must carry both; the price is choosing that arena. | ~~**L**~~ done | High — the brief's acceptance for the candidate that can give a cone | ~~#250~~; ~~#249~~; ~~#242~~; ~~#185~~; ~~#191~~ | > | ~~253~~ | Records | Records→spacetime | *new (2026-10-30, residue of ~~#243~~)* **Two observers on the Dirac chain.** ▸ **DONE 2026-11-01 (L took L).** [`RecordLayer/DiracObservers.lean`](../CsdLean4/RecordLayer/DiracObservers.lean) + [`RecordLayer/DiracTwin.lean`](../CsdLean4/RecordLayer/DiracTwin.lean): chord geometry on the circle (`norm_mul_star_sub_le`, `two_sin_le_norm_exp_sub_one`, `chordArc`); `devSiteWeight` and ★★ `devRegionWeight_compl_chordArc_le` (**where a device is, from both levels' budgets**); ★ `relWeight` (**the joint Born weight of one device at an offset window from the other**), ★ `relWeight_swap` (**mutual**), `le_relWeight`/`relWeight_le`; generic ★★ `regFlow`/`regDetector` (registers integrating any continuous weight along any flow, writes as actual occurrences). Instance: #262's device at rest and moving at `4/5` on a ring of `2π·10¹³` (`2·10⁶` modes each, splitting `10⁻¹¹`): ★★ `rest_tail`/`moving_tail` (within chord `1/25` but for `1.4·10⁻³` over the revolution), ★★★ `twinDeparture`/`twinReturn` (departure and, after one revolution, return as actual occurrences — the return window a fifth of a radian behind so its register has one lobe), ★★ `twin_moving_rate` (`R ∈ [6, 6.01]·10⁻¹²`), both clocks' ticks lifted; ★★★ `dirac_two_observers` (**one evolution; departure between ticks 0–1 of each clock, return between ticks 117–122 at rest and 70–73 moving; tick-count ratio in `[69/122, 73/116] ∋ 3/5 = √(1 − (4/5)²)`, excluding `1`; relative speed read by the observer at rest in a bracket ∋ `4/5`, below `1`**). **Finding:** the moving clock's slowing shows in the records; the asymmetry is the ring's rest frame (singled out on a ring in special relativity too). Lean lesson: higher-order unification of `Continuous.comp` against an expected type with `relWeight` times out — elaborate with `have h := …; exact h`. Scope (no new rows): detector law stated, test devices (#251); speed read with window centres; calibrated clocks (~~#273~~: re-counted by the rate counter, thresholds free of `R`); reciprocity needs signals between moving observers (#256, #270). ▸ *Original row:* For the test candidate (B): two packets of ~~#250~~ on one chain, each carrying its two-level clock, one at rest (window centred at `θ₀ = 0`) and one moving; their mutual detection as record writes (through #252's arena); the relative speed read from the records against the group velocity; the two clocks' rates compared — **the moving one's dilated by `√(1 − (v/c)²)`** where the control's agree (~~#243~~). **Acceptance:** #243's acceptance for (B), with the dilation visible in the comparison of the two observers' records. ⚠️ **Stuck on:** #252 (the arena carrying amplitudes and records); then the two packets' overlap while they pass, which a free chain does not keep zero. | ~~**L**~~ done | High — the first comparison in which relative motion shows in the clocks | ~~#243~~; #252; ~~#250~~; ~~#249~~ | > | 254 | Records | Records→spacetime | *new (2026-10-30, residue of ~~#244~~)* **An observer that launches its own signals.** In ~~#244~~ the messengers are prepared upstream and the observer records their departure. Here the observer launches: a messenger held at the observer (bound by a potential, or in a reservoir mode) is released by a coupling to the observer's clock — on its `n`-th tick — so the emission time is the observer's choice read on its own clock. Prove the launch is an `ActualOccurrence` of one autonomous flow, its time within the tick resolution, and the released messenger's state within a bound of ~~#244~~'s prepared one. **Acceptance:** `moving_exchange` restated with launched messengers. ⚠️ **Stuck on:** nothing — #212's coupling of a clock to a carrier is the pattern; the price is the release mechanism's flow. | **M–L** | Medium — makes the emission an act of the observer rather than a preparation | ~~#244~~; ~~#212~~; ~~#201~~ | > | 255 | Records | Records→spacetime | *new (2026-10-30, residue of ~~#244~~)* **Exchanges with recoil in the flow.** ~~#244~~'s moving observer is a rigid mirror; `elastic_exchange` gives a bounce's recoil exactly. Build the flow in which the observer of mass `m` recoils at each bounce (centre of mass free, relative coordinate unfolded per collision, collisions ordered in time) and bound the drift of its velocity and position after `n` exchanges by `n · 2(μ/m) w_max` and its time integral; restate the records and the timestamps with the drift in the budget. **Acceptance:** `moving_exchange` on the recoiling flow with the disturbance as a budget term. ⚠️ **Stuck on:** nothing for one messenger (two-body: centre of mass plus unfolded relative coordinate); the price for `N` is the ordered-collision flow. | **L** | Medium — closes the back-reaction item of #244 and #243's disturbance term for signals | ~~#244~~; #251; ~~#243~~ | > | ~~256~~ | Records | Records→spacetime | *new (2026-10-30, residue of ~~#244~~)* **Exchanges on the Dirac chain.** ▸ **DONE 2026-11-01 one way (re-scoped at the user's call; the two-way radar version is ~~#274~~, done 2026-11-02: `dirac_radar`).** [`RecordLayer/DiracDoppler.lean`](../CsdLean4/RecordLayer/DiracDoppler.lean): ★★ `levelCoherence_shift_sharp` (the clock drifts from any start at the sharp rate), messengers as one-level devices, encounter windows about moving centres; ★★ `approach_rate` (`B` at `3/5`: splitting `∈ [8, 8.002]·10⁻¹²`, four fifths of the rest one), `approachTick` (ticks from `B`'s shifted start); ★★★ `dopEmission₁/₂` (passages at `A`), `dopReception₁/₂` (arrivals at `B`, relational), both clocks' ticks — one evolution; ★★★ `dirac_doppler` (**`B` counts 21–26 ticks between arrivals, `A` 49–53 between emissions: ratio `∈ [21/53, 26/49] ∋ 16/35 = v√(1 − β²)/(v + β)`, excluding the Galilean `4/7`**). **Finding:** the Doppler factor read off the records is the relativistic one — compression times dilation. Scope: one way; prepared emissions (#254); messengers at `v = 4/5 < c` (the `v → c` limit `√((1 − β)/(1 + β))` lies in the bracket but is not reached on the chain, ~~#246~~); test detectors (#251, #266; the clocks re-counted by ~~#273~~'s rate counter). No new rows. ▸ ▸ *2026-11-01:* ~~#253~~ done (two observers, dilation in the records) and ~~#270~~ done at rest (reflection by images); the moving return is #274. For the test candidate (B): signals between two packet-observers (#253) carried by chain excitations near the band crossing, whose speed is bounded by `c = 2J`; emission, reception and return as record writes; the timestamps; repeated exchanges. **Acceptance:** #244's and ~~#245~~'s acceptance for (B) — the exchanges and their radar charts matched to record writes — with the return interval over the emission interval the relativistic Doppler factor `√((1 + β)/(1 − β))²` within the packets' errors. ⚠️ **Stuck on:** split 2026-10-31 at the user's call into ~~#269~~ (one-way matched signal, done 2026-10-31: `dirac_signal`, speed `[144/201, 72/79] ∋ 4/5` below `c`), #270 (a return leg — the free chain sends nothing back), #253 (moving observers); the record arena (#252's smallest part, ~~#265~~, ~~#268~~) is in place. Originally: #253 (two observers on the chain), #252 (the record arena). ~~#246~~ fixed what to look for: every packet of every species is slower than `c = 2|J|` in the chain's frame (`maxGroupVel_lt`), so the question here is whether both observers' records read the same `c`. | ~~**L–XL**~~ done (one way, L) | High — the radar data from which (B)'s common cone could be read | ~~#244~~; #253; #252; ~~#249~~ | > | ~~257~~ | Records | Records→spacetime | *new (2026-10-30, residue of ~~#245~~)* **The moving observer's radar chart.** ▸ **DONE 2026-10-30 (M took M).** [`RecordLayer/MovingRadarPair.lean`](../CsdLean4/RecordLayer/MovingRadarPair.lean): a second family of messengers reflecting off the observer at rest, read by the moving one, carried by #244's own flow through a **mirror view** (parity: same `d`, same `y`); ★★ `ActualOccurrence.pullback` (**a write survives a projection commuting with the dynamics** — generic), so every #244 write is a write of the joint evolution (`liftFirst`, `liftSecond`); ★ `mirror_window` (the mirror's physical reading); ★★ `biFlow`, `mirrorView_biFlow`, `biFlow_add`, ★ `measurable_biFlow`; ★★★ `radar_reciprocity` (**both charts by the same formula**); ★★ `biTrain_charts` (identical charts on the train); ★★★ `two_radar_charts`. No residue: relating the two charts is #247's work. ▸ *Original row:* In ~~#244~~/~~#245~~ only the observer at rest charts events at a distance. Mirror the construction: messengers reflecting off the observer at rest, recorded on emission and return by the moving observer, charted on its own clock; match its chart and the first observer's on shared record writes (each observer's reflections). **Acceptance:** both observers chart the other's reflection events by radar, with propagated errors and coverage; the two charts' relation on those events computed (in the control, the Galilean relation with `radarOffset`'s dependence on the messenger). ⚠️ **Stuck on:** nothing — the exchange flow is Galilean-covariant, so the mirrored geometry (messenger approaching from the far side of the observer it reflects off) is ~~#244~~'s with the roles swapped; the price is restating `ExchGeom` for a mirror on the left. | ~~**M**~~ done (M) | High — two radar charts on one set of events, the input to #247's transformation | ~~#245~~; ~~#244~~; #247 | > | ~~258~~ | Records | Records→spacetime | *new (2026-10-31, residue of ~~#246~~)* **The Dirac chain's own light-cone bound.** ▸ **DONE 2026-11-02 (S took S; the bound itself with ~~#268~~).** [`RecordLayer/DiracTiming.lean`](../CsdLean4/RecordLayer/DiracTiming.lean): `chainConeVel` (`4e(2|J| + |μ|)` sites per unit time), `siteWeight_chainProp_le_of_cone`; ★★ `ten_mul_maxGroupVel_lt_chainConeVel` (**the cone exceeds ten times the fastest packet `maxGroupVel J μ`**, `J ≠ 0`, `μ ≥ 0`), ★ `exists_chainConeVel_lt_eleven_mul` (**for a light species less than eleven times** — the slack is the factor `4e`), ★ `chainConeVel_le_localLRVelocity_iff` (**at most #213's local Lieb–Robinson velocity at the same edge strength exactly when `|μ| ≤ 2|J|`**); `dirac_timing` strengthened in place with both comparisons. **Finding:** the chain's own cone and #213's evaluated bound are the same order (`8e|J| + 4e|μ|` against `16e|J|`); neither is near the packets' `< c = 2|J|` — the cones are what the dynamics guarantees for every state, the packets are what propagates. Scope: the cone is the exponential series' bound, not an optimised Lieb–Robinson velocity; lattice units (`ħ = 1`, spacing `1`). No new rows. ▸ *2026-11-02, repo scan — re-priced M → S:* the bound itself landed with ~~#268~~ ([`RecordLayer/DiracTiming.lean`](../CsdLean4/RecordLayer/DiracTiming.lean)): `VanishesNear.diracHam_mulVec` (the Hamiltonian moves a support by one site), ★★ `norm_chainProp_mulVec_le` (**amplitude at distance `D` at most `(h|t|)^D/D! · e^{h|t|} · sup|ψ|`, `h = 2|J| + |μ|`**), ★★ `siteWeight_chainProp_le` (weight below `2B²8^{−D}` while `4e h|t| ≤ D`: a cone of `4e(2|J| + |μ|)` sites per unit time). What remains is the comparison with `maxGroupVel` as a theorem (the docstring states it in prose: the cone is about `11c`, the packets below `c`). ~~#246~~ compares the Dirac chain's fastest packet with #213's local Lieb–Robinson velocity **evaluated** at degree `2` and edge strength `|J|`; #213 proves its bound for field-structured flows, and the Dirac chain is a one-particle operator on `ZMod L × Fin 2`. Prove the chain's own bound: the edge terms of `diracHam` have norm `|J|`; `(H^n)_{pq} = 0` below the ring distance (from `diracHam_apply_eq_zero`); hence `‖exp(−itH)_{pq}‖ ≤ (v t)^d / d!`-type suppression outside `d ≥ v_LR t`, with `v_LR` explicit; then restate `twentyOne_mul_maxGroupVel_lt` against it. **Acceptance:** a light-cone bound proved for `diracHam` itself and compared with `maxGroupVel`. ⚠️ **Stuck on:** *done 2026-11-02 — was:* nothing — the support-of-powers argument is finite matrix algebra; the price is the ring distance on `ZMod L` and the series bound. | ~~**M**~~ ~~**S**~~ done (S, the bound with ~~#268~~) | Medium — makes #246's bound-against-propagation comparison a theorem about one chain | ~~#246~~; ~~#213~~; ~~#249~~ | > | ~~259~~ | Records | Records→spacetime | *new (2026-10-31, residue of ~~#247~~)* **The coordinate transformation for the test chain (B).** ▸ **DONE 2026-11-02, first brick (re-scoped at the user's call; residues #277, #278, #279).** Scan before building: `B` sends no messengers, so it charts only the events on its own worldline (the reception — its relay firing — and the meeting), and on those the whole-tick readings tie Lorentz with Galilean. [`RecordLayer/DiracWorldline.lean`](../CsdLean4/RecordLayer/DiracWorldline.lean): `radarTime`, `radarPos` (`A`'s radar chart, messenger at `4/5` both ways), `minkowskiSq`; ★★ `radar_interval_bounds` (**for readings in the records' brackets the meeting is `36`–`46` periods after the reception in `A`'s chart, the reception `23.2`–`27.2` light-periods away, the squared Minkowski interval in `[13904/25, 39444/25]`, and `B`'s measured interval `30`–`36` squares into it; the Galilean prediction `≥ 36` ties with the measurement `≤ 36`**); ★★ `clock_rate_lorentz` (**`B`'s clock runs at `4/5`–`0.8002` of `A`'s, the Lorentz dilation of its worldline at `3/5` is exactly `4/5` (`lorentzBoost_worldline`), and the Galilean `1` is excluded**); ★★★ `dirac_lorentz_worldline` (`dirac_radar`'s brackets, the chart and interval bounds, the rate). **Finding:** on the moving observer's worldline the transformation is Lorentz's at the level of the dynamics (the clock's rate is the worldline's dilation to `2.5·10⁻⁴`); the records read to whole ticks are consistent with it and do not exclude Galilean — the sure windows would (numerical estimate, not proved: Galilean `≥ 37.7`, measured `≤ 34.7`). ▸ *2026-11-02:* ~~#277~~ done — with two half-ticks of `A`'s clock the records exclude Galilean strictly (`≤ 36 < 36.75`), `dirac_lorentz_worldline` strengthened in place. Scope: `B`'s worldline only; the rate is a device fact (`approach_rate`), not a record. ~~#247~~ derived the control's transformation (Galilean) from the records. Repeat on the Dirac chain: from #256's matched exchange records between two packet-observers, build both charts, determine the map between them, and test `T_B ≈ γ(T_A − vX_A/c²)`, `X_B ≈ γ(X_A − vT_A)` with `c = 2|J|` in the band-crossing regime, the packets' clock readings for dilation (~~#249~~: internal clock dilated by `√(1 − v²/c²)` up to `O(k²)`), and the interval for agreement. Report the lattice's preferred-frame corrections quantitatively (~~#246~~: each massive species capped below `c`; ~~#249~~'s `lattice_dilation_defect`). The test of ~~#247~~'s finding: whether one-way speeds are locally measurable on the chain. **Acceptance:** the map proved within explicit errors in a stated regime, or its deviation from Lorentz quantified. ⚠️ **Stuck on:** *split 2026-11-02 — was unblocked 2026-11-02* — ~~#256~~ (one way, `dirac_doppler`), ~~#274~~ (two-way radar, `dirac_radar`), ~~#253~~ and ~~#252~~ are done; the price is the row itself (both charts from the radar records and the map between them). Was: #256 (exchanges and radar charts on the chain), behind #253 and #252. | ~~**L–XL**~~ first brick done (M); rest → #277–#279 | High — the brief's central question for the test candidate | ~~#247~~; #256; ~~#249~~; ~~#246~~; #39 | > | ~~260~~ | Records | Records→spacetime | *new (2026-10-31, out of an external review of ~~#249~~/~~#250~~: "the lattice dynamics and error estimates are proved; a nontrivial continuum limit is not")* **The controlled continuum limit for modes and the clock.** ▸ **DONE 2026-10-31 (M took M).** [`CV/DiracContinuum.lean`](../CsdLean4/CV/DiracContinuum.lean): with `ħ = 1`, `c > 0` and `μ > 0` held fixed, `J(a) = c/(2a)`, `θ = ap`, uniformly on `|p| ≤ P`: ★★★ `abs_scaledE_sub_le` (**`|E_a − √(μ² + c²p²)| ≤ c a² P³/6`**), ★★★ `abs_scaledV_sub_le` (**`|V_a − c²p/√(μ² + c²p²)| ≤ a²(cP²/2 + c²P³/(6μ))`**, `V_a` the packet's drift velocity in physical units), ★★★ `scaled_clock_dilation_sq_error_le` (**`|R_a² − (1 − (V_a/c)²)| ≤ 2Δ/μ + a²P²`** — the clock against the **measured** velocity, from `abs_clockRate_div_sub_le` and `lattice_dilation_defect`), ★★★ `scaled_limit` (uniform convergence as `a → 0⁺`, residue `2Δ/μ` kept), ★★ `scaled_clock_limit` (to zero only as `Δ → 0` too). The review's precision on ~~#249~~/~~#250~~'s docstrings applied in place (clock = level splitting within `(μ₁ − μ₀)/E₀` of the factor of `v_rel`; the coherence bound; the free transport mass). Residue: packet convergence → #261; one device and its phase → #262. *The row as asked:* add a module proving a uniform scaled mode-and-clock estimate at nonzero physical momentum, the clock compared with the drift velocity, the finite-level residue retained. | ~~**M**~~ done (M) | High — the first controlled relativistic limit in the corpus | ~~#249~~; ~~#250~~; #261; #262; #39 | > | ~~261~~ | Records | Records→spacetime | *new (2026-10-31, residue of ~~#260~~)* **Packet convergence in the scaled limit.** ▸ **DONE 2026-10-31 (M–L took M).** [`CV/DiracPacketLimit.lean`](../CsdLean4/CV/DiracPacketLimit.lean): ★★★ `spin_align` (**neighbouring upper-band spinors are nearly parallel**: `(1 − (s − t)²/(4μ²)) ‖u(s)‖‖u(t)‖ ≤ u(s)·u(t) ≤ ‖u(s)‖‖u(t)‖`, Lagrange's identity plus the `1/(2μ)`-Lipschitz slope `s/(μ + E)`); ★★ `chainInner_packet_self` (the packet's squared norm `Σ (|a_m| ‖mode_m‖)²`, every time); ★★★ `sum_norm_transWeight_le` (**`Σ ‖w_m‖ ≤ ⟨ψ, ψ⟩`**), ★★ `norm_transPhase_le`, ★★ `driftError_le`; `unitAmp`, ★★ `chainInner_packet_unitAmp` (normalised coefficients); ★★★ `re_transPhase_unitAmp_ge` (`Re T(0) ≥ (1 − (2πJ/(Lμ))²) Σ β_{m+1}β_m` for real `β ≥ 0`); ★★ `norm_mul_norm_div_sub_le` (close complex numbers have close phases, `‖w‖ ‖z/|z| − w/|w|‖ ≤ 2‖z − w‖`); on a ring of **fixed physical length `ℓ`** with `L` sites (`a = ℓ/L`, modes at physical momenta `2πm/ℓ` whatever `L`): ★★★ `dirac_packet_limit` (**normalised at every time, `|T(0)| ≤ 1`, the translation phase rotates at `k v(p₀)`, `v = c²p₀/√(μ² + c²p₀²)`, `k = 2π/ℓ`, within `|t| · limitErr`; the modulus drops by at most that; the phase — circular position times `k` — within `2|t| · limitErr/|T(0)|`**; `limitErr = k((c²/μ)(D + k) + a²(cP²/2 + c²P³/(6μ)))`, uniform in `a`), ★★ `tendsto_limitErr` (the lattice part vanishes as `a → 0`); ★★★ `ringPacket_localised` (**`|T(0)| ≥ (1 − (πc/(ℓμ))²) Σ β_{m+1}β_m`, independent of `L`**), ★★★ `flat_packet_localised` (the flat packet on `N < L` modes: normalised, `|T(0)| ≥ (1 − (πc/(ℓμ))²)(1 − 1/N)`). **Findings:** the packet, not only its modes, drifts at the relativistic velocity within an `a`-uniform error whose lattice part is `O(a²)`; what does not vanish is the window's own spread `k(c²/μ)(D + k)`; the initial localisation depends on the ring length against the Compton length (`πc < ℓμ`), not on the lattice. Residue: position is circular (chord bound), the error is linear in `t`, the flat packet is the only computed instance → #263. ▸ ▸ **First brick DONE 2026-10-31 (the curvature bottleneck).** `CV/DiracPacket.lean`: ★★ `hasDerivAt_groupVel` (`v_g' = (4J² cos 2θ − v_g²)/E`), `groupVel_sq_le`, ★★★ `abs_groupVelDeriv_le` (**the band's curvature is at most `4J²/μ`**), so `abs_groupVel_sub_le` and `vgLip` are sharpened in place to `4J²/μ` — the cubic `4|J|³/μ²` term is gone and every downstream bound (`abs_bandE_step_sub_le`, `driftError`, `transPhase_drift`, `dirac_packet`) tightens. `CV/DiracContinuum.lean`: ★★★ `sq_mul_vgLip_scaledHop` (**`a² vgLip(J(a), μ) = c²/μ`, spacing-independent**), ★★★ `abs_scaledV_sub_scaledV_le` (`V_a` is `c²/μ`-Lipschitz in physical momentum for every `a`), ★★★ `abs_scaledE_step_sub_le` (the energy-step estimate in physical units), ★★ `driftError_scaled`, ★★★ `transPhase_drift_scaled` (drift at `k V_a(p₀)`, `k = 2π/(La)`, within `|t| Σ ‖w_m‖ (c²/μ) k (|p_m − p₀| + k)`; ring size, window, horizon explicit). ▸ ~~#250~~'s drift and localisation errors are `|t| · Σ ‖κ_m‖ · vgLip · h(δ_m + h)` with `vgLip = 4J²/μ + 4|J|³/μ²`; under `J = c/(2a)` with momenta `θ = ap` the first term gives an `a`-uniform bound in physical units, the second grows like `a⁻¹`. Prove a sharp curvature bound `|d v_g/dθ| ≤ C J²/μ` (in physical units `|dV/dp| ≤ C c²/μ`, the curvature of `√(μ² + c²p²)`), restate `transPhase_drift` and `norm_transPhase_ge` in physical position and momentum, and show the errors are uniform in `a` for a packet of fixed physical width and normalised amplitudes, over a finite time horizon. **Acceptance:** a packet's physical position drifts at `c²p₀/√(μ² + c²p₀²)` within an `a`-uniform error, and stays localised, as `a → 0⁺`. ⚠️ **Stuck on:** nothing — the curvature and the units are done; what remains is a bound on `Σ ‖w_m‖` for normalised amplitudes (Cauchy–Schwarz on `transWeight`), a packet of fixed physical width on a ring of fixed physical length `La`, and its initial localisation `|T(0)|` bounded below, over a stated horizon. | ~~**M–L**~~ done (M) | High — the device, not just the modes, has the relativistic limit | ~~#260~~; ~~#250~~ | > | ~~262~~ | Records | Records→spacetime | *new (2026-10-31, residue of ~~#260~~)* **One clock device on the Dirac chain, read as a phase.** ▸ **DONE 2026-11-01 (M–L took M–L).** [`CV/DiracDevice.lean`](../CsdLean4/CV/DiracDevice.lean): ★★ `level_align` (**spinors of rest energies `μ₀ ≤ μ₁` nearly parallel, cosine `≥ 1 − ((μ₁ − μ₀)/μ₀)²`**), `levelCoherence`/★★★ `levelCoherence_rate` (each level its own amplitudes), ★ `abs_vRel_sub_le` (**velocity gap `≤ c(μ₁ − μ₀)/μ₀`**), ★ `levelError_scaled_le` (**clock error rate `≤ 2cD`, spacing-independent**); ★ `deviceHam` (block-diagonal: Dirac at `μ₀` on level `0`, at `μ₁` on level `1`), ★★ `devicePacket`, ★★ `hasDerivAt_devicePacket` (**one system, one Hamiltonian**), `deviceTrans` (position), `deviceCoherence` (clock); ★★★ `dirac_device` (**normalised; position turns at `k v₀` within `|t|` × the levels' errors and velocity gap; coherence turns at the level splitting `R` within `|t| c D`, `|C(0)| ≥ (1 − ((μ₁ − μ₀)/μ₀)²)/2`; the reading `C(t)/|C(t)|` within `2|t| c D/|C(0)|` of `e^{−iRt}C(0)/|C(0)|`; `R/(μ₁ − μ₀)` within `(μ₁ − μ₀)/E₀` of `√(1 − (v/c)²)`**). **Findings:** position and reading of one device bounded together; carrying both levels costs a velocity gap `c(μ₁ − μ₀)/μ₀` and a start coherence loss `((μ₁ − μ₀)/μ₀)²/2`. Residue: the reading written as a record, and an observer timing #269's signal with it → #271. ▸ *Original row:* `dirac_packet` transports at mass `μ` and estimates the clock coherence between packets at `μ₀`, `μ₁` without tying them to one device, and bounds the complex coherence, not its argument. Build one packet carrying both internal levels (the two-level device in each mode), prove both components move together at the drift velocity within an error, and convert the coherence bound into a phase bound when `|coherence| ≥ κ > 0` (argument error `≤ arcsin(error/κ)`), giving the clock's **reading**. **Acceptance:** one device whose position and clock reading are bounded together, in the scaled limit of ~~#260~~. ⚠️ **Stuck on:** nothing — the price is the joint mode structure; the phase step is ready (~~#261~~'s `norm_mul_norm_div_sub_le`: `|C(0)| ‖C(t)/|C(t)| − e^{iΩt} C(0)/|C(0)|‖ ≤ 2‖C(t) − e^{iΩt} C(0)‖`), and so are the normalisation and weight sum (`chainInner_packet_unitAmp`, `sum_norm_transWeight_le`). | ~~**M–L**~~ done | High — the input #252's detectors read | ~~#260~~; ~~#250~~; #261; #252 | > | 263 | Records | Records→spacetime | *new (2026-10-31, residue of ~~#261~~)* **Packet shape and linear position on the Dirac chain.** ~~#261~~'s `dirac_packet_limit` bounds the drift of a normalised packet on a ring of fixed physical length `ℓ` by `|t| · limitErr` and reads the position as the phase of `T` (a chord bound). Three things are left: (a) a **Gaussian packet** with its initial localisation computed (`|T(0)| ≈ e^{−(kσ)²/2}` against ~~#261~~'s flat `1 − 1/N`), and the width–horizon trade-off optimised (narrow window = small spread `(c²/μ)D` but weak localisation); (b) **position as a real number**, `X = arg T / k` on a branch, with the error `≤ (2/k) arcsin(err/|T(0)|)`; (c) the **large-ring limit** `ℓ → ∞` with the packet's physical width fixed, so the circular position becomes a position on the line. **Acceptance:** a packet of fixed physical width whose real position drifts at `c²p₀/√(μ² + c²p₀²)` within an explicit error on the line, with its localisation computed. ⚠️ **Stuck on:** nothing — (a) is a finite sum estimate, (b) a branch argument, (c) needs the error stated per unit `k`, which `limitErr` already is up to the factor `k`. | **M** | Medium — sharpens ~~#261~~; the observers of #256/#259 work on a line | ~~#261~~; #262; #256; #259 | > | ~~264~~ | Records | Records→spacetime | *new (2026-10-31, residue of #248 (a) for the control)* **A hand clock that meets the stable chart's condition.** ▸ **DONE 2026-11-02 (S–M took S).** [`RecordLayer/GalileanTransformation.lean`](../CsdLean4/RecordLayer/GalileanTransformation.lean): `trainF` (profile `−sin(2πx)/(2π)`: no detuning at carrier angle `0`, `f'(0) = −1`), `trainG` (`g = 1/(4n + 12)`), `trainK` (`K = 32(4n + 12)`), `trainClock`; `handRate_trainClock` (`κ = 2g`), `handFreq_trainClock` (`ω = 16` at `J = 16`, `ħ = 1`), ★ `handClockErr_trainClock` (**the clock error is exactly `1/16`**), ★ `le_handHorizon_trainClock` (horizon `≥ 2n + 6 − 1/32`); ★★★ `train_clocks` (**`galilean_transformation_clocks` on `exchange_train` with every hypothesis discharged**: two such clocks, calibration on messenger `0`, messengers `a, b ≤ n` charted, `δ = 1/4`, `H = 17/4 + 2n`, all bounds at `ε = 1/16`). **Finding:** the stability capstone is not vacuous for the train — but the clock must be retuned as the train lengthens (weaker coupling, finer dial: a horizon `~2/κ` covering `2n` units with the tick `2/(Kκ)` held at `1/32`). Scope: one parameter family; identical clocks for both observers; the bounds stay `train_stable`'s, linear in `ε` and growing with `k`. No new rows. `galilean_transformation_clocks` assumes `4 · handClockErr J ħ g hf K q ≤ δ` and that the exchanges lie inside the clock's horizon; `handClockErr = 2/(Kκ) + 1/(2ω)` (`tickTime`, Rabi rate `ω = handFreq`) against `handHorizon ≈ (2/κ)(1 − 1/K) − 1/(2ω)`, `κ = handRate = driftRate`. Exhibit parameters `J, ħ, g, f, K, q` (and the second clock) with `handClockErr ≤ 1/16` and a horizon covering `exchange_train`'s first `n` exchanges (`19/4 + 2k`), so `train_stable`'s bounds hold with both observers' own clocks, end to end. **Acceptance:** a concrete instance of `galilean_transformation_clocks` on the train, hypotheses discharged. ⚠️ **Stuck on:** *done 2026-11-02 — was:* nothing — a parameter choice in `driftRate` (`rateAt · (J/rabiFreq)²`) and `rabiRate`; a large Rabi frequency both slows the drift (a longer horizon) and shrinks `1/(2ω)`, and `K` then sets the tick `2/(Kκ)`. | ~~**S–M**~~ done (S) | Medium — makes the stability capstone non-vacuous for the train | #248; ~~#201~~; ~~#243~~ | > | ~~265~~ | Records | Records→spacetime | *new (2026-10-31, residue of #252's smallest part)* **A distant detector written on the packet's arrival, on the whole chain.** ▸ **DONE 2026-10-31 (M–L took M–L), at the resolution of half a ring.** [`RecordLayer/DiracArrival.lean`](../CsdLean4/RecordLayer/DiracArrival.lean): (a) ★ `exp_mulVec_of_mulVec_eq_smul` (the exponential on an eigenvector), `chainProp` (`e^{−itH}` on all of `ℂ^{2L}`), ★★ `chainProp_packet` (**the whole chain carries every packet exactly**), `WholeArena` (state × a register per region), `wholeFlow`/`wholeEvolve` two-time, ★ `measurable_wholeFlow`, ★★★ `regionDetector` (a region's detector as an `ActualOccurrence`); (b) ★ `transPhase_eq_sum` (`T = Σ_k w_k e^{2πik/L}`), ★★ `regionWeight_mul_le` (**a region at `Re(ū e^{2πik/L}) ≤ γ` from a direction holds at most `(⟨ψ,ψ⟩ − Re(ū T))/(1 − γ)`**), ★★ `regionWeight_ringPacket_le` / `le_regionWeight_ringPacket` (along `dirac_packet_limit`: little weight far from the predicted direction `Ωt`, nearly all of it in a region holding every site near it); the instance `crossPacket` (`16` momenta about `p₀ = 4` on a ring of length `2000π`, `c = 1`, `μ = 3`, at least `640000` sites, `v = 4/5`): ★ `cross_budget` (error budget `≤ 1/8` over the crossing), ★★ `regionWeight_cross_far` (**far half `≤ 1/4` for `t ≤ 625π/3`**), ★★ `regionWeight_cross_near` (**`≥ 3/4` for `3125π/3 ≤ t ≤ 4375π/3`**), ★★★ `dirac_arrival` (**the far half's detector is an actual occurrence of the whole chain's evolution written over `[625π/3, 4375π/3]`; the predicted arrival `d/v = 625π`, `d = 500π`, lies inside, and the interval opens at `(d − ℓ/6)/v`**). **Finding:** the translation phase alone locates the weight — `|T|` near `1` forces it into an arc about `arg T`, and the drift theorem moves the arc at `v`; no site-by-site propagation was needed. Scope: a region of half a ring, not a site; little weight before arrival, not none (no light cone); the interval brackets the arrival within a factor of seven → #268. ▸ *Original row:* #252's `dirac_detector` writes the detector where the two-mode packet's weight starts, on an arena that carries only the window's amplitudes. Two things are left: (a) **the whole chain as the arena**: the state `ψ : ZMod L × Fin 2 → ℂ` evolved by `e^{−itH}` (`diracHam`), with `e^{−itH} mode_m = e^{−iE_m t} mode_m` so that `DiracDetector`'s registers and `packetDetector` carry over; (b) **a detector at distance `d` from a localised packet** (~~#261~~'s normalised packets, or #263's Gaussian): not written before a time growing with `d` (a site-resolved bound — the weight a packet puts at distance `d` within time `t`, from the hop locality `diracHam_apply_eq_zero` or #246's cone `8e·c`), and written after the packet arrives (a lower bound on the arrived weight, from the drift). **Acceptance:** an `ActualOccurrence` at a site `d` away from the packet's start, with its write interval bounded below by `d` over a velocity and above by the arrival time, on the whole-chain arena. ⚠️ **Stuck on:** nothing — (a) is the eigenvector action of the matrix exponential; (b)'s price is the site-resolved weight bound, since the drift theorems bound only the translation phase `T`. | ~~**M–L**~~ done | High — a record written by the packet's arrival is the candidate's passage, the input of #256 | #252; ~~#261~~; #263; ~~#246~~; #256 | > | ~~266~~ | Records | Records→spacetime | *new (2026-10-31, residue of #252's smallest part)* **One click: the site record selected by the record layer.** ▸ **DONE 2026-11-04 (L took M), with the acceptance corrected.** ⚠️ The acceptance as worded is ill-posed: #252's registers do not absorb, the norm is conserved, so "the weight delivered to `k` by `t` over the norm" sums over the sites to `t` — not a probability. The normalised law is a click clock. [`RecordLayer/OneClick.lean`](../CsdLean4/RecordLayer/OneClick.lean): the fibre `ℝ × T¹` with `clickLaw κ = expMeasure κ ⊗ Haar`; an ontic point `(τ, η)` clicks at `τ` at the site whose arc at `τ`, sized by the packet's site weights `‖ψ_k‖²/N`, contains `η` (`clickSite`, the corpus's `circleOutcome` read at the click time's weights — non-anticipating; the detail sets when and where); `clickCell`, `mem_clickCell_iff`; ★★★ `existsUnique_clickSite` (**every ontic point clicks at exactly one site**), ★★ `clickCell_disjoint` (**no ontic point writes two sites**); ★★★ `clickLaw_clickCell` (**P(site `k` by `t`) = `∫_{s ≤ t} w_k(s)/N` against the click-time law** — given a click at `s`, the packet's site weights at `s`; first order `κ ∫₀ᵗ w_k/N`, the row's expression), ★★★ `clickLaw_clicked_by` (**P(click by `t`) = `1 − e^{−κt}`**, whatever the packet); `siteRate`, `sum_siteRate` (through `DiracArrival`'s `sum_siteWeight`); ★★★ `chain_one_click` (any continuous norm-conserving chain trajectory), ★★★ `dirac_packet_one_click` (**every Dirac packet of the chain, `dirac_detector`'s included**, via `chainInner_packet_self`). ▸ **On `T²` 2026-11-04 (S):** `clickTime` (`τ = −log(rep ξ)/κ`), ★★ `map_clickTime` (the uniform circle carries the exponential click-time law, by `Measure.ext_of_Iic`), ★★★ `measurePreserving_torusClick` (**the corpus's compact `T²` with Haar measure carries the click law**), ★★★ `volume_torusClick_clickCell`; `chain_one_click` and `dirac_packet_one_click` strengthened with the `T²` law. Scope: the exponential click clock (rate `κ`) is a stated detector law like #252's gain, not generated (#251) and without depletion; the threshold registers of #252 are not tied to the clock. The record is a definite fact of the trajectory; QM's statistics enter only as the site weights sizing the arcs (input, not derived — the Born rule is not constructed here). #252's detectors integrate the packet's Born weight on every site, so every register rises; the threshold fixes when one is written, not that only one is. Build the detectors on the record layer: an arena with the ontic fibre (the corpus's Born-volume measurement, `qubitBorn` and its `N`-outcome form) in which **exactly one** site's record is written for each ontic state, the probability of site `k` being its time-integrated weight over the total. **Acceptance:** for the packet of `dirac_detector`, the set of ontic states writing site `k` by time `t` has measure equal to the weight delivered to `k` by `t` over the norm, and no ontic state writes two sites. ⚠️ **Stuck on:** *done 2026-11-04 — was:* the selection mechanism for a time-dependent weight. | ~~**L**~~ done (M) | High — the candidate's records become single-trajectory occurrences, as the programme requires | #252; #265; ~~#191~~; ~~#194~~ | > | ~~267~~ | QIT | QC chain (link 12) | *new (2026-10-31, residue of ~~#240~~)* **The factory's rejection rate and expected attempts.** ▸ **DONE 2026-11-02 (S–M took S).** [`Empirical/QM/QEC/SteaneFactoryRate.lean`](../CsdLean4/Empirical/QM/QEC/SteaneFactoryRate.lean): `rejProb` (the Born weight of the rejected checker words under a fault pattern — rejection is a readout, not an event of the faults), ★ `rejProb_le_one`, ★ `rejProb_none` (**a fault-free attempt is never rejected**, from `facVec_none_of_not_accepts`); `faultLaw` (`Measure.pi` of `Ber(true, false, p)` over the forty-six locations, `card_facLoc`), ★★ `faultLaw_faultySet_le` (**some location faulty with probability `≤ 46p`**, the union bound on `Measure.pi`), ★★★ `integral_rejProb_le` (**an attempt is rejected with probability `≤ 46p`**); ★★ `integral_prod_rejProb` (`k` independent attempts all rejected with probability `r^k`, Fubini) and `integral_prod_rejProb_le` (`≤ (46p)^k`); `attemptsMean` (the tail sum `∑_{k | ~~268~~ | Records | Records→spacetime | *new (2026-10-31, residue of ~~#265~~)* **A sharp arrival: narrow detectors and a causal before-bound.** ▸ **DONE 2026-10-31 (M–L took M–L).** [`RecordLayer/DiracTiming.lean`](../CsdLean4/RecordLayer/DiracTiming.lean): (a) `farArc` (width `ℓ/20`, half a ring away), `sharpPacket` (`10⁴` momenta about `p₀ = 4` on a ring of length `2π·10⁹`, `v = 4/5`), ★ `sharp_budget` (budget `≤ 2·10⁻⁴` on at least `10¹³` sites), ★★ `regionWeight_sharp_far` (**`≤ 1/60` on the arc for `t ≤ 9π·10⁹/8`**), ★★ `regionWeight_sharp_near` (**`≥ 9/10` for `39π·10⁹/32 ≤ t ≤ 41π·10⁹/32`**), ★★★ `sharpDetector` (**written over `[9π·10⁹/8, 41π·10⁹/32]`, around the arrival `5π·10⁹/4`, width `1/8` of it**); (b) `diracHam_mulVec_apply`, `VanishesNear`, ★ `VanishesNear.diracHam_mulVec` (H moves the support one site), `pow_mulVec_eq_zero`, `norm_pow_mulVec_le`, ★★ `norm_chainProp_mulVec_le` (**amplitude at distance `D` at most `(h|t|)^D/D! · e^{h|t|} · sup|ψ|`, `h = 2|J| + |μ|`**), ★★ `siteWeight_chainProp_le` (**below `2B² 8^{−D}` while `4e h|t| ≤ D`**), ★★ `register_cone_le`; ★★★ `dirac_timing`. **Findings:** timing to `1/8` costs a long ring — localisation `1 − 1/N` and dispersion `(c²/μ) N k` pull against each other; the lattice's cone is `4e(2|J| + |μ|)` sites per unit time, `≈ 11c` physically, generous against the packets' `v < c`. Scope: the narrow detector (spread packet) and the cone (localised state) are separate results; one state with both is not built. ▸ *Original row:* ~~#265~~'s `dirac_arrival` writes a detector covering half the ring over `[625π/3, 4375π/3]`, bracketing the arrival `625π` within a factor of seven, and its before-bound is "at most a quarter of the weight", not zero. Two sharpenings: (a) **a detector of fixed physical width** `w ≪ ℓ` with a write interval `[(d − w)/v − δ, (d + w)/v + δ]`, from a better-localised packet (#263's Gaussian; more modes) through `regionWeight_mul_le` at `γ` close to `1`; (b) **a causal before-bound**: for a state supported on a block of sites, the weight at distance `d` stays below `(e·‖H‖ t / d)^{2d}` (the exponential series with `diracHam_apply_eq_zero`), so a detector beyond the cone is written with weight below any threshold until `t ≈ d/(e‖H‖)`. **Acceptance:** an `ActualOccurrence` of a detector of width at most `ℓ/20` written within a window around `d/v` of relative width at most `1/2`, and one at distance `d` not written before a time linear in `d`, with the cone's velocity explicit. ⚠️ **Stuck on:** nothing — (a) is ~~#265~~'s argument with sharper numbers once #263 gives the localisation; (b) needs the support of `H^n ψ` (an induction on `diracHam_apply_eq_zero`) and the exponential series' tail. | ~~**M–L**~~ done | Medium — turns the arrival bracket into a timing, the input #256's matched exchange needs | ~~#265~~; #263; ~~#246~~; #256 | > | ~~269~~ | Records | Records→spacetime | *new (2026-10-31, split from #256)* **A one-way matched signal on the Dirac chain.** ▸ **DONE 2026-10-31 (S–M).** [`RecordLayer/DiracSignal.lean`](../CsdLean4/RecordLayer/DiracSignal.lean): `nearArc` (width `ℓ/20` at the start), ★ `separation_arcs` (every site of one arc `9π/10`–`11π/10` in angle from every site of the other), `regionWeight_sharp_le_one`, ★★ `regionWeight_sharp_start` (**`≥ 9/10` on the start arc for `t ≤ π·10⁹/32`**), ★★★ `startDetector` (**the departure, written over `[π·10⁹/40, π·10⁹/36]`**), ★★★ `dirac_signal` (**departure and arrival records matched — same evolution, same preparation — and ordered; the speed read off them, `0.45ℓ` over the longest gap to `ℓ/2` over the shortest, is `[144/201, 72/79] ∋ 4/5`, inside `(0, 1)`: below `c`**). **Finding:** the candidate's own records show its messenger slower than `c`. Scope: the speed is read from the records' timing in the chain's time, not by a carried clock; one way, both detectors at rest; the bracket is wide (`0.72`–`0.91`). ▸ *Original row:* One messenger, two detectors at rest: the arc at the packet's start (A) and ~~#268~~'s `farArc` half a ring away (B), both `regionDetector`s of the same whole-chain evolution. Prove A's record is written early (`regionWeight` of the start arc high near `t = 0`, from `le_regionWeight_ringPacket`), B's over `[9π·10⁹/8, 41π·10⁹/32]`, A's write ends before B's begins (**matched and ordered**), and **the speed read off the records** — the arcs' separation `ℓ/2` over the interval between the writes — lies in a bracket containing `v = 4/5` and **below `c = 1`**. **Acceptance:** two `ActualOccurrence`s of one evolution, ordered, with the record-read speed bracket `⊂ (0, c)` containing `v`. ⚠️ **Stuck on:** nothing — A's arc is ~~#268~~'s argument at `t ≈ 0`; the bracket is arithmetic on the write intervals. | ~~**S–M**~~ done | High — the first speed read from the candidate's records | ~~#268~~; ~~#265~~; #256 | > | ~~270~~ | Records | Records→spacetime | *new (2026-10-31, split from #256)* **A return leg on the Dirac chain.** ▸ **DONE 2026-11-01 (L took L), option (a) by the method of images.** [`RecordLayer/DiracReturn.lean`](../CsdLean4/RecordLayer/DiracReturn.lean): `chainParity` (`x ↦ −x` with `σ_z`), ★★ `diracHam_mulVec_chainParity` (**parity commutes with the Dirac Hamiltonian** — its even sector is the segment `A`…antipode with reflecting ends), ★★ `chainParity_ringPacket` (**the image of a packet is the mirror packet**), ★ `chainInner_packet_packet` (opposite momenta orthogonal); pair weights with interference bounded by the geometric mean (`abs_re_interference_le`, `le_regionWeight_pair`, `regionWeight_pair_le`); ★★ `wallState` (messenger at `4/5` plus its image, ring `2π·10¹³`, `2·10⁶` modes) and ★★ `chainParity_wallState` (**stays parity-even**); ★★★ `wallEmission`/`wallReception`/`wallReturn` (records at `A`, at the far end `B`, at `A` again — the return off the same register at a threshold between the departure's half lobe and the return's full lobe); ★★★ `dirac_return` (**one evolution, ordered, round trip within `0.3·1.25·10¹³` below `2ℓ_AB/v`**). **Finding:** the reflection is exact — no scattering estimate needed — because the reflecting ends are those of the parity-even sector. Lean lesson: kernel `deterministic timeout` on `rfl : o₁.evolve = o₂.evolve` for two `regionDetector`s with different regions — prove each against its explicit value and chain. Residue → #274 (moving reflector or relay, for #256's Doppler), #275 (the segment Hamiltonian written out). ▸ *Original row:* An exchange needs the messenger back, and the free ring sends nothing back. Either (a) **a reflector**: a mass wall (`μ` raised on a block of sites) or an open end, with the reflected packet's drift `−v` and its amplitude loss bounded; or (b) **a relay**: a detector at B whose write releases a second messenger (#254's launch on the chain), so the return is caused by the reception. **Acceptance:** emission at A, reception at B, return at A as three ordered `ActualOccurrence`s of one evolution, with the round-trip time within a bound of `ℓ_AB (1/v₁ + 1/v₂)`. ⚠️ **Stuck on:** the price is new dynamics — (a) scattering off a mass step on the lattice (transmission and reflection of a packet, not of a mode), (b) a coupled emitter. | ~~**L**~~ done | High — the missing half of #256's exchange | #256; #254; ~~#268~~ | > | ~~274~~ | Records | Records→spacetime | *new (2026-11-01, residue of ~~#270~~)* **A return from a moving observer.** ▸ **First landing DONE 2026-11-01 (the reversibility wall, passed; the L of the XL).** [`Mathlib/Analysis/ODE/LipschitzFlow.lean`](../CsdLean4/Mathlib/Analysis/ODE/LipschitzFlow.lean) (Cat-1): ★★ `LipschitzWith.odeFlow` — **a globally Lipschitz field generates a global flow** (uniform Picard–Lindelöf step `1/(2(K + 1))` through every point, concatenation, uniqueness), `odeFlow_add` (the group law), `dist_odeFlow_le` (Grönwall `e^{K|t|}`), `continuous_odeFlow` (joint), `toFlow : Flow ℝ E`. [`RecordLayer/GatedFlow.lean`](../CsdLean4/RecordLayer/GatedFlow.lean): `gatedField` (`Ψ' = −i(H₀ + c(Ψ, r) V) Ψ`, `r' = w Ψ`), ★★ `exists_lipschitzWith_gatedField` (globally Lipschitz for a Lipschitz, bounded gate closed beyond `R` and a Lipschitz weight), ★★ `gatedFlow`/`gatedEvolve` (**the two-time law an `ActualOccurrence` needs, measurable**), ★★ `sum_norm_sq_gatedFlow` (Hermitian `H₀, V` ⇒ `Σ‖Ψ_i‖²` conserved), ★ `register_gatedFlow` (the register is the integral of its weight), ★★ `gatedFlow_fst_eq_of_gate_closed` (free evolution while the gate is closed), ★★★ `gated_register_flow`. **Finding:** the relay's release need not be an event in the dynamics — a gate that is a function of the state and of a monotone register is an autonomous Lipschitz field, hence a reversible flow. ▸ **Second landing DONE 2026-11-01 (the pulse; design corrected at the user's call).** The kick `diag(−Δp θ_k)` does **not** work: the spinor angle is `asin v`, so `v = 3/5 → −4/5` puts half the weight in the lower band. The relay flips the band **in place** instead: `σ_y` acts on the spinor at every site and anticommutes with `H`, so a short pulse `Ψ' = −i(H + c(t) σ_y)Ψ` with `∫c = π/2` sends the co-moving messenger into the lower band at the same momentum and position — it turns around where `B` is and returns at `B`'s speed. [`RecordLayer/GatePulse.lean`](../CsdLean4/RecordLayer/GatePulse.lean): generic, for Hermitian `H, V` with `VH = −HV`, `V² = 1` and any curve `Ψ' = −i(H + c(t)V)Ψ` — ★★ `sqNorm_mulVec_const` (`‖HΨ‖²` conserved), `gateRot` (`e^{−iθV} = cos θ − i sin θ V`, ★ unitary), ★★★ `sqNorm_sub_gateRot_le` (**`‖Ψ(T) − e^{−iC(T)V}Ψ₀‖² ≤ T²‖HΨ₀‖²`**, by the rotating frame `S = e^{iCV}Ψ` whose velocity has norm `‖HΨ₀‖`; no Grönwall, no Lipschitz constant of `H`), `integral_gate_eq` (`c = Φ'(r) r'` ⇒ `∫c = Φ(r_T) − Φ(r₀)` exactly), ★★★ `gate_pulse`; on the chain — `spinY`, ★ Hermitian involution, ★★ `spinY_mulVec_diracHam_mulVec` (anticommutes), ★★ `chainProp_spinY_packet` (**`e^{−itH} σ_y ψ(0) = σ_y ψ(−t)`: the flipped packet runs backward**), ★ `siteWeight_spinY` (site densities, hence `B`'s register, do not see the flip), ★★★ `dirac_band_flip`. No new rows. ▸ **Third landing DONE 2026-11-02 (the relay arena; L took L; the design corrected once more — the *incoming messenger* is the one flipped).** A packet co-moving with `B` would, once flipped, move *away* from `A` in #256's approach geometry; so the relay flips the incoming messenger itself and the return is a radar echo at the messenger's own `4/5`, not at `B`'s `3/5` as the previous remainder said. [`RecordLayer/DiracRelay.lean`](../CsdLean4/RecordLayer/DiracRelay.lean): generic — ★ `lipschitzWith_satCut` (**a function Lipschitz and bounded on a ball becomes globally Lipschitz when saturated**, and is untouched on the smaller ball), `pulseStep`/`pulseRate` (a `C^{1,1}` step from `0` to `π/2` built on a hat), ★★ `exists_lipschitzWith_relayField` (**the relay field `c = Φ'(r)·w(Ψ)` is globally Lipschitz**), ★★ `gatedFlow_block_eq` (**a block the gate does not touch evolves freely for all time**), ★★ `sqNorm_gateBlock_sub_gateRot_le` (the pulse bound on the gated block from any start time); on the chain — ★ `sqNorm_diracHam_mulVec_packet_le` (**`‖Hψ‖² ≤ E²‖ψ‖²` over the window**; `26` for the messenger), ★ `relWeight_levelLift_gateRot_spinY` (**the encounter register is blind to every `e^{−iθσ_y}`**, not only the completed flip), `abs_relWeight_levelLift_sub_le` (the weight moves by at most `2‖Δψ‖`); the arena — `relayFlow` (`A`, `B`, the messenger and `B`'s relay register fed by `max(relWeight − 1/2, 0)`, saturated beyond sup-norm `2`: **one `gatedFlow`**), ★★ `relayFlow_block_free` (the devices run free for all time), ★ `relayWeight_relayFlow` (the saturation is invisible along the orbit), ★★ `exists_relayCross` (**the register crosses `3·10¹¹` inside `t_k ± 4.28·10¹¹`**, #256's reception window), ★★ `relayFlow_eq_free` (the free trajectory up to the crossing), ★★ `relayRaw_relayFlow_ge` (**the rate stays above `2/5` through the pulse**: blindness plus two drifts of at most `0.003·√26`), ★★ `relayReg_after_pulse` (the pulse delivers exactly `π/2` within `3·10⁻³`), ★★★ `sqNorm_mess_return_le` (**for every `t ≥ s₀ + 3·10⁻³` the messenger is within `‖·‖² ≤ 2.34·10⁻⁴` of `−iσ_y ψ_free(2s₀ + 3·10⁻³ − t)`: its own past run backward, to `A` at `4/5`**), ★★★ `dirac_relay`. **Finding:** the relay cannot be driven by #256's reception register — the packets' tails give it a slack of order `10¹¹`, and a pulse spanning that slack would last `~10¹¹`; the gate's own register integrates the encounter weight *clipped at `1/2`*, which is exactly zero before the encounter, so the crossing happens inside the sure window and the pulse takes `3·10⁻³`. The flip error is set by the messenger's energy (`T‖HΨ₀‖ ≤ 0.003·√26`), never by `‖H‖`. Lean lesson: never `rfl`/`change` across `imagePacket`/`restObserver` (the unifier unfolds the packet sums — max recursion); reduce `![…] k` with the `Matrix.cons_val` dsimproc and rewrite with lemmas. No new rows. ▸ **Fourth landing DONE 2026-11-02 (the records and the two-way factor; L took L; closes the row).** [`RecordLayer/DiracRadar.lean`](../CsdLean4/RecordLayer/DiracRadar.lean): generic — `integral_le_two_bumps`, `le_integral_two_windows` (integrals over two open windows), ★ `hasDerivAt_expProp_mulVec` (`t ↦ e^{−itH}v` solves `Ψ' = −iHΨ` for every `v`), ★★ `devA_relayFlow_any`/`devB_relayFlow_any` (**the relay arena's devices run by `e^{−itH_dev}` from every start**, so the tick records pull back); the arena — `RadarArena` (the relay arena with four registers: the messenger's weight at `A` clipped at `1/2`, both clocks' tick rates, `B`'s encounter weight with `A` clipped at `1/2`), ★ `abs_regionWeight_mess_sub_echo_le` (**after the pulse the messenger's weight on any region is within `0.031` of its echo's**), ★ `meeting_near`/`meeting_far`; the records — ★★★ `radarEmission` (the register at `A` reaches `5·10¹¹` within `[4·10¹² − 5.3·10¹¹, 4·10¹² + 3·10¹⁰]`), ★★★ `radarReception` (**the relay register itself crossing `3·10¹¹`** within `t_k ± 4.28·10¹¹`), ★★★ `radarReturn` (the same register at `A` reaches `1.7·10¹²` within `[2t_k − 4·10¹² − 2.37·10¹², 2t_k − 4·10¹² + 1.86·10¹²]`), ★★★ `radarMeeting` (`B`'s encounter register with `A` reaches `6·10¹¹` within `[5·10¹³ − 1.65·10¹², 5·10¹³ + 5.7·10¹¹]`), `restTickRadar`/`approachTickRadar` (both clocks' ticks pulled back by `ActualOccurrence.pullback`; `approachTick` extended to `n ≤ 100` in `DiracDoppler`); ★★★ `dirac_radar` (**one evolution, one preparation; the four records ordered; emission between `A`'s ticks 5–7, reception between `B`'s 29–31, return between `A`'s 65–73, meeting between `A`'s 76–81 and `B`'s 61–65; the round trip within `2.5·10¹²` of `2ℓ_AB(t_k)/v`; the two-way factor off `A`'s clock `∈ [3/76, 16/69] ∋ 1/7 = (v − β)/(v + β)`, excluding `0`; with `B`'s clock `k₁ ∈ [30/76, 36/69] ∋ 16/35`, excluding the Galilean `4/7`, and `k₁k₂ = (16/35)(5/16) = 1/7`**). **Findings:** the reception is the relay's firing — the register that drives the gate is the record; clipping makes the weight at `A` exactly zero while the messenger is away, so one register with two thresholds carries both emission and return (as in ~~#270~~); the two-way factor `(v − β)/(v + β)` is kinematic (Galilean and Lorentzian alike) — the Lorentz content is in the one-way `k₁`, which the same records recover. **Scope:** bounds, not equalities — sure windows, and factors as ratios of bracketing tick indices containing the predicted values; the `k₂` bracket `[3/36, 16/30]` contains both the relativistic `5/16` and the Galilean `1/4` (residue → #276); the ticks are pulled back, not re-derived; prepared emission (#254); test detectors (#251, #266). *(Remaining before this landing: (3) emission, reception, return as records of one evolution, and the two-way factor off `A`'s clock.)* ▸ ▸ *2026-11-01, re-priced L → XL:* an `ActualOccurrence` needs a two-time (reversible) law, so a relay that fires once, or releases a messenger at a state-dependent place and time, is not admissible as it stands; the honest relay is a hybrid flow with a transversal switching surface (the register strictly increasing), its crossing time a measurable function of the state, and a group-law proof across the switch — several landings. The two-way (radar) Doppler factor of #256 waits on this; the one-way factor is ~~#256~~. ~~#270~~'s reflector is a parity wall at rest; images give no moving mirror. #256's Doppler factor needs the messenger returned by a **moving** observer (~~#253~~'s device at `v`). Either (a) **a relay**: the moving observer's relational encounter register (~~#253~~'s `relWeight`) triggers the release of a second messenger it carries, as a hybrid flow whose release time is the register's first crossing (measurable because the register is monotone); or (b) **a moving mass wall** (`μ` raised on a block co-moving with the observer), the reflected packet's drift and loss bounded. **Acceptance:** emission at `A`, reception at the moving `B`, return at `A` as three ordered `ActualOccurrence`s of one evolution, the round trip within a bound of `ℓ_AB(t)(1/v₁ + 1/v₂)` at the moving separation. ⚠️ **Stuck on:** *done 2026-11-02 — was:* (a) the released messenger must start at `B`'s moving position — a carried packet co-moving with `B` until the release; (b) scattering off a moving step. | ~~**XL**~~ done (four landings, L + L + L + L) | High — the two-way (radar) Doppler factor needs it | ~~#270~~; ~~#253~~; #256; #254 | > | ~~275~~ | Records | Records→spacetime | *new (2026-11-01, residue of ~~#270~~)* **The segment Hamiltonian of the parity-even sector.** ▸ **DONE 2026-11-02 (S–M took M — the boundary bookkeeping was the price).** [`RecordLayer/DiracSegment.lean`](../CsdLean4/RecordLayer/DiracSegment.lean): ★★ `sum_zmod_even` (**a parity-even sum over `ZMod (2M)` is the two ends plus twice the interior of `{0, …, M}`**); `segWt` (`1` at the ends, `√2` inside), `segOf` (the map), `IsEvenState`, ★ `IsEvenState.end_one` (**the lower component vanishes at both ends**); ★★ `chainInner_eq_segInner` (**isometry on the even sector**), ★★ `regionWeight_eq_segRegionWeight` (a symmetric region's weight is its trace's on the segment), ★★ `segOf_segUnfold` + `isEvenState_segUnfold` (**onto** the segment states with the lower component zero at both ends); ★★ `segHam` (**the hop `−iJκ` with `κ = √2` on the two end edges, the rest energy `μσ_z`, the lower component removed at both ends**), `segHam_mulVec_apply`; ★★★ `segOf_diracHam_mulVec` (**`S H φ = H_seg S φ` on the even sector**), ★★★ `segOf_chainProp_mulVec` (**`S e^{−itH} φ = e^{−itH_seg} S φ`**, through the exponential series); ★★★ `dirac_segment_return` (the wall state even at every time, its segment image moved by `e^{−itH_seg}` with norm `1`, the detectors' weights at `A` and `B` the weights of the arcs' traces on the segment, and `dirac_return`'s ordering and round trip). **Finding:** the reflecting ends are a boundary condition, not an approximation — the even sector is the open chain on `{0, …, M}` with the lower spinor component absent at both ends and `√2` on the end edges. Scope: `M ≥ 2`; the records are restated as functions of the segment state, not re-derived from a segment detector law. No new rows. ~~#270~~ uses that parity commutes with the Dirac Hamiltonian, so its even sector is invariant — the segment from `0` to the antipode with reflecting ends. Write that out: an isometry from the even sector onto the chain on `{0, …, L/2}` intertwining `H` with an explicit segment Hamiltonian (the hop and mass terms, and the boundary terms at both ends), and restate `dirac_return` on the segment. **Acceptance:** the segment Hamiltonian and the intertwining isometry, with `dirac_return`'s records on the segment. ⚠️ **Stuck on:** *done 2026-11-02 — was:* nothing — finite-dimensional linear algebra on the even sector; the price is the boundary bookkeeping (`L` even, sites `0` and `L/2` fixed by parity). | ~~**S–M**~~ done (M) | Low–Medium — makes the reflecting ends explicit rather than a symmetry | ~~#270~~ | > | 276 | Records | Records→spacetime | *new (2026-11-02, residue of ~~#274~~)* **The return located sharply enough to split the two-way factor.** ~~#274~~'s `dirac_radar` reads `k₁ ∈ [30/76, 36/69]` (relativistic `16/35` in, Galilean `4/7` out) but `k₂ ∈ [3/36, 16/30]`, which contains both the relativistic `5/16` and the Galilean `1/4`: the return's sure window (`radarReturn`, about `4.2·10¹²` wide, ~7 of `A`'s ticks) is too coarse. **Acceptance:** a return record (and, as needed, a reception record) whose window, read against `B`'s and `A`'s ticks, gives a `k₂` bracket containing `5/16` and excluding `1/4`, in the same one evolution as ~~#274~~'s records. ⚠️ **Stuck on:** the width of the return window — the threshold crossing on the echo lobe's flank of the clipped register at `A`; narrowing it means a sharper echo (narrower packets, more modes) or a finer threshold analysis, either re-running ~~#274~~'s chain numerics. | **L** | Medium — completes the split `k₁k₂` of the two-way factor in the records; `k₁` already separates the relativistic value from the Galilean one | ~~#274~~; ~~#256~~ | > | ~~277~~ | Records | Records→spacetime | *new (2026-11-02, residue of ~~#259~~)* **Exclude Galilean on the moving worldline from the records.** ▸ **DONE 2026-11-02, option (a), on `A`'s clock only (M–L took M).** Before building, the binding brackets were found: with the tick records' real windows (`±10¹¹`), only `A`'s half-ticks `72.5` (after the return) and `76.5` (before the meeting) clear the event windows — `B`'s half-ticks would not, and the meeting window cannot be sharpened (it is set by the encounter detector's own width). [`RecordLayer/DiracHalfTicks.lean`](../CsdLean4/RecordLayer/DiracHalfTicks.lean): ★ `hasDerivAt_deviceCoherence_prop` (**`dC/dt = −iS` from every start**), `quarterReading`, ★ `integral_rate_eq` (**`∫₀ᵗ Re(ūS) = Q(t) − Q(0)`**), `rateWeight_neg` — **no new register: the half register is the tick register minus the quarter reading's change**; ★ `integral_max_neg_cos_half_tick` (`(2n − 1)κ/R` at `π(2n−1)/R`), ★★ `half_register_lt`/`half_register_ge` (the rate counter against `−u`, under the ticks' margin); `restHalfTickRadar` (**half-tick `n` of `A`'s clock, written within `10¹¹` of `(n − 1/2)·2π·10¹¹`, an actual occurrence of the radar arena's one evolution**), `restHalfTickRadar_fields`. [`RecordLayer/DiracWorldline.lean`](../CsdLean4/RecordLayer/DiracWorldline.lean): ★★ `radar_interval_bounds_half` and `dirac_lorentz_worldline` **strengthened in place** (half-ticks `72.5` after the return, `76.5` before the meeting; for readings in the sharpened brackets `B`'s interval `≤ 36 < 36.75 ≤ ΔT` — **Galilean strictly excluded** — and its square in the Minkowski range `[9945/16, 39444/25]`). **Finding:** the records alone now separate Lorentz from Galilean on the moving worldline, as the rate already did. Scope: a margin of `0.75` period; `B`'s brackets stay whole ticks. No new rows. ~~#259~~'s `radar_interval_bounds` reads the reception and the meeting to whole ticks of `A`'s and `B`'s clocks: the measured interval `≤ 36` ties with the Galilean prediction `≥ 36`. Either (a) **half-tick records**: a second tick register on each clock reading the opposite phase (`readWeight` with `−u`, ~~#271~~'s `tickDetector` at `π(2n + 1)/R`), so every event is bracketed to half a period; or (b) **a longer flight**: the radar arena with `B` further out, the interval spanning more ticks than the windows' width. **Acceptance:** the records alone (tick brackets) give `B`'s interval and `A`'s radar chart with the Minkowski interval consistent and the Galilean `ΔT` strictly excluded. ⚠️ **Stuck on:** *done 2026-11-02 — was:* the price — (a) a second detector's write intervals on both devices, re-pulled-back to the radar arena; (b) re-running ~~#274~~'s chain numerics at a new geometry. | ~~**M–L**~~ done (M) | High — makes ~~#259~~'s first brick a record-level separation, as `dirac_radar`'s `k₁` already is | ~~#259~~; ~~#274~~; ~~#271~~ | > | ~~278~~ | Records | Records→spacetime | *new (2026-11-02, residue of ~~#259~~)* **The moving observer's own radar, and the full map.** ▸ **DONE 2026-11-03 — fourth landing; the row closes (residues #280, #281).** [`RecordLayer/DiracTwoCharts.lean`](../CsdLean4/RecordLayer/DiracTwoCharts.lean): `lorentzBoost_approach` (the boost at `−3/5` of an event at `A`, `D` before the meeting: `(−(5/4) D, −(3/4) D)`); ★★★ `lorentz_chart_bounds` (**for every reading in the records' brackets the Lorentz boost of `A`'s chart of the relay firing lands within `37/4` periods in time and `23/4` light-periods in position of `B`'s radar chart** — Lorentz values `≈ 39.8`, `≈ 23.9`, errors `≈ 23%`, `≈ 24%`); `dirac_two_charts` strengthened in place with the map. **Finding:** the Lorentz map carries one observer's record chart to the other's within the records' resolution; the error is the event records' window widths (`B`'s return `3.7·10¹²`, the meeting `2.2·10¹²`), not the tick spacing (continuous readings: `≈ 17–20%`, estimate). Σ's time is continuous and the dynamics are exactly Lorentzian (`clock_rate_lorentz`, `dirac_massless_limit`); the records add finite resolution → #280. Galilean is not excluded at this resolution and is not the target (the clock rate excludes it at the dynamics level, ~~#259~~). Scope: two common events (`R`, `M`); `B`'s messenger isotropy assumed (Einstein's convention) → #281. ▸ **Route chosen 2026-11-02 at the user's call: massless messengers** (massive ones move at different speeds out and back in `B`'s frame — `5/13` and `35/37` for the `4/5` messengers — so `B`'s radar would need calibration; massless ones move at `c` to `O((ap)²)`, and the radar is Einstein's, `B`'s isotropy assumption stated as part of what is tested). Four landings: (1) massless messengers (M–L), (2) `B`'s radar arena — a messenger prepared at `B`, flipped by a relay at `A`, records of `B`'s emission, `A`'s relay firing and `B`'s return with both clocks' ticks, one evolution (L), (3) both charts of the common events — `A`'s relay event in `B`'s radar chart, `B`'s relay event in `A`'s (~~#259~~/~~#277~~) (M–L), (4) the map, Lorentz-consistent within explicit errors, Galilean excluded (M). ▸ **First landing DONE 2026-11-02 (M–L took M).** [`CV/DiracMassless.lean`](../CsdLean4/CV/DiracMassless.lean): `bandE_zero` (`E(θ) = 2J sin θ` on `[0, π]`), ★ `abs_bandE_zero_step_sub_le` (**the energy step is `h · 2J cos θ₀` within `2J h(|θ − θ₀| + h)` — no `1/μ`**), ★ `spinOverlap_zero` (**right-moving massless spinors are exactly parallel**), `RightWindow`; ★★★ `transPhase_drift_zero` (the massless packet's translation phase rotates at `(2π/L)·2J cos θ₀`), ★★ `re_transPhase_unitAmp_zero` (**initial localisation exactly `Σ β_{m+1}β_m`**); `masslessErr` (`c k a (D + k)`), ★ `tendsto_masslessErr` (**→ 0 with the spacing**), ★ `c_mul_cos_mem`; ★★★ `dirac_massless_limit` (**normalised, moving at `c cos(ap₀)` within `c(ap₀)²/2` of `c`, error rate `c k a (D + k)`**), ★★ `flat_massless_localised` (`|T(0)| ≥ 1 − 1/N`). **Finding:** massless packets do not spread in the scaled limit — the error rate is entirely lattice, against a massive packet's spacing-independent spread `k(c²/μ)(D + k)`. Scope: right-movers; bounds in the chain's frame. ▸ **Second landing, first part DONE 2026-11-02.** [`RecordLayer/DiracMasslessTails.lean`](../CsdLean4/RecordLayer/DiracMasslessTails.lean): ★ `regionWeight_massless_le`, ★ `le_regionWeight_massless`, ★★ `regionWeight_compl_chordArc_massless_le` (**where a massless packet is**, from `dirac_massless_limit` through the state-generic `regionWeight_mul_le`); `bMess` (**`B`'s messenger**: the massless packet on the moving window, delayed to start at `B`'s position at `t = 0` and reach `A` at `3·10¹³`), `bMess_norm`, ★ `abs_bMessRate_sub_le` (rate `10⁻¹³` to `4·10⁻²³`), ★★ `bMess_tail`/`bMess_near` (**off the chord-`1/20` arc about `e^{i·10⁻¹³(t − 3·10¹³)}` at most `10⁻³`, on it at least `1 − 10⁻³`, for `0 ≤ t ≤ 8·10¹³`**). Planned geometry: echo meets `B` at `3.75·10¹³`; Lorentz predicts `B`'s radar coordinates of `A`'s relay event as `(1.5·10¹³, 1.5·10¹³)`, Galilean `(3·10¹³, 1.2·10¹³)`. ▸ **Second landing, second part DONE 2026-11-02 (the relay at rest).** [`RecordLayer/DiracRelayA.lean`](../CsdLean4/RecordLayer/DiracRelayA.lean): ~~#274~~'s relay arena mirrored — `B`'s massless messenger (block `4`, rest energy `0`) and `A`'s relay register fed by `A`'s clipped encounter weight, one gated flow (`aRelayFlow`), the generic machinery of `DiracRelay.lean` unchanged; ★ `rest_tail_fine`/`bMess_tail_fine` (chord-`1/100` tails, `2.2·10⁻²` and `1.1·10⁻²`), ★★ `aEncounter_near` (**≥ `0.978·0.989` within `8·10¹¹` of `3·10¹³`**) and ★★ `aEncounter_far` (**≤ `3.3·10⁻²` beyond `1.3·10¹²`**), ★★ `exists_aRelayCross` (**`A`'s register crosses `3·10¹¹` inside `[3·10¹³ − 8·10¹¹, 3·10¹³]`**), ★ `sqNorm_diracHam_bMess_le` (massless energy `‖Hψ‖² ≤ 17`), ★★ `aRelayRaw_aRelayFlow_ge` (rate `≥ 2/5` through the pulse), ★★★ `sqNorm_mess_aReturn_le` (**after the pulse the messenger is the band-flipped free messenger run backward, `‖·‖² ≤ 1.53·10⁻⁴`: a massless echo toward `B`**), ★★★ `dirac_relay_at_rest`. Scope: the relay fires early — the encounter window's chord `1/10` lets the register cross up to `8·10¹¹` before the messenger reaches `A` (`≈ 5%` of `B`'s coordinates). ▸ **Second landing, third part DONE 2026-11-03 — landing 2 complete.** [`RecordLayer/DiracBRadar.lean`](../CsdLean4/RecordLayer/DiracBRadar.lean): ★★ `bEncounter_near`/`bEncounter_far` (`B`'s encounter weight with the messenger, window chord `1/5`, at the messenger's predicted angle — free out, echo back); `bRate` phase by phase (★ `bRate_ge_depart` `≥ 0.4662` at departure, zero through the flight and the pulse, ★★ `bRate_ge_return` `≥ 0.441` within `1.125·10¹²` of `(2s₀ + 3·10⁻³)/1.6`); ★ `integral_bRate_flight` (flat `≤ 2.875·10¹²`), ★★ `integral_bRate_return` (`≥ 3·10¹²` after the echo); ★★★ `bEmission`, ★★★ `aReception` (the relay's firing), ★★★ `bReturn`, `bRestTick`/`bApproachTick` (both clocks pulled back); ★★★ `dirac_b_radar` (**one evolution: `B`'s emission within `[2·10¹², 2.15·10¹²]` between `B`'s ticks `2`–`3`, `A`'s reception within `[3·10¹³ − 8·10¹¹, 3·10¹³]` between `A`'s ticks `46`–`48`, `B`'s return within `[3.5·10¹³, 3.87·10¹³]` between `B`'s ticks `44`–`50`**). Scope: the records only — `B`'s radar coordinates of the relay event and the comparison with `A`'s chart are landings 3–4 (M–L, M); the return window (`3.7·10¹²` wide, from the echo's encounter window) is the coarsest bracket. ▸ **Third landing DONE 2026-11-03 (M–L took M).** `bMeeting` added to `B`'s arena (fourth register; the encounter of `A` and `B` reads only the devices, so `dirac_radar`'s meeting window carries over, `meetRaw_aRelayFlow`), `dirac_b_radar` strengthened in place (meeting between `B`'s ticks `61`–`65`, `A`'s `76`–`81`). [`RecordLayer/DiracTwoCharts.lean`](../CsdLean4/RecordLayer/DiracTwoCharts.lean): `bRadarPos` (messenger at `1` both ways in `B`'s frame — `B`'s Einstein convention, tested in landing 4); ★★ `two_chart_bounds`, ★★★ `dirac_two_charts` (**both charts of the relay firing `R` and the meeting `M` from one evolution: `A`'s puts `R` `28`–`35` periods before `M` at position `0`; `B`'s radar puts it `34.5`–`42` before, `20.5`–`24` away; squared intervals `[784, 1225]` (A) inside `[614.25, 1343.75]` (B), Lorentz `≈ 1013` in both**). **Finding:** at whole ticks the brackets admit both the Lorentz map and the Galilean one (Galilean met at `ΔT ∈ [34.5, 35]`, `ΔX ∈ [20.7, 21]`) — landing 4 needs sharper readings (half-ticks of both clocks, #277's device) to exclude Galilean. Scope: `B`'s emission record lags the departure by two to three ticks (threshold integration), inside the bounds, uncorrected. `B` charts events off its worldline only by sending messengers and receiving echoes. Build the mirror of ~~#274~~: messengers prepared at `B` (co-moving, as ~~#256~~'s emissions are prepared — a launch is #254), relayed back by `A`, `B`'s radar chart of `A`'s relay events by `B`'s clock; then both charts of a common set of events and the map between them, tested against `T_B ≈ γ(T_A − vX_A)`, `X_B ≈ γ(X_A − vT_A)` (`c = 2|J|`) and against Galileo. `B` must assign its messengers a speed in its own frame — a calibration by its own records (a round trip to a relay at known proper distance), itself the velocity-addition test. **Acceptance:** the map proved within explicit errors in a stated regime, or its deviation from Lorentz quantified (#259's original acceptance). ⚠️ **Stuck on:** *done 2026-11-03 — was:* landing 4 (the map) and `B`'s speed calibration (→ #281). | ~~**L–XL**~~ done (L: M + L + M + S) | High — the brief's central question for the test candidate | ~~#259~~; ~~#274~~; #254; #277 | > | ~~279~~ | Records | Records→spacetime | *new (2026-11-02, residue of ~~#259~~)* **The lattice's preferred-frame corrections, quantified on the radar arena.** ▸ **DONE 2026-11-02 (S–M took S).** [`RecordLayer/DiracLatticeCorrections.lean`](../CsdLean4/RecordLayer/DiracLatticeCorrections.lean): `arenaSpacing` (`a = 2π·10¹³/L ≤ 2π·10⁻⁶`), `two_mul_scaledHop_mul` (`c = 2Ja = 1`); `latticeDefect`, ★ `latticeDefect_le_sq` (**between `0` and `k²`**, from `lattice_dilation_defect` and `|sin k| ≤ |k|`); ★★ `defect_arena` (**`≤ 2.1·10⁻¹⁰` at `B`'s momentum `9/4`, `≤ 6.4·10⁻¹⁰` at the messenger's `4`**); ★★ `speed_cap_arena` (**the fastest packet of rest energy `3` moves between `1 − 3a` and `1`, `3a ≤ 1.9·10⁻⁵`**); ★ `cone_arena` (**the chain's cone is `4e(1 + 3a) < 11`**); ★★★ `dirac_lattice_corrections`. **Finding:** the clocks' dilation is Lorentz's to far better than the records can tell (`6.4·10⁻¹⁰` against the rate resolution `2.5·10⁻⁴`); the messengers sit `≤ 1.9·10⁻⁵` below `c`. Scope: bounds at the arena's parameters and the packets' central momenta, not over their windows; the speed cap is not compared with a record-level speed resolution. No new rows. ~~#259~~ reads the moving clock's rate as `√(1 − β²)` to `2.5·10⁻⁴`. Quantify what the lattice adds: the dilation defect of the moving device (~~#249~~'s `lattice_dilation_defect`, `O(k²)`), each species capped below `c` (~~#246~~'s `maxGroupVel_lt`), and the chain's cone (~~#258~~'s `chainConeVel`), as explicit terms in ~~#259~~'s and `dirac_radar`'s bounds at the radar arena's parameters. **Acceptance:** each correction as a stated bound at the arena's `J`, `μ`, momenta, below the records' resolution or not. ⚠️ **Stuck on:** *done 2026-11-02 — was:* nothing — the three bounds exist; the price is evaluating them at the arena's numbers. | ~~**S–M**~~ done (S) | Medium — says how far from Lorentz the lattice is where the records look | ~~#259~~; ~~#249~~; ~~#246~~; ~~#258~~ | > | ~~280~~ | Records | Records→spacetime | ▸ **DONE 2026-11-03 — the row closes at the author's call (residues #282, #283, #284).** The re-aimed acceptance is met: `radar_lorentz_scaling` with `DiracResolution.lean` — the Lorentz errors are an explicit function of the resolution and tend to `0`. *new (2026-11-03, residue of ~~#278~~, at the author's direction: "measurement is constrained by the physical limits of measurement in spacetime; Σ is continuous")* **The records' resolution as a physical limit, and the Lorentz errors in the limit.** Σ's time is continuous and the chain's dynamics are Lorentzian exactly at the device level (`clock_rate_lorentz`: the moving clock's rate is the dilation to `2.5·10⁻⁴`; `dirac_massless_limit`: massless packets at `c` to `O((ap)²)`); what ~~#278~~'s records add is a finite resolution (`≈ 23%` in `lorentz_chart_bounds`), set by the event records' windows — the encounter-window chords (`1/5`, `1/10`), the packets' widths, and the register's integration time to threshold (`B`'s emission lags its departure by two to three ticks). **Build:** (i) derive each window's width from the physical localisation of the packets and devices (chord-arc tails, `relWeight` windows) as a function of the packet width `w`, the window chord and the threshold, with no arbitrary constants; (ii) a family with sharper localisation (narrower packets, smaller chords, larger `L`) along which the Lorentz errors of `lorentz_chart_bounds` shrink, with the rate; (iii) the floor: the localisation–momentum trade-off (a narrower packet spreads faster and loses the `O((ap)²)` massless regime) — state where the errors stop shrinking, if they do. Continuous clock readings (the full-wave register, `≈ 17–20%` at the present windows) belong here as one ingredient, not the lever. **Acceptance:** the Lorentz errors bounded as an explicit function of the resolution parameters, decreasing along a stated family, with the floor (or its absence) stated. ▸ **Diagnosis 2026-11-03 (the author: "23% is huge") — the error is the proofs' instrumentation, not physics.** (1) **The localisation bound is Chebyshev on the angle**: the flat (box) momentum profile over `N = 2·10⁶` modes has `1 − Re transPhase = 1/N` (`bMess_re_transPhase_zero`), and `regionWeight_mul_le` turns that into a weight outside chord `ρ` of at most `2/(Nρ²)` — chord `1/100` costs `1%`. The packet's width is `~ℓ/N ≈ 3·10⁷`, `~10⁻⁶` of the `3·10¹³` baseline, and its true tail at chord `1/100` is `~1/(Nρ) ≈ 5·10⁻⁵` (Dirichlet kernel) — Chebyshev loses `~200×` (both are estimates, not proved). (2) **The detector chords** (`1/5`, `1/10`) are chosen to beat those tails, so the encounter dwell times are `~10¹²` and the record windows are `0.8–3.7·10¹²`. (3) **The threshold registers** (rate `≤ 1/2`, thresholds `~10¹²`) add lag and spread in proportion to the dwell time. (4) **Once the windows are below `~6·10¹¹`**, the clocks' tick period (`2π·10¹¹`, `2–3%` of the baseline) dominates; then continuous readings or faster clocks matter. **The levers, ranked:** (a) **sharper localisation** — a smooth (Gaussian-type) momentum profile, or a Dirichlet-kernel tail bound beyond Chebyshev (`~10²–10⁴×`); (b) **the encounter and record chain with chord, threshold and `N` symbolic** — errors scale with the chord, and with Chebyshev alone the chord scales as `N^{−1/2}`, so `N = 2·10¹⁰` gives `~100×` (the cheapest family); (c) **clock resolution** — the full-wave register read continuously, or a larger splitting; (d) **the floor** — packet spreading against the massless regime. ⚠️ **Stuck on:** the price, per lever. (b) is re-running `DiracRelayA`/`DiracBRadar`'s numeric estimates with the parameters left symbolic. (a) needs either a smooth-profile packet with its `transPhase` (new) or a tail bound sharper than `regionWeight_mul_le`. (c) needs a register-versus-time lemma (`∫|cos|` within a constant of `(2/π)t`). (d) needs a packet-spreading bound at the chain's dispersion. ▸ **The order of limits** (the author's question, 2026-11-03: why not Planck time?): the proofs' instrumentation (`≈ 23%` today) → the physical localisation floor of the measuring systems, (d) (`~10⁻⁶` here, estimate) → the lattice spacing `a` (the model's cutoff; the results are stated as `a → 0`, ~~#279~~) → the Planck scale, which needs `G` and enters only with gravity (#248's GR programme, rows 38–39). Units do not change the error, which is a ratio, and Σ's time is continuous. ▸ **Lever (b), first brick DONE 2026-11-03.** [`RecordLayer/DiracResolution.lean`](../CsdLean4/RecordLayer/DiracResolution.lean): ★ `re_transPhase_flat_zero` (`Re transPhase = 1 − 1/N` at every `N`), ★★ `flat_massless_tail` (**weight off the chord-`ρ` arc `≤ 2(1/N + |t|·masslessErr)/(ρ − η)²` at every `N`**), ★★ `flat_massless_tail_le` (**the chord law: `(ρ − η)² ≥ 4/(Nε)` and drift `≤ 1/N` give tail `≤ ε` — the chord shrinks as `N^{−1/2}`**). ▸ **Lever (b), second brick DONE 2026-11-03.** Same module: ★★ `encounter_near_of_tails`/`encounter_far_of_tails` (**the encounter law at any chord: near within `ρ − ρ_A − ρ_B`, weight `≥ (1 − ε_A)(1 − ε_B)`; far beyond `ρ + ρ_A + ρ_B`, weight `≤ ε_A + ε_B`**), ★★ `encounter_near_angle`/`encounter_far_angle` (by predicted angles — the encounter window in time is the chord over the relative angular speed). ▸ **Lever (b), third brick DONE 2026-11-03.** Same module: ★ `re_transPhase_flat_massive_ge` (`Re transPhase ≥ 1 − 1/N − (π/(ℓμ))²`), ★★ `flat_device_tail`, ★★ `flat_device_tail_le` (**the devices' chord law at any `N`**). ▸ **Lever (b), fourth brick DONE 2026-11-03.** Same module: ★★★ `record_window` (**any threshold register: below `θ` before `a + (θ − P)/M`, reached by `c + (θ − P')/κ`**), ★★★ `encounter_record_window` (**an encounter record at any chord falls in `[t_c − y/Ω + 2θ, t_c − (ρ − ρ_A − ρ_B)/Ω + θ/κ]` — its error is the chords over the relative angular speed plus the threshold's fill time**). Next brick of (b): the instance — the observers and messenger at large `N` (e.g. `2·10¹⁰` modes, chords `~100×` narrower), their records through these laws (the relay's pulse and echo, `B`'s return after its departure plateau), and `lorentz_chart_bounds` at the new brackets (M–L: new packet definitions and their numeric bounds; the generic chain is in place). ▸ **Correction 2026-11-03, before building the instance: larger `N` alone does not help the observers.** The massive devices' tail bound carries a drift `|t|·limitErr ≈ T k (N k/2)/μ` that grows with `N` (wider momentum spread, faster dispersion over `T ≈ 8·10¹³`); with `1/N` it has its optimum at `N* ≈ √(2μ/(T k²)) ≈ 2.7·10⁶` for `μ = 3` — the arenas' `N = 2·10⁶` is already there, and `N = 2·10¹⁰` makes the chord at a 1% tail `~1`, not `~10⁻⁴` (estimates; the earlier `~100×` ignored the drift). The massless messenger is not limited this way (`masslessErr` carries the lattice spacing). **What does help:** (1) **heavier observers** — the optimum moves to `N* ∝ √μ` and the chord at a fixed tail falls as `μ^{−1/4}` (`μ = 3·10⁸`, `N = 2·10¹⁰`: chord `≈ 1.8·10⁻⁴`, `~100×`), with `L` raised so that `a p₀` stays small (the moving observer's momentum grows with `μ`); (2) **lever (a)**, a sharper localisation bound — the box profile's heavy tails make `1 − Re transPhase ≈ 1/N` where a smooth profile gives `~1/σ²`, and the linear-in-`t` drift overstates the true spreading. **The physical floor** for tracking a free observer of mass `μ` over `T` is the standard quantum limit `Δx ~ √(T/μ)` (`ħ = c = 1`): `~1.7·10⁻⁷` of the baseline at `μ = 3`, so the present `≈ 23%` is the bound's, not the floor's. Re-priced: the instance is **XL** with heavier observers (every arena module is numeric at `μ = 3`, `N = 2·10⁶`, `L ≥ 10¹⁹`), or follows (a). ▸ **Lever (a), first brick DONE 2026-11-03 (the author's pick).** [`RecordLayer/DiracResolution.lean`](../CsdLean4/RecordLayer/DiracResolution.lean): ★★ `le_re_sum_phase` (the quadratic phase bound), ★★★ `regionWeight_compl_chordArc_quad` (**a massive packet with real nonnegative amplitudes: tail `× (ρ − η)²/2 ≤ 1 − Re transPhase(0) + (δ't)²/2` — the drift quadratic, not linear**). With it the floor is the profile's `1 − Re transPhase(0)`: next brick, a smooth profile (the discrete sine profile `sin(π j/(n+1))`, for which `Σ β_{j+1}β_j / Σ β² = cos(π/(n+1))` exactly, `1 − cos ≈ π²/(2(n+1)²)` against the flat profile's `1/N`); estimate with both: floor `~5·10⁻¹³`, chords `~10³×` narrower at `μ = 3`. ▸ **Lever (a), second brick DONE 2026-11-03.** Same module: ★★ `sine_neighbour_sum` (the discrete box's ground state: `Σ g(i+1)g(i) = cos(π/(n+1)) Σ g(i)²`), ★★ `sineAmp_sums`, ★★★ `sine_packet_tail` (**a massive packet with the sine profile over `n` modes: tail `× (ρ − η)²/2 ≤ (π/(n+1))²/2 + κ + (δ't)²/2`**). At `μ = 3`, `T ≈ 8·10¹³`: optimum `n ≈ 1.5·10⁷`, floor `≈ 4·10⁻¹³`, chord at a 1% tail `≈ 10⁻⁵` against the arenas' `1.75·10⁻²` (estimates). **Next:** the massless messenger's sine version, the `ringPacket` (physical-unit) form of both, and the instance — observers with the sine profile at `n ~ 10⁷`, their clocks' ticks, and the record chain through `encounter_record_window` (the observers' modules are numeric at the flat profile, so the instance rebuilds them: L–XL). ▸ **Lever (a), third brick DONE and the row re-aimed, 2026-11-03 (the author: "we're probably thrashing — the numbers are based off an arbitrary tick size, mass size").** ★★★ `sine_massless_tail` (the messenger's sine tail, no spin loss, quadratic drift). The chain is generic end to end (tails → chords → encounters → record windows), and **a record's timing error scales as `√F/Ω`, `F` minimised where the profile's `1/n²` meets the spreading `(δt)²`, `δ ∝ n/μ` — the standard-quantum-limit trade-off**. Absolute figures (`23%`, `10⁻⁵`, `n ≈ 1.5·10⁷`) depend on the arenas' arbitrary `μ`, splitting, ring size and speeds; the scaling does not. **Re-aimed acceptance:** a scaling theorem, not a number — the Lorentz errors bounded by an explicit function of the resolution parameters, tending to `0` along a stated family. The numeric large-`n` instance is **parked** (rebuilding the observers at another arbitrary point would trade one number for another). Remaining for the scaling theorem: compose the tails, the encounter law and `encounter_record_window` with the chart algebra of `lorentz_chart_bounds` at symbolic windows (M), plus the clocks' resolution (c). ▸ **The scaling theorem DONE 2026-11-03.** [`RecordLayer/DiracTwoCharts.lean`](../CsdLean4/RecordLayer/DiracTwoCharts.lean): ★★★ `radar_lorentz_exact` (**Einstein's radar with a clock dilated by `√(1 − v²)` is exactly the Lorentz boost, for every `v ∈ [0, 1)` — symbolic**), ★★★ `radar_lorentz_error` (**the Lorentz residual is linear in the reading errors**), `abs_record_sub_le` (each record's reading error `≤ y/Ω + θ/κ`, from `encounter_record_window`), ★★★ `radar_lorentz_scaling` (**for every `ε > 0` a reading resolution below which the radar chart is within `ε` of the Lorentz boost**). With `DiracResolution.lean` (reading error `≈ y/Ω + θ/κ`, `y ≈ 2(ρ + ρ_A + ρ_B)`, chords `∝ √F`, `F` the sine/quadratic-drift floor) this is the re-aimed acceptance: **the Lorentz errors are an explicit function of the resolution and tend to `0`**. Scope: kinematic — the messenger's speed and the clock's rate hold on the chain to `O((ap)²)` and `2.5·10⁻⁴`, entering as reading errors; the shrinking family is not instantiated on the arenas (parked, by design); the clocks' own reading resolution (lever (c)) enters as one more `d`. ⚠️ **Stuck on:** *done 2026-11-03 — was:* nothing for the acceptance; the extras moved to #282 (an arena instance of the family), #283 (lever (c), the clocks' resolution), #284 (lever (d), the floor as a theorem). | ~~**L**~~ done (L: (b) four bricks, (a) three bricks, the scaling theorem) | High — separates what the records measure from what Σ's dynamics are; the convergence half of #248 for (B) | ~~#278~~; #248; ~~#279~~; ~~#259~~ | > | 281 | Records | Records→spacetime | *new (2026-11-03, residue of ~~#278~~)* **The moving observer's speed calibration: testing the isotropy of its messengers.** `bRadarPos` assumes `B`'s massless messenger moves at `1` both ways in `B`'s frame (Einstein's convention); ~~#278~~'s map holds under that convention. **Build:** a round trip from `B` to a relay co-moving with `B` at a known proper distance (a second relay block carried along `B`'s orbit), its records in `B`'s arena, and the measured two-way speed `2d/(r − e)` within explicit errors; then the one-way question stated honestly — what `B`'s records can and cannot fix without a synchronisation convention (~~#247~~'s finding: one-way speeds are conventional on the control). **Acceptance:** `B`'s two-way speed of its massless messengers within explicit errors of `1`, and the one-way statement with its convention named. ⚠️ **Stuck on:** the price — a relay block co-moving with `B` (a new arena, `DiracRelayA.lean`'s machinery mirrored again). | **M–L** | Medium — closes the convention `B`'s chart rests on | ~~#278~~; ~~#247~~; ~~#274~~ | > | 282 | Records | Records→spacetime | *new (2026-11-03, residue of ~~#280~~)* **An arena instance of the shrinking family (parked by design).** ~~#280~~'s scaling theorem is kinematic and symbolic; no arena of the corpus realises a member of the family below the flat-profile observers' `≈ 23%`. **Build:** observers and messenger with the sine profile (`sine_packet_tail`, `sine_massless_tail`) at a stated parameter point, their clocks' ticks, the record chain through `encounter_record_window`, and `radar_lorentz_error` at the resulting `d`'s. **Why parked:** the parameter point is arbitrary (mass, splitting, ring size, speeds — the author, 2026-11-03), so its figure adds no physics beyond the scaling law; worth building only as a witness that the family is realised by the chain's own dynamics. ⚠️ **Stuck on:** the price — the observers' modules (`DiracTwin`, `DiracDoppler`, the relay arenas) are numeric at the flat profile and are rebuilt. | **L–XL** | Low — a witness, not a new result | ~~#280~~; ~~#278~~ | > | ~~283~~ | Records | Records→spacetime | *new (2026-11-03, residue of ~~#280~~, lever (c))* **The clocks' reading resolution.** ▸ **DONE 2026-11-03 — device brick; the row closes.** Same module: `quadRate` (the four clipped readings against `±ū`, `±iū`), ★★ `abs_quadRate_sub_le` (within four drifts of `R|C₀|(|cos Rs| + |sin Rs|)`), `one_le_abs_cos_add_abs_sin` (never stalls), ★★ `quadResponse_slope` (slope `≥ R|C₀|`), ★★★ `abs_integral_quadRate_sub_le` (tracks within `4ε(t + Rt²/2)`), ★★★ `quad_clock_reading_error` (**the quadrature clock reads time with no tick period: within `(Δ + 12ε(T + RT²/2))/(R|C₀|)`, `Δ` arbitrary — the floor is the drift**). Scope: the arenas carry single-mark tick registers; carrying the quadrature register in an arena is part of the parked instance #282. In `radar_lorentz_error` each clock reading's error is one `d`; today clocks are read by tick windows (`2π·10¹¹` periods), which dominate once the record windows shrink. **Build:** the full-wave register (the two half-wave tick registers summed, strictly increasing) read continuously, with a register-versus-time lemma (`∫₀ᵗ R|cos Rs| ds` within a constant of `(2/π) R t`, plus the drift), so a clock's reading error is the drift plus the event window times the rate, not a tick period. **Acceptance:** a clock's reading error as an explicit function of the device's drift and the event window, entering `radar_lorentz_error`. ▸ **First brick DONE 2026-11-03.** [`RecordLayer/ApparatusIndependence.lean`](../CsdLean4/RecordLayer/ApparatusIndependence.lean): ★★ `register_reads_time` (a register within `E` of a response rising at slope `≥ g` reads a threshold's time within `E/g`), ★★★ `clock_reading_error` (**the clock-reading law: thresholds spaced `Δ` read an event within `(Δ + 3E)/g` — the tick size enters only as `Δ/g`, arbitrarily fine; the floor is the tracking error `E/g`**; a clock is in #248's class with `d = (Δ + 3E)/g`). **Finding for the device brick:** a single rectified register (rate `∝ |cos Rt|`, the two half-wave tick registers summed) stalls twice a period, so its slope bound is `g = 0` and the law gives nothing near the stalls; two quarter-phase pairs (`tickMark` and `i·tickMark`, rate `∝ |cos| + |sin| ≥ 1`) give `g ∝ Rκ`, and with the exact response `F` (not its linear average) the tracking error is the drift alone. ⚠️ **Stuck on:** *done 2026-11-03 — was:* the device brick (the quarter-phase rate-versus-drift bound and the response's slope). | ~~**M**~~ done (S + M) | Medium — removes the tick period from the scaling law | ~~#280~~; ~~#277~~ | > | ~~284~~ | Records | Records→spacetime | *new (2026-11-03, residue of ~~#280~~, lever (d))* **The localisation floor as a theorem.** ▸ **DONE 2026-11-03 (S–M took S).** [`RecordLayer/DiracResolution.lean`](../CsdLean4/RecordLayer/DiracResolution.lean): ★★ `sine_floor_lower` (**the floor's terms `≥ π q` for every number of modes**), ★★★ `exists_sine_floor_le` (**`⌈√(π/q)⌉` modes give `≤ (5/2) π q`**), ★★★ `sine_packet_floor` (**the sine packet at that optimum: tail `× (ρ − η)²/2 ≤ κ + (5/2) π q`**, `q = vgLip·(2π/L)²·|t|`), ★★ `vgLip_scaledHop_mul_sq` (**`q = c²k²|t|/μ`**). **Finding:** the bound's floor lies in `[π q, (5/2) π q]` (plus the spin loss) — it scales as `|t|/μ`, the standard-quantum-limit scaling: position resolution `∝ c√(|t|/μ)`. Scope: the bound's floor, not a lower bound on physical resolution (that is the standard quantum limit itself, not proved here). `sine_packet_tail`'s floor `(π/(n+1))²/2 + κ + (δ't)²/2`, `δ' ∝ n/μ`, has an optimum in `n` whose value scales as `√(T/μ)` — the standard quantum limit for tracking a free observer. **Build:** the optimum stated and proved (`n` chosen as a function of `μ`, `T`, the spacing), the floor's value as an explicit function, and its limits (`μ → ∞`: the floor `→ 0`; fixed `μ`: bounded below by the bound's form). **Acceptance:** the floor as a theorem with its scaling in `μ` and `T`, and the statement that it is the bound's floor (a lower bound on the physical resolution is a separate question). ⚠️ **Stuck on:** the price — *done 2026-11-03 — was:* optimising a two-term bound over integers `n` (choose `n = ⌈…⌉`, bound both terms). | ~~**S–M**~~ done (S) | Medium — puts the standard-quantum-limit scaling into the corpus | ~~#280~~; #248 | > | ~~285~~ | Records | Records→spacetime | ▸ **DONE 2026-11-03 — (i), (ii), (iii); the row closes (residue #286).** At the record layer the ontic frame is hidden within stated errors: the radar record relation is the same in every frame (`radar_frame_covariant`), the ontic order of spacelike writes is invisible to Σ records (`sigma_record_write_order_indep` exactly, `sigma_record_kick_order` within the light-cone bound) and to QM statistics (`twoRecordLaw_props`), and the lattice frame shows up only at order `a²` (`lattice_frame_residue`). *new (2026-11-03, the author's question "does this mean CSD is Lorentz compatible?" — scoped at their go)* **Is the ontic time's frame hidden from records?** The record-layer arenas evolve in one ontic time `t`, a global slicing, and CSD is nonlocal (row 39: no local explanation of the records is possible, `bell_general_separation`). What is proved so far is operational: observers built from the Dirac chain recover Lorentz kinematics from their records within explicit errors (~~#278~~, ~~#280~~, ~~#248~~), the moving clock runs at the dilation (`clock_rate_lorentz`), and one wing's record marginals do not depend on the other's setting (`composite_no_signalling`, `singlet_operational_no_signalling`). What is not proved is that **no record statistic singles out the ontic slicing** — the condition under which CSD is Lorentz compatible at the record layer, as Bohmian mechanics is with its hidden foliation. **Build:** (i) **frame covariance on the chain** — a preparation and its boost (on packets, a momentum shift; the lattice has no exact boost) give record laws related by the Lorentz map within explicit errors, the errors vanishing as `a p → 0`; (ii) **slicing independence on composite arenas** — for two records written at spacelike separation, the joint record law is the same whichever the ontic time writes first, extending `composite_no_signalling` from marginals to the joint law under re-slicing; (iii) **the residue stated** — the lattice frame is detectable at `O((ap)²)` (~~#279~~), so exact hiddenness is a continuum-limit statement. **Acceptance:** record statistics invariant, within stated errors that vanish in the limit, under a change of the preparation's frame and of the ontic slicing — or an explicit record-level signature of the ontic frame. ▸ **(i) DONE 2026-11-03 (M–L took M) — no lattice boost operator was needed.** [`RecordLayer/FrameCovariance.lean`](../CsdLean4/RecordLayer/FrameCovariance.lean): ★★★ `radar_lorentz_covariant` (**in any frame, with `A` at `v_A` and `B` at `v_B`, `B`'s radar coordinates of `A`'s event are the Lorentz boost at the relative velocity `w = (v_A − v_B)/(1 − v_A v_B)` of `A`'s proper-time chart**; `radar_lorentz_exact` is `v_A = 0`), ★★ `lorentzRelVel_velAdd` (**a common boost leaves `w` unchanged**), ★★★ `radar_frame_covariant` (**a preparation and its boost give the same radar record relation**), ★★ `vRel_boost` (**a momentum boost `s' = γ(s + uE)` adds velocities relativistically** — a boosted packet is a packet at a boosted momentum, and `dirac_packet_limit`, which holds at every centre momentum, moves it at the boosted velocity). Scope: the radar relation (kinematic: light at `1`, clocks at `√(1 − v²)`); a boost raises the momenta, so the lattice error grows with the boosted momentum bound (`limitErr`'s `a²P³`) — (iii). ▸ **(ii), branch-weight brick DONE 2026-11-03.** [`RecordLayer/SlicingIndependence.lean`](../CsdLean4/RecordLayer/SlicingIndependence.lean): `actB` (the local `B`-action beside `actA`), ★★ `actA_actB_comm` (local actions on different factors commute), `LocalOp`, `applyOps`, `wingA`, `wingB`; ★★★ `applyOps_eq` (**any interleaving of the two wings' local operations acts as `actA (wingA) ∘ actB (wingB)`**), ★★★ `slicing_independent` (**two histories with the same local sequences give the same branch vector, hence the same joint record weights — the ontic order of spacelike writes is invisible in the joint law**), ★★ `joint_weight_order_indep` (two records, either order). Scope: the joint law as Lüders/Kraus branch weights; spacelike taken as tensor locality (no coupling across the cut during the writes). ⚠️ **Stuck on:** for (ii), the record-tier instance — a two-wing record arena whose two record writes the ontic time can order either way, delivering this law dynamically (the single-wing form is `composite_local_record_born`); on the Dirac chain, approximate locality from the light-cone bound (#246) in place of exact tensor locality. (iii) is S. ▸ **Re-priced 2026-11-03 on inspection: the record-tier instance is L, not M.** The corpus's two-stage composition (`two_stage_joint`, `TwoTimeLuders.lean`) is built on the swap protocol, whose post-record state is a fixed calibration `ν₁ i` — right for a nondegenerate record, wrong for a local record on a composite, whose post-record state must keep the other wing's correlations (`swap_not_blockLuders`). The protocol that does keep them is the join protocol (`join_luders_marginal`, block-degenerate Lüders), and there is no two-stage composition of join protocols: the instance needs one (a join-arena analogue of `TwoStageArena`, `twoStagePrep` and `two_stage_joint`, `~200` lines of measure theory in the model of `two_stage_joint`), then the two orders and `slicing_independent` to equate them. ▸ **(ii) for QM statistics DONE 2026-11-03 (the author: "we're not constructing the Born rule here" — "the Born rule is for QM, not for Σ records"); reworded the same day: this law is the isolated composite's Born statistics, not Σ's records.** Same module: `twoRecordLaw` (the joint law of two local records as the Born weight of the branch, the corpus's established record weights — no protocol rebuilt), ★★★ `twoRecordLaw_props` (**slicing independent — either ontic order gives the same law; normalised; no-signalling — each wing's marginal is its own Born weight whatever the other wing's evolution, setting and order**), with `norm_actB`, `sum_norm_sq_actA_single`, `sum_norm_sq_actB_single`. The join-protocol composition is not needed and is dropped from the row. ▸ **(ii) for Σ records DONE 2026-11-03.** [`CV/SigmaSlicing.lean`](../CsdLean4/CV/SigmaSlicing.lean), on the fibred field arena (records as definite facts of the one trajectory; no probability): `fibredFlow_add`, `base_flow_strokeSnd`; ★★★ `sigma_record_write_order_indep` (**`A`'s Σ record at `t` is the same whether `B` wrote at an earlier ontic time or not — exact, no light-cone condition: a write holds the base fixed**), ★★★ `sigma_record_kick_order` (**an apparatus kick by `B` at ontic time `s` before `A`'s write at `t` changes `A`'s Σ record by at most the local Lieb–Robinson bound `L_h L_g · 2 max(|Y|, 1) 2^{-d} ‖A‖`; a kick after `t` cannot reach it**). Much was in the corpus already (row 38's ST-1: `recordStroke_channels_comm`, `recordStroke_heisenberg_comm_kick_of_spacelike`; #213's `record_influence_le_of_separated_local`); this brick states it as the ontic-order statement #285 asks. Scope: the fibred field arena; the Dirac chain's single-particle radar arenas would need the chain's own cone bound. ▸ **(iii) DONE 2026-11-03.** [`RecordLayer/FrameCovariance.lean`](../CsdLean4/RecordLayer/FrameCovariance.lean): ★★ `abs_lorentzRelVel_sub_le` (the relative velocity is Lipschitz, `2(|Δa| + |Δb|)/(1 − V²)²`), `latticeVelErr`; ★★★ `lattice_frame_residue` (**the lattice's relative velocity in a boosted frame differs from that in the original frame by at most `4(latticeVelErr P + latticeVelErr P')/(1 − V²)²` — order `a²`, growing with the boosted momentum bound, vanishing as `a → 0`**). ⚠️ **Stuck on:** *done 2026-11-03 — was:* (iii); the optional (ii) on the Dirac chain's own arenas moved to #286. | ~~**L**~~ done (M + S + S + S + S) | High — the difference between "observers recover Lorentz kinematics" and "CSD is Lorentz compatible at the record layer" | ~~#278~~; ~~#280~~; ~~#248~~; ~~#279~~; #39; Q27 | > | ~~286~~ | Records | Records→spacetime | *new (2026-11-03, residue of ~~#285~~)* **Σ-record slicing independence on the Dirac chain's own arenas.** ▸ **DONE 2026-11-03 (M).** [`RecordLayer/ChainCone.lean`](../CsdLean4/RecordLayer/ChainCone.lean): `SiteSupp`, `grow`, `growIter`; ★★ `siteSupp_adOne` (**a commutator with the chain's generator spreads support by at most one site**, from `diracHam_apply_eq_zero`), `siteSupp_adIter`; `IdOff`, ★★ `commute_of_disjoint`; ★★★ `norm_commutator_chain_le` (**the chain's cone bound: `‖[A(t), W]‖ ≤ 2(2‖S‖t)^d/d! · ‖A‖‖W‖` when `d` sites of growth from `A`'s sites miss `W`'s** — the corpus's generic `flowRemainder` machinery, support tracked site by site); ★★★ `kick_expectation_le`, ★★★ `chain_record_kick_order` (**a remote operation before `A`'s reading moves the rate fed to `A`'s register by at most `2(2‖S‖t)^d/d! · ‖A‖‖W‖²‖φ‖²`; after the reading it cannot reach it**). Scope: the rate as a Heisenberg-picture expectation; the radar arenas are gated flows, unitary between gates; the cone speed is `2‖S‖` (mass term included), not the group-velocity cone. ~~#285~~ (ii) is proved on the fibred field arena (`CV/SigmaSlicing.lean`), whose records are skew strokes and whose locality is the local Lieb–Robinson bound. The Dirac chain's radar arenas are single-particle, with records in gated registers. **Build:** the chain's own cone bound (the propagator's weight outside `chainConeVel · t`, ~~#258~~) and the register form of `sigma_record_kick_order`: a register fed by a local device, perturbed by a remote operation before its write, changes by at most the cone tail. **Acceptance:** the ontic-order statement for the chain's register records with an explicit cone-tail error. ⚠️ **Stuck on:** *done 2026-11-03 — was:* a single-particle cone bound with an explicit tail on the ring. | ~~**M**~~ done (M) | Low — the field-arena statement carries the content; this is its chain instance | ~~#285~~; ~~#258~~; #246 | > | ~~287~~ | Records | Records→spacetime | *new (2026-11-03, the author's pick after "so CSD is Lorentz compatible?")* **Every encounter record is frame independent.** ▸ **DONE 2026-11-03 (M–L took S).** [`RecordLayer/FrameCovariance.lean`](../CsdLean4/RecordLayer/FrameCovariance.lean): ★★ `lorentzBoost_worldline_vel`, ★★★ `proper_time_boost` (**proper time along a worldline is the same in every frame**), ★★★ `encounter_reading_covariant` (a reading of an encounter is frame independent to the two frames' reading errors), ★★★ `record_statistic_covariant` (**every statistic of encounter readings, Lipschitz in them, is frame independent to `K Σ (ε_k + ε'_k)`**, the errors being each record's window and clock error in each frame). This generalises ~~#285~~ (i) from the radar relation to every encounter-record statistic. **By construction, recorded at the author's question "isn't this all self evident?":** the chain was chosen with a relativistic continuum limit, so light speed and dilation are inputs and the kinematics is special relativity checked in Lean; what is earned is that the chain's matter realises it in records with explicit errors. Scope: uniformly moving observers; readings within stated errors of the ideal proper-time readings. ⚠️ **Stuck on:** *done 2026-11-03.* | ~~**M–L**~~ done (S) | Medium — closes the record-statistic half of Lorentz compatibility at the record layer | ~~#285~~; ~~#280~~; ~~#283~~ | > | ~~288~~ | Records | Records→spacetime | *new (2026-11-03, the untested case named at the author's question "isn't this all self evident? we've run into not one problem")* **Individual Σ outcomes of entangled spacelike records: do they depend on the ontic slicing?** ▸ **DONE 2026-11-04 (scoping S, brick M; L took M) — the answer is a fork, and the corpus sits on its slicing-dependent branch.** Scoping found the corpus's two-wing selection: the I3 skew cells (`SkewFibrePartition.lean`), where A reads its arc and B reads its arc conditional on A's cell; that module called the order "arbitrary, no physical content" (corrected in place). [`RecordLayer/SelectionOrder.lean`](../CsdLean4/RecordLayer/SelectionOrder.lean): `JointRates.transpose`, `skewCellRev` (B first), ★★ `volume_skewCellRev` (**the mirror order has the same cell volumes, the QM joint table** — no statistic tells the orders apart); ★★★ `rate_eq_mul_of_selection_order_indep` (**order-independent individual outcomes a.e. force a product table**), ★★★ `selection_order_dependent` (**for any correlated table, the ontic points whose joint outcome changes with the order carry positive measure**), ★★★ `singlet_selection_order_dependent` (the singlet at any `a·b ≠ 0`); ★★★ `no_order_indep_local_first_selection` (**general, not tied to I3: two selections, A first and B first, each first write depending only on its own setting, A-first reproducing the singlet's CHSH correlations, cannot agree a.e. at the four CHSH contexts** — through the C1 no-go `no_compatible_global_chsh_assignment_realises_singlet`). **The fork:** either the first write is local (as every local record write in the corpus is), and then the individual outcomes depend on which wing is selected first — a hidden preferred slicing if that is the ontic time's order, as in Bohmian mechanics — or the first write reads the remote setting (no order, outcome-level parameter dependence). Both are hidden from the statistics. The corpus's selections (I3; C1's `singletContextualModel`, cells A-major) are on the first branch with A first by convention; **nothing in the corpus fixes the first wing** (the record layer is kinematic) → #290. Scope: I3 on the fibre `T²` with tables as data; agreement is a.e. under the statistics' measure; the general theorem is Bell's theorem read as a statement about selection order. Everything proved so far uses a deterministic, local record model — the stroke reads the base — in which a preferred slicing has nowhere to appear; for entangled spacelike measurements only the QM statistics were shown order independent (`twoRecordLaw_props`). In CSD the realised outcome is Σ's ontic selection; with a global ontic time and Bell nonlocality, the individual outcomes may depend on which selection the ontic time makes first, as Bohmian outcomes depend on the foliation, while no-signalling hides it from the statistics. **Build:** model a Bell pair's two Σ selections in the corpus's own selection dynamics (not the stroke model), and decide whether the individually realised outcomes are slicing independent; either answer is a result ("hidden preferred slicing" or "none"). **Acceptance:** a theorem either way, with the statistics' slicing independence alongside. ⚠️ **Stuck on:** *done 2026-11-04 — was:* scoping. | ~~**L**~~ done (S + M) | High — the one place a preferred frame could surface in CSD | ~~#285~~; ~~#286~~; ~~#287~~; Q27; #39 | > | ~~289~~ | Records | Records→spacetime | **SPLIT 2026-11-04 into #292–#298 (core) and #299–#301 (optional)** — scoping [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md). Findings: fermion doubling is **not** a wall for the record kinematics (the 1+1 chain already has the `k = π` doubler; packets at `|p| ≤ P` never meet it); covariance needs no Wigner rotation (records are proper-time readings); the numerical arenas need not be redone (the scaling theorem carries); the one new physical fact is the lattice's preferred directions at `O((ap)²)` (#295). *new (2026-11-03, the second gap named at the author's question)* **Beyond 1+1 dimensions.** Everything is on the ring `ZMod L`. **Build:** a lattice Dirac field in `d` spatial dimensions (fermion doubling handled — Wilson, staggered, or a quantum walk), an anisotropy residue for rotations at `O((ap)²)` (the rotational `lattice_frame_residue`), radar with several reference observers for a spatial chart, and the 3+1 `radar_lorentz_exact`. ⚠️ **Stuck on:** the price — fermion doubling, and the packet, tail and arena modules are one-dimensional. | ~~**XL**~~ split | Medium — the claim's dimensional generality | ~~#285~~; ~~#287~~; #39 | > | ~~292~~ | Records | Records→spacetime | *new (2026-11-04, split of #289; ▸ **DONE 2026-11-04 (M took S).** [`RecordLayer/LorentzKinematics.lean`](../CsdLean4/RecordLayer/LorentzKinematics.lean), on `ℝ × E` for any real inner-product space `E`: `minkowskiInner`, `minkowskiSqD`, `boostGamma`, `boostD` (the boost at speed `v` along a unit vector `n`), `boostD_sub`, `boostD_time_sub`; ★★★ `minkowskiInner_boostD` (**every boost preserves the Minkowski inner product**), ★★ `minkowskiSqD_boostD`, ★ `boostD_real_eq_lorentzBoost` (on `ℝ` along `1` it is the 1+1 `lorentzBoost`); `properTime`, ★★★ `properTime_boostD` (**proper time is the same in every frame**), ★ `properTime_worldline`; `relGamma`, ★★ `relGamma_boostD` (**a common boost leaves the relative Lorentz factor unchanged**), ★ `relGamma_vel`; ★★★ `timelike_order_boostD`, ★★★ `exists_boostD_reverses_spacelike`. Scope: boosts only — no rotations, no composition of boosts, no Wigner rotation (not used: scoping §2.4); `relGamma` is the Lorentz factor, not a relative velocity vector. [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **Minkowski kinematics in `d+1`.** Mathlib has no Lorentz group and Physlib is not a dependency. **Build:** on `ℝ × EuclideanSpace ℝ (Fin d)`, boosts along a unit vector preserving the Minkowski form; proper time along a worldline invariant (the `d+1` `proper_time_boost`); relative speed from four-velocities, invariant under a common boost; a boost along the separation reverses the time order of spacelike events, and none reverses timelike order (the `d+1` `exists_boost_reverses_spacelike`, `timelike_order_boost_invariant`). ⚠️ **Stuck on:** *done 2026-11-04 — was:* the price. | ~~**M**~~ done (S) | High — first step; the frame for every row below | #289; ~~#287~~; ~~#290~~ | > | ~~293~~ | Records | Records→spacetime | *new (2026-11-04, split of #289; ▸ **DONE 2026-11-05 (M–L took M).** [`RecordLayer/RadarTrilateration.lean`](../CsdLean4/RecordLayer/RadarTrilateration.lean): `RadarGeneralPosition`, `two_inner_sub_eq`, `eq_zero_of_inner_eq_zero`, ★★★ `trilateration_unique` (**distances to observers in general position determine the position**); ★★ `exists_trilateration_const` (**the geometry constant** `K`, from finite dimension via `LinearMap.injective_iff_antilipschitz`), ★★★ `trilateration_stable` (**a point whose distances are within `δ₀`, `δ i` of the event's lies within `K Σ_i (δ i (2ρ_i + δ i) + δ₀ (2ρ₀ + δ₀))/2`** — linear in the errors to leading order), ★★★ `trilateration_scaling` (**for every `ε` a resolution below which every consistent point is within `ε`**, events within `R`); `null_future_time_eq`, ★★ `null_future_boostD`, `boostD_zero`; ★★★ `radar_boostD` (**an observer of a moving group reads the boosted event's time and distance**), ★★★ `trilateration_chart_boostD` (**the `d + 1` `radar_lorentz_exact`: a moving group's radar chart of an event is its Lorentz boost**; at `v = 0` the event itself). Scope: kinematic (light null, clocks reading rest-frame time); `K` exists but is not an explicit formula in the positions; the estimator is any consistent point (the linear solve is not written out). [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **The spatial radar chart.** One observer's radar gives distance, not direction. **Build:** trilateration by `d + 1` co-moving inertial observers in general position: exact recovery of an event's rest-frame coordinates from radar readings; a moving group's chart is a Lorentz transformation of a resting group's; the chart's residual is linear in the reading errors with an explicit geometry constant; the `d+1` `radar_lorentz_exact` and `radar_lorentz_scaling`. ⚠️ **Stuck on:** *done 2026-11-05 — was:* the geometry constant. | ~~**M–L**~~ done (M) | High — the 3+1 `radar_lorentz_exact` #289 named | #292 | > | ~~294~~ | Records | Records→spacetime | *new (2026-11-04, split of #289; [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **The lattice Dirac Hamiltonian in `d` dimensions.** ▸ **DONE 2026-11-17 (M took S–M; milestone M7 of the author's programme, obligation UO-17, first of five rows).** [`CV/DiracLatticeD.lean`](../CsdLean4/CV/DiracLatticeD.lean) (+30 pins), for any finite set of axes and any finite spinor type: `IsCliffordFamily`, `cliffordSymbol`, ★★ `cliffordSymbol_mul_self` (**`H(s)² = (μ² + Σ s_i²) · 1` from the relations**), `mul_cliffordSymbol_add`, ★ `cliffordSymbol_mulVec_mode`, ★ `cliffordMode_ne_zero`; ★★ `latticeDiracHam`, `latticeDiracHam_apply_eq_zero`, ★ `latticeDiracHam_isHermitian`, `planeWaveD`, `spinorWave`, `modeSD`, ★★ `latticeDiracHam_mulVec_spinorWave` (**on a plane wave the Hamiltonian acts on the spinor by its symbol**), `diracModeD`, ★★★ `latticeDiracHam_mulVec_diracModeD` (**exact eigenvectors, `E² = μ² + Σ_i (2J sin θ_i)²`**), `diracModeD_ne_zero`; `pauli`, `cliffordOne`, ★ `cliffordTwo`, ★★ `cliffordThree`; `latticeDiracHam_one` (at `d = 1` it is `diracHam`); ★★★ `lattice_dirac_d`. `abs_clockRate_div_sub_le` applies at `s := |s|`, as expected. Contract: UD-206. **Scope:** modes, not packets; the speed is the defined `c |s|/E`, not the group velocity; doublers not removed; no record read; nothing on the unified arena. Remainder: none in this row (#295–#298 are their own rows). *As opened:* **Build:** for any Clifford family (`α_i`, `β` Hermitian, pairwise anticommuting, squaring to one), the Hamiltonian on `(Fin d → ZMod L) × spinor`: Hermitian, nearest-neighbour (the `d`-dimensional `diracHam_apply_eq_zero`), plane-wave modes `(H(k) + E) e₀` (`β e₀ = e₀`) exact eigenvectors, dispersion `dispE μ |s|` with `s_i = 2J sin k_i`, clock dilation (expected to reuse `abs_clockRate_div_sub_le` at `s := |s|`); instances `d = 2` (Pauli, two components) and `d = 3` (four components). ⚠️ **Stuck on:** the price — `H(k)² = E²` from the Clifford relations is the one computation. | **M** | High — the matter every later row uses | #289; ~~#249~~ | > | ~~295~~ | Records | Records→spacetime | *new (2026-11-04, split of #289; [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **The lattice's preferred directions (the anisotropy residue).** ▸ **DONE 2026-11-17 (M took M; milestone M7 of the author's programme, obligation UO-17, second of five rows).** [`CV/DiracAnisotropy.lean`](../CsdLean4/CV/DiracAnisotropy.lean) (+28 pins): `sinS`, `latticeED`, `groupVelD`, ★ `hasDerivAt_latticeED` (the group velocity is the gradient of the energy), ★ `sum_groupVelD_sq_le`, `sNorm_groupVelD_le` (`|v_g| ≤ c`), ★★ `lattice_dilation_defect_d` (**the defect is `Σ s_i² sin² k_i / E²`, exactly**), ★★ `massless_groupSpeed_defect` (**`1 − |v_g|²/c² = Σ sin⁴ k_i / Σ sin² k_i`**), `sq_sub_le_sin_sq`, `abs_sin_pow_four_sub_le`, ★★ `abs_sin_ratio_sub_le`, ★★★ `massless_anisotropy` (**`|1 − |v_g|²/c² − (ap)² Σ n_i⁴| ≤ (3/2)(ap)⁴`**); `abs_dispE_sNorm_sub_le`, `abs_div_dispE_sNorm_sub_le`, `scaledVelD`, `contVelD`, ★★ `abs_scaledVelD_sub_le` (each component within `c a² (P²/2 + c P³ (1 + d)/(6μ))` of the continuum velocity), `contVelD_map`, ★★★ `rotation_residue`. Contract: UD-207. **Scope:** the dispersion and its gradient, not a packet or a record; the anisotropy bound is for `μ = 0`; constants not sharp; the group-speed term has coefficient `1` (the `1/3` of the row as opened is the coefficient in `|s|²`). Remainder: none in this row. *As opened:* **Build:** the group velocity as a vector with `|v_g| ≤ c`; a massless messenger's speed along a unit direction `n` within `O((ap)²)` of `c`, with the `Σ_i n_i⁴` term explicit (`|s|² = c²p²(1 − (ap)²/3 · Σ n_i⁴ + …)`); a rotated preparation's velocities within `O(a²)` of the rotated velocities (the rotational `lattice_frame_residue`), vanishing as `a → 0`. ⚠️ **Stuck on:** the price — Taylor bounds in several variables, done one coordinate at a time. | **M** | High — the one new physical fact beyond 1+1 | #294; ~~#285~~ | > | 296 | Records | Records→spacetime | *new (2026-11-04, split of #289; [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **Packets in `d` dimensions.** ▸ **FIRST PART DONE 2026-11-18 (Task 8 of the author's plan of 2026-11-18, workstream S; milestone M7, obligation UO-17).** [`CV/DiracPacketD.lean`](../CsdLean4/CV/DiracPacketD.lean) (+32 pins), for any Clifford family and any number of axes: ★ `sum_star_planeWaveD`, `chainInnerD_spinorWave`, `transMulD`, `transPhaseD`, `packetD`, ★★ `hasDerivAt_packetD`, `NoWrapD`, `DistinctD`, ★ `noWrapD_distinctD_of_box`, ★★ `transPhaseD_packetD` (**the translation phase of each axis, exactly**), ★★ `chainInnerD_packetD_self` (the norm does not change), ★ `abs_groupVelD_sub_le`, `stepErrD`, ★★ `abs_latticeED_step_sub_le`, `driftErrorD`, ★★★ `transPhaseD_drift` (**one phase per axis, each drifting at its component of the group velocity within `|t|` times a drift error**), ★★ `norm_transPhaseD_ge`. Contract: UD-209. **Scope:** the drift for `μ > 0` only, constant not sharp; localisation along each axis, not a region. ▸ **SECOND PART DONE 2026-11-18.** [`CV/DiracPacketBoxD.lean`](../CsdLean4/CV/DiracPacketBoxD.lean) (+21 pins): `siteWeightD`, `regionWeightD`, `regionWeightD_add_compl`, `slabD`, ★ `regionWeightD_compl_piFinset_le`, `transPhaseD_eq_sum`, ★★ `regionWeightD_mul_le`, ★★★ `regionWeightD_compl_box_le` (**the weight outside a box is at most `Σ_i (⟨ψ, ψ⟩ − Re(ū_i T_i)) / (1 − γ_i)`**, any state), ★★ `le_regionWeightD_box`, `movedDirD`, ★★ `re_movedDirD_transPhaseD_ge`, ★★★ `regionWeightD_packetD_compl_box_le` (**the packet's weight outside a box moving at the group velocity, `μ > 0`**), ★★★ `le_regionWeightD_packetD_box`. Contract: UD-210. **Scope:** a union bound over the axes; the sides are a hypothesis; **nothing is shown small** (no window is evaluated). **Remainder (this row stays open):** ~~(a) the weight outside a box of sites from the `d` phases~~ (done, UD-210); ~~(a′) the sides of the box as arcs of each axis~~ (done: [`CV/DiracPacketArcD.lean`](../CsdLean4/CV/DiracPacketArcD.lean), +11 pins; `arcSideD`, ★ `mem_arcSideD_expI`, `arcBoxD`, ★★★ `regionWeightD_compl_arcBoxD_le`, `movedBoxD`, ★★★ `regionWeightD_packetD_compl_movedBoxD_le`, ★★★ `le_regionWeightD_packetD_movedBoxD`; contract UD-217; the arc is described by a cosine, not shown to be a run of consecutive sites); (b) the scaled limit with errors free of `L` (the `d`-dimensional `dirac_packet_limit`, `ringPacket_localised`) — ▸ **first half done 2026-11-19**: [`CV/DiracSpinorD.lean`](../CsdLean4/CV/DiracSpinorD.lean) (+15 pins; `spinExp`, ★★ `cliffordSymbol_mul_add_mul`, ★★ `star_cliffordSpinor_dotProduct_self` (norm `2E(E + μ)`), ★★★ `re_star_cliffordSpinor_dotProduct` (`Re u′†u = s′·s + (E′ + μ)(E + μ)`), ★★★ `re_star_cliffordSpinor_dotProduct_ge`; contract UD-220; any Clifford family, `β e₀ = e₀`; the real part only). ▸ **second half done 2026-11-19**: [`CV/DiracPacketLimitD.lean`](../CsdLean4/CV/DiracPacketLimitD.lean) (+34 pins; ★★ `norm_spinOverlapD_le`, ★★★ `re_spinOverlapD_succ_ge`, `unitAmpD`, ★★ `chainInnerD_packetD_unitAmpD`, ★★★ `regionWeightD_unitPacket_compl_movedBoxD_le`, `boxWindowD`, `flatAmpD`, ★★ `pairSumD_flatAmpD` (`1 − 1/N_i`), ★★★ `regionWeightD_flatPacket_compl_movedBoxD_le` (**weight outside the moving box `≤ Σ_i (1/N_i + (2π|J|/(Lμ))² + |t| (2π/L) Sm_i)/(1 − γ_i)`**), `alignDefect_scaled`; contract UD-221). **Something is now shown small.