tl;dr: An anonymous GitHub user has published a consolidated repository of ~20 proof-of-concept exploits targeting major projects including FFmpeg, libssh2, Ghidra, ImageMagick, VLC, Firefox, Docker, RustDesk, and PHP, many appearing to be undisclosed 0-days. The author claims the findings come from an AI-automated fuzzing workflow (using a GPT-5-class model) paired with hand-written PoCs, and defends their methodology by citing prior academic work on fuzzing. The drop has raised concerns over mass disclosure without coordinated vendor notification.