Yeah, that's my point. If they do end up surpassing OpenAI then it would seem likely that they aren't just relying on copying from o1, or whatever model is the frontier model at that time.