Accelerating scientific discovery with Co-Scientist.

Co-Scientist, an AI system with multi-agent architecture, generates novel biomedical hypotheses, leading to validated drug candidates and therapies for acute myeloid leukemia.

• Why it matters: Accelerating scientific discovery is crucial for addressing complex biomedical challenges, yet current hypothesis generation relies heavily on manual effort and intuition.
• What they did: The system uses a scalable, asynchronous multi-agent framework combined with tournament evolution to produce and refine hypotheses based on scientific evidence and research goals, validated in three biomedical applications.
• The result: Co-Scientist improved hypothesis quality through test-time compute scaling, identified new drug-repurposing candidates, and demonstrated potential to speed up discovery processes, enabling AI-empowered scientific breakthroughs.