confusion in that if you treat lambda calculus as computation instead of "mathematical truths", then following the proof in the opposite direction requires conjuring out of thin air a complicated construction that isn't obvious from the state you'd have