I think that people figure out what LLMs are capable of doing well pretty quickly after they start using them. I'd say that their capabilities are underestimate

I think that people figure out what LLMs are capable of doing well pretty quickly after they start using them. I'd say that their capabilities are underestimated by many people who once tried to use non-reasoning LLMs to do things that reasoning LLMs can do very well today.