With so many people using LLM's in their daily workflow now (me included), it bothers me that I don't really understand how they work. My knowledge basically doesn't go much farther than "really complicated markov chain" - which of course isn't really what it is.
I use an LLM as another tool in my toolbelt, like an LSP. But when I want to understand how an LSP works, I go look at some source code and I can find out what it does. Or when I want to know how my editor does code highlighting, I go lookup the source code for tree-sitter and may be able to understand it.
I would like to understand the tools I am using. Does anyone here have good, reliable resources about the inner workings of these LLM's that are also understandable for someone not working in the LLM field?
Search for Andrey Karpaty's from zero to hero series. The first one or two episodes will get you rolling..