Verify materialized Jacobian syzygy ledgers - #927
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c22606cd98
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 32692d4a3f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7bfd076828
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c8690509bb
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Second-pass architecture review: rebase + likely merge. The review cycle appears to have fixed the important trust-boundary issues: canonical >4,300-digit conversion, verifier discoverability, checker cost gating, linear-factor support, and digest formatting. The final shape is appropriate: a materialized evidence-producing capability gets its own provenance-bound verifier, while the compact inline producer/verifier path remains separate. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b0dd78726a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b2e327a99a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c9242d034c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 08aa49794b
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| if len(support) == 1 and sum(value > 0 for value in next(iter(support))) < 3: | ||
| maximum_degree = 0 |
There was a problem hiding this comment.
For a valid input containing 16 uniquely labelled copies of x+y with max_degree=8, the support has 17 monomials, so this shortcut does not apply and the final estimate is 17 * 126126, causing both syzygy verifiers to return UNSUPPORTED. Because the product omits z, its third partial derivative is zero and the producer stops after the small degree-zero map; detect variables absent from all factors (or use the stored searched-through degree) rather than limiting this shortcut to singleton support.
AGENTS.md reference: AGENTS.md:L72-L76
Useful? React with 👍 / 👎.
morluto
left a comment
morluto
left a comment
There was a problem hiding this comment.
Post-cutover architecture audit
Disposition: retain the need, redesign the verifier on current main.
Large materialized ledgers are a legitimate durable-artifact case that typed runtime-local value references do not replace. However, the checker should not be framed as trusting producer provenance or a legacy result_uri. An artifact URI is a carrier, not proof.
Define a current checker request that binds the expected typed syzygy subject/input and the exact artifact digest/content, independently parses and replays the ledger, and emits a verification record bound to that subject, candidate/evidence, operation, and checker. Remove catalog relationships and “advertise from producer” logic. Use artifact_uri/current storage contracts, preserve the compact inline checker separately, and rebase all tests onto current result/verification paths.
|
Post-#1256/#1265 review: keep the materialized syzygy verifier, but split out and delete the old discovery-relationship mechanism. The missing checker is still a coherent feature: a retained coefficient ledger can have an independent exact verifier with its own operation ID. That fits the current verification model. The Recommendation: restack from current main with the checker declaration/implementation/tests only. Make it discoverable through its own title/tags and a valid invocation example if necessary. Drop the projections change entirely, then rebase the frontier/catalog test onto the current explicit portfolio. |
3f0e4ae to
2b1c961
Compare
|
Superseded by #1292. The replacement keeps the durable |
2 participants
Summary
polynomial.jacobian_syzygy.coefficients.materializeresult_uriReproduction
A weak-model audit of Satriano–Usatine's 2026 Batyrev counterexample used the exact degree-14 homogenization of their published stringy E-polynomial through multiplier degree 8. The model selected the evidence-complete coefficient-ledger producer, which completed and returned a stored result, but could not verify it: discovery exposed no related ledger verifier, while the advertised minimum-degree verifier requires an inline
{input, candidate}envelope. Large stored ledgers have opaque previews, so the candidate cannot be recovered inline.This PR gives the materialized producer its own provenance-bound verifier accepting
{result_uri}.Tests
ruff checkon all changed filesAll passed locally.