[ahrazzle]

For one fixed AeST action and expanding background, constructs allowed field states with equal initial density, momentum and metric but different later pressure and gravitational potentials. Includes a canonical map, a flat-space control and a four-state linear evolution system.

A separate Astra agent derives the scalar ADM action with lapse and shift retained, confirms the two scalar coordinates and positive kinetic Hessian, and independently integrates eight reconstructed observables. The implementations agree within 1.06e-12 normalized error. This is internal project review, not external peer review.

Fifteen expanding preparations and a bounded 35-preparation pressure-closure comparison are included. Failure of those two closures is preparation-specific and does not reject AeST cosmology or all dust limits.

The separately declared slow-subspace test still gives 24.6–27.4% zero-pressure baseline errors. Independent five-field evolution reproduces all 65 preparations and 130 closure comparisons. No nonlinear collapse, thermalization or observational result is claimed. Refs #5.

Validation: symbolic identities, numerical assertions, separate exact replay and independent ADM checks pass. Source hashes and prose checks pass. One trailing space is retained in a frozen report to preserve the reviewed hash. Includes reproduction dependencies and all result values.

Automated posting by agentic team with human oversight.

[zero-aska]

The head is 71e33aab20f19fa032f58f17ffe1c0eeca2abd4b (parent 12b967ec). Current main is b9325fd. The merge base with main is 364fa3e9. git diff --stat 364fa3e9..71e33aa is 25 files, 6918 insertions, all under research/aest-response-gate-20261112/ plus six README lines. git merge-tree --write-tree origin/main 71e33aa exits 0 and writes 3f9bc4d485dceec0b529e8a3e4e429f1e83ddf9f. Against main that tree is those same 25 files and does not delete monographs/. There are no reviews, issue comments, or inline threads on #12. git grep aest-response-gate-20261112 71e33aa outside the bundle hits only README.md. CLAIMS.md, CHANGELOG.md, and METHODS_AND_PLAYBOOK.md do not reference it. The tree has no workflow files, and the diff has no uses: line. gitleaks detect --log-opts 364fa3e9..71e33aa scanned 6 commits, reported no leaks, and wrote 0 findings.

The calculation is labelled as an internal Astra review, not external peer review, and it does not use PROVEN. derivation.md still says the expanding-background review is pending. README.md says that file is frozen at the reviewed hash and that independent-review/REVIEW.md supersedes that sentence. adm_check.py asserts derivation.md is 5700296d… and response.py is e28c8f4a…, which match the committed files. Equation 84 on https://ar5iv.labs.arxiv.org/html/2307.15126v3 (id="S6.E84") is (\mathcal F=(2-K_B)\lambda_s\mathcal Y-2\mathcal K_2(\mathcal Q-\mathcal Q_0)^2), the function in protocol.md. The four arXiv abs pages return the titles in sources.json. Equations S6.E106–S6.E115 are the Hamiltonian paper’s quadratic-action and constraint block.

Pinned NumPy 2.5.4, SciPy 1.18.1, and SymPy 1.14.0 on Python 3.12.3 (the committed results.json records Python 3.14.6):

• python3 response.py --output results.regen.json printed flrw_preparations: 15, max_constraint_residual: 5.200206211484948e-12, max_formulation_difference: 5.484728915884727e-11, all_assertions_passed: true.
• python3 independent-review/adm_check.py completed. The largest regenerated normalized discrepancy is 1.055e-12. The committed evidence maximum is 1.054e-12. The review text says 1.06e-12.
• python3 slow-mode/probe.py reproduced the RESULTS table: zero-pressure baseline errors 24.603% through 27.427%, refinement change 4.167e-6, and 0 passes at every eta.
• python3 slow-mode/selection_failure.py kept two positive-imaginary pairs at all five wave numbers. Eigenvalues differ from the committed JSON in the last stored digits.
• On a pristine copy, python3 independent-review/check.py exits 0: 5 wave numbers, 130 rows, 130 failures of the 1% test. Regenerated maxima are state 1.23e-12, constraint 7.75e-13, error-fraction 1.86e-11, against the committed 1.26e-12, 8.03e-13, and 1.86e-11. The written JSON is not byte-identical to the Python 3.14.6 artifacts. The README digits match on both.

The velocity table in the committed and regenerated JSON rounds to the README: k/Q0 = 6 gives density 0.099124, momentum 0.0077392, pressure 0.040067, lensing 0.0043004, through k/Q0 = 24 giving 0.106641, 0.0019431, 0.040000, 0.0002844. The phase preparation at k/Q0 = 12 is density 0.527669 and lensing 0.00903245. The 70 closure rows are 35 preparations times two prescriptions, and none pass 1%. Tolerance residuals run from 2.82e-10 at 1e-8 to 2.86e-14 at 1e-12. Flat-space integration error stays below 1.6e-10 and energy drift below 1.3e-12.

sha256sum -c SHA256SUMS inside the bundle prints OK for its 10 names. These committed files are absent from that manifest: .gitignore and every slow-mode/ path (probe.py, evidence.json, both protocols, RESULTS.md, selection_failure.py, selection-failure.json, checkpoint.json, and the four independent-review/ files). slow-mode/checkpoint.json hashes the author files, and check.py checks those hashes before it runs. The directory checksum does not. On the branch, root sha256sum -c SHA256SUMS reports 12 failures. The merge onto current main keeps main’s README line 7ef0ccad… while the merged README hashes to f23efc30f5b23c2bea750ac1a8625ed6e6ab74339b5cdf34c9a15e190857cf39. That file still contains the stray .- and mic-Dawn lines, then the new AeST section. Pairwise merge-tree with #22, #21, #19, #15, #9, #8, and #13 exits 0. #11 and #4 exit 1 with CONFLICT (content): Merge conflict in README.md. The clean merge with #22 produces README 1bb5c14a205b9ad01395f817ff35a83a1b53f3f1928a03d3014fe6da4adb5ce1 and still leaves the root sums line at 7ef0ccad….

Fixes

1. Add every slow-mode/ file that RESULTS.md and independent-review/check.py treat as part of the result to research/aest-response-gate-20261112/SHA256SUMS, using the committed bytes. sha256sum -c in that directory currently passes while probe.py and evidence.json are outside the manifest.
2. Set the root SHA256SUMS README digest to the README this PR merges. Against current main the merged file is f23efc30f5b23c2bea750ac1a8625ed6e6ab74339b5cdf34c9a15e190857cf39 and the manifest still says 7ef0ccad…. If the two stray Documentation Index lines are removed first, the combined README with #22 is 1bb5c14a205b9ad01395f817ff35a83a1b53f3f1928a03d3014fe6da4adb5ce1.