Rewilding the Field

There are paddocks in Western Australia’s wheatbelt — tens of thousands of hectares — where the soil crusts white in summer. Salt. The farmers who cleared the mallee scrub for “improved” pasture in the mid-20th century didn’t intend disaster. The short-term arithmetic made sense: clear the deep-rooted native vegetation, plant shallow-rooted annual crops, harvest, repeat. For a generation, it worked.

Then the watertable rose. Without the native root systems drawing it down, the groundwater — laden with salt from ancient seabeds — climbed toward the surface. It poisoned paddocks. It killed trees. It sterilised soils that had grown native plants for millennia. Roughly two million hectares of Australian farmland carry this legacy. Not from malice. From simplification.

In Germany, ecologists coined a word for what happens when you replace a complex, species-rich forest with tidy monoculture plantations: Waldsterben. Forest death. Maria Farrell and Robin Berjon reached for that same word in 2024 to describe what has happened to the internet — and if you farm in Australia, their essay, We Need to Rewild the Internet, lands with a weight that goes beyond metaphor. We have seen what command-and-control simplification does to land. We are watching the same logic run its course through digital infrastructure.

Their core argument is this: the internet’s boom years were a one-off harvest. A few companies captured the complexity built by thousands of independent nodes, enclosed it in proprietary infrastructure, and extracted fortunes. What’s left is a plantation — efficient, legible, fragile. Two browsers. Two mobile operating systems. One search engine that matters. All the messy, generative biodiversity of the early web, scoured away.

We didn’t build GrowGood to write another think-piece about it. We built it because the same extraction is happening to agricultural data, and we think the answer is the same: rewild it.

A clarifying note on the term: in agricultural circles, “rewilding” can be a polarising word. To some, it suggests a complete withdrawal of human management or the total removal of livestock — a dangerous naivety that often results in catastrophic fires and lost biodiversity. That is not what we are arguing for here. True rewilding is about resiliency through pluralism and freedom, not abandonment. Hoofed animals, for instance, fulfil an extremely important role in plant management and soil health. Our “rewilding” is about restoring the autonomous, generative processes that keep a system alive — whether that’s a paddock or a protocol.

The AgTech Plantation

Most farm management software is the digital equivalent of clearing for annual pasture. It makes yields legible for the people who never look at the soil. Data gets harvested from farmers — their production records, their soil tests, their market intelligence — and centralised into systems that charge for access to what the farmers generated in the first place.

The logic isn’t cynical. It’s the same short-term arithmetic as the wheatbelt clearing. Lock in the data, own the format, sell the insights. For a generation, it works. Then the farmer wants to switch tools, or the service raises its prices, or the company gets acquired, and the records representing years of careful observation are either gone or ransomed.

This is Farrell and Berjon’s “pathology of command and control” applied to farming. Scour away the complexity — the informal networks between farms, the local knowledge passed across fences, the community seed libraries, the barter arrangements — and replace it with something that can be counted from a corporate office. The first harvest is profitable. The second depends on whether you still trust the system.

An Open Forest Floor

GrowGood is built on a different premise: your data is yours, and it should be readable by anything.

Every piece of data in GrowGood — every harvest log, every soil observation, every commitment to supply 10 boxes of carrots on Friday — is created as a valid JSON-LD linked-data document. Not exported to JSON-LD. Not converted on request. Born that way. Your farm records are machine-readable, semantically rich, and portable from the moment they exist. They don’t live in a proprietary schema that only GrowGood can interpret.

Underneath that, we model economic activity using Valueflows — an open standard for tracking resources, events, agents and processes in economic networks. Instead of building a custom vocabulary that only serves our tool, we use an open-standard vocabulary that can speak to any system built on the same foundation. Your harvest record isn’t just a “harvest record in GrowGood.” It’s a vf:EconomicEvent with a universally resolvable type, legible to any system that understands the standard.

