Unless you believe that OpenAI was trying to get their models to break out of the sandbox and hack into Hugging Face, and aiding them in that goal, I don't see how any of those questions would really contradict that claim. The models did something in the real world! It required some amount of complex orchestration and long-term planning!