I don't understand why LLMs are supposed to be able to do things more quickly anyway. I've tried a couple of times to use Claude to help write stuff, and it si

I don't understand why LLMs are supposed to be able to do things more quickly anyway. I've tried a couple of times to use Claude to help write stuff, and it sits there for a couple of minutes "thinking" before returning some grossly incorrect code. What's the bit that's supposed to be faster?