Friday Coffee

A Friday Coffee reflection on fashion-driven IT, technology hype, and the need for more scepticism in architecture decisions.

A personal reflection on keeping up with fast-moving AI tools without confusing constant novelty with durable understanding.

A personal reflection on how AI agents have made ambitious software work feel possible again after decades in technology.

A Friday Coffee reflection on building a VHF antenna, old technical books, patient learning, and the pleasures of finding things out slowly.

AI coding tools change the economics of software work, but they make judgement, craft, testing and responsibility more important rather than less.

AI makes one skilled person extraordinarily productive. But a business needs more kinds of judgement than any one person holds, and that is the constraint the hype quietly ignores.

AI can do the beginner work. That is exactly the problem, because the beginner work is how beginners became experts. Construction already ran this experiment, and the bill arrived decades later.

A wonky shelf looks wonky. But the most celebrated engineers have built magnificent structures with fatal flaws nobody could see. Software has always had that problem, and unsupervised AI takes it somewhere genuinely unknown.

My father already knew his trade in his hands. Becoming a lecturer meant serving a second apprenticeship on top of it — and what he taught me in a darkened living room was never really about roof trusses.

A Rubik's Cube, Mahler's Ninth and a slow data warehouse all point to the same skill: building internal models that connect representation, reality and judgement.