Same here The best I've seen are loose explanations that fine-tuning is better for setting things like the tone of the response or teaching the model new styles of output – or helping with problems like questions-to-SQL I want deeper understanding than that

→ View original post on X — @simonw,