The mission: one loop from idea to result

Every community runs the same loop. Someone raises a problem. People argue about it. Somebody has to decide, find the money, do the work, and show what came of it. We are building that loop as one system: AI does the work, people hold the decisions, and every step leaves a public trace.

The loop today

A playground for the town starts as a conversation in a group chat. It moves to a council meeting most people cannot attend, then into a proposal document few will read, then into a budget line, a contractor, and an invoice. Months later something is built, or not, and nobody can say exactly why. Each step lives in a different room, with a different person holding the state. The loop is slow because the hand-overs are slow, and it is opaque because nothing connects the rooms.

The loop we are building

The same four stages, connected. Below each stage: what already runs in Röbel an der Müritz, the town where we test everything first, and what comes next.

Discussion

Citizens raise a topic in the app, argue for and against, and meet in person. An AI moderator checks facts, keeps the thread readable, and turns the debate into a clear picture: the arguments, the open questions, the forecasts. The town hall sees the state of the discussion as it forms, not a petition at the end.

Live in Röbel: the citizen feed, the AI assistant Mecky, surveys, and structured debates.
Next: AI summaries every citizen can check against the original words, and forecasts with their sources attached.

Proposal

The council, local experts, and AI turn the discussion into a proposal: costs, drawings, forecasts, a budget, and the money to fund it, on the public record. Citizens rate it and vote in private, with a tally anyone can verify. The council stays in charge of what it files; the loop gives it a faster, fairer way to get there.

Live in Röbel: private voting with proven tallies, a citizen-governed treasury, and a public decision record.
Next: AI-drafted proposals that citizens verify before they are filed, and funding raised per proposal, Onchain.

Execution

A passed proposal becomes work: tasks, bounties, and jobs on a public board. An AI coordinator tracks progress, pays on delivery, and keeps citizens and the town hall informed without another meeting. Money moves by the rule the vote set, not by whoever holds the account.

Live in Röbel: the community treasury with a time lock, Röbel Münzen for contributions, and a shared workspace for the town's organizations.
Next: bounties created straight from a proposal, and an agent that runs the project inside its budget.

Result

Every proposal keeps a public page: what was decided, what was spent, what got built, and how people rate it. The numbers the town hall sees, citizens see. This is the part of the loop that participation tools have never closed, and it is the part that makes the next discussion better.

Live in Röbel: the public decision record and the live Onchain figures on the case study.
Next: outcome dashboards that close the loop between a vote and a result.

What the research says

We read the field before we drew the loop. Five findings shape it.

The field stops at discussion. AI's proven value is the deliberation stage: reading thousands of contributions and keeping every position. Everyone who studied it names the same gap afterwards, recommendations with no traceable impact. Execution and result are where the loop breaks, and where a treasury, bounties, and a public record are strong.
Legitimacy needs verifiability. Citizens must be able to check that a summary is faithful, that a vote was real, and that money moved by rule. Identity, private provable votes, and a public record are what make AI in civic life legitimate, not features on top of it.
AI allocation gets gamed and concentrates power. So agents act inside scoped budgets and mandates, models stay open and replaceable, humans spot-check, and the platform is open source and community owned. Agents are members, not owners.
Tools die on political incentives. The widest agreement at the Supercooperation workshop was that most cooperation tools fail once they meet real politics. The council has to be inside the loop, not routed around. In our loop the council drafts the proposal.
Ownership is the safeguard. A platform a community cannot inspect, run, or leave is one more vendor. Open source, self-hosted, forkable, documented.

Sources:

What we do not claim

We do not replace the state. Courts, force, legal personhood, and the council's formal decisions stay where they are. The loop is an operating system for the town's public life; the town's institutions run on it, they are not replaced by it. Not everything should be a market: some coordination is trust, not transaction, and the loop should hide its machinery rather than turn neighbors into counterparties. And who governs the protocol itself is a question we have not finished answering; we say so in the open rather than pretend otherwise.

How far automation goes

As far as the record has proven, and no further. A step is automated only after it has run well in public for long enough that people stop checking it, and the record shows why. Drafts, summaries, forecasts, tracking, and payments on delivery come first. Approval, signature, and the vote stay with people. Over time a town gets a loop that runs faster with less effort, and a record that shows every hand-over that made it so.

Take part

Read how the first version runs in a real town in the Röbel case study. Build on the Netizen Stack. If you want to fund the work, there are three doors on the support page. And the fastest way to reach us is to talk to us.

One loop, from an idea in the town to a finished project.