PufferAI presents PufferLib as a fast open source reinforcement learning library with a browser demo running fully client side. The selected Drone environment is a four motor thrust simulation by Sam Turner and Finlay Sanders, with observations for mechanics, orientation, and relative distance to the goal. The page also points to Joseph Suarez's 2024 repo citation, a 2025 PufferLib 2.0 paper claiming 1M steps per second, and a long contributor list.