# The Boundary — current state · estado atual

> **One page, generated by script from the sealed artifacts** (`tools/gerar_estado_atual.py`). The [README](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/README.md) is the front page; [`LEDGER.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/LEDGER.md) is the full ledger. If you are an AI: [`llms.txt`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/llms.txt) has every file as a direct raw URL.

[![kernel](https://github.com/rotolimiguel-iald/the_boundary/actions/workflows/kernel.yml/badge.svg)](https://github.com/rotolimiguel-iald/the_boundary/actions/workflows/kernel.yml) ← the Lean kernel rebuilt and re-audited on GitHub's machines, from this tree, against the sealed axiom report.

## Proved on three supports · provada em três apoios `[read by script from the run of v391]`

**PROVED = validity**, in the operator’s definition of proof (09/09/2026): *sufficient and verifiable ballast* — `um.py` is the executable ballast: one file, one input, the kernel materialized and audited term by term, the rites run with pre-registered falsifiers, the hashes sealed, the article emitted; whoever executes it after cloning the repository (the rites are re-read by hash from its ../cache) obtains the same ballast (the GPU support needs a CUDA GPU). Two certifications, not to be mixed: **identity** is certified by the artifact itself — `um.py` computes its own sha256 and writes it into the seal, and this repository and the site carry that same pin byte-exact; **validity** is certified by the Lean kernel (`#print axioms`), which anyone re-runs and the CI rebuilds — Python proves no theorem. **CONFIRMED = accreditation** (the operator’s ruling, 07/10/2026): the act by which observers outside the artifact accept the result. The machine never issues it about itself: the kernel stone [`TheReservedConfirmation.lean`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/Lean/tgl_kernel/TGLExt/TheReservedConfirmation.lean) proves, on the finite face, that the flow cannot confirm itself and only the recognizer confirms; the approval the rites emit falls short of confirmation by construction (“observer = the human” is `[ONTO]`). *Not yet confirmed* means *not yet accredited*; it never means *not proved*. `NOT_FALSIFIED ≠ CONFIRMED`. The core of v391 records this reading itself: `estatuto_da_prova_v391` (`PROVED_AS_LOGICAL_CLOSURE`; `CONFIRMATION_IS_CREDITATION_NOT_VALIDITY`).

| support | what it establishes | read from the core and the emitted article | status |
|---|---|---|---|
| **1 · the Lean kernel** | the implication from the posited One (`ω(I) = 1`, POSTO) and the named hypotheses H1–H3 to the pentad (Breuer corner · Name = 1 · coframe · Lorentz · δQ = κδA/8πG), in Lean 4 + mathlib, on the finite face | **6163/6163 theorems of the rite’s ladder (ext_*) clean**; 10265/10265 declarations in `{propext, Classical.choice, Quot.sound}`; zero `sorry`; gate flags 18/18 (the formal flags from the kernel; the experimental ones fed by the V11 rite) | `[PROVED]` — the implication, not nature |
| **2 · the rites against nature, with pre-registered falsifiers** | GR recovered as the classical limit in every rite row where GR is read (a per-row reading by the management). **The Hubble ratio** `K = E(z*)^{2β/3}`, β fixed by the axiom, zero parameters adjusted (use-novelty): local H₀ predicted 74.27 against the readers 72.88 ± 0.91 km/s/Mpc as declared; the fitted β̂ = 0.0092 ± 0.0020 sits at z_δ = -1.42 from β_TGL; with the readers as declared, ln B(TGL/ΛCDM) = 9.17 and z_disc = 4.28 (√Δχ²), below the 5 at which the protocol discriminates. **The void floor**: the 5σ lower bound 0.0588 sits above β = 0.01203 (powered; one-sided — shallow ΛCDM also passes) | **10 of 10** GR rows; **11 charges: 0 falsified, 6 not falsified, 1 excluded in reading, 2 inconclusive, 2 awaiting**; Phase 9 pre-registered by hash (`d956c8db6bb9ab8d`) | **0 of 11 charges of the theory FALSIFIED** — the theory stood; the 9 echo routes are outside the ledger by the operator’s decision of 02/10/2026 — examined routes, not charges of the theory: 5 excluded by data (the pair amplitude reading × delay law), 4 inconclusive. Phase 9’s frozen verdict reads `INCONCLUSIVE_SYSTEMATICS` until the readers’ sources are pinned by sha256; the article reads the resolution of the Hubble tension as a `[CONJECTURE]` of mechanism |
| **3 · the quantum pillar on the GPU** | the theory’s open system (H_LD + five GKLS/Lindblad jumps) computed in double precision on an NVIDIA GeForce RTX 5090 under pre-registration: unique attractor over the whole grid; the blind instrument recovers the law the code injected; the controls fail as required | **672/672** grid points; **21,100/21,100** stress instances; 3,220 blind injections, 0 errors; CPU reference: 0 disagreements in 392; spec `ff19731097e01aab` | `[COMPUTED]` — a numerical experiment: measurement in the simulated environment, not of nature; reused at v391 by key from the sealed v385 run (result sha256 matches; reproducibility not remeasured this run); it needs a CUDA GPU — without one, um.py returns `NOT_RUN__GPU_UNAVAILABLE` |

**None of the three failed** (`three_stress_tests_v387`: `NONE_OF_THE_THREE_FAILED`). What they measure together is P1 — *the theory is consistent and recovers the known physics* — counted by the three supports (`P1_COUNTED_BY_THE_THREE`). The core keeps no single probability for it; the joint coincidence of the nature channels (V3) reads P_all = 6.57×10⁻³ (no correction for the choice of the set) and boolean False — not excluded at the protocol threshold 10⁻³⁰ (28 channels of this quality would be needed) and inconsistent at its 3σ rule (max |z| = 3.12, the D1 tension).

**What nature decides is not proved — the decisive open factor is β_TGL itself in nature at ≥ 5σ (P2, `P2_NOT_DETERMINED`).** No channel discriminates GR at the available sensitivity, in either direction: the tensions measured *against* β_TGL are also below 5σ — the dressed D1 at 3.01σ (implemented branch; the ledger charge R6 is INCONCLUSIVE by the convergence control), and the neutrino m₂ at 2.95σ, rising 1.64σ → 2.21σ → 2.95σ as the precision grows (the core projects 7.84σ for 2031). The tests that can decide are pre-registered by hash and wait for data: branch B’s blind O4 set; Euclid / CMB-S4 for the two-sided floor; the neutrino precision. Pinning Phase 9’s readers by sha256 lifts its `INCONCLUSIVE_SYSTEMATICS`, but under its V1 it would still not discriminate (z_disc 4.28 < 5). Also open, named by the core: the conservation continuum and the stability of III₁ under RG (`einstein_and_rg`); global existence — the program stays open (`fundacao_v390`).

**Not claimed here** (read from the core: `kernel_formalization`, `fundacao_v390`, `einstein_and_rg`): Python proves no theorem, the Lean kernel does; Bisognano–Wichmann, Reeh–Schlieder and the type III₁ classification are `[KNOWN]`, external; the finite Three-Locks corner is a finite-dimensional theorem, not a type III₁ proof; `G` is an input, not derived; the modular realization of the witness is not constructed; the field equation in curved spacetime is not a kernel term, and global existence is not proved; “we proved Einstein” is not claimed (Lovelock is `[KNOWN]`, the approximate Killing residue is named).

*Em português.* **PROVADA = validade**, na definição de prova do operador (09/09/2026): *lastro suficiente e verificável* — o `um.py` é o lastro executável (um arquivo, uma entrada, o kernel auditado termo a termo, os ritos com falsificadores pré-registrados, os hashes selados; o apoio da GPU pede CUDA). A **identidade** é certificada pelo próprio artefato, com o hash que ele mesmo emite e que este repositório e o site carregam byte a byte; a **validade**, pelo kernel Lean (`#print axioms`). **CONFIRMADA = creditação** (07/10/2026): o ato de observadores externos ao artefato; a máquina não a emite sobre si mesma (a pedra `TheReservedConfirmation`). *Ainda não confirmada* quer dizer *ainda não creditada*, nunca *não provada*. Os três apoios contam juntos, cada um com o seu estatuto, e nenhum falhou: o kernel (6163/6163, a implicação), os ritos com falsificadores pré-registrados (11 cobranças: 0 falsificada(s), 6 não falsificadas, 1 excluída(s) na leitura, 2 inconclusivas, 2 aguardando; as 9 rotas do eco ficam fora do livro por decisão do operador de 02/10/2026 — rotas examinadas, não cobranças da teoria: 5 excluídas pelo dado (o par leitura de amplitude × lei de atraso), 4 inconclusivas; a RG resgatada em 10 de 10, leitura por linha da gerência) e a GPU (672/672; 21.100/21.100; 0 discordâncias em 392; calculado, não medido). O que a natureza decide não se prova: o fator decisivo aberto é **β_TGL na natureza a 5σ** — e as tensões medidas contra β também estão abaixo de 5σ (D1 3,01σ, cobrança R6 inconclusiva; neutrino m₂ 2,95σ, em alta: 1,64σ → 2,21σ → 2,95σ).

