Let's tell OpenAI we don't want a "faster GPT-4" ...if this means to completely slaughter its performance like they did this week