You describe what you want. The AI writes the code. It works — mostly. You do not know why. You do not need to. Until you do.
Literal meaning: A vibe coder is someone who creates software primarily through natural language instructions to AI coding assistants — describing what they want in words, accepting the AI’s generated code, and iterating on the description until the output works, without necessarily understanding the code the AI has written.
Origin: The term was popularised by Andrej Karpathy, former Director of AI at Tesla and co-founder of OpenAI, in a February 2025 tweet: “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.” Karpathy described the experience as productive, fun, and fundamentally different from traditional software development. The term spread immediately, capturing an experience many AI-assisted developers recognised.
A practice of software creation through AI instruction that produces working outputs without conventional programming knowledge — whose risks emerge later.
The Appeal: Vibe coding genuinely lowers the barrier to building software. People without formal programming backgrounds can produce working applications. Experienced developers can prototype faster. The experience is described as liberating — creative energy that was previously blocked by technical friction is suddenly unblocked. Karpathy’s enthusiasm was not ironic.
The Friction: The gap between working code and understood code has consequences that are deferred rather than avoided. Cognitive Surrender — the gradual transfer of thinking to digital systems — applies specifically here: the vibe coder does not develop the underlying skill, and the working code is a black box they cannot debug, maintain, or extend independently. Deskilling is the systemic consequence: as vibe coding normalises, the pool of developers who can reason about code without AI assistance shrinks. AI Dependency is structural: the vibe coder’s work exists only as long as the AI system they depend on is available, affordable, and producing correct outputs. The crisis arrives when the code fails in production and no one can explain why.
Why This Matters: Vibe coder names a practice that is already common and rarely examined. The question is not whether AI coding assistance is useful — it clearly is — but what you are trading when you accept outputs you cannot explain. Understanding the trade is the beginning of making it intentionally.
Related terms: Prompt Engineer · Deskilling · AI Dependency · Cognitive Surrender · AI Literacy · Red Teamer
Read more:
- There’s a new kind of coding I call “vibe coding” — Karpathy, A. (2025). X / Twitter
- Not all AI-assisted programming is vibe coding (but vibe coding rocks) — Willison, S. (2025). Simon Willison’s Weblog
- Vibe coding: programming through conversation with artificial intelligence — Sarkar, A. & Drosos, I. (2025). arXiv
- Vibe Coding as a Reconfiguration of Intent Mediation in Software Development: Definition, Implications, and Research Agenda — Meske, C., Hermanns, T., von der Weiden, E., Loser, K.-U. & Berger, T. (2025). arXiv
- Good Vibrations? A Qualitative Study of Co-Creation, Communication, Flow, and Trust in Vibe Coding — Pimenova, V., Fakhoury, S., Bird, C., Storey, M.-A. & Endres, M. (2025). arXiv