If I'm reading this correctly, this person had an LLM make up a bunch of code, then had another LLM make up some text purporting to analyze the code. The post doesn't mention actually running any of the code to see if it functions, nor even reading it, much less actually analyzing it. Somehow, this recursive LLM-ness is supposed to tell us something interesting about architectural style?

I am sure you can persuade an LLM to tell you anything you like about a piece of code, especially if it's not even real code, just something made up by a previous LLM.