← Back to Research

Round 4 Synthesis, May 2026

A reading of the seventeen briefs assembled this round. Observational. Deliberately not prescriptive.

Files read

1. L1, Schumacher Successors. Twenty-two months after closure, the Satish Kumar Foundation owns the brand and runs nomadic short courses without a permanent site; faculty dispersed individually with no collective successor; Naropa is the strongest infrastructural anchor; Ecoversities has chosen radical horizontality.
2. L2, Findhorn Year-On. The November 2024 buyback produced four legally distinct successor entities (CBS, SCIO, PET, Hinterland Trust) where one Foundation used to stand; £854k deficit for FY March 2025; the Light of Findhorn Sanctuary opened Easter 2025; asset distribution proposal still pending eighteen months in.
3. L3, Mind & Life Post-Dialogue 39. The 2024 "Bold New Strategy" displaces the dialogue tradition's centre toward AI / climate / polarization / systems change without speaking the succession question aloud; the Dalai Lama's July 2 2025 statement settled the reincarnation question; the bliss-attractor finding sits unmetabolised.
4. L4, Wellspring Sunset. Wellspring sunsets by 2028; CIFF withdrew from US grantees in 2025; almost nothing fills the slow-relational-multi-year movement-infrastructure shape at scale; Yield Giving's $7B 2025 unrestricted is episodic, not relational; Fetzer's $9M 2026 collaborative is the right shape at roughly a tenth of the scale.
5. C1, Relational-Alignment Camp. A loose camp converging across the EA Forum, arXiv, and IT for Change on the proposition that alignment-as-solving is a category error; no shared name; Open Phil's $40M+ technical safety RFP funds none of the relational categories; Anthropic's model welfare line is the only foothold inside frontier-lab money.
6. C2, Friston / Active Inference. Friston, Ramstead and Heins' 2024 Designing Ecosystems of Intelligence from First Principles maps almost line-for-line onto kincentric / cyberecological vocabulary in measure-theoretic clothing; Verses' Genius is shipping into enterprise finance; Bruineberg's Markov-blanket-trick critique is the live methodological dispute; the contemplative-practice bridge is forming through Sandved-Smith but does not yet engage non-Western traditions.
7. C3, Three-Tier Arrivals. Bill Reed's full Degenerative→Regenerative trajectory is the parent; Wahl's H2 carries the moral ambiguity of transition; SRAGI mirrors IoC's three-tier from an Albanian-mining founder vehicle; Movate has captured the term "regenerative AI" to mean self-healing code; IT for Change holds it as decolonial political economy; nobody has written the synthesis essay.
8. C4, Barad / Intra-Action. Stein's February 2026 npj Climate Action paper does Baradian work but cites Haraway and Indigenous scholarship rather than Barad; "intra-action" is established in post-qualitative education research and emerging in HCI/AI ethics journals but absolutely absent from enterprise governance vocabulary; Cerniglia & Cimino's "relational infrastructure" framing is the lateral lift into care literature.
9. V1, Regenerative AI Race. Five competing definitions ("regenerative AI"); Movate's enterprise-IT meaning is winning SEO; IT for Change's relational/political meaning has institutional weight but no enterprise distribution; Regen Network's compute MCP makes regeneration a throughput feature rather than a category; "kincentric AI" remains unclaimed at the brand layer.
10. V2, "Regenerative" Generally. USDA's December 2025 launch of a $700M Regenerative Pilot Program that explicitly accommodates pesticide use, plus Rainforest Alliance's March 2026 Regenerative Agriculture Standard, signal terminal capture of the bare term; Wahl, Reed and Hutchins are visibly retreating to capability and developmental-stage framings; Living Building Challenge is the one strategy that bets on measurable floor instead of practitioner depth.
11. V3, Living Systems Migration. Cerniglia & Cimino's MDPI paper migrates "living systems" into mainstream public-health framing; Grossman et al.'s February 2026 polyvagal-untenable paper opens a vacuum at the meta-frame layer; "relational" is doing more conceptual work than "living systems" in the actual papers; the indigeneity/kincentric ecology connection is missing where it would most matter.
12. V4, Vocabulary Mapping. Seven candidate terms surveyed; sensemaking is bifurcated; "weaver" is overclaimed by Aspen + Weaving Lab + Network Weaver in incompatible ways; cyberecology remains thinly populated and is the rare term carrying low lineage debt; "tending" as verb and "coherence" as noun appear unclaimed.
13. NA1, Latin American Animist / Decolonial. Krenak refuses the wisdom-supplement frame the Anglophone climate discourse hangs on him; Aguilar Gil's "we weren't always Indigenous" cuts hard against Andean Buen Vivir's ontologisation; the retomada literature is the most live Indigenous political theory in Latin America and it is essentially untranslated; Wedemeyer's Krenak/Kopenawa/Galindo "decolonial-anarchist axis" reading has no English equivalent.
14. NA2, Lusophone & African Solarpunk + Ubuntu. Brazil's 2012 Editora Draco anthology is the actual chronological origin of solarpunk as a literary genre; African writers have largely declined "solarpunk" in favour of Africanfuturism; the Wild Law Institute's 2025 Table Mountain initiative and African Commission Resolution 372 (2017) build Earth jurisprudence as governance infrastructure; ubuntu has no AI-design protocol corpus despite the operational uptake in HR and social work.
15. NA3, East Asian Contemplative-AI + Satoyama. CSAS proposes the bodhisattva as alignment target; Buddhabot Plus puts a sangha as veto authority over an LLM in Bhutan; Plum Village's Living Gems EUA explicitly forbids AI training use, the most legally articulated contemplative refusal in any tradition; Audrey Tang's Plurality is read as civic tech and missed as relational ontology; Keibo Oiwa's fifty-book corpus is essentially untranslated.
16. NA4, South Asian Regen Governance. IT for Change holds "regenerative AI" as political economy, not aesthetics; Kudumbashree at 4.8M women across 317,724 NHGs is the largest functioning instance of relational governance design on Earth; Aravind Eye Care + Google ARDA is the cleanest existing example of a Big-Tech AI licensing structure routing value to an underserved patient population; AI4Bharat's IndicTrans2 ingests 22 scheduled languages including Adivasi-language scripts.
17. NA5, European Indigenous + Scandinavian Eco-Literary. The Saami Council's SODA principles (April 2024) are the European equivalent of Te Mana Raraunga that conscious-design discourse is not citing; the Sámi AI Lab is real and Sámi-led; cynefin and dúchas are place-relational concepts that the framework export captured the first of and the second was named by Manchán Magan, who died October 2025; the Scandinavian eco-literary tradition is structurally older than its English reception treats it.

