You should read more work of Francois Chollet, you will understand what reasoning actually is.

What o1 did is that it took Reaoning CoT and baked it as memorization into fine tuning, along with CotT during inference.

Now we as humans also use memorization and heursitics to answer most of the time. If I ask you 4x4, you will answer that from memory, but for bigger numbers you can easily extrapolate the pricnicpl, but that's not what even o1 does.

Every form of reasoning can be memorized as CoT, but the fact that o1 doesn't generalize well on ARC challenge is a proff that it is definitely not reasoning.

Let me give you one more examples, the performance of LLMs changes in MCQs, if i change the order of the options, we a good reasomningsystem won't have problems like this.