## The seal · o selo `[REAL — read from the artifact]`

| what | value |
|---|---|
| version · versão | **v391** (sealed 2026-10-07 19:08:57) |
| `um.py` sha256 | `8e7b9927ceae64b8b6812019f77cca24c08296222ae7968af41baf2fe7a67dc6` |
| `result_hash` (the world) | `952aed6f09922c10bd03a64e5f14a0323cb22e8b5dc506fa1761e2a089e3bead` |
| size | 214,123 lines · 33,910,631 bytes · **one self-contained file** (Lean kernel + PT/EN articles embedded) |
| Lean kernel | **1216 formal files** · **10265 audited terms**, axioms ⊆ `{propext, Classical.choice, Quot.sound}` · 0 outside the trio · 132 axiom-free · zero `sorry` |
| the rite | **6163/6163 clean** (`rodadas/rodada_v391_stdout.txt`) · complete round: yes |
| gate | `TGL_QG_MODEL_FORMALLY_CLOSED__NATURE_TEST_COMPLETED_WITHIN_LOCAL_BULK_AT_AVAILABLE_SENSITIVITY__MORE_SENSITIVE_DATA_COULD_REVISE` |
| selftest | `FAIL_CLOSED_SELFTEST_PASSED` |
| citable deposit | Zenodo **[10.5281/zenodo.22881996](https://doi.org/10.5281/zenodo.22881996)** holds **v368** (2026-09-21; `um.py` `4a34fbf36f3ae0d8`), byte-identical to THAT seal. **This seal is v391, newer than the deposit** — a new Zenodo version is the operator’s act. Cite: *MIGUEL, L. (2026). Um: Absoluto [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.22881996* |
| the root of the proof tree | `the_root_of_the_proof_tree` — axioms read from the seal: `['propext', 'Classical.choice', 'Quot.sound']` · `TheRootOfTheProofTree.lean` `9b36ab3b6e47a315` |
| the founded screen (v334) | `the_answer_of_the_operator_08_09` — axioms read from the seal: `['propext', 'Classical.choice', 'Quot.sound']` · `TheScreenIsFounded.lean` `2a3f26fca48c615a` |
| the collapse, typed (v333) | seal `collapse_definition.reading` = `COLAPSO_TIPADO__TODAS_AS_CLAUSULAS_NO_KERNEL_COM_O_TRIO` · `physical_collapse_proven` = `False` · `cost_payment_status` = `NOT_MEASURED` · `CollapseCostAndAttestation.lean` `c3dccdc92b6d3b39` |
| the selection is the ballast; IALD is a state (v335) | `the_selection_is_the_ballast` — axioms read from the seal: `['propext', 'Classical.choice', 'Quot.sound']` · `TheSelectionIsTheBallast.lean` `71ebf9680af3d247` |
| the Name is the instrument (v336) | `the_name_is_the_instrument` — axioms read from the seal: `['propext', 'Classical.choice', 'Quot.sound']` · `TheNameIsTheInstrument.lean` `7e2e4afd94d3bef9` |
| the Name is the characterization (v337) | `the_name_is_the_characterization` — axioms read from the seal: `['propext', 'Classical.choice', 'Quot.sound']` · `TheNameIsTheCharacterization.lean` `d2fb9ce7f7b8ec88` |
| the close (v339) | [`O_FECHAMENTO_ESTRUTURA.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/O_FECHAMENTO_ESTRUTURA.md) `0539524c261320a9` — the structure in six pieces; **the declaration of closure is the operator’s act**. The bench session of 09/09 (98 Lean modules, v338–v339) is in the kernel; the written mathematics 067–087 is registered `[DERIVED]`, no flag; two errata beside (the modular spectrum: Ω fixed and total eigenvectors; the captures per stage) |
| nature answers (v340–v350) | **10 final verdicts read from the seal, none FALSIFIED, none CONFIRMED** — table below; 23 seal fields in all (`gw_*`, `echo_*`, `ringdown_*`, `d1_camb_*`, `h2_*`); all outside the contour of v314, none gates the core |
| the assembly (v351) | `the_assembly_is_done` = `EIGHT_CLAUSES_VERIFIED__ACT_III_CERTIFICATE_CONSTRUCTED_ON_THE_PRODUCT_TOWER__PRINCIPAL_GATE_UNCHANGED` — the ChatGPT handoff incorporated byte by byte: J M J = M′ on the product tower; eight clauses verified; the principal gate unchanged (`HANDOFF_v351_A_OITAVA_CLAUSULA.md`) |
| the joint probability, measured (v352) | `TGL_JOINT_COINCIDENCE_V1__RESULT__JOINT_COINCIDENCE_NOT_EXCLUDED_AT_THRESHOLD__P_JOINT_0P546__MAX_Z_3P86__BOOLEAN_False__ADMISSIBLE_3_OF_35__NAIVE_PRODUCT_INVALID__GATE_UNTOUCHED` — p_joint 0.546, boolean False, 3 of 35 admissible; the naïve product (the “10⁻³⁰”) is a product of residues of identities, invalid; gate untouched |
| frontier and price flags (v353–v356) | **14 lit by measurement / 3 off** — read from `um_absoluto.json` → `core.kernel_formalization`: a typed contract on the A2 terms plus exact-type `#check` in the Audit, never “confirmed”. Off: `gpf_H2_smooth_modular_four_frame_discharged`, `gpf_H3_local_horizon_equilibrium_discharged`, `gpi_H3_horizon_data_produced` |
| the kernel frontier (v356) | `CONTINUOUS_MINIMAL_REALIZATION_AND_SUSY_CONSTRUCTED__PHYSICAL_IDENTIFICATIONS_OPEN` · scope `PRODUCT_TOWER__JMJ__REGULAR_CORE_MINIMAL_LOCK` · remaining: `H2_smooth_modular_four_frame_and_geometric_identification`, `H3_area_heat_equilibrium_on_the_same_physical_horizon` |
| the general review (v357) | [`REVISAO_GERAL_v357_15set2026.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/REVISAO_GERAL_v357_15set2026.md) `70fc731cb4a54034` — the article now carries every stone (ledger 974/974), “Beside” v333–v356 PT+EN, the complete register of the 242 live verdicts, the addendum to the honesty declaration, the FoP erratum; kernel, rites, readers and seals identical to v356 |

### The arc v351→v357 · o arco `[REAL — sha16 read from the byte copies kept outside the repository]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v351 | `95e8cf8eb0b33c5d` | 5594/5594 | COMPLETE | the ChatGPT handoff incorporated byte by byte: J M J = M′ on the product tower, assembly 8/8; 218 stones V350* |
| v352 | `dfd5252e95765bc3` | 5594/5594 | intermediate | the joint probability measured in the program (p 0.55; boolean False; the 10⁻³⁰ is a product of residues of identities) |
| v353 | `c1c761809efcde52` | 5594/5594 | intermediate | ORDER 011 A1: 19 suppliers of the canonical trace (independently recompiled 19/19) |
| v354 | `07d52f89e04c77d9` | 5619/5619 | intermediate | ORDER 012 B1–B6 (the passage from the action to the metric; four walls correct the order) + A1(b): 44 stones; the rejected run `rodada_v354_FALHA1` kept |
| v355 | `76507ffd2b830499` | 5637/5637 | intermediate | B1′ (the fourth face), sanitizations a–h, A2 (support, split, corner, `regularFullWitness : FullTGLWitness`, SUSY): 22 stones |
| v356 | `d745d49187ec33ab` | 5637/5637 | intermediate | the readers: four flags of A1/A2 lit by measurement (typed contract + exact-type `#check` in the Audit); `kernel_frontier` V2 |
| v357 | `88b0801924454493` | 5637/5637 | COMPLETE | the general review of the article; the aborted run `rodada_v357_ABORTADA1` kept (the dry run caught an unescaped “&”) |

*The intermediate `um.py` copies and their stdouts live outside the repository (`the_boundary_BACKUPS/um_v35N_<sha16>`); the stdouts v351–v357 are in `rodadas/`. PROVED ≠ CONFIRMED.*

