Rust builds 7MB AI terminal

A terminal app is one of the oldest pieces of developer software around. An AI terminal is one of the newest. Terax tries to smash those two worlds together without dragging in the usual desktop-app bloat — and that’s the whole story here. In a YouTube demo published this week, the developer behind the project, crynta, showed an open-source AI terminal built in Rust and Tauri that he says comes in at about 7 MB, with the GitHub repo describing it as under 10 MB on disk. ### What is Terax, exactly? Terax is basically a desktop terminal emulator with extras developers normally piece together from separate tools. The repo lists multi-tab terminals, a code editor, file explorer, AI side panel, web preview, and agent-style workflows in one app shell. Underneath, it uses Tauri 2 and Rust for the native layer, React 19 for the UI, xterm.js for the terminal, CodeMirror 6 for editing, and portable-pty for process handling. (youtube.com) ### Why does the 7 MB claim matter? Because most AI desktop tools have started to feel like mini operating systems. The developer says Terax began as a reaction to seeing Warp’s app size north of 400 MB, and that gives the project its angle — not “can we add AI to a terminal?” but “can we do it without hauling around hundreds of megabytes of framework overhead?” Tauri helps here because it leans on the operating system’s own web renderer instead of bundling a full browser engine, which is why Tauri says apps can get very small. (github.com) ### How does it stay small? The trick is architectural, not magical. Rust handles the native backend and system-facing work. Tauri wraps a web UI without shipping Electron-style bulk. Then the app keeps AI flexible instead of embedding one giant model by default — the repo says users can bring their own API keys or connect fully local models through LM Studio. That means the app itself stays light, and the heavy model weights live elsewhere if you want them. (news.ycombinator.com) ### Is this really local-first? Mostly, yes — with an asterisk. Terax says there’s no telemetry, and API keys are stored in the OS keychain, which is the kind of detail that matters more than splashy “privacy” claims. But “local-first” here does not mean “everything runs locally all the time.” It means the app can work with local models, and it avoids forcing users into a hosted stack. If you plug in cloud APIs, you’re still using the cloud. (github.com) ### Why build this in Rust and Tauri? Because the project is making a systems argument as much as a product argument. Rust gives tight control over performance, memory behavior, and native integration. Tauri gives desktop reach without the usual packaging penalty. That combo has become a favorite for developers who want native-feeling tools but don’t want to maintain a giant C++ app or ship Electron-sized binaries. (github.com) ### Is this a big product launch? Not really — at least not yet. This looks more like an early open-source project getting attention through a demo video and a fresh GitHub release cadence. The repo is only a couple of weeks old, the video landed two days ago, and the project has started popping up on Hacker News. So the news is less “a new incumbent has arrived” and more “a clear proof point just showed up for a different way to build AI desktop tools.” (github.com) ### What’s the real takeaway? Terax matters because it pushes back on a lazy assumption — that AI software has to be huge, slow to start, and glued to someone else’s cloud. Turns out a lot of the heft is optional. If this design pattern sticks, the interesting shift won’t just be smaller terminals. It’ll be a broader return to compact, local-capable developer tools that feel more like software and less like a browser tab in a trench coat. (youtube.com)