All projects
3D-Grounded World Model
Lead. Preliminary research on a new world-model architecture: a first bridge between 3D models and video models.
- Type
- Exploratory research
- Context
- Stealth Startup
- Year
- 2026 – Present
Highlights
- Explore joint training of a VGGT-Omega 3D geometry model with a video diffusion model, so generated video is anchored to an explicit 3D scene state that can be checked against geometry rather than pixels alone.
- Prototype Spatia-style spatial-memory operations (scene read/write, camera-conditioned rendering) as the first step toward a 3D-grounded, verified world model.