Vibe coding is a recent software development approach where programmers rely on artificial intelligence—usually large language models (LLMs)—to generate most, or all, of the code for a project, guiding the process through plain language descriptions and iterative requests.
Definition and Origin
Vibe coding was popularized by Andrej Karpathy, a co-founder of OpenAI, in early 2025. Instead of manually writing lines of code, developers (and even non-technical users) interact conversationally—often speaking or typing broad requests such as “create a login form,” “reduce sidebar padding,” or “fix this bug”—and the AI generates, modifies, and debugs the code in response. The process is characterized by minimal inspection of the code itself; users routinely accept AI-suggested changes without reviewing them in detail, and may apply further prompt-based fixes if something isn’t working.
Workflow Differences
Traditional coding requires deep knowledge of programming languages and painstaking control over each function and line. Vibe coding, by contrast, focuses on the desired outcomes and creative intentions, delegating technical implementation to the AI and reviewing results by testing rather than reading code.
- You describe what you want your application to do in English or plain speech, and the AI crafts the code.
- Debugging and refinement involve asking the AI for improvements or new solutions, iterating not with written code, but with repeated conversational requests.
Practical Applications and Reception
Vibe coding has made it possible for amateurs and professionals alike to build apps quickly, sometimes for personal use or MVPs, though critics note risks regarding code quality, maintainability, and security vulnerabilities. Startups and individual creators have begun using this method for rapid prototyping, and large-scale projects can involve codebases that are almost entirely AI-generated. However, careful review and manual fixes may still be required for mission-critical systems due to limitations in current AI code generation.
Tools and Platforms
Common vibe coding tools include:
Key Characteristics
- Minimal technical barrier to entry
- Fast idea-to-app cycles
- Conversation and feedback-driven iteration, not code-centric editing
- Often little to no manual code review or understanding
Vibe coding represents a shift: the programmer’s role moves from writing code, to specifying outcomes and goals, while the AI “just handles the syntax”.