Near-Autonomous AI Chemist Improves a Challenging Drug-Making Reaction

OpenAI and Molecule.one have developed a near-autonomous AI chemist that uses GPT-5.4 to improve a challenging reaction in medicinal chemistry. The AI system, which combines large language models with robotic experimentation, can plan and conduct chemical experiments with minimal human intervention. In this case, the AI optimized a specific reaction known as "dearomatization"—a difficult process used to create complex 3D molecular structures important for drug discovery.

This matters because designing and testing new drugs is traditionally slow and expensive. The AI chemist can dramatically speed up this process by autonomously suggesting modifications, running experiments, and analyzing results. Think of it like a super-fast lab assistant that learns from every attempt and refines its approach without needing constant human guidance.

Unlike earlier AI systems that only offered theoretical suggestions, this system can physically execute experiments using robotic lab equipment. It was able to improve the yield of the dearomatization reaction by exploring conditions that human chemists had not yet tried. This represents a significant step toward fully autonomous scientific discovery.

For a detailed look at how the system works and what it achieved, refer to the original OpenAI blog post.