Rust powers 7MB AI terminal

Desktop AI tools usually get big fast. You add a web stack, an embedded browser, model connectors, update plumbing, and suddenly the “lightweight” helper is a chunky app with cloud assumptions baked in. Terax is interesting because it pushes the other way. A new demo from crynta shows an AI terminal built with Rust and Tauri that wraps terminal tabs, a code editor, file browsing, and an AI side panel into a package the project describes as roughly 7 MB. ### What actually shipped? What showed up this week was a YouTube demo — “I built my 7MB Full AI Terminal in Rust & Tauri” — plus a public GitHub repo for Terax-AI. The repo describes Terax as an open-source AI terminal emulator built on Tauri 2, Rust, and React 19, with a native PTY backend and a modern desktop UI. ### What is Terax trying to be? Basically, not just a chatbot in a window. Terax is aiming at the space between a terminal, a lightweight editor, and an AI assistant. (youtube.com) The current feature list includes multi-tab terminals, an integrated code editor, a file explorer, and a side-panel for AI agents and chat. That matters because most developer AI tools split those jobs across several apps — terminal here, editor there, browser tab somewhere else. ### Why does the 7 MB number matter? Because distribution is product design. A tiny binary means faster downloads, faster cold starts, lower memory overhead, and less friction for someone who just wants to try the tool. Tauri helps by using the system webview instead of bundling a full Chromium runtime, and Rust helps by keeping the backend native and lean. The result is closer to a pocketknife than a toolbox. (github.com) ### Is this fully local AI? Not automatically — but it is local-first in the useful sense. The repo says Terax works with your own API keys and can also use fully local models through LM Studio. So the app itself runs on-device, keeps credentials in the OS keychain, and avoids telemetry, while model execution can be either remote or local depending on how you wire it up. ### Why use Rust and Tauri for this? (github.com) Because this is one of the cleaner stacks for “native feel without native-app bloat.” Rust is good at long-running system tasks like terminal handling, process management, and file operations. Tauri gives you a desktop shell with a smaller footprint than Electron-style packaging. For a terminal app, that tradeoff is especially attractive — users care about startup time, responsiveness, and not having the UI eat the machine. ### What’s the catch? The catch is that tiny app size does not mean tiny total system footprint once local models enter the picture. A 7 MB shell can still sit on top of gigabytes of model weights if you want fully on-device inference. So the real trick here is separation — keep the interface, terminal, and workflow layer lean, then let users choose whether intelligence comes from an API or a local model server. (github.com) ### Why are people paying attention? Partly because the project hit Hacker News almost immediately after the video went up. But more importantly, it taps into a real mood among developers: AI tools do not have to be giant, cloud-tethered workspaces. There is still room for fast, opinionated utilities that do one workflow well and stay easy to install. (github.com) ### So what’s the bigger point? Terax is not important because it proves every AI app should be 7 MB. It is important because it shows the interface layer can stay small even when the AI layer gets complicated. That opens a lane for builders — especially solo developers — to make local-first desktop tools that feel sharp, ship fast, and do not require an entire platform strategy on day one. (youtube.com) (hn.algolia.com)