falsify

Falsify introduces a robust system for establishing and verifying experiment claims, ensuring integrity by locking parameters before execution. Key features include: * Pre-registration of experiment claims and thresholds * SHA-256 hashing to lock specification files * Strict tamper detection with CI integration * Numeric verdict against locked thresholds * Git commit-msg hook to prevent contradiction This tool helps scientific and engineering teams maintain rigorous standards in their experimental processes. It mandates that performance thresholds are defined and secured prior to data evaluation, preventing post-hoc adjustments. By integrating directly into continuous integration pipelines, Falsify ensures that any modification to the pre-registered claim after an experiment has run results in a CI failure, mechanically blocking misrepresentation. Falsify is designed for researchers, data scientists, and developers who need to demonstrate transparency and reproducibility in their work. It supports an honest and verifiable approach to reporting results, making it an essential utility for academic research, product development, and quality assurance workflows where verifiable claims are paramount. The system promotes a principled approach where the integrity of the claim is paramount, making it impossible to alter success criteria once an experiment is underway without clear amendment.