Tech With Tim

I Spawned 10,000 AI Agents to Predict the Future (MiroFish Is Insane)

The video introduces MiroFish, an open‑source platform that orchestrates tens of thousands of autonomous AI agents to build a knowledge graph and simulate social‑media‑style interactions for forecasting future events. By feeding historical data—such as Dubai real‑estate trends and geopolitical news—the system generates detailed predictions, exemplified by a ten‑year price outlook for two‑bedroom apartments in downtown Dubai. Key technical insights include a five‑stage pipeline: graph construction, agent persona creation, dual‑world simulation (Twitter vs. Reddit dynamics), report generation, and interactive exploration. The project has attracted 51,000 GitHub stars, $4 million in funding, and over 7,600 forks, highlighting rapid community adoption. Running the swarm requires OpenAI‑compatible LLMs and a Zepp Cloud API key; a typical GPT‑4 run costs about $3, while higher‑tier models can quickly become expensive. The presenter demonstrates real‑world use cases—ranging from market‑prediction bots to narrative forecasting—and showcases a deployment shortcut via Hostinger’s one‑click Docker service for as little as $6.50 per month. He also references auxiliary tools like Claude for data gathering and Whisper Flow for voice dictation, underscoring the ecosystem needed to operationalize MiroFish. For businesses, MiroFish democratizes large‑scale predictive modeling, allowing analysts to simulate diverse stakeholder opinions without building custom infrastructure. While cost and setup complexity remain considerations, the platform’s scalability and open‑source nature could accelerate data‑driven decision‑making across finance, real‑estate, and media sectors.