The rapid adoption of AI-powered coding assistants since 2022 has produced a striking empirical contradiction: controlled studies report individual-level productivity gains of 20-56% on well-scoped tasks, while the most rigorous randomized controlled trial to date documents a 19% slowdown for experienced developers on mature codebases. This paper argues that these findings constitute a systematic phenomenon—the Productivity-Reliability Paradox (PRP). Through a multivocal systematic literature review encompassing 67 sources (2022-2026), the paper makes four contributions: (1) it formally defines the PRP with three moderating variables; (2) it proposes the AI-Augmented Methodology Taxonomy (AAMT); (3) it introduces the Specification Governance Model (SGM) grounded in Transaction Cost Economics; and (4) it evaluates GitHub Spec Kit and TDAD as SGM instantiations, supported by an exploratory field evaluation across three development teams.
Sabry E. Farrag (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: