Notes from two months of vibe coding: what shipped, what broke, and why I’ll never hear a “spec-driven development” pitch the same way again
This summer, I took a brief hiatus from writing The AI Economy. Instead, I focused my energy on building—software, that is. Ever since I caught the vibe-coding bug this year, I’ve been thinking about how to turn ideas I’d been carrying around for years into working software. Summer used to be a slow season for tech news, back before AI made the idea of a slow season obsolete. I knew the news wouldn’t pause for me, and I chose to step away for a couple of months regardless. With the season over, I wanted to share what I built, what I learned, and how I expect the experience to shape the way I cover AI and tech from here.
My objective was straightforward: build apps that solved my problems and served others. The software needed to have real value—I didn’t want to create just for creation’s sake. I also wanted to smartly integrate tools and services I found interesting and wanted to try. That included open-source tools and developer platforms I previously only wrote about.
The result is two launches: OnCue, a liveblogging plugin for WordPress, and Dossier, a job-tracking application.