This study evaluates Vibe Coding, an emerging AI‑led conversational programming paradigm that lets developers generate software through natural‑language interaction with large language models. A mixed‑methods design recruited 30 participants—professional developers and advanced computing students—to complete equivalent programming tasks under three conditions: traditional coding, AI‑assisted coding, and Vibe coding. Quantitative data were analyzed with descriptive statistics and repeated‑measures ANOVA, while qualitative data underwent thematic analysis.
The results indicate that Vibe coding reduced task completion time by roughly 27% ($$\frac{T_{traditional}-T_{Vibe}}{T_{traditional}}\times100\% = 27\%$$) compared with traditional coding and by about 12% ($$\frac{T_{AI}-T_{Vibe}}{T_{AI}}\times100\% = 12\%$$) relative to AI‑assisted coding. However, maintainability indices declined and security vulnerabilities increased, revealing a trade‑off in software quality. Usability scored SUS=71.4 and cognitive workload measured NASA‑TLX=55.5, suggesting reduced syntactic effort but heightened linguistic reasoning.
Thematic analysis surfaced four key constructs: trust calibration, loss of control, cognitive adaptation, and prompt‑engineering strategy. Notably, perceived loss of control correlated with higher security risks due to reduced transparency and validation of AI‑generated outputs. Building on these findings, the paper proposes a three‑pillar framework for responsible adoption: hybrid integration of human and AI capabilities, human oversight with transparent accountability, and context‑aware deployment. Overall, Vibe coding boosts productivity but demands critical supervision to ensure its transformative yet transitional role in software development is exercised safely.
Review