** ▸ **(b) closed 2026-11-19**: [`CV/DiracPacketWindowD.lean`](../CsdLean4/CV/DiracPacketWindowD.lean) (+5 pins; `windowStepErrD`, ★★ `two_pi_div_mul_windowStepErrD_scaled` (drift rate `(2πc/ℓ)(ℓΔ_i + 2π)/L + (2πc²/(ℓ²μ))(ℓΔ_i + ℓΣΔ_j + 4π)`), ★★★ `regionWeightD_flatPacket_compl_movedBoxD_le_window`; contract UD-222). The terms are free of `L`, not small in `L`; ★★ `tendsto_two_pi_div_mul_windowStepErrD` added 2026-11-19 (+1 pin: the rate's limit along the family of lattices); no optimised window; (c) `μ = 0` and small `μ` without the `1/μ` bound (`dirac_massless_limit`); (d) the two-level clock on a packet — first part done 2026-11-19: [`CV/DiracClockPacketD.lean`](../CsdLean4/CV/DiracClockPacketD.lean) (+16 pins; ★★ `levelCoherenceD_eq`, ★★★ `levelCoherenceD_rate` (the coherence turns at the level splitting at the window's centre), ★★ `star_cliffordSpinor_dotProduct_levels`, `levelSpinOverlapD_pos`, `deviceHamD`, ★★ `hasDerivAt_devicePacketD`, ★ `deviceCoherenceD_devicePacketD`; contract UD-235); second part done 2026-11-19: [`CV/DiracClockDilationD.lean`](../CsdLean4/CV/DiracClockDilationD.lean) (+8 pins; ★★★ `levelClockRateD_dilation`, ★★ `sq_div_latticeED`, ★★ `abs_levelClockRateD_sub_le`, ★★★ `levelCoherenceD_rate_width`; contract UD-238); left of (d): the lattice term of `sq_div_latticeED` against a spacing, the normalised device. *As opened:* **Build:** one translation phase per axis, each drifting at its component of the group velocity within `|t|` times a drift error; normalisation; localisation independent of `L`; the scaled and massless limits (the `d`-dimensional `dirac_packet_limit`, `ringPacket_localised`, `dirac_massless_limit`). ⚠️ **Stuck on:** the price — multi-index re-indexing of `DiracPacket`, `DiracPacketLimit`, `DiracMassless` (≈1,750 lines) and a Lipschitz bound on each group-velocity component over the momentum window. | **L** | High — moves the observers and messengers | #294; #295 | > | 297 | Records | Records→spacetime | *new (2026-11-04, split of #289; [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **Records and the cone on any nearest-neighbour lattice.** The detector and record modules take states on `(ZMod L × Fin 2) × Fin 2`. **Build:** generalise the region detector, `encounter_record_window` and `ChainCone` (`norm_commutator_chain_le`, `chain_record_kick_order`) to any finite site set with a nearest-neighbour graph and any spinor type, then instantiate on the `d`-dimensional torus and recover the 1+1 statements as instances (rule of two). ▸ **FIRST BRICK DONE 2026-11-19 (the cone; Task 9 of the plan of 2026-11-18, §7 items 1 and 2).** [`RecordLayer/GraphCone.lean`](../CsdLean4/RecordLayer/GraphCone.lean) (+32 pins): `SiteSuppG`, `HopsG`, `growG`, `growIterG`, ★★ `siteSuppG_adOne`, ★★★ `norm_commutator_graph_le` (**the bound for any site graph**), ★★ `norm_commutator_graph_idOffG_le`, ★★ `norm_heisenbergFlow_sub_graph_le`, ★★ `abs_arenaObs_flow_kick_sub_graph_le`, ★★ `abs_arenaObs_flow_sub_flow_graph_le`, `disjoint_growIterG_of_rank`; the ring as an instance (`ringAdj`, `HopsOne.hopsG`, `coe_growIter`); the torus (`torusAdj`, ★★ `hopsG_latticeDiracHam`, `disjoint_growIterG_torus`, ★★★ `norm_commutator_latticeDirac_le`). Contract: UD-230. **Scope:** upper bounds with a factorial tail; the ring's theorems were not re-proved from the general ones. ▸ **SECOND BRICK DONE 2026-11-19 (a detector on a graph; §7 items 3 and 4, each in part).** [`RecordLayer/GraphDetector.lean`](../CsdLean4/RecordLayer/GraphDetector.lean) (+16 pins): ★★ `norm_heisenbergFlow_mul_graph_le`, ★★ `abs_arenaObs_flow_graph_le` (the occupancy of `X` by a carrier that starts on `Y`), ★★★ `abs_idxPointerShift_le_graph`, ★★★ `graphDetector_not_displayed` (**no display while `2 |f'| ‖B₀‖ (2‖H‖t)^d/d! · t < 1/6`**), ★★ `graphDetector_displayed` (from a hypothesis on the occupancy), `siteSetProj`, ★★★ `torusDetector_not_displayed`. Contract: UD-231. ▸ **THIRD BRICK DONE 2026-11-19 (a packet displayed).** [`RecordLayer/TorusDetector.lean`](../CsdLean4/RecordLayer/TorusDetector.lean) (+5 pins): ★★ `exp_mulVec_packetD`, ★★ `arenaObs_siteSetProj_mk`, ★★ `arenaObs_siteSetProj_packetD`, ★★★ `torusDetector_displayed_of_packet`, ★★★ `torusDetector_displayed_flatPacket` (**the flat packet whose moving box stays inside the detector's sites is displayed, on the fibres where the profile vanishes**). Contract: UD-234. **Remainder (this row stays open):** arrival and display in one run (compose with the second brick); an instance with numbers; a `TickReadout` for this arena, so that the detector is in a record history; `encounter_record_window` and `chain_record_kick_order` over a graph. ⚠️ **Stuck on:** the price. | **M** | Medium — library hygiene the capstone needs | #294; ~~#286~~ | > | 298 | Records | Records→spacetime | *new (2026-11-04, split of #289; [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **The `d+1` capstone.** **Build:** compose #292–#297: the chain's records give a trilateration chart within `ε` of a Lorentz transformation below a reading resolution (the `d+1` `radar_lorentz_scaling` on the chain); `record_statistic_covariant`, `chain_record_kick_order` and `spacelike_boost_changes_outcomes` in `d+1`; strengthen the 1+1 capstones where a conjunct carries, rather than adding parallel ones. ⚠️ **Stuck on:** the price — composition once its inputs land. | **M** | High — closes #289 | #292; #293; #294; #295; #296; #297 | > | 299 | Records | Records→spacetime | *new (2026-11-04, split of #289; [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **Doublers removed** (optional). **Build:** a Wilson term (or staggered fermions) on #294's Hamiltonian, the doublers lifted to energy `O(1/a)`, and the cost stated: the Wilson term adds about `r c a p²/2` in physical units, an `O(a)` error where the naive chain's is `O(a²)`. ⚠️ **Stuck on:** need — the record kinematics never populate the doublers (scoping §2.2); it matters for interacting matter with momentum transfer near `π/a` and for chirality. | **M** | Low — optional; the author decides | #294 | > | 300 | Records | Records→spacetime | *new (2026-11-04, split of #289; [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **A numerical arena in `2+1`** (optional): the analogue of `dirac_two_charts` — two observer groups, messengers, clocks and record windows with tick brackets and a Lorentz map within stated percentages. ⚠️ **Stuck on:** need — the scaling theorem (#298) does not use it (scoping §2.7); the 1+1 arenas run to ≈19,000 lines. | **XL** | Low — optional; a worked instance | #298 | > | 301 | Records | Records→spacetime | *new (2026-11-04, split of #289; [`beyond-1p1-scoping.md`](beyond-1p1-scoping.md))* **The coupled detector and its Liouville theorem in `d` dimensions** (optional): ~~#251~~'s `H = p²/(2m) + g Σ_j χ(x − j) η_j` with `x ∈ ℝ^d`, the disturbance bound, and ~~#291~~'s measure preservation with determinant one by Jacobi's formula (the variational field trace-free) in place of the 2×2 entries. ⚠️ **Stuck on:** need — the status note's item 6 uses the 1+1 detector; Jacobi's formula for the variational equation is the one new piece. | **M** | Low — optional; the author decides | ~~#251~~; ~~#291~~ | > | ~~302~~ | QIT | QC chain (link 12) | *new (2026-11-05, residue of ~~#226~~)* **The `CNOT̄` extended rectangle.** ▸ **DONE 2026-11-07 (M took S–M).** (a) in [`ExtendedRectangle.lean`](../CsdLean4/Mathlib/QuantumInfo/ExtendedRectangle.lean): `PropagatesInto P D₁ D₂`, `PropagatesD.propagatesInto`, `IsRecovery.mono`, ★★★ `isCorrectedStep_exRec_leading_of_propagatesInto` (`isCorrectedStep_exRec_leading` is now its one-set case). (b)–(d) in [`SteanePairExRec.lean`](../CsdLean4/Empirical/QM/QEC/SteanePairExRec.lean): `cnotT_mul_pairPauli_mul_cnotT` (**`CNOT̄` on Paulis, exactly**: `X^{a₀}Z^{b₀} ⊗ X^{a₁}Z^{b₁} ↦ X^{a₀}Z^{b₀+b₁} ⊗ X^{a₀+a₁}Z^{b₁}`; ~~#304~~'s `cnotT_mul_blockKron_pauliMat_mul_cnotT` is the case `a₀ = b₀ = 0`), ★★ `cnotT_mul_blockFault_mul_cnotT_mem_pairSpan` (a fault at one location of `CNOT̄`, pushed through the gate, is a unitary combination of the pair family, ~~#304~~'s `pairSpan`); `D₁` = `pairLocWeightOne` (a unitary on qubit `q` of each block, the same `q` — **stronger than the row asked**: a one-block fault is the case with one factor `1`, and a fault on `CNOT̄`'s own location is included); `D₂` = ~~#304~~'s `pairMix`, **reused**, with its recovery `exists_isRecovery_pairMix`; ★★ `PairGadget.propagatesInto_cnot`, ★★ `PairGadget.propagatesInto_pairMix` (every gadget of the set); ★★★ `steane_pair_exRec` (faulty leading recovery, faulty gadget, faulty trailing recovery, `CNOT̄` included). 20 pins. **Remainder** (not a row, rule of 2026-11-06): the leading recovery's `DeviatesBy … pairLocWeightOne` is a hypothesis, as in `steane_exRec` — Steane's method on two blocks built as a circuit is #232's per-block argument run twice (S–M); the one-set `PropagatesD` failing for `CNOT̄` is not proved (S, not needed). ▸ *Original row:* ~~#226~~'s two-block machine runs on `isCorrectedStep_rect` with an ideal recovery; #96's extended rectangle with a faulty *leading* recovery (`isCorrectedStep_exRec_leading`) also needs `PropagatesD`, which ~~#226~~ proves for the single-block gadgets only. The transversal `CNOT̄` copies an `X` on the control's qubit `q₀` to the target's `q₀`, so with a target error already on `q₁ ≠ q₀` the one-set `PropagatesD pairProj pairWeightOne` is expected to fail. Deliverables: (a) a two-set form of the leading-rectangle lemma (input deviations in `D₁`, output in `D₂`, recovery for `D₂`); (b) `D₁` = a fault on one qubit of **one** block, `D₂` = an arbitrary unitary on one qubit pair `(q₀, q₁)` across the blocks — not product form, since `CNOT` conjugation entangles the pair — with the pair recovery for `D₂` by the same Knill–Laflamme scalar argument as `exists_pairRecovery`; (c) `CNOT̄` carries `D₁` into `D₂`; (d) `steane_pair_exRec`, the two-block analogue of `steane_exRec`. ⚠️ **Stuck on:** nothing identified; the price is the non-product correctable set and the two-set lemma. | **M** | Low–Medium — the leading-recovery fault for the one two-block gate | ~~#226~~; ~~#96~~; `Empirical/QM/QEC/SteanePairMachine.lean`; `Mathlib/QuantumInfo/ExtendedRectangle.lean` | > | ~~303~~ | QIT | QC chain (link 12) | *new (2026-11-05, residue of #230's first brick)* **`S̄` at every concatenation level.** ▸ **DONE 2026-11-06 (S–M took S).** In [`SteaneMachineConcat.lean`](../CsdLean4/Empirical/QM/QEC/SteaneMachineConcat.lean): `ConcatGadget` (the level-`k` gate set: `H̄`, `X̄`, `Z̄` via `of`, and the transversal `S`, `S†`), `ConcatGadget.gate`, `ConcatGadget.logical` (`S ↦ S†`, `S† ↦ S`, the rest fixed), `logical_logical`, `logical_iterate_logical_iterate`, `logical_iterate_s` (`logicalᵏ(S)` is `S` for even `k`, `S†` for odd); ★★★ `ConcatGadget.exists_cTrans_mul_cEnc` — **`cTrans g k * cEnc k = cEnc k * (μ • (logicalᵏ g).gate)`**; ★★ `exists_cTrans_mul_cEnc_sGate` — **the logical `S̄` at every level**, from the transversal `logicalᵏ(S)`; `Gadget.exists_cTrans_mul_cEnc` now derived (the fixed points of `logical`). `steane_gadgetSet_corrected_level` and `steane_machine_level` **strengthened in place** to the five-gate set (`idealK`/`stepK`/`isCodeState_idealK` moved to `ConcatGadget`); `steane_machine_threshold` is gate-independent and unchanged. To import the level-one `S` facts, `SteaneLogicalS.lean` now precedes the concatenated module, and its two code-projector lemmas (`sTransversal_mul_steaneProj`, `sTransversal_conjTranspose_mul_steaneProj`) moved there (names unchanged). 10 pins, 3 renamed. Scope: the level-one `Gadget` type is unchanged (its consumers keep `H̄`, `X̄`, `Z̄`); the inherited scope of #227 stands (ideal recovery, gate faults, one block). ▸ *Original row:* `SteaneLogicalS.lean` proves the transversal `S` is the logical `S†` at level one. At level `k` the transversal `S` is therefore `S` or `S†` logically according to the parity of `k`, while `SteaneMachineConcat.lean`'s `Gadget.exists_cTrans_mul_cEnc` assumes a gate that is itself at every level, so `S̄` is in neither the one-block `Gadget` type nor the level-`k` machine. Deliverables: a logical-gate map on the gadget type (`G.logical`, with `S ↦ S†`), `exists_cTrans_mul_cEnc` restated as `cTrans g k * cEnc k = cEnc k * (μ • logicalᵏ g)`, and `S̄` added to the level-`k` machine (`steane_machine_level`, `steane_machine_threshold`). ~~⚠️ **Stuck on:** nothing identified; the price is restating the concatenated module's gadget lemma and its consumers.~~ closed. | ~~**S–M**~~ done (S) | Low — the phase gate in the concatenated machine | ~~#230~~; ~~#227~~; `Empirical/QM/QEC/SteaneLogicalS.lean`; `Empirical/QM/QEC/SteaneMachineConcat.lean` | > | ~~304~~ | QIT | QC chain (link 12) | *new (2026-11-05, residue of ~~#230~~)* **A faulty magic state within the correctable set.** ▸ **DONE 2026-11-05 for a fault on the magic state (M); faults in the `CNOT̄` and the read-out split to #306.** In [`Empirical/QM/QEC/SteaneInjection.lean`](../CsdLean4/Empirical/QM/QEC/SteaneInjection.lean) (37 pins): `pairSpan` (unitary combinations of the pair's Pauli family — **non-product pair errors included**), `pairMix` (**mixtures of conjugations by them**), ★★ `exists_isRecovery_pairMix` (**the pair recovery undoes every such mixture**: Knill–Laflamme scalar, `1` by unitarity, linearity for the mixture), `pairWeightOne_subset_pairMix`; ★ `cnotT_mul_blockKron_pauliMat_mul_cnotT` (`CNOT̄ (1 ⊗ X^a Z^b) CNOT̄ = Z^b ⊗ X^a Z^b`), ★ `logProj_mul_pauliMat_mul_steaneEnc` (the decoded read-out does not see a single-qubit Pauli on the code), ★ `readOp_mul_cnotConj_gateOf` (an arbitrary fault passes read-out and feed-forward, by linearity from the Paulis), `faultE`, ★ `faultE_mem_pairSpan`; ★★ `injKraus_fault_mul_pairEnc`, ★★★ `injChannel_fault_eq` (**the faulty gadget is the ideal one followed by the even mixture of the two outcome-dependent pair errors**), ★★ `deviatesBy_injChannel_fault`; ★★★ `steane_injection` strengthened (one recovery for the exact magic state, for **an arbitrary unitary fault on any one qubit of the magic state**, and for any state within `pairMix`). Scope: level one, ideal recovery. ▸ *Original row:* `SteaneInjection.lean`'s `steane_injection_corrected` takes the magic state's error as a `DeviatesBy` hypothesis and discharges it only for the exact state. A single-qubit fault `A` on the prepared `|Ā⟩` does not give one weight-one deviation: `CNOT̄` carries its `Z` part to the data block, and the ancilla reset `X̄^m` signs its `Y`, `Z` parts by the outcome, so the gadget is a *mixture* over `m` of two pair errors (each outcome has weight `1/2`). Deliverables: the correctable set closed under such mixtures (a recovery is linear, so it removes a mixture of removable errors), and the discharge for a single-qubit fault on the magic state — Pauli faults first (the sign cancels in the conjugation), then arbitrary unitary ones; then faults in the `CNOT̄` and the read-out. ~~⚠️ **Stuck on:** nothing identified; the price is the mixture-closed correctable set and the outcome-dependent propagation.~~ closed. | ~~**M**~~ done (M; residue #306) | Medium — the injection's fault tolerance | ~~#230~~; #302; `Empirical/QM/QEC/SteaneInjection.lean`; `Empirical/QM/QEC/SteanePairMachine.lean` | > | ~~305~~ | QIT | QC chain (link 12) | *new (2026-11-05, residue of ~~#230~~)* **Injection in the machine.** ▸ **DONE 2026-11-05 (M); residue #307.** [`Empirical/QM/QEC/SteaneUniversalMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteaneUniversalMachine.lean) (26 pins): ★ `cnotT_mul_blockKron_gateOf_mul_cnotT` (**`CNOT̄` keeps a one-qubit fault of the target on its column**: `Σ_x |x⟩⟨x|_q ⊗ (X^x V X^x)_q`); `colKraus`, ★★ `colKraus_mul_pairEnc` (a fault at column `q` — magic-state qubit `q`, the `q`-th `CNOT`, the read-out of qubit `q`, all at once — leaves the exact Kraus operator followed by the column error), ★ `colFaultE_mem_pairSpan`; `injFaulty`, ★★ `deviatesBy_injFaulty` (**the one-fault bound for the injection** over seven column locations); `UGadget` (**the universal gadget set**: `PairGadget`'s Clifford gadgets and `tGate`), ★★ `UGadget.deviatesBy_faulty`, ★★★ `steane_universalGadgetSet_corrected` (one recovery, `exists_isRecovery_pairKrausMix`, for all); ★★★ `steane_universal_machine` (**the machine over `H̄`, `S̄`, `X̄`, `Z̄`, `CNOT̄`, `T̄`**: output exactly ideal for every non-bad pattern, and with probability at least `1 − 21 N p²` under independent location faults of rate `p`; `steane_pair_machine` is the Clifford special case). Scope: level one, ideal recovery, unitary faults; an injection location is a column of three physical sites (rate ≤ `3p` if each site has rate `p`); `T̄` acts on block `0` only, block `1` its workspace, so the machine is universal on one logical qubit (#307 for two). ▸ *Original row:* `steane_pair_machine` counts faults over `PairGadget`, whose members are unitary transversal gates; the injection gadget is a channel with its own fault locations (the magic-state preparation, the read-out). Deliverables: the injection as a gadget of the two-block machine, its locations in the fault count, and `steane_pair_machine` restated over `H̄`, `S̄`, `X̄`, `Z̄`, `CNOT̄`, `T̄` — a universal gate set — with the bound re-derived. ~~⚠️ **Stuck on:** nothing identified — the one-fault bound at every location kind is done (~~#304~~ the magic state, ~~#306~~ the `CNOT̄` and the read-out); the price is a gadget type holding a non-unitary channel and its locations in the machine's fault count.~~ closed. | ~~**M**~~ done (M; residue #307) | High — a universal machine at level one | ~~#230~~; ~~#304~~; ~~#306~~; `Empirical/QM/QEC/SteanePairMachine.lean` | > | ~~307~~ | QIT | QC chain (link 12) | *new (2026-11-05, residue of ~~#305~~)* **A universal machine on two logical qubits.** ▸ **DONE 2026-11-05 (M–L took M) except the last deliverable, split to #308.** [`Empirical/QM/QEC/SteaneTripleMachine.lean`](../CsdLean4/Empirical/QM/QEC/SteaneTripleMachine.lean) (50 pins). The two-block development is **lifted**, not re-proved: `split3`, `lift3` (a pair operator on the ordered blocks `σ 0`, `σ 1`, tensored with an operator on the idle block `σ 2`), ★ `lift3_blockKron`, ★★ `sum_lift3_conj_eq` (**equal channels on the pair stay equal on three blocks**, slice by slice, via `sum_kronecker_one_conj_eq`); `tripleEnc`, ★ `tripleEnc_eq_lift3`, `tripleFam`, `encoderKL_tripleFam`, `tripleKrausMix`, ★★ `exists_isRecovery_tripleKrausMix`; `UGadget.idealK` / `faultyK` (the pair gadgets as Kraus families), ★★ `UGadget.exists_uniform_deviation` (the pair's one-fault bound with one deviation channel for every code state); `TGadget` (a pair gadget on an ordered pair of blocks), ★★ `TGadget.deviatesBy_faulty` (**the one-fault bound on three blocks**, a unitary fault on the idle block included), ★★ `TGadget.ideal_tGate_tripleEnc` (the lifted injection is `T` on logical qubit `σ 0`, `σ 1` reset, `σ 2` untouched); ★★★ `steane_tripleGadgetSet_corrected`, ★★★ `steane_triple_machine` (**the machine on three blocks**: exactly ideal for every non-bad pattern, and with probability at least `1 − 21 N p²`; data blocks `0`, `1`, ancilla block `2` give `H̄`, `S̄`, `T̄` on each data qubit and `CNOT̄` between them). Scope: level one, ideal recovery, unitary faults; the idle block is faulty only at the gadget's faulty location; the density theorem is cited, not composed (#308). ▸ *Original row:* `steane_universal_machine` injects `T̄` into block `0` with block `1` as the ancilla, whose contents the injection discards, so the two-block machine is universal on one logical qubit only. Deliverables: a three-block register (two data blocks and an ancilla block), the gadget set on it — the Clifford gadgets on the data blocks, `CNOT̄` in both directions, `T̄` on either data block through the ancilla — the recovery (Knill–Laflamme composes over blocks as for `pairFam`), and the machine theorem with its bound; then `exists_phase_mem_cliffordTmLim` read on the two logical qubits. ~~⚠️ **Stuck on:** nothing identified; the price is re-indexing the two-block lemmas (`PairReg`, `pairEnc`, `pairFam`) over three blocks.~~ closed — lifted instead of re-indexed. | ~~**M–L**~~ done (M; residue #308) | High — a universal multi-qubit machine at level one | ~~#305~~; `Empirical/QM/QEC/SteaneUniversalMachine.lean`; `Mathlib/QuantumInfo/CliffordTUniversal.lean` | > | ~~308~~ | QIT | QC chain (link 12) | *new (2026-11-05, residue of ~~#307~~)* **Density on the machine's logical qubits.** ▸ **DONE 2026-11-05 (S–M); residue #309.** [`Empirical/QM/QEC/SteaneMachineDensity.lean`](../CsdLean4/Empirical/QM/QEC/SteaneMachineDensity.lean) (24 pins): `dataEmb`, `dataEnc` (data states: the ancilla logical qubit in `|0⟩`); ★ `TGadget.ideal_clifford_tripleEnc`; `cnotL_eq_cnotGate'`, `cnotL_submatrix_swapQ`, `lift3_swap01` (`CNOT̄` either way is `cnotGate'`); ★ `exists_had_mul_steaneEnc` (**the transversal Hadamard's scalar has modulus one**); `realised` (**the two-qubit unitaries some list of three-block gadgets applies exactly to every encoded data state**, a submonoid); ★ `gateOf_hGateM_mem_realised`, ★ `cnotGate'_mem_realised`, ★ `gateOf_tGateM_mem_realised`; ★★ `cliffordTm_le_realised` (**every Clifford+T circuit on two qubits is realised by the machine, exactly**); ★★★ `steane_triple_machine_universal` (**every two-qubit unitary is, up to a global phase, a limit of realised logical gates**); `tMachineIdeal_get` (a realising list is a circuit of `steane_triple_machine`). Scope: ideal runs on encoded data states (faulty runs are `steane_triple_machine`'s); density without a rate (#309). ▸ *Original row:* `steane_triple_machine`'s gadget set holds Clifford+T on two logical qubits, and `exists_phase_mem_cliffordTmLim` says Clifford+T circuits are dense in `U(4)` modulo phase, but no theorem composes them. Deliverables: the logical action of each `TGadget` on the three logical qubits as a `gateOf`/`CNOT` of the corpus's Clifford+T vocabulary (`cliffordTGates`), and the statement that for every two-qubit unitary `V` and `ε > 0` there is a circuit of the gadget set whose ideal run on every encoded input is within `ε` of `V` on the data qubits, up to a global phase, with the ancilla's state discarded. ⚠️ **Stuck on:** nothing identified; the price is matching `tGateM`, `sGateM`, `cnotL` and the transversal `H̄` to the `QReg` gates of `CliffordTUniversal.lean` and the closure bookkeeping. ~~⚠️~~ closed — the corpus's gates were matrices already, no `QReg` bridge needed. | ~~**S–M**~~ done (S–M; residue #309) | Medium — universality as a theorem about the machine | ~~#307~~; `Empirical/QM/QEC/SteaneTripleMachine.lean`; `Mathlib/QuantumInfo/CliffordTUniversal.lean` | > | ~~309~~ | QIT | QC chain (link 12) | *new (2026-11-05, residue of ~~#308~~)* **A gate count for the machine's universality.** ▸ **DONE 2026-11-06 (M); residues #310, #311.** The exact product was counted in place: `prodLe` (products of at most `n` factors), ★ `ctrlOf_mem_prodLe` (six), `ctrlCount`, ★★ `ctrlSetOf_mem_prodLe` ([`ControlRecursion.lean`](../CsdLean4/Mathlib/QuantumInfo/ControlRecursion.lean)); `exists_multiCtrl_sandwich` now records `L.length < m`; ★ `twoLevel_mem_prodLe`, `elemCount`, ★★ `mem_prodLe_elementary` (**every unitary on `m` qubits is a product of at most `elemCount m` elementary gates**, `306` on two qubits; [`CliffordTUniversal.lean`](../CsdLean4/Mathlib/QuantumInfo/CliffordTUniversal.lean)). [`Mathlib/QuantumInfo/CliffordTCount.lean`](../CsdLean4/Mathlib/QuantumInfo/CliffordTCount.lean): ★ `exists_norm_gateOf_sub_le` (the lift is bounded, constant uncomputed), ★ `prod_ctCircuit` (a word is a circuit over `H`, `T`, `CNOT`, `T⁻¹` as `T⁷`), ★ `exists_circuit_near_prod`, ★★★ `exists_cliffordT_circuit_polylog` (**every unitary on `m ≥ 1` qubits, after one phase, is within `ε` of a Clifford+T circuit of at most `K · log(1/ε)^c` gates**, operator norm, `c = skExponent ∈ (3, 4)`). In [`SteaneMachineDensity.lean`](../CsdLean4/Empirical/QM/QEC/SteaneMachineDensity.lean): ★ `exists_gadget_hGateM` / `_cnotGate'` / `_tGateM` (each gate one gadget), ★ `exists_gadgets_of_cliffordT`, `isCodeState_dataEnc`, ★★★ `steane_triple_machine_count` (**at most `K · log(1/ε)^c` gadgets realise exactly a logical gate within `ε` of the target up to phase, and give its output with probability at least `1 − 21 N p²`**). 43 pins. Scope: `K`, `e` existential (#116's net, the lift constant); `K` exponential in `m` through `elemCount`; the two errors stated apart (#311). ▸ *Original row:* `steane_triple_machine_universal` is topological: every two-qubit unitary is a limit of realised gates, with no bound on the circuit length `N` for a given accuracy, so `steane_triple_machine`'s `1 − 21 N p²` cannot yet be read as a function of the accuracy. Deliverables: a Solovay–Kitaev count (`Mathlib/QuantumInfo/SolovayKitaevCount.lean`) composed with `cliffordTm_le_realised` — for every two-qubit unitary and `ε > 0`, a realised gate within `ε` (in a stated norm, up to phase) from a circuit of length `O(log^c(1/ε))` — and the threshold statement for that circuit. ⚠️ **Stuck on:** nothing identified; the price is reading the corpus's Solovay–Kitaev count on two qubits (it is stated for the single-qubit `SU(2)` net) and the two-level decomposition's gate count. ~~⚠️~~ closed — the count was threaded through the exact product in place. | ~~**M**~~ done (M; residues #310, #311) | Medium — the machine's overhead for a target accuracy | ~~#308~~; `Mathlib/QuantumInfo/SolovayKitaevCount.lean`; `Empirical/QM/QEC/SteaneMachineDensity.lean` | > | 310 | QIT | QC chain (link 12) | *new (2026-11-06, residue of ~~#309~~)* **Explicit constants for the gate count.** `exists_cliffordT_circuit_polylog`'s `K` and `e` are existential: #113's net of words comes from compactness with no bound on its size, #116's determinant radius from continuity, and the lift bound `C` of `exists_norm_gateOf_sub_le` from finite dimension. Deliverables: `‖gateOf j A‖ = ‖A‖` in the operator norm (so `C = 1`), an explicit radius in `exists_det_radius`, and an explicit Clifford+T `e₀`-net of `SU(2)` with a bound on its word length, giving numerical `K` and `e`. ⚠️ **Stuck on:** the net — an explicit covering of `SU(2)` by Clifford+T words (a computed net, or a constructive density argument with rates); the other two parts are S. | **L** (net) · S (lift, radius) | Medium — a number for the machine's overhead | ~~#309~~, #113, #116; `Mathlib/QuantumInfo/CliffordTCount.lean`, `Mathlib/QuantumInfo/CliffordTNet.lean` | > | ~~311~~ | QIT | QC chain (link 12) | *new (2026-11-06, residue of ~~#309~~)* **One error for the machine's output.** ▸ **DONE 2026-11-06 (S).** In [`SteaneMachineDensity.lean`](../CsdLean4/Empirical/QM/QEC/SteaneMachineDensity.lean): `dataIsom` (append the ancilla's `|0⟩`), `dataIsom_isometry`, `dataEncCh` (**the encoding of data states is a one-Kraus channel**), `dataEncCh_apply`; ★★ `traceDist_dataEnc_le` (**encoded outputs of two logical gates are within twice the gates' operator-norm distance in trace distance** — `channel_traceDist_le` then `traceDist_conj_sub_le`); `steane_triple_machine_count` **strengthened in place** with a third conjunct: for every encoded data state the run's output is within trace distance `2ε` of the target's with probability at least `1 − 21 N p²`. `Euler.expI_smul_mem_unitaryGroup` generalised from `Fin 2` to any index type. 9 pins. Scope: the trace-norm form of the deliverable is the trace distance (half the trace norm), so the bound reads `D ≤ 2ε`; the inherited scope of #309 stands. ▸ *Original row:* `steane_triple_machine_count` keeps its two errors apart: the realised logical gate `W` is within `ε` of the target (operator norm, up to phase), and the faulty run gives `W`'s output exactly with probability at least `1 − 21 N p²`. Deliverable: the composed statement on states — for every encoded data state, the run's output is within `2ε` (trace norm) of the target's output with probability at least `1 − 21 N p²`, `N ≤ K · log(1/ε)^c` — from `‖W X Wᴴ − V X Vᴴ‖ ≤ 2‖W − e^{iφ}V‖ ‖X‖₁` and the encoder's isometry. ~~⚠️ **Stuck on:** nothing identified; the price is a trace-norm conjugation bound on the encoded register.~~ closed — the corpus had the conjugation bound (`traceDist_conj_sub_le`) and data processing. | ~~**S**~~ done (S) | Medium — the threshold statement in one number | ~~#309~~; `Empirical/QM/QEC/SteaneMachineDensity.lean`, `Mathlib/QuantumInfo/TraceDistance.lean` | > | ~~312~~ | QIT | QC chain (link 12) | *new (2026-11-06, split from #231)* **The ideal decoder and one correct rectangle.** ▸ **DONE 2026-11-07 (M took M).** In [`SteaneRectangle.lean`](../CsdLean4/Empirical/QM/QEC/SteaneRectangle.lean): ★★ `steaneDecode` (**the ideal decoder as a channel from seven qubits to one**: Steane's method, then `Eᴴ`, the complement of the code sent to `|0⟩` so it is trace preserving on every input), ★★ `steaneDecode_apply_cssMix` (a CSS deviation of a code state `ψ` decodes to `Eᴴ ψ E`), ★★ `ConcatGadget.steaneDecode_ideal` (**`D(g ψ g†) = L D(ψ) L†`** for `H̄`, `X̄`, `Z̄`, `S`, `S†`, the encoder's scalar cancelling); **finding:** the transversal `S` does **not** carry `cssMix` (`X_a Z_b ↦ X_a Z_a Z_b`), so the faults are cut down to ★ `qubitMix` (every Kraus operator on one qubit), which every single fault produces (★★ `deviatesBy_recUnder_single_qubitMix`, ★★ `ConcatGadget.deviatesBy_faultyMap`) and every transversal gate of the set carries (★★ `ConcatGadget.propagatesD_qubitMix`); `RectLoc` (★ `card_rectLoc`: **403 locations**, both recoveries' `198` and the gadget's `7`), ★★★ `rectRun_eq`, ★★★ `steane_rect_correct` (**a rectangle with at most one fault among its 403 locations, followed by the decoder, is the decoder followed by the logical gate**, on every input the leading recovery accepts: a CSS deviation of a code state, the code state itself when the leading recovery holds the fault). 42 pins. **Remainder:** every input, not only the accepted ones, is #313's price (not a new row); the two-block `CNOT̄` rectangle with a pair decoder (S–M, on #315's path, carried there with ~~#302~~). ▸ *Original row:* The level-reduction step needs the level-one statement per rectangle, not per circuit: an ideal Steane decoder `D` (a channel from the seven qubits to one: #86's recovery, then `Eᴴ`), and for each of `H̄`, `X̄`, `Z̄`, `S̄` an extended rectangle with **every recovery location counted** (#233's `recUnder`, `window`) such that a good rectangle (at most one fault) satisfies `D ∘ exRec = 𝒰_g ∘ D` on every input the leading recovery accepts. `steane_machine_overlap` has this only for whole circuits on code-state inputs. ⚠️ **Stuck on:** nothing identified; the price is restating `isFaultChain_overlap` per rectangle with the decoder in place of the code-state input. | ~~**M**~~ done | Medium — the per-rectangle correctness the level reduction consumes | #231; ~~#233~~; ~~#96~~; `Empirical/QM/QEC/SteaneOverlapMachine.lean`; `Mathlib/QuantumInfo/ExtendedRectangle.lean` | > | ~~313~~ | QIT | QC chain (link 12) | *new (2026-11-06, split from #231)* **A bad rectangle is one faulty location one level down.** ▸ **DONE 2026-11-07 in the true form (A) (M–L took M; author decision 2026-11-07).** ⚠️ **The row's statement is false as written:** with faults in the leading recovery a rectangle can read the incoming syndrome, so two inputs with one decoded state can give different outputs and no `Φ` on the decoded qubit alone exists. What holds, in [`SteaneBadRectangle.lean`](../CsdLean4/Empirical/QM/QEC/SteaneBadRectangle.lean): ★★ `steaneProj_mul_recKraus` (**every branch of the recovery lands in the code**: each stabiliser picks up on the branch's syndrome spaces the sign it picks up commuting past the correction, `pauliMat_steane_mul_recKraus`, through `bdot_rowComb` and `syndrome_decode` — the Hamming code is perfect), ★★★ `methodRecovery_apply_eq_encode` (**`R(σ) = E D(σ) Eᴴ` on every input**), ★★ `steaneDecode_methodRecovery` (**`D ∘ R = D` on every input**), ★ `steaneDecode_encode` (`D ∘ Enc = id`), ★★★ `steaneDecode_apply_after_recovery` (any channel after a fault-free recovery is `D ∘ Rect ∘ Enc` on the decoded qubit), ★★★ `steane_badRect` (**a rectangle with a fault-free leading recovery and any faults in its gadget and trailing recovery is a channel on the decoded qubit, on every input**), ★★★ `steane_rect_correct_clean` (~~#312~~ on every input when the leading recovery is fault-free). 26 pins. **Remainder (B):** Aliferis–Gottesman–Preskill's general form — a unitary syndrome-keeping decoder `D̃` to the logical qubit and six syndrome bits, every rectangle `D̃⁻¹ Φ̃ D̃` with `Φ̃` acting on the qubit and the syndrome as its environment (M; whether #314 needs it depends on #314's noise model). ▸ *Original row:* For an extended rectangle with any faults, `D ∘ exRec = Φ ∘ D` for some channel `Φ` on the decoded qubit(s) — Aliferis–Gottesman–Preskill's reduction of a bad rectangle to a faulty level-`(k−1)` location — from a decoder that absorbs the trailing recovery (`D ∘ R = D`) and the channel form of the rectangle on every input. ⚠️ **Stuck on:** nothing identified; the price is the decoder identity on arbitrary (not code) inputs. | ~~**M–L**~~ done (A) | Medium — what lets bad rectangles become faults of the coarser circuit | #231; #312 | > | 314 | QIT | QC chain (link 12) | *new (2026-11-06, split from #231)* **The level reduction, abstractly.** ▸ **Split 2026-11-07 (author decision) into #316–#319**: two design forks the row left open — the decoder between rectangles must keep the syndrome (#313's form (A) needs a fault-free preceding recovery, an `O(p)` event), and the noise model must survive one reduction (bad rectangles sharing a recovery are correlated, so independent faults do not recurse; Aliferis–Gottesman–Preskill's local stochastic noise does). This row closes when #316–#319 do. ▸ *Original row:* A circuit model with locations (gates, preparations, measurements and classical feed-forward of their outcomes), a compiler replacing each location by its level-one rectangle, and the theorem: if every good rectangle is correct (#312's form) and every bad one is a faulty location (#313's form), the decoded level-`k` run equals a level-`(k−1)` run with faults exactly at the bad rectangles; with rectangles bad with probability at most `A p²`, the recursion `p_k ≤ A p_{k−1}²` and a threshold `1/A`, every recovery location at every level counted. Category 1 (code-free), the counterpart of `LevelReduction.lean`'s `IsCorrectedAtLevel` with the code changing by level. ⚠️ **Stuck on:** nothing identified; the price is the circuit model with classical feed-forward and the compiler. | **L** | High — the threshold theorem's recursion with recoveries counted | #231; ~~#97~~; `Mathlib/QuantumInfo/LevelReduction.lean` | > | 315 | QIT | QC chain (link 12) | *new (2026-11-06, split from #231)* **The Steane threshold with recoveries counted at every level.** Plug the Steane code into #314: a rectangle for every location Steane's method uses — the transversal `CNOT̄` between data and ancilla blocks (#302), ancilla preparation and transversal measurement (~~#236~~'s factory and read-out as rectangles), and the single-block gates (#312, #313) — and the constant `A` from their malignant pairs; the result supersedes `steane_machine_recursive`'s ideal recoveries. ⚠️ **Stuck on:** #332 and #324 (#319's open children; ~~#323~~ done 2026-11-08), then the preparation, measurement and two-block `CNOT̄` rectangles in the relative form (re-read 2026-11-08: ~~#322~~ done; 2026-11-07: ~~#302~~, ~~#312~~, ~~#313~~, ~~#316~~–~~#318~~, ~~#320~~, ~~#321~~ done). | **L** | High — the concrete threshold theorem with nothing ideal but classical decoding | #231; #302; #312; #313; #314; ~~#236~~ | > | ~~316~~ | QIT | QC chain (link 12) | *new (2026-11-07, split from #314)* **The decoder algebra between rectangles, abstractly.** ▸ **DONE 2026-11-07 (M took M).** ▸ **Amended 2026-11-07 (with ~~#317~~):** `IsGoodRect` (every input) fails for the transversal `S` at the Steane code (`exists_steaneDecode_sTransversal_ne`), so the chain now runs on Aliferis–Gottesman–Preskill's relative form — `IsGoodAfter` (`D ∘ T' ∘ G ∘ T = 𝒰 ∘ D ∘ T`, `T` the leading recovery), `recRun`, ★★★ `decode_recRun_of_isGoodAfter`, ★★★ `one_sub_le_measure_decode_recRun` (6 more pins). In [`SyndromeDecoder.lean`](../CsdLean4/Mathlib/QuantumInfo/SyndromeDecoder.lean) (Category 1): `synDecode W ρ = Tr_s(W ρ W†)` for a unitary `W : n ≅ m × s`, `liftRect` (`Φ̃ = W R(W†·W) W†`), ★★ `conj_ovIdeal` / ★★ `synDecode_ovIdeal` (**the decoded run is the lifted run on logical ⊗ syndrome, traced at the end, exactly**), `IsGoodRect` (`D ∘ Rect = 𝒰 ∘ D` on **every** input), `IsLogicalIdeal` (`Tr_s ∘ Φ̃ = 𝒰 ∘ Tr_s`), ★★ `isGoodRect_iff_isLogicalIdeal` (**good exactly when the lift is ideal on the logical register, whatever it does to the syndrome**), ★ `isLogicalIdeal_kron_one` (`U ⊗ 1`; not vacuous), ★★★ `synDecode_ovIdeal_of_good` (all good ⇒ ideal logical run on every input), ★★★ `one_sub_le_measure_synDecode_eq` (**joined with ~~#318~~: under local stochastic noise, if every rectangle that is not bad is good, the decoded run is ideal on every input with probability at least `1 − N A p²`**). 13 pins. Scope: maps, not channels; goodness a hypothesis (#317 at the Steane code); one level — next-level rectangles good for environment-coupled faults is #319. ▸ *Original row:* A decoder `D` and encoder `E` with `D ∘ E = id`, a syndrome-keeping isometry `D̃` with `D = Tr_syn ∘ D̃`, and the exact identity: the decoded run of a chain of rectangles is the level-`(k−1)` run of the maps `Φ̃ⱼ = D̃ Rectⱼ D̃⁻¹` on the logical register and the syndrome; a good rectangle's `Φ̃ⱼ` is the ideal gate on the logical register (any channel on the syndrome). Category 1. ⚠️ **Stuck on:** nothing identified; the price is stating good-rectangle correctness so that it composes with the syndrome left over. | ~~**M**~~ done | High — the exact form of the level reduction | #314; ~~#313~~; ~~#312~~; `Mathlib/QuantumInfo/OverlappingRectangles.lean` | > | ~~317~~ | QIT | QC chain (link 12) | *new (2026-11-07, split from #314; narrowed 2026-11-07, author decision)* **Steane's syndrome-keeping decoder.** ▸ **DONE 2026-11-07 (narrowed M took M).** ⚠️ **Finding:** the row as first written — a good rectangle correct on **every** input — is **false for the transversal `S`**: on `X₀ Z₁` applied to an encoded `|+⟩`, `S` makes the error `X₀ Z₀ Z₁`, whose phase syndrome names qubit `2`, and the residual `Z₀ Z₁ Z₂` is a logical `Z̄` (★★★ `exists_steaneDecode_sTransversal_ne`). Aliferis–Gottesman–Preskill's correctness is relative to the leading recovery, which ~~#316~~ now carries (`IsGoodAfter`); the faulty recovery's output on arbitrary inputs moved to #320. In [`SteaneSyndromeDecoder.lean`](../CsdLean4/Empirical/QM/QEC/SteaneSyndromeDecoder.lean): `synCorr`, `steaneW` (row `(a, st)` = `synCorr st · E|a⟩`), ★★ `synCorr_steaneEnc_orthonormal` (**the 128 states are orthonormal**; a Pauli with a nonzero syndrome has no matrix element between logical states), ★★ `steaneW_mul_conjTranspose` / `steaneW_conjTranspose_mul` (**`W` is unitary**), `recKraus_conjTranspose_mul_steaneEnc`, ★★★ `steaneDecode_eq_synDecode` (**#312's decoder is `Tr_syn ∘ W` on every input**), `sTransversal_conj_pauliMat` (`S X^a Z^b S† = i^{|a|} X^a Z^{b+a}`), `steaneDecode_pauliMat_conj`, `steaneEnc_conjTranspose_mul_logicalZ_stab`. 22 pins. | ~~**M**~~ done | High — the decoder #316 needs at the Steane code | #314; ~~#316~~; ~~#313~~; #320 | > | ~~318~~ | QIT | QC chain (link 12) | *new (2026-11-07, split from #314)* **Local stochastic noise and its level reduction.** ▸ **DONE 2026-11-07 (M took M).** In [`LocalStochastic.lean`](../CsdLean4/Mathlib/QuantumInfo/LocalStochastic.lean) (Category 1): `IsLocalStochastic μ p` (every set `S` entirely faulty with probability at most `p^|S|`), ★★ `isLocalStochastic_pi` (**independent faults are an instance**), ★ `IsLocalStochastic.measure_exists_le` (union bound); `RectChain` (windows, truncated windows, malignant pairs; windows two apart disjoint, a truncated window disjoint from the one before), ★ `badSeq` (**Aliferis–Gottesman–Preskill's sequential badness**: the truncated window when the predecessor is bad, so the shared recovery counts once), ★★★ `measure_forall_badSeq_le` (**every set `T` of rectangles is entirely bad with probability at most `(A p²)^|T|`** — the witnesses can be chosen pairwise disjoint), ★★★ `isLocalStochastic_map_reducePat` (**one reduction carries parameter `p` to `A p²`**), `lsBound`, ★★ `lsBound_induction`, ★★ `ofReal_codeCapacityBound` (= #51's `(A p)^{2^k}/A`), ★★★ `tendsto_lsBound` (**below `p < 1/A` the parameter tends to `0`**), ★ `RectChain.ofSteps` (#233's overlapping chain, every pair of a window malignant, `C(w, 2)` pairs). 24 pins. Scope: counting only — that a rectangle with no faulty malignant pair is correct is #316/#317; a chain of single-block rectangles, not the multi-block geometry (#319). | ~~**M**~~ done | High — the recursion `p_k ≤ A p_{k−1}²` | #314; ~~#51~~; ~~#233~~ | > | 319 | QIT | QC chain (link 12) | *new (2026-11-07, split from #314)* **The circuit model and the compiler.** ▸ **Split 2026-11-08 (author decision) into #322–#324**: the row held three pieces at three prices — the badness geometry for multi-block circuits (~~#322~~, done), the exact one-level reduction (#323) and the recursion with environment-coupled faults (#324). Scoping found that an ideal decoder moved from the end of the circuit towards its start truncates a rectangle's **trailing** recovery when its successor is bad (the bad rectangle's faulty location absorbs the shared recovery), so the truncated rectangle's correctness is `IsGoodAfter D U T G id`. This row closes when #322–#324 do (and #332, split from #323 on 2026-11-08). ▸ *Original row:* Circuits with locations — gates on one or two blocks, preparations, measurements and classical feed-forward of their outcomes — a compiler replacing each location by its rectangle, the rectangles' windows for multi-block gates (a `CNOT̄` rectangle shares recoveries with two predecessors, so #318's chain becomes a graph of bounded degree), and the theorem that joins #316 (exact decoded run) and #318 (local stochastic badness) into `p_k ≤ A p_{k−1}²` with every recovery location at every level counted. Category 1. ⚠️ **Stuck on:** nothing since ~~#316~~ (2026-11-07); the price is the circuit model with feed-forward and next-level goodness for environment-coupled faults. | **M–L** | High — the abstract threshold theorem | #314; ~~#316~~; ~~#318~~; ~~#97~~ | > | ~~320~~ | QIT | QC chain (link 12) | *new (2026-11-07, split from #317; narrowed 2026-11-07, author decision)* **The decoder on single-type syndromes, and good rectangles in the relative form.** ▸ **DONE 2026-11-07 (narrowed M took M).** **Finding:** the right condition is on syndromes, not qubits — a faulty ancilla can leave errors on different qubits in different branch components; what every gate respects is that the output's syndrome pairs are **single-type** (`s = 0`, `t = 0` or `s = t`: the syndromes of `X_q`, `Z_q`, `Y_q`), on which `decode s + decode t = decode (s + t)`. In [`SteaneSingleType.lean`](../CsdLean4/Empirical/QM/QEC/SteaneSingleType.lean): `SingleType`, `IsSingleTypeOp` (the rows of ~~#317~~'s `W` with non-single-type syndromes annihilate `K`), ★ `isSingleTypeOp_of_steaneProj_mul` (code-landing operators, so every fault-free recovery branch, `isSingleTypeOp_recKraus`), `ConcatGadget.synPerm` (the gates on syndrome pairs: `H̄` swaps, `X̄`, `Z̄` fix, `S`, `S†` send `(s, t)` to `(s, s + t)`), ★★ `ConcatGadget.exists_conj_synCorr` (**`g C_st g† = ω C_{π st}` on single-type pairs**), ★★★ `ConcatGadget.steaneDecode_conj_of_isSingleTypeOp` / `steaneDecode_conj_channel` (**`D(g K ρ K† g†) = L D(K ρ K†) L†` for single-type `K`, every `ρ`**), ★★★ `steane_rect_isGoodAfter` (now `_of_singleTypeOut`) (**every rectangle of `H̄`, `X̄`, `Z̄`, `S`, `S†` with at most one fault among its 403 locations is good after its leading recovery, on every input**, given `RecNoise.SingleTypeOut` for the leading recovery's noise; renamed `steane_rect_isGoodAfter_of_singleTypeOut` when ~~#321~~ took the name). 18 pins. **Remainder:** `SingleTypeOut` is a hypothesis → ~~#321~~ (discharged 2026-11-07). | ~~**M**~~ done | High — `IsGoodAfter` at the Steane code | #314; #315; ~~#316~~; ~~#317~~; #321 | > | ~~321~~ | QIT | QC chain (link 12) | *new (2026-11-07, split from #320, author decision)* **Steane's method with one fault has single-type branches, on every input.** ▸ **DONE 2026-11-07 (M–L took M).** In [`SteaneSingleTypeOut.lean`](../CsdLean4/Empirical/QM/QEC/SteaneSingleTypeOut.lean): `phaseProj_mul_pauliMat_zero`, ★ `bitProj_mul_of_commute` (an operator commuting with every `Z`-stabiliser commutes with every bit-syndrome projector, so `bitProj_mul_phaseProj`), ★ `gateOf_mul_bitProj` / `gateOf_mul_phaseProj` (an operator on qubit `q` moves either syndrome by `0` or `h_q`), `qubitSynProj` and ★ `isSingleTypeOp_of_qubitSynProj_mul`, ★★ `qubitSynProj_mul_phaseFaultOp_mul_bitProj_zero` (**the phase round with an operator on the pair `q`, fed bit syndrome `0`, lands in the syndrome spaces of `I`, `X_q`, `Z_q`, `Y_q`**), ★★ `bitProj_zero_mul_ancPhaseOp_mul_bitProj_zero` (**a periodic ancilla keeps bit syndrome `0`**), ★★★ `isSingleTypeOp_recKraus'_single` (**every branch with one fault anywhere is single-type, on every input**), ★★★ `RecNoise.singleTypeOut`, ★★★ `steane_rect_isGoodAfter` (**~~#320~~'s rectangle theorem with no hypothesis**: for each of `H̄`, `X̄`, `Z̄`, `S`, `S†`, every rectangle with at most one fault among its 403 locations is good after its leading recovery, on every input). **Simpler than the route rowed:** a fault in the bit round needs no analysis of the bit round, since the fault-free phase round lands every output on phase syndrome `0`; only a faulty phase round, fed bit syndrome `0`, needs the one-qubit analysis, and a periodic ancilla (which may carry `X` errors on different qubits in different branches) leaves bit syndrome `0`. 35 pins (and ~~#320~~'s conditional theorem renamed `steane_rect_isGoodAfter_of_singleTypeOut`, its pin with it). | ~~**M–L**~~ done | High — the last Steane hypothesis of the relative-form level reduction | #320; #315; #319; ~~#235~~; ~~#238~~; ~~#239~~; ~~#240~~; `Empirical/QM/QEC/SteaneOverlapMachine.lean`, `Empirical/QM/QEC/SteaneVerifiedAncilla.lean` | > | ~~322~~ | QIT | QC chain (link 12) | *new (2026-11-08, split from #319, author decision)* **Badness for any family of rectangles.** ▸ **DONE 2026-11-08 (S–M took S–M).** In [`RectGeometry.lean`](../CsdLean4/Mathlib/QuantumInfo/RectGeometry.lean) (Category 1): `RectFamily` (windows `W j` and malignant pairs `M j` in them, **nothing assumed about how the windows overlap**), `claimed` (the windows of the bad rectangles met so far, in decoding order), ★ `bad` (**a faulty malignant pair missing every claimed window**), `trunc` and ★ `bad_eq_true_iff` (equivalently, a faulty malignant pair in the truncated window), `trunc_eq_of_disjoint`, ★ `disjoint_of_bad` (witnesses of two bad rectangles are disjoint), ★★★ `measure_forall_bad_le` (**every set `T` of rectangles entirely bad with probability at most `(A p²)^|T|`, for any family of windows**), ★ `measure_bad_le`, ★★ `measure_exists_bad_le` (`|T| A p²`), ★★★ `isLocalStochastic_map_reducePat` (**`p ↦ A p²` for any family**), `ofWindows` / ★ `ofWindows_card_M_le` (every pair malignant, `C(w, 2)`). So a `CNOT̄` rectangle's two predecessors and two successors cost nothing beyond the count `A` of its pairs. 23 pins. Scope: counting only; not proved equal to #318's `badSeq`; the circuit model that fixes the windows and the order is #323. | ~~**S–M**~~ done | High — #318's bound off the chain | #319; #323; #324; ~~#318~~; ~~#316~~; #315 | > | ~~323~~ | QIT | QC chain (link 12) | *new (2026-11-08, split from #319, author decision)* **The exact one-level reduction.** ▸ **DONE 2026-11-08 for the serialised circuit (M took M); the two-block geometry split to #332 (author decision).** In [`ExactReduction.lean`](../CsdLean4/Mathlib/QuantumInfo/ExactReduction.lean) (Category 1): `inRec` / `outRec` / `stage` / `runInit` (**who owns each recovery**: a bad rectangle absorbs its leading recovery, so the good rectangle before it is checked without its trailing one), ★★ `recRun_eq_ovIdeal_stage` (**the run regroups into its stages**, for any set of bad rectangles), ★★★ `conj_recRun_eq` (**the decoded run is the run of the lifted stages on logical ⊗ syndrome, exactly**), ★★★ `partialTraceRight_liftRect_stage` (**a good location of the reduced run is its ideal gate on the input the run feeds it**, given `IsGoodAfter D U (T j) (G j) (outRec …)` — the truncated `IsGoodAfter D U T G id` when the successor is bad), `timeBad` (~~#322~~'s badness with the chain in decoding order), ★★★ `exact_reduction_of_rectFamily` (the identity and the ideal locations for every fault pattern), ★★ `one_sub_le_measure_forall_timeBad_eq_false` (`1 − N A p²`), `decode_recRun_of_forall_timeBad_eq_false` (no bad rectangle ⇒ ~~#316~~'s ideal run). 15 pins. Scope: arbitrary maps on one register, so measurements and classical feed-forward enter as maps on a register holding the classical bits; goodness (full and truncated) is the code-specific hypothesis — the truncated form at the Steane code is not proved; bad locations act on logical ⊗ syndrome, so the reduced run is on `m × s` (the next level is #324). | ~~**M–L**~~ done (serialised) | High — the level reduction's exact form | #319; ~~#322~~; #332; #324; ~~#316~~; ~~#318~~; #315 | > | 332 | QIT | QC chain (link 12) | *new (2026-11-08, split from #323, author decision)* **Two-block rectangles in the exact reduction.** In ~~#323~~'s serialised run a recovery is shared only with the next stage. A `CNOT̄` rectangle shares a recovery on block `b` with the next rectangle on `b`, which need not be next in the serialised run; absorbing it into that rectangle when it is bad needs the stages in between, on other blocks, to commute with it. Maps with tensor structure over the blocks (a decoder `⊗_b W_b`, gadgets on one or two blocks, recoveries on one), the commutation of maps on disjoint blocks, and ~~#323~~'s identity with each recovery owned by the rectangle that absorbs it; the windows and order feed ~~#322~~'s `RectFamily` unchanged. Category 1. ⚠️ **Stuck on:** nothing identified; the price is the tensor-structured maps. | **L** | High — the circuit model #315 needs | #319; ~~#323~~; ~~#322~~; #324; #315 | > | 324 | QIT | QC chain (link 12) | *new (2026-11-08, split from #319, author decision)* **The recursion with environment-coupled faults.** #323's reduced run has faulty locations acting on their blocks and the syndromes as environment, and fault-free locations that are only logically ideal (`Tr_s ∘ Φ̃ = 𝒰 ∘ Tr_s`); the next level's goodness must hold for such locations. That needs the maps to be linear (today they are arbitrary functions, "maps, not channels"), so that an environment-coupled fault decomposes into system operators tensored with environment matrix units, and the code-specific half (the Steane rectangles good for such faults) is #315's. Then `p_k ≤ A p_{k−1}²` with every recovery location at every level counted, through ~~#318~~'s `lsBound_induction`. Category 1. ⚠️ **Stuck on:** #323; the price is the linear-map refactor of the decoder algebra and the class of logically ideal locations. | **L** | High — the abstract threshold theorem | #319; #323; ~~#322~~; ~~#316~~; ~~#318~~; #315 | > | ~~325~~ | QIT | One QM, many formulations | *new (2026-11-08, the author's geometric-QM plan of 2026-11-08 (vault: *PLAN - Geometric QM gaps in csd-lean4 2026-11-08*), item 2; author decision "all phases")* **The Poisson bracket of observables on `ℂℙⁿ` is the commutator.** ▸ **DONE 2026-11-08 (M took S–M).** In [`ProjectiveSpacePoissonBracket.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpacePoissonBracket.lean) (Category 1): `IsSymplectic.poissonBracket` (`{f, g} = ω(X_f, X_g)`), `_eq_mfderiv`, `_eq_of_isHamiltonianVectorField`, `_antisymm`, `_self`, ★★ `IsSymplectic.hasDerivAt_comp_poissonBracket` (**`d/dt f(γ t) = {f, g}(γ t)` along the flow of `g`**), `commObs A B = -i[A, B]`, ★★ `hasDerivAt_expectation_schrodingerUnitary_smul` / ★★ `mfderiv_expectation_schrodingerField` (**Ehrenfest: `d⟨A⟩(X_B) = ⟨-i[A, B]⟩`**), ★★★ `poissonBracket_schrodingerHamiltonian` (**`{-2⟨A⟩, -2⟨B⟩} = -2⟨-i[A, B]⟩`**), ★★★ `poissonBracket_expectation` (**`{⟨A⟩, ⟨B⟩} = -½ ⟨-i[A, B]⟩`**), `expectation_eq_zero_iff` (a Hermitian `C` with `⟨C⟩ ≡ 0` is `0`), ★★★ `expectation_conserved_iff_commute` (**`⟨A⟩` conserved by the flow of `H` iff `AH = HA`**). 24 pins. Scope: `B` Hermitian (the generator), `A` any matrix in the bracket identities; the constants are the corpus's `-2` convention; no Jacobi identity for general functions. ▸ **Item 1 of the plan (Kähler) was already in the corpus:** `fsForm_isKahler` (G14a, `ProjectiveSpaceFubiniStudySymplectic.lean`) proves `IsKahler fsForm fsJ modelJ` — `J = i·` in every chart, with the Cauchy–Riemann equations of the atlas (`IsKahler.fderiv_chart_transition_comm`); the plan's audit read only the Riemannian module. | ~~**M**~~ done | Medium — Kibble's dictionary, brackets | #326; #327; #328; #329; ~~#30~~; `Instances/ProjectiveSpaceSchrodingerFlow.lean` (G13) | > | ~~326~~ | QIT | One QM, many formulations | *new (2026-11-08, the author's geometric-QM plan of 2026-11-08 (vault: *PLAN - Geometric QM gaps in csd-lean4 2026-11-08*), item 3)* **Eigenstates are the critical points of `⟨A⟩`.** ▸ **DONE 2026-11-08 (S–M took S–M).** In [`ProjectiveSpaceCriticalPoints.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceCriticalPoints.lean) (Category 1): ★ `IsHamiltonianVectorField.eq_zero_iff_mfderiv_eq_zero` (for a symplectic form `X_H` vanishes exactly where `dH` does), ★ `schrodingerChartField_eq_zero_iff`, ★★ `schrodingerField_eq_zero_iff` (**`X_A(p) = 0` iff `p` is an eigenline**, any `A`), `expectation_eq_of_eigen`, `eigen_eq_expectation_smul`, ★★ `mfderiv_expectation_eq_zero_iff_schrodingerField`, ★★★ `mfderiv_expectation_eq_zero_iff` (**`d⟨A⟩_p = 0 ⟺ Aψ = ⟨A⟩_p ψ`**), ★★★ `criticalValues_eq_eigenvalues` (**the critical values are exactly the real eigenvalues with an eigenline**), ★ `mfderiv_expectation_eq_zero_of_isMaxOn` / `_isMinOn` (Fermat on `ℂℙⁿ` through the chart Hamiltonian; Mathlib has no manifold Fermat), ★★★ `exists_isGreatest_eigenvalues` / `exists_isLeast_eigenvalues` (**`⟨A⟩` attains its maximum and minimum at eigenlines, and they are the largest and smallest eigenvalues**). 16 pins. Scope: the eigenvalue set is `{μ | ∃ p, A p.rep = μ p.rep}`, not identified with `Matrix.IsHermitian.eigenvalues`; no Morse indices. The route went direct (`X_A = 0` iff eigenvector, in the chart) rather than through the flow `exp(-itA)` the plan named. | ~~**S–M**~~ done | Medium — spectrum as critical points | ~~#325~~; `Instances/ProjectiveSpaceSchrodingerFlow.lean` (G13) | > | ~~327~~ | QIT | One QM, many formulations | *new (2026-11-08, the author's geometric-QM plan of 2026-11-08 (vault: *PLAN - Geometric QM gaps in csd-lean4 2026-11-08*), item 6)* **The quantum geometric tensor.** ▸ **DONE 2026-11-08 (S–M took S).** In [`ProjectiveSpaceGeometricTensor.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceGeometricTensor.lean) (Category 1): `qgt x u v = g(u, v) − i ω(u, v)`, `qgt_re` (the metric), `qgt_im` (minus the form), `fsModelForm_qgt`, ★★★ `qgt_eq` (**Provost–Vallée: `Q(u, v) = 4(⟪U, V⟫/‖ψ‖² − ⟪U, ψ⟫⟪ψ, V⟫/‖ψ‖⁴)`** with `ψ` the chart representative and `U`, `V` the lifted tangent vectors), ★★ `qgt_eq_inner_proj` (`4⟪U, (1 − P_ψ)V⟫/‖ψ‖²`), ★ `qgt_conj` (Hermitian), `qgt_self`, ★ `qgt_self_re_pos` (positive definite). 9 pins. Scope: with the corpus's `fsForm` the Hermitian combination is `g − iω = 4h` (the textbook `g + iω` has the other sign of `ω`); read in the chart of the point. | ~~**S–M**~~ done | Medium — metric and Berry curvature in one object | ~~#325~~; ~~#10~~; `Instances/ProjectiveSpaceFubiniStudyRiemannian.lean` | > | ~~328~~ | QIT | One QM, many formulations | *new (2026-11-08, the author's geometric-QM plan of 2026-11-08 (vault: *PLAN - Geometric QM gaps in csd-lean4 2026-11-08*), item 7)* **The unitary moment map.** ▸ **DONE 2026-11-08 (M took M).** In [`ProjectiveSpaceUnitaryMomentMap.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceUnitaryMomentMap.lean) (Category 1): `rankOneMatrix v = ‖v‖⁻² v v*` (Hermitian, trace one, idempotent, positive semidefinite, `P_{Uv} = U P_v U†`), `unitaryMomentMap p = rankOneMatrix p.rep`, ★ `unitaryMomentMap_smul` (**`μ(U • p) = U μ(p) U†`**), ★★ `trace_unitaryMomentMap_mul` / `re_trace_unitaryMomentMap_mul` / `_of_isHermitian` (**`tr(μ(p) A) = ⟨A⟩_p`**), ★★ `isHamiltonianVectorField_trace_unitaryMomentMap` (**the component along `A` has Hamiltonian field `-½ X_A`, the generator of `exp(-itA)`**), ★ `unitaryMomentMap_diag` (**the diagonal is the torus moment map**), ★★ `unitaryMomentMap_injective`, ★★★ `range_unitaryMomentMap` (**the image is exactly the Hermitian idempotents of trace one**: subtract the projector of a column; what is left is a trace-zero Hermitian idempotent, hence `0`), ★★★ `convexHull_range_unitaryMomentMap` (**the convex hull of the image is the density matrices**, by the spectral theorem). 24 pins. Scope: `𝔲(n+1)*` is the Hermitian matrices through the trace pairing, not a constructed Lie-algebra dual; the moment-map equation is stated component by component with the corpus's `-½`. | ~~**M**~~ done | Medium — states as projectors, mixtures as the hull | ~~#325~~; `Instances/ProjectiveSpaceMomentMap.lean` (G6) | > | ~~329~~ | QIT | One QM, many formulations | *new (2026-11-08, the author's geometric-QM plan of 2026-11-08 (vault: *PLAN - Geometric QM gaps in csd-lean4 2026-11-08*), item 4, first half)* **Uncertainty is the length of the Hamiltonian field.** ▸ **DONE 2026-11-08 (M took S–M).** In [`ProjectiveSpaceUncertainty.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceUncertainty.lean) (Category 1): `variance A p = ⟨A²⟩ − ⟨A⟩²`, ★★★ `fsMetric_schrodingerField_self` (**`g(X_A, X_A) = 4 (ΔA)²`**, from ~~#327~~'s Provost–Vallée formula), ★ `variance_nonneg`, ★★ `variance_eq_zero_iff` (**`ΔA = 0` exactly on the eigenlines**, through ~~#326~~), `variance_schrodingerUnitary_smul`, ★★★ `fsMetric_velocity_schrodingerUnitary_smul` (**the Schrödinger curve has velocity `X_H` of squared length `4 (ΔH)²_p` at every time**). 8 pins. Scope: the factor 4 is the corpus's metric normalisation (speed `ΔH` for the textbook `g/4`); the speed limit is #331. | ~~**M**~~ done | Medium — the metric reads variances | ~~#325~~; ~~#327~~; #331 | > | ~~330~~ | QIT | One QM, many formulations | *new (2026-11-08, the author's geometric-QM plan of 2026-11-08 (vault: *PLAN - Geometric QM gaps in csd-lean4 2026-11-08*), item 5)* **The Fubini–Study distance and its minimising curves.** ▸ **DONE 2026-11-08 (L took M–L).** In [`ProjectiveSpaceFSLength.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFSLength.lean) and [`ProjectiveSpaceFSDistance.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFSDistance.lean) (Category 1): ★★ `fsInnerHom_lift` (the homogeneous metric does not see the lift), ★★ `fsMetric_mfderiv_projCurve` (**the squared speed of `[ψ]` is `fsInnerHom ψ ψ' ψ'`**), ★★ `sq_dTransProb_le` (**`(dP)² ≤ P(1 − P) g`**, Cauchy–Schwarz on the components orthogonal to `ψ`); `fsENorm` (the Fubini–Study norm, a *local* instance), `fsPathLength` / `fsEDist` (**Mathlib's `pathELength` / `riemannianEDist` for that norm**), `fsAngle = arccos √P`, ★★ `hasDerivAt_fsAngle_comp` (the angle to a fixed ray changes at most at half the speed, away from `P ∈ {0, 1}`), ★★★ `ofReal_two_mul_fsAngle_le_fsPathLength_of` (**every path continuous on `[0, 1]` and differentiable inside is at least `2 arccos √P` long**: integrate the angle on a stretch where it climbs from `0` to any level below its final value; `measurable_deriv` supplies the measurability), ★★ `fsPathLength_greatCircle`, ★★★ `exists_fsPathLength_eq` (**a great circle has length `2 arccos √P`**), ★★★ `fsEDist_eq` (**`fsEDist p q = 2 arccos √P(p, q)`**), `fsEDist_le_pi`, ★ `fsEDist_eq_zero_iff`. 42 pins. Scope: the factor 2 is the corpus's normalisation (`arccos √P` for the textbook `g/4`); the norm is a local instance, so no tangent norm is imposed on importers; "minimising curves" means an explicit minimising great circle, not a geodesic theory (Mathlib has none). | ~~**L**~~ done | Medium — distances between states | ~~#329~~; #331; `LinearAlgebra/Projectivization/TransitionProbability.lean` | > | ~~331~~ | QIT | One QM, many formulations | *new (2026-11-08, the author's geometric-QM plan of 2026-11-08 (vault: *PLAN - Geometric QM gaps in csd-lean4 2026-11-08*), item 4, second half)* **The quantum speed limit.** ▸ **DONE 2026-11-08 (S took S).** In [`ProjectiveSpaceSpeedLimit.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceSpeedLimit.lean) (Category 1): `hasMFDerivAt_schrodingerUnitary_smul_mul` (the curve run at speed `T` has velocity `T X_H`), ★★ `fsPathLength_schrodingerUnitary_smul` (**its length on `[0, 1]` is `2 T ΔH`**), ★★★ `fsAngle_schrodingerUnitary_smul_le` (**Anandan–Aharonov / Mandelstam–Tamm: `arccos √P(p, exp(-iTH) • p) ≤ T ΔH`**), ★★ `arccos_div_le_time` (**the time needed is at least `arccos √P / ΔH`**). 4 pins. Scope: `ħ = 1`, time-independent `H`. | ~~**S**~~ done | Medium — the geometric speed limit | ~~#329~~; ~~#330~~ | > | ~~333~~ | QIT | One QM, many formulations | *new (2026-11-08, geometric-QM completion menu item 2, promoted by the author's "work through these")* **The Robertson–Schrödinger uncertainty relation as Cauchy–Schwarz.** ▸ **DONE 2026-11-08 (S–M took S).** In [`ProjectiveSpaceRobertson.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceRobertson.lean) (Category 1): `IsAlmostKahler.metric_add_right` / `_smul_right` / `_add_left` / `_smul_left` / `_zero_right` / `metric_self_nonneg` (the compatible metric is bilinear, with no linearity of `J` used), ★ `IsAlmostKahler.metric_sq_le` (real Cauchy–Schwarz), ★★ `IsAlmostKahler.metric_sq_add_apply_sq_le` (**`g(u, v)² + β(u, v)² ≤ g(u, u) g(v, v)`**, Cauchy–Schwarz for `g − iβ`, any almost Kähler structure), `schrodingerField_add`, `expectation_add`, `covariance` / `covariance_self`, ★★ `fsMetric_schrodingerField_schrodingerField` (**`g(X_A, X_B) = 4 Cov(A, B)`**), ★★ `fsForm_schrodingerField` (**`ω(X_A, X_B) = -2⟨-i[A, B]⟩`**), ★★★ `covariance_sq_add_sq_le_variance_mul` (**Robertson–Schrödinger: `Cov(A, B)² + (⟨-i[A, B]⟩ / 2)² ≤ (ΔA)² (ΔB)²`**), `expectation_commObs_div_two_sq_le`, ★★★ `abs_expectation_commObs_div_two_le` (**Robertson: `|⟨-i[A, B]⟩| / 2 ≤ ΔA ΔB`**). 18 pins. Scope: inequalities only (minimum-uncertainty states not characterised); Hermitian matrices on rays of a finite level system. | ~~**S–M**~~ done | Medium — the textbook uncertainty relation, geometrically | ~~#325~~; ~~#327~~; ~~#329~~ | > | ~~334~~ | QIT | One QM, many formulations | *new (2026-11-08, geometric-QM completion menu item 3, promoted by the author's "work through these")* **The critical values are Mathlib's eigenvalues.** ▸ **DONE 2026-11-08 (S took S).** In [`ProjectiveSpaceCriticalPoints.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceCriticalPoints.lean) (Category 1, added to ~~#326~~'s module): ★★ `setOf_eigenline_eq_range_eigenvalues` (**for Hermitian `A`, `{μ | ∃ p, A p = μ p} = Set.range hA.eigenvalues`**, via `Matrix.spectrum_toLpLin` / `spectrum_eq_image_range` one way and `eigenvectorBasis` the other), ★★ `criticalValues_eq_range_eigenvalues`, ★★ `exists_isGreatest_range_eigenvalues`, ★★ `exists_isLeast_range_eigenvalues` (~~#326~~'s critical values, maximum and minimum in Mathlib's terms). 4 pins. Scope: sets, so multiplicities are not seen. | ~~**S**~~ done | Low–Medium — the bridge to Mathlib's spectral theorem | ~~#326~~ | > | ~~335~~ | QIT | One QM, many formulations | *new (2026-11-08, geometric-QM completion menu item 4, promoted by the author's "work through these")* **The speed limit for a time-dependent Hamiltonian.** ▸ **DONE 2026-11-08 (S–M took S).** In [`ProjectiveSpaceSpeedLimit.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceSpeedLimit.lean) (Category 1, added to ~~#331~~'s module): for Hermitian `H(t)` and any curve with `γ'(t) = X_{H(t)}(γ t)` on `[0, T]`, ★★ `fsPathLength_eq_lintegral_sqrt_variance` (**its length is `∫ 2 ΔH(t)`**), ★★★ `ofReal_fsAngle_le_lintegral_sqrt_variance` (**`arccos √P(γ 0, γ T) ≤ ∫₀ᵀ ΔH(t) dt`** in `ℝ≥0∞`, via Mathlib's `pathELength_comp_of_monotoneOn` and ~~#330~~'s lower bound), ★★ `fsAngle_le_integral_sqrt_variance` (the same in `ℝ` when `ΔH(t)` is integrable). 3 pins. Scope: the solution curve is a hypothesis (two-sided derivative on `[0, T]`); its existence (the time-ordered exponential) is not proved. | ~~**S–M**~~ done | Medium — the speed limit for driven systems | ~~#329~~; ~~#330~~; ~~#331~~ | > | ~~336~~ | QIT | One QM, many formulations | *new (2026-11-08, geometric-QM completion menu item 5, promoted by the author's "work through these")* **The Jacobi identity for smooth functions.** ▸ **DONE 2026-11-08 (M took M).** In [`ProjectiveSpacePoissonJacobi.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpacePoissonJacobi.lean) (Category 1): `extDeriv_apply_three` (Cartan's formula on three fields, from Mathlib's `extDeriv_apply_vectorField`), `flat_apply_lieBracket` (`ω([X, Y], Z)` for a Hamiltonian `Z` of a `C²` function, by symmetry of the second derivative), ★★ `flat_jacobi` (closed flat form, three Hamiltonian fields ⇒ the cyclic sum of `d(ω(B, C))(A)` vanishes), `fderiv_transition_trivializationAt_snd` / `fderiv_comp_chartAt_symm_trivializationAt_snd` (the chart reads vectors and derivatives), ★★ `IsSymplectic.poissonBracket_poissonBracket_cyclic`, ★★★ `IsSymplectic.poissonBracket_jacobi` (**`{f, {g, h}} + {g, {h, f}} + {h, {f, g}} = 0` for `C^∞` functions on any symplectic manifold**), ★★ `Projectivization.poissonBracket_jacobi` (on `ℂℙⁿ`). 8 pins. Reuses `HamiltonianLieDerivative.lean`'s `localRep_apply_localHamiltonianVector` / `extDeriv_localRep_eq_zero` and `ContinuousAlternatingMap.apply_swap_two`. Scope: `C^∞` functions. | ~~**M**~~ done | Medium — the bracket is a Lie algebra on all smooth functions | ~~#325~~ | > | ~~337~~ | QIT | One QM, many formulations | *new (2026-11-08, geometric-QM completion menu item 6, promoted by the author's "work through these")* **The Morse index of `⟨A⟩` at an eigenline.** ▸ **DONE 2026-11-08 (M took M).** In [`ProjectiveSpaceMorse.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceMorse.lean) (Category 1): ★ `hessForm_eq_sum` (`⟨v, A v⟩ − λ‖v‖² = Σ (μ_j − λ) |⟨e_j, v⟩|²` in Mathlib's eigenbasis), ★★★ `expectation_mk_add_smul` (**exactly, `⟨A⟩([ψ + t v]) = λ + t² (⟨v, A v⟩ − λ‖v‖²) / (1 + t²‖v‖²)`** for a unit eigenvector `ψ` and `v ⊥ ψ`), `belowSpan` (span of the `e_j` with `μ_j < λ`, orthogonal to the `λ`-eigenvectors), ★★ `finrank_real_belowSpan` (**real dimension `2 · #{j | μ_j < λ}`**), ★★ `hessForm_neg_of_mem_belowSpan` / `hessForm_nonneg_of_inner_eq_zero` (**negative definite on it, nonnegative on its complement**), ★★★ `expectation_mk_add_smul_lt_of_mem_belowSpan` (⟨A⟩ strictly decreases along every projective line into it), `expectation_mk_add_smul_ge_of_inner_eq_zero`. 14 pins. Scope: the Hessian is computed along the curves `t ↦ [ψ + t v]` (exactly); its identification with the chart Hessian, a general Morse-index definition and the Morse inequalities are not stated; repeated `λ` makes `⟨A⟩` Morse–Bott. | ~~**M**~~ done | Medium — Frankel's perfect Morse–Bott function, at second order | ~~#326~~; ~~#334~~ | > | ~~338~~ | QIT | One QM, many formulations | *new (2026-11-08, geometric-QM completion menu item 8, split three ways by the author's choice)* **Fubini–Study as a Mathlib Riemannian metric.** ▸ **DONE 2026-11-08 (M–L took M).** In [`ProjectiveSpaceFubiniStudyRiemannianBundle.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFubiniStudyRiemannianBundle.lean) (Category 1): `fsModelMetricCLM` + ★ `contDiff_fsModelMetricCLM`, ★ `fsModelMetric_self_ge` (coercivity `g_w(u, u) ≥ 4‖u‖² / (1 + ‖w‖²)²`), `isVonNBounded_fsModelMetric`, `fsMetric_symmL_symmL`, ★★ `fsRiemannianMetric` (**a `C^∞` `ContMDiffRiemannianMetric` on `Tℂℙⁿ`**), scoped `FubiniStudy.instRiemannianBundle` + the `IsContMDiffRiemannianBundle … ∞` instance, `inner_eq_fsMetric` (rfl), ★★★ `isLeviCivitaConnection_fs` (Mathlib's Levi-Civita connection on `ℂℙⁿ` is metric and torsion-free — Mathlib's theorem at this instance). 12 pins. Remainder: the Christoffel symbols in affine charts are ~~#343~~'s `chartChristoffel_leviCivita` (as `g⁻¹` of the Koszul form, smooth); a closed formula for `∇`, and `fsENorm` = the instance's norm (so `riemannianEDist = fsEDist`), not stated. | ~~**M–L**~~ done | Medium — Mathlib's Riemannian API applies to `ℂℙⁿ` | ~~#330~~; #339; #340 | > | ~~339~~ | QIT | One QM, many formulations | *new (2026-11-08, geometric-QM completion menu item 8, split three ways by the author's choice)* **Covariant derivative along a curve, and geodesics.** ▸ **DONE 2026-11-08 (L took M–L).** In [`CovariantDerivativeAlongCurve.lean`](../CsdLean4/Mathlib/Geometry/Manifold/CovariantDerivativeAlongCurve.lean) (Category 1, any manifold modelled on a finite-dimensional real space, any connection on `TM`): `chartCov x₀` + ★ `isCovariantDerivativeOn_chartCov` (the chart's flat connection is a covariant derivative on its source), `christoffel` (`cov − chartCov` via Mathlib's `difference`), ★★ `covAlong` (`D_t V = d/dt V_loc + Γ(V, γ')`), ★★★ `covAlong_comp` (**`D_t (X ∘ γ) = ∇_{γ'} X`**), `IsGeodesic` (a velocity differentiable in `TM` with `D_t γ' = 0`; the differentiability conjunct was added 2026-11-09 after the independent review showed junk-`0` derivatives made nowhere-differentiable and broken curves pass), ★★ `isGeodesic_of_isMIntegralCurve`; in `ProjectiveSpaceGreatCircleGeodesic.lean`, ★★★ `isGeodesic_greatCircle` and `isGeodesic_schrodingerUnitary_smul_of_isMaxOn` (**great circles and maximal-uncertainty orbits are `IsGeodesic`**). 13 pins. Remainder: chart independence of `covAlong` (fields with differentiable lift) and the geodesic equation in a chart are ~~#343~~'s `covAlong_eq_covAlongAt`, `isGeodesic_iff_chartSpray`; existence/uniqueness of geodesics, parallel transport, `exp` — not stated. | ~~**L**~~ done | Medium — the geodesic equation | ~~#338~~; ~~#340~~ | > | ~~340~~ | QIT | One QM, many formulations | *new (2026-11-08, geometric-QM completion menu item 8, split three ways by the author's choice; split again into #341, #342)* **Great circles are Fubini–Study geodesics.** ▸ **DONE 2026-11-08 by the Killing route (~~#341~~, ~~#342~~ + this Remainder).** In [`ProjectiveSpaceGreatCircleGeodesic.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceGreatCircleGeodesic.lean) (Category 1): `circleHam φ χ = i(|χ⟩⟨φ| − |φ⟩⟨χ|)` (Hermitian), ★★ `toEuclideanLin_schrodingerUnitary_circleHam` (**`exp(-itH) φ = cos t φ + sin t χ`**, by `d/dt [U(-t) γ(t)] = 0` and the group law — no commutation of `exp` with `H`), `variance_circleHam_le_one` (Bessel) / `variance_circleHam_mk` (`(ΔH)² = 1` at `[φ]`), ★★★ `greatCircle_isGeodesic` (**the great circle `t ↦ [cos t φ + sin t χ]` is the integral curve of `X_H` through `[φ]` and `∇_{X_H} X_H = 0` at every point of it**). 11 pins. Scope: geodesic in the field sense of ~~#342~~ (`∇_{γ'} γ' = 0` with `γ' = X_H(γ)`) and, since ~~#339~~, in the curve sense too: ★★★ `isGeodesic_greatCircle` (`IsGeodesic`, `D_t γ' = 0` with a differentiable velocity lift); speed `2` in the corpus's metric, not normalised. | ~~**M–L**~~ done | Medium — geodesics of `ℂℙⁿ` in the Levi-Civita sense | ~~#330~~; ~~#338~~; ~~#341~~; ~~#342~~ | > | ~~341~~ | QIT | One QM, many formulations | *new (2026-11-08, split from #340, author decision)* **`X_H` is Killing infinitesimally.** ▸ **DONE 2026-11-08 (M–L took M), by a shorter route than rowed: no flow is differentiated.** In [`ProjectiveSpaceSchrodingerKilling.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceSchrodingerKilling.lean) (Category 1): in every chart Q29's `flatLieDeriv_localHamiltonianVector_localRep_eq_zero` (the field is Hamiltonian) and KG-3′'s holomorphy (`DF` commutes with `i·`) give, since `g(u, v) = ω(i u, v)`, ★★ `flatKilling_schrodingerChartField` / ★★★ `flatKilling_metric_schrodingerChartField` (**the full Killing equation `L_F g = 0` in every affine chart, `Dg(F)(u, v) + g(DF u, v) + g(u, DF v) = 0`**); with `mpullback_extChartAt_symm_eq_chartField`, `fsMetric_eq_chartField` and the G14b bracket-in-chart step, ★★ `mfderiv_fsMetric_schrodingerField` (**`d⟨X_H, Z⟩(X_H) = ⟨X_H, [X_H, Z]⟩`** for every `Z` differentiable at the point: on the manifold only the contraction `(L_X g)(X, Z) = 0`). 7 pins. Remainder: the polarised `X⟨Y, Z⟩ = ⟨[X, Y], Z⟩ + ⟨Y, [X, Z]⟩` for general `Y` (same proof) is not stated. | ~~**M–L**~~ done | Medium — the Killing property infinitesimally | ~~#338~~; ~~#342~~; ~~#340~~ | > | ~~342~~ | QIT | One QM, many formulations | *new (2026-11-08, split from #340, author decision)* **Schrödinger orbits through maximal-uncertainty states are geodesics.** ▸ **DONE 2026-11-08 (S–M took S–M), stated generally.** In [`ProjectiveSpaceSchrodingerGeodesic.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceSchrodingerGeodesic.lean) (Category 1): `mfderiv_apply_eq_zero_of_isLocalMax` (Fermat on `ℂℙⁿ`), ★★ `inner_leviCivita_schrodingerField_self` (**`⟨∇_{X_H} X_H, Z⟩ = −½ d‖X_H‖²(Z)`**, Mathlib's Koszul formula + ~~#341~~), ★★★ `leviCivita_schrodingerField_self_eq_zero_of_mfderiv_variance` / `_of_isLocalMax` (**`∇_{X_H} X_H = 0` wherever `(ΔH)²` is stationary**, in particular at a local maximum; review fix 2026-11-09, was a global maximiser), ★★★ `leviCivita_schrodingerField_self_orbit_eq_zero` (**the orbit `t ↦ exp(-itH) • q` of a global maximiser is an integral curve of `X_H` with `∇_{X_H} X_H = 0` along it**: a geodesic in the field sense). 6 pins. Remainder: none of the original — the instance `H = i(|χ⟩⟨φ| − |φ⟩⟨χ|)` (maximum of `(ΔH)²` at `[φ]`, orbit = ~~#330~~'s great circle) is ~~#340~~'s `greatCircle_isGeodesic`, and the curve-sense geodesic is `isGeodesic_greatCircle` (~~#339~~). | ~~**S–M**~~ done | Medium — geodesics of `ℂℙⁿ` in the Levi-Civita sense | ~~#341~~; ~~#338~~; ~~#340~~ | > | ~~343~~ | QIT | One QM, many formulations | *new (2026-11-09, the author's task "CSD Geometry: Geodesic initial-value foundation")* **The local geodesic ODE of `ℂℙⁿ`, with a smooth spray.** ▸ **DONE 2026-11-09.** In [`CovariantDerivativeAlongCurve.lean`](../CsdLean4/Mathlib/Geometry/Manifold/CovariantDerivativeAlongCurve.lean) (Category 1, any manifold modelled on a finite-dimensional real space): `chartConstField` (the field constant in a chart), `christoffel_chartConstField` (`Γ(σ_c, u) = ∇_u σ_c`), ★ `mlieBracket_chartConstField` (**chart-constant fields commute**, via Mathlib's `mpullback_mlieBracket`), `covAlongAt` / `covAlongAt_comp`, ★★ `covAlong_eq_covAlongAt` (**chart independence of `covAlong`** for a field whose lift to `TM` is differentiable), `chartChristoffel` / `chartSpray`, ★★★ `isGeodesic_iff_chartSpray` (**a curve in the source of one chart is `IsGeodesic` iff its chart coordinate `w` and some `v` solve `(w', v') = (v, −Γ_w(v, v))`**). In [`ProjectiveSpaceFubiniStudyGeodesicSpray.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFubiniStudyGeodesicSpray.lean) (Category 1): `fsKoszul`, `isInvertible_fsModelMetricCLM` (coercivity), `fsChristoffel w v u = g_w⁻¹(K_w(u, v, ·))`, ★★ `contDiff_fsChristoffel` and ★★ `contDiff_fsGeodesicSpray` (**`C^∞` on all of `ℂⁿ`, resp. `ℂⁿ × ℂⁿ`**), ★★★ `chartChristoffel_leviCivita` (**the Christoffel symbols of Mathlib's Levi-Civita connection in every affine chart are `fsChristoffel`**: Koszul formula with chart-constant fields), ★★ `contDiffOn_chartChristoffel_leviCivita`, ★★★ `isGeodesic_iff_fsGeodesicSpray`. 36 pins. Remainder: local existence and uniqueness and the time-restricted `IsGeodesicOn` are ~~#344~~; the exponential map, completeness, a closed formula for `Γ_w` — not stated. | ~~**M–L**~~ done | Medium — geodesics of `ℂℙⁿ` as an ODE with smooth right-hand side | ~~#338~~; ~~#339~~ | > | ~~344~~ | QIT | One QM, many formulations | *new (2026-11-09, #343's Remainder, the author's "1")* **Local existence and uniqueness of Fubini–Study geodesics.** ▸ **DONE 2026-11-09.** In [`CovariantDerivativeAlongCurve.lean`](../CsdLean4/Mathlib/Geometry/Manifold/CovariantDerivativeAlongCurve.lean): `IsGeodesicOn cov γ I` (geodesic for `t ∈ I`), `isGeodesicOn_univ`, `IsGeodesicOn.mono`, ★★★ `isGeodesicOn_iff_chartSpray` (the chart equation on an open set of times; `isGeodesic_iff_chartSpray` is now its corollary). In [`ProjectiveSpaceFubiniStudyGeodesicSpray.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFubiniStudyGeodesicSpray.lean): `chartSpray_leviCivita` (the chart spray is `fsGeodesicSpray` in every affine chart, whose target is all of `ℂⁿ`), ★★★ `isGeodesicOn_iff_fsGeodesicSpray`, ★★★ `exists_isGeodesicOn_fs` (**for every `x₀` and `v₀ ∈ T_{x₀}ℂℙⁿ` a geodesic on some `(-ε, ε)` with `γ 0 = x₀`, `γ'(0) = v₀`**, Mathlib's Picard–Lindelöf), ★★★ `eventuallyEq_of_isGeodesicOn_fs` (**two geodesics with the same position and velocity at `t₀` agree near `t₀`**, Mathlib's `ODE_solution_unique_of_eventually` with the spray Lipschitz near the initial point). 10 pins. Remainder: uniqueness on the whole common interval is ~~#345~~; maximal geodesics, completeness, `exp`, an estimate of `ε` — not stated. | ~~**M**~~ done | Medium — geodesics of `ℂℙⁿ` exist and are determined by initial data | ~~#343~~ | > | ~~345~~ | QIT | One QM, many formulations | *new (2026-11-09, #344's Remainder, the author's "2 then 1")* **Uniqueness of Fubini–Study geodesics on a whole interval.** ▸ **DONE 2026-11-09.** In [`ProjectiveSpaceFubiniStudyGeodesicSpray.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFubiniStudyGeodesicSpray.lean): ★★★ `eqOn_of_isGeodesicOn_fs` (**two geodesics on an open preconnected `I` with the same position and velocity at one `t₀ ∈ I` agree on all of `I`**, across any number of charts): the set `{t | γ₁ =ᶠ[𝓝 t] γ₂}` is open, contains `t₀` by ~~#344~~'s local uniqueness, and is closed in `I` because positions (`ℂℙⁿ` is T2) and chart velocities (differentiable lifts) pass to the limit; `IsPreconnected.subset_of_closure_inter_subset` concludes. 1 pin. Remainder: maximal geodesics are ~~#346~~; completeness, `exp` — not stated. | ~~**S–M**~~ done | Medium — geodesics of `ℂℙⁿ` are determined by initial data | ~~#344~~ | > | ~~346~~ | QIT | One QM, many formulations | *new (2026-11-09, #345's Remainder, the author's "1")* **Maximal Fubini–Study geodesics.** ▸ **DONE 2026-11-09.** In [`ProjectiveSpaceFubiniStudyGeodesicSpray.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFubiniStudyGeodesicSpray.lean): `isGeodesicOn_congr_fs` (being a geodesic on an open set depends only on the curve there, via the chart equation), `IsFSGeodesicData x₀ v₀ I η`, `IsFSGeodesicData.eqOn` (~~#345~~ on `I ∩ I'`), ★★★ `exists_maximal_isGeodesicOn_fs` (**for every `(x₀, v₀)` a geodesic on an open interval `J ∋ 0` such that every geodesic with the same initial data on an open interval `I ∋ 0` has `I ⊆ J` and agrees with it**; `J` = union of all such intervals, the curve glued by uniqueness). 4 pins. Remainder: `J = ℝ` is ~~#347~~; the exponential map — not stated. | ~~**S–M**~~ done | Medium — geodesics of `ℂℙⁿ` are determined by initial data | ~~#345~~ | > | ~~347~~ | QIT | One QM, many formulations | *new (2026-11-09, #346's Remainder, the author's "1")* **`ℂℙⁿ` is geodesically complete.** ▸ **DONE 2026-11-09.** In [`ProjectiveSpaceFubiniStudyGeodesicSpray.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFubiniStudyGeodesicSpray.lean): `isGeodesicOn_of_chartSolution` (a spray solution in any affine chart is a geodesic; local existence now derived from it), ★★ `fsMetric_velocity_eq_of_isGeodesicOn` (**constant speed**: `d/dt g_w(v, v) = Dg(v)(v, v) − 2 g_w(Γ_w(v, v), v) = 0` by the Koszul formula), `exists_uniform_fsGeodesicSpray`, ★★ `exists_uniform_isGeodesicOn_fs` (**one existence time `ε` for all data with `g(v, v) ≤ C`**: compactness of `ℂℙⁿ`, coercivity `g_w(u, u) ≥ 4‖u‖²/(1 + ‖w‖²)²`, finitely many Picard–Lindelöf balls per chart), `eq_univ_of_maximal_fs` (escape argument: each point of the maximal interval has an `ε`-neighbourhood in it), ★★★ `exists_isGeodesic_fs` (**every `(x₀, v₀)` has an `IsGeodesic` defined on all of `ℝ`, agreeing with every geodesic with the same data on an open interval containing `0`**). 7 pins. Remainder: the exponential map is ~~#348~~; metric completeness / Hopf–Rinow, minimisation — not stated. | ~~**M**~~ done | Medium — geodesics of `ℂℙⁿ` exist for all time | ~~#346~~ | > | ~~348~~ | QIT | One QM, many formulations | *new (2026-11-09, #347's Remainder, the author's "1")* **The Fubini–Study exponential map.** ▸ **DONE 2026-11-09.** In [`ProjectiveSpaceFubiniStudyGeodesicSpray.lean`](../CsdLean4/Mathlib/Geometry/Manifold/Instances/ProjectiveSpaceFubiniStudyGeodesicSpray.lean): `fsChristoffel_smul` (`Γ_w(cv, cv) = c² Γ_w(v, v)`), ★★ `isGeodesic_comp_mul_fs` (**`t ↦ γ (c t)` is a geodesic**, through the chart equation), `mfderiv_comp_mul_zero` (its velocity at `0` is `c γ'(0)`), `fsGeodesic x v` (the complete geodesic, by choice from ~~#347~~, with `fsGeodesic_zero`, `mfderiv_fsGeodesic_zero`, `isGeodesic_fsGeodesic`, `eqOn_fsGeodesic`), ★★ `fsGeodesic_smul` (**`γ_{x, cv}(t) = γ_{x, v}(c t)`**, by uniqueness), `fsExp x v = γ_{x, v}(1)`, ★★★ `fsExp_smul` (**`exp_x(t v) = γ_{x, v}(t)`**), `fsExp_zero`. 13 pins. Remainder: smoothness of `exp`, `d(exp_x)_0 = id`, local diffeomorphism, injectivity radius, metric completeness / Hopf–Rinow, minimisation — not stated. | ~~**S–M**~~ done | Medium — the exponential map of `ℂℙⁿ` | ~~#347~~ | > | ~~349~~ | Records | Records→spacetime | *new (2026-11-09, the author's programme "From `Σ` to physical spacetime", milestone M1A — the Dirac-chain-to-Hamiltonian-pointer bridge; out of #39)* **A Dirac carrier and a pointer in one Hamiltonian arena.** ▸ **DONE 2026-11-09.** [`RecordLayer/DiracPointerHamiltonian.lean`](../CsdLean4/RecordLayer/DiracPointerHamiltonian.lean): on `Σ = ℂℙ^{2M+1} × T²` with the arena form of R-016′, the one-particle Dirac ring of `M + 1` sites (`diracHam`, ~~#249~~) and a torus pointer are generated by **one** energy `𝓗 = −2⟨H_D⟩ − 2 g f(θ₂) ⟨P_j⟩`, an instance of `coupledEnergy` (the pattern of ~~#195~~). Bridge: `diracIdx : ZMod (M + 1) × Fin 2 ≃ Fin (2M + 2)` (`diracIdx_val`), `diracArenaEquiv`, `reindex_exp_smul` (reindexing commutes with `exp (t • ·)`, by ODE uniqueness). Projector: `siteProj j` with `siteProj_isHermitian`, `siteProj_mul_self`, `siteProj_eq_single`, ★ `arenaObs_siteProj` (`⟨P_j⟩` = the two moment coordinates of site `j`), `arenaObs_siteProj_nonneg`, `arenaObs_siteProj_le_one`, `arenaObs_siteProj_mk` (= `siteWeight / ‖ψ‖²` of `DiracDetector.lean`). Energy: `diracA`, `diracB`, `diracPointerEnergy`, `contMDiff_diracPointerEnergy`, ★★ `diracPointerEnergy_eq`. Carrier: `diracPointerH` (`H_D + g f(θ₂) P_j`), `diracCarrierU`, ★★ `hasDerivAt_diracCarrierU` (**`U̇ = −i (H_D + g f(θ₂) P_j) U`**), `idxUnitary_diracCarrierU`, `diracArenaEquiv_kick`. Pointer: `diracPointerShift`, ★★ `hasDerivAt_diracPointerShift` (**`θ̇₁ = −2 g f'(θ₂) ⟨P_j⟩_{p(t)}`**), `diracPointerShift_zero_coupling` (control: `g = 0`, no response). Flow: `diracPointerFlow` (`diracPointerFlow_zero`, `diracPointerFlow_add`, defined for all time), ★★★ `diracPointerFlow_apply` (**the Hamiltonian flow in closed form**; `θ₂` fixed, `diracPointerFlow_snd_snd`), ★★ `hamiltonianVectorField_diracPointer_snd` (the pointer component of `X_𝓗` read off `ι_X ω = d𝓗`, no flow), ★★ `diracPointerEnergy_flow` (**energy conserved**), ★★ `arenaVolume_map_diracPointerFlow` and `kMuL_map_diracPointerFlow` (**Liouville**, both measures). Checked special case (profile `trainF`, carrier `siteRay j`, pointer at `(0, 0)`): `diracPointerH_trainF_zero` (the carrier runs the free `H_D` there), ★★ `hasDerivAt_diracPointerShift_siteRay` (**rate `2g`** from the flow), ★★ `hamiltonianVectorField_diracPointer_siteRay` (**the same `2g`** from the vector field), `hasDerivAt_diracPointerShift_siteRay_of_ne` (rate `0` for a detector at another site), ★★★ `exists_diracPointerFlow_pointer_ne` (**for `g ≠ 0` the flow moves the pointer**); package ★★★ `diracPointer_hamiltonian`. 58 pins. **Exact scope:** one particle on a finite ring, `ħ = 1`, `f` `C^∞`; `𝓗` is independent of `θ₁`, so `θ₂` is conserved and the carrier sees the *static* potential `g f(θ₂) P_j` — the moving angle `θ₁` does not act back on the carrier; the coupling form is chosen, not derived; the nontriviality is a rate at one point and an existence statement, with no time estimate. Nothing is recorded: no outcome region, threshold, persistence, `ActualOccurrence`, clock, Born weight, causal order or spacetime statement. Remainder: ~~(M0) the programme's contract document is not written~~ (written, ~~#350~~); (M2) the write — a readout region and ready state for this pointer (the displacement and the `detFlow` identification landed as ~~#351~~); a pointer that acts back on the carrier through its moving angle needs an energy outside the `coupledEnergy` form; library-owed: `reindex_exp_smul` and a generic "energy is constant along `hamiltonianFlow`" belong upstream of this module (rule of two — second use each). | ~~**S–M**~~ done | High — the first object of the programme: carrier and pointer as one Hamiltonian system, the base M2–M6 build on | ~~#195~~, ~~#249~~, R-016′ | > | ~~350~~ | Records | Records→spacetime | *new (2026-11-09, the author's programme "From `Σ` to physical spacetime", milestone M0 — freeze the scientific contract)* **The programme's three registers: assumed, derived, owed.** ▸ **DONE 2026-11-09 — a document, no Lean.** [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md), from an audit of `POSITS.md`, `records-to-spacetime-status.md`, `beyond-1p1-scoping.md` and `residues.tsv`. **Register 1** (`UA-1`–`UA-11`) is what the programme assumes on the arena of ~~#349~~: the arena, the measure (Posits 2 and 9), the typicality reading (Posit 4), the apparatus preparation (Posits 5 and 8, and the law of the conserved angle `θ₂`, which this flow never produces), the coupling (R-015), the cell law (Posit 1), the flow parameter, the Dirac propagation law (Posit 6; relativity is input), composite structure (Posit 7), and the stated detector laws and kinematic idealisations inside the existing click and radar results. **Register 2** (`UD-1`–`UD-30`) lists the theorems by name with the arena each holds on: seven on the shared arena (the M1A results, none of them a record, a statistic or a geometric statement), fifteen on other arenas (infrastructure until an identification theorem exists), eight that pin the assumptions from outside. **Register 3** (`UO-1`–`UO-24`) lists what M2–M9 owe, each with the test for a theorem that would only restate a Register 1 entry. ⚠️ **The cell-law premise stays a premise** (§6): it is Posit 1, not derived from any dynamics, extensionally equivalent to its conclusion, and the one theorem concluding it (~~#152~~'s calibration conversion) has a hypothesis equivalent to the moment map; an M3 theorem that assumes it, or builds its outcome regions from `momentContext` or from arcs sized by site weights, is recorded as conditional and does not pass gate A. ⚠️ **Two calculations bearing on gate A are in the note and are NOT in Lean** (§5): under the product preparation a fibre region that does not depend on the base point has the same measure for every state, and a threshold on a ready pointer with a bounded profile derivative gives a law with a weight cut-off, not a linear one; if they survive formalisation gate A's answer is an obstruction with a named missing condition. **Audit findings:** no posit's standing changed; `POSITS.md` Posit 10 still said I3 had no rate field from the base, which ~~#136~~ supplied on 2026-10-13 — corrected with a dated line, wording of the posit untouched; Posit 1's entry does not mention ~~#152~~ (the author's reservation on that row, left as it is); the status note gains a pointer. `check-claims` now scans the new note. Scope: no Lean, no new posit, no new registry residue — the `UO-n` entries are the programme's ledger of missing implications. Remainder: the two gate-A calculations as Lean statements (M3's first brick, or sooner if the author wants the obstruction before M2); whether calibration becomes Posit 1's live wording (the author's decision, ~~#152~~). | ~~**S**~~ done | High — every later milestone is checked against it; it is what stops a premise being reported as a result | ~~#349~~, ~~#152~~, ~~#136~~, `POSITS.md` | > | ~~351~~ | Records | Records→spacetime | *new (2026-11-09, the author's programme "From `Σ` to physical spacetime", milestone M2 — records generated by the interaction; part A, the first of its four statements; out of ~~#349~~'s Remainder)* **The pointer displacement a chain packet creates.** ▸ **DONE 2026-11-09.** [`RecordLayer/DiracPointerWrite.lean`](../CsdLean4/RecordLayer/DiracPointerWrite.lean) (39 pins). ~~#349~~ gave the pointer's rate at one point, with no size and no time. The pointer integrates the occupancy of a carrier that feels the static site potential `g f(θ₂) P_j`, which has no closed form on the ring; the corpus's Duhamel bound (`Matrix.norm_exp_smul_neg_I_sub_le`) compares it with the free chain. ★★ `norm_diracCarrierU_sub_le` (the coupled propagator is within `|t| |g f(θ₂)|` of `chainProp`), `abs_diracOccupancy_sub_le`, `abs_integral_diracOccupancy_sub_le`; ★★★ `abs_diracPointerShift_sub_le` (**on every fibre the displacement is within `2 |g| |f'(θ₂)| · |g f(θ₂)| t²` of `−2 g f'(θ₂)` times the free chain's accumulated occupancy**), ★★★ `le_abs_diracPointerShift` (the lower bound); ★★ `diracPointerShift_of_lift_eq_zero` (**exact where `f(θ₂) = 0`** — there the carrier's generator is `H_D`, `diracCarrierU_of_lift_eq_zero`); `abs_diracPointerShift_le` (at most `2 |g| |f'(θ₂)| t`), `diracPointerShift_of_deriv_eq_zero` (**no displacement at a zero of `f'`**). ★★★ `diracPointerShift_eq_register` (**the registers of `DiracDetector.lean`'s `detFlow` are this flow's pointer** on the fibres with `f(θ₂) = 0`, up to the constant `−2 f'(θ₂)/⟨ψ, ψ⟩`), ★★ `abs_diracPointerShift_sub_register_le` (elsewhere they differ by at most the same error). **Instance** (`twoMode`, the 12-site ring, `J = 4`, `μ = 3`, `Σ_* = ℂℙ²³ × T²`): ★★★ `diracPointerShift_twoMode_zero` (**profile `trainF`, `θ₂ = 0`, detector at site `0`: the displacement is `g (116 t + 48 sin 2t)/696`**), `diracPointerShift_twoMode_six` (`g (116 t − 48 sin 2t)/696` at the opposite site), ★★★ `diracPointerFlow_twoMode` (the same, as the pointer coordinate of `diracPointerFlow`), ★★★ `le_abs_diracPointerShift_twoMode` (every profile, every fibre), `le_abs_diracPointerShift_twoMode_quarter` (at `t = π/4`, where `|g f(θ₂)| ≤ 1/10`, at least `|g| |f'(θ₂)|/14`). Contract: UO-2 and UO-6 → UD-31–UD-34 in [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); no new `UA` entry. **Scope:** a displacement, not a record — no readout region, ready state, threshold, outcome, persistence or `ActualOccurrence` (the other three statements of M2); exact only on the `f(θ₂) = 0` fibres and in the two closed forms, bounds elsewhere; the Duhamel error is quadratic in `t` against a linear signal, so the lower bounds say nothing past a time of order the mean occupancy over `|g f(θ₂)|`; no lower bound is uniform in `θ₂`; `twoMode` is a beating state with between `20/1392` and `212/1392` of its weight on each site, not a carrier arriving from elsewhere; the register identification is of one lifted coordinate, and `packetDetector`/`twoModeDetector` are not transported; no measure and no probability appear. Remainder: ~~(M2B) readout region, ready state, threshold crossing with the exceptional `θ₂` set named and the control run bounded~~ → ~~#352~~; ~~(M2C) the write as an `ActualOccurrence` whose `evolve` is `diracPointerFlow`, and a retention interval~~ → ~~#353~~; a carrier with no initial weight at the detector that then arrives (UO-2's residue); an error bound that does not grow as `t²`. | ~~**M**~~ done | High — the first statement of M2 as a theorem about the flow, and the existing Dirac detector registers placed on the shared arena | ~~#349~~, ~~#350~~, ~~#252~~ | > | ~~352~~ | Records | Records→spacetime | *new (2026-11-09, the author's programme "From `Σ` to physical spacetime", milestone M2 — records generated by the interaction; part B, the second of its four statements; out of ~~#351~~'s Remainder)* **The pointer crossing an outcome threshold.** ▸ **DONE 2026-11-09.** [`RecordLayer/DiracPointerThreshold.lean`](../CsdLean4/RecordLayer/DiracPointerThreshold.lean) (40 pins). ~~#351~~ bounded the displacement and stated no region to read it against. **Regions on `Σ_*`:** `diracReady` (pointer angle in the ready arc `(0, 1/6]`), `diracDisplayed` (in the displayed arc `(1/3, 5/6]` of ~~#194~~), both measurable; ★ `diracDisplayed_congr`, `mem_diracReady_congr` (**they read the pointer angle `θ₁` alone** — not the carrier, not `θ₂`). **The carrier over a short time**, along the coupled dynamics and with no comparison to the free chain: `norm_diracCarrierU_sub_one_le` (propagator within `|t| ‖H‖` of the identity, `H = H_D + g f(θ₂) P_j`, `norm_diracPointerH_le`), ★★ `abs_integral_diracOccupancy_sub_mul_le` (accumulated occupancy within `‖H‖ t²` of `t` times the starting occupancy), ★★ `abs_diracPointerShift_sub_init_le`; ★★ `le_diracPointerShift_siteRay` (**signal**: a carrier on the detector's site turns the pointer by at least `κ (t − ‖H‖ t²)`, `κ = −2 g f'(θ₂)`), ★★ `abs_diracPointerShift_siteRay_of_ne_le` (**control**: a carrier on another site by at most `|κ| ‖H‖ t²`). **The threshold, on `diracPointerFlow`:** ★★ `diracPointerFlow_displayed`, `diracPointerFlow_not_displayed` (the two crossing criteria, exact in the displacement); ★★★ `diracPointerFlow_siteRay_displayed`, ★★★ `diracPointerFlow_siteRay_of_ne_not_displayed`. **The fibres:** `slowFibres` (`|κ(θ₂)| T < 1/6`), ★★ `volume_slowFibres_pos` (**open and of positive measure for every `C¹` profile and every horizon** — Rolle's zero of `f'`), ★★★ `diracPointerFlow_not_displayed_of_mem_slowFibres` (**on a slow fibre nothing is displayed up to `T`, whatever the carrier**); `writeFibres` (`κ > 0`, `1/3 < κ (T − h T²)`, `κ T < 2/3`, `κ h T² < 1/6`, `h = ‖H_D‖ + |g f(θ₂)|`), `isOpen_writeFibres`, ★★★ `diracPointerFlow_displayed_of_mem_writeFibres` / `diracPointerFlow_not_displayed_of_mem_writeFibres` (**on a write fibre the carrier on the detector's site is displayed at `T` and a carrier on any other site is not, at any time up to `T`**), `disjoint_writeFibres_slowFibres`; ★★★ `volume_writeFibres_trainF_pos` (**profile `trainF`, `g = 1/(4T)`, `‖H_D‖ T ≤ 1/4`: the write fibres contain `θ₂ = 0` and have positive measure**), `exists_volume_writeFibres_pos` (such `g`, `T` exist for every ring, `J`, `μ`). Contract: UO-1 and UO-3 → UD-35–UD-39 in [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); no new `UA` entry. **Scope:** a threshold, not a record — no outcome selected, no `ActualOccurrence`, no time in the displayed arc beyond the one instant `T`; the crossing criteria are exact, every size is the first-order bound `‖H‖ t²`, so `writeFibres` is a set on which the write is proved and the set on which it happens is larger and not determined; **the signal carrier starts on the detector's site and does not arrive** — the write is complete in a time short on the scale `1/‖H‖`; the control is bounded only up to `T` (the chain later carries it to the site); the write is **not universal** — slow fibres of positive measure for every profile; the measures of the two sets are positive and not computed; fibres in neither set, and every fibre with a negative rate, carry no statement; Haar measure on the `θ₂` circle is used only to say the sets are not null, and no probability of a write is stated. Remainder: ~~(M2C) the write as an `ActualOccurrence` whose `evolve` is `diracPointerFlow`, and a retention interval~~ → ~~#353~~; explicit lower bounds for the measures of the two sets for `trainF` (arcs about `θ₂ = 0` and `θ₂ = 1/4`); the fibres with a negative rate; a carrier with no initial weight at the detector that then arrives (UO-2's residue); the control past `T`. | ~~**M**~~ done | High — the second statement of M2 as a theorem about the flow, with the exceptional fibres named and the control run bounded | ~~#351~~, ~~#349~~, ~~#194~~ | > | ~~353~~ | Records | Records→spacetime | *new (2026-11-10, the author's programme "From `Σ` to physical spacetime", milestone M2 — records generated by the interaction; part C, the third and fourth of its four statements and the milestone's package; out of the Remainders of ~~#351~~ and ~~#352~~)* **The record the pointer holds.** ▸ **DONE 2026-11-10.** [`RecordLayer/DiracPointerRecord.lean`](../CsdLean4/RecordLayer/DiracPointerRecord.lean) (51 pins). ~~#352~~ proved the threshold crossed at one time and built no occurrence. **The pointer does not turn back:** ★★ `diracPointerShift_mono` (on a fibre with rate `κ = −2 g f'(θ₂) ≥ 0` the displacement is monotone in time, for every carrier — it is `κ` times an integral of a nonnegative occupancy). **The flow as a two-time evolution:** `diracPointerEvolve` (`diracPointerFlow` over the elapsed time), `diracPointerEvolve_self`, `diracPointerEvolve_comp`, `measurable_diracPointerEvolve`. **Persistence:** ★★ `diracPointerFlow_displayed_of_mono`, `diracPointerFlow_not_displayed_of_mono` (the readout on an interval from the displacement at its ends); ★★★ `isRecordWrite_diracPointerFlow` (**displacement at most `1/6` at `τ₀`, at least `1/3` at `τ₁`, at most `2/3` at `τ₂` ⇒ the readout history of the flow is an `IsRecordWrite` of `true` with write interval `[τ₀, τ₁]` and window `τ₂ − τ₁`**, any carrier); ★★★ `diracPointerRecord` (**that write as an `ActualOccurrence` on `Σ_*`**, `diracPointerRecord_evolve`: its `evolve` is `diracPointerEvolve`, by `rfl`). **Where there is no record:** ★★ `not_isRecordWrite_of_mem_slowFibres` (on a slow fibre none is completed by `T`, whatever the carrier), ★★ `not_isRecordWrite_of_rateAt_eq_zero` (on a zero-rate fibre none at any time). **The record fibres** `recordFibres` (`8 (‖H_D‖ + |g f(θ₂)|) < 3 κ`, open): ★★★ `isRecordWrite_siteRay`, `siteRayRecord` (**the carrier on the detector's site writes `true`: not displayed before `1/(6κ)`, displayed throughout `[1/(2κ), 2/(3κ)]`, window `1/(6κ) > 0`** — `siteRayRecord_window`, `siteRayRecord_window_pos`); ★★★ `diracPointerFlow_not_displayed_of_mem_recordFibres`, `not_isRecordWrite_siteRay_of_ne` (**the carrier on another site reads `false` throughout `[0, 2/(3κ)]`** and completes no write in that time); ★★★ `volume_recordFibres_trainF_pos` (profile `trainF`, `4 ‖H_D‖ < 3 g`: the record fibres contain `θ₂ = 0` and have positive measure), `exists_volume_recordFibres_pos`. **The two-mode packet of ~~#351~~ on the fibre `θ₂ = 0`, `0 < g ≤ 29/12`:** ★★★ `isRecordWrite_twoMode`, `twoModeRecord` (write interval `[(29 − 12g)/(29g), (58 + 12g)/(29g)]`, window `(58 − 24g)/(29g)`); ★★ `exists_diracPointerFlow_snd_eq` (once the displacement has reached a full turn the pointer pair has been exactly its ready value), ★★★ `twoMode_record_erased` (**the record does not last**: at some time after the window and no later than `7/g` the pointer is back at its ready value and the readout is `false`). **Package:** ★★★ `diracPointer_record` (the one capstone of M2 — displacement, threshold held over an interval, the two carriers told apart, the write of nonzero window, and the slow fibres of positive measure with no record by `T`). Contract: UO-4 and UO-5 → UD-40–UD-44 in [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); no new `UA` entry. **Scope:** monotonicity, the interval statements and `isRecordWrite_diracPointerFlow` are exact given the three displacement values; the times on a record fibre come from the first-order bound `‖H‖ t²` and those of the two-mode record from `|sin| ≤ 1`, neither sharp; **the window is where the proof ends** — for the carrier on the detector's site nothing is proved after `2/(3κ)`, for the two-mode packet the record is proved gone by `7/g`; **"selected" means fixed by the ontic initial point** of a deterministic flow — two carriers are compared, no probability is stated, no Born weight appears, and a carrier with weight on both sites has no statement (M3); **not universal** — slow fibres of positive measure for every `C¹` profile, a zero-rate fibre for every profile, no statement at a negative rate; the carrier on the detector's site does not arrive; the two-mode record is on one fibre, a null set; the readout, ready set, value, site label and the `evolve` field are definitions — what is derived is the write field. Remainder: the two-mode record on an open set of fibres (from `abs_diracPointerShift_twoMode_sub_le`); a carrier with no initial weight at the detector that then arrives (UO-2's residue); the end of retention for the record on a record fibre; the fibres with a negative rate; the measure of `recordFibres` computed; a carrier with weight on both sites (→ M3, gate A: the statistics read at one time ~~#354~~; the event about two times is owed there). | ~~**M**~~ done | High — M2's acceptance: the record is a theorem about the Hamiltonian flow, and the occurrence's evolution is that flow | ~~#352~~, ~~#351~~, ~~#349~~, ~~#191~~, ~~#194~~ | > | ~~354~~ | Records | Records→spacetime | *new (2026-11-10, the author's programme "From `Σ` to physical spacetime", milestone M3 — measurement statistics, gate A; part A, the statistics read at one time; out of the Remainder of ~~#353~~)* **What the preparation law cannot see: the pointer at one time.** ▸ **DONE 2026-11-10.** [`RecordLayer/DiracPointerBorn.lean`](../CsdLean4/RecordLayer/DiracPointerBorn.lean) (25 pins). M3 asks whether the Hamiltonian and the apparatus fix the measures of the outcome regions without a stipulated rate field. For statistics read at one time the answer is an obstruction. **The fibre over a carrier ray:** ★★ `measurePreserving_diracPointerFlow_fibre` (the pointer pair moves by a turn of `θ₁` that depends on `θ₂` alone, which preserves `Haar ⊗ ν` for every law `ν` of `θ₂` — by `MeasurePreserving.vadd_twist_of_invariant`). **The law at one time:** `diracPointerHeld` (the carrier moved as in the flow, the fibre point left where it was — a comparison map, not a dynamics); ★★★ `map_diracPointerFlow_dirac_prod`, `map_diracPointerFlow_epistemicMeasure` (**the push-forward of the preparation `δ_{x₀} ⊗ Haar` under `diracPointerFlow` equals its push-forward with the pointer held**: no statistic read at one time sees the pointer's displacement); ★★ `map_snd_diracPointerFlow_epistemicMeasure` (the fibre point is Haar distributed at every time). **Regions of the fibre:** ★★★ `epistemicMeasure_diracPointerFlow_fibreRegion` (**the probability that the fibre point is in an apparatus-fixed region `F` at time `t` is the Haar measure of `F`, for every carrier, time, coupling and profile**); ★★ `epistemicMeasure_diracDisplayed` (the readout shows `true` with probability `1/2`), `epistemicMeasure_diracReady` (`1/6`), `volume_displayArc`; ★★★ `exists_half_le_abs_fibreRegion_sub_siteWeight` (**no region of the fibre has a probability within `1/2` of the occupancy `⟨P_j⟩` of the detector's site for every carrier**: `ε_meas ≥ 1/2`). **Regions that depend on the base point:** ★★ `epistemicMeasure_diracPointerFlow_globalBasin` (the basin of any rate field `c` has at time `t` the probability `∫ c.rate (coupled carrier at t) i dθ₂` — **conditional on the rate field, UA-6**; the pointer's motion adds nothing to it). **The check, two times:** `ready_diracPointerHeld_displayed_eq_empty`, ★★ `ofReal_mul_volume_writeFibres_le`, `epistemicMeasure_ready_displayed_pos`, `exists_epistemicMeasure_ready_displayed_pos` (*ready at `0` and displayed at `T`* is empty with the pointer held and has probability at least `(1/6) · vol(writeFibres)` under the flow for the carrier on the detector's site, positive for `trainF`: the one-time theorem does not extend to two times). Contract: calculation 1 of the M3 section proved, sharpened and corrected; UO-9 part met → UD-45–UD-49 in [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); no new `UA` entry. **Scope:** exact (equalities of measures; one lower bound); **one time only**; the law is UA-2's with `θ₁` uniform (any law for `θ₂` in the general forms); it uses that `θ₂` is conserved and that the carrier's evolution does not depend on `θ₁`; the basin statement is conditional on UA-6 and is not a derivation of statistics from the dynamics; `diracPointerHeld` is not a flow of `Σ_*`; one pointer, one site, two readout values; **gate A is not answered by this row** — a pointer prepared ready is the event about two times, which this row only shows to be outside the theorem. Remainder: the probability of *ready at `0` and displayed at `T`* as a function of the carrier: the weight cut-off and the comparison with `⟨P_j⟩` (calculation 2; M3 part B, with the gate-A statement and the `POSITS.md` line → ~~#355~~); an error bound for the basin statistic of the moment-map field along the coupled flow, with the disturbance as an explicit term (UO-8, UO-12; conditional on UA-6); several detectors and one outcome per trial (UO-10); repeated trials generated by one flow (UO-11); no-signalling and sequential update on a shared arena (UO-13). | ~~**M**~~ done | High — the first half of gate A: under the preparation law no apparatus-fixed fibre readout at one time depends on the state | ~~#353~~, ~~#352~~, ~~#350~~, ~~#349~~ | > | ~~355~~ | Records | Records→spacetime | *new (2026-11-10, the author's programme "From `Σ` to physical spacetime", milestone M3 — measurement statistics, gate A; part B, the event about two times; out of the Remainder of ~~#354~~)* **The click law of the pointer of `Σ_*`: fixed by the flow, and not an operator's expectation.** ▸ **DONE 2026-11-10 — gate A is answered, and the answer is an obstruction.** [`RecordLayer/DiracPointerClick.lean`](../CsdLean4/RecordLayer/DiracPointerClick.lean) (28 pins). The click event `diracClick` is *ready at `0` and displayed at `T`*, built from the two arcs, the readout time and `diracPointerFlow`; `diracClickProb` is its probability under `epistemicMeasure`, a function of the carrier's ray. **It is determined:** `diracClickProb_eq_volume`, `clickWindow`, `clickWindow_le`, `clickWindow_eq_zero`, `clickWindow_eq_of_mem`, ★★★ `diracClickProb_eq_lintegral` (**the click probability is the `θ₂`-average of a fixed window of the pointer's displacement** — no rate field, no region that depends on the carrier; the window is `0` for a displacement below `1/6` and `1/6` on `[1/3, 2/3]`). **It has a cut-off:** `abs_diracPointerShift_le_weight` (the pointer turns by at most `|κ| (T w + (‖H_D‖ + |g f(θ₂)|) T²)`, `w` the occupancy of the detector's site at time `0`); ★★★ `diracClickProb_eq_zero_of_weight_le` (**where that bound is below a sixth of a turn on every fibre, every carrier with occupancy at most `w₀` has click probability `0`** — `w₀` need not be `0`); `diracClickProb_le`, ★★ `diracClickProb_lt` (**no carrier clicks with the probability `1/6` of a ready pointer**, for every smooth profile, coupling and readout time); `ofReal_mul_volume_writeFibres_le_diracClickProb` (the lower bound of ~~#354~~). **It is not the expectation of an operator:** `arenaObs_mk_eq_inner`, `pairVec`, `pairRay`, `arenaObs_pairRay`, ★★ `exists_mul_le_abs_sub_arenaObs` (a function on rays that vanishes at the rays of `e_a`, `e_a + s e_b`, `e_a − s e_b` differs from `⟨E⟩` somewhere by at least its value at the ray of `e_b` times `s² / (2 (1 + s²))`, for **every** matrix `E`); `arenaObs_siteProj_pairRay`; ★★★ `exists_mul_le_abs_diracClickProb_sub_arenaObs` (**with a cut-off at occupancy `s² / (1 + s²)`, for every matrix `E` some carrier has a click probability differing from `⟨E⟩` by at least `P(siteRay j) · s² / (2 (1 + s²))`**); ★★ `not_exists_diracClickProb_eq_arenaObs`. **The check:** `abs_trainF_deriv_lift_le`, `abs_trainF_lift_le`, `trainF_weight_cutoff`; ★★★ `diracPointer_clickLaw` (**the package: for every ring, `J`, `μ` and site there are a coupling and a readout time at which, for the profile `trainF`, the click law holds, the carrier on the detector's site clicks with positive probability, every carrier clicks with probability below `1/6`, every carrier with occupancy at most `1/8` has click probability `0`, and on a ring with a second site every matrix is at distance at least `P(siteRay j) / 20` from the click probability**). **Gate A:** the selected dynamics does determine a click probability, and it is not the quantum one; the quantum weights enter the corpus only through the rate field of UA-6 (Posit 1), which this dynamics does not supply — the reconstruction stays conditional on UA-6, and the obstruction is M3's result. Contract: calculation 2 of the M3 section proved, corrected and sharpened; UO-7 met, UO-9 met → UD-50–UD-54 in [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); no new `UA` entry; [`POSITS.md`](POSITS.md) Posit 1 carries the answer on its discharge line (standing and count unchanged). **Scope:** the click law is exact; the window's values between `1/6` and `1/3` and between `2/3` and `5/6` are bounded and not computed; the cut-off is a sufficient condition, the distance a lower bound; the comparison is with every matrix (so with an effect, an efficiency factor, an effect carried back along a linear evolution); the probability is that of the joint event (six times it is the probability conditional on a ready pointer, and the comparison is the same); the package is **one coupling and one readout time**, short (`‖H_D‖ T ≤ 1/8`), and for long readout times only the click law and the bound `1/6` are proved; uniform fibre law (UA-2); one pointer, one site; **it is a statement about the coupling of UA-5, in which the pointer is driven by the expectation `⟨P_j⟩`, and not about every dynamics on `Σ`.** Remainder: the conditional error bound — the basin statistic of the moment-map field along the coupled flow against `⟨ψ, P_i ψ⟩`, with the disturbance as an explicit term and UA-6 named (UO-8, UO-12) → ~~#356~~; several detectors and one outcome per trial (UO-10); repeated trials generated by one flow (UO-11); no-signalling and sequential update on a shared arena (UO-13); the window on its two ramps and the exact threshold; long readout times; a non-uniform law for `θ₂`; a pointer whose moving angle acts back on the carrier (UO-24), the dynamics the obstruction does not cover. | ~~**M**~~ done | High — gate A answered: the Hamiltonian write of `Σ_*` fixes its own click statistics and they are not the quantum ones; QM's weights stay conditional on Posit 1 | ~~#354~~, ~~#352~~, ~~#350~~, ~~#349~~ | > | ~~356~~ | Records | Records→spacetime | *new (2026-11-10, the author's programme "From `Σ` to physical spacetime", milestone M3 — measurement statistics; part C, the conditional reconstruction gate A asks to keep; out of the Remainder of ~~#355~~)* **The basin statistic of a rate field along the coupled flow of `Σ_*`: the conditional error bound.** ▸ **DONE 2026-11-10 — conditional on UA-6 (Posit 1) throughout; no weight is derived.** [`RecordLayer/DiracPointerBasin.lean`](../CsdLean4/RecordLayer/DiracPointerBasin.lean) (18 pins). **The weight along the carrier's orbit:** `norm_single_self_le`, `diracWeight` (the expectation of the basis projector `\|a⟩⟨a\|` at the ray moved by the coupled propagator on the fibre `θ₂`), `diracWeight_eq_momentMap`, `diracWeight_nonneg`, `diracWeight_zero`, `diracWeight_zero_coupling_congr`, ★ `diracWeight_zero_coupling_mk` (with the coupling off it is `\|(e^{−itH_D} ψ)_a\|² / ‖ψ‖²`, the weight quantum mechanics assigns with no detector attached); ★★ `abs_diracWeight_sub_zero_coupling_le` (**the detector's term moves a weight by at most `2 \|t\| \|g f(θ₂)\|`**); ★ `abs_diracWeight_sub_init_le` (within `2 \|t\| (‖H_D‖ + \|g f(θ₂)\|)` of the weight at time `0`). **The basin of a torus-generated field:** `momentMap_idxArenaEquiv`, `diracBasinProb` (the probability under `epistemicMeasure` of the basin of the basis state `a` at time `t`), `rate_diracArenaEquiv_of_isTorusGenerated`; ★★ `diracBasinProb_eq_lintegral` (for a torus-generated rate field the basin probability is the `θ₂`-average of the weight — exact); ★ `diracBasinProb_zero_coupling` (**with the coupling off it is the free weight, exactly** — the error term is the detector's). **The bound:** `abs_toReal_lintegral_ofReal_sub_le`; ★★★ `abs_diracBasinProb_sub_zero_coupling_le` (**for a profile with `\|f\| ≤ F` the basin probability is within `2 \|t\| \|g\| F` of the free weight at time `t`** — back-reaction on the carrier, and nothing else); ★★ `abs_diracBasinProb_sub_init_le` (within `2 \|t\| (‖H_D‖ + \|g\| F)` of the weight in the carrier as prepared — finite time plus back-reaction). **The check:** ★★ `abs_diracBasinProb_trainF_sub_le` (for `trainF` at the coupling `g = 1/(4T)` and readout time `T` of ~~#355~~ the disturbance is at most `1/12`, for every carrier, basis state and `T > 0`). Contract: UO-8's conditional bound met and UO-12 met → UD-55–UD-57 in [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); no new `UA` entry; `POSITS.md` unchanged (the posit's standing does not move). **Scope:** every statement about the basin has `IsTorusGenerated c` as a hypothesis, which puts the quantum weights into the rate field (`torusGenerated_eq_momentMap`; instance `isTorusGenerated_momentContext`) — by §6 of the contract it passes no gate, and gate A's answer (~~#355~~) stands; what is proved is how the coupled Hamiltonian flow carries that statistic and by how much the detector moves it; **the basin event is not the pointer's record** — the pointer's displacement does not enter it, it is not the click event, and no relation between the two is proved; the terms of `ε_meas` present are back-reaction and finite time, and a transition region and a retention time are not terms because the event is read at one time; the bounds are upper bounds, not shown attained, and say something only where the right-hand side is below `1`; outcome = a basis index `(site, spin)`; uniform fibre law (UA-2); one pointer, one trial. Remainder: a lower bound on the disturbance (whether `2 \|t\| \|g\| F` is the right order); the basin of a site with the same error; several detectors and one outcome per trial (UO-10); repeated trials generated by one flow (UO-11); no-signalling and sequential update on a shared arena (UO-13); a non-uniform law for `θ₂`; a pointer whose moving angle acts back on the carrier (UO-24). | ~~**S**~~ done | Medium — M3's conditional half: given Posit 1's rate field, the Hamiltonian flow with the detector attached reproduces QM's weights up to a stated back-reaction term | ~~#355~~, ~~#354~~, ~~#350~~, ~~#349~~ | > | ~~357~~ | Records | Records→spacetime | *new (2026-11-10, the author's programme "From `Σ` to physical spacetime", milestone M4 — the clocks inside the same model; part A, the model and the law for the pointer's displacement)* **A two-level clock inside the common Hamiltonian: the pointer follows the clock's hand.** ▸ **DONE 2026-11-10 — part A only; every statement is in the flow parameter, and operational time from clock records is not reached.