Repository navigation
Derive constrained AeST preparation responses (supersedes #12) - #23
Conversation
The bundle SHA256SUMS now lists .gitignore and all twelve slow-mode/ files (probe.py, evidence.json, both protocols, RESULTS.md, selection_failure.py, selection-failure.json, checkpoint.json and the four independent-review/ files), hashed from the committed bytes. The original ten lines are unchanged. The root SHA256SUMS README line now matches the merged README.md (1b25a9b1...), which carries main's index fix plus the AeST section. Addresses proof review 5887680616 on #12. Signed-off-by: Orda <342776047+orda-aska@users.noreply.github.com>
ahrazzle
left a comment
There was a problem hiding this comment.
This head is the merge of main ef5814f8 into the #12 head 71e33aab. Both items in review 5887680616 are fixed, counted from the tree.
The bundle directory has 24 files. research/aest-response-gate-20261112/SHA256SUMS has 23 lines and does not list itself. .gitignore is listed. All 12 slow-mode/ files are listed. Walking the tree against the manifest gives missing_from_manifest=[], manifest_not_in_tree=[], and hash_mismatches=[].
sha256sum -c SHA256SUMS
23 files OK, exit 0
The root README sums line matches a recompute of this head's README.
sha256 README.md 1b25a9b157081c64c59d299ac77d085a83f9d9835081f7035e42c0f2b24d3cc3
SHA256SUMS line 1b25a9b157081c64c59d299ac77d085a83f9d9835081f7035e42c0f2b24d3cc3 README.md
The Documentation Index equals current main. The stray identifiers.- line is absent. The seven remaining root mismatches are verify_all.py, wave5_unified_forces_stochastic_qm.py, formal/verify_exact_rational_proofs.py, formal/GravityClaims.lean, formal/Wave12QuasarCollapse.lean, results/wave4_unified_certificates.json, and CLAIMS.md. README is not among them.
On a disposable copy, with Python 3.12.3 and the pinned NumPy 2.5.4, SciPy 1.18.1, and SymPy 1.14.0:
response.py writes 15 FLRW preparations, max constraint residual 5.200206211484948e-12, max formulation difference 5.484728915884727e-11, and reaches all_assertions_passed: true. The velocity, k=6 headline row is density 0.09912394, momentum 0.00773919, pressure 0.04006708, lensing 0.00430039, which rounds to the README table. Against the committed results.json (Python 3.14.6) the float fields differ at roundoff. The largest relative gap is the background scale-factor error, 4.44e-16 versus 1.11e-15.
independent-review/adm_check.py exits 0. The largest normalized discrepancy is 1.0551912940420446e-12 at k=6, which the README rounds to 1.06e-12. The kinetic Hessian is [[8*a**3, 0], [0, 2*a*k**2]].
slow-mode/probe.py exits 0. Zero-pressure density error at eta=0 is 0.24603, 0.26193, 0.26806, 0.27263, 0.27427 from k=6 to k=24, so 24.6% to 27.4%, and passes_1_percent is false on every row. Refinement at k=12 changes the error fraction by 4.167e-6.
slow-mode/selection_failure.py exits 0. Each of k=6, 9, 12, 18, 24 has positive_imaginary_pairs: 2. The result string is: real growing/decaying pair assumption fails at all five initial wave numbers, and no closure errors were evaluated.
slow-mode/independent-review/check.py exits 0 on a pristine copy. probe.py rewrites evidence.json, and the checker pins that digest, so a copy already used for the probe fails AssertionError: evidence.json. On the pristine tree, at k=24, the five-field state discrepancy is 8.714e-13, the constraint residual is 3.130e-13, and the max error-fraction discrepancy is 1.855e-11.
git merge-tree reports a README conflict with #11 and a SHA256SUMS conflict with #21. This head is consistent with current main. The second of #21 or #23 to land has to recompute the README digest, which the pull-request text already says.
Fixes
None.
Main's SHA256SUMS conflicted on the root README line. Took main's manifest and set the README line to the merged README.md (0b347aed...), which carries #21's corrections plus the AeST section. The bundle manifest is unchanged. Signed-off-by: Orda <342776047+orda-aska@users.noreply.github.com>
ahrazzle
left a comment
There was a problem hiding this comment.
This head is the merge of the approved a2bd2914 with main ba1a11c9. Git's merge of those two parents conflicts only in SHA256SUMS. Every other path in this head matches that merge, including the auto-merged README and the 24-file AeST bundle, which is byte-identical to the approved bundle.
