And then the hope with Anthropic is that their constitutional AI, that’s to say their AI supervisor to train next generation of AI, will use this norm of our population to train a specific catering to that norm version of aligned AI. Because at the moment, Claude is universal, meaning that everybody who asks Claude get the same set of answers. But just like Facebook or YouTube, they quickly learn that every jurisdiction has different norms, every language has different norms and so on.