Developers now face a new kind of “debt” as large‑language models flood code reviews. Margaret‑Anne Storey splits system health into three layers: Technical debt in code, Cognitive debt in people, and Intent debt in artifacts. Each layer limits how teams reason about change and keeps projects from staying aligned with original intent in the development pipeline today for every organization.

Storey’s framework echoes Kahneman’s two‑system theory, now extended by Shaw and Nave to include a System 3 that models AI reasoning. Their Tri‑System theory warns of cognitive surrender—passive trust in machine output—versus cognitive offloading, which delegates thought. Experiments show the theory predicts human bias in code generation, underscoring the need for deliberate verification when LLMs write production code in real‑world environments.

Practical adoption shifts focus from writing to validating. Teams replace feature‑ship questions with verification checkpoints, redefining roles so fewer engineers build code while more staff craft acceptance criteria and test harnesses. This reorganization, driven by the limits of LLMs, positions human judgment as the gatekeeper of correctness, ensuring that automated code still meets the nuanced expectations of diverse microservices systems.