Models that predict how the world evolves, not just what a single frame looks like.
๐ฅ Video Generation
Frame-by-frame plausibility isnโt the same as a coherent trajectory - why per-frame models fail at video, and what actually has to couple frames together.