# AI use cases for Pooled (Nehanth/pooled) Repo: [Nehanth/pooled](https://github.com/Nehanth/pooled) | Use cases: 4 | Page: https://dedupedai.com/use-cases/nehanth-pooled/ Pooled runs one open language model across the browsers and devices in a room, so people can chat with it or use it from coding agents. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Chat with a big open model Pooled across your devices: https://dedupedai.com/use-cases/nehanth-pooled/chat-with-model-across-friends-devices/ - Point a coding agent at a Pooled room's model: https://dedupedai.com/use-cases/nehanth-pooled/serve-room-model-to-coding-agent/ - Have a room's model build and fix a small app: https://dedupedai.com/use-cases/nehanth-pooled/build-small-app-in-code-mode/ - Run the experimental 125B Flash model from disk: https://dedupedai.com/use-cases/nehanth-pooled/run-flash-125b-from-disk-experimental/ ## How to chat with a big open model Pooled across your devices with Pooled Repo: [Nehanth/pooled](https://github.com/Nehanth/pooled) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Open one room link on your laptop and phone to run a 27B or 35B open model together, then chat with it in the tab. **Who it is for:** A person with a laptop, a desktop or a phone who wants to chat with a large open model that no single device can hold, and has friends or family willing to open the same link. **The problem:** A large open model needs more GPU memory than one personal device has, and setting up a shared inference stack usually means installing software and running a server. **The outcome:** Each device holds some of the model's layers and the room answers chat messages, with no install and no account. ### Steps 1. Open https://pooled.run and choose Start a room, or open a room link someone shared with you. 2. On each extra device (laptop, desktop or phone), open the same room link in a browser tab. 3. Pick a model in the room's model picker that fits the total GPU memory of the devices in the room. 4. Wait while each device downloads only its own layers, then type a message in the chat. ### Prompt to paste into your agent ```text Using the Nehanth/pooled repository, open the room link in two browser tabs on different devices, pick the Qwen3.6 35B MoE model, and send the message 'Hi'. Expected output: a reply from the model in the chat. Verify that both tabs list the room's devices and that the reply appears in both. ``` ### Caveat Model weights are downloaded per device, so the first run takes a long time, and the room's speed depends on how many devices hold layers and on the connection between them. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/nehanth-pooled/chat-with-model-across-friends-devices/ ## How to point a coding agent at a Pooled room's model with Pooled Repo: [Nehanth/pooled](https://github.com/Nehanth/pooled) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Run the Pooled bridge on your computer so Codex or Claude Code sends requests to the room's GPUs. **Who it is for:** A developer who already uses a coding agent such as Codex CLI or Claude Code and wants it to use an open model hosted by a Pooled room instead of a paid API. **The problem:** Coding agents expect an OpenAI or Anthropic compatible endpoint, and the room's model is only reachable through the browser tabs in the room. **The outcome:** A local endpoint on 127.0.0.1 forwards the agent's requests to the room, so the agent's prompts and answers run on the room's GPUs. ### Steps 1. Open the room in the browser and click the black Serve API button in the header to get the command for that room. 2. Run `npx @pooled/cli serve ""` on your own computer with Node 22 or newer. (inferred, check the README) 3. Set OPENAI_BASE_URL or ANTHROPIC_BASE_URL to the address the command prints, and set any API key value. 4. Point the coding agent at that base URL and send a short test request with curl or the agent itself. ### Prompt to paste into your agent ```text In the Nehanth/pooled repository, run `npx @pooled/cli serve ""` and then send a chat request with curl to http://127.0.0.1:8080/v1/chat/completions using model 'pooled'. Expected output: a JSON chat completion from the room's model. Verify that the room's chat shows the request. ``` ### Caveat Prompts go to the room's host and, unless the host limits who sees answers, to everyone in the room, so do not send private code or data through a room you do not control. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/nehanth-pooled/serve-room-model-to-coding-agent/ ## How to have a room's model build and fix a small app with Pooled Repo: [Nehanth/pooled](https://github.com/Nehanth/pooled) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Switch a room to Code mode, describe an app, and let the model write, preview and fix it in the tab. **Who it is for:** A hobbyist or a developer who wants a small browser game or app built quickly, without setting up a project locally first. **The problem:** Starting a small app usually means setting up files, a local server and a debugging loop before any code runs. **The outcome:** The room's model writes the files, serves them in a sandboxed preview, reads its own console errors and fixes them, and everyone in the room can run the same preview. ### Steps 1. Open a room in the browser and switch the room to Code mode. 2. Ask for the app in plain words, for example 'build a tetris game'. 3. Watch the preview on the virtual localhost port 5173 and the files the model writes. 4. Ask for changes or fixes in the same chat, and approve any edits to a real folder the host picked. ### Prompt to paste into your agent ```text Using the Nehanth/pooled repository, open a room in a browser tab, switch to Code mode, and ask for 'build a tetris game'. Expected output: a set of files and a running preview on port 5173. Verify that the preview runs without console errors after the model's fix step. ``` ### Caveat Files live in the host's browser unless the host picks a folder on disk, and what the agent can and cannot touch is described in SECURITY.md, so read that before letting it edit a real folder. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/nehanth-pooled/build-small-app-in-code-mode/ ## How to run the experimental 125B Flash model from disk with Pooled Repo: [Nehanth/pooled](https://github.com/Nehanth/pooled) | Difficulty: Advanced | Time: 3 hours | Basis: From the README Load Qwen3.8 Flash, which keeps most of its experts on disk, in one Chrome tab and speed it up by joining a second PC. **Who it is for:** An experienced user with a large SSD and a capable GPU who wants to test a 125B mixture-of-experts model that is much larger than the GPU memory of most computers. **The problem:** A 125B model does not fit in typical GPU memory, so running it normally needs far more hardware than a single personal computer. **The outcome:** The model runs with its experts stored on disk and cached on the GPU as needed, and a second computer joined with the command line lends memory to make it faster. ### Steps 1. Open a room in Chrome on the computer that has the SSD space, and pick Qwen3.8 Flash in the room's model picker. 2. Accept the prompt to store the first download, which is about 99 GB, on the device. 3. On a second PC, run `npx @pooled/cli join` followed by the room link or code, as described in cli/README.md. (inferred, check the README) 4. Send a short question and compare the decode speed to the figures in the README. ### Prompt to paste into your agent ```text In the Nehanth/pooled repository, open a room in Chrome, select Qwen3.8 Flash, accept the roughly 99 GB storage prompt, and send 'Hi'. Expected output: a reply after a few seconds on a single machine. Verify that the room shows the model as loaded and the reply uses the 64K context setting. ``` ### Caveat The README marks this model as experimental, with context up to 64K, slower decode than the other models, and loading and requirements that may change between releases. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/nehanth-pooled/run-flash-125b-from-disk-experimental/