[habitue]

> ChatGPT cannot give an answer that is the result of a "reasoning" before laying out the "reasoning".

This is slightly too strong a statement: it can give an answer before reasoning it out, but it only gets a single forward pass of the network to calculate that answer, so it has to be a simpler kind of answer or very obvious reasoning (intuitively, imagine it can only take "one logical step" in a forward pass).

Its answers get much better if it uses the context as a scratch pad to write down its thinking from previous passes, this is where Chain of Thought (CoT) comes in. The way language models work is they pass the output to the input over and over, each time generating one token. This means the context is really like a scratch pad recording its previous thoughts.