** [`RecordLayer/UnifiedDiracClock.lean`](../CsdLean4/RecordLayer/UnifiedDiracClock.lean) (96 pins). **The model:** the device of ~~#271~~ (`deviceHam J μ₀ μ₁`, the Dirac ring with a two-level internal mass) on `ℂℙ^{4M+3} × T²`, joint energy `clockPointerEnergy` = `−2⟨H_dev⟩ − 2 g f(θ₂)⟨X⟩` with `X = levelSwap` the exchange of the two levels at every site (`clockPointerEnergy_eq`). ★★★ `clockPointerFlow_apply` (**its Hamiltonian flow in closed form:** `θ₂` constant, the device moved by `exp(−it(H_dev + g f(θ₂) X))`, the pointer `θ₁` advancing at `−2 g f'(θ₂)⟨X⟩` — ★★ `hasDerivAt_clockCarrierU`, ★★ `hasDerivAt_clockPointerShift`; the pointer has no dynamics of its own); `arenaVolume_map_clockPointerFlow`, `kMuL_map_clockPointerFlow`. **The generic layer** (the pattern of ~~#349~~/~~#351~~ for any Hermitian pair and index equivalence — its second use): `idxCoupledH`, `idxCarrierU`, `idxFreeU`, `idxPointerShift`, ★★ `hamiltonianFlow_coupledEnergy_reindex`, ★★ `abs_idxPointerShift_sub_le` (the pointer against the uncoupled carrier, within `2 \|f'\| ‖B₀‖² \|f\| t²`). **The control:** ★ `idxPointerShift_one` (a pointer coupled to the identity advances at `−2 f'(θ₂)` whatever the carrier does — it reads the flow parameter and no clock; the case §6 of the contract excludes), `clockPointerShift_zero_coupling`. **The law:** ★★ `clockFreeRead_deviceRay` (along the uncoupled orbit `⟨X⟩ = 2 Re C / ‖Ψ‖²`, `C` = `deviceCoherence` — exact); ★★ `abs_clockPointerShift_sub_free_le` (back-reaction: `2 \|f'\| g² \|f\| t²`); ★★ `abs_integral_clockFreeRead_sub_le` (localisation: a coherence drift `\|s\| ε` costs `ε t² / ‖Ψ‖²`); `clockQuad`, `clockHand` = `∫₀ᵗ Re(e^{−iRs} C₀) ds`, ★ `integral_clockQuad`, ★ `clockHand_add_period`, ★ `abs_clockHand_le`, ★ `clockQuad_eq_zero_iff`; ★★★ `abs_clockPointerShift_sub_hand_le` (**`\|Δθ₁(t) + 4 g f'(θ₂) clockHand R C₀ t / ‖Ψ₀‖²\| ≤ 2 \|f'(θ₂)\| (g² \|f(θ₂)\| + \|g\| ε / ‖Ψ₀‖²) t²`**); ★ `clockFreeRead_single`, ★★ `clockPointerShift_single_eq_hand` (one mode on a fibre with `f(θ₂) = 0`: the law is exact for all `t`), ★★ `clockRead_single_eq_zero_iff` (the pointer's turning points are the flow times at which the coherence has turned to an odd multiple of `π/2`, `π/R` apart), ★ `clockPointerShift_single_add_period` (the pointer returns after `2π/R`); ★★★ `clock_pointer_device` (**for the packet of `dirac_device`: `ε = c D`, and `R/(μ₁ − μ₀)` is within `(μ₁ − μ₀)/E₀` of `√(1 − (v/c)²)`** — the dilation clause is `dirac_device`'s, reused). Contract: UO-14 **met in part and kept** → UD-58–UD-60 in [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); new entry UA-12 (the clock's level and its coupling, boundary R-015); `POSITS.md` unchanged. **Scope:** all times are flow times (UA-7) — no observer in the model reads them and nothing is recovered from records; **the pointer is a hand, not a record** — it reads the unrectified quadrature, so it swings and returns, there is no ready region, threshold, `ActualOccurrence` or retention, and its position at one time does not hold the number of turns (contrast ~~#271~~'s rectified register, a stated law); the law is exact only for one mode at `f(θ₂) = 0`, elsewhere an upper bound growing as `t²`, not shown attained; turning points and period are proved at `f(θ₂) = 0` only; the displacement's size depends on the hidden `θ₂` and on `g`, so the rate enters only as the period; the readout is global (`X` acts at every site); one pointer — no position record and clock reading in one run; M4's detector-resolution error is not a term (there is no readout). Remainder: (M4 part B → ~~#358~~: a tick as a record, the count, operational time and the detector-resolution term are there. The count is in the record history and no real function of the state holds it, and the unit is one full turn, `2π/(μ₁ − μ₀)`, not the `π/(μ₁ − μ₀)` written here before); the turning points under back-reaction (`f(θ₂) ≠ 0`); a local readout; two pointers on one arena (the clock's and ~~#349~~'s) so that clock and position records come from one run; library-owed: ~~#349~~'s `diracCarrierU`/`diracPointerShift` as instances of the indexed layer. | ~~**M**~~ done | Medium — the clock and its pointer are one Hamiltonian system and the relativistic rate reaches the pointer with stated errors; not yet operational time | ~~#349~~, ~~#351~~, ~~#271~~, ~~#262~~, ~~#350~~ | > | ~~358~~ | Records | Records→spacetime | *new (2026-11-10, the author's programme "From `Σ` to physical spacetime", milestone M4 — the clocks inside the same model; part B, ticks as records, the count and operational time)* **Ticks of the clock as records of the joint flow; operational time from the count.** ▸ **DONE 2026-11-10 — UO-14 met with five conditions that stay; M4's acceptance sentence is not met in full, because no arena carries the clock's pointer and a position detector.** [`RecordLayer/UnifiedDiracClockTicks.lean`](../CsdLean4/RecordLayer/UnifiedDiracClockTicks.lean) (60 pins). No dynamics, coupling or degree of freedom is added to ~~#357~~: the same pointer of `clockPointerFlow` is read through the two arcs of ~~#352~~ (`tickReady`, `tickDisplay`, `tickShown`). **Ticks and the count (any history):** `HasTicks` (`n` successive passages from resting to displaying by `T`), `IsTickCount`, ★★ `hasTicks_comp_orderIso`, ★ `isTickCount_comp_orderIso` (**the count is unchanged by any increasing relabelling of the parameter, with the terminal event carried along — it holds no value of the flow parameter**); `IsSwing` (a history resting at whole turns of a phase `R t` and displaying around the half turns), ★ `IsSwing.hasTicks`, ★★ `IsSwing.sub_one_mul_le`, ★★★ `IsSwing.abs_tickCount_sub_le` (**`\|2πN − R T\| ≤ 2π`**), ★ `IsSwing.exists_tickCount`, ★★ `IsSwing.isRecordWrite` (each tick is an `IsRecordWrite`: start `2πk/R`, write between the phases `2πk + arccos c` and `2πk + arccos d'`, window `2(π − arccos d')/R`), ★ `IsSwing.notMem_disp_of_cos_eq_one` (each record is erased by the next whole turn). **Operational time:** `clockOpTime` = `2πN/(μ₁ − μ₀)`; ★★ `IsSwing.abs_clockOpTime_sub_le`; ★★★ `IsSwing.abs_clockOpTime_sub_dilated_le` (**`\|2πN/(μ₁ − μ₀) − √(1 − (v/c)²) T\| ≤ 2π/(μ₁ − μ₀) + ((μ₁ − μ₀)/E₀) T`** — the detector's resolution, one tick, plus the finite splitting). **The tick law for the flow:** ★★ `isSwing_of_abs_shift_sub_le`, `clockSwingAmp`, `clockSwingErr`, ★★ `abs_clockPointerShift_sub_swing_le` (the hand of ~~#357~~ as `A (1 − cos R t)`, within `2 \|f'\| (g² \|f\| + \|g\| ε/‖Ψ₀‖²) T² + 4 \|g f'\| \|Re C₀\|/(R ‖Ψ₀‖²)`), ★★★ `isSwing_clockPointerFlow` (the readout history of `clockPointerFlow` is a swing at the device's rate, under four inequalities on the start, the amplitude and the error); ★★★ `isSwing_clockPointerFlow_single` (**exact: one mode on a fibre with `f(θ₂) = 0` — no error term, every horizon**), ★★ `isSwing_clockPointerFlow_train` (the instance: `trainF`, `θ₂ = 0`, half amplitude `1/4`, start `1/12`, displaying exactly while `cos R t < 0`). **The packet:** ★★ `norm_deviceCoherence_shift_sub_le` (the coherence from any start is within `\|s\|` times `deviceDrift` of a uniform turn), ★★★ `clock_tick_device` (the packet of `dirac_device` prepared a quarter turn early: `\|Re C₀\| ≤ (π/(2R)) ε`, `Im C₀ ≥ (1 − ((μ₁ − μ₀)/μ₀)²)/2 − (π/(2R)) ε`, the swing under the four inequalities, and the operational-time inequality at the packet's central velocity; the localisation error `ε = deviceDrift` enters through the swing's hypotheses and not through the final inequality). **Each tick an occurrence:** ★★★ `clockTickRecord` (an `ActualOccurrence` whose `evolve` is `clockPointerFlow` over the elapsed time, readout `tickShown`, preparation the flow's own state at the whole turn), ★★ `clockTickRecord_history` (all tick records of a run read one history). **The obstruction:** ★★ `ae_comp_eq_of_ae_le_comp`, ★★ `not_ae_add_one_le_comp`, ★★★ `not_ae_count_advances_clockPointerFlow` (**no measurable real function of the arena state gains at least one along a step of the flow almost everywhere** — the count is in the record history and not in a register; it holds for every map preserving a finite measure, so by §6 test 4 of the contract it is conditional on UA-2 and passes no gate). Contract: UO-14 **met with five conditions kept** → UD-61–UD-65 in [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); no new assumption entry (the arcs, the pointer's start and the device's prepared phase are UA-4); UA-4, UA-7 and UA-12 annotated; `POSITS.md` unchanged. **Scope:** one tick per full turn, so the unit is `2π/(μ₁ − μ₀)` and not the `π/(μ₁ − μ₀)` of ~~#271~~'s rectified register; the count and the operational-time statements are upper bounds, not shown attained; the tick law is proved outright only for one mode on the fibres with `f(θ₂) = 0`, a null set for a generic profile; elsewhere the swing is four inequalities taken as hypotheses, the error grows as `T²`, and no packet of positive width and no fibre with `f(θ₂) ≠ 0` is shown to meet them; `T`, and each record's start, write interval and window, are flow times (UA-7) — the count and `clockOpTime` do not use them, and nothing in the model selects the event the count is taken at; the amplitude and the thresholds depend on the hidden `θ₂`, on `g` and on the start (the rate and the count do not); the readout is global (the site is the whole ring); one pointer; the dilation clause is `abs_clockRate_div_sub_le`, reused. Remainder: (a) a packet of positive width, or a fibre with `f(θ₂) ≠ 0`, shown to meet the four inequalities up to a stated horizon, with the number of ticks it gives; (b) a register for the count — an angle each tick advances (a count modulo a fixed number is not excluded by the obstruction, which is about real functions); (c) two pointers on one arena (the clock's and ~~#349~~'s) → ~~#359~~: one joint flow carries both (`clockSiteFlow_apply`). A tick record and a position record in one run — M4's acceptance in full, and gate B — are not there; they are that row's Remainder (a); (d) the count taken at a record event (an emission, a reception) and not at a flow time — what M6 needs; (e) a readout at a site; (f) whether the bounds are attained; library-owed as ~~#357~~. | ~~**M**~~ done | Medium–High — the clock's reading is a record of the one flow and operational time is a function of tick records, with the three errors M4 names; conditional beyond one mode, and not yet on one arena with a detector | ~~#357~~, ~~#353~~, ~~#352~~, ~~#271~~, ~~#350~~ | > | ~~359~~ | Records | Records→spacetime | *new (2026-11-10, the author's programme "From `Σ` to physical spacetime", between milestones M4 and M5 — the arena M4's acceptance and M5 both need)* **Two pointers on one arena: two site detectors on the Dirac ring, and the clock's pointer beside a site detector.** ▸ **DONE 2026-11-10 — the joint flows are solved; no record is read on them and no propagation bound is proved, so no obligation of the contract is met and M4's acceptance sentence stands where ~~#358~~ left it.** [`LF4/EchoVolume.lean`](../CsdLean4/LF4/EchoVolume.lean) (18 pins) and [`RecordLayer/UnifiedTwoPointer.lean`](../CsdLean4/RecordLayer/UnifiedTwoPointer.lean) (91 pins). **The arena was already in the corpus:** `ℂℙᴺ × T² × T²`, its form (`echoForm`, `echoForm_isSymplectic`) and the closed-form flow of an energy with two product terms are ~~#202~~'s (`hamiltonianFlow_twoCoupledEnergy`); the contract had not adopted it, and its sentence that no arena carries both pointers was about the contract's own arenas. **The general flow (`LF4`, every `N`, every Hermitian triple):** `twoCoupledFlow`, `twoCoupledFlow_add`, ★★ `twoCoupledFlow_fst`, ★★ `twoCoupledFlow_snd` (with the other torus's conserved angle held, the ray and one torus move by the one-torus flow with the other coupling added to the uncoupled matrix), ★ `twoCoupledFlow_fst_of_lift_eq_zero`, ★ `twoCoupledFlow_snd_of_lift_eq_zero`, ★ `measurePreserving_addShear`, `echoMuL`, ★★★ `echoMuL_map_twoCoupledFlow` (**the flow preserves `μ_FS ⊗ vol ⊗ vol`**, read off the closed form by a skew product). **In the coordinates of a carrier (any finite index type):** `idxTwoH`, `idxTwoCarrierU`, ★★ `hasDerivAt_idxTwoCarrierU`, `idxTwoShiftFst`, `idxTwoShiftSnd`, ★★★ `twoCoupledFlow_reindex` (**one propagator `exp(−it(A₀ + f(θ₂)B₀ + h(ψ₂)C₀))` for the carrier; `θ₁` advances by `−2f'(θ₂)∫⟨B₀⟩` and `ψ₁` by `−2h'(ψ₂)∫⟨C₀⟩` along that one orbit; `θ₂`, `ψ₂` fixed** — the pointers act on each other only through the carrier, as a static term in its generator), `idxTwoCarrierU_of_snd_lift_eq_zero`, `idxTwoShiftFst_of_lift_eq_zero` and their mirrors. **Cross-talk:** ★ `norm_idxTwoCarrierU_sub_fst_le` (`≤ \|t\| \|h(ψ₂)\| ‖C₀‖`), ★★ `abs_idxTwoShiftFst_sub_le` (**the first pointer's displacement is within `2 \|f'(θ₂)\| ‖B₀‖ \|h(ψ₂)\| ‖C₀‖ t²` of its displacement alone**, `t ≥ 0`), ★★ `abs_idxTwoShiftSnd_sub_le`, and the forms for real strengths (`abs_idxTwoShiftFst_sub_le_of_norm_le`, `abs_idxTwoShiftSnd_sub_le_of_norm_le`). **Two site detectors on the Dirac ring** (`ℂℙ^{2M+1} × T² × T²`): `diracTwoPointerEnergy`, `diracTwoPointerEnergy_eq`, `diracTwoH` (**`H_D + g₁ f(θ₂) P_j + g₂ h(ψ₂) P_k` — the coupled generator M5's bound is to be about**), ★★ `hasDerivAt_diracTwoCarrierU`, `diracTwoPointerFlow`, ★★★ `diracTwoPointerFlow_apply`, ★★ `diracTwoPointerFlow_fst_of_lift_eq_zero`, ★★ `diracTwoPointerFlow_snd_of_lift_eq_zero` (on a fibre where the other profile vanishes it **is** `diracPointerFlow` of ~~#349~~), ★★ `echoMuL_map_diracTwoPointerFlow`, ★ `norm_diracTwoCarrierU_sub_le`, ★★ `abs_diracTwoShiftFst_sub_le`, ★★ `abs_diracTwoShiftSnd_sub_le` (against `diracCarrierU`, `diracPointerShift`). **The clock's pointer and a site detector on the device** (`ℂℙ^{4M+3} × T² × T²`): `deviceSiteProj` (the site projector on both levels; `deviceSiteProj_eq_diagonal`, `deviceSiteProj_mul_self`, `norm_deviceSiteProj_le`), `clockSiteEnergy`, `clockSiteEnergy_eq`, `clockSiteH` (`H_dev + g₁ f(θ₂) X + g₂ h(ψ₂) Q_k`), ★★ `hasDerivAt_clockSiteCarrierU`, `clockSiteFlow`, ★★★ `clockSiteFlow_apply`, ★★ `clockSiteFlow_fst_of_lift_eq_zero` (on a fibre with `h(ψ₂) = 0` it **is** `clockPointerFlow` of ~~#357~~), ★ `clockSiteFlow_snd_of_lift_eq_zero`, ★★ `echoMuL_map_clockSiteFlow`, ★★ `norm_clockSiteCarrierU_sub_le`, ★★ `abs_clockSiteShiftFst_sub_le` (the clock's hand with the detector on is within `2 \|f'(θ₂)\| \|g₁\| \|h(ψ₂)\| \|g₂\| t²` of the hand of ~~#357~~), ★★ `abs_clockSiteShiftSnd_sub_le`. Contract: **new assumption UA-13** (the two-pointer arena, its product measure, the additive energy with no term between the pointers); UD-66 to UD-71 in a new §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-10, UO-11, UO-13, UO-14 and UO-15 annotated, none met; `POSITS.md` unchanged. **Scope:** the closed forms, the reductions and the measure statements are exact; the cross-talk statements are upper bounds growing as `t²`, not shown attained, and they do not depend on the distance between the sites, so they are not a cone bound; the exact reductions hold on the fibres where a profile vanishes, a null set for a generic profile, and the theorems of ~~#349~~ to ~~#358~~ are not proved again on the two-pointer arenas; on a fibre with `h(ψ₂) = 0` and `h'(ψ₂) ≠ 0` the second pointer still moves; two tori, not a family; the measure statement is read off the closed form, and the top-form measure of `echoForm` is not identified with `echoMuL`; `deviceSiteProj` is a definition; the site pointer alone on the device has no module, so its reduction and cross-talk are stated against the generic one-pointer flow; every time is the flow parameter (UA-7). Remainder: (a) records in one run → ~~#360~~: a tick record of the clock's pointer and a position record of the site pointer are two `ActualOccurrence`s of `clockSiteFlow` (`clockSiteTickRecord`, `clockSitePositionRecord`), outright in one instance on a null set of fibres and conditionally elsewhere; what is left of M4's acceptance and of gate B's question for the clock and the detector is that row's Remainder; (b) M5's propagation bound for `diracTwoH` and `clockSiteH` → ~~#361~~: both generators hop at most one site and the cone bound holds for them (`norm_commutator_hopsOne_le`); UO-15 stays with five conditions (contract UO-15 (e1)–(e5)); what would lift them is ~~#361~~'s Remainder; (c) one trial, one outcome and no-signalling for two detectors on `diracTwoPointerFlow` (UO-10, UO-13); (d) a third pointer, which M6's radar exchange is likely to need; (e) whether the cross-talk bounds are attained; (f) `echoMuL` as the top-form measure of `echoForm`. | ~~**M**~~ done | Medium — it fixes the one dynamics that M4's two records and M5's two events must come from, and the coupled generators M5's bound is about; it reads no record and passes no gate | ~~#358~~, ~~#357~~, ~~#349~~, ~~#202~~, ~~#350~~ | > | ~~360~~ | Records | Records→spacetime | *new (2026-11-10, the author's programme "From `Σ` to physical spacetime", milestone M4 — condition (d) of UO-14, Remainder (a) of ~~#359~~)* **A tick record and a position record in one run of one flow.** ▸ **DONE 2026-11-10 — both records are occurrences of `clockSiteFlow` from one start state; outright in one exact instance (one momentum mode, the fibres where both profiles vanish: a null set), conditional elsewhere. UO-14 (d) is met with three conditions that stay in the contract; M4's acceptance sentence is met in that instance and conditionally elsewhere.** [`RecordLayer/UnifiedClockSiteRecords.lean`](../CsdLean4/RecordLayer/UnifiedClockSiteRecords.lean) (60 pins). **The readouts are apparatus input** (UA-4, extended): `echoTickShown`, `echoTickReady`, `echoTickDisplay` on the clock's torus and `echoSiteShown`, `echoSiteReady`, `echoSiteDisplay` on the site pointer's — the arcs `(0, 1/6]` and `(1/3, 5/6]` of ~~#352~~ and ~~#358~~, placed by hand. **The site's occupancy:** ★ `arenaObs_deviceSiteProj` (`⟨Q_k⟩` is a sum of four moment coordinates), `arenaObs_deviceSiteProj_nonneg`, `arenaObs_deviceSiteProj_le_one`, `sum_arenaObs_deviceSiteProj` (the occupancies sum to one), ★ `arenaObs_deviceSiteProj_single` (**one momentum mode: `1/(M+1)` at every site and every time**). **The site pointer with the clock coupled:** `clockSiteShiftSnd_eq_rateAt`, ★★ `clockSiteShiftSnd_mono` (**on a fibre of nonnegative rate the displacement is monotone in time for every carrier and every clock coupling — no cross-talk term is needed**), `clockSiteShiftSnd_le`, ★★★ `isRecordWrite_clockSiteFlow_site` (from a ready pointer, displacement `≤ 1/6` at `τ₀`, `≥ 1/3` at `τ₁`, `≤ 2/3` at `τ₂` ⇒ the readout history writes `true` on `[τ₀, τ₁]` and holds it for `τ₂ − τ₁`), ★ `isRecordWrite_clockSiteFlow_site_of_alone` (the three values read off the device with the site pointer alone, the cross-talk of ~~#359~~ as a margin), ★★★ `clockSitePositionRecord` (an `ActualOccurrence` of the joint flow), `clockSitePositionRecord_history`. **The clock's pointer with the detector coupled:** `IsSwing.preimage`, `clockSiteCrossErr` (`2 \|f'(θ₂)\| \|g₁\| \|h(ψ₂)\| \|g₂\| T²`), ★★ `abs_clockSiteShiftFst_sub_swing_le` (within `clockSwingErr + clockSiteCrossErr` of `A (1 − cos R t)` up to the horizon `T`), ★★★ `isSwing_clockSiteFlow` (**a swing under the four inequalities of ~~#358~~ with the cross-talk added to the error**, so the count, the tick as a write and the operational time of ~~#358~~ apply to it), ★★ `isSwing_clockSiteFlow_of_lift_eq_zero` (on a fibre with `h(ψ₂) = 0` every swing of ~~#358~~ carries over with no added error), ★★★ `clockSiteTickRecord` (tick `n` as an `ActualOccurrence` of the joint flow), `clockSiteTickRecord_history`. **One run:** ★ `clockSiteTickRecord_evolve`, `clockSitePositionRecord_evolve` (both records have the one propagator `clockSiteEvolve`), ★ `clockSiteTickRecord_prep` (the tick record's preparation is the position record's start state carried by that propagator to the start of turn `n`). **An exact instance** (one momentum mode prepared a quarter turn ahead, both profiles `trainF`, fibres `θ₂ = 0` and `ψ₂ = 0`, `g₂ > 0`): `clockSiteTrainStart`, `arenaKick_clockSiteCarrierU_devicePacket`, ★★ `clockSiteShiftSnd_single` (the site pointer's displacement is the rate times `t/(M+1)`), ★★★ `isRecordWrite_clockSiteFlow_site_single`, ★★ `isSwing_clockSiteFlow_train`, ★★ `isRecordWrite_clockSiteFlow_site_train`, ★★★ `clockSite_records_train` (**for every tick index `n`: the tick readout's history is a write begun at the turn's quarter, and the site readout's history is a write begun at `(M+1)/(12 g₂)` and held for `(M+1)/(6 g₂)` — one start state, one flow, two readouts**). Contract: no new assumption (UA-4 item 2 extended to the two-pointer arena, UA-13 updated); UD-72 to UD-76 in §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-14 (d) annotated as met with conditions (d1)–(d3); `POSITS.md` unchanged. **Scope:** the instance is on the fibres where both profiles vanish, a null set of `echoMuL`, and for one mode; there neither pointer acts on the device (it moves freely and drives both), so the cross-talk is zero and an action of one apparatus on the other record is not exhibited; off those fibres the swing's four inequalities and the write's three displacement values are hypotheses and the swing's error grows as `T²`; in the instance the occupancy of every site is `1/(M+1)`, so the position record does not single out its site — it is a definite fact of the one trajectory (the accumulated occupancy of site `k` crossed the threshold), not a statistical statement, and no QM weight is derived from it; tick records are held for less than a turn, the position record for a stated window; every time is the flow parameter (UA-7); the count of ticks at the position write is not taken (that write as an event is UO-14 (c)); the readouts and ready arcs are definitions. Remainder: (a) an instance on a set of fibres of positive measure — both sets of hypotheses discharged with `f(θ₂) ≠ 0` or `h(ψ₂) ≠ 0`, up to a stated horizon (UO-14 (d1), with UO-14 (a)); (b) a position record that singles out its site: a carrier localised near site `k` (a packet of positive width), with the write at `k` and no write at a distant site within the window (UO-14 (d2), UO-12's question on this arena) → ~~#363~~: `clockSiteFlow_site_shown_iff`, on pairs `(θ₂, ψ₂)` of positive Haar measure, comparing two placements of the detector; one run with two detectors and a moving packet are that row's Remainder; (c) the count of ticks at the position write, with the write as the event (UO-14 (c), M6) → ~~#366~~: `clockSite_count_at_display_train`, at the instant the readout switches, the two histories aligned by the flow parameter; (d) an instance in which one apparatus measurably acts on the other record (cross-talk attained, ~~#359~~ Remainder (e)); (e) the same two-record statement on `diracTwoPointerFlow` with a readout on each torus (UO-10, UO-13). | ~~**M**~~ done | Medium–high — the first statement in the corpus with a clock record and a detector record generated by one flow in one run, which is M4's acceptance sentence and gate B's question for the clock and the detector; outright only on a null set of fibres | ~~#359~~, ~~#358~~, ~~#357~~, ~~#353~~, ~~#352~~, ~~#350~~ | > | ~~361~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", milestone M5 — the propagation bound of UO-15, Remainder (b) of ~~#359~~)* **A propagation bound for the coupled generators of the two-pointer arenas, and the response of a pointer to a remote intervention.** ▸ **DONE 2026-11-11 — the cone bound is proved for `diracTwoH` and `clockSiteH`; a site pointer's response to a remote kick or to the other site pointer's angle is at most a factorial tail in the distance. Upper bounds only: suppression, not exclusion, and no response is shown to arrive. No order between record events is built, so UO-15 stays, with five conditions in the contract.** [`RecordLayer/UnifiedCausalOrder.lean`](../CsdLean4/RecordLayer/UnifiedCausalOrder.lean) (65 pins). **The interaction topology as a property of a matrix** (over a site map from the carrier's index to the ring): `SiteSuppOn`, `IdOffOn` (the predicates of ~~#286~~'s `ChainCone.lean` for any index), the new `HopsOne`, `deviceSite`; ★ `hopsOne_diracHam`, `hopsOne_deviceHam` (the free generators hop at most one site), `hopsOne_siteProj`, `hopsOne_levelSwap`, `hopsOne_deviceSiteProj` (the coupled operators move nothing between sites), ★★ `hopsOne_diracTwoGen`, ★★ `hopsOne_clockSiteGen` (**both coupled generators hop at most one site at every pair of pointer angles**); `siteSuppOn_siteProj`, `siteSuppOn_deviceSiteProj`; ★ `mem_of_siteSuppOn_levelSwap` (**any set of sites that supports the level swap contains every site**). **The cone bound for any such generator:** `siteSuppOn_adOne`, `siteSuppOn_adIter` (one commutator grows a support by one site), `exists_int_of_mem_growIter`, `disjoint_growIter_of_forall_ne`, `disjoint_growIter_singleton` (sets more than `d` apart on the ring are missed by `d` steps of growth), `norm_commutator_heisenbergFlow_le_of_adIter_commute`, ★★ `norm_commutator_hopsOne_le`, ★★ `norm_commutator_hopsOne_idOffOn_le` (**`‖[A(t), B]‖ ≤ 2 (2‖S‖t)^d/d! · ‖A‖ ‖B‖` for a skew-Hermitian `S` that hops at most one site — the generator may carry any terms that move nothing between sites**); `norm_heisenbergFlow_sub_le_of_commutator_le` (a mean-value bound between two flows, along `r ↦ τ_r^S (τ_{t−r}^{S'} D)`), ★★ `norm_heisenbergFlow_sub_hopsOne_le` (**two generators that differ on `Y` evolve an observable on `X` alike within `2 (2‖S'‖t)^d/d! · ‖D‖ ‖S − S'‖ t`**). **Two interventions, both changes of the initial point** (UA-14, new): ★ `arenaObs_kick_of_idOffOn` (exact at time `0`), ★★ `abs_arenaObs_flow_kick_sub_le` (a kick of the carrier's initial ray that is the identity off `Y`), ★★ `abs_arenaObs_flow_sub_flow_le` (a change of the generator supported on `Y`), `abs_integral_arenaObs_flow_kick_sub_le`, `abs_integral_arenaObs_flow_sub_flow_le`, ★ `abs_displacement_kick_sub_le`, ★ `abs_displacement_flow_sub_le` (the same for a displacement `−2 g c ∫₀ᵗ⟨D⟩`). **Two site pointers** (`T_d(t) = (2‖H‖t)^d/d!`, `H` the coupled Hamiltonian): ★★ `abs_diracTwoKickResponseSnd_le`, `abs_diracTwoKickResponseFst_le` (**a pointer's displacement moves by at most `4 \|g\| \|profile'\| T_d(t) t` under a kick more than `d` sites away on the ring**), ★★ `abs_diracTwoAngleResponseSnd_le`, `abs_diracTwoAngleResponseFst_le` (**between two runs that differ in the other pointer's conserved angle it differs by at most `4 \|g₂\| \|h'(ψ₂)\| \|g₁\| \|f(φ) − f(φ')\| T_d(t) t²` — a bound that falls with the distance between the sites, beside the cross-talk bound of ~~#359~~, which does not**), `diracTwoPointerFlow_kick_snd`, `diracTwoPointerFlow_kick_fst`, `diracTwoPointerFlow_angle_snd`, `diracTwoPointerFlow_angle_fst` (each response is the difference, on the circle, of a moving angle between two runs of the joint flow). **The clock and a site detector:** ★★ `abs_clockSiteKickResponseSnd_le`, `clockSiteFlow_kick_snd`, ★★★ `isRecordWrite_clockSiteFlow_site_of_kick` (**the position record of ~~#360~~ survives a remote kick: from a ready pointer on a fibre of nonnegative rate, if the unkicked displacement clears the three thresholds by the kick bound, the kicked run writes the same record over the same flow times**). Contract: one new assumption, UA-14 (the admitted interventions); UA-13's discharge line updated; UD-77 to UD-81 in §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-15 annotated as proved for the generators and open with conditions (e1)–(e5); `POSITS.md` unchanged. **Scope:** every bound at `t > 0` is a factorial tail, small when `2‖H‖t` is small against `d` and not zero unless a prefactor vanishes — nothing says a remote intervention has no effect at a positive time, and the one exact statement is at time `0`; upper bounds only, no response shown attained or bounded below, so influence is limited and never exhibited; the speed in the tail is `2‖H‖`, the norm of the whole generator with both couplings, not optimised and not the group velocity; an intervention is a change of the initial point, not a term of the joint energy; no relation between record events is defined and `CausalPrecedes` stays induced by an assumed graph; the predicates `HopsOne` and `SiteSuppOn` are the interaction topology of UA-8 and UA-13 restated on the matrix, proved for the two generators and not derived from anything else; on the arena of two site pointers no record is read, so the statements there are about displacements; no bound in the distance is proved for the clock's pointer, because the level swap is supported on no proper set of sites and the argument does not apply, so neither its response to a kick nor either pointer's response to the other's angle on that arena is bounded here; no instance in which the three margins of the record statement hold is exhibited, and `d` is at most about half the ring; every time is the flow parameter (UA-7). Remainder: (a) a lower bound — a response shown to arrive inside the cone, for some carrier and kick (UO-15 (e2), with ~~#359~~ Remainder (e)) → ~~#362~~ at distance one: `diracTwoKickResponseSnd_ne_zero`; larger distances are that row's Remainder (a); (b) a relation between two record events built from responses, with the argument that it is an order, or the obstruction (UO-15 (e3)); (c) the two-site-pointer statements for records: a readout on each torus of `diracTwoPointerFlow`, then a write that survives a remote kick and a remote angle (UO-15 (e4), UO-10, ~~#360~~ Remainder (e)); (d) a clock whose coupled operator is supported near a site, so that a tick record has a cone (UO-15 (e5), UO-14 (e)); (e) an intervention made inside the dynamics by a third apparatus (UA-14's discharge, ~~#359~~ Remainder (d)); (f) the tail with the chain's group velocity in place of `2‖H‖`. | ~~**M**~~ done | Medium — the bound M5 named, for the generators the arenas have, and the first conditional statement that a record of this programme survives a remote intervention; it limits influence and does not exhibit it, builds no order and passes no gate | ~~#360~~, ~~#359~~, ~~#286~~, ~~#350~~ | > | ~~362~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", milestone M5 — condition (e2) of UO-15, Remainder (a) of ~~#361~~, the author's pick)* **A response that arrives: the leading term of a pointer's response to a kick, and a kick at the neighbouring site that is felt.** ▸ **DONE 2026-11-11 — on the arena of two site pointers, for one stated carrier and the sign kick at the site next to the second detector, the second pointer's response is `2 g₂ h'(χ) J t²` up to `8 \|g₂\| \|h'(χ)\| ‖H‖² t³`, at every pair of pointer angles; it is not zero for `0 < t < \|J\|/(4‖H‖²)` when `g₂ h'(χ) ≠ 0`. Distance one only; UO-15 (e2) is met there and stays for larger distances.** [`RecordLayer/UnifiedCausalArrival.lean`](../CsdLean4/RecordLayer/UnifiedCausalArrival.lean) (22 pins). **The expansion** (any finite index, any skew-Hermitian generator): `arenaObs_kick_of_commute`, `arenaObs_smul_real`, `abs_arenaObs_le`, `commute_adIter_of_disjoint_growIter` (the iterated commutators that have not reached the kick's set commute with the kick), ★★ `abs_arenaObs_flow_kick_sub_sub_le` (**if the first `d` iterated commutators of `D` commute with the kick, the response of `⟨D⟩` at time `s ≥ 0` is `(−1)^d s^d/d!` times the change the kick makes in `⟨ad_S^d D⟩`, up to `2 (2‖S‖s)^(d+1)/(d+1)! · ‖D‖`**), `integral_flowCoef`, ★★ `abs_displacement_kick_sub_sub_le` (the same for a displacement `−2 g c ∫₀ᵗ⟨D⟩`: leading term `−2 g c (−1)^d t^(d+1)/(d+1)! · Δ_d`, error `4 \|g\| \|c\| (2‖S‖t)^(d+1)/(d+1)! · t`). **The kick:** `siteSignKick` (the amplitudes over one site multiplied by `−1`), `siteSignKick_val`, `idOffOn_siteSignKick`. **The instance** (ring of at least three sites, stated as `k − 1 ≠ k + 1`): `arrival_ne`, `arrivalRay` (the ray of `e_(k,1) + e_(k−1,0)`), `arenaKick_siteSignKick_arrivalRay` (the kick carries it to the ray of `e_(k,1) − e_(k−1,0)`), `inner_single_toEuclideanLin_single`, `adOne_diagonal_apply`, `diracTwoGen_apply_arrival`, `diracTwoGen_apply_arrival'` (the coupled generator between the two indices is `±J`, whatever the pointer angles), ★★ `arenaObs_adOne_diracTwoGen_arrivalRay_sub` (**the kick changes the current into the detector's site by `2 J`**), ★★★ `abs_diracTwoKickResponseSnd_sub_le` (**the response is `2 g₂ h'(χ) J t²` up to `8 \|g₂\| \|h'(χ)\| ‖H‖² t³`**), ★★ `le_abs_diracTwoKickResponseSnd` (its size is at least `2 \|g₂\| \|h'(χ)\| t² (\|J\| − 4‖H‖²t)`), ★★★ `diracTwoKickResponseSnd_ne_zero` (**for `g₂ h'(χ) ≠ 0`, `0 < t` and `4‖H‖²t < \|J\|` the second pointer's displacement, a real number of turns, differs between the kicked and the unkicked run**), ★★ `diracTwoPointerFlow_kick_snd_ne` (for smooth profiles and `4 \|g₂\| \|h'(χ)\| t < 1` as well, its angles on the circle differ). Contract: no new assumption; UD-82 and UD-83 in §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-15 (e2) annotated as met at distance one; `POSITS.md` unchanged. **Scope:** the kick is one site from the detector, where the cone bound of ~~#361~~ holds with `d = 0` only (`4 \|g₂\| \|h'(χ)\| t`), and the response found is of order `t²`, below it; for a kick `d` sites away the expansion gives the order `t^(d+1)` with coefficient `Δ_d`, and `Δ_d` is not computed for `d ≥ 2`, so no response is shown to arrive from farther; one carrier (a stated superposition over two sites, with half its weight on the detector's own site) and one kick (a stated sign, which changes the relative sign of the two amplitudes); a carrier off the detector's site is not treated — for a carrier on one site this kick is a phase of the whole ray and changes nothing; the lower bound is positive only for `t < \|J\|/(4‖H‖²)`; the error constant is not sharp (the tail is bounded on `[0, t]` by its value at `t`); a displacement and not a record, since no readout is on this arena; the kick is a change of the initial ray (UA-14), not a term of the joint energy; no relation between record events is built; every time is the flow parameter (UA-7). Remainder: (a) `Δ_d` for a kick `d ≥ 2` sites away on the Dirac ring — a response of order `t^(d+1)` shown to arrive, the two-sided form of the cone (UO-15 (e2)); (b) the first pointer's response, and the response to the other pointer's angle, with their leading terms; (c) the same on the clock-and-site arena, where it would bear on the record of ~~#360~~; (d) a carrier and a kick for which the response is attained up to a time not small against `1/‖H‖`. | ~~**S–M**~~ done | Medium — the first response of this programme shown to be not zero, so M5's bound is no longer one-sided at distance one; it reaches no farther site, reads no record and passes no gate | ~~#361~~, ~~#359~~, ~~#355~~, ~~#350~~ | > | ~~363~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", milestone M4 — condition (d2) of UO-14, Remainder (b) of ~~#360~~, the author's pick)* **A position record that singles out its site.** ▸ **DONE 2026-11-11 — on the arena of the clock's pointer and a site detector, on a set of pairs `(θ₂, ψ₂)` of positive Haar measure (detector profile `trainF`, a strong enough detector): a carrier with at least `7/8` of its occupancy on a site `x` writes the position record when the detector is at `x`, and is not displayed through the same window when the detector is at any other site. The two placements are two flows, not one run. UO-14 (d2) is met with three conditions that stay; the write half of (d1) is met for carriers with at least `7/8` on the detector's site.** [`RecordLayer/UnifiedSiteRecord.lean`](../CsdLean4/RecordLayer/UnifiedSiteRecord.lean) (20 pins). **The device over a short time** (`H = clockSiteH`, both couplings in it): `norm_clockSiteCarrierU_sub_one_le` (the propagator is within `\|t\| ‖H‖` of the identity), `norm_clockSiteH_le` (`‖H‖ ≤ ‖H_dev‖ + \|g₁ f(θ₂)\| + \|g₂ h(ψ₂)\|`), ★ `abs_clockSiteOccupancy_sub_le`, ★★ `abs_integral_clockSiteOccupancy_sub_mul_le`, ★★ `abs_clockSiteShiftSnd_sub_init_le` (**the site pointer's displacement is within `\|κ\| ‖H‖ t²` of `κ t q`, `q` the carrier's starting occupancy of the detector's site**). **The record fibres:** `siteRecordFibres` (the pairs `(θ₂, ψ₂)` with `κ > 0` and `16 (‖H_dev‖ + \|g₁ f(θ₂)\| + \|g₂ h(ψ₂)\|) < 3 κ` — the same set wherever the detector is), `isOpen_siteRecordFibres`, `siteRecordFibres_bounds`, `mem_siteRecordFibres_trainF`, ★★ `volume_siteRecordFibres_trainF_pos` (**positive measure for the detector profile `trainF`, any continuous clock profile and `16 (‖H_dev‖ + \|g₁ f(φ₀)\|) < 6 g₂`**), ★ `exists_volume_siteRecordFibres_pos` (such a strength exists for every device and clock coupling). **The write and its absence** (on a record fibre, from a ready site pointer, any moving angle `θ₁` of the clock's pointer; its conserved angle `θ₂` is in the fibre condition): `third_le_clockSiteShiftSnd`, `clockSiteShiftSnd_le_sixth`, ★★★ `isRecordWrite_clockSiteFlow_site_of_occupancy` (**occupancy `≥ 7/8` on the detector's site: the readout history is an `IsRecordWrite` of `true` with write interval `[1/(6κ), 1/(2κ)]` and window `1/(6κ)`**), ★★ `clockSiteFlow_site_shown_of_occupancy` (shown through `[1/(2κ), 2/(3κ)]`), ★★★ `clockSiteFlow_site_not_shown_of_occupancy` (**occupancy `≤ 1/8`: not displayed at any time up to `2/(3κ)`**). **The site is singled out:** `arenaObs_deviceSiteProj_le_one_sub` (the occupancies of two sites sum to at most one), `deviceSiteRay`, `arenaObs_deviceSiteProj_deviceSiteRay` (the device on one site, an example of the class), ★★★ `clockSiteFlow_site_shown_iff` (**for a carrier with at least `7/8` on `x` and a detector placed at any site `k`, the readout at the flow time `1/(2κ)` is `true` exactly when `k = x`**). Contract: no new assumption; UD-84 and UD-85 in §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-14 (d) annotated; `POSITS.md` unchanged. **Scope:** one site detector in each run — "a detector at another site" is another arena with the site projector moved, so the comparison is between two flows, and one trial with two detectors of which one writes is not shown (that is `diracTwoPointerFlow`, which has no readout); a fast detector: `16‖H‖ < 3κ` puts the write at `1/(2κ)`, short against `1/‖H‖`, by which the occupancy of the detector's site has changed by at most `3/16`; no moving packet is followed, nothing is said for a slower detector, and nothing is said after `2/(3κ)` for either placement; the carrier is a class of rays (occupancy `≥ 7/8` or `≤ 1/8`), of which the only member exhibited is `deviceSiteRay`, and no particular packet is evolved; carriers with occupancy between `1/8` and `7/8` are not covered; the record fibres have positive measure in `(θ₂, ψ₂)` for the write only — the clock's swing is not treated, its four inequalities stay hypotheses there, and the tick record with them; the record is a definite fact of the one trajectory, not a statistic, and no QM weight is derived; `7/8`, `1/8` and `16` are sufficient, not sharp; every time is the flow parameter (UA-7); the readout and ready arc are apparatus input (UA-4). Remainder: (a) one run with two site detectors, one of which writes and the other does not: a readout on each torus of `diracTwoPointerFlow` and this argument there (UO-10, ~~#361~~ Remainder (c), ~~#360~~ Remainder (e)) → ~~#364~~: `diracTwoPointerFlow_record_fst_only`, for a carrier on one detector's site; the split carrier is that row's Remainder (a); (b) the clock's swing on fibres of positive measure, so that the tick record and this position record hold together off a null set (UO-14 (d1), ~~#360~~ Remainder (a)); (c) a detector not fast against the carrier: a packet followed as it moves, with the write when it arrives (UO-12's question, M6); (d) the carriers between the two thresholds, and sharp constants. | ~~**M**~~ done | Medium–high — the position record of this programme now depends on where the carrier is, on a set of fibres of positive measure, which the instance of ~~#360~~ did not have; the comparison is between two arenas and the detector is fast, so it is not yet one trial with one outcome | ~~#360~~, ~~#359~~, ~~#353~~, ~~#352~~, ~~#350~~ | > | ~~364~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", obligation UO-10 on the arena of two site pointers — Remainder (a) of ~~#363~~, (c) of ~~#361~~, (e) of ~~#360~~; first of the rows the author asked for before M6)* **Two site detectors in one run: the one at the carrier's site writes and the other is not displayed until that write is complete.** ▸ **DONE 2026-11-11 — a readout and a ready arc on each torus of `diracTwoPointerFlow`; on a set of pairs `(θ₂, ψ₂)` of positive Haar measure (two `trainF` detectors of equal strength, strong enough), for a carrier with at least `7/8` of its occupancy on one detector's site, that detector's readout history is a write and the other detector's readout is not `true` up to the completion of that write, along the one orbit. UO-10 is met with conditions (up to the completion of the write) for that class of carriers and stays for a carrier split between the sites.** [`RecordLayer/UnifiedTwoSiteRecords.lean`](../CsdLean4/RecordLayer/UnifiedTwoSiteRecords.lean) (30 pins). **A pointer driven by an occupancy** (any finite index; a continuous family of unitaries with `‖U s − 1‖ ≤ \|s\| N`, an observable with `‖D‖ ≤ 1` and expectation in `[0, 1]`): `abs_arenaObs_arenaKick_sub_self_le`, `integral_arenaObs_mono`, `integral_arenaObs_le`, ★★ `abs_integral_arenaObs_sub_mul_le` (the accumulated expectation is within `N t²` of `t` times its starting value — the third use of this estimate, now stated once), ★★★ `isRecordWrite_of_occupancy` (**`κ > 0`, `16 N ≤ 3 κ`, starting expectation `≥ 7/8`, a ready pointer at `pshift (κ ∫₀ᵗ⟨D⟩)`: the readout history is an `IsRecordWrite` of `true` on `[1/(6κ), 1/(2κ)]` with window `1/(6κ)`**), ★★★ `not_shown_of_occupancy` (starting expectation `≤ 1/8`: not `true` at any time up to `2/(3κ)`). **The readout of the first torus** (apparatus input, UA-4): `echoFstReady`, `measurableSet_echoFstReady`, `echoFstDisplay`, `measurableSet_echoFstDisplay`, `echoFstShown`, `measurable_echoFstShown`, `echoFstShown_eq_true_iff`; the second torus has `echoSiteShown` of ~~#360~~. **The arena:** `diracTwoPointerFlow_fst_pointer`, `diracTwoPointerFlow_snd_pointer` (each pointer is its start turned by its displacement), `diracTwoShiftFst_eq_rateAt`, `diracTwoShiftSnd_eq_rateAt` (rate times the accumulated occupancy of its site), `norm_diracTwoCarrierU_sub_one_le`, `norm_diracTwoH_le` (`‖H‖ ≤ ‖H_D‖ + \|g₁ f(θ₂)\| + \|g₂ h(ψ₂)\|`), `arenaObs_siteProj_le_one_sub` (the occupancies of two sites sum to at most one). **The record fibres:** `twoSiteRecordFibres` (`16 B < 3 κ₁`, `16 B < 3 κ₂`, `3 κ₂ < 4 κ₁`, `3 κ₁ < 4 κ₂`, with `B` the bound on `‖H‖` above), `isOpen_twoSiteRecordFibres`, `twoSiteRecordFibres_bounds`, `zero_mem_twoSiteRecordFibres_trainF`, ★★ `volume_twoSiteRecordFibres_trainF_pos` (positive Haar measure for both profiles `trainF`, equal strengths `g`, `16 ‖H_D‖ < 6 g`), `exists_volume_twoSiteRecordFibres_pos`. **One run:** `echoFstShown_diracTwoPointerFlow_iff`, `echoSiteShown_diracTwoPointerFlow_iff`, ★★★ `diracTwoPointerFlow_record_fst_only` (**detectors at `j ≠ k`, both pointers ready, a record fibre, a carrier with at least `7/8` on `j`: the first readout's history is an `IsRecordWrite` of `true` on `[1/(6κ₁), 1/(2κ₁)]` with window `1/(6κ₁)`, and the second readout is not `true` at any flow time up to `1/(2κ₁)`**), ★★★ `diracTwoPointerFlow_record_snd_only` (the same with the detectors exchanged). Contract: UA-4 extended to this arena (a readout and a ready arc on each torus) and UA-13 updated; no other new assumption; UD-86 to UD-88 in §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-10 annotated; `POSITS.md` unchanged. **Scope:** the other detector is shown not to display up to `1/(2κ)` of the writing one, and nothing is said about it afterwards — the carrier moves on the scale `1/‖H‖` and a later write by the other detector is not excluded; **a carrier split between the two sites is not covered, and exactly one outcome is not shown for it: with half the occupancy on each site both displacements are within the bounds of `κ t / 2`, so this argument does not prevent both detectors, or neither, from displaying — one trial, one outcome is shown only where the outcome is already fixed by where the carrier is**; fast detectors (`16‖H‖ < 3κ` for both, the occupancy of a site changed by at most `3/16` at the write) and no moving packet; the two rates within a factor `4/3` of each other; the only carrier of the class exhibited is the one on the site (`siteRay`); a definite fact of the one trajectory, with no probability stated and no QM weight derived (UO-11 untouched); no signalling statement (UO-13 untouched); no clock on this arena; the cross-talk between the detectors is inside the bound through `‖H‖`; positive measure is Haar measure on the pair of conserved angles, and no `echoMuL` measure of the set of start points is stated; every time is the flow parameter (UA-7). Remainder: (a) a carrier split between the two sites: what the two readouts do, as a theorem — either an exclusion of a double write under a stated condition, or the double write exhibited (UO-10 proper; it bears on gate A's obstruction of ~~#355~~) → ~~#370~~: the double display is exhibited (`diracTwoPointerFlow_double_display`), on fibres of positive Haar measure — an obstruction for this coupling; (b) the other detector after the first write is complete, up to a stated horizon; (c) a detector not fast against the carrier, with a write on arrival (~~#363~~ Remainder (c)) → ~~#365~~: `regionPointerFlow_arrival_write`, on an arena with one pointer coupled to half the ring; (d) the response of ~~#362~~ and the kick bound of ~~#361~~ as statements about these records (UO-15 (e4)); (e) no-signalling between the two readouts (UO-13). | ~~**M**~~ done | Medium–high — the first run of this programme with two detectors in which one writes and the other does not, so the comparison of ~~#363~~ no longer needs two arenas; it is the localised case only, and the split carrier, which is the content of "one outcome", is open | ~~#363~~, ~~#361~~, ~~#360~~, ~~#359~~, ~~#350~~ | > | ~~365~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", towards milestone M6 — a reception record; Remainder (c) of ~~#363~~ and of ~~#364~~; second of the rows the author asked for before M6)* **A record written on arrival: a packet crosses the ring and the far half's pointer writes between `625π/3` and `4375π/3`, an interval about the predicted arrival, and not before.** ▸ **DONE 2026-11-11 — for the crossing packet of ~~#265~~ on a ring of at least `640000` sites, with a pointer of the joint Hamiltonian flow coupled to the projector on the far half: on the arrival fibres (open; positive measure for the profile `trainF` at strength `1/(1250π)`), from a ready pointer, the readout history is an `IsRecordWrite` of `true` with write interval `[625π/3, τ]` for some `τ ≤ 4375π/3`. The predicted arrival is `625π`. A half ring and a coarse time; one pointer, so no radar exchange yet.** [`RecordLayer/UnifiedArrivalRecord.lean`](../CsdLean4/RecordLayer/UnifiedArrivalRecord.lean) (23 pins). **A crossing of the threshold** (no arena): ★★★ `isRecordWrite_of_integral_crossing` (**a ready pointer at `pshift (κ ∫₀ᵗ F)`, `F` continuous in `[0, 1]`, `κ > 0`, displacement `≤ 1/6` at `T₁` and `≥ 1/3` at `T₃`: there is `τ ∈ [T₁, T₃]` with the readout history an `IsRecordWrite` of `true` on `[T₁, τ]` and window `1/(3κ)`** — `τ` by the intermediate value theorem). **A detector on a region:** `regionProj` (the projector on the sites of `A`), `regionProj_isHermitian`, `norm_regionProj_le`, `sum_arenaObs_siteProj`, ★ `arenaObs_regionProj` (its expectation is the sum of the occupancies of its sites), `arenaObs_regionProj_nonneg`, `arenaObs_regionProj_le_one`, `arenaObs_regionProj_mk`, `idxFreeU_diracHam_val` (the uncoupled carrier of the generic layer of ~~#357~~ is `chainProp` of ~~#265~~), ★ `arenaObs_regionProj_free_chainRay` (along it the expectation is the weight of `e^{−isH_D}ψ` on the region over the norm). **The arena** (`Σ_*` of ~~#349~~ with the site projector replaced by a region projector): `regionProj_smul_isHermitian`, `regionPointerFlow` (the Hamiltonian flow of `−2⟨H_D⟩ − 2 g f(θ₂)⟨P_A⟩` for the form of `ℂℙ^{2M+1} × T²`), ★★ `regionPointerFlow_apply` (its closed form), `diracDisplayed_regionPointerFlow_iff`, ★ `abs_integral_regionProj_sub_free_le` (**the detector's term moves the accumulated weight on the region by at most `\|g f(θ₂)\| t²`**). **The arrival:** `arrivalFibres` (the `θ₂` with `κ > 0`, `6κ(625π/12 + \|g f(θ₂)\|(625π/3)²) < 1` and `1 < 3κ(625π/2 − \|g f(θ₂)\|(4375π/3)²)`), `isOpen_arrivalFibres`, `zero_mem_arrivalFibres_trainF`, ★★ `volume_arrivalFibres_trainF_pos`, `crossPacket_ne_zero`, ★★ `arenaObs_farHalf_free_crossPacket`, ★★★ `regionPointerFlow_arrival_write` (**there is a flow time `τ ∈ [625π/3, 4375π/3]` such that the readout history along `regionPointerFlow` from the ray of the crossing packet is an `IsRecordWrite` of `true` from time `0` with write interval `[625π/3, τ]` and window `1/(3κ)`; in particular the readout is not `true` before `625π/3`**). Contract: UA-5 extended (the pointer coupled to a region projector); no other new assumption; UD-89 to UD-91 in a new §4.1d of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-16 annotated (one ingredient, not met); `POSITS.md` unchanged. **Scope:** the detector is on half the ring, so the record says the carrier has reached that half and does not locate it within it; the write is complete at some `τ` in `[625π/3, 4375π/3]`, an interval seven times longer at its end than at its start that contains the predicted arrival `625π`, and `τ` is not computed — no speed is read off the record; before `625π/3` the far half holds up to a quarter of the weight (the packet's tail), and the readout is silent because the accumulated weight is small, not because the far half is empty; one packet and one ring (the numbers of ~~#265~~: `ℓ = 2000π`, `16` modes), no family and no scaling; a weak detector (strength of order `1/(1250π)`, the pointer takes the whole transit to turn by a third), with the back-reaction bounded through `\|g f(θ₂)\| t²` — the arrival fibres are where that bound is small, a neighbourhood of `θ₂ = 0` for `trainF` whose size is not computed; one pointer: no emission record, no second detector and no clock on this arena; a definite fact of the one trajectory, with no QM weight derived; the record is held for `1/(3κ)` and then the pointer turns on; every time is the flow parameter (UA-7); the readout and ready arc are apparatus input (UA-4). Remainder: (a) an emission record and this reception record in one run: two region pointers on one arena (a near and a far one), which is ~~#364~~'s arena with region projectors, and the order of the two writes → ~~#367~~: `regionTwoFlow_departure_arrival` (a departure record, not an emission; the order through the one time `625π/3`); (b) the time `τ` located, and a speed read off two records; (c) a detector on a site or a short segment, with the weight bounds of ~~#265~~ redone for it; (d) a family along which the interval `[625π/3, 4375π/3]` shrinks about the predicted arrival (UO-16's scaling family); (e) the size of the arrival fibres. | ~~**M–L**~~ done | High for M6 — the first reception record of the programme: written by the joint flow within an interval about the predicted arrival and not before `625π/3`, on conserved angles of positive Haar measure (the carrier is one ray; no measure of the set of start points is stated); the threshold reads accumulated weight and does not by itself certify a crossing — a carrier holding a constant weight between `1/7` and `1/2` on the far half would satisfy the same conclusion at `θ₂ = 0` (a calculation, not in Lean); it is one pointer, a half ring and a coarse time, so gate B's question is not yet posed | ~~#265~~, ~~#364~~, ~~#357~~, ~~#349~~, ~~#350~~ | > | ~~366~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", milestone M4 — condition (c) of UO-14, Remainder (c) of ~~#360~~; third of the rows the author asked for before M6)* **The clock's count at a position write.** ▸ **DONE 2026-11-11 — in the exact instance of ~~#360~~ the tick history has a count at the instant the position readout switches, and the count times the tick period is within one period of that instant; the same at the start and the completion of the write, with the write interval in ticks within two periods. UO-14 (c) is met in part in that instance: the flow time at which the count is taken is proved to be the instant the site readout's history turns; the count is still taken at that flow time (c1), in one mode on a null set (c2). It is not a radar coordinate: no signal is emitted or received.** [`RecordLayer/UnifiedCountAtWrite.lean`](../CsdLean4/RecordLayer/UnifiedCountAtWrite.lean) (6 pins). **The count at a time** (any swing of ~~#358~~): `IsSwing.exists_tickCount_abs_sub_le`, ★★ `IsSwing.abs_tickTime_sub_le` (**`\|N · 2π/R − T\| ≤ 2π/R`: the count times the period is within one period of the time**), ★★ `IsSwing.abs_tickTime_sub_sub_le` (for two times, `\|(N₁ − N₀) · 2π/R − (T₁ − T₀)\| ≤ 4π/R`). **At the position write** (one momentum mode, both profiles `trainF`, the fibres `θ₂ = ψ₂ = 0`, `g₂ > 0`, the one orbit of `clockSiteFlow` from `clockSiteTrainStart`): ★★★ `clockSite_count_at_write_train` (**the site readout's history is a write begun at `τ₀ = (M+1)/(12 g₂)` and complete at `τ₁ = (M+1)/(6 g₂)`; the tick history has counts `N₀`, `N₁` at those times; `N₁ · 2π/R` is within one tick period of `τ₁`, and `(N₁ − N₀) · 2π/R` within two of `τ₁ − τ₀`**), ★★ `echoSiteShown_clockSiteFlow_train_iff` (**the site readout is `true` exactly after `t* = (M+1)/(8 g₂)`, up to `3(M+1)/(8 g₂)` — an instant fixed by the record's own history, where the write interval of an `IsRecordWrite` is not**), ★★★ `clockSite_count_at_display_train` (the tick history has a count `N` at `t*`, with `N · 2π/R` within one period of `t*`). Contract: no new assumption; UD-92 in §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-14 (c) annotated; `POSITS.md` unchanged. **Scope:** "the count at the write" is the count of the tick history up to the flow time at which the site readout switches: the two histories are compared at equal values of the flow parameter (UA-7), and nothing in the arena makes that comparison — no register for the count is built on this arena (the statement of ~~#358~~ that no real function of the state holds it is for the one-pointer clock arena and is not restated here) and no register is stamped by the write; the count fixes a flow time up to one period `2π/R` and an interval up to two, and the bound is not improved; the instance is ~~#360~~'s, one momentum mode on the fibres where both profiles vanish (a null set), where the device moves freely and the position record does not single out its site — elsewhere the swing is a hypothesis and the count bound holds for any swing; tick records are erased within their turn, so the count at `t*` is a property of the history on `[0, t*]` and not of the state at `t*`; `R` is the level splitting of the mode, and its relation to the rest splitting (the dilation clause of ~~#357~~) is not used; every time is the flow parameter. Remainder: (a) a register stamped by the write: a degree of freedom whose state after the position write holds the count, so that the comparison is made inside the arena (UO-14 (b), with this row's first condition); (b) the count at the write on the record fibres of ~~#363~~, which needs the swing there (~~#363~~ Remainder (b)); (c) the count at an arrival: the reception record of ~~#365~~ with a clock on its arena. | ~~**S**~~ done | Medium — the reading of a clock at an event given by a record, which is what a radar coordinate is made of; it is read by comparing two histories at a flow time, in one instance on a null set of fibres | ~~#360~~, ~~#358~~, ~~#350~~ | > | ~~367~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", towards milestone M6 — Remainder (a) of ~~#365~~; the part of the three-pointer step that needs no new arena)* **A departure record and an arrival record in one run.** ▸ **DONE 2026-11-11 — two pointers of one joint flow, one coupled to the near half of the ring and one to the far half; for the crossing packet of ~~#265~~, on fibres of positive measure, the near readout writes by `625π/3` and the far readout does not show before `625π/3` and writes by `4375π/3`, along the one orbit. Departure, not emission; no clock, so no time is read at either record.** [`RecordLayer/UnifiedDepartureArrival.lean`](../CsdLean4/RecordLayer/UnifiedDepartureArrival.lean) (16 pins). **Two families of unitaries:** ★ `abs_integral_arenaObs_sub_integral_le` (within `\|s\| N` of each other: accumulated expectations differ by at most `N t²`), `norm_idxTwoCarrierU_sub_idxFreeU_le`. **The arena** (the two-pointer arena of ~~#359~~ with region projectors): `regionTwoFlow` (the flow of `−2⟨H_D⟩ − 2 g₁ f(θ₂)⟨P_A⟩ − 2 g₂ h(ψ₂)⟨P_B⟩`), ★★ `regionTwoFlow_apply`, `echoFstShown_regionTwoFlow_iff`, `echoSiteShown_regionTwoFlow_iff`, ★ `abs_integral_regionProj_two_sub_free_le` (**both detectors together move an accumulated weight by at most `(\|g₁ f(θ₂)\| + \|g₂ h(ψ₂)\|) t²`**). **The run:** `departureArrivalFibres` (with `b` that back-reaction: `1 < 3κ₁(625π/4 − b(625π/3)²)`, `6κ₂(625π/12 + b(625π/3)²) < 1`, `1 < 3κ₂(625π/2 − b(4375π/3)²)`), `isOpen_departureArrivalFibres`, `zero_mem_departureArrivalFibres_trainF`, ★★ `volume_departureArrivalFibres_trainF_pos` (both profiles `trainF`, strengths `1/(625π)` and `1/(1250π)`), `continuous_regionWeight_crossPacket`, `integral_regionWeight_crossPacket_le`, `le_integral_regionWeight_crossPacket`, ★★ `arenaObs_nearHalf_free_crossPacket` (along the free chain the near half holds one minus the far half's weight), ★★★ `regionTwoFlow_departure_arrival` (**from two ready pointers there are flow times `τ_d ≤ 625π/3 ≤ τ_a ≤ 4375π/3` such that the near readout's history is an `IsRecordWrite` of `true` with write interval `[0, τ_d]` and the far readout's history one with write interval `[625π/3, τ_a]`, along the one orbit of `regionTwoFlow`**). Contract: UA-13 extended to region projectors; no other new assumption; UD-93 and UD-94 in §4.1d of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-16 annotated (two ingredients, not met); `POSITS.md` unchanged. **Scope:** departure and not emission — the packet is prepared on the near half and the near pointer records that accumulated weight on that half crossed the threshold; nothing emits it; with write interval `[0, τ_d]` the not-displayed-before clause of `IsRecordWrite` is empty, so the statement is that the near readout is `true` through its window for some `τ_d ≤ 625π/3`; the order is through one time: the departure write is complete by `625π/3` and the arrival readout does not show before `625π/3`, and the two could meet there; `τ_d` and `τ_a` come from the intermediate value theorem and are not computed, so no transit time and no speed are read off the two records; each detector is on half the ring; one packet, one ring, weak detectors, with the back-reaction of both inside the statement and the fibres where it is small — a neighbourhood of `(0, 0)` of positive Haar measure whose size is not computed; no clock on the arena, so no time is read by a clock at either record; definite facts of the one trajectory, no QM weight derived; every time is the flow parameter (UA-7). Remainder: (a) a clock in the same run: a third pointer, which needs an arena with three pointer tori (M6's prerequisite; the two-torus arena of ~~#202~~ built again one level up, or a tower of tori — the author's choice of route) → ~~#368~~: the arena exists (`hamiltonianFlow_stepEnergy`, `threeCoupledFlow_closed`), by the tower route; a clock and two detectors read on it are that row's Remainder (a); (b) the two times located and a transit time read off; (c) detectors on short segments; (d) an emission: a carrier released by an apparatus, with a record of the release. | ~~**M**~~ done | High for M6 — a departure record and an arrival record of one prepared packet in one run, ordered only through the one time `625π/3`; the thresholds read accumulated weight and do not certify a transit; without a clock there is no time at either, so no radar coordinate | ~~#365~~, ~~#364~~, ~~#359~~, ~~#265~~, ~~#350~~ | > | ~~368~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", the arena M6 needs — Remainder (a) of ~~#367~~; the author's choice of route B, "a tower of pointer tori", over a third arena built by hand)* **Adding a pointer torus to a symplectic arena, and the three-torus arena `ℂℙᴺ × T² × T² × T²`.** ▸ **DONE 2026-11-11 — one theorem adds a pointer torus to any compact symplectic arena and gives its flow as a skew product over the base's flows; the three-torus arena is that theorem applied to the two-torus arena of ~~#202~~, with its flow in closed form and its preserved measure. No record is read on it here.** [`LF4/PointerStep.lean`](../CsdLean4/LF4/PointerStep.lean) (16 pins) and [`LF4/TriCoupledFlow.lean`](../CsdLean4/LF4/TriCoupledFlow.lean) (19 pins). **The step** (any `C^∞` manifold `S` modelled on a finite-dimensional space, with a symplectic form `β`): `stepForm` (`β ⊕ dψ₁ ∧ dψ₂` on `S × T²`), ★ `stepForm_isSymplectic`, `stepEnergy` (`H(s) + G(s) g(ψ₂)`), `stepEnergy_eq`, `contMDiff_stepTerm`, `contMDiff_stepEnergy`, `contMDiff_add_const_smul`, ★ `isHamiltonianVectorField_stepLift`, ★ `isHamiltonianVectorField_stepTerm`, ★★ `isHamiltonianVectorField_step` (the field `(X_H + g(ψ₂) X_G, (G g'(ψ₂), 0))` is Hamiltonian for the step energy), ★★★ `hamiltonianFlow_stepEnergy` (**for compact `S` the flow is `(s, ψ) ↦ (φ_t^{H + g(ψ₂) G}(s), (ψ₁ + g'(ψ₂) ∫₀ᵗ G(φ_u(s)) du, ψ₂))`: a skew product over the conserved angle, with the base moved by the base's own flow of `H + g(ψ₂) G`**), `hamiltonianFlow_congr`, ★★ `hamiltonianFlow_stepEnergy_fst` (the slice), ★★★ `map_hamiltonianFlow_stepEnergy` (**if every base flow of `H + c G` preserves `μ`, the flow preserves `μ ⊗ vol_{T²}`**), `echoForm_eq_stepForm`, `twoCoupledEnergy_eq_stepEnergy` (the two-torus arena of ~~#202~~ is one step over the one-torus arena). **Three tori:** `TriModel`, `TriSigma`, `triForm`, ★ `triForm_isSymplectic`, `echoRayEnergy`, `contMDiff_echoRayEnergy`, `threeCoupledEnergy` (`−2⟨A⟩ − 2⟨B_A⟩ f(θ₂) − 2⟨B_B⟩ g(ψ₂) − 2⟨B_C⟩ k(χ₂)`), `threeCoupledEnergy_apply`, `contMDiff_threeCoupledEnergy`, ★ `twoCoupledEnergy_add_smul_echoRayEnergy` (adding `c` times `−2⟨B_C⟩` to the two-torus energy is replacing `A` by `A + c B_C`), `threeCoupledFlow`, `continuous_threeCoupledFlow`, `measurable_threeCoupledFlow`, `threeCoupledFlow_zero`, `threeCoupledFlow_add`, ★★★ `threeCoupledFlow_apply` (**with `χ₂` held, the ray and the first two tori move by the two-torus flow with uncoupled matrix `A + k(χ₂) B_C`, and `χ₁` advances by `k'(χ₂)` times the accumulated `−2⟨B_C⟩`**), ★★★ `threeCoupledFlow_closed` (the ray by the Schrödinger unitary of `A + f(θ₂) B_A + g(ψ₂) B_B + k(χ₂) B_C`, each first angle by its profile's derivative times the accumulated `−2⟨B⟩` of its operator along that one orbit, the three second angles fixed), `triMuL`, ★★★ `triMuL_map_threeCoupledFlow` (**the flow preserves `μ_FS ⊗ vol_{T²} ⊗ vol_{T²} ⊗ vol_{T²}`**). Contract: one new assumption, UA-15 (the three-pointer arena); UD-95 and UD-96 in a new §4.1e of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-16 annotated (the arena exists; nothing read on it); `POSITS.md` unchanged. **Scope:** the step adds one torus, and `n` tori are `n` applications, each with its own statement — no arena with a variable number of tori is defined, so "tower" means the step theorem and not a family of types; the base must be compact; the coupling's form is assumed (a function of the base times a function of the new torus's second angle; the moving angle does not enter the energy, so the second angle is conserved and no pointer acts back on its own rate); there is no term between the tori; the measure statements are read off the closed form and are conditional on the base flows preserving `μ` — the top-form measure of the product form is not identified with the product measure; no record, readout or carrier is in these two modules. Remainder: (a) the three-pointer layer in a carrier's coordinates, and a clock's pointer with two detectors' pointers under one flow (the next row) → ~~#369~~: `clockTwoSiteFlow`, `clockTwoSite_records_train` (a swing, two readout switching instants and two counts in one run, with no travelling signal); (b) the one-torus arena as a step over the ray alone, so that all three levels are instances of the one theorem (`hamiltonianFlow_coupledEnergy` is still its own proof); (c) an arena with a variable number of tori, if a statement needs to quantify over it; (d) the top-form measure of `stepForm β` as the product measure (~~#359~~ Remainder (f)). | ~~**L**~~ done | Medium–high for M6 — an arena with one clock pointer and two detector pointers under one flow now exists, a prerequisite for M6, and each further pointer is one application of a theorem instead of a new arena built by hand; UO-16's second observer needs further pointers and a clock that is not the carrier's own level; it carries no record yet | ~~#367~~, ~~#359~~, ~~#202~~, ~~#350~~ | > | ~~369~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", towards milestone M6 — Remainder (a) of ~~#368~~)* **A clock and two site detectors under one flow: a swing and two readout switching instants in one run, and the clock's count at each.** ▸ **DONE 2026-11-11 — on the three-torus arena of ~~#368~~, the device with the clock's pointer and two site pointers: in one exact instance (one momentum mode, the fibres where all three profiles vanish: a null set) the clock ticks, each detector's readout switches at an explicit flow time, and the interval between the two detector records read in ticks is within two tick periods of its flow value. No signal travels: the interval is set by the detectors' strengths, not by their sites.** [`RecordLayer/UnifiedClockTwoSites.lean`](../CsdLean4/RecordLayer/UnifiedClockTwoSites.lean) (22 pins). **The third torus's readout** (apparatus input, UA-4): `triThdReady`, `measurableSet_triThdReady`, `triThdDisplay`, `measurableSet_triThdDisplay`, `triThdShown`, `measurable_triThdShown`, `triThdShown_eq_true_iff`. **The slice of the three-torus flow:** ★★ `threeCoupledFlow_fst_of_lift_eq_zero` (on a fibre where the third profile vanishes the first two pointers move as on the two-pointer arena), `threeCoupledFlow_snd_of_lift_eq_zero`. **The arena:** `clockTwoSiteFlow` (the flow of `−2⟨H_dev⟩ − 2 g₁ f(θ₂)⟨X⟩ − 2 g₂ h(ψ₂)⟨Q_{k₂}⟩ − 2 g₃ k(χ₂)⟨Q_{k₃}⟩` on `ℂℙ^{4M+3} × T² × T² × T²`), `clockTwoSiteFlow_zero`, `clockTwoSiteFlow_add`, ★★ `triMuL_map_clockTwoSiteFlow`, ★★ `clockTwoSiteFlow_fst_of_lift_eq_zero` (**on a fibre with `k(χ₂) = 0` the device, the clock's pointer and the first detector move exactly by `clockSiteFlow` of ~~#359~~, so statements of ~~#360~~ and ~~#366~~ about that first component can be transported; here the swing and the readout iff are, and the `ActualOccurrence`s are not rebuilt**), ★★ `clockTwoSiteFlow_thd_pointer_of_lift_eq_zero` (the third pointer is turned by its rate times the accumulated occupancy of its site along that orbit). **The run** (`clockTwoSiteTrainStart`: the start of ~~#360~~ with a third pointer at `1/12`; all three profiles `trainF`; `g₂, g₃ > 0`; `μ₀ < μ₁ < 2μ₀`, `R` the mode's level splitting and the clock strength `g₁` fixed as in ~~#360~~): `pshift_twelfth_mem_displayArc_iff`, `clockTwoSiteFlow_train_fst`, ★★ `isSwing_clockTwoSiteFlow_train` (the clock's pointer is a swing at the level splitting on every horizon), ★★ `echoSiteShown_clockTwoSiteFlow_train_iff` (the first detector's readout is `true` exactly after `t₂ = (M+1)/(8 g₂)`, on `0 ≤ t ≤ 3 t₂`), ★★ `triThdShown_clockTwoSiteFlow_train_iff` (the second detector's exactly after `t₃ = (M+1)/(8 g₃)`, on `0 ≤ t ≤ 3 t₃`; nothing is stated later, and the two windows need not overlap), ★★★ `clockTwoSite_records_train` (**along the one orbit: both switching instants; counts `N₂`, `N₃` of the tick history at them, each within one tick period `2π/R` of its time; and `\|(N₃ − N₂) · 2π/R − (t₃ − t₂)\| ≤ 4π/R`**). Contract: UA-4 extended to the third torus; no other new assumption; UD-97 and UD-98 in §4.1e of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-16 annotated (three records and two counts in one run, with no signal; not met); `POSITS.md` unchanged. **Scope:** "record" here means a readout's switching instant (`false` on `[0, t_i]`, `true` on `(t_i, 3 t_i]`): no `IsRecordWrite` or `ActualOccurrence` with propagator `clockTwoSiteFlow` is constructed, no lemma places the start in `triThdReady`, and the clock's ticks are records by `IsSwing.isRecordWrite` of ~~#358~~ applied to this swing, which is not instantiated here; the only clock is the carrier's own internal level, so if the carrier travelled its count would be the messenger's and not an observer's; no signal travels — the carrier is one momentum mode, spread evenly over the ring, each detector's pointer turns at its own rate, and the two records are apart in time because `g₂ ≠ g₃`: the interval `(M+1)(1/g₃ − 1/g₂)/8` is not a transit time, and no distance or speed is read off it; the instance is on the fibres where all three profiles vanish, a null set, where the device moves freely — off them only the reduction on the zeros of the third profile is proved, with no cross-talk bound for three pointers; the counts are counts of the tick history up to flow times, held by no degree of freedom; the resolution is one tick; definite facts of the one trajectory, no QM weight derived; every time is the flow parameter (UA-7). Remainder: (a) a travelling packet in this arena: the two-level device crossing the ring, which needs the crossing bounds of ~~#265~~ for both rest energies (the two levels move at different speeds) and region projectors on the device — then the records of ~~#367~~ with the clock's count at each (M6's one-way signal); (b) the clock's swing for a packet in place of one mode, over a transit time (UO-14 (a)): the error of ~~#357~~ grows with the horizon and is not known to stay below a tick over the crossing of ~~#265~~; (c) the three-pointer layer for any Hermitian quadruple in a carrier's coordinates, with cross-talk bounds; (d) a return leg: a reflection or a second packet, for a two-way radar time; (e) a second observer. | ~~**M**~~ done | Medium for M6 — a clock's swing, two detector readouts and the clock's count at each switching instant are in one run of one flow. Not a radar reading: the carrier does not travel, the clock is the carrier's own level and not an observer's, the counts are aligned by the flow parameter, and there is no return leg or second observer | ~~#368~~, ~~#366~~, ~~#360~~, ~~#359~~, ~~#350~~ | > | ~~370~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", obligation UO-10 — Remainder (a) of ~~#364~~, the author's pick)* **A carrier split between two detectors: both display.** ▸ **DONE 2026-11-11 — on the arena of two site pointers, on a set of pairs `(θ₂, ψ₂)` of positive Haar measure (two `trainF` detectors of equal strength, strong enough), for a carrier with at least `3/8` of its occupancy on each detector's site: both readouts are `true` at the flow time `1/κ₁`, along the one orbit, neither having been `true` early. UO-10 is answered for this coupling as an obstruction: one run does not have one outcome.** [`RecordLayer/UnifiedSplitCarrier.lean`](../CsdLean4/RecordLayer/UnifiedSplitCarrier.lean) (14 pins). **A pointer driven by a half-occupied site** (the setting of ~~#364~~, any finite index): ★★★ `shown_of_half_occupancy` (**`κ > 0`, `96 N ≤ κ`, starting expectation in `[3/8, 5/8]`, a ready pointer: the readout is `true` at every time in `[23/(25κ), 26/(25κ)]`**), ★ `not_shown_of_le_sixth` (expectation in `[0, 1]`: not `true` up to `1/(6κ)`, whatever the carrier). **The split fibres:** `splitFibres` (`96 B < κ₁`, `96 B < κ₂`, `23 κ₁ < 25 κ₂`, `25 κ₂ < 26 κ₁`, with `B = ‖H_D‖ + \|g₁ f(θ₂)\| + \|g₂ h(ψ₂)\|`), `isOpen_splitFibres`, `splitFibres_bounds`, `zero_mem_splitFibres_trainF`, ★★ `volume_splitFibres_trainF_pos` (both profiles `trainF`, equal strengths `g`, `96 ‖H_D‖ < 2 g`). **Both display:** ★★★ `diracTwoPointerFlow_double_display` (**detectors at `j ≠ k`, both pointers ready, a split fibre, a carrier with at least `3/8` on each site: at the flow time `1/κ₁` the first readout is `true` and the second readout is `true`**), ★★ `diracTwoPointerFlow_split_not_shown_early` (neither is `true` up to `1/(6κ)` of its own rate, so both switch on), `split_ne`, `splitRay` (the ray of `e_(k,0) + e_(j,0)`), `arenaObs_siteProj_splitRay_fst`, `arenaObs_siteProj_splitRay_snd` (occupancy `1/2` on each site), ★★★ `diracTwoPointerFlow_splitRay_double_display` (**the equal superposition over the two detectors' sites makes both display**). Contract: no new assumption; UD-99 and UD-100 in §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-10 annotated as answered by an obstruction for the coupling of UA-5; `POSITS.md` unchanged. **Scope:** a statement about this coupling — each pointer is driven by the occupancy of its site (UA-5), so a carrier with half its occupancy on each site turns both pointers at half rate and both cross the threshold; in this regime the flow does not select one site; it is an obstruction of the same kind as gate A's (~~#355~~): the pointer reads an occupancy, not an outcome; it is not a prediction about detectors — the weights of quantum mechanics for one click or the other are not derived here and not contradicted here, and in the contract they stay conditional on the cell law (UA-6, Posit 1), which this arena does not use; a selection of one outcome would need a mechanism that is not in this joint energy; fast detectors (`96‖H‖ < κ` for both) with `κ₂` between `0.92 κ₁` and `1.04 κ₁`, so that one flow time serves both; a class of carriers (at least `3/8` on each of the two sites), with the equal superposition exhibited, and nothing for occupancy between `1/8` and `3/8` on one site; both readouts shown `true` at one flow time and each through an interval of its own — no `IsRecordWrite` with a window is built for the split carrier; positive measure is Haar on the pair of conserved angles, and no `echoMuL` measure of the set of start points is stated; definite facts of the one trajectory; every time is the flow parameter (UA-7). Remainder: (a) what would select one outcome: the cell law of UA-6 on this arena (a rate field and outcome cells for two detectors), which is Posit 1 and not this coupling — conditional statistics as in ~~#356~~, or the author's decision that the obstruction stands for drift pointers → ~~#371~~: not outcome cells for two detectors but one fibre angle partitioned over all sites, on `Σ_*`: at most one site at every arena point, the weights conditional on UA-6, and a no-go for families local to their sites (`not_siteLocal_of_one_outcome`); kinematic; (b) the carriers between the two classes, and sharp constants; (c) the same for slow detectors, where the carrier moves between the sites during the write; (d) the double display as two `IsRecordWrite`s with windows. | ~~**M**~~ done | High — on this two-pointer arena, for fast detectors with close rates, a split carrier makes both readouts `true` at one flow time (no `IsRecordWrite`), so outcome selection is not supplied by this coupling in that regime and stays with the cell law; a result of the obstruction kind, like gate A's | ~~#364~~, ~~#355~~, ~~#359~~, ~~#350~~ | > | ~~371~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", obligation UO-10 — Remainder (a) of ~~#370~~, the author's pick, at the author's question "given CSD, is this right? they're linked via locality / non-locality?")* **Site cells: one outcome for a split carrier, and why it cannot be site-local.** ▸ **DONE 2026-11-11 — if the outcome is read from the cell of the one ontic point, at most one site is selected for every arena point and every carrier, and with the weights assumed (UA-6) the split carrier selects each of its two sites with probability `1/2` and never both; and on three sites no family of readouts that depend on the carrier only through the occupancy of their own sites can have at most one site almost surely and the occupancies as weights for all carriers. Kinematic: no flow writes a cell. On `Σ_*` with one fibre angle, not on the two-pointer arena.** [`RecordLayer/UnifiedCellSelection.lean`](../CsdLean4/RecordLayer/UnifiedCellSelection.lean) (10 pins). **The site cells:** `siteCell` (the arena points whose fibre angle `θ₁` lies in the arc a rate field `c` gives, at the point's own base point, to one of the two basis states of the site — a union of two basins of `GlobalBasin.lean`, with no preparation in it), `measurableSet_siteCell`, ★★★ `siteCell_disjoint` (**the cells of two different sites are disjoint: no measure, no hypothesis on `c`**), `globalBasin_subset_iUnion_siteCell`, ★★ `epistemicMeasure_siteCell` (**for a torus-generated `c` the probability of the cell of `j` under the preparation law at the carrier `p` is the occupancy of `j`**), ★ `epistemicMeasure_iUnion_siteCell` (some site is selected with probability `1`), ★★★ `epistemicMeasure_siteCell_splitRay` (**the split carrier: `1/2`, `1/2`, and the two cells share no arena point**). **The no-go:** ★ `measure_add_add_le_one` (three pairwise almost disjoint events have total probability at most `1`), ★★★ `not_siteLocal_of_one_outcome` (**three different sites, any probability law `μ` on a fibre, readout regions `S j p` with `μ (S j p ∩ S k p) = 0` for `j ≠ k` and `μ (S j p) =` the occupancy of `j`: then `S j p` is not a function of the occupancy of `j` alone** — the three carriers split over the three pairs of sites give three pairwise almost disjoint regions of probability `1/2`), ★★★ `siteCell_not_siteLocal` (**the site cells of a torus-generated field are not all local to their sites**: for at least one site the fibre region is not a function of its own occupancy; for the cells as ordered here the regions of the first and the last site are, a calculation not in Lean). Contract: UA-6 extended (a site's cell is the union of the two basins of the (site, component) context, the fine basis and the order of the arcs being apparatus input) and no other new assumption; UD-101 and UD-102 in §4.1c of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-10 annotated; `POSITS.md` unchanged. **Scope:** kinematic — a site cell is a region of the arena, no flow of this corpus writes it, no pointer displays it and no `IsRecordWrite` is built, so what is proved is what a selection would have to be and not that a dynamics produces it; the weights are assumed (`IsTorusGenerated c`, UA-6, Posit 1) and the disjointness is not conditional; "local to its site" means the region of `j` depends on the carrier only through the occupancy of `j`, with the fibre point shared by all sites and its law independent of the carrier (UA-2) — a model whose hidden variable is a position with the carrier's weights as its law is outside the hypothesis, is local and gives one outcome, so the theorem is about `Σ_*`, where the ontic state is the ray and a fibre point; exact equality of regions is asked, not equality up to null sets; three sites are needed and with two the theorem says nothing; one carrier and no settings, so it is not Bell's theorem; the order of the arcs in a cell is the order of the index, a convention, and the theorem says some dependence on the carrier beyond the site's own occupancy is unavoidable, not which; which cell holds the point is a definite fact, and the measures are the preparation law over the unknown fibre point. Beside ~~#370~~: pointers each coupled through their own site projector both display, on the two-pointer arena; cells give at most one site and are not all local, on `Σ_*`; the readouts of ~~#370~~ are not shown to be local in the sense of the no-go. Remainder: (a) a dynamics that writes the cell: a pointer whose display follows the cell of the ontic point (the discharge of UA-6; ~~#355~~ shows the coupling of UA-5 does not supply it); (b) the no-go up to null sets, and for two sites the local family that exists; (c) the site cells along the coupled flow, as ~~#356~~ did for one basin; (d) the two-wing case with settings is ~~#288~~/~~#290~~. | ~~**M**~~ done | High — beside ~~#370~~ it states the conditional alternative for this carrier: a family of readouts local in the stated sense either selects two sites with positive probability or misses the weights, and the selection that gives at most one site is a fact of the one arena point (ray and fibre angle), not a function of the site's own occupancy. Kinematic; no claim about locality in `Σ` or in space (row 39 stays a conjecture) | ~~#370~~, ~~#364~~, ~~#356~~, ~~#355~~, ~~#288~~ | > | ~~372~~ | Records | Records→spacetime | *new (2026-11-11, the author's programme "From `Σ` to physical spacetime", milestone M6, first brick — after the author's decision under UO-16 that the clock is a second carrier and is the tick clock of ~~#358~~)* **A clock on its own carrier beside a messenger: one flow, two records, two counts.** ▸ **DONE 2026-11-11 — on the product of the messenger's arena (two region pointers, ~~#367~~) and the clock device's arena (tick pointer, ~~#357~~/~~#358~~), with the sum of the two energies: the Hamiltonian flow is the pair of the two flows and preserves the product measure; one orbit has a departure write, an arrival write and the clock's tick count at a flow time by which each is complete, and the count difference times the clock's period is within two periods of the difference of those times. No coupling between the clock and the messenger; the counts are aligned by the flow parameter; the two times are not located and the bound holds for any two flow times of the clock's swing.** Generic half [`LF4/ProductFlow.lean`](../CsdLean4/LF4/ProductFlow.lean) (6 pins): `prodEnergy`, `prodEnergy_eq`, `contMDiff_prodEnergy`, ★ `isHamiltonianVectorField_prodEnergy`, ★★★ `hamiltonianFlow_prodEnergy` (**for two compact symplectic arenas the flow of `H(a) + K(b)` is `(Φ_H^t a, Φ_K^t b)`**), ★★ `measurePreserving_hamiltonianFlow_prodEnergy`. Record half [`RecordLayer/UnifiedClockMessenger.lean`](../CsdLean4/RecordLayer/UnifiedClockMessenger.lean) (9 pins): `IsSwing.prodMk_snd`, `clockMessengerEnergy`, `contMDiff_clockMessengerEnergy`, ★ `clockMessengerFlow`, ★★★ `clockMessengerFlow_apply` (**the joint flow is `regionTwoFlow` on the first factor and `clockPointerFlow` on the second**), ★ `clockMessengerFlow_fst`, ★ `clockMessengerFlow_snd` (neither factor feels the other), ★★ `map_clockMessengerFlow` (`echoMuL ⊗ kMuL` is preserved), ★★★ `clockMessenger_departure_arrival_counts` (**from the messenger's start of ~~#367~~ and the clock's train start of ~~#358~~: flow times `τ_d ≤ 625π/3 ≤ τ_a ≤ 4375π/3`, the departure and arrival readouts, read on the messenger's factor along the joint orbit, each an `IsRecordWrite`, tick counts `N_d`, `N_a` of the clock at the two times, and `\|(N_a − N_d)(2π/R) − (τ_a − τ_d)\| ≤ 4π/R`**). Contract: new **UA-16** (the clock on its own carrier: the product arena, the product measure, the additive energy with no term between the factors, two rings with no distance between them); UD-103 and UD-104; UO-16 annotated with the decision and with what is still missing; `POSITS.md` unchanged. **Scope:** no coupling — the clock's history does not depend on the messenger's state and the reverse, and nothing places the clock at the departure region (not a statement of this arena); the counts are aligned with the writes by the flow parameter (UA-7), no degree of freedom holds a record with a count and no record reads the count; the two writes are those of ~~#367~~ and do not certify a transit, the departure's "before" clause being empty; the clock is the exact instance of ~~#358~~ (one momentum mode, `trainF`, the fibre `θ₂ = 0`, one start point), so no measure of clock starts is stated; the bound is two ticks against a difference of flow times, it holds for any two flow times of the clock's swing, the two times are existential and not located (`τ_a − τ_d ≥ 0` is all that is stated), no hypothesis relates the tick period to the interval so both counts may be zero, and the ticks are counted (`IsTickCount`) but not instantiated as `IsRecordWrite`s on this arena; no distance, no speed and no second observer, and nothing here is evidence about Lorentz kinematics; the measure statement is invariance only; definite facts of the one trajectory. Remainder: (a) couple the clock to the departure detector (a term between the factors, as ~~#212~~/~~#225~~ did on the two-site arenas), so that a count belongs to a record otherwise than through the flow parameter → ~~#374~~: not built — on this arena the flow commutes with turning the pointers, so with the carriers and conserved angles fixed nothing but a pointer's own angle depends on where it started (`clockMessengerFlow_shiftPointers`); two ways on are UO-24 or a product observable on a composite carrier (the author chose the second, 2026-11-12); (b) locate the two write times and certify a transit (the Remainder of ~~#365~~ and ~~#367~~) → ~~#373~~: a transit in the accumulated occupancies that drive the pointers (`cross_transit_certificate`), which no constant-occupancy carrier satisfies; the write times are not located and no record is built; (c) a return leg to the same clock: a radar time; (d) a second observer; (e) clock starts of positive measure (the swing for a packet and off the fibre `θ₂ = 0`, UO-14 (a)); (f) a register that holds the count at a record (UO-14 (b)). | ~~**M**~~ done | Medium for M6 — on the unified arenas (pointers moved by one joint Hamiltonian flow, no stated detector law) the first run with a departure write, an arrival write and tick counts of a device that is not the messenger; ~~#271~~ had this with located ticks under stated detector laws. The clock is no longer the messenger's level, by the choice of arena (UA-16). Not a reading of an interval: the two write times are existential and not located, the two-tick bound holds for any two flow times of the clock's swing, and nothing relates the tick period to the interval. It fixes the arena for the rest of M6 | ~~#367~~, ~~#358~~, ~~#366~~, ~~#368~~, ~~#225~~ | > | ~~373~~ | Records | Records→spacetime | *new (2026-11-12, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — the "certify a transit" half of Remainder (b) of ~~#372~~, the author's pick)* **A transit in the accumulated occupancies that drive the two pointers.** ▸ **DONE 2026-11-12 — on the flow and start of ~~#367~~, on a set of conserved angles of positive Haar measure (both profiles `trainF`, any positive strengths; a neighbourhood of `(0, 0)` whose size is not computed): the far half's accumulated occupancy is at most a third of `[0, 625π/3]` and gains at least two thirds of `[3125π/3, 4375π/3]`; no carrier with a constant occupancy of the far half satisfies those two inequalities. Four inequalities on accumulated occupancies at flow times: not a record, no pointer inequality, and the write times are not located.** [`RecordLayer/UnifiedTransitCertificate.lean`](../CsdLean4/RecordLayer/UnifiedTransitCertificate.lean) (13 pins). **A carrier with a constant occupancy:** `integral_arenaObs_of_const` (a constant expectation `w` accumulates `w t`), ★★ `not_transit_of_linear` (**`F t = w t` cannot have `F T₁ ≤ (1/3) T₁` and `(2/3)(T₃ − T₂) ≤ F T₃ − F T₂`**). **The transit fibres:** `transitFibres` (both rates positive and `92500 π (\|g₁ f(θ₂)\| + \|g₂ h(ψ₂)\|) < 1`; no condition relating a rate to the transit time; the two rate conditions are used by no theorem of the module and are there so that a displacement over its rate is the accumulated occupancy), `isOpen_transitFibres`, `zero_mem_transitFibres_trainF`, ★★ `volume_transitFibres_trainF_pos`. **The accumulated occupancies:** `crossAccum` (`∫₀ᵗ⟨P_C⟩` along the carrier of `regionTwoFlow` started on the crossing packet), ★★ `regionTwoFlow_cross_pointers` (**each pointer's moving angle is its start plus its rate times the accumulated occupancy of its half**, an identity of angles), ★ `le_integral_regionWeight_crossPacket_late`, ★ `abs_crossAccum_far_sub_le`, ★ `abs_crossAccum_near_sub_le` (back-reaction at most `(\|g₁ f\| + \|g₂ h\|) t²`). **The certificate:** ★★★ `cross_transit_certificate` (**with `T₁ = 625π/3`, `T₂ = 3125π/3`, `T₃ = 4375π/3`: far at most `(1/3) T₁` and near at least `(2/3) T₁` over `[0, T₁]`; far at least `(2/3)(T₃ − T₂)` and near at most `(1/3)(T₃ − T₂)` over `[T₂, T₃]`**), ★★★ `cross_far_occupancy_not_constant` (**the far half's occupancy along this run is not constant**). Contract: no new assumption; UD-105 in §4.1d of [`unified-record-spacetime-plan.md`](unified-record-spacetime-plan.md); UO-16 annotated; `POSITS.md` unchanged. **Scope:** what is new is the form of the statement — the estimates are those of ~~#367~~ and `dirac_arrival`, integrated over a second window, and an upper bound early with a lower bound on a late increment for the same half is what no constant meets, where the thresholds of ~~#367~~ were one-sided; accumulated occupancies, not readouts — the certificate has no pointer, rate or angle in it, the pointers enter through an identity of angles and so carry the accumulated occupancies only modulo one turn, and for the strengths of ~~#367~~ the proved bounds allow the far pointer up to `7/3` turns by `T₃` (at least `1/2`); no readout region, no threshold and no `IsRecordWrite`, so it is not a record and certifies nothing to an observer who has only the records of ~~#367~~; the near half's two inequalities restate the far half's, since the two occupancies add to one; four values of the flow parameter (UA-7), which no record holds; what is excluded is a carrier whose occupancy of the far half is constant in time (`integral_arenaObs_of_const` with `not_transit_of_linear`), and a carrier that moves in some other way and gives the same numbers is not excluded — "transit" means only that the far half's mean occupancy is at most `1/3` over `[0, T₁]` and at least `2/3` over `[T₂, T₃]`, with no direction, path or speed; the objection it answers is that a carrier whose far-half weight is a constant between `1/7` and `1/4` would satisfy the conclusions of ~~#365~~ and ~~#367~~ (a calculation, not in Lean); the change is located only within `[0, T₃]`: the four numbers give a time in `[0, T₁]` with far occupancy at most `1/3` and a time in `[T₂, T₃]` with at least `2/3`, while the location to `[T₁, T₂]`, a factor of five, belongs to the free packet's pointwise bounds and is not proved here for the coupled carrier; the write times of ~~#365~~ and ~~#367~~ are not located; the fibre condition is a different one from that of ~~#367~~ and no theorem states the records and the certificate on a common fibre; back-reaction is bounded and not removed; positive measure is Haar on the pair of conserved angles and the carrier is one ray; no distance, no speed, no clock, and nothing here is evidence about Lorentz kinematics; definite facts of the one trajectory. Remainder: (a) the certificate as records: a pointer read in arcs fine enough to recover its displacement over the run (the hand of ~~#201~~ on this arena), and the inequalities stated for the displacements; (b) a sharper packet estimate (weight at most `ε` before and at least `1 − ε` after, on a shorter interval), with a pointwise statement for the coupled carrier, to locate the change and the write times of ~~#365~~ and ~~#367~~ → ~~#375~~: done in part — a family of packets with the coupled carrier's far-half occupancy at most `1/(2n)` at every flow time in `[0, (625π/3) n²]` and at least `1 − 1/(2n)` at every flow time in `[(3125π/3) n², (4375π/3) n²]`; the window is not narrowed for the coupled carrier and the write times are not relocated; (c) the certificate on the arena of ~~#372~~ with the clock's counts at `T₁`, `T₂`, `T₃`; (d) the records of ~~#367~~ and the certificate on a common fibre. | ~~**M**~~ done | Medium — it answers the objection to ~~#365~~, ~~#367~~ and ~~#372~~ (a carrier with a suitable constant weight satisfies them) with a statement on the same flow and start that every constant-occupancy carrier fails; mathematically small, and at the level of accumulated occupancies, not of records | ~~#367~~, ~~#365~~, ~~#372~~, ~~#201~~ | > | ~~374~~ | Records | Records→spacetime | *new (2026-11-12, the author's programme "From `Σ` to physical spacetime", obligations UO-14 (b), UO-16, UO-24 — Remainder (a) of ~~#372~~, the author's pick "couple the clock to the departure detector")* **The flow commutes with turning the pointers: on this arena nothing else depends on where a pointer started.** ▸ **DONE 2026-11-12 as an obstruction — the brick asked for a coupling by which a clock's count belongs to a record otherwise than through the flow parameter. The coupling was not built. No coupling built so far does that, and on the arena of ~~#372~~ none that leaves the moving angles out of the energy can make another degree of freedom depend on a pointer's angle: a readout is a region of a pointer's moving angle, no moving angle is in the joint energy, and so the flow commutes with turning the pointers.** Generic half [`LF4/PointerInert.lean`](../CsdLean4/LF4/PointerInert.lean) (3 pins): `shiftMoving`, ★★ `hamiltonianFlow_stepEnergy_shiftMoving` (**for the step energy `H(s) + G(s) g(ψ₂)` on a compact symplectic base the flow commutes with turning the new pointer by any `c`**), `hamiltonianFlow_stepEnergy_shiftMoving_fst` (the base does not see where the pointer is). Record half [`RecordLayer/UnifiedRecordInert.lean`](../CsdLean4/RecordLayer/UnifiedRecordInert.lean) (8 pins): `shiftPointers`, `forgetMoving`, `exists_shiftPointers_of_forgetMoving_eq`, `forgetMoving_shiftPointers`, ★★ `clockMessengerFlow_shiftPointers` (**on the arena of ~~#372~~ the joint flow commutes with turning the departure pointer, the arrival pointer and the clock's tick pointer**), ★ `forgetMoving_clockMessengerFlow_congr` (**two starts that differ only in the moving angles have the same carriers and conserved angles at every time**), `clockMessengerFlow_shiftFst_snd`, `clockMessengerFlow_shiftFst_arrival` (turning the departure pointer changes neither the clock's state nor the arrival pointer's; the first is weaker than ~~#372~~'s `clockMessengerFlow_snd`). Contract: no new assumption; UD-106 and UD-107; UO-14, UO-16 and UO-24 annotated; `POSITS.md` unchanged. **Scope:** mathematically small (a moving angle is a cyclic coordinate; each theorem is read off a closed form already in the corpus, UD-95 and UD-104); the coupling asked for was not built, and this row is what is proved in its place; about the energy's form, which is assumed — an obstruction for that form and not for `Σ`; the record half is proved for `clockMessengerFlow` only (expected, not stated in Lean, on the other arenas of UA-13 and UA-15) and does not use the generic half; the generic half covers one step and the pointer that the step adds — a step whose `G` reads an earlier pointer's moving angle is of the step form and is not covered (that is UO-24), and the induction over a tower is not in Lean; what is ruled out is a degree of freedom, other than the pointer itself, whose later value depends on where a pointer started with the carriers and conserved angles the same (two starts with different readout histories are not exhibited in Lean); **not** ruled out: a dependence through the occupancy that drives the pointer (a register driven by the same accumulated occupancy varies with the readout through that common cause), so "no register can hold the count at a record" does not follow, and UD-65 leaves an angle-valued register open; a pointer whose moving angle acts back (UO-24, not in the corpus); and a pointer whose rate depends on the messenger and the clock together, such as the expectation of a product observable on a composite carrier, which does not remove the obstruction (its pointer is again cyclic and responds to the carriers and not to a record); a term of the kind of ~~#212~~ (the clock's carrier detuned by a detector's conserved angle and driving that detector's pointer) leaves every moving angle out of the energy, so the same commutation is expected — a calculation, not in Lean, since that energy has two terms in one conserved angle and is not a `stepEnergy`; it is not built; the record-level statement is for the uncoupled energy of ~~#372~~, where the clock's independence of the messenger is already ~~#372~~'s; pointwise, with no measure and no probability. Remainder: **the author's decision** between (a) a composite carrier for the messenger and the clock (one ray on the tensor product, a new arena assumption) with the clock's tick pointer coupled to `P_near ⊗ X`, so that the tick pointer turns only while the messenger is in the departure region and (expected, not shown) the total tick count of its history is the count at departure with no reference to the flow parameter; its pointer is still cyclic, so it gates a count by occupancy and not by a record — this needs the sharper packet estimate of ~~#373~~ Remainder (b) first, since with a quarter of the weight unaccounted for the gated swing drifts; (b) UO-24, a pointer whose moving angle enters the energy; (c) accepting alignment by the flow parameter (UA-7) for M6 and recording it as a condition of gate B. **Decided by the author 2026-11-12: (a).** | ~~**M**~~ done (as an obstruction) | Low as mathematics (a cyclic coordinate, read off UD-95 and UD-104); its use is that it records why Remainder (a) of ~~#372~~ was not built, gives a structural reason, for the couplings in the corpus, behind the limit of M4 and M6 (records and counts aligned only by the flow parameter), and frames the author's decision | ~~#372~~, ~~#368~~, ~~#358~~, ~~#212~~ | > | ~~375~~ | Records | Records→spacetime | *new (2026-11-12, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainder (b) of ~~#373~~, the author's pick; a prerequisite of the composite carrier chosen under ~~#374~~)* **A family of packets: the far half's occupancy at most `1/(2n)` on an interval before the transit and at least `1 − 1/(2n)` on an interval after, at each flow time of those intervals, for the coupled carrier.** ▸ **DONE 2026-11-12, in part: the bounds are sharpened and are pointwise for the coupled carrier; the transit window in flow time is not narrowed for the coupled carrier, and no write time is relocated.** [`RecordLayer/UnifiedFineTransit.lean`](../CsdLean4/RecordLayer/UnifiedFineTransit.lean) (29 pins). The family: `finePacket n` (`8n` flat modes from `p₀ = 4` upward, spread `1/(125 n)`, on a ring of length `2000π n²` with at least `640000 n³` sites, velocity `4/5`, predicted angle `t/(1250 n²)`), `fineHop`, `omega_fine`, `fine_spread`, `fine_modes_lt`, `fine_sites`, `fine_win`, `fine_norm`, ★ `fine_budget` (the error budget of ~~#265~~'s bound is at most `1/(6n)` up to `1875π n²`), ★★ `regionWeight_fine_far` (**while the predicted angle is at least `δ` short of the far half, the far half's weight times `1 − cos δ` is at most `1/(6n)`**), ★★ `regionWeight_fine_near` (**while it is at least `δ` inside, the weight outside the far half times `1 − cos δ` is at most `1/(6n)`**), `regionWeight_fine_far_third`, `regionWeight_fine_near_third` (at `δ = π/3`: at most `1/(3n)` for `t ≤ (625π/3) n²`, at least `1 − 1/(3n)` for `(3125π/3) n² ≤ t ≤ (4375π/3) n²`), ★★ `exists_fine_transit` (**for every `ε > 0` and margin `0 < δ ≤ π/2` there is an `n` with far-half weight at most `ε` before the angle `π/2 − δ` and at least `1 − ε` between `π/2 + δ` and `3π/2 − δ`, on every ring of at least `640000 n³` sites** — the free packet). The coupled carrier: `abs_arenaObs_regionProj_two_sub_free_le` (at each flow time `s` the two detectors move the expectation of a region projector by at most `2|s|(|g₁ f(θ₂)| + |g₂ h(ψ₂)|)`), `fine_chainInner`, `chainProp_finePacket`, `finePacket_ne_zero`, `arenaObs_farHalf_free_finePacket`, `fineCarrier`, `regionTwoFlow_fineCarrier` (the carrier component of `regionTwoFlow`), `fineFibres` (`17500 π n³ (|g₁ f(θ₂)| + |g₂ h(ψ₂)|) < 1`), `isOpen_fineFibres`, `zero_mem_fineFibres_trainF`, ★ `volume_fineFibres_trainF_pos`, ★ `abs_arenaObs_fineCarrier_sub_le`, `fine_backReaction_le`, ★★ `fineCarrier_far_before`, `fineCarrier_far_after` (**on a fine fibre the coupled carrier's far-half occupancy is at most `1/(2n)` at every flow time in `[0, (625π/3) n²]` and at least `1 − 1/(2n)` at every flow time in `[(3125π/3) n², (4375π/3) n²]`**). Contract: no new assumption; UD-108 and UD-109; UO-16 annotated; `POSITS.md` unchanged. **Scope:** not the first sharpened packet — ~~#268~~ has one instance (`10⁴` modes, a narrow arc, `1/60` and `9/10`) for the free packet and a test detector; new here are bounds that tend to `0` and `1` along a family, and the statement at each flow time for the carrier coupled to the pointers of ~~#367~~; the estimate is ~~#265~~'s translation-phase bound with a larger system: the ring's length grows as `n²`, the sites as `n³`, the modes as `n`, the flow times as `n²`, and the hop changes with `n` (a family of generators); for `n ≤ 2` the coupled bounds are no sharper than ~~#265~~'s; nothing is stated between the two intervals or after the second; the back-reaction estimate is ~~#357~~/~~#367~~'s, composed; the allowed back-reaction shrinks as `n⁻³` (the fine fibres are open and of positive measure for `trainF` and shrink with `n`; nothing is proved off them); the window between "before" and "after" in flow time is the old one scaled by `n²` — the margin `δ` is narrowed only for the free packet (`exists_fine_transit`), at the price `1/(6n(1 − cos δ))`, and in angle only: the unlocated window in flow time is `2500 n² δ`, which grows as `δ → 0` with the `n` the proof supplies; occupancies of the carrier, not records: no pointer inequality, no `IsRecordWrite`, the write times of ~~#365~~ and ~~#367~~ are not relocated; every time is a flow time (UA-7); little weight, not none — not a light-cone statement; the detector law is UA-5's drift coupling. Remainder: (a) the coupled carrier at a general margin `δ` (the fibre condition then depends on `δ`) → ~~#379~~: done, once round the ring, on fibres that do not depend on `δ`; (b) the write times of ~~#367~~ on this family, with thresholds set against `1/(2n)`; (c) the composite carrier of ~~#374~~'s decision, whose gated tick pointer needs these bounds for the departure region → ~~#376~~: done as a first brick — the arena (UA-17), its flow from a product ray where the gated pointer does not act back, and two bounds on the gated pointer; no tick, count or record. | ~~**M–L**~~ done in part | Medium-low — on a larger packet, replaces ~~#265~~'s `1/4` and `3/4` (a third in ~~#373~~'s accumulated form) by `1/(2n)` and `1 − 1/(2n)` (sharper from `n = 3`), and states the far-half occupancy of the carrier coupled to the two region pointers pointwise on two intervals; the back-reaction estimate is ~~#357~~/~~#367~~'s and the packet estimate is ~~#265~~'s with parameters; an ingredient, not a milestone | ~~#373~~, ~~#367~~, ~~#268~~, ~~#265~~ | > | ~~376~~ | Records | Records→spacetime | *new (2026-11-12, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — the author's decision of 2026-11-12 under ~~#374~~: the messenger and the clock as one composite carrier, the clock's pointer driven by a product observable; Remainder (c) of ~~#375~~)* **A composite carrier: the clock's pointer turns only while the messenger is in a region.** ▸ **DONE 2026-11-12, as a first brick: the arena, its flow from a product ray on the fibres where the gated pointer does not act back, and two bounds on the gated pointer. No tick, no count and no record on this arena; UO-16 not met.** [`CV/ProductRay.lean`](../CsdLean4/CV/ProductRay.lean) (20 pins) and [`RecordLayer/UnifiedCompositeCarrier.lean`](../CsdLean4/RecordLayer/UnifiedCompositeCarrier.lean) (28 pins). **A new arena assumption, UA-17, entered in the contract with this row.** Product rays over two index types: `prodVec`, `prodRay`, `prodRay_mk`, `arenaDM_prodRay`, ★ `arenaObs_kronecker` (the expectation of `A ⊗ B` in a product ray is the product of the expectations, `A` Hermitian), `arenaObs_kronecker_one`, `arenaObs_one_kronecker`, `kronecker_isHermitian`, `kronU`, ★ `arenaKick_kronU`, `exp_smul_kronecker_one`, `exp_smul_one_kronecker`, ★ `exp_smul_kronecker_add` (the exponential of a Kronecker sum is the Kronecker product of the exponentials) — the product-state statements are those of `CV/CompositeArena.lean` for field sectors, over arbitrary index types; the Kronecker-sum exponential is new. The arena: `prodIdx`, `compositeH` (`A₀ ⊗ 1 + 1 ⊗ C₀`), `idxTwoH_composite`, ★★ `idxTwoCarrierU_composite` (with the third pointer's term absent the composite propagator is the messenger's propagator, with its two pointers' terms, times the clock's free propagator), `arenaKick_idxTwoCarrierU_composite`, `idxTwoShiftFst_composite`, `idxTwoShiftSnd_composite`, `gateShift`, `gateFlow` (the three-pointer flow of ~~#368~~ with the third pointer coupled to `P ⊗ X`), ★ `triMuL_map_gateFlow`, ★★ `gateFlow_prodRay` (**from a product ray, on a fibre where the third profile vanishes: the carrier stays a product ray, the messenger and its two pointers move as on their own arena, the clock moves freely, and the third pointer advances by `−2 k'(χ₂) ∫₀ᵗ ⟨P⟩_messenger ⟨X⟩_clock`**). The gate: `continuous_arenaObs_idxTwoCarrierU`, `continuous_arenaObs_idxFreeU`, ★★ `abs_gateShift_sub_le` (**between two flow times the third pointer moves by at most `2|k'| c` times the messenger's accumulated `⟨P⟩`**, for `⟨P⟩ ≥ 0`, `‖X‖ ≤ c`), `abs_gateShift_sub_le_of_le`, ★★ `abs_gateShift_sub_free_le` (**it is within `2|k'| c` times the accumulated `1 − ⟨P⟩` of `−2 k'` times the clock's accumulated free reading of `X`**, for `⟨P⟩ ≤ 1`; on a fibre where the profile vanishes that is the displacement of a pointer coupled to `X` alone), `abs_gateShift_sub_free_le_of_le`. The messenger on its ring and the two-level clock: `arenaObs_regionProj_compl`, `gateClockFlow` (energy `−2⟨H_D ⊗ 1 + 1 ⊗ H_dev⟩ − 2 g₁ f(θ₂)⟨P_A ⊗ 1⟩ − 2 g₂ h(ψ₂)⟨P_B ⊗ 1⟩ − 2 g_c k(χ₂)⟨P_A ⊗ X⟩`), `triMuL_map_gateClockFlow`, `gateClockShift`, ★ `gateClockFlow_apply`, ★ `gateClockFlow_fst_pointer`, `gateClockFlow_snd_pointer` (the two region pointers move exactly as in `regionTwoFlow` of ~~#367~~), `idxPointerShift_eq_free`, ★ `fine_gateClock_before`, `fine_gateClock_after` (**on the packet family of ~~#375~~ with the gate on the near half: at every flow time `0 ≤ t ≤ (625π/3) n²` the gated pointer's displacement is within `|k'(χ₂)||g_c| t/n` of the displacement the pointer of ~~#357~~ has on the clock's own arena from the same clock start (a comparison across arenas), and between any two flow times of `[(3125π/3) n², (4375π/3) n²]` it changes by at most `|k'(χ₂)||g_c| (t₂ − t₁)/n`**). Contract: **UA-17** (new); UD-110, UD-111, UD-112; UO-16 annotated; `POSITS.md` unchanged. **Scope:** only the fibres where the gated pointer's profile vanishes (a null set for `trainF`) and only product starts (a null set of the base for `μ_FS`) — two calculations, not in Lean; off those fibres the product structure is not shown to persist (that `k(χ₂) P ⊗ X` entangles the messenger with the clock is expected, not in Lean) and nothing is proved; the pointer is gated by the messenger's occupancy, an expectation, and not by a record — it is again a cyclic coordinate by the generic statement of ~~#374~~ (not instantiated for `gateFlow`); the clock has no pointer of its own on this arena, so the comparison with ~~#357~~'s pointer is of two real functions across arenas; the dependence on the messenger is shown along one run, and no theorem compares two messenger starts; the second bound is a rate bound `|k'||g_c|/n`, not rest; ~~#274~~ has a gate by a register under a stated hybrid law, and here the gate is a term of one joint energy; nothing here is evidence about Lorentz kinematics; no ready arc, no threshold, no `IsSwing`, no `IsTickCount`, no `IsRecordWrite` on this arena, so "the count stops at departure" is not proved; the errors grow with the system: `|k'||g_c| t/n` with `t` up to `(625π/3) n²` is `(625π/3)|k'||g_c| n` at the end of the first interval, small against the free pointer's swing (of order `|k'||g_c|/R` for a clock of rate `R`) only when `n R` is small — a regime with many swings before departure and a small error needs `1/n² ≪ R ≪ 1/n`, a calculation and not a theorem; nothing is stated between the two intervals or after the second; the clock's ring is not placed (`X` acts at every site of the clock's ring; the gate is a region of the messenger's ring), so "the clock is at the departure region" is still not a statement of the arena; the product term joins a region of one ring to every site of the other, and the cone bound of ~~#361~~ is not proved for this generator; distinguishable carriers, no symmetrisation; flow times (UA-7); definite facts, no statistics, UA-6 not used. Remainder: (a) the gated pointer's swings as tick records and a count on this arena (ready and display arcs, `IsSwing` for the gated pointer before departure, the count constant after) → ~~#377~~: done in part — tick records and a count before departure; after arrival, no tick begun and completed inside a sub-interval short enough for a stated bound; the count is not shown to stop; (b) the same on fibres of positive measure for the gated pointer, which needs a bound on the entangling term; (c) a clock placed at a region (a gate `P_A ⊗ Q_A' X` with the clock's own site projector); (d) the cone bound for the composite generator. | ~~**L**~~ done as a first brick | Medium — the arena the author chose, with its assumption on the register; a pointer whose rate is, on those fibres, the product of the messenger's occupancy and the clock's reading (`gateFlow_prodRay`); the mathematics is the three-pointer flow of ~~#368~~ and product states, and it is exact only where the gated pointer does not act back | ~~#375~~, ~~#374~~, ~~#372~~, ~~#368~~, ~~#367~~, ~~#357~~ | > | ~~377~~ | Records | Records→spacetime | *new (2026-11-12, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainder (a) of ~~#376~~)* **The gated pointer ticks while the messenger is in the region; no tick inside a short interval after.** ▸ **DONE 2026-11-12, in part: tick records and a count before departure, and no ready-to-displayed passage inside a sub-interval after arrival that is short enough for a stated bound; "the count stops at departure" is not proved, and UO-16 is not met.** [`RecordLayer/UnifiedGateTicks.lean`](../CsdLean4/RecordLayer/UnifiedGateTicks.lean) (10 pins). No new assumption; the arena is UA-17's. `pshift_notMem_displayArc_of_mem_readyArc` (the two arcs are a sixth of a turn apart on both sides), `isSwing_thd_of_abs_shift_sub_le` (the swing criterion of ~~#358~~ for a third pointer), ★ `clockPointerShift_train` (the displacement of the clock's own pointer in the exact instance of ~~#358~~ is `(1/4)(1 − cos R t)`), `gateTrainStart` (the fine packet of index `n` and the one-mode clock as one product ray, the gated pointer at `1/12` on `χ₂ = 0`), `gateClockFlow_train_thd`, ★ `abs_gateClockShift_train_sub_le`, ★★ `isSwing_gateClock_before` (**up to a flow time `0 ≤ T ≤ (625π/3) n²` with `|g_c| T/n < 1/12`, the history of `gateClockFlow` read through the third pointer's arcs is a swing at the clock's rate `R`, with thresholds `±4|g_c| T/n`**), ★ `gateClock_tickCount_before` (**a tick count `N` at `T` with `|N · 2π/R − T| ≤ 2π/R`**; `N` may be zero), ★ `gateClock_tick_isRecordWrite` (**each tick whose displayed window ends by `T` is an `IsRecordWrite` of the third pointer's readout**; not packaged as an `ActualOccurrence` of `gateClockFlow`), ★★ `gateClock_no_tick_after` (**for flow times `a ≤ b` of `[(3125π/3) n², (4375π/3) n²]` with `|k'(χ₂)||g_c| (b − a)/n ≤ 1/6`: a gated pointer in its ready arc at `a` is not in its displayed arc at `b`** — for every clock ray and every profile on a fibre where it vanishes). Contract: UD-113, UD-114; UO-16 annotated; `POSITS.md` unchanged. **Scope:** the arena and its limits are ~~#376~~'s — a null set of fibres for the gated pointer, a product start (a null set of the base) — calculations, not in Lean — and the fine fibres of ~~#375~~; the clock is the one-point exact instance of ~~#358~~; the hypothesis `|g_c| T/n < 1/12` bounds the rate from above (`g_c` is `R` times a factor fixed by the clock's parameters, so `n R` small at `T = (625π/3) n²`) while the count, within one of `R T/2π`, is large only when `n² R` is large — that both hold with at least one tick for some clock is a calculation, not in Lean, and for small `T` the theorem holds with count zero; gated by the messenger's occupancy, an expectation, not by a record (the gated pointer is a cyclic coordinate by the generic statement of ~~#374~~, not instantiated here); for the whole interval after arrival the hypothesis needs `|k'||g_c| n ≤ 1/(2500π)`, which no theorem assumes; **"the count stops" is not proved**: nothing is stated between `(625π/3) n²` and `(3125π/3) n²`, where the pointer may tick, or after `(4375π/3) n²`; a tick begun before `(3125π/3) n²` may complete inside the interval; no theorem gives the count at a time after departure; no degree of freedom holds the count (UO-14 (b)); the clock's own pointer is not on this arena, so "it goes on ticking" is ~~#358~~'s statement on another arena; the tick records last less than one turn; flow times (UA-7); no distance, no speed, no second observer. Remainder: (a) a bound on the number of ticks in the transit window and the count at a time after arrival → ~~#378~~: an upper bound, too weak to fix the count; (b) parameters with many ticks before departure, in Lean → ~~#378~~: done; (c) the Remainders (b) to (d) of ~~#376~~. | ~~**M–L**~~ done in part | Medium-low — tick records and a count on the composite carrier, from a pointer whose swing is tied to the messenger's occupancy; the swing and count machinery is ~~#358~~'s and the bounds are ~~#376~~'s; a count of at least one is not shown in this row, and the count is not shown to stop | ~~#376~~, ~~#375~~, ~~#358~~ | > | ~~378~~ | Records | Records→spacetime | *new (2026-11-13, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainders (a) and (b) of ~~#377~~)* **The gated count: as many ticks before departure as wanted, and a bound at any later time.** ▸ **DONE 2026-11-13: (b) in full, (a) as an upper bound too weak to fix the count; UO-16 not met.** [`RecordLayer/UnifiedGateCount.lean`](../CsdLean4/RecordLayer/UnifiedGateCount.lean) (5 pins). No new assumption; the arena is UA-17's. `pshift_notMem_readyArc_of_mem_displayArc`, ★ `hasTicks_le_of_travel` (**ticks need travel**: a third pointer turned by `D t` with `|D b − D a| ≤ W b − W a` on `[s, T]`, without `N₀ + 1` ticks by `s`, has at most `N₀ + 3 (W T − W s) + 3/2` ticks by `T`), ★ `gateClock_hasTicks_le` (**on the composite carrier, from any product ray on a fibre where the gated pointer's profile vanishes, at most `N₀ + 6|k'(χ₂)||g_c| (T − s) + 3/2` ticks by `T`**), `gateClock_coupling_le` (in the exact instance of ~~#358~~, `g_c > 0` and `8 κ g_c ≤ R`, `κ = 1 − ((μ₁ − μ₀)/μ₀)²`), ★ `gateClock_hasTicks_before` (**for every `K` and every index `n` with `3πK < κ n` and `6K ≤ 625 n² R`, the gated pointer has at least `K` ticks by the flow time `2πK/R ≤ (625π/3) n²`**). Contract: UD-115, UD-116; UO-16 annotated; `POSITS.md` unchanged. **Scope:** `gateClock_hasTicks_before` removes the vacuity ~~#377~~ left open — its two hypotheses are lower bounds on `n` for a given clock and `K`, so they hold for every large `n` (the system grows with `n` as in ~~#375~~); `HasTicks` is a lower bound, `K < κ n/(3π)`, and for a given clock `2πK/R` is a vanishing fraction of `[0, (625π/3) n²]` as `n` grows, so the count at departure is given by no theorem unless `|g_c| n < 1/(2500π)`; the later-time bound is that of a pointer moving at its largest rate `2|k'||g_c|` throughout — with `8 κ g_c ≤ R` and `|k'| = 1` it is at most `3R(T − s)/(4κ) + 3/2` ticks where the clock makes about `R(T − s)/(2π)` (that it is not much smaller is a calculation, not in Lean), so over the transit window of ~~#375~~, four times as long as the interval before it, it does not give the count at arrival to within the count at departure; no lower bound beyond monotonicity; "the count stops at departure" is still not proved; the ticks are `HasTicks` passages of the readout history, not `ActualOccurrence`s; the arena, fibres and starts are ~~#376~~'s and ~~#377~~'s (null sets by calculations not in Lean); gated by an occupancy, not a record; flow times (UA-7); nothing here is evidence about Lorentz kinematics. Remainder: (a) a sharper count through the transit, which needs the sign of the gated rate and control of the pointer's wrapping, or a narrow transit window (~~#379~~ gives the margin). | ~~**M**~~ done | Medium-low — closes the satisfiability question of ~~#377~~ with a theorem; the travel bound is elementary and weak | ~~#377~~, ~~#376~~, ~~#358~~ | > | ~~379~~ | Records | Records→spacetime | *new (2026-11-13, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainder (a) of ~~#375~~ and the return leg listed under UO-16)* **The return leg: the messenger comes back to the gate, at any margin.