It was actually much less anti LLM than I was expecting from the beginning.

But I agree that it is self limiting to not bother to consider the ways that LLM inference and human thinking might be similar (or not).

To me, they seem do a pretty reasonable emulation of single- threaded thinking.