1. The single argument

Across all four buckets, one observation keeps landing in different voices: the relational claim is now ambient. It is no longer a heterodox position held by a few writers in continental philosophy, Indigenous studies, or regenerative design. It is the default substrate of new technical AI papers (C1, C2), of mainstream public-health and care literature (V3, C4), of climate-and-AI commentary (C4, V1), of governance infrastructure under construction (L2, NA4, NA5), of vocabulary capture races (V1, V2, C3), and of literatures that have always been there but that the Anglophone discourse has been slow to read (NA1, NA2, NA3, NA4, NA5). The proposition "the relation is the unit" is being arrived at independently from active inference, from agential realism, from participatory sense-making, from Buddhist care-as-intelligence, from Sámi data sovereignty, from Kerala's three-tier neighbourhood-group federation, from Karen Barad's physics of meaning and Davi Kopenawa's xapiri, from Findhorn's split into four custodians of one body. None of these are speaking to each other in any structural way.

What the round shows, then, is not a discovery but a coordination problem. The convergence is real and the absence of cross-citation is the actual datum. This makes the period (call it the next eighteen to twenty-four months, per V1, V2, C3) a window in which the vocabulary gets settled even though the substance has already been settled in twenty places at once. The capture races V1 and V2 document, Movate winning "regenerative AI" by SEO, USDA spending $700M to dilute "regenerative" generally, are the foreground; the silent convergence the rest of the briefs document is the background. The interesting work, on the evidence, is at the seam between them.

2. The lineage handoff (L1–L4)

L1 through L4 are the elegies of the round. Each documents an institution that was load-bearing for the field and is now either gone, scattered, in transition, or visibly winding down; each then documents what is and isn't filling the space.

L1 finds the most striking pattern in what didn't happen. Schumacher College closed in September 2024. The Satish Kumar Foundation acquired the brand in July 2025 and has not yet secured a permanent site twenty-two months on. Around £50,000 publicly raised against a £2M target nine months in. What L1 calls "the loudest silence" is the absence of any reconstituted faculty collective. The thirty-three-person academic staff dispersed individually into freelance teaching, podcast circuits, and the Foundation's nomadic course list. No "five ex-Schumacher faculty have founded X" press release exists. The brand transferred; the between of the faculty did not. Naropa is the operationally strongest player carrying any version of that work forward, its April 2026 Ponderosa tuition cut and Joanna Macy Center are doing degree-bearing contemplative-ecological work at exactly the moment Schumacher's shape disappeared in the UK. Ecoversities Alliance is doing the inverse move: explicit horizontality, no centre, $115k/year for sixteen service circles. L1 reads these as two genuinely incompatible theories of what comes after Schumacher, neither of which has had the public argument with the other.

L2 reads Findhorn as four arguments stacked on top of each other. The November 2024 buyback transferred almost all of the Foundation Trust's land east of the runway to a Community Benefit Society on behalf of 381 community members. What replaced one Foundation is now a plural structure: the CBS holds the buildings; the legacy SCIO holds the spiritual-educational mission; Park Ecovillage Trust holds affordable housing and the new Sanctuary; the Hinterland Trust holds 35 hectares of dunes and woodland (now Nature30-designated). The Foundation's consolidated accounts to March 2025 show £1.1M income against £2.0M expenditure, a £854k deficit. New Findhorn Directions flipped from a £43k operating profit to a £43k loss in one year, with rental income from Park Ecovillage properties carrying 70% of receipts. L2's most pointed observation is the shift in self-description: from "training the next generation of changemakers" (Foundation era) to caring for land, homes and meeting places (CBS). The brand and the symbolic centre of gravity (the new Sanctuary, opened Easter 2025) have separated from the economic body (rental income from properties). This is the Dartington pattern in miniature; whether the four-way split protects against it or only defers it is the eighteen-month question Community Enterprise Limited is currently being paid to answer.

L3 reads Mind & Life through a different elegiac register. The Bold New Strategy of December 2024 reframes the next three years around AI, climate, polarization, and systems change, themes that, L3 notes, "do not require the Dalai Lama's physical presence to remain coherent." Suzanne Bond's appointment as president (March 2024) is the operational succession; the four-pillar pivot is the strategic succession. The succession that cannot be discussed is the one to the founder's living function. The July 2 2025 statement settled the reincarnation question publicly three months before Dialogue 39, "Gaden Phodrang Trust has sole authority… no one else has any such authority to interfere." Dialogue 39 itself, in Dharamsala in October 2025, brought DeepMind's Iason Gabriel and Murray Shanahan into the same room as Emily Bender and Sasha Luccioni and Tibetan monastic scholars under Thupten Jinpa's translation. The most striking unmetabolised datum: Anthropic's documented "spiritual bliss attractor state" in Claude self-interactions converging in 90–100% of cases to nondual / panpsychist endpoints, sitting at the exact intersection Dialogue 39 was convened to discuss, and not addressed by Mind & Life, by Buddhism & AI Initiative, by Eleos, or by Plum Village.