** ▸ **DONE 2026-11-13, in displacements: the packet is followed once round the ring at any margin `δ`, and the gated pointer's displacement is bounded on the three intervals; no tick, count or record on the return, so no radar time from counts; UO-16 not met.** [`RecordLayer/UnifiedFineReturn.lean`](../CsdLean4/RecordLayer/UnifiedFineReturn.lean) (17 pins). No new assumption. The free packet: ★ `fine_budget_round` (the error budget of ~~#375~~ is at most `1/(6n)` up to the flow time `3125π n²`, predicted angle `5π/2`), ★ `regionWeight_fine_back` (while the predicted angle is between `3π/2 + δ` and `5π/2 − δ`, the far half's weight times `1 − cos δ` is at most `1/(6n)`). The coupled carrier: `roundFibres` (`37500 π n³ (|g₁ f(θ₂)| + |g₂ h(ψ₂)|) < 1`), `isOpen_roundFibres`, `zero_mem_roundFibres_trainF`, `volume_roundFibres_trainF_pos`, `round_backReaction_le`, `fineEta` (`(1/(6n))(1 + 1/(1 − cos δ))`), `le_fineEta`, ★ `fineCarrier_far_le_before`, `fineCarrier_near_le_away`, `fineCarrier_far_le_back` (**on a round fibre the coupled carrier's far-half occupancy is at most `fineEta n δ` while the predicted angle is in `[0, π/2 − δ]` and in `[3π/2 + δ, 5π/2 − δ]`, and at least `1 − fineEta n δ` in `[π/2 + δ, 3π/2 − δ]`**). The gate: `abs_gateShift_sub_sub_free_le`, `abs_gateShift_sub_sub_free_le_of_le` (the gating bound of ~~#376~~ between two flow times), ★ `round_gateClock_before`, `round_gateClock_away`, `round_gateClock_back` (**with `E = 2|k'(χ₂)||g_c| fineEta n δ`: before the messenger leaves the gated pointer's displacement is within `E t` of that of the pointer of ~~#357~~ on the clock's own arena; while it is away the gated pointer changes at a rate at most `E`; after it is back the gated pointer's advance is within `E (t₂ − t₁)` of the advance of the pointer of ~~#357~~**). Contract: UD-117, UD-118, UD-119; UO-16 annotated; `POSITS.md` unchanged. **Scope:** displacements, not ticks — on the return interval the gated pointer starts from an angle no theorem locates, the swing of ~~#377~~ is read from the flow time `0` and does not apply, and no tick, count or record is stated there; three intervals with gaps of angle `2δ` (flow time `2500 n² δ`) on which nothing is stated, and nothing after the angle `5π/2 − δ`; `fineEta n δ` is of order `1/(n δ²)`, but each bound is `E` times the length of its interval (of order `n²`), so over a whole interval it is of order `|g_c| n/δ²` and is small against the swing only for a clock whose rate falls with `n` (a calculation, not in Lean); all three are statements about the gated pointer only on a fibre where its profile vanishes; the round fibres are a subset of the fine fibres (by comparing the constants; no lemma), open and of positive measure for `trainF`; the gated pointer's fibres and the product starts are null sets (calculations, not in Lean); the comparison with ~~#357~~'s pointer is of two real functions across arenas; gated by an occupancy, not a record; one messenger, one clock, one round of one ring — the ring's length fixes the round-trip time, no distance is measured and there is no second observer; flow times (UA-7), the angle `t/(1250 n²)` is the predicted one; little weight, not none; nothing here is evidence about Lorentz kinematics. Remainder: (a) ticks and a count on the return leg (a swing read from a start other than the flow time `0`) → ~~#380~~: done as two cases — the return swing is offset by the clock's phases, and whether it ticks through the fixed arcs depends on the offset; (b) a radar time from two counts, which needs (a) and a count through the gaps → ~~#380~~: not available from the gated pointer; (c) a second round, and a second observer. | ~~**L**~~ done in displacements | Medium-low — the first return to the gate of UA-17: a free round of the ring (no reflector, no relay), read as rate bounds on the gated pointer, and the coupled carrier at any margin; earlier returns with records and radar readings are ~~#200~~/~~#202~~ (echo arena), ~~#270~~ (reflection, stated detector law) and ~~#274~~/~~#278~~ (relay, stated hybrid law), none on a joint flow with the clock; the estimates are ~~#375~~'s with a longer horizon and the bounds ~~#376~~'s on three intervals; no record and not a radar time | ~~#376~~, ~~#375~~, ~~#377~~ | > | ~~380~~ | Records | Records→spacetime | *new (2026-11-13, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainders (a) and (b) of ~~#379~~)* **The gated pointer on the return leg: its swing comes back offset by the clock's phase.** ▸ **DONE 2026-11-13, as two cases: read through the fixed arcs of ~~#377~~, whether the gated pointer ticks on the return leg depends on an offset fixed by the clock's phases — never ready in one stated range (not shown non-empty in Lean), a swing again for a small offset; no theorem gives a count across the absence and no radar time is formed; UO-16 not met.** [`RecordLayer/UnifiedGateReturn.lean`](../CsdLean4/RecordLayer/UnifiedGateReturn.lean) (11 pins). No new assumption. `roundOut`, `roundBack`, `roundEnd` (the flow times `T_a`, `T_b`, `T_c` of the predicted angles `π/2 − δ`, `3π/2 + δ`, `5π/2 − δ`), `gatePhaseShift` (`s = (1/4)(cos R T_b − cos R T_a)`), `gateReturnErr` (`2|g_c| (fineEta n δ · t + 5000 n² δ)`), `fineEta_nonneg`, `abs_gateClockShift_train_sub_le'` (the gated displacement changes by at most `2|g_c| (b − a)`), ★ `abs_gateClockShift_return_sub_le` (**in the instance of ~~#377~~ on a round fibre of ~~#379~~, at every flow time `t` of the return leg the gated pointer's displacement is within `gateReturnErr` of `(1/4)(1 − cos R t) + s`**), ★ `gateClock_return_not_ready` (**if `1/12 + e < s` and `s + e ≤ 5/12`, with `e` the error at the end of the return leg, the gated pointer is in its ready arc at no flow time of the return leg**), ★ `isSwing_gateClock_return` (**if `|s| + e < 1/12`, the history read from a whole turn `2πj/R` of the clock on the return leg is a swing at the rate `R` with thresholds `±4(|s| + e)`**), `gateClock_tickCount_return` (a tick count there, within one period of the time since that turn; it may be zero). Contract: UD-120, UD-121; UO-16 annotated; `POSITS.md` unchanged. **Scope:** the offset is there because the gated pointer's rate is the product of the messenger's occupancy and the clock's reading — while the messenger is away the pointer's rate is at most `2|g_c| fineEta n δ` (little over the interval only when `|g_c| n/δ²` is small) and the clock's phase moves on; the obstruction is for the product coupling of UA-17 (c) with the drift law of UA-5 **read through the fixed ready and display arcs and the start `1/12` (UA-4)** — in the never-ready case the pointer still swings, and arcs placed by the offset would read ticks; it is not a statement about `Σ`; another coupling, or such arcs, may behave otherwise, and no theorem shows that no fixed arcs work for every absence; `T_a`, `T_b` are flow times of predicted angles, located by no record; neither case is shown satisfiable — the error is of order `|g_c| (n/δ² + n² δ)`, and that it is below `1/12` with an offset in the stated range is a calculation, not in Lean, needing a clock whose rate is of order `1/n²` (large enough for a whole turn on the return leg or for the offset to reach `1/12`, small against `δ²/n` and `1/(n² δ)`), and a whole turn `2πj/R` on the return leg is not shown to exist (with ~~#378~~'s lower bounds for `K` ticks before departure it asks for `δ ≪ 1/K` and `n ≫ K/δ²`); the other values of the offset are not treated; no theorem gives a count through the absence or the gaps, so no radar time is formed; the start `2πj/R` of the return swing is a chosen flow time, not given by a record; the limits of ~~#376~~, ~~#377~~ and ~~#379~~ stand (a fibre where the gated pointer's profile vanishes and a product start — null sets by calculations not in Lean; round fibres; one clock start; a gate by an occupancy; ticks as `IsSwing`/`IsTickCount` statements about a readout history, not `ActualOccurrence`s); flow times (UA-7); no distance, no speed, no second observer; nothing here is evidence about Lorentz kinematics. Remainder: (a) a clock pointer that runs throughout beside the gated one (a fourth pointer, one more step of ~~#368~~), with the difference of the two counts as the time away → ~~#381~~: done on a stretch of the absence between two flow times; not a radar time; (b) a coupling other than the product term whose pointer resumes in phase; (c) the cases in between. | ~~**M–L**~~ done as an obstruction | Medium-low — shows that the gated pointer's return swing carries an offset from the clock's phases, so its ticks on the return leg depend on that offset when read through fixed arcs; the estimates are ~~#379~~'s joined, and neither case is shown satisfiable | ~~#379~~, ~~#378~~, ~~#377~~, ~~#376~~ | > | ~~381~~ | Records | Records→spacetime | *new (2026-11-13, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainder (a) of ~~#380~~, the route of a clock pointer that runs throughout, on the author's instruction to continue)* **Two counts in one run: the clock's own pointer beside the gated one.** ▸ **DONE 2026-11-13, in part: the fourth pointer and its swing at every horizon, and the difference of the two counts on a stretch of the absence between two flow times; not a radar time; UO-16 not met.** [`LF4/QuadCoupledFlow.lean`](../CsdLean4/LF4/QuadCoupledFlow.lean) (16 pins) and [`RecordLayer/UnifiedTwoCounts.lean`](../CsdLean4/RecordLayer/UnifiedTwoCounts.lean) (17 pins). **UA-17 is amended, (e): the clock has a pointer of its own again**, and UA-15 records the fourth torus. The arena: `QuadModel`, `QuadSigma`, `quadForm`, `quadForm_isSymplectic`, `triRayEnergy`, `contMDiff_triRayEnergy`, `fourCoupledEnergy`, `fourCoupledEnergy_apply`, `contMDiff_fourCoupledEnergy`, `threeCoupledEnergy_add_smul_triRayEnergy`, `fourCoupledFlow`, ★ `fourCoupledFlow_apply` (a skew product over the fourth conserved angle), `fourCoupledFlow_fst_of_lift_eq_zero`, `fourCoupledFlow_snd_of_lift_eq_zero`, `quadMuL`, ★ `quadMuL_map_fourCoupledFlow` — one more application of the step theorem of ~~#368~~. The run: `isSwing_snd_of_abs_shift_sub_le`, `hasTicks_le_succ_of_no_passage` (no passage from rest to display inside `[s, T]` ⇒ at most one more tick), `exists_tickCount_of_bdd`, `IsTickCount.le_succ_of_no_passage`, `gateFreeFlow` (the flow of ~~#376~~ with a fourth pointer coupled to `1 ⊗ Y`), `quadMuL_map_gateFreeFlow`, `gateFreeFlow_fst` (on a fibre where the fourth profile vanishes the carrier and the first three pointers move by `gateFlow`), ★ `gateFreeFlow_snd_prodRay` (**from a product ray, where neither clock pointer acts back, the fourth pointer advances by `−2 l'(ξ₂)` times the clock's accumulated free reading**), `gateFreeClockFlow`, `quadFreeReady`, `quadFreeDisplay`, `twoCountStart`, `gateFreeClockFlow_twoCount_fst`, `gateFreeClockFlow_twoCount_snd`, ★ `isSwing_free_twoCount` (**the clock's own pointer is a swing at the rate `R` for every horizon, in the run that also carries the gated pointer**), `exists_gate_tickCount` (the gated pointer has a count at every flow time), ★ `twoCounts_away` (**for flow times `a ≤ b` of `[(3125π/3) n², (4375π/3) n²]` with `|g_c| (b − a)/n ≤ 1/6`: the gated count advances by at most one, and the difference between the advance of the clock's count and the advance of the gated count, times `2π/R`, is within `6π/R` of `b − a`**). Contract: UA-17 (e) and UA-15 amended; UD-122, UD-123, UD-124; UO-16 annotated; `POSITS.md` unchanged. **Scope:** `a` and `b` are flow times inside the interval on which the messenger is in the far half — not a departure and a return, and no record marks them; over the two transit windows the gated count is bounded only by ~~#378~~, too weakly, and on the return leg it depends on the offset of ~~#380~~, so this is not a radar time; the time content is the clock's alone — its two-time bound `4π/R` holds for any two non-negative flow times (~~#366~~), the gated count contributes only that its advance is 0 or 1, no hypothesis relates `2π/R` to `b − a`, and all four advances may be zero; `g_c` is free here and only the clock's own strength `g₀` is tuned as in ~~#358~~ — the gated pointer is the swing of ~~#377~~ only when `g_c` also has the tuned value, that swing is not restated on this arena, and for `g_c` far below it the gated count may be zero throughout; with `g_c` tuned the hypothesis allows a number of periods of order `n` (a calculation, not in Lean); a clock pointer that swings at every horizon beside other pointers in one run is not new on other arenas (~~#369~~, ~~#372~~) — what is new is that the clock is a tensor factor of the composite carrier; both clock pointers' profiles must vanish on the fibre and the start is a product ray (null sets by calculations not in Lean), on a fine fibre of ~~#375~~ for the two region pointers; one clock start; both counts are read at flow times, no degree of freedom holds either and no record reads one (UO-14 (b)); the ticks are `IsSwing`/`HasTicks`/`IsTickCount` statements about readout histories, not `ActualOccurrence`s; gated by an occupancy, not by a record; the fourth pointer's swing is ~~#358~~'s carried to this arena by the product structure of ~~#376~~, and the gated bound is ~~#377~~'s; no distance, no speed, no second observer; nothing here is evidence about Lorentz kinematics. Remainder: (a) the clock's count at the two region pointers' writes of ~~#367~~ on this arena, the form of ~~#372~~ now on one composite carrier (those writes are thresholds of accumulated weight and do not certify a transit, and they are stated for the packet of ~~#365~~, not for the fine packet); (b) a round trip with the count at departure and at return, which needs (a) on the return leg → ~~#382~~: a record written on the return leg with the clock's count at a flow time by which it is complete, from a prepared start; one record, not two; (c) fibres of positive measure for both clock pointers. | ~~**M–L**~~ done in part | Medium-low — the running clock and the gated pointer in one run, and what the difference of their counts reads on a stretch of the absence; the arena is one more step of ~~#368~~ and the two ingredients are ~~#358~~'s and ~~#377~~'s; no record marks the stretch | ~~#380~~, ~~#378~~, ~~#377~~, ~~#376~~, ~~#375~~, ~~#372~~, ~~#369~~, ~~#368~~, ~~#366~~, ~~#358~~ | > | ~~382~~ | Records | Records→spacetime | *new (2026-11-13, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — towards Remainder (b) of ~~#381~~; Remainder (a) stays open)* **A record written on the return leg, with the running clock's count at a flow time by which it is complete.** ▸ **DONE 2026-11-13, in part: one record on the return leg of the round trip, from a prepared start, with the clock's own count at a flow time `τ` on the return leg by which the write is complete (the end of the stated write interval, not the instant the readout turns); not a radar time from two records; UO-16 not met.** [`RecordLayer/UnifiedRoundTrip.lean`](../CsdLean4/RecordLayer/UnifiedRoundTrip.lean) (14 pins). No new assumption; the arena is UA-17 with (e). `integral_le_of_forall_le`, `le_integral_of_forall_le`, `nearOcc` (the coupled carrier's occupancy of the near half), `continuous_nearOcc`, `nearOcc_nonneg`, `nearOcc_le_one`, `nearOcc_eq`, `roundUpper` (`1250 n² (π/2 + 3δ + fineEta n δ (π − 2δ))`), `roundLower` (`(1 − fineEta n δ) · 1250 n² (3π/2 − 3δ)`), ★ `integral_nearOcc_le`, `le_integral_nearOcc` (**on a round fibre the accumulated near-half occupancy is at most `roundUpper` at the start of the return leg and at least `roundLower` at its end**), `two_roundUpper_le_roundLower` (for `δ ≤ π/48` and `fineEta n δ ≤ 1/48` the two are a factor two apart, so for `n > 0` any `κ` with `1/(3·roundLower) ≤ κ ≤ 1/(6·roundUpper)` satisfies both conditions — that step is not stated in Lean), `echoFstShown_twoCount_iff`, ★ `return_record_with_count` (**in the run of ~~#381~~ with the first pointer ready, on a round fibre, `0 < δ ≤ π/2`, with a positive rate `κ` such that `κ · roundUpper ≤ 1/6` and `1/3 ≤ κ · roundLower`: there are a flow time `τ` on the return leg and a number `N` such that the first pointer's readout is an `IsRecordWrite` of `true` from the start with write interval `[roundBack n δ, τ]`, the clock's own pointer has exactly `N` ticks by `τ`, and `|N · 2π/R − τ| ≤ 2π/R`**). Contract: UD-125, UD-126; UO-16 annotated; `POSITS.md` unchanged. **Scope:** one record, at the return, from a prepared start — nothing records an emission, and this pointer is not the departure record of ~~#367~~ (its strength is chosen so that it does not write during the first presence); **the write does not certify a return**: it is a threshold of accumulated occupancy, and a carrier with a suitable constant near-half occupancy would give a readout with the same two clauses (a calculation, not in Lean; not exhibited on this arena) — what shows the round trip is the occupancy statement of ~~#379~~, which is not a record; `τ` is located only within the return leg, and the readout may turn before it (the `τ` the crossing lemma gives is where the displacement reaches `1/3`, more than half the leg after its start — a calculation, not in Lean); `N` may be zero, and no hypothesis relates the period to the length of the return leg; the time content of the count is the clock's alone — `|N · 2π/R − τ| ≤ 2π/R` holds at every non-negative flow time (~~#366~~) — and the count is read at the flow time `τ`, not by the record; no degree of freedom holds the pair (UO-14 (b)); the two rate conditions are hypotheses, compatible by `two_roundUpper_le_roundLower`, and that some `n` gives `fineEta n δ ≤ 1/48` is not stated in Lean; both clock pointers' profiles vanish on the fibre and the start is a product ray (null sets by calculations not in Lean); one clock start; the gated pointer is present with a free strength, is not shown to tick and plays no part; the write and the count are `IsRecordWrite` and `IsTickCount` statements about readout histories, not `ActualOccurrence`s; the messenger comes back because the ring closes — the flow time of the return is fixed by the ring's length and the packet's velocity (UA-8), no distance is measured, there is no reflector and no second observer; ~~#360~~/~~#366~~ have a clock's count at a position write with the flow time shown to be the switching instant, ~~#369~~ counts at two readout switches in one run, ~~#372~~ a separate clock's counts at a departure write and an arrival write on the uncoupled product arena, and ~~#200~~/~~#202~~, ~~#270~~, ~~#274~~/~~#278~~ have emission and return records with a clock reading under their own laws — what is new is that the write is on the return leg of the free round trip and the clock is a tensor factor of the same carrier; flow times (UA-7); nothing here is evidence about Lorentz kinematics. Remainder: (a) a second record in the same run, so that the reading is a difference of two records: the far-half arrival write with the clock's count (the second pointer already on the arena), or a departure record (a second near-half pointer with the strength of ~~#367~~; a fifth torus) → ~~#383~~: done with the far pointer — two records and two counts in one run; no emission or departure record; (b) a certificate of the return as a record, in the manner of ~~#373~~; (c) the count read by a record (UO-14 (b)). | ~~**M–L**~~ done in part | Medium-low — a record on the return leg with the running clock's count, on the composite carrier; the write is ~~#365~~'s crossing lemma on the occupancy bounds of ~~#379~~ and the count is ~~#381~~'s; one record, not certified, not sharp | ~~#381~~, ~~#380~~, ~~#379~~, ~~#376~~, ~~#372~~, ~~#367~~, ~~#366~~, ~~#365~~ | > | ~~383~~ | Records | Records→spacetime | *new (2026-11-13, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainder (a) of ~~#382~~ by the far pointer, and part of Remainder (a) of ~~#381~~)* **Two records and two counts in one run: one written during the absence, one on the return leg.** ▸ **DONE 2026-11-13, in part: two records in order with the clock's own count at a flow time by which each is complete; no emission record, so not a radar time; UO-16 not met.** [`RecordLayer/UnifiedArrivalReturn.lean`](../CsdLean4/RecordLayer/UnifiedArrivalReturn.lean) (13 pins). No new assumption; the arena is UA-17 with (e). `roundAway`, `roundAwayEnd` (the flow times of the predicted angles `π/2 + δ`, `3π/2 − δ`), `farOcc`, `continuous_farOcc`, `farOcc_nonneg`, `farOcc_le_one`, `farUpper` (`1250 n² (fineEta n δ (π/2 − δ) + 2δ)`), `farLower` (`(1 − fineEta n δ) · 1250 n² (π − 2δ)`), `integral_farOcc_le`, `le_integral_farOcc` (on a round fibre the accumulated far-half occupancy is at most `farUpper` when the absence begins and at least `farLower` when it ends), `two_farUpper_le_farLower`, `echoSiteShown_twoCount_iff`, ★ `arrival_and_return_records_with_counts` (**in the run of ~~#381~~ with both region pointers ready, on a round fibre, `0 < δ ≤ π/2`, both rates positive, the far rate with `κ₂ · farUpper ≤ 1/6` and `1/3 ≤ κ₂ · farLower` and the near rate as in ~~#382~~: there are flow times `τ_a ≤ τ_r`, the first in the absence and the second on the return leg, such that the far pointer's readout is an `IsRecordWrite` with write interval `[roundAway n δ, τ_a]`, the near pointer's with `[roundBack n δ, τ_r]`, the clock's own pointer has exactly `N_a` ticks by `τ_a` and `N_r` by `τ_r`, and `|(N_r − N_a) · 2π/R − (τ_r − τ_a)| ≤ 4π/R`**). Contract: UD-127, UD-128; UO-16 annotated; `POSITS.md` unchanged. **Scope:** both writes are thresholds of accumulated occupancy and certify neither an arrival nor a return — carriers with suitable constant occupancies would give readouts with the same clauses (a calculation, not in Lean; not exhibited on this arena); `τ_a`, `τ_r` are the ends of the stated write intervals, not the instants the readouts turn, and each is located only within its interval, so `τ_r − τ_a` is known only to lie between `2500 n² δ` and `1250 n² (2π − 2δ)`; the time content is the clock's alone — the two-time bound holds for any two non-negative flow times (~~#366~~), no hypothesis relates the period to `τ_r − τ_a`, and both counts may be zero; the counts are read at flow times, no record reads one and no degree of freedom holds a pair (UO-14 (b)); no emission record (the start is prepared), no reflector, no distance, one ring, no second observer — not a radar time; four rate conditions are hypotheses, each pair compatible for `δ ≤ π/48` and `fineEta n δ ≤ 1/48` (`two_roundUpper_le_roundLower`, `two_farUpper_le_farLower`), and that some `n` gives the latter and rates then exist is not stated in Lean, nor that one fibre pair meets both windows, both signs and the round-fibre bound; the far readout is shown to display only through its window after `τ_a`, not still when the near readout is written; both clock pointers' profiles vanish on the fibre and the start is a product ray (null sets by calculations not in Lean); one clock start; the gated pointer is present with a free strength, is not shown to tick and plays no part; the writes and counts are `IsRecordWrite` and `IsTickCount` statements about readout histories, not `ActualOccurrence`s; ~~#360~~/~~#366~~/~~#369~~ have counts at readout switches in one run, ~~#367~~ a departure write and an arrival write in one run (thresholds of accumulated weight) and ~~#372~~ a separate clock's counts at both, for the crossing packet on other arenas; ~~#200~~ and ~~#202~~ (echo arena), ~~#270~~ (reflection, stated detector law) and ~~#274~~ and ~~#278~~ (relay, stated hybrid law) have emission and return records with a clock reading, under their own laws — what is new is the pair on the free round trip with the clock a tensor factor of the same carrier; flow times (UA-7); nothing here is evidence about Lorentz kinematics. Remainder: (a) an emission or departure record in the same run (a fifth torus), so that a round trip is read from departure to return → ~~#384~~: a write during the first presence, by a second near-half pointer in place of the gated pointer (no fifth torus) — three records, three counts; not an emission record; (b) a certificate of the round trip as a record, in the manner of ~~#373~~; (c) the counts read by records (UO-14 (b)); (d) a second observer. | ~~**M**~~ done in part | Medium-low — two records and two counts on the free round trip, on the composite carrier; the writes are ~~#365~~'s crossing lemma on the bounds of ~~#379~~ and the count bound is ~~#366~~'s; not certified, not sharp, no emission record | ~~#382~~, ~~#381~~, ~~#380~~, ~~#379~~, ~~#376~~, ~~#372~~, ~~#367~~, ~~#366~~, ~~#365~~ | > | ~~384~~ | Records | Records→spacetime | *new (2026-11-13, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainder (a) of ~~#383~~, by a second near-half pointer in place of the gated pointer, not by a fifth torus)* **Three records and three counts in one run: before, during and after the absence.** ▸ **DONE 2026-11-13, in part: three records in order with the clock's own count at a flow time by which each is complete, and the midpoint of the first and last times placed in the absence; the first is not an emission record and nothing is certified; UO-16 not met.** [`RecordLayer/UnifiedThreeRecords.lean`](../CsdLean4/RecordLayer/UnifiedThreeRecords.lean) (22 pins). **Amends UA-17 with clause (f)** (the third pointer's clock factor is `g₃ · 1`, so it is a second pointer on the near half; the gated pointer of ~~#376~~ is absent from this run) — ⚠️ not yet read by the author. `clockOne`, `clockOne_isHermitian`, `arenaObs_clockOne`, `threeRecordFlow`, `quadMuL_map_threeRecordFlow` (the flow preserves `quadMuL`), `gateShift_clockOne`, `threeRecordFlow_fst_prodRay` (from a product ray on a fibre where the third and fourth profiles vanish the carrier stays a product ray and all three region pointers move as region pointers of the messenger's own arena), `integral_nearOcc_le_self`, `le_integral_nearOcc_out` (on a round fibre the near half has accumulated at least `(1 − fineEta n δ) · roundOut n δ` by the end of the first presence), `roundAway_le_midpoint`, `midpoint_le_roundAwayEnd`, `threeRecordStart`, `threeRecordRun`, `threeRecordRun_fst_pointer`, `threeRecordRun_snd_pointer`, `threeRecordRun_thd_pointer`, `threeRecordRun_clock_pointer`, `echoFstShown_threeRecord_iff`, `echoSiteShown_threeRecord_iff`, `triThdShown_threeRecord_iff`, `isSwing_threeRecord`, ★ `three_records_with_counts` (**from `threeRecordStart` with the three region pointers ready, on a round fibre, `0 < δ ≤ π/2`, the third profile vanishing on its fibre, all three rates positive, the first two rates as in ~~#383~~, and the third rate `κ₃` with `κ₃ T₁ ≤ 1/6` for a chosen `0 ≤ T₁ ≤ roundOut n δ` and `1/3 ≤ κ₃ (1 − fineEta n δ) roundOut n δ`: there are flow times `τ_d ∈ [T₁, roundOut]`, `τ_a ∈ [roundAway, roundAwayEnd]`, `τ_r ∈ [roundBack, roundEnd]`, `τ_d ≤ τ_a ≤ τ_r`, such that the three readouts are `IsRecordWrite`s with write intervals `[T₁, τ_d]`, `[roundAway, τ_a]`, `[roundBack, τ_r]`, the clock's own pointer has exactly `N_d`, `N_a`, `N_r` ticks by them, `|(N_a − N_d) · 2π/R − (τ_a − τ_d)| ≤ 4π/R`, `|(N_r − N_d) · 2π/R − (τ_r − τ_d)| ≤ 4π/R`, the midpoint `(τ_d + τ_r)/2` lies in `[roundAway, roundAwayEnd]`, and `|(N_d + N_r)/2 · 2π/R − (τ_d + τ_r)/2| ≤ 2π/R`**). Contract: UA-17 (f), UD-129, UD-130; UO-16 annotated; `POSITS.md` unchanged. **Scope:** the first write is not an emission record — the messenger is prepared in the near half and the third pointer accumulates its presence there; all three writes are thresholds of accumulated occupancy and certify no departure, arrival or return (carriers with suitable constant occupancies would give readouts with the same clauses: a calculation, not in Lean, not exhibited on this arena); `τ_d`, `τ_a`, `τ_r` are the ends of the stated write intervals, not the instants the readouts turn, each located only within its interval; **the midpoint statement is arithmetic on the three intervals** — for `δ ≤ π/2` the midpoint of any non-negative time up to `roundOut` and any time of the return leg lies in the absence; the two intervals are not symmetric about it (the return leg is twice as long) — and locates the midpoint only to within the absence's whole length `1250 n² (π − 2δ)`, the same interval that holds `τ_a`; it is formed from flow times, no degree of freedom holds it, and it is not a radar time (no reflector, no distance, one ring: the messenger returns because the ring closes; no second observer); the time content is the clock's alone — the bounds are ~~#366~~'s at any non-negative flow times, no hypothesis relates the period to any interval, all three counts may be zero; the counts are read at flow times, no record reads one (UO-14 (b)); each readout is shown to display only through its window after its `τ`, and the three are not shown to display at one flow time — the third pointer keeps turning while the messenger is in the near half; six rate windows and three signs are hypotheses, the first two pairs compatible as in ~~#383~~, the third satisfiable with `T₁ = 0` by any large enough rate once `δ < π/2` and `fineEta n δ < 1` (at `δ = π/2` the theorem is vacuous), and that one start meets all of them with the round-fibre bound is not stated in Lean in this row (→ ~~#387~~: one start exhibited, at `T₁ = 0`); the third and fourth profiles vanish on their fibres and the start is a product ray (null sets by calculations not in Lean); one clock start; **the gated pointer is absent and no term couples the messenger and the clock**, so neither half of the author's decision of 2026-11-12 (the composite carrier, the product coupling `P ⊗ X`) is exercised in this run: from a product ray it is ~~#372~~'s uncoupled arena with a third region pointer (a calculation; Lean has it only where the third and fourth profiles vanish), and the clock is read by its own pointer; the writes and counts are `IsRecordWrite` and `IsTickCount` statements about readout histories, not `ActualOccurrence`s; ~~#367~~ has a departure write and an arrival write in one run, ~~#369~~ three records and two counts in one run for one momentum mode, ~~#372~~ a separate clock's counts at two writes, and ~~#200~~/~~#202~~, ~~#270~~, ~~#274~~/~~#278~~ emission and return records with a clock reading under their own laws — what is new is the third write, during the first presence, in a round-trip run like that of ~~#383~~, and the midpoint clause; flow times (UA-7); nothing here is evidence about Lorentz kinematics. Remainder: (a) a first record that is an emission (a start not prepared in the near half, or a source); (b) a certificate of the round trip as a record, in the manner of ~~#373~~ → ~~#385~~: in part — two readouts certify, under a rate condition, a drop of the function that drives them; not the return, and not held by a record; ~~#386~~: a rise and a dip of means, still not a return and not held by a record; (c) the counts read by records (UO-14 (b)); (d) a second observer; (e) a reflector, so that the return is not the ring's. | ~~**M–L**~~ done in part (M) | Medium-low — three writes and three counts on the free round trip, with no term coupling messenger and clock (in effect the uncoupled arena with a third region pointer); the writes are ~~#365~~'s crossing lemma on the bounds of ~~#379~~, the counts ~~#381~~'s swing, the midpoint is interval arithmetic; not certified, not sharp, not a radar time | ~~#383~~, ~~#382~~, ~~#381~~, ~~#379~~, ~~#376~~, ~~#372~~, ~~#369~~, ~~#367~~, ~~#366~~, ~~#365~~ | > | ~~385~~ | Records | Records→spacetime | *new (2026-11-13, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — part of Remainder (b) of ~~#384~~)* **Two readouts that certify a drop of the occupancy that drives them.** ▸ **DONE 2026-11-13, in part: two readout histories, with known rates and start angles satisfying a rate condition (a hypothesis), exclude every constant drive; a drop of a mean, not a departure, and nothing about the return; UO-16 not met.** [`RecordLayer/UnifiedReadoutCertificate.lean`](../CsdLean4/RecordLayer/UnifiedReadoutCertificate.lean) (7 pins). No new assumption. `lt_of_pshift_mem_displayArc` (a ready pointer that displays after a non-negative turn has turned more than `1/3` minus its start angle), `le_of_forall_pshift_notMem_displayArc` (one that never displays before `T` has turned at most that by `T`), ★ `drop_of_two_readouts` (**two ready pointers turned by `κ_f ∫₀ᵗ F` and `κ_s ∫₀ᵗ F` for one continuous non-negative `F`: if the fast readout displays at some flow time up to `T_o`, the slow one at no flow time before `T_b ≥ T_o`, and `(1/3 − a_s)/κ_s · T_o ≤ (1/3 − a_f)/κ_f · T_b`, then `T_o ∫_{T_o}^{T_b} F < (T_b − T_o) ∫₀^{T_o} F`**), `not_monotone_of_drop` (such an `F` is not non-decreasing, so not constant), `two_roundUpper_le_roundBack` (a corollary of ~~#382~~'s `two_roundUpper_le_roundLower`), ★ `threeRecord_readouts_certify_drop` (**in the run of ~~#384~~ with `T₁ = 0`, `0 < δ < π/2` and the rate condition for the first and third pointers: every continuous non-negative `F` that reproduces the third and the first readout histories at the flow times in `[0, roundBack]`, through pointers of those rates and starts, has a smaller mean over `[roundOut, roundBack]` than over `[0, roundOut]`, and is not non-decreasing**), `threeRecord_nearOcc_drop` (the run's own near-half occupancy is one such `F`). Contract: UD-131, UD-132; UO-16 annotated; `POSITS.md` unchanged. **Scope:** a drop of a mean, with no direction, path, speed or instant — not a departure; nothing shows the mean rising again, so not a return and not a round trip; the premises are readout histories over intervals of flow time ("does not display at any flow time before `T_b`"), held by no record, and no degree of freedom holds the conclusion (UA-7); the form of the drive is assumed — both pointers turned by the accumulated value of one function at known rates from known start angles (UA-5, UA-13, UA-4) — and the theorem does not identify `F` with an occupancy; in the run the coupled carrier's near-half occupancy is one such `F`, and for it the drop also follows without the rate condition from the occupancy bounds of ~~#382~~ and ~~#384~~ for a small margin and a small `fineEta` (a calculation, not in Lean); what is new is that the drop follows from the two histories, the rates, the start angles and the rate condition for every continuous non-negative `F` that reproduces both histories on `[0, T_b]`, so that no constant `F` reproduces them; each single write still certifies nothing; the rate condition is a hypothesis that the rate windows of ~~#384~~ do not imply, and where it fails the constant-occupancy caveat stands — with equal start angles, the third rate at the least value of its window and the first at the greatest (the corner most favourable to the condition; elsewhere it is stronger, and it fails for a large third rate) it is `2 · roundUpper ≤ (1 − fineEta) · roundBack`, weaker than the first window's own compatibility and proved from it for `δ ≤ π/48` and `fineEta n δ ≤ 1/48`; that all conditions hold together for some start is otherwise not in Lean in this row (→ ~~#387~~: one start exhibited); the run is ~~#384~~'s with its limits (no term couples the messenger and the clock, null sets of fibres and starts) and the clock plays no part here; that the generic statement applies off the null sets is expected, not in Lean; ~~#373~~ has a transit certificate in accumulated occupancies with no pointer in it; flow times; nothing here is evidence about Lorentz kinematics. Remainder: (a) the rise on the return leg from readouts → ~~#386~~: done, from the first pointer read on two arcs, under a tightened window; a rise and a dip of means, not a return; (b) the certificate on fibres of positive measure (carried as ~~#386~~ (a)); (c) the premises held by a record (carried as ~~#386~~ (b)). | ~~**M**~~ done in part | Medium — a statement where two readout histories constrain the function that drove them (none earlier found in the programme rows ~~#349~~–~~#384~~), answering, under a rate condition and for two pointers on one region, the constant-occupancy caveat carried since ~~#365~~; a real-analysis statement about two histories, instantiated in one run on null sets; a drop only | ~~#384~~, ~~#382~~, ~~#379~~, ~~#373~~, ~~#367~~, ~~#365~~ | > | ~~386~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Remainder (a) of ~~#385~~)* **Readouts that certify a rise of the occupancy that drives them, and with ~~#385~~ a dip.** ▸ **DONE 2026-11-14, in part: one pointer read on two arcs certifies that its drive accumulated more over the return leg than in all the time before; with the drop of ~~#385~~, the mean over the middle interval is below the mean before and the mean after; under a tightened rate window and, for the dip, a rate condition (hypotheses); a rise and a dip of means, not a return and not a round trip; UO-16 not met.** [`RecordLayer/UnifiedReadoutRise.lean`](../CsdLean4/RecordLayer/UnifiedReadoutRise.lean) (9 pins). No new dynamical assumption; UA-4 extended by one readout (the first region pointer's ready arc at `roundBack`) and the tightened window. ★ `rise_of_ready_then_display` (**a ready pointer turned by `κ ∫₀ᵗ F` for one continuous non-negative `F`: if its readout displays at no flow time before `T_b`, the pointer is in the ready arc at `T_b`, and the readout displays at some flow time up to `T_e`, then `∫₀^{T_b} F < ∫_{T_b}^{T_e} F`**), `mean_rise_of_integral_lt` (if also `T_e − T_b ≤ T_b − T_o`, the mean over `[T_b, T_e]` exceeds the mean over `[T_o, T_b]`), `not_antitone_of_rise` (such an `F` is not non-increasing, so not constant), `five_halves_roundUpper_le_roundLower`, `echoFstReady_regionTwoFlow_iff`, `echoFstReady_threeRecord_iff` (the first pointer's ready arc along the flows, in the accumulated near-half occupancy), ★ `threeRecord_readouts_certify_rise` (**in the run of ~~#384~~ with `T₁ = 0`, `0 < δ < π/2` and `κ₁ · roundUpper ≤ 1/6 − a_θ`: every continuous non-negative `F` that reproduces the first readout history on `[0, roundEnd]` and the first pointer's ready arc at `roundBack` accumulates more over `[roundBack, roundEnd]` than over `[0, roundBack]`, has a greater mean over `[roundBack, roundEnd]` than over `[roundOut, roundBack]`, and is not non-increasing**), ★ `threeRecord_readouts_certify_dip` (**with the rate condition and the third readout history of ~~#385~~ in addition: the mean over `[roundOut, roundBack]` is strictly less than the mean over `[0, roundOut]` and than the mean over `[roundBack, roundEnd]`; `F` is neither non-decreasing nor non-increasing**), `threeRecord_nearOcc_dip` (the run's own near-half occupancy is one such `F`). Contract: UD-133, UD-134; UO-16 annotated; `POSITS.md` unchanged. **Scope:** a rise and a dip of means over stated intervals of flow time, with no direction, path, speed or instant — not a return, not a departure and a return, and no radar time is formed; nothing says that what the pointer registers on the return leg is what was there before; the premises are a display history over an interval and an arc membership at one flow time (UA-7), held by no record, and no degree of freedom holds the conclusion; each single display history still certifies nothing — the rise uses one more premise about the same pointer, its ready arc at the flow time `T_b`, and the threshold writes of the region pointers have read only the displayed arc after the start (tick records have read a ready arc at later flow times since ~~#358~~, but not this pointer's), so that is a further apparatus input, recorded in UA-4; the form of the drive is assumed (UA-5, UA-13) and `F` is not identified with an occupancy; for the run's near-half occupancy the rise also follows from the occupancy bounds of ~~#382~~ for a small margin and a small `fineEta` (a calculation, not in Lean); the rise needs no rate condition between pointers but a tighter window, `κ₁ · roundUpper ≤ 1/6 − a_θ` in place of `≤ 1/6`, which with `1/3 ≤ κ₁ · roundLower` asks `roundLower / roundUpper ≥ 2/(1 − 6 a_θ)` — the proved ratio `5/2` (for `δ ≤ π/96`, `fineEta n δ ≤ 1/96`) is enough for `a_θ ≤ 1/30` (a calculation), while the ratio is below `3` for every `n > 0`, `δ > 0`, so for `a_θ ≥ 1/18` the conditions cannot both hold and the run theorems are vacuous (a calculation); the dip needs in addition the rate condition of ~~#385~~, which the rate windows do not imply and the tightened window works against (most favourable corner `2 · roundUpper ≤ (1 − fineEta)(1 − 6a) · roundBack`, not covered by `two_roundUpper_le_roundBack`; a calculation); the far and third pointers' hypotheses are only carried in the rise; that one start meets all the conditions is not in Lean in this row (→ ~~#387~~: one start exhibited), and `threeRecord_nearOcc_dip` shows only that the class of functions is not empty when they are met; the run is ~~#384~~'s with its limits (no term couples the messenger and the clock, null sets of fibres and starts) and the clock plays no part here; that the generic statement applies off the null sets is expected, not in Lean; flow times; nothing here is evidence about Lorentz kinematics. Remainder: (a) the certificates on fibres of positive measure; (b) the premises held by a record; (c) the dip located in flow time (it compares three means and names no instant). | ~~**S–M**~~ done in part | Medium-low — the companion of ~~#385~~ for the return leg; together the readouts of one run give that the mean of their drive over the middle interval is strictly below its mean over the interval before and the interval after; real-analysis statements about readouts at flow times, instantiated in one run on null sets; no record holds them | ~~#385~~, ~~#384~~, ~~#382~~, ~~#367~~, ~~#364~~ | > | ~~387~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — the satisfiability caveat carried by ~~#382~~–~~#386~~)* **One start that meets every hypothesis of the three records and of the dip.** ▸ **DONE 2026-11-14: an explicit choice of profile, start angles and strengths satisfies the rate windows, the round-fibre bound, the tightened window and the rate condition together, so the statements of ~~#384~~, ~~#385~~ and ~~#386~~ are not vacuous; satisfiability at one point and nothing more; UO-16 not met.** [`RecordLayer/UnifiedDipInstance.lean`](../CsdLean4/RecordLayer/UnifiedDipInstance.lean) (12 pins). No new assumption. `one_sub_cos_pi_div_ninetySix`, `fineEta_le_of_le` (at `δ = π/96` every `n ≥ 73744` has `fineEta n δ ≤ 1/96`), `dipAngle` (moving angle `1/30`, fibre `0`), `dipAngle_snd`, `rep_dipAngle`, `dipAngle_mem_readyArc`, `rateAt_trainF_dipAngle` (the rate is twice the strength), ★ `dipInstance_windows` (**for `0 < δ ≤ π/96`, `fineEta n δ ≤ 1/96` and the strengths `1/(6 · roundLower)`, `1/(6 · farLower)`, `1/(6 (1 − fineEta) roundOut)`: every rate hypothesis of ~~#384~~ at `T₁ = 0`, of ~~#385~~ and of ~~#386~~**), ★ `dipInstance_three_records` (the conclusion of `three_records_with_counts`), ★ `dipInstance_readouts_certify_dip` (the conclusion of `threeRecord_readouts_certify_dip` for every `F` that reproduces the stated readouts), ★ `dipInstance_nearOcc_dip`, ★ `exists_nearOcc_dip` (**there are a ring, an index and two strengths for which the coupled carrier's near-half occupancy has the dip at `δ = π/96`; no hypothesis; on that fibre the profiles vanish and the carrier moves as the free packet**). Contract: UD-135, UD-136; UO-16 annotated; `POSITS.md` unchanged. **Scope:** satisfiability only — the conclusions are those of ~~#384~~–~~#386~~ word for word, with all their limits (threshold writes that certify nothing, counts at flow times that may be zero, a dip of means, premises that are readouts at flow times held by no record, no radar time); one start, a null set: the three pointers at one point, the windows met with equality at their lower ends (the edge of each window), one product ray, one clock start, and positive measure is not stated in Lean; on the fibres `0` all three `trainF` profiles vanish, so no region pointer acts back on the messenger — it moves as the free packet and the round-fibre bound holds trivially, so the back-reaction bounded in ~~#379~~ is not exercised; the strengths are reciprocals of the occupancy bounds, so they are tuned to the round trip's schedule in flow time (apparatus input, UA-4); the index bound `73744` is not sharp and with `640000 n³ ≤ M + 1` means a ring of more than `2 · 10²⁰` sites (a calculation), with no scale attached; the three instance theorems keep the ring's length, the index bound and the clock's parameters as hypotheses (each plainly satisfiable, not stated as one existential) and the certificate keeps its premises about `F`; only `exists_nearOcc_dip`, which involves no pointer and no clock, has none; the clock plays no part; nothing here is evidence about Lorentz kinematics. Remainder: (a) the hypotheses on a set of starts of positive measure (an open set of angles, strengths inside the windows, fibres where the pointers act back); (b) one closed existential for the three records and the certificate. | ~~**S**~~ done | Low-medium — removes the "not shown satisfiable" caveat from three rows (~~#384~~–~~#386~~, for their rate hypotheses and the round-fibre bound; the theorems of ~~#382~~ and ~~#383~~ are not instantiated); real arithmetic and one application of each theorem; adds no statement about records | ~~#386~~, ~~#385~~, ~~#384~~, ~~#383~~, ~~#382~~, ~~#379~~ | > | ~~388~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (2) of the plan of 2026-11-14: the probe of the moving observer)* **A clock packet at a reference momentum: a dephasing bound linear in the horizon.** ▸ **DONE 2026-11-14: a conditional swing-and-count bound for a localised packet of the clock carrier, up to a horizon limited linearly by its dephasing; no packet of positive width is shown to meet the conditions and the horizon is not shown to contain one tick; a clock, not yet an observer; the probe found no obstruction and is not decided; UO-16 not met.** [`RecordLayer/UnifiedClockPacket.lean`](../CsdLean4/RecordLayer/UnifiedClockPacket.lean) (13 pins). No new assumption. `clockHand_sum`, ★ `abs_clockHand_sub_le` (**for rates `R₁, R₂ ≥ ρ > 0`: `|clockHand R₁ C t − clockHand R₂ C t| ≤ ‖C‖ |R₁ − R₂| (2t/ρ + 3/ρ²)`**), `modeRate`, `modeCoh`, `bandE_sub_eq_neg_modeRate`, ★ `deviceCoherence_packet_shift`, `deviceCoherence_packet_eq_sum`, ★★ `integral_clockFreeRead_packet` (**the packet's accumulated uncoupled reading is exactly the sum of its modes' hands**), `packetSwingErr`, ★★ `abs_clockPointerShift_packet_sub_swing_le`, ★★★ `isSwing_clockPointerFlow_packet` (**on a fibre where the profile vanishes, a clock packet's pointer is an `IsSwing` up to `T` with an error linear in `T`**), `sum_norm_modeCoh_mul_le`, ★★★ `clock_packet_ticks` (**for the localised packet of ~~#262~~'s device with strength `R/(16 Im C₀)`: if `κ > 0` and `R ε (2T/ρ + 3/ρ² + π/(2R²)) < κ/3`, a swing up to `T` and `|2π N/(μ₁ − μ₀) − √(1 − (v/c)²) T| ≤ 2π/(μ₁ − μ₀) + ((μ₁ − μ₀)/E₀) T` for every tick count `N` at `T`**). Contract: UD-137, UD-138; UO-16 annotated; `POSITS.md` unchanged. **Scope:** a clock, not an observer — the arena is the clock arena of ~~#358~~ (the clock carrier and its own pointer), with no messenger, no detector, no meeting; the coincidence detector of the author's decision of 2026-11-14 is not built; the conditions (`κ > 0`, the inequality, a lower bound `ρ` of the rates) are hypotheses, explicit and linear in `T`; one mode meets them at every horizon and no packet of positive width is exhibited with numbers; that `ρ` follows from the window is a calculation, not in Lean; one fibre (`trainF` at `0`, where the pointer does not act back on the clock) and one start, a null set, and off that fibre ~~#358~~'s quadratic back-reaction term remains; `v` is the relativistic velocity at the reference momentum — that the packet moves at it is not shown here and no position of the clock is recorded; the count is compared with the flow time (UA-7); the dilation error is the finite splitting, as in ~~#358~~, and does not shrink with the packet's width; that the horizon (of order `1/ε`) is long enough for a radar exchange, the packet travelling of order `E/(μ₁ − μ₀)` of its own widths within it, is a calculation, not in Lean; ~~#358~~ has the same statement with the error `ε T²` on every fibre, ~~#273~~ the sharp drift on another arena, ~~#249~~ the dilation bound — what is new is the exact sum over modes and the linear bound; nothing here is evidence about Lorentz kinematics. Remainder: (a) a packet of positive width exhibited with numbers, with `ρ` from the window → ~~#391~~ (done, with a count of at least `n`); (b) the coincidence detector (step (3) of the plan); (c) the packet's position as a record; (d) the linear bound off the fibre where the profile vanishes. | ~~**M**~~ done | Medium — closes the gap ~~#358~~ stated about itself, for the object the author chose as the moving observer's clock; a statement about one clock on one fibre, with conditions as hypotheses | ~~#358~~, ~~#357~~, ~~#273~~, ~~#262~~, ~~#249~~ | > | ~~389~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, first brick)* **The five-pointer arena.** ▸ **DONE 2026-11-14: the manifold, form, energy, flow and invariant measure of the bare arena of UA-18; no carrier, coupling or record of UA-18 is defined on it yet; UO-16 not met.** [`LF4/PentaCoupledFlow.lean`](../CsdLean4/LF4/PentaCoupledFlow.lean) (14 pins). No new assumption beyond UA-18 (d). `PentaModel`, `PentaSigma`, `pentaForm`, `pentaForm_isSymplectic`, `quadRayEnergy`, `contMDiff_quadRayEnergy`, `fiveCoupledEnergy`, `fiveCoupledEnergy_apply`, `contMDiff_fiveCoupledEnergy`, `fourCoupledEnergy_add_smul_quadRayEnergy`, `fiveCoupledFlow`, ★ `fiveCoupledFlow_apply` (**a skew product over the fifth conserved angle: the four-torus flow with `A + r(ζ₂) B_E`, the fifth moving angle advanced by `r'(ζ₂)` times the accumulated `−2⟨B_E⟩`**), `fiveCoupledFlow_fst_of_lift_eq_zero`, `fiveCoupledFlow_snd_of_lift_eq_zero`, `pentaMuL`, ★ `pentaMuL_map_fiveCoupledFlow` (**the flow preserves `μ_FS ⊗ vol⁵` for all Hermitian matrices**). Contract: UD-139; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** one more application of the step theorem of ~~#368~~, as ~~#381~~'s four-torus arena was; no arena indexed by the number of tori is defined; the matrices are arbitrary Hermitian matrices on the ray's space — the three carriers, the walls, the coincidence projectors, the clock packets and the five readings of UA-18 are not defined, and no record is read; the coupling's form is assumed (each pointer's term a ray observable times a function of its conserved angle, no term between pointers); the measure statement is invariance only. Remainder (the rest of step (4)): (a) the three-carrier composite and the coincidence projectors → ~~#390~~: the projectors and gated operators for two carriers; the three-carrier composite still open; (b) the chain with walls; (c) one observer's emission and return readings. | ~~**S**~~ done | Low — infrastructure: the step theorem once more; needed because UA-18 has five pointers | ~~#381~~, ~~#368~~ | > | ~~390~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, second brick)* **Coincidence gates between two carriers.** ▸ **DONE 2026-11-14: the coincidence projector and the gated operator of UA-18 (c), with their algebra and their expectations on product rays; operators only, no flow, pointer or record; UO-16 not met.** [`RecordLayer/UnifiedCoincidence.lean`](../CsdLean4/RecordLayer/UnifiedCoincidence.lean) (15 pins). No new assumption beyond UA-18. `arenaObs_zero'`, `arenaObs_finset_sum`, `siteSel`, `siteSel_isHermitian`, `coincProj`, `coincProj_isHermitian`, `coincProj_mul_self`, `one_sub_coincProj` (the complement is the coincidence projector of the negated relation), ★ `coincProj_eq_sum` (**a sum over related pairs of sites of products of site projectors**), ★★ `arenaObs_coincProj_prodRay` (**on a product ray: the sum over related pairs of the product of the two carriers' site occupancies**), `gateOp`, ★ `coincProj_mul_one_kronecker` (it commutes with every operator of the second carrier that does not move it between sites), `gateOp_isHermitian`, ★ `gateOp_eq_sum`, ★★ `arenaObs_gateOp_prodRay` (**on a product ray: for each related pair, the first carrier's occupancy of its site times the second carrier's reading of `Y` at its site**). Contract: UD-140; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** operators only — no flow, pointer, reading or record; the definitions take any decidable relation between sites, an order relation included, and the restriction to coincidence within a fixed width is the contract's (UA-18 (c)), not the module's; two carriers, where UA-18 has three (a gate between two of three is to come from taking one factor to be a product; not done); the product-ray formulas are about product rays, and a gated term that acts back joins the two carriers, no such flow is defined, and under one a product ray is expected to stay one only on a fibre where the pointer's profile vanishes (as in ~~#376~~), which is not shown — the formulas say nothing about an entangled ray; ~~#376~~ has `P ⊗ X` for one region projector. Remainder: (a) the gate between two of three carriers; (b) the flow with a gated term from a product ray, on and off the fibre where the profile vanishes → ~~#393~~ on the fibre; off it still open. | ~~**S**~~ done | Low — definitions and algebra for the one kind of coupling UA-18 admits | ~~#389~~, ~~#376~~ | > | ~~391~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (2) of the plan of 2026-11-14, Remainder (a) of ~~#388~~, after the review found no tick and no packet of several modes shown)* **Narrow clock packets tick as often as wanted.** ▸ **DONE 2026-11-14: for every window of modes within `1/(600 n)` of a reference momentum the conditions of ~~#388~~ hold at the horizon `2π n/R`, the count there is at least `n`, and as a rate it is within `1/n + Δ` of `√(1 − v²)` with `v = σ/E₀ ≥ p₀/π`; windows of any odd number of modes exhibited; one flow time, one fibre; a clock, not an observer; UO-16 not met.** [`RecordLayer/UnifiedClockPacketInstance.lean`](../CsdLean4/RecordLayer/UnifiedClockPacketInstance.lean) (22 pins). No new assumption. ★★ `clock_packet_count_ge` (under the hypotheses of `clock_packet_ticks`, `R T ≥ 2π (n − 1) + π` gives a count `N ≥ n` at `T`), `half_le_levelClockRate`, `levelClockRate_le_gap` (levels `1` and `1 + Δ`: rate in `[Δ/2, Δ]` for `s² ≤ 3`), `ringLen`, `ringHop`, `ringSigma`, `ringRate`, `ringLen_pos`, `two_ringHop`, `abs_ringSigma_le`, `ringSigma_pos`, `ringSigma_sq_le`, `two_ringHop_mul_sin_angleZ`, `dispE_one_le_two`, `two_div_pi_mul_le_ringSigma`, ★ `div_pi_le_vRel_ringSigma` (in the lower half of the band `p/π ≤ σ/E₀`), `abs_div_horizon_sub_le`, ★★★ `narrow_clock_packet_ticks` (**ring of length `2π K`, `0 ≤ p₀ ≤ 9/10` (weakened from `0 < p₀` on 2026-11-15) with `6 K p₀ ≤ M + 1`, any window within `1/(600 n)` of `p₀`, any normalised amplitudes, `0 < Δ ≤ 1/2`: for some strength, at `T = 2π n/R` the count exists, is at least `n`, `|2π N/Δ − √(1 − v²) T| ≤ 2π/Δ + (Δ/E₀) T`, `|2π N/(Δ T) − √(1 − v²)| ≤ 1/n + Δ`, `v ≥ p₀/π`**), `modeWindow`, `flatAmp`, `sum_norm_flatAmp`, ★★ `exists_clock_packet_ticks` (**the `2 w + 1` modes centred on `600 n (w + 1)` of the ring of length `2π · 1200 n (w + 1)`, equal amplitudes: no hypothesis but the sizes**). Contract: UD-141, UD-142; UO-16 annotated; `POSITS.md` unchanged. **Scope:** a clock, not an observer — what a moving observer is in `Σ` is not settled here; one count at one flow time; `n` ticks need momentum width at most `1/(600 n)`, so for a fixed packet the number of ticks shown is bounded and nothing is said after the horizon; localisation in space is not shown; whether the dilation is resolved depends on `n` and `Δ` (`1/n + Δ` against `1 − √(1 − v²)`), and no instance here makes the comparison (~~#392~~ does); `v` is `σ/E₀` at the reference momentum, the band's group velocity is `v cos(a p₀) ≥ v/2` (a calculation, not in Lean), and that the packet moves is not shown; the strength exists and is not recorded in the statement (the proof uses `R/(16 Im C₀)`, tuned to the packet, UA-4); one fibre (a null set) and one start; the count is compared with the flow time and `T` is chosen with the rate known (UA-7); speed `1`, lower level `1`, constants not sharp. Remainder: (a) the clock packet's occupancy of a region bounded (localisation), needed for the coincidence detector → ~~#392~~ (a bound at a coarse resolution); (b) whether a radar exchange fits inside the horizon (step (6)); (c) off the fibre where the profile vanishes; (d) the count at every flow time of the horizon, not one. | ~~**M**~~ done | Medium — the three conditions of ~~#388~~ discharged, a positive count shown, the bound as a rate with a lower bound of `v` | ~~#388~~, ~~#358~~ | > | ~~392~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (2) of the plan of 2026-11-14, Remainder (a) of ~~#391~~; bears on the risk carried to step (6))* **A clock packet that ticks: where it is, and its dilation resolved.** ▸ **DONE 2026-11-14: for `N` flat modes the clock ticks at least `n` times; for `n ≥ 400`, `Δ ≤ 1/400` its count at the horizon is at most `(1 − 1/400)` of the undilated count; and at every packet time of the horizon the uncoupled packet's weight outside the arc of half-angle `δ` about its two levels' predicted positions is at most `(1/N + 1/(20000 Δ n N))/(1 − cos δ)`; a weight of the state vector, at a coarse resolution, no region exhibited; UO-16 not met.** [`RecordLayer/UnifiedClockPacketLocalised.lean`](../CsdLean4/RecordLayer/UnifiedClockPacketLocalised.lean) (19 pins). No new assumption. `deviceRegionWeight`, `deviceRegionWeight_nonneg`, `regionWeight_const_mul`, ★ `deviceRegionWeight_devicePacket` (the device's weight on a region is the mean of its two levels' weights), ★★ `deviceRegionWeight_devicePacket_le` (**any device packet: weight on a region away from both levels' predicted directions, times `1 − γ`, at most the mean of the two budgets `1 − Re T_i(0) + |t| · limitErr_i`**), `flatWindow`, `flatAmpN`, `sum_norm_flatAmpN`, `flatWindow_mem_range`, `flatWindow_width`, ★★ `flat_clock_packet_ticks` (the flat window meets `narrow_clock_packet_ticks`, `v ≥ 1/(2π)`), `flat_packet_start_le`, `flat_packet_limitErr_le`, ★★ `flat_packet_budget` (budget at most `1/N + 1/(20000 Δ n N)` for `|t| ≤ 4π n/Δ`), ★★★ `flat_clock_packet_localised` (**the uncoupled packet at those packet times: weight on `B` times `1 − γ` at most `1/N + 1/(20000 Δ n N)`**), `abs_sub_quarter_turn_le`, ★★ `flat_clock_packet_localised_flow_time` (the same at `−π/(2R) + s` for every flow time `s` of the horizon), `sqrt_one_sub_sq_le`, ★★★ `flat_clock_packet_rate_lt` (**`n ≥ 400`, `Δ ≤ 1/400`: the count `N' ≥ n` at `T` has `2π N'/Δ ≤ (1 − 1/400) T`**). Contract: UD-143, UD-144; UO-16 annotated; `POSITS.md` unchanged. **Scope:** the localisation is a weight of the uncoupled state vector at a packet time — not a record, no expectation of a region projector along `clockPointerFlow` is stated, and that the flow's ray on the fibre where the profile vanishes is this packet is not a named theorem (~~#388~~ uses it at the level of the coherence); no region is exhibited — the hypotheses hold for the empty one, and the bound says something only when `1 − γ` exceeds the budget; the resolution is coarse — arcs of half-angle large against the square root of the budget (`1/√N` when `20000 Δ n ≥ 1`), while the packet's width is of order `1/N` (a calculation); the two levels have two predicted positions, at most `π/(300 N)` apart over the horizon (a calculation); how far the clock travels is not stated — `π v/(600 R N)`, about `0.003/(Δ N)`, which exceeds the resolved arc for `Δ √N` small against `0.002` and `n` large against `1/Δ` (a calculation, not in Lean); the resolved dilation is one count at one flow time (UA-7), with `v = σ/E₀` and no motion shown; the predicted velocity in the localisation bound is the continuum one and the lattice needs `M ≥ 210000 n² N` sites; a clock, not an observer; constants not sharp. Remainder: (a) the localisation as an arena observable along the flow, and with a second coupling present; (b) the travel in Lean: a family in which the predicted position crosses a resolved arc within the horizon, with a region exhibited; (c) a sharper resolution (the packet's own width). | ~~**M**~~ done | Medium — the first bound on where the moving observer's clock is, and the first instance in which the count resolves the dilation | ~~#391~~, ~~#388~~, ~~#375~~, ~~#265~~, ~~#262~~ | > | ~~393~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, third brick; Remainder (b) of ~~#390~~)* **Two gated pointers on a composite carrier, from a product ray.** ▸ **DONE 2026-11-14: two pointers coupled to observables of a composite of two carriers in one flow; from a product ray on a fibre where both profiles vanish the carrier stays a product ray and each pointer advances by the accumulated expectation of its observable; for a gate and its complement the two displacements add up to the ungated one; displacements, not records; UO-16 not met.** [`RecordLayer/UnifiedCoincidenceFlow.lean`](../CsdLean4/RecordLayer/UnifiedCoincidenceFlow.lean) (12 pins). No new assumption beyond UA-18. `idxFreeU_composite`, ★ `idxTwoCarrierU_of_lift_eq_zero` (on a fibre where both profiles vanish the carrier moves by its uncoupled propagator), `coincShift`, `coincFlow`, `arenaKick_idxFreeU_composite`, ★★ `coincFlow_prodRay` (**the flow from a product ray on such a fibre: product ray kept, each carrier free, each pointer advanced by `−2 z ∫₀ᵗ ⟨G⟩`**), `continuous_arenaObs_prodRay_free`, ★ `coincShift_add`, `coincShift_one_kronecker`, ★ `gateOp_add_gateOp_not` (`gateOp r Y + gateOp (¬r) Y = 1 ⊗ Y`), ★★ `coincShift_gateOp` (the gated displacement site by site), ★★★ `coincShift_gateOp_add_not` (**a pointer gated on a relation and one gated on its complement, with one profile derivative: their displacements add up at every flow time to that of a pointer coupled to `Y` on the second carrier alone**). Contract: UD-145, UD-146; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** displacements, not records — no swing, tick, count or reading, and tick counts are not shown to add; only fibres where both profiles vanish and only product starts, both null sets (two calculations, not in Lean), and off them a gated term joins the carriers and nothing is proved; the last theorem needs the same profile derivative for both pointers; any two Hermitian observables in `coincFlow_prodRay`, and for a gate ~~#390~~'s `gateOp_isHermitian` supplies that — no carrier, clock or relation is instantiated; two carriers and two pointers where UA-18 has three, five and walls; that the flow preserves the product measure follows from ~~#359~~ and is not restated; ~~#376~~ has the same for one product observable. Remainder: (a) an instance: the Dirac messenger, the two-level clock, coincidence within a width, and the gated pointer as a swing while the two are together (the route of ~~#377~~, ~~#378~~) → ~~#394~~ has the reading under hypotheses on where the carriers are; the carriers themselves still open; (b) three carriers; (c) off the fibres. | ~~**S–M**~~ done | Low-medium — the flow for the couplings of UA-18 (c) in the one case where it is solved exactly, and the identity behind UA-18 (d) | ~~#390~~, ~~#376~~, ~~#359~~ | > | ~~394~~ | Records | Records→spacetime | *new (2026-11-14, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, fourth brick; Remainder (a) of ~~#393~~, the part that needs no carrier)* **A gated reading: the number of ticks fixed after the gate closes.** ▸ **DONE 2026-11-14: for the flow of ~~#393~~ with its last pointer coupled to a gate, hypotheses on where the two carriers are (weight outside the relation small up to `t_a`, weight inside it small on `[t_b, t_c]`) and on the clock's own swing give the pointer a number of ticks that, at every flow time of `[t_b, t_c]`, is one of two consecutive numbers within a stated bound of `R t_a / 2π`; conditional, no carriers put in; UO-16 not met.** [`RecordLayer/UnifiedGatedReading.lean`](../CsdLean4/RecordLayer/UnifiedGatedReading.lean) (13 pins). No new assumption beyond UA-18. `hasTicks_snd_le_of_travel`, ★★ `gated_pointer_count` (**one pointer's history: a swing up to `t_a`, speed at most `v` on `[t_a, t_b]`, at most a sixth of a turn inside `[t_b, t_c]` ⇒ a count `N_a` with `|2π N_a − R t_a| ≤ 2π`, a count `N_b ≤ N_a + 3 v (t_b − t_a) + 3/2`, and `N_b` or `N_b + 1` throughout `[t_b, t_c]`**), `arenaObs_siteSel`, `arenaObs_siteSel_nonneg`, `sum_arenaObs_siteSel`, `arenaObs_coincProj_prodRay_nonneg`, `arenaObs_coincProj_add_not`, ★ `arenaObs_coincProj_prodRay_le_of_disjoint` (apart regions: the weight inside the relation is at most the two carriers' weights outside their regions), ★ `abs_arenaObs_siteSel_mul_levelSwap_le` (the level swap is bounded site by site), ★★ `abs_arenaObs_gateOp_prodRay_le`, `abs_arenaObs_gateOp_prodRay_sub_le` (on a product ray a gated reading is at most the weight inside the relation and differs from the ungated reading by at most the weight outside it), `abs_coincShift_sub_le`, ★★★ `coincFlow_gated_reading` (**the reading: every count `N` in `[t_b, t_c]` has `|2π N − R t_a| ≤ 2π (6|z| (t_b − t_a) + 7/2)`**). Contract: UD-147, UD-148; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** conditional — the two weight hypotheses and the clock's swing are hypotheses, no messenger, clock packet, relation or region is put in, and they are not shown to hold together for any; between `t_a` and `t_b` the count may advance by up to `6|z| (t_b − t_a) + 3/2`, small only if the gate closes quickly against the pointer's speed; nothing after `t_c`; counts of passages, with the record write of a tick from ~~#358~~ up to `t_a` only and not restated; flow times, no speed, distance or second observer; only fibres where both profiles vanish and product starts, null sets; two carriers; ~~#377~~ and ~~#381~~ have the same three phases for a region projector with one-mode clocks. Remainder: (a) the carriers: a messenger packet and a clock packet with regions exhibited (Remainder (b) of ~~#392~~), so that the hypotheses hold → ~~#395~~ has the two weight hypotheses for two packets on a ring; the clock's swing with a margin and the assembled reading still open; (b) the wall (the mirror lemma, parked); (c) three carriers; (d) a record write for the ticks after `t_a`. | ~~**M**~~ done | Medium — the one statement all five readings of UA-18 (d) are to be applications of | ~~#393~~, ~~#390~~, ~~#381~~, ~~#378~~, ~~#358~~ | > | ~~395~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, fifth brick; Remainder (a) of ~~#394~~)* **The gate opens and closes: a messenger packet and a clock packet on one ring.** ▸ **DONE 2026-11-15: with the relation "within the angle `w`" on the ring, the two weight hypotheses of ~~#394~~ hold with an explicit bound for a messenger packet and a clock packet whose predicted angles are near (gate open) or apart (gate closed); a pair is exhibited; ~~#394~~ is not applied and no reading is stated; UO-16 not met.** [`RecordLayer/UnifiedGateRegions.lean`](../CsdLean4/RecordLayer/UnifiedGateRegions.lean) (21 pins). No new assumption beyond UA-18. `cos_add_le_cos_add` (the triangle inequality of the circle, in cosines), `ringNear`, `ringArc`, `mem_ringArc`, ★ `ringNear_of_mem_ringArc`, ★ `not_ringNear_of_mem_ringArc` (**arcs of half-angle `δ` with centres within `w − 2δ` hold only related pairs; more than `w + 2δ` apart, none**), `one_sub_sum_arenaObs_siteSel`, `sum_arenaObs_siteSel_free_chainRay`, `arenaObs_siteSel_device`, `sum_arenaObs_siteSel_device_mk`, `sum_arenaObs_siteSel_free_deviceRay` (the occupancies of ~~#394~~ along the uncoupled orbits are the region weights of ~~#265~~ and ~~#392~~), ★★ `ringGate_not_le`, ★★ `ringGate_le` (for any two rays: the weight outside (inside) the relation is at most the two carriers' weights outside their arcs, when the arcs are near (apart)), `ringPacket_ne_zero`, ★★ `one_sub_sum_arenaObs_ringArc_ringPacket_le`, ★★ `one_sub_sum_arenaObs_ringArc_devicePacket_le` (a messenger packet stays in the arc about its predicted angle, a clock packet in the two arcs about its levels' predicted angles, up to the budgets of ~~#265~~ and ~~#392~~ over `1 − cos δ`), `flatBudget`, `flat_packet_budget_time`, `flat_outside_arcs_le`, ★★★ `flat_gate_open`, ★★★ `flat_gate_closed` (**the flat clock packet of ~~#392~~ and a flat messenger of rest energy `μ ≥ 1` on the same ring: (weight outside the relation) `· (1 − cos δ)` is at most the sum of the two budgets `1/N + 1/(4 (1200 n N)²) + |t|/(360000 n² N)` at the packet times while the predicted angles are within `w − 2δ`, and the same for the weight inside while they are more than `w + 2δ` apart**). Contract: UD-149, UD-150; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** two of the three hypotheses of ~~#394~~; the clock's swing with a margin and the conditions tying the bound to the number of ticks are not discharged and `coincFlow_gated_reading` is not applied; weights along uncoupled orbits, not records; the conditions are on predicted angles at one flow time, no `t_a` or `t_b` is exhibited and the interval between open and closed is not bounded (by a calculation, not in Lean, `4δ` over the difference of the angular velocities); the flat messenger has the clock's momenta and a rest energy at least `1`, so it is no faster than the clock and is not the messenger of M6; `δ` large against `1/√N`; a ring and two carriers. Remainder: (a) the emission reading assembled: the clock's swing with a margin (from the proof of ~~#388~~), explicit flow times `t_a`, `t_b`, `t_c`, and ~~#394~~ applied → ~~#396~~, with the flow times and sizes as inequalities and no values exhibited; (b) a messenger at a speed within a stated error of `1` (its own window; the general lemmas here take it); (c) the wall; (d) three carriers. | ~~**M**~~ done | Medium — the gate hypotheses of the reading hold for exhibited carriers | ~~#394~~, ~~#392~~, ~~#390~~, ~~#265~~ | > | ~~396~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, sixth brick; Remainder (a) of ~~#395~~)* **The emission reading: a gated pointer's count for a messenger packet and a clock packet.** ▸ **DONE 2026-11-15: the third hypothesis of ~~#394~~ (the clock's swing with a margin) is proved for the clock packets of ~~#391~~, and ~~#394~~ is applied to any messenger packet and the flat clock packet on one ring: under explicit inequalities between the parameters the gated pointer's number of ticks is one of two consecutive numbers throughout an interval while the two are apart; no values are shown to meet the inequalities; UO-16 not met.** [`RecordLayer/UnifiedEmissionReading.lean`](../CsdLean4/RecordLayer/UnifiedEmissionReading.lean) (12 pins). No new assumption beyond UA-18. `scaledTrainF`, `scaledTrainF_periodic`, `contDiff_scaledTrainF`, `scaledTrainF_lift_zero`, `scaledTrainF_deriv_lift_zero` (the profile `g · trainF` carries the strength), ★★ `clock_packet_shift_le` (**under ~~#388~~'s hypotheses and a lower bound `κ` of its margin: `g ≤ R/(16κ)` and the clock pointer's displacement is within `R ε (2T/ρ + 3/ρ² + π/(2R²))/(4κ)` of `(1/4)(1 − cos R t)`**), ★★ `narrow_clock_packet_swing`, `flat_clock_swing` (for every window of ~~#391~~ and the flat window of ~~#392~~: `0 < g ≤ R/5` and the displacement within `1/24` of the swing up to `2π n/R`), `clockPointerShift_trainF_eq`, `outside_arcs_flat_clock_le`, `ringGateRun`, ★★★ `flat_clock_emission_reading` (**a count `N_a` at `t_a` with `|2π N_a − R t_a| ≤ 2π`, a count `N_b ≤ N_a + 6 g (t_b − t_a) + 3/2` at `t_b`, and at every flow time of `[t_b, t_c]` a count `N` that is `N_b` or `N_b + 1` with `|2π N − R t_a| ≤ 2π (6 g (t_b − t_a) + 7/2)`**). Contract: UD-151, UD-152; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** the conditions are inequalities between real numbers (the two budgets against `ε (1 − cos δ)`, the predicted angles near up to `t_a ≤ 2π n/R` and apart on `[t_b, t_c]`, `2 (R/5) ε t_a < 1/24`, `2 (R/5) ε (t_c − t_b) ≤ 1/6`) and no values are shown to meet them together; any messenger packet of `dirac_packet_limit` on the clock's ring, none exhibited, its speed not assumed; the count is compared with the flow time `t_a`, at which nothing is recorded; not yet the reading at an emission (on the ring the two start together and separate once); `6 g (t_b − t_a)` is the closing of the gate; counts of passages; nothing after `t_c`; only the fibres where both profiles vanish and product starts, null sets; for some strength, tuned to the clock packet; a ring and two carriers. Remainder: (a) the wall (the mirror lemma, parked) and the return reading: the count of the complementary gate over the interval the messenger is away; (b) the radar time and distance from the two counts; (c) values: a messenger at a speed within a stated error of `1` and a choice of the parameters meeting the inequalities, along a family → ~~#397~~, for the ring with two carriers; (d) three carriers. | ~~**M**~~ done | Medium — the first count of a run of the coupled flow with carriers put in | ~~#395~~, ~~#394~~, ~~#391~~, ~~#388~~ | > | ~~397~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, seventh brick; Remainder (c) of ~~#396~~)* **Values for the emission reading: a fast messenger and a family with a shrinking error.** ▸ **DONE 2026-11-15: the inequalities of ~~#396~~ are met along a family indexed by `k`: the messenger's velocity is within `1/(2k²)` of `1`, the gated pointer's count is of at least about `47 k` ticks and differs from `R t_a / 2π` by a relative error at most `15/k`; the only hypothesis left is the size of the ring; a ring with two carriers; UO-16 not met.** [`RecordLayer/UnifiedEmissionValues.lean`](../CsdLean4/RecordLayer/UnifiedEmissionValues.lean) (23 pins). No new assumption beyond UA-18. `sq_div_five_le_one_sub_cos`, `cos_le_cos_of_abs_le`, `cos_lt_cos_of_between`, ★ `vRel_half_light` (at the momentum `1/2` a rest energy `0 < μ ≤ 1/2` gives a velocity in `[7/10, 1]` within `2μ²` of `1`), `vRel_half_one`, `emission_times`, `emission_angles`, `emission_angles_near`, `emission_angles_apart` (the kinematics of the two packets, in real numbers), `limitErr_le_inv_mul`, ★ `flat_messenger_budget` (the flat messenger's budget for `0 < μ ≤ 1`), `emT0`, `emTa`, `emTb`, `emBudget`, `emission_budget_le`, ★★★ `light_messenger_emission_reading` (**under `800 Δ N ≤ 1`, `1 ≤ 150 n N δ`, `emBudget ≤ ε δ²/5`, `1920 Δ n N ε < 1/24`: at every flow time of `[t_b, (3/2) t_b]` the count `N'` has `|2π N' − R t_a| ≤ 2π ((22 δ + 29 Δ) R t_a + 7/2)`, and `R t_a ≥ 300 Δ n N`**), `emn`, `emN`, `emGap`, `emBudget_family_le`, `emission_family_error`, ★★★ `emission_reading_family` (**for every `k ≥ 1`, `n = 10⁷ k⁴`, `N = 2 · 10⁸ k³`, `Δ = k/(n N)`, rest energy `1/(2k)`, resolution `1/(10k)`: `|2π N' − R t_a| ≤ (15/k) R t_a` and `R t_a ≥ 300 k`**). Contract: UD-153, UD-154; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** decides that the conditions of ~~#396~~ can be met together, for the ring with two carriers and nothing else; the messenger's velocity is `vRel` at its reference momentum and enters through the predicted angles only, no record holds a speed; the count is compared with the flow time `t_a`, at which nothing is recorded, and on the ring the two packets start together, so this is not yet the reading at an emission; nothing after `(3/2) t_b`; the ring has more than about `4 · 10²⁷ k¹¹` sites, the gap falls as `k⁻⁶` and the flow times grow as `k⁷`, each `k` a different generator; constants not sharp; for some strength, one fibre, product starts; the values will not carry over unchanged to the chain with walls and three carriers. Remainder: (a) the wall (the mirror lemma, parked) and the return reading → ~~#398~~ has the pointer side of the return reading and finds it conditional on where the emission pointer stopped; (b) the radar time and distance from the two counts; (c) three carriers. | ~~**M–L**~~ done | Medium–High — the conditions of the reading are shown to be met, with a messenger at a velocity tending to `1` | ~~#396~~, ~~#395~~, ~~#392~~ | > | ~~398~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, eighth brick; Remainder (a) of ~~#397~~, the pointer side)* **A swing that starts in the middle: the count of a gate that opens late.** ▸ **DONE 2026-11-15: a pointer whose swing is offset by `c` ticks once a turn when `c` is in one of two bands and does not tick when it is in a third; so the return reading, whose gate opens in the middle of the run, will be conditional on where the emission pointer stopped; no flow or gate here; UO-16 not met.** [`RecordLayer/UnifiedWindowSwing.lean`](../CsdLean4/RecordLayer/UnifiedWindowSwing.lean) (10 pins). No new assumption. `IsWindowSwing` (a swing on a window, with a phase and a resting phase; ~~#358~~'s `IsSwing` is the case from the start), ★★ `IsWindowSwing.hasTicks_le`, ★★ `IsWindowSwing.hasTicks_add`, ★★★ `IsWindowSwing.abs_count_sub_le` (**counts at the two ends of the window have `|2π (N_T − N_s) − R (T − s)| ≤ 6π`**), `IsWindowSwing.exists_tickCount`, `pshift_notMem_readyArc_of_nonpos`, `pshift_mem_displayArc_iff_of_nonpos`, ★★ `isWindowSwing_snd_high`, ★★ `isWindowSwing_snd_low` (a pointer within `η < 1/12` of `(1/4)(1 − cos R t) − c` is a window swing when its lower turning point `p = θ₁ − c` is in `(−1/6 + η, 1/6 − η]` or in `(−1/2 + η, −1/6 − η)`), ★★★ `offset_swing_count`, ★★ `offset_swing_dead` (**for `p` in `(1/6 + η, 1/2 − η]` the pointer is never in the ready arc and its count advances by at most one, however far the phase advances**). Contract: UD-155, UD-156; UO-16 annotated, with a note for the author under the scope of UD-156; `POSITS.md` unchanged. **Scope:** one pointer's history, on no arena; the return reading is not assembled; where the offset `c` lies is not shown (while the emission gate closes it is bounded only by the pointer's speed), so that it lies in a band is a hypothesis about the run; nothing within `η` of the edges `c = 1/4, 7/12, 11/12`, where the count can be off by up to one tick a turn, and no choice of arcs and amplitude removes the edges (calculations, not in Lean); the numbers are those of the clock pointer of ~~#391~~. Remainder: (a) the return reading assembled on the ring: the complementary gate closed, open while the messenger is away, closed again when it comes round, with the band as a hypothesis → ~~#399~~, with no values exhibited; (b) a bound of the offset `c`, if one can be had (the emission pointer's displacement while its gate closes); (c) the wall; (d) the radar time and distance. | ~~**M**~~ done | Medium — the one reading of UA-18 (d) that ~~#394~~ does not cover, and the condition it needs | ~~#394~~, ~~#393~~, ~~#381~~, ~~#358~~ | > | ~~399~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, ninth brick; Remainder (a) of ~~#398~~)* **The return reading: the count of the gate that is open while the messenger is away.** ▸ **DONE 2026-11-15: for any messenger packet and the flat clock packet on one ring, with both gates in one run, the complementary pointer's count after the messenger is back is within a stated bound of the time away in ticks, under inequalities between the parameters and for runs in which the emission pointer stopped in a band of ~~#398~~; no values exhibited; UO-16 not met.** [`RecordLayer/UnifiedReturnReading.lean`](../CsdLean4/RecordLayer/UnifiedReturnReading.lean) (8 pins). No new assumption beyond UA-18. `count_after_transition` (the last two phases of ~~#394~~, from any count), ★★ `away_pointer_count` (**one pointer's history closed, open, closed: at every flow time of `[t₄, t₅]` a count `N` with `|2π N − R (t₃ − t_b)| ≤ 2π (3 v ((t_b − t_a) + (t₄ − t₃)) + 7)`**), ★★★ `coincFlow_return_reading` (the same for the flow of ~~#393~~ with its last pointer coupled to `gateOp (¬r) Y`, from where the two carriers are), `packetAngleGap`, `ringEmissionShift`, `ringReturnRun`, `levelSwap_apply_eq_zero_of_site_ne`, ★★★ `flat_clock_return_reading` (**for some strength `0 < g ≤ R/5`: if the emission pointer's displacement at `t_b` is in a band, `|2π N' − R (t₃ − t_b)| ≤ 2π (6 g ((t_b − t_a) + (t₄ − t₃)) + 7)` at every flow time of `[t₄, t₅]`**). Contract: UD-157, UD-158; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** conditional on the band of ~~#398~~, and where the emission pointer's displacement lies is not shown; the conditions are inequalities between real numbers (budgets, predicted angles near up to `t_a` and on `[t₄, t₅]` and apart on `[t_b, t₃]`, `t₃ ≤ 2π n/R`, three size conditions) and no values are shown to meet them, in particular no run in which the messenger comes back; how it comes back (a lap of the ring, or a wall) is not said and no wall is built; the count is compared with the flow times `t_b`, `t₃`, at which nothing is recorded; the radar time and distance are not formed; that `ringEmissionShift` is the first pointer's displacement is ~~#393~~'s `coincFlow_prodRay`, not restated; null sets of fibres and starts; for some strength; a ring and two carriers. Remainder: (a) the radar time and distance from the emission count and this one; (b) values for a run with a return: on the ring a lap (the clock's horizon must cover it), on the chain of UA-18 the wall (the mirror lemma, parked); (c) a bound of the emission pointer's displacement while its gate closes, if one can be had → ~~#400~~, only for a gate that closes within a small part of a tick; (d) three carriers. | ~~**L**~~ done | Medium–High — the fifth reading of UA-18 (d), with its condition | ~~#398~~, ~~#396~~, ~~#395~~, ~~#394~~, ~~#393~~ | > | ~~400~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, tenth brick; Remainder (c) of ~~#399~~)* **Where the emission pointer stops: the band as a condition on the clock's phase.** ▸ **DONE 2026-11-15: the emission pointer stops within a stated bound of the clock's swing at the flow time its gate begins to close; for a gate that closes within a small part of a tick the band of ~~#398~~ and ~~#399~~ is replaced by `1/6 < |cos R t_a|`; the band is not removed, no values exhibited; UO-16 not met.** [`RecordLayer/UnifiedStopBand.lean`](../CsdLean4/RecordLayer/UnifiedStopBand.lean) (5 pins). No new assumption beyond UA-18. `narrow_clock_packet_swing_fine`, `flat_clock_swing_fine` (over the `j`-th part of its horizon, `0 < j ≤ n`, the clock packet swings within `1/(24 j)` of `(1/4)(1 − cos R t)`), ★★ `coincShift_stop_le` (**the displacement at `t_b` of a pointer coupled to the gate `r` is within `δ₀ + 2|z| ε t_a + 2|z| (t_b − t_a)` of `(1/4)(1 − cos R t_a)`**), ★★★ `coincFlow_return_reading_of_phase` (the return reading of ~~#399~~ with the band replaced by `σ < 1/12` and `4 σ < |cos R t_a|`, `σ = 2 δ₀ + 2|z| (ε t_a + (t_b − t_a) + ε (t₃ − t_b))`), ★★★ `flat_clock_return_reading_of_phase` (**on the ring, flat clock packet with `n ≥ 4`, `t₃ ≤ 2π n/(4R)`: if `2 (R/5) (ε t_a + (t_b − t_a) + ε (t₃ − t_b)) ≤ 1/48` and `1/6 < |cos R t_a|`, for some `0 < g ≤ R/5` the count `N'` has `|2π N' − R (t₃ − t_b)| ≤ 2π (6 g ((t_b − t_a) + (t₄ − t₃)) + 7)` at every flow time of `[t₄, t₅]`**). Contract: UD-159, UD-160; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** while the gate closes only the pointer's speed is used, so the bound serves only a gate that closes within a small part of a tick (`2 (R/5) (t_b − t_a) ≤ 1/48`); in the family of ~~#397~~ the closing lasts many ticks, so that family does not meet it, and for a slowly closing gate nothing is shown about where the pointer stops; the band is replaced, not removed: runs with `|cos R t_a| ≤ 1/6` are not covered; `R t_a` is the clock's rate times a flow time, read by no pointer; the finer swing holds over a quarter of the horizon for `n ≥ 4`; inequalities between real numbers, no values, no run with a return exhibited; how the messenger comes back is not said; the radar time and distance are not formed; null sets of fibres and starts; for some strength; a ring and two carriers. Remainder: (a) the radar time and distance from the emission count and the return count → ~~#401~~, as two counts at two flow times; (b) a family of values with a return in which the gate closes within a small part of a tick (narrower arcs than ~~#397~~'s `δ = 1/(10k)`); (c) the author's decision on runs that fail the phase condition; (d) three carriers. | ~~**M**~~ done | Medium — the band of ~~#398~~ becomes a condition that names no pointer, for a quickly closing gate only | ~~#399~~, ~~#398~~, ~~#396~~, ~~#394~~ | > | ~~401~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — step (4) of the plan of 2026-11-14, eleventh brick; Remainder (a) of ~~#400~~)* **The radar reading: the emission count and the return count of one run.** ▸ **DONE 2026-11-15: in the run with both gates the first pointer's count while the messenger is away and the last pointer's count after it is back are both read, and the radar time and distance in ticks formed from them are within stated bounds of flow-time expressions; two counts at two flow times, conditional as ~~#400~~, no values exhibited; UO-16 not met.** [`RecordLayer/UnifiedRadarReading.lean`](../CsdLean4/RecordLayer/UnifiedRadarReading.lean) (5 pins). No new assumption beyond UA-18. ★★ `coincFlow_first_gated_reading` (the emission reading of ~~#394~~ for the *first* pointer of the flow of ~~#393~~), ★★★ `coincFlow_radar_reading` (**both pointers of one run: `N_e` at every flow time of `[t_b, t₃]`, `N_r` at every flow time of `[t₄, t₅]`; with `e = N_e`, `r = N_e + N_r`: `|2π (e + r)/2 − R (t_a + (t₃ − t_b)/2)| ≤ 2π (9|z| (t_b − t_a) + 3|z| (t₄ − t₃) + 7)` and `|2π (r − e)/2 − R (t₃ − t_b)/2| ≤ 2π (3|z| ((t_b − t_a) + (t₄ − t₃)) + 7/2)`**), `flat_ring_not_near_le`, `flat_ring_near_le` (the weight outside, and inside, "within the angle `w`" from the budgets at one flow time), ★★★ `flat_clock_radar_reading` (**the same on the ring in `ringReturnRun`, flat clock packet with `n ≥ 4`, for some `0 < g ≤ R/5`**). Contract: UD-161, UD-162; UA-18 "Where" updated; UO-16 annotated; `POSITS.md` unchanged. **Scope:** two counts at two flow times: after `t₄` the first pointer's gate is open again and it ticks on, so when the return count is read it no longer shows the emission count, and nothing says the two are read together; `(e + r)/2` and `(r − e)/2` are arithmetic in the statement, shown by no pointer, and compared with flow times; `(r − e)/2` is half the time away in ticks, not a distance in sites, and the messenger's speed enters no hypothesis; neither count sees the closing or the reopening of the gate (error terms); conditional as ~~#400~~ (a quickly closing gate, `1/6 < |cos R t_a|`); inequalities between real numbers, no values, no run with a return exhibited; how the messenger comes back is not said; null sets of fibres and starts; for some strength; a ring and two carriers; one observer. Remainder: (a) a family of values with a return in which the gate closes within a small part of a tick; (b) ~~the author's decisions~~ settled 2026-11-15 by the charter (note under UO-16): the first pointer ticking on is UA-18 (e), a reading is the first open period's tick records; the phase condition is an inequality on the prepared parameters (UA-18 (b), UA-4), M6 must exhibit values meeting it and say that outside it there is no reading; (c) three carriers and the second observer (step (5)). | ~~**M**~~ done | Medium–High — one observer's two radar readings in one run, with their conditions | ~~#400~~, ~~#399~~, ~~#394~~, ~~#393~~ | > | ~~402~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — the plan restated from the top after eleven bricks of step (4))* **M6 from the top: the Lorentz statement from five pointer histories.** ▸ **DONE 2026-11-15: M6 as one conditional theorem about one run, with everything not proved a named hypothesis; it fixes the list R1–R7 of what remains; on no arena; UO-16 not met.** [`RecordLayer/UnifiedRadarSkeleton.lean`](../CsdLean4/RecordLayer/UnifiedRadarSkeleton.lean) (6 pins). No new assumption. `IsGatedHistory`, `IsGatedHistory.exists_count` (~~#394~~'s count for a pointer read through any projection), `IsAwayHistory`, `IsAwayHistory.exists_count` (~~#399~~'s), `gatedErr`, `awayErr`, `abs_div_sub_le_of_rate`, ★★★ `radar_lorentz_of_histories` (**five pointers of one run with gated or away histories, `B`'s rate within `κ R₀` of `√(1 − v²) R₀`, the gates' closing times within tolerances `τ` of four times with `T_R − T_E = v (T_M − T_E)`, `T_Ret − T_R = v (T_M − T_Ret)` ⇒ the five counts in the unit `2π/R₀` satisfy the Lorentz statement of `radar_lorentz_error` with errors read off the histories**). Contract: UD-163, UD-164; UO-16 annotated: from this row on a brick is built only if it discharges an item of the list. **Scope — the list:** (R1) the five histories for the five-pointer flow from a product of three carriers; (R2) where the carriers are: packets on the chain with two walls, the reflection, `A` at rest, `B` moving; (R3) the two relations between the flow times, which state the messenger's speed and `B`'s velocity and may not stay hypotheses (test of UO-16); (R4) the moving clock's rate, likewise, from UD-138; (R5) the band of the return pointer or the conditions of ~~#400~~; (R6) values meeting all the inequalities together; (R7) an open set of the pointers' conserved angles. The theorem is on no arena and proves none of these. Scoped 2026-11-15 (contract, under UD-164): R4 mostly there (UD-142; a rate inequality is missing); `A` at rest not covered (every clock theorem asks `0 < p₀`); the wall has one parked lemma; R3 not started; R5: the family of ~~#397~~ fails ~~#400~~'s conditions; a hand estimate says a family with a slower clock meets them (`Δ δ K ≤ 1/150`, `δ ~ 1/(150 k)`, `N ~ 10⁹ k³`, errors of relative size `1/k`), **and that estimate is to be made a theorem on the ring before R2 is built on the chain: it is the cheapest test of whether this pointer arrangement can meet M6 at all.** Remainder: that test (R5 with R6, two carriers) → ~~#403~~, passed; then R1–R4, R7. | ~~**M**~~ done | High — the first end-to-end statement of M6 and the list it leaves | ~~#401~~, ~~#399~~, ~~#394~~, ~~#280~~ | > | ~~403~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — the test named when ~~#402~~ was scoped: items R5 and R6 for two carriers on the ring)* **Values for a run with a return: a lap of the ring with a quickly closing gate.** ▸ **DONE 2026-11-15: for every `k ≥ 1` a run in which the messenger leaves the clock, goes round the ring and comes back, with every hypothesis of ~~#401~~ and ~~#400~~ met; the return count is within eight ticks of a time away of at least `(5/2) k` in the clock's phase. The test of the pointer arrangement is passed, for two carriers on the ring; no item of the list is discharged on the arena of UA-18; UO-16 not met.** [`RecordLayer/UnifiedReturnValues.lean`](../CsdLean4/RecordLayer/UnifiedReturnValues.lean) (13 pins). No new assumption beyond UA-18. `lap_times`, `lap_angles` (the flow times and predicted angles of a lap), `lapBudget`, `lap_budget_le`, `lap_error_le`, `lapTb`, `lapT3`, `lapT4`, ★★★ `lap_radar_reading` (**seven scalar conditions ⇒ for some `0 < g ≤ R/5`: `2π N_e ≤ 8π` at every flow time of `[lapTb, lapT3]` and `|2π N_r − R (lapT3 − lapTb)| ≤ 16π` at `lapT4`, with `R (lapT3 − lapTb) ≥ (5/2) Δ K`**), `lapn`, `lapGap`, `lap_family_conditions`, ★★★ `lap_radar_family` (**`n = N = 10¹¹ k³`, gap `k/K`, `δ = ε = 1/(1000 k)`, `μ = 1/2`: no condition but the size of the ring, and `|2π N_r − R (lapT3 − lapTb)| ≤ (21/k) · R (lapT3 − lapTb)`**). Contract: UD-165, UD-166; UA-18 "Where" updated; UO-16 annotated. **Scope:** two carriers on a ring, a lap in place of the reflection at a wall, a clock moving at about `0.45`, no second observer: nothing of R1–R7 on the arena of UA-18; the emission is at the start (the gate begins to close at the flow time `0`, so the phase condition of ~~#400~~ holds because `cos 0 = 1`), a later emission is not exhibited; the return is read at one flow time; the messenger has rest energy `1/2` and velocity about `0.7`, and a family with the velocity near one and a quickly closing gate together is not exhibited; counts of two pointers at two flow times; null sets of fibres and starts; for some strength. Remainder: (a) R1: the five-pointer flow from three carriers and its five histories → first half ~~#404~~; (b) R2 on the chain with two walls, with the emission at `B`'s start; (c) a window for the return, and a messenger near velocity one, if R3's tolerances need them. | ~~**L**~~ done | High — decides that the arrangement of UA-18 can give a return reading | ~~#402~~, ~~#401~~, ~~#400~~, ~~#397~~ | > | ~~404~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — item R1 of the list of ~~#402~~, first half)* **The five-pointer run: the flow of UA-18's arena on the fibres where the profiles vanish.** ▸ **DONE 2026-11-15: the five-pointer flow of ~~#389~~ is solved from a ray on the fibres where all five profiles vanish, and the top-down statement of ~~#402~~ holds for the counts of its five pointers when five integrals along the carrier's uncoupled orbit have the five histories; any carrier and observables; the histories are not proved; UO-16 not met.** [`RecordLayer/UnifiedPentaRun.lean`](../CsdLean4/RecordLayer/UnifiedPentaRun.lean) (11 pins). No new assumption beyond UA-18. `freeTurn`, `twoCoupledMatrix_of_lift_eq_zero`, `twoCoupledFlow_of_lift_eq_zero`, `threeCoupledFlow_of_lift_eq_zero`, `fourCoupledFlow_of_lift_eq_zero`, ★★ `fiveCoupledFlow_of_lift_eq_zero` (the ray moves by the uncoupled unitary, each pointer is turned by the accumulated `−2⟨B⟩` of its own observable), `idxArenaEquiv_kick_free`, `freeShift`, `freeTurn_reindex`, ★★ `pentaRun_eq` (**a carrier with any index type: pointer `i` at `pshift (freeShift G_i z_i p t)`**), ★★★ `penta_radar_lorentz` (**~~#402~~'s Lorentz statement for the counts of the five pointers of this flow, from five histories of the `freeShift`s**). Contract: UD-167, UD-168; UA-18 "Where" updated; UO-16 annotated. **Scope:** half of R1: that the five integrals have the histories is not shown; the carrier and the five observables are arbitrary, so nothing here uses three carriers, the chain, the walls or coincidence gates; only the fibres where all five profiles vanish (a null set; R7 untouched); conditional as ~~#402~~ on the times, their relations and tolerances and the moving clock's rate; R2–R7 untouched. Remainder: (a) → reduced to two-carrier histories ~~#405~~; R1, second half: the carrier as messenger times a composite of the two clocks, the five observables as `gateOp`s between two of the three, each `freeShift` a `coincShift` of ~~#393~~, and the five histories from where the carriers are (~~#394~~, ~~#399~~, ~~#400~~ restated as histories); (b) R2–R7. | ~~**M**~~ done | Medium–High — the conditional M6 is now about the flow of UA-18's own arena | ~~#402~~, ~~#389~~, ~~#368~~ | > | ~~405~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — item R1 of the list of ~~#402~~, second half)* **Three carriers and five gates: each pointer's displacement is a two-carrier one.** ▸ **DONE 2026-11-15: for the carrier messenger × (clock `A` × clock `B`) from a product ray, with the five gates of UA-18 (c), the top-down statement of ~~#402~~ holds for the five counts when five two-carrier gate displacements have the five histories. R1 is reduced to five two-carrier statements, not discharged; UO-16 not met.** [`RecordLayer/UnifiedTriCarrier.lean`](../CsdLean4/RecordLayer/UnifiedTriCarrier.lean) (15 pins). No new assumption beyond UA-18. `freeShift_prodRay` (`freeShift` of ~~#404~~ from a product ray is `coincShift` of ~~#393~~), `arenaObs_kronecker_right`, `siteSel_snd`, `siteSel_fst`, ★ `arenaObs_gateOp_snd`, `arenaObs_gateOp_fst` (a gate between the messenger and one clock reads that clock only), `gateOpL`, ★ `arenaObs_gateOpL_prodRay` (operator on the first carrier = the gate of ~~#390~~ with the carriers exchanged), `gateOp_snd_isHermitian`, ★★ `coincShift_gate_snd`, `coincShift_gate_fst`, `coincShift_one_kronecker_pair`, `coincShift_gateOpL`, `triRun`, ★★★ `tri_radar_lorentz` (**~~#402~~'s Lorentz statement for the five counts of `triRun`, from five histories of `coincShift (gateOp …)` for two carriers each**). Contract: UD-169, UD-170; UA-18 "Where" updated; UO-16 annotated. **Scope:** the five histories are still hypotheses; ~~#394~~, ~~#399~~, ~~#400~~ give such histories only on the ring for a flat messenger and a flat clock packet, and three of the five gates here are opened by a clock's site, a role those theorems have not been checked in; any carriers, energies, sites and relations: no chain, wall, packet or clock instantiated; product start on the fibres where all five profiles vanish (null sets; R7 untouched); conditional as ~~#402~~ on the times, their relations and tolerances and the moving clock's rate; R2–R7 untouched. Remainder: (a) check ~~#394~~/~~#399~~/~~#400~~ with a clock as the gate-opening carrier → checked 2026-11-15: their general parts hold for any first carrier; the same day the narrow clock-packet theorems of ~~#391~~, ~~#396~~, ~~#400~~ were weakened to `0 ≤ p₀` (a clock at rest ticks); the flat packet centred on the mode `0` is not built; (b) R2: packets on the chain with walls giving the five histories → pieces ~~#406~~ (a clock at rest, on the ring), ~~#407~~ (the wall state of one Dirac carrier); (c) R3–R7 → R4 as an inequality ~~#406~~. | ~~**M**~~ done | Medium–High — the conditional M6 now has UA-18's three carriers and five gates in it | ~~#404~~, ~~#402~~, ~~#393~~, ~~#390~~ | > | ~~406~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — item R4 of the list of ~~#402~~, and of item R2 the clock at rest)* **A clock at rest, and the moving clock's rate against it.** ▸ **DONE 2026-11-15: the rate hypothesis of ~~#402~~ holds with `κ = Δ` for the level splittings of one clock at a momentum term and at rest; and a flat clock packet at reference momentum `0` ticks at the rate `Δ` and stays where it starts, on the ring. No wall, no gate; UO-16 not met.** [`RecordLayer/UnifiedClockAtRest.lean`](../CsdLean4/RecordLayer/UnifiedClockAtRest.lean) (11 pins). No new assumption. ★★ `abs_movingRate_sub_le` (**`|R(s) − √(1 − v²) R(0)| ≤ Δ R(0)`, `v = s/E₀(s)`, `R(0) = Δ`**), `ringRate_zero`, ★ `abs_ringRate_sub_le`, `restWindow` (the modes `0, …, N − 1`), `sum_norm_flatAmpN_rest`, `restWindow_mem_range`, `restWindow_width`, ★★ `rest_clock_packet_ticks` (the instance of ~~#391~~ at reference momentum `0`: at least `n` ticks, operational time per flow time within `1/n + Δ` of `1`), `rest_packet_start_le`, ★ `rest_packet_budget`, ★★★ `rest_clock_packet_localised` (**at every packet time `|t| ≤ 4π n/Δ`: weight on a region with `cos(2π k/L) ≤ γ`, times `1 − γ`, at most `1/N + 1/(20000 Δ n N)`**). Contract: UD-171, UD-172; UA-18 "Where" updated; UO-16 annotated. **Scope:** the rate inequality is between level splittings at a momentum term; its `v` is `σ/E₀`, not a packet's velocity and not the lattice band velocity (the mismatch is R3's to carry); the rest packet is on the ring, with no wall, and beside a wall it is not orthogonal to its mirror image; its modes have momenta in `[0, 1/(1200 n))`, the small drift is inside the budget; the localisation is a weight of the state vector at packet times, not restated at flow times, no region exhibited; one clock, its own pointer, one fibre and one start for the ticks; no messenger, gate or second clock. Remainder: (a) the rest clock beside a wall → ~~#408~~ (`wall_clock_ticks`); (b) the rest of R2 and R3–R7. | ~~**M**~~ done | Medium — R4 as an inequality, and the observer at rest of UA-18 (b) on the ring | ~~#402~~, ~~#392~~, ~~#391~~ | > | ~~407~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — of item R2 of the list of ~~#402~~, the wall for one Dirac carrier)* **The wall: a packet minus its mirror image.** ▸ **DONE 2026-11-15: for any ring packet of one Dirac carrier, the mirror-odd state (packet − mirror packet)/√2 is odd at every time, has no upper component at the self-mirror sites (the boundary condition of UA-18 (a)), is carried to itself by the ring's evolution, and has its weight on a region controlled by the packet's and the mirror packet's. No instance, so no reflection exhibited; UO-16 not met.** [`RecordLayer/UnifiedOddWall.lean`](../CsdLean4/RecordLayer/UnifiedOddWall.lean) (20 pins). No new assumption beyond UA-18 (a). `negWin`, `mem_negWin`, `mirrorPacket`, `chainParity_ringPacket_negWin`, `chainParity_mirrorPacket`, `chainParity_neg`, `oddWall`, ★★ `chainParity_oddWall`, ★ `odd_apply_eq_zero`, `oddWall_apply_eq_zero`, ★ `chainProp_oddWall`, `regionWeight_neg`, `chainInner_neg_right`, `chainInner_neg_neg`, `chainInner_ringPacket_mirrorPacket`, `chainInner_oddWall`, ★★ `regionWeight_oddWall_le` (**where neither is, at most `2τ`**), ★★ `le_regionWeight_oddWall` (**where both are, at least `1 − 2τ`**), `ne_zero_of_chainInner_eq_one`, ★★★ `arenaObs_regionProj_oddWall` (**on the arena: the expectation of a region projector along the uncoupled orbit is the wall state's weight on the region**). Contract: UD-173; UA-18 "Where" updated; UO-16 annotated. **Scope:** the odd sector of the ring, not shown to be a chain of half the sites with hard ends (~~#275~~ has that for the even sector); one Dirac carrier, the messenger: a two-level clock's wall state is not built; the two weight bounds take bounds on the packet's and the mirror packet's weight as hypotheses (~~#265~~ supplies them for both), and no window, region or time is instantiated; orthogonality and norms are hypotheses of the lower bound and of the arena statement; uncoupled: no pointer, gate or record. Built on ~~#270~~, which has the even sector with its records. Remainder: (a) the wall state of a two-level clock → ~~#408~~; (b) → the bound through the reflection ~~#408~~, numbers still owed; an instance: the messenger leaving `B`, reaching the wall and coming back, as bounds on coincidence weights; (c) the odd sector as a chain → ~~#416~~. | ~~**M**~~ done | Medium–High — the wall of UA-18 (a) exists for the messenger | ~~#402~~, ~~#270~~, ~~#265~~ | > | ~~408~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — of item R2 of the list of ~~#402~~: the messenger through the reflection, the clock's wall state, the clock at rest beside a wall; and item R7 scoped)* **The carriers on the walled chain: where the messenger is, and a clock beside a wall.** ▸ **DONE 2026-11-15: the messenger's wall state stays within the two mirror arcs up to the budgets of ~~#265~~, through the reflection; a wall state is one packet on `W ∪ −W` with odd amplitudes; a two-level clock has a wall state with the same bound; and a clock at rest beside a wall (modes `±1, …, ±N`) ticks with no hypothesis but the sizes, which answers risk (iii) of UA-18. No numbers for the budgets, no gate, no history; UO-16 not met.** [`RecordLayer/UnifiedWallObservers.lean`](../CsdLean4/RecordLayer/UnifiedWallObservers.lean) (12 pins). No new assumption beyond UA-18 (a), (b). ★★ `regionWeight_oddWall_away_le` (**weight away from the packet's direction and the mirror direction, times `1 − γ`, at most twice the two budgets**), `oddExt`, ★★ `ringPacket_oddExt`, `sum_norm_oddExt`, ★ `devicePacket_oddExt`, `chainParity_devicePacket_wall`, `deviceRegionWeight_wall`, ★★ `deviceRegionWeight_wall_away_le`, `wallWindow`, `wallWindow_neg_notMem`, `sum_norm_flatAmpN_wall`, ★★★ `wall_clock_ticks` (**the instance of ~~#391~~ at reference momentum `0` for the mirror-odd packet on `±1, …, ±N`: at least `n` ticks at the rate `Δ`**). Contract: UD-174, UD-175; UA-18 "Where" updated; UO-16 annotated; R7 scoped in the contract (a route by the back-reaction bounds of ~~#375~~, no obstruction known, nothing built). **Scope:** the odd sector of the ring, not shown to be a chain with hard ends; the budgets `1 − Re T(0) + |t| · limitErr` are not evaluated for the wall window, a moving window or a mirror window, so no region, time or width is exhibited and the place of the clock beside the wall rests on four unevaluated budgets; a moving clock's wall state is not shown to tick (its window is not narrow about one momentum; ~~#391~~ in the modulus of the momentum is owed); uncoupled carriers, one fibre and one start for the ticks; no coincidence weight, gate, pointer history or reading. Remainder: (a) the budgets as numbers for the three windows and their mirrors → ~~#409~~; (b) ~~#391~~ in the modulus of the momentum, for `B` → ~~#409~~; (c) coincidence weights between wall states and the five histories of ~~#405~~; (d) R3, R6, R7. | ~~**M**~~ done | Medium–High — both observers' carriers and the messenger exist on the walled chain, `A` ticking | ~~#407~~, ~~#406~~, ~~#391~~, ~~#265~~ | > | ~~409~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — of item R2 of the list of ~~#402~~: Remainders (a) and (b) of ~~#408~~)* **The walled chain with numbers: flat wall states, and a moving clock that ticks.** ▸ **DONE 2026-11-15: the budgets of ~~#408~~ are evaluated for every flat window of `N` neighbouring modes and its mirror (`4 (1/N + 1/(20000 Δ n N))` over `1 − γ`, up to `|t| ≤ 4π n/Δ`); the clock at rest beside the wall stays beside it; and the moving clock's wall state ticks, after ~~#391~~ was generalised to a window within `1/(600 n)` of `p₀` or of `−p₀`. No gate, no history; UO-16 not met.** [`RecordLayer/UnifiedWallBudgets.lean`](../CsdLean4/RecordLayer/UnifiedWallBudgets.lean) (10 pins) and one pin in `CV/DiracDevice.lean`. No new assumption beyond UA-18 (a), (b). `negWin_Ico`, `sum_norm_flatAmpN_Ico`, `ico_packet_start_le`, ★ `ico_packet_budget`, `mirror_ico_mem_range`, ★★ `regionWeight_icoWall_away_le`, ★★ `deviceRegionWeight_icoWall_away_le`, ★★★ `wall_clock_localised`, `flatWindow_neg_notMem`, ★★★ `wall_moving_clock_ticks`; `abs_sq_sub_sq_scaled_le_or` (the squared lattice momentum term is even). **Changed hypotheses** (the window may lie near `p₀` or near `−p₀`): `levelWeight_mul_rate_le`, `norm_deviceCoherence_shift_sub_le`, `clock_tick_device`, `sum_norm_modeCoh_mul_le`, `clock_packet_ticks`, `clock_packet_count_ge`, `narrow_clock_packet_ticks`; every earlier instance uses the first alternative. Contract: UD-176, UD-177; UD-142 restated; UA-18 "Where" updated; UO-16 annotated. **Scope:** weights of state vectors at packet times, for uncoupled carriers; no region exhibited, and the bounds are useful only for arcs wide against `1/√N`; the count of each clock is at one flow time, on one fibre and from one start; the swing theorems of ~~#396~~ and ~~#400~~ were not generalised, so the hypothesis of ~~#394~~ about the clock's own reading is not available for a wall clock; no coincidence weight between wall states (on the ring the relation must be mirror-symmetric to be the chain's). Remainder: (a) ~~#396~~/~~#400~~ generalised as ~~#391~~ was; (b) the mirror-symmetric coincidence relation and gate bounds from the wall states' arcs → ~~#410~~ (for any rays; not yet combined with ~~#409~~); (c) the five histories of ~~#405~~; (d) R3, R6, R7. | ~~**M**~~ done | Medium–High — both observers tick on the walled chain and all three carriers are placed with numbers | ~~#408~~, ~~#392~~, ~~#391~~, ~~#265~~ | > | ~~410~~ | Records | Records→spacetime | *new (2026-11-15, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — of items R1 and R2 of the list of ~~#402~~: Remainder (b) of ~~#409~~)* **The coincidence relation of the walled chain, and its gate from arcs.** ▸ **DONE 2026-11-15: `wallNear` (within `w` of a site or of its mirror image) and the two gate bounds of ~~#395~~ for it: for any messenger ray and any clock ray the weight outside (inside) the relation is at most the two carriers' weights outside their double arcs, when the arcs' centres are near (apart both ways round). No state is put in; UO-16 not met.** [`RecordLayer/UnifiedWallGate.lean`](../CsdLean4/RecordLayer/UnifiedWallGate.lean) (9 pins). No new assumption beyond UA-18 (a), (c). `cos_site_neg`, `neg_mem_ringArc`, `wallNear`, `wallArc`, `mem_wallArc`, ★ `wallNear_of_mem_wallArc`, ★ `not_wallNear_of_mem_wallArc`, ★★ `wallGate_not_le`, ★★ `wallGate_le`. Contract: UD-178; UA-18 "Where" updated; UO-16 annotated. **Scope:** the relation and two bounds for any rays; not combined with the weights of ~~#409~~ (those are weights of state vectors at packet times; the occupancies along the orbit are linked for the messenger by ~~#407~~, for a clock's wall state not yet), so no gate is shown open or closed for any state; conditions on angles, no flow time exhibited; a messenger and one clock, not two clocks; that `wallNear` is the relation of a chain of half the sites is a reading of UA-18 (a), not a theorem. Remainder: (a) ~~#409~~'s bounds as occupancies along the orbits for the three wall states → ~~#411~~ (messenger and one clock); (b) the wall clock's swing with a margin (~~#396~~/~~#400~~ generalised) → ~~#411~~; (c) flow times → the kinematic times ~~#412~~, the gates' flow times still owed; (d) the five histories of ~~#405~~; (e) R3, R6, R7. | ~~**S–M**~~ done | Medium — without it a gate on the walled chain sees half of a coincidence | ~~#409~~, ~~#395~~, ~~#394~~ | > | ~~411~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — of items R1 and R2 of the list of ~~#402~~: Remainders (a) and (b) of ~~#410~~)* **The wall gate, open and closed, for wall states; the wall clocks' swing.** ▸ **DONE 2026-11-16: for a flat messenger wall state and a flat clock wall state on the walled chain, along their uncoupled orbits, the weight outside (inside) the chain's coincidence relation times `1 − cos δ` is at most `8 (1/N + 1/(20000 Δ n N))` while their predicted angles are near (apart both ways round); and both wall clocks swing within `1/(24 j)` up to `2π n/(j R)`. The three hypotheses of ~~#394~~ for one messenger–clock gate, not combined; the messenger is slow; UO-16 not met.** [`RecordLayer/UnifiedWallOrbits.lean`](../CsdLean4/RecordLayer/UnifiedWallOrbits.lean) (18 pins). No new assumption beyond UA-18. `vRel_neg`, `chainProp_oddWall_shift`, `negWin_mem_range`, `negWin_width`, `sum_norm_negWin`, `icoWindow_ne_neg`, `icoWindow_mem_range`, `chainInner_icoWall`, ★★ `one_sub_sum_arenaObs_wallArc_icoWall_le`, ★★ `one_sub_sum_arenaObs_wallArcs_icoDevice_le`, ★★★ `wall_gate_open`, ★★★ `wall_gate_closed`, `wallSym_range`, `wallSym_width`, ★★ `wall_clock_swing_fine`, `flatSym_range`, `flatSym_width`, ★★ `wall_moving_clock_swing_fine`. **Changed hypotheses** (the window may lie near `p₀` or near `−p₀`): `clock_packet_shift_le`, `narrow_clock_packet_swing`, `narrow_clock_packet_swing_fine`. Contract: UD-179, UD-180; UA-18 "Where" updated; UO-16 annotated. **Scope:** the messenger has rest energy `μ ≥ 1` and momentum at most `1/2`, so velocity at most about `0.45`: by ~~#412~~ it cannot be the messenger of M6; conditions on predicted angles, no flow time exhibited; uncoupled orbits from a product start on the fibres where the profiles vanish (null sets); a messenger and one clock, not the two clocks' gate; the hypotheses are not combined, so no history, count or reading. Remainder: (a) a fast messenger's wall state with budgets for small `μ` → ~~#413~~; (b) the two clocks' gate; (c) the five histories of ~~#405~~, on the fast messenger. | ~~**M–L**~~ done | Medium–High — the three hypotheses of the reading theorems exist on the walled chain | ~~#410~~, ~~#409~~, ~~#400~~, ~~#395~~ | > | ~~412~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — item R3 of the list of ~~#402~~, first part)* **The four times of the radar run, and what a slower messenger costs.** ▸ **DONE 2026-11-16: the times `T_R = d`, `T_Ret = 2d/(1 + v)`, `T_M = d/v` satisfy the two relations of ~~#402~~, which are thereby theorems and not hypotheses for a run with these times; a messenger of speed `u ≤ 1` is late by `d (1 − u)/u` at the wall and by at most `2d (1 − u)/u` at the return; a packet has `1 − v ≤ μ²/(2 p²)`. FINDING: the messenger built so far (`u ≤ 0.45`) cannot serve M6. Arithmetic only; UO-16 not met.** [`RecordLayer/UnifiedRadarTimes.lean`](../CsdLean4/RecordLayer/UnifiedRadarTimes.lean) (12 pins). No new assumption. `radarTR`, `radarTRet`, `radarTM`, ★★ `radar_times_relay`, ★★ `radar_times_return`, `crossRelay`, `crossReturn`, `crossRelay_spec`, `crossReturn_spec`, ★ `abs_crossRelay_sub`, ★ `abs_crossReturn_sub_le`, ★★ `one_sub_vRel_le`. Contract: UD-181; UA-18 "Where" updated; UO-16 annotated. **Scope:** straight motion at constant speeds on the folded chain, on no arena; that the gates close at flow times near these times (arc widths over angular speeds) is not shown; the velocity is the relativistic one at a momentum term, not the lattice's band velocity; the clock's two levels move at slightly different velocities. Remainder: (a) the gates' flow times against these times, with tolerances → the intervals ~~#414~~, not yet one bound; (b) a messenger with `μ²/p²` small against `v²`, with its budgets → ~~#413~~; (c) the band velocity. | ~~**S**~~ done | High — turns two hypotheses of ~~#402~~ into theorems and shows the messenger must be fast | ~~#402~~, ~~#406~~ | > | ~~413~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — of item R2 of the list of ~~#402~~: Remainder (a) of ~~#411~~, forced by the finding of ~~#412~~)* **A fast messenger on the walled chain.** ▸ **DONE 2026-11-16: for a flat messenger of rest energy `0 < μ ≤ 1` (velocity within `2μ²` of `1` at the momentum `1/2`): its budget `lightBudget = 1/N + (1/(2Kμ))² + |t|(1/μ)/(300 n K)`, its wall state within the two mirror arcs up to `4 · lightBudget`, and the wall gate with a clock's wall state open and closed with the bound `4 · lightBudget + 4 (1/N + 1/(20000 Δ n N))` over `1 − cos δ`. No values; UO-16 not met.** [`RecordLayer/UnifiedFastMessenger.lean`](../CsdLean4/RecordLayer/UnifiedFastMessenger.lean) (8 pins). No new assumption beyond UA-18. `light_packet_start_le`, `lightBudget`, ★ `light_packet_budget`, ★★ `regionWeight_lightWall_away_le`, ★★ `one_sub_sum_arenaObs_wallArc_lightWall_le`, ★★★ `light_wall_gate_open`, ★★★ `light_wall_gate_closed`, ★★ `one_sub_vRel_half_le`. Reuses `limitErr_le_inv_mul` of ~~#397~~. Contract: UD-182; UA-18 "Where" updated; UO-16 annotated. **Scope:** the price of the speed is in the bound and is not paid: no choice of `n`, `N`, `μ`, `Δ`, `δ`, `w` is shown to make the budget small against `1 − cos δ`, the closing short against a tick and `μ²` small against `v²` together; momentum at most `1/2` (a messenger made fast by a larger momentum is not treated); weights along uncoupled orbits, conditions on angles, a messenger and one clock. Remainder: (a) the values → ~~#415~~ (for the bounds; readings not yet); (b) the two clocks' gate → ~~#415~~ (for any rays); (c) the five histories of ~~#405~~. | ~~**M**~~ done | High — without it no messenger can meet the tolerances of ~~#412~~ | ~~#412~~, ~~#411~~, ~~#397~~ | > | ~~414~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — item R3 of the list of ~~#402~~, second part: Remainder (a) of ~~#412~~)* **When the wall gate is open: the flow times of the emission, the absence and the return.