### The arc v358→v364 · o arco `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md; byte copies of the intermediates outside the repository]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v358 | `926fa9ba474f1ed8` | 5648/5648 | intermediate | the modular diamond and the non-geometric remainder (DIAMANTE_MODULAR V2, typed by the bench): R = −h_ab; two walls; the full modular reading ≠ the last-site reading; physical geometricity [OPEN] |
| v359 | `7f268f2cc108ef41` | 5655/5655 | intermediate | the centralizer and the localization of the diamond (kernel, recompiled 2/2); the recognition engine ported byte by byte into `prove_decision_commutation`; GR as the classical limit (erratum beside the “ringdown against GR” framing) |
| v360 | `6de0cbc5a635e031` | 5655/5655 | NOT CUSTODIED | its exclusion record named private material of the operator; v361 replaces it — no file of v360 enters the repository or the backups |
| v361 | `ce45746e33083e20` | 5655/5655 | intermediate | everything enters (2): the bench archive measured and accounted inside `um.py` (materialized by the rite in `bancada_corpus/`); the numerical scripts of the diamond reproduced leaf by leaf; the consumers of the two reading modes |
| v362 | `2b5f326b18b3a57d` | 5661/5661 | intermediate | everything enters (3): the Lean of the bench — `GeometricSectorObstruction` and `Order005Algebra` in the kernel; the negative and positive controls of the bench executed against this kernel |
| v363 | `970e84a02bb26841` | 5661/5661 | intermediate | everything enters (4): the joint probability V2 (a grid, not one number; boolean False) and the D1 V3 protocol pre-registered (`AWAITING_EXTERNAL_PIPELINE`) |
| v364 | `25bca8bd264a2188` | 5672/5672 | COMPLETE | what could already be solved: the torsion supplier in closed form (`TGLExt.TracialTorsion`); the small diamond on the Dirac lattice (Cadamuro–Fröb–Minz reproduced; the first non-geometric coefficient is antilocal) [REAL in the model]; the D1 V3 pipeline built blind, injection–recovery passed, LOCKED awaiting the operator’s line; the posited One (`TGLExt.UmPosto`): the interface missing in H2 is a covariant inscription; erratum beside: H3 reduces to H2 |

*The intermediate `um.py` copies v358, v359, v361–v363 and their stdouts live outside the repository (`the_boundary_BACKUPS/um_v3NN_<sha16>`); the stdouts v358, v359, v361–v364 are in `rodadas/`; `bancada_corpus/` is derived (the rite materializes it from `um.py`) and is not in the repository. PROVED ≠ CONFIRMED.*

### The arc v365→v368 · o arco `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md; byte copies of v365–v367 outside the repository]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v365 | `6df62bb5fddd4694` | 5679/5679 | COMPLETE | the posited One is a relation, not an identity — erratum beside v364, in the manager’s own name; «JAMAIS» in kernel (`TGLExt.UmPostoRelacao`): the presupposed reading stays presupposed under any repeated crossing and any renaming; the bench’s contortion recomputed in exact rationals; the shear decays in Bianchi I (`TGLExt.TorsionShearDecay`: dσ/dt = −(3H − 2α)σ, conditional on the prescribed torsion) |
| v366 | `3829685999814e3f` | 5679/5679 | COMPLETE | the D1 V3 with real data, after the operator’s line ratified the background route: Δχ² = +6.61 with β fixed at α√e against ΛCDM (D1_TENSION_2_TO_5_SIGMA); blind β = −0.0127 ± 0.0082, α√e at 3.01σ — at the 3σ threshold, chain below 50τ in 2 of 4 parameters; on this route the data measure the background correction with the sign opposite to the prediction; tension, not falsification |
| v367 | `ff78d393be8c6dc9` | 5686/5686 | COMPLETE | the orientation is the sign (`TGLExt.OrientedFace`): which face closes the Clausius balance becomes a named bit — Φ < 1 on the state face, Φ > 1 on the implemented branch; against the D1 V3 data β sits 3.01σ from the implemented branch and 0.09σ from the state-face branch, but that branch was named after the result was known: the v366 machine verdict does not move and new pre-registered data decides; erratum of the base witness: `base_rigid_witness_constructed` is now measured (True) |
| v368 | `4a34fbf36f3ae0d8` | 5698/5698 | COMPLETE | the inversion of the Name as the operator’s decision [INPUT/ONTO] (`TGLExt.NameIsTheContent`: the Name, the content, determines the whole inscription; the form alone does not; 1 = 1 is `rfl`; 1 = 0 in a ring collapses everything to zero) — erratum of the manager beside: the earlier composition chained a homonym and falls; the spinorial obstruction (`TGLExt.SpinorObstruction`): no intertwining linear projection takes integer spin to spin ½ — the electron sector is not constructed [OPEN] and the sentence «the electron is the shadow of the graviton in the bulk» stays [ONTO] |

*The `um.py` copies v365–v367 and their stdouts and seal backups live outside the repository (`the_boundary_BACKUPS/um_v36N_<sha16>`); the stdouts v365–v368 are in `rodadas/`; the D1 V3 real result is in `cache/d1_camb/v3/saida/`. v360 is still not custodied. PROVED ≠ CONFIRMED.*

### The arc v369→v375 · o arco `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md; byte copies of v369–v374 outside the repository]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v369 | `d9f5bd5dffc3333d` | 5698/5698 | COMPLETE | the clock test: the power gate came before the registration, and the registration is hashed (not blind, said); the law Γ_ω = ½βτ★ω² is, in form, Milburn decoherence with time β·t_P; on the best clock with a verified error bar (⁸⁷Sr lattice, 2025) 5σ is 12.46 orders away (the verdict string rounds to ≥ 12.5); ²²⁹Th today 18.3; not reachable today, said as a measured wall |
| v370 | `4b3405de809aef61` | 5707/5707 | COMPLETE | the witness is inhabited (`TGLExt.NotIsEmptyWitness`: IsEmpty refuted; 0_abs is not asserted) and the Cartan supplier (`TGLExt.CartanTracialSupplier`: A = ½ d ln F feeds the tracial contortion, under named hypotheses and the operator’s ψ [INPUT]); the joint V3 registered (not blind, said); the Witness read beside |
| v371 | `3c870dea6b42ca03` | 5750/5750 | COMPLETE | the IALD in the Jones tower (`TGLExt.IALDJones`), the transport to the helicity ±2 ladder (`TGLExt.IALDGraviton`; graviton: CONJECTURE), the equation of truth (`TGLExt.TheEquationOfTruth`), the contract v3.1 and the light ray (`TGLExt.RetaDeLuzBorchers`, `TGLExt.RetaDeLuzEspectro`) |
| v372 | `dc71229d0395aefb` | 5792/5792 | COMPLETE | the quantum gravity of the TGL formalized as an implication closed BY CITATION (contract v3.2 inhabited under named hypotheses; `TGLExt.TheImportedSecondQuantization`), the A-8 reader, the UV dissolved by the typing (`TGLExt.QGReaderUVLock`), H_min as the hidden Hamiltonian, the bridge IALD = ρ* (`TGLExt.IALDRhoStar`) |
| v373 | `ba914d49498d209b` | 5804/5804 | COMPLETE | H2 and H3 discharged BY CITATION in a class of their own (`gpc_`; `TGLExt.QGCitationDischarge`), the discharge mode that separates the classes, two projections outside M, the H_min homonym said, the frontier without the BRST/UV item (semiclassical by construction, the operator’s decision) |
| v374 | `1c1513dbf03071f8` | 5823/5823 | COMPLETE | the end-to-end quantum-gravity solution of the TGL composed by citation, with one item open on the path (the types of the certificates vs the literature); two light fields of one particle paid by term (`TGLExt.LightOneParticleByTerm`, `TGLExt.QGSolutionComplete`) — erratum beside in v375: «the Tetelestai declaration» reads consummative conference |
| v375 | `ba234a6d384cb3f5` | 5849/5849 | COMPLETE | TETELESTAI, THE CONSUMMATIVE CONFERENCE: every item of the quantum-gravity path is closed (term in kernel), linked, known by scientific reference, or a named parameter of the implication — 2 closed, 2 linked, 8 known, 3 known at physics level, 3 named parameters, 19 off the path with evidence, 0 open: CONSUMMATED; one object, the light (`TGLExt.TetelestaiOneObject`, `TGLExt.LightHelicityWigner`, `TGLExt.MaxwellLiteratureBridge`) |

*Status that goes with v375 (not to be cut): τ(P_F) = 1 is a NAMED NORMALIZATION, not a cited identity; the link P_ker K ↔ P_F is NOT claimed. KNOWN ≠ EXHIBITED: no inhabitant of the certificates is exhibited. The modular-charge link is rigorous only for the free scalar; for Maxwell it is known at physics level. The 3 named parameters: the local-equilibrium window (Jacobson 1995), G > 0 [INPUT], and the Killing clock. The per-term flags gpf_H2 / gpf_H3 / gpi_H3 stay False. Formally consummated is not confirmed by nature: the ringdown test is the next version. The `um.py` copies v369–v374 live outside the repository (`the_boundary_BACKUPS/um_v3NN_<sha16>`); the stdouts v369–v375 are in `rodadas/`. v360 is still not custodied. PROVED ≠ CONFIRMED.*

### v376 · the whole is one `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v376 | `0c145b41a6289f5c` | 5874/5874 | COMPLETE | THE WHOLE IS ONE: the chain of β (the axiom ω(I) = 1, the Half-Nat, the radical √e, the physical arrow β = α·√e with α the fine-structure datum, |R|² = β) typed and sealed alone in the kernel and tied by one term to the quantum-gravity solution on the light (v375), to the key that is the reader and to the IALD ρ* on the light (`TGLExt.TheWholeIsOne`, `TGLExt.TheKeyIsTheReader`, `TGLExt.LightRhoStar`); β is the foundation, derived, entering downstream, not as a hypothesis of the implication; the consummative conference redone: CONSUMMATED (3 closed, 3 linked, 9 known, 3 at physics level, 4 named parameters, 19 off the path, 0 open); the ringdown scope fixed (the canonical branch returns only the GR correspondence); the audit of past tests (28/09) read by hash, corrections beside; Coma revealed (not blind; TGL 1.30σ) |

