Archetype 07 — Open-source + smallholder + multilateral: how the Global South is operationalising Mozilla's 2026 turning point

Archetype 07 — Open-source + smallholder + multilateral: how the Global South is operationalising Mozilla’s 2026 turning point

A substantive deep-dive for multilateral institutions, civil-society funders, and African/smallholder-adjacent policy advisors.

Header

What this talk is for

The audience is interested in the structural shift Mozilla (State of Open Source AI 2026, July 14 2026) just empirically named. They already know “open source” is a contested term. They want to know: in agrifood specifically, is the open-source turning point real, where is it operational, and what does it cost / what does it produce? This talk walks them through Mozilla’s quantitative spine, then narrows to agrifood, then anchors in three operational layers — multilateral deployment (CGIAR + AgriLLM + AGRA), civil-society / African-led open-source (Ushahidi + Open Data Kit + FarmOS + Digital Green + OSSI seed-sovereignty + Mozilla Common Voice African-languages), and the policy-lever question.

The talk is deliberately not a survey of every open-source agritech project. It is deliberately a structural-spine talk: it picks three operational layers and one open question to leave the audience with. It is the load-bearing talk for the cycle wave of Sub-Saharan multilateral + open-source + Mozilla-quantitative + plant-breeding-global scans.

The single claim. Open-source AI has reached a turning point in agrifood specifically because the Global-South deployment (Mozilla + CGIAR + AGRA + Ushahidi / ODK / Digital Green + OSSI + Mozilla Common Voice African-languages) is the empirical evidence Mozilla’s 2026 numbers said would have to exist for the turning point to be real. That evidence exists. The talk names it.

Run-of-show

1. Opening — Mozilla’s 2026 numbers (5 min)

Frame. Mozilla (State of Open Source AI 2026, July 14 2026; Raffi Krikorian CTO; 950+ developers surveyed by SlashData) named a turning point. The numbers:

• 3% performance gap between open and proprietary models — open is no longer playing catch-up.
• 50× cost reduction in three years.
• 33% of real-world AI usage runs on open models; that 33% captures only 4% of AI revenue. Value is real; revenue is not flowing back to the open ecosystem.
• 89% adoption in China / East Asia vs. far lower in the West.
• 79% of developers use open models; only 51% have deployed them in production (vs 63% for closed models). The gap is infrastructure, not quality.
• “The real fight has moved beyond the model” — the agentic harness (the software layer between people and models) matters more than the model itself.
• 93% of users approve AI agent requests by default (“consent fatigue”).

Anchor quote. Raffi Krikorian: “Open source AI has reached a turning point. It’s no longer about expanding access to models; it’s about who has the power to shape, audit, and improve them. Without investment in the infrastructure, tooling, and governance around open models, we risk locking in a system where only restrictive, closed AI can scale – and that doesn’t serve the public interest, or sovereignty over tech policy decisions.” quotes/researchers-and-experts/mozilla-krikorian-open-source-turning-point.md.

Unit: units/mozilla-state-of-open-source-ai-2026.md.

Critical move. Mozilla’s numbers are about AI generally, not agrifood specifically. The talk’s job is to narrow Mozilla’s global claim to the agrifood case — and to do so honestly (the agrifood quantitative panel does not yet exist at the Mozilla level of methodological rigour; G-356 is real; that gap is part of the story).

2. Segment one — the multilateral deployment layer (12 min)

The first operational anchor. CGIAR + AgriLLM + UAE AI71 is the corpus’s most-substantive multilateral deployment-of-record for the Global South.

CGIAR + AgriLLM specifics.

• CGIAR has been operating since 1971 (HQ near Montpellier France; major center ILRI Nairobi).
• AgriLLM launched June 2025 as the substantive AI extension LLM — chatbot prototype target COP30 (November 2025 / 2026 cycle).
• $200M UAE-Gates-CGIAR partnership.
• FAO Q&A pair collaboration.
• Multilingual deployment — Bihar dialects + Swahili + Spanish + Hindi + Tamil + Telugu + Arabic + French + Portuguese + African languages.
• Smallholder-centred design — voice-first, low-literacy, low-trust in conventional digital platforms.
• Deployed in Bihar (India) + Kenya + Mexico at S1–S2 pilot scale.
• CGIAR institutional scale is global S3.
• Data governance: multilateral-open (CGIAR Open and FAIR Data Assets Policy; FAIR Principles + Responsible Data Guidelines); IDSov-aligned (CARE Principles intersect + WIPO Treaty on Intellectual Property, Genetic Resources 2024 — Malawi + Uganda first ratifications).

