Google's neues Gemini hat die UI in einer Nacht ersetzt: Ein einzelne HTML-Datei kann macOS exakt nachbilden, mit einer Erfolgsrate von 100%.
The work of front - end UI has been eliminated overnight by Google AI.
Just in the latest capability demonstration, Gemini 3.0 Pro actually "created" a macOS by itself.
With just a few lines of prompt words, Google's AI replicated Apple's proud UI in the browser.
Prompt: Design and create a web os like mac os with full - functional features from text editor, to file manager, to paint, to video editor, and all important pre - bundled software in macOS. Use whatever libraries to get this done, but make sure I can paste it all into a single HTML file and open it in Chrome. Make it interesting and highly detailed, show details that no one expected. Go full creative and full beauty in one code block.
Apple - style animations, window minimization, toolbars, browsers... Everything is available, and all functions work properly.
There is even an Easter egg —
Enter "matrix" in the terminal, and it will trigger the same cool special effects as in "The Matrix".
The key lies in the success rate!
Since it's impossible to generate repeatedly in A/B testing, all you see above is completely the result of a single generation, and the source code has been made public on CodePen.
Netizens exclaimed that if the official version can really reach such a level, Gemini will undoubtedly become the most powerful programming model in history.
Everything Can Be Turned into a UI
Within just one day, Gemini 3.0 Pro has covered the PC operating system market dominated by Apple, Microsoft, and Linux.
Some netizens tried to let Gemini 3.0 Pro generate a web - based Windows, and it succeeded on the first try.
Moreover, it not only has Python built - in the terminal, but also can play games and run code. Its functions are incredibly complete.
Prompt: Design and create a web os like windows os with full - functional features from text editor, terminal with python and code editor, and a game that can be played, to file manager, to paint, to video editor, and all important pre - bundled software in windows os. Use whatever libraries to get this done, but make sure I can paste it all into a single HTML file and open it in Chrome. Make it interesting and highly detailed, show details that no one expected. Go full creative and full beauty in one code block.
No comparison, no harm —
Some netizens showed the "masterpiece" of Claude 4.5 Sonnet with the same prompt words. Even the application icons can't be clicked.
Linux is naturally not spared.
The result is still good. You can visit Wikipedia, call the calculator, and change the wallpaper.
Prompt: Create a fully functional Linux desktop environment (Ubuntu/GNOME style) as a complete web operating system in a single HTML file with embedded CSS and JavaScript. All applications must be fully functional. Use whatever libraries to get this done, but make sure I can paste it all into a single HTML file and open it in Chrome. Make it interesting and highly detailed, show details that no one expected. Go full creative and full beauty in one code block.
Seeing these, the silicon - based life faction is very satisfied:
Who says LLMs can only parrot others?!
After all, there isn't even a single line of source code for macOS and Windows on the Internet.
But some netizens pointed out that this is just a simulated environment and is still far from a real OS.
It is indeed impressive, but we must distinguish the difference between "simulation" and "implementation". These are actually just "glue code", and the underlying technologies all rely on existing ones. Just as a programmer can't take credit for calling a library, neither can AI.
One More Thing
Although Google has not announced the official release date of Gemini 3.0 Pro, based on the previous leakage rhythm and release rules, some industry insiders speculate that it may be unveiled in the next few months.
An obvious sign is that there are more and more demonstration videos from various influencers on the Internet.
Somehow, we are seeing more demos of Gemini 3 before its official release than we did for 2.5 Pro.
The successful marketing of NanoBanana in the previous round obviously made Google taste the sweetness of "pre - heating", and this time it may want to take another gamble.
However, the higher the expectations, the greater the gap. Just like what happened to GPT - 5 back then.
Friends who are interested can go and operate on CodePen~
macOS:https://codepen.io/ChetasLua/pen/LEGzZaQ
Windows:https://codepen.io/ChetasLua/pen/yyezLjN
Linux:https://codepen.io/ChetasLua/pen/LEGzZaQ
Reference links:
[1]https://x.com/chetaslua/status/1977936585522847768
[2]https://www.ainewshub.org/post/google-gemini-3-pro-rumors-release-date-features-and-what-to-expect-in-late-2025?utm_source=chatgpt.com
This article is from the WeChat official account "QbitAI". Author: Jay. Republished by 36Kr with authorization.