Designing molecules that have never been made
We proposed ten bivalent degraders our pilot lab had never made. One of them, the lab had gone on to synthesize independently, and our potency call was right. Cortex got there by learning how the lab's primary chemist reasons, turning their structure-activity table into eight explicit axioms it could apply and combine. This included the judgment that one warhead binds its target well and fails in every molecule built from it. That rule appears in no paper. It was unpublished judgment that lived in one person's head. Cortex captures that judgment and applies it at a scale one person never could.
