GSC: A Pre-Registration Reproducibility Stack for Falsifiable Cosmological Models

Le résumé fourni par la source

An open-source pre-registration reproducibility stack demonstrated on a scale-covariant cosmology framework. The stack combines deterministic computational pipelines, cryptographically-signed numerical predictions, and a layered four-tier claim hierarchy (kinematic frame, phenomenological fit, RG ansatz, speculative extensions) to make speculative model-building falsifiable in operational practice. Ten worked predictions (P1-P10) cover BAO standard-ruler shifts, 21cm Cosmic-Dawn signals, neutron-lifetime experiments, CMB cosmic birefringence, strong-CP θ-bounds, Kibble-Zurek defect spectra, gravitational-wave-memory atomic-clock signatures, redshift drift, proton-electron mass-ratio constancy, and TeV blazar dispersion. Developed and audited via multi-LLM iterative hostile-review cycles (Gemini, Claude, ChatGPT). After two AI hostile-audit sprints, the framework's predictions mostly fail current observational data — which is itself the value proposition: pre-registration discipline catches errors before submission, retracts them explicitly, and updates the framework status transparently. The methodology is the primary contribution; the cosmology framework is a working case study. The stack is reusable for any model whose predictions can be expressed as numerical functions of well-defined parameters. Author affiliation: Monster Cleaning Ltd. — https://monstercleaning.com Repository: https://github.com/monstercleaning/gsc

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.