AI-Native Discovery Engines

Science still runs on a slow hypothesize-experiment-interpret loop. YC wants AI-native systems that run that loop end to end alongside researchers, not just research copilots.

The scientific loop of hypothesis, experiment and interpretation works, but it is slow and every step takes significant human effort. Frontier models now reach PhD-level performance on many scientific reasoning benchmarks, and the frontier is moving from research copilots to systems that run closed discovery loops. In drug discovery, materials science and protein engineering, models already propose candidates, automated labs make and test them, and the results feed back in.

The ask: AI-native discovery engines that partner with scientists to generate and test hypotheses.

Read the full request on YC →