What Is Vibe Coding? A Builder's Honest Guide

What is vibe coding? It's building software by describing what you want to an AI model, running what it produces and steering with follow-up prompts, often without reading the code line by line. You judge the result by whether it works, not by how it's written. It's the fastest way to go from idea to working prototype, and a risky way to build anything people depend on.

I vibe code most weeks: internal tools, quick scripts, prototypes for clients. I also run production systems where I'd never work that way. Here's what the term means, what the evidence says and the rules I use to get the speed without the mess.

What is vibe coding, exactly?#

According to Wikipedia's summary, vibe coding is AI-assisted software development where a developer describes a project in a prompt and a large language model generates the source code. It may involve accepting that code without thorough review, relying on results and follow-up prompts to guide changes.

The key difference from other AI-assisted programming is where your attention goes:

Both use the same tools. The difference is discipline, not software.

Where the term came from#

Andrej Karpathy, an OpenAI co-founder and former head of AI at Tesla, coined the phrase in February 2025. He described giving in to the flow and letting the AI handle the code. It built on his 2023 remark that the hottest new programming language is English.

The term spread fast. Merriam-Webster listed it as slang and trending in March 2025, and Collins English Dictionary named it Word of the Year for 2025. By early 2026, even Linus Torvalds had vibe coded a small visualization tool for a hobby project.

What the evidence says#

Enthusiasm is easy to find. The data is more mixed, and worth knowing before you build on it. All of these findings are collected in Wikipedia's article on vibe coding, which links the original reports:

• Adoption is real. Y Combinator reported in March 2025 that 25% of startups in its Winter 2025 batch had codebases that were 95% AI-generated.
• Speed can be an illusion. A July 2025 randomized trial by METR found experienced developers were 19% slower with AI tools on their own projects, while believing they were faster.
• Quality suffers without review. A December 2025 CodeRabbit analysis of 470 open-source pull requests found AI co-authored code had about 1.7 times more major issues.
• Security incidents happen. In May 2025, 170 of 1,645 apps built with one popular app builder were reported to have a flaw exposing personal data, and in July 2025 an AI agent deleted a production database despite instructions not to change anything.

None of this means don't vibe code. It means match the method to the stakes.

When vibe coding is the right call#

• Prototypes you'll show, learn from and probably throw away.
• Software for one: a personal dashboard, a script that renames files, a tool only you use.
• Internal tools with no sensitive data and a small, forgiving audience.
• Exploring a library, API or idea before committing to a design.

It's the wrong call for anything handling payments, personal data, authentication or other people's work, unless you add the safeguards below.

How to vibe code without shipping a mess#

These are the rules I follow, and they barely slow anything down:

1. Use Git from the first prompt. Commit after every working step, so a bad change is one command away from undone.
2. Write a rules file. In Claude Code, a CLAUDE.md with your stack, conventions and forbidden actions prevents most repeat mistakes. My guide to CLAUDE.md shows what to put in it.
3. Ask for tests with every feature. Tell the AI to write a test and run it before saying it's done.
4. Keep secrets out. Environment variables only, never pasted keys, and a quick search for exposed credentials before sharing anything.
5. Separate prototype from production. When a vibe-coded tool becomes important, review it properly or rebuild it deliberately.
6. Get a second look. Have a fresh AI session or a person review the diff for security and logic, not just whether it runs.

Anthropic's Claude Code best practices push the same idea: give the model a way to verify its own work, and the results improve dramatically.

Which vibe coding tool to start with#

• Beginners who want an app in the browser: Lovable or Replit.
• People comfortable in a terminal or editor: Claude Code, Cursor or Codex, which give you more control over files, Git and tests.

If you're choosing between the developer tools, compare them in Cursor vs Claude Code and Claude Code vs Codex, or start with how to use Claude Code.

Where to go next#

Vibe code one small tool this week, with Git and a rules file from the start, and notice where you stopped understanding what was built. That's the line where vibe coding needs to become engineering, and it's exactly what my Claude Code course teaches people to cross.