[amaybaum]

Draft — preparation only. Not authorized for merge or workflow dispatch (owner decision #918 c.6650823924, D2 and D12).

What this is

This PR holds one history-only merge commit, 4a90ac2f0ca535db71228ecea7a28e9b4da640a9:

• first parent: main f5c281ae10a913ec68fa41b3b839ad0b164979cd;
• second parent: ce978ce5db4165dd1e618e9f44bb67281ac152df, the head of closed PR Research: nonvacuous coherent permutation entropy profiles #989;
• tree: 236b68cf64510cc4b88eef003732b93d61f63e61, identical to the first parent's tree. The diff against main is empty.

It makes reachable from main the commits that the Introduction (#918) cites as [CE] (bdcbd32) and [CR] (2624d49), under AGENTS.md §A.42 items 2, 3 and 8. The owner accepted reachability from main as the durable home for this specifically reviewed evidence (D2), with the limitations below.

This is not a content integration. No file enters or leaves main's working tree.

Commits that become reachable from main

All four sit on top of 20ca05fe, which is already on main. Paths are under verification/audits/manuscript/page-coherent-recovery-20261028/.

Commit	Files it changes
a7cba26d3fa82372c5fd464c7def0cc326b783fc	adds README.md
bdcbd32bb55a5464d72b61682721f0b4bfaf446c	edits README.md; adds entropy-profile-proof.md, exact-results.json, exact_controls.py
2624d49cffc0794b300517719c879de6b57d395e	edits README.md; adds sequential-release.md, sequential-release-results.json, sequential_release_controls.py
ce978ce5db4165dd1e618e9f44bb67281ac152df	edits README.md; adds independent-core/manifest.json, independent-core/review.json, independent-core/review.md, independent-core/independent_controls.py, independent-core/independent-results.json, and .github/workflows/page-coherent-entropy.yml

Across the four commits, 13 files change, with +1170 lines in total. None of these files is in main's tree after the merge.

Retrieving the evidence

git fetch origin main            # a full clone; see "Limitations"
git show bdcbd32b:verification/audits/manuscript/page-coherent-recovery-20261028/entropy-profile-proof.md
git show 2624d49c:verification/audits/manuscript/page-coherent-recovery-20261028/sequential-release.md
git show ce978ce5:verification/audits/manuscript/page-coherent-recovery-20261028/independent-core/review.md
git ls-tree -r --name-only ce978ce5 -- verification/audits/manuscript/page-coherent-recovery-20261028/
# replay a control at its own commit:
git worktree add --detach /tmp/pcr ce978ce5 && (cd /tmp/pcr && python3 verification/audits/manuscript/page-coherent-recovery-20261028/exact_controls.py)

Ancestry check after landing: git merge-base --is-ancestor bdcbd32b origin/main && git merge-base --is-ancestor 2624d49c origin/main.

Limitations, stated explicitly

• GitHub browsing: the files do not appear in main's file tree. They are visible only by browsing the commits above.
• Default git log main -- <path>: it shows nothing for these paths, because history simplification follows the first parent, whose tree never had them. Use git log --full-history main -- <path>, or the commit ids directly.
• Shallow clones: a shallow clone of main cannot reach these commits. A full clone, or git fetch --unshallow, is needed.
• Research: nonvacuous coherent permutation entropy profiles #989: after landing, GitHub may show Research: nonvacuous coherent permutation entropy profiles #989's head as contained in main. That does not mean Research: nonvacuous coherent permutation entropy profiles #989's file changes were integrated. The merge took main's tree in full and discarded the second parent's changes, so Research: nonvacuous coherent permutation entropy profiles #989's workflow .github/workflows/page-coherent-entropy.yml is not on main and never runs.
• b0affb21: cited in independent-core/review.md:7, it is not an object in this repository. Under §A.42 item 8 it is provenance, not support. It becomes neither reachable nor resolvable through this merge.
• [CR]: its status is unchanged: proved on paper and not independently reviewed. Its independent review remains with Research backlog: advance downgraded and conditional scientific claims from corpus audit #970. Retention strengthens no claim (§A.42 item 6).

