Registered falsifiable predictions from a single-coupling counting framework: the waxion (263.74 MeV), the atmospheric mixing angle, and reproduction code

Version 2.1 (2026-09-24). Paper-text corrections to version 2.0; the corpus, formula register (same SHA-256), predictions register and receipts are unchanged. The abstract of paper P-II now states that the waxion mass is unexcluded, that its canonical coupling (2.0×10⁻⁴ GeV⁻¹) is excluded by preliminary Belle II 408 fb⁻¹ data, and that the lower half of the coupling band remains allowed; the previous wording ("currently unexcluded", citing a BESIII limit retired on 2026-08-01) contradicted the paper body and the register. P-I's corpus size is corrected from 536 to 541. Version 2.0 (2026-09-24, corpus v541). New version of this record. Version 1.0 (2026-06-13, corpus v528) was the only earlier public version. This version adds the full corpus snapshot (541 formula entries under append-only governance: 13 added, none removed; 37 existing fields changed value, mostly status downgrades from self-audits, all listed in the package), a DOCTOR report (13 PASS / 0 WARN / 0 FAIL), current drafts of papers P-I and P-II, and every analysis receipt from 2026-06-13 to 2026-09-24. One forward card outside the public register (m_Omega_cc) was killed on 2026-08-11 by LHCb’s announced observation of the Omega_cc+. Three third-party full-text extracts are excluded and replaced by citations with SHA-256 hashes. The predictions register for this version is REGISTERED_PREDICTIONS_20260801.md (P1–P12); where it differs from the version 1.0 summary below, the register governs. Start with README.md. Version 1.0 description: This record registers, with explicit kill conditions, the predictions of a phenomenological framework in which flavour-sector observables take the closed form X = X0 + n·alpha(mu), with zero continuous free parameters. The point of the deposit is to place the numbers, the dates, and the reproduction code on the public record before the relevant experimental analyses appear. Five already-matched constants (zero fitted parameters; PDG 2024 / CODATA): sin^2(theta_W) at 0.003 sigma, sin^2(theta_12) at 0.023 sigma, alpha_s at 0.04 sigma, sin^2(theta_13) at 0.15 sigma, and m_tau/m_e at 0.79 sigma. A full look-elsewhere accounting gives a joint chance probability of roughly 2-3 x 10^-5. Three forward predictions with kill conditions:(A) A 263.74 MeV pseudoscalar (the "waxion"), g_agamma-gamma = 2.019 x 10^-4 GeV^-1, c-tau = 53 microns, decaying to two photons. Currently unexcluded; testable by the Belle II full-luminosity three-photon analysis.(B) sin^2(theta_23) = 1/2 + n·alpha with n in {3, 6, 8} only; n = 7 is forbidden.(C) sin^2(theta_12) = 0.307297 at JUNO final precision. Every quoted number is reproducible to 30 digits from the included scripts. Fifteen further derivation scripts are deposited in encrypted form with their SHA-256 digests committed in HASHES.txt. See REGISTERED_PREDICTIONS.md for the full self-contained note.