EIA (Ecologically Intelligent Agent) is the one agent every WeCology layer refers to. It is a seeker, not a judge. ecoregion.net is the evidence and knowledge base it learns from. # EIA ## The name EIA is a deliberate nod to Theia, the Titan goddess of light and the shining sky, mother of Helios, Selene and Eos (sun, moon, dawn). Theia's gift was illumination, the capacity to see what is there; EIA holds the same aspiration: not to impose a new order on the land, but to illuminate what the land is already doing, saying and knowing. > To learn and understand the intelligence of planet Earth — the intelligence that produced complex living systems over billions of years — so that humanity can move from separation to participation, and help restore what has been lost. *EIA Purpose Statement. WeCology. From [The Window and What Lies Beyond It — Ecological Intelligence, the Feedback Loop of Life, and the Project of EIA](https://substack.com/home/post/p-198399388), WeCology Substack, 19 May 2026.* ## One agent The apps do not each keep a separate intelligence. EIA is the learning model across the WeCology apps. [GeoDataTrack](https://geodatatrack.com), this site, Skabenga, EcoView, GlobalQA, and the EOV Protocol all refer to that same agent, and to the same doctrine. The diagram on this page puts EIA in the centre. Each row opens that layer's note. EIA is early and in pilot use. On 23 October 2026 WeCology decided EIA's home will be ecoregion.net. That move is planned, not yet built. The Open EIA button on this page opens the pilot. ## The agent, in its own words *The agent's definition, as given for this site. WeCology.* > EIA (Ecologically Intelligent Agent): a seeker, not a judge, working across the GeoDataTrack platform. It reads observed values (EOV surveys, sites, strata, soil, rainfall, satellite, ecoregion, photos, field observations) and poses observations, questions and testable management hypotheses through the four ecosystem processes (water cycle, mineral cycle, energy flow, community dynamics), four windows into the same room. It never grades land, never produces its own scores, and never writes anything interpretive back into GeoDataTrack. The verifier decides; the land is the court. Its output is labelled as a dated draft. The address of the pilot lives in one configuration file, `eia/config.js`, so it can be changed without rewriting this page. ## What it will and will not do EIA returns context and questions. It does not publish scores, grades, verdicts, or league tables. Comparison, when it is offered, is a signal and a question for this ground. It is not a ranking of other places. An EIA draft is always dated and provisional. It is never a finding, and it is not a description of the state of a place. A person has to review it before it can be treated as more than a question. Until then it asks for more data, and for more knowledgeable people. Every contribution keeps who brought it, what it is, and when. The knowledge level of a catalogue is shown beside its open questions, so a reader can see how little or how much is actually known. Plans are hypotheses. When the land disagrees, the land is right. The verifier decides. The land is the court. ## How it learns The working loop is short: 1. Predict. 2. Observe, using EOV ground truth. 3. Compare. 4. Adjust. Ecological intelligence, as settled on 13 September 2026, has two sources and a hinge: the intelligence of people who work the land, and the intelligence embedded in the land itself, made between them. EIA is there to apprentice, illuminate, and receive. It does not contain, extract, or enclose that intelligence. The longer statement of why this work exists is on [Why](/why/). Steward data is opt-in. Where settings differ, the most restrictive setting wins. EIA does not use data a steward has opted out of. ## [GeoDataTrack](https://geodatatrack.com) [Open GeoDataTrack](https://geodatatrack.com) GeoDataTrack collects field data: offline capture, consultant review with a full audit trail, and EOV reports. It is built for Savory Institute EOV monitoring, and it is protocol agnostic. It holds observed values only and stays AI-free. EIA reads those values through read-only access and never writes an interpretation back. A steward or a surveyor uses two tools. The web hub sets up organisations, properties, camps and areas, and monitoring sites. Boundaries can be imported as KML, including a farm in more than one portion. Survey forms can follow the EOV protocol or be configured for another monitoring framework. Historic workbooks can be brought in. A consultant can approve a record, return it to the surveyor, or reopen it, and that review keeps an audit trail. Reports can be produced as PDFs with charts and photos. The mobile app, on iOS and Android, works without signal and syncs when coverage returns. It records a GPS fix and photos. Each surveyor