Vibe coding
Building software by describing to an AI in plain language what it should create.
Explanation
Vibe coding means building software by describing to an AI in natural language what it should create, while barely reading the generated code. The main check is whether the result works.
How it works
You describe a feature or app in a tool such as Lovable, Bolt, v0, Cursor or Claude Code. The AI generates the code, runs the application and fixes errors based on feedback. The human steers through instructions and trying things out, not through the code itself.
Example
A sales manager builds an internal tool in one afternoon that summarizes quotes from the CRM and filters them by status, without knowing how to program.
Why it matters
Vibe coding makes prototypes and internal tools possible in days instead of months. Without checks on security, data model and operations, however, it quickly becomes a risk once real users and real data are involved.
What is vibe coding? Origin and meaning
The term was coined by AI researcher Andrej Karpathy in February 2025. He used it to describe a way of working in which you fully give in to the AI, accept changes without reading them closely and simply paste error messages back into the chat. Since then, vibe coding has come to mean software development through natural-language descriptions in general.
It is important to distinguish it from AI-assisted development by professionals. Experienced developers also let AI write large parts of their code today, but they review it, define architecture and standards and secure the result with tests. Vibe coding in the narrower sense skips exactly those steps.
Vibe coding: opportunities and risks
The opportunities are real. Business teams can show ideas as working prototypes instead of describing them in requirement documents. Internal tools get built without waiting in the IT queue, and founders can test product ideas with real users before investing in development.
The risks usually surface once the prototype becomes a system people rely on. AI optimizes for “it works”, not for “it is secure”. Veracode's 2025 GenAI Code Security Report found a vulnerability from the OWASP Top 10 in 45 percent of the AI-generated code samples it tested. Typical issues are missing access rules in the database, API keys in frontend code, missing input validation and no backups. As soon as personal data is involved, a bug can quickly become a data protection incident.
From vibe-coded prototype to production
Getting to production does not have to mean a rewrite. A structured check before rollout has proven effective: review authentication, permissions and database rules, scan for exposed keys, clean up the data model, separate test and production environments, set up automated backups and monitoring, and define a clear update process. What such a check looks like and which vulnerabilities we find most often is covered in Taking vibe-coded apps to production.
Frequently asked questions
What does vibe coding mean?
Vibe coding means building software by telling an AI what to create and judging the result mainly by whether it works, without reviewing the generated code in detail.
Which tools are used for vibe coding?
Popular options are app builders such as Lovable, Bolt and v0, as well as AI-powered editors and coding agents such as Cursor or Claude Code.
Is vibe coding suitable for businesses?
Yes for prototypes, experiments and simple internal tools. Before an application goes live with real customer data or many users, it should be reviewed for security, data protection and operational readiness.
Will vibe coding replace software developers?
No. It shifts the work from typing code to clarifying requirements, defining architecture, reviewing and securing. Those skills decide whether an application runs reliably in production.