A Quick Note on AI Scaling Asymptotes — LessWrong