C.26.1:4.1 - Activation boundary

This pattern is active only when the interaction both participates in the represented state and its output is being used as evidence, export, comparison input, or decision input as if it were passive. A passive-read, unjustified lossless-transfer, one-way-message, or ordinary-bridge treatment must be materially false for the current decision.

Ordinary influence is not enough. A meeting that changes attention is ordinary work unless the meeting output is later used as a passive reading of alignment. An API call that is a mutating operation by its interface semantics is ordinary service/API semantics unless the call result is used as a neutral state export. A feature flag that changes behavior is ordinary intervention unless the flag readout is being used as evidence of the state it changes.

Performative prediction is also an important ordinary rival. If a prediction, score, or metric changes behavior because people act on it, but no incompatible probe frame, order-sensitive reading, contextual-probability cue, or instrument-like state/export support load remains, try performative-prediction analysis first; use C.16 for measurement, A.10 for source recovery and bounded reliance, and B.3 only for an actual named assurance claim. Keep C.26.1 only for the residual probe admissibility question.