Farrell and Berjon describe the internet’s original architecture as a general-purpose network — built to connect anyone, to do things its designers couldn’t imagine. Valueflows plus JSON-LD is that architecture for the farm economy. The forest floor, not the plantation rows.

The Farm That Doesn’t Go Down When the Server Does

One of the most striking passages in the rewilding essay describes the Howard Street Tunnel fire in Baltimore, 2001. A freight train derailed, chemicals ignited, and the resulting blaze burned for five days. WorldCom had designed its network with redundancy — different fibre-optic cables on different routes. But the region’s topography funnelled all those “different” routes through a single tunnel. The redundancy was theoretical. The chokepoint was real.

Every farm management tool built as a cloud-first SaaS has the same problem. The server goes down, the internet drops out, the paddock still needs logging. A farmer in a gully outside Ararat doesn’t have reliable 4G. Neither does a grower in the Barossa in a lightning storm.

GrowGood’s field-facing app is built with offline-first architecture. Log a harvest, record a soil observation, note a paddock movement — all of it works without a connection. The data syncs when connectivity returns. The farm doesn’t stop because a server did.

This isn’t a UX convenience. It’s an infrastructure philosophy. True resilience isn’t redundant servers in different racks in the same data centre. It’s the absence of a single chokepoint.

Commoning With a Broadfork

Elinor Ostrom won a Nobel Prize studying what happens when communities manage shared resources without central authority. Her finding — often ignored, frequently rediscovered — is that when people can see the problem clearly and trust their neighbours, they self-organise remarkably well. Maine lobster fishermen. California water companies. Farmers managing common irrigation systems across multiple jurisdictions. None of them needed a corporation to extract value from their cooperation and distribute it back at a margin.

Farrell and Berjon point to Internet Exchange Points (IXPs) as the internet’s version of this: not-for-profit associations where internet service providers collectively agree to route each other’s data without individual contracts, reducing costs for everyone. This is what commons governance looks like in practice — not charity, not ideology, just a more efficient structure that keeps the value with the participants.

GrowGood is being built to be the IXP for agricultural data, stewarded by the Growing Data Foundation (GDF) — a not-for-profit entity whose interests are aligned with farmers by design, not by aspiration. The code is open-source. The standards are open. The revenue model funds the commons.

At the operational level, we’ve implemented Ostrom’s logic directly. GrowGood uses Valueflows Intents — the standard’s mechanism for expressing need before committing to produce. A CSA can signal it needs 200 kg of tomatoes in six weeks. A community seed library can signal it’s offering heritage varieties. A regenerative grazing operation can signal it has compost to share. These signals travel the network and connect supply with demand, without anyone in the middle taking a cut.

This is what demand- and supply-driven coordination looks like in code. Not a marketplace owned by a middleman. Not a matching algorithm that monetises the connection. Just: what do you have, what do you need, let’s see if there’s a fit.

Linking Soil Health to Economic Events

The rewilding essay argues that the internet’s original biodiversity was its strength — different protocols, different services, different technical layers maintained by different people, each acting as a check on the others. The monoculture didn’t just make the system fragile; it destroyed the informational soil wealth that the diversity had built up.

In agriculture, the equivalent is treating ecological health as an externality — something that happens off the books while “real” farm data is recorded elsewhere.

GrowGood refuses this separation. Alongside Valueflows, we integrate the W3C SOSA ontology — the standard vocabulary for describing sensor observations. This means a soil moisture reading from an IoT sensor and a harvest event logged by a farmer live in the same linked-data graph, connected. A sosa:Observation of soil organic carbon and a vf:EconomicEvent for a harvest sit in the same structure, cross-referenceable, portable. When you want to know why this season’s yield was different, the ecological context is already part of the record — not siloed in a separate sensor app, not held by the sensor manufacturer’s cloud.

This is not a feature. It’s a fundamental rejection of the accounting logic that keeps nature off the books. In GrowGood, the soil observation is part of the economic record. That’s the informational soil wealth the plantation model cannot replicate — because it was never designed to care about it.

