Gemini 2.5 Flash-Lite for instant UI! 🚀 Just vibe coded a chat app that only responds in dynamically generated UI to your prompts. No generation time was sped up. Enjoy five quick components in a matter of seconds, fun to see what fast models will enable. Demo + code below!

Fork it and remix the code here!

this is so cool, Flash-Lite is already able to generate UI's at basically the speed of thought

funny part is the slowest part in the video is my typing!

AI-generated UIs should wrap the internet 😎

I think they just might 👀

Using this just to make snake games

Humanity's last exam

Very cool Ammaar :)

thanks Kiri!

This is exactly the kind of latency breakthrough that will unlock real-time AI UX.

100%!

Lets connect!

This is amazing! Forked. Will experiment with the prompt for better ui mockups and will see how to handover these mockups to Cursor or Claude Code so agents can plug in the back end. Thanks

I use Gemini Flash Lite to summarize news articles. It's almost instant.

Right? I'm swapping to it in a bunch of apps that I needed faster latency on.

Google Team Keeps Cooking!🧑‍🍳

really cool, man.

thanks man!

this in vr and you truly have a pretty novel experience

That could be a fun experiment 👀

Damn. Use case unlocked.

Killing it !

thanks brother!

👀

Awesome, thanks for sharing the code!

anytime!

This is cool Ammaar 👍

Thanks Jay!

🙌

I like it very much. Good job!

Thanks!

Awesome

interesting concept, but let's not forget that pure speed doesn't guarantee quality. still, good to see practical use cases emerging. keep it focused, that's where the magic lies.

The speed is unreal. Imagine this baked into client tools or SaaS dashboards real-time adaptive UIs are going to become the norm.

Always cook’n!!

Very cool! Nice work

@grok which is best ai agent that can create smooth ui ux designs based upon the documentation. A complete user flows, screen design components styling css react native all of these together.

slow is smooth and smooth is fast

I like the app build in aistudio but it's still to much unreliable on changes (it says it applies changes but it doesn't) and you have to remind it a lot of times, any eta on this bug fix? any way I can monitor or give proper feedback? I'm a full stack dev (25+y) and I'm impressed by current level of result, even if I have constantly to remind him to apply refactor, follow best practice, separation of concerns and so on. I'd have a lot of feedback, any discord channel or some proper user group ?

i call taht a stock UI lib element now make it a full GUI with lite 🤭

Imagine if @PerplexityComet or @diabrowser supported something like typing /UI before a request to trigger a custom interface. That would be such a game-changer!

Played around with something similar last month:

I was so inspired by @ammaar 's Gemini UI generator, but I wanted more control over the final look. So I forked it to bridge the gap between AI speed and a deliberate design process. 🎨📷 (thread 📷

its circle

Hey @korablev perhaps something you want to play with for inspiration. This is the kind of workflows I was preaching you to explore

Impressed.. but real value unlocks when UI adapts to user behavior, not just prompts.. thoughts?