Notes on thinking, debugging, and building software with an AI duck in the loop.
I went in with low expectations. Over the past couple of years I’ve tried a steady stream of AI tools. Some were clever. A few were useful. Almost none changed how I actually work on a normal day. Whispr Flow did. The bottleneck was never…
Continue reading...When you first start wiring AI agents into your repo, the temptation is predictable. You have a single instruction file that boots with the agent, so you begin putting everything there. CLAUDE.md. AGENTS.md. Whatever gets loaded at startup becomes the obvious home for conventions, notes,…
Continue reading...I didn’t set out to build “multi-agent infrastructure.” I just wanted Claude and Codex to coexist without stepping on each other. That turned into something more deliberate. Two agents. One repository. Zero ambiguity about who is responsible for what. If you let both read and…
Continue reading...Doing a bit of pair coding today this week with Codex vs. Claude. I have to say, Codex was really easy to set up, comparatively, but.... that's because Claude already helped me fix a bunch of shit in my PATH and .zshrc and stuff like…
Continue reading...It’s funny. I’ve been watching the gen AI space for a couple of years now. I’ve dabbled with a dozen apps, built a few integrations using APIs for work, and experimented with various tools to try to improve my coding or project management workflows. I…
Continue reading...