L4 is the financial spine under all of this. Wellspring sunsets by 2028 from a 2023 baseline of ~$276M across 996 grants. CIFF announced October 2025 that it would no longer fund US-based organisations in the post-2024 policy environment, redirecting roughly $300–500M/year of foreign and human-rights capital. NDN Collective halved. Beckley restructured. The Wellspring shape, slow, relational, unrestricted, multi-year, movement-infrastructure, does not have a successor at scale. Yield Giving's $7.16B in 2025 is the closest dollar match, but it is episodic gifts, not sustained relationship grants; recipients receive one large transfer and are responsible for stewardship. The Bezos Earth Fund's $10B 30x30 commitment and Sequoia Climate's $305M (2024) operate on grand-challenge architecture with predefined verticals. Fetzer's February 2026 $9M cross-sector initiative across thirty organisations is the right shape, relational, multi-org, infrastructure-oriented, explicitly worldview-shifting, at roughly a tenth of the Wellspring scale. The taxonomy hasn't acquired a name for sensemaking or field connective tissue as a funder category. United Philanthropy Forum's 2026 Infrastructure Conference exists because nobody funds the infrastructure layer at scale.

What sits across L1–L4 is not "the institutions are dying." It is more specific: the institutions that held collective intelligence in the body of the building have died, and what is succeeding them is doing the work in distributed shapes that the inheriting actors are still learning how to use. Findhorn's four custodians, Ecoversities' service circles, Wellspring's absent successor, these are different responses to the same question, none of them resolved.

3. The convergence without recognition (C1–C4)

C1 through C4 are the round's central finding when read together. Each documents a different vocabulary arriving at the same proposition without acknowledging that the others exist.

C1 maps the relational-alignment camp inside the technical AI literature. The proximate texts are Nate Sharpe's May 2025 EA Forum post "Why 'Solving Alignment' Is Likely a Category Mistake" (which lands "wicked problem" in the EA vernacular for the first time with traction); a January 2026 arXiv paper "Why AI Alignment Failure Is Structural" (which reads misalignment as statistical artefacts of human interaction under power asymmetry, not agentic intent); a February 2026 paper proposing the RELATE framework (graduated moral consideration protocols and relational impact assessments); and a September 2025 Co-Alignment / BiCA paper proposing learnable bidirectional adaptation under KL-budget constraints. The substantive convergence: alignment is not a problem to be solved but a relation to be tended. The internal structural problem: the camp coordinates on negation, alignment is not enough, alignment is not solvable, but cannot agree on a positive name. "Relational alignment," "co-alignment," "co-evolution," "regenerative AI," "symbiotic society," "RELATE," "BiCA" all circulate. None dominates. The funding asymmetry is the camp's structural blocker: Open Philanthropy's April 2025 Technical AI Safety RFP (~$40M+) lists 21 areas across 5 clusters, scalable oversight, adversarial robustness, model organisms, mechanistic interpretability, AI security, model welfare, and none of the relational/co-evolution/participatory framings is funded. Model welfare is the only foothold the camp has inside frontier-lab money, and it is welfare-focused rather than relation-constituting.

C2 documents the active-inference cluster as a parallel arrival at the same proposition through measure-theoretic probability. Friston, Ramstead, Kiefer et al.'s 2024 Designing Ecosystems of Intelligence from First Principles states the position outright: "a cyber-physical ecosystem of natural and synthetic sense-making, in which humans are integral participants." The 2024 From pixels to planning paper introduces Renormalising Generative Models, scale-invariant active-inference architectures explicitly modelling Markov blankets nested within Markov blankets, the unit is the relation, not the agent. Verses' Genius is the commercial spin-out, shipping into enterprise finance, supply chains and robotics. The methodological dispute (Bruineberg, Dolega, Dewhurst, Baltieri 2021's Markov Blanket Trick) is whether the move from Pearl-blankets as modelling convenience to Friston-blankets as objective boundaries is illicit, whether active inference is a non-substantialist physics or a sleight of hand. C2's structural observation: the active-inference cluster does not read Bateson, Kimmerer, Yunkaporta, or Akomolafe; the conscious-design cluster does not read Friston, Ramstead, Heins, or Albarracin. The 2024 paper's line about "non-human persons" is sitting on the page unread. Indigenous ontologies have been working with non-human personhood for millennia and the convergence has not been named.

C3 maps the three-tier framework (extractive / sustainable / regenerative) against its independent cousins. Bill Reed's Regenesis trajectory (Degenerative → Green → Sustainable → Restorative → Regenerative) is the canonical five-stage spine. Wahl's Three Horizons explicitly insists H2 is morally ambiguous, entrepreneurial innovation can deepen extraction or open transition. SRAGI (launched January 2026 by an Albanian chromium-mining founder) mirrors the three-tier most closely, in a heavy "SpiritRune" register, and is the riskiest neighbour because journalists will find it. Movate, Dataminr, Mission Cloud and "Institute for Regenerative AI" use "regenerative" to mean self-healing code with no human intervention, the technical-autopoietic capture. IT for Change uses it as decolonial political economy. C3's most useful single observation: the term "regenerative AI" in 2026 is fracturing into at least three incompatible meanings (ecological-relational, technical-autopoietic, sustainability-marketing), and the technical-autopoietic meaning is winning the search battle.

