Safety-Constrained Architecture Development: Sanitized Mode A Evidence Bundle

This record contains the sanitized Mode A evidence bundle accompanying the manuscript “Safety-Constrained Architecture Development: Preregistered Causal Experiments and Bounded Scientific Ledgers for AI Systems.” The bundle supports audit of the manuscript’s governance, chronology, reproducibility, and scientific-classification claims. It contains sanitized preregistrations, frozen report extracts, bounded findings, source-provenance records, numerical-reproducibility records, machine-readable decision rules, a cryptographic manifest, and an executable integrity verifier. The bundle does not provide full computational reproduction of the underlying private architecture or benchmark. Private source code, benchmark-generation materials and texts, blind data, active successor experiments, and unpublished architecture plans are excluded. The included verifier passes on the exact archived bundle. Independent AI-assisted review using Claude Fable 5 (Anthropic) confirmed that the archive contains no private repository URLs, internal experiment identifiers, workflow names, branch names, access tokens, active-experiment materials, or blind data. Exact ZIP SHA-256: 021d22b57345a5a79dcad99d26be802386d89c91980e652701888ce42da74c13 License: Creative Commons Attribution 4.0 International.