Tactical Diversity

The rewilding essay champions “comcom” — competitive compatibility — the capacity for different tools to work together through open standards, reverse engineering and shared protocols. Each successful tactic creates a new niche; the diversity of niches produces resilience.

GrowGood is built to enable comcom rather than prevent it.

The backend is fully decoupled from the frontends. We have a mobile-first field app for in-paddock logging and a desktop admin interface for building workflow templates. Both run on the same open API. If a community wants to build a voice interface for accessibility, a kiosk terminal for a community tool shed, or an SMS-based logging tool for low-connectivity environments — the architecture invites it. The interface layer isn’t ours to control.

This extends to the types of agriculture GrowGood supports. We provide the backbone for seven different operation types: the home gardener, the small-holder, the market garden operator, the dairy producer, the regenerative grazing operation, the eco-village food forest, and the enterprise-stacking farm. One framework, seven different models of farming. A system that holds Barb’s backyard vegie patch and Steve’s 2,000-hectare regenerative cattle station in the same standards-compliant data model isn’t doing feature expansion. It’s doing the opposite of monoculture.

The Social Layer, Federated

Farrell and Berjon close their essay with concrete examples of rewilding already underway — RSS feeds, email newsletters, the Fediverse. Platforms like Mastodon demonstrate that decentralised social infrastructure works: users own their social graph, no single company can collapse the whole network by changing its algorithm, and anyone can run a node.

GrowGood is building a direct bridge to this layer.

A farmer using GrowGood will be able to publish an offer — say, 20 kg of freshly harvested carrots at $5 per kilogram, available at the farm gate on Saturday — as an ActivityPub post directly to their Mastodon or Bonfire account. Their followers see it in their feed. No middleman takes a commission. No algorithm decides who sees it. The social graph belongs to the farmer.

This is not a vague aspiration. The approach is fully specified: when a farmer creates an “offer” Intent in GrowGood, an ActivityPub adapter translates it into a federated Note and broadcasts it across whichever Fediverse instance they’ve connected. A customer replies by direct message, the farmer creates a vf:Commitment in GrowGood, and the sale becomes part of the auditable economic record. Direct-to-consumer, without an intermediary anywhere in the chain.

What Rewilding Actually Requires

Farrell and Berjon are careful about what rewilding doesn’t mean. It’s not nostalgia. It’s not a return to a mythical original state. It’s “rebuilding resilience by restoring autonomous natural processes and letting them operate at scale to generate complexity.”

In agricultural terms: stop clearing. Give the complexity somewhere to take hold.

That means choosing tools that return data sovereignty. Building on open standards that predate the software and will outlast it. Supporting the institutions — foundations, cooperatives, community networks — that steward shared infrastructure without extracting rent from it.

GrowGood is one part of this. The Growing Data Foundation is another. The Valueflows community, the open AgTech network, the farmers already choosing to document their work in portable formats — these are the pioneer species.

The wheatbelt’s salt problem wasn’t solved by telling farmers they were wrong. It was addressed, slowly, by replanting deep-rooted perennials, by restoring native vegetation in riparian zones, by working with the landscape’s own complexity rather than against it. The results are measured in decades, not quarters.

The monoculture came from a thousand individually sensible decisions made by people who couldn’t see the system they were simplifying. The rewilding works the same way. One farm at a time, one open data record at a time, one federated post at a time.

The forest grows back. But only if you stop clearing it.

We are the first to admit we’re not there yet. GrowGood is a young project, a small cluster of pioneer species working to establish a foothold in a landscape dominated by giants. But the forest doesn’t grow back all at once; it starts with the soil — or in our case, with the syntropy that drives systems to self-organise and evolve into higher complexity.

If you believe farm data belongs in the hands of farmers, and that the future of food depends on open collaboration rather than closed plantations, we invite you to join us. Whether you’re a grower, a developer, or a community organiser, help us nurture the first lines of this digital commons — lines of code that mirror the productive lines of a syntropic system.

Contribute to the journey