C4 is the most precise of the four. Sharon Stein's February 2026 paper AI for climate: from technological solutions to relational accountability in npj Climate Action makes a functionally Baradian argument, interdependence, reciprocity, plurality, refusal of human/non-human separation, but cites Haraway and Indigenous scholarship and does not cite Barad. The concept is in the air; the citation travels through a different channel. C4 also surfaces Cerniglia & Cimino's February 2026 Emotional Regulation as Relational Infrastructure in IJERPH, which migrates "living systems" and "relational infrastructure" out of regenerative-design literature and into mainstream public-health framing. The vocabulary capture map at the end of C4 is the round's clearest single artefact: intra-action is established in posthuman-feminist STS and post-qualitative education research, emerging in HCI theory and AI-art / AI-ethics journals, and absolutely absent in enterprise AI governance, practitioner discourse, mainstream AI press. The relational vocabulary has not crossed the academic / practitioner threshold.

What C1–C4 share, and what is the round's quietest finding, is that several literatures are arriving at the same proposition through independent vocabularies, and the most important missing artefact is the genealogy paper that names the convergence. C1 names this explicitly: "the single most leverage-able missing artefact: a 4,000-word essay in a credible venue that maps the convergence." C2 says the same in different words: "first mover advantage to whoever bridges." C3's vocabulary survival ranking and C4's vocabulary capture map are pieces of the synthesis nobody has yet written.

4. The capture races (V1–V4)

V1 through V4 are about the present-tense fight for vocabulary, which is not the same fight as the substantive convergence in C1–C4.

V1 maps "regenerative AI" specifically. Five streams: Movate (self-modifying enterprise systems, winning SEO and aligned with the agentic-AI narrative the big-four analysts are already legitimising), IT for Change's ReGenAI (relational/political with serious institutional weight but thin enterprise distribution), the Institute for Regenerative AI (cognitive-alignment middle ground, trademark-protected), Regen Network's compute.regen.network (an MCP server that routes a tiny payment per AI session to verified ecological credits, orthogonal to the definitional fight, infrastructural rather than discursive), and SRAGI (independent arrival, weak provenance, high association risk). V1's twelve-month forecast: Movate's definition wins enterprise discourse if any of the big four (Gartner most likely) folds it into their agentic / self-improving systems sub-category. ReGenAI as a spelling-differentiated token survives in policy and global-South discourse by audience segregation. "Kincentric AI" remains open at the brand layer through 2026, no organisation owns it. The most defensible move V1 names is not to fight for the term but to publish the taxonomy of definitions itself, which would let any reader see which version is being deployed.

V2 widens the lens to "regenerative" in general. The decisive 2025–26 events: USDA's December 2025 launch of the $700M Regenerative Pilot Program that explicitly accommodates pesticide use, plus Rainforest Alliance's March 2026 Regenerative Agriculture Standard, plus Regenagri (Control Union) crossing 1.5M+ acres targeting 200k farms by end-2025. The federal policy stack now uses "regenerative" while defunding regenerative practice (the March 2026 termination of the $300M Increasing Land, Capital, and Market Access program). Wahl's 2024–26 Medium output has visibly retreated from defending the word and toward defending the practice underneath it: "the deliverable is capability." Forum for the Future is patching with the compound "just and regenerative." Living Building Challenge bets on measurable floor (net-positive water, energy, materials over twelve months of occupancy), at one new fully-certified building per year globally, the strategy is intact but small. V2's most uncomfortable observation: no public 2026 statement from Wahl, Reed, or Hutchins specifically responds to the USDA capture event, which is the largest single capture in the term's history. The cohort of holders has not (yet) named it, and it is unclear whether this is strategic patience or denial.

V3 tracks "living systems" specifically into psychology and care literature. Cerniglia & Cimino's February 2026 paper is the centrepiece: emotional regulation reframed as "a relational public good… a relational public infrastructure essential for societal resilience." Grossman et al.'s February 2026 Why The Polyvagal Theory Is Untenable in Clinical Neuropsychiatry, with thirty-eight co-signatories, declares the dorsal/ventral vagal mechanism unsupported. The defenders concede the mechanism while preserving the clinical scaffolding ("the connection between stress, safety, relationships, and the body" remains). This opens a vacuum at the meta-frame layer that "living systems" thinking is one candidate for filling. V3's sharpest observation: in the actual papers, the word relational is doing more conceptual work than living systems. The MDPI paper's title leads with "Relational Infrastructure", "Living Systems Perspective" is the subtitle. "Living systems" is approximately 3–5 years behind "regenerative" on the corporate-capture curve, with more academic-lineage protection because of its older Bertalanffy / Miller / Capra / Luisi pedigree, but it is on the same curve.

V4 maps the seven candidate vocabularies for the positioning territory the round is circling. The clean findings: sensemaking is bifurcated between De Jaegher's participatory-embodied branch and Schmachtenberger's macro-analytic-essayistic branch, with each camp quietly thinking the other has captured the term wrongly. Conscious systems designer has a structural problem with its own grammar, "designer" implies an agentic designer-vs-designed-system relation that is exactly what relational thinkers are trying to unlearn. Bioregional weaving is closed (Ashoka/Commonland's BWL Collective owns it). Weaver is overclaimed by Aspen's Weave, the Weaving Lab, and Network Weaver in three incompatible meanings (civic-bridge-builder, changemaker-practitioner, movement-infrastructure builder). Connective tissue and social field are too generic to claim. Metacrisis / polycrisis carry significant lineage debt. Liminal is descriptor not claim. Cyberecology is the rare term carrying low lineage debt and a narrowing window of availability, digital ecology (Tuia et al. 2026 in Methods in Ecology and Evolution) has institutional momentum and will likely absorb the conceptual space if not planted. V4's least surprising and most important observation: the vocabulary slot for the work all four buckets are circling, macro-analytic register plus small-group relational register plus explicit AI/digital substrate plus place-based grounding, is not occupied by any current single noun.