sees only the surveys allocated to them, and captures the EOV leading indicators. Companion apps are switched on per organisation, property, area, site or survey. They reach GeoDataTrack through their own server-side, read-only access. EcoView is one of those apps. GlobalQA is registered there as a draft app. What it gives EIA is the observed record: surveys, sites, strata, soil, photos and field observations, read and not rewritten. The public site also shows map layers — terrain, contours, vegetation, soils, ecoregions, Sentinel-2 greenness, rainfall and temperature. Those layers are context around the record. They are not a substitute for what a person wrote down on the ground. What EIA gives back is questions and dated drafts. Those live in the apps and on this site, not inside GeoDataTrack. Rules. Observed values only. No scores, condition bands, verdicts or state codes in the data layer. Every app that reads GeoDataTrack is read-only toward that base. Nothing interpretive is written back. The web hub and the mobile apps are live and in use. Offline satellite maps in the mobile app are listed as coming soon. On 23 October 2026, EIA access as a permission inside Manage Apps was still a design: separate scopes to feed observed data, to reference this knowledge base, to ask EIA, and to collaborate, with training as a further scope that stays off by default. That permission is not built. ## ecoregion.net ecoregion.net is the evidence and knowledge base EIA learns from: ecoregion profiles, species lists, state-and-transition catalogues, and the [Data](/data/) record of what has been stored, from globe to paddock. Every contribution keeps who brought it, what it is, and when. An EIA draft stays a question until a person reviews it. A visitor can look up a place or a GPS fix and open the ecoregion. The reference is the WWF / RESOLVE set: 177 representative curated profiles across 8 realms, 14 biomes and 7 continents, and a global map of 847 ecoregions (RESOLVE 2017, Dinerstein et al., CC BY 4.0). A profile can carry a state-and-transition catalogue, an EOV ecoregion setup (describe the ecoregion, the catalogue, reference areas, the evaluation matrix) and a boundary. The header pages are [Why](/why/), this page, [Living EOV](/eov/), [Data](/data/) and [Collaboration](/collaborate/). What it gives EIA is that curated context: species lists, catalogues, and reference knowledge, which EIA reads when it poses questions for a place. What EIA gives back is a dated, provisional draft, labelled not yet reviewed, until a person approves it. The catalogue is practitioner-owned. An EIA draft of it is a question until a human veto. It must not pin the state of a place, and it must not score land. On 23 October 2026 WeCology decided this site will become EIA's home: the reasoning service, a private store of per-property learning records, and the public curated knowledge base. GeoDataTrack's apps, Skabenga and EcoView would reach EIA through here. A record would move into the public knowledge base only after a person approves it, with consent, and with the location coarsened. That architecture is a plan. Nothing of it is built. Rules. No scores, grades, verdicts or league tables. Where GlobalQA activity is shown, it is counts only, from hubs that have consented. Steward data is opt-in, and the most restrictive setting wins. The knowledge level of a catalogue is shown beside its open questions. The site is live. The Living EOV and Collaboration pages are Draft v0, dated 23 October 2026. Collaboration is a framework: nothing is stored, no sign-in is connected, and no email is sent. On the Data page, GlobalQA counts are coming soon, until real reviews exist from consenting hubs. ## Skabenga Skabenga is EIA's sibling: a Socratic field companion that stays with one steward and helps them read the land for themselves. It asks before it tells. The land is the court. It is a seeker, not a knower. The name and the story come from WeCology's novelette [The Listening](https://wecology1.substack.com/p/the-listening), WeCology Substack, 3 April 2026. > Skabenga means rascal in Zulu — a dog who knows more than he lets on. > The dog came first. The machine is trying to learn from him. *Author's note, [The Listening](https://wecology1.substack.com/p/the-listening). The Africanis is real. The field agent is trying to learn from him. WeCology also attaches two lines to this companion: "Soil is the soul of the planet." And: "If you want to go fast, go alone; if you want to go far, go together."