Broader course for all
Daniel Thomas Murnane
Abstract
Agents are quickly becoming an invaluable tool for practicing scientists across all components of their workflow: organizational, software, career management, experimentation, publication and theory building. In this one-day course, we will build a personalized scientific agentic workflow from first principles: how LLMs reason and use tools, what loops look like and how long can they run, how to manage the knowledge base and results base of an agentic workflow, and ultimately what is the most targeted setup for each attendee. We will look at which research tasks are essentially solved by agents in late 2026 (debugging code, running simulations, literature compilation, and general automation), which are likely to become solved in 2027, and where is there still value to invest deep expertise. The course is aimed at all KSETA fellows, whether in theory, experiment, astroparticle physics or technology development. No AI/ML experience is needed, nor any coding skills.