FIELD STATION experiment #001: running our own slice of an AI co-op

Most people using AI tools every day are, often without thinking about it, sending their questions and tasks to a handful of huge companies' data centres. FIELD STATION is an action lab for digital sovereignty and civic tech, and this experiment is about trying a different pattern in practice: could a group of people run AI for one another instead, on hardware they already own, with a public record of who did what for whom?

This is the first in a series of experiments where new ideas get tested in the open and the results get written up, whether they work smoothly or not. The aim is not just to talk about alternatives to platform dependence, but to build small, inspectable examples of different ways of working.

There is already a project for this kind of thing: co/core, an AI inference co-operative where people with spare compute can let their machines do AI work for other people, and where completed jobs are publicly recorded and signed. This FIELD STATION experiment runs on top of co/core, but it also sits in a wider tradition that helps explain why this matters.

An older tradition

Part of that tradition is new mutualism, which looks for practical, democratic ways to organise shared infrastructure so that value circulates among participants instead of being extracted by platforms. Part of it is Ivan Illich's idea of convivial tools: tools people can understand, shape, and use without surrendering control to distant systems. Part of it is the solidarity economy, with its emphasis on co-operation, shared governance, and building institutions that serve communities rather than investors. And part of it is systems thinking, which asks what changes when a small intervention shifts relationships, feedback loops, and power across a wider system.

Seen that way, this is not just a technical demo. It is a small civic-tech experiment in digital sovereignty: a test of whether AI can be treated as shared infrastructure rather than a private service, and whether people can organise that infrastructure around reciprocity, transparency, and trust.

What the experiment involves

A small group of publicly named people take on two roles at once. At some points, their own computer does AI work for someone else in the group; at others, they ask someone else in the group to do work for them. Every completed job produces a signed public record, while the prompts and outputs themselves stay private.

That matters because the point is not anonymity or abstraction. The point is to see whether a group of real people can run useful AI work between themselves, using infrastructure they can inspect and relationships they can govern. If it works, that tells us something about what mutualist digital infrastructure might look like in practice; if it fails, that is useful too, and worth documenting.

Anyone who wants to take part should be comfortable doing this in public. Participants need a verified Bluesky handle, because identity in the system is tied to that, and anyone who wants their machine to do the work needs an Apple silicon Mac they are happy to leave on and connected. The machine can be paired either through the co/core menu bar app or through a one-line terminal command.

How to take part

You'll need a verified Bluesky account — that's how co/core authenticates — and, if you're acting as a provider, an Apple silicon Mac. That last part is ironic for a sovereignty experiment, but it's what co/core currently runs on, and we're pragmatic about using what works today.

1. Sign up at cocore.dev with your Bluesky handle.
2. Pair an Apple silicon Mac if acting as a provider, via the menu bar app or a one-line terminal command.
3. Get in touch with FIELD STATION to receive the current friends list — email experiment@fieldstation.xyz or message us on Bluesky (@dougbelshaw.com / @tomcw.xyz).
4. Add the other participants as friends inside co/core.
5. Once added, requests route only between the people in the experiment, and every completed job produces a signed public record.

What happens next is the real point of the exercise. FIELD STATION will publish what is learned from running this experiment in the open: what worked, what broke, what felt convivial, what created friction, and what this suggests about digital sovereignty, civic tech, and different ways of working. The hope is not to produce a grand theory, but to contribute one careful, public, practical example to a wider tradition of mutualist and solidarity-based experimentation.