The Story Behind TMAI

I’ve been keeping an eye on AI coding assistants for a long time. I have to say, being "lazy" has actually become the driving force behind many things I do. Even before ChatGPT was released, I found an IntelliJ IDEA plugin (I’ve even forgotten the name now) that did some pretty smart code completion. As one of their early paid users, I was lucky enough to be selected by the dev team for a 30-minute chat. When I asked them about their future plans, they said that generating code using natural language was technically quite difficult, and that they might instead build a tool that translates code into natural language to explain it to humans.

Later, at re:Invent, I saw that Amazon had actually developed a tool called Amazon CodeWhisperer to autocomplete code. I just remember the tool’s performance was so-so, but the guy presenting the session was really handsome. I never expected that a few years later, ChatGPT would burst onto the scene, followed by GitHub Copilot, Cursor, Codex, Claude Code, and all the way to Google's Antigravity—each one better than the last.

Ever since I realized that AI coding was improving at an unbelievable speed, I knew that with the desperate efforts of major tech companies, the barrier to app development would get lower and lower. Sure enough, "vibe coding" started to catch on fast. Even at work, everyone is tacitly using AI tools on the sly, though no one admits exactly where they’re using them. Finally, one day, I felt that AI had become smart enough to write a complete project. So, I made up my mind to start researching the various tech stacks for independent development and how to use different vibe coding tools.

At first, I thought that even with vibe coding, I still needed to understand the code myself. So, based on my experience writing ReactJS, I chose to write my first app using React Native. I never expected this thing to be so hard to use. To be compatible with multiple platforms, you have to configure a ton of stuff just to get it running, and Xcode takes several minutes to compile every time. Later, when I wrote my first app with SwiftUI and found it compiled in tens of seconds—or even just a dozen—I was genuinely shocked. I also discovered that React Native’s third-party libraries are full of pitfalls, and using various native iOS features involves even more headaches, like system notifications, push notifications, and so on. I finally gritted my teeth and finished two apps, exhausting myself in the process. The upside is that I used this opportunity to familiarize myself with various cloud service platforms, development tools, and the App Store submission process. To this day, those two apps still have some weird bugs, and I hope one day I can rewrite them entirely in SwiftUI.

After finishing the first React Native app, I started brainstorming the next one. I have to admit, I really enjoy the process of enslaving AI to write apps for me every night. I often lose track of time and suddenly realize it’s 2 AM... As an ENTP debater, I have a huge number of new ideas. After thinking it over, I felt that a travel-related app was a great idea, but it was also much more complex than the apps I had written before, so I started working on a very careful architecture design.

Just as the design was coming together, I suddenly remembered that we were going to New York for a trip in a week. I had admired the Metropolitan Museum of Art (The Met) in New York for a long time but never had the chance to visit. Since we didn't have a packed itinerary this time, I felt we could go check it out. But the problem was, my partner really dislikes visiting museums. While I was secretly calculating how to persuade him, I had a lightbulb moment: isn't this exactly one of the features of the travel app I wanted to write?! It didn't seem too complicated, and it looked like I could whip up an MVP (Minimum Viable Product) in a week!

So, the happy times of enslaving AI every night began again. Sure enough, I managed to finish the MVP just before we went to New York. While wandering around The Met, I was excitedly snapping away with TMAI, and I also discovered a lot of issues. So after coming back, I continued to make AI do my bidding... After a long while, I finally patched up many interesting details and necessary features bit by bit, started beta testing with friends and family, and finally, after battling a few rounds with the App Store reviewers, it successfully launched.

In short, this app is a piece of work I really like, and it solves a curiosity-driven problem I often encounter in my life. If you, like me, have an intense curiosity and a thirst for knowledge, why not download this app and give it a try? You’ll definitely fall in love with it! If you have any good suggestions, please don't hesitate to let me know!