Cognitive Debt

"The space between those two lines is cognitive debt. There are two lines worth paying attention to right now. The first is code volume. We're producing more code than ever. Code you'd normally spend time designing for a bit, work out its interface, what it needs to do, figure out the right responsibilities, maybe write a test first, now turns up fully formed from a prompt."

"Cognitive debt is different. It's code you don't even know is bad, because nobody understands it well enough to judge. You didn't choose to take it on. It accumulated silently, one accepted suggestion at a time. Technical debt lives in the codebase. Cognitive debt lives in your team's heads, in the gaps between what the codebase does and what anyone can actually explain."

"When you write code by hand, you build understanding as a side effect. You wrestle with the problem, try things that don't work, and arrive at a solution you can explain because you lived through finding it. When AI writes code for you, the code appears fully formed. It compiles. It passes tests. You read it, it looks reasonable, you merge it. But the understanding was never constructed."

Cognitive debt emerges as development velocity increases through AI-assisted coding, creating a divergence between code production and team understanding. Unlike technical debt—which involves known shortcuts—cognitive debt consists of code nobody fully comprehends because understanding was never constructed during development. When developers write code manually, they build expertise through problem-solving struggle. AI-generated code arrives fully formed, compiles, and passes tests, but skips the learning process that creates mastery. This silent accumulation of incomprehensible code represents a distinct problem from traditional technical debt, residing in knowledge gaps rather than in the codebase itself.

Read at Andrew Cairns