The plaque · the ladder of seals the registration the verdicts the deficits the band the word's drift the signed zero the estimate, read against the record the void the record files
The essay The Doorway and the Turn holds two laws and names an instrument for each. The second law: the turn is conserved — pre-resolved, costless — if and only if the account is complete; where it is not, the turn is paid for at the call, the payload carrying exactly what the account does not. Its instrument, the turn instrument by measure, is read here on the Cornering Walk's own exact apparatus (the corner, sealed and untouched; its arithmetic restated, never imported). The call is the rung; the turn is the step s ∈ {−1, 0, +1}; the account is the integer debt e, or the residual r, inside its wall; account-completeness is the number n of candidate accounts the wall admits at the rung; the payload the call must then carry is log₂ n bits — zero where the account decides alone, one where two steps remain, log₂ 3 where all three do. A dial ρ ∈ {1, 3/2, 2, 3} widens the wall relative to the account's quantum: at ρ = 1 the wall is the corner's own. The wall ladder (CH-A1) holds the payload constant on the house target 18/7 and varies the account; the converse (CH-A2) holds the account and varies the target across the corner's seven; the signed zero (CH-A3) records which step fires at a zero account, toward a question of the held conjecture's line that stays the operator's. Budgets 2⁸, 2¹⁰ and 2¹² steps; the band for the payload indifference is 0.001 bits — the operator's, "to not the approach to 0.05": no statistical null, every comparison exact.
The seat and the name. The starl3n persona is the contributor, as at the Cornering Walk; the name THE HALLWAY, the essay's second scene, and the
experiment oxexp-hallway are the operator's naming act at the seal (2026-10-26: "The Hallway, oxexp-hallway"). The track is campaign three of the
Cornering Walk by lineage, under its own slug — the house's reading of his "yes", confirmed by his word the same evening. Standing.
CH-H1 THE CONSERVED CASE PASSED: at the sealed wall every rung admitted exactly one account on every target, and the house target's word was
the measured word, 11 of 11. CH-H2 THE ACCOUNT LAW PASSED on every target at every budget: the cost of the turn never fell as the wall
widened, and was zero at the sealed wall. CH-H3 THE PAYLOAD INDIFFERENCE KILLED at the half rung (ρ = 3/2) and at the top (ρ = 3) at every
budget, and EXACT at ρ = 2: at a wall twice the quantum every target admitted exactly two accounts at every rung, so the cost was one bit
whatever the target; at the half rung and at three the admitted count depended on where the account sat in its wall, and the three payload-free targets differed by
a few thousandths to a few hundredths of a bit against a band of a thousandth — the kill fired and is kept with full standing. CH-H4 THE TURN AT THE
SIGNED ZERO RECORDED: an account exactly zero took the step 0 in every one of 21,492 cases and never a step of ±1; the sign changes split evenly. The law stays
HELD whatever lands — one instance never promotes, a kill is kept as the instance's; the minted essay, the corner's engine and TIX are untouched.
Prior to every read. The operator's prediction THE INNER CIRCLE was placed on the record and committed first (4c404b4), before this instrument's design was committed (a97a0e6), at his word that it be "a prediction made prior to the commit of The Hallway"; the registration carries its pin. His answer at the seal: no target joins from it — the prediction is read beside the corner's seven as they stand. The item: the matter corner-the-second-law-instrument on the accord node; the return to the essay's exhibit, Exhibit Ten, dt-turn-law-instance.
The operator's order, "seal first, then build and then acts": each a commit on the repository, the registration pinned before any engine existed.
