Why the same codebase should always produce the same audit score

There is a failure mode in AI-powered analysis tools that does not get talked about enough, and we...