AI_SLANG_ENTRY
What Does Vibe Coding Mean?
A loose style of AI-assisted programming where a person describes what they want, lets the model write most of the code, and judges progress mainly by running or seeing the result rather than understanding every line.
What does Vibe Coding mean?
A loose style of AI-assisted programming where a person describes what they want, lets the model write most of the code, and judges progress mainly by running or seeing the result rather than understanding every line.
Vibe coding is prompting an AI to build the software, trying what it produces, and steering by whether the result feels right instead of reviewing every implementation detail.
Origin and usage
Andrej Karpathy coined the phrase in a February 2025 post describing a workflow of speaking or typing requests, accepting model-written changes, running the result, and often ignoring the underlying code.
Source type: community. Last checked: 2026-08-26.
Karpathy's original post described an intentionally code-light, result-driven workflow for casual projects. Current usage is broader, so this entry distinguishes vibe coding from any careful use of an AI coding assistant.
Where the term vibe coding came from
Karpathy used the phrase for a workflow in which the model handled almost all of the code while he supplied requests, ran the program, pasted errors back, and evaluated what appeared on screen.
The phrase spread because it captured both the speed and the deliberate looseness of building by feel. It was not originally a formal software-engineering method or a claim that review no longer matters.
Vibe coding versus ordinary AI-assisted coding
Using autocomplete, asking for a test, or reviewing a model-generated patch is AI-assisted coding. It becomes vibe coding when the developer mostly steers outcomes and stops treating understanding the code as a requirement.
The boundary is cultural rather than standardized. One person may use the term proudly for rapid prototyping while another uses it critically for unreviewed production code.
Where vibe coding fits
- Fast prototypes, experiments, personal tools, and disposable weekend projects.
- Interfaces where a person can quickly see whether the generated result matches the request.
- Early exploration before maintainability, security, accessibility, and operational requirements are fully known.
- Work that will later receive normal tests, review, ownership, and production hardening if it becomes important.
The production risk
A result that looks correct can still contain insecure dependencies, broken edge cases, inaccessible UI, invented APIs, or code nobody on the team can maintain. Prompting faster does not remove those obligations.
For production software, keep the speed but restore the engineering loop: inspect the diff, run relevant tests, verify security boundaries, document ownership, and make sure someone understands how the system fails.
Examples
- I vibe coded the prototype, then spent Friday making it real.
- My app broke because the model's vibes changed after the last vendor update.
FAQ
What is the definition of vibe coding?
Vibe coding is an informal style of AI-assisted development where a person prompts for changes, runs the result, and steers by the outcome while reading or understanding little of the generated code.
Who coined the term vibe coding?
AI researcher Andrej Karpathy coined the phrase in a February 2025 post about letting an AI model handle most of the code for casual projects.
Is using an AI coding assistant always vibe coding?
No. Developers can use AI while still reviewing every change, writing tests, understanding the architecture, and owning the result. Vibe coding specifically emphasizes steering by outcomes rather than code comprehension.
Is vibe coding safe for production software?
Not by itself. Production work still needs code review, tests, security checks, accessibility review, observability, and a maintainer who understands the system.