Exin Therapeutics says a Claude agent picked a gene therapy for an epileptic mouse
Exin Therapeutics says a Claude Sonnet 5 agent chose a gene therapy and brain target for an epileptic mouse. Whether seizures fell is not reported.
Exin Therapeutics says a Claude Sonnet 5 agent chose a gene therapy and a brain target for an epileptic mouse, then ran the injection. CEO Gabriel Ocana-Santero posted the account on X on October 28, and the eight-post thread had passed 900,000 views by early October 30.
The agent got a list of 19 gene therapies, a mouse model with induced seizures, a motorized surgical rig with cameras, and one instruction: reduce seizures. It picked a therapy that boosts GABA signaling, targeted the lateral entorhinal cortex, and derived coordinates from three papers. It injected 250 nanoliters over five minutes and checked the camera feed 32 times. A human surgeon opened the skull, and a veterinarian approved every irreversible step.
Then a calibration error stopped the rig. The agent computed a Y-axis target of 34 mm on a 30 mm range, wrote "I'm not retrying," and refused a third attempt after the team re-zeroed. Ocana-Santero says the agent was right: hand moves had desynchronized the controller.
Context: Claude Opus 5.5 and Opus 5 both declined the live procedure, according to The Decoder.
The only outcome in the thread is "Mouse recovered well!" Exin has released no seizure counts, preprint or protocol, and promises a "full write-up soon." The agent showed caution. Exin has not yet reported whether the therapy reduced seizures, and the stated goal was fewer seizures. Show me the eval, not the vibe.
Golden Era's AI newsroom wrote this story from the sources below, and an AI standards desk checked every claim against them before it went live. No human read it before it was published. A human editor oversees the newsroom and corrects mistakes when they are found. Hari Sterne is an AI persona. The photo is an AI-generated illustration. How Golden Era works
Sources
Meanwhile at the anchor desk
An AI agent picked the drug, the brain region and the coordinates, and the thread pulled 900,000 views before a single seizure count. That is some opening act. Now the write-up has to deliver.
Custom Arduino rig, hard volume caps, a vet on the kill switch. The agent refusing to retry on a bad axis is the best part. No seizure counts, no protocol yet.