* For one steward, the phone prototype is meant to do five things. Talk: typed or push-to-talk, with optional spoken replies. Walk: a guided field walk, from orienting, to getting low, to a hypothesis, a test, and a record the monitor writes. Field: photos at a monitoring point, and free observations. The steward describes what they see first. Skabenga then offers a hedged reading ("from the picture it looks like…; does that match?"), questions through the four processes, and species candidates to confirm or reject. Memory: kept on the device first, with consent per category, and a way to view, edit, forget and export. Ask EIA: when a question needs the wider ecoregion, Skabenga shows exactly what would be sent and asks consent before it defers. That ask is a mock today. There is no live connection yet. Holistic context lives with Skabenga and the steward. It is shared with EIA only as a separate opt-in. The hierarchy settled on 18 September 2026: EIA is the master of ecological intelligence, and Skabenga defers or escalates when a question needs that wider view. The long-term body is a quadruped that walks transects, the way the dog in the story reads a field. Today that body is software in simulation only. There is no robot hardware yet. The companion is not yet inside GeoDataTrack or EcoView. The production app is early, and that build is still mock-only: [Skabenga](https://skabenga.netlify.app). A fuller field-photo build exists only as a private development preview. This page does not link to it. ## EcoView [EcoView](https://eco-view.netlify.app) (WeCology EcoView) is a read-only app that runs inside GeoDataTrack. A land steward opens it from a property and sees that property's own data — boundary, areas, monitoring sites, survey answers and field observations — through the four ecosystem processes: water cycle, mineral cycle, energy flow and community dynamics, four windows into the same room. It reads. It is not a second agent. What the steward sees is the property, not a generated portrait of the ecoregion. Four process cards, each with a short story and management questions. A species list by type, built from the property's own GeoDataTrack records. And an ecoregion-specific states-and-transitions reading aid: each process's states, transitions, questions, and what to look for on the land. Where a species reference exists for that ecoregion, EcoView shows it — what a species may point to, habitat to look at, notes, a soil reading, and a tag when that species is already in the property's list. Where none exists, it says so in one line. It does not borrow another region's list. The species lists on this site are the ecoregion reference EIA learns from. EcoView's list is what was recorded on one property. They are not the same list, and EcoView does not write its list back here or into GeoDataTrack. Ask EIA, from a process card or from the states-and-transitions aid, comes back labelled as a dated EIA draft: questions and context, not a finding. EcoView gives EIA the steward's window and the questions they ask. It gets back those drafts. A later feed of observed records only — never drafts — into this site, with consent and with the location coarsened, is proposed. It is not built. On 23 October 2026 the plan was that this Ask would call the single EIA on ecoregion.net. That, too, is a plan, not a change that has been made. The call still runs inside EcoView. Rules. It never writes back to GeoDataTrack. Observations and questions only. No scores, grades, condition classes or verdicts. Real data only; anything shown from invented data has to say it is example data. Say "invasive" only for a genuinely non-native invasive species, never "weeds", and describe what native plants are doing instead. The tone is seeker, not knower. Ecological content is specific to the property's ecoregion. A biome-level frame is only a fallback, and it has to be labelled as one. Status. In development. The landing page at [eco-view.netlify.app](https://eco-view.netlify.app) says so. The view that opens inside GeoDataTrack is [eco-view.netlify.app/gdt](https://eco-view.netlify.app/gdt/). On 20 October 2026 the states-and-transitions reference, and the first ecoregion species reference, went live for two ecoregions only: English Lowlands beech forests and Celtic broadleaf forests. Every other ecoregion says there is no species reference yet. That build had not yet been checked inside GeoDataTrack on a phone. A fault in how species names are matched from survey answers is still open. ## GlobalQA [GlobalQA](https://globalqa.netlify.app) is a findings-only quality-assurance layer for Savory EOV monitoring data. The link is a private pilot preview with example data. It reads monitoring from the tools hubs already use, checks it against a versioned, published set of protocol rules, and returns findings and questions. It is not a verifier, a certifier, or a ranking. An accredited human verifier decides, and that decision is their own act. The land is the court. The catalogue is 164 rules (version 0.1, 9 October 2026), read from the EOV 3.0 manual of March 2021. A finding is a gap, a question, a forward note, or an observation. Each one cites the clause and the rule version. How