AI로 진짜 앱을 만들어봤더니, 걸린 시간은 무려 1년

I Tried Building a Real App With AI. It Took a Year. 17 Jul 2026 #ios #ai #programming Alex Hyett 75.3k subscribers Watch on PeerTube Watch on YouTube Subscribe so you don't miss out on new videos Subscribe on PeerTube Subscribe on YouTube My wife and I have been looking for an app to track things like habits, hobbies, and when we last did certain chores. To be honest there are probably hundreds of different habit trackers on the App Store. Next to todo list apps they are supposed to be the easiest apps to create so it is not really that surprising there are so many options. But we must have tried dozens of different apps but none of them quite fit what we were looking for. Of course now we have AI, so this sort of thing should be easy to create. You just right a couple of prompts and AI will build it for you, while you sleep, or sip cocktails on the beach. As we shall see it didn't quite turn out that way. Requirements and Existing Apps # So for this app I had a few requirements: We wanted to have the habits in a long list and not have to go through multiple pages to find them. We didn't want any data be stored in the cloud or on other peoples servers. We wanted to be able to use the gen-emojis from iOS 18 for some of the icons. It is a great option when the predefined icons don't fit. We needed to set goals on a daily, weekly, or monthly basis. It should be possible to have no goal and just count the number of occurrences. My wife sometimes gets headaches so she wanted to keep track of all of those as well. We wanted the ability to add notes. I wanted to be able to write down what exercises I did in each of my workouts and what speed I had got a song to when practicing guitar. We wanted to set reminders on habits, either every day or on specific days of the week or month. There is nothing in this list that is particularly groundbreaking so I had a look at the various options to see if something would work to save me having to develop something. We first tried the Streaks app. This is a completely free app on iOS and even has an Apple Health integration. This suited most of my needs but it couldn't do some of the requirements on this list: The apps were in pages instead of a long list. When you have 20 things you want to track this can get pretty tedious flicking through all the pages. This app has won design awards so that is probably more of a me issue. My wife couldn't find suitable icons for some of the things she wanted to do. Hence wanting the ability to create your own. There was no way to count occurrences. You had to set how many times you wanted to do something every day. Which if you are tracking headaches isn't what you want as ideally you don't want any. She also realised she needed a way of turning off streaks. Streaks aren't always helpful for everything that you are tracking. We tried a few more apps as well. There is a popular one called HabitKit and another one called Grit. I quite liked the GitHub style contribution charts of HabitKit but my wife, not being a developer wasn't really fussed by this. Grit was probably the closet to what we were looking for but still lacked a few features. To be honest I would have been quite happy with either HabitKit or Grit for my needs if it wasn't for one thing, the price. Now I don't mind paying for applications. I have lots of apps on my phone that I have bought as well as desktop software I have paid for. I am all for supporting indie developers but I also remember a time when an iPhone app was the price of a coffee. Grit on the other hand costs £9.99/month or £29.99/year. There is an option to buy a lifetime licence but it costs a whopping £44.99. That is nearly the cost of a new console game. It is the same story with HabitKit as well but not quite as much at £1.99/month or £11.99/year with a £29.99 lifetime option. Maybe I am just a stingy git but I don't want to add even more subscriptions to my budget I have enough as it is. So I set out to build one myself with the help of AI. I started this project in March last year. Cursor were giving away free credits and it was the perfect opportunity to see what AI was capable of. Up until this point I had mostly avoided using AI beyond simple chat questions. Given I had never developed anything in Swift before I needed all the help I can get. We decided on the name HabitTed and my wife even designed a cute bear logo. AI Development Process # If you haven't developed an iOS app before you are pretty much limited to using XCode if you want to test your application on a simulator and debug it. So there was a lot of back and forth between cursor and Xcode to try and get everything set up. This was just before Claude Code was released so this wasn't full agentic hands-off engineering. Instead, I broke down the development into lots of small tasks and went through them manually with the AI, testing each change as I went along. I started with a simple hello world application. Then added in the habit list, the ability to add a habit, edit a habit and so on. To be fair after about 6 hours over the course of a weekend, I had an app that appeared to work. It didn't do everything I wanted yet but you could create a habit with a custom icon, you could click on it to complete it, and it supported various goal options and even had iCloud syncing. Even though I had something working and only took 6 hours, I can't say it was a pleasant 6 hours. Usually when I complete a project, I am proud of what I have accomplished. Even if I had problems and things took longer than needed, I would finish it having learned something new. Those failures taught me something and any frustrations I had during development were usually worth it by the end and at least I learned not to make the same mistake again. However, after 6 hours I can't say I learnt much about Swift development. Whenever there was an error I would find myself copy and pasting the error message and asking AI to try and fix things. I was getting annoyed with the AI for not being good enough to solve the issue, but also annoyed with myself that I couldn't do it either. The Problems # Even though I hadn't programmed in Swift before I have programmed in many other languages and looking at the code I could tell it wasn't good code. Each of the views was nearly a 1000 lines long. The compiler on XCode couldn't handle the length and was always telling me to break things down. Styling for things like buttons that looked the same had completely different code in different views. The whole thing was a bit of mess. So instead of relying on AI to try and fix the mess I decided to refactor it myself, manually, the old school way. I moved everything into separate views that were no more than 100 lines long and I fixed all the inconsistent code. I then tested the app some more to make sure everything worked properly and found even more problems. I found that the iCloud sync didn't actually work. It may have been syncing to iCloud but if you uninstalled the application and reinstalled it, it would just lose all your data. I also found that once I had added a lot of completions to my habits, things started to become really slow. I realised the stats I had on the details page were looping through all the habits to calculate them every single time you went on the page. I created some properties on the habits to store the stats instead, and moved the calculation to be done on habit completion, and then I realised the data model was overly complex and had some really weird naming. This led me to coming up with a data migration plan, otherwise habits would break whenever I changed the schema. Once everything was fixed, I added in some of the missing features such as reminders and different goal options. Then iOS 26 came out. I am not sure if the used AI a lot for development in Apple, but iOS 26 had a lot of bugs. I am really not a fan of the glass effect, so I turned on the reduce transparency accessibility option and this caused