My Midjourney sessions often require a ton of iterations (image is one session). There's a fun back-and-forth between tweaking prompts, making manual edits, and seeing what the randomness of the models produces. You can evaluate an image in a split-second, which is not the same with code or text. But it should be doable to an extent with generated UI, but the tools for that don't seem to embrace this type of quick iteration with multiple suggestions from what I've seen. Might be helpful.
