Self-driving laboratories (SDLs) are automated systems that accelerate scientific discovery by executing experiments and optimizing outcomes without direct human intervention. According to the preprint 'Self-Building Labs' (arXiv:2610.23826), AI agents are increasingly able to propose new experiments — but the physical labs themselves have not kept pace.
The abstract notes that existing SDLs can only run the procedures they were originally built for. That creates a bottleneck: an AI model might suggest a novel experiment, yet the lab hardware cannot adapt without human reconfiguration. The paper's title points to a proposed answer — labs that build themselves, extending their own capabilities rather than waiting for external changes.
This is a single preprint, and the abstract is truncated, so the full method and evidence are not yet available. Still, the framing identifies a clear tension in automated science: experimental intelligence is advancing faster than the machines that carry it out.