The Community AI Project
Public access media centers built the civic infrastructure of the broadcast era. In the 21st century, they can do the same for AI — the Community AI Project introduces a new generation of open-source tools designed + developed by Stephen Walter and produced + maintained by BIG, with the help of AI*, aiming to put civic power back in people’s hands.
A demonstration of some of our apps, including Neighborhood AI — the local AI maker that can start it all, outside the confines of Big Tech.
*Built using an agile — sometimes adversarial — interplay of Claude Code (primary AI for coding), Gemini (primary AI for in-app API function calls), OpenAI (original source of some visual style mockups, as well as some API calls, but later no longer used), open-source local AI models running via LM Studio and Ollama on a Macbook Pro M1 Max — including Deepseek, Qwen, and Gemma — and finally some old-fashioned human designing, coding, and software management.
We find that AI is powerful but indiscriminate, like a wild rushing river; a new AI-infused tool is infinitely more useful, targeted, ethical when a clever and creative human hand guides it along — each intricate decision point like the intricate roots of the trees and floor of the forest, and the dams and homes of the beavers, which contain and guide a river along in a direction looking something like peace and coexistence.
AI as Community Infrastructure
Each app solves a problem community media centers, city halls, and public access TV outlets face every week: from captioning live streams, to transcribing municipal meetings, to translating civic documents. But they are also each an attempt to offer a new kind of infrastructure to the public. Something new to grab onto, to take hold of, or to find one’s footing on in order to better engage with civic life: to make better sense of decisions, to unearth patterns and hidden histories, to hold not only government officials, but each other, accountable, for making our neighborhoods more humane and thriving places to live.
Neighborhood AI
Fine-tune and launch constitutional local AI models for your community. A governance-first platform for stations that want AI on their own terms.
Community Highlighter
Paste a YouTube URL of any public meeting and get a searchable, speaker-diarized transcript synced to the video. Built for the meetings your station already covers.
Community Captioner
Free, browser-based open captions for OBS live streams — no expensive hardware, no subscription. Real-time accessibility for any station’s broadcast.
Civic Documenter
Turn messy PDF attachments from municipal agendas into structured, readable data. AI parsing that makes local government documents legible.
Civic Translator
Instantly translate complex civic documents into the languages your community actually speaks — while preserving formatting and legal fidelity.
Commit
A localized AI agent for cities and towns — an experimental agentic chatbot that helps residents take real-world actions in their community, not just get answers.
Artificial: The Game of AI
A three-part game for learning, practicing, and critiquing AI — a media-literacy tool stations can run in classrooms, workshops, and community programs.
Anti-AI
A tool for critiquing and counteracting the negative impacts of AI in civic contexts. Because serving your community also means naming what could harm it.
Community media is the right home for civic AI.
PEG stations and community media centers already do the unglamorous work the AI industry now claims it wants to support: covering municipal meetings, training residents in media literacy, filling the gaps left by shrinking local newsrooms, and serving communities in multiple languages. We already have the trust, the civic relationships, and the broadcast infrastructure.
What we haven’t had — until now — is software that matches the moment. Corporate AI platforms are built for advertising and enterprise productivity, not for captioning a Tuesday-night zoning board meeting or translating a school committee agenda into Haitian Creole. The Community AI Project is what happens when you build AI with those use cases as the starting point, not an afterthought.
We believe community media centers (like BIG) represent one of the best places for this new type of civic engagement infrastructure. There is nothing quite else like them. With over 1,500 of these places in the US alone — whose funding comes in part from laws enacted by Congress to divert money away from cable companies toward the public benefit, and who have the technical infrastructure few libraries or Town Halls could maintain — there is no better place to start a movement for local AI as a new form of public works than at your neighborhood public access station.
owner: “The Public”,
model: “Open Source”,
profit: false,
origin: “Brookline, MA”,
deployableAt: “any PEG station”,
stack: [
“FastAPI”,
“React”,
“Qdrant”,
“Ollama”
],
data: function() {
// Stays in the community
return local.storage;
},
license: “CC BY-SA 4.0”
};
Principles that guide every line.
These aren’t marketing copy — they’re commitments embedded in every tool we build, every architecture decision we make, and every deployment we ship.
Climate Justice Seeking
We try to prioritize locally-run models over frontier AI to reduce carbon footprint. Right-sized models, efficient design, honest about our transition timeline. While we are still a ways away from making local AI publicly-accessible, that is the current goal we are striving toward. Help us by donating
to BIG to pay for the local servers needed to do this!
Radical Openness
No black boxes. Every line of code is open source and available for inspection. Trust is earned through transparency, not marketing.
Human Centric
AI should amplify human connection and civic action, not replace it. We reduce administrative friction so people can focus on each other.
Community Owned
Neighborhoods should own their digital infrastructure as much as their physical spaces. No data sales, no ads, no extraction. Ever.
Accessible by Default
We design for disabilities and language barriers first — screen readers, keyboard nav, eight-language translation, plain language alternatives.
Sustainable Tech
Lightweight tools that run on old hardware. Minimal dependencies, backwards compatible, documented for community maintenance.
Code meets concrete.
We apply the speed of modern software development to slow-moving civic problems — iterating faster than bureaucracy allows, while maintaining the safety and inclusivity required for public life.
Why AI?
Corporate AI extracts value; community AI creates it. These tools analyze public data at the scale of a government to empower the individual — translating dense municipal documents, surfacing patterns in public meetings, and connecting residents to the decisions that shape their lives.
Our Means
We build with open-source models that can be audited by anyone. Our stack is designed for local-first deployment: low cost, high privacy, community-maintainable. No vendor lock-in, no corporate dependencies, no surprise subscription hikes.
Our Ends
Our goal isn’t “smarter” cities — it’s stronger stations and stronger citizens. We reduce the tedious admin that keeps volunteers from acting, translate jargon that keeps residents from understanding, and give community media the tools to serve at scale.
Public media makers × creative technologists.
A collaboration between community media, civic design, and local data infrastructure — the three things this work needs to travel.
Brookline Interactive Group
Nonprofit community media center and civic hub in Brookline, MA. BIG incubates hyperlocal storytelling and serves 500+ residents annually through classes, coverage, and partnerships.
Weird Machine
The R&D lab of Stephen Walter — creative technologist, civic designer, and BIG’s Director of Innovation. Fifteen years building at the intersection of design, democracy, and emerging tech.
Neighborhood AI
Local data models and policy expertise focused on hyperlocal, community-governed AI for dignified civic infrastructure. Based in Boston and New York.
Community media deserves tooling built for community media.
Whether you’re at a PEG station, a community media center, an ACM affiliate, or just a neighbor with a municipal meeting to cover — these tools are yours to deploy, fork, and adapt. Get in touch to bring The Community AI Project to your station.