Why was Tanenbaum wrong in the Tanenbaum-Torvalds debates?

So in the debate microkernel vs monolithic the correct answer was - hybrid.

In the debate CISC vs RISC the correct answer was both, in different niches (x86 and ARM), and both groups included many features of the other (I've heard ARM isn't really RISC nowadays).

In the debate free vs commercial the answer is also both.

Seems like "truth is always in the middle" is a good heuristic.