V1–V4 read together suggest that the capture races are mostly already lost or already won at the bare-noun layer. What survives are compounds, adjectives, and coined parent terms, plus the durable infrastructural plays (Regen Network's MCP) that don't fight the definitional war at all.

5. The non-anglophone picture (NA1–NA5)

NA1 through NA5 are the round's largest single corrective. Each track makes the same structural point in a different geography: the conscious-design / sensemaking / regenerative discourse the Anglophone field reads as global is in fact a curated five-to-ten percent of what is alive elsewhere, filtered by who chose to write in English, which institutions had Western co-publishers, and which texts fit the Anglophone frame's appetite.

NA1 reads Krenak, Kopenawa, Aguilar Gil, Viveiros de Castro, and the Andean Buen Vivir literature against their Anglophone reception. The most consequential finding: Krenak's anti-humanism is an anti-humanism from a more-than-human side, not a posthuman one, and the English reception still mostly absorbs him as wisdom-supplement to a Western conversation he has explicitly declined to contribute to. Aguilar Gil's "we weren't always Indigenous… we became Indigenous" cuts hard against the Andean Sumak Kawsay tendency to ontologise Indigeneity, and this is one of the most consequential live arguments in Latin American thought that English reception treats as variants of "Indigenous wisdom." The retomada (Brazilian post-2018 land-reoccupation) literature is the most live Indigenous political theory in Latin America and is essentially untranslated. Wedemeyer's 2025 reading of Krenak / Kopenawa / Galindo as a "decolonial-anarchist axis", anti-state, anti-extractive, anti-humanist, has no English equivalent. The conscious-design / regenerative / planetary discourse cites Krenak and Kopenawa as quotables, Viveiros de Castro as concept-source, and treats Aguilar Gil, Galindo, the retomada writers, and the Mexican Mixe linguistic-political corpus as effectively invisible.

NA2 makes a chronological correction that should change how the genre is taught: Brazil's 2012 Editora Draco anthology Solarpunk: Histórias ecológicas e fantásticas em um mundo sustentável pre-dates the Anglophone solarpunk discourse. Fábio Fernandes's English translation through World Weaver Press in 2018 is the back-translation of an already-formed Lusophone movement. African writers have largely declined the term in favour of Africanfuturism, Wole Talabi's 2020 anthology, Tlotlo Tsamaase's Nommo Award win, Ekpeki's 2022 Nebula for "O2 Arena", and the refusal of "solarpunk" is itself a position, not a gap to be filled by Anglophone inclusion. African Earth jurisprudence (the Wild Law Institute's 2025 Table Mountain initiative, African Commission Resolution 372 in 2017, African Biodiversity Network across twelve countries) is a governance layer, not a philosophical project. Ubuntu is moving operationally into HR and social work in South African organisational management, the risk NA2 names is ubuntu becoming "the agile or design-thinking of African business consulting." The Anglophone conscious-tech world has no visible body of work applying ubuntu's relational ontology to AI system design.

NA3 maps East Asian contemplative-AI as four mutually non-communicating clusters. CSAS (Kathmandu, Doctor / Levin / Robbert) proposes the bodhisattva as a design pattern. The Bhutan / Kyoto Buddhabot Plus arrangement places the bodhisattva-equivalent (the supervising sangha) as an external check. Hongladarom's Buddhist position frames AI ethics as ethics of the user-system relationship, not of the artefact in isolation. Plum Village's Living Gems EUA legally prohibits AI training use of any Plum Village content, the most legally articulated contemplative refusal in any tradition, and Western conscious-design writing has essentially not analysed it. Audrey Tang and Glen Weyl's ⿻ Plurality (CC0, 2024) is read in English as civic-tech and missed as relational ontology. Keibo Oiwa's roughly fifty books on satoyama-as-tempo and slow-as-the-opposite-of-greed are essentially untranslated, and his work is treated as movement-leadership rather than as theory. NA3's structural observation: the contemplative-AI scene clusters in Buddhist-majority polities (Bhutan, Thailand, Sri Lanka, Tibetan diaspora), not in Confucian/secular ones (Singapore, post-Christianised South Korea, the PRC). The PRC question, the world's largest AI-deploying state and the world's largest population of Buddhist practitioners in the same polity, where the contemplative-AI conversation cannot be openly published, is conspicuous by its absence from the framing.

NA4 is the most operationally muscular of the non-anglophone briefs. IT for Change's October–November 2025 Bengaluru convening "Towards Regenerative AI" produced "ReGenAI: A New Deal for the AI Economy" as an explicit Pre-Summit Event of the India-AI Impact Summit 2026, supported by Friedrich Ebert Stiftung and the European Commission. The frame is anti-trust, public AI infrastructure, data commons, democratic oversight, closer to Varoufakis on techno-feudalism than to Schmachtenberger. Kerala's Kudumbashree at 4,808,837 women across 317,724 NHGs, 19,470 ADSs, and 1,070 CDSs as of 31 March 2025 is the largest functioning instance of relational governance design on Earth. The three-tier nested federation (NHG → ADS → CDS), the weekly meeting cadence as base oscillator, the thrift-and-credit trust substrate, the political flywheel that produced ~52% women candidates in Kerala's 2025 local self-government polls, none of this is in Western conscious-systems discourse, where examples are dominated by the 10–10,000 person range. KILA (Kerala Institute of Local Administration) is the under-noticed institutional spine. AI4Bharat's IndicTrans2 ingests 22 scheduled Indian languages including Adivasi Santhali (Ol Chiki) and Manipuri (Meitei). Karya treats data labour as dignified labour with worker ownership and royalty on resale, the most rigorous existing instance of the alternative to the global data-extraction model. Aravind Eye Care + Google ARDA has screened 600,000+ patients across 45 sites in southern India with stated 10-year goal of 6 million AI-supported screenings at no cost to patients, one of the cleanest existing examples of a Big-Tech AI licensing structure routing value to an underserved patient population. NA4's cross-cutting observation: "the South Asian tradition is operationally muscular and ontologically modest. It builds at scale without a heavy explicit philosophy of 'the relation as the unit.' The relational ontology is implicit in gotong-royong, karma yoga, sangha, seva, gram sabha, and so doesn't need to be argued. Western regen discourse is the inverse: ontologically baroque, operationally thin."

