Falsification-First AI for Biological Target Assessment: What Four Prospective Experiments Taught Us About Scientific Agents

TL;DR: A technical report on four prospective experiments: why scientific AI is more useful when it preserves negative evidence and stops when data runs out. Why autonomous scientific systems should