[authorfly]

I'm someone with a weakness for design (luckily, my other half is the opposite). I was so excited to try gpt-4-turbo and others in their vision mode to try and get design feedback.

Some was ok. But I soon realised, the key feed back, was about the content or amount of content/actions in relation to the rest of the app.

I have tried feeding in multiple steps as images. This hasn't really helped (maybe it's too complicated)

The overall journey of the user seems to be the most important part. If they slow down at one stage, it reduces their enthusiasm.. and a great later design only feels good

Is there any way your tool in the future could do that? More to the point, and very much an anti-thesis of your starting with an extension here, the use case I see is connecting this with screen recorders of users to analyze, which would be the opposite of an extension. But much more useful for me. I have thousands of sessions a day; tens of errors. I can't watch them all (and stay sane - I suppose I could watch moments before users purchase, but that's weird).

Any thoughts?

[purple-leafy]

The way I get around these issues is atomic design feedback, so my tool allows you to select a rectangle area, and when I use my tool I basically select a component sized area and get feedback component-by-component