Most of what I publish starts with something I did. An automation I built and then broke. A process I redesigned. A pattern I noticed while working.
The ideas, the context, and the opinions come from that work. So do the mistakes.
I start with notes, screenshots, thought dumps and rough ideas. AI helps me structure them into a draft. I use Claude, Gemini, and Claude Code.
It's good at structure. It can't know what I experienced, so that part stays with me.
I do the work and figure out the messy parts. AI turns my notes into a first draft. I rewrite it and cut what I wouldn't say. Then I publish.
A post has to sound like me and match what I built or tested.
“If I didn't do it, I don't write about it as if I did.”
Google's guidance on AI content says to review and fact-check before it goes live. I already work that way.
I don't prompt a model to write a post from scratch. Photos on this site are mine, and no imagery is AI-generated.
Things change fast, and I'd rather share something practical now than polish every sentence by hand and publish it weeks late.
It would also be odd to work in AI and automation, and refuse to use the tools myself.
If you spot a mistake, tell me and I'll fix it.
Last updated October 2026