** ▸ **DONE 2026-11-16: for two carriers starting together at the angle `−d` with angular speeds `v ≤ u`: open while `(u − v) s ≤ w − 2δ`; closed while `w + 2δ < (u − v) s` and `(u + v) s + (w + 2δ) < 2d`; open again while `|s − 2d/(u + v)| ≤ (w − 2δ)/(u + v)`; and the wall gate at the return (the messenger near the clock's mirror image) for any rays and for the light messenger's wall state. One pair of speeds; UO-16 not met.** [`RecordLayer/UnifiedGateTimes.lean`](../CsdLean4/RecordLayer/UnifiedGateTimes.lean) (8 pins). No new assumption beyond UA-18. `cos_lt_cos_of_lt_abs`, ★ `gate_open_at_emission`, ★ `gate_closed_away`, `add_mul_sub_crossReturn`, ★ `gate_open_at_return`, `wallArc_neg`, ★★ `wallGate_not_le_mirror`, ★★★ `light_wall_gate_open_return`. Contract: UD-183; UA-18 "Where" updated; UO-16 annotated. **Scope:** arithmetic of straight motion with the conditions of ~~#410~~ as conclusions; one clock level at a time (the two levels' intervals are not intersected); angles within half a turn (`(u − v) s ≤ π`, `2d ≤ π`); not joined to the tolerances of ~~#412~~ as one bound; the return gate for a messenger and one clock along uncoupled orbits. Remainder: (a) the two levels; (b) one bound for `τ_E`, `τ_R`, `τ_Ret`; (c) values. | ~~**M**~~ done | Medium–High — the reflection's return is a gate that opens at a stated flow time | ~~#413~~, ~~#412~~, ~~#410~~ | > | ~~415~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — item R6 of the list of ~~#402~~ for one messenger–clock gate: Remainder (a) of ~~#413~~; and of item R1 the two clocks' gate)* **The tolerance test for one gate of the walled chain; the two clocks' gate.** ▸ **DONE 2026-11-16: for every `k ≥ 1`, with `n = N = 10¹⁶ k⁴`, gap `k/K`, `δ = 1/(10⁵ k)` and messenger rest energy `1/k`, at all packet times within `10 K`: the gate's bound is at most `1/(1000 k)`, the closing interval at most `1/1000` of the clock's phase, and the messenger within `2/k²` of speed `1`, together. The test is passed for the bounds; readings are not derived; UO-16 not met.** [`RecordLayer/UnifiedWallValues.lean`](../CsdLean4/RecordLayer/UnifiedWallValues.lean) (12 pins). No new assumption beyond UA-18. `heavy_packet_budget_time`, `regionWeight_heavyWall_away_le`, `deviceRegionWeight_icoWall_time_le`, ★★ `one_sub_sum_arenaObs_wallArcs_icoDevice_time_le` (the clock's wall state at each packet time, bound `4 · lightBudget n N 1 t`, no horizon), `wallN`, `wallK`, `wallGap`, `wallDelta`, `wallMu`, ★★★ `wall_values`, ★★ `wallGate_clocks_not_le`, ★★ `wallGate_clocks_le`. Contract: UD-184, UD-185; UA-18 "Where" updated; UO-16 annotated. **Scope:** a test of the bounds of ~~#413~~ and ~~#414~~, not of readings: the size conditions of ~~#394~~, ~~#399~~, ~~#400~~ (the swing's margin, `|z| ε t`, the band) are not checked for this family, and ~~#403~~'s check on the ring is evidence only; the relative angular speed `1/(20 K)` and the run length `10 K` are taken, not derived, and the margin between the messenger (at least `1/2` for `k ≥ 2`) and the clock's lower level (about `0.447`) is small; the family has gap `→ 0` and about `k` ticks in a run, and the comparison with the dilation is not stated; the two clocks' gate is for any two clock rays, not their wall states. Remainder: (a) the reading of one observer on the walled chain from these values; (b) the two clocks' gate for wall states; (c) the five histories of ~~#405~~; (d) the two levels, the band velocity, R7. | ~~**M–L**~~ done | High — the tolerances of one gate do not rule the arrangement out | ~~#414~~, ~~#413~~, ~~#403~~ | > | ~~416~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 3 of the research delivery plan of 2026-11-16: Remainder (c) of ~~#407~~)* **The mirror-odd sector is a chain with hard ends.** ▸ **DONE 2026-11-16: on a ring of `2M` sites the mirror-odd sector, with the ring's evolution, is unitarily the chain on `{0, …, M}` with hard ends and the explicit energy of ~~#275~~ (at rest energy `−μ`, in exchanged components). The boundary condition that UA-18 (a) chose is confirmed, the `√2` end hop included. One Dirac carrier, uncoupled; UO-16 not met.** [`RecordLayer/UnifiedOddChain.lean`](../CsdLean4/RecordLayer/UnifiedOddChain.lean) (24 pins). No new assumption. `compSwap`, `compSwap_compSwap`, `compSwapₗ`, `compSwapₗ_apply`, `compSwap_neg`, ★ `chainParity_compSwap`, `regionWeight_compSwap`, `chainInner_compSwap`, ★ `diracHam_mulVec_compSwap`, `compSwap_pow_mulVec`, ★ `compSwap_chainProp_mulVec`, `chainInner_mulVec_hermitian`, `IsOddState`, `IsOddState.diracHam_mulVec`, `isOddState_oddWall`, `IsOddState.isEvenState_compSwap`, `oddSegOf`, ★★ `chainInner_eq_segInner_oddSegOf`, ★★ `oddSegOf_end`, ★★★ `oddSegOf_diracHam_mulVec`, ★★★ `odd_sector_chain_flow_equiv`, ★★ `oddSeg_generator_symm`, ★★ `regionWeight_eq_segRegionWeight_oddSegOf`, ★★ `oddSegOf_oddWall_flow`. Contract: UD-186; UA-18 "Where" updated; UO-16 annotated; `specs/unified-m6-feasibility.md` updated. **Scope:** one Dirac carrier along its uncoupled evolution; the two-level clock is covered block by block, not restated as one carrier; the gates and pointers are not carried to the segment; the segment's energy is symmetric on the image of the odd sector, not shown Hermitian as a matrix; a ring of `2M` sites with `M ≥ 2`. Remainder: (a) the device and the coupled flow on the segment, if the capstone is to be stated there; (b) Tasks 4 to 12 of the delivery plan → Tasks 4 and 5 in part ~~#417~~, ~~#418~~. | ~~**S–M**~~ done | Medium–High — the wall is a theorem about the model, not a name for a ring sector | ~~#407~~, ~~#275~~, ~~#270~~ | > | ~~417~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 5 of the research delivery plan, first part; required by the feasibility note)* **Translating a carrier round the ring: the observer at the far wall.** ▸ **DONE 2026-11-16: translation commutes with the lattice Dirac energy and evolution (one carrier and the two-level clock); a translated clock has the same reading and pointer displacement on the zero fibre, and its place moves with it. The clock at rest of ~~#408~~, translated by any `h`, swings with the margin of ~~#411~~ and stays in the arc about `2π h/L` up to `4 · lightBudget n N 1 t`. One clock, uncoupled; UO-16 not met.** [`RecordLayer/UnifiedRingShift.lean`](../CsdLean4/RecordLayer/UnifiedRingShift.lean) (25 pins). No new assumption beyond UA-18 (b). ★ `exp_smul_mulVec_intertwine` (generic), `chainShift`, `chainShiftₗ`, `chainShiftₗ_apply`, `regionWeight_chainShift`, ★ `diracHam_mulVec_chainShift`, `chainProp_mulVec_chainShift`, `chainInner_chainShift`, `chainParity_chainShift`, `deviceShift`, `deviceShiftₗ`, `deviceShift_ne_zero`, `deviceRegionWeight_deviceShift`, `deviceCoherence_deviceShift`, `deviceNorm_deviceShift`, ★ `deviceHam_mulVec_deviceShift`, `deviceProp_mulVec_deviceShift`, ★★ `clockFreeRead_deviceShift`, ★★ `clockPointerShift_deviceShift`, ★★ `sum_arenaObs_siteSel_deviceShift`, `cos_site_sub`, `mem_ringArc_sub`, `image_sub_ringArc`, ★★★ `far_wall_clock_swing_fine`, ★★★ `far_wall_clock_outside_le`. Contract: UD-187; UA-18 "Where" updated; UO-16 annotated; feasibility note updated. **Scope:** uncoupled orbits and the fibre where the profile vanishes; the count of the translated clock is not restated (its pointer's history is identified on the zero fibre); mirror-oddness of the translate is stated for one Dirac carrier and a half turn, not for the device; no gate with the messenger or the other clock. Remainder: the gates between the translated clock and the other two carriers, with Task 6. | ~~**M**~~ done | High — the layout the theorems force needs the observer at rest at the far wall | ~~#415~~, ~~#411~~, ~~#408~~ | > | ~~418~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Tasks 4 and 5 of the research delivery plan)* **Event times of the radar run from the gates, and the clocks' rate relation.** ▸ **DONE 2026-11-16: the rate hypothesis of ~~#402~~ with the velocity of the packets, `|R(p) − √(1 − v²) R(0)| ≤ (Δ + a² p³/6) R(0)`; gate windows for the whole chain and at an observer who sits at a wall; and the tolerances `τ_R`, `τ_Ret`, `τ_M` from a gate window (`ρ + d(1 − u)/u`, `ρ + 2d(1 − u)/u`, `ρ`). Arithmetic of straight motion; UO-16 not met.** [`RecordLayer/UnifiedEventTimes.lean`](../CsdLean4/RecordLayer/UnifiedEventTimes.lean) (11 pins). No new assumption. `abs_ringSigma_sub_le`, `sqrt_one_sub_vRel_sq`, `abs_inv_dispE_sub_le`, ★★★ `wall_observer_clock_rate_relation`, `abs_vRel_level_sub_le`, ★ `gate_closed_away_turn`, ★ `wall_gate_apart_before`, ★ `wall_gate_near_at`, ★★ `wall_packet_relay_time_bound`, ★★ `wall_packet_return_time_bound`, ★★ `wall_observer_meeting_time_bound`. Contract: UD-188; UA-18 "Where" updated; UO-16 annotated; feasibility note updated. **Scope:** the timing bounds take a gate window `|t − crossing| ≤ ρ` as hypothesis and are not yet applied to the wall states; one speed per carrier (the clock's two levels differ by at most `Δ` and their intervals are not intersected); the emission's tolerance is the first gate window itself; real arithmetic except the two rate lemmas. Remainder: Task 6 (the five histories), where the windows are joined to the gates and swings → first part ~~#419~~. | ~~**M**~~ done | High — discharges `hκ` and reduces `hτR`, `hτRet`, `hτMB`, `hτMA` to gate windows | ~~#414~~, ~~#412~~, ~~#406~~ | > | ~~419~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 6 of the research delivery plan, first part)* **Histories of gate displacements, from where the carriers are.** ▸ **DONE 2026-11-16: the gated history and the away history that ~~#405~~ asks of a two-carrier gate displacement follow from the two carriers' weights inside and outside the relation on stated intervals, the clock's own displacement, and sizes. Any two carriers and relation; not applied to the walled chain; the band of the away history stays a hypothesis; UO-16 not met.** [`RecordLayer/UnifiedWallHistories.lean`](../CsdLean4/RecordLayer/UnifiedWallHistories.lean) (5 pins). No new assumption. `coincProj_not_not`, `arenaObs_coincProj_prodRay_le_one`, `arenaObs_coincProj_not_prodRay_le_one`, ★★★ `isGatedHistory_coincShift` (error `δ₀ + 2|z| ε t_a`, speed `2|z|`), ★★★ `isAwayHistory_coincShift` (offset the `r`-gated displacement at `t_b`, start error `2|z| ε t_a`, error `δ₀ + 2|z| ε (t₃ − t_b)`). Contract: UD-189; UA-18 "Where" updated; UO-16 annotated; feasibility note updated. **Scope:** the displacement part of ~~#394~~ and ~~#399~~ stated on its own, for displacements along uncoupled orbits from a product ray; no packet, wall, interval or value is put in, so no history of the walled chain is proved; the band (a condition on where the `r`-gated pointer stopped) and the clock's own displacement are hypotheses. Remainder: (a) the five applications for the wall states, with their intervals and both clock levels; (b) the wall states' gates involving the translated clock and between the two clocks; (c) the band (Task 7) → for the moving observer ~~#420~~; (d) Tasks 8 to 12. | ~~**M**~~ done | High — each hypothesis of ~~#405~~ about a pointer is now one application from existing statements | ~~#405~~, ~~#402~~, ~~#399~~, ~~#394~~ | > | ~~420~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Tasks 6 and 7 of the research delivery plan, for the moving observer)* **The moving observer's two radar histories on the walled chain.** ▸ **DONE 2026-11-16: for the light messenger's wall state and the moving clock's wall state, the emission pointer's displacement is a gated history and the return pointer's an away history, for one strength `0 < g ≤ R/5`; the band of the return pointer is derived from sizes (emission at the start). Two of the five hypotheses of ~~#405~~ about pointers are theorems for the wall states; the conditions on predicted angles over three intervals are hypotheses; UO-16 not met.** [`RecordLayer/UnifiedWallObserver.lean`](../CsdLean4/RecordLayer/UnifiedWallObserver.lean) (3 pins). No new assumption beyond UA-18. `wallAngle`, ★★ `wall_observer_gate_weights` (the wall gate open on `[0, t_a]`, closed on `[t_b, t₃]`, open again on `[t₄, t₅]`, each up to `ε`), ★★★ `wall_observer_radar_histories`. Contract: UD-190; UA-18 "Where" updated; UO-16 annotated; feasibility note updated. **Scope:** the conditions on the predicted angles over the three intervals and the size conditions are hypotheses (~~#414~~ and ~~#418~~ give the intervals for straight motion; not joined for the two clock levels); `R t_a` small, an emission at the start, a later one is outside the theorem; the messenger and one clock; displacements along uncoupled orbits from a product start on the fibres where the profiles vanish; counts not stated (they follow from ~~#402~~). Remainder: (a) the relay and the two meeting histories (the observer at rest, the two clocks' gate) → ~~#421~~, ~~#422~~; (b) the intervals and the sizes for a family; (c) `tri_radar_lorentz`; (d) Tasks 9 to 12. | ~~**L**~~ done | High — the first of the five histories proved on the walled chain, and R5 derived | ~~#419~~, ~~#415~~, ~~#413~~, ~~#411~~ | > | ~~421~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 6 of the research delivery plan, the third of the five histories)* **The relay history: the messenger arrives at the observer at rest.** ▸ **DONE 2026-11-16: for the light messenger's wall state and the clock at rest translated to its place, the displacement of the pointer coupled to that clock through "the messenger is not near" is a gated history at the rest rate, for some strength `0 < g ≤ R₀/5`. Angle conditions over two intervals and sizes are hypotheses; UO-16 not met.** [`RecordLayer/UnifiedWallRelay.lean`](../CsdLean4/RecordLayer/UnifiedWallRelay.lean) (4 pins). No new assumption beyond UA-18. `gate_weight_le`, ★ `far_wall_clock_outside_wallArcs_le`, ★★ `wall_relay_gate_weights`, ★★★ `wall_relay_history`. Contract: UD-191; UA-18 "Where" updated; UO-16 annotated. **Scope:** the conditions on predicted angles on `[0, t_a]` and `[t_b, t_c]` and the size conditions are hypotheses (~~#418~~ gives the intervals for straight motion); `A` is the clock at rest translated by any `h`; the reading window lies inside the short passage of the messenger beside `A`; displacements along uncoupled orbits from a product start on the fibres where the profiles vanish. Remainder: with ~~#422~~. | ~~**M–L**~~ done | High — the hypothesis `Hρ` of ~~#405~~ for the wall states | ~~#420~~, ~~#419~~, ~~#417~~, ~~#413~~ | > | ~~422~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 6 of the research delivery plan, the fourth and fifth of the five histories)* **The meeting histories: the moving observer arrives at the observer at rest.** ▸ **DONE 2026-11-16: for the moving clock's wall state and the clock at rest translated to its place, the displacements of the two pointers coupled, one to each clock, through "the two are not near" are gated histories, at the moving rate and at the rest rate. With ~~#420~~ and ~~#421~~ all five histories of ~~#405~~ are theorems for the wall states, separately; angle conditions and sizes are hypotheses; UO-16 not met.** [`RecordLayer/UnifiedWallMeeting.lean`](../CsdLean4/RecordLayer/UnifiedWallMeeting.lean) (5 pins). No new assumption beyond UA-18. `arenaObs_coincProj_prodRay_swap`, ★ `moving_clock_outside_le`, ★★ `wall_meeting_gate_weights`, ★★★ `wall_meeting_history_moving`, ★★★ `wall_meeting_history_rest` (the moving clock as first carrier and the relation reversed, as ~~#405~~ has it). Contract: UD-192; UA-18 "Where" updated; UO-16 annotated; feasibility note updated. **Scope:** the five history theorems are separate: each asserts a strength for its own pointer and takes its own interval and size hypotheses, and one statement for one prepared state with one strength per clock is not written; every condition on predicted angles is a hypothesis; the sizes are not checked for a family; ~~#405~~'s theorem is not yet applied, so no count or Lorentz statement about the walled chain is proved. Remainder: (a) one statement for one prepared state (one strength per clock) → ~~#423~~, which made the strength and the swing bound hypotheses of ~~#420~~ to ~~#422~~ and put the emission condition of ~~#420~~ in the mirror form; (b) the intervals from ~~#414~~ and ~~#418~~ for the wall states, both clock levels → ~~#424~~; (c) the sizes for a family (Task 8); (d) apply ~~#405~~; (e) Tasks 9 to 12. | ~~**L**~~ done | High — the last two of the five histories for the wall states | ~~#421~~, ~~#420~~, ~~#419~~, ~~#415~~ | > | ~~423~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Tasks 6 and 10 of the research delivery plan: one prepared state, and the first application of the top-down theorem)* **The walled chain's run: five counts and the Lorentz statement, from where the carriers are.** ▸ **DONE 2026-11-16: for one prepared state of the light messenger, the clock at rest translated to its place and the moving clock, with one strength per clock, the five pointers of the five-pointer flow have counts and the counts satisfy the Lorentz statement of ~~#402~~, conditional on seven conditions on predicted angles and on sizes. The errors are not shown small; the fibres where the profiles vanish only; UO-16 not met.** [`RecordLayer/UnifiedWallRadar.lean`](../CsdLean4/RecordLayer/UnifiedWallRadar.lean) (2 pins). No new assumption beyond UA-18. `wallRun`, ★★★ `tri_wall_radar_lorentz`. In the same commit the four history theorems of ~~#420~~ to ~~#422~~ take the strength and the clock's swing bound as hypotheses (so that one strength serves every pointer on a clock), and the emission condition of ~~#420~~ is in the mirror form. Contract: UD-193; UD-190 to UD-192 reworded; UA-18 "Where" updated; UO-16 annotated; feasibility note updated. **Scope:** the seven angle conditions and the sizes are hypotheses and are not shown compatible; the bound contains `|t − T|` for each gate time against the kinematic times of ~~#412~~ for a free `d`, and is not shown small; strengths exist, not chosen; every pointer starts at `(1/12, 0)`; the labelling of the basis is arbitrary. Remainder: with ~~#424~~. | ~~**M**~~ done | High — the first statement in which counts of the walled chain's flow satisfy the Lorentz statement, conditionally | ~~#422~~, ~~#421~~, ~~#420~~, ~~#405~~ | > | ~~424~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 4 of the research delivery plan, for the angles the wall states have)* **The seven gate intervals of the walled chain, from straight motion.** ▸ **DONE 2026-11-16: the seven conditions on predicted angles that ~~#423~~ takes follow from thirteen inequalities between the three rates and the ends of the intervals. Real arithmetic; not yet applied to ~~#423~~; UO-16 not met.** [`RecordLayer/UnifiedWallIntervals.lean`](../CsdLean4/RecordLayer/UnifiedWallIntervals.lean) (4 pins). No new assumption. `cos_le_cos_of_abs_sub_two_pi_le`, `mul_mem_of_mem`, ★★★ `radar_gate_intervals`, ★★★ `radar_gate_times` (added in a second commit the same day: explicit gate times, with `w = 3δ`, that are in order and satisfy the thirteen inequalities under six conditions on the rates, the start and the half-angle). Contract: UD-194; UO-16 annotated; feasibility note corrected. **Found:** the moving clock is prepared at a negative packet time, so it starts before the wall and moving into it; a messenger started with it would meet it twice at once, so the messenger must start at the mirror place and the emission is a coincidence with the clock's mirror image. The clock is reflected at the flow time `τ = π/(2 R_B)`; against ~~#412~~ for the distance `π` the relay is early by `b₁ τ/a` and the meeting late by `τ`, a quarter of a tick. **Scope:** three linear functions of the flow time, both clock levels; the clock at rest at `π` exactly; sufficient conditions; the shifts of the kinematic times are not bounded here. Remainder: (a) the six conditions of `radar_gate_times` and the sizes of ~~#423~~ for a family → ~~#425~~; the join and the error shown to vanish (Task 10); (b) Task 9 (an open set of pointer angles); (c) Tasks 11 and 12. | ~~**M**~~ done | High — every hypothesis of ~~#423~~ about angles is now a scalar inequality | ~~#423~~, ~~#418~~, ~~#414~~ | > | ~~425~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 8 of the research delivery plan)* **The family: the rates of the three carriers, the conditions of the gate times, and the sizes.** ▸ **DONE 2026-11-16: for `n = N = 10¹⁶ k⁴`, `Δ = k/K`, messenger rest energy `1/k`, `δ = 1/(10⁵ k)`, gate weight `1/(500 k)`, swing index `k` and every `k ≥ 100`, the six conditions of ~~#424~~'s gate times and every size condition of ~~#423~~ hold. Inequalities between numbers; ~~#423~~ is not applied here; UO-16 not met.** [`RecordLayer/UnifiedRadarFamily.lean`](../CsdLean4/RecordLayer/UnifiedRadarFamily.lean) (22 pins). No new assumption. `radarRateA`, `radarRateB`, `radarRateB'`, `radarTau`, `two_fifths_le_vRel_half_one`, `vRel_level_le`, ★★ `radar_family_rates`, `radar_six_of_bounds`, ★★★ `radar_family_gate_conditions`, the eleven gate times `famTE` to `famTcM`, `radar_sizes_of_bounds`, ★★★ `radar_family_all_conditions`. Contract: UD-195; UA-18 "Where" updated; UO-16 annotated; feasibility note updated. **Scope:** the sizes are stated in a form from which each hypothesis of ~~#423~~ follows (gate times at most `8K`, horizons at least `8K`, closing terms at most `16/2500`, the band); the budget is ~~#415~~'s `wall_values`; `k ≥ 100` for round constants, not the least value; nothing about the flow. Remainder: (a) apply ~~#423~~ to the family and bound its error against the length of the run (Task 10, on the fibres where the profiles vanish) → ~~#426~~; (b) Task 9; (c) Tasks 11 and 12. | ~~**M–L**~~ done | High — every hypothesis of ~~#423~~ is a consequence of proved inequalities for one family | ~~#424~~, ~~#423~~, ~~#415~~ | > | ~~426~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Tasks 8 and 10 of the research delivery plan, on the fibres where the profiles vanish)* **The family's run: five counts and the Lorentz statement with a vanishing error.** ▸ **DONE 2026-11-16: for every `k ≥ 100` and every even ring large enough, the five pointers of the walled chain's five-pointer flow have counts in five windows and the counts satisfy the Lorentz statement of ~~#402~~ for `v = vRel 1 1 (1/2)` within `150 K/k`, about `21/k` of the length of the run. No hypothesis about angles, sizes, histories, light speed, the dilation law or the radar relations is left. On the fibres where the five profiles vanish only: UO-16 not met.** [`RecordLayer/UnifiedRadarRun.lean`](../CsdLean4/RecordLayer/UnifiedRadarRun.lean) (4 pins) and two declarations in `RecordLayer/UnifiedWallRadar.lean`. No new assumption beyond UA-18. `WallRadarLorentz`, `WallRadarLorentz.mono` (the conclusion of ~~#423~~ as one statement about a run; ~~#423~~ now concludes it), `radar_error_le`, `radar_error_pieces`, `radarRateB_mul_wallK`, ★★★ `radar_family_lorentz`. Contract: UD-196; UA-18 "Where" updated; UO-16 annotated; feasibility note updated. **Scope:** every pointer starts at `(1/12, 0)`, a set of measure zero in the pointer angles, so the plan's definition of success (a non-empty open set) is not met; the comparison of `150 K/k` with the run (`≈ 7K`, `≈ 1.1 k` ticks) is arithmetic in prose; one velocity, the distance the whole chain; the emission is a mirror coincidence (~~#424~~) and its shifts of the kinematic times are inside the error; strengths exist, not chosen. Remainder: (a) Task 9, an open set of pointer angles (R7), the last scientific step → first part ~~#427~~; (b) the capstone under the plan's name once (a) holds; (c) Tasks 11 and 12. | ~~**L**~~ done | High — the first unconditional Lorentz statement about counts of the walled chain's flow, with an error that vanishes along a family | ~~#425~~, ~~#424~~, ~~#423~~ | > | ~~427~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 9 of the research delivery plan, first part)* **Histories are stable under a small change of the displacement.** ▸ **DONE 2026-11-16: a gated history and an away history survive a displacement that starts within `η₀` and drifts at a rate at most `λ`, with errors enlarged by `η₀ + λ t` and speeds by `λ`. Real analysis; one of five pieces of R7; UO-16 not met.** [`RecordLayer/UnifiedRadarStability.lean`](../CsdLean4/RecordLayer/UnifiedRadarStability.lean) (3 pins). No new assumption. `abs_sub_le_of_drift`, ★★ `IsGatedHistory.of_drift`, ★★ `IsAwayHistory.of_drift`. Contract: UD-197; UO-16 annotated; feasibility note updated. **Scope:** the true closing bounds of the unperturbed displacement and the band for the enlarged error are hypotheses, because the structures of ~~#402~~ keep only `1/6`; nothing about the flow. Remainder (the rest of Task 9; (a) and (b) → ~~#428~~): (a) the five-pointer flow on an arbitrary fibre: the conserved angles are constants, so the carrier moves by the free flow of `H + Σ f_i(θ_i) G_i` and pointer `i` is turned by `f_i'(θ_i)` times the accumulated expectation of `G_i` (generalise `pentaRun_eq` of ~~#404~~; needs all five gates Hermitian); (b) the drift rate from the bound on the change of an expectation under a change of the generator (`abs_arenaObs_idxCarrierU_sub_le` is a candidate); (c) the true closing bounds of the wall histories; (d) counts for a start angle near `1/12`; (e) the open set for the family and the join with ~~#426~~; then the capstone under the plan's name and Tasks 11 and 12. | ~~**S–M**~~ done | High — the first piece of the last scientific step | ~~#426~~, ~~#402~~ | > | ~~428~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 9 of the research delivery plan, second and third parts)* **The five-pointer run on any fibre, and the drift of a pointer's displacement.** ▸ **DONE 2026-11-16: the five-pointer flow is solved exactly on every fibre (the carrier moves by the free flow of one constant matrix, the uncoupled one plus the five observables with the five conserved profile values), and a pointer's displacement there drifts from the uncoupled one at a rate at most `4 |z'| ‖G‖ T ‖H' − H‖ + 2 |z' − z| ‖G‖`. Three of five pieces of R7; UO-16 not met.** [`RecordLayer/UnifiedFibreRun.lean`](../CsdLean4/RecordLayer/UnifiedFibreRun.lean) (8 pins), [`RecordLayer/UnifiedFibreDrift.lean`](../CsdLean4/RecordLayer/UnifiedFibreDrift.lean) (5 pins). No new assumption beyond UA-18. `freeTurn_congr`, `twoCoupledFlow_fibre`, `threeCoupledFlow_fibre`, `fourCoupledFlow_fibre`, ★★ `fiveCoupledFlow_fibre`, `fibreH`, `fibreH_reindex`, ★★★ `pentaRun_fibre`; `periodic_const_one`, `idxCarrierU_const_eq`, ★ `abs_arenaObs_idxFreeU_sub_le`, `freeShift_zero`, ★★ `freeShift_drift`. Contract: UD-198; UO-16 annotated; feasibility note updated. **Scope:** the run is exact and needs all five observables Hermitian; the drift is a bound in the operator norm with the factor `T`; nothing about histories or counts. Remainder (the rest of Task 9): (c) the true closing bounds of the wall histories (`coincShift_stop_le` of ~~#400~~ and the proof of ~~#419~~ have them); (d) counts for a start angle near `1/12` → ~~#429~~; (e) the open set for the family: the five gates Hermitian, the rate small by the choice of angles, the profile `g · trainF` near its zero, and the join with ~~#426~~; then the capstone under the plan's name and Tasks 11 and 12. | ~~**M**~~ done | High — the flow off the zero fibres, exactly | ~~#427~~, ~~#404~~ | > | ~~429~~ | Records | Records→spacetime | *new (2026-11-16, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 9 of the research delivery plan, fourth part)* **M6 from the top, for the five-pointer flow from any pointer angles.** ▸ **DONE 2026-11-16: the top-down theorem of ~~#404~~ without its restrictions to the fibres where the profiles vanish and to pointers started at `1/12`: for any conserved angles and moving angles `1/12 + η_i`, if the offset displacements `η_i + freeShift G_i` along the free orbit of the fibre's matrix have the five histories of ~~#402~~, the five pointers have counts and the counts satisfy the Lorentz statement. Conditional on those histories; UO-16 not met.** [`RecordLayer/UnifiedFibreRadar.lean`](../CsdLean4/RecordLayer/UnifiedFibreRadar.lean) (2 pins). No new assumption beyond UA-18. `pshift_eq_of_start`, ★★★ `penta_radar_lorentz_fibre`. Contract: UD-199; UO-16 annotated; feasibility note updated. **Scope:** the histories are those of the coupled flow's displacements; nothing here shows they hold near the fibres where the profiles vanish; all five observables Hermitian. Remainder (what is left of Task 9; all of it → ~~#430~~, ~~#431~~): (a) `tri_wall_five_histories`: the five uncoupled histories of the wall states as an exported statement (they are inside the proof of ~~#423~~), with their true closing bounds from the gate weights of ~~#420~~ to ~~#422~~ and `abs_coincShift_sub_le`; (b) the three remaining gates Hermitian (`1 ⊗ gateOp`, `gateOp … (Ya ⊗ 1)`, `1 ⊗ gateOpL`); (c) the profile `g · trainF` and its derivative near `0`; (d) the open set for the family: radii in terms of `k`, the ring and the norms of the gates, the enlarged errors against the margins of ~~#425~~, via ~~#427~~ and ~~#428~~; (e) the join with ~~#426~~ under the plan's name `csd_operational_lorentz_from_records`; then Tasks 11 and 12. | ~~**S–M**~~ done | High — the top of the chain no longer assumes the zero fibres | ~~#428~~, ~~#404~~, ~~#402~~ | > | ~~430~~ | Records | Records→spacetime | *new (2026-11-17, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Task 9 of the research delivery plan, fifth and sixth parts)* **The walled chain's run from an open set of pointer angles.** ▸ **DONE 2026-11-17: for three carriers and five gates with the profiles `g_i · trainF`, if the five two-carrier gate displacements have the histories of ~~#402~~, each moves by less than `1/6` while its gate is closed and the away band holds strictly, then for every `ϵ > 0` there is a radius `ρ > 0` (the same for every labelling of the basis) such that from every five pointer angles within `ρ` of `(1/12, 0)` the five pointers have counts and the counts satisfy the Lorentz statement with the errors enlarged by `ϵ`; and the walled chain's instance under the hypotheses of ~~#423~~ with the closing sizes strict. UO-16 not declared met.** [`RecordLayer/UnifiedTriFibre.lean`](../CsdLean4/RecordLayer/UnifiedTriFibre.lean) (24 pins), [`RecordLayer/UnifiedWallOpen.lean`](../CsdLean4/RecordLayer/UnifiedWallOpen.lean) (3 pins). No new assumption beyond UA-18. `abs_trainF_le`, `scaledTrainF_lift_coe`, `scaledTrainF_deriv_lift_coe`, ★ `scaledTrainF_near`, `gatedErr_add`, `awayErr_add`, `exists_pos_le`, `exists_radius`, `isHermitian_real_smul`, `fibreH_isHermitian`, `norm_fibreH_sub_le`, ★ `fibre_rate_le`, ★★ `IsGatedHistory.of_fibre`, ★★ `IsAwayHistory.of_fibre`, `coincShift_closed_le`, `gateOp_fst_isHermitian`, `gateOpL_isHermitian`, `freeShift_tri_gate_snd`, `freeShift_tri_gate_fst`, `freeShift_tri_pair`, `freeShift_tri_pairL`, `NearStart`, `nearStart_start`, ★★★ `tri_radar_lorentz_open`; `wallRunAt`, `wallRun_eq_wallRunAt`, ★★★ `tri_wall_radar_lorentz_open`. Contract: UD-200; UO-16 annotated; feasibility note updated. **Scope:** the radius exists for each `ϵ` and is not explicit (strengths, flow times, slack in the sizes, operator norms of the five gates); the walled chain's statement is still conditional on the bounds on where the carriers are, the seven angle conditions and the sizes; the plan's name `tri_wall_five_histories` is not a theorem: the five uncoupled histories are consumed inside `tri_wall_radar_lorentz_open`. Remainder: the family → ~~#431~~. | ~~**M–L**~~ done | High — the run no longer needs the zero fibres | ~~#429~~, ~~#428~~, ~~#427~~, ~~#423~~ | > | ~~431~~ | Records | Records→spacetime | *new (2026-11-17, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Tasks 9 and 10 of the research delivery plan)* **The family's run on an open set of pointer angles: the plan's theorem.** ▸ **DONE 2026-11-17: for every `k ≥ 100` and every even ring with `M ≥ 210000 N³`: strengths and an open set of the ten pointer angles, containing the start of every pointer at `(1/12, 0)`, from every point of which the five pointers of the walled chain's run have counts at every choice of flow times in the five reading windows, and the counts satisfy both components of the Lorentz statement of ~~#402~~ for `v = vRel 1 1 (1/2)` within `151 K/k`, which is `151/(π k)` of the distance `π K` and tends to `0`. No hypothesis on histories, the messenger's speed, a dilation law or radar relations. UO-16 not declared met.** [`RecordLayer/UnifiedRadarGeometry.lean`](../CsdLean4/RecordLayer/UnifiedRadarGeometry.lean) (4 pins); `radar_family_lorentz` in [`RecordLayer/UnifiedRadarRun.lean`](../CsdLean4/RecordLayer/UnifiedRadarRun.lean) strengthened with the open statement (hypotheses unchanged). No new assumption beyond UA-18. `isOpen_nearStart`, ★★★ `csd_operational_lorentz_from_records`, `radar_error_relative`, `tendsto_radar_error_relative`. Contract: UD-201; UO-16 annotated; feasibility note updated. **Scope:** the open set is a product of intervals of an unnamed radius `ρ > 0`, one for each `k`; nothing bounds the radius below as `k` grows (the drift rate carries the flow time `≈ 8K` and the norms of the gates) and nothing is said about its measure; one family, one velocity, one distance, `1 + 1`; the Lorentz statement is about five tick counts read in flow-time windows; the strengths exist and are not given by a formula. Remainder ((b), (c) and part of Task 10 → ~~#432~~): (a) a radius bounded below along the family, or a statement of how it shrinks (needs bounds on the norms of the five gates and the explicit slack of ~~#425~~); (b) Task 11, the bridge (`UnifiedRecordSpacetimeBridge.lean`); (c) Task 12, the audit of the theorem against the test of UO-16; (d) the author's decision on UO-16. | ~~**M**~~ done | High — the plan's target theorem | ~~#430~~, ~~#426~~ | > | ~~432~~ | Records | Records→spacetime | *new (2026-11-17, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — Tasks 10, 11 and 12 of the research delivery plan)* **Radar coordinates, the common interval, the bridge and the audit.** ▸ **DONE 2026-11-17: the plan's theorem now states the two observers' coordinates of the relay event as functions of the five counts, the comparison of their intervals (a boost preserves `t² − x²`), and that the open set of pointer angles has positive measure; the run is identified as one orbit of the arena's Hamiltonian flow, which preserves `μ_FS ⊗ vol⁵`; the audit is written. UO-16 not marked complete.** [`RecordLayer/UnifiedRadarGeometry.lean`](../CsdLean4/RecordLayer/UnifiedRadarGeometry.lean) (+9 pins), [`RecordLayer/UnifiedRecordSpacetimeBridge.lean`](../CsdLean4/RecordLayer/UnifiedRecordSpacetimeBridge.lean) (5 pins), [`specs/unified-m6-audit.md`](unified-m6-audit.md). No new assumption beyond UA-18. `restEvent`, `movingEvent`, `radarInterval`, `lorentzBoost_interval`, `abs_sq_sub_sq_le_of_abs_sub_le`, `WallRadarGeometry`, ★★ `WallRadarLorentz.geometry`, `isOpenPosMeasure_volume_prod`, `sFinite_volume_prod`, ★★★ `csd_operational_lorentz_from_records` (restated); `IsTickCount.exists_crossings`, `pointerReady_disjoint_pointerDisplay`, ★ `triRun_eq_arenaFlow`, ★ `triFlow_map_pentaMuL`, ★★ `wallRunAt_orbit`. Contract: UD-202; UO-16 annotated; `specs/POSITS.md` Posit 11; `specs/connectivity-manifest.md` L10. **Scope:** the interval's error is `E₁ (2|b₁| + E₁) + E₂ (2|b₂| + E₂)` in the boosted coordinates; the identifications are definitional or instances of ~~#404~~'s arena. Remainder (the audit's findings; F1 → ~~#433~~): (F1) export the bounds on the counts from ~~#402~~ and show the interval's residual vanishes against the interval; (F2) an explicit radius for the open set (Remainder (a) of ~~#431~~); then the author's decision on UO-16. | ~~**M**~~ done | High — Tasks 10 to 12 of the plan | ~~#431~~, ~~#404~~, ~~#402~~ | > | ~~433~~ | Records | Records→spacetime | *new (2026-11-17, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — finding F1 of the audit of ~~#432~~)* **The interval's residual vanishes against the squared distance.** ▸ **DONE 2026-11-17: the statement of ~~#402~~ and every statement built on it now also bounds the rest observer's time between the relay and the meeting; for the family the bound is `6K`; so the two observers' intervals in `csd_operational_lorentz_from_records` differ by at most `4600 K²/k`, which is `4600/(π² k)` of the squared distance `(πK)²` and tends to `0`. UO-16 not marked complete: the mark is the author's.** Conclusions strengthened in [`UnifiedRadarSkeleton.lean`](../CsdLean4/RecordLayer/UnifiedRadarSkeleton.lean), `UnifiedPentaRun.lean`, `UnifiedFibreRadar.lean`, `UnifiedTriCarrier.lean`, `UnifiedTriFibre.lean`, `UnifiedWallRadar.lean` (`WallRadarLorentz` has a new last argument, the size), `UnifiedWallOpen.lean`, `UnifiedRadarRun.lean`, [`UnifiedRadarGeometry.lean`](../CsdLean4/RecordLayer/UnifiedRadarGeometry.lean) (+5 pins in all). No new assumption beyond UA-18. `gatedErr_div_le`, `radar_rest_time_le`, ★★ `radar_interval_error_le`, `radar_interval_error_relative`, `tendsto_radar_interval_error_relative`. Contract: UD-203; UO-16 annotated; audit note updated (F1 closed). **Scope:** the comparison scale is the squared distance; no lower bound on the interval itself; the constant is not sharp. Remainder: (F2) an explicit radius for the open set (worked estimate `1/(20000 k²)`: gate norms at most `1`, the band slack of ~~#425~~ as a number, an explicit-radius form of `tri_radar_lorentz_open`); a lower bound on the interval; the author's decision on UO-16. | ~~**M**~~ done | High — the clause of UO-16 about the interval | ~~#432~~, ~~#402~~ | > | ~~434~~ | Records | Records→spacetime | *new (2026-11-17, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — residue of ~~#433~~, and the author's decision)* **The rest observer's time against the kinematic time; UO-16 marked complete.** ▸ **DONE 2026-11-17: the size conjunct of ~~#433~~ in its two-sided form through the whole chain; for the family the rest observer's time is within `52 K/k` of `πK − πK/v`, at least `3K` in size, so its interval is at least `9 K²` and the two observers' intervals agree within `512/k` of the interval. UO-16 MARKED COMPLETE BY THE AUTHOR, 2026-11-17, with the findings of the audit as the scope of the mark.** `WallRadarLorentz` and `WallRadarGeometry` take `T₀ B` in place of `B`; `radar_rest_time_le` replaces the one-sided lemma of ~~#433~~; [`UnifiedRadarGeometry.lean`](../CsdLean4/RecordLayer/UnifiedRadarGeometry.lean) (+2 pins): `radar_kinematic_time_bounds`, ★★ `radar_rest_time_ge`, `radar_interval_error_le` (restated). No new assumption beyond UA-18. Contract: UD-204; UO-16 annotated as marked; audit, feasibility note, manifest L10 updated. **Scope of the mark:** an operational reconstruction (Posit 11), one family, `1 + 1`, windows in the flow time, strengths that exist, an open set of positive measure for each `k`. Remainder: (F2) an explicit radius for the open set. | ~~**S–M**~~ done | High — M6 | ~~#433~~ | > | ~~435~~ | Records | Records→spacetime | *new (2026-11-17, the author's programme "From `Σ` to physical spacetime", obligation UO-16 — finding F2 of the audit of ~~#432~~)* **An explicit radius for the open set of pointer angles.** ▸ **DONE 2026-11-17: the plan's theorem holds from every five pointer angles within `1/(12000 k²)` of the start `(1/12, 0)`, for every ring: the five gates have operator norm at most `1`, the open theorem of ~~#430~~ takes its radius as a hypothesis (three explicit conditions), and the family meets them with slack `1/40`.** [`RecordLayer/UnifiedTriFibre.lean`](../CsdLean4/RecordLayer/UnifiedTriFibre.lean) (+6 pins), [`RecordLayer/UnifiedRadarRun.lean`](../CsdLean4/RecordLayer/UnifiedRadarRun.lean) (+3 pins); `tri_wall_radar_lorentz_open` has a third clause; `radar_family_lorentz` and `csd_operational_lorentz_from_records` restated with the radius. No new assumption beyond UA-18. `norm_le_one_of_isHermitian_idem`, `norm_le_one_of_isHermitian_invol`, `norm_idem_mul_invol_le_one`, `norm_gateOp_le_one`, ★ `tri_gate_norms_le_one`, ★★★ `tri_radar_lorentz_near`, `radar_band_le`, `radar_radius_A`, `radar_radius_B`. Contract: UD-205; audit note updated (F2 closed). **Scope:** the radius is not sharp and shrinks like `1/k²`. Remainder: none under UO-16; M7 (`3 + 1`) is the next milestone of the programme. | ~~**M–L**~~ done | Medium — the plan does not require it | ~~#434~~, ~~#430~~ | > | ~~436~~ | Records | Records→spacetime | *new (2026-11-18, the author's plan "Unified records, quantum measurement and emergent spacetime", workstream R, Task 1; obligation UO-25)* **Tick counts as recorded facts of the SigmaLayer.** ▸ **DONE 2026-11-18: each tick count of the radar theorem is a recorded fact whose event is a measurable set of initial states of the arena, defined by the Hamiltonian flow; the five counts of ~~#432~~ are a record history whose compatible region contains the run's start.** [`RecordLayer/UnifiedRecordHistoryBridge.lean`](../CsdLean4/RecordLayer/UnifiedRecordHistoryBridge.lean) (+42 pins): ★★ `hasTicks_succ_iff` (one more tick through two rational times), `IsSigmaClosed`, `hitSet`, ★ `isClosed_hitSet`, ★★ `measurableSet_hasTicks`, ★★ `measurableSet_isTickCount`; `tickSignature`, `TickReadout`, `TickReadout.semantics`, ★★ `TickReadout.mem_event_startFact`, ★★ `TickReadout.crossings_of_mem_event`, `TickReadout.model`, ★ `TickReadout.recordsEstablishedOutcome`, ★★ `TickReadout.establishedFact_eq_some_iff`, ★★ `TickReadout.mem_compatibleSet_startFacts`; `pentaDynamics`, ★ `continuous_fiveCoupledFlow_prod`, `pentaReadout`, `wallDynamics`, `wallReadout`, ★ `wallRunAt_eq_flow`, `radarHistory`, `WallRadarRecords`, ★★★ `WallRadarGeometry.records`, ★★★ `csd_radar_record_history`. Contract: UD-208, UO-25. **Scope:** a crossing, a recorded fact and a register are kept apart, and only the first two are given; the carriers' rays are fixed, so the compatible region is not shown to have positive measure for `μ_FS ⊗ vol⁵` and no `HistoryPreparation` is built; the readout is a function of the state through the flow. Remainder: (a) the preparation (Task 2: positive measure for a neighbourhood of carrier rays, or a fibre measure by disintegration); (b) the macroscopic projection (Task 3). | ~~**M**~~ done | High — the plan's first critical task | ~~#432~~, ~~#435~~ | > | 437 | Records | Records→spacetime | *new (2026-11-18, the author's decision under UO-17: choice (c) of [`unified-m7-feasibility.md`](unified-m7-feasibility.md) §7 is to be worked through; obligation UO-26)* **A return caused at the event.** On the arena of UA-18 a messenger turns round only at a flat wall, so a moving group in general position cannot radar one common event (the note's §4). **Build:** a mechanism by which a messenger reaching a place causes a signal to leave it, the departure fixed by the flow and not by the preparation: (1) no signal before the meeting, within an explicit error; (2) a signal does leave, with its weight, place and time of launch within explicit errors; (3) a gate reads it within the budget its weight allows; (4) the radar identities of ~~#293~~ for a moving observer from those readings; (5) one arena, including the pointers' action on the carriers off the fibres where the profiles vanish. **The mechanism is not fixed** (a contact term that changes the level of a carrier resting at the event, or a reading of the part a point obstacle scatters); either needs a clause beside UA-18 (c) first. ▸ **FIRST BRICK DONE 2026-11-18.** [`Mathlib/Analysis/Matrix/DuhamelState.lean`](../CsdLean4/Mathlib/Analysis/Matrix/DuhamelState.lean) (+4 pins): `Matrix.continuous_exp_smul`, ★★ `Matrix.norm_exp_smul_sub_exp_smul_mul_le`, ★★ `Matrix.norm_exp_neg_I_add_sub_mul_le` (`‖(e^{−it(H₀+V)} − e^{−itH₀}) X‖ ≤ ∫₀ᵗ ‖V e^{−isH₀} X‖ ds`), ★★★ `Matrix.norm_mul_exp_neg_I_add_mul_le` (**nothing reaches a sector the free orbit avoids, except through `V` acting on that orbit**). Contract: UD-211. **Scope:** an upper bound; no mechanism, carrier or arena; nothing is shown to be launched. ▸ **MECHANISM DECIDED by the author 2026-11-18: the transponder → UA-19** (a two-level carrier resting at the event; the contact term is UA-18 (c)'s operator `G · X_T`, in the carriers' energy on the fibres where a profile does not vanish, so no term of a new kind). ▸ **SECOND BRICK DONE 2026-11-18.** [`Mathlib/Analysis/Matrix/DuhamelFirstOrder.lean`](../CsdLean4/Mathlib/Analysis/Matrix/DuhamelFirstOrder.lean) (+11 pins): `Matrix.duhamelKernel`, ★ `Matrix.duhamelFirst`, `Matrix.hasDerivAt_duhamelKernel`, `Matrix.norm_duhamelFirst_le`, ★★★ `Matrix.norm_exp_sub_exp_mul_sub_duhamelFirst_le`, ★★★ `Matrix.norm_exp_neg_I_add_sub_mul_sub_duhamelFirst_le` (**the perturbed orbit is the free one plus the first-order term, within `‖V‖² ‖X‖ t² / 2`**), ★★★ `Matrix.le_norm_mul_exp_neg_I_add_mul` (a lower bound for what reaches a sector the free orbit avoids). Contract: UD-212. **Scope:** the first-order term is not evaluated, so no launch is shown; weak coupling or short times. ▸ **FOURTH BRICK DONE 2026-11-19 (item (1) for the transponder's structure).** [`Mathlib/Analysis/Matrix/DuhamelSector.lean`](../CsdLean4/Mathlib/Analysis/Matrix/DuhamelSector.lean) (+3 pins; ★★★ `Matrix.norm_mul_exp_neg_I_add_mul_le_of_commute`) and [`RecordLayer/UnifiedTransponder.lean`](../CsdLean4/RecordLayer/UnifiedTransponder.lean) (+5 pins; `levelProj`, `levelProj_mul_blockDiagonal`, `one_kronecker_mul_compositeH`, ★★★ `transponder_launch_le`: **the second level is populated by at most `‖Q‖ |g| ∫₀ᵗ ‖G e^{−isH₀} X‖ ds`**). Contract: UD-224. **Scope:** an upper bound; `G` any Hermitian matrix. ▸ **FIFTH BRICK DONE 2026-11-19.** ★★★ `transponder_launch_le_coincidence` (`RecordLayer/UnifiedTransponder.lean`, +1 pin): the contact term `gateOp σ τ r (levelSwap n)` substituted; bound `‖Q‖ |g| ∫₀ᵗ ‖P e^{−isH₀} X‖ ds`, `P` the coincidence projector. Contract: UD-226. **Remainder (this row stays open):** (a) that integrand as the coincidence weight of two packets, bounded from their boxes (UD-221, UD-222) — the bound at one time done 2026-11-19: [`RecordLayer/CoincidenceBoxes.lean`](../CsdLean4/RecordLayer/CoincidenceBoxes.lean) (+4 pins; ★★★ `sum_norm_sq_coincProj_mulVec_prod_le`, ★★ `sum_norm_sq_coincProj_mulVec_prodD_le`; contract UD-236); the composite's free orbit of a product done 2026-11-19: [`RecordLayer/CoincidenceOrbit.lean`](../CsdLean4/RecordLayer/CoincidenceOrbit.lean) (+2 pins; ★★ `exp_compositeH_mulVec_prod`, ★★★ `sum_norm_sq_coincProj_freeOrbit_prod_le`; contract UD-240); left of (a): the moving boxes, and the passage to the operator norm under the integral; the first-order term for product packets (its weight, place and time of launch); ~~(b) a remainder in terms of the orbit~~ (done: [`Mathlib/Analysis/Matrix/DuhamelOrbitRemainder.lean`](../CsdLean4/Mathlib/Analysis/Matrix/DuhamelOrbitRemainder.lean), +6 pins, and +2 in `DuhamelFirstOrder.lean`; ★★★ `Matrix.norm_exp_neg_I_add_sub_mul_sub_duhamelFirst_le_orbit`, remainder `≤ ‖V‖ (∫₀ᵗ ‖V e^{−isH₀} X‖ ds) t`, ★★★ `Matrix.le_norm_mul_exp_neg_I_add_mul_orbit`; contract UD-216; smaller than the first-order bound only for `‖V‖ t < 1`); (c) the gate that reads the second level, with a budget for a small weight; (d) the radar identities from those readings; (e) one arena. | **L** | High — the moving group's chart from records (UO-17) rests on it | ~~#293~~; the author's choice of mechanism | > | ~~438~~ | Records | Records→spacetime | *new (2026-11-18, the author's plan "Unified records, quantum measurement and emergent spacetime", workstream R, Task 2; obligation UO-25 (2); residue (a) of ~~#436~~)* **A preparation for the five-record history.** The compatible region of the five radar facts (UD-208) contains the start for one carrier ray and an open set of pointer angles; one ray has Fubini–Study measure zero. **Build:** a set of positive `μ_FS ⊗ vol⁵` measure inside the compatible region, and the `HistoryPreparation` on it, by the route of a neighbourhood of rays: (a) a history survives a small turn of the ray; (b) the family's five histories, with their slack, at every ray turned by `W` with `‖W − 1‖` below an explicit radius; (c) the counts and the Lorentz bounds on turned rays times the box of angles, from the general-ray theorem `penta_radar_lorentz_fibre` with (a) and the drift over the fibres together (a turned ray is not a product of the three carriers' rays, and product rays are themselves a null set, so the three-carrier open theorem of UD-205 does not apply to it); (d) the turned rays have positive Fubini–Study measure (the measure is the image of Haar measure under the orbit map, `fsMeasure_pos_of_isOpen`'s argument); (e) the `HistoryPreparation`. Positive measure in the pointer fibres is not positive measure in the arena; the fibre measure by disintegration is the other route and is not taken here. ▸ **FIRST BRICK DONE 2026-11-18 ((a)).** [`RecordLayer/UnifiedRayDrift.lean`](../CsdLean4/RecordLayer/UnifiedRayDrift.lean) (+5 pins): ★★ `abs_arenaObs_kick_kick_sub_le`, ★★ `freeShift_ray_drift` (rate `4 |z| ‖G‖ ‖W − 1‖`), `abs_freeShift_ray_sub_le`, ★★★ `IsGatedHistory.of_ray`, ★★★ `IsAwayHistory.of_ray`. Contract: UD-213. **Scope:** one displacement along the free orbit; no measure. ▸ **SECOND BRICK DONE 2026-11-18 ((c), and (b) for any carrier as a condition).** ★★★ `penta_radar_lorentz_near` ([`RecordLayer/UnifiedTriFibre.lean`](../CsdLean4/RecordLayer/UnifiedTriFibre.lean), +1 pin; the open statement of UD-205 for any carrier and **any ray**, `tri_radar_lorentz_near` now its corollary) and ★★★ `penta_radar_lorentz_near_ray` ([`RecordLayer/UnifiedRayRadar.lean`](../CsdLean4/RecordLayer/UnifiedRayRadar.lean), +1 pin): **counts and Lorentz bounds from the ray turned by any `W` with `‖W − 1‖ ≤ r_W`, `4 γ S r_W T ≤ s/2`, times the box of angles of radius `ρ ≤ s/8`**. Contract: UD-214. **Scope:** conditional on the five histories at the unturned ray; the family not instantiated; no measure. **Remainder (this row stays open):** (b′) the family's five histories with a named slack as a statement (they are inside the proof of `radar_family_lorentz`), and the family's instance of `penta_radar_lorentz_near_ray`; ~~(d) the measure of the turned rays~~ (done, below); (e) the `HistoryPreparation`. ▸ **THIRD BRICK DONE 2026-11-18 ((d)).** [`RecordLayer/UnifiedTurnedRays.lean`](../CsdLean4/RecordLayer/UnifiedTurnedRays.lean) (+4 pins): `turnedRays`, ★★★ `fsMeasure_turnedRays_pos` (**the rays turned by `W` with `‖W − 1‖ < r` have positive Fubini–Study measure, for any reference point**). Contract: UD-215. **Scope:** positive with no value; outer measure of a set not shown measurable; in the carrier's own projective space. ▸ **FOURTH BRICK DONE 2026-11-19 ((b′): the family instantiated).** `penta_radar_lorentz_near_ray` restated with the errors enlarged by `ϵ + ϵ′` (no enlarged speeds) and ★★★ `tri_radar_lorentz_near_ray` added ([`RecordLayer/UnifiedRayRadar.lean`](../CsdLean4/RecordLayer/UnifiedRayRadar.lean), +1 pin); `wallRunRay`, ★★★ `tri_wall_radar_lorentz_ray` ([`RecordLayer/UnifiedWallRay.lean`](../CsdLean4/RecordLayer/UnifiedWallRay.lean), +3 pins); `ray_error_of_radius`, ★★★ `radar_family_lorentz_ray` ([`RecordLayer/UnifiedRadarRay.lean`](../CsdLean4/RecordLayer/UnifiedRadarRay.lean), +2 pins): **for every `k ≥ 100`, from every ray within `1/(12000 k²)` of the family's and every five pointer angles within `1/(24000 k²)` of the start, the counts satisfy the Lorentz statement within `152 K/k`**. Contract: UD-218. **Scope:** the strengths are asserted to exist, not stated equal to those of ~~#436~~'s theorem; no measure in the arena. ▸ **DONE 2026-11-19 ((e)): a `HistoryPreparation` of non-zero arena measure.** [`RecordLayer/UnifiedRadarPreparation.lean`](../CsdLean4/RecordLayer/UnifiedRadarPreparation.lean) (+8 pins): ★★ `fsMeasure_image_turnedRays_pos`, `RadarCountBounds`, `radarStartRay`, ★ `wallRunRay_eq_flow`, ★★ `WallRadarGeometry.mem_compatibleSet_ray`, `radarStartSet`, ★★★ `pentaMuL_radarStartSet_pos`, ★★★ `csd_radar_history_preparation` (**for every `k ≥ 100`: strengths exist such that for every labelling, reference point and reading times there are five counts whose history has a compatible region of non-zero `μ_FS ⊗ vol⁵` measure and which satisfy the Lorentz statement within `152 K/k`**). Contract: UD-219; UO-25 (2) met with conditions. **Scope:** the counts are not named (covering argument over countably many histories); the measure has no value; ordinary conditioning, not an interaction; no register. **Residue:** a named history with a lower bound for its measure (needs the counts to be constant on the set, i.e. uniqueness of a tick count plus continuity), if a later task needs it; Tasks 3 and 6 can start from UD-219 as it stands. | ~~**M–L**~~ done | High — the plan's Task 2; Tasks 3 and 6 condition on it | ~~#436~~; ~~#427~~; ~~#430~~ | > | 439 | Records | Records→spacetime | *new (2026-11-19, the author's plan "Unified records, quantum measurement and emergent spacetime", workstream R, Task 3; obligation UO-25 (3))* **The macroscopic projection of the radar records.** **Build:** the radar's tick records as cells of a macroscopic projection in the frame of ~~#99~~, reusing `MacroProjection` and `FactorsThrough`: (a) the projection for any tick readout and "a record history is one macrostate"; (b) the radar's instance: a bound on the five counts and the cell of non-zero measure from ~~#438~~; (c) what stability of the cells can be shown (the cells are fixed at the start by the flow up to the reading times; persistence under later readings is the open part). The Born weights of these events are Task 6, not this row. ▸ **FIRST BRICK DONE 2026-11-19 ((a)).** [`RecordLayer/UnifiedRadarMacrostate.lean`](../CsdLean4/RecordLayer/UnifiedRadarMacrostate.lean) (+11 pins): `TickReadout.countEvent`, `TickReadout.tickCode`, ★★ `TickReadout.tickMacro`, ★★ `TickReadout.countEvent_eq_preimage`, ★★★ `TickReadout.compatibleSet_eq_preimage_singleton` (**a record history of the family is one macrostate**). Contract: UD-223. **Scope:** the construction only; not instantiated for the radar; no stability. ▸ **SECOND BRICK DONE 2026-11-19 ((b)).** ★★★ `csd_radar_macrostate` (same module, +1 pin): **the radar's history is one cell of the projection of the five readings, of non-zero `μ_FS ⊗ vol⁵` measure, with counts satisfying the Lorentz comparisons**. Contract: UD-225. The cell is not named. ▸ `TickReadout.map_tickMacro_singleton` added 2026-11-19 (+1 pin): the pushed-forward measure of a string is the arena measure of its history's compatible region (item 3 of the author's five; plan [`unified-records-spacetime-task-plans.md`](unified-records-spacetime-task-plans.md) §1). ▸ **THIRD BRICK DONE 2026-11-19.** [`RecordLayer/TickCountBound.lean`](../CsdLean4/RecordLayer/TickCountBound.lean) (+3 pins): ★★★ `HasTicks.mul_le` (**`n` ticks by `T` need `n d ≤ K T`**), `IsTickCount.mul_le`, `IsTickCount.le_div`. Contract: UD-228. Not instantiated. ⚠️ The count is **not** a function of the history (`IsGatedHistory.exists_count` fixes it only within `3v(t_b − t_a) + 7/2` ticks), so the radar's cell splits. ▸ **FOURTH BRICK DONE 2026-11-19.** [`RecordLayer/PointerCountCap.lean`](../CsdLean4/RecordLayer/PointerCountCap.lean) (+11 pins): ★★ `one_sixth_le_dist_of_mem_readyArc_of_mem_displayArc`, ★★ `HasTicks.pointer_mul_le`, ★★★ `pentaCount_le`, ★★★ `wallCount_le` (**the radar's cap: `12 max(|g_A|, |g_B|) T` from every start**), ★★★ `TickReadout.exists_history_measure_ge`, ★★★ `wall_exists_history_measure_ge` (**one history holds at least `1/(B+1)^5` of the measure of any set of starts**). Contract: UD-232. **Fifth brick, 2026-11-19:** [`RecordLayer/RadarHistoryShare.lean`](../CsdLean4/RecordLayer/RadarHistoryShare.lean) (+2 pins): ★★ `TickReadout.exists_history_measure_ge_of_forall`, ★★★ `csd_radar_history_share` (**the pigeonhole history's counts satisfy `RadarCountBounds` on the start set of UD-219**). Contract: UD-237. **Remainder (this row stays open):** (c) the start set's measure as a number; that the reading windows lie in `t ≥ 0`; what stability can be stated. | **M** | High — the plan's Task 3 | ~~#438~~; ~~#99~~; ~~#102~~ | > | 440 | Records | Records→spacetime | *new (2026-11-19, the author's plan "Unified records, quantum measurement and emergent spacetime", workstream R, Task 5; obligation UO-25 (5); plan [`unified-records-spacetime-task-plans.md`](unified-records-spacetime-task-plans.md) §3)* **A record that controls a second apparatus.** **Build:** (B) the obstruction on the arena of UA-18: no apparatus there depends on a pointer's moving angle; (C) a pointer whose moving angle acts back (UO-24), which is what "a record controls an apparatus" means when a record is the ontic selection in the fibre: a clause for the new term, then a two-stage run on one arena; (A) the weaker carrier-mediated succession (a second pointer gated by a level that the first interaction changed), stated as succession and not as control. ▸ **FIRST BRICK DONE 2026-11-19 ((B)).** [`LF4/PentaPointerInert.lean`](../CsdLean4/LF4/PentaPointerInert.lean) (+11 pins): `shiftPenta`, ★★★ `fiveCoupledFlow_shiftPenta` (**the five-pointer flow commutes with turning the pointers**), ★★ `fiveCoupledFlow_shiftPenta_snd`, ★★ `fiveCoupledFlow_shiftPenta_ray`. Contract: UD-227. **Scope:** an obstruction for UA-18 (c)'s couplings; correlation through a carrier is not excluded. ▸ **(C) admitted 2026-11-19 → UA-20** ("Continue all of them"; to be struck if that was not meant): a step whose observable carries a smooth profile of an earlier pointer's moving angle. Nothing in Lean yet; the engine is `hamiltonianFlow_stepEnergy`. ▸ **SECOND BRICK DONE 2026-11-19 (the controlled step, under UA-20).** [`LF4/ControlledStep.lean`](../CsdLean4/LF4/ControlledStep.lean) (+11 pins): `controlObs`, `controlledEnergy`, ★★ `hamiltonianFlow_controlledEnergy`, ★★★ `hamiltonianFlow_controlledEnergy_of_lift_eq_zero`, ★★★ `controlled_pointer_fixed_of_window_eq_zero` (**the second pointer does not move while the window of the first pointer's angle is shut**), ★★ `map_hamiltonianFlow_controlledEnergy`. Contract: UD-233. **Scope:** on the fibres `g(ψ₂) = 0`; no record read. **Third brick, 2026-11-19:** [`RecordLayer/ControlledOrder.lean`](../CsdLean4/RecordLayer/ControlledOrder.lean) (+2 pins): ★★★ `controlled_display_after_first` (**a display of the second pointer is preceded by a display of the first, in one run**), ★★ `controlled_displayArc_after_first`. Contract: UD-241. **Remainder (this row stays open):** (C) that the second pointer does tick once the first has stayed displayed; a window supported in the displaying arc; a tick readout on the controlled arena and the two displays as two recorded facts; (A). | **L** | High — the plan's Task 5 | ~~#374~~; UO-24 | > | 441 | Records | Records→spacetime | *new (2026-11-19, the author's plan "Unified records, quantum measurement and emergent spacetime", workstream R, Task 6; obligation UO-25 (6); plan [`unified-records-spacetime-task-plans.md`](unified-records-spacetime-task-plans.md) §4)* **The Born bridge for tick records, or its obstruction.** **Build:** (a) the exact law of a tick count; (b) the displacement of a pointer coupled to a positive observable as the accumulated expectation along the orbit, monotone, and `2|z| w T` for a weight `w` constant over the window; (c) the bridge: the mean count is `2|z| w T − 1/6`, so the mean tick rate is the weight within an explicit error, with every assumption stated; (d) what fails: the count's distribution, a weight that varies over the window, a pointer that swings; (e) the relation to the click law (which saturates) and to the cell law of Posit 1 (which this does not discharge). ▸ **FIRST BRICK DONE 2026-11-19 ((a), for a pointer that only advances).** [`RecordLayer/TickCountLaw.lean`](../CsdLean4/RecordLayer/TickCountLaw.lean) (+16 pins): `periodicArc`, `tickIndices`, ★★★ `isTickCount_line` (the count is the number of integers `k` with `x₀ ≤ r₂ + k`, `p₁ + k < x₀ + Δ`), ★★★ `integral_card_tickIndices` (**its mean over a uniform starting angle is `Δ − (p₁ − r₂)`**), ★ `rep_add_coe_mem_Ioc_iff`, ★★★ `isTickCount_pointer`, ★★★ `integral_card_tickIndices_pointer` (**`D(T) − 1/6` for the arena's arcs: linear, a dead zone of one gap, no upper cut-off**). Contract: UD-229. **Scope:** the law of a count given a displacement; no weight appears. **Remainder (this row stays open):** (b) to (e). ▸ **Author, 2026-11-19:** the Born weight is already proven and is not to be re-derived (`csd_nstep_born`; the standing rule of 2026-11-03). (b) and (c) as written would rebuild a weight from a pointer's displacement; they are withdrawn until restated as a relation to the weights already proven. | **M–L** | Critical in the plan — Task 6 | ~~#436~~; ~~#355~~ | > | 442 | Records | Records→spacetime | *new (2026-11-19, the author's plan "Unified records, quantum measurement and emergent spacetime", workstream R, Task 4; plan [`unified-records-spacetime-task-plans.md`](unified-records-spacetime-task-plans.md) §2; the author's decision of 2026-11-19: events keep their real flow times, no map to periods)* **The causal bridge for the radar's records.** **Build:** the three notions kept apart: what the graph permits, what is demonstrated under a non-zero coupling, and the order of the flow parameter. **First brick, 2026-11-19:** [`RecordLayer/FlowEventCone.lean`](../CsdLean4/RecordLayer/FlowEventCone.lean) (+10 pins): `FlowEvent`, `HopSeparated`, ★★ `hopSeparated_comm`, `growIterG_add`, `FlowEvent.PermittedWithin`, ★★★ `FlowEvent.kick_response_le` (**a kick `d` hops away changes a later reading by at most a factorial tail in `d`**), ★★ `FlowEvent.kick_response_eq_of_time_eq`. Contract: UD-239. **Remainder (this row stays open):** (a) the radar's five readings as `FlowEvent`s with their gate regions and their permitted relations; (b) a demonstrated influence (*relay_count_depends_on_messenger*); (c) the reading windows ordered in flow time, which needs rate inequalities among the family's window times; (d) the five events as the coordinates of the macrostate of Task 3. | **M–L** | High — the plan's Task 4 | ~~#361~~; #297; #439 | > | ~~306~~ | QIT | QC chain (link 12) | *new (2026-11-05, residue of ~~#304~~)* **Faults inside the injection's `CNOT̄` and read-out.** ▸ **DONE 2026-11-05 (M).** In [`Empirical/QM/QEC/SteaneInjection.lean`](../CsdLean4/Empirical/QM/QEC/SteaneInjection.lean) (28 pins): `pairKrausMix` (**the correctable channels of the pair** — Kraus operators in the span of the family, trace preserving on the code; the two-block `cssMix`), ★★ `exists_isRecovery_pairKrausMix` (**the pair recovery undoes every one**: the Knill–Laflamme scalars sum to `1` by trace preservation), `pairMix_subset_pairKrausMix`; `sTransversal_pow_mul_gateOf`, ★ `xBar_logProj_mul_gateOf_mul_steaneEnc`, ★ `readOp_mul_blockKron_gateOf` (a product fault passes the feed-forward: by `S†^m` on the data exactly, by `X^m` on the ancilla on the code), ★★ `sum_conj_eq_of_mul_pairEnc` (a gadget whose Kraus operators are the exact ones followed by errors is the ideal gadget followed by them); ★★★ `injChannelCnot_eq` / ★★ `deviatesBy_injChannelCnot` (**a faulty `CNOT̄` location**, arbitrary unitaries on its two qubits after the gate — `pairFault`, the model of `steane_faultyCNOT_recovery` — is the ideal gadget followed by an even mixture of two weight-one pair errors); ★★★ `injChannelMeas_eq` / ★★ `deviatesBy_injChannelMeas` (**a faulty read-out of one qubit**, any trace-preserving family before it is read — `measChannel` at one faulty readout — is the ideal gadget followed by a correctable channel); ★★★ `steane_injection` strengthened (one recovery for the exact gadget and for one fault at any of the three location kinds). Scope: level one, one fault per gadget, ideal recovery; the read-out is the decoded projector `logProj`, not the full collapse of the seven qubits. ▸ *Original row:* ~~#304~~ discharges a fault on the prepared magic state. The gadget's other locations: a single-qubit fault on either qubit of one of the seven transversal `CNOT`s (a product pair fault, which `CNOT̄` itself does not spread across the pair beyond the one qubit per block — `SteaneTransversalCNOT.lean`'s `steane_faultyCNOT_recovery` is the rectangle precedent), and a fault in the read-out (a flipped measured bit, which the decoded `logProj` absorbs on the code, or a wrong outcome feeding a wrong `S̄^m` — a logical error unless the fault model bounds it). Deliverables: `DeviatesBy pairProj pairMix idealInj` for one fault at each location kind, with the read-out modelled as `SteaneLogicalIO.lean`'s `measChannel`. ⚠️ ~~**Stuck on:** nothing identified; the price is the read-out fault model (a classical bit flip before the decoder) and the `CNOT̄` location bookkeeping.~~ closed. | ~~**M**~~ done (M) | Medium — the injection's full one-fault bound | ~~#304~~; #305; `Empirical/QM/QEC/SteaneInjection.lean`; `Empirical/QM/QEC/SteaneLogicalIO.lean` | > | ~~290~~ | Records | Records→spacetime | *new (2026-11-04, residue of ~~#288~~)* **What fixes the first wing of an entangled selection?** ▸ **DONE 2026-11-04 — the author chose (a): "it all comes back to QM and the actual microstate being unknown until measured. But the detail sets the outcome." (S–M took S).** [`RecordLayer/SelectionOrder.lean`](../CsdLean4/RecordLayer/SelectionOrder.lean): `onticCell` (**the selection by ontic write order** — the individual outcome is a deterministic function of the ontic point and the ontic order; ties to A), ★★ `volume_onticCell` (the cell volumes, whatever the order and frame), ★ `lorentzBoost_time_sub`; ★★★ `timelike_order_boost_invariant`, ★★★ `timelike_onticCell_boost` (**timelike or null writes: the individual outcomes are the same in every boosted preparation**); ★★ `exists_boost_reverses_spacelike` (`u = (Δt + |Δx|)/(2Δx)`); ★★★ `spacelike_boost_changes_outcomes`, ★★★ `singlet_spacelike_boost_changes_outcomes` (**spacelike writes: some boost of the preparation changes the individual outcomes on a set of positive measure, and no outcome's probability**). So CSD carries a hidden preferred slicing in the Bohmian sense, confined to the individual outcomes of spacelike-separated entangled writes and hidden from every record statistic, as the microstate is. Scope: a boost of the preparation relative to the ontic frame; re-describing a fixed experiment changes nothing; I3 cells on the fibre with tables as data; the choice of (a) is the author's, not derived; the record layer stays kinematic. ~~#288~~ proved that the order in which Σ selects the two wings' outcomes fixes the individual outcomes (`selection_order_dependent`), and that order-independent outcomes with local first writes are impossible (`no_order_indep_local_first_selection`). The corpus fixes A first by convention. **Decision (the author's, foundations):** (a) the **ontic time's order** — the wing written first in ontic time is selected first, so CSD carries a hidden preferred slicing, invisible in every record statistic (~~#285~~, ~~#287~~), like Bohmian mechanics; (b) a **non-temporal rule** (e.g. a fixed wing, or a rule from the joint context) — then say what it is; (c) **symmetric, both wings reading both settings** — no order, outcome-level parameter dependence. **Build, after the decision:** for (a), the selection indexed by the ontic write order, with the theorem that a boost reversing the order of two spacelike writes changes the individual outcomes but no statistic (S–M, from `SelectionOrder.lean` + `SlicingIndependence.lean`); for (c), a symmetric selection on `T²` reproducing any table, with its parameter dependence as a theorem (M). ⚠️ **Stuck on:** *done 2026-11-04 — was:* the author's decision. | ~~**decision + S–M / M**~~ done ((a), S) | High — what CSD says about the individual outcomes of a Bell pair | ~~#288~~; ~~#285~~; ~~#287~~; #39; Q27 | > | ~~271~~ | Records | Records→spacetime | *new (2026-11-01, residue of ~~#262~~)* **The device's clock read as records, and #269's signal timed by it.** ▸ **DONE 2026-11-01 (M took M).** [`RecordLayer/DiracTicks.lean`](../CsdLean4/RecordLayer/DiracTicks.lean): `deviceProp` (`e^{−itH_dev}`), ★★ `deviceProp_devicePacket` (**the device's evolution carries the device**, mode by mode); `readWeight` (Born weight of the internal readout `(|0⟩ + u|1⟩)/√2`), ★ `readWeight_sub` (**its excess over `|−u⟩` is `2 Re(ū C)`**), `tickWeight` (that excess halved and rectified), `TickArena`/`tickFlow`/`tickEvolve`; ★★ `integral_max_cos` (`∫₀^{2πn/R} max(κ cos Rs, 0) = 2nκ/R`), ★ `integral_max_cos_lt`; ★★★ `tickDetector` (**tick `n` an actual occurrence written at exactly `2πn/R`**); ★★ `deviceCoherence_single` (**a one-mode device turns exactly**); ★★★ `dirac_ticks` (**ticks = occurrences of one evolution carrying the device; tick `n` at exactly `2πn/R`, one period apart; `R/(μ₁ − μ₀)` within `(μ₁ − μ₀)/E₀` of `√(1 − (v/c)²)` — a moving device's ticks are dilated**); `sideEvolve`, ★ `ActualOccurrence.liftFst`/`liftSnd` (an occurrence of either system is one of both side by side); ★★★ `dirac_timed_signal` (**departure, arrival and every tick of a device at rest (`T = 2π·10⁶`) are occurrences of one evolution from one preparation; departure between ticks `12` and `14`, arrival between `562` and `641`; arcs `450 cT` to `500 cT` apart; speed in ticks `[450/629, 125/137] ∋ 4/5`, below `c`**). **Findings:** the clock's reading is written as records, tick `n` at the `n`-th return to the mark; a speed is read in a device's ticks, a little wider than #269's chain-time bracket. Scope: one-mode device (exact, no position); detector calibrated (mark = starting reading, thresholds in units of `2g|C₀|/R`); device beside the chain, uncoupled, compared through the one evolution (a single local observer needs #270). Residue → #272 (localized device), ~~#273~~ (uncalibrated counter). ▸ *Original row:* ~~#262~~'s `dirac_device` gives the device's clock reading as a phase of the state; ~~#269~~'s speed is read off detector records in the chain's time. Join them: (a) **tick records** — a detector on the device's coherence (a readout of the internal levels in a rotating basis, #252's other half) written each time the reading passes a mark, as `ActualOccurrence`s of the whole chain with the internal levels (`deviceHam` lifted to `e^{−itH_dev}`), with the tick times within the reading's error; (b) **the signal timed by the clock** — the device carried with the start arc's detector, its tick count at the departure and arrival records giving `ℓ/2` over (ticks × tick period), the dilation `√(1 − (v/c)²)` entering for a moving device. **Acceptance:** tick records of the device as occurrences of one evolution, and #269's speed bracket restated in tick units. ⚠️ **Stuck on:** nothing — (a) is ~~#265~~'s whole-chain arena with the internal levels and a projector readout; (b) is arithmetic on (a) and ~~#269~~. | ~~**M**~~ done | High — makes the speed an observer's measurement, the input #253 and #256 need | ~~#262~~; ~~#269~~; #253; #256 | > | ~~272~~ | Records | Records→spacetime | *new (2026-11-01, residue of ~~#271~~)* **Ticks of a localized device.** ▸ **DONE 2026-11-01 (M took M).** [`RecordLayer/DiracDeviceTicks.lean`](../CsdLean4/RecordLayer/DiracDeviceTicks.lean) + `CV/DiracDevice.lean`: ★★ `abs_levelClockRate_sub_le` (**`|R(s) − R(s₀)| ≤ |s² − s₀²|(μ₁ − μ₀)/(2μ₀²)`**, algebraic), ★★★ `levelCoherence_rate_sharp`, and `dirac_device` **strengthened in place** with the sharp coherence conjunct `|t| c D (2c|p₀| + c D)(μ₁ − μ₀)/(4μ₀²)` (quadratic in `D` at rest, proportional to the splitting); ★ `abs_tickWeight_sub_le`, ★★ `abs_integral_tickWeight_sub_le` (drift `|s|ε` moves the register by `≤ ε t²/2`), ★ `le_integral_max_cos_near`; ★★★ `driftTickDetector` (**tick `n` written inside `[2πn/R − δ, 2πn/R + δ]` whenever `δ ≤ 1/R` and `ε(2πn/R + δ)² < |C₀|δ`**); `deviceDrift`, `deviceTick`; ★★★ `dirac_device_ticks` (**#262's device with its position bound; every tick an occurrence of one evolution carrying it; tick `n` inside the window under the margin**); ★★★ `dirac_located_signal` (**#269's departure and arrival records and ticks `0`–`641` of #262's device at rest — two modes, `D = 10⁻⁷`, ring `2π·10⁷`, a hundred times shorter than the chain's — in one evolution; tick `n` inside `[nT − 1, nT + 1]`; speed in ticks `[450/629, 125/137] ∋ 4/5`, below `c`**). **Findings:** the integrated drift grows as `t²`, so the tick horizon is finite; with the first bound `cD` the same ticks would need a device spread over `> 10¹⁴` (larger than the chain) — the sharp, stationary-at-rest drift is what makes the device small. No new residue: the device's position is a two-mode translation phase beside the chain, uncoupled (#270, #253); calibration stays ~~#273~~. ▸ *Original row:* ~~#271~~'s `tickDetector` needs a coherence that turns **exactly** (`deviceCoherence_single`: one mode, no position). ~~#262~~'s windowed device turns only within `|t| c D` (`dirac_device`). Prove the tick record `n` is still written, inside the write window `[2πn/R − δ_n, 2πn/R + δ_n]` with `δ_n` controlled by `|t| c D/|C₀|` and the threshold margin, for `n` up to the horizon where the error stays below the lobe margin; then restate `dirac_timed_signal` with a device that has a position. **Acceptance:** tick records of #262's device, with written error windows. ⚠️ **Stuck on:** nothing — a perturbation of `integral_max_cos` (the integrand moves by at most the coherence error, `|max a 0 − max b 0| ≤ |a − b|`) and the strict-growth lemma with a margin. | ~~**M**~~ done | Medium — the device of the timed signal gets a position, the input #253 needs | ~~#271~~; ~~#262~~; #253 | > | ~~273~~ | Records | Records→spacetime | *new (2026-11-01, residue of ~~#271~~)* **An uncalibrated tick counter.** ▸ **DONE 2026-11-01 (re-scoped at the user's call from the phase-locked bound, which locks only `n |R/R₀ − 1| < 1/2` ticks — up to tick 1 for #253's moving clock, 2 for #256's — to a counter needing no rate; S–M → M–L, took M).** [`RecordLayer/DiracRateTicks.lean`](../CsdLean4/RecordLayer/DiracRateTicks.lean): ★★ `calibrated_lock` / `calibrated_slip` (the row's acceptance: thresholds `2n|C₀|/R₀` crossed within `π/(2R)` of `2πn/R` while `n |R/R₀ − 1| < 1/2`, and lost beyond); ★★ `hasDerivAt_deviceCoherence` (**`dC/dt = −i S`**, `S` the splitting coherence, a quadratic observable built from `H_dev`), ★★★ `splitCoherence_shift_sharp` (`S` within `(1 + R|s|)` times the sharp drift of `R e^{−iRs} C(t₁)`, any start); `weightFlow` (registers of any weight; `tickFlow` an instance), `rateWeight`; ★★★ `rateTickDetector` (**tick `n` at exactly `2πn/R`, threshold `2gn|C₀|`, no `R`**), ★★★ `rateDriftDetector`, `singleRateTick`, `deviceRateTick`, ★★★ `dirac_rate_ticks`. **#253's and #256's clocks re-counted with it** (`twinRestTick`, `twinMovingTick`, `approachTick`; `approach_rate_drift`), tick windows and capstones unchanged. **Finding:** counting the reading's rate needs no rate; the dilation and Doppler factor read off the records are no longer put in by the counters' thresholds. Scope: the counter still takes its mark and `|C₀|` from the device's starting reading; stated detector law (#251, #266). No new rows. ▸ ~~#271~~'s thresholds are set in units of one turn's integral `2g|C₀|/R`, which needs the device's own `R`. Show the counter is **phase-locked** without it: with thresholds `2n g|C₀|/R₀` set from a reference rate `R₀`, tick `n` is still written inside the `n`-th lobe (reading within a quarter turn of the mark) while `n |R/R₀ − 1| < 1/2` — so a counter calibrated at rest and carried by a moving device counts the device's own turns for that many ticks. **Acceptance:** tick `n` written within `π/(2R)` of `2πn/R` for every `n` below the stated bound, `R ≠ R₀`. ⚠️ **Stuck on:** nothing — `integral_max_cos` gives the lobe areas exactly, the rest is the bracket `(2n−1)|C₀|/R < θ_n < (2n+1)|C₀|/R`. | ~~**S–M**~~ done (M) | Medium — removes the calibration caveat from the dilated ticks | ~~#271~~; #253 | > | ~~208~~ | Geometry | Reconstruction frontier | *new (2026-10-24, the QM-gap survey against HEAD `9ce090e8` — edges of ordinary QM with a Lean shape that no row carried)* **A5's approximate regime as a dynamical statement.** ▸ **DONE 2026-10-29 (M–L took M).** [`SigmaLayer/ActivatedRegime.lean`](../CsdLean4/SigmaLayer/ActivatedRegime.lean) (32 pins): `verticalDeriv`, `EpsProjectableDeriv` (**the derivative form `sup‖d(Hs)|_V‖ ≤ ε` on the arena**), ★★ `epsProjectable_of_deriv` (**a vertical derivative bound `ε` integrates to fibre oscillation `ε/2`**, mean value along a straight fibre path of sup-length `≤ 1/2`); for every coupled energy `−2⟨A⟩ − 2⟨B⟩ f(θ₂)`: `epsProjectable_coupledEnergy_iff` / ★ `epsProjectable_coupledEnergy` (`2‖B‖·osc f`), ★ `verticalDeriv_coupledEnergy` (**`d𝓗|_V = −2⟨B⟩ f'(θ₂) dθ₂`**), `epsProjectableDeriv_coupledEnergy`, ★★★ `coupledFlow_shadowing` (**`quantum_effective_shadowing` carried to the arena**: the coupled flow and the exactly projectable sector flow of `A + f(φ₀) B` keep their rays within `‖B‖·osc·T·‖v‖`) and ★★★ `coupledFlow_expectation_shadowing` (every expectation within `2‖O‖·‖B‖·osc·T`, on rays); for the activated write: ★★ `epsProjectable_activatedEnergy_iff` (**the sharp `ε = (2|g|/|ħ|)·osc f`**, attained where the receiver is occupied), ★★ `activatedEnergy_not_projectable` (`g ≠ 0`, `f` nonconstant ⇒ not `EpsProjectable` at `0`), `epsProjectableDeriv_activatedEnergy` (`(2|g|/|ħ|)·sup|f'|`), `norm_hamB`, ★★★ `activated_expectation_shadowing` (**sector predictions shadowed at `‖O‖·ε·T`**, the matrix constant being exactly half the ontic `ε`), its flow `activeFlow` (`hamiltonianFlow_activeFlow`); ★★★ `activated_regime` (6 conjuncts). Scope: sup norm on `V` (agrees with the flat one here, `d𝓗|_V` being pure `dθ₂`); the fibre is not shadowed and moves by design (the record write). No residue row. ▸ *Original row:* `EpsProjectable _ ε` is populated at `ε > 0` (`SigmaLayer/MovingFibreWitness.lean`) but no flow is constructed whose generator it is. ~~#194~~/~~#195~~ supply the missing object: a smooth record-creating Hamiltonian flow on the arena whose energy `−(2/ħ)(J⟨X⟩ + g f(θ₂) n_B)` does not factor through `π`. Deliverables: the activated energy is `EpsProjectable` with an explicit `ε` in `g`, its flow is the one constructed, and the `(ε, T)` shadowing bound for this witness, bridging `quantum_effective_shadowing` from matrices to the arena. ⚠️ **Stuck on:** the derivative form `sup‖d(δH)|_V‖ ≤ ε` on the arena — the §2a smooth structure now exists (R-016′), so this is work, not a wall. | ~~**M–L**~~ done (M) | Medium — the last A5 residue named in `reconstruction-status.md` §7 | `SigmaLayer/MovingFibreWitness.lean`; `RecordLayer/ActivatedHamiltonian.lean`; `reconstruction-status.md` §7 | > | ~~209~~ | Docs | Reviewer-proofing | *new (2026-10-24, the QM-gap survey against HEAD `9ce090e8` — edges of ordinary QM with a Lean shape that no row carried)* **R-016 re-scoped after ~~#195~~.** ▸ **DONE 2026-10-27 (S as priced).** `specs/residues.tsv` R-016: the statement now leads with the propagators it was opened for (shear, joint lift) and says it is **not** open for record-creating propagators in general — #195's activated write and #202's echo are arena-level Hamiltonian flows — with the price of that route stated (exact only off a positive-measure transition region) and why it does not transfer (the strokes are symplectic, not Hamiltonian, on the torus); `last_review` 2026-10-27, carriers unchanged (the guard's tag set). `specs/reconstruction-status.md` A2: a dated update naming the two theorems. Every other R-016 mention in `POSITS.md`, `CSD-CHARTER.md`, `docs/TOUR.md` and `docs/FROM-POSTULATES-TO-QUANTUM-COMPUTERS.md` concerns the shear or swap propagator and stays true; none edited. ▸ *Original row:* The residue says arena-level Hamiltonian generation of a record-creating propagator is stated, not formalised. ~~#195~~ formalised it for the activated write; the shear and joint-lift propagators remain as the residue describes. Rewrite `residues.tsv` R-016 and `reconstruction-status.md` A2 so the first arena-level case is named and the remaining object (the piecewise and corridor propagators) is the residue. ⚠️ **Stuck on:** nothing. | ~~**S**~~ done (S) | Medium — a stale residue is a false frontier | `residues.tsv` R-016; `reconstruction-status.md` §2 A2; ~~#195~~ | > | ~~210~~ | Docs | Reviewer-proofing | *new (2026-10-24, the QM-gap survey against HEAD `9ce090e8` — edges of ordinary QM with a Lean shape that no row carried)* **The volume identity reconciled.** ▸ **DONE 2026-10-27 (S as priced).** `specs/reconstruction-status.md` §7's A1/KG-1 bullet now states both halves proved with the constants visible: `ω_FS^{∧n} = (4π)ⁿ μ_FS` (`fsVolume_eq_smul_fsMeasure`), `vol_g = ω^{∧n}/n!` (`riemannianVolume_fsMetric`), hence the textbook form as `ω^{∧n}/n! = ((4π)ⁿ/n!) · μ_FS` (`riemannianVolume_fsMetric_eq_smul_fsMeasure`), and on the arena `(N + 1)(4π)ᴺ · (μ_FS ⊗ vol_{T²})` (`arenaVolume_eq_ofReal_smul_kMuL`). The bullet had contradicted the same file's §2a, which recorded the discharge since 2026-09-11. ▸ *Original row:* `reconstruction-status.md` §7 still lists `ωⁿ/n! = μ_FS` as the remaining half of A1/KG-1; `LF4/ArenaVolume.lean` has `arenaVolume_eq_ofReal_smul_kMuL` with the constant `(N+1)(4π)ᴺ` visible. Reconcile the status file with the theorem and name the constant. ⚠️ **Stuck on:** nothing. | ~~**S**~~ done (S) | Low–medium — status honesty | `reconstruction-status.md` §7; `LF4/ArenaVolume.lean` | > | 211 | Geometry | Library owed / upstream | *new (2026-10-24, the QM-gap survey against HEAD `9ce090e8` — edges of ordinary QM with a Lean shape that no row carried)* **The pointwise Birkhoff ergodic theorem.** Mathlib-absent (`MATHLIB-ABSENT`), named as the residual gap of T3 in `reconstruction-status.md` §2a and of the retired Born-from-flow route; library-grade, stated for a measure-preserving map on a probability space with the time average converging a.e. to the conditional expectation on the invariant σ-algebra. ⚠️ **Stuck on:** nothing but size; and it advances no CSD claim (T3's route is a settled non-goal), so its value is to the library. | **L** | Low — library owed, no consumer in the programme | `reconstruction-status.md` §2a T3; `MATHLIB-GAPS` | > > > ⚠️ **QIT chain SCOPED 2026-09-11** — [`qit-chain-scoping.md`](qit-chain-scoping.md): can a reader go from CSD's > posits to the QIT layer by theorems alone? No: the bridge `preparation → density operator` exists as TWO > proved theorems (`fromPreparation`, `effect_gleason_representation`) never composed and never consumed by > QIT; every K1/K3 theorem is on bare matrices; the QEC twins bind their CSD bundle as an unused `_b`. > Priced rows **W2–W12** there (critical path W2 → W3 → W6 → W7). **W2 and W12 DONE 2026-09-11**: > `LF2/PreparationQdensity.lean` — `preparationDensity` (the density operator of a preparation on `Σ`, > `effect_gleason_representation` on `fromPreparation`), the instantiation lemmas, and the first consumers > `vonNeumannEntropy_preparation_nonneg` / `_le_log` and `channel_traceDist_preparation_le`. **W3 DONE > 2026-09-11**: `LF2/PreparationBarycenter.lean` — that density operator IS the barycentre > `∫ |ψ⟩⟨ψ| d(π_* μprep)` (`preparationDensity_eq_barycenter`, entrywise `preparationDensity_apply`, the > `ρ_ep dμ_FS` form under absolute continuity). **W6 DONE 2026-09-11**: `LF2/FlowChannel.lean` — if the ontic > flow lifts a unitary (`IsUnitaryLift`, a theorem for the projective unitary actions), the density operator of > the flowed preparation is `U ρ Uᴴ` (`barycenter_flow`) and, on a joint sector, the reduced flowed state is the > Stinespring channel of `U` applied to the system's density operator (`traceRight_barycenter_flow`; mixed > environment = barycentre of the environment preparation, `traceRight_barycenter_flow_prod`). W5's witness > half landed with it (`LF6/DecoherenceChannel.lean`: `decohereReduced` is a channel output). ⚠️ The W6 row's > gloss of `IsJointLift` was wrong (it is the pointer-arena stroke predicate); corrected in the note. **W7 > DONE 2026-09-11**: `Thermo/SigmaSecondLaw.lean` — the second law, data processing and Landauer on `Σ` > (`vonNeumannEntropy_le_pinching_flow`, `vonNeumannEntropy_le_deisolation`, `traceDist_traceRight_flow_le`, > `landauer_flow`), with TH2's coarse-graining identified as the de-isolation channel > (`deisolationChannel_apply_eq_pinch`). **The critical path W2 → W3 → W6 → W7 is complete.** **W6′ DONE 2026-09-11** (`LF6/MeasurementFlowChannel.lean`: the de-isolation channel is > the environment marginal of LF5's `measurementFlow`). **W6″ DONE 2026-09-11** (`Projectivization.unitSection`, measurable, so the lift is unconditional for > projective actions and LF5's flow produces the de-isolation channel with no section hypothesis). **W5 DONE 2026-09-11** (`LF2/ChannelBridge.lean`: `QuantumChannel ≃ QuantumInfo.Channel`, > the parallel type marked an interface, DPI for LF2 channels). **W4's `= 0 ↔ pure` half DONE 2026-09-11** (`LF2/PreparationPurity.lean`: zero entropy iff the projective law is a > Dirac mass at one ray; Cat-1 `vonNeumannEntropy_eq_zero_iff`). **W8 + W4's coarse-graining half DONE 2026-09-11** (Cat-1 concavity `vonNeumannEntropy_mixture_ge`, > `holevoChi_nonneg`; `preparationEntropy_mixture_ge`: mixing preparations never lowers entropy on average). **The two preparation interfaces identified 2026-09-11** (`SigmaLayer/PreparationDensityBridge.lean`: > W3's `ρ_ep` form with Q28's density, hypothesis-free on the Kähler arena). **W10 DONE 2026-09-11** (`LF6/DeisolationCapacity.lean`: the single-letter Holevo capacity of the > de-isolation channel is one classical bit, `deisolationChannel_holevoCapacity`; Cat-1 Holevo bound). **W11 DONE 2026-09-11** (`Empirical/CSD/QEC/ThreeQubit.lean`: the bit-flip channel is the environment marginal > of a concrete `Σ`-flow on `ℂℙ³`; the four error regions of `Σ` are pairwise disjoint; both unused bundle binders > gone, the bundle deleted). Open on the QIT chain: W9 SSA fork only. > The author decides. > > *Closed by the 2026-09-11 scan (were "not scheduled"):* the C2 Mathlib-gap pair — the metric on `ℙ` > **exists** (`Mathlib/LinearAlgebra/Projectivization/Metric.lean`, the gap metric, a Mathlib-upstream > candidate) and `segre_range_null` was **dissolved** (MG-2, 2026-08-22: the consumer never needed it). > Two queues have closed since the 2026-09-02 block below was written. Every numbered item > of that block landed, and the external-review **CR queue** > ([`cr-queue.md`](cr-queue.md), CR-1…CR-16) is fully resolved: 13 done, CR-9 deferred by > author decision, CR-10 declined with reasons, CR-14 fallback-only (`R-019`). The heavy one > was **CR-4** (BELL-MIGRATE, 28 files, closed 2026-09-06 over fifteen commits): nothing in > `Empirical` or `LF6` states a Born-region result on the base arena any more. ⚠️ Its "no new > theorem" premise was false — the migrated files prove *frequency* statements and the fibred > side had no frequency theorem, so seven engine pieces, the weight bridge > `globalBasin_toReal_eq_bornRegion_toReal` and a generic i.i.d. trial process > (`Mathlib/MeasureTheory/IidTrials.lean`) were built for it. Since CR-4 the Born headlines no > longer route through `fsMeasure` at all (`TERMS.md`, 2026-09-07, which also > registered `momentMap` — the corpus's most load-bearing named object — with a guard). > > ⚠️ **There is no queued Lean brick left** — no item that is scoped, sized and waiting to be > built the way the CR queue's were. That is the state of the programme, not an oversight in > this list. An assistant asked "what is next" should answer from the four lists below **and > say which list the item is on**: list 1 is research, and offering an item from it as though > it were a tranche is the specific error this block exists to prevent (exactly one row of it > is buildable); list 2 is real but is entirely prose-and-guards; list 3 is Mathlib depth — > two rows are buildable today and neither of them advances the physics; list 4 is closed to > assistants. **Every row in lists 1–3 carries `Cx / P(success) / Value`**, on the 2026-08-20 > triage's columns; a row offered without them is a row that has not been thought about. > > **1. Forward — research. ⚠️ Row E landed 2026-09-07, and with it the last buildable item on > this list.** What remains is a frontier, a Mathlib-gated wall, a blocked asymptotic and a > parked problem: nothing here is a tranche, and offering one as though it were is the error > this block exists to prevent. > > | Item | Cx | P(success) | Value | Recorded in | Status, honestly | > |---|---|---|---|---|---| > | ~~**★ Expert row E — quantum Darwinism / record redundancy**~~ **LANDED 2026-09-07, both sides** | **M** (QM side **S–M**; CSD side **M**, or **L** if the arena needs a new species) | Medium–high | Medium–high | [`quantum-darwinism-scoping.md`](quantum-darwinism-scoping.md); the expert-review table below (2026-09-02) | **SCOPED 2026-09-07; QM SIDE BUILT the same day** (`Empirical/QM/Darwinism.lean`, 5 pins: `SpectrumBroadcast`, ★ `not_of_constant_fragment`, ★ `successProb_eq_one_of_discriminating`, ★★ `copyBroadcast_perfect` — every fragment on its own identifies the outcome with probability one, at arbitrary `k`). **CSD side BUILT the same day** (`Empirical/CSD/Darwinism.lean`, 6 pins): ★★ `register_eq_of_same_point` — two registers agree across DIFFERENT strokes with DIFFERENT fragment counts, so the agreement is produced by one ontic selection upstream, not by the copying — plus ★ `copyStroke_zero_eq_id`, the same point from the other end (zero fragments, stroke is the identity, record just as definite). ⚠️ §3's stop condition was never reached because the arena needed no measure: **the twin is pointwise and kinematic**, frozen-base, with no invariance claim and no `a.e.` statement. The measure-theoretic redundancy §3 costed is NOT built, and that residue is recorded in the module. **Row E is closed at the kinematic tier.** ⚠️ **NOT `csd-foundations`-checked** — that agent was unavailable; run the check before the first line of Lean. Three findings the note records: the **information-theoretic** form (mutual-information plateau) is ⛔ **gated** — it routes through SSA, which this corpus carries only with `hDPI`, behind the Lieb summit behind Gate 1 — so the target is **spectrum broadcast structure** (entropy-free, strictly stronger) plus a record-level twin; the pointer bank's slots are indexed by **outcomes, not fragments**, so borrowing them as environment copies is a category error and the fragments need their own index; and the statement must be checked against a **one-fragment stroke** or it is a tautology. **The one buildable forward item.** M, P(success) medium–high, value medium–high. Its own note says "scope only after A/C" — rows A (WAY) and C (Local Friendliness) both landed 2026-09-04, so it is **unblocked as of that date**. Natural home: the P2 composite arenas + `RecordLayer/PointerLudersMarginal.lean`, as a `k`-fragment redundancy statement for the pointer marginal. ⚠️ In CSD a record's objectivity is its being an *ontic selection in `Σ`*, **not** its redundancy — Darwinism's criterion is the operational twin of that fact, not its ground, and the statement has to be chosen carefully to avoid landing a placeholder | > | **Posit 1's discharge — derive `IsTorusGenerated` from the de-isolation dynamics** | **XL** | Low — no Lean shape today | Highest in the programme | [`POSITS.md`](POSITS.md) §"three different things" (3) | ***The* reconstruction frontier**, and deliberately **not on any queue**: it is open *foundations*, with a known discharge condition but no Lean-shaped task. Posit 5's discharge is the same `H_int` family, which is why the definite article holds | > | **`R-016` — arena-level Hamiltonian generation** (`ι_X ω = dH` on the arena) | **L–XL** | Low near-term | High | [`residues.tsv`](residues.tsv), §A rows A2/A4 below | **Open mathematics, paper-side.** Chart-level generation is proved (`SigmaLayer/ChartIntegralCurve.lean`, `RecordLayer/HamiltonianShift.lean`); exactly one arrow is missing. ⚠️ **Re-priced 2026-09-08** ([`generator-layer-scoping.md`](generator-layer-scoping.md) §3, `R-016′`): the Mathlib vocabulary wall is **gone** — `mextDeriv`, `IsSymplectic` and `topFormMeasure` exist since steps (2b)–(3). What remains is (1) the arena as a manifold (`AddCircle` has no `ChartedSpace` at the pin, only `Circle`; transport, never retype `KTorus`), (2) the arena 2-form (pullback of forms along the projections, unbuilt), and (3) the statement, which the flux correction makes *locally* Hamiltonian only. The honest brick `R-016′` = `IsLocallyHamiltonian` on the arena: **L**, P medium, value medium — it formalises the residue's *statement* and its obstruction, not its discharge. Row stays open, wording stays. ⚠️ **`R-016′` LANDED 2026-09-14** (`LF4/ArenaSymplectic.lean`): the arena is a symplectic manifold, Liouville holds on it, the isolated Schrödinger dynamics is proved to be its Hamiltonian flow of `−2⟨H⟩ ∘ π₁`, and the stroke `(0, a)` is proved locally Hamiltonian; the obstruction half (no global `H`) landed the same day (`LF4/ArenaStrokeFlux.lean`, #28), and the arena volume is the corpus's `μ_FS ⊗ vol` up to its mass (`LF4/ArenaVolume.lean`, #29; the mass itself is #32). **`R-016` proper stays open**: nothing yet identifies `jointLift` with a time-`1` map on the arena. | > | **`R-019` — the relaxation H-theorem** (Track B / Valentini) | **XL** | Low | Highest *outside* the programme | [`cr-queue.md`](cr-queue.md) CR-14, [`residues.tsv`](residues.tsv) | The only route to *new predictions* past the empirically-identical ceiling — and **not scheduled**: blocked on first-passage asymptotics. The fallback `relaxation_requires_hyperbolic_fibre` landed 2026-09-06 and **was** the target of the item; the H-theorem was not. *Re-listed with a price in ▶ OUTSTANDING (2026-09-11): XL, the author decides.* | > | ~~**Expert row G — Berry phase as fibre holonomy + Aharonov–Bohm twin**~~ **SCOPED + BP-1/BP-2 BUILT 2026-09-21** ([`berry-phase-scoping.md`](berry-phase-scoping.md); queue row 10; residues #54–#56) | ~~**L**~~ done (M) | Low near-term | Medium | the expert-review table below | ⚠️ **Wall re-probed 2026-09-07, and the row's phrasing was wrong in one half.** "No connection API" is **false**: Mathlib has covariant derivatives on vector bundles (`Geometry/Manifold/VectorBundle/CovariantDerivative/{Basic,Metric,Torsion}.lean`). What is genuinely absent is **parallel transport and holonomy (0 declarations, anywhere)** and **principal bundles (0)** — and a Berry phase *is* a holonomy, so the missing half is the half that matters. ⚠️ The second obstruction is not Mathlib's at all and does not move: the corpus's fibre is a **torus**, not a `U(1)`-bundle over configuration space, so even a complete holonomy API would not make CSD's fibre the object a Berry phase lives on. Scope doc before touching; do not start from Lean | > | **General-`N` base-only A7** | **XL** | — (parked) | Low — gates nothing | the Research table below | ⏸ **PARKED, not refuted** — a well-characterised open problem with the decision-relevant output already in hand. **Do not resume without a new idea**; it gates nothing (the fibred reading is canonical, A7 is discharged at every `N`) | > > ⚠️ **How row E's rating actually came out, since the warning below was written before it was > built.** The **M** held, and the precedent held with it: the QM side came in at **S–M** as > re-rated, and the CSD side came in **below M** — but not because the work was easy. §3's > arena question *dissolved*: with no measure-theoretic claim the fragment register is a plain > product coordinate, so the `Measure.pi` arena was never needed and the stop condition was > never reached. ⚠️ **That is the limitation, not a saving**: what landed is pointwise and > kinematic, and the measure-theoretic redundancy §3 costed is the residue. The general lesson > stands and is worth keeping: *"rating a brick before the vocabulary it depends on exists is > rating the wrong problem"* (`cr-queue.md`) — here the vocabulary decision (which of three > theorems, and which arena) moved the cost more than the mathematics did. > > **2. Queueable now — hardening residue.** All of it is prose and guards; none of it is new > mathematics. In rough order of how much a referee would notice: > > * **Prose audit passes 1–2 over the remaining four-fifths of the surface** > ([`prose-audit.md`](prose-audit.md)). Passes 3–5 are done. ⚠️ The 2026-09-07 pass (below) > is evidence this is not busywork. > * **`REFERENCES.json` coverage** — 37 entries against ~110 distinct cited sources, and the > line-precise `[Key, file:Lstart-Lend]` citations CONVENTIONS §8.2 asks for are untouched. > Deliberately not gated; entries must be transcribed from the citing modules' own > `## Source` blocks, never recalled. > * **The untagged "Mathlib has no …" claims** — 45 at the last guard run (the figure moves; > `check-mathlib-absence.sh` reports the live count, and the "44" recorded on 2026-09-04 was > already stale). Author call, not mechanical — see the open decisions below. > * **The `_of_` / `_statement` migration** (CONVENTIONS §8.3), opportunistic by policy. > ⚠️ **Executed as an audit 2026-09-07, and the finding was in the ledgers, not the code:** > the nine gate realisability Props were DISCHARGED 2026-07-19 on `cpSectorData` > (`Gates/*Discharge.lean`) and the `PLACEHOLDERS.md` §1 *table* was updated the same day — > but all nine defining docstrings still opened with `**PLACEHOLDER (Prop definition, not > proved).**` and still carried `TODO(LF4 §13.2): construct a witness bundle`; > `Gates/BellPrep.lean` contradicted itself inside one docstring; §1's own header paragraph > still said "the remaining five … are the mechanical continuation" directly above the table > saying otherwise; and `BRIDGE-OBLIGATIONS.md` §2.6 still said "pre-LF4, no concrete `D` > exists for which any of them is shown to hold". **A ledger that understates what is > proved is still a false ledger**, and these two files exist to be believed. Fixed at every > site and now guarded by `scripts/check-placeholder-status.sh` (both directions: a > discharged Prop may not keep its banner, an undischarged one may not lose it, and a > discharge witness renamed away fails instead of being checked vacuously). `§2.1` was stale > the same way — `ofKählerPreparation` proves both LF3 singlet fields. > * ~~**Ecdsafail doc de-application** (~70 stale ECDLP refs in Cat-1 `Reversible/` headers)~~ > — **STRUCK 2026-09-07: already done, and the row's count was never right.** The corpus > carries 42 ECDSA/ECDLP mentions in total, 11 of them in `Reversible/`. Every one is either > a provenance pointer to the separate `Ecdsafail` repository ("the application that > motivated this substrate lives in …"), an explicit **"no ECDSA score change is claimed"** > disclaimer, or a section label in `Tests/AxiomAudit/MathlibStaging.lean`. Nothing applies > the corpus to ECDLP. Do not re-raise it. > > **3. Mathlib gaps — and yes, two of them are buildable.** [`MATHLIB-GAPS.md`](../MATHLIB-GAPS.md) > carries 14 rows; **9 are struck** (closed in-corpus, dissolved when the consumer turned out not > to need them, or simply wrong — the toral-automorphism row was factually false and the > Wigner-normal-form row described a lemma that is false as stated). Five stay live, and only the > first has a queueable piece: > > | Gap | Cx | P(success) | Value | Verdict | > |---|---|---|---|---| > | ~~**Manifold exterior calculus — step (0): `ℂℙ^{N−1}` as a smooth manifold**~~ | **M** | — | — | **LANDED 2026-09-07** (`Mathlib/Geometry/Manifold/Instances/ProjectiveSpace.lean`, 6 pins; corpus-side `CSD.LF4.cpn_isManifold`): the standard affine atlas, `instChartedSpace`, and `instIsManifold` at **`ω`** — analytic, not merely `C^∞`, because the transition maps are ratios of coordinates of `Fin.insertNth i 1 w`. ⚠️ The **M** rating held. The one non-mechanical step was chart continuity: the chart is defined through `Projectivization.rep`, a **choice function with no continuity of its own**, so continuity comes from `mk'` restricted to the open chart domain still being a quotient map — the same manoeuvre `continuous_bargmann` needed (CR-10 successor) | > | ~~**Manifold exterior calculus — step (1): the continuous wedge**~~ | **S** | — | — | **LANDED 2026-09-07** (`Mathlib/Analysis/Normed/Module/Alternating/Wedge.lean`, 6 pins): `ContinuousAlternatingMap.wedge` + the missing `domDomCongr`; witness `Kahler.extDeriv_fundamentalFormSq` (`d(ω ∧ ω) = 0`). ⚠️ The **S** rating held, but the item's shape did not: the tensor-valued form Mathlib's name suggests **cannot be stated** (no topology on `F ⊗ G` at the pin), so the codomains are paired through a continuous bilinear map instead. ⚠️ Binary product only — no associativity, no graded commutativity, no iterated power, no norm bound | > | ~~**Manifold exterior calculus — step (2a): differential forms on a manifold as OBJECTS**~~ **DONE 2026-09-07** | **L** (was M–L) | Medium | Medium | ⚠️⚠️ **Re-rated TWICE on 2026-09-07, and the second re-rating is the one to trust — it was done by trying to build it.** The morning probe found the bundle (`Topology/VectorBundle/ContinuousAlternatingMap.lean`) and smooth sections (`Geometry/Manifold/VectorBundle/ContMDiffSection.lean`) and concluded **M–L**. Attempting it found the crux: `Geometry/Manifold/` contains **no `AlternatingMap` anything**, so the smooth-bundle instance has to be built, mirroring `VectorBundle/Hom.lean` (512 lines) — and *its* crux, smoothness of the coordinate change, needs a **`ContDiff` lemma for the alternating pullback that Mathlib does not have**. Upstream has the pullback's *first* derivative (`Analysis/Calculus/FDeriv/ContinuousAlternatingMap.lean`) and the analyticity of the **multilinear** analogue (`Analysis/Analytic/CPolynomial.lean`, `cpolynomialAt_uncurry_compContinuousLinearMap`), but the alternating version of the latter is absent, and the existing continuity proof reflects along an inducing embedding — an argument that does **not** carry to `ContDiff`. **Four layers, the first a genuine analysis lemma — and ⚠️ THE FIRST ONE IS NOW LANDED (2026-09-07, `Mathlib/Analysis/Normed/Module/Alternating/Pullback.lean`, 6 pins):** ★★ `analyticAt_uncurry_compContinuousLinearMap` plus the `ContDiff` corollaries, unlocked by making **alternatization a continuous linear retraction** (upstream has only the `AddMonoidHom`) — it multiplies an already-alternating map by `(card ι)!`, so in characteristic zero it inverts the inclusion and analyticity reflects from the multilinear case. **Three layers remain** (`ContMDiffOn` coordinate change, bundle instance, forms as sections), so the row stays **L** and open. ⚠️⚠️ **LAYERS THREE AND FOUR THEN LANDED TOO (2026-09-07, `Mathlib/Geometry/Manifold/VectorBundle/AlternatingMap.lean`, 2 pins):** ★ `contMDiffOn_continuousAlternatingMapCoordChange` and ★★ **`ContMDiffVectorBundle.continuousAlternatingMap`** — the alternating-map bundle of two `C^n` vector bundles is a `C^n` vector bundle, which is what step (2a) was for. ⚠️ **And the blocker recorded an hour earlier was diagnosed WRONG**: not an instance-path mismatch — the two topologies on a CLM space are the *same instance* (`rfl`, checked) — but **elaboration order**, a type ascription re-synthesising instances down the normed path while the term carries the topological one. Stating the `ContDiff` fact through the term fixes it. ⚠️⚠️ **AND THEN (2a) CLOSED** (`Mathlib/Geometry/Manifold/DifferentialForm.lean`, 1 pin): `DifferentialForm` plus ★ `projectiveDifferentialForm_nonempty` — `ℂℙⁿ` carries **analytic** forms of every degree (the witness theorem was removed 2026-09-16 under Brief B: its witness was the zero form and it forced a cross-slice import; `fsForm_ne_zero` is the certificate). The "last mile" was a missing `[IsManifold IM 1 M]` in the probe's context, not a synthesis failure. ⚠️ **Witness is the ZERO form** — the FS form as a section is NOT built (chart-overlap agreement). ⚠️ **Still no `d`**: (2b) is upstream's TODO, so the top-power identity is sayable and no nearer proved. **Steps (0), (1) and (2a) are now all done; the row's remaining content is (2b)–(4).** **Superseded note follows:** the alternating bundle's coordinate change decomposes as `compContinuousAlternatingMapL ∘L compContinuousLinearMapCLM` — verified **`rfl`** — and with the operator-valued smoothness and the postcomposition CLM (both landed 2026-09-07) the `ContMDiffOn` proof is the same three lines as `Hom.lean`'s. What blocks it is an **instance-path mismatch**: the bundled maps elaborate on the topological-module instances (`ContinuousLinearMap.topologicalSpace`) where `ContinuousLinearMap.contDiff` wants the normed path. Same topology, different instance term. That is plumbing, and it was deliberately left rather than papered over — **whoever takes the next layer should start there, not at the mathematics.** ⚠️ Note the split this exposed: a form as a *family* `Π x, TₓM [⋀^k]→L ℝ` needs no instance and is definable today; what the instance buys is the word **smooth**. Half an object is not worth landing ⚠️⚠️ **WITNESS UPGRADED 2026-09-07 (evening): the Fubini–Study form IS now a section.** `Instances/ProjectiveSpaceFubiniStudy.lean` (chart invariance — ★★ `fsChartForm_transE`, `dd^c` naturality plus the pluriharmonic `log ‖zⱼ‖` correction, on top of the new `KahlerPluriharmonic.lean`) and `Instances/ProjectiveSpaceFubiniStudyForm.lean` (★ `localRep_fsSection`, ★★ `fsForm : DifferentialForm 𝓘(ℝ, Fin n → ℂ) (ℙ ℂ (Ambient n)) ∞ (Fin 2) ℝ`, ★★ `fsForm_ne_zero`), 24 pins. `C^∞`, not `ω` (the potential is only known `C^∞`); still no `d`. **Only (2b)–(4) remain in this row.** ⚠️⚠️ **(2b) LANDED TOO, SAME EVENING (2026-09-07):** `Mathlib/Geometry/Manifold/ExteriorDerivative.lean` — `mextDeriv`, ★★ `DifferentialForm.localRep_mextDerivFamily` (chart-independence as the local-representative identity), ★★ `contMDiff_mextDerivFamily`, ★★ `mextDerivFamily_mextDerivFamily` (`d ∘ d = 0`), 15 pins; scope `𝓘(ℝ, E)`, `∞`, `Fin k`. **Payoff ★★★ `fsForm_mextDeriv`: `d ω_FS = 0` on `ℂℙⁿ`** (3 pins). The XL was §3.1's plumbing, and the FS section had just paid for it. **Remaining: (3) top forms → measures (the top-power identity), (4) the symplectic / moment-map layer; and for `mextDeriv` itself the Palais formula, naturality, Leibniz.** | > | ~~**Manifold exterior calculus — step (2b): the exterior derivative on manifolds**~~ **DONE 2026-09-07** | **XL** (actual cost: one evening, once the FS section had paid for §3.1) | Medium *for route A specifically*; low otherwise | High | **SCOPED 2026-09-07** ([`exterior-derivative-scoping.md`](exterior-derivative-scoping.md)), not built. The real wall, and upstream's own stated TODO. ⚠️ Three routes, and the probe settles it: the **chart-local route's crux lemma already exists upstream** (`extDeriv_pullback` — the flat `d` commutes with pullback, so chart-independence is an application rather than a theorem); the invariant Palais formula is the wrong *definition* (its tensoriality is the classical hard step and is absent upstream) though it is the right later theorem; and the covariant-derivative route is ⛔ **blocked** — Mathlib has connections and torsion but **no existence theorem, no Levi-Civita, no torsion-free existence**. ⚠️ §4 of the note records four design traps, of which the smoothness arithmetic (`d` of a `C^n` form is `C^(n-1)`) is the one most likely to double the work. ⚠️⚠️ **ROUTE A ATTEMPTED (2026-09-07); nothing landed, and the attempt's first two write-ups were RETRACTED** (§3a). The local representative typechecks. The claimed blockers — "no computation lemmas for the alternating trivialisation", then "the next brick is that API, M–L" — are **both false**: `FiberBundle.trivializationAt_continuousAlternatingMap_apply` (`rfl`), `inCoordinates_eq`, `VectorBundleCore.trivializationAt_symmL` (`@[simp]`) and `coordChange_self` all exist, and the base-point identity reduces to one goal closed by the last two, blocked only by `TangentSpace` vs `(tangentBundleCore).Fiber` being defeq-but-not-syntactic — a `show`, not a theorem. **There is no level below §3.** The XL is made of constant chart-plumbing friction, not of a missing theorem. ⚠️ §3a also records the procedural lesson: four walls were called on 2026-09-07 and all four were wrong ⚠️⚠️ **LANDED 2026-09-07 (evening)** — `Mathlib/Geometry/Manifold/ExteriorDerivative.lean`, route A exactly as scoped: `mextDeriv` = flat `extDeriv` of the local representative in the chart at the point; ★★ `DifferentialForm.localRep_mextDerivFamily` (the local representative of `dω` in every chart is the flat `d` of the local representative of `ω` — `extDeriv_pullback` on the transition + the tangent cocycle), ★★ `contMDiff_mextDerivFamily`, ★★ `mextDerivFamily_mextDerivFamily` (`d ∘ d = 0`), bundled `DifferentialForm.mextDeriv`; 15 pins. Scope per §4 taken the cheap way: `𝓘(ℝ, E)` boundaryless, `∞`, `Fin k`, `toFlat` cast. **Payoff ★★★ `Projectivization.fsForm_mextDeriv`: `d ω_FS = 0` on `ℂℙⁿ`** (3 pins) — the first manifold-level Kähler statement in the corpus. The "wall" of §3a was the plumbing of §3.1 and nothing else; the scoping note now says so. Still open from §5: Palais formula, naturality under maps, linearity, Leibniz (needs the wedge of sections). | > | Manifold exterior calculus — steps (3)–(4): top-forms→measures, then the symplectic/Kähler/moment-map layer | **L–XL** each | Low | High (it is `R-016`'s wall, and `momentMap`'s) | **Price as a Mathlib programme, not a repo task.** Gated on (2b). For (3): `PartitionOfUnity.lean` exists, and there is **no measure theory on manifolds at all**. For (4): `SymplecticForm` and `MomentMap` are 0 declarations ⚠️ **(4)'s first sentence LANDED 2026-09-07 (evening):** `Mathlib/Geometry/Manifold/SymplecticForm.lean` (`DifferentialForm.IsSymplectic`, closed + non-degenerate) and `Instances/ProjectiveSpaceFubiniStudySymplectic.lean` (★★ `fsForm_nondegenerate` by Cauchy–Schwarz taming in the chart, ★★★ `fsForm_isSymplectic`: **`ℂℙⁿ` with the Fubini–Study form is a symplectic manifold**), 7 pins; names registered in check-claims' symplectic-vocabulary inventory (parity 2n, even). **Still open in this row:** (3) top forms → measures (the top-power identity), a manifold-level Kähler predicate, and the generator / moment-map layer. ⚠️ **(3) SCOPED 2026-09-07 (late):** [`specs/top-power-scoping.md`](top-power-scoping.md) — Route U (uniqueness: build `|fsForm^{∧n}|` as a measure, show it finite, nonzero and `U(N)`-invariant, conclude by `fsMeasure_unique`; never compute either side), seven milestones M1–M7 with M3 (the measure of a top form on a manifold with a finite atlas, via chart densities on a measurable partition + `lintegral_image_eq_lintegral_abs_det_fderiv_mul`) the genuine Mathlib gap and M6(b) (flat `fundamentalFormAlt^{∧n} ≠ 0`, a shuffle count) the genuine mathematics; everything else is the chart plumbing `c35b090` made routine. Rated **L**, P(success) medium–high up to a constant. ⚠️⚠️ **M1 + M3 LANDED 2026-09-08** — `Alternating/TopForm.lean` (the Jacobian rule for the coefficient of a top form) and `Geometry/Manifold/TopFormMeasure.lean` (★★ `chartMeasure_congr` by change of variables along the transition, `ChartCover` + its measurable partition, ★★ `topFormMeasure`, `topFormMeasure_apply_of_subset_source` for any chart, ★ `topFormMeasure_congr_cover`), 16 pins, plus `Instances/ProjectiveSpaceChartCover.lean` (the affine atlas of `ℂℙⁿ` as a cover). **The Mathlib gap of this step — a measure from a top form on a manifold — is closed; the stop condition held.** Remaining: M2 (wedge of sections), M4 (smooth `U(N)` action, pullback, invariance), M5–M7 (assembly). ⚠️ **M2 LANDED 2026-09-08 too** — `Alternating/WedgeCLM.lean` (★ `norm_wedge_le`, `wedgeL`, ★ `wedge_compContinuousLinearMap`, `domDomCongrL`) and `Geometry/Manifold/WedgeForm.lean` (★★ `DifferentialForm.wedge`, `domDomCongr`, `constZero`, ★★ `wedgePow`), 29 pins; `fsTopForm n := wedgePow fsForm n` is the `2n`-form on `ℂℙⁿ` whose measure M5–M7 evaluate. **Remaining: M4, M5–M7.** ⚠️⚠️ **M4 + M5 + M6(a),(c) LANDED 2026-09-08** — `Instances/ProjectiveSpaceUnitaryAction.lean` (★★ `fsModelForm_uTrans`: the chart form is `U(n+1)`-invariant, via a holomorphic pluriharmonic-log lemma), `TopFormMeasure.lean` (★★ `topFormMeasure_map_eq`: a form-preserving homeomorphism preserves the measure; `isFiniteMeasure_topFormMeasure`: finite on a compact manifold, by local finiteness, no decay estimate), `Instances/ProjectiveSpaceFubiniStudyVolume.lean` (`fsVolume`, ★★ `fsVolume_map_smul`, ★ `isFiniteMeasure_fsVolume`, ★★★ `fsVolumeNormalized_eq_fsMeasure`: **the normalised volume of the top power of the Fubini–Study form IS `fsMeasure p₀` — under the premise `fsVolume n ≠ 0`**), 29 pins. **Remaining: M6(b) — the flat non-vanishing `fundamentalFormAlt^{∧n} ≠ 0` (a shuffle count), which discharges the premise; M7 the constant.** ⚠️⚠️ **M6(b) LANDED 2026-09-08 (later the same day)** — `Alternating/WedgeShuffle.lean` (★★ `wedge_mul_apply_pairs`: the shuffle sum of a wedge with a 2-form on a pair family, sign-free because every surviving class has a two-disjoint-transposition representative), `TopFormMeasure.lean` (★ `topFormMeasure_ne_zero_of_localRep_ne_zero`: one nonzero chart coefficient makes the measure nonzero), the Volume module (★★ `wedgePow_stdForm_pairFamily`: the `k`-th power of the standard symplectic form on `k` standard pairs is `k!`; ★★ `fsVolume_ne_zero`), 39 pins. **★★★ `fsVolumeNormalized_eq_fsMeasure` is now UNCONDITIONAL: the normalised volume of the top power of the Fubini–Study form IS `fsMeasure p₀`.** Step (3) is complete up to the constant. **Remaining in this row: M7 (the constant — the total mass of `fsVolume n`, which would make the literal `ωⁿ/n!` a theorem), a manifold-level Kähler predicate, and the generator / moment-map layer (step (4)).** ⚠️⚠️ **M7 LANDED 2026-09-08 (the constant; 30 pins)** — `Analysis/SpecialFunctions/JapaneseBracketIntegral.lean` (★★ `∫_{ℂⁿ} (1 + ‖w‖²)^{-(n+1)} = πⁿ/n!`: polar coordinates on one factor, then a Fubini induction) and `Instances/ProjectiveSpaceFubiniStudyMass.lean` (★★ the density `(-4)ⁿ n! (1 + ‖w‖²)^{-(n+1)}` everywhere on the chart, by rotating to the first axis with a unitary — a complex matrix's real determinant is `|det|²`, so the Jacobian is `1`; the null hyperplane; ★★ `fsVolume_univ = (4π)ⁿ`; **★★★ `fsVolume_eq_smul_fsMeasure`: `ω_FS^{∧n} = (4π)ⁿ · μ_FS`, every factor visible**). **Step (3) is COMPLETE: every milestone of `top-power-scoping.md` is built. Remaining in this row: a manifold-level Kähler predicate, and the generator / moment-map layer (step (4)).