Show HN: Decentralised AI Inference Network

A project called rootmode launched a decentralized AI inference network that routes prompts directly to independent GPU operators running vLLM or ComfyUI, with no accounts, no central company server, and encrypted prompts, according to its Show HN post. The project claims 0 accounts required, 0 companies in the middle, support for 3 platforms (Mac, Windows, Linux), and 100% open source, with workers setting their own prices and competing for requests. The network's viability depends on recruiting enough worker machines, which the project acknowledges as its main dependency.

AI Without The Middleman Every model. No account. All your prompts are encrypted. Download rootmode https://rootmode.ai/download Free forever. Nothing to sign up for. By downloading you agree to the Terms https://rootmode.ai/terms and the Privacy policy https://rootmode.ai/privacy . 0 Accounts needed 0 Companies in the middle 3 Platforms 100% Open source The same models. A fraction of the price. Whoever answers your question sets their own price, and they're all competing for the work. That's what keeps pushing it down. Owned by no one. Open to everyone. rootmode isn't a company's service. It's people's computers talking to each other directly — the way files once moved between people, with no shop in the middle. All your prompts are encrypted They travel encrypted to the computer that answers. There's no company server in the middle to collect them. That machine has to read the question to answer it — nobody else does. Nothing to sign up for No email, no password, no card. Open the app and start typing — there's no account for anyone to close. Nobody can switch it off No one owns the network, so no one can raise the price, change the rules, or take it away from you. You're 3 steps from using it. None of them involve a form. Download the app It installs like any other program, on Mac, Windows or Linux. Nothing to fill in. Open it It finds computers that are free to help and shows you what each one can do. Ready the moment it opens. Ask for something Type a question and get an answer. Ask for a picture and it lands in your downloads. Got GPUs? Anyone can serve. No application, no approval queue. Run the worker on your machine, set your own price, and answer as much or as little as you like. Setup takes minutes Same compose file on every box. Point .env at the vLLM or ComfyUI you already run. The container needs no GPU access of its own. You set the price Charge what you like, or nothing at all. Serve a request, get paid — no invoicing, no approval. Your box stays yours Nothing a client sends is ever executed. They fill in the parameters you declared, and nothing else reaches your machine. Fair questions. Do I need to be good with computers? No. If you can open an app and type, you can use it. There's nothing to set up and nothing to configure. Isn't this the same as ChatGPT or OpenRouter? The apps look similar; underneath they're opposites. Those are one company taking your request through their own computers. rootmode has no company in the middle — your app talks straight to whoever is answering. Why is it cheaper? Nobody's paying for enormous data centres. People use computers they already own, and they compete with each other on price. The app itself is free. Who can see what I type? Only the computer answering it, and it isn't kept afterwards. If you'd rather keep everything at home, you can run it entirely on your own machine. What's the catch? It depends on people running workers. The more machines on the network, the faster and cheaper it gets for everyone. What if you disappear? It keeps working. There's no central service to switch off — it's people's computers helping each other, and that carries on without us. Start using AI that nobody owns. The button picks the installer for this computer. No account, nothing to fill in. By downloading you agree to the Terms https://rootmode.ai/terms and the Privacy policy https://rootmode.ai/privacy .