Oh yes, of course! The method is nice, but yeah, it would be a bit weird if it could outperform LLMs on things that LLMs are good at.

→ View original post on X — @rasbt