Yeah, that's basically at the intersection of pretrained LLM and tool-use.

→ View original post on X — @rasbt