Digital Protein Twin

Curiosity Loop

The engine that decides what HER2 question is most worth investigating next — by scoring the Twin's own uncertainty. It reads Reality Memory but never overwrites it, and every priority score is derived from real recorded drivers (low confidence, model disagreement, missing evidence, open contradictions, benchmark misses). The reasoning is used only to phrase the question; it introduces no new facts.