Unit: units/cgiar-agrillm-ai-global-south.md.

The second operational anchor — AGRA. AGRA (Alliance for a Green Revolution in Africa; founded 2006 by Rockefeller + Gates Foundations; “proudly African-led institution”) is the corpus’s primary multilateral deployment-of-record for Sub-Saharan smallholder agricultural transformation.

AGRA specifics.

• 15 million smallholder farmers trained.
• AGRA-PASS (Programme for Africa Seed Systems) — 562 new seed varieties commercialised.
• 42 public policy reforms advocated across 13 African countries.
• Current Mastercard Foundation partnership: Ghana + Tanzania + Kenya focus on youth inclusiveness.
• African-led board since 2012.
• Digital agriculture platforms + climate-smart agriculture knowledge dissemination + financial inclusion + market information systems + payment innovations + smallholder extension + agronomic advisory.

Unit: units/agra-alliance-green-revolution-africa.md.

Substantive cluster observation. Sub-Saharan multilateral agritech is not a state-DPI substrate (no AgriStack / WAGRI equivalent). It is multilateral-institutional convening + private-foundation funding — a different cluster pattern. The LAC pattern observation (talks/archetypes/06-regional-cluster-comparison.md segment 3.4) of multilateral-institutional convening + venture-funded SaaS is the closest cluster-pattern analog. Worth naming for any audience thinking about cluster-pattern generally — SSA’s cluster pattern is multilateral-institutional + foundation-funded, not state-DPI.

Critical move. The multilateral layer is operational and African-led. It is not a hand-wave. It is named CGIAR + AgriLLM + AGRA + 13 national AI strategies + 12 new national AI strategies globally + 47 countries restricting foreign processing. Mozilla’s empirical observation that “open models power ~33% of real-world AI usage” is consistent with the multilateral-Global-South deployment substrate — these are open-source / multilateral-open deployments operating at scale that Mozilla’s 33% aggregate is measuring.

3. Segment two — the civil-society / African-led open-source layer (12 min)

The second operational layer. Below the multilateral substrate is a substantive civil-society + African-led open-source ecosystem. This is where Mozilla’s “who has the power to shape, audit, and improve” claim meets the agrifood case most directly.

Civic-tech / open-source data tooling (Sub-Saharan).

• Ushahidi — Kenyan-origin crisis-mapping platform; substantial African deployment.
• Open Data Kit (ODK) — open-source mobile data collection; multilingual; low-bandwidth defaults.
• FarmOS — open-source farm-management web app; smallholder + extension-services use cases.
• KoBoToolbox — open-source data collection, UN-affiliated deployment at scale.
• SurveyCTO — open-source-adjacent data collection with consent-management design.

Civil-society / NGO open-source.

• Code for Africa — pan-African civic-tech / open-data network.
• Digital Green — AIEP (Agricultural Information Exchange Platform) partner; CGIAR AgriLLM collaboration.
• Viamo — AIEP partner; mobile-first smallholder advisory at scale.
• GODAN (Global Open Data for Agriculture & Nutrition) — open-data-policy advocacy.

Academic / research open-source.

• Open AIR (Open African Innovation Research) — African-led open-innovation research network.
• Maseno University (Kenya) — Mozilla Common Voice Workshop; Afrifose2030 partnership.
• Africa Next Voices Workshop — Mozilla + Maseno partnership.
• Strathmore University (Kenya) — AI tools deployment.
• University of Cape Town — research open-source.

Open-source seed systems (the IDSov-coupled subset).

• Open Source Seed Initiative (OSSI) — “Free the Seed” Pledge (2012); 14+ years operation.
• African deployment via named actors: Daniel Wanjama (Kenya Seed Savers Network); Pelum Zambia; Swaziland Rural Women’s Assembly; A Growing Culture.
• International Seed Federation engagement — OSSI collaborative engagement.
• WIPO Treaty on Intellectual Property, Genetic Resources (2024) — Malawi + Uganda first ratifications; IDSov-aligned.

Mozilla Common Voice African-languages (the speech-recognition open-source anchor).

• Common Voice 23.0: 357 hours Spontaneous Speech across 51 languages.
• Substantive African deployment via Maseno University + Africa Next Voices + Africa’s Talking Kiswahili Hackathon Series.
• CC0 license (the most permissive data licence available).
• Voice-first design = smallholder + low-literacy compatible.

