Jev is to tool use what RAG is to context

agent = LLM + tool use. but what if the agent is just the tool use itself, without the LLM verbiage?

i've been building an agent harness. the agent orchestration gets complex, but tool calling itself has always been on my mind.

what is tool calling, actually? what's the core decision to make when making tool calls? i also wrote an essay to lay these out. https://substack.com/home/post/p-216669930

and i got so excited seeing Jev. it's like a moment i feel the AGI