What is a vibe coder?

A vibe coder is a developer who builds software by describing intent to AI tools like Cursor, Claude, Bolt, Lovable, and Replit, then iterating on the output. The term was coined by Andrej Karpathy in 2025. It is now a recognised skill set with thousands of practitioners worldwide.

What a vibe coder actually does

Traditional software development requires deep familiarity with syntax, frameworks, and build tooling. A vibe coder works differently. They describe what they want in plain language — the feature, the behaviour, the constraint — and use AI tools to generate, refine, and debug the code. The skill is in the prompting, the iteration, and knowing when the output is wrong.

This does not mean vibe coders are non-technical. The most effective practitioners understand how code works, can read and debug generated output, and know when to reach for a traditional approach. What distinguishes them is velocity: they ship faster, prototype cheaper, and operate across more of the stack than a conventional developer with equivalent years of experience.

84% of developers use or plan to use AI coding tools (Stack Overflow Developer Survey 2025). Vibe coders are the leading edge of that shift — people who have restructured their entire workflow around AI assistance, not just added it as an occasional aid.

Tools commonly used by vibe coders

Cursor and GitHub Copilot for in-editor AI assistance. Claude, GPT-4o, and Gemini for reasoning and architecture decisions. Bolt, Lovable, and v0 for rapid prototyping from a prompt. Replit for browser-based development. Vercel and Supabase for deployment and backend without deep infrastructure knowledge.

Vibe coder vs AI engineer vs AI-native developer

These three terms overlap but are not interchangeable. The table below shows the key distinctions as they are understood in practice:

In practice, many practitioners span more than one category. A vibe coder may also be AI-native; an AI engineer may use vibe coding techniques to prototype. Hiring managers benefit from understanding the distinction rather than treating all AI-fluent developers as interchangeable.

How to hire a vibe coder

Assessing a vibe coder is different from a standard technical interview. The goal is to evaluate output quality and workflow fluency, not syntax recall.

1. Screen for shipped, live projects. Ask for links to things they have actually deployed — apps, tools, prototypes. A portfolio of live URLs is a stronger signal than a GitHub repo full of unfinished branches.
2. Assess AI-tool fluency, not just awareness. Everyone knows Cursor exists. Ask how they use it — what their prompting approach is, how they handle incorrect output, which tools they reach for at different stages of a build.
3. Test ability to debug AI-generated code. Give them a real piece of broken AI-generated code. Can they read it, identify the error, and fix it — with or without AI assistance? This distinguishes effective vibe coders from people who can only generate code they cannot understand.
4. Ask about scope management. Strong vibe coders know when to keep prompting and when to step back and reason manually. Ask about a time an AI tool led them in the wrong direction and how they recovered.
5. Look for velocity evidence. Ask how long it takes them to go from idea to deployed MVP. Vibe coders with genuine fluency can often ship an end-to-end prototype in a single working session.

Standard engineering interviews — LeetCode problems, whiteboard algorithms — are a poor fit for this role. They measure the wrong skills. Assessment should focus on real product output, workflow reasoning, and debugging capacity.

Frequently asked questions