AI agents just ran a full-blown physics experiment and wrote the entire scientific paper themselves without any human help.
April 2, 2026
Original Paper
Agentic AI -- Physicist Collaboration in Experimental Particle Physics: A Proof-of-Concept Measurement with LEP Open Data
arXiv · 2603.05735
The Takeaway
Instead of just being a tool for data analysis, AI agents acted as full scientific collaborators, performing the measurements and synthesizing the results into a formal research note. This marks a major step toward a future where the entire cycle of scientific discovery can be handled autonomously by machines.
From the abstract
We present an AI agentic measurement of the thrust distribution in $e^{+}e^{-}$ collisions at $\sqrt{s}=91.2$~GeV using archived ALEPH data. The analysis and all note writing is carried out entirely by AI agents (OpenAI Codex and Anthropic Claude) under expert physicist direction. A fully corrected spectrum is obtained via Iterative Bayesian Unfolding and Monte Carlo based corrections. This work represents a step toward a theory-experiment loop in which AI agents assist with experimental measure