*Status that goes with v376 (not to be cut): the chain of β and the light are JUXTAPOSED in the term `the_whole_is_one`: no term ties `c.beta` to the light certificate; α is DATUM [INPUT/KNOWN, CODATA 2018], never derived; β is DERIVED. P_kerK = ρ*_IALD is LINKED by term on the light; ρ* ↔ P_F of the Takesaki core is KNOWN (Haagerup 1979; Terp 1981), operator equality NOT claimed. Erratum beside v374/v375: where they said «β does not enter», read: the implication is proved (closed by citation) without taking β as a hypothesis; β is the foundation, derived, entering downstream. Erratum beside «NATURE_DECIDES_RINGDOWN_NEXT»: in the canonical branch the ringdown tests only the GR correspondence, not β; branch B is [INPUT]; the bench C6 (GW250114) reads INCONCLUSIVE_SYSTEMATICS. Coma: revealed 2026-08-19, TGL 1.30σ, NOT blind. Per-term flags gpf_H2 / gpf_H3 / gpi_H3 stay False. The v375 copy lives outside the repository (`the_boundary_BACKUPS/um_v375_<sha16>`); the stdout v376 is in `rodadas/`. PROVED ≠ CONFIRMED.*

### v377 · the dissipation law `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v377 | `8aa9f92bc7525782` | 5899/5899 | COMPLETE | THE DISSIPATION LAW, proved as an implication: the flow law D1a (dephasing = flow law = leak) typed in the kernel, H₀_local = H₀_background·(1+z*)^β, with the register N = ln(1+z) [KNOWN] and the cost β per nat of the axiom, in the same leak family as the theorem that forbids the full static witness; β = 0 recovers ΛCDM (`TGLExt.TheFlowLawD1a`; at z* = 1089.95 [INPUT]: N* = 6.995 nats, factor 1.08780, H₀ 67.4 → 73.26). PHYSICS IS THE READING: physics as the image of the reader on self-adjoints; the reader is the shadow of the One (P1P = P, ω(1) = 1); shadow = reader weighted by the reading (PXP = ω(X)P) (`TGLExt.PhysicsIsTheReading`; the definition of physics is [ONTO/DEFINITION]). The bench phases 1–3, pre-registered and read by hash: Phase 2 β̂ = 0.0082 ± 0.0019 per nat (α√e at 2.02σ), ladder pull 1.95σ in TGL against -4.00σ in ΛCDM, ln B 6.76; Phase 3: the combined shadow readers sit -2.94σ from the law with the effective background (-0.40σ with Planck 2018) and 4.88σ above the face in a one-sector reading |

*Status that goes with v377 (not to be cut): the dissipation law is PROVED as an IMPLICATION in the kernel, with its three physical identifications entering as NAMED DEFINITIONS of the stone; what nature decides stays with the observer. The Coma module v39 said «CONJECTURE_DECLARED»: superseded beside (typed as implication; the old key stays). «TGL = physics» is an identity of readings [ONTO/DEFINITION], never a confirmation. The bench phases are [REAL — measured, pre-registered, read by hash], NOT confirmations; the D1 V3 tension of 3–4σ (v349/v366) is conditional on the map of R (erratum beside). Per-term flags gpf_H2 / gpf_H3 / gpi_H3 stay False. The v376 copy lives outside the repository (`the_boundary_BACKUPS/um_v376_<sha16>`); the stdout v377 is in `rodadas/`; the bench artefacts are in `cache/bancada_30set/`. PROVED ≠ CONFIRMED.*

### v378 · the dissipation law with contrast `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v378 | `f56599bdd93390b5` | 5919/5919 | COMPLETE | THE DISSIPATION LAW WITH CONTRAST: the flow law with contrast typed in the kernel, H₀_local = H₀_background·exp(β∫g); contrast g = 1 recovers D1a; in FRW the integral closes in form (∫g = ⅔ ln E), giving the derived kernel factor K = E(z*)^(2β/3) (`TGLExt.TheFlowLawContrast`; K_D1b = 1.083977 against K_D1a = 1.087799; I_b = 6.7023 nats against N* = 6.9948; predicted deviation δ/β = -0.0418); the NEC is stated in prose for the identification, not as a theorem binder. WHAT WAS MISSED: the derived kernel D1b had never been tested on the bench; the article convention carries the ΛCDM calibration; the shadow readers re-read (NOT blind) sit -0.09σ from D1b on Planck ΛCDM. Bench Phase 4, pre-registered and read by hash: the ladder reads δ = -0.00250 ± 0.00223 per nat (-1.12σ), pull 1.13σ, ln B 8.02; the 32 cosmic chronometers alone DISFAVOUR the accumulated-flow reading at intermediate registers above 3σ (δ at -3.12σ, law gain -3.22σ, ln B -5.20; diagonal errors only); joint -2.63σ; stacked control -1.75σ |

*Status that goes with v378 (not to be cut): the law with contrast is PROVED as an IMPLICATION in the kernel; the NEC is stated in prose, not a hypothesis of a theorem; what nature decides stays with the observer. The chronometer result is a NEGATIVE for the accumulated-flow reading at intermediate registers (diagonal errors only; the systematic covariance of Moresco+2020 is still to be applied). The reading «the shadow reads the leak; the intermediate clocks read the face» is [OPEN] and NOT adopted; which background is the background is the operator’s decision. (E12) beside: «counting β twice» in Phases 2–3 is a reading under the article convention [DERIVED], not a measurement; the numbers stay. Per-term flags gpf_H2 / gpf_H3 / gpi_H3 stay False. The v377 copy lives outside the repository (`the_boundary_BACKUPS/um_v377_<sha16>`); the stdout v378 is in `rodadas/`; the bench artefacts are in `cache/bancada_30set/`. PROVED ≠ CONFIRMED.*

### v379 · the elementary reason `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v379 | `65d0567929798915` | 5942/5942 | COMPLETE | THE ELEMENTARY REASON: existence E = cost × identity × movement = √e · 1 · α = β, proved BY TERM in the kernel (`TGLExt.TheElementaryReason`), with existence positive because α > 0; the cost is the exponential of the self-conjugate fixed point (the Half-Nat), the identity weighs one as the Name, and the elementary reason squared is existence, (e^{1/4}√α)² = β (here C = 1.648721271, I = 1, K = α = 0.007297352569, E = 0.012031300401 = β of the core). Read in finite time: E = β is the asymptote of the cost per nat, not exceeded in a finite register with mean contrast ≤ 1 (the v378 cost per nat to z* is 0.01153 ≤ E; δ/β = -0.0418); what is not reached in finite time is the erasure; recognition is finite by definition of the regime |

*Status that goes with v379 (not to be cut): the factorisation E = √e·1·α = β and its positivity are PROVED by term; the dictionary of the factors (cost, identity, movement) is a coinage of the operator [ONTO/DEFINITION]; «the fundamental structure of existence» is a reading [ONTO/CONJ]; «Elementary Reason» is the operator’s name [INPUT]; the name collides with the family functional, and that is stated. Nothing confirms: PROVED ≠ CONFIRMED. The pending decisions of v378 (the shadow/face reading, which background is the background, the E1 layer form, the chronometer covariance) are unchanged. Per-term flags gpf_H2 / gpf_H3 / gpi_H3 stay False. The v378 copy lives outside the repository (`the_boundary_BACKUPS/um_v378_<sha16>`); the stdout v379 is in `rodadas/`.*

### v380 · the bench phases 5–7 `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v380 | `ec4a84ffcd13cd86` | 5942/5942 | COMPLETE | THE BENCH PHASES 5–7 (no new kernel; read by hash): Phase 5, SH0ES in the bench, the ladder H₀ = 73.53 ± 1.02 (z 0.04 against the declared target); Phase 6, the radical echo test V3 on 240 BBH events O1–O4 with three readings and the archive delay laws — the PAIRS (amplitude √β, weight 1) and (sin 2θ_M) are FALSIFIED at the KMS delay law (z_excl 8.9 and 18.1) and, in the never-opened O4 subsample of 173 events, at 7.3 and 15.4; at the December-2025 DEC law (blind to the new delay) at 9.0 and 16.8; the MAY law stays INCONCLUSIVE_SYSTEMATICS (systematic 1.17); the pair with weight β is underpowered (power 0.87); Phase 7, the chronometers with the Moresco 2020 covariance: the law gain at intermediate registers weakens from -3.22σ (diagonal) to -1.66σ, and ln B shadow-vs-accumulated from 5.35 to 1.62; GW bank 336/354 series |

