Writing
Studies
Short pieces from the work. What I learned, how I failed, how I got better, and what I’d do next time. AI-assisted, and mine.
Want to see the real work behind these? My client portfolio is password-protected. Here’s what’s inside, and how to get access.
-
How I configure, automate and measure AI work
One page that points to the rest.
-
Automation that runs while the lid is closed
Routines, a catch-up job, and a packager that refuses to ship.
-
My setup didn’t make the model better
I tested my own safety layer by taking it away. Here’s what it’s actually for.
-
Design for the engineer who wasn’t in the room
How I work with engineering when I don’t own delivery.
-
Turn the argument into a number
How I make product decisions when the room disagrees, and how I keep them made.
-
What’s inside the client work, and how to see it
Three case studies open now, two more coming: what each shows, what I owned, and how to get access.
-
Skills, connections, and the checkers between them
How I configure the parts of an AI setup that aren’t agents.
-
None of my agents can edit a file
Four checkers, each one tied to the moment I’m most likely to be wrong.
-
I configured three agents. They sat idle for two months.
The rule that benched them was mine.
-
How I work with AI
Most of what gets built on my desk now is AI-assisted. Some of it is code I can read but couldn’t write. The way I see it, that…
-
The five rungs
Where AI entered one feature, and what each step cost.
-
Directing code I couldn’t write
I ship AI-written code into a repo that senior engineers review. I can read it — I couldn’t have written it.