Vibe Coding
Noun · Verb · AI & Machine Learning · Origin: 2025
Definitions
Vibe Coding is a software development approach where the programmer describes what they want in natural language and lets an AI assistant generate the actual code, iterating through conversation rather than writing syntax directly. Coined by Andrej Karpathy in early 2025, the term captures the shift from precise manual coding to a more intuitive, conversational workflow where the developer focuses on intent and direction while the AI handles implementation details. Practitioners may accept code they do not fully understand, trusting the AI output and testing through execution rather than line-by-line review. Proponents see it as democratizing software creation and dramatically accelerating prototyping. Critics worry about code quality, security vulnerabilities, and developers losing fundamental skills. The approach works best for prototypes, personal projects, and well-understood domains where the developer can quickly validate correctness through testing.
In plain English: Programming by telling an AI what you want in plain English and letting it write the code — trusting the output based on vibes rather than fully understanding every line.
Used critically to describe accepting AI-generated code without review or comprehension, leading to subtle bugs and unmaintainable systems. The dark side of AI-assisted development: code that works but nobody understands.
Example: 'The intern vibe-coded the entire auth module. It passes tests but none of us can explain why it makes three redundant database calls.'
Source: pejorative usage
Etymology
- 2025
- Andrej Karpathy coins 'vibe coding' in a February tweet, describing writing code by describing intent to an AI and accepting whatever it produces
- 2025
- The term goes viral, sparking debate about whether vibe coding is the future of programming or a recipe for unmaintainable software
- 2025
- Merriam-Webster and other dictionaries add 'vibe coding' to their watch lists within weeks of coining
Origin Story
When AI turns programming into a conversation
In February 2025, Andrej Karpathy — former Tesla AI director and OpenAI researcher — posted on X (formerly Twitter) about a new way of programming he called 'vibe coding.' The concept was simple but radical: instead of writing code manually, you describe what you want in natural language and let an AI coding assistant generate it.
Karpathy described it as 'fully giving in to the vibes, embracing exponentials, and forgetting that the code even exists.' The programmer becomes more of a director than a typist, steering the AI through high-level descriptions and accepting or rejecting its output.
The term instantly resonated with developers who had been using AI coding tools like GitHub Copilot, Cursor, and Claude. It captured a genuine shift in how software is being written — one where understanding and intent matter more than syntax and memorization. Critics worry about developers who can't debug AI-generated code they don't understand; proponents argue it democratizes programming.
Coined by: Andrej Karpathy
Context: X (Twitter) post, February 2025
Fun fact: Within weeks of Karpathy's post, 'vibe coding' appeared in job descriptions, conference talk titles, and at least three Y Combinator applications.