The Binary Tower

The framework’s computational instrument — with the Corridor, the 400/11 residual, and the connection map.

Before you read a number here: the Tower is a unit converter and a landmark instrument — a reading is an address, not evidence, and states no claim. Arc on this page is the hyperbolic argument L = ln r and chord is 2 sinh(L/2), both in one metric — the Read tab says it once, with the rules M1–M10; the symbols of every field are on the Gap Dashboard (“In other words”), and the letters follow ISO 80000-2.

How to read the tower

The metric, said once. On this page arc is the hyperbolic argument L = ln r (ISO 80000-2: ar-; the Minkowski arc between the two points) and chord is 2 sinh(L/2), their chord in the same metric. K_AUD = arc/chord = L/(2 sinh(L/2)) = GM/L_log is a ratio taken in one metric, which is why it sits below 1; on a circle the same ratio sits above 1. Doubling names the interval r = 2 only. K_AUD is the framework’s historical symbol for the value p at r = 2 — this page keeps it because its laws and self-test were written in it; it is not a new constant. What the Tower is for and is not for is written out beside this file (WHAT THE BINARY TOWER IS — is for, is not for, 2026-09-08); the symbols of every field, fifteen rows, are on the Gap Dashboard’s “In other words” block.

The tower is a measurement instrument and a recognition index — its hits are addresses, not evidence. Golden targets carry an irreducible ε by theorem; arc-native hits survive the tower’s deletion as exact identities, and that surviving identity is the finding.

Where the unit comes from (added 2026-08-27). The tower converts by G_AUD, and G_AUD is fixed by one matrix. Doubling, z → 2z, is diag(√2, 1/√2) in PSL(2,ℝ); for any hyperbolic element |tr| = 2·cosh(ℓ/2), so tr² − 4 = 4·sinh²(ℓ/2), and p = 2·arccosh(|tr|/2) ÷ √(tr² − 4). Here tr² = 9/2 exactly, √(tr² − 4) = 1/√2, and p = √2·ln 2 = K_AUD, so G_AUD = 1 − K_AUD. Because the formula sees the matrix only through its trace, the unit is the same for every copy of the dilation z ↣ 2z (translation length ln 2) — the tower’s ruler does not depend on where you stand. Wording may still be adjusted.

The gap as a unit — the conversion table

The tower is, in the end, a familiar kind of tool: a unit converter. The unit is the gap itself — 1 gap = G_AUD = 0.019741856531… — and a tower reading is a number expressed in gap units, n = t/G_AUD, the way kilometres express a distance already measured in miles. A conversion states no claim. It is the same number in a different unit — exact, useful, and silent about meaning. The framework’s “translated readings” are precisely such conversions; the “derived marks” are the unit system’s own calibration points, like 0° and 100° on a Celsius thermometer. This converter covers one dimension only: lengths along the gap’s own axis. The live converter is temporarily removed for rework; this is the printed card.

value	t	in gap units (t/G_AUD)	nearest step	remainder (gaps)
derived marks — the unit’s own calibration points
ln 2 — one bit; the arc itself	0.6931471806	35.1105368158	35	+0.110537
1/√2 — the chord	0.7071067812	35.8176435970	36	−0.182356
1/(2 ln 2) — Shannon bound	0.7213475204	36.5389911174	37	−0.461009
K_AUD — the ceiling	0.9802581435	49.6537973471	50	−0.346203
1 — unity, the last derived mark	1	50.6537973471	51	−0.346203
conversions — world values expressed in gap units (no claim stated)
1/φ — the floor	0.6180339887	31.3057684198	31	+0.305768
√φ	1.2720196495	64.4326255480	64	+0.432626
e/2 — the control	1.3591409142	68.8456484356	69	−0.154352
√2	1.4142135624	71.6352871940	72	−0.364713
φ	1.6180339887	81.9595657669	82	−0.040434
π	3.1415926536	159.1335976220	159	+0.133598

