Exin Therapeutics Lets Claude Operate Brain Surgery
Exin Therapeutics says a Claude Sonnet 5 agent selected a gene therapy, brain target, coordinates, and injection settings before operating a stereotaxic rig on an epileptic mouse. A veterinarian supervised every irreversible step, but the experiment remains self-reported, unreviewed, and its therapeutic effect is unknown.
This is a striking demonstration of agentic laboratory control, but the headline is ahead of the evidence: the real milestone is closed-loop experimental decision-making, not proven gene-therapy efficacy.
- –The agent selected from 19 therapies and translated scientific literature into a treatment plan and physical coordinates.
- –Human safeguards remained central: a vet performed the craniotomy, approved irreversible actions, and held an emergency kill switch.
- –A rig-coordinate failure required human code intervention, underscoring the fragility of current lab-agent interfaces.
- –No seizure outcomes, independent validation, full logs, or peer review are available yet.
- –Exin’s broader “in vivo lab-in-the-loop” model could accelerate neurotherapeutic discovery if reproducibility and animal-welfare controls hold up.
DISCOVERED
1h ago
2026-11-01
PUBLISHED
1h ago
2026-11-01
RELEVANCE
AUTHOR
mark_k