Verification

Done locally:

• the merge tree equals main's tree (236b68cf); git diff f5c281ae 4a90ac2f is empty;
• the second parent's ancestry is exactly the four commits above, on top of 20ca05fe.

Simulated earlier by the independent plan review (#918 c.6650329205), on full-history clones:

• no path-filtered workflow matches;
• legacy_records_check, citation_roles_check and v3_verifier --receipts give results identical to main (52 receipts hold);
• the [CE]/[CR] controls replay as PASS.

Pending:

• independent review of this exact commit, repeating those checks on 4a90ac2f;
• ordinary draft-PR CI (a synthetic-merge run, not an exact-head attestation);
• exact-head verify via workflow_dispatch on 4a90ac2f, which needs owner approval of the head SHA and inputs;
• owner merge approval naming the exact SHA. It must land as a merge commit: squash or rebase would discard the second parent and defeat the purpose.

Links

• Plan: Introduction paper: reconcile sources and integrate scoped theorem recoveries #918 c.6650052943
• Plan review: Introduction paper: reconcile sources and integrate scoped theorem recoveries #918 c.6650329205
• Owner decisions: Introduction paper: reconcile sources and integrate scoped theorem recoveries #918 c.6650823924
• Planning: Introduction full-corpus scientific consistency audit (gate for #918) #941. Evidence source: Research: nonvacuous coherent permutation entropy profiles #989 (closed, branch kept). [CR] review: Research backlog: advance downgraded and conditional scientific claims from corpus audit #970.

---

Generated by Claude Code

[amaybaum]

Independent review of exact candidate 4a90ac2f0ca535db71228ecea7a28e9b4da640a9

Verdict: ACCEPT. This is a read-only review by a reviewer that did not write the candidate. It used its own full-history clone (git clone --no-local, not shallow, no alternates) and stated exit codes and structured CI records. There are no blocking findings. The four notes at the end are for the owner's dispatch and merge decisions.

1. Tree identity and parents

• The parents are exactly f5c281ae10a913ec68fa41b3b839ad0b164979cd (= origin/main) and ce978ce5db4165dd1e618e9f44bb67281ac152df.
• 4a90ac2f^{tree} = f5c281ae^{tree} = 236b68cf64510cc4b88eef003732b93d61f63e61.
• git diff f5c281ae 4a90ac2f is empty.
• No path under page-coherent-recovery-20261028/ is in the merged tree (0 hits), and neither is .github/workflows/page-coherent-entropy.yml.

2. Reachability

• git rev-list f5c281ae..4a90ac2f lists exactly 4a90ac2f, ce978ce5, 2624d49c, bdcbd32b, a7cba26d.
• 20ca05fe is an ancestor of f5c281ae.
• Each commit's diff-tree --name-status matches the PR body's table exactly.
• Totals for 20ca05fe..ce978ce5: 13 files changed, +1170 lines, nothing deleted.
• Before the merge none of the four commits is an ancestor of main; on a local branch at 4a90ac2f, all four are.

3. Integration consequences (full history; exit codes)

Each tool was run with the HEAD detached at the commit under check. "Post" is a simulation: refs/heads/main and refs/remotes/origin/main were moved to 4a90ac2f and then restored.

Check	f5c281ae	4a90ac2f	post-merge sim.
legacy_records_check.py <c>	0: 303 records, all intact	0: identical	0: identical
citation_roles_check.py <c>	0: 0 manifests	0: identical	0: identical
v3_verifier.py --receipts <c>	0: 52 receipts hold	0: identical	0: identical
SC-42 check_sc42_integration.py (bell-operational-reconciliation-20261029)	0	0: output byte-identical	n/a

• Self-tests at 4a90ac2f: legacy_records_check --self-test exits 0, and v3_verifier --self-test exits 0.
• tools/release_gate.py at 4a90ac2f: every step passes except lean-axioms. That step fails only because no local Lean/lake toolchain is installed. The CI release-gate step is green (below).
• History simplification behaves as expected for a TREESAME first parent. The two rev-list -n 1 <c> -- <path> lookups (legacy records, receipts) follow the first parent, so their results do not change.

