ScienceHuman-AI collaboration reportedly cut diagnostic workload by 35% in CRISP testing
A September 14 science brief describes CRISP’s results during surgery. It also says MatBrain generated 30,000 catalyst structures and narrowed them to 38 promising candidates within 48 hours.
TLDR
According to a September 14, 2026 science brief, researchers reported high diagnostic accuracy for CRISP in a prospective cohort of more than 3,000 patients. The brief says human-AI collaboration reduced diagnostic workload by 35%, while the system informed surgical decisions in 92.6% of cases. It identifies testing across more hospitals, scanners and clinical teams as the next step.
The same brief describes MatBrain combining specialized reasoning and tool-use models for materials research. In a catalyst-design demonstration, it generated 30,000 structures and narrowed them to 38 promising candidates within 48 hours, the brief says. One candidate underwent experimental synthesis and testing.
Combined views
120
1 Source, first seen 10d ago
Human-AI collaboration reportedly cut diagnostic workload by 35% in CRISP testing
A September 14 science brief describes CRISP’s results during surgery. It also says MatBrain generated 30,000 catalyst structures and narrowed them to 38 promising candidates within 48 hours.