You call it "highly accurate", referring to your own Bloomberg piece. In that piece, you recognize its iterative improvements, but for v4, your main source is their internal testing and your dad's essay. There is a big thing you're overlooking, which is generative adversarial activity. ↵