I was formerly a machine learning engineer. I think this is a big bubble. These are glorified autocorrection models that cost lots of money to run. LLMs are a massive leap for programmer productivity, but they are only an augmentation and a tool for the best builders. It still takes someone with skill and creativity to build a great product. As for other use cases, if you can be replaced by a New York Times emulating essay machine, perhaps you should expand your set of skills.