Würzburg Seminar on Quantum Field Theory and Gravity

I use LLM-based tools daily in my actual research pipeline, for proof exploration, computation scaffolding, literature triage, and writing. This workshop and Q/A session is a hands-on attempt to transfer some of that practice. We build from basic agents, confront context management head-on, and orchestrate multi-agent workflows on problems relevant to research in theoretical physics. I will share what fails, what remains fragile, and where the serious open challenges are. The central thesis is uncomfortable for both skeptics and enthusiasts: these tools can genuinely work for research, but only after sustained and deliberate practice.