Why would data and training environment lead to frontier intelligence but slow?
Because the environment did not reward being as efficient as other models?
Because the environment did not reward being as efficient as other models?