Position PaperThe work we think the field should be doing
A world model owes its world.
Generative video looks right for five seconds and then forgets. We argue the next generation of world models has to be physics-based, 3D-driven, and channel-committed — models that don't just predict pixels but maintain a world a robot, a creator, and an audience can all rely on.
Position Paper · MMXXVINo. 01