An open model predicting a robot's actions from a control signal. The corner panels are the action and hand pose it was given, everything else is imagined. Is this a world model, or just a video generator?
LingBot-Video, from Robbyant, open weights. You give it a first frame and an action signal, and it rolls out what it thinks happens next. World models are this sub's favorite argument, so I will just ask it plainly: does predicting future frames from actions make something a world model, or is that still far from the Dreamer or JEPA idea of one?