Maybe im overlooking something obvious but how do you 'simply' scan the content of Claude users their prompts?

GP was making a joke, but Anthropic could implement this if they wanted to. Not a bad metric actually if you can measure it cheaply enough.