I had ChatGPT critique my app like it was a real user.

I had ChatGPT critique my app like it was a real user and it instantly told me I had no idea who my users were. See, I thought I had two core personas, college roommates and couples. They share bills and expenses, so the app would help them manage that. And chat goes, those aren't personas, those are buckets. So I said, deep research on those buckets, then interrogate me with trade-off questions. After a few minutes, it's grilling me with this or that questions. Focus on cash flow coordination or fairness perception? Prioritize equal splits or flexible rules? Choose one, roommates or couples? Now I thought that we would move on from there, but it kept pushing me. Because roommates have low trust and high churn, while couples are the exact opposite. So they needed completely different features. So I was super nice back to ChatGPT and I finally picked a lane. Time to critique. It hit me with breakdowns of thoughts that these users would have going through my onboarding. It was clear that my designs needed some work and the key takeaway was that I needed to switch to an outcome-first approach. But I'm a visual learner, so I did two things. First, I had it generate updated designs to address the critique. Wow, that's beautiful. But I wanted it more minimal. So I refined it a few times with the new images 2.0 and then I told it to use those mockups as inspiration and the chat as reference to make me a prompt for Codex, which let me import all of the screens into my design tool and then, we're designing. But doing it for a much clearer user persona that's made my entire product better and now it's actually ready to test on real people, or buckets.