Yes, if there are already many worked out examples of a problem existing on the Internet, then a large language model can emit one of those examples.

But you see, no one pays me to write tic-tac-toe games, precisely for that reason.

A year ago, when this article was written, I would have been less skeptical. But so far, every time I use an LLM to write anything non-trivial, it is simply wrong, and it takes significantly longer to correct than to do it right from the start.