Half of the AI patches don't work with Keith Hoodlet and @joshbressers

This paper proves exactly what I have been saying in numerous talks: fixing vulnerabilities properly still requires a clever human involved.

opensourcesecurity.io/2026/202