Comment by theteapot

> It helps to know that LLMs don’t “reason”. They predict ..

Semantics. Prediction is the training objective. The ability to reason can be, and very arguably is, an emergent property of that.