Developers can keep gameplay explicit and debuggable while using generative visuals to vary a world's appearance. Core capabilities include Explicit engine-owned world state, Coding-agent rule updates, Low-poly proxy conditioning, and Dynamic captions.
Public code or model weights are linked from the project, with setup instructions for running or adapting the system. Additional characteristics include Multiple appearance references, Camera-indexed spatial memory, One- or two-step rendering, and Released code and game assets.
