AI drug discovery becomes a bottleneck trade: Twist orders up triple digits, GenScript hits 8,000 designs a day
QuanquanGu · x · 2026-10-05
Quanquan Gu highlights a Robonomics essay arguing AI drug discovery (AIDD) is becoming a bottleneck trade, alongside his own vision at Geodesic: design in silico, validate in the wet lab, feed experimental data back as semi-synthetic training data, and loop — making the wet lab part of the learning algorithm.
Key points:
- Frontier labs are piling in: Anthropic formed a life-sciences group with its own wet lab, using Claude agents to identify an uncharacterized enzyme system tied to CRISPR-like DNA repeats; OpenAI launched GPT-Rosalind, a model family built for biology and drug discovery; ByteDance's former AIDD unit, now Anew Labs, runs foundation models plus a real internal drug pipeline (ByteDance still holds a majority stake); Isomorphic Labs is explicitly aiming to take AI-designed medicines into human trials.
- Supply chain is moving: Twist ($TWST) expects triple-digit percentage growth in AI-enabled discovery orders in FY26 and again in FY27; GenScript's AIDD business doubled YoY in 1H26 — officially 4,000+ designs/day validated, but channel checks suggest 8,000/day now, on a path to 16,000/day by end of 2026. Sequencing ($ILMN, $TXG), lab automation and CROs all benefit.
- Core logic: AI doesn't shrink wet-lab demand — it floods the funnel with candidates, shifting the bottleneck to experimental validation and boosting the picks-and-shovels layer.
More from AGI Musings
- Should AI tell you if a core belief is psychologically necessary but false? — PierceLilholt · 2026-10-05
- AI agents are about to flood the workforce, and no one's ready: WIRED — Dr_Singularity · 2026-10-05
- Eric Horvitz Revisits His 2001 Human-AI Vision: Core Capabilities Now Exist — erichorvitz · 2026-10-05
- Sam Altman: the biggest mistake talented people make is not thinking about what to work on — _AustinCalvert_ · 2026-10-05
- AI decompilation era: PS5 is 80% ported to PC and 'closed source' is ending — almmaasoglu · 2026-10-05
- The Curve talk to tackle acceleration emergency and superintelligence policy agenda — jachiam0 · 2026-10-05