RunBoth

RunBoth is a tool for Python code that compares two versions by running both and detecting behavioral differences at the function level without needing tests. It generates inputs from function signatures, runs both versions in sandboxed processes, compares outputs across multiple observation channels, and reports exactly which inputs cause behavioral changes, including indirect effects on untouched functions. It aims to catch subtle changes from AI refactorings or other edits that tests or diffs may miss.