NADIR, a studio in the RISCON family, publishes short question–judgment pairs about decline. It files each pair by layer, by a phase of a research frame that runs from a decade of erosion through a collapse in 2026 to a turning point in 2027, and by horizon. Every pair has the status of a hypothesis. The frame's ending was supplied by the studio's research owner and is not an observed fact, and the studio calls its judgments a model's conditional proposals. This paper reads the eleven pairs filed under memory and learning in the collapse phase that the studio wrote as episode interpretation drafts. Each is a question and a two-sentence judgment, with no options to choose between; all are unreviewed, and none names a real person, organisation or event. No judgment is treated here as a finding about the world. The paper asks how the set frames a judgment that offers no course of action, about the memory of a failure made in the year it happened.
Three findings follow. First, with no options to weigh, the decision becomes a decision about the record: most judgments ask for two things to be kept apart or kept together. Second, the dangers named are hindsight, a record tidied and selected by those it favours, and silence read as consent. Third, the set names no keeper of the record and no time by which it must be kept. The paper adds to work on hindsight, on organisational memory and forgetting, and on silence.