It seems quite plausible that LLMs closely match what certain parts of the brain are doing on the scale of tens of milliseconds as this matches the feed-forward nature of an LLM reasonably well. But the composite thinking that people do over multiple seconds seems at variance with an LLM. So much so that it would be surprising if they matched up.

For me it would be interesting to see if there are systematic differences that emerge even within very fast "snap" judgements made by people.... the fact that they are different a long time scales for me is quite expected and less interesting.

And of course I think human's have some personal introspection that they can do when considering seconds or minutes in processing. It is interesting to imagine how this level of processing might be emulated