I Want to Use Loops. I Can't Yet.
I keep wanting to build with agentic loops. Every time I do the math on real work, I close the tab — and I'm still trying to figure out what I'm missing.
I keep wanting to build with agentic loops. Every time I do the math on real work, I close the tab — and I'm still trying to figure out what I'm missing.
Generation is becoming a commodity. Trust isn't. Here's what I noticed in the trend and what I changed in mx-workflow because of it.
I keep spending hours building real things from my phone — Claude Code running on my Mac mini, Tailscale tunneling me in. The laptop barely shows up. I'm starting to wonder if this is just the shape…
Three things have to hold before I let the agent rebase end-to-end and push without reading the diff: solo code, code I know, and real QA plus automation behind me. Pull any one out and the workflow…
On consecutive days, one of the most craft-obsessed programmers alive argued for the craft and predicted the market would stop rewarding it. I think I land somewhere in between.
AI is fast. Humans are slow. The joke writes itself. But the question underneath it — where do I fit in this? — keeps following me around.