I'm building again

I never thought I’d be the person who builds software. In college I paid someone to do my C++ homework because I assumed my brain just wasn’t built for this. For most of my life, the most advanced stuff I did felt like MySpace-level coding. Changing colors, hiding a music player, adding a cursor glitter trail. That was my ceiling.

Now I’m sitting here on my phone, vibe coding apps with AI and connecting APIs like it’s normal. What’s strange is how different this feels from how I normally work. I usually get stuck in analysis paralysis and overthink every detail. With vibe coding I don’t hit the brakes. I just move. It feels almost feral. No bottlenecks. No overthinking. Just forward motion.

Andre Hegge reignited my curiosity almost a year ago when he showed me what DeepSeek could do with property data. That was the spark. I did not realize it right away, but that moment pushed me back into the idea of building my own tools instead of waiting for someone else to get around to them.

One of the first things I revisited after that was AI Personas. I opened it again, talked to AI like it was Jarvis, and suddenly I was generating code, connecting APIs, and everything was actually working. Not a demo. A real tool. And the wild part is how it tied back to something I had written about before. I had posted a breakdown of a long running mystery in real estate data on my Substack in my piece titled A Real Data Case Study. That problem lived in my head for almost ten years. AI Personas is the first time I saw the entire pattern laid out in one place without jumping between tools. Finally the puzzle made sense.

That was the moment I realized I was not experimenting anymore. I was building real products.

That momentum carried me directly into SodaCloud, which became its own rabbit hole. It started because SoundCloud annoys me and I wanted something that behaved the way I needed it to. I did not write a plan. I did not map anything out. I opened a blank project and went feral with it. In a couple days I had uploads, timeline comments, waveform previews, mobile layouts, everything. Then I got obsessed with waveform accuracy and spent hours tightening scrubbing behavior and visual fidelity. I swapped themes, fixed seeking, fought with storage, battled CORS until it finally listened, and kept going. Eventually it turned into a full platform. The kind of thing I used to assume only big teams could build.

It still blows my mind that I made that.

I am not trying to act like a developer. I am not pretending. I am just building. And it is actually working.

I am grateful to the people who nudged me in the right direction.
Michael Bergin, Tony McGibbon, Brian Warren, and the Realforce crew who let me build internal tools which are classified. And shoutout to YouTubers Alex Finn and Corbin for explaining things in a way my brain could follow. Their videos helped me move from prototype mode to production.

I am still figuring out what this version of me looks like. But this is the start.