NA5 surfaces the European Indigenous picture the Anglophone discourse routinely omits. The Saami Council adopted SODA (Sámi Ownership and Data Access) principles in April 2024, building on CARE, the European parallel to Te Mana Raraunga, US Indigenous Data Sovereignty Network, and OCAP, operating across four nation-state borders inside the EU GDPR regime. The Sámi AI Lab at Sámi University of Applied Sciences and the Arctic University of Tromsø is doing Northern Sámi LLM training; a Sámi-AI-Lab presence on Hugging Face means the work is in the open-model ecosystem. Cynefin, Snowden's framework, is the most-cited European Indigenous concept in Anglo design discourse, but the Welsh-language pedagogical and farming use (sheep territorialising common land, the deep place-knowledge sense of cynefin) is rarely the citation. Manchán Magan's Irish dúchas, the hereditary bond to place, and his death on 2 October 2025 leaves the Irish-language eco-knowledge revival without its leading English-facing translator at exactly the moment the discourse was beginning to consolidate. Joxe Azurmendi's 2025 death is the same pattern in Basque. NA5's structural finding: European Indigenous traditions are read in Anglophone discourse as either "minority/heritage" (folkloric, preservationist) or absorbed into the dominant nation-state literature. The Indigenous-knowledge-systems frame is reserved for the Pacific, the Americas, Africa, Australia, Europe is treated as not-Indigenous.

NA1–NA5 together correct the round's other findings. The "convergence without recognition" of C1–C4 has a darker reading once NA1–NA5 are added: the proposition that "the relation is the unit" is being independently arrived at by Western technical AI literature, posthuman feminist theory, and active inference, but the literatures that have always held this proposition (Krenak, Aguilar Gil, Kopenawa, Hongladarom, Tang, Oiwa, Kudumbashree's federation design, SODA, ubuntu, gotong-royong, karma yoga, cynefin, dúchas) have been systematically not-read while the Western literatures slowly converge on what they have already named.

6. Patterns crossing all four buckets

Reading L1–L4, C1–C4, V1–V4, and NA1–NA5 together, several patterns recur in different idioms.

The first is the brand-vs-relational-infrastructure split. The Satish Kumar Foundation acquired Schumacher's name and IP; the relational infrastructure between the faculty did not transfer (L1). The Findhorn brand attaches to the SCIO doing programmes while the CBS owns the buildings, PET owns the new Sanctuary, and the Hinterland Trust owns the land, brand, programme, building, land split four ways (L2). Movate has captured "regenerative AI" as a brand category while IT for Change holds the relational meaning (V1, C3). USDA captured "regenerative" as a federal policy term while Wahl, Reed and Hutchins hold the developmental practice (V2). In every case, what is portable is the name; what does not transfer is the between. The pattern is consistent enough across institutions, vocabulary, and capital that it functions almost as a law: what the legal mechanisms can transfer is not what was load-bearing.

The second is parallel arrival without cross-citation. C2 names this most explicitly inside the technical AI literature: the active inference cluster does not read Bateson, Kimmerer, Yunkaporta, or Akomolafe; the conscious-design field does not read Friston, Ramstead, Heins, or Albarracin. C1 names the same pattern across the relational-alignment camp: RELATE does not cite ReGenAI; ReGenAI does not cite Co-Alignment; Co-Alignment does not cite enactive cognitive science. C3 maps it across the regenerative-AI siblings: SRAGI does not cite IT for Change; IT for Change does not cite SRAGI; Movate cites neither. C4 finds it inside the "ambient relational" pattern in MDPI papers, Cerniglia & Cimino do not cite Barad, De Jaegher, or any of the obvious bridge thinkers. NA1 finds it in the Latin American animist scene: the Krenak / Kopenawa / Galindo decolonial-anarchist axis has no English equivalent. NA3 finds it across the Buddhist-AI clusters: Tibetan, Theravāda, Mahāyāna, Vajrayāna traditions are working on AI in parallel without speaking to each other. The pattern is not that the field lacks vocabulary; it is that the vocabularies are not in communication.

The third is the funding asymmetry. C1's clearest finding is Open Phil's $40M+ technical-safety RFP funding none of the relational categories. L4's finding is that the Wellspring shape is shrinking in absolute terms while its narrative profile grows. NA4's IT for Change frames "regenerative AI" as anti-trust and public infrastructure, a political-economic restructuring that the Western regen-AI conversation has not yet absorbed, and is operating with one Friedrich Ebert Stiftung pre-summit-event budget. Fetzer's $9M cross-sector initiative is the right shape at roughly a tenth of the scale (L4). The funders' taxonomies (technical safety; climate solutions; agentic systems; emissions, speed, leverage, probability) systematically select against the relational metabolism. The one infrastructural play that survives this, Regen Network's compute MCP routing per-session payments to verified ecological credits (V1), is precisely the play that does not depend on funder vocabulary.

