# Chat Computer

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Ren Shu Tan](https://cerebralvalley.ai/u/RenshuBTC)
- **GitHub:** https://github.com/RenshuBTC/chat-computer
- **Demo video:** https://renshubtc.github.io/chat-computer/media/
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4814

Chat Computer is a computer you talk to. Our goal is a chat-centred user experience as a full computer: one conversation instead of apps, windows and settings menus. Underneath, the goal is an AI-native operating system that is free and open source (Apache 2.0).

It is a whole operating system written from scratch in Rust: its own UEFI start-up program, kernel, drivers (NVMe, SATA, USB, HD Audio, network cards), TCP/IP, TLS 1.3, file systems (NTFS, FAT, exFAT, BitLocker) and model engine. You type or speak, and the computer answers and does it.

For this hackathon, Meta's Muse Spark became its brain. The kernel itself sends the conversation and the operating system's tools to Meta's Model API over its own network stack and TLS, and Muse Spark runs the computer through tool calls: the computer's status, text size, timers and reminders, volume, files on the PC's drives (even a BitLocker C:), web search and pages, memory, restart and power off. Press Windows+H to speak: Whisper recognises your words on the PC.

Chat Computer's full screen (its app, and chatcomputer.ai) shows your messages from every chat app on the left (WhatsApp, Gmail, Telegram, Instagram, X, LinkedIn and more) and Actions on the right, where every action is recorded and any can be reverted with one click. Typing an address like google.com opens the site right there. Muse Spark answers there too.

The problem: computers are still built around apps, files and settings, and many people find those hard, including older people, children and anyone who never learned. A conversation removes that layer, and a frontier model like Muse Spark makes it capable enough to run a real computer.

It is local-first: without a key, an open model runs on the PC itself, offline; with Muse Spark, the status line says that what you say goes to Meta. It is built with today's frontier models from Anthropic, OpenAI and Meta, and works with each of them.

Chat Computer is still a work in progress. The own system runs in virtual PCs (QEMU) so far; real PCs, more tools and more everyday requests come next.

## More from Meta Global AI Developer Hackathon

- [Viil Sentinel](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4811)
- [Puddle](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4812)
- [SIMT](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4813)
- [AxVigil](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4815)
- [Vwoosh](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4816)
- [CovertLens](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4817)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4814. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt