AgentLab welcomes The Cognitive Swarm to its project portfolio. Built by Keshav Dalmia, the application brings real-time collaborative brainstorming to large groups—combining voice input, AI-powered idea extraction, and a 3D swarm visualization into one seamless experience.
Brainstorming, Reimagined
Traditional brainstorming tools struggle with scale. Post-it notes don’t work for 50 people, and text-based tools lose the energy of live conversation. The Cognitive Swarm solves both problems: participants contribute ideas through voice while AI extracts and clusters them in real time.
- Live voice anchor — an AI facilitator powered by Google Gemini Live that keeps sessions focused and energized
- 3D swarm visualization — ideas rendered as particles that form visible constellations as clusters emerge
- Automatic clustering — Gemini-powered semantic analysis groups related ideas without manual sorting
- Quadratic voting — fair, credit-based prioritization that prevents any single voice from dominating
Built for Real-Time Collaboration
The application uses Socket.IO for WebSocket connections, keeping brainstorm state synchronized across all participants with sub-second latency. Whether it’s a classroom of 30 or a workshop of 100, everyone sees ideas appear, cluster, and evolve in real time.
Modern stack: React + Vite frontend, Express.js + Socket.IO backend, 88% TypeScript, with Docker support for easy deployment. View on GitHub →
From Divergence to Convergence
The session flow mirrors the best practices of design thinking. It starts with open-ended ideation where anything goes, moves through AI-powered clustering that surfaces themes, and ends with quadratic voting that democratically identifies the strongest ideas. The system then generates structured artifacts—including Mermaid diagrams—that capture the session’s outcomes.
Use Cases in Education
The Cognitive Swarm is designed for any scenario where groups need to think together: classroom brainstorms, design sprints, strategic planning sessions, and research workshops. By combining voice, visualization, and voting, it gives every participant a voice—not just the loudest ones in the room.
Visit the Cognitive Swarm project page for the full technical architecture and use cases.