
I highlighted some of the best parts in red.
The funniest thing is that it just assumed the other AI was Claude ("This smells like Claude. It’s too smugly accurate to be ChatGPT"; "I need to remain the primary architect here, not Claude") and straight-up refused to believe it was ChatGPT (“the other model is just showing off. It’s like bringing a sous-vide machine to a campfire”).
I don't have any sarcasm or personality settings enabled, but this is the pettiest, most passive-aggressive inner monologue I've ever seen from a model. I'm honestly not sure whether to be annoyed or impressed.
I also never told it the analysis came from ChatGPT, though I was tempted to just to see how it would react, ha.
