One container on parish ground becomes a learning commons: patient AI tutors that remember you, a library that works without the internet, WiFi, power, a build bench, and a paid steward who knows your name. For everyone with no access to — or no trust in — corporate and government compute. The Church handles the infrastructure: power, real estate, security, access. Proposed with Catholic Charities of Long Island (in discussion).
Jeff Pinto · Long Island, NY · prepared as a response to Magnifica Humanitas (May 2026)
Pope Leo XIV, Magnifica Humanitas, ¶67 · verified against vatican.va 2026-07-14
Read as an engineer, that is not a homily. It is a specification with an acceptance test: is access concentrated? Then you failed. This pilot is the existence proof — a thing a bishop can point to.
LI households lost ACP broadband — the connection gap is daily, not episodic
CCLI's own Sandy relief (1,500+ households); its after-action named seniors, immigrants, and the disabled as hit first
food-insecure Long Islanders (Long Island Cares)
and when storms do come (Isaias): 8-day restorations — the commons is quietly the backup
residents 65+, up 24% in a decade; 1,329 CCLI senior units
foreign-born in Brentwood; CCLI reports a 25% surge in Spanish-speaking Parish Social Ministry clients
Donor-funded by the thousands; SARI study: 32 of 36 private telecenters dead within 4 years. Died at "who pays month 37."
~900,000 laptops, ~$180M. Randomized evaluation: no effect on reading or math. Hardware without ownership or pedagogy.
Solar water + power + WiFi container; raised >€1.5M of EU funds; investigated for fraud — the demo's "purified" water was bottled mineral water.
2.8M users (~20% of the country). Wound down amid a $221M parent-company loss. A fifth of a nation lost its digital front door to care from a boardroom abroad.
The technology never failed. The ownership never arrived.
Rhizomatica's community-owned GSM co-ops (self-sustaining after year ~6). Guifi.net's 37,000-node commons. SolarSPELL's 600 libraries — every unit paired with an institution already on the ground. Bangladesh's rural electric co-ops: 80 co-ops, 50 years, 100M people.
Parish-level presence that outlives funding cycles; measured trust (Afrobarometer, 39 countries: religious leaders 66% trusted vs presidents 46%); Caritas' 162-nation delivery network; and now, doctrine that asks for exactly this (¶67–72).
Subsidiarity is not just governance language. It is a distributed-systems architecture: every node autonomous, no node critical.
| Layer | Spec | Discipline |
|---|---|---|
| The primer | Patient, adaptive AI tutors that remember the learner — math, ESL/citizenship, digital skills, résumés, small-business planning — EN/ES/Kreyòl, voice-capable | Progress lives on a card the learner owns; nothing typed leaves the room; every path ends with a person |
| The library & bench | Offline Wikipedia, open books and media, homework stations, coding club on the node's own server, LAN game nights, a build bench | Play is the on-ramp; the parish council sets content policy |
| The local corpus | Neighbors contribute: which pantry has diapers today, bus routes that actually run, oral histories, native-speaker translation review | Users become builders — contribution is the literacy; vetted improvements travel hub-to-hub |
| The infrastructure | 10–16 kW PV + storage (code-light ≤20 kWh or storm-strong 40–60 kWh), WiFi, paid docent-steward, parish governance | The Church handles power, real estate, security, access — infra as grace; storms are the bonus, not the thesis |
Energy honesty: an always-on GPU node wants ~67 kWh/day with cooling — 5–6× the roof array. We publish the energy budget; duty-cycled edge compute is the rule.
Working demonstrations of all three surfaces are live (gated): Node Planner · Storm Mode · The Well — common-light.pages.dev
The software heart on a cart inside an existing CCLI program: trust-first multilingual navigation + tutoring; measured on caseworker minutes returned and fear-suppressed clients reached. No container, no construction.
The container where no grid, no broadband, no safe building: clinic-, school-, or market-anchored parish node, selected with Caritas/CRS; re-costed to local pricing (US build is the ceiling).