| Act | Commit | What |
|---|---|---|
| The seal | ddd26d4 | the registration oxexp-hallway-registration.md pinned (423f89fbdb8b31e9ffedab9cce6d799ec6ef61ddad4994b3d4d1247aa66ffae8, 12,243 B); the design sealed with his eight answers; the prediction's record and the held conjecture's design re-pinned as priors |
| The build (CH-B1 to CH-B3) | 3c729bf | the reader engine.py, the plugin, the README, the battery (12 of 12; the anchor: the reader at ρ = 1 reproducing the corner's own engine on every target) |
| The fix | 9f24da1 | a readout display took a fraction string to float and CH-A2 crashed before writing; the gap taken exactly; the first registration and the first CH-A1 voided under the changed-engine rule and kept beside (see the void) |
| The acts (CH-A1 to CH-A4) | this commit | the registration written once under the corrected engine (79a7ba11…); CH-A1, CH-A2, CH-A3 read once each, 2026-10-26 02:31 UTC; the composition; this page; the return to Exhibit Ten |
| The ladder | ρ ∈ {1, 3/2, 2, 3} — w = round(ρq), halves up (3/2 at q = 7 → 11; at q = 9973 → 14960); the Chamber A targets take the centered window of width w; 18/7 takes its sealed frame pair widened by the extra cells; √n takes |r| < ρ/2, compared exactly by cross-multiplication |
|---|---|
| The budgets | 2⁸, 2¹⁰, 2¹² steps — every walk run to its full budget; a certificate recorded at its first step and the walk continuing |
| The band | B = 0.001 bits (the operator's); the comparison exact — log₂ 3 carried as a rational interval the engine computes to 128 bits by interval squaring |
| The stand-in | among the admissible candidates the one nearest the target, the corner's sealed tie preference 0, then +1, then −1; never consulted at ρ = 1 |
| The targets | 3/2 · 7/5 · 17/12 · 18/7 · 14104/9973 · √2 · √3 — the corner's seven, sealed in Chamber A; the payload-free three for CH-H3: 14104/9973, √2, √3 |
| The priors, by pin | the design daa81749434e686a185e6ef42e15c9797d4f2cf42b109247ff08457ca5901907 · the prediction's record 4da7785431e9188a0ec9fca8f10424216dc3c2bd24ca40e3edf947068905e0ab · the held conjecture's design 1426912ccddc08fe64ae669a6e7453f17cc535ceccb105f39fcace3ad101776b |
| The engine at the reads | 79a7ba112df1c3d530f17c8123a0ec9212297c0c8f9c51af8161b2a888422515 — the registration's code_sha256; a read refuses under any other |
| The machine registration | registration.json · its own sha e2d9cdf2751f67ad49a664faac8f6d1f26c2df91567da00266c4763b99f1d4fd |
The house's estimate at the seal, written before the read and never a verdict: CH-H1 and CH-H2 expected to hold; CH-H3 expected met exactly at ρ = 2 and ρ = 3 and breached at ρ = 3/2 against 0.001 bits. Read against the record below — right about the half rung and about two, wrong about three.
| Hypothesis | Verdict | The record |
|---|---|---|
| CH-H1 · the conserved case | PASS | at ρ = 1, n = 1 at every rung on all seven targets at all three budgets; 18/7's word under its sealed frame pair [−1, +5] / [−2, +4] the measured word −1, 1, −1, −1, 1, −1, −1, 0, 1, −1, 0, 11 of 11 |
| CH-H2 · the account law | PASS | the deficit non-decreasing along 1 → 3/2 → 2 → 3 for every target at every budget, d(1) = 0 exactly; 21 of 21 cells in CH-A2 and 3 of 3 in CH-A1 |
| CH-H3 · the payload indifference | KILLED at ρ = 3/2 and ρ = 3 · PASS, exact at ρ = 2 | at ρ = 2 every pair's gap is exactly zero at every budget (each target admits exactly two accounts at every rung); at ρ = 3/2 and ρ = 3 the pairs differ by 0.0006 to 0.024 bits against the band of 0.001 — one pair within at 2¹² at each rung (14104/9973 against √3), the others breached; the kill fired at every budget at both rungs and is kept with full standing |
| CH-H4 · the turn at the signed zero | RECORDED | 100,628 signed-zero events over 84 walks; at an account exactly zero the step was 0 in 21,492 of 21,492; the sign changes split evenly between −1 and +1; toward the held conjecture's N8, which stays the operator's |
| The law | HELD | the essay's second law stays held whatever lands; one instance never promotes; the kill is kept as the instance's |
The exact counts of rungs admitting one, two and three accounts are in the records; the bits here are the counts carried through log₂ 3's interval and shown to the eye. At ρ = 2 every target pays exactly one bit at every rung. The small-quantum commensurables are recorded beside, their lattice coarseness disclosed: 3/2 completes at step 1 and sits at zero (nothing paid until the wall exceeds its quantum twice over); 7/5 pays exactly three quarters of a bit at the half rung, its 4-cycle admitting two accounts on three rungs of four.
