Vibe coding
What is vibe coding? A plain-English answer
Part of the vibe coding guides.
Short answer
Vibe coding is building software by describing what you want to an AI in plain English and letting it write the code. You judge the result by whether it works, not by reading the code yourself.
That’s why so many people who never thought of themselves as builders are suddenly building things, and why some of those things break the moment real customers touch them.
What does vibe coding mean?
Before, building a website or an app meant writing code, line by line, in a language the computer understands. With vibe coding, you write what you want in the language you already speak: “add a booking form under the prices and send me an email when someone fills it in”. The AI writes the code, runs it and shows you the result. You react, ask for changes, and repeat.
So the skill shifts. You don’t need to know the syntax anymore. You do still need to know what you want, and whether what you got is actually right.
Why is it called vibe coding?
The name came from AI researcher Andrej Karpathy in early 2025. He described a way of working where you “give in to the vibes”: you describe, accept what the AI gives you, and keep going without reading every line. The name stuck because it was honest about the trade: speed now, understanding later.
How does vibe coding work?
In practice it’s a loop:
- You describe the thing. One page, one feature, one fix. Small asks work far better than “build me an app like Airbnb”.
- The AI writes the code. Tools like Claude, Cursor, Lovable or Replit turn your description into working files.
- You look at what it made. Click through it like a customer would.
- You ask for changes. “The button’s hidden on my phone.” “Send the email to me, not the customer.”
- You save a version that works. This is the step beginners skip, and it’s the one that saves you later.
What is vibe coding in AI terms?
There are three parts. The model is the AI that writes the code, like Claude or Gemini. The tool is the app you type into, which gives the model access to your files and a way to run them. And you are the part that decides. The model can suggest any number of answers, but it doesn’t know your business, so it can’t tell you which answer is right for you.
Vibe coding vs agentic coding, prompt engineering and AI-assisted coding
- Prompt engineering is getting better answers out of a chat window by writing better questions. You still copy and paste the result.
- AI-assisted coding is a developer writing code with AI suggestions alongside. The developer reads every line.
- Vibe coding is describing what you want and mostly trusting the output. You test the result, not the code.
- Agentic coding is the next step: the AI plans the work, edits many files, runs commands and checks its own results, while you review. That’s closer to how I work now. I haven’t typed a Git command by hand in a year or two, because I tell Claude what I want and it runs them.
Most people are still at the chat-window stage, copying answers out one at a time. The jump from there to letting AI work inside your actual files is the biggest one, and it’s where knowing what not to break starts to matter.
Vibe coding examples
Things a small business owner can realistically vibe code:
- A simple website with your services, prices and a contact form.
- A booking page that sends you an email when someone picks a slot.
- An internal checklist or calculator your team uses every day.
- A landing page for one offer, to test if anyone wants it.
Things I’d be careful with: anything that takes payments, stores health or personal data, or has lots of users logging in. Not because AI can’t write it, but because the decisions behind it cost real money when they’re wrong.
Where vibe coding stops working
Code was the fast part, and vibe coding proved that. What it didn’t solve is the decisions: where your data lives, what happens when something breaks at a bad moment, how it copes when more people use it, and what’s actually secure. AI will make all of those choices for you if you let it. It just won’t tell you which ones were guesses.
That’s usually when people hire a developer, and it’s worth knowing what you’re paying for: not more code, but someone to undo the guesses. I wrote more about that in my guide to vibe coding tools and the decisions they can’t make.
Questions people ask
Who coined vibe coding?
Andrej Karpathy, an AI researcher, in early 2025.
Will vibe coding replace programmers?
I don’t think so. People have predicted the end of programmers several times before, and each time the tools got better and the people who understood how the whole system fit together pulled further ahead. What’s changing is who gets to build the first version.
Is vibe coding the future?
For first versions, yes, it already is. For things that hold customer data or money, the future looks more like vibe coding plus someone who understands the decisions underneath.
Why do people say vibe coding is bad?
Because it’s easy to ship something that works in a demo and fails with real users: exposed keys, no backups, no way back when a change breaks everything. None of that is a reason not to do it. It’s a reason to save versions and check the basics before you launch.
Can I vibe code if I’ve never written code?
Yes. That’s the point. Start with something small and real, use one tool, and save every version that works.
Can vibe coding get you a job?
It can get you a working thing to show, which helps. The people who get hired are the ones who can explain why they built it that way.
Next steps
Want the next guide by email, or want it built for you?
Join the waitlist for a done-for-you build.
One short email every Friday, plus a full guide once a month.