00 / CORPUS
Coverage
CHECKING
Every Brain node gets an explicit decision.
A11oy / model operations / M1
Evidence before inference.
The M1 adapter is an experimental, local-only candidate. The gate independently checks exact weights, offline load evidence, evaluation receipts, GPU admission, and provider identity before a single token can run.
NOT_PROMOTEDQuality remains NOT_ESTABLISHED. This surface never converts compatibility evidence into a production or safety claim.
Operational chain
CHECKING
01 / ARTIFACT
Weights
CHECKING
Exact base, adapter, and tokenizer digests.
02 / RUNTIME
Load
CHECKING
Offline reload, GPU admission, local provider.
03 / EVIDENCE
Evaluation
CHECKING
Receipt integrity with limitations retained.
04 / EFFECTOR
Inference
DISABLED
Experimental local-only. Production blocked.
Verified identity
Candidate
—Base revision
—License evidence
—Provider
—GPU admission
—Brain decisions
—Distinct artifacts
—Quarantined / excluded
—Formula curriculum
—Formula status
—Corpus receipt
—Evaluation receipt
—Unrun suites
—Bounded local turn
No inference attempted.