MedOS brings agentic AI, XR, robotics to surgery

MedOS is attempting the "full stack" play that surgical AI has long promised but rarely delivered: perception, reasoning, and physical action in one system rather than bolted-together point solutions.

• –The dual System 1/2 architecture is the real technical bet — distilling slow chain-of-thought reasoning into fast edge inference, bringing AI latency from 1-2 minutes down to near real-time for intraoperative use
• –The cross-specialty uplift numbers are the boldest claim: a cardiologist reaching 89% accuracy in oncology is either a genuine breakthrough or the most optimistic preprint framing in recent memory — peer review will be the test
• –The MedSuperVision dataset (85k+ hours, annotated by 1,882 clinical experts) is a meaningful moat — data quality at this scale is the primary bottleneck for surgical AI
• –Combining XR tremor elimination (0.5 px static jitter vs. 13.6 px baseline) with robotic guidance puts MedOS in direct competition with established incumbents like Intuitive Surgical and Caresyntax
• –NVIDIA GTC 2026 presentation scheduled for March 18 signals the team is actively courting infrastructure partnerships