Experiment-as-Code Labs: Declarative Stack for AI-Driven Discovery

A new paradigm called "Experiment-as-Code (EaC) Labs" proposes encoding experiments as declarative configurations to bridge AI agents and automated lab equipment. The approach aims to untether AI from purely digital environments, enabling real-time exploration and control of physical lab instruments. The concept draws insights from computer systems to integrate AI with programmable lab APIs, allowing agents to adapt to unexpected clues during experiments. This framework could accelerate scientific discovery by combining AI's analytical power with hands-on laboratory work. The research is detailed in arXiv preprint 2605.04375.