AI education should go much deeper than prompts or product literacy. The roots of AI are more durable; they're the approaches, architectures, mindsets, and concepts associated with "thinking" and learning. Teaching about AI doesn’t require teaching with AI. It doesn’t mean every new AI advance completely changes the conversation.  Though not identical to human versions, there are common aspects that any intelligence must deal with. By contrasting AI and human versions, AI education can be part of the fabric of school. Each subject, for each grade level, is at heart about thinking and learning. These meta-principles are critical for intuiting what AI might do, like understanding the drivers of another person. They help anticipate errors. They inform what examples to offer the AI. Most importantly, they move the mindset toward teaching AI, not just using it. I offer an example learning progression for K-16 at the end of each chapter, and suggestions for the environment and pedagogy for teaching such meta-principles. Drawing on decades of experience in both AI development and human learning and judgment, I explain: - AI processes information like economies do - AI does “know,” but it may not be what we want - You’re teaching AI, not just using it - Education should adopt the way AI learns - Most AI “bias” isn’t what you think In understanding the AI roots, educators gain not just insight into a transformative technology, but a powerful new framework for teaching students about themselves.