back to table · edit · history · where entries came from · files · download
1781 bytes, as of the version from 2026-10-28 22:59 (current). Recorded here, not run.
# T558 build environment findings, 2026-10-28
The canonical and archived skills matched. Computation used agents/sage.sh on
the local AWS builder with its configured memory limit; no website-container
computation or SSH command was used. Python 3.12.3, Sage 10.9, numberdb 0.1.10.
PDF inspection used the shared browser dependency directory (PyMuPDF 2.0,
pypdf 6.19.0); browser checks use its Playwright 1.64.1 and shared runtime.
Authenticated GET /T558 returned the empty draft successfully. Anonymous GET
returned 404. This run did not require a browser login to inspect the draft.
POST /api/table/T558 refused the prepared prose-only document because Numbers
was empty: HTTP 400, "A table needs at least one entry." The write endpoint
calls _document_of without allow_empty even for an existing private draft;
the error's suggestion to send X-Draft applies to creation, not this endpoint.
No change resulted. The authorized path is to fill the existing draft using
the generator's entry upsert, then update its complete document preserving
all read-back entries. No new draft, identifier, or parameter was introduced.
This is a site behavior finding, not a change to the table-contribution skill.
One guessed arXiv URL returned an unrelated scheduling paper. Its first-page
identity was checked and it was rejected, not used as mathematical evidence.
Those downloads are retained under unrelated-2306.16669-*; the correct article
was retrieved through the journal's verified arXiv identifier 1609.00495.
Transient format-check error: ExactPolynomial.variables is a method, not a
property. The check was corrected to variables() and rerun successfully; the
initial failed log remains preserved. These were research/checker issues, not
failures of the generated polynomials.