4. Workflows

• verify.yml: pull_request (no filter), push: [main] and workflow_dispatch. It is the only workflow that fires.
• bg2-orientation-design.yml and bridge-compatibility-research.yml: both have path-filtered push/PR triggers on verification/programmes/oi-qm/reconstruction/... and their own files. None of the five commits touches those paths, under any diff basis.
• hidden-sector-replay.yml: main only, and its paths are not touched.
• evidence-replay.yml, oret-phasea.yml, pe01-bounded-tests.yml and project-board.yml: dispatch only.
• page-coherent-entropy.yml: absent from the merge tree, so it cannot run on the PR merge ref or on main.
• Observed: the only run on this branch is verify run 41976644832 (pull_request, head_sha 4a90ac2f). Its conclusion is success. Of its 26 jobs, 25 succeeded and one was skipped:
  • The skipped job is A42 exclusion, whose if: requires workflow_dispatch.
  • In Mathlib bridge, the Release gate step succeeded. The KTRANS-SEP-2 marker step was skipped by its condition, which is the expected state.
  • The head's check-runs list only verify jobs; the three required contexts are all green.
• This is a synthetic-merge run, not an exact-head attestation.

5. [CE]/[CR] control replays (each in its own worktree at its own commit)

Commit	Control	Exit	Status	vs committed JSON
bdcbd32b	exact_controls.py	0	PASS	equal to exact-results.json
2624d49c	sequential_release_controls.py	0	PASS	equal to sequential-release-results.json
ce978ce5	independent-core/independent_controls.py	0	PASS	equal; the rewritten independent-results.json is byte-identical (clean git status)
ce978ce5	exact_controls.py and sequential_release_controls.py	0, 0	PASS	equal

The [CE]/[CR] files are unchanged from bdcbd32b/2624d49c to ce978ce5 (empty diff --stat).

6. PR body against the D2 safeguards (#918 c.6650823924)

• All four commits and their files are named: correct.
• "No file enters or leaves main's working tree": correct.
• Retrieval commands: tested. All three git show commands resolve, returning 129, 48 and 69 lines. The ls-tree lists the 12 files, and the worktree replay runs PASS.
• Default git log limitation: confirmed on a local branch at 4a90ac2f. Default git log <br> -- <dir> returns 0 commits. --full-history returns 4a90ac2f, ce978ce5, 2624d49c, bdcbd32b and a7cba26d.
• Shallow-clone limitation: confirmed. A --depth 1 clone of that branch cannot resolve bdcbd32b.
• Research: nonvacuous coherent permutation entropy profiles #989 explanation: correct. The workflow is absent from the tree and did not run.
• b0affb21: the id is not an object here (fatal: Not a valid object name). In review.md:7 it is the reviewer's local HEAD with an identical tree; the review pins bdcbd32b. Provenance is the right classification.
• Other ids cited by the retained files: every other hex id resolves. These are 20ca05fe, tree 700fa72b, and blobs 16bdf99a, 908608ae, a1888fbb, ca0d37ad and dc6c3483. All of them are reachable from 4a90ac2f; ca0d37ad is already reachable from main. The 0510232 and 9305007 hits are arXiv numbers.
• [CR] status: stated as unchanged, and retention claims nothing stronger (§A.42 item 6). Correct.

