The biggest challenge was translating the messy, human experience of trip planning into something a machine could understand. Flight search feels simple until you realize how many variables matter: Should it prioritize price or convenience? How do you rank the savings? I spent hours in Figma boards iterating on how to present overlapping results without overwhelming users, eventually settling on a card-based layout that felt like browsing rather than calculating. Prompting Claude required precision. I had to describe not just what buttons did, but why users would click them, what they'd expect to see next, and how the app should handle edge cases. Base44's strength was its speed in developing, but I needed to get the prompt right the first time. Vague instructions produced vague results, so I learned to be specific about interaction patterns and the exact flow between searching, comparing, and saving.