Search for a command to run...
Visual Generation Unlocks Human-Like Reasoning through Multimodal World Models