what do you mean with the comparison to tool calls? like what changes if tokens are cheaper than tool calls? would you for example put the files into the llm instead of making it call grep?
sorry i really didn't understand that section 😅
added these couple sentences, do they help?
> Once models are cheaper than a tool, it becomes attractive to put models *in* tools.
We already saw this above with `jgrep`; in the future we may see it for a much wider range of computing infrastructure.
putting files in an LLM is a interesting idea!