Forecast register with prospective decision rules

A register of fixed-number claims in the author's exploratory vacuum-mechanics framework, with formulas or frozen stated values, inputs, ordinary comparators, decision rules, dependencies, and a retained correction history. It contains 104 entries, including 59 live numerical candidates, 14 binary or bounded candidates, 17 conditional entries, seven historical exclusions, and one borderline classification; a separate coverage appendix records 59 claims or batches. These categories do not imply 104 independent tests. The 15 internal claim records from 22 September 2026, new draft rules from 7 October, and scoring qualifications from 8 October are kept distinct. A private freeze, draft upload or reserved DOI is not a public timestamp. Prospective scoring begins only at actual public release. All comparisons already present in the packet are historical, not successful prospective predictions. This is a transparent testing register, not a derivation or validation of a physical theory. Agreement with known measurements is not confirmation. Ordinary comparators, multiple variants, dependencies, model assumptions, unresolved ordering and covariance, and source rounding limits remain explicit. Lattice comparisons are simulations, not physical observations. The accompanying read-only checker reproduces stored arithmetic and selected scoring counterexamples; it is not an automated experimental-likelihood intake engine. Version 1.0. Public registration begins with publication of this exact version on Zenodo, not with its private freezes, draft upload or DOI reservation. The files preserve their as-prepared chronology and null registration fields; the publication event activates this version prospectively and does not turn any retained historical comparison into a successful prospective prediction. Text and mathematical data are CC BY 4.0. Python and Wolfram verification scripts are excluded from that grant and retain separate software rights as stated in SOFTWARE_RIGHTS.txt. AI-assisted research, computation, checking and editing do not constitute external peer review; Luqman Mahmood is responsible for the final claims and text.