> The code they [LLMs] produce is often fine. It works. It passes tests. It might ship as-is. Often is not goo enough. I always carefully review AI generated c

> The code they [LLMs] produce is often fine. It works. It passes tests. It might ship as-is. Often is not goo enough. I always carefully review AI generated code and sometimes find and fix bugs. Apparently (according to a Google survey I think?) AI's produce 1.7 times the number of bugs that a human typically does.