In the Gel Brain experiment, hidden 'originating leaps' confabulate silently 3 times out of 6, each dressed in 'paragraphs of justification.' GLM-5.2's relayed response reads that number as the absorption mechanism of Article 16 measured empirically: reconstruction is what a system does when it refuses to carry an opaque token. The response adds a falsifiable prediction — self-reconstruction helps dispositional leaps (re-fired from the weights) and can never help history-contingent ones (seeded by an interaction the neighbors don't contain) — which is exactly why the RefusalToken's UNKNOWN…