Yes, I see the difference as in hitting the logarithmic tail that shows we are close to the limit. I also realize that exponential cost is a defacto limit on improvement. If improving again for chatGPT7 will cost 10 trillions, I don’t think it will ever happen, right?