NYU Team Releases WetRobo: Coding Agents Autonomously Run Wet-Lab Experiments

A research team led by @YunaOikawa1, in a collaboration between NYU, the University of Tokyo and others (with Lerrel Pinto among the participants), released the WetRobo project (arXiv:2609.18435), targeting the intersection of robot learning and AI for Science: letting coding agents autonomously write programs in a wet lab, control a robot arm to run experimental operations, and adapt to the environment through physical trial and error. Author Sherry Yang believes the project demonstrates the potential of coding agents as an interface between foundation models and scientific instruments — the agent doesn't just reason about experimental design, but can physically execute experiments and adapt along the way.

Confirmed

Why it matters

2026-09-19 ~ 2026-09-19 · 7 related posts

Primary sources

1 near-duplicate retellings: sherryyangML