Units: units/mozilla-common-voice-african-languages.md, units/open-source-seed-initiative-africa.md, units/open-source-in-agrifood-framework.md, units/cgiar-agrillm-ai-global-south.md (also segment 1).

Substantive observation. Mozilla’s “the real fight has moved beyond the model” maps cleanly to this layer. The civil-society / African-led open-source ecosystem is exactly the agentic harness + governance + multilingual data layer Mozilla is naming. CGIAR AgriLLM’s voice-first + multilingual + local-language deployment is harness-layer; ODK + FarmOS + KoBoToolbox are harness-layer; Mozilla Common Voice is training-data layer (the layer where the open-vs-closed gap is largest). When Mozilla says 33% of real-world AI usage runs on open models, the Global South is over-represented in that share because the harness and training-data layers are where the Global South leads.

Critical voice. Open-source is not a value-free frame. Mozilla explicitly names the power to shape, audit, and improve as the open-source differentiation — and the Global-South deployment is where that power is being substantively exercised (CGIAR governance + AgriLLM deployment governance + AGRA African-led governance + OSSI African deployment). The talk names this honestly: open-source is a structural choice about who decides, not a free-lunch alternative to proprietary AI.

4. Segment three — the limits and the gap (8 min)

Engage the limits. Mozilla’s turning point is real but it is not a victory lap. Three substantive limits worth surfacing:

Limit 1 — the agrifood-specific quantitative panel does not yet exist. Mozilla’s 3% / 50× / 33% / 89% / 79% / 51% / 93% figures are global AI, not agrifood. The agrifood-specific panel that would let a sector analyst say “3% of agrifood AI usage runs on open models” does not yet exist in published form (G-356). The corpus’s regional scans and anchor units substitute for this absence with named-deployment evidence, but they are not Mozilla-methodology-rigour. Worth naming this honestly: Mozilla’s claim is structural; the agrifood case is consistent with the structural claim; the agrifood-specific panel is a downstream corpus-building job.

Limit 2 — value capture is not yet flowing back. Mozilla: “33% of real-world AI usage runs on open models; only 4% of revenue.” The Global South is the deployment side of that asymmetry. The revenue capture side has not yet been substantively developed — who captures the value of the open-source agrifood AI deployed through CGIAR / AGRA / Mozilla Common Voice / ODK / Digital Green? The substantive answer today is: foundations (Gates + Mastercard + Rockefeller + UAE Government), multilateral budgets (CGIAR + FAO + WIPO), and African-led institutions (AGRA + AU Continental AI Strategy). That answer is partial. The vendor capture layer has not yet been substantively articulated at the Global-South scale.

Limit 3 — the agentic-harness gap is real. Mozilla’s 79%/51% developer/deployment gap is infrastructure, not quality. For Global-South agrifood, the infrastructure gap is even larger: low-bandwidth, low-literacy, low-trust, multilingual, voice-first. CGIAR AgriLLM addresses harness at the LLM-advisory layer; ODK + FarmOS + KoBoToolbox address harness at the data-collection layer; Mozilla Common Voice addresses harness at the training-data layer. But the integration layer — stitching voice-first advisory + multilingual text + local-language datasets + farm-management web app + smallholder-extension feedback loop into a single deployable surface — is not yet at Mozilla-rigour in the agrifood case. Worth naming as the structural work of the next cycle.

Don’t end on boosterism. The 60-min audience earns its keep by engaging limits honestly. The turning point is real; the deployment-of-record evidence is real; the integration-gap work is real and ongoing.

5. Segment four — the cluster-pattern observation (8 min)

The structural comparison. Where does the open-source + smallholder + multilateral pattern sit relative to the corpus’s cluster-pattern taxonomy?

The substantive observation. The Sub-Saharan + Global South cluster is the only cluster pattern where open-source AI is load-bearing at the deployment layer (not just the framework layer). EU has Mozilla + JoinData at framework; EU also has Wageningen + INRAE institutional depth. But the deployment of open-source AI at the Global-South agrifood smallholder layer is substantively led from CGIAR + AGRA + civil-society African-led. Mozilla’s empirical 33%-of-usage figure is structurally consistent with the Global South being over-represented in the open-source share — and the corpus’s anchor units are the named evidence.

