A near-autonomous AI chemist improves a challenging reaction in medicinal chemistry
Summary
OpenAI researchers, in collaboration with Molecule.one, used a near-autonomous system featuring GPT-5.4 and the Maria laboratory to optimize the Chan–Lam coupling reaction. By autonomously generating hypotheses, designing experiments, and analyzing data, the AI identified TEMPO as a key additive that significantly improved reaction yields for primary sulfonamides—a critical but historically difficult process in drug discovery. While human chemists provided oversight and validation, the study demonstrates the potential for AI to serve as a high-level scientific partner capable of accelerating research through high-throughput experimentation.
(Source:OpenAI)