From AI Agents to Flexible Automated Research Pipelines

AI agents can connect activities that are usually distributed across research software, development tools and laboratory instruments. Drawing on my experience with hardware and software development, remote computing and instrument integration, I describe a practical route towards end-to-end research automation. The central requirement is a working connection between readable documentation, executable tools and observable results. With suitable robotic capabilities, this connection could extend to physical operations and allow laboratories to reconfigure workflows around changing research goals. Codex is used as one practical example of a general class of AI agents. This perspective outlines the implementation conditions, feedback loop and implications for engineering, while distinguishing personal experience from proposed extensions to robotic research pipelines.