Google’s C2S‑Scale 27B foundation model, developed with Yale and based on Gemma, generated an in silico hypothesis for a drug acting as an interferon‑conditional amplifier of antigen presentation; researchers report in vitro confirmation of the model’s prediction. Google and collaborators simulated thousands of drugs across immune contexts and prioritized unexpected candidates, with multiple in vitro validations. The work—published as a blog and preprint—illustrates emergent capabilities of large foundation models to prioritize biologically plausible, experimentally testable hypotheses. A conditional amplifier here refers to a compound that boosts antigen presentation only in an immune‑context‑positive environment (e.g., low interferon), a concept relevant to turning immunologically “cold” tumors “hot.”