Why Your AI IDE Shouldn't Weigh 800MB

Most AI editors are Electron apps that eat your RAM. Origin is built on Tauri and ships as an 8MB binary. Here is why that matters.

Right now, open your activity monitor and look at your editor. Not building. Not running tests. Just sitting there with two files open. On a lot of machines, that idle editor is sitting on 400 to 600MB of RAM. Doing nothing.

Let that sink in. Half a gigabyte to display some text and blink a cursor.

It got that way because your "editor" is a full copy of the Chrome browser wearing a hoodie. That is the dirty secret of the AI editor market, and I got tired of paying for it. So did everyone who built Origin. A lightweight AI code editor should not cost you more memory than the app you are actually building.

The numbers nobody wants you to look at

Let's stop being polite about this. Here is the comparison, idle, with an AI assistant loaded:

• VS Code with an AI extension: 200MB+ on disk, 400 to 600MB RAM at idle. Add a language server and a couple of extensions and you can watch it climb past 800MB before you have typed a single character.
• Origin: 8MB download. A small fraction of that idle memory, because there is no browser engine running in the background pretending to be an app.

Eight megabytes. Not eighty. Not eight hundred. Eight.

That is not a clever optimization or a setting you can toggle. It is the difference between two completely different ways of building a desktop app. One of them ships a browser inside every single program. The other does not. That single decision drives everything else.

Why this actually costs you

The "RAM is cheap, who cares" crowd misses the point. Nobody runs just an editor. A real session is the editor, plus a dev server, plus a database, plus three containers, plus forty browser tabs, plus Slack quietly eating another 500MB in the corner.

Every megabyte your editor wastes at idle is a megabyte your compiler, your tests, and your actual program never get to touch. On a 16GB laptop you can lose a meaningful chunk of your machine before you start working. Then the symptoms show up, and most people blame the wrong thing:

• Sluggish cold starts. A big binary loads slowly and initializes slowly. You wait.
• Swapping under pressure. RAM gets tight, the OS starts paging to disk, and suddenly everything stutters at once.
• Battery you can feel draining. A heavy runtime burns more power doing the same job.
• Fans. The audible tax. Your editor should not have a soundtrack.

None of these are dramatic alone. Stacked across an eight-hour day of thousands of tiny interactions, they are death by a thousand paper cuts. And it is the one tool you touch more than any other.

Electron ships you an entire browser. Every time.

Here is what is happening under the hood, plainly.

Electron apps bundle a full copy of the Chromium browser engine and a Node.js runtime inside every single app. That is genuinely convenient for the people building the app, because it runs identically everywhere. But the cost lands on you. Every Electron app is, functionally, a separate full browser installed on your machine. Run ten Electron apps and you are running ten copies of Chrome. You would never install ten browsers on purpose. You just did it by accident.

That is why the binaries are huge. That is why the idle memory is brutal. There is a browser sitting there doing browser things even when all you wanted was a text editor.

What Tauri does instead

Origin is built on Tauri, and Tauri makes the opposite bet.

Instead of shipping a browser, it borrows the one your operating system already has. On Windows that is the WebView2 runtime. On macOS it is WebKit. On Linux it is WebKitGTK. The frontend renders in a webview that is already on your machine, and the actual application logic runs in Rust, a compiled language with no garbage-collected runtime to drag around.

The wins fall straight out of that:

• No bundled browser engine. That is why it is 8MB instead of 800.
• No Node.js runtime sitting in memory. That is why startup is instant and idle RAM stays low.
• Rust on the backend. The heavy lifting runs as compiled, memory-efficient native code, not interpreted JavaScript.

Origin is not a website cosplaying as an app. It is a real native shell wrapped around a web frontend, using parts your OS already shipped. The browser was always there. We just stopped making you install another one.

The honest tradeoff

No architecture is free, and I am not going to pretend otherwise. Leaning on the system webview means rendering can differ slightly between platforms, and Tauri's ecosystem is younger than Electron's decade-long head start. For some teams those are real concerns.

For a code editor, the math is not close. You do not need pixel-identical rendering across every OS to write code. You need a fast, quiet tool that gets out of the way and leaves the machine to your actual work. The lighter path wins that argument every time.

A light tool makes you faster in ways you stop noticing

Productivity is mostly the absence of friction. Every tiny delay between wanting something and getting it is a micro-tax on your focus, and those taxes compound across hundreds of actions a day.

A lightweight AI code editor pays that debt down constantly:

1. It opens instantly. Quick edit to a config file? Jump in, jump out. No spinner, no splash screen, no waiting for a runtime to wake up.
2. It leaves the machine for your real workload. The RAM the editor was hogging goes back to your build, your tests, your app.
3. It stays calm when things get heavy. Low idle usage means real headroom when the work piles up, so the editor never becomes the bottleneck.
4. It travels anywhere. An 8MB download installs in seconds on a cloud VM, a borrowed machine, or the ancient laptop in your bag.

There is a quieter payoff too. When a tool is this light, you stop thinking about it. You are not closing it to free up RAM. You are not waiting for it to load. It just sits there, ready, costing almost nothing. That is what a tool is supposed to do, and most of us forgot it was even an option.

Light because of how it is built, not how it is marketed

Origin's footprint is not a stat we chase for the homepage. It is the direct, unavoidable result of choosing Tauri over Electron, Rust over a bundled runtime, and the system webview over a shipped browser. We made those calls so the editor could disappear into your workflow instead of dominating your resource graph.

And you lose nothing for it. Agent Mode, Plan Mode, inline AI diffs, model switching, the whole modern AI editor experience runs on top of that lean base. You get all of it without the weight.

Your editor should be the lightest thing in your stack, not the heaviest. Eight megabytes, not eight hundred. Work this way for a week and going back feels like strapping on a backpack full of bricks you forgot you were allowed to take off.

Origin is free, open source, and runs on Windows, Linux, and now macOS. Download it on GitHub and give your RAM back to the work that actually matters.