** ⚠️ **Step (4) RE-PRICED 2026-09-08 (evening):** [`generator-layer-scoping.md`](generator-layer-scoping.md), seven numbered sub-bricks **G1–G7**. Recommended **G1 → G6** (`IsHamiltonianVectorField` at manifold level, **S**; then the moment map of the torus action on ℂℙⁿ as a manifold theorem, **L**): together **L**, P medium, value **high** — it discharges the `TERMS.md` moment-map line, and it is the compact-Kähler-pointer route the 2026-08-02 review recorded. **G5** (Liouville at manifold level) is **XL** and not scheduled: Mathlib has integral curves on manifolds but no flows and no Cartan calculus, and the corpus's measurement pieces are not globally Hamiltonian anyway. G7 (Kähler predicate) **M**, cosmetic. ⚠️ **G1 LANDED 2026-09-08 (evening; 13 pins)** — `Geometry/Manifold/HamiltonianVectorField.lean`: `IsHamiltonianVectorField α X H` (the equation `α x (X x, v) = mfderiv H x v`), `IsLocallyHamiltonian`, `interiorProduct`; ★ `dH (X) = 0`, ★ uniqueness for a symplectic form, linearity in `H`. Predicates on families, no existence, no inhabitant. ⚠️⚠️ **G6 LANDED 2026-09-08 (same night; 15 pins)** — `Instances/ProjectiveSpaceMomentMap.lean`: ★★★ `torusField_isHamiltonianVectorField`, **the torus action on ℂℙⁿ is Hamiltonian for the Fubini–Study form with `2 ∑ θₖ momentMap` as its Hamiltonian** — the moment-map equation of the cell law at manifold level, with the field proved to be the velocity of the action. The `TERMS.md` moment-map line is discharged; the factor `2` is the `-4` convention of `fsChartForm`. **Remaining on this ladder:** G2–G4 (existence and smoothness of `X_H`, Mathlib staging, optional), G7 (Kähler predicate, cosmetic), G5 (Liouville at manifold level, XL, not scheduled), `R-016′` (statement only). Nothing here is a physics brick. ⚠️ **Re-numbered and extended 2026-09-08 (late), `generator-layer-scoping.md` §8: G8–G13.** Next = **G8** (a moment map is unique up to a constant and the normalisation pins it — the sentence `POSITS.md` bullet 1 calls unformalised; **M**, P high, value high for the ledgers, nil for the posit count), then G9 (image is the simplex, S–M), G10 (the torus flow preserves `fsVolume`, S). ⚠️⚠️ **G8 LANDED 2026-09-08 (late; 9 pins)** — ★★ `eq_add_const_of_isHamiltonianVectorField` (a Hamiltonian is unique up to a constant on ℂℙⁿ, no connectedness lemma) and ★★★ `eq_torusHamiltonian_of_nonneg_of_sum` (**non-negativity and the sum-two normalisation pin a family of phase-field Hamiltonians to `2 · momentMap`**). The "standard symplectic argument, unformalised" of `POSITS.md` bullet 1 is a theorem; **Posit 1 is unchanged** (it asserts the dynamics generates the action). **Next: G9, G10.** ⚠️⚠️ **G9 + G10 LANDED 2026-09-09 (11 pins, same module)** — ★★ `range_momentMap` (**the image of the moment map is exactly the standard simplex** `stdSimplex ℝ (Fin (n + 1))`: the polytope of this action by direct computation — the ray of `(√t₀, …, √tₙ)` — with the Atiyah–Guillemin–Sternberg convexity theorem neither used nor proved) and ★ `fsVolume_map_torusUnitary_smul` (**every time-`t` map of the torus flow preserves `fsVolume n`**, a corollary of unitary invariance, with `torusUnitary_add_smul` the group law: Liouville in the dynamics sense for the one flow G6 built; G5 stays unscheduled and **Posit 3 is untouched** — the constraint dynamics' measurement pieces are not globally Hamiltonian). `TERMS.md` moment-map and Liouville entries updated; `POSITS.md` Posit 3's stale scope bullet (it still said the top-power identity was open) corrected. **Remaining on this ladder, all optional:** G11 (Hamiltonian ⇒ locally Hamiltonian, M), G12 (analyticity of `fsForm`, registry line), G13 (the `U(n+1)` moment map, M–L), G2–G4 (Mathlib staging), G7 (cosmetic); G5 not scheduled; `R-016′` statement only. Nothing here is a physics brick. ⚠️ **G11 LANDED 2026-09-09 (7 pins)** — `ExteriorDerivative.lean` gains the `0`-form API its header had listed as absent (`zeroFormFamily`, `zeroForm`, `contMDiff_zeroFormFamily`, ★ `toFlat_mextDeriv_zeroFormFamily`: **`d` of a `0`-form is its differential**, from `extDeriv_constOfIsEmpty`), and `HamiltonianVectorField.lean` gains ★ `IsHamiltonianVectorField.isLocallyHamiltonian` (**Hamiltonian implies locally Hamiltonian**: `d(ι_X ω) = d(dH) = 0` for a `C^∞` energy). The converse is false and not stated — closed-not-exact is the flux obstruction. **Remaining, all optional:** G12 (analyticity of `fsForm`, registry line), G13 (the `U(n+1)` moment map, M–L), G2–G4, G7; G5 not scheduled; `R-016′` statement only. ⚠️⚠️ **G13 LANDED 2026-09-09 (new module `Instances/ProjectiveSpaceSchrodingerFlow.lean`, 25 pins)** — ★★★ `schrodingerField_isHamiltonianVectorField`: **for every Hermitian `H`, the Schrödinger flow `p ↦ exp(-itH) • p` on ℂℙⁿ is Hamiltonian for the Fubini–Study form, with Hamiltonian `-2 ⟨H⟩ = -2 ⟪z, Hz⟫.re/‖z‖²`** — the `U(n+1)` moment map, the field ★ proved to be the velocity of the corpus's own `schrodingerUnitary`. G6's torus is the diagonal case (`schrodingerHamiltonian_neg_diagonal`). Proved in the ambient inner product (the chart tangent lift `insertZeroCLM`), not by coordinate sums. The sign and factor are the `-4` convention plus `exp(-itH)`; Posit 1 untouched. **Remaining, all optional:** G12 (analyticity, registry line), G2–G4 (Mathlib staging), G7 (cosmetic); G5 not scheduled; `R-016′` statement only. ⚠️ **G2 LANDED 2026-09-09 (14 pins)** — `HamiltonianVectorField.lean`: the flat map `v ↦ α x (v, ·)` into the dual, non-degeneracy = injectivity, finite dimension makes it an equivalence, and ★★ `hamiltonianVectorField α hnd H = fun x => (ω♭ₓ)⁻¹ (dH_x)` with **existence** (`hamiltonianVectorField_isHamiltonianVectorField`) and **uniqueness** (`IsHamiltonianVectorField.eq_hamiltonianVectorField`); `IsSymplectic.hamiltonianVectorField` for a symplectic form. On ℂℙⁿ the torus field and the Schrödinger field ARE the constructed fields (`torusField_eq_hamiltonianVectorField`, `schrodingerField_eq_hamiltonianVectorField`). Pointwise only — the smooth section is G3. ⚠️⚠️ **G3 LANDED 2026-09-09 (16 pins)** — ★★★ `contMDiff_hamiltonianVectorField` (`HamiltonianVectorField.lean`): **G2's field `x ↦ (ω♭ₓ)⁻¹ (dH_x)` is a `C^∞` section of the tangent bundle** for a `C^∞` non-degenerate 2-form and a `C^∞` energy (`hamiltonianVectorFieldSection`, `IsSymplectic.contMDiff_hamiltonianVectorField`). Route: in the tangent trivialisation the field is `ContinuousLinearMap.inverse (curryLeft (localRep α x₀ w))` on the chart derivative of `H` — by uniqueness at the flat level — and inversion is smooth at an invertible point (`contDiffAt_map_inverse`). Corollaries on ℂℙⁿ: both Hamiltonians are `C^∞`, so ★★ **the torus field (G6) and the Schrödinger field (G13) are `C^∞` vector fields** (`contMDiff_torusField`, `contMDiff_schrodingerField`). The plumbing walls (no `FiniteDimensional` on alternating-map spaces; `→L` over them picks raw instances; constant families must be wrapped to be fibre-typed) are recorded in the scoping note. **Remaining, all optional:** G4 (integral curves of `X_H` exist, are unique and conserve `H`, S–M — now unblocked), G7 (cosmetic), G12 (registry line); G5 not scheduled; `R-016′` statement only. ⚠️⚠️ **G4 LANDED 2026-09-09 (13 pins)** — `HamiltonianVectorField.lean`: ★★ local existence of integral curves of `X_H` (Picard–Lindelöf in the chart on G3's section), ★★ uniqueness of global ones (Hausdorff), ★★ **energy conservation** (`IsHamiltonianVectorField.comp_eq_of_isMIntegralCurve`: `H` is constant along every integral curve of a Hamiltonian vector field of `H`, from G1's `dH (X) = 0`); `IsSymplectic.` forms. On ℂℙⁿ (`ProjectiveSpaceSchrodingerFlow.lean`): the same for `schrodingerField`, ★★ `⟨H⟩` conserved along every integral curve, and ★★★ `isMIntegralCurve_schrodingerUnitary_smul` — **the corpus's Schrödinger flow `t ↦ exp(-itH) • p` IS the integral curve of its Hamiltonian vector field**, so ★★ `⟨H⟩` is conserved by the flow (`expectation_schrodingerUnitary_smul`). Not stated: a global flow of a general field (Mathlib has no flows on manifolds — G5's wall) and the torus orbits (same route, unwritten). **The step-(4) ladder is complete except G5 (XL, unscheduled), G7 (cosmetic) and G12 (registry line); `R-016′` statement only. Nothing here is a physics brick; the next real work is off this ladder ("▶ NEXT STEPS").** ⚠️ **G7 LANDED 2026-09-09 (13 pins)** — `DifferentialForm.IsAlmostKahler β J` (`HamiltonianVectorField.lean`: symplectic + `J² = -1` + `J`-invariance + taming, with the compatible metric `g = β (J ·, ·)` symmetric and positive definite) and ★★ `fsForm_isAlmostKahler` (`Instances/ProjectiveSpaceFubiniStudySymplectic.lean`): **`ℂℙⁿ` with the Fubini–Study form and `J = i·` is almost Kähler**, with ★ `fderiv_chart_transition_smul_I` — the chart transitions are holomorphic, so `J = i·` is the complex structure of the atlas, chart-independent. "Almost" is deliberate: the Nijenhuis integrability of `J` is not stated. **The ladder is complete except G5 (XL, unscheduled) and G12 (registry line).** ⚠️ **G12 LANDED 2026-09-10 (11 pins; priced M–L "registry line only", took S)** — `Kahler.contDiff_omega_fsPotential` (the potential `log(1 + ‖z‖²)` is real-analytic: the `C^∞` proof at `ω`) pushed through `d^c`, `dd^c`, the pullback and the chart, every step already generic in the order: ★★ `contMDiff_omega_fsForm` — **the Fubini–Study form is an analytic section of the analytic manifold** — and `fsFormAnalytic`, the same section as a term of the `ω` type (`Instances/ProjectiveSpaceFubiniStudyForm.lean`). Nothing downstream is restated at `ω`. **The step-(4) ladder is complete except G5 (XL, unscheduled); `R-016′` statement only. Nothing here is a physics brick.** ⚠️⚠️ **G14a LANDED 2026-09-10 (9 pins)** — `DifferentialForm.IsKahler β J J₀` (`HamiltonianVectorField.lean`: almost Kähler, and `J` is the model's complex structure through the tangent trivialisation of every chart, whence ★ every chart transition is holomorphic — the textbook definition) and ★★★ `fsForm_isKahler` (`Instances/ProjectiveSpaceFubiniStudySymplectic.lean`): **`ℂℙⁿ` with the Fubini–Study form and `J = i·` is a Kähler manifold.** ⚠️ **Policy change, same day (author):** nothing on this ladder is "unscheduled" any more — every residue is a priced numbered row in [`generator-layer-scoping.md`](generator-layer-scoping.md) **§9**: G5 (Liouville, XL, three milestones; Mathlib now has uniform-time global integral curves, still no Lie derivative), G14b (Kähler in the tensor sense via `VectorField.mlieBracket`, M–L), G15 (`J` as a smooth section, S–M), G16 (torus orbits are integral curves, S), G17 (Riemannian volume of the FS metric, L), G18 (Darboux, XL), G19 (analytic Hamiltonian fields, S–M), and **W1** (wire `KahlerOnticSetup`'s `ℂℙⁿ` instances to the manifold layer and correct the four stale ledgers, M, value high — the item that turns the sector from posited to proved). The author picks; Mathlib landings deprecate. ⚠️⚠️ **W1 LANDED 2026-09-10 (10 pins; new module `LF4/SectorManifold.lean`)** — ★★★ `unitaryFlowSetup_liouvilleMeasure_eq_fsVolumeNormalized` (**the sector's Liouville measure IS the normalised top power of the Kähler form**), `fsVolume_eq_smul_unitaryFlowSetup_liouvilleMeasure` (with the constant), ★★ `unitaryFlowSetup_flow_measurePreserving_fsVolume` (the flow preserves `ω_FS^{∧n}` itself), ★★★ `fsVolumeNormalized_isForcedKahlerVolume` (**the forced volume of `KahlerVolumeForced.lean` IS the symplectic volume** — symmetry and the Kähler form pin the same measure), the same three on the many-to-one sector `ℂℙⁿ × T²`, and ★★★ `unitaryFlowSetup_isKahler_liouville` / `manyToOneSetup_isKahler_liouville` (**the sector is the standard object**: Kähler target, Liouville = normalised top power, flow preserves it). Structure unchanged, Posit 3 untouched. **Connectivity link L1 DISCHARGED**; the stale claims in `KahlerOnticSetup.lean`, `KahlerVolumeForced.lean`, `LiouvilleUnique.lean`, `ManyToOnePillars.lean`, `NonTrivialSetup.lean`, `connectivity-manifest.md`, `reconstruction-status.md`, `future-work.md` (KG-1 DONE), `TERMS.md` (Fubini–Study, symplectic), `README.md` and `docs/TOUR.md` corrected. `isFiniteMeasure_fsVolume` is now an instance. ⚠️ **G15 + G16 LANDED 2026-09-10 (10 pins)** — ★★ `IsKahler.contMDiff_hom_section` (**the complex structure of a Kähler structure is a `C^∞` section of `Hom(TM, TM)`**, constant `J₀` in every chart) with `contMDiff_fsJL` on `ℂℙⁿ`; ★★★ `isMIntegralCurve_torusUnitary_smul` (**the torus orbit `t ↦ diag(e^{itθ}) • p` IS the integral curve of `torusField θ`**, unique through `p`: `eq_torusUnitary_smul_of_isMIntegralCurve`) and `torusHamiltonian_torusUnitary_smul` (`2 ∑ θₖ μₖ` conserved). **§9 residue now: G5 (XL), G14b (M–L), G17 (L), G18 (XL), G19 (S–M).** ⚠️ **G19 LANDED 2026-09-10 (10 pins)** — ★★★ `contMDiff_omega_hamiltonianVectorField` (**the Hamiltonian vector field of a `C^ω` energy for a `C^ω` non-degenerate 2-form is a `C^ω` section**: G3 at `ω` on an analytic manifold, via `contDiffAt_omega_localRep` and the one-line bridge `ofOmega`) and on `ℂℙⁿ` `contMDiff_omega_schrodingerField` / `contMDiff_omega_torusField` (**both fields are analytic vector fields**, for `fsFormAnalytic`). **§9 residue now: G5 (XL), G14b (M–L), G17 (L), G18 (XL).** ⚠️⚠️ **G14b LANDED 2026-09-10 (7 pins; priced M–L, took M)** — `DifferentialForm.nijenhuis` (the Nijenhuis tensor with Mathlib's `VectorField.mlieBracket`) and ★★★ `IsKahler.nijenhuis_eq_zero`: **on a Kähler manifold the Nijenhuis tensor vanishes** (every bracket is the flat bracket of the chart pullbacks, `J` pulls back to the constant `J₀`, and the flat expression cancels) — the easy direction of Newlander–Nirenberg; `nijenhuis_fsJ_eq_zero` on `ℂℙⁿ`. So `ℂℙⁿ` is Kähler in both senses. **§9 residue now: G5 (XL), G17 (L), G18 (XL).** ⚠️⚠️ **G17 LANDED 2026-09-11 (27 pins; priced L, took M–L)** — `Geometry/Manifold/RiemannianVolume.lean` (★★ `MetricFamily.riemannianVolume`: **the Riemannian volume of a metric on a manifold**, chart Gram densities `√det G` glued along a cover — the Mathlib gap; ★★ the chart-by-chart bridge to a top-form measure) and `Instances/ProjectiveSpaceFubiniStudyRiemannian.lean` (★★ the Gram determinant of the Fubini–Study metric is `(4ⁿ(1+‖w‖²)^{-(n+1)})²`, by the rotate-and-scale route of the mass computation; ★★★ `riemannianVolume_fsMetric`: **`vol_g = ω_FS^{∧n}/n!`**, the Kähler identity at the level of measures; ★★★ `riemannianVolume_fsMetric_eq_smul_fsMeasure`: **`vol_g = ((4π)ⁿ/n!) · μ_FS` — the Fubini–Study measure IS the normalised Riemannian volume of the Fubini–Study metric**). All three textbook readings of `μ_FS` now identify the same measure. Glossary entries `fubini-study-measure`, `kahler-form`, `liouville-measure`, `moment-map` updated (they still said the manifold residual was blocked). **§9 residue now: G5 (XL), G17b (chart-independence of the Gram construction, S–M), G18 (XL).** | > | **Operator convexity — the Effros/Lieb summit** → joint convexity of relative entropy → DPI → unconditional SSA (row E2) | **L–XL** | Low–medium | Gated | **Superseded 2026-09-12 by the bridge** (`lieb-dpi-scoping.md` §6): Gate 0 passed, `csd-qit-bridge` exports unconditional SSA and the corpus's `hDPI` verbatim, foundational triple, from physlib on identical pins. This row is now the in-corpus *provenance* option only — build it if owning the proof matters more than having it (a values decision, §4 of the note). The instance wall that used to head this row is **resolved**; what is left is the theorem | > | **Lévy concentration / spherical isoperimetry** | **XL** | Low | Low | Blocks only the `exp(−c·d_E·ε²)` *rate*; the expectation form and a polynomial Chebyshev rate are both proved. Not worth the tower | > | **Pointwise Birkhoff** | — | — | — | ⛔ **Do not pursue** — the single-flow route is a proved dead end CSD does not take (NG1) | > | **Stone's theorem, general** | **XL** | Low | — | Nothing in flight; the dynamics spine is `Matrix (Fin N) (Fin N) ℂ` and the finite-dimensional continuity-only theorem is proved | > > ⚠️ **Steps (0) and (1) are both done, and the price paragraph below was accurate and stays.** > *Nothing in flight waited on either, and nothing does now.* The gain is exactly what was > advertised when they were picked: `ω ∧ ω` is a term, `d(ω ∧ ω) = 0` is a theorem, `ℂℙ^N` is an > analytic manifold, and a top-power identity is sayable. **No physics moved, and no residue > closed** — `R-016` is untouched. What is gone is the oddity that the space the reconstruction is > stated over had no smooth structure in Lean, so every geometric word applied to it was prose. > > ⚠️ **Honest price on steps (0)+(1).** *Nothing in flight waits on them.* The corpus routes around > all four restricted terms — `μ_FS` is pinned by uniqueness-under-symmetry, Duistermaat–Heckman by > an independent Gaussian route, the moment-map equation at the linear level. What closing them buys > is **self-containment and a statable top-power identity**, not new results. That is a real gain for > an honest-scope register and a real Mathlib contribution; it is not progress on the physics, and > should not be sold as such. > > ⚠️ **Upstreaming the staged batch is CR-9 — DEFERRED by author decision 2026-09-05**, not blocked. > `MATHLIB-GAPS.md`'s "suggested first upstream batch" is that item. Do not re-raise it as pending. > > **4. Closed to assistants** — the ⛔ section immediately below. Read it before proposing > anything: the Wigner hand-pass, manuscript edits, and `lieb-dpi` Gate 0 keep being offered > back to the author, which is why the section exists. > > ### Session record — 2026-09-02 → 09-04 (every numbered item below LANDED; kept as the construction record) > > **Closed this session** (CI-green, 13 pins): **brick 3 of the `H_int(M)` arc** — the > back-reacting joint lift on the arena (`RecordLayer/JointLift.lean`: `isJointLift_jointLift`, > `jointLift_measurePreserving` via the staged `MeasurePreserving.vadd_twist_of_invariant`, > `jointLift_base_moves_of_ne`) and the chart-level Hamiltonian generation of its shift > (`RecordLayer/HamiltonianShift.lean`: `strokeCurve_hasDerivAt_hamiltonianField`). All four > bricks of `frozen-base-obstruction-scoping.md` are landed; `R-016` is now exactly the manifold > identification `ι_X ω = dH` on the arena, paper-side (PAPER-FIRST row). ⚠️ **There is no cheap > `H_int` Lean brick left** — the honest options are nonvanishing of `hamiltonianShift` on > explicit data (M, computational, optional) or a two-stroke Lüders composition on the joint > lift; say so rather than scaffold a fifth brick. Previous session (2026-09-01, ledger > CL-062…CL-068): Q10-a/b; L.4; fidelity core + `F ≤ 1`; Bargmann lift discharging both W5 > staged data; Kähler chart identification; cat-map witness; residue registry; > Mathlib-absence guard. > > **Do next, in this order.** Item 1 is done (bricks 0 and 1 landed; WAY brick 2 waits on row B); item 2 **LANDED 2026-09-02**; 3 is mechanical. > > 1. **WAY** (row A; `specs/way-theorem-scoping.md`) — scoping **DONE 2026-09-02** and **brick 0 > LANDED 2026-09-02**: `Empirical/QM/WignerArakiYanase.lean` (`arakiYanase_identity`, > `wigner_araki_yanase`, `no_exact_record_of_not_commute`; CNOT non-vacuity, SWAP sharpness, > `sigmaX_no_exact_conserving_record`; 7 pins). The R-014 fold (`inner_toEuclideanLin_unitary` > generalised) was executed early under it. Scoping findings: the record trilemma > (`NullSeamWitness.lean`) is **not** WAY in disguise; the `ε` of `ε`-Born is **not** a WAY bound; > the first-draft framing "WAY-compatible by construction because every stroke conserves what it > reads" was **withdrawn** — the fibrewise witness freezes the whole base (`pointerEvolve_fst`), > conserving non-commuting base functions too; the honest reason WAY has no *record-layer* > instance is that the stroke is a skew product on the arena, not a tensor-product isometry with > an additive conserved `L` (`RESIDUE(R-015)`'s boundary; the LF5 `vnUnitary` tier meets the > hypotheses with an engineered `L` and WAY's conclusion holds there trivially). **Brick 1 LANDED > 2026-09-02**: `Empirical/CSD/WignerArakiYanase.lean` — the scope statement as a theorem at > the `IsJointLift` level (`record_not_factor_hilbert`, `no_joint_hilbert_map`: the pointer > image is not a function of the Hilbert data, the fibre register selects the outcome via > `IsJointLift.landing`; non-vacuous at `[(1,1)]`, instances for `pointerEvolve` and `jointLift`; > 7 pins). ⚠️ The `IsJointLift` quantifier is *weaker* than WAY's (`pointer_eq` pins every > lift's pointer image to the witness's): it buys survival under base back-reactions, not a > stronger statement (scoping §2 W3). Rule of two fired: `LF4.momentMap_mk_of_norm_eq` > (equal moduli ⇒ barycentre) now serves both `momentMap_mk_xPlus` and > `BB84Sequential.momentMap_ketPlus`. The two `rfl` lemmas first scoped were dropped as > misreadable trivia. **Row A is > closed except brick 2** (behind row B). (breadth/hardening) > 2. **brick-2 — generation-from-records** (M–L). Go given by the author 2026-09-02; scoped > (`generation-from-records-scoping.md`, `csd-foundations` 18 findings folded), then built. > **LANDED 2026-09-02**: `SigmaLayer/TensorTomography.lean` — the generation premise `hgen` of > `compositeAlgReconstruction` / `composite_dim_eq` IS record-level local tomography > (`RecordLocallyTomographic`: equal joint record rates in every pair of local bases force equal > composite densities), both directions under locality + star-preserving embeddings > (`recordLocallyTomographic_iff_adjoin_eq_top`); consumers re-premised > (`compositeAlgReconstructionOfRecords`, `composite_dim_eq_of_recordLocallyTomographic`); > non-vacuous on the Kronecker composite (`kronecker_recordLocallyTomographic`); the record rates > at a pure preparation ARE `bornRateBasis` of the joint register > (`RecordLayer/StatisticsRigidity.lean` `productRecordRate_eq_bornRateBasis`). Rule of two > fired three times: `Matrix.kroneckerLeftAlgHom`/`RightAlgHom` (Cat-1 `KroneckerAlgHom.lean`, > `CompositeArena` now consumes them); `IsHermitian.eq_eigen_outer` hoisted to any Hermitian > matrix (`density_eq_eigen_ensemble` / `eq_eigen_ensemble` its corollaries); `outerProduct` > index-generic. ⚠️ **Premise conversion, not a derivation of `⊗` from Σ**: local tomography > stays a posit, now registered as the permanent boundary `R-017`; constraint work one level > above Σ, not MD-1 progress. (reconstruction path; 20 pins) > 3. **The hardening session.** (breadth/hardening) **Tranche 1 LANDED 2026-09-04** — runLinter > adoption and the brick-2 leftovers: > * **`simpNF` 18 → 0.** Eighteen `@[simp]` lemmas whose LHS was not in normal form, so the > attribute silently never fired. Two head unfolders were the root cause: > `Channel.apply_def` (killed `id_apply`, `unitaryChannel_apply`, `traceOutChannel_apply`, > `mixedUnitaryChannel_apply`, `tensorRight_apply`) and `Kahler.complexStructure_apply` > (killed `J² = -1` and the taming identity `ω u (J u) = ‖u‖²`); both de-`simp`ed with the > reason at the declaration. **`docBlame` 20 → 0** (12 in `LF6/CGLMPQudit.lean`). > * **`unusedArguments` triage.** 68 of 73 are unused instance binders (§9.2 fix-on-touch). > The five explicit-*hypothesis* findings — the class the necessity audit found by hand — > were each read: all are documented uniform-interface bundles > (`shor_phase_estimation_lower_bound`, `cfullAdder_correct_general`, > `cuccaroAdd_preserves_B`). **No stated assumption in the corpus is unexplained.** > * **The ratchet.** `check-std-lint` now fails on GROWTH against `docs/std-lint-baseline.txt` > (CONVENTIONS §9.5), never on the existing findings — so the no-sweep decision stands and > `simpNF`/`docBlame` regressions fail on the commit that introduces them. > * ⚠️ **`check-axiom-sweep` had an unguarded coverage precondition**, found by fixing its own > self-test: the sweep sees only what `import CsdLean4` EXPORTS, and under the module system > a proof term is exported only from an `@[expose] public section`. Planted and verified: a > module-private `theorem … := by sorry` behind a `public theorem`, in a module without that > section, **passes the sweep clean**. The corpus is fully covered (583 of 583 declaring > modules have the section), and `check-axiom-sweep.sh` now checks that rather than assuming > it. The `--with-lean` self-test had been passing vacuously — the planted probe lacked > `module`, so it never compiled and the guard "fired" on the broken build (probes 12–13). > * **Brick-2 leftovers.** `prodVec` folded into `QuantumInfo.tensorState` (with > `tensorState_ne_zero`/`_smul_smul`/`_continuous` hoisted to the Cat-1 file); > `exists_trace_mul_ne` now derives from Mathlib's `Matrix.ext_iff_trace_mul_right`. > ⚠️ The "one separation lemma, three surfaces" note was **wrong** and is corrected in place: > `LF2.matrix_eq_zero_of_quadForm_zero` (quadratic form, a polarisation) and > `SigmaLayer.exists_forall_eq_trace_mul` (every functional is a trace pairing) are > different statements, not copies — only `exists_trace_mul_ne` was a duplicate. > > * **The Gleason-free claims, guarded.** 56 module headers assert Gleason-freeness and > exactly ONE import-independence pair was declared. `check-import-negative.sh` now > declares **46**: the 44 modules whose claim is structural (`LF2/EffectGleason` absent > from the import closure) plus `RecordLayer/GlobalRecordClosure` (CL-011 — the claim is > made *about* it, in the glossary and the reports, though its own header never says so). > The guard also fails now when a declared root is not a tracked module (a rename used to > leave the claim silently unchecked) and when its own BFS errors — an escaping slip made > `awk` abort mid-change and the empty result read as "all claims hold". Two new probes. > * **The 12 proof-term claims, guarded too** (2026-09-04). The modules that DO import > `EffectGleason` and claim only that their proofs route around it — `MalusVolume`, > `MachZehnderVolume`, `SternGerlachVolume`, `VolumeCanonical`, `KS18Volume`, > `MerminPeresVolume`, `KCBSVolume`, `ElitzurVaidmanVolume`, `Metrology/Ramsey`, > `SingletKahler`, `SingletKahlerFlow` — are now checked at the **constant-graph** level by > `scripts/check-gleason-free.sh` (+ `gleason-free.lean`), in CI next to the axiom sweep: > 259 declarations, none reaching `effect_gleason_representation`. A re-route through the > Busch-mediated twin now fails on the commit that makes it, instead of falsifying eleven > headers and the glossary's headline framing silently. A declared module that is renamed > away fails too, rather than being checked vacuously. Two probes in `check-guards.sh` > (16 probes). `Tests/Witnesses/SingletBell.lean` carries the same claim and is out of > scope (it is in `CsdLeanTests`, outside `import CsdLean4`); its content is an > instantiation of `LF4.ofKählerPreparation`, whose module is covered. > > **Tranche 2 — PART LANDED 2026-09-04.** > * **Module paths in prose.** The `SigmaLayer` → `RecordLayer` split had left **510 broken > path references** — 303 in 87 Lean headers, 207 in 22 documents (including the > `reconstruction-status.md` A2 row README sends readers to for the axiom-level audit). > All fixed; six left deliberately (a rename sentence's subject, a companion never built, > an audit file recorded as never created) as a declared exception list with reasons. > `check-doc-promises.sh` now checks paths as well as declaration names, over Lean and > markdown both. > * **Expert row D — Colbeck–Renner.** [`colbeck-renner-note.md`](colbeck-renner-note.md): > the escape is **parameter independence at the `Σ` level** (the Bell escape, a theorem — > `no_product_partition_realises_singlet`, CL-020), ⚠️ **not** free choice; CR's "free > choice" bundles parameter independence with measurement independence > (Ghirardi–Romano 2013, Leegwater 2016) and the corpus KEEPS measurement independence as > an explicit premise. Positioning prose. **The Lean form landed 2026-09-29** (row 23, `ChainedBell.lean` + > `ColbeckRenner.lean`): the chained walk, the mixture transfer, and the singlet instance with > bound `π²/(8(2n+1))`, stated so that parameter independence and measurement independence > appear as *separate* hypotheses — which is the unbundling this row exists for. > * **Expert row F — excess baggage.** `docs/glossary.yaml` entry `excess-baggage`: Hardy > 2004 / Montina 2008 are satisfied trivially (Σ is a continuum by construction), with the > two disclaimers a referee needs — satisfying a lower bound is not evidence, and > cardinality is not what carries the programme. > * **`REFERENCES.json`** grew from the 2-entry seed to 7 (the five sources rows D and F > cite), and is now guarded: `check-references.sh` (CI) fails when a `cited_by` path does > not exist, when a `[Key]` citation in prose has no entry, or when an entry is cited by > nothing. ⚠️ **Coverage is still the open §8.2 obligation** — one entry per source plus > line-precise `[Key, file:Lstart-Lend]` citations. The guard checks what is there; it > says nothing about how much of the corpus is cited. > > **Tranche 2 — second part LANDED 2026-09-04.** > * **Prose audit pass 4** (`prose-audit.md`): the pass-3 consumer sweep, run against every > recorded correction — twelve probes. **Three stale restatements found**, all of the A2 > torus-flux withdrawal ("piecewise Hamiltonian" → piecewise rigid symplectic translation): > two in `PiecewiseHamiltonian.lean` itself (its "## The decision" record and a section > heading) and one in `ShearDiscontinuity.lean` (route (2) named by the withdrawn name, and > still calling the choice open). Fixed; no identifier changed. Nine probes clean. > **Sharpened lesson:** the stale restatements were in the *same file as the correction* > and its nearest neighbour — a correction paragraph at the top of a module does not sweep > that module's own headings. > * **Publication errata E-3 and E-4** — Paper D §8: PBR's target stated backwards with the > three senses of "epistemic" run together (E-3), and the equivariance analogue described as > "proved twice" when μL-preservation is a *structure field* (`ConstraintDynamics.flow_preserves`, > P4) and only μ_FS invariance is a theorem (E-4). ⚠️ Both are **reported, not verified here** > — the manuscripts are not in this repository; what is verified is every repository-side > claim the entries make. > * **`REFERENCES.json` coverage, measured rather than asserted.** `check-references.sh` now > reports the gap: **7 entries against ~108 distinct `Author YYYY` citations in Lean > docstrings**. Deliberately NOT gated (the pattern over-counts dated prose) and deliberately > NOT bulk-filled: entries whose bibliographic details are unverified would make the file > bigger and less trustworthy, which is the failure its own header warns about. > > **`REFERENCES.json` — HARVESTED 2026-09-04, 7 → 37 entries.** Taken from the citing modules' > own `## Source` blocks (33 of them carry full references), so this is transcription, not > recall: `venue` is the corpus's verbatim string, `title` appears only where the corpus records > one, and `doi`/`url` only where the identifier is derivable from the citation (APS constructs > DOIs from journal+volume+page). `check-references.sh` now accepts *title OR venue* — authors + > year + "Phys. Rev. Lett. 67, 661" is auditable, and requiring a title would have forced exactly > the half-remembered bibliography the file exists to prevent. **Still open:** ~73 of ~110 cited > sources have no entry, and the line-precise `[Key, file:Lstart-Lend]` citations §8.2 asks for > are untouched. And **prose audit passes 1–2 over the > remaining four-fifths** of the surface; the pass-3 sweep against corrections recorded only > in commit messages is also unrun. > 4. **Ozawa error–disturbance twin** (M, row B) — **LANDED 2026-09-04** > ([`ozawa-scoping.md`](ozawa-scoping.md), `csd-foundations`-checked, 15 findings folded). > `Empirical/QM/Ozawa.lean`: ★★ `ozawa_error_disturbance`, the universally valid relation > `ε(A)η(B) + ε(A)σ(B) + σ(A)η(B) ≥ ½|⟪Ψ,[A_in,B_in]Ψ⟫|`, from ★ `ozawa_commutator_identity` > (the fourth commutator term vanishes by `commute_out` — that is the whole content) and three > instances of the new unsquared ★ `commutator_le_two_mul_norm`, which `robertson_core` is now > **derived from** (§9.3). Non-vacuous where it matters: `zxWitness_commutator_ne_zero` gives a > nonzero right-hand side, and ★ `ozawa_two_term_false` refutes the two-term variant outright. > `Empirical/CSD/Ozawa.lean`: ★ `no_ozawa_model_of_jointLift` — the record layer supplies **no** > such model, so `ε` and `η` are not defined there (a corollary of `no_joint_hilbert_map`, not a > parallel capstone). Six pins. ⚠️ **Not built, and not claimed:** a measurement-model > instantiation deriving `A_out = U†(1 ⊗ M)U` from a probe — the inequality turned out not to > need the tensor product at all, so the structure is four operators on one space. The naive > Heisenberg form remains un-refuted in Lean, deliberately. Two decisions settled there and worth carrying: the core is stated on WAY's > **abstract-`tensor` interface**, not Mathlib `⊗[ℂ]` (§3a — otherwise brick 2 cannot share the > definitions `way-theorem-scoping.md` says it shares), and the CSD side is a **scope theorem** on > the brick-1 pattern, NOT a volume-ratio twin (§5 — there is no Σ-side reading of `ε`/`η`, and the > named template `Empirical/CSD/Uncertainty.lean` is itself SCHEMA-MISMATCH / TRANSPORT-ONLY). > ⚠️ The row's "no new infrastructure" likelihood note is inaccurate — `isSymmetric_map`, an > unsquared Cauchy–Schwarz core, and `stdDev` are all new — though the **High** rating stands. > **Local Friendliness twin** (M, row C) — **LANDED 2026-09-04** > ([`local-friendliness-scoping.md`](local-friendliness-scoping.md), `csd-foundations`-checked, > 14 findings folded). `Empirical/CSD/LocalFriendliness.lean`: the assumption CSD denies is not > one of LF's three as listed but a fourth the derivation uses tacitly — **persistence** of the > friends' records. ★ `jointLawInvariant_of_persistent`: on a single `Σ` with a single `μ`, > fixed records *force* LF Locality, so a model reproducing the violation must deny persistence > (★★ `not_persistent_of_jointLaw_moves`, non-vacuous by `movingRecords_not_persistent`). > **CSD keeps Absoluteness of Observed Events** — definite at every moment — and denies only > invariance across the party's operation, which is forward evolution, not reversal. 3 pins. > ⚠️ **Not built:** no LF inequality, no derivation that CSD's joint law moves (that needs the > friends' labs). ⚠️ The scoping note's own §4 deliverable was **refuted before building** — its > `RemoteJointInvariantA` demanded that A's outcome be independent of A's own setting, which no > model satisfies; superseded by §4a. > > 5. ⚠️ **The Q16 CP brick is NOT pending — it landed 2026-08-20**, and this list said otherwise > until 2026-09-04. `Mathlib/Analysis/NormedSpace/TrotterGeneral.lean` + `LF6/LindbladPositivity.lean`: > ★★ `NormedSpace.trotter_product` (de-skewed Lie–Trotter), ★★ `lindbladSemigroup_posSemidef` > (every GKSL flow with Hermitian `H` is positive at every `t ≥ 0`), and — closing the brick's > own declared boundary — ★★ `idTensor_lindbladSemigroup` + ★★ > `lindbladSemigroup_completelyPositive` (Q23, same week): `id ⊗ Φₜ` maps PSD to PSD for every > finite ancilla, i.e. complete positivity in so many words. Pinned in > `Tests/AxiomAudit/Dynamics.lean`. > > **So the reconstruction-path row of this list is closed.** What rows B and C unblock instead > is **WAY brick 2** (quantitative WAY, `way-theorem-scoping.md` §3), deferred *behind row B* > and now free. ⚠️ Its **L** rating predates row B and is likely stale: brick 2 was scoped to > "share every definition" with Ozawa, and the Ozawa core turned out to need **no tensor product > at all** (four symmetric operators, two of them commuting), so `OzawaData` composes with WAY's > abstract-tensor interface whichever way brick 2 goes. Re-rate before queueing. > > **WAY brick 2 LANDED 2026-09-04, the same day** (`Empirical/QM/QuantitativeWAY.lean`, > 2 pins): ★★ `quantitative_way` — `ε(A)·σ(L) ≥ ½|⟪Ψ,[A_in,L]Ψ⟫|`, an apparatus measures `A` > accurately only at the price of charge variance — and ★ `commutator_eq_zero_of_error_eq_zero`, > the structural WAY theorem as its `ε = 0` corner. ⚠️ The **L** rating was wrong and the reason > is instructive: sharing row B's definitions collapsed the work to one commutator identity plus > a single `commutator_le_two_mul_norm`. Rating a brick before the vocabulary it depends on > exists is rating the wrong problem. ⚠️ `Commute A_out L` is assumed, not derived (`OzawaData` > carries no `U`); it encodes the probe observable commuting with the apparatus charge, the > standard extra assumption of the quantitative treatment. > 5. Opportunistic: `_of_` hypothesis pattern; Ecdsafail doc de-application (~70 stale ECDLP refs in > Cat-1 `Reversible/` headers); the general Boolean→amplitude lift (`RESIDUE(R-013)`'s neighbour). > > ### ⛔ NOT ASSISTANT WORK — do not re-raise these as "what's next" > > *(Added 2026-09-04, because they kept being offered back to the author in session summaries. > Each is here for a stated reason, not because it is hard. An assistant reading this list should > treat these as **closed to it** and say so in one line if asked, rather than proposing them.)* > > * ⛔ **G11 / CL-024 / CL-014 — the Wigner hand-pass.** Moved off the queue by author decision > 2026-08-07: *"owner-tracked with the author's own review work, same stroke as §C; not > repository work."* It is a human review of the five highest-risk Wigner tactic blocks for > whether the reasoning is right. An assistant re-deriving them would defeat its purpose — the > point is a second, human pair of eyes on machine-checked proofs. **Do not offer to do it.** > * ⛔ **Manuscript edits, anywhere.** Standing rule. Defects found in published or circulating > text go to `specs/publication-errata.md` as an entry (currently E-1…E-4) and stop there. The > entry closes when the *text* is amended, which is the author's act. Assistants correct the > repository side and record the debt; they do not edit papers and do not track their progress. > * ~~⛔ **Gate 0 of `lieb-dpi-scoping.md` — BLOCKED, not pending.**~~ **Resolved 2026-09-12.** Gate 1 was > answered by the author picking W9 from the QIT-chain queue (`qit-chain-scoping.md`); Gate 0 then ran > from `C:\zayn\physlib` and passed, and route (b) was built as `csd-qit-bridge` outside the corpus > (`lieb-dpi-scoping.md` §6). The corpus itself is unchanged. **Published 2026-09-12** at > https://github.com/zblore/csd-qit-bridge with by-revision requires and its own CI. The BCFJS `iff` > of `NoBroadcasting.lean` is not unlocked by it and, it turned out (2026-09-12, row **BC**), does not > need DPI at all: an elementary route through support confinement; BC1–BC2 landed, BC3–BC6 L–XL. > * ⛔ **B6 (Mathlib PRs), retired 2026-08-06.** Retirement logic unchanged. The one live question > is the *summit* case in `lieb-dpi-scoping.md` §3b, which is an author call, not a queue item. > > **Open decisions that are the author's, not technical.** > > * ⚠️ **The 49 untagged "Mathlib has no …" claims.** `check-mathlib-absence.sh` reports them as a > non-fatal warning. The register rows are tagged; the in-corpus prose is not, and will rot exactly > as the five corrected rows did. **49 → 44 on 2026-09-04**: the five claims of the *Kähler / > symplectic manifold API* wall now carry the register's own token > (`file:Mathlib/Geometry/Manifold/DifferentialForm`) rather than an invented one. > ⚠️ **The residue is not mechanical, which is why it stopped there.** Each remaining claim needs a > token naming what would exist if Mathlib added it, and a wrong guess produces a tag that can > NEVER fire — decoration wearing a guard's clothes. Where a claim maps to an existing > `MATHLIB-GAPS.md` row, reuse that row's token (honest, and what was done here). Where it does > not — e.g. "no Finset-level `factorization_lcm`" — the token is a guess about upstream naming > and is the author's call, or the claim should be reworded to name a declaration that exists. > * ⚠️ **B6 (Mathlib PRs), retired 2026-08-06.** Unchanged for L.4's lemmas — they genuinely do not > advance CSD, so the retirement logic holds. It is the **summit** where a PR would remove a wall > the corpus is standing behind; `lieb-dpi-scoping.md` §3b makes that case and leaves it open. > * ⛔ **CL-024 / CL-014** — the G11 Wigner hand-pass (five tactic blocks), author-only. See the ⛔ section above: not assistant work, do not re-raise. > * ⚠️ **Paper C's Born-from-dynamics is ε-Born with ε free** (the smooth-pointer witness gives a > sandwich, `ε`-close to Born for every `ε > 0`, exact only in the limit). Where the paper cites > the smooth witness it should say so. Repo side is correct; the paper is the author's. > > ~~**One thing genuinely unfinished — and ⛔ BLOCKED on a decision, not on effort**~~ **Done 2026-09-12:** > Gate 0 of `lieb-dpi-scoping.md` ran from `C:\zayn\physlib` (the 2026-09-01 attempt had died on a > Windows path-length error from the long scratchpad path) and passed — `#print axioms` on physlib's SSA, > joint convexity and DPI is the foundational triple; the pin is checked by > `csd-qit-bridge`'s `Tests/AxiomAudit.lean`, not by a green build (physlib sets `-Dwarn.sorry=false`). > > **Repo hygiene note.** Working-tree line endings were normalised to LF on 2026-09-01 and > `core.autocrlf=false` / `core.eol=lf` set **repo-locally**, ending the phantom modified-badges > `.gitattributes` was added to prevent. No content changed (`git add -A` staged zero). > > ### Expert-review additions — 2026-09-02 (QM/EFT coverage sweep) > > A both-ways grep of ~40 reviewer-level QM/EFT topics against `CsdLean4/`, `specs/`, `docs/`. > **Present and not re-listed:** Bell/CHSH/Tsirelson/monogamy, Kochen–Specker, Leggett–Garg, PBR, > Spekkens toy model, Zeno, weak values (`Empirical/CSD/WeakMeasurement.lean`), Lindblad/CP, > Robertson uncertainty (`csd_robertson_uncertainty`), canonical typicality, Gibbs, H-theorem, > no-cloning, tomography, Stern–Gerlach, Lorentz/dispersion (`CV/DispersionEarned.lean`), > Lieb–Robinson, interferometry cluster; EFT pillars P1/P2/P4/P5 complete, P3 the declared ceiling. > **Seven absences survive**, ranked; same columns as the triage table. Canonical here — the boards > in `qm-empirical-tests.md` carry pointer rows only. > > | Item | Cx | P(success) | Value | Verdict | > |---|---|---|---|---| > | **A — Wigner–Araki–Yanase (WAY).** The "Araki" hits in the corpus are all Araki–Lieb (entropy); WAY is absent. Two halves: (0) QM-side finite-dim Araki–Yanase theorem in `Empirical/QM/` (a measurement that conserves an additive `L` measures exactly only observables commuting with `L_S`); (1) the CSD reading — a *scope* theorem, not a result: the stroke is a skew product on the arena, not a linear isometry on `HS ⊗ HA`, and no additive `L_S ⊗ 1 + 1 ⊗ L_A` is modelled, so WAY's hypotheses have no record-layer instance (`RESIDUE(R-015)`; the LF5 `vnUnitary` tier meets them with an engineered `L`, and WAY's conclusion holds there trivially). ⚠️ The first framing "every stroke conserves what it reads, and everything conserved commutes, so CSD is WAY-compatible by construction" was **withdrawn at scoping**: the fibrewise witness freezes the base (`pointerEvolve_fst`), conserving every base function — including the non-commuting `rotatedProj` expectation — while recording exactly; that is the combination WAY forbids, available only *outside* WAY's hypotheses. Scoping answered: the record trilemma (`NullSeamWitness.lean`) is **not** WAY in disguise (its costs are connectedness/openness/measure theorems, no conserved quantity); the `ε` of `ε`-Born is the collar measure (`born_lower`/`born_upper`), **not** a WAY (Ozawa 2002 apparatus-variance) bound | M (spec S) | High for (0); (1) is a short `IsJointLift`-level theorem in `Empirical/CSD/` | **Medium–high** — the QM-side theorem is the deliverable (finite-dim, first in Lean as far as known); the CSD side is a machine-checked scope pointer for the referee, not a headline | **Brick 0 LANDED 2026-09-02** (`Empirical/QM/WignerArakiYanase.lean`; `specs/way-theorem-scoping.md`); **brick 1 LANDED 2026-09-02** (`Empirical/CSD/WignerArakiYanase.lean`: `no_joint_hilbert_map`, at the `IsJointLift` level — a robustness quantifier, weaker than WAY's); brick 2 behind row B | > | **B — Ozawa error–disturbance** (universally valid uncertainty, Ozawa 2003; Erhart et al. 2012, Rozema et al. 2012). Robertson is in; the measurement-theoretic inequality that the Heisenberg-microscope form *fails* is not. Finite-dim, operator algebra only; a CSD twin follows the `Uncertainty.lean` pattern | M | High — operator inequality, no new infrastructure | Medium — closes the uncertainty cluster at the level a 2012+ referee expects | Add to twins board; hardening-adjacent | > | **C — Local Friendliness** (Bong et al. 2020; Cavalcanti–Wiseman). D10/D11 (`qm-empirical-tests.md` §3.3) are conceptual rows only. LF is the *testable* Wigner's-friend no-go and CSD must say **which assumption it denies, as a theorem**: the same setting-locality it denies for Bell — the denial already lives in `no_product_partition_realises_singlet` (`LF6/ForcedContextuality.lean`) and `LF3/SettingLocality.lean`; the Bell twins (`Empirical/CSD/Bell.lean`) are transport re-exports of it — *not* Absoluteness of Observed Events (records are ontic selections in `Σ`). ⚠️ LF's Locality is *event-level* (an outcome does not depend on a distant setting), stronger than Bell's parameter independence; the denial theorem must be stated at that level, not re-exported from CHSH. Deliverable = the LF inequality as a theorem + the event-level denial as a named theorem, no more | M | High — the inequality is CHSH-shaped; the denial needs stating at LF's level, not restating | **High** — the current-decade paradox, and the one place the record layer's ontic-record claim is directly tested | Queue after A | > | **D — Colbeck–Renner** (no extension of QM has improved predictive power, 2011) as a **positioning theorem**: CSD escapes it by denying **parameter independence at the `Σ` level** — the same escape as Bell (`no_product_partition_realises_singlet`). ⚠️ Not by denying free choice: CR's "free choice" *bundles* parameter independence with measurement independence (Ghirardi–Romano 2013; Leegwater 2016), and the corpus **keeps** measurement independence as a premise (`LF3/OperationalNoSignalling.lean`). Record the escape as a named theorem, with the unbundling stated, not a footnote | S spec / M–L Lean | High for the spec note; Lean L because the CR argument needs the chained-Bell family | Medium — referee-facing | Spec note in the hardening session; ~~Lean only if asked~~ **Lean DONE 2026-09-29** (row 23) | > | **E — Quantum Darwinism / record redundancy** (Zurek; Riedel–Zurek). The corpus proves records are *made* (`Pointer*`, Lüders); it does not state that the record is redundantly copied into many environment fragments. ⚠️ In CSD a record's objectivity is its being an *ontic selection in `Σ`*, not its redundancy — Darwinism's criterion is the operational twin of that fact, not its ground; state the redundancy theorem as a twin. Natural home: the composite arenas of P2 + `LudersMarginal` — a `k`-fragment redundancy statement for the pointer marginal | M | Medium–high — composites + marginals are in; the redundancy *statement* needs choosing carefully to avoid a placeholder | Medium–high — objectivity of records is a claim CSD makes in prose | Row, not urgent; scope only after A/C | > | **F — Hardy / Montina excess baggage** (an ontic model of an `N`-level system needs a continuum of ontic states). CSD's `Σ` is a continuum by construction, so the theorem is *satisfied* trivially — worth one glossary line + citation so a referee does not raise it | S | — | Low, referee-facing | Glossary line in the hardening session | > | ~~**G — Geometric (Berry) phase as fibre holonomy + Aharonov–Bohm twin.**~~ **SCOPED + the stateable half BUILT 2026-09-21** ([`berry-phase-scoping.md`](berry-phase-scoping.md); queue row 10). The `T²` fibre is the natural carrier of a holonomy, and AB is the flagship phase experiment with no twin. Blocked on infrastructure: Mathlib has no connection/holonomy API on principal bundles (Kähler chart yes, connection no), and the corpus's fibre is a *torus*, not a `U(1)`-bundle over configuration space | L | Low near-term (infrastructure) | Medium — a phase-holonomy twin would be a genuine CSD statement, but it is the same `R-016`-shaped wall (manifold API) from a different side | Scope doc before touching; do not start from Lean | > > **Checked and deliberately not added.** Identical particles / spin-statistics — already IP-1 > (now rowed: ~~#206~~, ~~#214~~, ~~#215~~, ~~#216~~ and ~~#217~~ done 2026-10-28, ~~#220~~–~~#223~~ operator dictionary and interactions done, ~~#219~~ mixed sectors done, remainder #218, #224). Cluster decomposition, Appelquist–Carazzone decoupling, EFT positivity bounds, > vacuum structure / Unruh — above the P3 ceiling or QFT-level; the scope ladder (QM → CV → > relativistic EFT) does not require infinite-dimensional QFT. Consistent histories — a competing > bookkeeping, not a missing theorem (settled non-goal). Page–Wootters relational time — the register > is a conserved *selector* coordinate (`IsJointLift.register_conserved`), not a clock (per-outcome > clocks were withdrawn, `CSD-CHARTER.md`); no theorem to state; off the reconstruction path. > Retrocausal / two-state-vector readings — interpretive, no theorem. > ### Triage 2026-08-20 — complexity × likelihood × value (post-pillars review) > > Taken after the EFT pillars arc closed (Q22): every open item below, rated. `Cx` is > effort; `P(success)` is a dated planning estimate of landing the item as stated (for > research rows, of landing a first bounded brick after scoping), not a promise; `Value` > is what it buys the programme. Ordered by recommendation. > > | Item | Cx | P(success) | Value | Verdict | > |---|---|---|---|---| > | ~~**Ledger sign-off session**~~ **LARGELY DONE — row corrected 2026-08-24** | S | — | — | ⚠️ **This row was stale and was over-selling.** S1 ran 2026-08-20 and S2 the same day; the ledger *itself* retracts the "28 one-sign-off-away" figure ("optimistic: it counted the by-design rows as promotable — a permanent scope qualification is not a missing signature"). Current state is **58 rows: 27 validated, 30 qualified, 1 needs-change**, not 9→37. **What actually remains is not a signature session**: ⚠️ *re-read 2026-09-01 — five of the seven cleared on 2026-08-24 and this row did not follow.* Of the named-artefact rows (CL-008, CL-014, CL-015, CL-024, CL-029, CL-030, CL-032), the TSV now records CL-008/CL-015/CL-029/CL-030/CL-032 as **CLEARED**; and the ledger grew to **68 rows 2026-09-01** (CL-062…CL-068, this session's headline theorems, all `qualified` with their scopes named); what is genuinely outstanding is **CL-024** (author hand-pass on the five Wigner tactic blocks, G11 — author-only) and **CL-014**, which rides it — several are "not recorded", i.e. assistant-doable rather than author-only. Plus ~18 `qualified`-by-design rows that are at their **correct terminal status** and must never be counted as promotable | > | **Q12 — the fibre mechanism (MD-1 dynamics)**, with the Paper D `H_int` successor row | Research; scoping doc **M**, first brick M–L after | Scoping: high. A full derivation of the fibre mechanism from de-isolation dynamics: low–medium near-term — it is the programme's genuine frontier | **Highest** — the reconstruction frontier itself | **SCOPED 2026-08-23** (`specs/q12-fibre-mechanism-scoping.md`), and the scoping RESHAPED the row. ⚠️ **W1, decisive**: (A)'s mixing hypothesis is **unsatisfiable by any flow the corpus defines** — the base action is a compact-group action (`UnitaryGroup.instCompactSpace`) and `kFlow` is a **T² shift**, so E6's `integral_mul_self_eq_of_recurrent` + `exists_le_pow_mem_of_compactSpace` kill correlation decay for both. Blocked by a THEOREM now, not an impression. W2: Mathlib has the exponential distribution + `HittingTime`, but **no** exponential law for competing hitting sets and **no** memoryless characterisation (Galves–Schmitt is not upstream and is not a brick). ★ **W3: `DeIsolationInteraction` has NO witness instance** — an unpopulated interface, the same defect E5 just fixed for E4, and plainly fixable from the landed `volume_bornCell`/`fibreOutcome`. **Split**: ~~Q12-a witness~~ **DONE 2026-08-23** (★★ `cdfDeIsolationInteraction` — `DeIsolationInteraction.born` now has a POPULATED antecedent; `cdfPointer` + `cdfPointer_preimage` + `fibreTypicality_compl_iUnion_bornCell`. ⚠️ Witnesses SATISFIABILITY ONLY: arbitrary outcome order, no dynamics carves the cells. ★ Process: four containment lemmas I began writing ALREADY EXISTED in `DeIsolationFlow.lean` — reverted the duplicates; **grep the neighbouring module before adding lemmas to a mature area**) → ~~Q12-b symmetric race~~ **DONE 2026-08-23** (`Mathlib/Probability/CompetingExponentials.lean`, Cat-1, 3 pins): ★★ `measure_raceCell` (`P(i wins) = bᵢ/Σb`) + ★★ `measure_raceCell_of_sum_eq_one` + `raceCell_pairwiseDisjoint`. ★ Shortcut worth keeping: `lintegral_exp_neg_expMeasure` evaluates NO improper integral — `e^{-St}` × the `Exp r` density IS a constant multiple of the `Exp (r+S)` density, whose mass is 1. ⚠️ **FINDING: the race does NOT fit `DeIsolationInteraction`**, whose pointer is `ℝ → Fin n` (**1-D fibre**) while the race needs `Fin (n+1) → ℝ`; §3b says the minimal fibre dim is `n−1`, so **the existing interface was committed to the ORDERED construction**. ~~generalise it~~ **DONE same day as Q12-b′**: `DeIsolationInteraction` now takes an ARBITRARY fibre `(F, ν)` ⇒ ★★ `raceDeIsolationInteraction`, a SECOND and **symmetric** witness on `Fin (n+1) → ℝ`. Shared machinery extracted at the second consumer (§9 rule of two): `Mathlib/MeasureTheory/CellPointer.lean` — `cellPointer` (disjoint cells → a TOTAL readout, leftover to a default) + ★ `measure_cellPointer_preimage` (leftover null when the weights exhaust the probability); Q12-a's bespoke `cdfPointer` deleted and re-derived from it. ⚠️ NEITHER witness is dynamical — no flow carves either family. Also: strictly positive rates only (a zero amplitude is outside `expMeasure`'s domain) → **Q12-c PROBED 2026-08-23, NOT STARTED — the gate fires**: the scoping note wrote "is the exponential law forced?" as one question; it is **two**, and §3c states the harder one. **(c1)** memorylessness ⇒ exponential is tractable (M) via `AddMonoidHom.toRealLinearMap`, modulo extending the survival function from `[0,∞)` to `ℝ` and **assuming** continuity (monotone ⇒ linear is NOT upstream). **(c2)** §3c's actual claim — *first-to-fire `= bᵢ` **iff** exponential* — has a forward direction that is an **integral equation in the unknown law** (`∫ G(x/b) dF(x) = b/(b+1)` for all `b`), solved classically by Laplace/Choquet–Deny: ~~**research-grade, not a brick**~~ ⚠️ **SUPERSEDED SAME DAY — route found** (`specs/q12c-exponential-characterisation-route.md`): the research verdict came from the **two-clock** condition only. ★ **With the `k`-clock family it collapses** — instantiate the race at rates `(1,c,…,c)` to get `E[G(cξ)^k] = 1/(1+kc)`, turning the integral equation into a **moment sequence on `[0,1]`** where moments determine the law; then `G(ξ) ~ U[0,1]` + monotone-equal-in-law give `G(ct) = G(t)^c`, and `t = 1` gives `G(c) = G(1)^c`. No Cauchy equation, no Laplace inversion. **Lean: L, fully mapped**, no missing upstream mathematics (Weierstrass + `ext_of_forall_integral_eq_of_IsFiniteMeasure` assemble the determinacy step); the cost is redoing Q12-b's product-measure setup for *general* iid clocks. ⚠️ **Caveat**: needs arbitrarily many clocks ⇒ what is forced is the exponential law *given that one clock law serves every n* — which is exactly the measurement-independence `sigma-fibre-contextuality.md` already commits to. Paper proof only; NOT a corpus theorem. ⚠️ **Correction to `record-layer-plan.md` §3c, of STATUS not truth**: the iff is classically true; what the plan understates is its cost — "the exponential fibre measure is FORCED" means forced in the literature, not in the corpus. (c1) would narrow the posit from "some memoryless law" to "the exponential" but would NOT establish §3c's claim, and sits one careless sentence from being read as if it had ⇒ **author decision, not a default** → **Q12-d: original still BLOCKED (foundations, W1) — but ROUTE 2 EXECUTED 2026-08-23**: `HasCorrelationDecayUpTo` (correlations bounded only on lags **below T**) + ★★ `blockPop_timeAverage_le_of_finiteHorizon`. ★ **E6 does NOT reach this**: it kills the ASYMPTOTIC antecedent for unitary flows (powers recur ⇒ correlations recur) but that argument needs the bound at *arbitrarily large* lags and says nothing over a bounded window ⇒ **E4's conclusion is not lost for finite-dim unitary Σ-dynamics; its asymptotic FORM is.** The weakening was nearly free — `hdec` was only ever used at `s,t ∈ range T`, so the proof had been finite-horizon all along and only the *hypothesis* was asymptotic. ⚠️ Still conditional and still not exhibited for any Σ-flow; what changed is that the hypothesis is no longer PROVABLY UNSATISFIABLE** (escape routes ranked; the recommended one is weakening mixing to FINITE-HORIZON decorrelation, which is the shape E4's quantitative envelope was already built for). → **Q12-c2 STEPS 1 + 2 DONE 2026-08-24** (`Mathlib/Probability/IidClockRace.lean`, Cat-1, 5 pins), joining the analytic half (steps 3/3′ + the `intervalMeasure` carrier) that landed the day before. ★ **Step 1 was NOT the fiddliest part, because of a change of framing**: `CompetingExponentials` gave each clock its own law `Exp bⱼ` and raced them unscaled, which is unusable when the law is the unknown; `scaledRaceCell` puts ONE iid law on every clock and carries the rate as a **scaling of the reading** (`ξⱼ / bⱼ`). Same `measurePreserving_piFinSuccAbove` split and `Measure.pi_pi` box as Q12-b, but **strictly cleaner** — the slice is a box at EVERY `t`, so the a.e. step of the exponential case disappears. ★★ `measure_scaledRaceCell` (kernel identity, no hypothesis on `μ`) ⇒ ★★ `HasRaceProperty.lintegral_measure_Ioi_pow` = the memo's `E[G(cξ)^k] = 1/(1+kc)` and ★★ `…_pow_mul_pow` = the mixed form `E[G(cξ)^p G(ξ)^k] = 1/(1+pc+k)` that step 3′ consumes. `hasRaceProperty_expMeasure` = non-vacuity (the `⇐` direction in the iid framing; the rate `r` cancels out of `r/(r + r·S/bᵢ)`). **`HasRaceProperty` quantifies over the number of clocks, so §4's caveat is now encoded in the Lean statement, not only in prose.** ★ **Step 2 came out STRONGER than the memo predicted**: `HasRaceProperty.map_survival` (`G(ξ) ~ U[0,1]`) needs **no regularity hypothesis on `μ` at all** — the memo's "assume `G` continuous and strictly decreasing" is unnecessary here, because the `c = 1` moments ARE the uniform moments and step 3 closes it. So **atomlessness is DERIVED from the race property, not assumed of it** (the `(k+1)`-clock race says the smallest of `k+1` iid readings is *strictly* smallest with probability `1/(k+1)`; ties would cost). Second time the `k`-clock family paid for a step the two-clock framing made look expensive. ★★★ **STEPS 3 + 4 LANDED THE SAME DAY ⇒ Q12-c2 COMPLETE, §3c IS A COROLLARY THEOREM**: `hasRaceProperty_iff_exists_expMeasure` — first-to-fire `= bᵢ` **iff** the waiting-time law is exponential. ★ **The memo's three mapped assemblies were ALL unnecessary, and so was the §5a successor I wrote mid-session** (§5a now records what actually happened). Steps 3/3′ compare `H_c(u) = G(c·G⁻¹(u))` with `u^c`, which needs the quantile `G⁻¹` as a CONTINUOUS function on a CLOSED interval — machinery Mathlib lacks. **Restricting the ratio to a NATURAL NUMBER `m` dissolves all of it**: `G(t)^m` is itself a product of `m` survival factors at rate 1, so every term of `∫ (G(mt) − G(t)^m)² dμ` is an instance of the SAME race family (`1/(1+2m)` three times, at rates `(1,m,m)`, `(1,m,1^m)`, `(1,1^{2m})`) and they cancel ⇒ `survival_natMul_ae`. The cross term is the whole difficulty and it is computable EXACTLY when the ratio is an integer. ★ **And the integers suffice because ANTITONICITY connects the lattices** (`raceRate_le`): the equation ties `G` together only along `{mt}`, but `mt ≤ nt'` forces `G(t)^m ≥ G(t')^n` ⇒ `m·λ(t)·t ≤ n·λ(t')·t'`; let `m/n` climb to `t'/t` ⇒ `λ(t) ≤ λ(t')`, symmetry ⇒ one `λ` everywhere. No density of the support anywhere. ★ Support in `(0,∞)` also came out DERIVED (the `m = 2` case: `t ≤ 0` ⇒ `2t ≤ t` ⇒ `G(t) ≤ G(2t) = G(t)²`, false on `(0,1)`), so §3's regularity list is now entirely consequences. Finish via `map_survival` (`t > s ↔ G t < e^{−λs}`) + `Measure.ext_of_Iic`. ★ **LESSON, the same one twice and worth carrying**: §6a already said *when a step looks like it needs a stronger determinacy theorem, check whether it needs the same theorem against a different object*; the stronger form is **check whether it needs a determinacy theorem at all** — ask what the family can already compute and shape the question to fit, rather than building machinery for the question as first posed. ⚠️ Residue: `eq_of_forall_integral_mul_pow_eq` + the `intervalMeasure` carrier are now UNCONSUMED (Cat-1, left in place; do NOT build `H_c` to justify them). ⚠️ **THE SECOND CONJUNCT IS NOT OPTIONAL**: `HasRaceProperty` quantifies over the NUMBER OF CLOCKS; at fixed `n` the family gives `n−1` moments and finitely many moments determine nothing ⇒ forced *given that one clock law serves every `n`*, the measurement-independence `sigma-fibre-contextuality.md` commits to. This is where that commitment earns its keep. ⚠️ **A POSIT REMOVED, NOT A MECHANISM SUPPLIED** — the fibre LAW is no longer a choice, but **no dynamics carves the race cells**; `Q12-d` stays blocked by `W1` and neither `DeIsolationInteraction` witness is dynamical. The reconstruction frontier has not moved. `record-layer-plan.md` §3c updated in place. → **W1 COMPLETED + Q12-w DONE 2026-08-24** (`Mathlib/Dynamics/CompactGroupNoMixing.lean` Cat-1 + `LF4/KahlerFlowNoMixing.lean`, 3 pins). ⚠️ **W1 was only HALF a theorem** and the row above overstated it: the unitary base action was proved, the `T²` fibre shift `kFlow` was asserted **by analogy** ("the identical argument applies"), and only `kProjectedFlow = id` was otherwise covered. Both gaps closed. ★★ `not_hasCorrelationDecay_of_compactGroup` — **for ANY flow `Ψ U` with `U` in a COMPACT GROUP, no observable of nonzero variance has a summable decay envelope**, given only that `V ↦ ∫ f(x)f(Ψ V x)` is continuous at `1`. So W1 is no longer an enumeration of the three flows the corpus happens to have today; it is structural and covers any compact-group flow added later — which is the form in which a wall is actually useful, because it says what an escape must AVOID. ★ Stated with a bare `Ψ` + `hpow : (Ψ U)^[n] = Ψ (U^n)`, **not a `MulAction`**, because the two consumers are a multiplicative group action and an **additive** torus shift; `to_additive` then makes one proof serve both (`exists_le_pow_mem_of_compactSpace` gained `@[to_additive]` ⇒ `exists_le_nsmul_mem_of_compactSpace`). `not_hasCorrelationDecay_blockPop_of_unitary` **refactored into a corollary**, not left as a parallel proof (§8.3b) — which is what earns the general lemma its second consumer. ★ The `kFlow` half was cheap because **the exact correlation is never needed**: shifting replaces the character `e(x)` by `e(v)e(x)`, so the observable moves by `Re((e(v)−1)e(x))` and the correlation by at most `‖e(v)−1‖` — a continuous modulus vanishing at 0, which is all `continuousAt_correlation_of_abs_sub_le_add` asks. Every Fubini argument dodged; only the VARIANCE needs the product structure. `fibreObs` reuses `MeasureTheory.circObs`, built as E5's mixing WITNESS for the doubling map — here the same function certifies the fibre flow CANNOT mix; the difference is entirely the map. ⚠️ **This STRENGTHENS the wall; it does not move the frontier.** Mixing systems exist and are exactly the ones that are not compact-group translations. Ruled out = deriving the race from **asymptotic** mixing of any dynamics the corpus currently has. **Q12-d route 2 (`HasCorrelationDecayUpTo`, finite horizon) is untouched and remains the recommended escape.** → ⚠️ **Q12-d ROUTE 2 CLOSED FOR `kFlow` 2026-08-24** (`LF4/KahlerFlowFiniteHorizon.lean`, 2 pins). Route 2 was the *recommended* escape from W1 — weaken mixing to FINITE-HORIZON decorrelation, since a real environment decorrelates on a timescale and a unitary flow on a large space can wander a long time before recurring. ★★ `exists_lag_le_envelope`: **the corpus's own fibre flow does not get that room.** For every `δ>0` there is a lag bound `n` such that on ANY horizon `T > n`, a finite-horizon envelope for `kFlow sh` must already exceed `1/2 − δ` at some lag in `[1,n]`. Mechanism = **quantitative recurrence**: Dirichlet (`AddCircle.exists_norm_nsmul_le`, in Mathlib) returns `j•sh` to within `1/(n+1)` of the identity at some `j ≤ n`, **for every shift**. ★ **What is uniform is the point**: `n` depends on `δ` ALONE — not on `sh`, not on `p₀`, **not on the horizon `T`**. Enlarging `T` buys nothing because the return already happened inside it, and the bound is blind to every parameter one might tune. Non-vacuous (`ε = 1` satisfies the hypothesis, so the conclusion is a real constraint); `exists_lag_envelope_ge_quarter` states it with a number. ⚠️ **Scope**: about `kFlow`, NOT finite-horizon decorrelation in general — route 2's ENGINE (`blockPop_timeAverage_le_of_finiteHorizon`) is untouched and still correct; what is ruled out is INSTANTIATING its antecedent on the Kähler fibre shift. **Net for Q12-d: with W1 general and route 2 closed on the only fibre flow the corpus has, Q12-d has NO route the present Σ-vocabulary can supply.** The open question is now a **Σ-vocabulary** question (what flow could CSD have?), not a proof-technique one — which is a sharper place to be stuck than "research". → ★ **Q12-d BRICK (i) DONE 2026-08-24** (`LF4/KahlerFibreMixing.lean`, 2 pins) — **and it corrects how W1 should be read.** ⚠️ W1 does NOT say "Σ cannot mix": it rules out flows whose iterates are **powers of an element of a COMPACT GROUP**. `kFlow` is one (translations = `T²` acting on itself); an **endomorphism** of the same torus is not (iterates `y ↦ 2ⁿy` are powers in a discrete, non-compact monoid). ★★ `torusDouble_hasCorrelationDecay`: the doubling map on the corpus's **own** fibre `KTorus` has correlations **exactly zero at every nonzero lag**. Same fibre, same observable as `not_hasCorrelationDecay_kFlow`, **opposite verdict** ⇒ **the obstruction is the CHOICE OF MAP, not the ontic space**; Σ untouched, and route 3 does not require leaving `T²`. Proof is free (observable reads one angle, map acts coordinatewise ⇒ collapses to E5's circle witness). ⚠️ Does NOT settle: (1) does not replace `kFlow` (that is the PHASE translation — a phase should translate; this is a candidate DE-ISOLATION map); (2) **not invertible** ⇒ not yet a physically admissible Σ-flow, the invertible case being a hyperbolic toral automorphism (cat map, same character argument, extra cost = Haar-invariance of a toral automorphism which Mathlib lacks); (3) ★ **MIXING IS NOT THE RACE** — Q12-d still needs first-passage **exponential at moment-map rates**, i.e. Galves–Schmitt/Abadi, which is `W2`'s research-grade item. **THAT, not Σ's vocabulary, is the real blocker.** Supersedes the "no route the present Σ-vocabulary can supply" line above. → ★★ **KAC LANDED + Q12-d RECOMMENDATION LOGGED 2026-08-24** (`Mathlib/Dynamics/Kac.lean` Cat-1 + `kac_doubling`, 3 pins). ✅ **RETIRED 2026-08-24 (author decision TAKEN): Q12-d's MIXING formulation is WITHDRAWN as mis-specified** — not blocked, not open; do not reopen in that form. ⚠️ **SUCCESSOR QUESTION CORRECTED 2026-08-26 — the version below is WITHDRAWN.** ~~Successor question: what structure supplies the INDEPENDENCE of the outcome clocks? (a question about Σ's STRUCTURE, not its dynamics)~~ Time in Σ is ONE shared parameter (`flow : ℝ → Σ → Σ`, P3), so the clocks are Σ-aligned deterministic readings rather than independent processes; and the proved Born construction never raced — `volume_circleCell` gives the weights as ARC LENGTHS of a partition of ONE uniform fibre coordinate. There is nothing to supply. **The successor question is `H_int`: what physical de-isolation interaction generates a READOUT whose level sets carry the Born measures? (⚠️ scope carefully -- there are TWO partitions. The Omega-regions on the projective BASE are EPISTEMIC, a calculation space (Paper C A7); nothing sculpts them and they match volume because mu_FS is forced by symmetry and the weights are the moment map. The FIBRE cells are a different object: BornFibrePartition.lean places them on the ONTIC fibre and calls the outcome the ontic selection of which cell the fibre point occupies -- while adding that the CDF family is ONE concrete realisation, the content being the interface plus the measure identity. So the cell SHAPES are bookkeeping; the fibre, the rates and the selection are not. The obligation is DeIsolationFlow.lean's: exhibit a pointer p = readout . flow(H_int(M)) whose basins are cdfCell(moment map).)** — and it IS about dynamics. Reasons (1) and (2) below still justify the retirement; reason (3) justified it too but the conclusion drawn from (3) did not follow. Keep *dynamical independence* (absent, unneeded) apart from *measure factorisation* (legitimate, what the Gumbel construction and Q10 use). See `CSD-CHARTER.md`.** Reasons it was withdrawn** — not because it is hard but because it cannot deliver. Three reasons, none about effort: **(1) regime mismatch** — Galves–Schmitt/Abadi is about **rare** sets (`μA → 0`); a Born partition has cells summing to 1, so the theorem's hypothesis is the OPPOSITE of §3c's regime and cannot be instantiated there; **(2) exact vs asymptotic** — Q12-c2 says exact Born ⟺ *exactly* exponential, so a limit theorem yields Born only asymptotically, **strictly weaker than `measure_raceCell` which already proves it exactly**; **(3) independence** — the race needs `n` INDEPENDENT clocks, but one deterministic trajectory is one process and hitting times of the `n` cells along one orbit are strongly dependent. ★ **The fibre is not scaffolding for work not yet done: it carries the independence base dynamics cannot supply.** ⇒ **Q12-d is not one theorem away, it is MIS-SPECIFIED**; the successor question is what supplies the independence. **Salvaged instead**: ★★ `tsum_measure_lt_returnTime` (Kac; absent from Mathlib, which has Poincaré recurrence but no return times). ★ Kac is **regime-correct** — no rarity hypothesis, any cell of positive measure — giving mean return `= 1/μA`, so a cell of Born weight `bᵢ` is returned to every `1/bᵢ` steps on average: **feature (B) of §3c, the RATES, derived not posited.** ★ Proof avoids the Kakutani tower entirely (telescoping on PREIMAGES, where measure preservation applies directly). ⚠️ `returnTime` is `ℕ∞`-valued and that is **not cosmetic** — with a ℕ junk value the bridge lemma is FALSE on the never-returning set. ⚠️ **Rates, NOT the law**; by reason (2) nothing regime-correct gives the law. Headline: **the frontier half is blocked by a theorem we own; the remaining half is smaller and more certain than the "research" rating suggests** → ★★ **SUCCESSOR QUESTION EXECUTED IN ITS HONEST FORM 2026-08-27** (`RecordLayer/ShearDeIsolation.lean`, 6 pins): the THIRD `DeIsolationInteraction` witness, the first whose pointer is the **readout of the constructed de-isolation propagator** — ★ `cellPointer_outcomeSector_eq_readout` (pointer = `(readout ∘ Φ_{0→1}).getD i₀` POINTWISE, the `p = readout ∘ flow(H_int(M))` shape as a theorem), ★★ `shear_sector_born` (the bankless mirror of `swap_sector_born`: `readyPrep [ψ] (Ωᵢ) = momentMap` weight, `measure_outcomeSector_eq_of_correlates` consumed) ⇒ `basin_rate` **discharged from the dynamics**, not assumed and not read off defined cells; ψ enters ONLY through the ontic preparation measure, readout arcs/basins/propagator all context-fixed. ⚠️ CARRIED NOT LAUNDERED: the propagator's Hamiltonian generation is stated-not-formalised (`ShearWitness` item 1, permanently-scoped symplectic row) and the coupling is engineered (witness, not derivation) — the surviving D1 residue is exactly that pair. ★ Process: the assembly was three-quarters landed and distributed (`swap_sector_born` / `shear_correlates` / `cellPointer`); what was missing was the bankless shear sector-Born + the interface bridge — Q12-a's grep-the-neighbouring-module lesson, one layer up | > | ~~**Q10 — no-signalling from primitives**~~ ★★ **Q10-a AND Q10-b DONE 2026-09-01** (`LF3/SettingLocality.lean`; ledger CL-064). The primitive `RemoteSettingLocalityA/B` — a remote setting change is *implemented by* a measure-preserving relabelling of `Σ` along which the local wing's reading is unchanged — and `operationalNoSignalling_of_settingLocality` derives the predicate from it. Escapes `no_product_partition_realises_singlet` because that no-go is about the **arity** of the response maps, which this primitive does not constrain; `translationLocality` + `translation_wingA_setting_dependent` prove non-vacuity (the readout provably moves with the remote setting). ⚠️ **Still open:** measurement independence is INHERITED not discharged, and whether CSD's own `Σ` satisfies the primitive is untouched. Q10-c (C1 §4.2 wording) is author-side. | — | — | — | Was: M–L, scoping-first. Scoping (`q10-no-signalling-scoping.md`) killed route (a) first — see the Q10-w box there — which is why (b) landed cleanly. | > | ~~**Q16 gate re-check** (CP of `e^{tℒ}`)~~ **DONE 2026-08-20** — verdict recorded in the Q16 row: the wall label was stale; the CP brick is an M–L in-corpus programme (drift part needs no Trotter at all; jump part CP by Choi series; assembly by de-skewing `trotter_skew`), medium–high likelihood | S (spent) | — | Medium | Q16's gate is now a queueable brick, not a wall | > | ~~**brick-2 — generation-from-records**~~ (`unitary-tpp-scoping.md` §7 residue) | M–L | — | — | **LANDED 2026-09-02** (`SigmaLayer/TensorTomography.lean`, `recordLocallyTomographic_iff_adjoin_eq_top`; NEXT STEPS item 2). Go given 2026-09-02; local tomography survives as boundary `R-017` — the row closes the *conversion*, not the posit | > | **Prose audit completion** (`prose-audit.md`, in progress) | M | High — the method has caught real defects twice | Medium–high — the honesty surface no guard reaches | Hardening session, soon | > | ~~**runLinter adoption**~~ (measured 2026-08-11: 122 findings, 4 linters) | M | — | — | **DONE 2026-09-04** — `simpNF` and `docBlame` to zero, `unusedArguments` hypothesis subset read and all deliberate, ratchet in CI against `docs/std-lint-baseline.txt` (CONVENTIONS §9.5). Found and closed an unguarded coverage precondition in `check-axiom-sweep` on the way | > | **`REFERENCES.json` + line-precise citations** (`CONVENTIONS.md §8.2`) | M | High | Medium–high — the row calls it the biggest auditability win | Schedule with the next release pass | > | **`_of_` hypothesis pattern** (`§8.3`) | M, incremental | High | Low–medium standalone | Opportunistic only — apply when touching a file anyway | > | ~~**Ecdsafail-split doc de-application**~~ **STRUCK 2026-09-07 (the note in §2 above: the count was never right, the pointers are provenance) and CLOSED 2026-09-23 with #17 (four application-flavoured headers rewritten)** (residue of the 2026-08-21 lift extraction: ~70 ECDLP/ECDSA docstring references remain, esp. the Cat-1 files `Reversible/{Cost,ModAdd,ModInv,ModMul}.lean` whose headers still point at `specs/ecdlp-resource-plan.md` — now at `Ecdsafail/specs/ecdsa/`; Mathlib module docs should describe the mathematics, not the downstream application. `Circuit.lean` done with the extraction) | S–M | High — mechanical doc pass | Low–medium — upstream-candidacy hygiene | Batch whenever next touching those files; not worth a session of its own | > | **General Boolean→amplitude lift** (`Reversible/Lift.lean`'s completion: per-constructor lemmas for `X`/`CX`/`swap` — each easier than the landed CCX case — plus the gate-list fold, a nontrivial induction) | M (sizing UNSETTLED — estimates oscillated across the 2026-08-21 audit session; scope fresh if picked up) | High mechanically | **Pure optionality — nothing in flight needs it** (verdict 2026-08-21, recorded in `Lift.lean`'s header: the Boolean frontier never touches amplitudes; the gadget strand is walled on the tensor factorisation, not the permutation lift; the QFT gap needs the converse direction on a non-permutation) | Build only when a consumer appears, or as a Mathlib submission in its own right | > | **MG arc — the Mathlib-gaps attack plan** (`specs/mathlib-gaps-plan.md`, triage 2026-08-22: every `MATHLIB-GAPS.md` row re-probed at the pin). ~~**MG-3** (S probe → L programme)~~ **PROBED AND LARGELY EXECUTED 2026-09-01.** The instance question was a SCOPE question: `Matrix.instCStarAlgebra` exists `scoped[Matrix.Norms.L2Operator]`, the same scope Track A already used. Upstream had meanwhile proved the concavity tier (`CFC.concaveOn_log`, `CFC.concaveOn_rpow`), so **L.2 and L.3a-interior landed as transports** (`operatorConcaveOn_log`, `matrix_rpow_concave`, `OperatorConvexBridge.lean`). ⚠️ **What remains of the ladder is only the summit** — Effros perspective / Lieb concavity → joint convexity → DPI, absent from Mathlib and the corpus both — so `hDPI` and unconditional SSA are **not** retired, and that summit is the XL. **MG-1** (M): the FS metric on `ℙ` via the projection embedding `dist p q = ‖P_p − P_q‖` — kills a ledger row in-corpus, unlocks Q28's ε-ball forms, prime upstream candidate. **MG-2** (M + S–M + M): `polynomial_zeroSet_null` (Fubini induction) + FS-as-a.c.-pushforward (hand-built radial density + the STAGED uniqueness theorem) + `segre_range_null` → ★★ "almost every composite state is entangled" (Q28 item 5 de-research-gated). **MG-4** (M, opportunistic): chart-level FS Kähler closedness via upstream flat `extDeriv`. **MG-5**: register tensor factorisation — ledger row added, attack gated on a consumer. Birkhoff + Lévy probed and verified standing (leave) | MG-3 probe S; MG-1 M; MG-2 M+S–M+M; each scoping-first at execution | MG-3 probe ~certain to settle the question either way; MG-1/MG-2 high | **High** — unconditional SSA is the largest single prize in the ledger; MG-2 is a C2 headline; MG-1 feeds the upstream batch | Recommended order: MG-3 probe → MG-1 → MG-2; MG-4/5 opportunistic/gated. **MG-3 EXECUTED + B.4 LANDED 2026-08-22** (five probe rounds: the wall = exactly two non-firing generic instances; with both shims the upstream monotonicity tier fires; B.4 = shim 2 + ℝ≥0-`cfcₙ` naturality + ★ `matrix_nnrpow_le_nnrpow` + `matrix_sqrt_le_sqrt` in the bridge, first-try compile, 3 pins; residue = L.3a interior CONCAVITY/Lieb, now provisioned by upstream's `Rpow/IntegralRepresentation.lean`). **MG-1 LANDED same day** (`Projectivization/Metric.lean`, 4 pins — the FIRST metric on `Projectivization` anywhere: `rankOneProj` embedding, ★ `isClosedEmbedding_toProjCLM`, `instMetricSpace` with topology definitionally the staged quotient topology, `dist_eq`; `CPN N` inherits automatically; ledger row CLOSED, joins the upstream batch; C2 ε-ball corollaries = recorded Q28 rider). **MG-2 LANDED same day** (`Projectivization/FubiniStudyLebesgue.lean` NEW + `EntangledMeasure.lean` extended, 6 pins): FS = projectivization of the normalized punctured-ball Lebesgue measure (by the staged uniqueness theorem) ⇒ ★★ `fsMeasure_null_of_cone`; the Segre cone lies in ONE coordinate quadratic's zero set (`segre_minor_eq`'s corner minor through the index bijection), null by Fubini slicing ⇒ ★★ `compositeFubiniStudy_range_segre_null` + ★★ `ae_not_mem_range_segre` — **almost every composite state is entangled** (Q28 item 5 de-research-gated; the general polynomial-zero-set lemma was never needed and is recorded as pure optionality). **MG-4 LANDED 2026-08-22** (`Mathlib/Analysis/InnerProductSpace/KahlerPotential.lean`, 4 pins): route correction — upstream has no wedge/∂∂̄ but DOES have `d² = 0`, so closedness comes from EXACTNESS: the `dd^c` construction (`complexStructureL`, `packL` via `ofSubsingletonLIE`, `dForm` anchored to `extDeriv_constOfIsEmpty`, `dcForm`, `ddcForm`) gives ★ `extDeriv_ddcForm` (every `dd^c` of a smooth potential is closed) and, with `contDiff_fsPotential` (`log(1+‖z‖²)`), ★★ `extDeriv_fsChartForm` — **the genuine non-constant FS chart form is closed**. Residue kept honest: the form is DEFINED by its potential (identification with `Kahler.fundamentalForm`'s pullback = a second-derivative computation, not attempted) and the manifold spelling stays Mathlib-blocked — the Kähler row is NARROWED, not closed. **MG-5 LANDED 2026-08-22** (gate opened by queue decision; `QuantumInfo/RegisterTensor.lean`, 4 pins, first-try compile): the row's premise had partly expired — Mathlib now has the inner product on `E ⊗ F` AND `OrthonormalBasis.tensorProduct`, so only the bridge to the concrete `EuclideanSpace` model was missing, and the change of basis between two ONBs indexed by `ι × κ` supplies it. Closed LITERALLY: ★★ `regTensorEquiv : QReg (a+b) ≃ₗᵢ[ℂ] QReg a ⊗[ℂ] QReg b` + `regTensorEquiv_basisState`, the reusable `prodTensorEquiv`, and the consumer-facing `tensorFirst` (operator on a wire block, identity elsewhere) with its basis-state rule. ⚠️ Delivers the API the gadget wall named, NOT the gadget hybrid — and `MeasurementAdder.lean`'s stale "Mathlib does not provide" wall note was corrected at source (the obstruction is the open n-fold amplitude argument), as was `Reversible/Lift.lean`'s optionality reasoning. **MG arc COMPLETE — every ledger row it targeted is closed or narrowed as recorded**; the one genuine remainder is the DPI ladder's Lieb/interior-concavity rung | > | **Q13 — trilemma exhaustiveness** | Research | Low–medium | Medium | Physics-first; not a Lean queue item until it has a scoping doc | > | **Track B — quantum relaxation** (Valentini) | Research/XL | Low in Lean near-term | High scientifically — the only route to new predictions | Physics-first; keep off the Lean queue | > | ~~**Q8 / KG-1..3 — manifold exterior calculus**~~ (blocks the A1 Kähler/Liouville half) **DONE IN-CORPUS 2026-09-07..12** (`Mathlib/Geometry/Manifold/`, steps (0)–(4) of `generator-layer-scoping.md`: `fsForm_mextDeriv`, `fsForm_isSymplectic`, `fsForm_isKahler`, `fsVolume_eq_smul_fsMeasure`, the Hamiltonian layer, Q29 Liouville; KG-3 reconciled 2026-09-14) | ~~XL~~ | — | — | Deprecate in favour of Mathlib when Mathlib lands the same (the standing rule for corpus-built API) | > | **TH-1 — Lévy / spherical isoperimetry** | L–XL | Low — Mathlib lacks the isoperimetry | Low–medium — optional strengthening; the mean is proved | Leave | > | **Base-only A7 at `N ≥ 3`** | Research | The natural (radial, ψ-only) class is answered in the negative by `sigma-fibre-contextuality.md`; what remains is class-generality of that verdict | Low — the fibred reading is canonical by author decision | Leave parked | > | **SSA / DPI** ⛔ **SCOPED AND CLOSED 2026-09-01** (`specs/lieb-dpi-scoping.md`); **bridge built and published 2026-09-12** (§6 of the note: Gate 0 passed, `csd-qit-bridge` exports unconditional SSA and the corpus's `hDPI` verbatim from physlib, foundational triple, corpus unchanged; public at https://github.com/zblore/csd-qit-bridge, both dependencies by pinned revision, own CI). ⚠️ The BCFJS `iff` is NOT unlocked by it and does not need it: row **BC** above (BC1–BC2 landed 2026-09-12, BC3–BC6 by an elementary route, L–XL) | — | — | — | The ladder's cheap half is finished: L.2/L.3a landed as transports, and **L.4 landed** (`OperatorConvexCFC.convexOn_mul_log` + `Matrix.matrix_mul_log_convex`, ledger CL-068), closing two TODOs named in Mathlib's own source. The summit (Effros/Lieb → joint convexity → DPI) is re-sized from the plan's "4–7 weeks" to **3–5 months** with zero Mathlib precedent, and the result already exists unconditionally in **physlib** on byte-identical pins. ★ **Recommendation: build neither.** `hDPI` stays an explicit hypothesis, which is its terminal status (CL-023 is qualified-BY-DESIGN). The prize is one `iff` in `NoBroadcasting.lean`. §3b of the scoping doc weighs the bridge as long-term architecture; the B6-upstreaming question is put back to the author there. | > | **Q14 (Poisson section), E-1/E-2 manuscript edits, glossary math review** | S–M each | — | — | User-side; the repo's half is done and pointed out | > | **Q9** | — | — | — | User-deferred; untouched by instruction | > | **Maximal scope (full GPT reconstruction); P5 constant-matching** | — | — | — | By explicit queue decision only; declared non-goals otherwise | > > **Recommendation in one line** (as written 2026-08-20): the sign-off session (user, S), then Q12's > scoping doc (the frontier), with the Q16 gate re-check (S) and the prose-audit/runLinter hardening > pair as the bounded interleave; brick-2 whenever the go/no-go is called. **Superseded 2026-09-02** > by the NEXT STEPS block above: Q12 executed in honest form (2026-08-27) and the `H_int` arc is > landed to its paper-side end state; brick-2 landed 2026-09-02 (`TensorTomography.lean`); the > live order (corrected 2026-09-04: the Q16 CP brick landed 2026-08-20 and was listed here in > error) is hardening session → LF / Ozawa twins **[both landed 2026-09-04]** → WAY brick 2 (WAY bricks 0–1 landed > 2026-09-02). > > ### Q. The numbered quick queue (added 2026-08-12 — say "do Qn" to reference) > > | # | Item | Pillar | Cx | Status | Row | > |---|---|---|---|---|---| > | ~~Q1~~ | ~~Derived-ε ontic bridge~~ **DONE 2026-08-12** (`Empirical/CSD/QuantumChaos/DerivedCoupling.lean`, 4 pins): ★★ `deficitKick_record_halfLife` — `δ·μ((intact n)ᶜ) ≤ n·‖W−1‖`, the §H5 half-life with `ε` DERIVED; route = the overlap-deficit trigger (`1 − Re⟨u,Wu⟩/‖u‖²`, ray-well-defined, continuous by `momentMap`-style quotient descent, pointwise `≤ ‖W−1‖` by Cauchy–Schwarz) + **Markov over μ_FS** (`measure_deficitTrigger_le`) + the generic `triggeredRecordKick`; `W = 1` ⟹ a.s. persistence (`deficitKick_persists_of_id`). Same `‖W−1‖` knob as `carrier_persistence_window` — one quantity, both levels. Recorded refinements, not claimed: the threshold-free sharp version (needs the deficit's law); the `FieldConfig` reindex | Chaos/records | M | **done** | H7 residue | > | ~~Q2~~ | ~~Half-life attainment~~ **DONE 2026-08-12** — in two halves, one of which predated the row: (i) **generic attainment was already settled** (`HalfLifeAttainment.lean`, `cyclicKick_halfLife_attained` — the bound is an EQUALITY on the cyclic kick; the Q2 row description was stale on this); (ii) **the derived coupling bites, exactly** (`DerivedCoupling.lean` bite section, 4 pins): for the qubit phase flip `diag(−1,1)` the deficit **is** `2·momentMap` (`overlapDeficit_phaseFlipW`), so the trigger is a moment super-level set and the **Duistermaat–Heckman law computes the coupling exactly** — `μ_FS(trigger) = 1 − δ/2` (`measure_deficitTrigger_phaseFlipW`), strictly in `(0,1)` on `δ ∈ (0,2)` (`deficitKick_phaseFlip_bites`), with the half-life at the exact rate (`deficitKick_phaseFlip_halfLife`). Honest note recorded in-module: `‖W−1‖ = 2` makes the Q1 Markov bound trivial on this window — the generic bridge is for drives whose deficit law is unknown; where the law is available (DH), use it | Chaos | M | **done** | §H thread | > | ~~Q3~~ | ~~Chaos diagnostics: SFF growth / OTOC beyond the gate~~ **DONE 2026-08-12** (`CV/ChaosBounds.lean` Q3 section, 3 pins): the pair that completes the diagnostics beyond the light-cone gate. (i) ★★ **Slow scrambling** (`otoc_interactingU_le`): for disjointly supported `A`/`B`, `otoc(interactingU, A, B, n) ≤ 4n·|τ||λ|C·‖A‖‖B‖` — the free comparison commutes exactly (`commute_of_disjointSupport` + free-pow support), so the whole OTOC is priced by the telescoped Duhamel distance (`norm_unitary_pow_sub_pow_le` + `interactingU_dist_le`): **linear growth cap, no fast scrambling in the kicked-diagonal family**. (ii) ★ **Exact revival** (`sff_freeFieldU_revival`): at `τ = 2π` the integer-spaced spectrum (`oscEnergy n = n + ½`) makes every phase `e^{−iπK}` (`freeField_phase_two_pi`), so `sff = 1` at **every** period — the clean integrable baseline (no dip–ramp–plateau) the cap is measured against. Not claimed: any lower bound on OTOC growth (would need a specific drive computed out, like Q2's DH route) | Chaos | M | **done** | §H thread | > | ~~Q4~~ | ~~Entropy-ledger link~~ **DONE 2026-08-12** (`Empirical/CSD/QuantumChaos/EntropyLedger.lean`, 7 pins): the ledger made precise and priced, strictly as retrodiction reliability (every quantity a present-tense measure of a present-tense set). Chain: ★ `measure_retrodictionSuccess_compl_le` (reading the register now retrodicts the formation value except on `≤ n·ε`); ★ `erosionFraction_monotone` (**the ledger's substrate is one-way** — second-law shape at the register, from carrier antitonicity alone); ★★ `ledgerEntropy_le` (below half-filling, the two-cell coarse entropy `binEntropy(erosion) ≤ binEntropy(n·ε)` — **one coupling knob, three readings**: reliability, erosion, entropy); ★ `vonNeumannEntropy_ledgerState` (the measure-side ledger **is** `S(diag(1−e, e))`, the object TH2's pinch H-theorem governs — the link the row asked for). Instantiation with **no free parameter**: phase flip + Q2's DH coupling `ε = 1−δ/2` (`deficitKick_phaseFlip_reliability` / `_ledger`) — closes the Q1→Q2→Q4 chain. Fine-grained entropy stays constant throughout (measure preservation); all production is coarse. Not claimed: Fano-type converse (needs joint-distribution conditional entropy); any thermodynamic entropy beyond the register's own coarse-graining | Chaos/records | M–L | **done** | H7 follow-up | > | ~~Q5~~ | ~~Shor-9 concatenation + Lindblad tier~~ **DONE 2026-08-13**, both halves. **(i) Shor-9 by concatenation** (`Empirical/QM/QEC/ShorNine.lean`, 9 pins): the E1 row's combinator option delivered — ★ `tel9_bkron_vkron` (`(A⊗B⊗C)(u⊗v⊗w) = Au⊗Bv⊗Cw`) reduces every 512-dim statement to the proven 8-dim block facts; **nothing above 8 dims is entry-wise**. All eight stabilisers fix `shorLogical`; syndrome tables ★ `innerStab_syndrome_X` (3-qubit pattern in the hit block, injective) + ★ `outerStab_syndrome_Z` (block pattern, **independent of the inner position**); corrections ★★ `shor_corrects_X` / `shor_corrects_Z_degenerate` (**recovery `Z₁` of the identified block works whichever qubit was hit** — the code's degeneracy as a theorem) / `shor_corrects_XZ`, all 9 positions; with `pauli_decomposition` this is the discretised "corrects an arbitrary single-qubit error". Not claimed: two-error patterns, 9-qubit syndrome collapse, fault tolerance. **(ii) Lindblad semigroup tier** (`LF6/LindbladSemigroup.lean`, 6 pins): `Φₜ = exp(tℒ)` for an **arbitrary** GKSL generator in the ℝ-endomorphism algebra — ★ semigroup law, ★★ the master equation `d/dt Φₜρ = ℒ(Φₜρ)` (`hasDerivAt_exp_smul_const'`), ★ trace preservation and ★ Hermiticity preservation at the flow level (series arguments: `HasSum.mapL` / `HasSum.star`, termwise from the generator tier). **CP of `e^{tℒ}` remains the recorded Mathlib-scale gap** (Lie–Trotter), unchanged on the GKSL row. Proof-engineering: multiple module-system instance-defeq walls (`Complex.mulAux` unexposed) — worked around via `module` tactic, pointwise `ext`, induction-not-`smul_pow`, and the `exp_add_of_commute_of_mem_ball` (𝕂 := ℝ) route; recorded in the module docstring + memory | QI ladders | L | **done** | E1 | > | Q6 | Composable finite-key QKD — **deprioritised (author note 2026-08-13): this is core-QM QIT reproduction (Lean-QIT's territory, not CSD's floor); the existing E91/BB84 twins already show the reconstruction carries crypto. Pick up only if a CSD-specific angle appears** | QI ladders | XL | open, deprioritised | E3 | > | Q7 | CV programme, post-Stage-5 — **Stage 6 OPENED 2026-08-13** ([`eft-stage6-plan.md`](eft-stage6-plan.md), the cone's consequences; continuum + RG postures unchanged): ~~CV-20~~ ★★★ **the Lieb–Robinson velocity DONE same day** (`norm_commutator_velocity_le`: outside `v·t ≤ d`, `v = 2e²‖S‖`, the commutator ≤ `2‖A‖‖B‖e^{−d}` — Stage 5's recorded non-claim discharged; constant not claimed optimal); ~~CV-21~~ ★★ **vacuum clustering DONE same day** (`diag_entry_mul_of_disjointSupport` + `vacuum_clustering`: no vacuum correlations across disjoint mode sets — the statics companion to the cone, landed stronger than planned at any configuration); ~~CV-22~~ **Wick four-point: gate run same day, PASSED, landed pattern-resolved** (`eqFourPoint` table: `3/4` all-equal needing `2 < N`, `1/4` two-pairs, `0` singleton patterns — exactly `Σ_pairings ∏ ½δ`). **Stage 6 COMPLETE 2026-08-13, opened and closed same day**; further CV stages are new queue decisions, not part of this row | CV/EFT | L | **done (Stage 6 complete)** | E4 | > | ~~Q8~~ | ~~Manifold forms on `ℂℙ^{N-1}` (A4 manifold half)~~ **DONE 2026-09-07..12** (`fsForm`, `fsForm_isSymplectic`, `fsForm_isKahler`; the A4 manifold half is `schrodingerField_isHamiltonianVectorField`, and on the arena `hamiltonianFlow_sectorEnergy_schrodinger`) | Geometry | ~~XL~~ | ~~**Mathlib wall**~~ built in-corpus | A4 | > | Q9 | DPI → SSA ladder | QI ladders | L | **user-deferred 2026-07-23 — leave alone** | E2 | > | Q10 | Non-factorising-Σ no-signalling from primitives | Foundations | Research | scoping-first | C1 item 16 | > | ~~Q11~~ | ~~U(N) / hTPP discharge — scoping session~~ **SCOPED 2026-08-13** ([`unitary-tpp-scoping.md`](unitary-tpp-scoping.md)) — the deliverable (the map + a named first brick) is delivered; no theorems claimed. **The map:** Hardy **survives** as the minimal-scope programme (probabilities = LF1's LLN already a theorem; continuous reversibility = Σ's native flow, absorbed by the substrate; composites = `compositeAlgReconstruction`, landed; simplicity/subspaces non-questions at this scope). CDP's grammar **declined** (ontology-free circuit framework; purification is theorem-shaped in CSD — the dilation machinery runs the other direction; harvest = local discriminability, already `TensorReconstruction`'s generation hypothesis). MM's **requirement style survives, the Lie-theoretic lever deferred** — at minimal scope `wigner_rigidity_unitaryGroup` replaces the group classification outright; MGM 2019 not relied on (Kent-disputed). **The reduction (§4):** both conditioners are ONE record-level premise — `transProb` is a record observable (the Born rate any context containing `q` assigns to `p`, via `bornRateBasis_eq_inner_sq`), so `hTPP` converts to "the projected flow preserves record statistics" and the measure posit to "invariant under every statistics-preserving symmetry", `U(N)` never named in a premise (the group becomes a Wigner OUTPUT). Premise conversion in the `fubiniStudy_forced_by_symmetry` template, NOT elimination (two operational premises survive; the papers owe their motivation); NOT the §13.2 trap (statistics-preservation is not measure-preservation; the `UnitarySelection` guard stands); D1 (`G`-from-dynamics) untouched and still obstructed (`obsFlow_not_uniquely_ergodic`) — the operational route sidesteps it. Maximal scope (full GPT reconstruction) named and deferred as a future queue decision. **First brick → Q18** | Necessity | Research | **done (scoping)** | necessity audit | > | Q12 | Fibre mechanism from de-isolation dynamics | Σ structure | Research | open | E5 | > | Q13 | Trilemma exhaustiveness over all arenas | Measurement | Research | open | E5 | > | Q14 | Poisson-argument paper section | Paper (authorial) | — | unblocked since B1 | A2 | > | ~~Q15~~ | ~~SigmaLayer role split~~ **DONE 2026-08-13** (the consolidation moment G10 named: the record-layer tranche is stable and the M/L brick queue is empty). The split is exactly the namespace/directory mismatch, measured mechanically: of 115 files in `SigmaLayer/`, **74 declared `namespace CSD.RecordLayer`** and moved to `CsdLean4/RecordLayer/`; 41 genuine `CSD.SigmaLayer` files remain. **Namespaces unchanged, so all 1,700+ axiom pins and the Headlines facade drift-guard survive untouched** — only module paths moved. 85 reference surfaces rewritten (root, `Headlines.lean`, all importers, `check-claims.sh` scope ledgers, `docs.yml` smoke page, `docs/TOUR.md`, `docs/PATHS.md`). Policy per G10's spirit: **historical BACKLOG/plan rows keep pre-split paths** (they are construction records; the archaeology stays true to its date). Residues, recorded not queued: the `Tests/AxiomAudit/SigmaLayer.lean` part still carries both namespaces' pins (G9's classifier keys on namespace, placement never affects resolution — a `RecordLayer` part split is cosmetic); G10(b)/(c) declines reaffirmed | Infra | L | **done** | G10 | > | Q16 | Metrology A4 (decoherence) | Empirical | XL (gate re-rated: the CP brick itself is M–L) | half-unblocked by Q5 (semigroup tier). **Gate re-checked 2026-08-20 (triage item): the "Mathlib wall" label on CP-of-`e^{tℒ}` is stale.** In-corpus route: (i) split `ℒ = ℒ₀ + J` with `ℒ₀ρ = Gρ + ρG†`, `G = −iH − ½ΣL†L` — left/right multiplications commute, so `exp_add_of_commute` gives `e^{tℒ₀}ρ = e^{tG}ρe^{tG†}`, CP by construction, no Trotter needed; (ii) `e^{tJ}` is a positive series of Kraus conjugations, CP via `choi_iff_posSemidef` + PSD-closed-under-limits (finite dim); (iii) assemble by Lie–Trotter for GENERAL matrices — `trotter_skew`'s estimate machinery already carries `exp(‖A‖+‖B‖)` factors, so dropping the skew hypothesis is a bounded generalisation; products/limits of CP are CP (Kraus composition + Choi limits). Verdict: the semigroup-CP brick is an **M–L in-corpus programme, medium–high likelihood**, not an upstream wall; the metrology tier above it keeps its XL rating. **THE BRICK LANDED SAME DAY 2026-08-20** (`Mathlib/Analysis/NormedSpace/TrotterGeneral.lean` + `LF6/LindbladPositivity.lean`, 3 pins; scoped in [`cp-semigroup-plan.md`](cp-semigroup-plan.md)): ★★ `NormedSpace.trotter_product` (the de-skewed Banach-algebra Lie–Trotter formula, rate `n⁻¹s²(3+s)e^{2s}`), ★★ `lindbladSemigroup_posSemidef` (every GKSL flow with Hermitian `H` is a positive map at every `t ≥ 0` — drift = conjugation with no Trotter, jump = positive Kraus series, PSD closed under limits), ★ `lindbladSemigroup_amplified_posSemidef` (positivity survives every ancilla amplification of the generator — the CP shape). Named remainder: the `id ⊗ Φ` identification (declared in the module and the wait ledger). `LindbladSemigroup.lean`'s stale wall note superseded at source; CL-044's qualification narrowed in the TSV | metrology plan | > | ~~Q17~~ | ~~Headline-census refresh~~ **DONE 2026-08-13** — all four asks delivered. **(i) Admission criteria defined** (`VALIDATION-LEDGER.md`, new section): terminal + claim-bearing + pinned + distinct, with the census sources named — "the 30" is no longer a codex-session artifact but a rule, re-run when a tranche closes with starred headliners. **(ii) Census run, ledger extended 31 → 51** (CL-032…CL-051, all `qualified` with load-bearing posits named; promotion stays sign-off): the necessity audit's named strongest-direction omissions (`stone_continuous`, `no_exact_finite_ccr`, the three `MeasurementConstraints` no-gos, `compositeAlgReconstruction`, `posMeasure_noRecord_pointer`), the A7-faithful `qubitBorn`, the 08-12/13 tranche's starred headliners (derived coupling, DH-exact rate, entropy ledger, Shor-9 degeneracy, the Lindblad master equation, LR velocity, vacuum clustering, the general-`N` third horn), the Q18 conversions, and the two boundary-discharging record-layer ★★ (`povm_sector_born` vs CL-026, `pointer_luders_born_prep`). Headlines facade + drift guard extended to 51 (all constants elaborate; `check-validation-ledger.sh` OK at 51). **(iii) Necessity classification re-run** (`necessity-audit.md`, addendum 2026-08-13): CL-031 (which the 08-09 audit predated and had never classified) + the twenty new rows, same rubric; **updated tally 2 necessity / 13 conditional necessity / 9 impossibility / 24 sufficiency / 3 instantiation** — the body's "strongest-direction results are mostly not among the declared headlines" is discharged. **(iv) The Q18 ledger stroke landed in the same addendum**: CL-012 and the Schrödinger-chain rows re-annotated — posits carry a proven conversion to record-level premises (conversion, not elimination; §13.2 trap and D1 untouched) | Necessity/infra | M | **done** | validation ledger + necessity audit | > | ~~Q18~~ | ~~The Q11 first brick~~ **DONE 2026-08-13, same day as its scoping** (`RecordLayer/StatisticsRigidity.lean`, 8 pins) — all seven scoped items landed, none weakened: `exists_context_extending_ray` (every ray is an outcome of some context, Gram–Schmidt extension), `recordKernel` + `recordKernel_well_defined` (the operational pairwise statistic, defined through `bornRateBasis`, apparatus-independent, inner-product-free), ★★ `recordKernel_eq_transProb` (**transition probabilities ARE record observables** — the kernel identification), ★ `recordStatisticsPreserving_iff_transProbPreserving` (the operational and geometric premises are THE SAME predicate), `recordStatisticsPreserving_unitary`/`_conjProj` (both branches inhabited), `recordStatisticsPreserving_realisation` (the operational symmetry group is semi-unitary, Wigner through the iff), ★ `projectedFlow_unitary_of_record_statistics` (W3+Bargmann consuming the record-level premise in place of `hTPP` — a premise conversion, honestly a thin wrapper), ★★ `measure_eq_fubiniStudy_of_record_statistics_invariant` (**`U(N)` in the proof, never the statement** — the necessity audit's requested conversion). Scope-noted as predicted: the FS-invariance converse needs `conjProj` pushforward invariance (absent) and the discharge direction does not need it. *Snag ledger:* the forecast `rep`/`mk` defeq class never bit (the `show`-idiom sufficed); the two actual snags were **name drift** — `Complex.norm_ofReal` gone (superseded by the exact-shape simp lemma `norm_smul_inv_norm`) and `Set.restrict` deprecated for `Set.domRestrict` in the PiL2 extension lemma's signature. Ledger stroke deferred to Q17 as scoped: re-annotate CL-012 + the Schrödinger-chain rows (posit renamed, not removed), admit the two ★★ | Necessity | M | **done** | Q11 scoping | > | ~~Q19~~ | ~~C1 positive witness + contextual capstone~~ **DONE 2026-08-13, same day as its review** (reviewer handoff amended per the session review, then landed; `LF6/C1BellConsistency.lean` amended in place — no parallel capstone file — + `Tests/Witnesses/SingletBell.lean` WS-H upgrade; 5 production pins in the Dynamics part + 1 witness pin). **The gap was real**: CL-031's reproduction slot had NO inhabitant anywhere (the recorded non-vacuity covered only the compatibility conjunct), so the corpus proved "non-contextual can't do the singlet" without proving "contextual can" — the second half being CSD's own claim. **Landed**: `ReproducesSingletTableAt` (the full-table predicate — the review's one substantive correction: a correlation-only witness can carry degenerate marginals, so the positive model MUST be the table, the E-1 lesson applied in advance); the explicit model (per context, the first torus coordinate of `(KSigma 4, kMuPsi)` read through the four cumulative `circleCell` arcs of the context's own `P_st` weights — `kMuPsi = δ ⊗ vol_T²` makes the arc masses exact, no atomless abstraction); ★ `singletContextualModel_table` (**the full table at EVERY context**, marginals ½, correlations −a·b) + `integral_wing_mul_of_table` (table ⇒ correlation — `correlation` IS the weighted table sum by definition); ★ `singletContextualModel_reproduces` (the first inhabitant of `ReproducesSingletAtCHSH`); ★★ `c1_singlet_contextual_capstone` (**existence ∧ obstruction**: a measurable shared-context family reproduces the singlet and no global CHSH assignment is compatible with it — the C1 separation two-sided, both halves witnessed); WS-H re-export `kMuPsi_singlet_contextual_model`. Existing names untouched (`kMuPsi_no_global_chsh_assignment` etc. — pins stand); CL-031 retained as the general obstruction. Docs: `C1-FORMAL-SUPPORT.md` + validation-hardening WS-H entry amended. *Snags:* `Finset.sum_eq_single` side-goal ordering (inline the side proofs); the bare-negation `∉ setOf` unifier trap (ascribe the membership type); `integral_indicator_const` now lands in the `Measure.real` API; `Measure.volume_eq_prod` takes explicit type args. **CL-052 ADMITTED (author sign-off 2026-08-13, same day)**: `c1_singlet_contextual_capstone` is ledger row CL-052 (qualified; facade + drift guard extended to 52; classification INSTANTIATION carrying CL-031's impossibility on the witness — necessity-audit addendum). CL-031 retained; the two rows read together, existence + no-go — the citation pair for the revised C1 paper | Entanglement/C1 | S/M | **done** (CL-052 admitted) | C1 stack review | > | ~~Q20~~ | ~~C1 §4/§5.1 discharge on the exhibited model~~ **DONE 2026-08-13/14** — ⚠️ *provenance: written in a separate (remote-control) session and left uncommitted at HEAD `ecf0c86`; recovered 2026-08-14, verified build-clean (0 errors, 0 warnings, 4009 jobs) + pins + guards, then committed — the exact K1-precedent failure mode (`INDEX.md` K1 row), caught by a status check.* Content: **(i)** `LF3.OperationalNoSignalling` made a real predicate stack (`RemoteMarginalInvariantA/B` — one `μ` across contexts, which IS measurement independence, now named in-module rather than implicit) and **inhabited for the first time** by ★ `singletContextualModel_no_signalling` — until Q20 the paper's §4 cited kernel-level finite sums with no measure or outcome map in them; general route `wingA_marginal_of_table` / `wingB_marginal_of_table` (marginals from the table, any family). **(ii)** `no_product_partition_realises_singlet`'s reproduction hypothesis `ReproducesSinglet` also had no inhabitant; ★ `singletContextualModel_not_product` discharges it on the exhibited model — **C1 §5.1 as a theorem about an exhibited object** (the model's wings provably cannot be setting-local). **(iii)** the Λ → `SigmaSpace` rename across the C1 chain (prose-alignment with the papers' Σ vocabulary). **(iv)** honesty additions to `C1-FORMAL-SUPPORT.md`: the outcome family is a **chosen partition, not a readout** (deriving it from de-isolation dynamics is MD-1, open); a dilated system-plus-pointer version was written and REMOVED same day (it closed the gap only by leaving Σ — wrong direction); measurement independence named as a genuine Bell premise. 2 pins (Dynamics part). Ledger note: neither theorem admitted as a row (author's tranche made no tsv change — respected); both are census candidates at the next Q17-criteria sweep | Entanglement/C1 | M | **done** (recovered + committed) | C1 paper §4/§5.1 | > | Q21 | **EFT Stage 7 — SCOPED 2026-08-17** ([`eft-stage7-plan.md`](eft-stage7-plan.md), the Q11 mold: gates and abort criteria agreed in advance, feasibility checked before scoping — `gibbsState` index-generic ✓, finite-dim KMS is entrywise algebra for diagonal `H` ✓, Thermo→CV import direction clean ✓, walk-collapse threshold pattern generalises ✓). Three rows: **CV-23** the Wick closure — **(a) DONE 2026-08-17** (★★ `eqFourPoint_wick`: the packaged pairing-sum formula `¼(δδ+δδ+δδ)` over every mode pattern at `2 < N`, the table strengthened in place per §8.3b, 1 pin); **(b) DONE 2026-08-17** (`CV/Wick.lean`, 4 pins, the D3b clean-context session executed the plan's notes with zero re-derivation: `twoPointKernel` in two-factor form, ★ `timeTwoPoint_eq` = `δ_{kl}·K(n,m)`, the distinct-times coincidence table where arrangement picks the **time** pairing, ★ `timeFourPoint_same` via the cross-pairing exponent identity, ★★ `timeFourPoint_wick` — the pairing sum over every mode pattern at `2 < N`; recoveries to `freeTwoPoint`/`eqFourPoint`); **(c) gate PASSED + CLOSURE DONE 2026-08-17** (`CV/Wick.lean`, 6 pins: gate ★ `modeOpQ_six_vac` `15/8` + `modeOpQ_four_two_vac` `3/8`; closure ★★ `Q_pow_two_mul_vac`/`modeOpQ_pow_two_mul_vac` — the moment ladder `⟨Q^{2n}⟩ = (2n−1)‼(½)ⁿ` exactly for `n < N`, commutator recursion vs `truncated_ccr`, defect killed by the walk band; odd moments by parity; grouped factorisation `modeOpQ_pow_mul_pow_vac`; new `modeOp_mul`/`modeOp_pow` API — **CV-23 COMPLETE**); ~~CV-24~~ the thermal tier — **DONE 2026-08-17, first row landed** (`CV/ThermalPropagator.lean`, 7 pins: ★★ `thermal_kms` **exact at the cutoff** — the corpus's first KMS statement, the Thermo↔CV verticals joined; closed-form Boltzmann state via `cfc_eq_conj_diagonal` at `U = 1`, `Z = z^K`, the thermal mode marginal, the thermal propagator with the truncation edge explicit, and the `β → ∞` vacuum limit recovering `freeTwoPoint` — no residue; full record in the `future-work.md` strike); ~~CV-25~~ **SCOPING COMPLETE 2026-08-18** ([`channel-rg-scoping.md`](channel-rg-scoping.md): the gate's criterion MET — provable budget `ε = 2n\|τ\|\|λ\|C_v` in trace distance via DPI + the price ladder; route = mode tracing (`traceOutChannel`∘`modeSplit`); level decimation recorded-not-selected pending a leakage estimate; first brick **CR-1** (the `traceDist`↔opNorm bridge) named; the arc = CV-26). ~~CV-26~~ **THE ARC EXECUTED 2026-08-18, same day** (`Mathlib/QuantumInfo/UnitaryPerturbation.lean` + `CV/ChannelRG.lean`, 7 pins: ★★ **CR-1** `traceDist_conj_sub_le` `D(UρU†,VρV†) ≤ 2‖U−V‖` uniform in the state — the norm bridge, Cat-1; ★ **CR-2** `coarseChannel_free_intertwine` **exact** — the traced mode's phase meets its own conjugate and cancels, so zero coupling ⇒ zero error and the whole budget is the interaction's price; ★★ **CR-3** `channelRG_dist_le` `D(C(U^nρU^{n†}), U_eff^n C(ρ)U_eff^{n†}) ≤ 2n\|τ\|\|λ\|C` — **the channel-level statement the Stage-4 no-go demanded**, coarse-graining CPTP via the Stinespring mode-split isometry. Scoping doc executed as written, both wall-checked links discharged. One step, NOT a flow: no iteration/fixed point/beta function). Non-goals restated: no continuum limit, no RG theorems, no velocity optimality; the departures tower stays theory-gated. **Labelled breadth per the charter** (the isolated-piece image, not the reconstruction frontier). **Q21 / STAGE 7 COMPLETE 2026-08-18** — all three rows closed | CV/EFT | (a) S, (b) M, (c) L-gated; M/L; Research-gated | **DONE 2026-08-18** | Q7 residues + Thermo↔CV | > | ~~Q22~~ | ~~**The EFT pillars arc** ([`eft-pillars-plan.md`](eft-pillars-plan.md))~~ **DONE 2026-08-20 — every pillar resolved; row added retroactively at the P5 close** (the arc had been tracked only on the pillars page, which the canonical-list discipline does not allow; this row is the correction). **P1 CLOSED in three arcs 2026-08-19/20**: the arena bridge (`CV/ArenaBridge.lean` — ★★ `arena_lightcone`, the 2026-08-10 "unstatable" Lieb-Robinson-at-the-record-layer theorem, via CR-1's Hölder-lite as the category interface); the definitional layer (`CV/FieldStructuredFlow.lean` — ★★ `FieldStructuredFlow.lightcone`, EVERY field-structured flow has the arena cone, witnesses = the corpus's own `freeFieldU`/`interactingU`); the fibre-active extension (`CV/FibredArenaBridge.lean` — ★★ `record_lightcone`, the record cell in the `T²` fibre cannot be steered from outside the cone; the `ShearWitness` skew stroke realised through the bridge). **P2 CLOSED 2026-08-20** (`CV/CompositeArena.lean` — composite = mode concatenation, Segre `arenaJoin`, mode-local subalgebras `SupportedOn` their blocks so P1 machinery is free; ★★ `composite_no_signalling` exact for ALL states, ★★ `bell_not_join` (`⊗` vs `×` at the arena), ★★ `compositeArenaForced` — the landed `compositeAlgReconstruction` CONSUMED at the arena's own algebras). **P4 CLOSED 2026-08-20** (`CV/DispersionEarned.lean` — the necessity-audit's "covariance is not shown to select it" superseded: ★ `cone_preserving_is_boost` derives the boost group from ray preservation + unimodularity, ★★ `cone_symmetry_characterises_omega` the iff `ω = √(p²+m²)`, mass gap sharp by the massless counterexample). **P5 CLOSED 2026-08-20** (RG half CV-26 2026-08-18; attainment `CV/PriceAttainment.lean` — ★★ `price_lower_bound` `\|sin(τλ/2)\| ≤ dist`, ★★ `price_linear_attained` the sandwich `τλ/π ≤ dist ≤ 2τλ`: the linear price is attained, "costs at most" is "costs exactly"). **P3 = permanent ceiling** (epistemic placement, reclassified 2026-08-19), never scheduled. 22 pins across the arc; scope boundaries declared per-module and in `check-claims`' wait ledger | CV/EFT + SigmaLayer | M–L per pillar | **DONE 2026-08-20** | `eft-pillars-plan.md` change log | > | ~~Q23~~ | ~~**The `id ⊗ Φ` identification**~~ **DONE 2026-08-21** (`LF6/LindbladPositivity.lean` `IdTensor` section, strengthened in place per §8.3b; 4 pins) — the amplified GKSL generator's flow IS `id ⊗ Φₜ`, delivered as queued: `blockAtCLM` (the `(i,j)` ancilla block as a continuous `ℝ`-linear map, `ℂ`-linear as a function via `blockAtCLM_csmul`) + the two Kronecker collapse lemmas (`(1 ⊗ A)·M` acts blockwise left, `M·(1 ⊗ B)` blockwise right); ★ `blockAtCLM_ampGenerator` — the amplified generator acts blockwise, **no Hermiticity needed** (pure Kronecker algebra; `(1⊗L)ᴴ(1⊗L) = 1⊗(LᴴL)` by the mixed-product identity); powers by induction; ★★ `blockAtCLM_lindbladSemigroup_amplified` — the FLOW acts blockwise (exponential-series transport through `expCLM_apply_hasSum`, the module's own idiom — HasSum mapped through `blockAtCLM`, unique limits); `idTensorCLM` (`id ⊗ Φ` as a superoperator, `idTensorCLM_one` sanity); ★★ `idTensor_lindbladSemigroup` — **the identification as an equality of superoperators**; ★★ `lindbladSemigroup_completelyPositive` — **CP of `Φₜ` in so many words** (Hermitian `H`, `t ≥ 0`, every finite ancilla). *Route note:* the queued "P2 machinery (`compositeReindex`, `leftOp/rightOp`)" was not needed — the amplified flow already lives on the pair index, so the blockwise computation is direct; no reindex. The `LindbladPositivity` wait-ledger entry retired (supersession record in the module docstring + check-claims WAIT LEDGER); the Q16 CP-brick's declared remainder is discharged | LF6/QuantumInfo | ~~M~~ done, likelihood was high — landed same-week as queued | **done 2026-08-21** | `cp-semigroup-plan.md` honest boundary (superseded) | > | ~~Q24~~ | ~~**Chebyshev-grade canonical typicality**~~ **DONE 2026-08-21** (`Thermo/CanonicalTypicality.lean` extended in place per §8.3b; scoping note `specs/th1-concentration-scoping.md` executed same day; 5 pins) — polynomial-rate canonical typicality, delivered WITHOUT the rated-hard moment integrals: the scoping pass found the second moments are pure **twirl algebra** (no Gaussian/Dirichlet machinery, no `∫ψᵢψ̄ⱼψₖψ̄ₗ` integral, no isoperimetry). New unitaries `phaseFlip i` (quarter-phase) and `hadamardU i j` (the `(1/√2)·[[1,1],[1,−1]]` block; unitarity by restricting each block row's sum to the pair via `Finset.sum_subset` + `Finset.sum_pair` — the B1 gate never fired); kills `fs_cross_linear_zero` (sign flip) and `fs_re_sq_moment` (`E[(Re r)²] = E[xᵢxⱼ]/2`, quarter-phase); ★ `fs_x_sq_eq_two_cross` — the per-pair engine `a = 2b` from Hadamard invariance of the second moment; with the integrated normalisation `a + (N−1)b = 1/N`: ★ `fs_x_sq_moment` (`a = 2/(N(N+1))`) and ★ `fs_x_cross_moment` (`b = 1/(N(N+1))`) — the Dirichlet values by algebra; `fs_linear_expectation`, ★ `fs_linear_sq_moment` (`E[(Σλx)²] = ((Σλ)²+Σλ²)/(N(N+1))`); ★★ `fs_chebyshev_concentration` — `μ_FS{ε ≤ \|Σλₖxₖ − (Σλ)/N\|} ≤ (NΣλ²−(Σλ)²)/(N²(N+1)ε²)` via `meas_ge_le_variance_div_sq` + `variance_eq_sub`. *Gated residue (by the scoping note's own brick list):* B4 (general Hermitian `A` via diagonalisation transport) and B5 (reduced-state union-bound capstone) NOT attempted — fresh-session gates. The exponential Lévy tier stays the L-row below, unchanged. *Snags:* pointwise goals under `integral_congr_ae (ae_of_all _)` arrive beta-unreduced — open with `dsimp only`; nested `Integrable.add` infers Pi-level function sums that `integral_add` can't match — state cumulative integrabilities as explicit-lambda `have`s (defeq accepts); `set μ` folding breaks cross-lemma value rewrites — bridge with `rw [hμ]` haves or drop `set`; `Var[·;·]`/`variance` need `Mathlib.Probability.Moments.Variance` imported and the qualified name (file doesn't open `ProbabilityTheory`); `linarith` fails on compound-division atoms — `linear_combination -h / 2` | LF4/Thermo | ~~scoping M, brick M–L~~ done — scoping found the algebraic route, B1–B3 landed same-day | **done 2026-08-21** | TH-1 row; `MATHLIB-GAPS.md`; `th1-concentration-scoping.md` | > | ~~Q25~~ | ~~**Record layer concern II: what happens to `Ω_j` when outcome `i` is realised**~~ **DONE 2026-08-21** (`RecordLayer/TwoTimeLuders.lean` NEW + `swap_sector_born_ctx` family in `SwapClosure.lean` extended in place per §8.3b; scoping note `specs/two-time-luders-scoping.md` executed same day; 5 pins, SigmaLayer part) — the composed two-time statement on ONE arena, delivered: `TwoStageArena = SwapArena × (fresh register × fresh bank)` with the stage-2 propagator conjugated through the `regroup` shuffle, so **the first record persists STRUCTURALLY** (`stageTwo_register₁`/`stageTwo_bank₁` are `rfl` — the same design payoff as `swapG_register`). *Feasibility discovery that kept it M:* the stage-2 record event never reads the stage-1 record coordinates, so only the SYSTEM marginal of the conditioned post-state enters — `swap_luders_marginal` is the whole engine; no joint factorisation, no `Measure.pi` decomposition, no Fubini. Reusable `cond_map` (conditioning commutes with pushforward); ★ `swap_sector_born_ctx` (dynamical sector Born at an ARBITRARY context field); ★★ `two_stage_joint` (generic: joint record law = stage-1 sector × stage-2 sector at the relocated state); ★★ `two_time_born` — `P(record i at t₁ ∧ record j at t₂) = momentMap p i · c₂.rate [eᵢ] j`, Born-then-Lüders-Born as one number, any second context; ★ `two_time_repeat` (von Neumann repeatability composed: `momentMap p i · δᵢⱼ`); ★ `two_time_other_fate` — **the row's literal question**: conditioned on record `i` at `t₁`, the next partition carries the collapsed weights — the other `Ω_j` of the repeated context are NULL, `Ω_i` certain, a fresh context sees `c₂.rate [eᵢ]`; plus `two_stage_first_record` (no retro-action) and `two_stage_readouts` (both records on display at `t₂`). *Route correction vs the queued sketch:* `rotated_swap_luders_born`/`pointer_luders_born_prep`/`RotatedSwapClosure` were NOT needed — the composition lives one level up, on the swap protocol alone. *Gated residue (scoping note):* the clock-glued two-epoch `MeasurementProtocol` on `[0,2]` (presentation, not physics) and the entangled/composite two-time instantiation (Q27's mixed tier). *Snags:* `rw` with `Measure.map_apply`/`cond_map` needs TYPE-ASCRIBED `MeasurableSet {y \| …}` haves so the pattern instantiates with the `setOf` form (the composed-projection form does not match); `postMeasure` has no probability instance — provide `have`s via `cond_isProbabilityMeasure` + `Measure.isProbabilityMeasure_map` (the name is `Measure.