Blog

Notes from a workshop of one.

How I actually build: the harness around the AI, the zero-dependency habit, the tools that let one person ship like a team.

Post · Sep 16, 2026 · 10 min

It Took Me Months to Build 258 Pages by Hand. In 2026, AI Did It in Seconds.

The script would have taken an afternoon. Nobody was available to write it. Here is what that taught me about why large companies cannot automate simple things, and what AI actually changed.
Post · Sep 9, 2026 · 4 min

The Best Deal Washington Didn't Offer

Canada cannot put a tariff on the technology that runs its days. That empty shelf might be the best deal we get out of this trade war.
Post · Sep 4, 2026 · 7 min

What Happened When I Capped Claude's Context Window at 250K

One setting, 29,306 API calls before and after, and a bill that fell by half while the work stayed the same. A case study in why a long chat costs more per message, with the numbers I can stand behind and the parts I can only report as impressions.
Post · Aug 27, 2026 · 6 min

Same Song, Four Different Bills

Four Claude runs, one songwriting task, and the biggest gap in the whole benchmark came from a single model measured against itself twice. What a model-efficiency test looks like before it can name a winner.
Post · Aug 19, 2026 · 7 min

Running a Studio Session

Sometimes the enormous window is the point. How to spend that deliberately instead of leaking it.
Post · Aug 12, 2026 · 7 min

Building the Assembly Line

The default session shape in detail: folder layout, search commands, the prompts I paste, and two hooks that keep me honest.
Post · Aug 5, 2026 · 8 min

The Rent You Pay Every Turn

I audited 3 months and 16 billion tokens of my own AI sessions. The bill was not where I thought.
Post · Jul 22, 2026 · 6 min

How I Ship Thirteen Games Without Hand-Testing Any of Them

Shipping a lot alone is a tooling problem, and the tool that matters most is the one that checks your work while you sleep.
Post · Jul 15, 2026 · 6 min

Zero dependencies on purpose

Why almost everything I ship is one self-contained file, and the sentence I write at build time that keeps that from turning into a mistake.
Post · Jul 8, 2026 · 6 min

The Harness Beats the Prompt: How I Get Claude to Build Like a Reliable Co-Worker

You don't fix an unreliable AI by trusting it more. You fix it with three scripts that make its failures loud, caught, and impossible to ship.