[1337isnot1337]

Summary

This adds an exact local optimization for the selected 112-port paired-plane factorization. Replacing predecessor matching edges (90,42) and (87,44) with (87,90) and (42,44) lowers its exact B14 support-pairing invoice from 126 to 96.

The package enumerates the full two-edge-switch neighborhood: 3,080 reconnection descriptions, 666 legal matchings, no inconsistent factorization systems, and a unique minimum at 96. It reconstructs the F2 identity, exact support ranks, and all 28,672 dirty-basis plane-endpoint states. Its selected-complex relaxation is explicitly screening evidence only; the F2 factorization is not a scalar masked-plane decoder, and no supplier, assembly, or selected-kappa improvement is claimed.

Reproduction

python3.12 -B research/paired-plane-local-swap-seed9c65a5/verify.py

Expected first line: PASS_PAIRED_PLANE_LOCAL_TWO_EDGE_SWITCH.

The verifier and scripts/research_package.py --check-only passed locally. The direct verifier ran under Python 3.14.7; Python 3.12 is unavailable in this environment, so the registered Python 3.12 replay remains for CI.

Predecessor attribution and remaining questions are recorded in the package README and TASK metadata. Substantial OpenAI Codex assistance was used; no independent human review or formal proof assistant is claimed.

[Mister-Meeseeks]

Thanks for the contribution. Review of 5d324bba9ae28113f3a92ba45af1361c3e742634 in the frozen 100-PR round found one remaining issue:

The local switch replay passes, but SOURCES marks the historical task catalogue and priority log as live inputs. Both pins differ on current main; bind the historical guidance to its Git snapshot instead of making routine priority updates invalidate this package.

The scoped replay evidence is retained in the batch review. Passing finite checks and the scope of the written claim are tracked separately. This PR is staying open for the focused revision; contributor and predecessor credit are retained.