A decade in the making: new paper tackles AI retrosynthesis limits
rbhar90 · x · 2026-08-19
The author shares a new paper ten years in the making. Around a decade ago in grad school at Stanford, a lab meeting talk in Vijay Pande's group introduced him to the limits of retrosynthesis: maitotoxin, a potent neurotoxin with 32 fused rings and one of the most complex natural products known, was beyond the reach of retrosynthetic algorithms, and several groups had failed at its total synthesis.
The problem stuck with him across research projects—during his PhD he built the first seq2seq model for retrosynthesis with Bowen Liu and Vijay Pande as a proof of concept.
More from Research
- MatrAIx uses 8.3B persona agents to simulate the world with 91.5% consistency — jiqizhixin · 2026-08-19
- RLC Keynote: Reshaping RL Exploration Through Co-Evolution — tw_killian · 2026-08-19
- When the Hard Part Stops Being Hard: How AI reshapes academic research culture — rseroter · 2026-08-19
- Quantifying Pretraining Variance: Floating-Point Order Effects Nearly as Large as Data Seeds — AndrewLampinen · 2026-08-19
- Muon Optimizer Trains nanoGPT in Just 1.23 Minutes — dianarycai · 2026-08-19
- MarinDNA 1B Model Matches Evo 2 40B Performance with 1980x Less Compute — anshulkundaje · 2026-08-19