Note by @fastfinge

I do ironically enjoy it when a company releases a new and improved #LLM, and suddenly some extremely specialized and specific tasks that I use LLMs for perform drastically better. And suddenly they can complete the exact examples I used to provide in my prompt, all by themselves. And seem to perform the task in my exact style, even though I'm not giving them my prompt examples anymore. Hmmm, it couldn't be that they trained on my prompt data, could it? Even though they said they don't do that? Nah, of course not! They'd never!

Oh well, at least someone, somewhere, spent several billion dollars to make something I do once a week slightly easier.