What we learned building things, written up while it was still fresh. No thought leadership, no listicles.
234 articles · page 38 of 39
06 Jan 2026
Automated suites confirm what you already knew. The value of a human tester went up, not down, the moment code started arriving in bulk.
Quality · 6 min
05 Jan 2026
Four questions we ask about code a model wrote, in the order that catches the most problems soonest.
AI · 6 min
02 Jan 2026
The difference between a spec that a model implements correctly and one it guesses at is mostly structure, not length.
01 Jan 2026
A short account of the three habits we kept from last year, the two we dropped, and the one we are still arguing about internally.
Process · 5 min
18 Nov 2025
Most LLM features we are asked for are a chat box. Almost none of them should be. Here is the test we apply before writing any of it.
AI · 7 min
07 Oct 2025
Time and materials is easier for the studio and worse for the client. What a week of paid discovery actually buys.
Process · 6 min
