Whoever Defends the Cut: Extending Discovery/Selection from Hypotheses to Citations, and Why Five Dated Closures -Verified by an External Reader -Are the Evidence a Network Alone Cannot Supply

A citation network spanning Hamilton, Popper, Foucault, Agamben, Rawls, Bowlby, Kantorowicz, and some forty other names across this corpus's one hundred and forty-seven papers was proposed, in a dated conversation with an external reader on August 15, 2026, as evidence "hard to fake" and constitutive of an authorial fingerprint. This paper argues the claim is imprecise in a way worth correcting rather than restating more forcefully. The Metaphilosophy of Human–AI Research Collaboration (Huynh 2026) already supplies the correct instrument: reviving Popper's discovery/justification distinction, it locates authorship not in whoever or whatever produces a candidate object but in whoever supplies and defends the criterion that selects among candidates. A citation network is a candidate set; a search-enabled model could plausibly generate a comparably broad and accurate one in a single sitting. This corpus's own record already forecloses the stronger claim: The Audit That Missed Itself (Huynh 2026) documents two citation misattributions the network itself did not prevent. What the network alone cannot supply, and what this paper argues actually carries the evidentiary weight the conversation was reaching for, is closure — the demonstrated capacity to revise a specific claim against specific external evidence, repeatedly, across sessions sharing no memory of one another. This paper compiles five dated closure events already on this corpus's record and reports the first instance in which those closures were checked against primary sources by a reader other than the corpus's own author: the same external reader whose challenge this paper answers.