A simple and universal swarm intelligence engine that generates thousands of AI agents, drops them into a virtual world, and watches them predict the future.
View on GitHub DocumentationUpload any text — a news article, policy paper, market report, or research paper. MiroFish extracts every entity, relationship, and idea using GraphRAG.
Thousands of autonomous agents are created, each with their own personality, beliefs, emotional state, MBTI type, and behavioral patterns derived from the document.
Agents interact on a simulated social platform — posting, reacting, debating. Inject variables ("pass this law") and watch the entire society respond in real time.
Analyze the simulation results, interview focus groups of agents, and generate structured prediction reports with evidence from the knowledge graph.
Simulate voter populations reacting to campaign events, debates, and policy announcements.
Model how consumers react to product launches, pricing changes, or competitor moves.
Predict public sentiment before a press release drops. Test messaging with simulated audiences.
Test how new laws or regulations affect different population segments before implementation.
Simulate international relations, treaty impacts, and conflict scenarios with nation-state agents.
Run social science experiments at scale without real human subjects. Test hypotheses instantly.
20-year-old senior at Beijing University of Posts and Telecommunications. Previously built BettaFish, a multi-agent public opinion analysis tool that hit #1 on GitHub trending with 20,000 stars in one week.
Raised from Chen Tianqiao (Shanda Group) within 24 hours of launch
30 million yuan investment — before the project even had a demo video