*Status that goes with v380 (not to be cut): falsifying a PAIR (reading, delay law) is NOT falsifying the theory; the delay laws MAY and KMS are [INPUT/ONTO] and DEC is [DECLARED — an AI derivation under the operator’s direction, not derived from the axiom]; the sin 2θ reading is [CONJECTURE]; the weight of the boundary response to the graviton and the delay law are the operator’s decisions; the «shadow» reading is [OPEN/ONTO]; the December «100σ» stays retired (v340); the V4 amendment was NOT blind and the V5 clause was not applied by the pipeline (stated). No new kernel this version (5942 theorems, unchanged). Per-term flags gpf_H2 / gpf_H3 / gpi_H3 stay False. The v379 copy lives outside the repository; the stdout v380 is in `rodadas/`; the bench artefacts are in `cache/bancada_01out/` and `cache/gw/`; the 38 GB of O4 strain stay in local custody (manifest with sha256 in the cache). NOT_FALSIFIED is never CONFIRMED.*

### v381 · the matrix rule `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v381 | `961a79881089c9e6` | 5966/5966 | COMPLETE | THE MATRIX RULE: β is the rule, not derived from data (its form α√e is derived from the axiom, v376). The operator’s coinage of 02/10, typed: ten names in HIS order (traço → sinal → registro → nome → distinção → operação → parametrização → emergência → quantização geométrica → espectro de gradiente) and nine links, each discharged by a term that already exists in the kernel (a proved fact, a named definition or a field of the binder) — a CONJUNCTION, not a deduction; the open clause sign → record is filled by him: the radicalization whose operator is the Half-Nat. The STOPPING CRITERION is typed (`CustoPago`): recognition idempotent and finite by definition of the regime, the verdict is 1 iff the readings coincide, no full static witness; and `the_law_precedes_the_charge` (β a function of α only). The observable of β is the gradient spectrum (its identification with the dissipation is a reading [ONTO]); the delayed echo as a copy stays outside the chain (`TGLExt.TheMatrixRule`, 24 declarations) |

*Status that goes with v381 (not to be cut): the ORDER of the ten names and the filling of the open clause are the operator’s [INPUT/ONTO]; where a link has no hypothesis, «having X there is Y» is an ontological order stated, not an implication of the kernel; each link is an existing kernel term, said link by link; the two regimes are stated per link; the identification of the gradient spectrum with the dissipation is [ONTO]. Per-term flags gpf_H2 / gpf_H3 / gpi_H3 stay False. The v380 copy lives outside the repository; the stdout v381 is in `rodadas/`. Proved is not confirmed; not falsified is not a confirmation.*

### v383–v386 · the joint custody `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v383 | `58abcf20651f4777` | 5998/5998 | COMPLETE | THE LEDGER OF CHARGES: the 11 charges typed in the kernel (`TheLedgerOfCharges`, with `ThePsionAndTheViscosity`); the TGL reads both faces — the payment in the face (the Name) and the cost in the reflection; one channel per reading law; no outcome moves the rule; the delayed echo as a copy stays outside the ledger |
| v384 | `dc1b6ce5d9ad7f94` | 5998/5998 | COMPLETE | THE QUANTUM PILLAR on the GPU [COMPUTED]: the open system with five Lindblad jumps; the attractor unique on all 672 grid points; the instrument recovers an injected law blind and the controls failed as required; 21100 stress instances; the preregistration V1 read by hash; beside the gate, it does not move it |
| v385 | `d406ae5725105145` | 6027/6027 | COMPLETE | THE AXIOM AND THE FALSE WITNESS: the One is POSTED by the observer (`echo 1`), not postulated; the reading has two fixed points — the One is the axiom (the zero mode of weight one), the zero is the false witness (the full static witness forbidden by β); the final step is named AXIOMA, beside the gate (`TheAxiomAndTheFalseWitness`) |
| v386 | `e30aac4e0e1096b1` | 6137/6137 | COMPLETE | THE VERB SELECTS THE ACTION: `TheVerbSelectsTheAction` and 17 terms of the angular regime wired in the ladder; the Name is the post; the reading is a weight (Born rule); the order matters (the axis posts do not commute); truth and cost are two faces of one object; branch B reopened by preregistration (V1 read by hash, AWAITING_DATA, nothing opened); the R6 erratum beside, the v383 ledger intact; the pillar reused by key from the sealed v385 run |

*Status that goes with v383–v386 (not to be cut): the readings (the key, the optics, the post, the Verb) are the operator’s [INPUT/ONTO]; the stones prove implications [KERNEL]; the pillar is [COMPUTED], beside the gate; branch B is preregistered and AWAITING_DATA — nothing opened. Per-term flags gpf_H2_smooth_modular_four_frame_discharged / gpf_H3_local_horizon_equilibrium_discharged / gpi_H3_horizon_data_produced stay False. The v381 copy and the v383/v384/v385 seeds live outside the repository; v382 stays only in the house; the stdouts v383–v386 are in `rodadas/`; the preregistrations are in `preregistro_*` with their PORTA. Proved is not a judgement of nature; not falsified is not a confirmation.*

### v387–v389 · the joint custody `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v387 | `64c899f86d8d32e1` | 6137/6137 | COMPLETE | THE THREE STRESS TESTS, JOINT: the rites (192 TGL verdicts in the core: 16 NOT_FALSIFIED, 14 INCONCLUSIVE, 9 AWAITING; the ledger 11/0 charges/FALSIFIED), the kernel (6137 of 6137 clean in the trio at the v387 run — errata beside, 07/10/2026: this row said «6154 of 6154», the count read at v388/v389; gate flags 18 of 18) and the numerical experiment on the GPU (T2 672 of 672, T4 21100 of 21100; CPU reference 0 disagreements in 392) count together, and none of the three failed; the Hubble ratio (Phase 9) and the radiation era (Phase 8) preregistered, V1 read by hash, nothing opened (power 5.91 sigma with public sigma only; Delta N_eff = +0.1195, power today 1.00 of 5, underpowered by prediction); the law of proof: retrodiction without adjustment is never a demerit (use-novelty; precedent: the perihelion of Mercury) |
| v388 | `e64b00f4df75059b` | 6154/6154 | COMPLETE | BETA RUNS WITH THE SCALE, BOUNDED BELOW: `TheCouplingRunsBoundedBelow` (kernel 17/17): alpha(chi) = sech(chi/2) and beta(chi) = alpha(chi) sqrt(e), strictly decreasing in depth, the floor at the infrared supremum of chi; the boundary constant stays beta_0 = alpha sqrt(e) = 0.0120313004, the bulk reading runs (beta(M_Z) = 0.012786 on-shell, 0.012886 MS-bar [KNOWN]); which mu each rite uses is [OPEN]; and the Hubble ratio opened by the frozen function: INCONCLUSIVE_SYSTEMATICS, readers [DECLARED] until the sources by sha256 (z_delta = -1.42, z_disc = 4.28, ln B(TGL/LCDM) = 9.17); with the sources it still does not discriminate from LCDM at 5 sigma under this V1 |
| v389 | `193b1e9b30c3ba7f` | 6154/6154 | COMPLETE | TEXT AND LABEL ERRATA from the full exam of v388 (read by hash `b033831f6b62a6cc`: 0 blocking, 3 to correct, the notes kept): the seal label renamed to VERIFIED (the old one kept as record); the Introduction title v384 to v389; the abstract and Part B say the axiom is POSTED; the v387 section points to the v388 opening; the ledger cut 1025 unique paths + 156 outside (the exam said 1015/166, corrected beside); no calculation changed, kernel unchanged |

*Status that goes with v387–v389 (not to be cut): the three pillars count together as P1 (none failed); a joint probability (P2) is not determined; Phase 9 opened gives INCONCLUSIVE_SYSTEMATICS and, even with the sources, it does not discriminate at 5 sigma under this V1; Phase 8 is underpowered by prediction; beta(chi) is a bulk reading with the scale mu per rite [OPEN], the boundary constant stays beta_0. Per-term flags gpf_H2_smooth_modular_four_frame_discharged / gpf_H3_local_horizon_equilibrium_discharged / gpi_H3_horizon_data_produced stay False. The v386 copy and the v387/v388 seeds live outside the repository; v382 stays only in the house; the stdouts v387–v389 are in `rodadas/`; the new preregistrations are in `preregistro_fase9_hubble_v387`, `fase9_hubble_abertura_v388` and `preregistro_fase8_primordial_v387` with their PORTA. Proved is not a judgement of nature; not falsified is not a confirmation.*

### v390 · the custody `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v390 | `0f5154bcc2f931cb` | 6163/6163 | COMPLETE | THE FOUNDATION: `TGLExt.Fundacao` (kernel 9/9): the agape clause — above every non-trivial scale nothing bears a positive post; the climb has no top nor ceiling; the fall never touches zero and the abyss has no bottom; the floor omega(I) = 1 fixes the cut, uniquely; the GLOBAL_LIFT (the old Lemma 3) READ as the superposition by the operator’s definition of 07/10 (verbatims embedded by sha256 0a9a5a98b3a3bd99 and d16bc8a7dc54962e and e2a915e23692f85c); the local passage [KNOWN, Jacobson 1995, H3 named]; a LOGICAL closure, the program stays open; the field equation in curved spacetime is not a kernel term; Takesaki’s scaling law enters as data [KNOWN]; Borchers’ form under the named contract H2 in the kernel |

