OpenAI collaborated with Ginkgo Bioworks to connect GPT-5 to a fully automated cloud lab, creating an AI-driven system that dramatically lowered the cost of protein production. Cell-free protein synthesis (CFPS) is a rapid method for making proteins without living cells, but it’s notoriously hard to optimize and expensive at scale. Over six rounds of closed-loop experimentation, GPT-5 tested more than 36,000 unique reaction compositions across 580 automated plates. The AI used a computer, web browser, and scientific papers to design experiments, analyze results, and iterate. In just three rounds, it established a new state of the art, achieving a 40% reduction in overall protein production cost and a 57% improvement in reagent cost. The system even discovered novel reaction compositions that are more robust for autonomous lab conditions. This demonstrates that frontier models can now directly drive physical experimentation, removing the iteration bottleneck that has long slowed biological research. Why it matters: By slashing costs and accelerating discovery in protein synthesis, this approach could speed up the development of everything from new medicines and diagnostics to industrial enzymes and cleaner chemical processes, bringing the power of AI into the messy, physical world of biology.