{"id":257387,"date":"2025-11-02T10:43:10","date_gmt":"2025-11-02T10:43:10","guid":{"rendered":"https:\/\/www.newsbeep.com\/au\/257387\/"},"modified":"2025-11-02T10:43:10","modified_gmt":"2025-11-02T10:43:10","slug":"pewdiepie-dives-into-an-ai-side-quest-revealing-his-self-made-chatos-fueled-by-chinese-qwen-models-modded-rtx-4090s","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/au\/257387\/","title":{"rendered":"PewDiePie Dives Into an AI Side-Quest, Revealing His Self-Made \u2018ChatOS\u2019; Fueled by Chinese Qwen Models & Modded RTX 4090s"},"content":{"rendered":"

Well, the famous YouTuber PewDiePie has been found ‘tinkering’ with AI in a rather unique fashion, creating his own AI service that uses Chinese open-source models with NVIDIA’s GPUs.<\/p>\n

PewDiePie Actually Managed to Let AI Models Talk To Each Other, And Interestingly, They Colluded Against Him<\/p>\n

It seems like Felix has a knack for running AI models locally on machines, and we are surprised to see his expertise around this particular venture, to say the least. In a new video<\/a>, PewDiePie dives into how he leverages PCIe burification to essentially create a 10-GPU ‘mini-datacenter’, which he later utilized for a personalized AI service. Interestingly, his GPU rack features eight NVIDIA RTX Ada GPUs and two blower-style RTX 4090s, which appear to be similar to those that are ‘modded’ and widely used in China. He calls his service ‘ChatOS’, and it was found to be utilizing Chinese open-source models.<\/p>\n

PewDiePie just vibe-coded his own Chat UI, built an army of chatbots for majority voting and gave them all RAG, DeepResearch and audio output<\/p>\n

naturally, he only uses chinese Qwen models and runs them on his local PC with 8x modded chinese 48GB 4090s and 2x RTX 4000 Ada<\/p>\n

his army\u2026 pic.twitter.com\/vS6DlPFwdQ<\/a><\/p>\n

\u2014 Lisan al Gaib (@scaling01) October 31, 2025<\/a><\/p>\n

Based on what Felix says, his idea of building a capable AI machine was to assist individuals in medical research by running protein-folding simulations on it. However, he decided to experiment with AI model hosting by hosting models like the Llama 70B, and the LLM ran successfully. Interestingly, he took a step further by creating a web service to communicate with his local AI models, integrating functionalities such as web search, RAG, audio output, and memory. Felix was spotted utilizing Baidu’s Qwen open-source model to create a fully private and self-hosted system.<\/p>\n

\"AImage Credits: PewDiePie<\/p>\n

Now, what’s next is even more hilarious. PewDiePie managed to run multiple LLMs locally to see how they interacted with each other, naming them\u00a0“The Council”\u00a0and later expanding it into\u00a0“The Swarm.” The Council basically voted on a response to a prompt, to analyze whether it was the correct one, and the voting mechanism was operated by multiple AI instances. Over time, the Council members began to strategically favor each other strategically, exhibiting a kind of emergent cooperation, even though it was not part of the original program.<\/p>\n

To solve it, he switched the entire system to a more “dumber’ model, and this venture was pretty hilarious to witness in the first place. These models apparently evolved while interacting with each other, exhibiting a human-like behavior of colluding with the voting system. Well, PewDiePie’s AI side-quest was defintely a highlight of my day.<\/p>\n<\/p>\n

Follow Wccftech on Google<\/a> or add us as a preferred source<\/a>, to get our news coverage and reviews in your feeds.<\/p>\n

\t\t