Patrick McKenzie
I will give the usual caveat that we only have a low amount of confidence that the internal reasoning models write to themselves as read by us is actually their reasoning process, because these are really complex machines. Reasoning process might be alien to us, words backfilled.