"I don't think that the details of which particular Turing machine is or is not universal are all that significant. What is significant, though, is how widespread universality is in general. And currently the only way we have of establishing that is to do the hard work of proving the universality or non-universality of particular Turing machines."

I smell a rat. His "Universality" and how often it occurs is what he has yet to prove.

He proves it in all kinds of examples (I think NKS is full of them). Other than that, he is just putting forward a hypothesis that some people might find interesting to consider. What is wrong with that?