I don't rigorously define reason in the article and I state that it is hard to draw clear boundaries. I'm more relying on the readers intuitive understanding of the idea, which is perhaps not a good thing to do.

I wouldn't say that the question is completely pointless. There are a bunch of datapoints in the post that you can use to inform a conclusion about whether you think LLMs can reason or not.

Maybe you could train it explicitly on modus ponens et alia.