[vintagedave]

This is a great idea, but I noticed all the components I tried ended up looking very similar: a wide rectangle with (sometimes) an image on the left, and two sets of text captions, one pair left-aligned and one right-aligned.

My custom prompt,

> files in a file system

gave something almost identical to the predefined prompts:

> comedian shows in a comedy venue's booking application (which has an image too) > zany brainy stuff (has a centered label too) etc.

Is it more that it's not a UI component, but specifically only list items or cards, that it can generate? I did notice that it got the order of priority (eg filename first, size secondary) correct.

[designhat]

Thanks for the feedback! Yep it's only trained on examples of list items and cards, most of which follow the layout you described.

It does occasionally do different layouts, here's an example: https://designhat.ai/generations/7396ec47-ad9d-4f1e-90ce-ce9...

In the future I want to make it work for more types of components than just list items and increase the set of primitive elements it has to work with. Currently it's just arranging badges, labels, icons and images.