A model can predict what might happen next and still not know what to do.
That gap is where I keep circling: action-relevant state, context faithfulness, and whether a world model is honest about what it does and does not know.
fragment / 2026-06-30
A working note around robotics, world models, and the difference between predicting frames and choosing actions.
A model can predict what might happen next and still not know what to do.
That gap is where I keep circling: action-relevant state, context faithfulness, and whether a world model is honest about what it does and does not know.