Meta-Cognition Engine

Learning how to use AI to learn HER2

The cognition — the exact way LLM intelligence is applied to a problem — is a first-class, versioned, gradeable object. Many cognitions attack the same real blinded failure; their proposals are committed before the answer is revealed, then graded against held-out reality by the strongest available grader. The system learns which way of thinking produces predictive improvement per unit compute.