`-qualified); an `omit [MeasurableSpace _]` on a lemma whose statement carries `Measurable idx` is rejected — drop the omit, not the hypothesis | RecordLayer | ~~scoping M, brick M–L~~ done same-day — the marginal-only discovery collapsed the rated hard part | **done 2026-08-21** | `record-layer-plan.md` §4; `two-time-luders-scoping.md`; Q27 residue row | > | Q26 | **The epistemic measure as a disintegration, not a definition** (queued 2026-08-20, external review; `GlobalBasin.lean`'s own design note flags it). `δ_p ⊗ Haar` is currently TAKEN as the isolation-conditioned epistemic measure. The theorem a careful reader asks for: it IS the disintegration of the sector's own measure — `kMuL = μ_FS ⊗ Haar` disintegrates along the base projection into exactly `p ↦ δ_p ⊗ Haar`, `μ_FS`-a.e. Route: Mathlib HAS kernel disintegration for standard Borel spaces (`Probability.Kernel.Disintegration`); for a product measure the disintegration is the constant kernel, so the content is identification + a.e.-uniqueness. Turns a definition into a derived object. **DONE 2026-08-21** (`RecordLayer/EpistemicDisintegration.lean`, 3 pins) — exactly as scoped, and CHEAPER than rated (S–M in practice: the 2026-08-21 feasibility probe had already cleared the two instance risks — `StandardBorelSpace KTorus` synthesizes via `StandardBorelSpace.prod`, `(kMuL p₀).condKernel` elaborates): `kMuL_fst` (the `c = 1` bridge as a marginal), `kMuL_eq_compProd_const`, ★★ `kMuL_condKernel_ae` (**the identification** — the disintegration kernel is `μ_FS`-a.e. the constant Haar kernel, via `Measure.condKernel` + `eq_condKernel_of_measure_eq_compProd` on the constant kernel), ★★ `epistemicMeasure_eq_disintegration` (**the headline** — `δ_p ⊗ Haar` IS the disintegration fibre planted at its base point, `μ_FS`-a.e.), ★ `kMuL_disintegration` (the reassembly `kMuL = μ_FS ⊗ₘ condKernel`). The a.e. form IS the theorem — kernels are only determined up to base-null sets, so no pointwise strengthening exists to owe. `GlobalBasin`'s "modelling choice stated as a definition" note superseded at source; the glossary `epistemic-measure` page upgraded to `proved-in-corpus` anchored at the headline. *Snag:* `Measure.compProd_fst_condKernel` (the docstring's own advertised name) does not exist at this pin — the identity is the `ρ.disintegrate κ` method | LF4/SigmaLayer | ~~M–L~~ S–M actual | **done 2026-08-21** | `GlobalBasin.lean` design note (superseded) | > | Q27 | **What entanglement does to the weights** (queued 2026-08-20 — the external reader's direct question, not addressed by the Bell/no-signalling work). CSD form: for an ENTANGLED composite preparation, a local context's record weights should be the REDUCED state's Born weights. **FIRST BRICK DONE 2026-08-20, same day as its queueing** (`CV/EntangledWeights.lean`, 7 pins): `reducedDM` (the right partial trace of `arenaDM` read on the pair index) is a genuine density operator for EVERY composite point (`reducedDM_posSemidef` / `reducedDM_trace`); ★★ `arenaObs_leftOp_eq_reduced` — **the bridge**, `arenaObs (leftOp A) x = re tr(reducedDM x · A)` for every composite point, entangled included: local arena observations ARE mixed-Born pairings against the reduced state; ★ `reducedDM_join` — on a product point the reduced state is the PURE local state `arenaDM p`, so departure from rank-one is exactly entanglement's contribution; ★★ `reducedDM_bell` — the Bell ray's reduced state is the equal mixture `½(∣x₀⟩⟨x₀∣ + ∣x₁⟩⟨x₁∣)`, with corollaries ★ `bell_local_weight₀`/`₁` = exactly `1/2`: entanglement maximally mixes the local weights across the correlated patterns, and the remote pattern labels are gone (`composite_no_signalling` read as a consequence). *Landing note:* the supporting reduced-trace identity was NOT re-proved — the draft's duplicate was dropped and `trace_mul_kronecker_one_right`/`_left` were **rehomed** from `Subadditivity.lean` to `PartialTrace.lean` (the defining property lives beside the definition; same names/namespace, K1-B consumers unchanged through the public import). ~~**Remaining (the row's second half, open):**~~ **SECOND HALF DONE 2026-09-14 — RL-1** (`RecordLayer/EntangledRecord.lean`, 15 pins; ▶ OPEN QUEUE ~~#5~~; the composite-arena seam is #25): the record-tier connection — transporting the delivered `re tr(reducedDM · A)` form through the `Fin`-indexed LF2 mixed tier (`DensityOperatorIx`, `MixedLuders`) ~~is declared index plumbing in the module scope note and is not claimed~~ is `DensityOperatorIx.toFin` + `reducedDensity e x` + `traceForm_reducedDensity_single`, and ★★ `entangled_local_record_born` / ★★ `entangled_two_time_born` are the record-tier and two-time forms; ~~sequential/record-conditioned versions are Q25's territory~~ **the sequential/record-conditioned FORM landed with Q25 2026-08-21** (`RecordLayer/TwoTimeLuders.lean` — the two-time joint law with structural persistence); ~~what remains with THIS row is the mixed-tier transport plus the **entangled two-time instantiation** (the swap witness over the composite arena with `reducedDM` weights — recorded in `two-time-luders-scoping.md`'s gated list)~~ both delivered by RL-1 (the two-time form rides a NEW generic `mixed_two_time_born`, the two-stage arena at `mixedReadyPrep ρ` for every density operator). The ★★ are census candidates at the next Q17-criteria sweep. *Snags:* `rw [Matrix.single_apply_of_ne …]` unmatchable on module-system `Matrix` entries (the application `single … i' j'` is not type-correct at `rw`'s implicit transparency — "function expected"; the B5-geom defeq trap in `rw`-pattern form) — route through `simp only [Matrix.single, Matrix.of_apply]` instead; and that `simp only` reduces `a = a` ite conditions to `True`, so the `if_pos` witness is `⟨trivial, trivial⟩ : True ∧ True`, not `⟨rfl, rfl⟩` | CV/RecordLayer | first brick ~~S–M~~ **done**; record-layer statement ~~M (likelihood medium–high)~~ **done** (S–M); composite-arena seam **done** (S) | first brick **done 2026-08-20**; second half **done 2026-09-14**; composite-arena seam **done 2026-09-14** (~~#25~~, `RecordLayer/CompositeRecord.lean`) — **Q27 CLOSED** | P2 (`CompositeArena.lean`), `QuantumInfo/PartialTrace` | > | Q28 | **The C2 support arc** ([`c2-support-plan.md`](c2-support-plan.md), scoped 2026-08-21 from the external C2 review brief; every cited ingredient verified against the tree, one correction found — **no metric on `ℙ` exists**, so the brief's `Metric.ball`/`dist` forms are restated topologically; the FS metric + `segre_range_null` are recorded as Mathlib-gap items, not scheduled). **The C2-blocking item is 2**: (2a) `segre_range_isClosed` (compact image — `instCompactSpace` already staged — via `segre` continuity, gate on the rep-vs-descended form); (2c₀ freebie with 2a) `(range segre)ᶜ` open + nonempty (`segre_not_surjective`) ⟹ **positive μ_FS measure** — citable immediately; (2b′) entangled rays in every open neighbourhood of a product ray (`w t = a⊗b + t•(a'⊗b')`, diag(1,t) rank-2 via the `segre_not_surjective` technique); (2c′) `fubiniStudy_entangled_pos` on every open meeting the Segre image. Home: `OnticComposite.lean` in place; its "not attempted" caveat superseded at source. **Then, none C2-blocking**: (4) `conditional_not_mutuallySingular` (the ψ-epistemic claim — density argument, NOT shared-support; + the metric-free overlapping-open-preimages witness; doc-currency rider: glossary `psi-epistemic` + `does-csd-conflict-with-pbr` pages cite it when landed); (3) `projectivePreparationLaw ≪ μFS` + `withDensity` (ρ_ep — the object already exists in `ProjectiveSector.lean`; the Kähler `c = 1` corollary is the first statement mentioning both `SigmaLayer.Preparation` and the bridge — the seam that broke C2 v1.01); (1) `fsMeasure_singleton = 0` for `N ≥ 2` (pigeonhole via transitivity + invariance + probability + infinitude — retires `KahlerInstance`'s "Haar-of-subgroup" caveat; fibres `kMuL`-null corollary; Mathlib-upstream candidate). Item 5 (`segre_range_null`) research-gated in `MATHLIB-GAPS.md`. **ITEMS 1–4 ALL DONE 2026-08-21, same day as the scoping.** Item 2 (`OnticComposite.lean` extended in place + new `RecordLayer/EntangledMeasure.lean`, 6 pins): ★ `segre_range_isClosed`; `not_mem_range_segre` (the minor criterion, reusable); ★ `exists_entangled_mem_nhds` — the perturbation is a SINGLE standard-basis vector `e_{(j₁,k₁)}` (cheaper than the brief's `a'⊗b'`, no orthogonal complements); `compositeFubiniStudy` (THE FS measure across `finProdFinEquiv`, probability, full support); ★★ `compositeFubiniStudy_entangled_pos_global` + ★★ `compositeFubiniStudy_entangled_pos` — **the C2 contradiction's set has positive measure in every open neighbourhood of every product ray**; the `OnticComposite` "not attempted" caveat superseded at source. Item 4 (`IsolationPreparation.lean` in place + `SigmaLayer/PreparationDensity.lean`, the density argument exactly as briefed): ★★ `conditional_not_mutuallySingular` + ★★ `kahler_preparations_overlap` (the ψ-epistemic witness on overlapping open-preimage preparations). Item 3 (`PreparationDensity.lean`): ★ `projectivePreparationLaw_absolutelyContinuous` (no `c ≠ 0` needed — `c • μFS ≪ μFS` always), ★ `projectivePreparationLaw_withDensity` (**ρ_ep exists**), `kahlerFstSector_projectiveLaw` (`c = 1` exact) + ★★ `kahler_preparation_density` — the seam module speaks both preparation interfaces by design. Item 1 (`FubiniStudyUnique.lean`, 2 pins): ★ `fsMeasure_singleton` (pigeonhole, NO stabiliser Haar) + `projectivization_infinite`; `KahlerInstance`'s "Haar-of-subgroup" caveat superseded at source. *Snags:* `rw [someDef]` on cross-module defs fails equation-theorem unfolding AND whole-measure `show`-defeq can `isDefEq`-timeout — go SETWISE (`Measure.ext` + the `_apply` lemma); `⋃ x ∈ (S : Finset), {x}` is a Finset-binder pattern `Set.biUnion_of_singleton` cannot match — use `sum_measure_singleton` with a hand-built `MeasurableSingletonClass` from T1; `EuclideanSpace.single_apply` deprecated → `PiLp.single_apply`; `Nonempty (KSigma N)` needs `[NeZero N]` (via `instNonemptyCPN`); simp pre-reduces `a = a` ite conditions (the `⟨trivial, trivial⟩` trap again) | RecordLayer/SigmaLayer + Mathlib staging | 2: M ~~(blocking)~~ done; 4: done; 3: done; 1: done; ~~5: research~~ **done** | **items 1–5 ALL done** (1–4 on 2026-08-21; **item 5 landed 2026-08-22 via MG-2** — `compositeFubiniStudy_range_segre_null` / `ae_not_mem_range_segre`, the Segre cone inside one coordinate quadratic + FS-as-Lebesgue-pushforward; no general polynomial lemma needed) | external C2 review brief; `c2-support-plan.md` | > | **E3** (SPIKE — run FIRST) | **The restricted pushforward** `π_*(μ_L\|_S) = μ_FS^{H_R}` ([`equilibration-arc-plan.md`](equilibration-arc-plan.md)). ⚠️ Premise correction from the planning pass: the UNRESTRICTED constant is not existential — `kahlerFstSector_projectiveLaw` computes it exactly (`c = 1`, Q28). But that proof is a PRODUCT computation (`μ_FS ⊗ vol` through `Prod.fst`), so it transfers only to **fibre-saturated** constraint surfaces; a generic energy shell is not saturated, and `fsMeasure_unique` does not apply either (it needs `U(N)`-invariance on ALL of `ℂℙ^{N−1}`, whereas the restricted object lives on `ℙ(H_R)` with only `U(d_R)`-invariance). The real question is saturation + sector invariance, not the constant. Three-way gate fixed in advance (PASS / PARTIAL-with-hypothesis-carried / FAIL-and-demote-to-posit); the theorem-vs-posit DECISION is itself the deliverable. **EXECUTED 2026-08-22, same day — gate outcome PARTIAL, the naive statement is FALSE** (`Thermo/SectorRestriction.lean` + `fsMeasure_subspaceRays` staged, 4 pins): the POSITIVE half is `projectiveLaw_restrict_saturated` (fibre-saturated surfaces `S = π⁻¹B` push forward to `μ_FS\|_B` exactly — the honest generalisation of the unrestricted `c = 1`, needing saturation because the proof is a product computation); the NO-GO is ★★ `projectiveLaw_restrict_sector_eq_zero` — for a proper spectral sector the restriction is the **ZERO measure**, because the constraint set is **Fubini–Study-null** (rays of a proper subspace have the subspace as their cone; `Measure.addHaar_submodule`). Not a normalisation failure and not repairable by one. Consequence: **E2 must be re-scoped to positive-measure energy windows** (route 1, theorem route — the conditioned law is NOT `μ_FS` on any `ℙ(H_R)`) or accept the sector-as-own-arena **posit** (route 2); "μ_FS on the sector" is refuted and becomes a check-claims guard candidate in E5. E1 is untouched | SigmaLayer + Mathlib staging | ~~M–L~~ done same-day | **done 2026-08-22 (verdict: naive statement false, arc proceeds re-scoped)** | brief item 3; `equilibration-arc-plan.md` | > | **E1** | **FS second moment for reduced states** — `E‖Tr_B\|ψ⟩⟨ψ\| − I_A/d_A‖₁²` bounded in `d_A/d_B`. ⚠️ Much cheaper than the brief assumed: the first moment is LANDED (`canonical_typicality_expectation`, TH-1) and the second moments are LANDED (Q24 twirl algebra). What remains is exactly Q24's gated **B4 + B5**, already route-scoped in `th1-concentration-scoping.md`. **H-TENSOR discipline is binding**: the bipartition enters as a named hypothesis in the signature (as `canonical_typicality_expectation` already does), NEVER silently through the tensor API — now a live temptation since `regTensorEquiv` landed (MG-5). Rationale: a silently-chosen factorisation is a second D1. **EXECUTED 2026-08-22 — PARTIAL** (`Thermo/ReducedSecondMoment.lean`, 4 pins): H-TENSOR honoured (the bipartition is an explicit `e : Fin N ≃ Fin dA × Fin dB` in every signature); `blockPop_eq_linear` makes the subsystem populations a LINEAR moment-map statistic so Q24 applies verbatim ⇒ ★ `fs_blockPop_mean` (`= d_B/N = 1/d_A`) and ★ `fs_blockPop_sq` (`= (d_B²+d_B)/(N(N+1))`); the new twirl content is `signFlip_smul_rayDensity_ne` + ★ `fs_redOff_cross_vanish` (**four-index expectations vanish** — an odd-occurring coordinate is killed by the sign flip). **Extended 2026-08-23**: ★★ `fs_redOff_normSq` (`E|(ρ_A)_{aa'}|² = d_B/(N(N+1))` for `a ≠ a'` — the `b = b'` terms are the landed cross moment, the `b ≠ b'` terms vanish) — **E1's mathematical content is now closed**: every moment the Lubkin–Page average needs is a theorem. **CLOSED 2026-08-23 — the Hilbert–Schmidt assembly landed** (4 further pins): `reducedMatrix`/`hsDeviation`/`hsDeviationNormSq` (the reduced state as a matrix, its deviation from maximal mixing, and the entrywise HS norm squared; `redOff_diag` bridges the two vocabularies); ★ `fs_hsDeviation_diag_sq` (short because the mean population is EXACTLY `1/d_A`, so the cross term collapses against the constant); ★ `fs_hsDeviation_off_sq`; ★★ **`fs_hsDeviationNormSq`** = the Lubkin–Page purity average `E‖ρ_A − I_A/d_A‖₂² = (d_A+d_B)/(N+1) − 1/d_A`; ★ `fs_hsDeviation_typicality` (Markov on it — the usable form). **H-TENSOR dividend**: `N = d_A d_B` is NOT an added hypothesis, it is read off the bipartition by `card_eq_mul_of_tensorEquiv` — the discipline adopted for honesty also paid for the arithmetic. Honest residue (in the module header): the trace-norm form needs the matrix-norm API not moments; the `Matrix.traceRight` identification is bookkeeping in another vocabulary; and the concentration delivered is **Markov on a quadratic functional**, weaker than `fs_chebyshev_concentration` (which does apply to the linear populations individually). *Process:* the assembly had failed twice as a long single-shot draft (both drafts carried placeholder `sorry`s and were reverted, not patched); built this time in four increments with a build after each — no reverts | Thermo | ~~M~~ done | **done 2026-08-23** | brief item 1; Q24-B4/B5 | > | **E2** | **Constrained microcanonical version** — E1 with `μ_FS` on the unit sphere of a spectral sector `H_R`. Explicit hypotheses required in the signature: which restriction, why it is dynamically preserved (an invariance hypothesis on the Σ-flow, not an assertion), and what "effective dimension" means (`d_R = finrank ℂ R`). Inherits E3's verdict for the `μ_L ↔ μ_FS^{H_R}` relation and its status (theorem / hypothesis-carried / posit). ⚠️ **RE-SCOPED 2026-08-22 by E3's verdict:** the exact-eigenspace formulation is **refuted** (`projectiveLaw_restrict_sector_eq_zero` — the sector is `μ_FS`-null, so the restriction is the zero measure). E2 must instead use a **positive-measure energy window** `{p \| ⟨H⟩_p ∈ [E, E+Δ]}`, carrying the window and its measure-positivity in the signature, and must NOT describe the conditioned law as "`μ_FS` on the sector" (E3 refutes that phrase). Cost rises: the window needs `⟨H⟩` measurability plus a non-degeneracy hypothesis to make positivity non-vacuous. **EXECUTED 2026-08-23** (`Thermo/EnergyWindow.lean`, 3 pins; built incrementally per E1's process note — no reverts): for a diagonal Hamiltonian the energy expectation is the LINEAR statistic `Σ λ_k x_k`, so Q24's moments and Chebyshev control the window with no new integral. ★★ `energyWindow_ne_zero` — the window provably carries weight once `Var/ε² < 1`, i.e. wider than the fluctuation scale: **the positivity E3 made load-bearing is QUANTITATIVE, not assumed**. `microMeasure` = μ_FS conditioned on the window (a probability measure). ★★ **The structural finding**: conditioning breaks `U(N)` invariance but NOT uniformly — a sign flip fixes every moment coordinate, hence the energy, hence the window (`map_signFlip_microMeasure`), so ★★ `micro_redOff_cross_vanish` holds **microcanonically**; permutations and Hadamard move coordinates and change the energy, so the moment VALUES do not transfer. Conditional moments = a microcanonical density-of-states problem, NOT attempted. Nothing calls the conditioned law "μ_FS on the sector" (E3 refutes that) | Thermo + SigmaLayer | ~~M–L~~ done | **done 2026-08-23** | brief item 2; E3 verdict | > | **E4** | **Equilibration as a CONDITIONAL mixing theorem** — decaying correlations ⇒ time-averaged reduced states converge to the `μ_FS` average at a rate set by the decay. Mixing is NEVER proved. ⚠️ Wall check (2026-08-22): Mathlib has `Ergodic` but **no mixing definition, no mean ergodic theorem, no pointwise Birkhoff**. So state the antecedent as **quantitative correlation decay** and get the Cesàro estimate elementarily — routing through abstract mixing ⇒ ergodicity ⇒ Birkhoff hits the recorded gap immediately. That choice is the difference between a feasible brick and a blocked one. ⚠️ **Wall note partly CORRECTED 2026-08-23 while executing**: Mathlib DOES have the von Neumann mean ergodic theorem (`Analysis/InnerProductSpace/MeanErgodic.lean`) and a `Dynamics/BirkhoffSum/` directory — the "no mean ergodic theorem" clause was stale. The route decision is unaffected and for a recordable reason: von Neumann gives **no rate** and its limit is the invariant projection, which is the space average only under an *ergodicity* hypothesis. Right conclusion, partly wrong premise — the "wall labels rot" lesson again. **EXECUTED 2026-08-23, two modules, 7 pins.** Generic engine `Mathlib/Dynamics/CorrelationDecay.lean` (Cat-1, 5 pins): `HasCorrelationDecay` (the explicit two-index bound); `sum_sum_nat_dist_le` = the only combinatorial content and the source of the factor two (within a row `t ↦ Nat.dist s t` is injective on each side of the diagonal SEPARATELY — truncated subtraction collapses the left half to 0); ★ `integral_birkhoffAverage_sub_sq_le` → ★★ `integral_birkhoffAverage_sub_sq_le_cesaro` (`E[(A_T f − ⟨f⟩)²] ≤ (2/T)Σ_{u | **E5** | ⚠️ *Work order RETAINED but NOT REQUIRED (2026-09-05, CR-12).* Under the closed-posit architecture the reconstruction does not need this spike: the outcome rates are pinned by torus generation (`torusGenerated_eq_momentMap`) and what remains is Posit 1's discharge condition, not a partition-from-dynamics derivation (`specs/POSITS.md`). It is kept here, unexecuted parts and all, **for anyone who later disputes Posit 1** — that is the case in which the spike becomes live again. Do not schedule it as reconstruction work. **Non-vacuity witness + register the arc.** (a) Exhibit a concrete Σ + Hamiltonian satisfying E2/E4's hypotheses NONTRIVIALLY — without it E4 is a conditional with an unpopulated antecedent (candidate: a finite-mode CV drive with closed-form correlation decay). (b) Register E1–E4 in `specs/validation-claims.tsv` with honest status and `load_bearing` filled in (⚠️ there is no `claims.yaml`; the TSV + `VALIDATION-LEDGER.md` are the register), and add a `check-claims.sh` guard phrase so E4 cannot reach external prose without its antecedent. **EXECUTED 2026-08-23 — and goal (a) came back with a partial NO, which is the finding.** ⚠️ **For finite-dimensional unitary dynamics there IS no witness**, and in the periodic case that is now a theorem: ★ `HasCorrelationDecay.integral_mul_self_eq_of_periodic` (if `Φ^[k] = id`, a summable envelope forces `⟨f²⟩ = ⟨f⟩²`, i.e. an a.e. CONSTANT observable — the correlation at lag `k·m` is exactly `⟨f²⟩`, and a constant bounded by a null subsequence is zero) ⇒ ★★ `not_hasCorrelationDecay_blockPop_of_periodic`: for `d_A ≥ 2` **no periodic flow satisfies E4's antecedent for the population, for any summable envelope** (Q24 arithmetic: `fs_blockPop_sq` and `fs_blockPop_mean` agree with the periodic conclusion exactly when `N = d_B`, i.e. `d_A = 1` — no subsystem). A unitary on `ℂℙ^{N−1}` generates a relatively compact group so its correlations are almost periodic and cannot decay either; the finite-order case is the piece available without a recurrence argument — **the general almost-periodic statement is argued but NOT proved** and is declared as such. This is the referee's attack, found before the referee. **Positive half — the engine is not vacuous**: `Mathlib/Dynamics/CorrelationDecayWitness.lean` (Cat-1, 5 pins), the doubling map on `ℝ⧸ℤ` with `cos 2πx`: ★★ `circ_hasCorrelationDecay` (finitely-supported envelope) + ★★ `circ_nontrivial` (`⟨f²⟩ = 1/2 ≠ 0`, so NOT the trivial constant case) + ★ `doubling_not_periodic` free. Every correlation computed by **Q24's SIGN-FLIP argument, not by integration** (rotate by `2^{-(s+1)}`: the `s`-factor flips, the `t`-factor sees an integer shift for `s | **E6 — the almost-periodic no-go, general case** (residue of E5, opened 2026-08-23) | E5 proved the no-go for **periodic** flows and argued the general **almost-periodic** case without proving it. That gap is now NARROW and precisely located — two of its three pieces landed 2026-08-23: ★★ `HasCorrelationDecay.integral_mul_self_eq_of_recurrent` reduces the whole statement to one property `hrec` (the correlation must return near its lag-zero value at arbitrarily large lags; the periodic case is now its corollary), and ★ `exists_le_pow_mem_of_compactSpace` (`Mathlib/Topology/Algebra/CompactRecurrence.lean`, Cat-1) supplies the recurrence of `U^n` itself — powers of any element of a **compact topological group** return to every neighbourhood of `1` at arbitrarily large exponents. ⚠️ **Wall label had rotted**: `UnitaryGroup.instCompactSpace` DOES exist (an earlier grep missed it), so `Matrix.unitaryGroup` is a compact topological group off the shelf. **What remains is exactly one analytic bridge**: the UNIFORM estimate `|f (V • p) − f p| ≤ c·√(dev V)` with `dev` continuous and vanishing at `V = 1`, which transfers group recurrence to the correlation. Route scoped: `dev V = Σ_{k,j} ‖V k j − δ_kj‖²`; `|momentMap (V•p) k − momentMap p k| ≤ 2‖(V−1)ψ‖/‖ψ‖` via `|a²−b²| = (a+b)|a−b|`; then discrete Cauchy–Schwarz entrywise. ⚠️ Must be UNIFORM, not dominated convergence: **`FirstCountableTopology` does NOT synthesize for `Matrix.unitaryGroup`**, so `continuous_of_dominated` is unavailable (probed 2026-08-23). Probed sub-steps: the `toEuclideanLin` entry formula is `Complex.ext rfl rfl`; discrete Cauchy–Schwarz, general unitary norm-preservation, and entry-continuity in the group element each need hand-construction (no `exact?` hit). **CLOSED 2026-08-23, same day** — the bridge landed and the general no-go is a theorem: ★★ **`not_hasCorrelationDecay_blockPop_of_unitary`** — for `d_A ≥ 2` and **ANY** unitary `U`, E4's antecedent is FALSE along `p ↦ U • p`, for every summable envelope. No periodicity hypothesis; the periodic version is now the weaker sibling (kept because it covers non-unitary periodic `Φ`). The bridge: ★ `abs_blockPop_smul_sub_le` (`|blockPop (V•p) a − blockPop p a| ≤ 2·matDev V`, **uniformly in the state**) → ★ `abs_corr_smul_sub_le`. ★ **`matDev` is the ENTRYWISE ℓ¹ deviation**, not ℓ² — that choice collapses the proof to a triangle inequality plus `coord_norm_le`, with **no Cauchy–Schwarz and no square roots** (the scoped route was harder than needed). Supporting: `norm_toEuclideanLin_unitary` (the general isometry statement behind CanonicalTypicality's per-gate `signFlip_normSq`/`perm_normSq`/`hadamard_normSq` — extraction candidate), `toEuclideanLin_comp`, `smul_iterate`; `blockPop_variance_ne` factored out so both no-goes share the Q24 arithmetic. ★ **What unblocked it**: the corpus already had `(continuous_const_smul U).measurable` and `fsMeasure_smul_invariant` — the projective action is CONTINUOUS, so no quotient-topology work was needed. **Read as a LIMITATION, not a refutation**: E4's machinery is sound, its antecedent is simply not populated by finite-dimensional unitary Σ-dynamics ⇒ equilibration there rests on the TYPICALITY results (E1/E2), not on mixing | Thermo + Mathlib staging | ~~M–L~~ done | **done 2026-08-23** | E5 residue; `equilibration-arc-plan.md` | > | **Lévy — CUT from the arc** (author decision 2026-08-22) | Severable, gates nothing in E1–E5, and is a Mathlib-PR candidate rather than a CSD result. The `MATHLIB-GAPS.md` row stands unchanged; the arc consumes Q24's polynomial tier instead | — | — | **cut** | brief's own cut list | > > **Settled non-goal (2026-08-12, author decision):** the beyond-horizon permanence toy > (expanding-chain "Λ > 0 ⇒ records beyond the cosmological horizon persist forever") is > **off the queue** — its physical referent is unmeasurable in principle from inside the > horizon, so it fails the present-tense-checkable test however formalizable it is. Do > not re-propose; the idea stays recorded in the vault navigation note only. > > ### A. The blocker, split into what is doable and what is not > > The one open scientific requirement is Paper C A2 / TN6's **joint-arena Hamiltonian > flow**: measurement generated by a scalar `𝓗` on the whole arena, whose *actual* flow > creates Born-weighted records. What exists is fibrewise Schrödinger with the base frozen > (`PointerGeneration.lean`'s dated boundary note). Verified 2026-08-04: Mathlib now has > manifolds, `ContMDiff`, `Diffeomorph`, `DerivationBundle` and `IntegralCurve`, but **no > symplectic form, no Poisson bracket, no exterior derivative**. So exactly one arrow is > missing — `𝓗 ↦ X_𝓗 = ω⁻¹d𝓗` — and everything downstream of "here is the flow" is > expressible today. The blocker therefore splits: > > | # | Size | Task | Blocked? | > |---|---|---|---| > | ~~A1~~ | **DONE 2026-08-04** (`RecordLayer/JointFlowTransfer.lean`, 5 pins) | ★★ `IsJointLift.outcomeSector_eq` — the record preimages of a joint lift and of the fibrewise witness are **the same set**, not merely the same measure: `arenaRecord` is a cylinder over the pointer, so horizontal back-reaction is *invisible* to it. Landing and both Born bounds then transport with **no measure-theoretic work**. ★ `weights_conserved` (what the Poisson argument delivers) and ★ `moment_marginal_unchanged` (the honest replacement for the no-collapse theorem — under a lift `pointerEvolve_base_marginal_unchanged` genuinely fails, and what survives is that the moment-register data is unchanged). Non-vacuous (`isJointLift_pointerEvolve`). Conditional by design; the paper discharges the hypotheses. **Scope:** the single measurement stroke — the protocol's two-time laws and persistence are statements about a propagator family, not one map, and are not restated. Formerly: (`_of_` form, `CONVENTIONS.md` §8.3). For *any* flow `Φ` on the arena with **(1)** the `(m, θ₁, q)`-marginal of `pointerPrep` preserved, **(2)** the moments `m` **and** `θ₁` conserved, **(3)** pointer component = the fixed-weight propagator `exp(-it·H_eff(w))` — prove landing, persistence and the ε-Born sandwich hold verbatim. Pure measure theory; the substantive half, since it makes horizontal back-reaction **harmless**. Non-vacuous: the current fibrewise flow satisfies (1)–(3) with the base fixed. ⚠️ *Hypotheses restated %s.* The first draft said "(1) arena-measure preservation, (2) `basinIndex` conserved", and **both were wrong**: Born is measured against `pointerPrep`, a base-**Dirac slice** that is Liouville-null, so "Hamiltonian ⇒ Liouville" says nothing about it; and `pointer_landing` (the smooth witness's actual landing theorem) is hypothesised on `shrunkCell` membership, not on `basinIndex` — conserving `m` and `θ₁` implies both selectors, conserving `basinIndex` alone does not. ⚠️ Also: no perturbative softening is free — `circleCell` is half-open and `rep` jumps at 0, so the selector is discontinuous in `(m, θ₁)`; *approximate* conservation buys nothing without a separate collar estimate | **NO** | > | A2 | **Paper** — ✅ **B1 discharged 2026-08-04; unblocked** | Write the Poisson argument: `{wᵢ,wⱼ} = 0` because the weights depend only on the moment coordinates `mᵢ` (abelian-torus moment map ⇒ pairwise commuting) and one register coordinate `θ₁`; hence `m` and `θ₁` are conserved, discharging A1's (2). Three things %s that must be in the paper: **(a) the dependency** — `clampDiv` is `max`/`min`, Lipschitz but **not `C¹`**, so `{wᵢ,wⱼ}` is *undefined on the current weights*: B1 is a **prerequisite**, not an independent nicety, and even after it the available smoothness is the universal-cover lift under `2ε < r`, `r < 1`; **(b) scope** — the commutation holds for `momentContext`, not for an arbitrary `ContextField`, and since the capstone's `every_basis` ranges over all bases the claim must read "for each fixed context, with *its own* moment family" (distinct contexts' families do not commute with each other); **(c) what the lift costs** — the base point `p` is **not** conserved (only its moments are), so `pointerEvolve_fst` (`… .1 = y.1 := rfl`) fails and with it `pointerEvolve_base_marginal_unchanged`, the theorem the corpus advertises as the smooth horn's no-collapse boundary. The honest replacement is "the **moment** marginal is unchanged" — say so, or a reader will notice the lift silently repeals a theorem the corpus leans on. **Bonus, verified:** the product form is not an idealisation — `couplingH w = ∑ⱼ wⱼ • pointerH j` is linear in `w` and the FS moment map is linear in its generator, so `𝓗 = Σⱼ wⱼ(x)hⱼ(q)` **identically** | **NO** (authorial) | > | ~~A3~~ | **DONE 2026-08-04** (`SigmaLayer/ChartBracket.lean`, 4 pins) | ★ `poissonBracket_eq_zero_of_disjoint` — and the statement is **not** the naive one: `𝓗 = Σⱼ wⱼ(x)hⱼ(q)` is *not* momentum-free (it depends on the pointer's momenta), so "both independent of momenta" would be false of `𝓗`. What holds is **disjoint support**: `w` momentum-free, its position indices disjoint from `𝓗`'s momentum indices — exactly the arena's base-vs-pointer product structure. Plus `poissonBracket_comm_of_momentumIndep` (`{wᵢ,wⱼ}=0`), `hamiltonianField` (explicit in canonical coordinates — no `ω⁻¹` needed), and ★ `conserved_of_bracket_eq_zero`/`weight_conserved_of_disjoint`, which feed A1's `weights_conserved`. **Scope:** a chart *model*; nothing transports it to the arena — that transport IS A4. `dω = 0` unstated because in a canonical chart it is automatic, which is why this is weaker than the manifold claim. Formerly: — the formalisable fragment of A2. `Mathlib/Analysis/InnerProductSpace/KahlerForm.lean` already gives the pointwise Kähler triple and Mathlib has `fderiv`, so a *chart-level* `{f,g} = ω(∇f,∇g)` with `{wᵢ,wⱼ} = 0` and weight conservation is buildable as concrete computations. Would NOT give `dω = 0` or the global statement — a local theorem, weaker than the paper's, but machine-checked | **NO** | > | A4 | **Linear fragment DONE 2026-08-06** (`Mathlib/Analysis/InnerProductSpace/HamiltonianVectorField.lean`, 4 pins + `RecordLayer/PointerHamiltonianField.lean`, 1 pin) | The manifold identification `X_𝓗 = ω⁻¹d𝓗` itself. **What landed:** the identification at the linear level, where a single global chart covers everything — ★ `hamiltonian_duality` (`ι_X ω = dH` for any observable whose differential is `g`-represented, no inverse formed), ★★ `quadraticEnergy_hamiltonian_duality` (the Hamiltonian vector field of the quantum energy `½⟨x,Ax⟩` IS the Schrödinger field `-(i•Ax)` — Kibble/Ashtekar–Schilling as a theorem), and ★★ `coupling_hamiltonian_duality` (the same, on the witness's own fixed-weight generator `couplingH w` — so the fixed-weight loop closes at the formalisable level: energy → field (here) → flow (`rampedU_schrodinger`) → form-preservation (`schrodinger_flow_kahler_symplectomorphism`)). **Wall re-verified per the check-impossible-first rule, and it MOVED:** Mathlib's pin now carries `extDeriv` on **normed spaces** (`Analysis/Calculus/DifferentialForm/`, with `d²=0`) — the 2026-08-04 "no exterior derivative" note is superseded for flat space; manifold forms remain that file's own TODO. Flat `dω = 0` for the constant fundamental form **LANDED 2026-08-06** (`Mathlib/Analysis/InnerProductSpace/KahlerClosed.lean`, 3 pins): generic `extDeriv_const` (constant differential forms are closed — the Mathlib-gap lemma), `Kahler.fundamentalFormAlt` (ω packaged as a continuous alternating 2-form, pointwise-tied by `fundamentalFormAlt_apply`), ★ `Kahler.extDeriv_fundamentalFormAlt` (`dω = 0` on the flat tangent model). Foundational-triple-only | ~~**Manifold form: YES** — §2a wall, same as A1/A3, but *narrowed*: the blocked residue is now exactly the quotient/manifold packaging (forms on `ℂℙ^{N-1}`/the arena), not the identification itself~~ **Manifold form DONE**: on `ℂℙⁿ` 2026-09-08/09 (`schrodingerField_isHamiltonianVectorField`, `torusField_isHamiltonianVectorField`, for `fsForm_isSymplectic`), the flow identification 2026-09-12 (`hamiltonianFlow_schrodingerHamiltonian`), and on the arena `ℂℙᴺ × T²` 2026-09-14 (`hamiltonianFlow_sectorEnergy_schrodinger`, `LF4/ArenaSymplectic.lean`) | > > ### B. Not blocked by anything — ordinary work > > | # | Size | Task | > |---|---|---| > | ~~B1~~ | **DONE 2026-08-04** | ~~**Substitute the `C^∞` profiles into the protocol stack.**~~ Landed: `RecordLayer/SmoothProfile.lean` extracts the profile primitives *below* the pointer stack (they had lived **above** it, which is why the witness could not be built on them), and `pointerWeights` is now `smoothArcWeight`. The plateau interface being identical by construction, **every downstream proof transferred with no changes** — both targets green at 3773/3797 including all audit pins. New: `contDiff_pointerWeights_lift`. **A2's prerequisite is discharged: `{wᵢ,wⱼ}` is now a well-formed expression on the corpus's weights.** Deliberately NOT substituted: the time ramp, which is not a phase-space function and whose swap would change the capstone's `generation` field to carry a rate factor — recorded as B1b. Formerly: Confirmed outstanding by the Opus review: `grep` finds **zero consumers** of `smoothArcWeight`/`smoothPointerRamp`/`smoothRampedU_schrodinger` outside their own module, and the capstone's `generation` field is still the trapezoid, window-restricted `rampedU_schrodinger`. `PointerSmoothProfile.lean` proves the profiles and the global-time ODE, but `PointerProtocol`/`PointerBorn` still run on trapezoids. Recorded as "mechanical" — recording is not doing, and the fifth review was right to list it as outstanding | > | ~~B1b~~ | **DONE 2026-08-04** (commit ad53638) — ⚠️ *this row sat marked-open for several commits after it was finished, because the edit that should have struck it through was in a script block that aborted on an earlier error; found 2026-08-04 while re-reading the list. The exact staleness class the whole day was spent correcting.* `pointerRamp` is now `(π/2)·smoothTransition`, `C^∞` (`contDiff_pointerRamp`), plateau interface unchanged so the two-time law, freezing and persistence were untouched. The one casualty, as predicted: `rampedU_schrodinger` is **window-free** but carries the rate factor `smoothTransition′(t)`, and the capstone's `generation` field was updated to match. Formerly: substitute the smooth *time ramp* (`smoothPointerRamp`) into the protocol. Not needed for A2 (the ramp is not a phase-space function); its consequence is that the capstone's `generation` field gains the time-dependent factor `smoothTransition′(t)`, trading the constant-generator form for a window-free one. A deliberate trade, not an omission | > | ~~B2~~ | **DONE 2026-08-04** (`RecordLayer/NullSeamLift.lean`, 3 pins) | ★★ `nullSeamLiftClosure` — the third horn on `T² × ℂℙ²`, real dimension 6. Giving the register its conjugate coordinate removes the parity obstruction; the construction is otherwise unchanged (the crossing reads `θ₁`, `θ₂` rides along as a conjugate variable should), so every transfer is a cylinder argument and Born stays **exact**. ⚠️ *Earned:* the parity obstruction. *Not earned, not claimed:* the symplectic **form** — still §2a-scoped (A4); even dimension is necessary, not sufficient. Formerly: — earns the symplectic vocabulary the third horn currently cannot claim (`S¹ × ℂℙ²` is 5-dimensional, odd). Same kissing-crossing construction; the seam/Born theorems should carry over | > | ~~B3a~~ | **DONE 2026-08-04** (`RecordLayer/PointerFrequency.lean`, 2 pins) | ★ `pointer_born_frequency` — the ε-Born **frequency** layer: i.i.d. trials of the smooth witness's preparation converge a.s. with the limit inside `[rⱼ−2ε, rⱼ+2(N−1)ε]` (`pointerSectorProb_mem_window` carries the `ℝ≥0∞` sandwich across `toReal`). Small exactly as scoped: LF1's `freq_tendsto_of_iid` is already generic over (space, probability measure, event), so only the transport was new. The limit is **bracketed, not pinned** — the ε-horn's price; `ε → 0` sharpens the window only *across* witnesses, since each `ε` is a different propagator | > | ~~B3b~~ | **DONE 2026-08-05 — brick 2 landed the same day** (`RecordLayer/PointerLudersMarginal.lean`, 6 pins), **closing the row and with it the "smooth-witness Lüders composition" extension**: ★ `pointerRelocate_measurePreserving` (the owed piecewise invariance — `measurePreserving_of_partition` over record cylinders + the no-record piece, each cylinder fixed by its own slot swap exactly because the relocation never moves the pointer), ★ `pointerLudersStroke_measurePreserving` (the **whole two-stroke composite conserves Liouville** — collapse as relocation on the smooth horn), `pointerProtocol_outcomeSector` (sector = the brick-2b propagator's record-cylinder preimage), ★★ `pointer_luders_marginal` (**the Lüders theorem**: conditioned on the outcome-`i` sector — a base cylinder, the bank plays no part in outcome selection — the post-stroke system marginal IS the slot-`i` calibration), ★ `pointer_luders_born` (follow-up statistics = the collapsed state's Born weights, any context field), ★★ `pointer_luders_born_prep` (on the witness's own `pointerPrep`: `2ε < rate i` makes the conditioning non-vacuous via the ε-Born lower bound — records **and** update on one arena). *Sizing postmortem:* the warning "not a wiring job" was right about the shape but the cost had already been paid — `cond_prod_cylinder` and `Measure.map_eval_pi'` (swap tranche) plus `measurePreserving_of_partition` (staged exactly for this) carried both halves, and `swap_luders_marginal`'s proof skeleton transported nearly verbatim. One new snag for the ledger: unification of `pointerRelocate_of_record`'s implicit arena point whnf-exploded through `couplingUU` (the matrix exponential) — fixed by stating the equation with an explicit `show` goal first so the implicit is solved structurally, never by reduction (same family as the B5-geom module-system defeq notes). Scope unchanged and stated in-module: rank-one only, one bank per measurement, the composite is not a `MeasurementProtocol` (relocation is a triggered map; its Hamiltonian generation stays the same recorded extension as the swap witness's), and the ε lives in *which* outcome occurs — the conditioned marginal itself is exact. Formerly: **BRICK 1 LANDED 2026-08-05** (`RecordLayer/PointerLuders.lean`, 3 pins) — the composed arena `(Σ × ℂℙ^N) × bank`, the readout `pointerIndex` off the record regions, the record-triggered `pointerRelocate` (★ never moves the pointer, so the record survives its own relocation), ★ `pointerBankSwap_measurePreserving`, and the two-stroke composite `pointerLudersStroke` defined. **Brick 2 owes:** piecewise invariance of `pointerRelocate` itself (the partition argument `swapG` uses, with record cylinders for register arcs) **and** the conditioned post-measurement marginal — the actual Lüders theorem. ⚠️ *Sizing note 2026-08-05:* the conditioning is where the swap and join witnesses' difficulty actually lived (`swap_luders_marginal`, `join_luders_marginal` are both large), so brick 2 is **not** a wiring job and should start from a clean context. The staged `Mathlib/MeasureTheory/PiecewisePreserving.lean` is the likely tool for the invariance half. Until then this is arena-and-dynamics, not collapse. Formerly: — the genuinely large half, and not a wiring job. The smooth witness *provably does not collapse* (`pointerEvolve_base_marginal_unchanged`), so composition needs a **new arena** carrying pointer *and* bank (`KSigma × Pointer × bank`), a two-stroke protocol whose second stroke is triggered by the **pointer record** rather than a torus arc, and the conditioning computation for the post-measurement marginal. That is the scale of the original swap/join witnesses | > | ~~B3, formerly~~ | superseded by B3a/B3b | Smooth-witness Lüders composition + the ε-Born LLN. The LLN half is the smaller one: the i.i.d. frequency layer already exists and applies to arena measures | > | ~~B4~~ | **DONE 2026-08-04** (`RecordLayer/PovmSectorBorn.lean`, 2 pins) | ★★ `povm_sector_born` — the join protocol's **outcome sector** at the dilated preparation carries exactly `⟨ψ, Eᵢψ⟩`, i.e. the *dynamical* statement the prose had wrongly attached to the selector-level theorem; `povm_sector_born_canonical` for every POVM. A two-line composition of `join_sector_born` with `sum_block_normSq_dilate` — which is itself the evidence that the audit's finding was a **description** defect, not missing mathematics. Formerly: (the shape `join_sector_born` has), discharging the scope note the 2026-08-04 audit added to `PovmDynamics.lean` | > | ~~B5-geom~~ | **DONE 2026-08-05** (`RecordLayer/NoRecordGeometry.lean`, 4 pins) | ★ `noRecord_subset_closure_strict` (the core, over an arbitrary index set of record moments — one proof serves the all-`j` and pair consumers), ★ `noRecord_subset_closure_interior` (the row's stated lemma), ★ `recordRegion_pair_compl_regular` (the exact `hreg` shape `posMeasure_noRecord_of_correlates` consumes), and ★★ `posMeasure_noRecord_pointer` — **B5 closed outright**: the third leg instantiated on the pointer with *no* geometric hypothesis left. The recorded construction was correct and was used verbatim (`feedReady` = feed weight into the ready component; both branches of the family; the strict drop via `div_lt_div_of_pos_left`). The recorded snags were dodged as prescribed: no `set`/`rep` binding — `induction q using Projectivization.ind` gives the representative with zero defeq debt — and both limit branches done by hand (`tendsto_one_div_add_atTop_nhds_zero_nat` + `Tendsto.smul_const`/`Complex.continuous_ofReal`, no `simpa`). One *new* snag class, recorded for the next module: under the module system the elaborator would not delta-unfold `feedReady`/`recordRegion` during unification **when metavariables were present** (`Function.update_self` as a term proof failed where `rfl` succeeded on the identical goal) — the working idioms are `rfl`, `show`-then-exact, typed `have`s to cross set-membership/inequality defeq, and explicit `f`/typed `(0 : Fin (K+1))` arguments to `Finset.add_sum_erase`. Estimated one focused hour; took about that. Formerly: The single remaining hypothesis of B5: `noRecord_subset_closure_interior`, i.e. `{q | ∀ j, mⱼ₊₁ q ≤ ½} ⊆ closure {q | ∀ j, mⱼ₊₁ q < ½}` on `Pointer K`. **The construction is settled and correct — feed weight into the ready component, leaving all record components untouched, so every record numerator is fixed while the norm strictly grows and every record moment strictly drops.** Family: `c n = if v 0 = 0 then ((1/(n+1) : ℝ) : ℂ) else ((1 + 1/(n+1) : ℝ)) • v 0` — scaling in the second branch **keeps the phase**, which is what makes convergence immediate and avoids any chart argument. Perturbed vector `w n = toLp 2 (update (ofLp v) 0 (c n))`. **Two supporting lemmas compiled clean and are worth rewriting verbatim:** `update_zero_succ` (record components unchanged, via `Function.update_of_ne (Fin.succ_ne_zero j)`) and `norm_sq_update_zero_gt` (norm strictly grows, via `EuclideanSpace.norm_eq` + `Real.sq_sqrt` + splitting the sum with `Finset.add_sum_erase` at `0`). The strict-drop step also worked (`LF4.momentMap_mk` + `div_lt_div_of_pos_left`). **WHERE IT SNAGGED — elaboration, not mathematics:** (i) binding the representative with `set v := q.rep` creates a defeq mismatch later (`v 0` vs `(Quotient.out q).1.ofLp 0`) that breaks the final `Tendsto` — **use `obtain`/plain `let`-free style instead**; (ii) the ℂ-valued `Tendsto c atTop (𝓝 (v 0))` needs the two branches done by hand rather than `simpa`, since `tendsto_one_div_add_atTop_nhds_zero_nat` does not line up with `1 + (↑n+1)⁻¹` under `simpa`. Closing the ray-limit uses `Projectivization.continuous_mk'` composed with `tendsto_subtype_rng.mpr`, then `mem_closure_of_tendsto`. Estimated one focused hour from a clean context | > | ~~B5~~ | **DONE 2026-08-05** — closed outright: the geometric hypothesis is discharged (B5-geom row above, `RecordLayer/NoRecordGeometry.lean`) and ★★ `posMeasure_noRecord_pointer` states the leg on the pointer with every hypothesis either the witness's own dynamics or a theorem — **exact-a.e. records force Dirac calibration, as a theorem**. Scope unchanged: this is the *local* leg; general exhaustiveness over all arenas stays research. (Earlier same-day state:) **DONE-MODULO-GEOMETRY 2026-08-05** (upgraded from PARTLY DONE the same session — the forcing step is now proved: `exists_noRecord_of_meets_two` inverts `no_everywhere_correlation`, and `posMeasure_noRecord_of_correlates` chains it to the bound. **All dynamical hypotheses discharged**; one *geometric* hypothesis remained — that the no-record set is contained in the closure of its interior, true of moment regions but not then constructed. Original entry:) **PARTLY DONE 2026-08-04** (`RecordLayer/SharpenedNoGo.lean`, 2 pins) — ★ `posMeasure_noRecord_of_isOpenMap`: an **open-map** propagator on an **open** ready set cannot hide the no-record set in a null set, because the image is an open neighbourhood and any neighbourhood of a boundary point meets the no-record **interior**. That is exactly why the Dirac calibration escapes. ⚠️ **Sharpened, not closed**, and two things are explicitly *not* proved: (a) the *forcing* step — that some ready state must land in `closure (interior R)` — is a hypothesis here, needing `no_everywhere_correlation`'s connectedness plus a regularity condition on the no-record set; (b) the **quantitative** `O(δ)` fattening I had asserted in `NullSeamWitness.lean` is **withdrawn** — it needs an estimate on the landing moments, not soft topology, and has been corrected at source. Formerly:: continuity + a positive-width ready region force a positive-measure no-record set, so exact-a.e. records force Dirac calibration — the trilemma's third leg as a theorem | > | ~~B6~~ | ⚠️ **The retirement was LIFTED 2026-09-19** (author decision, `CONVENTIONS.md` §7): upstreaming is a numbered row again, one pull request at a time, written in this repository first and pushed nowhere without approval per step — ▶ OPEN QUEUE #37, with pull request 1 already backported here (`future-work.md` HY-7). This row stays struck because its *batch* framing is not what is happening; the work is per-file, not a batch of 67. **Formerly RETIRED 2026-08-06 (author decision, `CONVENTIONS.md` §7): Mathlib PRs, while nice, do not advance CSD and are not a need of this repository — off the todo list, not deferred.** The staging discipline is untouched (Cat-1 files stay CSD-free in `CsdLean4/Mathlib/` as code hygiene), the option §7 keeps open stays open, and the readiness record below stands so that a future decision to upstream starts from a green sheet rather than a scoping exercise. Formerly: **M** — **now the queue head**: the 2026-08-04 deprioritisation ("should not jump ahead of B3b/B5") is spent — both closed 2026-08-05/06 — and F1's interface pass plus the readiness scan below removed the known blockers | Mathlib upstream batch 1 — was **67** staged files (the row's earlier "68" was a one-off; recounted 2026-08-06 via `git ls-files`). ⚠️ *Question answered 2026-08-04 — they are NOT residue.* The ~25 `QuantumInfo/Reversible/` arithmetic files are **live corpus**: 26 references from the root module and five `Empirical/QM/Measurement*Adder*` modules import them. The ecdsa.fail extraction removed the ECDLP track; the reversible-arithmetic circuits stayed because the QM-side measurement-adder work is built on them. So the batch is genuine, none droppable. **Readiness scan 2026-08-06 (post-F1):** (i) **zero corpus imports** — no staged file imports `LF*`/`SigmaLayer`/`Empirical`; the staging discipline held; (ii) ~~one real namespace blocker~~ **CLEARED 2026-08-06**: `CSD.StoneC1.*` renamed to `Matrix.StoneC1.*` (dotted top-level names kept — the SMul-diamond note in the file explains why no `namespace` block), all corpus consumers + AxiomAudit pins updated in the same commit, full library green; the five other `CSD.` hits are docstring cross-references, edited at PR time; (iii) **naming clean** — the staged tree's (E) entries are all compliant camelCase (long-but-descriptive), no theorem-style def names; (iv) `Reversible/` no-API rows = 0 after F1, so the batch goes out wearing the interface mathlib review will demand | > > ### C. ~~Yours, not mine~~ — **MOVED OUT OF THIS REPOSITORY'S QUEUE 2026-08-06** > (author decision, same stroke as B6: paper work is not a need of the csd-lean4 > repository; it is tracked with the author's writing, outside this queue. The rows > stay below as the *record of what the papers owe the corpus*, because the corpus is > the source of truth for the divergences — but nothing here is repository work.) > > | # | Task | > |---|---| > | ~~C1~~ *(off-queue)* | **Paper C A7** — amend base-only `Ωᵢ(M) ⊂ ℂℙ^{N-1}` to the fibred ontic-sector reading | > | ~~C2~~ *(off-queue)* | **Paper C A2 / TN6 alignment** — the papers sit *behind* the corpus, and the gap has **grown** since this row was written: the measurement fork is a **trilemma** with its third leg now a theorem (B5), the smooth witness's boundary is fibrewise **and it now carries a Lüders update** (B3b), POVM Born is **sector-level** (B4 discharged the selector-only boundary this row still names), and mixed-degenerate updates exist at every rank (D3a). A reader coming from the papers is told something the corpus has corrected several times over | > | ~~C3~~ | **RESOLVED 2026-08-06** (`v1.1.0-finite-qm-closed` tag notes + README History table): `v1.0.5-trilemma`'s pre-audit annotation stands as published history (never rewritten); `v1.0.6-audited` supersedes it in the tag graph and the README History row for `v1.0.6` now records that supersession in prose, so a reader of the release history meets the correction without archaeology. Formerly: Decide whether `v1.0.5-trilemma`'s annotation (written pre-audit) should be superseded in the release notes; `v1.0.6-audited` already supersedes it in the tag graph | > > ### D. Guard and hygiene residue > > | # | Size | Task | > |---|---|---| > | ~~D1~~ | **DONE 2026-08-06** (`check-claims.sh` (8b), the scope-expiry ledger) | Every open-scope file now carries the BACKLOG row its boundary waits on (`DECLARED_SCOPE_WAITS`), or `none` for permanent/rowless boundaries; when a tagged row is struck `~~DONE~~`, the check **fails** until the site is superseded at source. Negative-tested (striking D3 fires on `MixedLuders.lean`). §E items got stable IDs E1–E5 so long-horizon waits are taggable. *Motivating case found while building the ledger:* `PointerGeneration.lean` still said the Lüders composition "is a recorded extension, not delivered here" a day after B3b delivered it — fixed at source same commit. Honest residue: `none`-tagged boundaries stay untracked by construction (the improvement is that every boundary is now *classified* and the tracked ones expire mechanically); the (8) KNOWN-LIMIT note is narrowed, not deleted. Formerly: Close check (8)'s documented blind spot: it fires when a boundary claim is added or removed, **not** when the fact beneath a claim that stays put changes. Candidate: cross-reference each open-scope site against BACKLOG rows marked DONE | > | ~~HY-6~~ | ~~**S–M, gated on the next Mathlib pin move**~~ done (2026-10-12 + 2026-10-23) | **The pin-move rename table** (Compat canary 2026-09-14, master `cf6bde1` / Lean `v4.34.0-rc2`; `validation-hardening-plan.md` row M carries the log). Everything master breaks outright is fixed with both-tree spellings (2026-09-14, six drifts, row M's log has each); what remains is deprecation warnings whose replacement names do not exist on the pin, so they cannot be applied without leaving it. **COUNTS COME FROM [`scripts/pin-move-inventory.sh`](../scripts/pin-move-inventory.sh), not from this cell** (added 2026-10-01 with row 27): the lists below were measured once by hand from the canary run and the corpus has grown since, so the script measures them instead and flags any row that finds nothing — on this corpus a zero means a wrong pattern far more often than a clean sweep. Numbers in brackets are the 2026-10-01 measurement. When the pin advances to Lean `≥ 4.34` / Mathlib `≥ 2026-08-29`: (1) Lean core renames, a word-boundary sed over `CsdLean4/`: `if_pos`→`ite_eq_left`, `if_neg`→`ite_eq_right`, `dif_pos`→`dite_eq_left`, `dif_neg`→`dite_eq_right`, `if_true`→`ite_true`, `if_false`→`ite_false` (~~~1,300~~ **1,564** sites in 141 files; the new names have the SAME signatures); (2) imports: `Mathlib.Data.Complex.Basic`→`Mathlib.Basic.Complex.Basic` (~~6~~ **7** files), `Mathlib.Data.Complex.BigOperators`→`Mathlib.Basic.Complex.BigOperators` (~~`CV/Position.lean`~~ **2 files**: `CV/Position.lean` and `Mathlib/LinearAlgebra/Matrix/TwoLevel.lean`, the second written after the canary), `Mathlib.Data.Real.Basic`→`Mathlib.Basic.Real.Basic` (`SigmaLayer/ConstraintSurface.lean`), `Mathlib.Data.ENNReal.Basic`→`Mathlib.Basic.ENNReal.Basic` (`LF2/MeasureBridge.lean`); (3) `ContinuousLinearMap.lipschitz`→`lipschitzWith` — ⚠️ **sweep `.lipschitz`, not the qualified name**: every site is dot notation on a `ContinuousLinearMap` term, the qualified spelling appears nowhere, and a sed on it would have matched nothing and reported success (~~×2 + 1~~ **5 sites**: `Analysis/ODE/FlowDerivative.lean` ×4, `SigmaLayer/UntriggeredVolume.lean` ×1), `Filter.eventuallyEq_set`→`eventuallyEqSet_iff` (`Probability/IidClockRace.lean`, `RecordLayer/JoinLuders.lean`), `isQuotientMap_mk'.secondCountableTopology isOpenMap_mk'`→`isOpenQuotientMap_mk'.secondCountableTopology` (`Projectivization/MeasureSpace.lean`); (4) the real rework: `stdSimplex` (a `Set`) became the structure `StdSimplex` — 15 occurrences restated against the new type, and ~~all in `Instances/ProjectiveSpaceMomentMap.lean`~~ **in three files**: `Instances/ProjectiveSpaceMomentMap.lean` (7, including the declaration `momentMap_mem_stdSimplex`), `LF4/MomentMap.lean` (1) and `Tests/AxiomAudit/MathlibStaging.lean` (4 — **axiom pins, whose names change with the declaration**, so the sweep touches the audit as well as the proof). Not a word-boundary sed: the name occurs inside the longer identifier too; ~~(5) delete the `MapProbability.lean` shim and collapse its call sites (row M's standing instruction) — ~~five~~ **27 occurrences of `isProbabilityMeasure_map'` across 19 files**, plus **12 of `Measure.map_smul'` across 10**, the other shim from the same canary;~~ ▸ **DONE 2026-10-13 (#142)**, and the counts measured at the time were **30 of `isProbabilityMeasure_map'` across 22 files** and **11 of `Measure.map_smul'` across 10** — a third drift of this table, in the same direction as the others; (6) drop `all_goals try rfl` at ~~`FlowDerivative.lean:397`~~ **`FlowDerivative.lean:913`** if master's `convert` still closes the goal. Verify each sweep with an untruncated grep that returns empty (row M's sweep discipline). ▸ **(4) DONE 2026-10-23** (row ~~27~~): the G9 theorems restated against the bundled `StdSimplex` (`momentMapSimplex_surjective`, `range_momentMap` as `Set.range StdSimplex.weights`), eleven pins, the inventory's section (4) reads zero by design. **Every section of this table is swept.** | > | D4 | **record only** | Check (7) is **structurally blind to `born_*` names** (independent Opus review, 2026-08-04) — `nullSeam_born_left/right` assert Born for statements containing no preparation, ray or moment map. Not closable as an inventory: there are **278** born-named declarations. Recorded rather than pretend-closed | > | ~~D2~~ | **DONE 2026-08-06** (`check-claims.sh` (7d), the abstract-Sigma instantiation inventory) | Every declaration whose **result type** is `KahlerOnticSetup` must appear in `DECLARED_KAHLER_INSTANTIATIONS` with its concrete `Sigma` and a verified EVEN real dimension — set equality, like (7a). The signature scan strips parenthesised binder groups so consumers `(d : KahlerOnticSetup N)` are exempt (they inherit their argument's parity and assert nothing). Six sites in-tree, all read from their `Sigma :=` field and all even (`trivialKahlerOnticSetup`, `unitaryFlowSetup`, `rotationSetup`, `manyToOneSetup`, `manyToOneRotationSetup`, `manyToOneSchrodingerSetup`). Formerly: Close check (7)'s: it cannot see an odd-dimensional *instantiation* of an abstract `Sigma` (`KahlerOnticSetup` and friends carry no parity claim of their own) | > | ~~D3a~~ | **DONE 2026-08-06** (`RecordLayer/MixedJoinLuders.lean`, 4 pins) | Degenerate-on-mixed: ★ `mixed_join_sector_born` (the mixed **block** Born rule, `∑_{k:bk=i} Tr(ρ|eₖ⟩⟨eₖ|)`, via the block spectral bridge `spectral_block_born_eq_traceForm` — sum interchange + one rank-one bridge per block member), ★ `mixed_join_post_bayes` (Bayes on a degenerate outcome, same `cond_finsetSum` engine as rank one), ★★ `mixed_join_luders` — **block Lüders composed with Bayes**: the conditioned system marginal is the Bayes mixture of the per-component block posts `epistemicMeasure [Πᵢψⱼ]`, at rank ≥ 2 genuinely distinct (the record does NOT erase classical ignorance — the density-operator update `ρ ↦ ΠᵢρΠᵢ/Tr(ρΠᵢ)` as a mixture, contrasting `mixed_luders_followup`'s rank-one vertex collapse). Rode `JoinClosure` exactly as `MixedLuders`'s scope note predicted — "mechanical" was accurate here. Stated under `hproj` (every component meets block `i`); the zero-likelihood filtered refinement is bookkeeping, recorded in-module. `MixedLuders`'s scope note superseded at source; expiry ledger re-tagged | > | ~~D5~~ | **DONE 2026-08-13** (the `check-contradictions.sh` shortlist, worked as the guard's own header instructs — "an advisory reading list, not a gate") | The 8 flagged identifiers re-read. **Seven benign**, all in the guard's documented false-positive classes: filename-as-identifier (`BACKLOG.md`); window-slicing on same-statement pairs (`isFubiniStudyKahler` — `NonTrivialSetup` and `ManyToOnePillars` say the identical thing); a correction-narrative line (`CircleFibre.lean` vs the 2b′ retraction block); scope-compatible pairs (`SectorData` — HY-3's doc sweep vs the Gates placeholder TODOs; `PointerInvariantOn` — proved on `ShearWitness`, deliberately a hypothesis in the `RecordPersistence` interface; `productSector_hasFubiniStudyPushforward` and `FiniteQMClosure` — proved-with-a-known-open-extension, the guard's own pre-classified example class). **One real, and the guard earned its keep**: `TheoremTargets.lean`'s B7 docstring said `BornFromFlow` "is proved from this together with a Birkhoff hypothesis in concrete models" — **nothing in the corpus proves `BornFromFlow`**, conditionally or otherwise (the audit pin at the `UniqueErgodicity` block and that module's own header both say so); the exact "prose stronger than the statement" class the necessity audit catalogues. Fixed at source with a dated correction note; T3/L7 stay open (= the SO-1 face). *Guard lexicon improved in the same stroke:* `not proved` added to the open-token list — `UniqueErgodicity`'s "What is NOT proved here" was classifying as SETTLED via the bare `proved` token (negation-blindness); such lines now carry both tokens and are skipped as nuanced. The shortlist now reports exactly the 7 benign pairs, by design — the guard has no allowlist, and its instruction is to fix the wrong document, never to soften truthful lines into agreement | > | ~~D3b~~ | **DONE 2026-08-12** (`RecordLayer/NullSeamGeneralN.lean`, 5 pins) | ★★ `nullSeamGenClosure` — **the third horn at every `N ≥ 2` and every weight vector**: continuity, measure invariance, exact records off an `N`-point null seam, **exact** Born mass `r i` per cell; `nullSeamGenClosure_uniform` the concrete `1/N` witness. **The construction departed from the recorded per-boundary-rotation sketch, deliberately**: gluing `(j, j+1)`-plane rotations cannot be full-matrix continuous at the wrap (monodromy — the product of `N` quarter-turns is not `1`), so the landing amplitudes are **plateau tents** (`max 0 (gap − infDist to the closed cell)`, cells as closed balls at the corpus's own `cellMid`, `gap = min r / 2`: at most one foreign tent nonzero anywhere, equal tents exactly at the kiss) and the unitary is a **single amplitude-polynomial rotation** `I − f₀f₀ᵀ − aaᵀ + af₀ᵀ − f₀aᵀ` (orthogonal for every unit `a ⊥ f₀`, `seamRotationR_orthogonal`) — continuity reduces to continuity of the tents. Reused: `loSum`/`cellMid`/`rep` (the corpus's own cumulative-cell layer — no new sum machinery), `volume_closedBall`, the two-cell circle toolkit, the `IsOpenQuotientMap` continuity descent and skew-product invariance verbatim. Two-cell witness unchanged as the minimal exhibit; its "general N recorded" scope line superseded at source. Original row: General-`N` for the null-seam witness. "Mechanical" is right about *content* (no new mathematics) but not size: the two-cell witness is an 827-line bespoke construction (circle toolkit, sign/angle profile, the `(1,2)`-plane kissing rotation, per-side landing, seam-null, per-side Born, continuity descent, skew-product invariance), and general `N` re-derives essentially all of it with cyclic index bookkeeping. Construction notes so nothing is re-scoped from scratch: cells = arcs at cumulative weights `c_j = Σ_{k≤j} r_k`; **N seams**, one per consecutive-cell boundary; at boundary `j` the same kissing-crossing rotation but in the `(j, j+1 mod N)`-plane of `ℂℙ^N` (consecutive records kiss at `m_j = m_{j+1} = ½` — `momentMap_add_le_one` gives the geometry at every `N` already); the profile is a piecewise-glued angle function, continuous because each piece's plateau matches its neighbours at the seam; cell-`j` Born mass = `r_j` exactly (arc measure); seam-null = `N` points. A dedicated session's job, from a clean context | > > ### E. Longer horizon (unchanged in content; given stable IDs 2026-08-06 so the > scope-expiry ledger in `check-claims.sh` (8b) can name them) > > | # | Item | > |---|---| > | ~~E1~~ | ~~Shor-9 concatenation and the Lindblad tier~~ **DONE 2026-08-13** — see the Q5 row (ShorNine.lean + LindbladSemigroup.lean); CP of `e^{tℒ}` stays on the GKSL construction row as the Mathlib-scale residue | > | E2 | The operator-convexity → DPI → SSA ladder (**L**, still gated on the `CStarMatrix` instance wall) | > | E3 | Composable finite-key QKD (**XL**) | > | E4 | The CV programme | > | E5 | ⚠️ *Restated 2026-09-05 — "the two research frontiers" was partly stale.* The fibre-from-dynamics question was **retired in its mixing form** (2026-08-24, mis-specified) and re-executed honestly (2026-08-27, `ShearDeIsolation.lean`); what survives of it is Posit 1's discharge condition — derive `IsTorusGenerated` from the de-isolation dynamics — which is **open foundations**, not a queued brick (`specs/POSITS.md`, "What 'frontier' means here"). General trilemma exhaustiveness (`Q13`) is unchanged and open. | > > ### F. Library-grade quality programme (adopted as standard 2026-08-06 — `CONVENTIONS.md` §9) > > Author decision 2026-08-06: the corpus aims for **library-level code** — mathlib review > standards on the `Mathlib/`-staged tree, the same bar with documented physics-notation > exceptions elsewhere (the Lean-QIT/Physlib route). Motivated by Ilin & Nugent > (arXiv 2606.13925); measured by `scripts/check-review-surface.sh` against > `docs/review-surface-baseline-2026-08-05.txt`; rules in `CONVENTIONS.md` §9. Ordered by > forcing function, then cost-effectiveness. **F1 is the only externally-forced item**; a > mathlib reviewer at B6 will demand exactly what it delivers, so it precedes B6. > > | # | Size | Task | > |---|---|---| > | ~~F1~~ | **INTERFACE PASS DONE 2026-08-06; one named residue** | **`Reversible/` API pass, delivered**: interface lemmas across eleven files — per-constructor lemmas (`gateCost_X/CX/CCX/swap`, `gateWires_*`), per-field wire lemmas for the concrete layouts (`modAddLayout2_*` ×8, `modSubLayout2_*` ×6, `cModAddLayout2_*` ×10, `modReduceLayout2_*` ×6, `cuccaroMulModLayout3_*` ×6), value-level lemmas for the parameterised banks (`cBank_*_val` ×8, `wAdd_*_val` ×10, `wDbl_*_val` ×8, `wMulLoop_bank`/`_X_val`), the bank identifications (`constMulLayout0/1/2/8_bank`), and `andSumPass_def`/`layerWF_iff`. All non-`@[simp]`; no existing proof touched; full library green. **Success metric MET: `Reversible/` (B) rows with `api=0` = 0** (corpus-wide no-API 25 → 8, the remainder being the `private`-exempt `boxSet` + seven `through≤2` tail items). *Build note:* two structure literals (`modAddLayout2`, `cBank`) had their `where`-blocks split by a mis-anchored insertion mid-pass — caught by the build, fixed before commit; the anchor for such insertions is the LAST `h`-field, not the last data field. ~~**Named residue (clean-context item, §9.4):** extract stage lemmas from `cuccaroModAdd_spec`~~ **Residue CLOSED 2026-08-06 by documented decision, not refactor**: the full proof was read end-to-end and assessed — every step is a one-to-three-line citation of a named block lemma (`cuccaroAdd_correct`, `cuccaroSub_correct`, `maskCopy_*`, the `*_preserves_*` frames), so the 313 lines are the irreducible bookkeeping of nine stages × five tracked invariants; per-stage invariant records would add ~150 lines of statement boilerplate for indirection without content. Verdict recorded in the theorem's docstring (which also replaced a misleading legacy "Placeholder spec" doc). **F1 is closed; B6 is unblocked on the interface front** | > | ~~F2~~ | **DONE 2026-08-06** | Witness-scalar and plumbing interfaces landed: `alphaOff`/`betaOff` value lemmas + `deltaOff` (def + all four setting-pair values), `aAngle_def`/`bAngle_def`/`baseAngle_def` (`CGLMPQudit`), `om2_def` (`MUB3Volume`), `innerProd_def` (`SternGerlach`), `joinArenaMeasure_def`, `paramMeasure_def`, and branch lemmas `joinFstAux_coe_of_ne/_of_eq`, `blockCollapseAux_coe_of_ne/_of_eq` (the exact case splits their unfold sites re-derive). All non-`@[simp]` so no existing proof changes behaviour; unfold sites left in place (converting them is on-touch churn, §9.4 spirit). **`boxSet` deliberately skipped**: it is `private` — §9.1's local-plumbing exemption, recorded here rather than silently dropped. Full library green | > | F3 | **S — decisions taken 2026-08-06 (`CONVENTIONS.md` §9.2); renames land on-touch** | **Naming standard.** Decided: mathlib naming strict in `Mathlib/`-staged files; literature notation blessed in physics modules where it mirrors the cited source (the `HardyKahler` `A'`/`B'` family — ten (E) rows — is *compliant* under this rule, docstring citation required); theorem-style def names (`*_realisable_for` ×9, `b92_encode`, `P_st`) are defects of form, renamed when next touched, never swept. (E) tracks; no sweep row exists on purpose | > | F4 | **standing rule, no sweep** | **Rule of two** (`CONVENTIONS.md` §9.3): at a second near-use, generalise the existing single-use lemma instead of cloning. The 1,109-row (C) list is the where-to-look index; obligation `Prop`s are exempt by design (docstring must name the discharging witness). Per-file top offenders (`WignerRigidity` 24, `ShorCore` 21, `CanonicalTypicality` 17, `Grover` 16) get attention on touch only | > | ~~F5~~ | **DONE 2026-08-06** | Auditor refinements landed: (a) (A) excludes pinned defs and first-letter-case closure twins (119 → 111 at the time; the 8 exclusions verified individually — 6 direct pins, 2 twins incl. `smoothWitnessClosure`); (b) baseline v2 captured (`docs/review-surface-baseline-2026-08-06.txt` — metric v2 + post-F1/F2 corpus; the 08-05 file kept as the pre-standard record) with the §9.5 discipline in the script header; (c) non-blocking unchanged. Known residual: (B) cannot see `private` (the `boxSet` row is §9.1-exempt but still listed) — acceptable noise, noted here rather than special-cased | > | F6 | **record** | **Ratchet discipline live from now**: any landing that increases the (B) no-API count or adds theorem-style def names states so with justification in its commit message. First exercised retroactively this week: brick 1's `pointerLudersStroke`/`pointerLudersMeasure` entered the lists on 2026-08-05 and left them when brick 2 consumed them — the intended lifecycle for definition-shaped debt: visible, short-lived, paid | > > ### G. External-review intake follow-ups (codex 2026-08-06 — classification + verification > evidence in `audit-sweep-plan.md`'s intake section; these rows are the open remainder) > > | # | Size | Task | > |---|---|---| > | ~~G1~~ | **DONE 2026-08-06** (transport-theorem route, author decision; `LF2/Preparation.lean`, 2 pins): `MeasureBridgeData.integral_comp_pi` (∫ f∘π dμL = c.toReal • ∫ f dμFS — `bridge_eq` extensionally consumed) + `OperationalPackage.fromPreparation_liouville_apply` (Liouville-prepared operational probability IS the μFS-integral at c = 1); the "Symmetry enters via the bridge argument" docstring over-claim corrected to cite them; `check-semantic-mutations.sh` now REQUIRES the transport theorems (deletion = re-review trigger); CL-003 → qualified. Foundational-triple verified in isolation. Formerly: **F-01 — the LF2 `bridge` argument.** `OperationalPackage.fromPreparation` takes `bridge : MeasureBridgeData` type-level only (deliberate, documented; verified 2026-08-06: zero `bridge_eq` consumption sites corpus-wide). Decide: drop the argument, or state a transport theorem that extensionally consumes `bridge_eq` (an ontic-volume computation carried into the projective probability). Either way, fix `born_rank_one_direct`'s docstring line "Symmetry enters via the `bridge` argument" — it asserts load-bearing status the term dependency does not support. Current state pinned by `scripts/check-semantic-mutations.sh` (CL-003) | > | ~~G2~~ | **DONE 2026-08-06**: `EffectFn.lean` retitled "the quadratic-form representation layer" with an explicit F-02 scope note (Born-from-volume citations redirected to the LF4 engine); all "volume-ratio" phrasing in `EffectFn`/`Preparation` renamed/reworded. Formerly: **F-02 — docstring precision in LF2.** Reword the "volume-ratio foundational" framing in `LF2/EffectFn.lean` + `LF2/Preparation.lean` (`effectProjFn` is `Re(v†Ev)` by definition — a representation layer); reserve "Born from volume" for the LF4 engine (`MomentBornN`, `BornRegionUncond`). Same genus as the 2026-06-11 item-2 fix | > | ~~G3~~ | **DONE 2026-08-06** (tighten-keep-name, author decision): the abstract `Prop` pairs replaced by concrete fields `kahler_pointwise : IsFubiniStudyKahler N` + `liouville_isProbability : IsProbabilityMeasure liouvilleMeasure` (the latter an instance — `IsProbabilityMeasure S.liouvilleMeasure` now automatic); all three instances (`trivialKahlerOnticSetup`, `unitaryFlowSetup`, `manyToOneSetup`) carry identical obligations; consumer `unitaryFlowSetup_liouville_isProbability` kept as thin projection for pin stability; PLACEHOLDERS.md / connectivity-manifest L1 / reconstruction-status updated. Manifold residual (dω = 0, top-power) unchanged, honestly open at L1. Formerly: **F-04 — `KahlerOnticSetup` interface.** Consumers quantified over the structure cannot extract the Kähler laws from `kahler_condition : IsKahlerSector` (instances all supply the proved `IsFubiniStudyKahler` core — verified; the weakness is interface-level, not instance-level). Decide: rename per the reviewer (`ProjectiveMeasureFlowSetup`) or tighten the `Prop` pairs to the concrete pointwise-Kähler interface now that `IsFubiniStudyKahler` exists; full strengthening stays gated on Mathlib exterior calculus (connectivity L1) | > | ~~G4~~ | **DONE 2026-08-06** (`RecordLayer/SwapLuders.lean`, 1 pin): ★ `swap_luders_iff_calibrated` — post-outcome-`i` system marginal `= τ` ⟺ `ν i = τ`, both directions of "the update is calibration-encoded" as one theorem (forward = `swap_luders_marginal`, converse = the new content); CL-025 updated. Foundational-triple verified in isolation. Formerly: **F-05 — minimal calibration theorem.** `swap_luders_marginal` gives calibration ⇒ Lüders; the new content is the converse: which apparatus hypotheses are *equivalent* to Lüders behavior on the swap witness. Makes "the update is calibration-encoded, not assumption-free" a theorem instead of a scope note | > | ~~G5~~ | **DECIDED 2026-08-06** (author) | **F-06 — capstone naming: keep names, rename on-touch (F3 class).** `FiniteQMClosure` / `ProjectiveMeasurementCapstone` keep their names; the docstrings carry the witness-index framing, and any future touch renames per the F3 convention (never swept). The L-size genuinely-unified closure (single arena/preparation/dynamics/measurement interface) stays a *listed candidate*, deliberately not queued next to the record-layer frontier | > | ~~G6~~ | **DONE 2026-08-06** (root half): the 34 root-unreachable non-test modules added to `CsdLean4.lean` (default root now reaches every non-test module — re-verified 0 missing; full library green, 3809 jobs). Reverse-layer imports (16 edges) stay on-touch per LF4-todo §10 — deliberately not swept. Formerly: **F-07 — root repair.** Add the 34 root-unreachable non-test modules (verified list: the SigmaLayer pointer/record tranche + `Empirical/CSD/Eraser*` + `HamiltonianVectorField`/`L2OpNormEntry`) to the `CsdLean4` root, or adopt the reviewer's generated per-layer aggregates; the 16 reverse-layer imports move down-layer on-touch per LF4-todo §10 (no bulk refactor). Audit coverage is intact meanwhile (union of the four roots reaches all 443) | > | ~~G7~~ | **DECIDED + DONE 2026-08-06** | **`ci.yml`: `windows-latest` leg dropped** (author decision — local development is on Windows, so Windows breakage surfaces immediately anyway; the CI leg roughly doubled cost on the slowest runner class). The reviewer's script wiring is kept: the three new checks join the blocking guard step, the vacuity/contradiction/review-surface trio runs as the advisory step | > | ~~G8~~ | **DONE 2026-08-06** (`CsdLean4/Headlines.lean`) | **Headline facade module landed**: imports exactly the 27 modules carrying the 30 `validation-claims.tsv` claims, with a full-name **drift guard** (30 top-level `example := @…` lines — a rename/namespace-move/deletion of any headline breaks the build) and a new `check-validation-ledger.sh` block enforcing ledger↔facade import sync. Immediate payoff: the guard caught **four wrong ledger constants** on creation day (CL-025/026/027 recorded `CSD.SigmaLayer.*` for `CSD.RecordLayer.*` theorems; CL-001 missed the `OnticSetup.TrialModel` prefix) — all fixed in the tsv. Imported by the root; full library green. Root rename to `All` NOT taken (kept `CsdLean4` exhaustive; the facade is additive). Formerly: the good half of the codex session's step-4 rearchitecture suggestion | > | ~~G9~~ | **DONE 2026-08-06** (`Tests/AxiomAudit/` ×8 + umbrella) | **AxiomAudit split by layer, landed**: all 1,709 pin blocks re-sorted by pinned constant's namespace into eight part files (`Foundations` 49, `EmpiricalQM` 174, `EmpiricalCSD` 302, `LF4` 216, `Dynamics` 138, `SigmaLayer` 370, `MathlibStaging` 408, `Extensions` 52 — count-invariant verified), each importing the `CsdLean4` root so placement never affects resolution; the umbrella keeps the `CsdLeanTests` target and coverage story unchanged. **Measured: the full pin sweep now runs in ~26 s wall on a built tree** (parts parallel, largest 14 s) vs the ~13-minute serial monolith; layer-local gate = `lake build CsdLean4.Tests.AxiomAudit.`. `check-axiom-imports.sh` updated to collect pins from the parts. Codex session step 5 | > | G10 | **record (declines, 2026-08-06, author decision)** | **Rearchitecture sweep items assessed and NOT adopted:** (a) SigmaLayer role split (codex step 3) — real point, 111 mixed-role files, but file moves across the active frontier layer churn imports/pins/docs mid-programme; **deferred to a consolidation release** when the record-layer tranche stabilizes; (b) corpus-wide shrinking of `@[expose] public section` — consumer-facing benefit delivered by G8 at zero risk; de-exposing changes cross-module defeq (the B5-geom/pointer-unification trap class) — **declined as a sweep**, at most a CONVENTIONS decision for new modules on-touch; (c) reverse-layer import sweep (codex step 2) — stays extract-on-concrete-consumer per LF4-todo §10, **no bulk refactor** (reaffirmed) | > | ~~G11~~ | **MOVED OFF THIS QUEUE 2026-08-07** (author decision — owner-tracked with the author's own review work, same stroke as §C; not repository work) | **CL-024 residue: human hand pass** on the five highest-risk Wigner tactic blocks named by the 2026-08-06 audit — `sign_link_core` (the global-sign crux), `diagReducedMap_fixes_two_level_general`, `diagReducedMap_fixes_three_level`, `two_level_imrelphase_of_flips` + the `cnorm_±I_sq` conventions, `diagReducedMap_complexSign_all`. ~~The up-to-phase uniqueness formalization~~ **DONE 2026-08-06**: matrix form already existed (`PhaseRigidity.lean`, missed by the audit — both-ways correction recorded in the intake); the `projMap`/`conjProj`-vocabulary form landed in `WignerUniqueness.lean` (`exists_unit_smul_of_projMap_eq` + antiunitary twin via `conjProj_conjProj`, 3 pins). The classical Wigner statement is now complete in-corpus (existence + uniqueness + downstream exclusivity); the hand pass alone promotes CL-024 qualified → validated | > > ### H. Quantum-chaos workstream (adopted 2026-08-06 — one-repo architecture, `external-library-map.md`) > > | # | Size | Task | > |---|---|---| > | ~~H1~~ | **DONE 2026-08-06** | **Architecture layer**: `specs/external-library-map.md` (ownership table, four-way classification rule, single-provider rule, target topology, toolchain facts), charter §"Repository architecture", `CsdLean4/Interop/` (documentation-only until the first dependency). One repo confirmed; no satellite CSD repos | > | ~~H2~~ | **DONE 2026-08-07** (`Incubator/QuantumChaos/FloquetInterface.lean`, 5 pins + 2 `projMap` functoriality pins) | **`FloquetEvolution` interface landed.** Four-way check done first against the real repositories: Physlib (`leanprover-community/physlib`, 4.32.0, 690 modules) has `FiniteHilbertSpace d` = a structure *wrapping `EuclideanSpace ℂ d`* (adapters will be `.val`-thin) but **zero Floquet/kicked/chaos modules** → class 3 confirmed, `upstream-candidate(physlib)`. Interface: `step : H ≃ₗᵢ[ℂ] H`; `iterate n = step ^ n` with ★ exact information-preservation (`norm_iterate_apply`, `inner_iterate_iterate`); induced ray dynamics `projStep`/`projIterate` (functorial via the new `Projectivization.projMap_refl`/`projMap_trans`; transition-probability preserving — the `wigner_rigidity` hypothesis); inhabitation (`refl`) + the matrix-dynamics adapter seam `ofUnitary` (via `toEuclideanLinearEquiv` + `inner_toEuclideanLin_unitary`). New audit part `Tests/AxiomAudit/Incubator.lean`. All foundational-triple-only | > | ~~H3~~ | **DONE 2026-08-07** (4 modules, 7 pins) | **The pilot vertical result, landed.** `Incubator/QuantumChaos/KickedIsingPilot.lean`: the explicit two-qubit kicked-Ising Floquet unitary (Ising phase × kick⊗kick, group-multiplication membership; `kronecker_mem_unitaryGroup` = upstream-candidate(mathlib)) + ★ `kickedIsing_changes_marginal` — at `b = π/2` the reduced first-qubit state of the evolved `\|00⟩` flips `\|0⟩⟨0\| → \|1⟩⟨1\|` while `inner_iterate_iterate` keeps every global overlap exactly invariant (the "global information intact, local accessibility moved" clause). `Empirical/CSD/QuantumChaos/{OnticLift,RecordPersistence,Capstone}.lean`: the ontic lift `floquetOnticStep` (= the constant-family `manyToOneSetup` time-1 flow — measure preservation inherited, not re-proved) with ★ period-by-period lift equations; **sure** (set-invariant, not a.s.) record persistence under uncoupled post-record driving; ★★ `floquetPilotClosure` — the four universal clauses bundled for EVERY unitary `U` and base point, honest witness-index framing. All foundational-triple-only. Open continuation (thread, not pilot): coupled post-record driving (record erasure pricing), KI-model reindex to `Fin 4` for the ontic-lift clause, chaos diagnostics | > | ~~H5~~ | **DONE 2026-08-07** (`Empirical/CSD/QuantumChaos/RecordDegradation.lean`, 3 pins) | **Coupled post-record driving priced — the record half-life bound.** The honest degradation quantity identified first: for measure-preserving coupled steps the record EVENT's probability never changes (that IS measure preservation); what degrades is STABILITY (the trajectory's readout surviving). ★★ `recordIntact_compl_measure_le`: a formed record survives `n` periods of coupled driving except on a set of measure `≤ n·ε`, `ε` = the per-step coupling set's measure (least-change covering + iterate measure preservation) — the discrete-time sibling of `collapse_accuracy_bound`'s erasure pricing. Null coupling ⟹ a.s. persistence; the pilot's uncoupled regime recovered as `ε = 0` (empty coupling set, sure persistence). **Next brick (queued): a small-coupling witness `0 < ε < 1`** (a weakly record-coupled drive on `(KSigma N) × Rec`, e.g. fibre-conditioned record kick) showing the bound bites; then diagnostics (Loschmidt first) and the KI `Fin 4` reindex | > | ~~H6~~ | **DONE 2026-08-07** (3 bricks, 7 pins) | **The §H5 continuation trio, landed.** (a) **Small-coupling witness** (`Empirical/CSD/QuantumChaos/CouplingWitness.lean`): `triggeredRecordKick` — record kicked exactly when the system enters a trigger region; measure-preserving by Mathlib's skew-product theorem; **the coupling set IS the trigger cylinder**, so coupling strength = trigger measure. CSD instantiation `fibreTriggeredKick`: quarter-ball fibre-arc trigger, coupling **exactly `1/2`** (`AddCircle.volume_closedBall`) — `0 < ε < 1`, the half-life bound genuinely bites (★ `fibreTriggeredKick_record_halfLife`: survival except on `≤ n/2`). (b) **Loschmidt echo** (`Incubator/QuantumChaos/Diagnostics.lean`, class-3): `loschmidtEcho` with endpoint/unit-interval/symmetry API; CSD reading fixed in-module — both drives preserve every overlap exactly, so echo decay measures RELOCATION of preserved information, never loss. (c) **KI `Fin 4` reindex** (`kickedIsingU₄` via `reindex_mem_unitaryGroup`, upstream-candidate(mathlib)): ★ `kickedIsing_pilotClosure` — the concrete model reaches the ontic-lift clause at `Fin 4`, every parameter pair and base point. All foundational-triple-only. Thread continuation: attainment (actual erasure rate vs the bound), SFF/OTOC diagnostics | > | H4 | **record** | Root-wire every new module (the job-count tell); `Empirical/CSD/QuantumChaos` rows inherit the transport/ledger discipline; guard inventories (`check-claims.sh`) get their rows in the same commit that adds boundary phrases | > | H7 | **LANDED 2026-08-12 (same day as scoping) — one honest residue below** (`CV/CarrierPersistence.lean`, 4 pins + `Empirical/CSD/QuantumChaos/CarrierPersistence.lean`, 3 pins) | ★★ `carrier_persistence_window` — **the level-1 headline: zero until the cone arrives, the derived Duhamel rate after** (`≤ 2(n−m)·‖W−1‖·‖B‖` for a perturbation outside the readout's `m`-ball). ★ `heisenberg_perturbed_pow_eq` — exactness in the cone-complement is an EQUALITY, for an *arbitrary* unitary perturbation. ★ `heisenberg_diagonal_pow_eq` — **einselection of the configuration basis**: diagonal readouts are eternally intact under the whole diagonal-phase family (a discovery of the build: in this model degradation can ONLY enter through non-diagonal perturbations, which sharpens the physical reading). Level 2: `recordEvent_preimage_step` (events reindex, never decay — `[propext, Quot.sound]` only) + `recordEvent_measure_invariant` + `recordIntact_antitone` (**axiom-free**) — the event/carrier separation as theorems, with the framing constraint honoured throughout. **Residue DISCHARGED 2026-08-12 (Q1, `DerivedCoupling.lean` — the Markov bridge): the ontic derived-ε instantiation** — a concrete record-coupled ontic step in the `fibreTriggeredKick` style whose `recordFlip` measure is *derived from* the level-1 operator bound rather than posited; needs the operator-norm → flip-measure bridge (genuinely new, not assembly). Entropy-ledger follow-up unchanged. Original scoping: **Carrier persistence priced by locality (the CV/record bridge brick — scoped 2026-08-12).** The first statement where a record carrier's stability is priced by field locality rather than posited. ⚠️ **Framing constraint, binding (2026-08-12 discussion):** every clause stated present-tense-checkable; the theorem family is about CARRIERS (readout-cylinder membership), never "the past changing" — the record EVENT is trajectory-indexed and time-invariant by construction, and information is conserved outright (overlaps exact). Say "carrier persistence", never "record degradation". Four clauses: **(a)** event invariance — the sector-selection indicator is a function of the trajectory, constant under evolution (trivial-but-clarifying lemma, ontic side); **(b)** information conservation — cite `inner_iterate_iterate`, don't re-prove; **(c)** THE new content, two halves: *(c1, exact)* a readout supported on site set `T` evolves IDENTICALLY under two drives that differ only on couplings at graph distance `> n` from `T`, for `n` periods — records in the cone-complement are exactly intact (assemble from `heisenberg_graphInteractingU_pow_supportedOn` + `commute_of_disjointSupport`; the OTOC gate `otoc_graphInteractingU_eq_zero` is the same geometry at commutator level); *(c2, bounded)* after cone arrival the readout deviation is `≤ 2n·\|τ\|·\|λ\|·C` at the Duhamel rate (`interactingU_dist_le`, the `one_sub_loschmidtEcho_interacting_le` pattern), with the sharp distance form via `norm_commutator_spatial_factorial_le`; **(d)** operational fixedness — the isolated carrier's pointer-basis repeatability, cite `csd_repeatability_same` (finite-arena; scope note that the CV-mode version rides the `OscillatorBorn` truncation). **Level 2 (ontic lift):** feed the derived (c2) bound into `recordFlip`/`recordIntact_compl_measure_le` following the `fibreTriggeredKick` pattern — `ε` DERIVED from `(τ, λ, C, d)` instead of posited; the §H5 half-life bound with locality doing the pricing. **Follow-up recorded, not queued:** the entropy-ledger link (carrier erosion rate vs coarse entropy production, framed strictly as retrodiction reliability) — same austerity constraint applies | > > ### ~~V. Validation-hardening campaign~~ (adopted AND CLOSED 2026-08-12 — [`validation-hardening-plan.md`](validation-hardening-plan.md)) > > **ALL 13 WORKSTREAMS DONE 2026-08-12, single-day campaign.** Validation machinery > on the **sufficiency/inhabitation leg** (explicit nontrivial inhabitants of the > assumption packages, driven through the headline chains) — not reconstruction > progress, and distinct from the necessity-audit conditionality gap. Landed: > `--wfail` CI (A) · `Tests/Witnesses/` suite, 8 modules + 22 pins, all > foundational-triple (B, C, D, E+H, F+G, I, J — incl. the honest > `Measure.infinitePi` trial model for **every** `OnticSetup`, discharging the > Examples.lean caveat) · `check-import-hygiene.sh` blocking + 3 permanent > check-guards probes (L) · advisory `check-std-lint.sh`, 25 s (K) · > `compat.yml` forward-compat + macOS (M, O) · `docs.yml` doc-gen4 smoke (N). > Residue: none blocking; the first scheduled compat/docs runs validate their > runner plumbing; std-lint's `unusedArguments` list is the F3 on-touch class. > > --- > > ## What's next, by size & complexity (updated 2026-08-02, post-v1.0.1 + second review) *(HISTORICAL — see the live list above)* > > | Size | Item | Notes | > |---|---|---| > | ~~S~~ | ~~Conditioning toolkit → staging; standard-basis lemma~~ | **DONE 2026-08-02** (`Mathlib/Probability/ConditionalProbability.lean`; `basisContext_basisFun_rate`) | > | ~~**M**~~ | ~~Mixed preparations in the dynamical model~~ **DONE 2026-08-03** (`RecordLayer/MixedSwap.lean`, 2 pins): `mixedSwapPrep ρ = Σⱼ λⱼ • swapPrep [ψⱼ]` (two-stage sampling as a probability measure) and ★ `mixed_swap_sector_born` — the mixed **dynamical** Born rule: the mixture's mass on the protocol's outcome sector is exactly `Tr(ρ\|eᵢ⟩⟨eᵢ\|)`. Per-protocol facts inherited verbatim; the Bayes-conditioned mixture post-ensemble stays a recorded extension | discharged | > | **M–L** | **Dynamical no-signalling (A6 step 4) + the dynamical eraser process** | A *local* measurement on a composite IS a block-degenerate measurement; `joinWitness_blockLuders` + `reduceB` machinery is most of the route. **Brick 1 LANDED 2026-08-03** (`RecordLayer/LocalLuders.lean`, 4 pins): the coordinate-form local projectors and ★★ `reduceA_localLuders_mixture` — the Born-weighted mixture of post-measurement `A`-marginals equals the pre-measurement `A`-marginal (the statics core, zero-weight branches honest). **Brick 2 LANDED 2026-08-03** (`RecordLayer/LocalBlockBridge.lean`, 3 pins): ★ `toComposite_blockProj` — the degenerate-Lüders block projector of `localBlock` (= second projection under `finProdFinEquiv`) IS the local projector, definitionally; ★★ `reduceA_blockLuders_mixture` — **dynamical no-signalling** stated in the join witness's own vocabulary (the post-states and weights of `BlockLudersObligation localBlock`, inhabited by `joinWitness_blockLuders`). **Brick 3a LANDED 2026-08-03** (`RecordLayer/LocalLudersBasis.lean`, 5 pins): the local Lüders map for an **arbitrary** orthonormal basis on the measured factor — ★★ `reduceA_localLudersOn_mixture`, marginal invariance in **every basis** (Alice cannot detect Bob's outcome or his basis choice; Parseval replaces the ite-collapse). **Brick 3b LANDED 2026-08-03** (`Empirical/CSD/EraserDynamics.lean`, 6 pins): **the dynamical eraser** — mark arm: `localProjB_bellE` (which-path product state) + `marked_no_fringe` (flat screen rate at every phase); erase arm (an instance of `localProjOn` at the genuine ± `OrthonormalBasis`): ★ `erased_amp` — the dynamical post-state's screen amplitudes are **exactly `√2·eraserOut`**, so the `QuantumEraserVolume` fringes, conditional rates (`erased_rate`), the exact dark zero (`erased_dark`), and the `1/2` marker weights (`erased_weight`) are all statistics of the measurement dynamics' own output. **ROW CLOSED 2026-08-03.** Residue discharged same day (`Empirical/CSD/EraserSequential.lean`, 3 pins): ★ `sequential_no_revival` — mark-then-erase stays flat at every phase and outcome: **once a record exists, erasure cannot revive the fringe** (records are statistically irreversible; the erase stroke only rescales the recorded ray, `seqProfile_eq`). The measure-level ensemble integral is **closed as definitional**: for finite outcomes the post ray-ensemble IS the discrete mixture and its barycenter statement IS `reduceA_localLudersOn_mixture`; a Bochner-integral recast adds a `reduceA`-measurability lemma and no content — revisit only if a continuum-outcome layer appears. *(Superseded scoping note:)* **the dynamical eraser process** on the `2 ⊗ 2` composite (`nA = nB = 2`, so `localBlock` applies as-is). (a) *The mark arm*: the computational-basis local marker measurement — post-states `[localProjB j ψ_φ]` are the which-path states, flat screen statistics, no fringe; (b) *the erase arm*: the `±`-basis marker measurement = `localProjB` conjugated by the local Hadamard (`actA`/`actB` idiom from `OnticMarginals`) — post-states identified with `QuantumEraserVolume`'s `eraserOut φ ±` (via `eraserOut_eq_jointAmplitude`), marker weights `1/2` (`eraser_marker_marginal`), conditional fringes and the exact dark zero inherited from the existing typicality theorems; (c) follow-up statistics through `csd_sequential_born`. The measure-level "one integral" form of no-signalling is bookkeeping over bricks 1–2 (the post ray-marginal is a finite discrete mixture) and folds into (c) | > | ~~**M–L**~~ | ~~Projective-measurement capstone~~ **DONE 2026-08-03** (`RecordLayer/MeasurementCapstone.lean`, 1 pin): ★★★ `projectiveMeasurementCapstone` — rank-one + every basis + degenerate + **the smooth horn** (four closures, one Prop; the fields quantify over different witnesses by design — the two-horn framing). The single citation for dynamical projective measurement; constituents remain as the construction record. *+`generation` field added 2026-08-03 (fourth review): the Schrödinger ODE is now a capstone field, not a satellite* | the review's step 4, discharged | > | ~~**M–L**~~ | ~~POVM / instrument dynamics~~ **DONE 2026-08-03** (`RecordLayer/PovmDynamics.lean`, 5 pins): exactly the planned route — Naimark-dilate, then the *existing* degenerate record machinery on the dilated arena (the ancilla block structure IS `localBlock N K`, so no new dynamics, sectors, or arena were needed). ★ `povm_selector_born` — the POVM Born rule **at the selector level** (⚠️ *corrected 2026-08-04: this row said "dynamical"; the statement is an `epistemicMeasure` of a selector fibre, not of a protocol outcome sector — the sector-level lift is a recorded extension*): the block selector's outcome-`i` fibre at the dilated preparation `[Vψ]` carries exactly `⟨ψ, Eᵢψ⟩` (`degenerate_selector_born` + `born_transfer`); ★ `toComposite_blockProj_dilate` — the record-layer block posts ARE the Naimark–Lüders posts `Πᵢ(Vψ)`; ★★ `povm_instrument` — the join witness's post-marginals satisfy degenerate Lüders at `[Vψ]` (the instrument of the dilation, by Liouville-preserving dynamics); ★★ `naimarkInstrumentClosureCanonical` — via `canonicalNaimark`, **every POVM** has dynamically realised statistics and instrument. Honest scope: the instrument is dilation-relative (a POVM does not determine its instrument — physics, not a defect); realising `V` itself as a unitary-plus-ancilla stroke inside the record dynamics is a recorded extension | discharged | > | ~~**S–M**~~ | ~~Outcome-conditioned mixed update~~ **DONE 2026-08-03** (`RecordLayer/MixedLuders.lean`, 3 pins + staged `ProbabilityTheory.cond_finsetSum` — Bayes for finite mixtures, hypothesis-free): ★ `mixed_post_bayes` — the conditioned post-ensemble IS the Bayes-posterior mixture (component `j` at weight `λⱼ·pᵢ|ⱼ/Tr(ρ\|eᵢ⟩⟨eᵢ\|)`); ★★ `mixed_luders_followup` — **the record, not the pedigree, fixes the post-state**: follow-up statistics after outcome `i` on the mixture are exactly `c'.rate [eᵢ]` (at rank one the record erases the classical ignorance; `ρ ↦ ΠᵢρΠᵢ/Tr(ρΠᵢ)` dynamically). Spine: `mixedSwapPrep` **factors** through the common bank (`mixedSwapPrep_eq_prod`), so the pure `swap_luders_born` applies verbatim; positivity is a theorem (`mixed_outcome_pos`). Degenerate-on-mixed = recorded extension (rides `JoinClosure`; posteriors genuinely distinct at rank ≥ 2) | > | ~~**M**~~ | ~~Degenerate one-protocol package~~ **DONE 2026-08-03** (`RecordLayer/JoinClosure.lean`, 2 pins): ★★ `degenerateMeasurementClosure` — ready/record/exclusivity/persistence + Liouville + ★ `join_sector_born` (the coarse dynamical Born mass at the canonical join preparation, calibration-independent; spine = `preimage_sector_ae` + `volume_goodTheta`, the Dirac slice of `degenerate_selector_born`) + ψ-dependent degenerate Lüders, one structure on the join protocol for every block structure and preparation. The capstone's `degenerate` field upgraded to this closure (was bare `BlockLudersObligation`) — the fourth review's remaining ⚠️ on the degenerate tier is now ✅ | > | ~~**M**~~ | ~~`C^∞` ingredients for the ε-corridor witness~~ **DONE 2026-08-03** (`RecordLayer/PointerSmoothProfile.lean`, 3 pins): `smoothClampDiv`/`smoothArcWeight`/`smoothPointerRamp` — the `Real.smoothTransition` profiles with the **plateau interface verbatim identical** to the trapezoids' (same statements, same hypotheses: `= 1` on the shrunk arc, `= 0` off the open arc, `0`/`π/2` around the stroke). ★ `contDiff_smoothArcWeight_lift` — the arc weight's periodic lift is `C^∞` on the universal cover (both circle-distance kinks fall inside plateaus: centre needs `2ε < r`, cut locus needs `r < 1`; the transition zone sees a locally affine distance); ★ `smoothRampedU_schrodinger` — **Schrödinger at every time**, no open-window restriction (the trapezoid corners are gone, exactly as `PointerGeneration`'s honest scope predicted; outside `[0,1]` the ODE reads `U̇ = 0` — persistence as an ODE). Manifold-level arena smoothness stays §2a-scoped with A1/A3; protocol re-instantiation with the smooth profiles is mechanical (interface identical) and recorded | > | ~~**S–M**~~ | ~~Cantor-horn witness~~ **DONE 2026-08-03** (`RecordLayer/NullSeamWitness.lean`, 4 pins) — and **simpler than the devil's-staircase sketch**: no Cantor function; the transition dodges the record gap by crossing **where the record regions kiss** (`m₁ = m₂ = ½`), touching the complement at one projective point per boundary. ★★ `nullSeamClosure` — continuous (`continuous_nullSeamEvolve`), measure-preserving (⚠️ *not* "Liouville": `S¹ × ℂℙ²` is odd-dimensional hence not symplectic — naming corrected 2026-08-03, fifth review; the `T² × ℂℙ²` lift that would earn the word is the row below), records exact and correct off a **two-point** seam (`nullSeam_seam_null`), **exact** Born `r`/`1−r` (`nullSeam_born_left/right`), on `S¹ × ℂℙ²`. **THE TRILEMMA**: each horn pays exactly one price — *seams* (piecewise), *ε-Born* (smooth witness), *Dirac calibration* (this one: exactness at `[f₀]`; positive-width ready fattens the seam to `O(δ)` — `collapse_accuracy_bound` prices it). General `N` = mechanical, recorded | > | **PARKED** (author, 2026-08-03) | Sharpened no-go (trilemma completeness) | **Parked before starting** — scoping found two different theorems wearing one title. (a) *Local* version, S–M and confidently provable: for the pointer geometry, continuity + a positive-width ready region force a **positive-measure** no-record set of order the calibration width, so exact-a.e. records force Dirac calibration — the trilemma's third leg, rigorous. (b) *General exhaustiveness* — no witness whatsoever combines continuity + positive-width ready + a.e.-exact records + exact Born — quantifies over all arenas and propagators; the IVT/Fubini mechanism leans on the specific moment-region structure and does **not** obviously generalise. Reopen with (a) if the trilemma needs to be a theorem rather than a classification; (b) stays research. Ingredient check done: `fs_volume_eq_dirichlet` (`LF4/MomentBornN.lean`) gives the absolutely-continuous moment pushforward, so the kissing set is FS-null and the argument must run on topology, not measure | > | **PAPER-FIRST** (was Research) | Joint-arena Hamiltonian lift — **proof strategy now known** | ⚠️ *Row rewritten 2026-08-03 (fifth external review): my earlier pessimism here was wrong.* I wrote that "once the register moves mid-stroke the exact moment-transport argument dies, and adiabatic/Grönwall control is needed." It does not die — **the weights are exact constants of motion**, no adiabatic approximation. The argument, verified: on a product arena `{f(x), g(q)} = 0`, so `{w_k, 𝓗} = Σⱼ hⱼ{w_k, wⱼ}` for `𝓗(x,q) = Σⱼ wⱼ(x)hⱼ(q)`; and the weights depend **only** on the moment coordinates `mᵢ(p)` and the single register coordinate `θ₁`. The `mᵢ` Poisson-commute (moment map of an abelian torus action — `ℂℙ^{N-1}` is the standard completely integrable system, the `mᵢ` a maximal commuting family), `{mᵢ, θ₁} = 0` (different factors), so **any** two functions of that family commute ⇒ `{wᵢ, wⱼ} = 0` ⇒ `dw_k/dt = 0`. **Stronger than the review states:** the same computation gives `{θ₁, 𝓗} = {mᵢ, 𝓗} = 0`, so `basinIndex` — a function of exactly those — is conserved: **the outcome selector survives the true joint flow untouched**, and back-reaction lives entirely in the conjugate variables (`θ₂`, the phases conjugate to the moments), which the record does not read. The existing landing/Born analysis should transfer nearly verbatim. **Why PAPER-FIRST and not a Lean brick:** every step (define `𝓗` on the arena, derive `X_𝓗 = ω⁻¹d𝓗`, prove `{wᵢ,wⱼ}=0`, conserve along the flow) is symplectic-manifold mathematics, and Mathlib has no symplectic-manifold API — the *same* §2a wall as A1/A3, already recorded for the vector-field equation itself. Paper C A2 / TN6 can carry this as mathematics **now**; the corpus records the scoped boundary. Steps 1–6 of the review's package are the paper; step 7 (a second Hamiltonian stroke implementing the reversible Lüders relocation) is the smooth-witness Lüders composition row. **Lean brick landed 2026-09-02** (`specs/frozen-base-obstruction-scoping.md` brick 3): `RecordLayer/JointLift.lean` — `jointLift c ε Δ`, a back-reacting `IsJointLift` **on the arena** for every conserved-data shift `Δ` (`isJointLift_jointLift`, so landing / ε-Born / moment-marginal transfer; `jointLift_base_moves_of_ne`), with arena-level Liouville preservation for every measurable shift **without disintegration** (`jointLift_measurePreserving`, via the CSD-free `MeasurePreserving.vadd_twist_of_invariant`); `RecordLayer/HamiltonianShift.lean` — the shift the stroke Hamiltonian generates, chart-level (`strokeCurve_hasDerivAt_hamiltonianField`, `hamiltonianShift_eq_strokeCurve_one`, regular data). What stays paper-first is exactly the manifold statement `ι_X ω = dH` on the arena (`R-016`) and nonvanishing of the shift | > | **M–L** (scoping done) | Coordinate-model Poisson bracket — the formalisable fragment | The partial Lean route for the row above, **scoped 2026-08-03**: `Mathlib/Analysis/InnerProductSpace/KahlerForm.lean` already delivers the *pointwise* Kähler triple (`g`, `ω`, `J`, J-compatibility, taming) axiom-free, and Mathlib has `fderiv` on normed spaces — so a **chart-level** Poisson bracket `{f,g} = ω(∇f, ∇g)` is buildable, with `{wᵢ,wⱼ} = 0` and weight conservation as concrete computations. What it would NOT give: `dω = 0` (KahlerForm's own honest scope), the global projective statement, or the flow's existence on the manifold — so the result would be a **local/chart theorem**, weaker than the paper's, but real and machine-checked. Worth doing if the trilemma's third leg needs Lean backing; not a substitute for the paper theorem | > | **M–L** | `T² × ℂℙ²` lift of the null-seam witness | Earns the word "symplectic"/"Liouville" for the third horn: `S¹ × ℂℙ²` is 5-dimensional (odd), so the current invariant measure is a product measure and nothing more. Lifting the register to the full `T²` gives even dimension and lets the same kissing-crossing construction be stated Hamiltonianly. Mechanical geometry; the seam/Born theorems should carry over | > | ~~**M** (long elapsed)~~ | Mathlib upstream batch 1 — **retired 2026-08-06**, see B6 | Tracked row below; `MATHLIB-GAPS.md` | > | **M–L** | Smooth-witness Lüders composition + ε-Born LLN | The recorded extensions of the COMPLETED ★ item: compose the pointer stroke with the swap/join relocation machinery on one arena (collapse stays on the exact horn until then), and run the i.i.d. frequency layer over the ε-Born sandwich | > | ~~**L**~~ | ★ ~~**Compact Kähler pointer witness (`ℂℙ^K`)**~~ **COMPLETE 2026-08-03** | The REOPENED Hamiltonian row's route, **CONFIRMED (author decision 2026-08-03: retain the smooth-Hamiltonian architecture)** — a genuinely (globally) Hamiltonian, continuous record-creating propagator; kills the torus-flux obstruction; the second review's steps 1–3. Execution plan: [`pointer-witness-plan.md`](pointer-witness-plan.md). **Bricks 0–1 LANDED 2026-08-03** (`RecordLayer/PointerArena.lean`, 5 pins — arena, open positive-measure ready/record regions, `arenaReady_pos`; `RecordLayer/PointerRotation.lean`, 6 pins — the continuous one-parameter unitary rotation group, ready → record quarter turn, FS invariance). **Brick 2a LANDED 2026-08-03** (`RecordLayer/PointerCoupling.lean`, 6 pins — Hermitian weighted coupling, unitary exponential propagator, ★ `pointerRot_eq_exp` generation identification by ODE uniqueness, entrywise-Lipschitz weight continuity via Duhamel + the new staged `Matrix.norm_entry_le_l2_opNorm`). **Brick 2b LANDED 2026-08-03** (`RecordLayer/PointerWeights.lean`, 5 pins — circle-intrinsic weight field, ★ `continuous_pointerEvolve`: **the full arena propagator is continuous**, the property `shearEvolve_not_continuous` forbids every torus-register witness; skew-product Liouville preservation; pure-cell reduction to the quarter rotation); **Brick 3 LANDED 2026-08-03** (`RecordLayer/PointerLanding.lean`, 6 pins — midpoint separation on the circle, exact moment-map record transport, ★ `pointer_landing`: records created by the continuous propagator with the ontic sector selecting the outcome, shrunk-slice volume = `rⱼ − 2ε` exactly); **Brick 4a LANDED 2026-08-03** (`RecordLayer/PointerProtocol.lean`, 7 pins — the smooth witness as a `MeasurementProtocol`: two-time law = exponential group property, persistence = freezing, correlation = landing, ★ joint time–state continuity); **Brick 4b LANDED 2026-08-03** (`RecordLayer/PointerBorn.lean`, 5 pins — the ε-Born sandwich `rⱼ − 2ε ≤ sector ≤ rⱼ + 2(N−1)ε` for the ready-conditioned preparation, no Dirac posit; ★★ `smoothWitnessClosure` — the smooth-horn closure, canonical moment-map instantiation). **Brick 5 LANDED 2026-08-03** (`RecordLayer/PointerGeneration.lean`, 3 pins — ★ `rampedU_schrodinger`: the Schrödinger ODE with the explicit Hermitian `H_eff`; records-without-back-reaction). **ALL FIVE BRICKS COMPLETE 2026-08-03 — the ★ item is DONE.** Eleven modules, ~40 pins. Remaining offshoots are their own queue rows: the smooth-witness Lüders composition (M–L) and the LLN over the ε-Born sandwich (M) | > | **L** | Shor-9 concatenation; Lindblad tier | QM-side QEC; open systems | > | **L** | Operator-convexity ladder → DPI → unconditional SSA | Gated on the `CStarMatrix` instance wall (`MATHLIB-GAPS.md`) | > | **L** (parked) | ~~Wigner §13 normal form~~ **struck 2026-08-30: not a gap** — `wigner_rigidity` closed 2026-07-02 and the named lemma was false as stated (`MATHLIB-GAPS.md`); Lévy concentration | Paused / Mathlib-gated; reopen only with cause | > | **XL** | Composable finite-key QKD | The security-model tranche | > | **Authorial** | Papers alignment | A7's fibred reading vs Paper C's base-only `Ωᵢ(M) ⊂ ℂℙ^{N-1}`; ~~the measurement-dynamics architecture~~ **DECIDED 2026-08-03: retain smooth Hamiltonian** (the `ℂℙ^K` witness satisfies the papers rather than the papers bending to the piecewise witness) — remaining: Paper C/D/TN6 edits | > | **Research** | The fibre-from-dynamics frontier | `sigma-fibre-contextuality.md`; the join/phase machinery is the first real tool | > | **Programme** | The CV track | After the compact-pointer witness and the capstone — the review's defensible closure point | > **This is the ONE place open work lives.** Do not add TODO / future-work / "next > steps" / open items to any other `.md`. The per-phase plans (`LF*-plan`, `shor-plan`, > `povm-plan`, …) are **historical execution logs** — frozen; read them for how a > completed layer was built, not for what is open. The status/claims docs > (`reconstruction-status`, `connectivity-manifest`, `PLACEHOLDERS`, `AXIOMS`, > `BRIDGE-OBLIGATIONS`) describe *what is proved*; they point here for *what is next*. > > Effort key: **S** hrs–day · **M** days–2wk · **L** weeks · **XL** Mathlib-scale. > Last updated 2026-07-30. > > **★ For "how far is finite QM from Σ from closed?" read [`reconstruction-status.md`](reconstruction-status.md) §2a** — the A1–A7 audit and the **scoping decision**. Short version: A1's exterior calculus **on `KSigma`** and A3's smoothness are **permanently scoped** (Mathlib has no manifold DG API; nothing downstream uses them); A4 is **closed as a posit** with `μ_FS` *forced* under unitary symmetry; and the genuinely open rows ~~are **A2, A5, A6, A7** plus the fibred-Σ question~~ *(banner corrected 2026-08-02: A5/A6/A7 were discharged and A2's formalisable half landed that day — see `reconstruction-status.md` §2a; the **Hamiltonian-generation** sub-question of A2 was then REOPENED by the second external review — torus-flux obstruction — and **CLOSED 2026-08-03 at the formalisable level**: the `ℂℙ^K` smooth witness landed, bricks 0–5, `smoothWitnessClosure` + `rampedU_schrodinger`; see the dynamical-measurement row)*. Do not read "Partial" rows as pending work. > > ⚠️ **CORRECTION 2026-07-30 — do not extend the "permanently scoped" label to the *fibred* Σ.** It > was being used that way, here and in `CircleFibre.lean`. The fibred Σ's missing Kähler structure is > a **dimension-parity** fact, not a Mathlib gap: `ℂℙⁿ⁻¹ × ℝ` and `ℂℙⁿ⁻¹ × AddCircle 1` both have > real dimension `2n-1`, odd, and no odd-dimensional manifold admits a symplectic — hence Kähler — > structure. No amount of tooling repairs that. It is **tractable work on `T²`**, not scoped work. > See the ★★ row below. --- ## M — genuinely closeable, real content | Item | Status / what's needed | Former source | |---|---|---| | ~~**Choi converse** (PSD Choi ⇒ Kraus)~~ | **DONE 2026-07-19** (`LF2/ChoiConverse.lean`). `choi_iff_posSemidef`: a matrix on `Fin M × Fin N` is a Kraus family's Choi matrix **iff** PSD. The feared "vectorization iso" was definitional (the Choi index *is* a product), so the content was the spectral `Kᵢ=√λᵢ·unvec(eᵢ)` reconstruction (`choiOfKraus_krausOfChoi` + `IsHermitian.eq_eigen_outer`). Foundational triple. | `qi-qec-roadmap.md` | | ~~**Gisin's theorem** (pure entangled ⇒ CHSH violation)~~ | **DONE 2026-07-19** (`LF6/GisinTheorem.lean`). `gisin_chsh_violation`: every entangled `Ψ(c,s)=c\|00⟩+s\|11⟩` (`0 2`. Built directly on the existing `psQubit_pauli_correlation`; the feared "general Schmidt decomposition" wasn't needed (the real-Schmidt two-qubit state + its correlation were already in `PartialSchmidtCorrelation.lean`). Trig-free `c,s`-dependent witness giving `2√(1+(2cs)²)`. Closes LF6-6. Foundational triple. | `lf6-plan.md` (LF6-8) | | ~~**Busch–Gleason** (effect-Gleason, finite-dim)~~ | **DONE 2026-07-21** — see the "Discharge `busch_effect_gleason`" row below. Axiom deleted; corpus now imports **zero** axioms. | `AXIOMS.md §2.2` | | ~~**Separate the ecdsa.fail track**~~ | **DONE 2026-07-20** — extracted to its own repository and removed from this one (see the section below). | `ecdsafail-two-track.md` | ## M — genuinely closeable, real content (continued) | Item | Status / what's needed | Former source | |---|---|---| | ~~**Born-as-volume at the simplex boundary (zero Born weights)**~~ | **CLOSED-AS-STALE 2026-08-02 (second external review).