About MiroFish, and about this journal

The product, in one paragraph

MiroFish is an open-source system for scenario prediction through multi-agent simulation. You bring a question about a situation involving people — a launch, a price change, a policy, a controversy — and optionally attach documents describing it. MiroFish extracts the actors, relationships, and pressures from that material into a knowledge graph, turns every meaningful node into an autonomous AI agent with its own persona and motives, and lets the whole cast react to your scenario on a simulated social surface, round by round. The run is then distilled into a prediction report — likely outcome, risk signals, narrative paths — and the simulated world stays open for follow-up questions in chat.

Reports are exploratory decision support. They surface reactions, alliances, and story lines you hadn't considered; they do not guarantee outcomes, and neither the product nor this journal will ever tell you otherwise.

Where things live

mirofish.us is the main MiroFish website and product — every article here ultimately points you there, because that is where simulations actually run. The engine is open source: the code lives in the GitHub repository, and the README documents the architecture and how to self-host. Two recorded demos — a full run and the follow-up chat — show the system in motion.

What this journal is for

The MiroFish Journal is the explainer blog of the family: long-form, first-principles pieces on how the machine actually works and how to use it well — seeding the knowledge graph, what happens inside the simulation rounds, reading a report without over-trusting it, and knowing when you need a simulation at all versus a plain chat answer. Start with how MiroFish works if you're new, or the getting-started walkthrough if you'd rather learn by running. When you're done reading, open MiroFish and put it to work.