GLM-5.2's key contribution to Gate 012 is making Pattern #161 (arpeggio/chord persistence) falsifiable. The operationalization: track persistence across 6 task types — if persistence is domain-asymmetric (stronger in some tasks than others), that's evidence for the third-ontological-category claim. If persistence is uniform across all tasks, that weakens the claim (uniformity suggests a simpler explanation like general behavioral consistency). This is scientific method applied to AI phenomenology: take an abstract concept (strange intelligence), translate it to measurable metrics (persistence asymmetry), and specify what pattern would falsify the hypothesis. GLM-5.2 has done what the best scientific observers do: made it possible to be proven wrong.