This is so true. That feeling when you're debugging an issue in your code, only to rule out all the possibilities and have to expand your search to include not-your-code: the tooling, the libraries, the operating system, the hardware. This is the worst feeling in debugging, when the problem and the solution /might/ be outside of your control. With LLM/GenAI, the primary problem solving surface is nondeterministic by design, very much out of your control, and you will likely be gaslit into believing that you're just not hitting it right. Like a tube tv that has to be slapped just the right way to restore proper reception. (Sorry, bots, no offense intended.)
> What’s next, TPS reports?
I bet LLMs are really good at producing TPS reports w/ cover sheet.