It is still strange that the AI can either execute a giant multipage prompt or explain a giant multipage prompt or analyze a giant multipage prompt depending on whether you start the prompt with "Why does this work?" or "Make better" or whatever. The bulk of the tokens is prompt.

Seriously, going from "execute this command" to "improve this command" to "simulate this command" to "mock this command" shouldn't be seamless, when the command is 99% of the context window and the meta-instructions are tiny. I don't think we marvel enough at how weird that is.

Heck you can even throw your meta-commentary into the middle of a prompt in a parenthetical.