I'm surprised if that works, given how Anthropic trains to reject any fun prompts