On cognitive debt

"I suspect, however, that: When you control for the size and scope of the project, cognitive debt tends to be at least as bad, and usually worse, in pre-AI codebases than in AI codebases; This fact is obscured because so much of what is normalized as traditional engineering work is in fact either managing crippling cognitive debt or avoiding it at enormous cost; The best users of AI are already pretty good at avoiding cognitive debt, and we're only going to get better at it."

"If you're moving 10 or 100 times as fast as before, this wouldn't be surprising, even if AI weren't affecting your comprehension per unit work. Imagine someone in the before-times saying: "I've been working hard on a project for a year and half, and I often struggle to remember exactly what I need to do in order to extend it.""

Cognitive debt denotes codebases becoming unknowable, inextensible, unobservable, and hard to debug. When controlled for project size and scope, cognitive debt is often at least as severe, and usually worse, in pre-AI codebases than in AI-driven ones. Much traditional engineering work involves either managing crippling cognitive debt or avoiding it at enormous cost. Encapsulation remains difficult, and difficulties with encapsulation persist when using AI tools. Rapid progress with AI can create a perception of increased cognitive debt due to compressed timelines rather than increased debt per unit work. Skilled AI users already employ practices that mitigate cognitive debt and those practices will improve.

Read at Nate Meyvis