Learn Something New Daily
I’ve been learning Swift, the Xcode enviornment, AI tools like Cursor and ChatGPT. The learning cruve is not too bad. I’m trying to build a habit-tracking app from scratch. But like anything worth doing, it got complicated. I feels good to learn about coding, but also the architecture and thought process that goes into makign scalable, reusable, and robust code. Of course, the frustrations started to mount.
Some of the easeir things to learn right away:
- Learning how to prompt AI with more specific requests
- Learning the differences between an Agent and an Ask request
- Understanding tokens, which models to use, and using context windows
- Memorizing Xcode shortcuts, cleaning up my work-flow
- Switching between projects that are vastly different (qmd vs XCode project)
- Multi-tasking both XCode being open and Cursor (to see the preview and build easier)
Instead of vague asks like “make this better,” try being specific: “Refactor this SwiftUI view using MVVM and preserve the binding logic.” Of course, you should get even more specific than that, but you get the idea.
An AI Agent is like giving complete control of your code and IDE to the AI, so it’s risky, but can speed up things (no more cut and paste). However, I think this is what gets me in trouble the most. The more I want to get the app done, the more I was trying to rely on the agent to do the work. In hindsight, I think having patience is the key. When you ask it just responds without taking control. This forces you to actually read the response before deciding to apply the logic.
One step at a time.
I Mean, I Knew Not to Trust the AI Too Much
Learning how to have a clean build, archive, setting up TestFlight and uploading your project to an actual Apple server to distribute to your friends is exiting. That being said, it’s easy to fall into that trap of trusting the AI too much because you’re in a hurry to launch and show your friends. You’re like, “Cool, I’ll just let the AI write that part. I’ll fix it later.”
I started going back through the code. I slowed down my pace a bit. I wanted my computer science brain back.
Back to the Basics
Algorithms, data structures, OOP concepts. Some of it I remembered, some I had to relearn. And AI would remind me, “Yeah, that’s called refactoring.” Or, “That’s a struct in Swift, not a class—and here’s why.”
I started asking smarter questions:
- Can I persist this data or not?
- What’s the best way to store this locally and scale later?
- Why is this UI state correct sometimes, but not always?
Another frustration was updating the logic and design without remembering to update the preview versions.
And when I didn’t know how to ask my IDE agent directly, I’d ask ChatGPT how to ask it, and it worked. I’d ping one AI to help me talk to another.
The thing is, if you don’t dive in and make these mistakes, you will never get better. You can’t just read a Reddit post about prompting better, you have to actually feel the pain yourself and starting prompting to fit your exact needs.
Are APPS For Me?
I’m not doing this to become a UI guru. I’m not that interested in making flashy little swipe widgets that bounce when you gesture at it. I want to get back to more mathematical, deeper work—stuff that requires actual thinking and research. Code that solves real problems, not just interfaces that look good in screenshots. No offense to those that do.
This app project is my warm-up I guess. A hobby to get my muscle memory back. Syntax, logic, system design. I want that stuff locked in, because once it is, programming becomes more fun and you can start tackling the hard problems.
The Coming Flood of Mediocre Apps
AI is going to supercharge a wave of bad apps. Millions of people will start creating them. Everyone will be shipping MVPs that don’t scale, don’t persist data, and definitely break and have bugs. AI will get them 60, 70, 80, or even 90% there, but they’ll hit the wall when something goes wrong.
That’s where people like me come in. Learning how to think through problems, not just generate code vibe coding with my talking mouth. Don’t get me wrong, I’ll never compete directly with the hard-core coders knocking out C++ projects in their spare time at age 12 through 40, but I have other skills to help me build a career (potentially) out of this.
I don’t think there will be fewer skilled programmers—maybe even more. But the signal-to-noise ratio is going to get rough. AI gives a lot of people a head start, but it can’t replace fundamentals. And fundamentals are what I’m getting back to.
I’m only a few weeks in, and I feel like I have some momentum.