Recent AI model progress feels mostly like bullshit — LessWrong
www.lesswrong.com
About nine months ago, I and three friends decided that AI had gotten good enough to monitor large codebases autonomously for security problems. We s…