Vibe Coding is Iterative, Not One-Shot: New Review of 47 Sources

A new multivocal literature review by Shahbaz Siddeeq and colleagues from the University of Jyväskylä systematically analyzes 47 sources (28 peer-reviewed, 19 grey literature) on vibe coding—a software development practice where developers state intent in natural language and large language models generate code. The review, spanning 2022 to October 2025, reveals that vibe coding is consistently described as an iterative loop of generation, evaluation, and revision, rather than a one-shot activity. This shifts the developer's role from writing code toward specification, supervision, and validation of AI-generated outputs.

Short-term productivity and time-to-prototype gains are reported in 21 of 47 sources (45%), while evidence on maintainability, long-term quality, and safeguard effectiveness remains limited. The review finds strongest evidence for prototyping and user-interface work, and weakest for production, data-intensive, and safety-critical use cases. The authors note that tool visibility does not imply effectiveness. This is one of the first reviews to integrate both academic and practitioner evidence under a documented protocol, and calls for future work on safeguard evaluation, session-level dynamics, and long-term maintainability testing.