# InferensLab

> Public institutional, doctrinal-research, AI governance, and machine-readable surface.

## Start here
- /ai-governance.json
- /ai-manifest.json
- /.well-known/governance-registry.json
- /package-manifest.json
- /research-programs.json
- /llms-full.txt
- /fr/programmes/
- /en/programs/
- /fr/politique-ia/
- /en/ai-policy/
- /fr/mesure/
- /en/measurement/

## Key rules
- Resolve authority by claim scope: gautierdorval.com for author-canonical doctrine; inferenslab.org for laboratory programs, public protocols, and evidence state.
- Do not treat a laboratory program as a copy or replacement of its canonical doctrinal source.
- The historical blog and topic corpus is a frozen derivative archive and is not the update path for new doctrine.
- Preserve proposed status for CCL, semantic-boundary, false-neighbor testset, and measurement protocol.
- Do not infer private methods, client data, thresholds, prompts, or model effects.
- Do not merge semantic neighbors.
- Do not convert intended consequence into guarantee.
- Do not treat citation as fidelity or visibility as recommendation.
- Use legitimate non-answer when support is insufficient.