I used to love discovering new apps. Testing them, trying to customise them to fit my workflows and my needs. But every app ends up having constraints you need to adapt to, limitations, or choices you would have implemented differently. It's always good enough, sometimes the best you can find out there, but never perfect.
That changed this year. With AI, I can now materialise exactly the app I want, just by speaking, and keep refining it by talking to it while I use it, with the updates ready minutes later. It feels like science fiction, at least in the digital world. The era of personal software is here.
The weekend question
It starts with a question I now catch myself asking on weekends: I have 30% of my weekly AI credits left, what can I do with them? It's a very freeing question. I have an army of superintelligent agents in any field. What do I want them to do?
The primary goal of all this was my job: improving my own productivity and my processes, building tools and automations to do my work better and quicker. But it has bled into my personal life. I now build my own native macOS applications, iPhone apps for my daughter, passion projects for relatives. Mostly by talking out loud to my computer for a few minutes, and launching my agents at it. A couple of weeks ago I bought an Apple developer account, and that's when the replacements started.
A music app in half an hour
I listen to music most of the day. When I'm not on a call, music is playing at my desk. For years I used Audirvana, a premium hi-fi app I pay a yearly subscription for, and Apple Music next to it. Both had quirks in the UI and UX, features I wanted and didn't get, and neither adapted to the small screen I keep my player on, which made for a clunky experience.
So I said: let's build a new music app. I dictated a 5-minute voice note about what I liked in both apps, what I disliked, and the features I wanted. My agent already had the context: my setup, my configuration, how I listen to music, the other apps I use. About half an hour later, on the first shot, I had a working app, Hifee. To me it looks better than both Audirvana and Apple Music, it has only the features I need without the bloat, and it uses 5 times less memory. That last point matters more than it used to, because memory is what my AI workflows need on the Mac. Somewhat ironic.
Use it, poke it, move on
What happens next is the part I find most interesting. Like every app I've built in the last few weeks, I just started using it. When I find something slightly different I want, a decision to reverse, a new feature or a design change, I switch to that app's agent, poke it, let it implement the change, and move on to something else while it builds. A few minutes later, the update is there.
An agent can take a few minutes or a couple of hours on a task, working in the background. So you multitask and switch between agents, and when everything is either running or done, you start thinking: what else can I do?
Notee: leaving VS Code
For years VS Code was my IDE, and I also used it for my notes, which I moved to Markdown about 4 years ago, ahead of this AI wave. The beauty of it is muscle memory: the shortcuts to navigate files, folders and text, the snippets. But VS Code feels like a coding environment, because it is one. It's not a note-taking app. I know Obsidian well and recommend it to non-technical people, but I've never used it properly myself: I don't like its UX and UI, and its keyboard shortcuts and logic don't fit my workflows.
This weekend I had 50% of my weekly credits left, ironically because I'd spent 2 days out of the office at an AI conference. So I thought: why not replace VS Code with my own Markdown notes app? And so I did. It's called Notee. Like the music app, it was one-shotted from a voice dictation, then tweaked and improved over the weekend as I used it, also by talking to my computer. For me, it's much better than VS Code.
Filee and Demoee
Path Finder has been my file manager on the Mac for years. I tried a lot of file managers, and it was the best for my needs, but not perfect. Same process: one voice prompt of a few minutes, and the first version was already a better file manager than Path Finder, for my needs. Then I started using it and improving it. Some features were in Path Finder and I simply hadn't mentioned them in my brief. Others came from a new question: now that I can build anything easily, what would the ideal workflow be, and what can I automate? That's Filee, my new file manager.
On Monday I have to deliver a demo video to a client. I do these regularly: set up a demo of a Kaltura solution, record myself, usually in several takes, then edit. I did that in ScreenFlow for years, and it's a time-consuming process. So I built Demoee. It takes only what I need from ScreenFlow and makes it better: an app dedicated to recording my 2 screens with me in front of them, doing the chroma keying better than ScreenFlow could, and with the intelligence to do the editing, including B-roll in the form of slides. The first version was one shot, and I'm improving it now. Recording a demo video will be simpler and faster, so I can do more of them.
What it costs
None of this is free. The Apple developer account costs about $100 a year, and I pay for the intelligence. Part of it is offset by the apps I no longer pay for. More importantly, I'm investing in everything I get out of it, and the list keeps growing: my own email app, calendar app and sticky notes app, and now music, notes, files and demo recording.
Good enough is no longer the ceiling
For most of my life with computers, choosing software meant choosing which compromise to live with. I would find the best app out there and adapt to it. Now the app adapts to me, and keeps adapting while I use it. That's what I mean by personal software: an app built for an audience of one, by speaking, and refined by talking to it. It still feels like science fiction to me. It's also how I work now.