Nuclear regulation runs on demonstrable safety cases, and the word doing the work is demonstrable. An empirical claim — we ran this controller for ten thousand hours and it behaved — tells a regulator about the past under the conditions that happened to occur. A certificate is a different kind of object: a proof that a stated property holds over a stated envelope, checkable by the reader without rerunning the experiment. That difference is why provable-by-construction is a licensing posture rather than a performance claim. It does not say the system is good. It says here is the property, here is the envelope, here is the check, run it yourself. For an unattended plant, where nobody was present to observe the past, an artifact that does not depend on having been there is worth disproportionately more.