fix(register): demote Z-3 (sixth broken discharge) + quarantine the five void scripts + honest-label the constant - #9872
Merged
Conversation
…ive void discharge scripts + honest-label the 1/(3sqrt2) constant
Z-3 VERIFIED BROKEN AND DEMOTED §A -> §B. Six independent counts, each checked against the code:
1. The "identity" is -ln(1/x) = ln(x), evaluated once. `computeZ3(kappa)` sets
tStar = kappa * TSIRELSON and returns -Math.log(tStar / kappa); the kappa cancels SYMBOLICALLY
before any float is evaluated, so the result is -ln(TSIRELSON) for every input. True for EVERY x,
therefore silent about Loewner entropy, SLE, or kappa.
2. "For all kappa" is vacuous: t/kappa is PINNED to a constant by construction, so kappa cannot
influence the outcome. Structurally the same defect as Z-7 (runDLA(seed) never received `compiler`).
3. NO LOEWNER EQUATION IS INTEGRATED. The panel header claims it "runs the Loewner equation for all 4
kappa presets"; the file contains no ODE solver, only -Math.log(t/kappa) inside a drawing loop.
4. The row contradicts its own cited evidence: row says kappa in {2.67, 4.0, 5.7, 8.0}; KAPPA_PRESETS
is {2.0, 2.67, 5.7, 6.0}. Two of four kappa values in the register do not exist in the code.
5. error < 1e-10 measures float round-trip precision of exp/log, not model-vs-measurement agreement.
The check could not fail.
6. The "certificate" is a React component that prints its own verdict ("This promotes Z-3 from
§B to §A"), and being a .tsx rather than a docs/research/*.json it EVADED the discharge-consistency
guard entirely (the guard skips rows citing no JSON certificate). Reported to shadow; covered by
the independent hardening in #9867 rather than duplicated here.
FABRICATED-CONSTANT SWEEP (non-demo half; demo/ is shadow's under the agreed scope split).
1/(3sqrt2) ~ 0.2357 is NOT a Tsirelson bound. Tsirelson's bound is S <= 2sqrt2 ~ 2.828 on the CHSH
CORRELATOR -- implemented correctly at src/Core/Tsirelson.fs (S^2 = 8, exact integer arithmetic) and
src/Core/BellTest.fs. There is no Tsirelson bound on a correlation COEFFICIENT, and quantum
correlations are not capped at 0.2357.
PROVENANCE (corrects the "relabeled DLA rendering parameter" framing): the constant did NOT originate
in the DLA renderer. It originates 2026-07-04/06 in YinYangEnsemble as rho_T = rho*/sqrt(2) -- the
Condorcet limit rho* = 1/3 pushed through the FREELY CHOSEN linear map rho = S/12 -- and that site
declared it honestly ("A DESIGN CHOICE, not a first-principles derivation"). It was adopted as the DLA
sticking probability 2026-07-16 (4b8dc2d) with the caveat dropped, then read as physics in the demo
site and Z-3/Z-5. The failure is not an invented number; it is that THE NAME TRAVELLED AND THE CAVEAT
DID NOT. Every definition site now carries the caveat.
Also found -- MoneyVelocityOracle.fs contradicted ITSELF: `tsirelsonAgeDays` docstring claimed it is
"the age at which the oracle enters the Classical regime" (1184 days, rho = 0.2357) while `moneyRegime`
in the same file puts that boundary at 1/(1+sqrt2) = sqrt2 - 1 ~ 0.414, i.e. ~517 days. Docstring
corrected; VALUE DELIBERATELY UNCHANGED (which threshold is wanted is a modeling call, left for a human --
nothing outside the file consumes it). Its regimeTable row also ASSERTED `SoundMoney` instead of calling
`moneyRegime`; now computed, so the row can disagree with the classifier.
QUARANTINE: the five discharge scripts + tests moved out of src/Core.TypeScript/verify/ to
docs/research/void-discharges-2026-08-01/ with every extension suffixed .void, plus a README stating
they are not evidence. They were ungated dead weight that read as proof -- and `bun test
src/Core.TypeScript/verify/` reported 13 pass / 0 fail, so the bare `bun test` in `test:typescript`
showed 13 green tests that looked like discharged conjectures. That is false-green CI. Now 0 tests
found there. Moved (git-tracked renames) rather than deleted so the register's "RETAINED FOR THE
RECORD" citations still resolve; register citations repointed.
§A #1-#28 IS CLEAN. Zero hits across the whole frozen core for: Tsirelson, 3sqrt2, 0.2357, any
discharge certificate, any z2/z4/z5/z6/z7 script, Z-2..Z-7, Z3DischargePanel, identity-dla, useDLA,
IdentityDLA, Loewner, SLE_, Halsey, FEP, YinYang, MoneyVelocity, Celegans, BusDelayTick, dla-meter.
