Published June 10, 2026
Vibe coding is building software by describing what you want in plain language and letting an AI write the code, often without reading it line by line. The term was coined by Andrej Karpathy in February 2025 and became Collins Dictionary's Word of the Year for 2025. This guide explains what vibe coding actually means, where the term came from, the tools people use, whether it is a good idea, and how to start, with links to each step.
What is vibe coding?
Vibe coding is a way of programming where you tell an AI what you want in natural language and accept the code it generates without fully reviewing or understanding it. You describe a feature, the model writes it, you run it, and if something breaks you paste the error back and ask the AI to fix it, guiding the project by results and prompts rather than by reading the code.
The key word is vibes: you are steering on intent and output, not on a careful read of every line. That is what separates vibe coding from ordinary AI-assisted programming, where you still review, test, and understand what the model produced.
Where did the term vibe coding come from?
Andrej Karpathy (a founding member of OpenAI and former head of AI at Tesla) coined "vibe coding" in a post on X on February 2, 2025:
"There's a new kind of coding I call 'vibe coding', where you fully give in to the vibes, embrace exponentials, and forget that the code even exists... I 'Accept All' always, I don't read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it."
He described it as fun and effective for small, throwaway projects, not as a method for serious production software, a caveat that often gets lost. The term spread fast: Merriam-Webster added it as a "slang and trending" entry in March 2025, and Collins English Dictionary named "vibe coding" its Word of the Year 2025 (announced November 6, 2025), defining it as "the use of artificial intelligence prompted by natural language to assist with the writing of computer code."
Is vibe coding the same as AI-assisted engineering?
This distinction matters and most articles skip it. Developer Simon Willison drew the cleanest line: vibe coding is "building software with an LLM without reviewing the code it writes." His test:
"If an LLM wrote the code for you, and you then reviewed it, tested it thoroughly and made sure you could explain how it works to someone else, that's not vibe coding, it's software development."
So the precise definition is narrow: vibe coding specifically means not reading or understanding the code. Using Claude Code or Cursor while still reviewing, testing, and understanding the output is just AI-assisted engineering, a different and generally safer practice. In everyday use the term has blurred to mean "any AI coding," but the precise meaning is the no-review version.
What tools do people use to vibe code?
Karpathy's original setup was Cursor (Composer) with Claude Sonnet, plus voice input. In 2025-2026 the tools commonly grouped under vibe coding are:
- Natural-language app builders (closest to pure vibe coding): Lovable, Bolt.new, v0 by Vercel, Replit Agent, Google AI Studio. You describe an app and get a working one.
- AI code editors and agents (closer to assisted engineering): Cursor, Claude Code, Windsurf, GitHub Copilot.
For a full breakdown of the tools see Best Vibe Coding Tools 2026. For the two most-used developer tools head to head, see Claude Code vs Cursor and Claude Code Pricing.
The breakout commercial story is Lovable: it reportedly went from $0 to about $100M in annual recurring revenue in roughly eight months, doubled to around $200M by late 2025, and raised a $330M round at a $6.6B valuation in December 2025, which is a large part of why "vibe coding" stopped being a meme and started being an industry.
Is vibe coding bad?
It depends entirely on what you are building. The honest answer:
Fine for: prototypes, throwaway tools, weekend projects, learning, validating an idea fast. This is exactly what Karpathy meant.
Risky for: anything real and long-lived. The documented problems are consistent:
- Security. Code you did not read can ship vulnerabilities. A 2025 Veracode analysis found AI-generated code showed little security improvement even as functionality improved, and there have been real incidents of vibe-coded apps leaking user data.
- Maintainability. Code you do not understand becomes technical debt you cannot service. You cannot fix what you cannot explain.
- The prototype-to-production trap. The biggest danger is not the throwaway app, it is the throwaway app that quietly becomes business-critical without anyone ever reading it.
The practical rule: vibe code to explore and prototype, then switch to reviewed, tested, understood code (AI-assisted engineering) the moment something is going to matter. If you cannot explain what your code does, do not put real users or real data behind it.
How do you start vibe coding?
A sensible first path:
- Pick a tool. For a no-setup "describe an app, get an app" start, try Lovable, Bolt.new, or v0. For coding in a real editor or terminal, use Cursor or Claude Code.
- Start with something throwaway. A small tool, a landing page, a script. Low stakes is where vibe coding shines.
- Prompt by outcome. Describe what you want, run it, paste errors back, iterate.
- Level up when it gets real. Add MCP servers to connect your tools, start reading and testing the code, and move toward reviewed engineering.
- Then ship it. Building is now the easy part; getting users is the hard part, which is a whole discipline of its own. See How to Promote Your App.
Frequently asked questions
What is vibe coding? Vibe coding is building software by describing what you want in natural language and letting an AI generate the code, accepting and iterating on it without fully reviewing or understanding every line. You steer by results and follow-up prompts rather than by reading the code.
Who coined the term vibe coding? Andrej Karpathy, a founding member of OpenAI and former head of AI at Tesla, coined it in a post on X on February 2, 2025. Collins Dictionary named vibe coding its Word of the Year for 2025.
Is vibe coding the same as using AI to code? Not exactly. The precise meaning of vibe coding is using an AI to write code you do not review or fully understand. If you use an AI but still review, test, and understand the output, that is AI-assisted engineering, a different and safer practice. In casual use the terms get blurred.
Is vibe coding bad? It is good for prototypes, throwaway tools, and learning, which is what the term's creator intended. It is risky for production software because code you have not read can carry security holes and becomes unmaintainable. The danger is a quick prototype quietly turning into a business-critical system no one understands.
What tools do people use for vibe coding? Natural-language app builders like Lovable, Bolt.new, v0, and Replit Agent, and AI code editors and agents like Cursor, Claude Code, Windsurf, and GitHub Copilot. Karpathy's original setup was Cursor with Claude Sonnet.
How do I start vibe coding? Pick a tool (Lovable or Bolt for no-setup, Cursor or Claude Code for a real editor), build something small and throwaway, prompt by describing outcomes and pasting errors back, and switch to reviewed code once a project starts to matter.