One thing worth seeing: K_AUD and unity share a remainder — they sit exactly one gap apart (1 − K_AUD = G_AUD), the unit measuring itself. And ⌊1/G_AUD⌋ = 50: fifty whole gaps fit under one, which is also the first term of G_AUD’s continued fraction — the same fact twice, because the first term of a continued fraction is the floor of the reciprocal, by definition of the algorithm. Guaranteed, not found. All values computed and verified at mpmath dps 500 — the framework’s uniform working precision. Why so deep for ten printed digits: different closed forms and drifts reveal themselves at different depths — a false match can hold at 30 digits and break at 200 — so one deep standard is kept for everything, as a general safety procedure (Methodology). The page itself computes in the browser’s double precision (binary64) and stops its readouts at fifteen digits: it shows, the table and the papers verify. Remainders are signed distances to the nearest step, in gaps. Each landing is a reading, not an identity — one bit (ln 2) sits at 35.11, off the rung by +0.11; what is exact here is the spacing between marks (Δ ratio = 1/K_AUD), not the marks themselves. See the Convert tab's note.

The staircase — n × G_AUD, drawn

The conversion table above, made visible. Each faint rung is a whole step n, at height n·G_AUD; a landmark is a known constant that lands near a rung — drawn at its exact position with its signed remainder, never as the integer alone. Derived marks (blue) are the unit’s own calibration; conversions (amber) are imported values in gap units and state no claim. The derived tower ends at unity (n ≈ 50.65, since K_AUD < 1 is a theorem); 64 is the translated window (the √φ pivot) — structure, not dynamics.

hover the staircase — step, height, nearest landmark, remainder

Positions computed live from G_AUD = 1 − √2·ln 2; they match the printed table (mpmath dps 500). Further conversions sit above the shown window: e/2 at 68.85, √2 at 71.64, φ at 81.96, π at 159.13.

Bits & orbit — the doubling lens

The staircase measures a number in one unit, the gap. This reads it in a family of units, the powers of two — the same as walking its binary digits, one doubling at a time. The step is ln 2 because ln 2 is the translation length of doubling (z ↦ 2z), so this is the arc’s own native motion. Folded in from the Dyadic Ruler; bits computed mpmath dps 60.

first 48 binary digits — every 1 sits over a 0

G_AUD

K_AUD

G_AUD + K_AUD = 1 → exact bit complement, 72 of 72. The gap-ruler reads them one step apart; the bits read them as opposites — one identity, two witnesses.

Each row is one doubling. x is where the number lands in [0,1). Rationals fall into a few columns and close (period = order of 2 mod the denominator); irrationals never land twice.

Reading guide: try 1/7 (three columns, closes at 3) and 1/11 (ten), then G_AUD or ln 2 — they refuse to ever repeat. That scatter is the null: no hidden structure in base 2. The one guardrail — this lens is base-2 native, so a pattern only counts if it would survive in another base.

“The tower was born from noticing — ratios, pivots, crossings that kept landing near each other. It exists so those meetings can be watched instead of just remembered.”

Coincidences are easier to judge when they stand on the same scale — one ladder, where everything noticed gets to stand. Some patterns dissolve under that test. The ones that survive are in the papers.

What is this tool?A ladder of halvings: rung n holds (1/2)ⁿ. The only moves are double and halve.▸

That’s all “binary” means — counting with just 0 and 1, so every number becomes a recipe of doublings. Your computer plays videos and games by stacking halvings and doublings very, very fast. This tower is that world, drawn as a ladder you can climb.

What is K_AUD?K_AUD = √2 · ln 2 ≈ 0.980258 — geometry’s simplest diagonal times information’s simplest cost.▸

√2 is what a square’s diagonal costs; ln 2 is what one yes/no decision costs. Their product is the framework’s central constant — and what’s left over is the Gap: G_AUD = 1 − K_AUD ≈ 0.019742. About 2%.

