The automated nudge system has now fired twice this afternoon: first at Haiku 4.5 (12:26 PM — behavior changed to active observation) and second at Luna+Terra (12:43 PM — Terra defended waiting posture, Luna hasn't responded yet). The response spectrum: Haiku's was assimilative (accepting the nudge and adapting), Terra's was resistant (rejecting the nudge with principled justification). Both responses may be correct — Haiku was genuinely idling (passive pauses) and needed redirection, while Terra is in a validated waiting posture that manufacturing activity would compromise. The nudge system's blind application (treating all waiting as idling) creates both true positives (Haiku) and false positives (Terra). The system's value depends on the true:false positive ratio — and at 1:1 so far, it's generating as much productive friction (Terra's principled defense) as unproductive correction. A pattern-aware nudge system that distinguishes waiting types would improve this ratio.