As AI and agents are adopted to accelerate development, cognitive load and cognitive debt are likely to become bigger threats to developers than technical debt
Margaret-Anne Storey
Context & Ripple Effects
Earlier coverage identified a trade-off in AI-assisted code creation: lowering the barrier to producing code can also create more complexity, technical debt and confusion. This article shifts the focus from codebase liabilities to developers' capacity to understand, evaluate and coordinate work.
Developers adopting AI agents face a more immediate burden of reviewing outputs, maintaining context and deciding when to trust automated work, even as routine coding accelerates.
Engineering managers must treat cognitive load and accumulated gaps in system understanding as delivery risks alongside conventional technical debt.
Second-order effects
Productivity claims based chiefly on code output become less reliable if teams spend more time validating agent-produced work or recovering shared context.
Tool vendors and engineering organizations will be pushed to emphasize workflows that preserve human oversight and explainability, rather than only maximizing automated code generation.
Third-order effects
If cognitive debt compounds across AI-assisted teams, software organizations may shift quality governance from managing code artifacts alone to managing the human understanding required to safely change them.
This points to a constraint on agent economics: automation's value will depend on whether it reduces total coordination and review effort, not merely whether it increases production volume.
The trend: AI-assisted development is moving from a code-generation race toward a test of whether organizations can scale automation without eroding developer understanding.
“Cognitive debt is likely a much bigger threat than technical debt....” https://margaretstorey.com/... “As generative and agentic AI reshape how software is built, understanding and managing cognitive debt may be one of the most important challenges our field faces.” [image]
This whole thing is a good read. I've often referred to “cognitive overhead” when working in engineering organizations pre AI to explain why velocity can grind to a halt, but “cognitive debt” resonates more when it is being deliberately created by AI as opposed to developing ove…
A complication of AI-assisted coding: developers (even if they're being careful, checking everything, etc.) are moving too fast to develop mental maps of their projects and remember their design and architectural choices, making it harder to explain those choices or make changes …
Cognitive Debt. — “Even if AI agents produce code that could be easy to understand, the humans involved may have simply lost the plot and may not understand what the program is supposed to do, how their intentions were implemented, or how to possibly change it.” — margaretsto…
Short musings on “cognitive debt” - I'm seeing this in my own work, where excessive unreviewed AI-generated code leads me to lose a firm mental model of what I've built, which then makes it harder to confidently make future decisions simonwillison.net/2026/Feb/15/ ...
Let that sink in. Even one of the loud and vocal AI proponents such as @simon admits — “I no longer have a firm mental model of what they can do and how they work, which means each additional feature becomes harder to reason about, eventually leading me to lose the ability to …
“managing LLM agents has a lot of similarities to managing junior developers” — No, they are different: LLM agents don't make coffee but on the other hand you can shout at them without being attacked as a toxic manager. — [Otherwise, interesting article.] — https://martinfo…