Hi Hackers! I am working on a problem on a code review tool for designers and have a use case I am trying to solve with GenAI but not able to solve well. LLM models are trained on huge corpus of text, and many multimodal models also on images with text caption that can identify objects in the images. However, models are not necessarily trained on more abstract visual concepts like aesthetics, contrast, symmetry or such concepts. If I want to take an app design and ask a model to provide feedback on it from a design perspective, what might be some ways to do it using GenAI, where it can also give pointers on design and aesthetics (like contrasts, font sizes, symmetry, etc..)?

I understand it's slightly open-ended, and "design is very subjective", but at the same time there are very good principles for design so it seems plausible that the design languages and principles can be codified?