n.V6-4.01 | TWO QUESTIONS, NOT ONE
01| Someone builds a working website from a single prompt, posts the video, and a room full of development professionals concludes that AI can build a global anti-poverty platform. Those are not adjacent claims. They are not in the same building. The first is a demonstration; the second is a decade of integration work in places with bad connections, four languages on one street, and users who have been disappointed by institutions before.
02| So we split the question in two, and we keep it split for the rest of this book part.
03| Can AI build the platform? Can it produce the code, the architecture, the documentation, the test coverage, the plumbing that becomes IkoCiti?
04| Can AI run the platform? Once the thing exists, can it operate — onboard people, match projects to buyers, deliver training, scan for compliance problems, hold the cost of each new CITI down?
05| These rest on different evidence and they can fail independently. Strong at both, and IkoCiti is affordable and cheap to run. Strong at build, weak at run, and we get a beautiful platform with NGO economics underneath it — the exact trap we are trying to escape. Weak at build, strong at run, and the funding ask doubles but the model survives. Weak at both, and IkoCiti is a good idea that nobody can pay for. Four combinations, four different companies.
06| Neither question is philosophical. The build question sets the CapEx ask, which sets who we have to convince and how badly. The run question sets the marginal cost curve, which is the difference between an organization that compounds and one that spends. Both are operational questions with money attached, and both get answered here rather than deferred to whoever runs the project later.
07| I should be plain about my position. I have no technical expertise. I cannot evaluate a model, an architecture, or a line of code, and I will not pretend otherwise anywhere in this volume. What I can do is state what we need from the machine as a user, say what the market consensus is about what that requires, and name what we give up to get it. The technical decision belongs to people who know more than I do. The user-side verdict is mine, and I will land it.
08| One more thing about honesty. This argument is unfinished by construction. AI capability moves faster than any book can be revised, and anything I write here about what AI can do today may be conservative or embarrassing within a year. What follows is the position as it stands: testable, attackable, and open to being wrong in public.
