The Community AI Project
The Community AI Project is a public-facing hub for building “digital public works” — open, accessible tools that help communities engage with civic information, govern their own data, and reduce dependency on big-tech black boxes.
I was the solo designer, developer, and project manager while in my position at Brookline Interactive Group. The six initial apps that are part of this have been implemented with community to fill real world needs, and continue to grow and get put to use.
Thesis: Corporate AI extracts value; Community AI creates value.
Live site: community.weirdmachine.org
Code / profile: github.com/amateurmenace
License & terms: CC BY-SA 4.0 + acceptable use + privacy terms
The GOALs
Community AI = AI as community infrastructure.
This project frames AI tools like libraries and roads: built for the public good, designed for accessibility, and owned by the people they serve. What we need are not just tool community AI tools, but a place that brings them together under a shared set of values. So the Community AI Project is:
A constellation of six initial apps coded with community in minds
A mission-driven hub for community-first civic tools
A values-forward “moral compass” baked into product decisions
A playful “AI Playground” that invites experimentation (visual editor + export)
The Challenges
Public-interest organizations and communities are increasingly asked to rely on opaque, corporate AI systems that:
create lock-in through black-box models,
weaken privacy by centralizing sensitive workflows, and
optimize for profit rather than civic outcomes.
Meanwhile, civic work has real constraints: limited budgets, legacy hardware, accessibility needs, and high stakes for accuracy and trust.
DEVELOPMENT STRATEGY
Phase 1 - Complete: Foundation with Frontier Models
We began by leveraging existing AI APIs (GPT, Claude, Gemini) to rapidly prototype and validate our concepts. This allowed us to focus on understanding community needs without building infrastructure from scratch.
Phase 2 - Current: Hybrid Approach
We're currently operating in a hybrid mode: using frontier models where necessary while actively migrating functionality to open-source alternatives. Our tools are being refactored to support multiple backends.
Phase 3 - In Progress: Local LLM Integration
We're testing local deployment of models like Llama, Mistral, and Phi on community hardware. The goal is to run AI that doesn't phone home to corporate servers.
Phase 4 - Future: Community-Owned AI
Our long-term vision: AI models trained on local data, running on local hardware, governed by local communities. True digital sovereignty.
Design Principles
These values are explicit in the interface and guide feature choices:
Climate justice: prioritize locally-run LLMs over frontier models to reduce environmental impact.
Radical openness: no black boxes; code can be audited.
Human-centric: amplify connection and civic action, not replacement.
Community-owned data: neighborhoods should own their digital data.
Accessible and equitable by default: design for disabilities and language barriers first.
Resilient + flexible tech: lightweight tools that can run on older or cheaper hardware.
Difference as strength: Our tools serve diverse communities with diverse needs, experiences, and strengths. We believe in plurality, care, collision, and connection as fundamental concepts to fighting totalitarianism and oppression, and must be designed into our civic systems.
Play: The best way to understand and make sense of something is to play with it. We embrace play for both ourselves and our users so we might better make sense of the role AI should play in democracy.
The Six Initial Apps
Finally, here is a civic AI vibe coder where you can make your own apps for community!