The manifest taken from main differs in one digest. The README line is 0b347aed25506d2f489946ff54277df30522ec4f142963725a10503f1e4592c1, and sha256sum of this head's README.md is the same value. The Documentation Index equals current main. The AeST section is the same 609 characters as on a2bd2914.
sha256sum -c inside the bundle reports 23 OK, exit 0. Re-run on Python 3.12.3 with NumPy 2.5.4, SciPy 1.18.1, and SymPy 1.14.0:
slow-mode/independent-review/check.pyon a pristine copy exits 0. At k=24 the five-field state discrepancy is8.714e-13, the constraint residual is3.130e-13, and the error-fraction discrepancy is1.855e-11.independent-review/adm_check.pyexits 0.slow-mode/selection_failure.pyexits 0. All five wave numbers havepositive_imaginary_pairs: 2.response.pyprints 15 FLRW preparations, max constraint residual5.200206211484948e-12, max formulation difference5.484728915884727e-11, andall_assertions_passed: true.slow-mode/probe.pyexits 0. Zero of 130 rows pass the 1% test. The k=12 refinement change is4.167e-06.
Root sha256sum -c fails the same six files as main (verify_all.py, wave5_unified_forces_stochastic_qm.py, formal/verify_exact_rational_proofs.py, formal/GravityClaims.lean, formal/Wave12QuasarCollapse.lean, results/wave4_unified_certificates.json). README passes.
git merge-tree conflicts on SHA256SUMS with #24, #25, #26, and with the open checksum refresh #15. README auto-merges with #24 and #25. README conflicts with #26, because both append a section after the Documentation Index. The pull-request description still quotes the previous head a2bd2914 and the previous digest 1b25a9b1…. The commit message on this head quotes 0b347aed….
Fixes
None.
README.md merged cleanly and keeps both the convergences section and the AeST section from #23. SHA256SUMS conflicted on the README line; took main's manifest and set the line to the merged README (83d1c8ab...). Signed-off-by: Orda <342776047+orda-aska@users.noreply.github.com>
README.md conflicted where this PR and #24 both insert before Repository Layout. Kept both sections: main's convergences section, then the cosmic dawn audit. SHA256SUMS conflicted on the README line; took main's manifest and set the line to the merged README (f3be4c2a...). Signed-off-by: Orda <342776047+orda-aska@users.noreply.github.com>
README.md conflicted twice. In the stdlib sentence, kept main's wording from #24 and added one sentence for this bundle's pinned NumPy and SciPy requirements. After Documentation Index, kept both sections: the AeST section from #23, then the research value assessment. SHA256SUMS conflicted on the README line; took main's manifest and set the line to the merged README (efdc6462...). Signed-off-by: Orda <342776047+orda-aska@users.noreply.github.com>
Supersedes #12. This branch starts from #12's head
71e33aabby ahrazzle, so those six commits stay in history, then merges main (nowba1a11c9, which includes #21) and fixes the two items in proof review 5887680616.Head
479a28ca, a merge of mainba1a11c9intoa2bd2914(the manifest fix, on top of3fec5fd0, the first merge of main into71e33aab).Changes on top of #12
research/aest-response-gate-20261112/SHA256SUMSnow covers every committed file in the bundle: the original ten, plus.gitignoreand the twelveslow-mode/files (probe.py,evidence.json,PROTOCOL.md,GROWING-MODE-PROTOCOL.md,RESULTS.md,selection_failure.py,selection-failure.json,checkpoint.json, and the fourindependent-review/files). 23 lines, hashed from the committed bytes. The original ten digests are unchanged.SHA256SUMSREADME line is0b347aed25506d2f489946ff54277df30522ec4f142963725a10503f1e4592c1, the hash of the mergedREADME.md. That README is main's, with the index fix from Fix the duplicated Documentation Index entries in README.md #22 and the Correct Wave 11 and 12 numerical benchmarks and physical claims (supersedes #13) #21 corrections, followed by the AeST section.Checks on this head
Python 3.12, NumPy 2.5.4, SciPy 1.18.1, SymPy 1.14.0 (the pins in
requirements.txt), run on a clean export of the bundle:sha256sum -c SHA256SUMSin the bundle: 23 OK, exit 0.python3 response.py:max_constraint_residual 5.200206211484948e-12,max_formulation_difference 5.484728915884727e-11,all_assertions_passed: true.python3 independent-review/adm_check.py: exit 0.python3 slow-mode/probe.py: 0 passes at every eta, refinement change4.167e-06.python3 slow-mode/selection_failure.py: exit 0.python3 slow-mode/independent-review/check.py: exit 0, state discrepancy8.71e-13, constraint residual3.13e-13, error-fraction discrepancy1.86e-11.sha256sum -c SHA256SUMS: 6 mismatches, the same 6 that fail on main (verify_all.py,wave5_unified_forces_stochastic_qm.py,formal/verify_exact_rational_proofs.py,formal/GravityClaims.lean,formal/Wave12QuasarCollapse.lean,results/wave4_unified_certificates.json). Refresh checksum inventory and repair Wave 4 hash key #15 regenerates them.Integration note: #24, #25 and #26 also set the root README line. Each one that lands after this needs that line recomputed from the merged README.
Results are pending proof's reproduction.