I Met the Plants Through an App
I should tell you where this project actually comes from, because it is not where you might expect. I did not grow up knowing the names of the plants around me. I learned them the way a lot of people my age have—by holding a phone up to a leaf. The summer I started paying attention to my own yard, the apps that cracked it open were Seek and iNaturalist. Point the camera, get a name, kneel down and check it against the plant in front of you, and slowly the green blur of the world resolves into individuals you can call by name. That was the doorway. Technology is how I met the living world, not what kept me from it.
So when I say I am building a plant tool, understand that I am building the thing that built me. projectG.A.I.A. is the question I kept turning over after those first summers: what if a tool like that could run on hardware you own, without quietly shipping everything you ask it back to a server farm—and what if it actually knew something deep about the world growing outside your door?
What G.A.I.A. Is
G.A.I.A. stands for Garden & Agricultural Intelligence Assistant. The name is borrowed, of course—in Greek myth Gaia is the primordial Earth, the ground everything else stands on—and I am aware that naming a piece of software after the Earth goddess is a large promise to make. I am making it anyway, because the whole point is to keep the thing pointed at the actual ground rather than at itself.
At its core G.A.I.A. is an AI system meant to run on modest, locally owned hardware: a small home server, a repurposed laptop, a community hub on a solar panel. No cloud subscription. No data harvested off the back of your questions. No constant internet connection required. It runs where you are, on what you own.
What You Could Actually Ask It
Picture the kind of question I was fumbling toward in my first season—“What can I plant in Zone 7a in March that will fix nitrogen for my summer tomatoes?”—and an answer grounded in real agronomic data rather than a confident guess. The knowledge base is built from publicly available sources: USDA datasets, university agricultural extensions, open-access botanical databases. The goal is a tool that helps you sort out companion planting for your specific zone, lay out a succession schedule across the season, recognize a disease on a leaf and reach for an organic remedy first, sketch a food forest to fit your space and soil, track down regionally appropriate seed, and read the medicinal and nutritional profile of something you found while foraging.
What it is not is a general chatbot reciting whatever it scraped. It is narrow on purpose—trained on curated, checkable material, built to be useful to one specific kind of person: someone with dirt under their nails trying to grow food.
Why Local, Why Offline
The thing that most separates G.A.I.A. from the AI everyone already knows is that it runs locally and off-grid. The dominant model lives in enormous data centers drinking electricity and water; a single query to a large cloud model can cost roughly ten times the energy of an ordinary web search, and the whole arrangement concentrates power—the literal kind and the other kind—in the few companies able to build at planetary scale.
G.A.I.A. goes the other direction. Using an efficient open-source model in the Hermes3-8B family—eight billion parameters, small enough to run on consumer-grade GPUs—the system fits on hardware a household or a community can own and keep running themselves. It does not need to phone home. Your questions about your garden, your health, your land stay on your machine. When your tool needs a corporate subscription and a fiber line to so much as answer you, it was never really yours; G.A.I.A. is meant to keep working on the day the grid doesn’t.
Where It Sits in the Solarpunk Frame
Solarpunk has always pictured technology and nature as collaborators rather than enemies—tools that serve the living world instead of mining it, communities holding real agency over what they depend on. G.A.I.A. is one attempt to make a piece of that concrete. It is decentralized, so there is no single gatekeeper to cut you off. Its privacy is structural rather than a setting you have to trust someone to honor. It is sized to run on a solar setup, open-source so the code belongs to the people using it, and grounded in real botanical and agricultural data rather than vibes. None of that is novel as a wish list. The wager is in building it small enough that ordinary people can actually run it—a piece of technology learning from how living systems already work: distributed, redundant, no central point that brings the rest down when it fails.
Where It’s Headed
I think of G.A.I.A. less as one app than as a first seed. The plant-ID and garden-planning work is the part I can see clearly right now; beyond it I want tools that help neighborhoods map and manage hyperlocal food production, coordinate the unglamorous logistics of sharing surplus and running seed swaps, make botanical and mycological knowledge easy to reach, and help a block think ahead about the kind of supply disruption a lot of us watched firsthand in 2020. Each piece should stand on its own or work alongside the others—local, private, in service of the people running it.
An Honest Note on Where This Is
I want to be plain that this is a project in progress, not a product with a ship date. We are living through a stretch where AI is being pointed, by default, toward more concentration, higher paywalls, and heavier resource use—and I do not think that default is the only road available. A tool can be built to help someone grow food instead of to sell them ads. It can be built to get stronger as communities share what they know rather than as a company hoards what it takes. G.A.I.A. is my attempt to build down that other road, and I am early on it. I will keep posting here as the pieces come online—the failures included, because there will be plenty.
Frequently Asked Questions
What is G.A.I.A.?
G.A.I.A. (Garden & Agricultural Intelligence Assistant) is Futurespore's off-grid, local-first AI system designed to support solarpunk practitioners — plant identification, foraging knowledge, garden planning, and ecological reference — running on local hardware without sending data to corporate clouds. It is solarpunk applied to AI itself.
Why build a local AI instead of using ChatGPT?
Because cloud AI is centralizing, energy-intensive, surveils its users, and disappears when the company changes its terms. A local AI runs on hardware you own, with data that stays yours, on energy you can source renewably. The solarpunk argument for local AI is the same as the solarpunk argument for backyard food: sovereignty over what you depend on.
How is G.A.I.A. different from other AI assistants?
G.A.I.A. runs locally (no cloud), is trained on ecological and herbal corpora rather than generalized internet text, and is purpose-built for solarpunk practitioners — gardeners, foragers, herbalists. Privacy is structural, not optional. The model is small enough to run on a laptop or Raspberry Pi cluster and is designed to work offline.
Is G.A.I.A. available yet?
G.A.I.A. is currently in development. Early prototypes focus on plant ID and foraging reference. The project follows a research-first roadmap rather than a startup release schedule. Updates publish to the Mycelial Grimoire as capabilities come online.
What does solarpunk AI even mean?
Solarpunk AI is locally hosted, energy-modest, open-source, and built to serve communities rather than extract from them. It runs on renewable power. It learns from regional knowledge instead of homogenizing it. It treats users as collaborators, not training data. G.A.I.A. is one attempt at making that real.
Written by E. Silkweaver, founder of Futurespore.