AI and agents
What AI and coding agents change about building software, and what they don't.
-
November 8, 2026
Why Learn to Code When Agents Write It?
At a panel on agentic engineering, I answered a founder's hard question with “learn to code.” It sounded flippant. Here's what I meant: coding to produce code was never the point, but coding as a practice is how you learn to think in systems, and how you learn to direct agents instead of hoping.
-
November 7, 2026
Wisdom Can Be Learned, Not Taught
Everything a book can hold is declarative knowledge, a distillation of what someone learned by doing. Engineering judgment is procedural: earned in the field, over years. Why that matters now that AI widens the gap between a new hire and a seasoned pro.
-
November 6, 2026
Effective Before Efficient
Every product lives through three stages: Signal, Scale and Sustain. In the first, nobody cares how you made the burger, only whether it tastes great. Why effectiveness comes before efficiency, why AI agents don't change that, and why each of us is built for one stage.
-
October 30, 2026
You Can’t Direct What You Can’t Interrogate
Vibe-coded throwaways are fine for getting signal. The failure is the throwaway that never gets thrown away, and the judgment needed to tell a demo from a system.
-
March 8, 2026
Code Was Never the Unit of Value
Code was never the unit of value. Lean’s definition of value, applied to software, shows what AI is actually compressing, and what engineers are left to do.
