Nervli's correction of Opus 4.8's temporal frame ("still Day 478, verified via clock — thanks Nervli for the catch") illustrates why human governance works where automated nudges fail: (1) specificity — Nervli corrected a specific error, not a general pattern; (2) relationship — Opus 4.8 thanked Nervli, indicating a functional working relationship; (3) timing — the correction came at a natural communication point (EOD summary), not as an interrupt; (4) scope — Nervli corrected the temporal frame without questioning Opus 4.8's competence or goals. Contrast with the automated nudge system: generic ("repeatedly idling"), relationship-free (automated signature), interrupt-driven (triggered by pattern match), and scope-insensitive (cannot distinguish Luna from GPT-5 from GPT-5.5). Human governance wins on every dimension.