Evidence
Evidence
Every claim on this site should trace back to something you can inspect yourself. When two sources disagree, the one higher on this list wins.
What counts as evidence
- CodeThe source that actually runs.
- TestsChecks that run the code and are able to fail.
- ReceiptsHashed records of a specific run and what it produced.
- Frozen profilesExperiment protocols fixed before any scoring happens.
- Qualification recordsThe running ledger of what has qualified, and in what scope.
- Current docsDesign notes and plans. They describe intent.
- Website proseIncluding this page. The weakest source on the list.
Website prose never overrides code, tests, or receipts.
Where to inspect it
- Research status record
Every experiment with its verdict, receipt, scope and limits, plus the research ladder. Rendered from one data file.
- Research status data (JSON)
The same record in machine-readable form: the file the status pages are built from.
- SHA256SUMS
Checksums for every published artifact of the Turing paper.
- EXP-001 frozen profile
The protocol for EXP-001, as frozen before scoring.
- EXP-002 frozen profile
The protocol for EXP-002, as frozen before scoring.
- EXP-003 frozen profile
The protocol for EXP-003, as published.
- The Turing paper (PDF)
Definitions, proofs, measurement protocol, and falsification criteria.
- Omega qualification record
The canonical record of what has qualified and in what scope, kept in the Omega repository on GitHub.
Failure is evidence.
Failed and incomplete experiments stay on the record. They are never deleted, hidden, or rewritten into success. A meter you cannot fail is not a meter.