Vibe Coding
In short
Vibe coding is building software by describing what you want to an AI and accepting the code it writes, mostly judging the result by whether it seems to work.
What is vibe coding?
The term was coined by the AI researcher Andrej Karpathy in 2025 for a style of programming where you talk to an AI coding tool in plain language, accept its changes without reading them closely, paste error messages back to it, and keep going until the program does what you want. The code itself fades into the background.
For prototypes, personal tools and quick experiments, vibe coding can be remarkably fast, and it lets people who don't program build working software. The trouble starts when vibe-coded software is meant to last: nobody understands the code, so bugs, security holes and duplicated logic pile up unnoticed, and changes get harder with every round.
Most professional developers use AI assistants differently: they still read, test and review what the model writes, and they own the result. The line is not whether an AI wrote the code, but whether anyone understands it well enough to vouch for it.
Key takeaways
- Vibe coding means describing what you want to an AI and accepting its code with little review.
- It is fast for prototypes and lets non-programmers build working software.
- Unread code hides bugs and security holes, which makes it risky for software that must last.
- Reviewing and testing AI-written code turns vibe coding back into ordinary AI-assisted programming.
Readers ask
Is vibe coding bad?
Not in itself. It is a good fit for throwaway prototypes, demos and personal scripts. It becomes a problem when the result goes into production or handles other people's data, because nobody has checked what the code really does.
What is the difference between vibe coding and using an AI coding assistant?
The difference is review. With an AI assistant, a developer reads, tests and understands the suggested code before keeping it. Vibe coding skips that step and judges the code only by whether the program seems to work.
See also
- LLMAI & Machine Learning, p. 25An LLM is a machine learning model trained on huge amounts of text that generates language by repeatedly predicting the next most likely piece of text.
- AI AgentAI & Machine Learning, p. 2An AI agent is a system that uses an LLM to plan and carry out multi-step tasks by deciding which tools to call, observing the results, and acting again.
- PromptAI & Machine Learning, p. 35A prompt is the input text or instructions you give an AI model, such as an LLM, to tell it what task to perform and what kind of answer you want.
- Generative AIAI & Machine Learning, p. 20Generative AI is artificial intelligence that creates new content, such as text, images, code, or audio, based on patterns learned from existing data.
- Code ReviewVersion Control, p. 4A code review is the practice of having other developers check code changes before they are merged, to catch bugs, improve quality, and share knowledge.
- Unit TestTesting & Quality, p. 35A unit test is a small, automated check that verifies one function, method, or class behaves correctly in isolation from the rest of the program.
- HallucinationAI & Machine Learning, p. 23A hallucination is when an AI model, such as an LLM, confidently produces information that sounds plausible but is false, invented, or unsupported by sources.
- DebuggingProgramming Fundamentals, p. 15Debugging is the process of finding out why a program misbehaves, locating the faulty code and fixing it, often with the help of a tool called a debugger.
Spotted a mistake or something missing on this page?Suggest an edit