> even having perfect prediction (which I contest with (1)) does not entail perfect control, just by continually computing the effect of, and refining, your X.