LeashBio's replayable experimental worlds run 10,000 drug-discovery campaigns in a weekend
madhavsinghal_ · x · 2026-10-16
Latent Space published an interview with LeashBio's Ian on "lab-in-the-loop" data:
- DNA-encoded libraries generate billions of protein–molecule measurements
- No team beat chance on an unseen chemical library in Leash's Kaggle competition
- Public datasets can teach AI to recognize the chemist rather than understand chemistry; protein mutation engineering exposes model memorization
- "Premeasured experimental worlds" let agents run 10,000+ medicinal chemistry campaigns in a weekend
- Also covers Hermes vs. Boltz and going from DEL hits to mouse studies in six months
More from Research
- Anima Anandkumar on neural operators: AI weather models beat supercomputers — AnimaAnandkumar · 2026-10-23
- Scaling GLM-5.3-Flash training to 1M token context, 64x over Megatron baseline — AndrewLBeam · 2026-10-23
- Experimental boolean-logic LM: microsecond forward pass on one CPU thread, but painfully slow training — cephaloform · 2026-10-23
- AdaDex turns seconds of human hand motion into generalizable dexterous robot skills — mangahomanga · 2026-10-23
- ReAPO paper: conditioning retries on failures beats GRPO in RLVR reasoning training — jasondeanlee · 2026-10-23
- Quantum-JEPA: first JEPA world model deployable on real quantum circuits (QPU) — DrMorganLevine · 2026-10-23