I said before and I'll say it again, there'll be the year in which ~100% of the code will be AI generated, immediately followed by at least 3 decades of desperately finding engineers to reverse/rewrite everything

There is a lot in how you word it though. "Double check for security gaps or exposed credentials" probably would have caught it. Claude is generally pretty good with interpreting even simple instructions like that. But yeah, always verify your outputs. This stuff isn't ready for blind trust yet.

Honestly, if we ever hit a skynet or I, Robot situation is going to be from someone like this who vibe codes and pushes into production without reviewing the changes that includes a fairly obvious new SubjugateHumanity method.

I hope this guy is not an engineer. Anyone who releases code that publicly exposes API tokens deserves to be sued.

womp womp