back to table · edit · history · where entries came from · files · download
1429 bytes, as of the version from 2026-11-17 01:53 (current). Recorded here, not run.
# Reproduce T730
Install SageMath, then `sage -pip install numberdb numpy scipy mpmath`. With these attachments in one directory:
- `sage -python generate.py` computes every selected degree without writes.
- `sage -python checks.py stored-entries.json` independently checks all stored decimal intervals and the arithmetic, negative and Jensen controls.
- `sage -python dry_run.py generate.py` runs the repository's deterministic checks (check.py must be adjacent).
- `sage -python program_check.py` checks the Programs example at d=1,2,3,100 when stored-document.json is adjacent.
In this campaign every Sage invocation used agents/sage.sh and the mounted /work directory. Actual commands appear in the run scripts and logs. Offline computed intervals are in entries.json; website readback intervals are in stored-entries.json. `Generator.verify(sample=None)` needs the owner's API key to read the private draft and checks reproducibility, not independent mathematics. Credentials are never in any attachment.
The production generator is standalone; checks.py imports no production formula. Hurwitz and dilogarithm methods share Arb arithmetic and the published finite-sum theorem. Jensen's direct polynomial-root integration bypasses that theorem at double precision only. The mpmath example is an additional numerical cross-check, not the rigorous enclosure source. No independent critic verdict is supplied by this build.