Mobile UX Prompting: Using Physical Constraints Instead of Design Requests

My latest AI app just feels off. It made me realize that the models I use to build these apps have never actually used a phone. They can't feel what's in my hands, so whatever they give me, they just think it's perfectly fine. And I realized, I actually have to tell it, "Here's my phone. Here are the problems with using it." And here's what I tell it. It should be much more of like a phone design because you have a lot of space on an iPhone Pro Max screen. So we have stuff at the bottom, and then you also want to be mindful of people's fingers because most people's thumbs can only reach a certain range. And so we have to like think about this and kind of decompose it apart a little bit further. And the way that I'm thinking about this is that whenever a person is using their left or right hand, there's a certain reachability aspect. So we should grade higher for things that are reachable within a thumb point of view, and then use the rest of the space that's on top as far as any information that the user needs to see, but doesn't necessarily need to interact with. So I want you to think a little bit deeper here about this problem and think about gestures and taps and things that we can do. But we want to keep it as simple as possible. We should pretend we only have like about one and a half to two seconds to do an interaction, and we shouldn't need the user to interact. It should be as simple as possible so that even a baby can pick this up and really intuitively understand what to do here. So that's it. A thumb reach, one and a half seconds, a baby can actually figure this out. None of that is special. I've been using a phone for all my life, and what's real is that the model actually has not. So I actually have to say that out loud. You can see the passive zone and the thumb zone. So we're kind of seeing how the model's now understanding what our intent is, and how we can design some UX interactions, and it gets us to start thinking about, "How do we want to do this here, right?" Like here in Spanish, so like you can, with one thumb, without reaching too far, you can do all these interactions here, right? So the top 65 has no important controls, and only shows what helps the user trust the system. So that's the detected input, transcript. Like all the other stuff that's peripheral can just like literally just be up here. You actually get more value taking things out of the interface than you do actually keeping them there. You don't want to have too many things in the way. So I'm not really asking the model to design here, I'm just telling it what's actually going on with my phone. There is a thumb, there is about a second and a half, and a baby should figure it out. That's the prompt. And if you want me to keep building this out, just drop a comment, and let's just go ahead and see where this goes.