Researchers describe their research programme and lab layout to a system of AI agents, which drafts step-by-step protocols for the lab’s robots, tests them in a virtual copy of the lab, and then instructs the real robots to carry out the experiments. The approach targets persistent limits of lab aut