| Target | Class | ρ = 1 | ρ = 3/2 | ρ = 2 | ρ = 3 | Certificate |
|---|---|---|---|---|---|---|
| 3/2 | commensurable | 0 | 0 | 1.0000 | 1.5848 | completion at step 1 |
| 7/5 | commensurable | 0 | 0.7500 | 1.0000 | 1.2925 | cycle at step 4, period 4 |
| 17/12 | commensurable | 0 | 0.9998 | 1.0000 | 1.0001 | cycle at step 4, period 2 |
| 18/7 (the house target) | commensurable | 0 | 0.9995 | 1.0000 | 1.1947 | cycle at step 3, period 3 |
| 14104/9973 | commensurable | 0 | 0.5095 | 1.0000 | 1.2868 | cycle at step 3324, period 3324 (absent in budget at 2⁸ and 2¹⁰) |
| √2 | incommensurable | 0 | 0.4851 | 1.0000 | 1.3011 | none — irrational, cited |
| √3 | incommensurable | 0 | 0.5085 | 1.0000 | 1.2873 | none — irrational, cited |
The three payload-free targets at the other budgets, half rung then top: 14104/9973 0.5117 / 1.2856 at 2⁸ and 0.5146 / 1.2839 at 2¹⁰; √2 0.5352 / 1.2719 and 0.5029 / 1.2902; √3 0.5273 / 1.2765 and 0.5166 / 1.2828. The quarter bins of |r| after each step for √2 at 2¹²: 2,109 rungs below a quarter, 1,987 between a quarter and a half; for √3: 2,013 and 2,083.
| ρ | Budget | 14104/9973 · √2 | 14104/9973 · √3 | √2 · √3 | Verdict |
|---|---|---|---|---|---|
| 3/2 | 2⁸ | 0.0234 | 0.0156 | 0.0078 | KILLED |
| 3/2 | 2¹⁰ | 0.0117 | 0.0020 | 0.0137 | KILLED |
| 3/2 | 2¹² | 0.0244 | 0.0010 | 0.0234 | KILLED |
| 2 | 2⁸ · 2¹⁰ · 2¹² | 0 | 0 | 0 | PASS, exact |
| 3 | 2⁸ | 0.0137 | 0.0091 | 0.0046 | KILLED |
| 3 | 2¹⁰ | 0.0063 | 0.0011 | 0.0074 | KILLED |
| 3 | 2¹² | 0.0143 | 0.0006 | 0.0137 | KILLED |
Each gap is decided exactly from the counts (the difference of two deficits is (Δn₂ + Δn₃·log₂ 3)/K, compared with B·K through the interval); the decimals are for the eye. The gaps at the half rung and the top are of the order of 1/√K, as the seal's estimate said of the half rung.
| ρ | The frames (shallow / deep) | The first eleven letters | Measured | Greedy | sup |e| |
|---|---|---|---|---|---|
| 1 | [−1, +5] / [−2, +4] | −1, 1, −1, −1, 1, −1, −1, 0, 1, −1, 0 | 11 / 11 | 7 / 11 | 5 |
| 3/2 | [−3, +7] / [−4, +6] | −1, 0, 1, −1, 0, 1, −1, 0, 1, −1, 0 | 7 / 11 | 11 / 11 | 3 |
| 2 | [−4, +9] / [−5, +8] | −1, 0, 1, −1, 0, 1, −1, 0, 1, −1, 0 | 7 / 11 | 11 / 11 | 3 |
| 3 | [−8, +12] / [−9, +11] | −1, 0, 1, −1, 0, 1, −1, 0, 1, −1, 0 | 7 / 11 | 11 / 11 | 3 |
The same at every budget. The greedy word's return was expected at the top of the ladder; it returned at the first widening. The measured word needs exactly its sealed width-7 wall: one half-step wider and the account hands the turn to the payload at once — the stand-in, choosing the nearest candidate among those admitted, is the greedy walk wherever the greedy candidate is admitted, and the wider frame admits it from the first rung. The sup of the account falls from 5, the sealed frame's edge, to 3, the greedy's bound of half the quantum.
