Did anyone solve this problem? Any good approaches?
Did anyone solve this problem? Any good approaches?
Perhaps it is much more basic: the words to describe really good user interfaces are inadequate to the task, and they are often emotional/psychological in nature. For example, look at how Mac enthusiasts talk about interfaces, talking about "delight", "surprise", "subtlety" on the positive side, "janky", "crass", etc. on the negative.
Even if an LLM had a deeper understanding of these words, they are vague, know-it-when-I-see-it things. It is often obvious to someone who sits down with an app for any time why Capture One or even Apple Photos feel nicer and more gentle to use than Lightroom or even DxO PhotoLab (made by people who do care about interfaces), why people still have a deep longing for Aperture even though it is lost, why Inkscape feels qualitatively better than many open source apps, or why the recent improvements to FreeCAD are just _obviously_ nicer, but the words to explain why, even to put screenshots or videos into context for a multimodal LLM, just aren't very precise. Apps are just good or bad in this regard, it is subjective, and tolerance to discomfort is subjective too.
For example, here is a really helpful linux app for converting STL to papercraft nets, a sort of open source alternative to Pepakura:
https://github.com/rodrigorc/papercraft
It's competent, it works, someone wrote it and gave it away; I have nothing bad to say about it. I am happy it exists.
Here is a Mac app that costs a bit of money but approaches the same task and is somehow a thing of genuine beauty:
The words to describe why this app has a better user experience are actually difficult to come by.
I have a book on my bookshelf that I look at occasionally, to help me clarify thoughts:
https://www.amazon.co.uk/Childrens-Technology-Kaufmann-Inter...