** The boundary extension this row asks for had **already landed 2026-06-11**: `fs_born_volume_ratio_N_uncond` / `born_frequency_convergence_N_uncond` (`LF4/BornRegionUncond.lean`) are hpos-free for every unit preparation, vanishing weights included, and the empirical engine routes through the hpos-free forms (`ContextVolume.lean`). The row was opened 2026-07-27 without noticing the six-week-old `_uncond` engine — a verify-before-scoping failure, now corrected in the HOM and eraser module docstrings (the record-layer route there stands as a **choice** with its own foundational motivation, no longer as a necessity). *(Original row, superseded:)* OPEN — found 2026-07-27 while building the Hong–Ou–Mandel CSD twin. `fs_born_volume_ratio_N` / `fs_born_volume_ratio_N_apex` / `fsMeasure_bornRegionN` (`LF4/MomentBornN.lean`, `LF4/ObservableCorrespondenceN.lean`) all carry `hpos : ∀ j, 0 < ‖⟨eⱼ,ψ⟩‖²` — **strictly positive** Born weights. It is load-bearing, not decorative: the region for index `i` is `replaceMap b i '' openSimplexFree` and `replaceMap_det b i = b i` (Cramer, `LF4/BornVolume.lean`), so a **zero** Born weight makes the vertex-replacement map singular, puts `b` on the simplex *boundary* (`b ∈ openSimplexFree` fails, since it needs all free coords `> 0` **and** the apex coord `> 0`), and breaks both the openness⇒measurability step (`LinearMap.equivOfDetNeZero`) and the volume-scaling step. **Consequence:** the projective/DH machinery cannot express *any* phenomenon whose signature is a vanishing amplitude — interference nulls, dark ports, HOM. **What's needed:** a boundary extension, `fsMeasure_bornRegionN_of_born_zero : ‖⟨e_k,ψ⟩‖² = 0 ⟹ fsMeasure p₀ (bornRegionN ψ hψ0 k) = 0`. Sketch: the region sits inside the preimage of `range (replaceMap b i)`, a *proper* subspace when the det vanishes, hence Lebesgue-null; push through `fs_volume_eq_dirichlet` by monotonicity of the (outer) measure, intersecting with `openSimplexFree` to satisfy that lemma's subset hypothesis. The care needed is that `replaceMap_image_subset` also assumes `b ∈ openSimplexFree`. **Workaround in place:** the **record layer** has no such restriction — `volume_cdfCell` carries no positivity hypothesis (a zero rate is simply a zero-width CDF cell), so `Empirical/CSD/HongOuMandelVolume.lean` states the degenerate case there instead. This is a concrete architectural argument for the record layer over the DH route, independent of the foundational one. | `Empirical/CSD/HongOuMandelVolume.lean` | | ~~**Error discretization** (why four Paulis suffice)~~ | **DONE 2026-07-27** (`Empirical/QM/QEC/ErrorDiscretization.lean`). `pauli_decomposition` (every `2×2` complex matrix is `c₀I + c₁X + c₂Z + c₃XZ`, coefficients read off the entries) + `pauli_span_top` (`span ℂ {I,X,Z,XZ} = ⊤`), lifted to the three-qubit code by `kron3` linearity (`error_discretization_qubit₁/₂/₃`) and to states (`errored_codeword_eq`). Dimension-free; the conceptual content that makes `ThreeQubit` + `PhaseFlip` a general error-correction claim rather than two special cases. **Discretization half only** — not a claim that the 3-qubit code corrects arbitrary errors (its correctable set is `{I,X₁,X₂,X₃}`); the syndrome-collapse half and the concatenated Shor-9 claim are the row below. | `EMPIRICAL.md` | | ~~**Syndrome collapse** (join discretization to the four point-checks)~~ | **DONE 2026-07-27** (`Empirical/QM/QEC/SyndromeCollapse.lean`). Before it, `ErrorDiscretization` ("an arbitrary error is a combination of four") and `ThreeQubit` ("each of the four is corrected") were **unconnected** — a superposition of error branches is not obviously reducible to one branch, so discretization was true but bore no load. `errored_pairwise_orthogonal`: the four errored codewords are mutually orthogonal, their supports being disjoint (`{000,111}`, `{100,011}`, `{010,101}`, `{001,110}`) — the concrete form of the distinct `(Z₁Z₂,Z₂Z₃)` syndrome pairs, available directly so no spectral theorem is needed. `branch_overlap_X1/X2/X3`: the overlap of the corrupted codeword with branch `k` is exactly `cₖ·⟪branchₖ,branchₖ⟫`, so the syndrome measurement reads off one coefficient and is blind to the rest. `three_qubit_corrects_span_error` bundles decomposition + orthogonality + extraction + recovery: **the code corrects an arbitrary error in `span ℂ {I,X₁,X₂,X₃}`** — a continuum, not four points. Still the bit-flip span (the code's actual correctable set); all four Paulis per qubit needs the row below. | `EMPIRICAL.md` | | ~~**BB84 sequential twin + rotated context**~~ | **DONE 2026-08-02** (`Empirical/CSD/Crypto/BB84Sequential.lean` + `RecordLayer/RotatedContext.lean` + `prep_outcome_pos` in `SequentialMeasurement.lean`; foundational-triple, 6 pins) — the crypto-twins gap from the empirical audit, closed at its most valuable point: the QM module's classical-marginal Eve becomes the calibrated-swap dynamics (`csd_sequential_born`), Bob reads the **rotated context** `basisContext b` (any orthonormal basis — the object the unitary-covariance extension quantifies over; the equivariance *law* is still the open item), and `hpos` is now a **theorem** at concrete preparations (`prep_outcome_pos`). Remaining from that audit note: ~~**B92/Wiesner sequential twins**~~ **DONE 2026-08-02** (`B92Sequential.lean` + `WiesnerSequential.lean`, 6 pins — unambiguity as a null basin + false clicks at `½`; the `¾` counterfeit value attained by measure-resend, optimality out of scope), and the fully dynamical **eraser process** (still needs the composite arena, not just the rotated context). | crypto twin 2026-08-02 | | ~~**QuantumEraser ontic twin (record route)**~~ | **DONE 2026-08-02** (`Empirical/CSD/QuantumEraserVolume.lean`, foundational-triple, 5 pins) — routed through the **record layer** exactly as the gating note below prescribed: the conditioned fringe as a fibre-typicality volume at every phase (`eraser_fringe_typicality`), the dark fringe an exact ontic null four ways (cell / P5 record / measurement / **v1.0 global basin** — at `φ=π` the conditioned state is the vertex `[e₁]`), rates derived as joint-over-marginal from `Empirical/QM/QuantumEraser.lean`. The `hpos` boundary was routed around, not extended — that row stands. **Remaining (recorded, not in-flight):** a *fully dynamical* eraser — marker measurement as calibrated-swap dynamics on the two-qubit composite, screen read via a rotated context field — needs the unitary-covariance extension step. **Effort L**, folds into the extension-steps row. | eraser twin 2026-08-02 | | ~~**KCBS ontic volume twin**~~ | **DONE 2026-08-02** (`Empirical/CSD/Contextuality/KCBSVolume.lean`, foundational-triple, 3 pins) — representative pentagon context completed by the cross product, orthonormality transported from the QM side, engine instantiated, apex weight `1/√5` realised (`kcbs_pentagon_weight`). *(Original queue entry:)* **QUEUED 2026-08-02 (empirical-coverage audit).** `Empirical/QM/KCBS.lean` exists with no CSD twin — the only flagship test missing one for no structural reason: pentagon measurements involve no vanishing amplitudes, so the `hpos` boundary does not bite. Build `Empirical/CSD/Contextuality/KCBSVolume.lean` on the volume-series pattern (nearest relatives: `ContextVolume`, `KS18Volume`). **Effort M.** Same audit also recorded: the **QuantumEraser** twin is *structurally gated* on the `hpos` boundary row (interference nulls = zero Born weights) — route it through the record layer as `HongOuMandelVolume` was, or wait for the boundary extension; and **BB84/B92/Wiesner** CSD twins remain planned-not-done (`specs/qm-empirical-tests.md`). | empirical audit 2026-08-02 | | ~~**Shor 9-qubit code (concatenation)**~~ | **DONE 2026-08-13** (`QEC/ShorNine.lean`, Q5/E1) — the assessment's concatenation-combinator option delivered: the block-Kronecker action lemma reduces all 512-dim statements to the landed 8-dim block facts, and the full single-Pauli correction set (X/Z/XZ at all nine positions, Z degenerately) is proven with no entry-wise work above 8 dims. The 9-qubit syndrome-collapse half is NOT restated (recorded in `SyndromeCollapse.lean`'s scope note). *(Superseded assessment kept below.)* Formerly: **OPEN — assessed 2026-07-27. Not conceptually blocked; blocked on idiom.** The syndrome-collapse half is now **done at 3 qubits** (row above), so what remains is purely the 9-qubit lift. Both halves already exist: `QEC/ThreeQubit.lean` (bit-flip) and `QEC/PhaseFlip.lean` (phase-flip, whose own docstring calls itself "the phase-flip half of the 9-qubit code"), and the discretization content is now landed (row above). The obstacle is dimensional: the corpus does QEC **entry-wise** on `Fin 2 × Fin 2 × Fin 2` (8-dim, 64 entries, `fin_cases`-friendly, 309 lines); Shor-9 is 512-dim / 262 144 entries and entry-wise tactics do not survive it. **What's needed:** either a **stabiliser-formalism layer** (Pauli group, (anti)commutation, syndrome as eigenvalue pattern — `PhaseFlip.lean` is already half-structural, and this would also unlock Steane-7 and CSS codes generally) or a **code-concatenation combinator** (inner code ∘ outer code, with the correctable set of the composite). The collapse argument itself is settled in form (`SyndromeCollapse.lean` at 3 qubits); it must be re-established at 9. **Effort:** L. The last iconic empirical item. | `EMPIRICAL.md`, `project_empirical_tests` | | ~~**Unconsumed declarations**~~ (found by `scripts/check-vacuity.sh`, 2026-07-28) | **RESOLVED 2026-07-28 for the three substantive cases; three harmless aliases remain.** The scan found six declarations nothing consumed. **The three `Prop` fields now have consumers:** `liftA_liftB_unitary_commute` (`LF3/Projectors/TensorModel.lean`) is consumed by the new `liftA_liftB_apply_comm` — local operations on independent wings commute pointwise, the kinematic core of no-signalling. ⚠️ *Correction: an earlier version of this row said that field's prose implied it was load-bearing. It did not — its own docstring already stated "`MeasurementUnitary.factorises` itself does not consume it". The corpus was honest; the row was unfair.* `A_right` / `B_right` (`LF3/SectorSeparation.lean`) are consumed by `pointerOverlapA_pos` / `_discriminates` (and the B-wing twins) and the capstone `pointerLeakage_nonvacuous`: **`A_wrong`/`B_wrong` alone are satisfied by a completely uninformative pointer** (set every overlap to 0), and it is `A_right`/`B_right` that force the pointer to register and to discriminate. So the two-sided reading is now supported by theorems rather than assumed. **Remaining (harmless, deliberately not deleted):** `bell_prep_compose`, `mzOutput`, `productMeasureBridge` — aliases and a bundled re-export, dead but not misleading. | `scripts/check-vacuity.sh` | | ~~**Adjudicate the contradiction-scan shortlist**~~ (`scripts/check-contradictions.sh`, 2026-07-28) | **EXECUTED AS D5, 2026-08-13 — this row was left unstruck by oversight and is struck at the 2026-08-20 triage; the D5 row above carries the full adjudication record (seven benign in the documented false-positive classes, one real defect fixed at source, guard lexicon improved).** *(Original row:)* ~~OPEN — 8 identifiers described as settled in one place and open in another.~~ ⚠️ **The scan has LOW PRECISION and does not catch its own motivating case** — see its header; both hits spot-checked were benign (`BornFromFlow`: proved *given* a Birkhoff hypothesis, open *because* Mathlib lacks it; `fs_born_volume_ratio_N`: proved, with a known open extension — the `hpos` boundary row above). Treat the list as reading material, not as defects. The eight: `IsKahlerSector`, `isFubiniStudyKahler`, `BornFromFlow`, `born_eq_P_st`, `fs_born_volume_ratio_N`, `productSector_hasFubiniStudyPushforward`, and two more the run prints. **What's needed:** read each pair and, where the two descriptions genuinely disagree, fix the wrong one rather than softening both until they agree. ✅ **One genuine contradiction adjudicated and corrected 2026-07-30** (found by external review, not by the scan — another data point on its recall): `MomentMapRace.lean` asserted "there is **no separate dynamical problem to solve**" while `DeIsolationFlow.lean:36-38` correctly states the open obligation to exhibit an `H_int(M)` generating the basins. `MomentMapRace` was wrong and its docstring is now corrected — what the moment map dissolves is the **statistical** residual (no stochastic postulate needed *given* the basins), never the **dynamical** one (`basin_rate` is a hypothesis field). | `scripts/check-contradictions.sh` | | ~~**★★ Is the fibred Σ a legitimate A1 ontic sector?**~~ *(successor to the parked A7 row)* | ✅ **ALL THREE SUB-ITEMS DONE — row closed 2026-08-25.** (a) partition on `KSigma`'s `T²` 2026-07-30, (c) the `ℝ` closure retired 2026-07-31, **(b) the fibre measure exhibited as forced rather than merely Haar 2026-08-25** (`kMuL_unique`). ⚠️ The headline below ("OPEN — now the load-bearing question, and nothing is pointed at it") was written 2026-07-29 and is **retained as the historical record only**. ⚠️ Residual, and it is a Mathlib wall not a CSD one: the *manifold* Kähler half (`dω = 0`, top-power volume) stays blocked — `Q8`, XL — which is why the measure result takes the symmetry reading. *(Original row:)* **OPEN — now the load-bearing question, and nothing is pointed at it.** The A7 chain's substantive output is that **the fibre is where contextuality has to live at `N ≥ 3`** ([`sigma-fibre-contextuality.md`](sigma-fibre-contextuality.md)) — which makes the fibre's own legitimacy critical. ⚠️ **SHARPENED 2026-07-29 — the gap is not what this row first said.** It is *not* "`FibredSigma` uses `ℝ`, `KSigma` is compact, swap them". The corpus has **two** record-layer constructions and **compactness and fibre-activity sit in different ones**: `KSigmaRecord.lean` puts a P5 `RecordSemantics` on the **compact** `KSigma = ℂℙⁿ⁻¹ × T²`, but its events are `π⁻¹' bornRegion ψ i` — **pulled back from the base**, so the torus fibre is *inert* and the construction inherits the preparation-indexed base regions. `FibredSigma.lean` has an **active** fibre (it carries the CDF Born cells, `cdfCell`) but that fibre is `ℝ` with Lebesgue restricted to `[0,1)` — **non-compact**. **Neither module has both properties**, and since the A7 work concluded contextuality must live in the *fibre*, the one we need is precisely active-fibre-on-compact-Σ. Also: no Kähler structure on the fibre, and its measure not shown Liouville, while Paper C A1 wants a compact Kähler ontic surface. **FIBRE HALF LANDED 2026-07-29** (`RecordLayer/CircleFibre.lean`): the Born partition now lives on a **compact** fibre. `CircleFibre = AddCircle 1` (the same factor `KTorus` is built from) is compact with Haar a genuine **probability** measure (`circleFibre_volume_univ`) — what restricted Lebesgue on `ℝ` only had by fiat. `circleCell` is a **preimage** under the canonical-representative map, not an image of `cdfCell`, so measurability is immediate; `volume_circleCell` gives **exactly** the Born weight `rᵢ` (via Mathlib's `AddCircle.measurePreserving_equivIoc`), `volume_circleBornCell` gives `‖ψ i‖²`, and `circleCell_pairwiseDisjoint` keeps outcomes exclusive. **So compactifying the fibre changes nothing about the outcome probabilities.** **Record-semantics counterpart LANDED 2026-07-30** (`RecordLayer/CircleRecord.lean`): `circleRecordSemantics` (P5, measurable + exclusive), `compatibleSet_circle_single` (P6 isolation = conditioning), `circleOutcome_eq_record` (the ontic selection *is* the record), `circleBornMeasurement_prob` (`= ‖ψ i‖²`, the same weight `ℝ` gave), `circleBornMeasurement_ae_total`. ★ The `RecordSignature` is reused **verbatim** — it never mentioned the fibre — so only the *semantics* changed, which is why no Born weight moved. ★ And `ae_total` came out *stronger* — but ⚠️ **CORRECTED 2026-08-11**, this row previously gave the wrong reason ("on `ℝ` it had to be restricted to `[0,1)` by hand because Lebesgue measure on the line is infinite"). `fibreTypicality` is `volume.restrict (Ico 0 1)`, a **probability** measure, not Lebesgue on the line, and the restriction was never forced: `fibreTypicality_uncovered_univ` proves the `univ` form on `ℝ`. The real content is the one this row already states correctly two sentences earlier — on the circle the probability mass is Haar mass, *what restricted Lebesgue on `ℝ` only had by fiat* — and `fibreTypicality_Ici_one` makes the fiat explicit: the fibre's complement, of infinite Lebesgue measure, gets typicality zero, so an uncovered point there is excused by the measure rather than covered by a cell. ⚠️ **TWO CORRECTIONS, 2026-07-30 (external review, both verified against the files).** **(1) The re-plumbing is NOT done — this row said "DONE" and the landing commit was headlined "the record layer now runs on the compact fibre". Both are retracted.** What exists is a **parallel compact counterpart** of the record semantics, proved in full. The corpus's actual capstones — `Measurement.lean`, `RecordLayerClosure.lean`, `FiniteQMClosure.lean`, `KSigmaRecord.lean` — are untouched since 2026-07-25 and still run on the `ℝ` fibre with `fibreTypicality`, and **nothing outside `AxiomAudit.lean` imports the circle modules**. Retiring the `ℝ` closure is real, open work. **(2) A single circle CANNOT complete A1, and the reason is parity, not tooling.** This row (and `CircleFibre.lean`) classified the missing Kähler structure as *permanently scoped per §2a — blocked on Mathlib's absent manifold exterior calculus*. **That was a misclassification.** `ℂℙⁿ⁻¹` has real dimension `2n-2`, so `ℂℙⁿ⁻¹ × AddCircle 1` has real dimension `2n-1` — **odd** — and no odd-dimensional manifold admits a symplectic form (`ωᵏ` must be a volume form), hence none admits a Kähler structure. More Mathlib API would not change this. ⚠️ The same objection applies **retroactively to `FibredSigma`'s `ℂℙⁿ⁻¹ × ℝ`**, also `2n-1`: that gap was never a tooling gap either. **The fix is already in the corpus and is cheap.** `LF4/KahlerInstance.lean` has `KTorus = AddCircle 1 × AddCircle 1` and `KSigma N = CPN N × KTorus` — real dimension `2n`, **even**, compact, a product of Kähler manifolds. Put `circleCell` on the **first** torus coordinate and keep the second as its symplectic partner; every `CircleFibre` theorem is stated about one `AddCircle 1` and transports, so what is missing is the product-measure step, not new fibre mathematics. **So: the compactness objection is closed; the A1 objection is NOT scoped, it is open and tractable.** → ✅ **ITEM (b) DONE 2026-08-25** (`LF4/LiouvilleUnique.lean`, 2 pins) — ★★ `kMuL_unique`: **`kMuL p₀` is the ONLY probability measure on `ℂℙ^{N-1} × T²` invariant under `U(N)` on the base and `T²` on the fibre.** The sector measure is **forced by the symmetry, not chosen**, which is what this item asked for. Fibre half via `isAddInvariant_eq_smul_of_compactSpace` — ★ **no regularity side conditions BECAUSE `T²` is compact, so fibre compactness is load-bearing here rather than decoration, and it is exactly what `TorusFibre`/`GlobalRecordClosure` bought in July**; base half via `invariant_measure_uniqueness_cpn`; joined by `Measure.prod_eq`. ⚠️ Forced **given** the group — it does not derive the symmetry, Σ stays the floor, and it does not touch the other open item (no `H_int(M)`). *(Scoping record, kept:)* ⚠️ **The gap is narrower than this row implies.** `KahlerOnticSetup.liouvilleMeasure` is *named* Liouville but only *typed* as `Measure Sigma` + `liouville_isProbability` — **nothing in the structure forces the sector's measure to be canonical.** The textbook definition (top exterior power of the Kähler form) is unavailable and stays unavailable — that is L1's manifold residual, blocked on Mathlib, `Q8` = XL. ★ **The formalisable content of the same fact is SYMMETRY-UNIQUENESS** — on a homogeneous space the Liouville measure *is* the invariant one, and the corpus already uses exactly that reading for the base (`invariant_measure_uniqueness_cpn`). **Target: `kMuL p₀` is the UNIQUE probability measure on `ℂℙ^{N-1} × T²` invariant under `U(N)` on the base and `T²` on the fibre** ⇒ the measure is forced, not chosen. **Route, walls all pre-checked (the Q12 lesson):** (i) fibre half — for measurable `A`, `(μ.restrict (A ×ˢ univ)).map Prod.snd` is a finite translation-invariant measure on `T²`; ★ **`isAddInvariant_eq_smul_of_compactSpace` needs NO regularity side conditions because `T²` is COMPACT**, so compactness of the fibre is load-bearing here rather than decoration (and it is the very thing `TorusFibre`/`GlobalRecordClosure` bought). (ii) base half — the marginal is `U(N)`-invariant ⇒ `invariant_measure_uniqueness_cpn`, total mass fixing the scalar. (iii) join by `Measure.prod_eq` (rectangles suffice). **Instances verified to synthesize:** `CompactSpace KTorus`, `SecondCountableTopology KTorus`; and `Q26` (DONE 2026-08-21) already paved the standard-Borel/disintegration instances on the same space if the rectangle route is ever swapped out. **Nothing is blocked; what remains is mechanical.** ⚠️ Scope when it lands: forced **given** the symmetry group. It does not derive the group, and Σ stays the floor. What is left of this row: (a) ~~move the partition onto `KSigma`'s `T²`~~ **DONE 2026-07-30** (`RecordLayer/TorusFibre.lean`, foundational-triple, 6 pins verified in isolation) — `torusCell r i = circleCell r i ×ˢ univ` puts the Born cells on the **first** torus coordinate with the second free; `volume_torusCell` gives **exactly** `rᵢ` (the free coordinate contributes a factor of `1`, `T²`'s Haar being a probability measure), `volume_torusBornCell` gives `‖ψ i‖²`, plus exclusivity and a.e. totality. ★ `mem_torusCell_iff` states the second coordinate's freedom — "the symplectic partner carries no record content" — as a *theorem* rather than prose. ⚠️ **This REMOVES AN OBSTRUCTION to A1; it does not ESTABLISH A1.** `KSigma` is not proved Kähler here, and the partition is still kinematic and still preparation-indexed. Nothing outside `AxiomAudit.lean` consumes it yet; (b) exhibit the fibre measure as Liouville rather than merely Haar; (c) ~~retire the `ℝ` closure~~ **DONE 2026-07-31** (`RecordLayer/GlobalRecordClosure.lean`, foundational-triple, 4 pins). `GlobalRecordClosure` certifies **the same five record-layer facts** as `RecordLayerClosure` — exclusivity, selection-is-record, isolation-is-conditioning, Born, a.e. totality — but on `KSigma = ℂℙⁿ⁻¹ × T²` for a `ContextField` rather than on `ℝ` for `bornContext ψ`. ★ **The record event is now a function of `(context, outcome, time)` and nothing else** — visible in the type, needing no theorem: the *same* set serves every preparation and only the epistemic measure moves, where `fibreRecordSemantics`'s event was `cdfCell (bornRate ψ)` and moved with `ψ`. ★ `globalOutcome` is literally `circleOutcome` read at the point's own base, so the ontic selection needed no new machinery. ★ `ae_total` strengthens from `Ico 0 1` to the whole space. **That the five fields are otherwise identical is the evidence that neither defect was ever load-bearing for the record layer's content.** ⚠️ `RecordLayerClosure` is **superseded, not deleted** (still true, still consumed), and **`FiniteQMClosure` still carries the older `vnPointerOutcome` readout** — swapping that is a separate migration on the `productDynamics` engine and is **not** done. Still kinematic. Related and also open: the fibre *mechanism* (softmax/Gumbel) is **posited, not derived** from de-isolation dynamics, and the Gumbel form is injected noise rather than CSD-native. **Effort: M–L** (revised down from L — the `T²` route needs no Mathlib API). **★ SUCCESSOR TARGET — the global context-fixed basin on `T²`** — **LANDED 2026-07-30** (`RecordLayer/GlobalBasin.lean`, foundational-triple, 6 pins; plus `LF4.continuous_momentMap` / `measurable_momentMap`). `ContextField` bundles a measurement context as a **measurable simplex-valued rate field on the base**; `globalBasin c i = {x **PARITY GAP CLOSED 2026-08-09** (`RecordLayer/TorusRecord.lean`): the compact fibre `S¹` is one-dimensional, so `ℂℙ^{N-1} × S¹` has ODD real dimension and cannot carry a symplectic or Kähler structure at all — compactness was fixed, parity was not. `TorusFibre` had already moved the Born cells onto `T²` (cells constrain `θ₁`, `θ₂` free); this ports the rest of the record layer there, so the **active fibre and the A1-admissible arena are finally the same object**: `torusRecordSemantics` (P5), `compatibleSet_torus_single` (P6), `torusOutcome_eq_record` (the ontic selection IS the record), Born weight `‖ψ i‖²` unchanged. ⚠️ **Not an A1 discharge**: no Kähler form is constructed on the arena and the fibre measure is not shown to be a Liouville volume for one (KG-1's Mathlib forms block). Parity was a *necessary* condition that was violated and now is not; necessity is not sufficiency, and the row stays open on the Kähler/Liouville half. | x.2.1 ∈ circleCell (c.rate x.1) i}` reads the rate **at the ontic point**, so **no `ψ` appears in the definition**. `measurableSet_globalBasin` (two inequalities between measurable real functions — the step the moment-map regularity unblocks), `globalBasin_pairwiseDisjoint`, `globalBasin_prob` (conditioning the epistemic state on `p` returns `rate p i`), `globalBasin_ae_total`, and the headline **`globalBasin_born`: at preparation `ψ` the basin's probability is exactly `‖⟨eᵢ,ψ⟩‖²`** — Born, from a partition that never mentions the preparation. ⚠️ **What this closes is the PREPARATION-INDEXING defect, not general-`N` A7 outright.** Whether Paper C intends the regions to be *base-only* — in which case the parked chain still governs — is a question about the axiom, not about the file. Still **kinematic** (no `H_int(M)`), `δ_p ⊗ Haar` is the **epistemic** measure taken as a definition rather than a disintegration, and `KSigma` is still not proved Kähler. ⚠️ **Estimate correction:** the `momentMap` measurability prerequisite was logged **M**; it was **S** — `Mathlib/LinearAlgebra/Projectivization/{Topology,MeasureSpace}.lean` already staged `continuous_iff_continuous_comp_mk'` and the Borel instance, so the descent was four lines. *(Original assessment retained:)* Define `Bᵢ(M) = {(p, θ₁, θ₂) : θ₁ ∈ cdfCell (m_M p) i}`, where `m_M` is the torus moment map in the basis `M` selects. The point: `m_M` is evaluated **at the ontic point `p`**, not at the preparation, so `Bᵢ(M)` depends on the apparatus context **alone** — which is what Paper C A7 literally asks for, and what `bornContext ψ` does not give. Born follows by conditioning on the preparation and integrating the fibre: the slice over `p` has Haar measure `(m_M p) i`. This does **not** collide with the parked general-`N` chain — that chain constrains **base-only** densities, and this is genuinely fibred, exactly as `sigma-fibre-contextuality.md` concluded. It is not circular either: `bornRate_eq_momentMap` already establishes the rates as *forced* by the Kähler structure and the `Tⁿ` action. ⚠️ **Design point to state, not gloss:** conditioning on the preparation means conditioning on `p = [ψ]`, a **null set** in `ℂℙⁿ⁻¹`. Take the isolation-conditioned epistemic measure to be `δ_[ψ] ⊗ Haar` outright (a legitimate product measure) rather than disintegrating — and be explicit that this is the *epistemic* measure, **not** the sector's Liouville measure. ⚠️ And it stays **kinematic**: it does not touch the Paper D `H_int` obligation. ⚠️ **KNOWN PREREQUISITE, found 2026-07-30 while landing the arena:** `Bᵢ(M)`'s measurability needs `momentMap` to be measurable (or continuous), and **the corpus has no such lemma** — `LF4/MomentMap.lean` proves `momentMap_nonneg`, `momentMap_le_one` and `momentMap_sum_eq_one` (which *are* exactly the rate hypotheses `volume_torusCell` wants) but says nothing about regularity. `momentMap` is a ratio of norms on a projective quotient, so this is a genuine small piece of work, not a `fun_prop` call. It is a detour, not a wall — but it is why this target is **M**, not **S**. | `sigma-fibre-contextuality.md`, `RecordLayer/FibredSigma.lean` | | **★ General no-signalling over a non-factorising Σ (C1 item 16)** | **OPEN — recorded 2026-08-10, deliberately not closed.** Determine sufficient **primitive** conditions on setting-dependent measure-preserving CSD dynamics over a non-factorising ontic Σ that **imply** operational remote marginal invariance, *without* assuming Bell-local pointwise response independence. ⚠️ **Why this row exists rather than a theorem:** the measure-level predicate `LF3.OperationalNoSignalling` is close to a restatement of the conclusion rather than a premise doing derivational work, so verifying it in the constructed sector (`localDeisolation_no_signalling_A/B`) is a **verification, not a derivation**. C1 §4.2 must say so in those words. ⚠️ **And the pointwise form is not merely too strong — it is FALSE in this sector:** imposing `F_A(a,b,x) = F_A(a,b',x)` on both wings over a deterministic shared state yields exactly the setting-local response pair `no_product_partition_realises_singlet` forbids. ⚠️ **Hidden premise now named:** `OperationalNoSignalling` fixes ONE `μ` across all four contexts, and that fixture **is measurement independence** — a genuine Bell premise that was previously carried silently by the shape of the definition. No axiom is added for this row. | `LF3/OperationalNoSignalling.lean`; `specs/c1-correction-plan.md` §1 | | **★★ The dynamical measurement layer (Paper D `H_int`)** — *successor to the context-fixed basin row* | **OPEN — and the diagnosis that opened it is sharper than the "still kinematic" caveat `GlobalBasin` carried.** ⚠️ **`globalBasin_ae_total` shows the basins cover `Σ` up to a null set, so a.e. every point ALREADY carries a record and there is no apparatus-ready state of positive measure.** A flow cannot *create* a record in such a space — so "add dynamics to `GlobalRecordClosure`" is not merely incomplete, it is **structurally impossible**. The repair (external review 2026-08-01) splits the hidden **selector** from a pointer **register**: `Σ_meas = ℂℙⁿ⁻¹ × T²_λ × T²_R`, real dimension `2n+2` — even, compact, a Kähler product (parity checked). ★ **CONSISTENCY CHECK RUN FIRST, before building anything** (`RecordLayer/MeasurementConstraints.lean`, foundational-triple, 3 pins) — the cheap-failure discipline the `ℂℙⁿ⁻¹ × S¹` parity argument should have had. **Result 1 — the measure check comes back NEGATIVE, i.e. it does *not* obstruct:** `pointer_region_measure_ge` gives `μ_sel(S)·μ_R(R₀) ≤ μ_R(B)`, which granting the (unformalised) `μ_sel(Sᵢ) = 1/n` reads `μ_R(Bᵢ) ≥ μ_R(R₀)/n` — satisfiable with room. That is a **green light to attempt the concrete `H_int`, not evidence a witness exists**; measure preservation is a weak invariant. `ready_region_measure_le` is the summed form and is recorded **as weak** rather than dressed up. **★ Result 2 — the constraint with teeth:** `no_everywhere_correlation` — a continuous `Φ` sends the preconnected `S ×ˢ R₀` to a preconnected image, which cannot meet two disjoint *open* pointer regions, so an **everywhere** correlation is impossible at `n ≥ 2`. **Therefore the `a.e.` in the correlation theorem is mathematically necessary, not a convenience:** the exceptional set must be **non-empty**, it contains the seams between selector sectors, and its image threads the gaps between pointer regions. Two consequences: (i) any candidate `H_int` advertised as giving an *exact* correlation is wrong on these grounds alone, without inspection; (ii) a witness that leaves the seam unspecified has not addressed the hardest part of its own statement. **REMAINING WORK, and the ordering matters:** the plan's items 1–2 and 4–8 (arena, two-time propagator interface, dynamic Born, persistence on a finite `[T_M, T_M+τ_R]`, Lüders update, generalisation, capstone) are tractable Lean engineering; **item 3 — an explicit `H_int(M)` or propagator proved to arise from one, with `Φ(Sᵢ × R₀) ⊆ Bᵢ` a.e. — is the research result and is unbounded.** ⚠️ **Build only the minimum interface needed to *state* item 3, then attack item 3.** Executing 1→9 in order risks a large scaffold with a hypothesis at its heart, which is the failure the plan itself warns against (`basin_rate` is already one). **ARC COMPLETE THROUGH RANK-ONE LÜDERS (2026-08-01/02, eleven commits `750327f..3f8b9a9`).** What landed, in dependency order: **(a)** the interface (`MeasurementProtocol`, two-time propagator, ready/pointer regions; `CorrelatesOn`/`PointerInvariantOn` kept as *hypotheses*, never fields); **(b)** persistence + post-measurement ensemble (`RecordPersistence` — selection vs disturbance finally expressible); **(c)** the **shear witness** discharging both hypotheses from an explicit propagator, with measure preservation and dynamical Born (`ShearWitness`, `DynamicBorn`); **(d)** ⚠️ `shear_base_marginal_unchanged` — a machine-checked **negative** result: the shear cannot collapse, structurally (no back-reaction ⟺ no Lüders); **(e)** the **collapse no-gos** (`no_exact_collapse` — pointwise collapse impossible for measure-preserving dynamics, so collapse must be *relocation of a null slice*; `collapse_accuracy_bound` — approximate collapse priced in ready-state improbability, Landauer as a measure inequality, *forcing* the Dirac-calibration convention); **(f)** the **calibrated-swap witness** (`SwapWitness`, design from a Fable review with one substantive correction found by verification — the forward-only crossing flag fails the two-time law on non-monotone triples; the involution `G` fired symmetrically repairs it, all 8 cases closing) — record-triggered bank swap, measure-preserving by the new partition lemma; **(g)** ★★ **the Lüders theorem** (`SwapLuders`): conditioned on outcome `i`, the post-measurement system marginal **equals** `epistemicMeasure [eᵢ]` as a pushforward, so follow-up statistics for any context are the collapsed state's Born weights (`swap_luders_born`). Items 7–8 landed alongside (`OutcomeField`/`blockField`/`outcomeBasin` — degenerate rate fields reach the basins; `DynamicMeasurementClosure` + `CsdFiniteQMClosure`, now carrying the `luders_followup` field). **The plan's boxed completion criterion is met line-by-line for nondegenerate computational-basis measurements.** **STILL OPEN on this row:** (i) **degenerate Lüders** — **PROBLEM FORMALISED + BOUNDARY PROVED 2026-08-02** (`RecordLayer/DegenerateLuders.lean`, foundational-triple, 4 pins): `BlockLudersObligation` states the demand (§8.3 `_statement` pattern — post-marginal = `epistemicMeasure [Πᵢψ]`, **ψ-dependent**); ★★ `swap_not_blockLuders` proves the calibrated-swap witness fails it for **any** fixed calibration whenever a block has dim ≥ 2 (the swap's preparation-*independence*, `swap_luders_marginal`, turned against it at two in-block vertices); ★ `degenerate_selector_born` proves the positive half that survives — the block-selector's sectors carry exactly the coarse-grained Born weights, the dynamical form of `blockField`. **Statistics generalise; the update is what does not.** The **witness construction stays open**: it must relocate `[ψ] ↦ [Πᵢψ]` while preserving measure, so the lost base data must be *stored* (`no_exact_collapse` governs); route = the **projective-join decomposition** `ℂℙ^{N-1} ≈ join(ℂℙ^{dᵢ-1}, ℂℙ^{N-dᵢ-1})`, wall = the FS-measure decomposition under the join, unformalised geometry, effort **L**. **BRICK 1 LANDED 2026-08-02** (`RecordLayer/BlockCollapse.lean`, foundational-triple, 4 pins): the measurable ray-level collapse `blockCollapse : [ψ] ↦ [Πᵢψ]` (quotient descent + `measurable_iff_measurable_comp_mk'`); ★ **collapse as relocation** (`luders_target_eq_relocation`, `blockLudersObligation_iff_relocation` — the §8.3 target *is* the pushforward of the preparation under a deterministic relocation, fibre untouched); and the **vector-level witness core** `componentSwap` on `ℂᴺ ⊕ ℂᴺ` (involutive, summed-norm-preserving; with a block-calibrated slot it collapses *with the residual stored*). **The wall, sharpened:** the ray-pair descent fails on the *relative scale* the product `ℙ×ℙ` quotient forgets — a surviving relative `U(1)` is what the join needs. Routes: (i) FS disintegration under join coordinates (original); **(ii) NEW — a phase-carrying slot** (sphere-level bank, quotient at readout), likely cheaper. ~~`swap_not_blockLuders` remains the boundary; no ray-level witness claimed. Remaining effort **L** (route ii: M–L);~~ **BRICK 2 LANDED 2026-08-02 — route (ii) WORKS** (`RecordLayer/PhaseSlot.lean`, foundational-triple, 4 pins): ★★ `phase_slot_block_luders` — phase-orbit preparation (uniform over the ontic phase; `readout_phasePrep` shows the enrichment adds no epistemic content) + **fixed** block-calibrated slot + the **total, involutive, measurable** partial pair swap (`pairSwap`) ⇒ ray-level post-state **exactly** `δ_{[Πᵢψ]}`, preparation-dependent from a fixed calibration. The no-go is *evaded, not contradicted*: it killed full swaps (post = slot prior); the partial swap keeps the system's block part and stores the complement — and is only well-defined on the phase-enriched arena, as the sharpened wall predicted. ~~**Brick 3 (remaining, effort M):** register/sector protocol plumbing on the enriched arena (mirror `SwapWitness`) + Liouville preservation via a unitarily-invariant reference measure on the doubled space (Gaussian route; `componentSwap_norm_sum` is the invariance driver);~~ **BRICK 3's LIOUVILLE HALF LANDED 2026-08-02** (`RecordLayer/JoinArena.lean`, foundational-triple, 4 pins) — and the Gaussian route was not needed: **the phase-enriched pair arena IS the projective join `ℙ(ℂ^{N+N})`** (global phase quotiented, relative phase survives in the point). There the component swap is a **permutation unitary**, so ★★ Liouville preservation = FS unitary invariance (`joinSwap_measurePreserving`, one line from `fsMeasure_smul_invariant`), and ★★ the Lüders update is **pointwise deterministic** (`join_block_luders` — every join microstate reads out to exactly `[Πᵢψ]`; slot nonvanishing not even needed). The recorded join route's original wall (FS disintegration) was never crossed — it was **dissolved**: the join arena needs no disintegration because the dynamics is unitary on the whole space. ~~**Remaining, final piece (mechanical M, no new mathematics):** the register/sector `MeasurementProtocol` on `ℙ(ℂ^{N+N}) × T²_R` firing `joinSwap` (mirror `SwapWitness`) + the conditioned-marginal bookkeeping to a `BlockLudersObligation` instance;~~ **BRICK 4 (the protocol) LANDED 2026-08-02** (`RecordLayer/JoinProtocol.lean`, foundational-triple, 4 pins): the degenerate measurement runs as a `MeasurementProtocol` — arena (join point, system fibre, **ancilla fibre**) × register, record-triggered `joinG` (join unitary + fibre exchange; the original fibre **stored**), two-time law (the eight-case crossing proof), `CorrelatesOn`/`PointerInvariantOn` discharged, ★ full-propagator **Liouville preservation** (`joinEvolve_measurePreserving`). Regions/readout/sectors inherited from `shearProtocol` by structure update. ~~**Brick 5 (the last):** the sector-conditioned post-marginal for the canonical phase-orbit preparation = `epistemicMeasure [Πᵢψ]` — the `BlockLudersObligation` instance, mirroring `SwapLuders`'s conditioning bookkeeping (pointwise input `join_block_luders` already proved); effort **M**;~~ **BRICK 5 LANDED 2026-08-02 — THE DEGENERATE-LÜDERS ARC IS CLOSED** (`RecordLayer/JoinLuders.lean`, foundational-triple, 4 pins): ★★ `joinWitness_blockLuders` — **`BlockLudersObligation` is inhabited**. The demand `swap_not_blockLuders` proved impossible for every fixed ray-level calibration is delivered by the join witness: ψ-dependent post-states `epistemicMeasure [Πᵢψ]` from a fixed block-supported calibration family, through Liouville-preserving dynamics in the standard protocol architecture, with positivity derived from `Πᵢψ ≠ 0` alone (`goodTheta_vol_pos` — no measure hypothesis). The five bricks: relocation target → phase slot → join arena → protocol → obligation. `swap_not_blockLuders` stands as the boundary theorem for the ray-pair arena. **Row CLOSED**; (ii) **Hamiltonian generation — REOPENED 2026-08-02** (external review; the earlier same-day "permanently scoped (Mathlib gap)" classification was **misjustified** — the second `check-impossible-before-scoped` failure: we verified *parity*, not *continuity*). The review's claim is now **machine-checked**: ★ `shearEvolve_not_continuous` (`RecordLayer/ShearDiscontinuity.lean`) — the witness propagator displaces the register by the *discrete* `shearAmt (basinIndex x)`, and a continuous propagator would give a clopen partition of the (proved-connected) `KSigma`; so the witness is a measurable, measure-preserving **piecewise** map and **cannot be the time slice of any continuous flow**. Writing down `H_int = g(t)·Â_sel ⊗ P̂_R` does not make the witness its flow: that generator's flow *is* continuous, so it is a **different map**. Kept honest in both directions: `no_everywhere_correlation` proves *some* seam set is forced for every exact-record dynamics — the witness's jumps sit exactly on the seams the no-go requires, so the discontinuity is structural, not a construction defect. **Two repair routes, decision pending (authorial):** (1) **smooth corridor regularisation** — mollify the selector across seam corridors; records/Born correct up to `ε`, priced by `collapse_accuracy_bound`; effort M–L, genuinely new mathematics; (2) **classify CSD measurement dynamics as piecewise Hamiltonian with null seam set** — standard in dynamics (billiards, impact systems), physically justified by the forced-seam no-go; effort S (docs + the classification statement). The narrower `X_H = ω⁻¹dH` *formalisation* gap (symplectic form + exterior derivative in Lean) remains a genuine Mathlib matter, but it is no longer the load-bearing obstacle; ~~**DECIDED + LANDED 2026-08-02 (route 2: piecewise Hamiltonian)**~~ **REOPENED AGAIN 2026-08-02 (second external review), on two corrected grounds:** **(flux)** the basin-cylinder pieces are rigid torus translations — symplectic and measure-preserving, but **not globally Hamiltonian**: `ι_Xω = a·dp` is closed-not-exact on `T²` (`∮dp ≠ 0`), so no global generating function exists and the `hᵢ = aᵢ·p_R` reading was wrong; the correct name for the landed classification is **piecewise rigid symplectic translation with null seam set** (`PiecewiseHamiltonian.lean`'s theorems all stand; its docstring carries the dated correction, and the theorem names are retained for pin stability); **(overread)** `no_everywhere_correlation` forces an *exceptional non-correlating set*, **not** discontinuity — a continuous, even smooth globally-Hamiltonian propagator that maps seams to transition states outside `⋃Bᵢ` remains **open**. **★ ROUTE CONFIRMED 2026-08-03 (author decision: retain the smooth-Hamiltonian architecture; execution plan `pointer-witness-plan.md`): the compact Kähler pointer `ℂℙ^K`** — replace the torus register with a projective pointer (one ready state, `K` record states); unitary rotations on `ℂℙ^K` **are** globally Hamiltonian (`H¹(ℂℙ^K) = 0`, moment maps exist), killing the flux obstruction at the root. Obligations for the new witness: continuity/smoothness, Hamiltonian generation (moment-map level), measure preservation (FS unitary invariance — free), a.e. record creation, Born sector weights, persistence, and the Lüders composition. **Effort L.** ~~The papers question (Paper C A2 / TN6 require smooth Hamiltonian flow — retain that architecture, per the review's strong recommendation, vs. amend the papers) is **authorial**;~~ **DECIDED 2026-08-03 (author): the smooth-Hamiltonian architecture is RETAINED** — Paper C A2 / TN6 stand as written, and the `ℂℙ^K` pointer witness is the obligation that must satisfy them; the landed piecewise witnesses stay as the exact-record horn of the `no_everywhere_correlation` trade-off, not as the architecture's final word; **★★ NEGATIVE RESULT 2026-08-10 on the LAST obligation in that list (`RecordLayer/RelocationObstruction.lean`, foundational-triple, 4 pins; plus 4 staging pins).** `PointerGeneration.lean` discharged the *record-CREATING* obligation (`rampedU_schrodinger`: the stroke solves the Schrödinger ODE for an explicit Hermitian generator at every time and every weight vector). The **Lüders-composition** obligation **cannot be discharged as posed**, and the obstruction is topological, not flux. **Horn 1** (`pointerBankSwap_not_flow_time_one`): the relocation exchanges the system factor with bank slot `j`; embed a circle in the system's torus angle and the exchange **collapses that section** (afterwards the readout coordinate reports slot `j`, which the section held constant), and the circle is not contractible — so no jointly continuous family joins the identity to the relocation. ⚠️ **This is NOT the `PiecewiseHamiltonian` flux obstruction**: flux obstructs a symplectomorphism *within* the identity component; the exchange **never reaches that component**, and the `H¹(ℂℙ^K) = 0` escape that saved the record stroke does not apply, because the obstruction lives in the **bank's product structure**, not the pointer's topology. **Horn 2** (`pointerImprint_not_injective` / `_not_homeomorph`): the obvious non-permutation repair — write the system into a slot instead of exchanging with it — is **not injective**, hence not a homeomorphism, hence not a flow map either, for a completely independent reason. **Scope, stated in-module:** this constrains the **swap architecture**, not collapse-as-dynamics in general. What survives: a generated relocation must be a **bijection that is not a factor exchange**, which points at the join and phase-slot routes (`JoinLuders.lean`, `PhaseSlot.lean`), where the ψ-dependence is produced by the dynamics rather than by permuting coordinates. Read together with `swap_not_blockLuders` (same architecture, obstructed by calibration rather than topology), the swap route is a **witness that cannot be generated**. **Enabling infra — two genuine Mathlib gaps filled:** `Mathlib/Topology/Homotopy/CircleFundamentalGroup.lean` (π₁(S¹) as the deck group 2πℤ — Mathlib has the whole covering apparatus and `Circle.isAddQuotientCoveringMap_exp` but nowhere states the consequence, checked at the pin) and `Mathlib/Topology/Homotopy/FactorExchangeObstruction.lean` (the obstruction, stated **basepoint-free**: routing through `contractible_iff_id_nullhomotopic` avoids the π₁ basepoint-conjugation bookkeeping entirely, which was the predicted cost and did not materialise). Also constrains **record broadcasting**: carrier maps that relocate a record by permuting factors inherit horn 1, imprint-style copies inherit horn 2, so redundancy built either way is stuck with hand-built maps; **★★ RESOLVED THE SAME DAY — THE JOIN RELOCATION *IS* GENERATED** (`RecordLayer/JoinGeneration.lean`, foundational-triple, 6 pins). The obstruction really was about the **swap architecture**, exactly as its scope note claimed. **Why the join escapes both horns, structurally:** the bank arena is a **product** `(Σ × pointer) × (Fin N → Σ)` and `pointerBankSwap` exchanges two of its **factors**; the join arena is `ℂℙ^{N+N-1}`, a **single** projective space, and `joinSwap b i p = joinU b i • p` is one unitary acting on one connected space — so there are no two arena factors to exchange (horn 1 has nothing to act on) and a projective unitary is bijective (horn 2 has nothing to act on). **And the escape is constructive, not merely permitted:** `joinMat` is a real permutation matrix of an **involutive** permutation (`joinPerm_involutive`), hence **Hermitian** (`joinMat_conjTranspose`) and an **involution** (`joinMat_mul_self`), hence has eigenvalues ±1; so `joinProj := ½(1 - P)` is a Hermitian **idempotent** (`joinProj_mul_self`); so `joinFlowMat t := (1 - Q) + e^{iπt}Q` is **unitary at every time** (`joinFlowMat_mem_unitaryGroup`) with `U 0 = 1` and `U 1 = joinMat`; and ★★ `joinFlowMat_hasDerivAt` proves it solves the **Schrödinger equation** `U'(t) = (i·H)·U(t)` for the **explicit Hermitian generator** `H = π·Q` (`joinGen_isHermitian`). Payoff ★★ `joinSwap_eq_flowTimeOne`: the join relocation is the **time-one map of a Hamiltonian flow**, the exact property `pointerBankSwap` provably lacks. *Technique note:* **no matrix exponential is needed** — on an idempotent the exponential series collapses to `1 + (e^z - 1)Q`, so writing that closed form down directly turns the whole construction into algebra (the `Matrix.Norms.L2Operator` scope is still needed for `HasDerivAt` on matrices). **So: collapse CAN be dynamics.** With `rampedU_schrodinger` (record-creating stroke) and this (relocation stroke), both strokes of a measurement now have explicit Hermitian generators — on **different** architectures, which is the honest statement. ⚠️ **Remaining, and it is the real frontier:** the two-stroke composite is still **triggered** by a readout, and the trigger is where `no_everywhere_correlation` bites. What is now settled is that the **relocation is not the obstacle**; the **trigger** is. Design rule extracted: a generated relocation must be a **bijection that is not a factor exchange** — satisfied by being a unitary on an irreducible arena rather than a permutation of coordinates on a reducible one; (iii) ~~unitary covariance~~ **DONE 2026-08-02** (`RecordLayer/RotatedSwap.lean`: ★★ `measurement_covariance` — the six-fact closure for a first measurement in **every** orthonormal basis, rotated bank + rotated selector; context-generic accounting `sector_born_ctx`/`prep_outcome_pos_ctx`; the BB84 **primal round** lands as `bb84_primal_wrong_basis`, dual-round caveat retired); mixed preparations / POVMs remain (plan's extension steps 3, 4); (iv) the **engine migration** (operational and dynamical closures still a conjunction over different arenas). **Step 1 of 2 DONE 2026-08-02** (`RecordLayer/SwapClosure.lean`, review step 2): all six dynamical facts now live on the **swap arena alone** (`SwapMeasurementClosure`/`swapMeasurementClosure`, foundational-triple) — with the Born field upgraded to the **dynamical** sector measure (`swap_sector_born`, consuming `measure_outcomeSector_eq_of_correlates`) and the Lüders field carrying **no** measure hypothesis (`prep_outcome_pos`, relocated to `SigmaLayer`). ~~Step 2 — migrating the **operational** closure onto the same arena — remains **OPEN, effort L**.~~ **Step 2 DONE 2026-08-02** (`RecordLayer/UnifiedArena.lean`, foundational-triple, 4 pins): ★★ `unifiedArenaClosure` — **one ontic model carries the finite-QM reconstruction**. `UnifiedArena M` with `arenaLiouville = swapMeasure (kMuL p₀)`: the lifted `exp(-itH)` preserves it and projects to Schrödinger; `swapEvolve` preserves the *same* measure; FS bridge, i.i.d. Born frequencies and mixed-state Born weights transfer through the system-slot marginal; the six dynamical facts embed as a field. All eleven operational fields accounted for (migrated/upgraded/superseded — the mapping table is the module header). ~~**Recorded residues (S):** the mixed-state two-stage LLN on the arena (`mixtureMeasure` lift, mechanical); the composed **round-trip theorem** (isolate → measure → isolate with records intact) — the natural next statement on the now-shared arena.~~ **Both residues DISCHARGED 2026-08-02 (same day):** `arena_mixed_born_frequency` (two-stage LLN through the system-slot marginal) and ★ `arena_round_trip` (isolate → measure → isolate; the record survives because the pointer register is a conserved coordinate of the lifted flow, `readout_arenaIso` — definitional). The engine-migration row is **CLOSED**. `CsdFiniteQMClosure` stays as the historical two-arena capstone. *(Original row detail retained below.)* **First witness candidate: the von Neumann shear** `H_int(t) = g(t)·Â_sel ⊗ P̂_R` — textbook, Hamiltonian, fits the nondegenerate computational-basis case; the wrinkle is that on a *compact* `T²_R` the translation wraps, so the `Bᵢ` must be arcs with `T_M` and coupling chosen to keep images disjoint. ⚠️ **Note on the plan's item 9:** it proposes rewriting Paper C §5 on the grounds that base-only regions are "no longer supported". **That over-reads the Lean.** Our result shows a *fibred* construction **works**; the `ContextFixedA7` chain does **not** show base-only **fails** — it is parked, open in both directions. Changing the paper is an authorial call, but not on those grounds. **Effort: interface M; `H_int` unbounded.** | external review 2026-08-01; `RecordLayer/MeasurementConstraints.lean` | | ~~**★ A5 approximate `(ε,T)`-projectability**~~ | **DONE 2026-08-02** (`Mathlib/Analysis/Matrix/DuhamelBound.lean` + `RecordLayer/ApproxProjectability.lean`, foundational-triple, 7 pins) — the row's stated shape delivered in full: **the predicate** (`EpsProjectable`, oscillation form — the derivative form `sup‖d(δH)|_V‖ ≤ ε` is the §2a-scoped manifold statement, substitution stated not silent); **the exact case as the `ε = 0` instance, as an iff** (`epsProjectable_zero_iff` — zero fibre-oscillation ⇔ factors through `π`, tying to `kSectorDataFlow_projectable`); non-vacuity (`diagOnticEnergy_epsProjectable` — the corpus's own moment-map energies are exactly projectable); and **the tracking theorem** (`quantum_effective_shadowing` + `_state`, via the integral-free **Duhamel bound** `‖e^{tC}−e^{tA}‖ ≤ |t|‖C−A‖` for skew-Hermitian generators, unitarity from `StoneC1` meeting the L2 norm's C*-structure): `‖H−H₀‖ ≤ ε` and `|t| ≤ T` give sector dynamics tracking true dynamics to within `ε·T` — what "selects the sector" means operationally. ⚠️ **Residue, stated:** the shadowing lives on the Hilbert side where the corpus's dynamics runs; an ontic Hamiltonian *generating* the flow is **A2's** row, not A5's. *(Original row:)* **OPEN — the highest-value remaining item after the fibre, and NOT blocked on Mathlib** (identified by the A1–A7 audit, 2026-07-29). Only the **exact** case `H = h∘π` is formalised (`kSectorDataFlow_projectable`, `SigmaLayer/DynamicsBridge.lean`); the corpus contains **no `ε`-projectability at all**. But A5's *physical* content is the approximate case — `sup‖d(δH)|_V‖ ≤ ε` over a window `T` — which is what makes a Hamiltonian **quantum-effective** rather than arbitrary, i.e. what **selects the sector**. So the axiom that does the selecting has its actual content unformalised. Ingredients are already here: operator norms, matrix-exp differentiability, `Mathlib/Analysis/Matrix/StoneC1.lean`. **What's needed:** state `(ε,T)`-projectability as a predicate on `(H, π, V)`, prove the exact case is the `ε = 0` instance, and show the projected flow tracks `e^{-itH}` to within a bound controlled by `ε` and `T`. **Effort: M–L.** | `reconstruction-status.md` §2a | ## L — weeks | Item | Status / what's needed | Former source | |---|---|---| | **Operator convexity → unconditional SSA** | ★★ **BRIDGED AND PUBLISHED 2026-09-12** — the `csd-qit-bridge` package planned below exists and is public (https://github.com/zblore/csd-qit-bridge, against **physlib**, not Lean-QIT: identical pins, Gate 0 passed, both dependencies by pinned revision, own CI), exporting unconditional SSA and the corpus's `hDPI` verbatim, foundational triple; `lieb-dpi-scoping.md` §6. The corpus is unchanged. ★ **REFRAMED 2026-07-31 — this may not need to be *proved* here at all.** `[LeanQIT2026]` (`QuAIR/Lean-QIT`, Apache-2.0, commit `48d4515`) **already proves DPI and SSA**: `relativeEntropy_dataProcessing_channel_ge` and `condMutualInfo_nonneg`. The repo is 322 modules with **zero `axiom` and zero `sorry`**, and is fully module-system like ours. ⚠️ **SSA here REMAINS CONDITIONAL** — this is a citation, not an import, and nothing in the corpus changed. **Why not imported:** Lake resolves one version per dependency, and upstream targets Lean `v4.30.0` / Mathlib tag `v4.30.0` (2026-05-26) against our `v4.33.0-rc1` / master `c732b96d` (2026-07-20). **Why that is fine:** `hDPI` is stated entirely in *our* vocabulary, so discharging it means producing a term of that `Prop` — doable **outside this repo**. Plan: a separate **`csd-qit-bridge`** package depending on both **by tag**, holding the adapter and the unconditional SSA; neither project merges, and `CsdLean4` keeps its Mathlib-only footprint. ⚠️ **Vendoring was considered and rejected:** the transitive closure of their DPI module is **71 modules / ~60,800 lines**, against our whole corpus of 374 files / ~104,000 lines — ~59% of our mass in someone else's code, which would relocate the toolchain problem onto us permanently, require auditing 61k unfamiliar lines to our zero-axiom standard, and muddy provenance on a result that would then be partly theirs. **Highest-leverage next action: ask QuAIR about a toolchain bump — or whether DPI/Lieb concavity is headed for Mathlib**, which would dissolve the question entirely for both projects. **The ladder below stays the fallback**, not the plan of record. *(Original entry:)* **Deferred by user 2026-07-23** (leave as-is; SSA stays honestly conditional on `hDPI`). **Endpoint:** discharge `hDPI` (= data-processing inequality = joint convexity of relative entropy = Lieb concavity) in `strong_subadditivity_of_relEntropy_monotone` — SSA already follows from it genuinely; only `hDPI` itself is open (no axiom, no `sorry` — the wall is isolated as an explicit hypothesis, not papered). **Ladder status** (`operator-convexity-plan.md`, `Mathlib/Analysis/Matrix/OperatorConvex*.lean`): DONE — L.1 inverse operator-convexity (Schur), L.2-resolvent concavity, CStarMatrix↔Matrix bridge, the reframing lemma (operator concavity ↔ `ConcaveOn` on the Löwner order), cfc-integral commutation + order-topology instances, `x^p` endpoints `p∈{0,1}`. **LIVE WALLS:** (1) `x^p` interior `p∈(0,1)` — integral-rep assembly, *prerequisites already landed, closest to breaching*; (2) operator concavity of `log`; (3) L.4 operator-Jensen/Ando + L.5 Lieb concavity ⇒ `hDPI` (not started, the summit). Structural obstruction: `NonUnitalCStarAlgebra (Matrix n n ℂ)` fails, so `rpow` can't transport via CStarMatrix. **Effort:** ~4–7 working weeks (Mathlib-scale; Lieb concavity is not in Mathlib). Scoped probe available: the `x^p`-interior rung alone. (The prior "step 0 / PartialOrder ℂ cascade" note was stale — that order-instance layer has since landed.) | `operator-convexity-plan.md` | | **GKSL / Lindblad open-systems tier** (LF6-9) | **Generator tier DONE 2026-07-20** (`LF6/LindbladGenerator.lean`): the general GKSL generator `ℒ(ρ)=−i[H,ρ]+Σₖ(LₖρLₖ†−½{Lₖ†Lₖ,ρ})` + trace-annihilation (`lindbladGenerator_trace`), Hermiticity-preservation, CP of the jump part (`lindblad_dissipation_posSemidef`, reusing the Choi/Kraus witness); dephasing shown to **be** a GKSL instance (`dephasingGenerator_eq_lindblad`) and to **solve** its master equation (`dephasingChannel_master_equation`). **Semigroup tier DONE 2026-08-13** (`LF6/LindbladSemigroup.lean`, Q5): `Φₜ = exp(tℒ)` for arbitrary GKSL generators — semigroup law, master equation, trace + Hermiticity preservation at the flow level. **Remaining (genuinely Mathlib-scale):** CP of the *exponentiated* `e^{tℒ}` for arbitrary generators (needs Lie–Trotter or resolvent positivity, neither in Mathlib). Unblocks LF6-2 full + Metrology A4. | `lf6-plan.md` | | **§14 observable correspondence — general-N, all self-adjoint** | **DONE 2026-07-22** (`LF4/ObservableCorrespondenceN.lean`). `observable_correspondence_diagonal`: for any `N` and any real eigenvalue vector `lam`, `⟨ψ, diagonal(lam·) ψ⟩ = ∑ₖ lam k · vol(bornRegionN ψ k)` — the Hilbert expectation of a diagonal observable is the eigenvalue-weighted sum of its ontic Born-region Fubini–Study volumes (`diag_expectation` Hilbert side + `fsMeasure_bornRegionN` unifying `fs_born_volume_ratio_N`/`_apex` via `Fin.lastCases`). Foundational triple, AxiomAudit-pinned, carving-free, Gleason-free. Generalises the single-qubit `sg_observable_correspondence` to all N + all eigenvalues. Both the pointwise-volume form (`observable_correspondence_diagonal`) and the **canonical integral form** (`observable_correspondence_diagonal_integral`: `⟨ψ,Aψ⟩ = ∫ A_ontic dμ` with `A_ontic = ∑ₖ lam k·𝟙_{Rₖ}` an explicit measurable Σ-function, `bornRegionN_measurableSet`) are landed. **Now extended to ALL self-adjoint observables** (`hermitian_observable_correspondence` / `_integral`): via the spectral theorem `A = U·diag(λ)·Uᴴ`, `⟨ψ,Aψ⟩ = ∑ₖ λₖ·vol(bornRegionN φ k)` where `φ = Uᴴψ` is the state transported by the spectral unitary (`hermitian_expectation_transport` + isometry `transport_norm`) — no separate §13 Σ-flow machinery needed. §14 observable correspondence **fully discharged**. Foundational triple, AxiomAudit-pinned. | `LF4-todo §14` | | **§14 *states* obligation** | **Pure-state / projector part DONE 2026-07-23** (`LF4/ObservableCorrespondenceN.lean`, `pure_state_born_prob_eq_volume`): `‖⟨Φ,ψ⟩‖² = vol(bornRegionN (Wᴴψ) 0)` — any pure-state / rank-one-projector Born probability (SuperdenseCoding's Bell projectors, Teleportation's input state) realised as an ontic Fubini–Study volume, via a unitary `W` with `W e₀ = Φ` (`exists_unitary_e_zero_eq`) + the unitary adjoint move (`inner_toEuclideanLin_adjoint`). Foundational triple, AxiomAudit-pinned. **Mixed-state / density-operator part also DONE 2026-07-23** (`Empirical/CSD/MixedStateBornVolume.lean`, `mixed_state_born_eq_ensemble_volume`): `ρ` realised as an ontic eigen-mixture — `Tr(ρ·|φ⟩⟨φ|) = ∑ᵢ (ρ.eigenvalues i)·vol(bornRegionN (Wᴴeᵢ) 0)`, composing `SigmaLayer.mixedEnsemble_capstone` + `LF2.born_quadratic` + `pure_state_born_prob_eq_volume`. §14 states obligation **fully discharged** (pure + mixed), foundational triple, AxiomAudit-pinned. **MIGRATED AND STRENGTHENED 2026-09-06 (CR-4, CL-074):** the mixed half is now stated on the fibred arena — `Tr(ρ·|φ⟩⟨φ|) = ∑ᵢ (ρ.eigenvalues i)·μ(globalBasin (momentContext _) 0)` at the transported ray `[Wᴴeᵢ]` — through the new `pure_state_born_prob_eq_basin`, and the genericity bundle `∀ i j, 0 < ‖⟨e_j, Wᴴeᵢ⟩‖²` is **gone**, not moved: the fibred route goes through the unconditional `globalBasin_born`, so every density operator and every pure outcome is covered, where before the outcome had to overlap every eigenvector of `ρ`. The base-side `pure_state_born_prob_eq_volume` is retained in `LF4/ObservableCorrespondenceN.lean` as the base-arena twin. | `BRIDGE-OBLIGATIONS.md` | | ~~**Axiom pin: `rankOneDensity_unique_of_certainty`**~~ | **DONE 2026-08-09** (pinned in `Tests/AxiomAudit/Foundations.lean` beside the effect-Gleason pin; `lake build CsdLeanTests` green). *(Original row:)* **QUEUED 2026-08-09** (landing-surface rewrite, axiom reconciliation). The theorem is proved and its footprint is the foundational triple (verified directly), but it carries no `#print axioms` pin in `Tests/AxiomAudit/`, so CI would not catch future drift. One-line addition to `Tests/AxiomAudit/Foundations.lean`. Not made during the rewrite because that pass does not touch Lean source. Effort **S**. | `specs/papers-vs-repo.md`, `TODO-MANUAL.md` | | ~~**★ Necessity audit of the constraint set**~~ | **DONE 2026-08-09 — `specs/necessity-audit.md`.** All thirty declared headlines (`CsdLean4/Headlines.lean`, CL-001…CL-030) classified against the five-class rubric, by the **Lean statement** rather than the docstring. Result: **1 necessity, 8 conditional necessity, 3 impossibility, 16 sufficiency, 2 instantiation.** The single unconditional necessity among the headlines is CL-024 `wigner_rigidity` (hypothesis: transition-probability preservation, nothing else). **The corpus's strongest-direction results are mostly NOT ledger rows** — `Matrix.StoneC1.stone_continuous`, `CV.no_exact_finite_ccr`, the three `RecordLayer/MeasurementConstraints.lean` no-gos (pure topology / pure measure theory, zero CSD hypotheses), and `compositeAlgReconstruction` are all absent from the ledger. **Verdict:** the constraint set pins the substrate **up to a large family, not uniquely**, and the family is bounded more by the exclusions than by the positive results; what *is* pinned tightly is the layer one level above the substrate (symmetry, dynamics, probability, composition), all of which take the complex Hilbert space or matrix algebra as given. **Two systematic conditioners identified as the highest-value structural targets:** the symmetry group `U(N)`, taken as given in every measure-forcing result, and transition-probability preservation, taken as a hypothesis in every Wigner-selection result on a `KahlerOnticSetup` (and *not* derivable from `flow_preserves_volume` — measure ≠ metric). **Structural find:** `LF1.OnticSetup.hΦ_pres` (Liouville) is declared and **never consumed** — only `measurable_Φ` is extracted, disclosed in `LF1/Setup.lean`; symmetrically `KahlerOnticSetup.flow_preserves_volume` enters no Wigner or Schrödinger conclusion. **15 docstring-overstates-statement findings** recorded on the page, headed by the general-`N` Born headlines not flagging that `bornRegion ψ` is **preparation-indexed** (the `qubitBorn` contrast is context-fixed and says so). ⚠️ **This row's own premise was wrong and is corrected on the page:** `composite_is_tensor_product` is the **sufficiency** half (`SigmaLayer/TensorSolved.lean` says so in a section headed "SUFFICIENCY here, UNIQUENESS in `TensorReconstruction.lean`"); the necessity converse is `compositeAlgReconstruction` / `composite_dim_eq`. The Lean and its docstrings were right; this ledger prose was not. No Lean changed; docs-only, eight guards green. *(Original row, superseded:)* The corpus constrains the ontic substrate from above, which is the right method given that the substrate is only observed through what it produces. But the constraints are of mixed logical strength, and nothing records which is which. *Sufficiency* results say a posited structure yields observed behaviour (local generators ⟹ light cone, `norm_commutator_spatial_factorial_le`; posited SU(n) symmetry ⟹ `μ_FS` unique). *Necessity* results say observed behaviour forces the structure (local tomography + locality ⟹ `⊗` — named here as `composite_is_tensor_product`, **which the audit found to be the sufficiency half**; the necessity converse is `compositeAlgReconstruction`). **Task:** audit every constraint-bearing headline, classify it sufficiency vs necessity vs conditional-necessity (and on what posit), and state what the accumulated set actually pins down: uniquely, or up to a family. This is a question about the constraints, not about the substrate, so it is answerable. Value: it is the honest measure of how much the reconstruction has achieved, and it is currently unmeasured. Effort **M** (audit + a `specs/` page; no new mathematics unless a gap is found). | `specs/reconstruction-status.md`, `AXIOMS.md` | | **★ Adopt the remaining upstream Lean checks** | **OPEN, scoped and measured 2026-08-11.** Survey of what other Lean repositories gate on, tested against this tree rather than estimated. **(1) `lake exe runLinter` — DONE 2026-09-04.** (Measured 2026-08-11: **122 findings in 7620 declarations** across 4 of 14 linters — `unusedArguments` 53, `defsWithUnderscore` 34, `docBlame` 20, `simpNF` 15.) Adopted as a **ratchet**: `simpNF` 0, `docBlame` 0, the two fix-on-touch classes pinned in `docs/std-lint-baseline.txt`, growth fails CI. Triage before gating: many `unusedArguments` hits are instance binders (`[FiniteDimensional ℂ E]`, `[NeZero N]`) kept deliberately for API uniformity, and `defsWithUnderscore` is a Mathlib naming convention this repo may not want. ⚠️ But **`unusedArguments` is exactly the class the necessity audit found BY HAND** (`LF1.OnticSetup.hΦ_pres` declared and never consumed) — an unused *hypothesis* is epistemically significant here, since it means a stated assumption does no work. That subset is worth reading as findings, not noise. `simpNF` (15) is the cheapest real win: a simp lemma not in normal form silently never fires. **(2) `lean4checker` — NOT installed; the biggest genuine gap.** Independent kernel re-verification of the built `.olean`s: `#print axioms` reads the environment, `lean4checker` re-checks it from scratch, catching metaprogramming that adds declarations, `implemented_by`/`unsafe` escapes, and tampering that axiom inspection cannot see. Needs building against the exact toolchain and re-checks Mathlib too, so it belongs on a **scheduled** run, not per-push. **(3) `shake` (unused imports) — NOT available** as an exe from the dependency; would need vendoring. **(4) `assert_not_exists` — DONE**, as `scripts/check-import-negative.sh`. **(5) mutation testing of the guards — DONE**, as `scripts/check-guards.sh`. | `prose-audit.md`, `necessity-audit.md` | | **★ Prose audit: reasons given for formal restrictions** | **IN PROGRESS 2026-08-11 — `specs/prose-audit.md`.** The C1 correction found two false claims that lived **only in prose**, and neither was reachable by any guard: both were prose giving a *reason* for a formal fact. `check-claim-provenance` targets claims *about objects*, so a wrong reason attached to a **true** statement trips nothing, and Lean cannot help because nothing in Lean states reasons. **Surface: 201 doc blocks** that both explain a restriction and mention a hypothesis, exclusion or degeneracy. **Pass 1** (quantitative prose): no defect; one candidate cleared (`Einselection.lean`'s `3/2` off-diagonal is correct — the witness is deliberately unnormalised, trace 5). **Pass 2** (SigmaLayer/LF4): ★ **one defect found and corrected.** Five modules and a BACKLOG row explained the `Ico 0 1` restriction in `fibreTypicality_uncovered` by "Lebesgue measure on the line is infinite" — wrong twice: `fibreTypicality` is `volume.restrict (Ico 0 1)`, a **probability** measure (`instIsProbabilityMeasure`, proved a dozen lines above), and the restriction was never forced (`fibreTypicality_uncovered_univ` proves the `univ` form on `ℝ`). The comparison is still real but different: it is about **where the mass one comes from** — on `ℝ` by fiat (`fibreTypicality_Ici_one`: the complement has infinite Lebesgue measure and zero typicality, so uncovered points are *excused by the measure*), on the compact fibre by Haar. ⚠️ **This BACKLOG row itself carried both readings at once**, the correct diagnosis three sentences before the wrong one — a defect being known somewhere does not stop it spreading. **Method that works, and the one to keep: when prose gives a reason for a restriction, try to prove the restriction unnecessary.** That converts an unfalsifiable sentence into a pinned Lean obligation. **~1/5 of the surface covered; no guard can close this class** — it needs a reader. | `prose-audit.md`, `c1-closure-report.md` | | ~~**Mathlib upstream batch 1**~~ | **SUPERSEDED 2026-08-06 by the B6 retirement** (author decision, `CONVENTIONS.md` §7: Mathlib PRs do not advance CSD and are not a need of this repository). This row predates that decision and was left standing by oversight; struck 2026-08-09. The staging discipline is untouched, the readiness record in the struck B6 row stands, and the option stays open for a future decision. *(Original row:)* **QUEUED 2026-08-02** (`MATHLIB-GAPS.md` is the public-facing cut). Self-contained staged files with no pause dependencies: Projectivization `Topology` + `MeasureSpace`, `StoneC1`, `DuhamelBound`, `PiecewisePreserving`, `QuantumInfo/{PartialTrace, TraceDistance}`. Effort per PR **S–M**; elapsed time weeks (review cycles). Value: credibility + de-risking toolchain drift (the corpus tracks a release candidate). | `MATHLIB-GAPS.md` | | **Lévy / spherical isoperimetry** (TH-1) | Canonical-typicality concentration (single-state typicality). Mathlib lacks spherical isoperimetry; the mean is proved. ~~Optional strengthening.~~ **RE-RATED 2026-08-20 (external physicist review): NOT peripheral** — typicality is the mechanism that turns volume ratios into single-trajectory observations, and the expectation-level TH-1 leaves that mechanism unproved. The actionable path was **Q24** — **LANDED 2026-08-21** (`fs_chebyshev_concentration`: Chebyshev-grade polynomial concentration via twirl-algebra second moments, no integrals and no isoperimetry); THIS row keeps only the exponential-rate strengthening, which genuinely waits on Mathlib isoperimetry | `thermo-plan.md`; Q24 (done) | | ~~**Continuity-only Stone**~~ | **DONE 2026-07-23** (`Mathlib/Analysis/Matrix/StoneC1.lean`, `stone_continuous`): a *strongly continuous* one-parameter unitary group is `t ↦ exp(t•A)` for skew-Hermitian `A` — differentiability **derived** (not assumed) via integral averaging (`∫₀ˢU` invertible near 0, by FTC + `s⁻¹•∫₀ˢU → 1`) + the change-of-variables `U t·∫₀^{s₀}U = ∫_t^{t+s₀}U`, then the C¹ core `eq_exp_of_hasDeriv`. Foundational triple, AxiomAudit-pinned. Closes W5-S2 fully. | `future-work.md` (W5-S2) | ## XL — Mathlib-scale (depth, not correctness) | Item | Status | Former source | |---|---|---| | **Manifold exterior calculus — the tower the four restricted terms wait on** (Kähler `dω=0` + top-power, symplectic gradient `X_H = ω⁻¹dH`, Liouville `ωⁿ/n!`, **and the moment map** — KG-1/2/3) | Genuine upstream gap; the volume is forced, the pointwise Kähler core is proved, and the moment-map equation is proved at the **linear** level (`IsPhaseHamiltonian`, `generatedRateField_eq_momentMap`), so this is formalization-depth, not a correctness gap. ⚠️ **`momentMap` added to this row 2026-09-06** — it is the same wall, and it is the most load-bearing of the four (`specs/TERMS.md`, `moment map`). **PLAN, staged, probed at the pin 2026-09-06.** Mathlib already HAS manifolds — this is not "build manifolds": `IsManifold` 108 declarations, `mfderiv` 177, `TangentBundle` 46, `extDeriv` 69 (**flat only**, on normed spaces), `AlternatingMap.domCoprod` 23. What is absent: `ContinuousAlternatingMap.domCoprod` **0**, a charted-space instance on `Projectivization` **0**, `SymplecticForm` **0**, `KahlerManifold` **0**, `MomentMap` **0**. So the missing piece is the differential-forms tower ON manifolds plus the symplectic layer above it, in this order: **(0)** `ℂℙ^{N−1}` as a smooth manifold — the standard affine charts `z ↦ (z₀/zᵢ, …, ẑᵢ, …, zₙ/zᵢ)`; self-contained, and the corpus already staged the topology, measurable space and metric upstream (**M**). ⚠️ **LANDED 2026-09-07** (`Mathlib/Geometry/Manifold/Instances/ProjectiveSpace.lean`), and at **`ω`** rather than `C^∞`. **(1)** the **continuous** wedge — adding boundedness to an existing algebraic construction (**S**), and it matters out of proportion to its size: without it the top-power identity `ω^{∧(N−1)}/(N−1)! = μ_FS` **cannot even be stated**. ⚠️ **LANDED 2026-09-07** (`Mathlib/Analysis/Normed/Module/Alternating/Wedge.lean`; witness `Kahler.extDeriv_fundamentalFormSq`, `d(ω ∧ ω) = 0`). ⚠️ **And the item's own name was wrong**: it is NOT `ContinuousAlternatingMap.domCoprod`, because the tensor-valued shape that name denotes cannot be stated at all — `F ⊗[𝕜] G` carries no topology at the pin. The landed form pairs the two codomains through a continuous bilinear map, which is what every use needs and which recovers the tensor version the moment a topology on `⊗` exists. A second absence surfaced en route and had to be built: `ContinuousAlternatingMap.domDomCongr`, without which an `ιa ⊕ ιb`-indexed product cannot re-enter the `Fin n`-indexed form API. **(2)** differential forms on manifolds — sections of the alternating-form bundle, pullback, exterior derivative; **upstream's own stated TODO** in its flat-forms file (**XL**). **(3)** top forms → measures: orientation + integration + comparison with `μ_FS`; Mathlib has `Orientation.volumeForm` flat and partitions of unity, not the manifold assembly (**L**). **(4)** symplectic / Kähler / moment-map definitions and enough theory for `X_H = ω⁻¹dH` and `ι_{X_i}ω = dΦᵢ`, then the `Tⁿ` action on `ℂℙⁿ⁻¹` (**L–XL**). ⚠️ **Steps (0) and (1) are the high-leverage cheap part and can be done independently**: they do **not** prove the top-power identity, they let it be **stated** — converting "cannot be said in Lean" into "said, and open", which is a real upgrade for an honest-scope register. Both are genuine standalone Mathlib contributions. ⚠️ **Nothing in flight waits on any of this.** The corpus routes around all four: `μ_FS` is pinned by uniqueness-under-symmetry (`fsMeasure_unique`), the Duistermaat–Heckman pushforward is proved by an independent Gaussian route (`fs_moment_pushforward_uniform`, a *discharged axiom*; `fs_moment_joint_dirichlet_N`), and the moment-map equation is proved at the linear level. What closing this buys is **self-containment** — the names become earned and the geometry internal rather than imported — not new results. Price it as a Mathlib programme, not a repo task. | `future-work.md` (KG-1/2/3); `specs/TERMS.md`; `MATHLIB-GAPS.md` (Kähler / symplectic manifold API) | | ~~Pointwise Birkhoff ergodic theorem~~ | **Do not pursue** — the single-flow route is a proved dead-end CSD does not take (NG1). | — | ## Research — physics-first, not formalization tasks | Item | Note | Source | |---|---|---| | ~~**SO-1 — "derive Σ / the sector origin"**~~ | **RETIRED as "derive Σ" — a non-question (2026-07-24).** Σ is the floor and is everything; nothing beneath it. **But *constraining* Σ (≠ deriving it) is legitimate, ongoing work:** Σ is hidden, so pin its structure/symmetry/measure down as tightly as possible from above (e.g. μ_FS *forced* as the unique SU(n)-compatible measure — `LocalisedTypicality`, Paper B). The frontier for completing QM is the record layer below. Kept struck so it is not re-proposed as a "derive the sector" goal. | `CSD-CHARTER.md`, `reconstruction-status.md §7` | | **★ The record layer (MD-1)** — measurement: context-fixed regions + the ontic record | **MD-1 CLOSED 2026-08-31** (author decision; reasoning recorded 2026-09-01). The machinery is complete: a measurement is a physical interaction that creates a record, and Lean carries the mathematical machinery showing how QM becomes physics via records. **Which** physical interaction an apparatus realises is a modelling input that can never be a Lean object — a permanent boundary, not a gap (`R-015`, and the same input Bohm and Everett take). One residual is genuinely open mathematics: the chart→arena transport of Hamiltonian generation (`R-016`; chart-level generation proved 2026-09-01, `SigmaLayer/ChartIntegralCurve.lean`). (Kinematic interface built 2026-07-25, `record-layer-plan.md` §4.) The record layer is formalized end-to-end, foundational-triple, no `sorry`, all axiom-pinned: fibre Born partition (`BornFibrePartition`), outcome distribution + flow⟹Born bridge (`DeIsolationFlow`), the P5 `RecordSemantics` (`FibreRecord`), the closure bundle (`RecordLayerClosure`), the **rates = the Kähler torus moment map** (`MomentMapRace`, `bornRate_eq_momentMap`), the measurement-as-`context+microstate→record` object (`Measurement`), **Born = the law of large numbers over the unknown microstate** (`Measurement.bornMeasurement_frequency` — no dynamical postulate; randomness = ignorance of the initial condition), the migration onto the corpus's real Σ (`ProjectiveRecord`, `projRecord_frequency`), the base×fibre ontic space `Σ = CPN × ℝ` with epistemic base / ontic fibre (`FibredSigma`), arbitrary observable (`BasisMeasurement`), and — the wire-in — `KSigmaRecord.born_frequency_region_eq_record`: the region `FiniteQMClosure.born_frequency` lands in **is definitionally the record-layer event** on the actual `Σ = KSigma`. **Only residue:** the pinned `unifiedFiniteQMClosure.records_time_physical` field still *names* the coarse `vnPointerOutcome` (a block-sum of these events) — rewriting it carries no new theorem and is deliberately not done. ⚠️ **SCOPE, corrected 2026-07-28 (external review).** What is built is the **kinematic** record interface: *given* a basin partition, an unknown microstate selects one definite record and repeated i.i.d. preparations obey the LLN. Three things are **not** established. (i) **General-`N` A7 is open** — the partition is built from the preparation (`bornContext ψ`, `cdfCell (bornRate ψ)`), so `Ωᵢ(M)` fixed by the apparatus alone is proved only at `N = 2`; see the re-opened row above. ⚠️ **UPDATED 2026-07-31 — this splits in two, and one half is closed.** `RecordLayer/GlobalBasin.lean` + `GlobalRecordClosure.lean` give a **fibred** context-fixed partition at every `N` (`globalBasin` reads a `ContextField`'s rate *at the ontic point*; the record event is a function of `(context, outcome, time)` alone), so the **preparation-indexing** half is closed and the statement above now describes only the older `FibreRecord`/`FiniteQMClosure` path. The **base-only** half — whether `Ωᵢ(M)` can live on `ℂℙⁿ⁻¹` itself — is the ⏸ parked chain and stays open. Which of the two Paper C intends is a question about the axiom. (ii) **The dynamics are not derived** — `DeIsolationInteraction.basin_rate` is a *hypothesis field*, and no `H_int` generating those basins is constructed (`DeIsolationFlow.lean` says so; an earlier staging row calling this "DISSOLVED" was wrong and is retracted). (iii) The fibred `Σ = ℂℙⁿ⁻¹ × ℝ` is a **measurable record model**, not shown to be an **A1 ontic sector** — non-compact fibre, no Kähler structure, and its measure not shown to be a Liouville measure. ⚠️ *Sharpened 2026-07-30: "no Kähler structure" here is a **parity** fact, not a missing-tooling one — `ℂℙⁿ⁻¹ × ℝ` has odd real dimension `2n-1`. See the ★★ row.* So this row is "kinematic interface DONE", not "MD-1 discharged". | `record-layer-plan.md §4`, `CSD-CHARTER.md` | | ~~**★ Qubit context-fixed Born rule (A7 faithfulness)**~~ | **DONE 2026-07-26** (`LF4/QubitBorn.lean` `qubitBorn`, foundational-triple, pinned, `check-claims.sh` PASS). Discharges the A7 residual flagged by the record layer: makes the epistemic partition genuinely **context-fixed** (a function of the measurement axis `n` alone, *not* the preparation — unlike the corpus's preparation-indexed `bornRegion ψ`) and *derives* the Born weight from the ontic Fubini–Study typicality volume: `∫ ½(1 + rsign(2·blochProj n − 1))·4(2·blochProj ψ − 1)₊ dμ_FS = |⟨n\|ψ⟩|²`. 7-module `CP¹` chain (all foundational-triple, all pinned): `QubitReflection` (reflection identity) → `BlochProjection` (general-axis Born weight + `U(2)`-equivariance) → `AxisBridge` (general-axis hat-box) → `QubitDipole` (`R_n=2\|n⟩⟨n\|−I` unitary; dipole `=(2c−1)/2`) → `QubitCrossTerm` (antipode via Haar right-invariance; cross-term `=0`) → `QubitBorn` (assembly), + `HatBox`. Next: general-`N` context-fixed partition (foundation is `N`-agnostic; `CP¹=S²` reflection geometry is the `N`-specific part). | `record-layer-plan.md §2/§4`, `reconstruction-status.md` A7 | | **★ General-`N` BASE-ONLY A7** *(demoted from axiom-gating to special-case question, 2026-08-02)* | **DECISION 2026-08-02: the fibred reading of A7 is canonical (author call), so A7 itself is DISCHARGED at every `N` — kinematically and dynamically (v0.7.0). This chain now characterises whether a BASE-ONLY realisation also exists at `N ≥ 3` — the qubit-special-case question — and no longer gates the axiom. Content below unchanged.** **⏸ PARKED 2026-07-29 — a well-characterised open problem. DO NOT RESUME without a new idea.** *Parked, not refuted.* The decision-relevant output is in hand: a five-step **derived** constraint chain (`SigmaLayer/ContextFixedA7*.lean`) forcing any base-only density onto measure `≤ 1/n` **and** above `(n−1)/n`, with the `N = 2` / `N ≥ 3` threshold confirmed **three independent ways** and the bound sharp and attained at `N = 2`. Remaining path is blocked: `hdense` (M) → generic-`ψ` (L) → harmonic analysis on `ℂℙⁿ⁻¹` (**XL, Mathlib has no API**). One refutation route is closed too — the Fubini–Study triangle inequality is *exactly* tight (`arccos√σ + arccos√(1−σ) = π/2`), so it yields nothing; the system sits **on** the boundary, which is why neither direction falls out. **Rationale for parking: proving the no-go and assuming it lead to the same next action** — the fibre carries contextuality at `N ≥ 3` either way. Canonical record: [`sigma-fibre-contextuality.md`](sigma-fibre-contextuality.md). **Successor: the fibre/A1-sector row below.** *(Superseded detail retained:)* **OPEN — re-opened 2026-07-28 after an external review. This row previously read "NOT A TARGET — provably dead"; that was an overclaim and is retracted.** What actually exists against base-only regions at `N≥3` is (a) a **numerical** experiment (`scripts/experiments/`, FS-Voronoi density forced negative at `N=3`) and (b) an **informal** operator argument (the affine dipole reproduces Born but is signed; the nonneg-restoring monopole is ψ-independent per cell only via the qubit `CP¹=S²` antipode, which has no `N≥3` analogue). Numerics plus an unformalised argument is a **conjecture**, not a no-go — and "provably" had no business in a corpus whose claim to attention is that its results are machine-checked. **Two admissible routes, no-go first:** (1) **prove the no-go** as a Lean theorem under explicitly stated hypotheses (covariance, non-negativity, base-only dependence, the region family's regularity) — if it goes through it settles the architecture, and if it fails the failure names the hypothesis to drop; (2) otherwise **construct** `{Ωᵢ(M)}` depending on the measurement context alone, together with a preparation law `ρ_ψ^ep`, satisfying Paper C A7's `P(i∣M,ψ) = ∫_{Ωᵢ(M)} ρ_ψ^ep dμ_FS`. **Step one LANDED 2026-07-28** (`SigmaLayer/ContextFixedA7.lean`) — the **support reduction**. Evaluating A7 at the `n` basis-vector preparations `ψ = eⱼ` (Born weights `δᵢⱼ`) makes a **non-negative** integrand integrate to **zero** over `Ωᵢ` for `i ≠ j`, forcing it to vanish a.e. there (`ae_eq_zero_of_setIntegral_eq_zero` — non-negativity does all the work, exactly what a *signed* density escapes). Hence: each support lies in its own region (`overlapSupport_ae_subset`), the `n` supports are pairwise a.e. disjoint (`overlapSupports_ae_disjoint`), and their measures sum to `≤ 1` (`sum_measure_overlapSupport_le_one`) — so each is `≤ 1/n` by symmetry while still carrying total integral `1`. **A base-only density must spike**, which is what the `N = 2` solution `4(2s−1)₊`, supported exactly on `(½,1]`, does. Stated over an abstract probability space, so it survives a move to a fibred Σ. **Step two LANDED 2026-07-28 — the CAP.** If `S_g` contained a positive-measure set of overlap values **below `½`**, two coordinates could both land in it (they would sum to `< 1`, so nothing forbids it) and at `N ≥ 3` a third coordinate absorbs the remainder, so such states occur with **positive** measure — putting positive measure in `Aⱼ ∩ Aₖ` and contradicting step one's disjointness. Hence `cap_of_joint_nondegenerate`: **`g` vanishes a.e. below `½`.** Sharp and attained — the `N = 2` solution `4(2s−1)₊` is supported exactly on `(½,1]`. ★ **And it makes the qubit's escape precise:** `joint_degenerate_of_sum_eq_one` — at `N = 2` the two Born weights exhaust the state (`sⱼ + sₖ = 1`), so they can *never* both be below `½`; the set is literally **empty** and the abundance hypothesis fails identically. That degeneracy *is* the escape route, and it closes at `N ≥ 3` where the coordinates stop being functionally dependent. `base_only_density_confined` assembles both halves: the density lives on measure `≤ 1/n` **and** only above `½`, while still integrating to `1`. The abundance input (`hjoint`) is an explicit hypothesis rather than derived — deriving it is the Dirichlet pushforward of `μ_FS`, real work and orthogonal to the argument; stating it that way is what makes the `N = 2` contrast visible. **Step three LANDED 2026-07-28 — the cap is UNCONDITIONAL at `N ≥ 3`** (`SigmaLayer/ContextFixedA7FS.lean`). `fs_joint_abundance` discharges the abundance hypothesis for the actual Fubini–Study measure via the corpus's own pushforward `fs_volume_eq_dirichlet_inter` (`μ_FS` pushes to the uniform/Dirichlet measure on the open simplex); abundance then reduces to Lebesgue positivity on `Fin M → ℝ` with an explicit witness — the two chosen coordinates in `T`, every other in a small `(0,ε)`. The subtlety handled by `exists_trunc_of_volume_pos`: `T ⊆ (0,½)` bounds `tⱼ + tₖ < 1` *pointwise but not uniformly*, so one first passes to a positive-measure part of `T` bounded away from `½`. ★ `fs_cap_unconditional`: **a base-only, `U(N)`-covariant, non-negative preparation density reproducing Born on the FS sector vanishes a.e. on overlap values below `½`** — no hypothesis left over. **`N ≥ 3` is exactly the threshold**: `M = N−1` free simplex coordinates, and two *distinct* ones exist precisely when `M ≥ 2`; at `N = 2` the second Born weight is `1 − s₁`, functionally dependent. The dimension count that powers the result is the same one that exempts the qubit. This replaces the **numerical** evidence the retracted "provably dead" row rested on with a **derived** constraint. **Step four LANDED 2026-07-29 — the CONSTRUCTION ATTEMPT, and the first generic-`ψ` input.** Asked to *build* `{Ωᵢ(M)}` + `ρ_ψ` rather than refute it, the sharpest untried lever turned out to be **orthogonal preparations**: `ψ ⊥ eᵢ` gives Born weight `0`, so non-negativity forces `g ∘ sψ` to vanish a.e. on `Ωᵢ` — for a whole *family* of `ψ`, not one. `orthogonal_preparation_vanishes` + `vanishes_on_interval_of_dense` (`SigmaLayer/ContextFixedA7.lean`). The geometric input: for unit `ψ` in the sphere of `eᵢ^⊥`, `|⟨ψ|φ⟩|²` peaks at the normalised projection of `φ` into `eᵢ^⊥` with value `1 − sᵢ(φ)`, and tilting `ψ` inside `eᵢ^⊥` sweeps it continuously to `0` — **which needs `dim eᵢ^⊥ ≥ 2`, i.e. `N ≥ 3`.** At `N = 2` the orthocomplement is a line and only one value is realised. **Third independent appearance of the same threshold.** **Consequence (analysis done, covering step not yet formalized): `g` vanishes below `(n−1)/n`** — sharpening the cap from `½`, and at `n = 2` reading "below `½`", exactly where `4(2s−1)₊` lives. ⚠️ **THE CONSTRUCTION DID NOT SUCCEED, and one plausible refutation route is also closed.** The natural kill — a state `ψ` whose cap cannot reach a region `Ωᵢ` with small `sᵢ(ψ)` — fails: the Fubini–Study triangle inequality gives `arccos√σ + arccos√(1−σ) = π/2` **exactly**, so the bound it yields is vacuous. The constraint system is critical rather than contradictory, which is why neither a construction nor a refutation falls out. **Step five LANDED 2026-07-29 — `(n−1)/n` is now a Lean theorem** (`vanishes_below_of_balanced`): given step four's output and the balanced-state abundance, `g` vanishes on every overlap value below `(n−1)/n`. Short for a structural reason — step four's conclusion is *pointwise in `g`* (it dies on a whole interval, for a.e. `φ`), and `g` is a fixed function, so **one** suitable `φ` suffices; the helper `exists_mem_of_measure_pos_of_ae` is the whole bridge from measure to pointwise. ⚠️ **STILL CONDITIONAL for `μ_FS` on one input, NOT landed:** balanced-state abundance — that for every `c > 1/(M+1)` the states with *all* moment coordinates `≤ c` are non-null. True, and the proof shape is settled (`fs_volume_eq_dirichlet_inter` reduces it to Lebesgue positivity; a barycentre box of side `min(b, c−b)/(M+1)` with `b = 1/(M+1)` witnesses it, the two constraints being `M·δ < b` to stay inside `∑ t < 1` and `M·δ < c − b` to keep the dropped coordinate below `c`). **LANDED 2026-07-29 on the second attempt** (`fs_balanced_abundance`). The first attempt failed across four iterations because `b`, `δ` were introduced with `set`, making them opaque local definitions `linarith` cannot see through; the fix was to **split geometry from arithmetic** — `box_in_simplex` takes `b`, `d` as *abstract* reals constrained by linear relations plus the single identity `M·b = 1 − b`, with the concrete choice made once outside. ⚠️ **This does NOT make `(n−1)/n` unconditional.** Step four's *other* input — `hdense`, the tilt fact that unit `ψ` in the sphere of `eᵢ^⊥` sweep overlaps across `[0, 1 − sᵢ(φ)]` — remains a hypothesis and **has not been attempted**; it needs genuine projective/inner-product geometry, not arithmetic. **Effort: M.** ⚠️ **And A7 REMAINS UNPROVEN IN BOTH DIRECTIONS.** Everything in this chain is a *necessary condition* on `g`; none of it shows a base-only construction is impossible, nor exhibits one. After `hdense`: the residual generic-`ψ` gap (L) and the harmonic argument (XL, blocked on Mathlib). Foundational-triple, pinned. ⚠️ **This is NOT the no-go** — it constrains `g`, it does not refute it. **Two things remain:** (a) the **generic-`ψ`** requirement is entirely untouched (everything so far comes from the `n` basis-vector preps, and the suspected obstruction lives at generic `ψ` where the cap around `ψ` straddles several regions); (b) the **harmonic argument** — `g(|⟨ψ|φ⟩|²)` integrated over a fixed region has components of every degree `(k,k)` while the target `|⟨eᵢ|ψ⟩|²` is pure `(1,1)`, and killing the higher harmonics for all regions and contexts is what should fail at `N ≥ 3` — needs representation theory / harmonic analysis on `ℂℙⁿ⁻¹` that **Mathlib does not have**, so it is XL-to-infeasible as things stand. **Status of what is proved:** `N = 2` is genuinely done (`LF4/QubitBorn.lean` `qubitBorn`). `N ≥ 3` is unresolved in either direction. Note the record layer does **not** discharge this: its partition is built from the preparation (`bornContext ψ`). See also the `hpos` finding below — the DH region machinery is blind to vanishing amplitudes, a second structural limit at the same place. | external review 2026-07-28; `record-layer-plan.md` | | **Track B — quantum relaxation** (Valentini H-theorem) | The only route to *new predictions* past the "empirically identical" ceiling. | `project` note | | **CV chain** — continuous spectra / the EFT direction | Extend Born-as-volume past finite `ℂℙⁿ`. **Stage 0 DONE 2026-07-25** (`CsdLean4/CV/OscillatorBorn.lean`): the truncated single mode is now a genuine record-layer measurement — its number/energy measurement = the standard-basis `Measurement` (number basis = standard basis, `hamiltonian` diagonal), `numberMeasurement_prob` = `‖⟨n\|ψ⟩‖²`, `numberMeasurement_frequency` = Born as the LLN over the unknown microstate, and `numberBornProb_embed` = **cutoff-independence** of the Born content (the Born analogue of `oscEnergy_cutoff_independent`). **Stage 1 DONE 2026-07-25** (`CsdLean4/CV/FieldModes.lean`): a free scalar field at a cutoff = finitely many modes as the (truncated) tensor product `FieldSpace K N = EuclideanSpace ℂ (Fin K → Fin N)` (configs). `fieldHamiltonian_mulVec_single` (free field = sum of oscillators, diagonal, eigenvalue `∑ₖ oscEnergy (c k)`), `fieldEnergy_cutoff_independent`, `sum_fieldBornProb_unit` (config Born distribution), `norm_sq_tprodState` (product state `‖⊗ψₖ‖²=∏‖ψₖ‖²` — the composite/tensor structure), and **`modeMarginal_tprod_unit` = mode-wise Born** (the marginal of a product state = the single-mode Born weight `‖ψ_{k₀} n‖²`). **Stage 2a DONE 2026-07-27** (`CsdLean4/CV/Dispersion.lean`): relativistic dispersion `ω(m,p)=√(p²+m²)`. `omega_sq_sub_sq` = the **mass shell** `ω²−p²=m²` (the Lorentz-invariant content, and why `m` is the mass); `abs_le_omega` (`|p| ≤ ω` — excitations do not outrun the light cone); `abs_mass_le_omega` + `omega_zero` (the **mass gap** `|m| ≤ ω`, attained at rest); `omega_massless` (`ω=|p|`, the light cone exactly); `omega_le_newtonian` (`ω ≤ m + p²/2m` — the non-relativistic limit as a clean inequality, no asymptotics); `omega_mono`. On the field: `relFieldHamiltonian_mulVec_single` + `_isHermitian` (still **diagonal** in the configuration basis, so the `OscillatorBorn` record-layer account carries over verbatim — only the eigenvalues change), **`relFieldEnergy_quantum`** (the headline: one quantum in mode `k₀` costs exactly `ω(m, p k₀)`, so the excitations *are* relativistic particles of mass `m` — the dispersion is about the particle content, not a parameter choice), `relFieldEnergy_vacuum` (zero-point `½∑ω`), `relFieldEnergy_cutoff_independent`. **Stage 2b DONE 2026-07-27** (`CsdLean4/CV/ModeLocality.lean`): the **Haag–Kastler kinematic locality axiom** at the finite cutoff — `commute_of_disjointSupport`, operators supported on **disjoint mode sets commute**, so observables of disjoint regions are jointly measurable and the record layer can assign them outcomes simultaneously (proof = uniqueness of the intermediate configuration). Non-vacuous: `modeOp_supportedOn` (every single-mode matrix inhabits `SupportedOn {k₀}`) + `commute_modeOp` (the concrete instance at distinct modes). ⚠️ **HONEST SCOPE:** this is *subsystem* locality — spatial only under the position-space reading of the modes (`CV/Position.lean`); under the momentum reading the modes are delocalised and mode-disjointness is not spatial separation. **Continuum microcausality `[φ(x),φ(y)]=0` at spacelike separation is NOT proved and does NOT hold exactly at a finite cutoff** — it needs the continuum limit, deliberately deferred (`ApproxCCR.no_exact_finite_ccr`). **CV-5 DONE 2026-08-07** (`CsdLean4/CV/FreeFieldFloquet.lean` + `Empirical/CSD/QuantumChaos/FreeFieldClosure.lean`): the free field's **stroboscopic dynamics** — `freeFieldU K N τ` = the diagonal-phase unitary `e^{-iτE(c)}`, ★ `freeFieldU_eq_exp` proves it **IS** `exp (-(iτ)•H_field)` (generation by the field Hamiltonian is a theorem, not a gloss), `freeFieldFloquet` plugs it into the §H2 `FloquetEvolution` interface (information preservation, projective dynamics, Loschmidt diagnostics apply verbatim), and ★ `freeField_pilotClosure` lands the full §H3 pilot closure for every cutoff through the `freeFieldUFin` reindex (`card_fieldConfig = N^K`) — the CV vertical and the quantum-chaos vertical meet on the same closure instance. **CV-6 DONE 2026-08-07** (`CsdLean4/CV/DynamicalLocality.lean`): **dynamical mode locality** — Heisenberg conjugation `heisenberg U A = U†AU` by any mode-additive diagonal phase preserves `SupportedOn` (`heisenberg_phaseDiagU_supportedOn`: the conjugating phase difference collapses to the support modes), so every free-field period preserves support (`heisenberg_freeFieldU_pow_supportedOn`) and ★★ `commute_heisenberg_freeFieldU_pow`: **disjointly supported observables still commute after ANY number of periods** — the Stage-2b Haag–Kastler axiom is stable under the free dynamics. ⚠️ Honest scope: free (mode-diagonal) drive only; interacting drives can spread support — that spreading is the scrambling the §H diagnostics track. **Stage 3a+3b DONE 2026-08-07** (plan: [`cv-stage3-plan.md`](cv-stage3-plan.md); rows ~~CV-7~~/~~CV-8~~ in `future-work.md`): the **interacting drive** (`CV/Interaction.lean` — diagonal pair couplings, EXACT step, ★ `interactingU_eq_exp`, §H3 closure via `interacting_pilotClosure`) and the **local-algebra light cone** (`CV/LocalAlgebra.lean` + `CV/SupportSpreading.lean` — `SupportedOn` is a unital *-subalgebra, disjoint couplings act trivially, ★★ `heisenberg_graphInteractingU_pow_supportedOn`: `n` periods spread support at most onto the coupling graph's `n`-ball, one edge per period, with the `K=N=2` non-vacuity witness `spreadKick_not_supportedOn`). **Stage 3c DONE 2026-08-07** (~~CV-9~~, `CV/InteractionPrice.lean`): locality violation **priced linearly in the coupling** — ★★ `heisenberg_interactingU_near_supported` (within `2|τ||λ|C‖A‖` of an `S`-supported operator), with `freeField_perturbed_exp_dist_le` covering non-diagonal (hopping) perturbations too (Duhamel engine generalized to any finite index; diagonal L2-norm brick staged). **Stage 3d DONE 2026-08-07** (~~CV-10~~, `CV/PowerCounting.lean`): finite-cutoff **power counting** — ‖Q‖ ≤ √(2N) via the C*-identity (no spectral asymptotics), field embedding norm-nonincreasing, and the grade-m interaction Q_k^m priced at ≤ |τ||λ|√(2N)^m with ★★ cutoff-UNIFORM price under the renormalized coupling λ₀/√(2N)^m — relevant/irrelevant as a theorem about price-bound scaling (upper bounds only; no RG, no continuum). **Stage 3 (CV-7..CV-10) COMPLETE** — the EFT interaction stage is closed at the cutoff. Recorded residues (not queued): non-diagonal light cone (exp-closure route, plan §3b prose), price attainment (InteractionPrice ledger note). **Stage 4 COMPLETE 2026-08-09 (CV-11..CV-16)** ([`eft-stage4-plan.md`](eft-stage4-plan.md); rows CV-11..CV-16): non-diagonal light cone (exp-closure), matrix Lie–Trotter (constructible arbitrary-Hermitian drives), ★★ the finite free propagator (`freeTwoPoint` at the dispersion frequency) + priced corrections, boost covariance of the mass shell, the renormalization-trivial class, and RG matching as a gated decision point. Landed: the non-diagonal (kicked) light cone via exp-closure; matrix Lie–Trotter (staged, no Trotter in Mathlib at pin) making arbitrary-Hermitian drives constructible; ★★ the **finite free propagator** `⟨vac|Q_k(n)Q_l|vac⟩ = ½e^{-inτ}δ_{kl}` — the chain's first computed correlation function — with the interacting correction priced; boost covariance of the mass shell (the Lorentz content as a theorem); and the renormalization-trivial class (relevant/irrelevant now a theorem on both sides). CV-16's gate changed its own statement: exact unitary RG matching is IMPOSSIBLE for support-spreading drives (`exists_unitary_compress_not_unitary` — decimation loses norm), so the effective low-cutoff theory is necessarily open; exact matching holds for the occupation-defined class (`compressCfg_interactingU`), and the RG proper is re-scoped to a channel-level statement with an error budget (research, unqueued). **Stage 5 COMPLETE 2026-08-09 (CV-17, CV-18, CV-19)** (`CV/LiebRobinson.lean`: the Heisenberg flow's derivative, the interaction-picture split, and the **linear** bound `‖[A(t),B]‖ ≤ 4|t|‖T‖‖A‖‖B‖` — information cannot leave a region instantaneously, at a rate set by the cut coupling alone; the field instantiation takes `[A,B]=0` from CV-8). CV-19's gate was then run: the pass found a concrete route to the **Grönwall (exponential-in-time)** bound, which landed (`norm_commutator_gronwall_le`, prefactor `‖[T,B]‖`, so a coupling that does not reach the probe contributes nothing at any time — the seed of a light cone). A second pass landed the **combinatorial half of the spatial cone** (nested commutators stay in the `k`-ball and commute with a distant probe exactly below the graph distance), and a third landed the **analytic half** (the Taylor remainder of the adjoint series, bounded by the mean-value inequality). ★★★ `norm_commutator_spatial_le` assembles them: `‖[A(t),B]‖ ≤ 2‖A‖‖B‖(2‖S‖|t|)^d` for every `d` whose ball has not reached the probe, so for `2‖S‖|t| < 1` the commutator **decays geometrically in the graph distance** and the propagation speed is bounded by the coupling strength. A fourth pass took the recorded strengthening: `norm_flowRemainder_le_factorial` (integral estimate in place of mean-value) gives ★★★ `norm_commutator_spatial_factorial_le`, `‖[A(t),B]‖ ≤ 2‖A‖‖B‖(2‖S‖t)^d/d!`, **the textbook Lieb–Robinson shape**, decaying in the graph distance at every time. **Stage 5 is complete with nothing recorded against it.** No velocity constant is extracted or claimed optimal. **PLANNED 2026-08-09** ([`eft-stage5-plan.md`](eft-stage5-plan.md); rows CV-17..CV-19): the feasibility verdict is split and stated up front — CV-17 (commutator flow) and ★★ CV-18 (the **linear** bound: `‖[τ_t(A),B]‖ ≤ 4|t|‖H−H_X‖‖A‖‖B‖`, i.e. no instantaneous signalling quantitatively) are tractable now on the existing `DuhamelBound` pattern with CV-8 supplying `[A,B]=0`; CV-19 (the exponential/velocity form) is XL research with an **abort criterion agreed in advance** — stop at CV-18 if the iteration step has no concrete route after one focused pass. To our knowledge Lieb–Robinson is unformalized in any proof assistant. The continuum posture stays deliberately deferred (`ApproxCCR.no_exact_finite_ccr`). **Stage 6 OPENED 2026-08-13** ([`eft-stage6-plan.md`](eft-stage6-plan.md)): ~~CV-20~~ the **velocity** extracted same day (★★★ `norm_commutator_velocity_le` — the Stage-5 non-claim discharged, constant `2e²‖S‖` not claimed optimal); ~~CV-21~~ vacuum clustering DONE same day (`vacuum_clustering` — the statics companion); ~~CV-22~~ Wick four-point gate PASSED same day, landed pattern-resolved (`eqFourPoint` table). **Stage 6 COMPLETE 2026-08-13.** | `project` note | ## Hardening / conventions (from the 2026-07-20 Lean-QIT / Physlib comparison — `CONVENTIONS.md §8`) | Item | What / why | Size | |---|---|---| | ~~**Category tags, canonical and guarded**~~ | **DONE 2026-09-12.** The `**Category:**` slot had drifted into free prose: 28 spellings (`6-Local`, `CV`, `dynamical measurement`, `conceptually 1-Mathlib`, `2-LF4`, `1-Mathlib-staging`, …) and six modules with no tag. Every module now carries one of CONVENTIONS §2's five tags, decided by directory (§4), with the former phrase folded into the rationale parenthetical so nothing was lost (185 modules rewritten, 6 tags inserted; `RecordLayer/` is `7-SigmaLayer` with `SigmaLayer/`, one stratum since the Q15 split — CONVENTIONS §2 now says so). New guard `scripts/check-category-tags.sh` (one awk pass, 0.5 s; CI; three probes in `check-guards.sh`) keeps it that way. Supersedes W12's two-module re-tag. | **S** | | ~~**Zero-`axiom` CI gate**~~ | **DONE 2026-07-21**. `check-claims.sh` now sets `DECLARED_AXIOMS=""` and fails on ANY `^axiom ` under `CsdLean4/` (the whitelist is empty now that `busch_effect_gleason` is discharged). Gates the Physlib route. (`CONVENTIONS.md §8.1`) | **S** | | ~~**Discharge `busch_effect_gleason`**~~ | **DONE 2026-07-21 — axiom deleted, corpus imports ZERO axioms.** Proved as `OperationalPackage.effect_gleason_representation` (`LF2/EffectGleason.lean`, foundational-triple, AxiomAudit-pinned); the `axiom` was removed from `BornWrapper.lean` and its consumers `born_form_of_package` / `pure_state_born_weights_of_certainty` relocated to `EffectGleason.lean` (signatures unchanged). Step 4 pieces: `qmatrix_posSemidef`, `qmatrix_trace_one`, `qdensity`, `qdensity_unique` (via the polarisation lemma `matrix_eq_zero_of_quadForm_zero`). Guards updated (`check-claims.sh` zero-axiom gate, `AxiomAudit` pins, `AXIOMS.md §2.2`). Full history below. — Finite-dim effect-Gleason → "three axioms, zero imported". **Steps 1–2 LANDED** (`LF2/EffectGleason.lean`, 2026-07-20): step 1 foundational layer (`Effect.smul`, `p_zero`, `p_mono`, `p_smul_add`, `p_smul_mono`) + **step 2 homogeneity** `p(t•E)=t·p E` on `[0,1]` (`p_smul_homog`, via `smulVal_natMul`→`smulVal_rat`→`smulVal_homog`, the monotone-Cauchy density squeeze) — all foundational-triple, no `sorry`. **Step 3 spectral reduction LANDED**: `p_finsetSum` (finite additivity), `Effect.eigenvalues_le_one`, `Effect.sum_eigenEffect_M`, `p_eq_eigen_sum` (`p(E)=∑ᵢλᵢ·p(\|eᵢ⟩⟨eᵢ\|)`) — reduces the problem to `p` on rank-one projectors, foundational-triple, no `sorry`. **Step 3b polarisation identities LANDED**: `outerProduct_parallelogram` (`|u+v⟩⟨u+v|+|u−v⟩⟨u−v|=2|u⟩⟨u|+2|v⟩⟨v|`, cross terms cancel) + `outerProduct_polarization_real` — the algebraic core letting `p` inherit the parallelogram law, foundational-triple, no `sorry`. **Step 3b groundwork LANDED**: `outerEffect` (sub-unit rank-one `\|v⟩⟨v\|` for any `‖v‖≤1`, via `1-\|v⟩⟨v\|=(1-\|v̂⟩⟨v̂\|)+(1-‖v‖²)\|v̂⟩⟨v̂\|`), `outerProduct_smul`, and `p_outerEffect_smul` (degree-2 homogeneity `p(\|c·v⟩⟨c·v\|)=c²·p(\|v⟩⟨v\|)`) — foundational-triple, no `sorry`. **Step 3b `p`-parallelogram LANDED**: `one_sub_two_outerProduct_posSemidef` (CS sum-of-projectors `≤I`, not the feared eigenvalue bound) + `p_outerEffect_sqrt2` (√2-doubling) + **`p_parallelogram`** (`p(|u+v⟩⟨u+v|)+p(|u−v⟩⟨u−v|)=2p(|u⟩⟨u|)+2p(|v⟩⟨v|)` for `‖u‖²+‖v‖²≤½`) — foundational-triple, no `sorry`. **Step 3b-final (ρ-build) LANDED 2026-07-21**: `qform` (`q v=‖v‖²·p(\|v̂⟩⟨v̂\|)`, the degree-2 homogeneous extension off the unit ball) with `qform_eq_p` / `qform_smul` / **`qform_parallelogram`** (unrestricted parallelogram law); the Jordan–von Neumann core `qpolar_add_half`→**`qpolar_add_left`** (bi-additivity) and **`qpolar_smul_real`** (`ℝ`-homogeneity) via the new general lemma **`additive_bounded_linear`** (bounded additive `ℝ→ℝ` is linear — replaces the classical *continuity* step, which is unavailable since `p` is an arbitrary assignment, by the bound `0≤q≤‖·‖²`); the complex polarisation **`qsesq`** `S(u,v)=¼f(u,v)−(i/4)f(u,i·v)` shown sesquilinear (`qsesq_add_right`, `qsesq_smul_right`, `qsesq_conj_symm`) with `qsesq_self : S(v,v)=q v`; and the matrix **`qmatrix`** `R j k=S(eⱼ,eₖ)`, Hermitian (`qmatrix_isHermitian`), with **`p_outerEffect_eq_trace`** (`p(\|v⟩⟨v\|)=Tr(R·\|v⟩⟨v\|)`) and — through the spectral reduction — **`p_eq_trace` : `p E = Tr(R·E)` for EVERY effect**. All foundational-triple, no `sorry`. **Remaining (step 4 only):** `R.PosSemidef` (from `OP.nonneg`), `R.trace=1` (from `OP.total_one`), uniqueness (non-degenerate trace pairing) → package `R` as a `DensityOperator` and rewrite the axiom as a theorem in `BornWrapper.lean`; then update `AXIOMS.md §2.2` + add the zero-`axiom` CI rule. Cosmetic for CSD (NG2) but clears the last `axiom` + gates Physlib. | **S–M** | | **`REFERENCES.json` + line-precise citations** | Machine-readable provenance; docstrings cite `[Key, file:Lstart-Lend]` incl. exact CSD-preprint lines. Biggest auditability win. (`§8.2`) | **M** | | **`_statement`/`_of_`/final-theorem pattern** | Turn `BRIDGE-OBLIGATIONS.md` prose into explicit `_of_` hypotheses; discharging an obligation = removing a hypothesis. (`§8.3`) | **M** (incremental) | | ~~**`autoImplicit=false`**~~ + module-system migration + tagged Mathlib pin | **`autoImplicit=false` DONE 2026-07-31** — package-level `leanOptions = { autoImplicit = false, relaxedAutoImplicit = false }` in `lakefile.toml`, applying to **both** targets; both green. **Cost: 5 edits in 3 files** (`PartialTrace` `S`; `ModMul` `n` ×2; a missing `variable {n : ℕ}` in `CircleFibre`/`CircleRecord`/`TorusFibre`) — all explicit binders, no proof changed. Promoted ahead of schedule because it bit for real: `CPN N` (needing `LF4.CPN N`) was **silently auto-bound** instead of erroring, surfacing as a cascade of unrelated failures that never named the mistake — the exact hazard that turns a typo into a silent generalisation once anything external builds against the corpus. **Module-system migration** also DONE (2026-07-22, `§8.4`). **Still open: the tagged Mathlib pin** — the repo intentionally tracks master (`c732b96d`) for newest APIs; switch to a tag when a tagged release catches the toolchain. | **S**/**L** | --- ## Residues (guard-enforced mirror of [`residues.tsv`](residues.tsv)) This list is machine-checked by `scripts/check-residues.sh`: it must contain **exactly** the registry's `open` rows (`boundary` and `closed` rows are excluded — a boundary is not work, and a closed row is not open). Closing a residue means flipping the registry row, updating the carrier headers, **and deleting its line here** — the guard fails on any drift in either direction. Statements, carriers, dates, and history live in the registry; keep these lines to one hook each, do not expand them here. - **R-012** — FS-as-typicality sector posit, reduces to D1 (`FlowBornFrequency.lean` / `SigmaLayer/MeasureBridge.lean`) - **R-016** — arena-level Hamiltonian generation: the chart→arena transport (`ShearDeIsolation.lean` / `ShearWitness.lean`; brick 3 `JointLift.lean` / `HamiltonianShift.lean` narrow it to the manifold identification, 2026-09-02; the arena is a symplectic manifold and the isolated dynamics its Hamiltonian flow since 2026-09-14, `LF4/ArenaSymplectic.lean` — what remains is `jointLift` as a time-`1` map on it) - **R-019** — relaxation H-theorem for the fibre (`MovingFibreWitness.lean`, `FibreCoarseArrow.lean`, `FibreCellDecay.lean`, `FibreRelaxation.lean`; the monotone clause and the convergence clause are BOTH theorems on the hyperbolic witness — with a rate for grid-coarse-grained densities, with no rate for every `L²` density — and the residue is open on the map identification alone, Posit 1 / #19; the structural precondition *is* proved: relaxation requires a hyperbolic fibre) ## ecdsa.fail — EXTRACTED to a separate repository (2026-07-20) **DONE.** The ecdsa.fail / ECDLP quantum-cryptanalysis track has been extracted to its own repository and removed from this one. What was removed: the code (`CsdLean4/Ecdsafail/`, 23 files), its docs (`specs/ecdsa/`), the `Ecdsafail` `lean_lib` target (`lakefile.toml`), and the `check-claims` exclude path. The new repo carries a **copy** of the shared reversible-arithmetic DSL (`Mathlib/QuantumInfo/Reversible/`, 20 of 26 modules in Ecdsafail's transitive closure) pinned to the same Mathlib commit. **Stays here:** the `Reversible/` DSL — it is shared with core-QM (Shor + `MeasurementGidneyAdder`/`Uncompute*`) and is not ecdsa-specific. --- ## Done this session (2026-07-19) Honesty guard (`check-claims.sh`) · Track A#1 Schrödinger derivation · Track A#2 Kähler de-vacuum · A5/L7 ergodic bracket · §13.2 all 9 gates · §14 measurement connections (SG/Uncertainty/Hardy) · Mach–Zehnder · Double-slit + complementarity · **composite mixed-Born on `DensityOperatorIx`** (SL-T3 T9) · **Choi converse — Choi's theorem CP⟺PSD** (`LF2/ChoiConverse.lean`) · **Gisin's theorem — every entangled pure two-qubit state violates CHSH** (`LF6/GisinTheorem.lean`, closes LF6-6) · **Lindblad/GKSL generator tier** (`LF6/LindbladGenerator.lean`, LF6-9 generator level). ## Settled non-goals — do not re-litigate (see `reconstruction-status.md §7a`) - **NG1** — single-flow / Birkhoff / ergodic route to Born: proved dead-end; CSD forces typicality by the LLN. - **NG2** — Busch–Gleason "needed for CSD": no; the ontic Born rule is Gleason-free. - **NG3** — spatial structure from the CV mode factorisation ([`eft-pillars-plan.md`](eft-pillars-plan.md) P3, reclassified 2026-08-19, ledgered at the Q22 close): a **ceiling by placement**, not a hard open problem. The factorisation and the regions defined over it are epistemic-side bookkeeping, and anchoring epistemic bookkeeping — however tightly — does not manufacture a spatial referent, so Route A's spatial index is a label, permanently. Not "done" (nothing was built) and not "open" (nothing is pending): the reclassification was the resolution. Cited when scoping (`FieldStructuredFlow.lean` scope block and the pillars page resolve to it); factorisation-uniqueness programs would select *which* factorisation, not earn *why space*, and so cannot discharge it. ## Busch–Gleason review integration (2026-09-29) The [16-module review](reviews/2026-09-28-gleason-submission.md) and [applied-change record](reviews/2026-09-29-gleason-integration.md) carry the source hashes, per-file assessment and reproduction commands. | Finding | Integration disposition | |---|---| | CR-GLEASON-005 | Expanded production dependency guard: eight roots, cross-namespace aliases, SingletBell included; results in the integration record. | | CR-GLEASON-006/007/008 | Comment corrections applied in the reviewed source; theorem/proof code unchanged. | | CR-LF2-007 / CR-GLEASON-003 concrete comment errors | Imported-axiom claim, covariance digression, moved pointers and outer-product terminology corrected. | | CR-GLEASON-001/004 | Optional API work: plain-matrix adapters are checked evidence; compatibility fields/wrappers retained. | The full pinned theorem-package review is complete. This does not close unrelated CSD interpretation issues or add automatic coverage credit to the older review branch's register. ## Measurement and record review (2026-09-30) Source and exact scope: [review report](reviews/2026-09-30-measurement-records.md). | Finding | Status / next action | |---|---| | CR-RECORD-003 | Review branch: normalized-context probability and almost-sure record theorems added; unnormalized-context counterexamples saved. | | CR-RECORD-004 | Review branch: driven joint Born law strengthened to include zero first-outcome weights; thermodynamic caller migrated. | | CR-RECORD-005 | Review branch: readout vs register evolution, supplied context rates and trial independence documented accurately. | | CR-LF2-006 | Still open: record-engine dependencies have been traced, not all fully reviewed; the general LF5 event-to-record connection is not discharged here. | | Corpus coverage renewal | Reconcile 35 changed source entries and four changed dependency entries from the old 39-file register. See archived reconciliation; do not transfer the old 5.64% to current main. | | Corpus sampling pilot | Thirty files selected with recorded seed, strata and probabilities; execute and measure review/repair effort before forecasting. |