Cognitive debt

Overview

Cognitive debt is a product and engineering concept that describes the hidden cost of AI-accelerated development when teams lose shared understanding of how a system works. Instead of debt showing up primarily as messy code, brittle architecture, or deferred cleanup, it accumulates in the gaps between what the system does and what the team can confidently explain, debug, and evolve. In AI-assisted environments, teams can ship faster with agent-written code and generated implementations, but that speed can create black-box systems that are difficult to reason about later.

For AI Product Managers, cognitive debt matters because delivery velocity is only valuable if teams can sustain and extend what they ship. When developers, PMs, and stakeholders no longer share a clear mental model of the product’s behavior, decision-making slows, debugging gets harder, and roadmap risk increases. Managing cognitive debt means treating system comprehension, documentation, explainability, and knowledge transfer as product delivery concerns—not just engineering hygiene.

Key Developments

• 2026-02-15 — Simon Willison highlighted Margaret-Anne Storey’s explanation of cognitive debt as a shift in accumulated cost from traditional technical debt toward the team’s shared understanding of AI-built systems.
• 2026-03-01 — In discussion of interactive explanations, the concept was extended to agent-written code becoming a black box, with the argument that interactive walkthroughs can reduce long-term friction and preserve understanding.

Relevance to AI PMs

• Balance speed with maintainability of understanding. AI PMs should evaluate not just how quickly teams can ship with coding agents, but whether the team can still explain core workflows, edge cases, dependencies, and failure modes a month later.
• Add comprehension to delivery criteria. For features built with AI assistance, require artifacts such as architecture notes, interactive walkthroughs, or implementation summaries so knowledge is not trapped inside generated code.
• Track risk in roadmap planning. Cognitive debt creates future drag on iteration, onboarding, incident response, and cross-functional alignment. PMs can reduce this by scheduling time for system explanation, refactoring of understanding, and explicit ownership transfer.

Related

• agent-written-code — A major driver of cognitive debt when generated implementations are accepted faster than teams can understand or review them.
• interactive-explanations — A proposed mitigation: walkthroughs and explainable interfaces that help teams understand how AI-generated systems work.
• simon-willison — Helped popularize the framing through commentary on how generative and agentic AI shift concerns from technical debt to cognitive debt.
• margaret-anne-storey — Credited with the clear articulation of the concept as debt accumulating in team cognition and system comprehension rather than only in code.