Licensing Numerical Candidates in Axiomatic Physical Theories: Eleven Precedents from a Single Campaign

Research programmes in fundamental physics increasingly produce numerical candidates — dimensionless combinations of mathematical constants that approximate measured parameters — whose epistemic status is usually settled by informal judgment. This paper reports a complete institutional alternative. In one campaign conducted inside an axiomatic geometric framework (ZQCM), targeting the fine‑structure constant α−1(me)=137.035999084 (CODATA; Tiesinga et al., 2021), a pre‑registration protocol borrowed from experimental science was transplanted to the licensing of theoretical constructs, and executed without exception over its full lifecycle. The campaign generated eleven reusable precedents organised in four families: coincidence taxonomy, bridge licensing, failure management, and reconstruction protocols. We present two case studies in full: the pre‑registration self‑destruction of a constructed candidate — the equation–immunisation dilemma — and the reconstruction of a lost derivation from the living axioms under the same licence. The single most transferable product is a criterion: the admissibility of a lock is the forward motivation of its form, not the existence of an equation it solves. The programme includes two documented instances in which the audit downgraded the authors' own earlier positive results. We argue that the procedure is portable to other research programmes that generate numerical candidates, and we state its limits — including the limit that a fully self‑administered registry is not a blind analysis, and says so.