EraCode glossary
Cognitive debt
Definition
Cognitive debt is the accumulated mental cost of skipping the act of understanding code in favor of accepting generated output. Like technical debt, it compounds quietly and is paid during incidents, reviews, and maintenance.Also called: AI cognitive debt, understanding debt
Cognitive debt is not about whether generated code works today. It is about whether the team still understands enough to change, explain, and repair it tomorrow.
What it looks like in practice
Cognitive debt appears when a change passes tests but no one can confidently explain the implementation. During review, the author may know what prompt produced the patch but not the invariants the patch relies on.
The debt is paid later: incidents take longer, refactors feel riskier, and teams become hesitant around modules that are technically new but mentally unowned.
What causes it
Simon Willison describes cognitive debt as one of the costs of leaning on AI without doing the comprehension work. The generated code may be useful, but the missing mental model remains.
The pattern compounds when teams optimize only for merged output and do not preserve time for reading, debugging, and explaining the code that shipped.
How to address it
Reduce cognitive debt by making understanding visible: ask authors to explain high-risk generated changes, read tests as documentation, and keep practice loops that require reasoning without an assistant doing every first draft.
EraCode can help by creating short reps around debugging, code reading, terminal work, and multi-part reasoning instead of only puzzle-style coding.