Nearest adjacency, and it does NOT contaminate: §A #15 cites CondorcetBoundary.Tests.fs, which also
contains COND-10 and RHO-STAR-5 asserting rho_T. Those two assertions are TRUE arithmetic
((1/3)/sqrt2 = 1/(3sqrt2)) and merely misnamed; #15's claim rests on rho*(N) = (N-3)/(3(N-1)) from
docs/research/rhostar-analytic-proof.md and is independent of them. Both annotated.
Gates: lint-typescript.ts green; lint-discharge-certificate-consistency.ts green (0 rows);
dotnet build Core / Bayesian / Bayesian.Tests -c Release all 0 warnings 0 errors.
Agency-Signature-Version: 1
Agent: soraya
Agent-Runtime: Claude Code (local)
Agent-Model: claude-opus-5
Credential-Identity: acehack
Credential-Mode: human-local
Human-Review: EXPLICIT
Human-Review-Evidence: Aaron "route to any format checks Soraya needs to clean things up"
Action-Mode: interactive
Task: none
Co-authored-by: Otto <noreply@anthropic.com>
…/(3sqrt2) as a Tsirelson bound FACTORY-RESUME.md is read on wake and math-physics-connections.md is the source doc for Z-2..Z-6, so both were still teaching the fabricated reading. Banners added; prose retained unedited below them (history-preserving correction, not deletion). The remaining ~16 docs/ hits are ferries, letters, memory and research notes -- historical record, deliberately left intact. Agency-Signature-Version: 1 Agent: soraya Agent-Runtime: Claude Code (local) Agent-Model: claude-opus-5 Credential-Identity: acehack Credential-Mode: human-local Human-Review: EXPLICIT Human-Review-Evidence: Aaron "route to any format checks Soraya needs to clean things up" Action-Mode: interactive Task: none Co-authored-by: Otto <noreply@anthropic.com>
AceHack
enabled auto-merge (squash)
August 1, 2026 13:19
|
Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits. |
Contributor
Resume claim-level diff — reviewer attention requestedThis PR touches one or both of the factory's Base SHA:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Soraya's cleanup sweep. She corrects my framing on two counts — both corrections are better than what I had.
Z-3 verified broken on six counts
Beyond the tautology I found: no Loewner equation is integrated at all (the header claims it "runs the Loewner equation"; there's no ODE solver in the file), and the row contradicts its own code — it lists κ ∈ {2.67, 4.0, 5.7, 8.0} while
KAPPA_PRESETSis {2.0, 2.67, 5.7, 6.0}; two of four don't exist.error < 1e-10merely measuresexp/loground-trip precision. Demoted, prior text retained NOT IN FORCE.Provenance — my framing was backwards
The constant did not originate as a rendering parameter. It was born 2026-07-04 in
YinYangEnsemble.fsas ρ_T = ρ*/√2 via a declared map — with an honest caveat ("A DESIGN CHOICE, not a first-principles derivation"). 2026-07-16: adopted as DLA sticking probability, caveat dropped. 2026-07-31: read as physics. The name travelled and the caveat did not — a better diagnosis, and a more useful lesson than "fabrication".My convicting argument was also weak
I argued "the real crossing is ρ* = √2−1". Soraya: the repo already adjudicated that 0.2357 and 0.414 are different quantities, not a contradiction. The argument that actually convicts: Tsirelson's bound is S ≤ 2√2 on the CHSH correlator — there is no Tsirelson bound on a correlation coefficient at all. Both numbers are images of 2√2 under freely chosen maps.
A real defect, not just naming
MoneyVelocityOracle.fscontradicted itself:tsirelsonAgeDaysclaims the Classical boundary at 1184 days (ρ=0.2357) whilemoneyRegimein the same file puts it at √2−1 ≈ 517 days — 2.3× off. ItsregimeTablealso assertedSoundMoneyinstead of callingmoneyRegime— the same could-not-fail pattern. Now computed.False-green CI found and killed
bun test src/Core.TypeScript/verify/reported 13 pass — andtest:typescriptis a barebun test. Anyone running the suite saw 13 green tests reading as discharged conjectures. All nine files →docs/research/void-discharges-2026-08-01/with.voidsuffixes + an 82-line README. Verified: 0 tests found there now. Moved not deleted, so register citations still resolve.§A #1–#28: CLEAN ✅
Explicitly verified — zero hits for the constant, any certificate, any z-script, or any of Z-2…Z-7 across the whole §A span. The frozen core was never contaminated.
Gates: tsc ✓ · discharge-consistency ✓ (0 rows) · dotnet build Core/Bayesian/Bayesian.Tests 0 warn / 0 err. 🤖 Generated with Claude Code