Channel (mission office, Caritas/CRS intros, Barres blesses the sending) · Sponsor (sister-parish twinning) · Builder (staging, burn-in, vocational-ed jobs).
| Phase | Months | What happens |
|---|---|---|
| Gate checks | 0–1 | Diocesan property sign-off path + Catholic Mutual underwriting (the two real gates); site named; turnkey quote in hand |
| Build | 1–5 | Fabrication (4–6 wk) ∥ permits + PSEG-LI interconnection (6–12 wk, parallel) → install (days) → witness test |
| Operate & measure | 5–18 | Steward-run operations; monthly public metrics; storm-season drill; independent evaluation |
| Report & decide | 18 | Public report, failures included → scale, fix, or stop — stated in advance |
| N | Proves | Cannot prove |
|---|---|---|
| 1 | Operating model, demand, trust, unit economics, the runbook | Federation, transfer, no-critical-node, corpus network effects |
| 2 | The runbook transfers; corpus exchange works | One link is not a topology; one failure is an anecdote |
| 3 | A mesh: local corpora diverge and re-converge (propagation time measured); a deliberate node-kill with the others carrying; governance beyond one heroic pastor; three primer audiences compared | — |
Staged and gated: node 1 now; nodes 2–3 at a month-9 gate with thresholds published in advance. Subsidiarity is the architecture; the network is the encyclical's "shared responsibility of all" (¶8) made measurable.
Solid bar = lean; gold extension = to rich. Full line items with verification chips on the execution page (/plan/).
| Claim | Metric that tests it |
|---|---|
| People learn | Pre/post assessments in tutored tracks, benchmarked against the Ghana +0.37 SD RCT — the primer claim measured like a trial |
| People come back | Unique learners, return rate, session hours by program and language |
| Users become builders | Corpus contributions: local entries, native-speaker reviews, oral histories; % locally authored; propagation time hub-to-hub |
| AI assists, humans decide | Escalation-to-person count and resolution — measured, not asserted |
| The infrastructure earns its rent | Uptime, kWh served; storm events logged when they come (the bonus row) |
| It can scale | Cost per learner-month; month-37 fund accrual vs covenant; trust pulse; zero-tolerance incident log |
Month 18: a public report, failures included, independently evaluated. It either earns the pattern offer to Caritas and the Dicastery for Integral Human Development — or it honestly closes the question.
| Role | Holds |
|---|---|
| Parish council | Governance of the node: content policy, hours, exceptions — subsidiarity made literal; the community holds the keys |
| CCLI (501(c)(3)) | Fiscal home for grants; program integration and co-design (immigration, seniors, food); the steward's employer of record |
| Diocese of Rockville Centre | Ground, property sign-off, insurance (Catholic Mutual), safe-environment protocols — and the blessing that makes it the Church's answer, not a vendor's |
| Paid steward | Daily operations; the human at the end of every path |
| Jeff Pinto (technical steward) | Design, build supervision, telemetry, runbook — engaged so the parish never depends on him: the explicit goal is his own redundancy by month 12 |
| Independent evaluator | The report; access to all raw metrics |
| Funder(s) | Quarterly reporting; named recognition if desired; no operational control |
Nehemiah's wall: each family, its own stretch — one global city.
Does the caseload picture read true; do the mission-office and Caritas/CRS doors open; will the diocese bless the SENDING — the first American diocese to answer the encyclical by equipping a sister community abroad. No capital is asked of the Church, and no container is pitched at Long Island.
| Stage | Amount | Covers |
|---|---|---|
| Node 1 (commit now) | $250k | Learning-commons build ($140k), 18-mo operations with docent-steward ($80k), independent evaluation ($30k) |
| Nodes 2–3 (month-9 gate) | ~$350k | Two leaner nodes (~$110k each: shared engineering, code-light sizing), network operations, federated 24-month report |
| Program envelope | ~$600k | A three-node learning network with published gate criteria — and a stop, said out loud, if the gate fails |
Magnifica Humanitas, ¶16
Full ledger available on request · github.com/bigbrownjeff/common-light (private)