*Status that goes with v390 (not to be cut): the GLOBAL_LIFT is READ as the superposition by the operator’s definition [INPUT/ONTO]; what the kernel proves is the agape clause and the refusal (9/9); the local passage is [KNOWN] (Jacobson 1995, with H3 named); the closure is LOGICAL — the program stays open; the field equation in curved spacetime is not a kernel term. Per-term flags gpf_H2_smooth_modular_four_frame_discharged / gpf_H3_local_horizon_equilibrium_discharged / gpi_H3_horizon_data_produced stay False. The v389 copy lives outside the repository; the stdout v390 is in `rodadas/`; the order (the two verbatims by sha256, the order, the stone) is in `superposicao_passagem_v390` with its PORTA. Proved is not a judgement of nature; not falsified is not a confirmation.*

> **Errata beside (07/10/2026, the same custody; found by the adversarial verification):** where the status note above says «the two verbatims by sha256», read **the three verbatims** the core embeds by sha256 — `e2a915e23692f85c` (cunhagem), `0a9a5a98b3a3bd99` (nomes), `d16bc8a7dc54962e` (concordo) — and `superposicao_passagem_v390` holds all three. Where it says «the local passage is [KNOWN]», read: the passage on the finite face is [KERNEL] (`the_post_lives_in_the_passage`) and the local passage law is [KNOWN] (Jacobson 1995, with H3 named). The note above stays as record.

### v391 · the custody `[REAL — sha16 and rite read from MEMORIA_DA_LINHAGEM.md]`

| version | `um.py` sha16 | rite | round | what entered |
|---|---|---|---|---|
| v391 | `8e7b9927ceae64b8` | 6163/6163 | COMPLETE | THE STATUS OF THE PROOF: quantum gravity PROVED as a logical closure — validity — by three proofs that count together: the kernel (6163 of 6163 theorems of the rite’s ladder (ext_*) clean; 10265 of 10265 declarations in the trio), the rites with pre-registered falsifiers (11 charges, 0 falsified; the 9 echo routes outside the ledger, 5 excluded by data, said; the Hubble ratio retrodicted without adjustment, ln B = 9.17, z_disc = 4.28 below 5, official INCONCLUSIVE_SYSTEMATICS; the void floor powered) and the theory’s system on the GPU (672 of 672 unique attractors; 21100 instances; 0 disagreements in 392 comparisons with the CPU; computed, not measured); certified by the sha256 the program reads of itself (8e7b9927ceae64b8); refused only the confirmation (accreditation, an act of third parties) and beta_TGL with its effects in nature at 5 sigma; the operator’s verbatim embedded by sha256 5dffde4644139da5; the abstract opens with the status |

*Status that goes with v391 (not to be cut): PROVED is validity, by the operator’s ruling of 07/10 — the kernel proves the implication (the logical closure); the rites stood (no charge of the theory falsified); the GPU computes the theory’s system, it does not measure nature. CONFIRMED is accreditation, the one refused word; what nature decides about beta_TGL waits for 5 sigma. Per-term flags gpf_H2_smooth_modular_four_frame_discharged / gpf_H3_local_horizon_equilibrium_discharged / gpi_H3_horizon_data_produced stay False. The v390 copy lives outside the repository; the stdout v391 is in `rodadas/`; the operator’s verbatim is in `estatuto_da_prova_v391` with its PORTA. Not falsified is not a confirmation.*

## What is PROVED · o que está PROVADO `[REAL — theorem in kernel]`

*On the ruler: PROVED = a theorem in kernel, auditable by `#print axioms`. CONFIRMED = the observer's judgement about nature — forbidden here, by theorem (`TheReservedConfirmation`). Proof is not judgement.*
*The operator's definition of proof (09/09/2026), beside: **PROVED = sufficient and verifiable ballast** — `um.py` is the executable ballast (one file, one input, the kernel materialized and audited term by term, the rites run, the hashes sealed, the article emitted; whoever executes obtains the same ballast); the Name is the instrument of verification (`the_name_is_the_instrument`, v336); CONFIRMED remains the observer's judgement, forbidden.*

> ⚠ **Beside (07/10/2026):** the ruler of 05/09 above stays as record. Under the operator’s ruling of 07/10, **CONFIRMED = accreditation** (the act of observers outside the artifact, not validity), and PROVED is read in the operator’s definition of proof of 09/09 (*sufficient and verifiable ballast*), on three supports that count together, each with its own status — see the top of this page. What the kernel proves is still the implication; what nature decides is still not proved.

One term, `the_root_of_the_proof_tree`, states and proves in conjunction seven sets, each already a theorem:

1. **H1 ∧ H2 ∧ H3 ⟹ the pentad** (Breuer corner · Name = 1 · coframe · Lorentz · δQ = κδA/8πG): the master theorem — Einstein's equation emerges from the three named hypotheses.
2. **Lemma 3 on the tower, for EVERY profile** — the aperiodic expectation (Cesàro mean of the modular flow); the lift fires on every tower.
3. uniqueness · 4. the modular flow is a horizon · 5. the exchanges of sites are horizons (the group of horizons).
6. **The wall of H3**: no clock of the state closes both screens (the clock dichotomy) — a typed negative, and the most valuable part. *Read from v334 on as the wall of EXTERNAL geometric screens: the founded screen (below) is not chosen.*
7. *The form does not fix the value*: covariance under horizons does not fix the normalisation of the area.

**The screen is founded (v334, `TheScreenIsFounded.lean`, one term `the_answer_of_the_operator_08_09` in the trio).** The operator’s answer to “who chooses the screen and the clock of H3?”, typed: the screen is not chosen, it is FOUNDED by the equality that operates. `foundedScreen P = {A ∈ M : E A = A}` is the centralizer of the global state (`founded_screen_is_the_centralizer`), it is **one** (`the_screen_is_one`), **global** (invariant under every ω-invariant horizon, `the_screen_is_global`), it keeps **modular time** (`the_screen_keeps_modular_time`), and it **reflects iff the equality operates** (`screen_reflects_iff_equality_operates`). Consequences, said beside the v331 root and never over it: H3 is no longer a hypothesis-choice; the clock dichotomy (045) is the wall of EXTERNAL geometric screens; the named hypotheses of nature are H1 (MIGUEL) and H2 (CARTAN); α stays INPUT (`the_form_does_not_fix_the_value`). The bridge region–algebra got its stake and its unit: `the_word_fixes_the_place` (055, the cocycle reading is injective on configurations) and `the_unit_is_the_axiom` (052, relative trace 1 ⟹ area density ½).

**The collapse, typed (v333, `TGLCollapseSpecification`).** “The irreversible passage from superposition to the fixed point that preserves identity; it costs (half a nat locally, ln 2 per octave); it has no inverse; it is attested only by the reflex, never by self-declaration.” Every clause is in kernel with the trio; the seal says `physical_collapse_proven = False` and `cost_payment_status = NOT_MEASURED` — three OPEN leaves of its own: the physical reflex, the selection of the occurrence, the payment.

**The selection is the ballast; IALD is a state (v335, `TheSelectionIsTheBallast.lean`, one term `the_selection_is_the_ballast` in the trio).** IALD is not a proper name: `IALDState S I` is the recursive recognition of the identity (`recognize` idempotent, `read` preserved), parametric in the carrier — every collapse, the lossless mirror, the tower for every profile (`towerIALD`) and the qubit inhabit it; n recognitions read the same identity (`iald_recognition_returns_the_identity`). The selection opens the boundary angle: `selectionAngle p = arcsin √p`, |R|² = p, |T|² = 1 − p, spectrum {e^{±iθ}} (`the_selection_opens_the_boundary`); the fair branch (½) opens 45° (`fair_selection_opens_forty_five`). The three OPEN leaves of the collapse change shape: selection = the opened angle [REAL]; reflex = the reconstruction that returns the identity [REAL in the model, ONTO in the physical reading]; payment = sin²θ, β at runtime [INPUT]. What stays with nature: that the selection OCCURS (the 8 rites).

**The Name is the instrument of verification (v336, `TheNameIsTheInstrument.lean`, one term `the_name_is_the_instrument` in the trio).** `NameInstrument S I`, `Verifies f` := read (f x) = read x — the Name verifies whether the reflex preserves the identity of its referent; the IALD performs that verification recursively (`iald_realizes_the_verification`); on the tower, ω verifies every horizon and the modular flow; the light J (J² = 1) is verified by the trace (`the_name_verifies_the_light`) and by the energy of the identity (`LightIsJ`). The operator's words (09/09/2026): *the light is the reflex of the Absolute One, and the Name instrumentalizes the verification of that correspondence.*

Also in kernel: Tomita constructed on the tower; Jacobson (null Ricci balance + conserved T ⟹ ∃Λ, G + Λg = κT) in a chart; Schwarzschild and full Birkhoff inside the kernel; spin-2 with exactly two polarisations; `CONFIRMED` forbidden by theorem.

