Nah I’m using it extensively, I know the limits. I do not think scaling is going to magically fix the fundamental limits of attention LLMs