[Clear1nsight]

Upstream publication and complete scope

Publishes fork PR #16, preserving exact verified head 0898b3a8faa5b809d2beedc99dce5d69574568e4. Publication comparison was refreshed against upstream f626db5c3a3d781d8f7695038776c621651f75e1 on 22 October 2026: 16 changed files, 3 contribution commits. The scientific package is isolated in its own directory, but the complete upstream diff also carries the shared historical-replay/source-role and navigation/CI repairs from its preparation base, plus the base documentation. Shared repair bytes overlap across fork #14–#18 and must be reconciled once; they are not independent scientific contributions. All five research heads have successful exact-head fork workflows. Opening this PR is not a new scientific replay, upstream CI pass, merge or theorem acceptance. Scientific source pins, attribution and scope limitations are retained.

---

Change and mathematical scope

Assignment D independently audits one quantified implication in zxv's PR4483 Section 4. The all-real proof corroborates the cancellation threshold and characterises its sharp universal guarantee at fixed net production and leaf mass. It handles zero mass without division and gives explicit rejected root/leaf omission examples. These are failures of weakened tests, not bugs in PR4483. No new kappa, compiler rank or full compiler admission is claimed.

For I=B+F+A, 0<=beta<eta<1 and nonnegative potentials with B<=beta I, the exact identity is

(1-beta)(eta I-B-F)
 = (eta-beta)A-(1-eta)F+(1-eta)(beta I-B).

The largest allowed exit fraction at fixed A,F is (beta*A+F)/(A+F), attained by the stated extremiser. Necessity concerns a guarantee over all admissible boundary masses, not each individual flow. PROOF.md binds A to the source's net production including its root debit.

Task and overlap

Canonical base f9f85c1e4fa1e955ff39f10fca447e3e2a4e4adc, priority revision 7, unchanged at the final refresh. Source PROOF.md is pinned by Git blob 75f451837204bd141f357e7f5e5d6c6f42e1423e and SHA256. TASK.json preserves the original card and declares the user-assigned changed proof-audit scope.

PR4529 covers finite gates, PR4532 broader gate existence, and PR4533 retuned witnesses. The final refresh also found PR4535's critical-root conditional transfer and PR4536/4537's arithmetic checks. Those do not substitute for the scoped logical argument here, and this package does not claim to audit their new critical-endpoint theorem. Targeted searches and bounded Discussion reading are not exhaustive overlap pagination. The historical baseline helper was not run.

The scientific contribution remains pinned to its isolated preparation base. This upstream submission preserves the exact verified head; no merge or selected-frontier change is requested.

Validation

Final-head hosted workflow 39724661544 passes at 0898b3a8faa5b809d2beedc99dce5d69574568e4. Onboarding, package metadata/source authentication, source-bound scientific replay, optimised controls and committed-input preservation pass on Ubuntu with CPython 3.12.15, standard library only.

RESULT.json retains 6,120 rational cases, 1,800 attained extremisers, 120 zero-mass cases, four invalid-premise controls, nine coefficient checks and both omission witnesses. Normal replay compares every retained result field. The universal conclusion comes from the written identity and extremal argument, not extrapolation of the grid. No predecessor acceptance function is imported.

python -B scripts/research_package.py --package research/adaptive-cancellation-audit
python -I -O -B research/adaptive-cancellation-audit/verify.py --self-test

Chernoff/root estimates, mirror certification, funded majorant, chronological native realisation, physical costs, precision/stock and outer transfer remain separate. No full selected cold replay, optional source-history preflight, formal-kernel or independent human verification was run.

Credit

Credit zxv for the cancellation argument already used by PR4483, Jacob Alan Sussman's compiler, Ryanson Jonathan's mirror/gate work, Dillon's broader gate-existence follow-up and the full retained Cinna Davis/native/outer lineage. The elementary extremal calculation carries no priority claim. Prepared for Chris R / Clear1nsight with substantial OpenAI assistance, Apache-2.0. Separate implementations in one session do not imply separately launched agents.

Open next questions

A better contraction must use stronger actual-flow information than these scalar premises, or a better mirror/leaf bound. First prove one source-bound joint occupation inequality and compare it with the extremisers. A separate audit of a Chernoff/root estimate or PR4532's critical-endpoint argument remains distinct. Refresh overlap before claiming it. README.md, PROOF.md, TASK.json and the public handoff record the exact boundaries.