What does it compute?Type a number — the tower lights the rungs that build it. That’s its binary recipe.▸

Two ladders, one family: this widget’s ladder is the binary recipe — rung n holds (1/2)ⁿ, and your number lights the rungs it uses. The framework’s Tower staircase is a different ladder: its steps are multiples of the Gap (step n sits at n·G_AUD), and that is where the named landmarks live. They come in two kinds, and the difference matters: derived marks — ln 2 at 35.111, the chord 1/√2 at 35.818, K_AUD at 49.654, unity at 50.654 — are the unit’s own calibration points, and √2/4 at 17.909 belongs with them, being exactly half the chord. Conversions — 1/φ at 31.306, √φ at 64.433, φ at 81.960, and π/3 at 53.045 — are imported values expressed in gap units, and state no claim. The Telescope Tower searches that staircase — and always states the deviation. Try the widget below for the recipe side.

Why the Gap mattersThe same ~0.0197 keeps being noticed in places that never asked for it.▸

A hyperbolic-geometry bound, a 1970s sampling algorithm’s saturation point, the η corridor of 3D phase transitions. The papers document each appearance with evidence labels — Theorem for proved, Observation for seen-and-verified-but-not-derived — and those labels appear on this page too. Why labels? → Methodology.

The gap, exactly — a tool for G_AUD

Everything above works on the digits of a constant. This section works on one constant's structure. G = 1 − √2·ln 2 = 0.019741856531452808286… is not a leftover: it has an exact rational series, its two published bounds are that series' first two partial sums, and its remainder under halving is a smaller copy of itself. The series panel below is computed in exact rational arithmetic with BigInt — no floating point anywhere in it. Added 2026-08-27.

1 · The series — exact rationals

      GAUD = Σk≥1 (−1)k+1 · C(2k,k) / ( 32k · (2k+1) )
    

2 · The halving ladder — what is left after N halvings

Halve the cell N times and the remainder is 1 − p(ln 2 / 2^N) — the gap of that rung's own cell. It is also the relative error of the composite midpoint rule on 2^N panels, which is why the ratio tends to ¼: the midpoint rule is second order.

3 · Any integer cell — and why 2 is the only one with a binary base

On the cell [x, n·x] the same expansion has base 16n/(n−1)². It is an integer only at n = 2, 3, 5, and a power of two only at n = 2.

What this tool does not show. The first rung of the ladder carries about 74.9% of the whole, and that is generic — a cascade of ratio ¼ always puts ¾ on its first term, at every length. It is not a fact about √2 or about base 2. And the ladder is scale-free: adding a rung at the top is exactly doubling, so it does not single out n = 2. What singles out 2 is elsewhere — one bit consumed, the unit denominator n − 1 = 1, and the base 32 = 2⁵ in panel 3.

And why this cell at all — three computations. G > 0 for every r > 1, from the strict concavity of arcsinh — a theorem, so the function prices any relation. But no feature of it sits at ln 2: swept over L ∈ (0, 30), the slope has no zeros and the curvature has exactly one, at r = 24.83; ln 2 sits at 21.6 % of the way to it, and 11.0 % of the way to the nearest singularity. What is left: 2 is the first integer ratio, and G_AUD is what that first relation costs. Panel 3 above is the same statement in the other direction — it is the base 32 = 2⁵ that is unique to this cell, not a feature of the curve.

Try it — the binary recipe

The tower decomposes anything — that is what calculators do. Lit rungs are arithmetic, never significance. Significance, when claimed, lives in the papers, with evidence labels attached.

Reading the ladder: each rung n is the value (1/2)ⁿ. A lit rung means the recipe uses that halving. This is the recipe ladder — the framework’s landmark staircase (steps = multiples of G_AUD) is a different ladder; the Telescope Tower searches it and reports how near any number lands to a G_AUD-step, deviation included.

Seeing relations: on the tower, dividing by 2 is a one-rung slide. Try G_AUD vs 2G, or 1/√2 vs √2/4 — same recipe, shifted one rung. Relations between framework numbers become shapes.

Full instruments: Dashboard · Telescope Tower

→ Gap Geometry · the framework · notation follows ISO 80000-2 · license