Vibe coding

Vibe coding, an approach to software development that emerged in the 2020s in which programmers rely on artificial intelligence (AI) to create or modify computer code through natural-language prompts. Unlike conventional programming, which typically requires detailed knowledge of programming languages and software, vibe coding involves describing the intended outcome and allowing an AI assistant to determine much of the implementation.

Origins and development

The development of large language models (LLMs) in the late 2010s and their subsequent introduction to the general public enabled users to enter prompts in chatbots, such as ChatGPT. These prompts could define specific functions, and the bot would generate snippets of code to achieve the necessary output. However, large software projects typically consist of codebases (collections of interconnected files that work together). Early AI coding systems could not reliably process entire codebases and were prone to errors when creating or modifying longer snippets of code.

Beginning in mid-2024, with models such as Anthropic’s Claude 3.5 Sonnet, AI-assisted coding became considerably more capable. LLMs began to translate conversational instructions into working software, in addition to debugging programs and creating scripts to automate tasks. AI agents could now analyze much larger codebases, allowing them to generate and edit code across multiple files. This made it possible for users with limited programming experience to develop working software primarily through natural-language prompts. Computer scientist Andrej Karpathy christened this programming style “vibe coding” in a post on the social media site X in February 2025. He described it as a type of programming in which “you fully give in to the vibes…and forget that the code even exists.” The term spread rapidly through software engineering communities and social media, becoming shorthand for AI-first programming workflows.

Discover how ordinary objects work, and expand your understanding of the familiar.

SIGN UP

" height="20" alt="HEW Logo">

Thank you for signing up!"

Back to homepage

By clicking "Sign Up", you consent to the Britannica Privacy Policy and Terms of Use.

A year after coining the term, Karpathy observed that vibe coding had evolved from a niche approach used mainly for hobby projects into a more mainstream software-development practice. He argued, however, that professional software development required greater oversight and code review than hobbyist use, which involves little intervention from the prompter and is strictly what vibe coding refers to. To distinguish these workflows, he introduced the term agentic engineering, describing it as a separate discipline with its own skills, methodology, and learning curve, with engineering emphasizing the technical judgment, oversight, and expertise required to use AI agents reliably in professional software development.

Benefits and drawbacks

Advocates for vibe coding state that the practice increases programmer productivity by automating repetitive programming tasks and reducing the time needed to produce working software. It also lowers barriers to software development by allowing users without extensive programming experience to build functional applications. Additionally, AI-assisted development can shorten deployment time, enabling applications to be published online much more quickly than in traditional software-development workflows—giving rise to the term vibe deployment.

Despite its popularity, vibe coding has attracted criticism, much of it reflecting broader concerns about AI-assisted creation. AI-generated code may contain logical errors, security vulnerabilities, outdated programming practices, or fabricated references. Because LLMs generate code by predicting likely sequences of text rather than verifying that programs are logically correct, AI-generated software requires careful human review and testing. Critics have argued that programmers who accept AI-generated code without sufficient verification risk introducing subtle defects into production systems. Legal and ethical questions have also arisen concerning copyright, licensing, attribution, privacy, and the use of proprietary source code in training AI models. Many organizations adopting AI-assisted programming have developed policies governing code review, security assessment, and responsible AI use.