back to table · edit · history · where entries came from · files · download
3043 bytes, as of the version from 2026-11-20 05:53 (current). Recorded here, not run.
# T772 build evidence
T772 is a private draft of primitive real periods for all 284 source rank-zero
curves through conductor 100, with stable N,c4,c6 parameter identities.
Independent scientific critique and argument-selection review are still pending.
This build did not offer, publish or review it.
To reproduce outside the campaign, download generate.py, roster.py,
allgens.00000-09999.gz, alllabels.00000-09999.gz build_check.py and skill-manifest.json together.
Decompress the two .gz source archives (gzip -dk FILE.gz); the original exact
source names must be beside the scripts. The archives reproduce the full pinned
source bytes, not a roster-only extract. Install numberdb in SageMath.
Run `sage -python generate.py` for a read-only preview, and
`sage -python build_check.py` for the complete numerical checks. Default checks
do not need a checkpoint and recompute the full roster. `--verify` is read-only
and needs your own key to see a private draft. `--publish` is an explicit
additive entry upload for an authorized contributor; it does not offer review.
Do not run the session-specific fill.py with somebody else's provenance.
numerical-evidence.json records every production outer enclosure, certified
direct integral, both PARI approximations and exact arithmetic metadata.
check-results.json is the raw successful builder report. The production target
is 100 significant decimal digits, with one uncertain last place per NumberDB's
bare-decimal semantics. All 284 stored intervals were read back and checked.
Sage 10.9, PARI 2.17.1, cypari2 2.2.4, gmpy2 2.3.0, numberdb 0.1.10 and Python
3.12.3 were used on the isolated builder. deterministic.log records versions,
native types/error checks and size measurements. The primary Sage source and
Cremona/PARI documentation locations and hashes are in source-downloads.json.
The archived canonical skill bundle manifest is skill-manifest.json.
The first 212 and following 67 completed numerical records were resumed using
checker archives v1/v2 and exact dependency hashes, not recomputed after unrelated
checker changes. The final five records complete the same exact roster.
77b3 required a rescaled stable defining-integral expression; see build_check.py
and probe_difficult_stable.log. The 99d3 guard assessment is accuracy-assessment.md.
The final generator differs from the numerical-run copy only in its module
docstring's measured accuracy statement (594 bits); documentation-change.json
records the unchanged computational AST. Final deterministic and stored-value
reproducibility checks use the attached final generate.py.
The source rank claims were checked with unconditional PARI algebraic rank
bounds [0,0] on every curve. This does not provide an independently verified
2-descent covering certificate, a Sha order, or a proof of BSD. The integration
and AGM checks share Sage algebraic roots and Arb arithmetic; PARI is a separate
period implementation. The fresh critic must make its own source, numerical,
precision, coverage and presentation assessment.