At first, AI didn’t make much of an impression on me. I mostly used it to look things up, chat, or help polish a Xiaohongshu post or make a cover image. That was about it. I hadn’t used it deeply enough to feel much about it.
Later, I started using Codex intensively and tried building apps that could solve problems of my own. That hands-on experience completely changed how I saw AI. Zeying was one of those projects.
It began with a specific question: which of these similar photos should I keep? I wanted a Mac tool that could put them together so I could compare and choose. So I started building the app myself.
Once I got to work, details that had not been clear when merely discussing requirements began to surface in real testing. Early tests with my own photos sometimes grouped different buildings or different paintings together. Some photos looked similar at first glance but showed different subjects. Conversely, shots of the same person in different poses, or in portrait and landscape compositions, might be worth comparing together. There were far more specific questions behind the simple phrase ‘duplicate photos’ than I had expected.
I had to look through those groups one by one and make adjustments. Zeying moved forward through that process of trying and testing. It is not yet on the App Store, and I am still checking it with my own photos.
That experience gave AI more uses in my everyday life. I made EchoLine to practice everyday speech with YouTube creators. I also tried using astra to make a highlights vlog from thousands of photos I took while living in Osaka two years earlier. Each time, I started with something specific I wanted to do and then explored how much the tools could help.
Of course, having more tools sometimes feels like getting new toys: I want to try them all. Later, I began asking myself: if copying and pasting would do the job, why have two agents pass messages back and forth? That extra round also takes resources and attention.
Zeying came from that sequence of building, trying and adjusting. Whether a tool I want to use is actually useful ultimately comes back to the photos waiting for me to sort them.