Vibe Coding XR: Google’s New Trick for Prototyping Spatial Apps in Under a Minute

The phrase “vibe coding” has been floating around for a while now, mostly describing the feeling of letting an LLM handle the boilerplate while you focus on the fun stuff. Google Research just took that concept and strapped it onto extended reality.

They’re calling it Vibe Coding XR. It’s a workflow that ties Gemini Canvas to the open-source XR Blocks framework. The result? You type something like “create a beautiful dandelion” and get back a fully interactive, physics-aware WebXR application that runs on Android XR headsets. The whole thing takes under 60 seconds.

Why this actually matters

XR prototyping has always been a pain. You need to stitch together perception pipelines, fiddle with game engines, and deal with low-level sensor integrations before you can even test a basic interaction idea. That’s a lot of friction for something that might get thrown out after a quick demo.

Vibe Coding XR sidesteps most of that. The system uses Gemini’s long-context reasoning, specialized system prompts, and curated code templates to handle spatial logic automatically. You don’t need to know anything about XR to get a working prototype. Just describe what you want.

The workflow is straightforward:

• You open the XR Blocks Gem in Chrome on an Android XR headset (or on desktop with the built-in simulator).
• Type or speak a prompt. It can be as simple as “create a beautiful dandelion.”
• Gemini designs and implements the XR experience using sample code from XR Blocks.
• You pinch the “Enter XR” button and see the result instantly — an animated dandelion that blows away when you interact with it.

If you’re on desktop, there’s a simulated reality environment so you can test interactions before deploying to a headset. Some features like depth sensing and hand tracking obviously work better on actual hardware, but the simulator is good enough for rapid iteration.

What’s under the hood

The technical side is less about new models and more about clever orchestration. Google’s team has been iterating on this for the past year, refining the prompts and templates so Gemini can consistently generate functional XR scenes. The system taps into XR Blocks’ physics engine and perception modules, so you get things like gravity, collision detection, and hand interactions without writing a single line of code.

The catch (there’s always one)

This is still a prototype tool. The demos look impressive, but I’m curious how well it handles complex prompts. A dandelion is one thing; a multi-room spatial layout with conditional interactions is another. Also, the output is WebXR, which means it’s tied to browser-based XR. That’s fine for quick demos, but production apps will still need proper native development.

Where to try it

The team is showing this live at ACM CHI 2026, but you can also try it right now. The XR Blocks framework is open-source on GitHub, and the Vibe Coding XR workflow is accessible through Gemini Canvas. Links are in the original announcement.

I’d love to see where this goes. If it matures, it could lower the barrier for spatial computing experimentation the way tools like Three.js did for 3D on the web. For now, it’s a fun way to validate an idea before committing to a full build.