My point is that code review or running things through a compiler is not sufficient to find bugs. At least when a person writes it you can ask them what they were thinking. An AI doesn't really let you do that.