Vibe coding tools have this weird framing problem where everyone talks about them as if they were strictly for developers. Which might make sense on the surface because the output is code and the interfaces of these tools often look like code editors. But the point of these tools is that, first and foremost, they are natural language: they are designed so that you can describe what you want in simple language and then you can create it. That’s the real tone. Vibration Coding Tools They are intended to make coding tasks more accessible to the general population.
And I think designers, especially, should snoop around in them. There is increasing overlap between designing a user interface and creating it, and tools that were built with AI first tend to handle that better than design tools that added AI later. Antigravity is something I’ve been curious about for a while, and I finally sat down with it to see if a designer with no development experience could actually use it for design work.
Do you want to stay up to date with the latest in AI? The XDA AI Insider newsletter is published weekly with deep dives, tool recommendations, and practical coverage you won’t find anywhere else on the site. Subscribe by Modifying your newsletter preferences.!
Configuring the agent first is doing something that the design tools haven’t figured out yet
Antigravity was launched in November 2025 as Google’s agent development platform. Technically, it’s a fork of VS Code with a heavily modified interface, but calling it an IDE understates what it’s actually doing. It has two modes: an Editor view that looks like a normal code editor, and an Administrator surface where it spawns agents that schedule, execute, and verify tasks on their own in the editor, terminal, and browser. Ultimately, you Don’t really touch the code unless you want to. You describe a goal, the agent writes a plan, approves it, and delivers what Google calls artifacts. Basically, these are just implementation plans, tutorial documents, screenshots, and a running preview that you can open in your browser.
For a designer, the agent testing its own output in a browser really matters. The agent can control a browser to test its own work, take screenshots, and iterate based on what it sees. This is a feedback loop that most traditional design tools don’t typically have, and even the native AI design tools I’ve been playing with don’t exactly do this. You can also choose the model: Gemini 3 Pro is the default, but you can change it to Claude Opus or even GPT-OSS mid-session. I ran mine on Gemini 3.5 Flash on the Medium setting, which is free, but I got more out of it than I expected.
Designing in Antigravity is like having a briefing team
You will receive a working version of your description.
The way you design in Antigravity is hardly design in the traditional sense. You open a new conversation, describe the UI you want, and the agent takes care of everything from the folder structure to CSS to running a local server so you can preview it. There’s no canvas to organize things, so the closest you get to an “iteration” is to go back to the chat and ask for changes or feedback.
For my test, I asked him to create a habit tracking dashboard. Five habit cards in a horizontal scroll, tap to check with animation and a streak counter that actually updates, a weekly progress points matrix, a floating action button with an add habit modal (name field, six icon options, six color options), and an empty state. The style direction was Things 3 and Duolingo with a warm off-white background.
He first drafted an implementation plan, which I had to approve before moving forward. From there, it ran in its own project folder on my machine, generating index.html, style.css, app.js, and a Node server. And it stopped to ask permission several times along the way because anything outside the workspace or involving the terminal defaults to manual approval. Approximately seven minutes later, the prototype was active on a localhost address. He chose Google Fonts (Outfit and Inter), added glass morph headers, wrote a confetti burst for the check animation, and even added a Duolingo-style pot mascot for the empty state. I didn’t put those parts in the message (I deliberately left it a bit open to see where Antigravity would improvise).
The results here open up a workflow that I don’t think designers have considered yet. You’re not just generating a static mockup, but something you can actually use and test, with real interactions and in real state. The verification artifact it produces (a walkthrough.md file with manual test steps) is also really useful. This way, you get an interactive prototype plus documentation on what it does and how to test it, in one pass.
For early-stage exploration, where you’d normally spend an hour wiring up Figma prototype connections just to fake an interaction, this seems like a much more efficient way to test something before finalizing the design. And since it runs locally, you can send someone a screen recording or invite them to access the same server. Overall, Antigravity gives you real usability tests in a draft, not a clickable illusion.
Figma still runs the design world
And probably always will
Nothing here is a case for getting rid of Figma. I’ve said this many times before, but Figma has the entire design stack that really matters for professional work: component libraries, design systems, developer handover, automatic layout, development mode, team collaboration, etc. None of that is what Antigravity is trying to do anyway, I don’t think Google intended to do it. And even the AI features that Figma has added, like Figma Make, aren’t bad, they just feel different from what a tool like Antigravity does because the starting point is different.
Figma is a manual editor who added AI. Antigravity is an artificial intelligence platform that does the work of designing results. Those two things approach the same problem from opposite ends and end up producing quite different results. When I want to build something quickly to see if the idea holds up, Antigravity is faster. When I really want to design something to a professional standard with the right components and specifications, Figma is still the default.
The best design workflow uses more than one tool
Design and AI can no longer be separated at this point. Every design tool now has some form of AI built in, even the odd open source standalone one. But tools that were first built with AI from the ground up will likely always feel a step ahead of those that add it as a feature. So while Antrigravity helps me iterate faster and also gives me insight into the development side, I think it’s best to use them together. There’s even a Figma MCP that you can use to connect your design environment directly to the Antigravity IDE.
