October 2025 Summaries

The initiative is expanding its open-source environments program to become a global hub for open evaluations and reinforcement learning (RL) environments, committing substantial funding through grants and seeking partners to support the mission of accelerating open superintelligence. In the past two months, the program has successfully crowdsourced over 400 RL environments and evaluations, with plans to scale this to thousands through a large-scale bounty program. The program offers two categories of tasks: Open Access for newcomers and Application-Only for more experienced builders, with bounties ranging from $100 to over $5,000. The effort invites organizations to sponsor bounties, suggest tasks, and contribute resources, aiming to establish the Environments Hub as a central platform for building, sharing, and accessing RL environments and evaluations. The program's ultimate goal is to enhance the open-source ecosystem, enabling the training of competitive and useful models while fostering collaboration among companies, academic researchers, and individual contributors.