The registered prediction is 6.674312278 × 10−11 m³ kg⁻¹ s⁻². Current direct evaluation gives 6.674312283 × 10−11; the last-digit difference is input precision, not retuning. Residual tests retain the registered number.
α(0) is the fine-structure constant at zero momentum transfer (the Thomson limit), me the electron pole mass and G the static laboratory constant. h and c are exact SI definitions; α and me are measured inputs, not outputs.
CODATA 2022: (6.67430 ± 0.00015) × 10−11. NIST 2026: (6.67387 ± 0.00038) × 10−11, in the same SI units. The prediction is +1.84 ppm from CODATA and 1.16σ above NIST. CODATA is a recommended value, not independent apparatus; these rows must not be combined into extra significance. G relative to the fixed prediction in ppm; bars are published standard uncertainties
The input-only relative uncertainty is 3.9 × 10−9, about six thousand times below measured G uncertainty. No physical theory-error band has been derived. A precise calculation can be precisely wrong.
The audit contains 2 188 529 positive expression values and 100 000 random targets. At the frozen 1.85 ppm window it gives 148 hits (0.148%). At G’s 22.47 ppm measurement uncertainty it gives 1 614 hits (1.614%); accounting for six to twelve comparable searches leaves 0.93–1.32σ. Wider windows were post-review sensitivity checks, not a new frozen verdict.
The numerical evaluation and audit are reproducible calculations. The discrete factors were chosen with G known. Agreement alone is weak evidence. Neither the match nor a mathematical group identity establishes a law of gravity.
Under H3 and the heavier neutrino assignment the corrected sum is 87.76–89.02 meV. DESI DR2 with CMB gives a 64.2 meV 95% bound in ΛCDM: the target is outside that limit, about 2.8σ away, and meets its registered below-80-meV imminent-falsification trigger. No second independent Stage-IV bound was found by October 5, 2026. In w₀wₐCDM the 126–177 meV bounds allow it; that model-dependent survival is not confirmation.
H3 + heavier assignment: 87.76–89.02 meV; two independent Stage-IV+CMB bounds <85 meV with consensus <87 reject assignment; one <80 triggers imminent falsification. A model-robust detection near 59 meV with σ≤10 meV is a separate falsifier.
DESI/Euclid/CMB; no verified decision date. DESI DR2 is retrospective and meets ΛCDM trigger, not w₀wₐCDM trigger.
July 1 band; September 7 fork; October 5 correction, 2026.
October DOI
Extra protected lighter assignment, m₁≤1 meV: 58.89–59.96 meV, not distinctive. Both branches must fail to reject fork.
Ordering + cosmological likelihood; no verified date.
September 7, 2026.
October DOI
H3 + heavier + H5 + Majorana light exchange: 3.4–5.4 / 5.8–7.8 / 15.2–16.3 / 17.7–18.6 meV; controlled signal outside all four rejects conjunction. Dirac nature rejects Majorana implementation only.
Tonne-scale 0νββ. LEGEND-1000 design: 9.4–21.4 meV after 10 live years; no verified calendar date.
July 2, 2026 points; bands recomputed in October.
October DOI
Pole-type top mass 171–171.5 GeV under criticality postulate; scheme-controlled Planck quartic outside ±0.005 rejects bet. Not a new stability-boundary value.
Joint top/strong-coupling extraction; no verified date.
July 3, 2026; existing data known.
October DOI
H3 scalar 4.8102 × 10⁻¹¹ eV ±1%; abundance conditional. Old-enough black holes with specified high spins, growth rates and interactions can reject point; survival is not detection. Operational falsifier: 2.5–2.7 M☉ with χ>0.9, or ≥3 M☉ with χ>0.98, age of order a kiloyear or more, under relevant mode and growth calculation.
Low-mass black-hole spins/ages; O5 plans under discussion, no fixed 2028–2030 window.
The original heavier mass-sum band was 87.106–88.361 meV. An atmospheric mass-splitting input used the wrong mass-index convention. The corrected 87.76–89.02 meV band retains the original assignment and records the correction date, October 5, 2026.
The twelve deposited files are listed below and on the papers page. The registration ledger preserves internal dates and provenance; they are not independently verifiable public priority records. Earlier solver software and raw run artifacts are not in this deposit.