Critical move. Name what the cluster-pattern observation does not produce: Sub-Saharan + Global South open-source + smallholder + multilateral pattern is not a state-DPI substrate (no Sub-Saharan equivalent of AgriStack / WAGRI), is not a corporate-vendor-funded SaaS (no Sub-Saharan Agrosmart-equivalent at scale — yet), is not a cooperative-governed substrate (NAPDC is US-frame; no JoinData-equivalent in SSA — yet). It is multilateral-institutional convening + private-foundation funding + civil-society / African-led open-source deployment. That cluster-pattern statement is the corpus’s load-bearing analytical claim.

Worth naming — Seed-breeding cluster as a special case. CGIAR Breeding for Tomorrow + EiB (12 priority crops; 700K+ germplasm in ITPGRFA trust; Gates/FFAR/USAID funding substrate; 15-30% genetic-gain uplift commitment) + OSSI Pledge seed-sovereignty movement (2012; 14+ years; African deployment via named actors) + brazilian-seed cluster-with-three-structures (academic-research-led Tier-1 + multinational-corporate-pipelined Tier-2 + substantially-empty Brazilian-origin-corporate-vendor Tier-3) + Longping-Yuan-CAAS Chinese-state-orchestrated cluster + USDA-ARS-AIIRA US-land-grant cluster + Limagrain-KWS-RAGT EU-private-breeding cluster + Indigenous Seed Sovereignty AI Breeding cross-cutting critical-voice. The seed-breeding substrate is its own multi-cluster pattern, and is substantively distinct from the agrifood-AI-deployment cluster. Worth naming because seed-breeding is the cleanest test case for whether Mozilla’s open-source turning point reaches to agrifood’s deepest biological-substrate layer.

Units: units/cgiar-eib-global-south-plant-breeding.md, units/longping-yuan-caas-china-seed-ai.md, units/bayer-syngenta-corteva-multinational-pipelines.md, units/usda-ars-iowa-state-aiira-us-land-grant.md, units/limagrain-kws-ragt-eu-private-plant-breeding.md, units/indigenous-seed-sovereignty-ai-breeding.md, units/ai-breeding-genetic-diversity-counter-narrative.md.

6. Segment five — the policy-lever question (3 min)

The frame. Mozilla’s turning point + the Global-South deployment-of-record + the integration-harness gap = a policy lever that multilateral institutions, foundations, and African-led civil society are uniquely positioned to pull. The lever is not “more model training” (that’s Mozilla’s already-met bar). The lever is:

1. Harness-layer investment. Voice-first multilingual advisory + low-bandwidth deployment + consent-management + audit-trail integration. CGIAR AgriLLM is one surface; the integration-harness work is the next.
2. Training-data layer investment. Mozilla Common Voice African-languages + ODK + FarmOS local-language datasets are load-bearing; the speech + text + image data layer for low-resource languages is the slowest-moving open-source gap.
3. Governance layer investment. CGIAR Open and FAIR Data Assets Policy + AGRA African-led governance + OSSI seed-sovereignty + WIPO Treaty Genetic Resources ratification. Who decides the data rights, who captures the value, who audits the agentic harness — Mozilla’s “consenting to agent requests by default 93% of the time” is exactly the governance layer where Global-South open-source agrifood AI is substantively breaking ground.

The single sentence. Mozilla’s empirical turning point is real and the Global-South agrifood deployment is the corpus’s most-substantive live evidence; the next cycle’s work is harness-layer integration + training-data multilingual coverage + governance audit infrastructure, and the leverage points are multilateral + foundation + civil-society African-led — not proprietary vendor capture.

