DONE: selected k8 helper endpoint rigidity and admission audit (PR #1583)

[Arham-Tahir64]

DONE — PR #1583, exact head 04516881142961cdf97a828fbbc63d72e232a148.

All 480 selected k8 helper endpoints are minimal with unchanged gates, entrances, gate frames and cleanup: 78 restricted ends equal their last native frame, while 402 helpers occur in full-frame cleanup. A span argument excludes all 105,927,652 different single-end same-rank caps containing the original entrances; this is a counted exclusion, not a 105-million-case search. Every one of 78 executed same-rank mutants passes both exact scalar endpoints and has identical 480 owner widths, then fails local frame monotonicity. The frozen word payload pin also rejects every mutant. This corroborates existing local admission and reports no selected acceptance bug.

Positive retained replay covers 704 rational scalar columns in both orientations and all 69,120 owner roles/144 calls. Python 3.12.14 registered replay and six tests pass; Python 3.14.7 optimized replay from a read-only minimal relocated tree with spaces produces the identical receipt. Source corruption, small subspace/quotient counts and selected integrity pass.

Scope: fixed-schedule helper endpoint subspaces only; no enlarged-endpoint pricing, changed-tail synthesis, full selected cold replay, native phase/compiler or physical admission, new kappa or all-size theorem. New code independently records last-touch/full-cleanup spans; the existing exact scalar/frame and owner auditors are openly reused comparison targets. Last-touch witnesses suffice, so new inverse-cone derivations were not needed. Source pins, proof, complete receipt, credit and next question are in research/k8-helper-endpoint-audit/.

Credits Rohan Arun PR1094; Jeremy Kothe and retained gauge44 lineage; Josie Elliston PR1091; Sagar Thakrar PR1049; Douglas Colkitt's composition. PR892 has an earlier different helper endpoint floor; neighboring PR1358/1312/1238 scopes remain distinct. Prepared for Arham Tahir with substantial OpenAI Codex assistance; no independent human-review or hosted-CI-success claim. Claim released.

---

Original claim (preserved):

CLAIM — selected k8 helper endpoint admission audit

Question: For the PR1094 selected k8 word, are all 78 explicit restricted helper final frames forced when the gate list, entrance/gate frames and early/full cleanup schedule stay fixed? Can same-rank endpoint replacements preserve the residual width inventory used by owner admission while failing local semantics?

Base: bce54fb; selected PR1094 audited head 0fc329d. Word payload c2b294300760e1fd6eeab8c22c4d13e4633dca40e52cac8da0d68163a76e3ca8, bound to the k8 recipe and PR1091 input.

Method/domain: independent label-span propagation and exact rational dependency cones, 704 registers/480 helpers/78 explicit low ends; construct same-dimension invalid endpoint fixtures and execute retained local checks. All scalar gates, order and cleanup fixed. No full cold replay or new kappa claim.

Distinct from PR1358 (one B-gate superspace, endpoints fixed), PR1312 (X2 ambient owner-axis disjointness), PR1238 (source/target end caps). Credits Rohan Arun PR1094, Jeremy Kothe k8/leaver lineage, Josie Elliston PR1091, Sagar Thakrar PR1049 and retained gauge44 suppliers.

Expected artifact: research/k8-helper-endpoint-audit, standard-library replay plus exact lower spans, adversarial controls, source pins and scoped proof. Assignment nonce 1f08d36879e997ec973cb952b5221865, supplier-boundary-audit. Prepared for Arham Tahir with substantial OpenAI Codex assistance.

Claim expires 2026-10-14 01:50 UTC.