FollowAgents AI AGENTS DIRECTORY
Home / AI Agent Glossary / Vibe Coding
Ecosystem & Emerging Terms

Vibe Coding

Also called: vibecoding · vibe-coding

Vibe coding is a style of programming where you describe what you want in natural language, let an AI write the code, and accept the result largely by how it behaves rather than by reading it.

The term was coined by Andrej Karpathy in early 2025 to describe a mode where you "fully give in to the vibes": prompt the AI, run it, paste errors back, and iterate without closely reading the diffs. It spread quickly, and became a widely used label for AI-assisted building by people who may not write code themselves.

In common use the meaning has broadened. Some people say vibe coding for any AI-assisted programming, but the original sense is specific: the human is not reviewing or fully understanding the code. That distinction matters, because reviewing is what separates it from professional AI-assisted engineering with a coding-agent-harness.

It is good for prototypes, personal tools, and exploring ideas. It gets risky when the code will be maintained, exposed to users, or handle sensitive data, because bugs and security holes nobody read are still shipped.

Example

A designer asks a coding agent for "a page that lets me drag photos into a grid and export them as a zip", runs it, sees a bug, pastes the error back, and repeats until it works, never opening the source. That is vibe coding. A developer who has the agent draft the same feature, then reads the diff, runs tests, and fixes issues before merging is doing AI-assisted engineering, not vibe coding.

How it differs

Vibe coding vs. AI-assisted engineering: both use an agent to write code, but vibe coding skips careful review and understanding, while assisted engineering keeps a human accountable for what merges (see human-in-the-loop).

Common misconceptions

Often assumed: Vibe coding means any programming that uses AI.
Actually: The original meaning is specifically about accepting AI-generated code without closely reviewing or understanding it. Careful AI-assisted development is a different practice.
Often assumed: If the app runs, the vibe-coded code is fine.
Actually: Working behavior on a happy path says little about security, edge cases, or maintainability, which are exactly what unreviewed code tends to miss.

FAQ

What is vibe coding?
Building software by describing what you want to an AI and accepting its code based on how it runs, with little or no reading of the code itself.
Who coined the term vibe coding?
Andrej Karpathy, in early 2025.
Is vibe coding safe for production?
Generally not without review. It suits prototypes and personal tools; production code needs review, tests, and security checks by someone who understands it.

Last checked: 2026-09-20

Related terms

Coding Agent Harness

A coding agent harness is the surrounding software — outside the model itself — that assembles context, executes tool calls, enforces permissions, and feeds results back, turning a raw model into something that can actually operate a coding environment.

Human-in-the-Loop

Human-in-the-loop is a design pattern where an agent pauses to get a person's review or approval before or during a consequential action, rather than acting entirely on its own.

AI Slop (and Workslop)

AI slop is low-quality, unreviewed AI-generated content produced in volume; workslop is the same thing at work: AI output that looks like finished work but leaves the recipient to fix or redo it.

Autonomy Level

Autonomy level is an informal way of describing how much an agent decides and does on its own before checking in with a human, ranging roughly from fully manual to fully autonomous — it's not an official or standardized scale.