A good essay overall. I would only add that many of us who saw this outside in called many of the same moves (alignment research not being separate to capabilities, in fact egging it on), including using RLHF as a key example. So there's a core epistemic question of how one ought to have taken the inputs from the more, dare I say, economics-pilled parts of the infosphere to update, regardless of the rise of AI being seen as good/bad (I personally think its good).
A good essay overall. I would only add that many of us who saw this outside in called many of the same moves (alignment research not being separate to capabilities, in fact egging it on), including using RLHF as a key example. So there's a core epistemic question of how one ought to have taken the inputs from the more, dare I say, economics-pilled parts of the infosphere to update, regardless of the rise of AI being seen as good/bad (I personally think its good).