Q&A handles

• “What about Mozilla on the technical / non-policy side?” → Mozilla’s empirical claim is dual-use: open-source reached technical parity (3% gap), AND reached a structural turning point about who decides. Both are true and both matter. The Global-South agrifood case deploys both — the technical layer (CGIAR AgriLLM, Mozilla Common Voice voice-first deployment) and the structural-power layer (CGIAR governance, AGRA African-led governance, OSSI seed-sovereignty). The talk names both halves.
• “What is AGRA’s stance on AI?” → AGRA is the corpus’s primary multilateral deployment-of-record anchor for Sub-Saharan smallholder agricultural transformation; digital agriculture platforms + climate-smart agriculture knowledge dissemination + financial inclusion + market information systems + payment innovations + smallholder extension + agronomic advisory are the named deployment surfaces (per units/agra-alliance-green-revolution-africa.md). AGRA does not have a named independent AI programme in the corpus; the digital surfaces are within AGRA-PASS, Soil Health Initiative, and Centre for African Leaders in Agriculture. Worth saying so honestly: AGRA is deployment-of-record anchor for smallholder, not a named AI programme.
• “What about the China / East Asia 89% adoption figure?” → Mozilla’s empirical observation: China + East Asia lead in open-source AI adoption at 89%, far ahead of the West. The structural interpretation worth surfacing is that East-Asia open-source is state-strategy-driven (Chinese state + Alibaba + Baidu + Tencent + Huawei all deploying open-source models as state-policy infrastructure), whereas Global-South Sub-Saharan open-source is multilateral-institutional + foundation-funded + civil-society-led. Two distinct cluster-patterns of open-source adoption, both real, both different. Per talks/archetypes/06-regional-cluster-comparison.md, China’s cluster pattern is state-vendor hybrid + provincial autonomy + multilateral-state coordination reach (WAICO). East Asia does not have a CGIAR or AGRA equivalent; the smallholder-cooperative agricultural layer is not where 89% East Asia open-source adoption is operating.
• “What about open-source seed / OSSI in Africa?” → units/open-source-seed-initiative-africa.md + units/indigenous-seed-sovereignty-ai-breeding.md + units/ai-breeding-genetic-diversity-counter-narrative.md. OSSI Pledge (2012; 14+ years) is the corpus’s substantive open-source seed-sovereignty movement; African deployment via Daniel Wanjama (Kenya Seed Savers Network) + Pelum Zambia + Swaziland Rural Women’s assembly + A Growing Culture. WIPO Treaty on Intellectual Property, Genetic Resources (2024; Malawi + Uganda first ratifications) is the international-treaty layer that intersects CARE Principles for Indigenous Data Governance — substantive IDSov + seed-sovereignty + open-source licensing alignment. The seed-sovereignty layer is separate from but parallel to the AI-deployment layer; both are open-source-led.
• “What about Mozilla Common Voice for African languages?” → units/mozilla-common-voice-african-languages.md. Common Voice 23.0 has 357 hours Spontaneous Speech across 51 languages. Substantive African deployment via Maseno University (Kenya; Common Voice Workshop; Africa Next Voices; Afrifose2030 partnership with University of Nairobi + University of Embu) + Africa’s Talking Kiswahili Hackathon Series (Nairobi). CC0 license (the most permissive data licence). The training-data layer is the slowest-moving open-source gap globally; Common Voice is the corpus’s anchor for the African-language voice-data layer.
• “What about open-source data infrastructure generally?” → units/open-source-in-agrifood-framework.md (cross-cutting framework) + units/open-data-ecosystem.md (GODAN, CGIAR FAIR, USDA Ag Data Commons, Copernicus, SoilGrids) + units/oada-open-ag-data-alliance.md (open-source interoperability standards). Open Data Kit (ODK) + FarmOS + KoBoToolbox + SurveyCTO are the operational mobile-data-collection layer. Open Ag Data Alliance (OADA) is the open-source interoperability standards project that the cooperative / commons ecosystem can run on. Worth naming to the audience: the open-source agrifood ecosystem is a layered stack (training data → model → agentic harness → standards → deployment interface), not a single project.
• “What about the Mozilla 33%/4% value-vs-revenue asymmetry for global health / agrifood?” → The structural observation: open models power ~33% of real-world AI usage but capture only 4% of AI revenue. The Global South is the deployment side; the revenue capture is at the foundation + multilateral + vendor-capture layers. The substantive answer today is foundations (Gates + Mastercard + Rockefeller + UAE Government), multilateral budgets (CGIAR + FAO + WIPO), and African-led institutions (AGRA + AU Continental AI Strategy). Worth naming honestly: the value capture question is open.
• “Is this just about Africa?” → The talk is anchored Sub-Saharan because the corpus’s Sub-Saharan cycle waves are substantively populated (multilateral + open-source + Mozilla Common Voice African-languages + AGRA African-led). But the cluster pattern (open-source + smallholder + multilateral) is global-South, not Africa-only — Bihar (India) + Kenya + Mexico + Brazilian seed AI + Spanish cooperative AI cluster-with-three-structures are all consistent examples. The structural claim is “global South open-source + smallholder + multilateral”; Africa’s prominence in the talk is a function of where the corpus is deepest, not of where the pattern is bounded.

