This is a follow-up to this piece, which is a good summary of the paper Apple published (also linked from there) talking about the issues with large reasoning models. Really good piece for anyone following AI.