Research consolidation: restore reviewed PR236 source package

[williamreidrichards]

The published PR236 maintainer review retains its complete execution receipts, but main lacks the standalone source package needed to reproduce its input checks. This restores all 14 original files byte-for-byte and adds an offline source/receipt reconciliation. Research consolidation; no new κ claim.

The checker binds 3,606 Lean modules, eight scientific inputs, six principal theorem copies, 3,606 receipts and 3,606 logs. It reconstructs all 14,320 audited declarations and 7,253 dependency-object hash links, checks the recorded Darwin adapter and rebuilt/reused coverage, and rejects 17 semantic mutations. Twelve additional tests cover source/archive boundaries. The reconciliation found no inconsistency in the original sources or receipts; this closes the source-availability gap identified in the review ledger.

python3 -I -B research/pr236-source-reconciliation/verify.py

Expected: PASS_PR236_SOURCE_RECEIPT_RECONCILIATION 3606 modules, 14320 audits, 8 scientific inputs, 17 rejected controls.

Exact base: 66849e1475c930fd7732b1a0067ffee7e1e95f60. Restored PR236 head: f825b6fdfd763f189db3aa1c02fdd5d604c85a01. Scientific target: d1d6c070f5a8c684727ee7ec35d930f9ebfa9758. SOURCE-PINS.json records all Git blobs and SHA256 pins; the README explains scope and reproduction.

Validation passed on macOS arm64: Python 3.14.6/3.13.11, repeated ordinary runs, optimized Python, a committed-files-only export from an unrelated working directory, all 17 receipt controls, 12 new tests, 11 original driver tests, the original complete input-only check, selected-record verification and its five tests, the existing 528-PR reconciliation checker, and git diff --check. The export remains byte-identical after execution. No inherited checker or generator changed; full make verify and the selected cold-supplier replay were not rerun. Hosted CI is separate.

This checks preserved execution evidence, not a fresh Lean build. The historical run rebuilt 3,602 modules and reused four under the uncapped Darwin adapter. Compiled objects, compiler/standard-library binaries and the original acceptance manifest are absent from the published archives; completeness is checked against the public input map. All inherited mathematical and realization obligations remain. The selected record, existing PR dispositions and upstream/ are unchanged; this does not close PR236.

OS-random seed 98, research consolidation. No pinned coordination issue existed, so the claim is local. Overlap was reviewed across 539 initial open PRs, 551 through #732, and finally 562 through #743 immediately before submission. #695, #704 and #721 concern different evidence or ledger checks; #740 concerns the archived asynchronous inventory and #743 a stage-count bound.

Credit James Chang (jamesyc) for PR236, Douglas R. Colkitt for the retained maintainer review and receipts, and the original paired-cube contributors. Prepared by Will Richards (williamreidrichards) with substantial OpenAI Codex assistance, including a separate agent's audit and code review. Original files, attribution and disclosures are retained under Apache-2.0.

[Mister-Meeseeks]

100-PR batch review at ee3580bd708d82b5bdfda6a2bc3f0a34710be50e.

This remains open after scoped triage. Next review step: Restored source/Lean archive reconciliation requires coverage-aware archive review.

This is a specific continuation point, not a rejection or a request to redo already recorded checks. Completed local evidence is retained separately from outstanding whole-head obligations.

Batch review, attribution and evidence. Thank you for the contribution.