It looks like a lot of their early attemps are early gpt outputs or claude design outputs, which have a lot of the same smell. People are discounting how good image->html flows are these days though.

I had gemini describe this webpage as prompts, then fed it into an image -> code pipeline. Note that this is with 0 input from me:

https://html.non.io/railcode-animated

This is with Opus 5.5 building it out on low effort, and yes the spec was already made from the original site, but 0->1 that doesn't result in purple gradients or claude brown is very much possible now.

Honestly looks better than the original