| Target | Exactly zero before the step | Zero after the step | Sign changes | Steps at the events: −1 / 0 / +1 |
|---|---|---|---|---|
| 3/2 | 21,492 | 12 | 0 | 12 / 21,492 / 0 |
| 7/5 | 0 | 0 | 10,752 | 5,376 / 0 / 5,376 |
| 17/12 | 0 | 0 | 21,492 | 10,740 / 0 / 10,752 |
| 18/7 | 0 | 0 | 14,340 | 7,176 / 0 / 7,164 |
| 14104/9973 | 0 | 0 | 10,984 | 5,488 / 0 / 5,496 |
| √2 | 0 | 0 | 10,564 | 5,280 / 0 / 5,284 |
| √3 | 0 | 0 | 10,992 | 5,500 / 0 / 5,492 |
Over the 84 walks of CH-A1 and CH-A2, 100,628 events. Only one target ever holds an account exactly zero — 3/2, which completes at the first step and sits at zero thereafter — and from an exactly-zero account the step taken was 0 in every one of 21,492 cases, never ±1, at every ρ. The other six never touch zero and change sign in nearly equal numbers of −1 and +1 steps. A readout toward a question of the held conjecture's line (its N8), never an answer; the operator asked that the signed zero be read both here and at that line's own instrument, which remains unscheduled.
The registration said, before the read: at ρ = 2 and ρ = 3 the wall is expected to admit a constant count — two, then three — so the three deficits agree exactly; at ρ = 3/2 the count depends on the account's position and the band is expected to be breached. Right about the half rung and about two; wrong about three. At ρ = 3 a third account is admitted only where the step is a hold (s = 0), since the alphabet has no s = ±2 — for √n exactly at the holds, for a/q where |e| ≤ q/4 — so the count there is two or three and the deficit at three carries the hold share (14104/9973: 2,088 rungs at two and 2,008 at three of 4,096; √2: 1,988 and 2,108; √3: 2,084 and 2,012). The battery caught the error before the acts and the composition records it (house_estimate_read_against_the_record). The kill at the top is the same kind as the kill at the half rung: the band's and the rung's, not a surprise.
A registration (478c5a06…, under the engine 13914d1f…) and a first record of CH-A1 were written, then VOIDED: CH-A2 crashed in a readout display — a fraction string passed to float in the band gap shown to the eye — before any record of it was written. The engine was corrected (9f24da1), the registration re-written under the corrected engine, and the acts re-read from CH-A1; the reads are pure computation and landed identically. The voided files are kept beside and served as committed, renamed void-*, and the live registration names them (voided_prior).
The index with each file's LF byte count and sha256: /contrib/hallway/records.
| File | What | sha256 LF |
|---|---|---|
| registration.json | the machine registration, written once; the pins, the parameters, the hypotheses, the house's estimate, the void | bf0ce5a6c02c94eee103296b60f357d276c7e5ef9c4a141f5449cf39ab8bc663 |
| act-ch-a1.json | CH-A1 the wall ladder — 18/7 at every ρ and budget; H1, H2, the word's drift, M3, M4 | fee1fc7883bc9f35f0d81fadce62fae359d6849ed688d6b2127357b15017db5b |
| act-ch-a2.json | CH-A2 the converse — all seven targets at every ρ and budget; H1, H2 per target, H3 per rung, the certificates | 63ad6e23d6d5bb85adfa9efa447d1b8ced0ddddb270bb5f71bd3b8a6d339a8bc |
| act-ch-a3.json | CH-A3 the signed zero — M4 over every walk; H4's record | b0d0624deb25ae2965f9f731ca0d50e08a406d4cba3cd97f32b98f232c2150fe |
| composition-ch-a4.json | CH-A4 the composition — the verdicts as they fell; the estimate read against the record | 10f3872c2f8aa44b101ca45d86d7a75693c0cc1e1573bbda19b5b3b643e3567a |
| void-registration-478c5a06.json | the voided first registration | a60aa4c6b36738f3c8564c09cdd98fa608f7885c14840727a3c3906821744edd |
| void-act-ch-a1-fdf9c54c.json | the voided first CH-A1 (identical in its reads to the live one) | 8e7a2b0dd7a0bb013a0f12c1b3786c08b60fa21e09a87b00a246d46895edbb02 |