{
  "txid": "e8954ad19bab6f51d73df61012f497134f6e568b4f7a00b62f0d82246ed2b753",
  "block_height": 0,
  "time": null,
  "app": "treechat",
  "type": "post",
  "map_content": "MiroFish: The God View Engine\r\nA developer in China built an AI engine that generates thousands of digital humans - each with their own personality, memory, and behavior - drops them into a virtual world, and watches them predict the future.\r\nIt's called MiroFish. It hit #1 on GitHub's global trending. The creator is 20 years old.\r\nThe creator\r\nHis name is Guo Hangjiang. Online he goes by BaiFu. Senior at Beijing University of Posts and Telecommunications. Writes Python. Obsessed with agent architectures and graph computing.\r\nLate 2025, his first project - BettaFish, a multi-agent public opinion analyzer - hit #1 on GitHub trending and pulled 20,000 stars in a week.\r\nThat's when a Chinese billionaire noticed him.\r\nThe billionaire\r\nChen Tianqiao. Founder of Shanda Group. Once the richest person in China. Built a gaming empire in the early 2000s, stepped away, moved to the US, and quietly transformed Shanda into a tech investment platform.\r\nHe'd been pushing a thesis he calls \"the super-individual\" - the idea that in the AI era, one person can build what used to require an entire company.\r\nGuo was the proof of concept.\r\nThe 10-day build\r\nChen invited Guo as an intern. Gave him total freedom. What followed.\r\nGuo built MiroFish in 10 days using what he calls \"vibe coding\" - fast, intuitive, no over-engineering.\r\nThe night he finished, he recorded a rough demo and sent it directly to Chen's desk. Within 24 hours, Chen committed 30 million yuan - roughly $4.1 million - to incubate the project.\r\nGuo went from intern to CEO overnight.\r\nWhat it actually does\r\nHere's what MiroFish does, stripped to the core:\r\nYou feed it a document - a news article, a policy draft, a financial report, even a novel. The system reads it, extracts every entity and relationship into a knowledge graph using GraphRAG, then generates thousands of autonomous AI agents. Each one gets a unique biography, personality type, social connections, and behavioral logic.\r\nThe killer feature: \"God's Eye View.\"\r\nAt any point you can inject new variables into the simulation:\r\n\"Fed suddenly cuts rates by 50bps\"\r\n\"CEO resigns\"\r\n\"Competitor launches a product\"\r\nThen watch the entire digital world reorganize in real time. Controlled experiments you can't run in reality.\r\nThe simulation engine\r\nThe simulation runs on OASIS - an open-source framework by CAMEL-AI. Agents don't just talk. They form groups, develop opinion leaders, create herd effects, shift positions over time. Long-term memory is maintained through Zep Cloud.\r\nThe key feature: \"God's Eye View.\" At any moment, you inject a new variable - a rate hike, a CEO resignation, a product launch - and the entire world recalibrates in real time.\r\nControlled [[experiments]] that are impossible in reality.\r\nUnder the hood:\r\nSimulation engine: OASIS (by CAMEL-AI)\r\nMemory: Zep Cloud for long-term agent memory\r\nKnowledge graphs: GraphRAG\r\nLicense: AGPL-3.0 (fully open source)\r\nDeploy: Docker Compose one-click setup\r\nThis isn't a toy. It's a serious multi-agent simulation framework.\r\nReal demos\r\nTwo demonstrations already shown:\r\nFirst - they loaded the first 80 chapters of \"Dream of the Red Chamber,\" a classical Chinese novel with a famously lost ending. MiroFish generated character agents with authentic personalities and relationships, ran the simulation, and produced alternative narrative branches predicting the missing conclusion.\r\nSecond - a Fed rate hike scenario. The system simulated how retail investors, institutional players, and analysts each react, tracked where group sentiment converges, and mapped the full opinion trajectory.\r\nThe honest part\r\nWorth noting what this is not:\r\nMiroFish has published no benchmarks comparing its predictions against real-world outcomes. The demos are illustrations of the approach, not evidence of accuracy. Running thousands of agents means massive LLM API costs. Agent personalities inherit whatever biases exist in the training data. And simulated humans, no matter how sophisticated, are not real humans.\r\nThe honest framing: it surfaces scenarios and dynamics you might otherwise miss. It doesn't deliver certainties.\r\nThe bigger picture\r\nOn March 7, 2026, MiroFish reached #1 on GitHub's global trending - above projects by OpenAI, Google, and Microsoft. 18,000 stars and 1,900 forks in days. Currently at 22,000+.\r\nOne undergraduate. Ten days of coding. A simulation engine that builds parallel digital societies from a single document.\r\nChen Tianqiao's bet wasn't on the software. It was on the thesis: that the era of the super-individual has already started, and most people haven't noticed yet.\r\nOpen source under AGPL-3.0. Ships with Docker one-click deployment.\r\nhttps://github.com/666ghj/MiroFish",
  "media_type": "text/markdown",
  "filename": "|",
  "author": "14aqJ2hMtENYJVCJaekcrqi12fiZJzoWGK",
  "display_name": "1Bitcoin_user",
  "channel": null,
  "parent_txid": null,
  "ref_txid": null,
  "tags": null,
  "reply_count": 0,
  "like_count": 0,
  "timestamp": "2026-03-14T18:37:14.000Z",
  "media_url": null,
  "aip_verified": true,
  "has_access": true,
  "attachments": [],
  "ui_name": "1Bitcoin_user",
  "ui_display_name": "1Bitcoin_user",
  "ui_handle": "1Bitcoin_user",
  "ui_display_raw": "1Bitcoin_user",
  "ui_signer": "14aqJ2hMtENYJVCJaekcrqi12fiZJzoWGK",
  "ref_ui_name": "unknown",
  "ref_ui_signer": "unknown"
}