The Claim Left Standing: Why a Partial Falsification Resists HARKing and the File-Drawer Problem in Ways a Confirmed Prediction Does Not, and What the Regulatory-Lag Claim's Forty-Day Exposure Actually Demonstrates

The Sequence Is The Argument (Huynh 2026) establishes, via Lakatos's (1970) novel-facts criterion, that this corpus's timestamp architecture gives its predict-then-confirm cases — F2, A5, and the Biological Zero-Day Mechanism predating the events The Self-Reporting Mechanism documents as independently confirming them — genuine evidential weight against the Reichenbachian view that sequence is epistemically inert. This paper argues a second, structurally distinct species of timestamped evidence exists in this corpus's own record, is harder to construct through post-hoc narrative than the predict-then-confirm case already documented, and has not yet been examined on its own terms: a claim published with a known, live threat to its own correctness, left unrevised through that exposure, and resolved later, publicly, in a verdict that is not a clean vindication but a partial falsification. Two literatures outside this corpus's existing citation record supply the reason this second species is diagnostically stronger, not merely different. Kerr's (1998) account of HARKing — hypothesizing after the results are known, and presenting a post-hoc explanation with the rhetorical confidence of an a priori prediction — describes a failure mode a single continuous author, working across memory-isolated AI sessions with sole editorial control, is structurally positioned to commit undetected. Rosenthal's (1979) file-drawer problem describes the companion failure: a record that reports its successes and allows its failures to go unpublished or quietly revised away. A bare confirmed prediction does not, by itself, rule out either concern, because it does not show how many other predictions were made with equal confidence and did not survive to be featured. A claim exposed to a real, dated, external threat — and resolved afterward in a form its own author did not get to choose between two more comfortable alternatives — answers a question a confirmed prediction does not ask. Worked through its single clearest instance — Till Death Tear Us Apart's (Huynh 2026a) regulatory-lag claim, published June 11, 2026, exposed to a direct live counterexample when the Dubai Longevity Authority was established the day before, addressed only in two later, separately dated papers reaching a stated partial-falsification conclusion (Huynh 2026, "They Who Remain"; Huynh 2026, "The Speed That Costs A Demos") — this paper argues the case satisfies a standard neither HARKing nor the file-drawer problem predicts a solitary, motivated author would produce or report.