AI Agents in Practice for Physics Research¶
KIT · 22 October 2026 · 09:00–17:00 · Daniel Murnane. A draft companion for particle and astroparticle PhD students and postdocs, across theory and experiment.
Working notes¶
- Show what agentic science can achieve today.
- Explain the machinery and confront the limits.
- Make professional scientific practice the centre.
- Separate research loops from implementation choices.
- Work through experiments, papers and programmes with live demos.
Draft course
About 70% of the day is live work. The ovens start early and are inspected at breaks and at 16:30. Placeholders are explicitly marked; a blank scoreboard is not a result.
Explore the course¶
- Setup and five environments
- Machinery: LLM basics and tool loop
- Evidence: possibilities and limits
- Professional practice
- Research loops: inner, middle, outer
- Patterns and live demos
- Landscape and cheatsheet
Programme¶
| Time | Block |
|---|---|
| 09:00 | Ovens, possibilities and evidence |
| 09:25 | Machinery and limits |
| 10:10 | Professional practice and two axes |
| 10:40 | Break; check ovens |
| 10:55 | Inner loop and fiddler benchmark |
| 12:15 | Lunch; ovens continue |
| 13:15 | Theory sprint and experiments |
| 14:00 | Project, paper and RANGE dashboard |
| 15:00 | Break; check ovens |
| 15:15 | Programme, fleet and volunteer setup |
| 16:00 | Landscape |
| 16:30–17:00 | Serving, trust tally and next steps |
Before the course¶
- [TODO: select a compact worked example and its verification artifact.]
- [TODO: rehearse the live walkthrough and prepare a fallback.]