Playbooks · Everyday tasks
A human-agent team ran X-ray experiments you can audit
Written November 12, 2026
Three physical synchrotron experiments, run by a human-agent team, with a complete transcript behind them. The record shows who proposed each step, who corrected each error, and where the humans stayed in charge.
What actually ran
Closed-loop autonomous experimentation promises faster science, but building the control software for a physical instrument is a project in itself, so few labs get there. A preprint by Gunik Luthra and eight co-authors, submitted on November 6, 2026, takes a shorter path: coding-agent infrastructure, grounded in validated scientific methods, running synchrotron X-ray experiments on nanoparticle superlattices. A synchrotron beamline is real, shared, expensive equipment, not a simulator, and the experiments ran under objectives that humans set and kept revising as results came in.
Across three experiments, the system kept operating on the physical instrument. That sustained operation is the paper's first claim, and it is the part most agent demonstrations skip: an agent that books beam time has to keep working when the sample, the beam, and the plan all change mid-run.
What auditable means here
The team kept complete transcripts of the human-agent and agent-agent exchanges. Those transcripts are the audit trail. From them, the authors reconstruct decision provenance: which participant proposed a step, whose input shaped the decision that was finally taken, and who caught which error. They use the same record to describe how the cognitive work split between people and agents. For a reader, that is the difference between "the AI did an experiment" and an account you can pull apart decision by decision.
The division of labour, in numbers
Between 61% and 67% of the messages across the three experiments involved interpretation or judgment, rather than routine execution. The character of the work also shifted as the team settled in: methodological and scientific contributions rose from 46% of messages in the first, setup-focused experiment to between 56% and 61% in the later ones. Error correction was mostly a team sport — between 59% and 68% of corrections involved multiple participants — and about half of all decisions incorporated the input of another participant, human or agent.
Read plainly, the agents were not left alone with the beamline and the humans were not supervising from a distance. Each decision that mattered tended to pass through more than one mind, and the transcript is what lets the authors say that with numbers instead of impressions.
Strengths
- The work is physical and inspectable at once. Three real experiments, plus a complete record of who said and decided what, is a stronger package than a larger claim with no trail behind it.
- The provenance numbers give other labs something to design against: if half of good decisions involve a second participant, the tooling around a lab agent should make that second look easy, not optional.
- The objectives stayed human-defined and kept evolving. That is a credible shape for lab automation: people steer the question, agents carry the load of running and interpreting.
Limits, and what the agents did not do
- Three experiments on one kind of sample, nanoparticle superlattices, at synchrotron beamlines. Nothing here shows the same setup surviving a different instrument, a wet lab, or a field site.
- The agents did not choose the scientific goals. Objectives were defined by humans and revised by humans; the autonomy on display is autonomy in execution and interpretation, inside a human-set frame.
- An audit trail shows how decisions were made, not whether they were right. Provenance can tell you who proposed a step and who corrected an error; it cannot grade the science.
- It is a preprint from the team that built the system. The transcript method is the contribution, and other groups have yet to run it on their own agent teams.
Related playbooks
- GeoVisA11y: maps a screen reader can explore
- Klad: you and your agent edit the same brush stroke
- Cheapest way into an always-on agent
- Chase an unpaid invoice with an agent
Sources
- Auditable Human-Agent Teams for Autonomous X-ray Experimentation — arXiv — the three experiments, the transcript audit, and the division-of-labour figures.
Back to: Everyday playbooks hub · Also: Contrastive Discovery: make hypotheses compete · Scientist World: 1,000 executable research tasks · AlphaGenome 2 reads variants per cell type