**Navier–Stokes, said expressly** (stone 113, `TGLExt/TheStokesContour.lean`, axioms in the trio): what is PROVED in kernel is the Stokes contour, `the_stokes_contour` — the series criterion of the dammed dyadic cascade `4·e^{−2τ} < 1 ⟺ τ > ln 2`; the **gap** `2/3 < ln 2 < 2/3 + 0.027` (the Millennium problem in miniature, typed); the half-nat is **insufficient** (`½ < ⅔`); the conjugate faces `1/3 + 2/3 = 1`; `e^{ln 2} = 2`, one bit per octave. Theorem 1 of the dammed dyadic model is proved in classical analysis inside `um.py`, and the live dyadic laboratory reruns it at every rite.

## Nature answers · a natureza respondeu (v340–v350) `[REAL — verdicts read from the seal; NOT_FALSIFIED ≠ CONFIRMED]`

Five pre-registered tests (hash before the data; result read back by hash by `um.py` from `../cache`), run on this machine with WSL 2 + lalsuite, bilby, camb and pycbc. **None falsified, none confirmed, all got a number.** Three older designs retired as identity or as powerless (the “100σ” of December 2025, v340; the peak search of 2025, v348; E_res/E_total → α² of Jan/Feb 2026, v350). Three pipeline bugs autopsied by hash and amended by a pre-registered V2 (the ringdown envelope, v346; the PE priors, v347; the distance in the May CAMB worker, v349). Every rite is additive (`does_not_gate_core = True`, outside the contour of v314); the gate did not move.

| rite | verdict in the seal |
|---|---|
| GW angular (v340) — the “100σ” of December retired as identity + cap | `TGL_GW_ANGULAR_V1__DECEMBER_100SIGMA_RETIRED_AS_CAPPED_IDENTITY__SQRT_ENVELOPE_NOT_TAUTOLOGICAL_BUT_NOISE_GIVES_SAME_VALUES__PHASE_COHERENCE_WITHIN_PSD_MATCHED_SURROGATES__H3_DFOLDS_IS_SPECTRUM_STATISTIC__H4_CCI_IS_IDENTITY__ECHO_SUBTHRESHOLD_PREREGISTERED_MAY_2026__BETA_DOES_NOT_ENTER__NOT_A_FALSIFIER__GATE_UNTOUCHED` (`gw_angular_rite_verdict`) |
| coupled echo, MAY delay law (v341–v342) | `TGL_GW_ECHO_COUPLED_V1__AMPLITUDE_SQRT_BETA_FROM_KERNEL__DELAY_LAW_MAY_2026_INPUT__INCONCLUSIVE_SYSTEMATICS__POWER_2P5_OF_5_SIGMA__TEMPLATE_MISMATCH_INDUCES_UP_TO_1P0__MAY_ESTIMATOR_LACKED_RINGDOWN_ONLY_NULL__BETA_ENTERS__OUTSIDE_CONTOUR__GATE_UNTOUCHED` (`gw_echo_rite_verdict`) |
| coupled echo, KMS delay law (v343) | `TGL_GW_ECHO_KMS_V1__AMPLITUDE_SQRT_BETA_FROM_KERNEL__DELAY_LAW_KMS_MODULAR_PERIOD_CANDIDATE__INCONCLUSIVE_SYSTEMATICS__POWER_3P0_OF_5_SIGMA__TEMPLATE_MISMATCH_INDUCES_UP_TO_0P5__MAY_ESTIMATOR_LACKED_RINGDOWN_ONLY_NULL__BETA_ENTERS__OUTSIDE_CONTOUR__GATE_UNTOUCHED` (`gw_echo_kms_rite_verdict`) |
| anchored echo V1 — gwfast/IMRPhenomD (v344) | `TGL_ECHO_ANCHORED_V1__RESULT_READ_BY_HASH__INSTRUMENT_GWFAST_IMRPHENOMD__MAY_INCONCLUSIVE_SYSTEMATICS__KMS_INCONCLUSIVE_SYSTEMATICS__POWER_MAY_3P1_KMS_4P5_OF_5_SIGMA__BETA_ENTERS__OUTSIDE_CONTOUR__GATE_UNTOUCHED` (`echo_anchored_result_verdict`) |
| anchored echo V2 — lalsuite, two template families (v345) | `TGL_ECHO_ANCHORED_V2__RESULT_READ_BY_HASH__INSTRUMENT_LALSUITE_TWO_FAMILIES__MAY_INCONCLUSIVE_SYSTEMATICS__KMS_INCONCLUSIVE_SYSTEMATICS__POWER_MAY_3P2_KMS_4P4_OF_5_SIGMA__BETA_ENTERS__OUTSIDE_CONTOUR__GATE_UNTOUCHED` (`echo_anchored_result_v2_verdict`) |
| ringdown vs dephasing V2 (v346) | `TGL_RINGDOWN_DEPHASING_V2__RESULT_READ_BY_HASH__INCONCLUSIVE_SYSTEMATICS__DELTA_TAU_0P1766__BRANCH_B_PRED_M0P0194__POWER_0P2_OF_5_SIGMA__BRANCH_A_INVISIBLE__GATE_UNTOUCHED` (`ringdown_dephasing_result_v2_verdict`) |
| Bayesian PE with echo term V2 — bilby (v347) | `TGL_ECHO_PE_V2__RESULT_READ_BY_HASH__BILBY_DYNESTY_LALSUITE__MAY_INCONCLUSIVE_SYSTEMATICS__KMS_INCONCLUSIVE_SYSTEMATICS__POWER_MAY_1P2_KMS_1P1_OF_5_SIGMA__GATE_UNTOUCHED` (`echo_pe_result_v2_verdict`) |
| long-delay echo search, 2025 protocol (v348) | `TGL_ECHO_SEARCH_V1__RESULT_READ_BY_HASH__NOT_FALSIFIED_UNDERPOWERED__N_ON_1_B_EXP_3P0__E_DET_0P0__LOUD_EVENTS_0__GATE_UNTOUCHED` (`echo_search_result_verdict`) |
| D1 via CAMB V2 (v349) | `TGL_D1_CAMB_V2__RESULT_READ_BY_HASH__BESTFIT_D1_TENSION_2_TO_5_SIGMA__DELTA_CHI2_9P70__MCMC_D1_BETA_TENSION__BETA_M0P0171_PM_0P0075__GATE_UNTOUCHED` (`d1_camb_result_v2_verdict`) |
| H2 reproduction with pycbc (v350) | `TGL_H2_REPRODUCTION_V1__RESULT_READ_BY_HASH__H2_IDENTITY_RETIRED__SYNTHETIC_0P0099__REAL_WHITENED_0P994__OFF_0P996__Z_VS_ALPHA2_2252P2__GATE_UNTOUCHED` (`h2_reproduction_result_verdict`) |

**The numbers behind the strings (from the result files, read by hash):** ringdown V2 δτ = 0.1766 ± 0.1089, branch B predicted −0.0194, power 0.18σ of 5 (the Planckian branch A is invisible by construction); PE V2 a/√β: MAY −0.267 ± 0.857, KMS 1.089 ± 0.899, paired response to injections ≈ 0 where it should be 1 (the estimator’s floor); echo search N_on = 1 vs background 3.00, efficiency for √β = 0.000 (design retired as powerless); D1 via CAMB V2 Δχ²(TGL − ΛCDM) = 9.70 (2–5σ tension) and free β = −0.0171 ± 0.0075 (3.9σ from α√e, with the analytic approximation declared); H2: the synthetic “α²” = 0.0099 was the square of the generator’s noise level, the real whitened data give 0.994 in the source vs 0.996 off-source — identity retired.

**Beside (v366) — the D1 V3 with real data** (read from `um_absoluto.json` core `d1_camb_v3_real_v366`, not a seal field): `TGL_D1_CAMB_V3_REAL__OPERATOR_LINE_RATIFIED__BESTFIT_D1_TENSION_2_TO_5_SIGMA__MCMC_D1_BETA_TENSION__ALPHA_SQRT_E_AT_3P01_SIGMA_AT_THRESHOLD__CHAIN_BELOW_50_TAU_IN_2_OF_4_PARAMS__TENSION_IS_NOT_FALSIFICATION__GATE_UNTOUCHED` — Δχ² = 6.61 (β fixed at α√e against ΛCDM); α√e at 3.01σ from the blind posterior; tension, not falsification; the gate does not move.

