Vibe Coding
Building software by describing what you want to an AI in plain language and accepting the code it produces, often without reading it closely.
What is vibe coding?
Vibe coding is a way of building software by telling an AI, in plain language, what you want and letting it write the code. The term was coined by Andrej Karpathy in February 2025. In its purest form the person accepts the generated code without closely reviewing it, and steers by running the result and asking for changes.
How does it work?
You describe a feature, the large language model writes the code, and you try it. If something is wrong, you describe the problem or paste the error and ask again. Tools such as AI copilots and coding agents support this loop. Careful engineers use the same tools but still read and test what the AI writes.
What is a simple example?
Someone with no programming background asks an AI to "make a page where I can log my workouts and see a weekly chart". They run what comes back, ask for a different colour and a delete button, and end up with a working tool.
Why does it matter?
It lets more people build working software quickly, which suits prototypes and personal tools. The risk is that unreviewed code can hide bugs and security problems, because models can produce confident mistakes (hallucinations). For anything important, review and testing still matter.