There is no magic to coding, any sort of coding. Humans learn how to do it by learning the rules. A coding agent can learn the rules, easily.
I have a suspicion that a lot of the stuff people use AI coding agents for would be served just as well by, say, the WYSIWYG editor from Visual Basic 6. But we threw away that kind of technology long ago and settled for Reactslop, and now you need an LLM to write the Reactslop.
The agent will already have seen enough super optimised code and read enough material about it. It will read the entire repo you're in, and understand how the code should be structured and written to work efficiently. And anything it doesn't get at first, you write about in the AGENTS.md file and tell it.
Then it works. If you can teach a human to write specialised 3d rendering code, you can distill the same teachings to an agent in text form, and it'll work.
I am supremely confident of this, and I bet that neither you nor the author of the article has bothered to try properly.
American Fuzzy Lop usually manages to generate valid files of any format you want, using a genetic algorithm that the author didn't even call ML. AlphaGo trained against itself.
Such things aren't impossible when you can automate the reward function, though you might have to come up with novel techniques.