Living white paper · v0.1-draft
Open modular farming in Vilanculos
Smallholder yields in Mozambique remain far below agronomic potential. Many interventions and large commercial attempts fail under logistics, cost, land conflict, thin markets, weak training, and theft. GroundYield proposes small modular units, public costs and failures, and local baselines before impact claims — with AI and connectivity only as support.
As of this version: 0 GroundYield units, 0 pilot baselines. The operator discloses a prior large-scale farming failure in the same geography.
1. Problem
Yield gap
Smallholder maize in Mozambique is commonly reported around 0.4–1.3 t/ha (often near ~1 t/ha nationally) — below what reliable water, improved seed, and basic fertility can deliver. Sources: SOURCES.md.
National figures justify why a pilot exists. They do not replace local baselines for impact claims.
Why projects fail
Recurring modes: logistics, cost ramps, land conflict, skills mismatch, tech drops without training, theft of solar/pumps, thin markets, mega-programme under-delivery. See WHY_OTHERS_FAILED.md.
The operator previously farmed at scale near Vilanculos, then left after COVID-era pressure and bad choices, with an operator-disclosed loss on the order of ~ZAR 10 million. Full scar: LESSONS_MZ.md. This pilot is the opposite bet: small, open, measured.
2. Thesis
A standardized modular package on community-legitimate small plots, with public measurement and stop rules, can raise household productivity and produce replicable cost/yield data — better than opaque large-scale or unmeasured gift projects.
Secondary claims (also to be tested):
- Few units done honestly beat many units on paper
- Theft and training are first-class design problems
- AI and Starlink support choice and access — they do not replace water and seed
- Open failure is an asset for the next community
3. Method
Physical package
Per unit (~0.25–0.5 ha): solar irrigation, improved seed, micro-dose fertilizer, hermetic storage, optional livestock. Planning cost band ~$1,270–$2,985 (until field quotes). Detail: UNIT.md.
Process (binding)
- Baseline before any % impact claim
- Public unit registry before “N units deployed”
- Truth marks F / E / H (fact · estimate · hearsay)
- Dated public corrections — no silent rewrites
- No household dependency we cannot support
Gates G0–G6: GATES.md · Integrity: INTEGRITY.md
Geography
Vilanculos area, Inhambane, Mozambique — community return focus Rumbacaca. Tourism offtake is a hypothesis to check on the ground.
4. Current evidence
| Question | Answer | Mark |
|---|---|---|
| GY units installed | 0 | F |
| Pilot baselines | 0 | F |
| Field quotes priced | 0 | F |
| Does the package raise yield here? | Unknown | — |
| Offtake prices local | Unknown | — |
| National maize context | ~0.4–1.3 t/ha | H/E |
This section is rewritten when data exists. Live registry: units.html.
5. Success (honest)
- Public baselines and season data — good or bad
- Real registry units with trained custodians
- Honest costs; community exit rights
- Others can learn from the autopsy
Charter aims (e.g. 30–60 units, large % uplift) remain hypotheses until measured — CHARTER.md.
6. Open questions
- Local staple yields (method + F/E) around Rumbacaca?
- Real water at candidate plots?
- Local theft pattern for panels/pumps?
- Who buys surplus, at what farm-gate price?
- Which apps do people already use for rain/planting?
- Does a unit clear costs under real Season 1 prices?
- When do sensors beat a weekly phone photo?
Measurement: KPI_AND_DATA.md · Risks: RISK_AND_GAPS.md
7. Closing
We are not asking anyone to believe a finished success story. We are publishing a testable method and a promise of measurement.
If the model fails under real Vilanculos conditions, this paper should become an honest autopsy — not a deleted website.
Phases & milestones: roadmap.html · Master plan: PLAN.md · Cite with version date (v0.1-draft, 27 Jul 2026).