Hello, World
The joy of pure productivity, of pure creation. It’s the reason I got into software development many (many) years ago. But as I moved away from full time coding at work (I’m a “business man” now), those skills get rusty. I still enjoy the thrill of creation, the high of pure focus, listening to my favorite music while just hammering away at the stone until the idea in my head emerges.
But as those skills atrophy those highs are harder to hit. I spend more and more time scratching my head wondering why something doesn’t work, reading the manual, reading message board posts, stumbling into dark corners.
Enter AI. Like a lot of builders out there, I’m rediscovering the joy of creation thanks to these amazing tools. And like them, I’ve re-built every personal project I’ve been working on for years. Every feature I dreamed about - my personal financial tracking system is super charged, my home monitoring system is humming - all done. I was looking for something to build next. And then I came across John Gruber’s commentary on personal media-tracking apps:
It’s a well-established cliché that no one ever finds the perfect to-do app or “task management system” unless they create it themselves. That’s certainly true for me (and resulted in my co-creating Vesper). Keeping track of things you want or need to do is too close to codifying how you think and remember things in your own mind, and we all think and remember in unique ways. We thus crave unique apps or systems to manage our tasks, ones that fit our minds just right. That’s why there are a zillion to-do apps, including a bunch that are actually good. And, these days, that’s why there are so many people creating their own personal to-do apps using AI coding systems.
And right then it hit me: I should build the media tracking app I’ve always wanted. Claude Design had just come out, so I wrote it a prompt describing what I had in my head, and it built it right before my eyes, nearly perfectly what I had in mind.
I handed that mockup to Claude and it went to work, turning out a near-perfect copy in native Swift for iOS. I upgraded my Claude plan and went to work. Dozens (hundreds?) of prompts, and hours of happily prompting away and the app I wanted started to take form.
We’re only a few years into this (sometimes scary) new world of AI. I watched The AI Doc (https://www.humanetech.com/landing/the-ai-doc) and I feel the same mix of fear and hope Daniel Roher talks about. A year ago when I tried to build tools like this the models got stuck a lot, but today they shrug off the hardest problems I can throw at them. They work in parallel, write clean, readable, documented code and barely make mistakes. I can’t imagine how they will work in another year, or 3. But today, I’ll just enjoy the pure thrill of creation and share my creation with all of you. Maybe some of you will enjoy it too.