Ω_t specifies a drift operator for the capture registry computed only over matched pairs — same address, same surface, two dates — and refuses two tempting alternatives on stated grounds: regression on source count is sign-ambiguous, and field-mean-by-date measures the sampler rather than the field.
Trial OMT-001 is pre-registered before the treated domains resolve, with three arms — treated, archive-control, field-control — and difference-in-differences required for attribution. The treated baseline was captured by chance hours before registration.
A retrospective prior is recorded explicitly as NOT-evidence, with five reasons it fails to be a finding: unmatched intervals, selection rather than sampling, and others. The instrument's value is that it can return nothing, and the pre-registration is what makes that outcome legible.