Two AI advances showcased at recent conferences highlight how generative models are being applied to biological discovery. KAIST researchers unveiled a generative AI that predicts and assembles cellular drug responses to guide combination strategies and mechanistic experiments. Separately, Google’s C2S‑Scale 27B foundation model — built with Yale and leveraging Gemma — generated an interferon‑conditional amplifier hypothesis that the team validated in vitro. Both groups emphasized emergent capabilities in larger models for context‑dependent biological reasoning: KAIST focused on modular ‘LEGO‑like’ assembly of cellular responses, while Google highlighted the model’s ability to prioritize candidate drugs that boost antigen presentation selectively in an immune‑context‑positive state. These demonstrations suggest generative AI can shrink hypothesis generation cycles and surface unexpected mechanistic candidates, but the field will need reproducible validation, transparent model benchmarks, and careful experimental design to ensure predicted hits translate reliably to in‑lab biology.