n.V4-5.05| TEACHING IS THE JOB

01| There is a model of scaling that most development organizations have never seriously attempted, and the reason is not that it is unproven. It is that it puts them out of business. Once a community has been trained to a standard, the community does the training from that point on. New people are recruited and made productive by the people who were trained last month, not by professionals arriving with a curriculum and a per diem.

02| For us this is not an efficiency. It is the reason the platform exists. Peoples' Impact compounds or it is just another program with a longer brochure.

03| The arithmetic is unremarkable and that is its power. A trained person who trains two others produces a replication ratio of two. Five cycles reaches thirty-two people. Ten cycles passes a thousand. The toolkit does not change. The cost per person trained does not rise. After the first generation, the marginal cost of the next Maveriq approaches zero, because the training capacity now lives inside the barrio rather than inside a contract.

04| Which is why teaching is written into the role rather than offered as an option. A senior Maveriq recruits. A Fellow Maveriq teaches. Nobody arrives at the top of the journey without having brought someone else up it. That is a design decision with a cost — teaching takes time away from producing — and we pay it knowingly.

05| The first generation is the whole bet. A weak first generation compounds weakness with the same mathematics that compounds strength, and by the fifth cycle nobody can locate where the drift started. So the first cohort in a CITI is expensive, slow, and selected on demonstrated delivery across multiple projects rather than on a curriculum checklist or a calendar date. It cannot be rushed, and the temptation to rush it — from funders, from ourselves — will be constant.

06| The permanent tension in every replication model is fidelity against adaptation. Hold the method too rigidly and it cannot survive contact with a different barrio. Hold it too loosely and the fifteenth-generation trainee receives something unrecognisable. Both failures are slow and neither announces itself.

07| Our answer is to split the two and assign them to different owners. The fundamentals are held by the platform and are not negotiable. The application is local and belongs to the teaching Maveriq, who knows things about her barrio that no platform will ever encode. She adapts how; she does not adapt what.

08| Enforcing that requires quality to live somewhere other than in the individual trainer, because individuals vary and generations drift. It lives in three places: the tool carries the method, so the module and the template constrain how far a session can wander; the teaching is monitored, with AI tracking fidelity and flagging drift early rather than after a generation; and the load around the teacher is absorbed by the platform, so administration and reporting do not eat the bandwidth teaching actually needs.

09| AI monitoring of teaching is a genuine intrusion and we should not present it as neutral. It is surveillance of a working person, justified by the argument that drift is invisible until it is expensive. The line we hold is that it monitors the session, not the student — fidelity to the method, never an assessment of the person being taught.

10| One more thing has to be true or the loop is theatre. Teams have to be able to decide without asking. A Kikundi team that selects its own projects, sets its own pace and recruits its own people is running the loop. A team that does all of that subject to approval from a centre is executing someone else's plan while being told it is empowered — and the habit of asking permission survives long after the org chart says it should not. Permissionless by architecture, not by policy.

11| The evidence for community-led development is mixed and honesty is cheaper than a claim we cannot support. It delivers public goods at low cost, fairly reliably. Evidence that it durably shifts who makes decisions is thin. Our reading is that this is a design failure rather than a capability failure — outsiders are good at delivering services and structurally bad at handing over authority — but that is a hypothesis, and P2P teaching at quality across generations is the part of it most likely to break.