Kosmos: An AI System for Automated Scientific Discovery Through Iterative Analysis and Literature Review

Summary

Kosmos is an AI scientist system that automates data-driven scientific discovery through iterative cycles of literature search, hypothesis generation, and data analysis. Unlike previous AI agents that lose coherence after limited actions, Kosmos uses a structured world model to share information between data analysis and literature search agents, enabling it to run for up to 12 hours and perform up to 200 agent rollouts. The system executes an average of 42,000 lines of code and reads 1,500 papers per run, producing scientific reports with traceable citations. Independent evaluation found 79.4% of statements accurate, and collaborators reported a single 20-cycle run performed the equivalent of 6 months of human research time. Kosmos has made seven discoveries across metabolomics, materials science, neuroscience, and statistical genetics, including three that independently reproduced unpublished findings and four novel contributions.