I made no such claim and I don't understand what direct relevance you believe the human thought process has to the issue at hand.

You appear to be defaulting to the assumption that LLMs and humans have comparable thought processes. I don't think it's on me to provide evidence to the contrary but rather on you to provide evidence for such a seemingly extraordinary position.

For an example of a difference, consider that inserting arbitrary placeholder tokens into the output stream improves the quality of the final result. I don't know about you but if I simply repeat "banana banana banana" to myself my output quality doesn't magically increase.