Google Research describes AI agents running closed-loop wet-lab experiments on aptamers and proteins

Google Research describes three projects that connect AI agents to physical wet-lab experiments. The VERID framework gives agents a live digital twin of a lab workcell and dry-runs protocols before robots execute them.

In aptamer design over three rounds, agent-driven hit rates reached 33.5% for myoglobin and 42.1% for α-thrombin, and the post reports that agents revised their own design rules after spotting a G-quadruplex artifact.