How a movement actually runs.

Layer 0 is the unglamorous foundation everything else sits on. We index the documents that ministries already publish — strategic plans, surveillance bulletins, programme reviews, HMIS extracts, EPI bulletins, climate-health briefs — and turn them into a stable, findable, citable record that universities, students and modellers in the same country can actually use.

The index lives on the AMB website with stable URLs, version dates, and short structured summaries. It is offered back to the ministry as a public good, and re-indexed twice a year.

Layer 1 is the *Lunch for Democracy* lens applied to public-health modelling. Sessions are intentionally small — 15 to 25 people — and run around a shared meal. Ground rules: active listening, respectful disagreement, no rank in the room, no exclusion by discipline or seniority.

Each session opens with a short ministry brief on a current decision; it closes with a working list of modelling questions and a named follow-up owner. Notes are summarised, scrubbed for attribution, and folded into the Layer 2 portfolio.

Layer 2 is where dialogue becomes work. The country anchor, the ministry, and the lead consultant draft a short portfolio of three to five modelling questions per country per year. Each question is feasible on a six-to-twelve-month student timeline, draws on publicly available data, and answers something a programme officer will recognise as their problem.

The portfolio is published openly on the site; it is the menu universities pick from for the next discovery cycle.

Layer 3 takes the Layer 2 portfolio onto campus. Two to four discovery sessions per country, hosted by partner universities, open to any student or staff member regardless of background who wants to engage in good faith.

Each session walks through one or two questions, presents the methods landscape, includes a peer panel with a working modeller from the country anchor's network, and ends with a concrete follow-up plan — who is taking which question forward, and who they will work with.

Layer 4 is the thin scaffolding around the student. AMB provides a short methods brief tailored to the question, helps the candidate co-formulate the concept note, attaches a named adviser from the modeller network, and opens a peer channel of students working on adjacent questions.

Each candidate also gets a profile on the platform, and — where appropriate — an *open-to-support* tag that lets funders see and back individual projects. The year ends with a showcase: every candidate presents what they learned, in front of their ministry, their supervisor, and the network.

Layer 5 is what the rest of the world sees. The website publishes the document index, the question portfolio, the project profiles and the output archive (with Zenodo DOIs for anything substantive). A monthly researcher spotlight surfaces one student, one method, one country.

A funder-facing index — separate from the public site — lists every open-to-support dossier with budget asks, timelines, and named contacts.