The prototype did the two jobs a deck never can. It was the experience design: every interaction decision was made by using the thing, not imagining it. And it de-risked the investment: usability testing ran on the functioning prototype with real users, and stakeholders drove it themselves rather than watching a walkthrough, all before committing an engineering team to the full build. What we learned was folded back in, then the vision was reconciled against version one; every ambitious idea annotated into the near-term spec, parked with a reason, or cut. The shipped direction kept what mattered: one engine, two surfaces, model identity grounded in an honest derivative, and a verdict that gives people permission to decide. Discovery, research, architecture, prompt design, working software, evidence. One person, end to end.