OpenAI partnered with Molecule.one to deploy an AI system that enhanced a notoriously difficult chemical reaction in medicinal chemistry. The team connected GPT-5.4 to Maria, an agentic chemistry AI integrated with a high-throughput laboratory, giving it an open-ended goal to improve Chan-Lam coupling for process chemistry.

Maria ran 10,080 reactions across two experimental cycles while GPT-5.4 proposed hypotheses, analyzed data, and suggested follow-up experiments. Human chemists provided steering prompts, selected proposals to test, and made limited corrections to experimental plans. The most promising approach, OAI-M1-03, identified primary sulfonamides as a challenging substrate class and proposed mild oxidants like TEMPO to boost reaction performance.

The optimized conditions dramatically improved yields: 88% of boronic acids and 83% of sulfonamides showed better results, with mean yield rising from 16.6% to 25.2%. Reactions above 30% yield increased from 15.6% to 37.5%. Independent bench-scale validation confirmed these improvements for 11 of 14 substrate pairs, demonstrating the system works beyond microliter screening.

This matters because synthesis often bottlenecks drug discovery—scientists can only test molecules they can make. Primary sulfonamides appear in anticancer and antimicrobial drugs, yet their Chan-Lam coupling historically gave low yields. Making this reaction more reliable could expand medicinal chemists' toolkit for exploring therapeutically relevant compounds.