Previously, I wrote about making a MacOS Menu Bar app using Gemini. In that post, I left a lot of loose ends left by the generated code that I never tidied up. With a bit more time over the last few weeks, well, I tidied them up.
First, I was using deprecated APIs. So I went back and updated those. I went down the tried and true path, starting with the Apple Developer Docs and then combing through various software forums to find examples on how to do these things. All culminating in an aptly named commit titled “refactor AI generated code”.
Next, I wanted a prettier UI. During my initial search for an app that did
what I created, I found “Default Browser” created by Sindre Sorhus
but damn it if that
button didn’t discourage me from taking a closer look. But Sindre’s app has
icons and I wanted icons. This ended up being a fun conversation with Gemini
where it gave me examples using NSworkspace to get an icon from a URL handler,
which I then implemented
as well as setting the main menu bar icon so I could tell, at a glance, what the
default was.
Around this time, Gemini 2.5 Pro was released and I wanted to retry the original prompt
with the new model. My own little way to benchmark these LLMs, you might say.
Gemini 2.5 Pro still didn’t account for changing the default browser from a different place in the OS
but it ended up implementing the UI using MenuBarExtra.
That was new to me, all the previous responses I got had used NSMenu and NSMenuItem,
but that’s cool, that’s something new to learn.
I tried MenuBarExtra out a bit, learning a bit of SwiftUI and a bit more of XCode.
I’m liking MacOS development more now. There’s not that many examples or tutorials so I ask Gemini a lot of questions to learn syntax and see examples. I’ve found that using AI as a partner for learning feels like actual development instead of throwing darts blindly. I first learned to program watching YouTube tutorials and reading example blog posts, not by reading documentation. Now, as a seasoned programmer, I’m more comfortable reading the manual but some learning styles, like seeing examples, just work better for me. I need a little bit of inspiration. And as always, you don’t know what you don’t know, and sometimes the AI chatbot can tell you something new.
Here’s a link to the project’s GitHub.