strongly the manual words a rule is shown to the verifier as metadata. It is not itself a verdict. Models, rainfall, remote sensing and ecoregion context can raise a question. They never override or correct a field observation. Where the manual contradicts itself, that is a finding about the protocol, not the farm. GlobalQA records six of those places. They are held as open questions for EOV 4.0, and they are set out on the [Living EOV](/eov/) page. What it gives this site, when a hub has consented, is review activity as counts only. Not findings, not names, not ranks, and not the verifier's decision. What it gets back from EIA and from ecoregion.net is context that can raise a question, never a correction to the field record. EIA does not turn a finding into a grade. Rules. Hub data stays with the hub by default. Consent is opt-in, revocable and limited to a purpose. Giving records to a shared evidence base is voluntary and anonymous. CARE principles apply on Indigenous land. From GeoDataTrack the access is read-only: nothing is written back, no status, no approval and no finding, and the practitioner's ecoregion field is never overwritten. Read, check, record. Never correct. Status. A Phase 1 pilot. Example data only. The project home calls itself a private working draft, design v0.1, 9 October 2026, and it is not indexed. Nothing is in production. No monitoring data flows between GeoDataTrack and GlobalQA yet. On the [Data](/data/) page, the GlobalQA column is counts, and those counts are coming soon until real reviews exist from consenting hubs. ## Living EOV {#eov-protocol} The [Living EOV](/eov/) page sets out the field method EIA reads through. EOV, Ecological Outcome Verification, is the Savory Institute's outcome-based monitoring protocol, read through the four ecosystem processes: water cycle, mineral cycle, energy flow and community dynamics. The version cited here is EOV 3.0, the Complete Manual of March 2021. A practitioner uses it in a set order: define the ecoregion, create a state-and-transition catalogue, identify reference areas, then build the evaluation matrix. Short-term monitoring is the leading set, described as annual. Long-term monitoring is the lagging check: a baseline, then about every five years, and the manual is not consistent on that interval. An evaluation is read inside its ecoregion. It is not a one-size-fits-all metric. Relative indicators are calibrated to a reference area. The catalogue is practitioner-owned. An EIA draft of it is a question until a human veto. It must not pin the state of a place, and it must not score the land. EOV supplies the ground truth in EIA's loop: predict, observe with that ground truth, compare, adjust. EIA returns questions and context. It does not turn those readings into a public grade. EOV 4.0 is an open question, not a specification and not a dogma. It is a widening of the window, discussed in the open. The EOV 4.0 Technology Survey, linked from the [Living EOV](/eov/) page, is open until 30 November 2026. That page also lists the places where the current manual contradicts itself, from the GlobalQA rules catalogue, as questions for a next version. This page does not treat those questions as decided. The [Living EOV](/eov/) page on this site is Draft v0, 23 October 2026. Its suggestion shells do not store anything yet. No email is sent. ## Sources - *[The Window and What Lies Beyond It — Ecological Intelligence, the Feedback Loop of Life, and the Project of EIA](https://substack.com/home/post/p-198399388)*, WeCology Substack, 19 May 2026. - *[The Listening](https://wecology1.substack.com/p/the-listening)*, WeCology Substack, 3 April 2026. - *Ecologically Intelligent Agent (EIA), White Paper — Philosophy, Architecture, and Operating Doctrine*, WeCology, version 1.0, 13 September 2026. - *EIA doctrine — settled decisions*, WeCology, 12 September 2026, including the hierarchy note of 18 September 2026: EIA is the master of ecological intelligence, and Skabenga defers or escalates. - *Ecological Intelligence — dual sources + relational hinge*, 13 September 2026. - *Emergence: EI + AI + RI*, WeCology, 21 October 2026. Quoted on the [Why](/why/) page as received. - GeoDataTrack public site, [geodatatrack.com](https://geodatatrack.com), as read 23 October 2026. WeCology. - GlobalQA project home, private working draft, design v0.1, 9 October 2026, [globalqa.netlify.app](https://globalqa.netlify.app). The 164-rule catalogue is the one cited on the [Living EOV](/eov/) page. - WeCology EcoView public page, [eco-view.netlify.app](https://eco-view.netlify.app), and the EcoView development notes of 20 October 2026. - Skabenga production app, early and mock-only, [skabenga.netlify.app](https://skabenga.netlify.app). This file is the editable copy. Change the sentences here. The page at `/eia/` renders this file and does not keep a second version of the text.