Summary
Y Combinator’s AI for Bio Club asks whether AI is now good enough to help cure cancer, mixing a live experiment with talks from researchers and founders working at the intersection of machine learning and biology.
The evening opens with an audience Turing test, vibe coded with Claude Code in about 12 hours, in which attendees chatted for two minutes with either a human or an AI agent. About a third of humans were mistaken for AI, and the models were ranked by how convincingly they passed, with Claude and Gemini among those tested.
The presentations then turn to biology. The host points to Evo 2, trained on 9 trillion base pairs, AlphaFold’s role in drug development, AI-designed protein binders, and AI-developed phages entering patients. Later speakers explain virtual cells, including the Universal Cell Embedding model published in Nature, which applies masked-prediction style self-supervised learning to gene expression data.
The talks cover why clinical failure often stems from poor target identification, how falling sequencing costs fuel exponential data growth, and where benchmarks like CASP and the Virtual Cell Challenge are helping the field improve.
📺 Source: Y Combinator · Published November 08, 2026
🏷️ Format: Keynote Launch
