Can AI learn to rate surgery skills by watching hand movements?

Surgical training depends on expert feedback, which doesn't scale. ExpOS watches hand-tool dynamics from video to predict skill level and explain *why*—identifying which movements matter most. Trained on 221 student surgery videos, it combines hand pose tracking with temporal attention networks to pinpoint informative moments and behaviors. The strongest results came on fascial closure (r=0.778), suggesting the approach could enable autonomous practice and scalable feedback without constant instructor oversight.