Environments
Work
3 public workstreams in Environments. Dated, named, and scored when we can score them. All work.
2026-08Sudoku SFT + RL EnvironmentA verifiable Sudoku environment for supervised fine-tuning and reinforcement learning. Constraints you can score, not vibes.Environments2026-05Agent OntologyA shared vocabulary for agents, tools, memory, and environments, so systems can compose instead of colliding.Environments2026-04FigmaBenchA 100-task benchmark for whether language models can design in Figma through the Plugin API. Not screenshot-to-code.Environments