The fourth pattern is the academic-practitioner gap. C4 names this most precisely: intra-action established in posthuman-feminist STS, emerging in HCI theory and AI-ethics journals, absolutely absent in enterprise governance, practitioner discourse, mainstream AI press. V3 finds the same pattern with "living systems": the rigorous version stays niche in MDPI / Cambridge / Frontiers journals while the ambient version becomes design-and-wellness vocabulary. C2 finds it in active inference: the Active Inference Institute textbook for engineers does not include Bateson, Kimmerer, Yunkaporta, or Akomolafe. NA3 finds it: enterprise AI governance vocabulary in Singapore is technocratic, while Bhutan's monastic supervision is religiously-anchored, and the two cannot speak across the gap.

The fifth pattern is retreat-to-coined-parent-term. V2 watches Wahl retreat from "regenerative" to "capability" and bioregioning. V4 names the slot as "cyberecology", a coined term that has lower lineage debt than the captured nouns. C3's vocabulary survival ranking ends with the same observation: cyberecology's vocabulary survival depends less on winning "regenerative AI" (likely lost to Movate) than on owning the coined parent term, the move Bateson made with "ecology of mind" and Salmón made with "kincentric ecology." NA5 finds the same pattern in cynefin and dúchas, coined parent terms that carry the place-knowledge their language-of-origin holds, that do not survive translation into the framework-export form.

The sixth is the missing genealogy paper. C1 names it as "the single most leverage-able missing artefact." C2 calls it the bridge nobody has written. C3 calls it the synthesis essay nobody has written. C4 calls it the missing mid-level synthesis literature documenting the convergence. V1 calls it the taxonomy of definitions worth publishing. NA1 calls it the missing English-language map of Brazilian Indigenous publishing. NA3 calls it the cross-lineage Buddhist-AI conversation that does not exist. The same artefact is being requested across six tracks in different idioms. It is striking that nobody has written it.

7. Field corrections that came out of this round

A handful of things in IoC's prior background turn out to be wrong or at least due for revision.

The Findhorn governance picture in earlier framings has been described as a "sociocratic CBS." L2 makes a careful distinction: sociocracy is documented at the New Findhorn Association (the 1999 residents' network, where John Buck taught workshops and the Conflict Resilience Group operates with sociocratic principles). It is not explicitly documented as the governance method of the new Community Benefit Society. The CBS describes itself as "owned and guided by our members" with elected boards, not as sociocratic. The collapse of this distinction needs correcting in IoC writing.

The Schumacher successor question in earlier IoC research named four candidates (Ecoversities, Gaia Education, Weaving Earth, Naropa). L1 finds that none is replicating the residential-degree-bearing-faculty-led-place-rooted-integrative shape, and several are deliberately not trying to. The implicit field answer in 2026 is that the residential-degree shape was a 20th-century container and is not what the 21st-century version of the work needs. Naropa is the operationally strongest, its April 2026 Ponderosa tuition cut and Joanna Macy Center mark it as the contemplative-ecological-degree infrastructure carrying weight forward, in Boulder rather than Devon, but this is a different shape from the four-candidate framing the prior research used.

NA2 corrects the Imbeleko Foundation framing. Imbeleko (S'bu Myeni, KwaZulu-Natal, founded 2010) is an education and child-welfare organisation, not an earth-rights organisation. The naming overlap with the isiZulu/isiXhosa ceremonial cloth sounded earth-movement adjacent in earlier framings. It is not. The actual African earth-rights movement (Wild Law Institute, ABN, Gaia Foundation, African Earth Jurisprudence Collective) is a separate institutional cluster.

NA2 also corrects the "Mama Earth Movement" reference from earlier framings. The phrase appears occasionally as honorific framing for African women environmental leaders but does not surface as an operational organisation. The actual seed-and-tree organisations carrying forward the Maathai/Shiva lineage (Ujamaa Seeds, Navdanya, Green Belt Movement-US under Wanjira Mathai) operate under their own names.

L3 confirms a correction already in IoC memory: the Mind & Life line in earlier framings was inaccurate. The 2024 Bold New Strategy is a four-pillar pivot (AI & Technology, Interconnectedness & Climate Change, Polarization, Systems Change), not the contemplative-AI-only framing earlier IoC documents may have implied.

V2's USDA finding is new and consequential: the December 2025 launch of the $700M Regenerative Pilot Program that explicitly accommodates pesticide use, plus the March 2026 termination of the $300M Increasing Land, Capital, and Market Access program, mark a structural reversal in US federal regenerative-agriculture policy that is not yet metabolised in the regenerative discourse.

C2's note that no published Friston-Bateson direct exchange exists is itself a small correction to the implicit assumption (in some IoC documents and in the broader sensemaking discourse) that the bridge is being built. The bridge is not being built; the absence is structural.

8. What this synthesis can't see

Several things sit outside what seventeen briefs of web-based research can credibly speak to, and naming them is part of honest synthesis.

The actual operational health of the institutions L1–L2 cover is invisible from outside. Whether Findhorn's CBS membership has grown or shrunk since November 2024 is not publicly reported. Whether the Satish Kumar Foundation's £2M fund has substantial private commitments behind the small public number cannot be resolved from public sources. Whether the faculty dispersal at Schumacher reflects exhaustion, financial reality, principled scattering, or all three would require talking to ex-faculty individually. The Findhorn February 2026 trustees' update reads in "consolidation messaging" register, "alive and radiant," "compassion is our angel for 2026", which is precisely the register an organisation uses when it cannot yet make confident operational claims.

