Kits, prompts, and frameworks I paste in my own work. Operator stack notes, weekly.
Sep 28, 2026
•
2 min read
The first one is the easy call. The second is where most setups quietly fall apart.
Sep 21, 2026
1 min read
Ninety minutes in Sandy, and you walk out with one thing running. Last time I bring it up.
Sep 14, 2026
It gave you something generic and slightly wrong, so you closed the tab. That was a setup problem, and it is fixable in an afternoon.
Sep 8, 2026
3 min read
Most owners never ask for this, because nobody has ever described it plainly. So here is the shape, the price, and the exit.
Sep 1, 2026
Pick one. Put a clock on it. That number is the whole argument, and it is the only one that counts.
Aug 25, 2026
Most owners point AI at the work they can see. The expensive leak is the quote, the reply, and the promise that went cold while you were busy.
Aug 12, 2026
I shipped a paid product this week. Almost nobody saw it. Here is the part of the job I keep underrating.
Aug 3, 2026
The math on replacing people with AI almost never survives reality. Here is what does.
Jul 27, 2026
The automation that ran 880 times and produced nothing, and the four-question test that would have caught it.
Jul 20, 2026
The bottleneck was never the tool. It is the seam between two people, and AI makes it louder.
Jul 15, 2026
Why I booked a room, invited strangers, and told them to watch me use AI live.
Jun 16, 2026
It was never the tool. It was adoption.
Jun 11, 2026
The model is fine. The map is wrong. Here is how to tell which one you have.
May 20, 2026
humanatwork.ai is live, multiple SKUs on the shelf, built from work that already shipped.
Apr 28, 2026
The most expensive mistake in AI adoption is not the wrong tool. It's skipping the diagnosis.
Constraints
Apr 21, 2026
