By Quick answer: What is vibe coding?
- Definition — describe, generate, run, iterate; don't read every line.
- Origin — coined by Andrej Karpathy in February 2025.
- Great for — prototypes, internal tools, learning, personal apps.
- Risky for — payments, personal data, security-critical or large systems.
- Do it safely — tests, reviews, secrets kept out of code, small steps.
Key takeaways
- Vibe coding lowers the cost of trying ideas dramatically.
- Speed to a demo is not the same as readiness for real users.
- Experienced developers still review, test and structure AI output.
- Developer surveys show high AI adoption but low trust in accuracy.
- Use it to explore; switch to careful engineering before you scale.
In February 2025, Andrej Karpathy described a new way he'd been writing software: describing what he wanted, accepting what the AI produced, pasting errors back in, and rarely reading the code at all. He called it vibe coding1. The phrase spread so fast that Collins Dictionary named it a Word of the Year for 20252. It captures something real: for the first time, many people can turn an idea into working software just by talking about it.
How vibe coding works
- Describe what you want in plain language — "a habit tracker with streaks and a weekly chart".
- Generate — an AI tool writes the code and often runs it for you.
- Try it — click around; see what works and what breaks.
- Iterate — describe changes or paste errors back to the AI.
- Repeat until it does what you want.
What it's great for
Prototypes
Test an idea with users in hours instead of weeks.
Personal and internal tools
A dashboard for your shop, a script for your team, a tool for your own workflow.
Learning
See how things are built; ask the AI to explain its code.
Unblocking
Get past the blank page; generate boilerplate quickly.
Where it goes wrong
| Risk | Why it happens | What to do |
|---|---|---|
| Security holes | AI code can include insecure patterns or exposed secrets | Review against the OWASP Top 105; scan dependencies |
| Fragile logic | Nobody fully understands the code | Write tests; keep changes small |
| Data loss or leaks | Weak database rules and access control | Use proven auth and storage services; check permissions |
| Unmaintainable code | Many quick patches pile up | Refactor regularly; involve an engineer before scaling |
| Overconfidence | Demos hide edge cases | Test with real users and real data volumes |
What the evidence says
AI coding is now mainstream: 84% of developers in Stack Overflow's 2025 survey said they use or plan to use AI tools — but more distrusted the accuracy of AI output (46%) than trusted it (33%)3. And speed is not guaranteed: in a 2025 randomised trial, METR found experienced open-source developers took 19% longer on real tasks when using AI tools, even though they believed they'd been faster4. The lesson: vibe coding shines for exploration, and careful engineering still matters for production.
Vibe code safely
- Keep API keys and secrets out of your code and prompts
- Use managed services for login, payments and storage
- Ask the AI to write tests, then run them
- Commit often so you can roll back
- Get a code review before real users or real money are involved
For a step-by-step build, see how to build an app with AI; to pick a tool, see the best AI coding tools.
Key terms, defined
Vibe coding — building software by prompting AI and iterating on its output. Prompt — the instruction you give an AI. AI app builder — a tool that generates whole apps from prompts. Code review — a person checking code for problems. Technical debt — future cost of quick, messy code.
Frequently asked questions
Who coined vibe coding?
Andrej Karpathy, an AI researcher and former OpenAI and Tesla leader, coined the term in a post in February 2025, describing programming where you 'fully give in to the vibes' and let AI write the code.
Is vibe coding real programming?
It's a way of producing software, and it can create real, working tools. But building reliable, secure software for many users still needs programming skills: understanding, testing and maintaining the code.
What tools are used for vibe coding?
AI app builders such as Lovable, Bolt and Replit for full apps from prompts, and AI coding tools such as Cursor, Claude Code, GitHub Copilot and OpenAI's Codex for working inside codebases.
Is vibe-coded software safe?
Not by default. AI-generated code can contain security flaws, leaked secrets and fragile logic. Review it, test it, scan it and keep humans responsible for anything handling money or personal data.
Can non-programmers vibe code?
Yes, for prototypes and simple tools. Learning basic concepts — how data is stored, how logins work, what an API is — makes results much better and safer.
Will vibe coding replace developers?
It changes what developers spend time on, shifting work toward specifying, reviewing, testing and designing systems. Demand for people who can make software reliable remains strong.
Sources
- Andrej Karpathy — Original 'vibe coding' post (February 2025)
- News On AIR — 'Vibe coding' named Collins Dictionary's Word of the Year 2025
- Stack Overflow Developer Survey 2025 — AI
- METR — Measuring the impact of early-2025 AI on experienced open-source developer productivity
- OWASP — Top 10 web application security risks
- Checked October 23, 2026. Figures that change — fees, salaries, prices — are reviewed every six months; if something is out of date, tell us.
Where to go from here
- How to build an app with AI — the practical walkthrough
- Best AI coding tools — choose your vibe-coding tool
- AI & vibe codingWhat is an AI agent? A plain-English explanation, with examples you already useWhat an AI agent is, how it differs from a chatbot and a workflow, the loop that powers it, everyday examples and the risks.
- Tech stackWhat is an API? A plain-English explanation with a restaurant analogy and real app examplesWhat is an API? A plain-English explanation using a restaurant analogy, with real examples from UPI, maps and login, plus REST, keys, webhooks and security.
- AI & vibe coding14 copy-paste AI skills for data analysts, for ChatGPT, Claude and Gemini14 free copy-paste AI skills for data analysts — SQL from plain English and more — each with an output preview and the exact text to paste.
- AI & vibe codingHow to build an AI agent without coding: a step-by-step guide with a worked exampleBuild an AI agent without coding: pick one job, choose a platform like n8n, Zapier or Make, connect tools, add guardrails and test.
- AI & vibe codingPrompt engineering guide: the principles that work, with 50 copy-ready examplesA practical prompt engineering guide: seven principles from the AI labs' own docs, a reusable template and 50 copy-ready examples.
- How to make an appHow to make an app with zero coding experience (2027 guide)You can make an app without coding in 2027: validate the idea, design it, then build with no-code tools, AI app builders or a hired developer. Step by step.
Get the ideas, not the noise.
Sign up with Google and we’ll send you short UX psychology tips by email — why people tap, stall and leave, and what to do about it. Unsubscribe anytime. Prefer something to keep? Get the free Laws of UX guide as a PDF.
Make your AI feature something users trust.
A behavioral UX audit of your app or SaaS product: why users hesitate, stall and leave, and a prioritized fix list. Fixed price, 2–4 weeks.