Imagine typing a web address into your phone and, instead of fetching files from a server, an AI model builds the whole page right on the spot. That is the exact concept behind Mirage, a new experimental browser project created by Vidy Thatte, a developer on the Google AI Studio team.
Instead of loading static HTML and running standard scripts, Mirage treats whatever URL you type in directly as a prompt. Vidy showcased the app in action on X, explaining that with Gemini 3.5 Flash Lite, “there’s almost no difference between loading a website and generating a new one on the fly.”
With gemini 3.5 flash lite, there’s almost no difference between loading a website and generating a new one on the fly.
I built a little browser that treats every url as a prompt and generates a site from scratch.
who wants the testflight? pic.twitter.com/fWIKSyD1GZ
— vidy (@vidythatte) July 22, 2026
As seen in the latest demo videos they posted, you can visit actual live websites and use Gemini to remix or edit them in real time. You can change dark modes, switch visual themes, modify text, or tweak layout elements on the fly.
new update: you can now visit any (real) website and edit it using gemini ✨
testflight in comments https://t.co/yUQbNsWzo3 pic.twitter.com/DsxfCyVy7W
— vidy (@vidythatte) July 27, 2026
For those who want to try out the browser, you’ll have to grab an API key and then download the TestFlight version of the app from here.
We’ve already seen just about every browser out there throw in an AI chatbot in the sidebar, but Mirage approaches the web from a completely different angle. It turns the browser from a passive document viewer into an active creation tool.
That said, I do not think this is going to replace your standard web browser anytime soon. I like the idea a lot, but I personally can’t imagine myself using it to generate random web pages, or even tweak existing ones for more than a day or so. The novelty factor will likely wear off quickly.
Ultimately, Mirage is a clever proof of concept rather than a serious browser replacement. But it is still pretty impressive to watch a phone generate page layouts almost instantaneously. Google is already trickling AI into browsing through smaller features like the upcoming AI virtual try-on feature and its new AI embedding API. Of course, these are just a scratch on the surface. We at PiunikaWeb have highlighted dozens of other AI-related additions in browsers lately.
So we wouldn’t be surprised if more such AI-powered browser experiments emerge in the future.
The post This experimental browser uses Gemini to let you edit real web pages in real time appeared first on PiunikaWeb.