Researchers have developed an AI system that can autonomously design and refine controllers for complex physical systems. The system uses large language models and code generation to create interpretable control policies. It was tested on a fluid-structure interaction problem, where it successfully designed a controller that could navigate an unsteady flow and reach a target. The process is fully traceable, with an auditable evolution log showing the development of the control architecture. --- Why it matters: This matters because it demonstrates a new approach to designing controllers for complex physical systems, which can be difficult to model and optimize using traditional methods. This could have significant implications for fields like robotics and aerospace engineering, where precise control is crucial. Source: https://arxiv.org/abs/2606.08405