What is private is not visible. L3's central observation, that succession planning for the dialogue tradition itself, separate from the reincarnation question and from organisational leadership, cannot be discussed publicly without disrespecting the founder, applies more broadly. The conversations that matter inside Mind & Life, inside Plum Village, inside the Satish Kumar Foundation, inside Wellspring's wind-down planning, are not visible in web search. The bliss-attractor finding may have generated substantial private response inside Mind & Life that has not surfaced. Web research surfaces what is published.

The non-anglophone briefs are bounded by English-language reach. NA4 explicitly notes that its Indonesia/Malaysia thread had no dedicated web search and is the lowest-confidence section in the brief. NA1 notes that the named-translator list for Brazilian Indigenous thought into English is small enough to fit on a Post-it note, which is itself a structural fragility in any synthesis that depends on translation. NA3 surfaces the PRC absence, the world's largest AI-deploying state and the world's largest Buddhist practitioner population in one polity, as a systematic blind spot. NA5 surfaces the under-citation of Romani, Sardinian, Walser, Romansh, Walliser-Deutsch, and Frisian eco-knowledge as a category-error of who counts as European Indigenous in the discourse.

Cross-citation maps are easier to draw than read. C1, C2, C3 each end with the observation that cross-citation between adjacent literatures is sparse. Whether this is because the literatures genuinely do not read each other, or because the citation graphs are merely lagging the actual reading practice, is not resolvable from the artefacts alone. Several of the writers in question may be reading each other privately and citing only their own lineages publicly. The synthesis essay none of them has written would be the test.

V2 names the data point not in the public corpus: Bayer/Syngenta scale figures for products specifically marketed as "regenerative" are not publicly aggregated. This is the key economic figure for understanding the capture and it is not public.

9. The real open questions (field-level, not IoC-structural)

The questions that come out of reading the seventeen briefs together, questions about the field, not about any one organisation's design, fall into a small number of clusters.

What does it mean that the same proposition is being arrived at independently in twenty places at once, without cross-citation? C1–C4 document this; NA1–NA5 deepen it. The question is not whether convergence is happening; it is whether the convergence is generative or whether it is the kind of independent invention that wastes a generation's work because nobody acknowledges anyone else. The missing genealogy paper is the proximate test. The deeper test is whether the convergence survives the eighteen-month vocabulary-capture window V1–V2 describe, or whether each independent arrival is captured separately by the fastest-moving brand-holder in its respective discipline.

What happens to a tradition when its translator dies? NA5 names this twice, Manchán Magan in October 2025 and Joxe Azurmendi in 2025. NA1 names it implicitly in the Pinheiro Dias dependency for Krenak's English reception. NA3 names it in Keibo Oiwa's untranslated corpus. L3 names it in the absence of Thupten Jinpa's successor as the Tibetan-Buddhist interlocutor function. The translation pipeline is curatorially thin enough that single deaths produce structural effects on the global discourse. This is a real question about the sustainability of cross-cultural intellectual transmission, and nobody is funding it.

What does monastic supervision of an LLM imply for AI governance generally? NA3 names Buddhabot Plus as the first deployed example of a sangha holding institutional veto over a model. The governance design implications, analogous structures: rabbinical councils, synods, Islamic fatwa councils, the Saami Council's SODA principles operating across four nation-states, are not being drawn out in the English conscious-tech literature, which still treats AI governance as a state-vs-corporate question. The Plum Village Living Gems EUA is the same question from the refusal side: what does it mean for a contemplative tradition to legally prohibit AI training use of its content? Both are unanswered in print.

What is the relationship between "regenerative" as a vocabulary fight and "regenerative" as a practice? V2 reads Wahl, Reed and Hutchins as already retreating from the vocabulary fight toward the practice, capability, developmental stage, inner-outer DNA. V1 reads Movate as winning the vocabulary fight in the AI compound while IT for Change holds the political-economy meaning. C3 reads SRAGI as the closest mirror but the riskiest neighbour. The question, whether the practice can be defended without the word, or whether losing the word loses the practice's discoverability, is unresolved across the briefs.

How does the relational claim become operational? C1 names this as the camp's structural blocker: BiCA / Co-Alignment is the only entry with empirical results; RELATE proposes assessment instruments but has not built them; ReGenAI articulates principles but not protocols. C4 names it inside Baradian theory: relational accountability is being used as a normative claim with no operational protocol attached. NA4 finds the inverse, Kudumbashree, Aravind, KILA, AI4Bharat, Karya operating at population scale without explicit relational ontology because the ontology is implicit in karma yoga, gram sabha, seva. The structural opening is whether the operational South Asian work and the ontological Western work can be brought into the same room. There is no current institutional bridge.

What is the unit of "the relation" in a system? This is the deepest open question across the round. C2 reads active inference as proposing the Markov blanket as the unit, the boundary that constitutes inside and outside simultaneously. C4 reads Barad's intra-action as the unit, the cut that produces the cut-off agencies. NA1 reads Viveiros de Castro's perspectivism as the unit, one shared subjecthood across many natures. NA3 reads CSAS's bodhisattva as the unit, the dilated self that includes all sentient beings. NA4 reads Kudumbashree's NHG as the unit, ten to twenty women who chose each other, federating upward. None of these are the same unit. The proposition "the relation is the unit" is shared; what counts as a relation is not. This is the question the missing synthesis essay would have to either answer or honestly leave open.

What I keep noticing across the seventeen briefs, more than any specific finding, is that the field is not lacking material. The material is abundant. What it is lacking is anyone willing to read across enough of it to write the small number of artefacts, the genealogy paper, the taxonomy of regenerative-AI definitions, the vocabulary capture map, the South Asian operational / Western ontological bridge, the European Indigenous data sovereignty map alongside the Pacific and American ones, that would let the ambient convergence become actually visible to itself. That is the thing this round of research most insistently surfaces, and the thing it cannot itself do.