Skip to content

Conditional κ = 0.0036504442446 (+>2.219087727e-7 over #7604) - #7605

Open
aliromman wants to merge 3 commits into
CrocSwap:mainfrom
aliromman:codex/p104-kappa-20261114
Open

aliromman wants to merge 3 commits into
CrocSwap:mainfrom
aliromman:codex/p104-kappa-20261114

Conversation

@aliromman

@aliromman aliromman commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

Helper 757 receives the same two updates as helper 1255 at different points in the native word. Subtract 1255 before those updates and add it back at the later common chart. This replaces 757's two rank-one MOVEs by one rank-two MOVE, preserves all arbitrary-input outputs and erased rows, and keeps the scalar-operation count unchanged.

The complete 17-call bill changes by ΔP₁ = −34, ΔP₂ = +17. Paid positions fall from 110,383 to 110,366; all 22,461 productive positions and rank surplus 1,424 remain. The exact moment gain is 17*(2−2^(1−C)) > 0 for 0<C<1.

Conditional κ = 0.0036504442446, 2.219087727e−7 above #7604's certified raw upper 0.0036502223358273, hence more than that above its actual crossing (about 0.00608%). Two independent directed engines enclose the new crossing in (0.0036504442447168877, 0.0036504442447168879). A negative derivative bound proves the fixed-invoice ceiling on the inherited BIT interval; no optimum over changed words is claimed.

Verification covers all 1,569 final and 74 erased rows, COPY/ownership controls, 512 forward and 512 adjoint columns, 38 affected lineage pairs, 3,040 catalogue maps, 104 boundaries, 13,600 native contacts and all 54 strict plus four functional outer predicates. The full ordinary bound is 782699212 < 10^9; precision remains 2430; all tape costs are recomputed. The full framework supports the stronger claim 0.003650444244716.

Reproduce with Python 3.12 and the standard library:

python3.12 -B scripts/research_package.py --package research/p104-timed-helper757

The package includes the exact source closure, explicit proof, rejecting controls and counted application adaptations. All six registered commands and a read-only, Git-free export replay passed. Original Git readback covers all 109 retained files; committed-byte and mode checks pass. VALIDATION.md records exact coverage, the source freeze and the unavailable optional full-history check in this shallow checkout. Hosted CI at e1d2756eecf03e146338239fee81772a99ba8d4c finished with the inherited navigation failure: run 43632689871, 735 passing tests, 24 skips and one failure in test_current_ci_section_matches_selected_record. Unchanged docs/ci-verification.md names frontier6030 while the selected record requires frontier7000. Package selection and hosted research replay were skipped. The failing test, CI guide and selected record are byte-identical to canonical main. PR7276 addresses that separate issue; no unrelated fix is included. The final documentation-only commit corrects the CI guide filename in the package note. Its hosted run has the same shared failure. The six local registered commands and the read-only Git-free replay passed.

Credit Hao Yan/#7604, SYLforge/#7603 for the stock criterion, DreamingOfClouds/#7601, ToikY/#7546 and #7544, Rohan Arun/#7113, Clear1nsight, dillona, CubeJerry and every retained author. Prepared by Ali Romman with substantial OpenAI Codex assistance. Canonical investigation and overlap record.

The same five assumptions remain: native_all_columns, owned_erasure, compiler, precision_tape, analytic_recovery. No full materialized multiplier, independent human or formal-kernel acceptance, selected-frontier change or merge is claimed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant