A Dated, Graded Ledger of the Testable Predictions of a Vacuum-Tied MOND Framework, with Its Confronted Record and Retracted Claims (Working Note)

Abstract

Working note, version 1.0 (2026-10-11). AI-assisted compilation; not peer reviewed. Nothing here says the data favour the framework; kappa is fitted, not derived; the cold energy's mass is still required; the theory is not closed. We collect, in one dated ledger, every testable prediction we could find in the working record of a programme built on the law a0 = kappa c sqrt(G rho_Lambda) with kappa = 1/2 fitted (two footings, 9.3603e-11 and 1.1312e-10 m/s^2, never pooled). For each of 98 future tests the ledger gives the framework's number, the rivals' numbers, the result that would count against the prediction, and a grade: K1 (fixed before the data exist, by frozen criteria or a sealed SHA-256 hash, with a date and a stated kill condition; 8 tests), K2 (sharp but capped by systematics, conditional, or nearly shared by the rivals; 11), K3 (open: no usable data, no date, or a forecast only; 65) and K4 (speculative or numerology-era predictions, including the primordial tensor-to-scalar ratio r = 0.0149, that could still be checked but have no derivation and are mostly retracted in the record; 14). The dated releases ahead are the ALMA cubes of 2026-11-06 (U4_27928) and 2026-11-28 (GS4_24110) and Gaia DR4 on 2026-12-02 (wide binaries; the Milky Way vertical force and outer curve, sealed in advance). A section sets out the neutrino sector in detail: the forced vacuum-energy scale (2.2395 meV), the unforced identification of the lightest neutrino mass with it (a fit by inversion that the record itself calls generic), the spectrum on both footings, the DESI DR2 checks, the Lean-certified implications of the lepton chain, and the closed doors. The paper also lists 51 confronted results, including the failures (structure growth on the gravitating field at 512^3, the Milky Way outer slope, SLUGGS ellipticals, Coma ultra-diffuse galaxies and others), 15 public deposits with their status, the claims that must not be cited, and what the framework does not predict. Rows are generated from one data file by a committed script and audited by a second (PAPER46_build.py, PAPER46_audit.py: 10 checks, including that the sealed-hash prefixes quoted match the committed hash files). The compilation drew on nine read-only sweeps of the programme's repository; some lane folders were only keyword-screened, so the list is thorough but not guaranteed complete, and release dates for Euclid, Rubin, JUNO, LISA and 4MOST come from the record's own inventories and were not independently verified. Rows taken from other AI models' exploration folders are labelled non-authoritative. AI assistance: Anthropic's Claude.