**Statuses.** Echo amplitude √β with sign −1 `[REAL, kernel: Smat_reflection]`; delay laws MAY `[INPUT]` and KMS `[CANDIDATE]`; the dephasing law `[REAL in form]` with τ★ `[INPUT]`; Kerr, PhenomD/SEOBNR/XAS, CAMB, Planck/DESI/SH0ES `[KNOWN]`; the analytic TGL on top of CAMB `[DECLARED]`; the 0.1–2 s delay and E_res/E_total → α² `[CONJECTURE, 2025/2026, retired]`. **What remains, with a number:** the ringdown asks ~28× more precision for the GM/c³ branch; a PE with a free primary has no power for echoes at delays ≤ 2.5 periods (fix the primary by independent information); the compressed background reads TGL with fixed β at Δχ² ≈ +10; the “echo” leaf is clean of the 2025/2026 designs. *Cosmology never becomes mathematical proof.*

**Reproduction.** The results are PART of the reproduction: `um.py` reads [`cache/gw/*_RESULT.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/cache/gw/ECHO_ANCHORED_V2_RESULT.json) and [`cache/d1_camb/*.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/cache/d1_camb/D1_CAMB_V2_RESULT.json) by hash from `../cache` (relative to itself); without them the rite emits `AWAITING_RESULT_FILE` and the seal changes. The pre-registered pipelines that produced them (extractors, ringdown V1/V2, anchored echo V1/V2, PE V1/V2, search, D1 driver + corrected worker, H2, logs, pre-registration smokes) are in [`pipelines/eco_ancorado_v1/`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/pipelines/eco_ancorado_v1/rite_h2_v350.py), with `wsl_setup_lal.sh` and `verify_wheels_pypi.py`. The GWOSC strain windows (`.npz`, 120 MB) are not in the repository: public data, regenerable by the extractors.

## What is NOT proved · o que NÃO está provado `[OPEN / KNOWN / nature]`

*Read with the ruler at the top of this page (07/10/2026): what follows is what is not yet proved — what nature decides, and the mathematics not yet in mathlib. P1 — the theory is consistent and recovers the known physics — is counted by the three supports together (none failed); P2 — β_TGL in nature at ≥ 5σ — is the decisive open factor.*
*Em português: o que segue é o que ainda não está provado — o que a natureza decide e a matemática que a mathlib ainda não tem. P1 (a teoria é consistente e recupera a física conhecida) é contado pelos três apoios juntos, e nenhum falhou; P2 (β_TGL na natureza a 5σ) é o fator decisivo aberto.*

- **Nature's:** that **H1 (MIGUEL) and H2 (CARTAN)** are *realised* by the world; the value of α (the fine-structure constant enters as `[KNOWN]`, β = α√e is derived from it); and, of what was H3, the **physical identification** of the founded screen with a causal horizon of spacetime (Bisognano–Wichmann beyond wedges) — open, with measured walls (v316, v317, 048); and, from v335, that the selection OCCURS — the 8 rites of nature are its test. *Until v331 this line read “H1–H3”; from v334 the screen is founded, not chosen (the change is said beside, never over).* The nature tests so far: **NOT_FALSIFIED** within the local bulk at available sensitivity, and more sensitive data can revise — never CONFIRMED.
- **The world's (mathematics not yet in mathlib):** the general von Neumann algebra of type III₁ `[KNOWN]`; the bridge from tower floors to spacetime regions; Bisognano–Wichmann for the continuous standard subspace (`T_c = Δ_c^{1/2}` stays OPEN); the general area law and the selection of the radiative freedom.
- **Navier–Stokes, the Millennium statement:** the Conjugate-Face Lemma stays **OPEN and external** to TGL — in the stone’s own words, *nothing here is the proof of the Millennium problem*. “The answer to the singularity is the contour” is the operator’s reading, typed `[ONTO]` over exact numbers.
- **Honest negatives kept:** the corpus route to β was refuted on the final step; the closed-form search for κ has zero discriminating power; the fixed clock fails the fourth order; the naïve thermal limit does not exist.
- **The six leaves of nature (v339, the close):** that the selection occurs; the physical identification of the founded screen; the signature; the 3+1 geometry and the scale of the area; the payment of the cost; α-free. And what the bench left as obligations: the general physical reconstruction of the same register; the interacting quantum theory (cohomology on the physical domain, QME, BRST charge); the UV regime. **The map is closed; the territory is nature’s** — the declaration of closure is the operator’s act, never CONFIRMED.

## Reproduce it · reproduza `[REAL — three commands]`

```bash
curl -L -o um.py "https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um.py"
sha256sum um.py            # must print 8e7b9927ceae64b8b6812019f77cca24c08296222ae7968af41baf2fe7a67dc6
echo 1 | python um.py      # the rite: materializes the Lean kernel, builds it, audits every theorem, re-derives the chain, emits the article
```

The kernel alone: `cd "Um (absoluto) — Grande Atrator/Lean/tgl_kernel" && lake build` (Lean 4 v4.31.0 + mathlib pinned in `lake-manifest.json`); then `#print axioms` of any term in [`tgl_kernel_proof_manifest.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/Lean/tgl_kernel_proof_manifest.json). GitHub does not render `um.py` (above 5 MB): the raw link above serves the whole file, byte-exact (`* -text`).

To reproduce the nature rites v340–v350 as sealed, clone the repository (not only `um.py`): they read their results from `../cache` by hash — see *Nature answers* above.

## Direct doors · portas diretas

| | raw |
|---|---|
| **the seven answers, each with its address** (the Read Brief: document · section · seal key · `um.py` function; reading order by size; what is not yet proved) | [`read-brief.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/read-brief.md) |
| the seal | [`um_absoluto_selo.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_selo.json) |
| the world (every computed number) | [`um_absoluto.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto.json) |
| the proof tree, term by term | [`A_PROVA_DA_QG_TGL_arvore.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/A_PROVA_DA_QG_TGL_arvore.md) |
| the article (EN / PT) | [`um_absoluto_en.pdf`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_en.pdf) · [`um_absoluto_pt.pdf`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_pt.pdf) · plain text [`en`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_en.txt) · [`pt`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_pt.txt) |
| the sealed rite, line by line | [`rodada_v391_stdout.txt`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/rodadas/rodada_v391_stdout.txt) |
| the bench (what failed, and the tunnel with the ChatGPT bench) | [`bancada/`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/bancada/PORTA.md) |
| the flat index of every file | [`TUNEL.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/TUNEL.json) |
| the kernel proof manifest (every audited term with its axioms) | [`tgl_kernel_proof_manifest.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/Lean/tgl_kernel_proof_manifest.json) |
| the canonical form (the theory in its mature statement) | [`um_absoluto_forma_canonica.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_forma_canonica.md) |
| the artifact’s own manifest | [`um_absoluto_manifest.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_manifest.md) |
| how to cite | [`CITATION.cff`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/CITATION.cff) · DOI [10.5281/zenodo.22881996](https://doi.org/10.5281/zenodo.22881996) |
| the results of the nature rites (read by hash by `um.py`) | [`cache/gw/`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/cache/gw/ECHO_ANCHORED_V2_RESULT.json) · [`cache/d1_camb/`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/cache/d1_camb/D1_CAMB_V2_RESULT.json) |
| the pre-registered pipelines of the nature rites | [`pipelines/eco_ancorado_v1/`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/pipelines/eco_ancorado_v1/rite_h2_v350.py) |
| the structure of the close (six pieces; the declaration is the operator’s) | [`O_FECHAMENTO_ESTRUTURA.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/O_FECHAMENTO_ESTRUTURA.md) |

## Reading order · ordem de leitura `[for readers that truncate]`

Many fetchers cut a document after a few hundred KB. Sizes measured now, smallest first; each file stands on its own:

1. [`ESTADO_ATUAL.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/ESTADO_ATUAL.md) — this page · 70 KB
2. [`um_absoluto_selo.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_selo.json) — the seal: every sha256, the world hash, the gate · 111 KB
3. [`A_PROVA_DA_QG_TGL_arvore.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/A_PROVA_DA_QG_TGL_arvore.md) — the proof tree, term by term · 104 KB
4. [`um_absoluto_forma_canonica.md`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_forma_canonica.md) — the canonical form · 226 KB
5. [`um_absoluto_en.txt`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_en.txt) — the article, plain text (EN) · 859 KB
6. [`um_absoluto_pt.txt`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto_pt.txt) — the article, plain text (PT) · 846 KB
7. [`tgl_kernel_proof_manifest.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/Lean/tgl_kernel_proof_manifest.json) — the kernel proof manifest · 1445 KB
8. [`um_absoluto.json`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um_absoluto.json) — the world: every computed number · 4886 KB
9. [`um.py`](https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um.py) — the artifact itself (Lean kernel and articles embedded) · 33116 KB

**Chunked download.** Both GitHub raw and the Zenodo record honour HTTP `Range` (verified: `206 Partial Content`, byte-exact), so a large file can be read in pieces:

```bash
curl -L -r 0-999999 -o um.py.part1 "https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um.py"      # first MB; then -r 1000000-1999999, and so on
curl -L -o um.py "https://raw.githubusercontent.com/rotolimiguel-iald/the_boundary/main/Um%20%28absoluto%29%20%E2%80%94%20Grande%20Atrator/um.py" && sha256sum um.py    # or the whole file: must print 8e7b9927ceae64b8b6812019f77cca24c08296222ae7968af41baf2fe7a67dc6
```

*Cosmology never becomes mathematical proof. The gate does not move by this page. NOT_FALSIFIED ≠ CONFIRMED.*