An AI agent selected a gene therapy, a brain target, and surgical coordinates for a mouse model of epilepsy, then performed the injection itself, according to Gabriel Ocana-Santero, cofounder and CEO of Exin Therapeutics and a PhD student at the University of Oxford. He described the experiment in a thread on X.
The team gave Claude Sonnet 5 a list of 19 possible gene therapies, an epilepsy mouse model, and access to a custom motorized surgical device with cameras. Its only instruction was to reduce the mouse's seizures.
The agent chose a therapy that boosts a specific inhibitory signaling pathway. It picked a brain structure that Ocana-Santero says the scientific literature identifies as the origin point for this seizure type. It rejected a second target region, arguing that its effects were unreliable and had failed in one study against this seizure type. It derived injection coordinates from three scientific papers, cross-checked reference points, and averaged the values. It also set the injection volume, speed, and wait times.
During the procedure, the agent reviewed the surgical field through a camera, made 32 observations, tracked the injection, and reported whether it saw bleeding, per Ocana-