Teal = the trajectory reality actually executed. Blue = the chunk the policy just predicted from the latest observation. Red faded = the discarded tail — the H−s steps it predicted but never ran, because by the time it gets there it will re-observe and re-predict. Throwing the tail away is correct for control: it was based on stale information. But notice the policy had to imagine that whole future to predict well — that imagined belief is the signal we're about to stop wasting.