AI coding tools have reached near-universal adoption among software engineers, and the speed of that shift is the problem. Gergely Orosz, writing from Craft Conference in Budapest alongside Kent Beck, Hillel Wayne, and Titus Winters, argues that the instinct to move faster with AI agents is exactly what causes teams to lose ground.
The core argument is not about slowing adoption. It is about deliberate pacing at the task level: understanding what an AI agent is doing before accepting it, not after. The piece draws on real patterns Orosz has observed in engineering teams, making it more diagnostic than prescriptive.
The full piece is worth reading for the specific behaviors Orosz flags as warning signs, not just the headline conclusion. The Craft Conference context also matters: these ideas are being stress-tested in front of an audience that includes some of the most rigorous thinkers in software engineering.
[READ ORIGINAL →]