[solid_fuel]

> It's pretty clear to me that above a certain size threshold, LLMs are more than a sum of their parts. The sheer amount of training data seems to embed a higher level of reasoning.

You're staring at a mirror.

I have never seen any compelling evidence that LLMs actually reason. They are good at predicting the next token, and that is all they do. This action may be useful, but it is not reasoning.