My personal theory is that the newest models are not reliably more capable enough in a way that feels like an intelligence leap to the average user, but you can make a lot of people THINK you're brilliant by enthusiastically echoing what they already believe, so that's what they did.
I wonder if this is similar to how that experiment[0] where they attempted to domesticate foxes led to traits like floppy ears.
Which is to say, even when attempting to objectively select for "well aligned" behavior, human tendency to favor non-material signals of "friendliness" still leaks in.