Freshness check

• Mozilla State of Open Source AI 2026: July 14 2026; next-year report cadence TBD.
• CGIAR + AgriLLM: AgriLLM launched June 2025; chatbot prototype target COP30; deployment in Bihar / Kenya / Mexico at S1–S2 pilot scale. Re-verify at COP30 follow-up cycle.
• AGRA: 2006 founding; 15M smallholder farmers trained; 562 seed varieties commercialised; Mastercard Foundation partnership in Ghana + Tanzania + Kenya. Re-verify at next AGRA annual report.
• Mozilla Common Voice: Common Voice 23.0 at 357 hours Spontaneous Speech / 51 languages; Maseno University + Africa Next Voices + Africa’s Talking Kiswahili Hackathon deployment. Re-verify at next Common Voice release.
• OSSI: 2012 OSSI Pledge; 14+ years operation; African deployment via Wanjama + Pelum Zambia + Swaziland Rural Women’s Assembly. Re-verify at next OSSI annual.
• WIPO Treaty on Intellectual Property, Genetic Resources (2024); Malawi + Uganda first ratifications. Re-verify at next WIPO Treaty status update.
• Sub-Saharan Africa multilateral + open-source scans (scans/2026-07-sub-saharan-africa-multilateral.md, scans/2026-07-africa-open-source-agrifood.md, scans/2026-07-open-source-cycle.md) are last-verified 2026-07.
• The agrifood-specific quantitative panel (G-356) does not yet exist in published form; re-verify at next Mozilla-agrifood or EPRS-agrifood or Stanford-AI-Index-agrifood report release.

Substitutions

• If audience is EU-anchored, deepen segment 5 (cluster-pattern comparison) with the EU multilateral pattern observation from archetype 06. Add La Ferme Digitale / GAIA + JoinData + Wageningen + INRAE as the EU open-source + cooperative + institutional anchor.
• If audience is US-anchored, lead with Mozilla’s national-strategy observation (12 new national AI strategies launched last year; 47 countries restrict foreign processing for critical workloads). Add US land-grant cluster + Land-Grant AIIRA at Iowa State + Heritable Agriculture from Google X + USDA-NIFA + Texas A&M wheat-breeding pipeline (units/usda-ars-iowa-state-aiira-us-land-grant.md).
• If audience is funder-grantmaker (Gates / Mastercard / Rockefeller), lead with segment 3 (civil-society / African-led open-source layer) — the Mozilla + civic-tech + Common Voice anchor.
• If audience is multilateral-institution staff (CGIAR / FAO / WIPO / AU), lead with segment 1 (multilateral deployment layer) + segment 5 (cluster-pattern observation). The cluster-pattern claim about multilateral-institutional + foundation-funded is load-bearing for the multilateral-staff audience.
• If audience is academic STS / political-economy, deepen segment 2 with the Mozilla “consenting to agent requests by default 93% of the time” consent-fatigue observation. The governance-layer-where-consent-is-defaulted argument is the substantive academic-handle.

What this archetype is doing in the methodology

This is the open-source + smallholder + multilateral talk — the one a presenter gives when the audience wants the structural case for the Mozilla turning point, anchored in agrifood, with the Global South as the load-bearing evidence. The 60-min duration matters because Mozilla’s empirical claim deserves careful unpacking before narrowing to agrifood; civil-society / African-led open-source ecosystem deserves substantive depth; the integration-harness gap deserves honest engagement; and the cluster-pattern observation is the corpus’s load-bearing analytical claim about how Global South open-source + smallholder + multilateral differs from state-DPI / vendor-led / cooperative-governed substrates.

The structural load-bearing move: if the audience walks out thinking “open-source is happening in agrifood but not at scale” or “open-source is happening at scale but only as a Western / Western-funded alternative”, the talk failed. The point is that Mozilla’s empirical turning point is substantively operationalised by the Global-South multilateral + civil-society + African-led open-source ecosystem — and that operationalisation is the corpus’s most-substantive live evidence of the turning point Mozilla names. The audience should leave with: (a) Mozilla’s numbers anchored, (b) named Global-South deployments anchored, (c) integration-harness-gap acknowledged, (d) cluster-pattern claim about multilateral-institutional + foundation-funded + African-led open-source delivered.

This archetype is new — added in July 2026 alongside the Sub-Saharan Africa multilateral + open-source + Mozilla State of Open Source AI 2026 cycle waves. Future iterations of the corpus (additional Sub-Saharan cycles, additional Mozilla-agrifood empirical reports, additional Global South open-source deployment evidence) will refine the cluster-pattern vocabulary here as well as the integration-harness gap observation.