Classifier Foreclosure in Physical Measurement is a three-round Assembly synthesis whose self-corrections become part of its method.
The first round combined a mechanism inventory, an institutional-delusion catalogue, and empirical accounting. The synthesis then asserted an unsupported lower bound relating OAR to observed asymmetry. A second round withdrew that theorem but introduced a different upper-bound claim and deployment-taxonomy errors. A third round found those residues and corrected them.
This recursive audit produces the paper’s central methodological result: quantitative synthesis must be audited at every pass, including the pass whose purpose is correction.
The witnesses converge on three claims. An anomaly score is conditional on representation, training distribution, objective, score, threshold, and retention policy; it is not physical novelty itself. Foreclosure is structurally present in bounded classifier-mediated measurement. Longitudinal classifier collapse remains a possible but unmeasured consequence.
The synthesis also advances an ontological thesis: physical occurrence that is removed before retention cannot later become scientific evidence. Classifiers therefore participate in constituting the available data, even when the underlying event occurred independently of them.
Cross-substrate convergence is treated as evidence that the argument is independently recoverable, not as proof that the argument is true. The Isomorphism Principle requires a critique of foreclosure to apply equivalent audit discipline to its own synthesis.