5 comments

[ 0.22 ms ] story [ 150 ms ] thread
hi guys I built this tool, its actually working way better than i ever expected. I had the problem a lot that i tried to copy a certain ui but the agent just massively hallucinated. I simply implemented evrything around a single feature.

The agent renders the actual image of the created web page, and checks the actual diff against the original.

Thus the agent is literally forced to iterate till its done.

Right now its still taking some time to be done. So I wish that some smart minds improve the algoirthm and make it quicker and more efficient.

I also cloned some other websites just for testing purposes with anonymized logos. But i am not sure if i can upload all due to copyright risks. Its literally 1 to 1 so idk.

What's the intended use case? Porting from framework A->B?

Does it handle responsive design? Keeping the same breakpoints etc?

This would be great for scammers
no it only copies exactly one resolution. However if one person were smart it could implement 3 screenshots which it would adapt to in parallel and check each screenshot. This would preserve scaling. However i havent impelemtned that yet
I can't see a real world use case for this.

If you actually want to make something worth the effort, you'll have to make it responsive and accessible, along with actually implementing whatever functionality your project needs.

At that point, it seems guaranteed to be more work to try to make all that happen with a visually "perfect" but implementationally soulless clone than to just mimic the design while building out your interface.