Beijing: AI scientists run the benches alone

A public lab in Beijing demonstrates autonomous agents that design, execute, and interpret material chemistry experiments.

The fact

The AI4S (AI for Science) program in Beijing presented, via Pandaily on July 10, 2026, a robotic platform coupled with LLM that mixes a conductive fluid in twenty seconds and automatically iterates variants. The agent proposes the synthesis, the robot executes, a model interprets the measurements and restarts the experiment - without human intervention on the short cycle.

Our analysis

The real signal is not "AI replaces the chemist". It's the iteration rate: an experimental cycle in a few tens of seconds, compared to several hours for a PhD student. The acceleration factor changes the landscape of materials R&D - batteries, catalysts, semiconductors. The usual question remains: what proportion of these iterations produces exploitable results, rather than additional points on an already known curve? Pandaily does not decide and no peer-reviewed paper is cited. We remain cautious: elegant demonstration ≠ scientific discovery.

To watch

The 2026 publications from AI4S. If the discoveries hold up to independent replication by other labs, the paradigm of materials R&D really shifts. Otherwise, we will have seen a nice showcase and another political influence tool.