Yes, but AGI will first emerge from keeping state between calls to multiple models and assessing how closely they resemble humans intelligence, and using a loop to keep it going and updating the state. Which is what they are basically doing here