Comment by gordonhart

Honestly curious, have you seen agents succeed at this sort of long-trajectory wide breadth task, or is it theoretical? Because I haven't seen them come close (and not for lack of trying)