TL;DR

Vibe coding is a term that describes a programming approach where individuals with minimal or no coding expertise use natural language prompts to instruct AI tools, specifically large language models (LLMs), to generate functional software.

Instead of manually writing code line-by-line, the user describes what they want in plain English (e.g., "create a to-do list app"), and the AI produces functional software.

The term captures a "go with the flow" mindset, accepting that the code may have bugs but valuing rapid development.

A hallmark of vibe coding is that users may accept AI-generated code without fully reviewing or comprehending it, trusting the AI to handle the technical details.

Origin

The concept was coined by Andrej Karpathy, a prominent computer scientist, AI researcher, ex-AI leader at Tesla, and former OpenAI co-founder.

In his post, he described vibe coding as “fully giving in to the vibes, embracing exponentials, and forgetting that the code even exists.”

He highlighted using tools like Cursor Composer and Claude Sonnet, where he’d issue casual commands (e.g., “decrease the sidebar padding by half”) and accept AI-generated changes without reviewing every detail. This marked a shift from traditional coding’s emphasis on control to a more intuitive, AI-driven process.

The coder’s role shifts to guiding, testing, and refining the AI’s output, often without deeply understanding the code itself. It’s a playful, experimental method prioritizing speed and creativity over precision, typically used for prototyping or small-scale projects.

Why It Sparked Recently?

LLMs had become highly proficient at generating accurate and functional code from natural language prompts.

Tools like OpenAI’s Codex, GitHub Copilot, and Anthropic’s Claude have evolved to generate reliable code from natural language. Earlier tools (e.g., Copilot in 2021) focused on code suggestions, but newer models handle entire projects, making vibe coding feasible.

Earlier AI coding assistants focused on autocompletion or suggesting code snippets, but newer models could build entire applications from high-level descriptions, making vibe coding feasible.

Vibe coding represents a natural evolution, where natural language itself becomes the "programming language". The rise of no-code and low-code platforms in the early 2020s (e.g., Zapier, Bubble, Webflow) had already lowered barriers to creating software.

This shift allowed non-technical creators, designers, entrepreneurs, or hobbyists to prototype apps or tools without learning syntax or hiring developers.

Concerns

While vibe coding excels for “throwaway” projects, some critics argue it’s risky for production code due to potential bugs, security issues, or lack of maintainability if coders don’t understand the output.

It’s less about replacing software engineering and more about empowering creativity for non-technical users or speeding up early development.

AI expert Gary Marcus also pointed out that much of vibe-coded software relies on LLMs reproducing patterns from existing code, raising questions about originality.

One thing is sure for the VIBE coding that English becomes the hottest programming language.