7. Notes for the owner (non-blocking)

1. The landed head will not be 4a90ac2f.
   • The ruleset requires a PR, so the GitHub merge button creates a new merge commit with first parent main and second parent 4a90ac2f. Its tree is still main's tree. The body's post-landing merge-base --is-ancestor check remains the right test.
   • The ruleset has strict_required_status_checks_policy: true. If main moves before the merge, the branch must be updated. That changes the head away from the reviewed SHA, which then needs re-review and a new exact-head run.
2. Merge method. The ruleset's allowed_merge_methods is merge, squash, rebase. Choose "Create a merge commit" explicitly; squash or rebase would land nothing, because the tree diff is empty and the second parent would be dropped.
3. Replay recipe wording. The body's "replay a control at its own commit" runs exact_controls.py at ce978ce5, not at bdcbd32b. The file is byte-identical, so the result is the same. Optional improvement: also compare the output with exact-results.json.
4. A dispatched verify run is larger than the PR run. It additionally runs the dispatch-only A42 exclusion matrix (15 shards).

Exact-head dispatch to approve

verify.yml's workflow_dispatch declares no inputs. The run would be dispatched as follows:

• gh workflow run verify.yml --repo amaybaum/incompleteness --ref land/918-b1-evidence-ancestry-20261105
• The branch must be held at exactly 4a90ac2f0ca535db71228ecea7a28e9b4da640a9.
• The attestation is the run with event == workflow_dispatch and head_sha == 4a90ac2f0ca535db71228ecea7a28e9b4da640a9, read from its structured job and step conclusions.

The reviewer's clone was independent, and it is deleted after this review. It made no commits, pushes or dispatches.

---

Generated by Claude Code

[amaybaum]

Exact-head attestation for 4a90ac2f0ca535db71228ecea7a28e9b4da640a9 (dispatch approved in #918 c.6651201185).

Run: verify run 41979200273 (run #3514).

• Event workflow_dispatch, ref land/918-b1-evidence-ancestry-20261105, head_sha 4a90ac2f0ca535db71228ecea7a28e9b4da640a9, no inputs.
• Conclusion success.

Structured job results: 40 of 40 jobs succeeded and none failed or was skipped. They include:

• all 17 jobs of the dispatch-only A42 exclusion matrix;
• the Mathlib bridge job, whose Build and Release gate steps succeeded.

State re-read after the run (via git ls-remote and the PR API):

• The branch is still at 4a90ac2f, and main is still at f5c281ae10a913ec68fa41b3b839ad0b164979cd.
• The PR is mergeable and clean.
• The tree of 4a90ac2f equals main's tree (236b68cf), and git diff f5c281ae 4a90ac2f is empty.
• a7cba26d, bdcbd32b, 2624d49c and ce978ce5 are all in the head's ancestry, and none of them is on main yet.

Exact-head review: ACCEPT (c.6650995543).

Merge: not authorized yet. It awaits separate owner authorization naming this SHA. It must use Create a merge commit.

---

Generated by Claude Code

[amaybaum]

Merge record. Merged under the owner's approval for exact head 4a90ac2f0ca535db71228ecea7a28e9b4da640a9, expected main f5c281ae, as a normal merge commit. Both SHAs, mergeability and tree identity were rechecked immediately before the merge.

• Merge commit: 53a3eb3d1485601982b28c11def5e7b2a0db21b9
• Parents: f5c281ae10a913ec68fa41b3b839ad0b164979cd (first) and 4a90ac2f0ca535db71228ecea7a28e9b4da640a9 (second).
• Tree: 236b68cf64510cc4b88eef003732b93d61f63e61, identical to the pre-merge main tree, so no file in main's working tree changed.
• Ancestry (git merge-base --is-ancestor <c> 53a3eb3d): all of these are reachable from main:
  • the evidence commits a7cba26d, bdcbd32b, 2624d49c and ce978ce5;
  • the head 4a90ac2f.
• Legacy records: 303 intact.
• Post-merge CI: verify push run 41982684989 on 53a3eb3d completed successfully, with 25 jobs passing and 1 skipped (the matrix placeholder). The jobs API shows Mathlib bridge Build and Release gate both succeeded.

This is a history-only merge. It makes the evidence reachable by ancestry and does not integrate any content, as the PR body documents. The #918 landing work now proceeds under D12, on current main. It has no dispatch or merge authorization.

---

Generated by Claude Code