Architecture, engineering, and AI, written with the same restraint I bring to the work itself. Reflections and astrology essays live under /reflections.
Why architecture work is mostly listening, and what that looks like in practice on a real codebase.
An honest sketch of when retrieval-augmented generation pulls weight, and when it's a workaround for a missing data model.
Why the cheapest engineering artefact in your codebase is often the one with the most compounding value.