Polymath

In order for AI to better serve people and create value in the economy, agents need to act more reliably and autonomously across long horizons. We're building the infrastructure to get there. Current AI agents excel at tasks in controlled settings under human supervision. However, the highest value work unfolds over days or weeks, requires fluency across dozens of tools, and the ability to navigate changing conditions. Polymath creates production-grade RL environments where agents tackle realistic tasks over extended time horizons, and success is measured by real outcomes. We work with frontier labs to develop and scale environments that expand the frontier of what agents are capable of. We're starting with software engineering environments, but plan to extend to other domains with the same level of depth and nuance. We envision a future where reliable, autonomous agents can amplify what people can accomplish alone, enabling humanity to tackle the most difficult problems and accelerate progress.