The Silent Erosion: Vibe Coding and the Decline of Programming Literacy in the Age of Generative AI

Abstract

The rapid integration of generative artificial intelligence (GenAI) into software development has given rise to a new programming practice commonly described as “Vibe Coding,” in which developers express high-level intentions in natural language and increasingly rely on AI systems to generate, modify, debug, and refine code. While Vibe Coding can reduce the effort required to produce software and accelerate routine development tasks, its implications for programming literacy and skill formation remain an emerging concern. This review synthesizes recent empirical evidence on AI-assisted programming and examines how extensive reliance on code-generating systems may influence productivity, comprehension, problem-solving, and cognitive engagement. A meta-analysis by Maier et al. (2026), covering 23 quantitative studies and 27 effect sizes published between 2019 and 2025, found a statistically significant, moderate positive effect of GenAI assistance on programmer productivity (Hedges' g = 0.33, 95% CI [0.09, 0.58]) but no statistically significant overall effect on learning outcomes (g = 0.14, 95% CI [-0.18, 0.47]). In addition, a randomized controlled trial by Shen and Tamkin (2026) found that developers using AI assistance scored 17 percentage points lower on a comprehension assessment than developers who coded without AI assistance. These findings suggest that the consequences of Vibe Coding depend not only on whether AI is used, but on how it is used: AI-supported explanation, exploration, and verification may preserve opportunities for learning, whereas extensive delegation may reduce independent cognitive engagement. The review therefore examines cognitive offloading as a plausible mechanism and discusses pedagogical strategies for integrating Vibe Coding into programming practice without undermining core programming competencies.