Well, before I dive into the cesspit of orange site comments, I want to say that I respect this guy for admitting publicly that he wasted a few weeks of time that his clients were waiting on him to dick around with AI and that ultimately that was a failure with little to show for itself.
But this plan... I've asked AI for some smaller, controlled code snippets. Not even this "give me a website" business some people seem to try, just "give me a priority heap" and it can't even build out the basics of the api. It takes some incredible arrogance to blindly trust AI code for days and days.
Especially in a language you don't know, and doubly especially in Go which has a beginner friendly reputation but is actually filled with footguns. Trying to generate a SAAS app using go, I'm sure it was filled with whacky channel usage that if you aren't careful and do proper global planning, you will tangle yourself all up in them. Not to mention it's nil-based error handling which requires you to write a very specific style if if-ladder error checks. Which LLMs are great at consistently applying rules, I'm sure you've noticed. I haven't used go in a few years so some of this may have been streamlined a little with syntax changes, but hey, guess who else was trained on years of slightly outdated Go code?