Building with AI
Field notes from running a production system on LLMs every day — what actually cut a four-minute boot, why the optimization you wrote isn't the one that runs, and where today's tooling debates have played out before.
-
A knowledge base is a graph, not a folder
An agent knew the manual's minimum radius cold and still got the curve wrong, because the number was only the start of the question. Why a domain agent reasons better from a graph of criteria than from a persona prompt, and what that structure caught in real road-geometry and brand reviews.
-
The optimization you wrote isn't the one that runs
I had the optimization written and approved, and a week later it still wasn't what ran. The fix wasn't a better design. It was making the trigger deterministic and leaving the judgment to the model.
-
Anatomy of a four-minute boot
Where my Claude Code sessions actually spent their first four minutes, and what cut it to thirty seconds without losing any context. Hooks were three seconds of the problem. The model's diligence was the rest.
-
Calibrate against your own voice
AI detectors flag non-native English as machine-written. callus scores your draft against your own voice instead. pip install callus.
-
The Pixar precedent: vibe coding has been here before
Debates over disruptive tools follow a pattern. The vibe coding fight is at a recognisable stage of it. Notes from someone who passed through an earlier version of the same argument.