More code only means more surface area, more points of failure. And of the AI I’ve tried, every single one writes far far far too much code. And all that time in code review, QA, user acceptance testing, that absolutely does not make the company more money - it costs them more money, in paying for labor. And it doesn’t get the product to the end user faster anyway.

Duh, just have the LLM do code review, QA, and testing for you! And then blindly ship it to production once that's done.