Overall: Gated
The pillar is open for reading and critique, closed for citation-as-fact. Access is deliberate.
Status & Remediation.
This is the page BCIA is proudest of. It records exactly where the field stands, which early claims failed audit, how we corrected them, and what remains unsolved. Nothing is published as fact until it clears the protocol below.
We publish our mistakes on purpose. A research pillar that hides its retractions cannot be trusted with its claims.
Where BCIA stands today.
Document Zero: in review
The seven bases are drafted and under open audit in the Atrium.
Volumes: not yet shipped
No volume publishes before Document Zero settles and each passes the protocol.
Claims that failed — and what replaced them.
Every retracted or corrected claim is logged here in public. This is the ledger of what we got wrong on the way to getting it right.
🚫 Banned Claim
The "DishBrain learned Pong 28,800× more efficiently than DeepMind" comparison is banned from all MetaX material as a like-for-like efficiency claim. The two systems solve different tasks, on different substrates, measured differently — it is a category error, not a result.
Advertisement
The gate every claim must pass.
Nothing leaves research-preview status until it clears all five steps.
Scope limits & what remains unsolved.
Measurement
We lack agreed, substrate-fair metrics. Until we have them, cross-substrate "efficiency" claims stay banned.
Reproducibility
Many cited results are single-lab. We require independent replication before any claim graduates.
Ethics envelope
The human-cell exclusion is fixed; the wider ethics envelope for embodied systems is still being drawn.