$ ls blog/
How I build software with AI agents, for clients and for my own products. What works, what broke, and how I check the result.
I haven't written code by hand in over a year
How I build client software, my own product and a SaaS side business with Claude Code: the loop I run every day, real numbers, and what went wrong.
The agent says it's done. Here's how I check
Coding agents report success in the same confident tone whether they succeeded or not. The checks I run before I trust a result, and the failures behind each one.
A ticket goes in, a reviewed merge comes out
The agent on my home server that takes client tickets all the way to a reviewed merge on staging: the pipeline, the gates, and the bugs in the pipeline itself.
Writing for an agent that starts from zero
Every agent session starts without yesterday. How I write the project docs and tickets an agent reads, and why every mistake ends up as a written rule.