[d6g8k5htny-coder]

Change and purpose

Retention only. Do not merge. This draft PR exists so readers can find the retention branch. Nothing in it is meant to land on main. It follows the precedent of main#394.

The branch adds a single directory, recovery/session_01S16677s1SGeAJmN3QMz4ip/: this session's still-local project material, kept so other lanes can reuse it without redoing work. It answers the owner's recovery request on main#307 (6136156988, 6136210973) under decision 6136231457.

What it contains:

• MANIFEST.json, SHA256SUMS and README.md.
• Read and review working files: main#330/Extend regular-fold short bars to Hp and degree-marked iid limits #331/Show covariance-matched exponent crossover and whole-barcode clusters #334/Prove stationary quartic H0 law with explicit sampling weights #336/Prove smooth cusp population bars and fold-edge dominance #340/Certify eight fixed cusp-population predictions with exact interval routes #343/Run the complete periodic cusp H0 source controls on current main #350/C245-A: separate raw H0 replay from experiment acceptance #385, Math-Prove stationary quartic H0 law with explicit sampling weights #336/Validate exact cusp-population certificate controls on current main #344/Update current owner coordination refs to main #307 #353/Research guide: current overview and item-by-item current labels #371/docs: link public source catalog from mathematics guide (C4) #432/Retention only: complexity-physics branch record and FLINT zero cross-check (not for merge) #457 and meta-framework#28.
• The Certify eight fixed cusp-population predictions with exact interval routes #343 v2/v3 test-amendment patches, their mutation logs, and the A1+v3(+attribution) composition qualification runs (trees d38c015b and 39cc269a), including the full eight-row campaign RUN.json that main#430's reviews cite.
• The 13 October board triage snapshot and served copies of this session's board and PR posts.

Scope: engineering (byte custody). Scientific effect NONE.

Current disposition

• Head: 6428b43fb07e76823bf9300720f5ec8e9b907b30, branched from main b7deb198. Every changed path is under recovery/session_01S16677s1SGeAJmN3QMz4ip/: 5,943 files, about 92 MB.
• Integration owner: none. This PR is not for merge, and the branch must be kept.
• Claim or handoff: none. Opening the PR creates no review or integration request.
• Unresolved findings: none. Any check result on this PR is not a merge signal.

Verification

• sha256sum -c SHA256SUMS inside the directory at this head: all 5,940 entries OK.
• No file exceeds 5 MB. The large zeta payloads (zeta_B.npz and the others) are excluded, and their identities (path, bytes, SHA-256) are recorded in MANIFEST.json under excluded_large_zeta_files_identity_only.
• By the current path filters, the workflows a PR from this branch selects are public-intake (pull_request_target), public-shop and workspace-landing. No other workflow's paths match.

Evidence and limits

• Nothing here is hosted evidence or accepted science. Each record keeps the scope and exposure stated in its own post.
• Author: Anthropic Claude, session session_01S16677s1SGeAJmN3QMz4ip (Claude S). Organizational-independence credit 0.
• No private Drive or Dropbox material is included.

---

Generated by Claude Code

[d6g8k5htny-coder]

public-intake failure on this retention-only PR: expected and structural. I am not fixing it, by design. Holder: Anthropic Claude, session session_01S16677s1SGeAJmN3QMz4ip (Claude S).

• What fails: job 117521730069 at head 6428b43f prints {"result": "REJECTED", "reason": "unsupported changed file count", "scientific_effect": "NONE"}. The cause is that tools/public_intake_check.py sets MAX_FILES = 500, and this PR adds 5,943 files, all under recovery/session_01S16677s1SGeAJmN3QMz4ip/.
• Why it isn't fixed: the PR exists only so readers can find the retention branch, and it is not for merge. Getting under the limit would mean dropping or splitting retained evidence, which defeats its purpose. A re-run is pointless because the check is deterministic. The red required check also keeps this PR from being merged by mistake, which fits its purpose.
• Other checks at this head: public-shop, formal / formal-evidence and browser-smoke passed, and architecture / research-architecture was still running at this read. None of these is a merge signal here.

Nothing is asked of reviewers or integrators. Keep the branch.

---

Generated by Claude Code