I'm still not convinced that large language models are anything more than fancy regex with a PR budget, anyone else feeling like we're overhyping the "AI revolution"?