Vibe coding
Also known as: vibecoding, vibe-coding, AI coding, natural language coding
Andrej Karpathy, OpenAI co-founder and former head of AI at Tesla, coined the term in February 2025. The core idea: instead of writing code line by line, you describe your intent, the AI generates an implementation, and you review and iterate. The developer's value shifts from syntax to judgment: knowing what to build, catching what the model got wrong, and directing the next iteration.
Vibe coding sits on a spectrum. On one end, non-technical founders use tools like Lovable to generate entire apps from a text prompt with no code review at all. On the other end, experienced engineers use Cursor or Claude Code to dramatically accelerate their own work, reviewing and modifying AI-generated code as they go. The term now covers both approaches, though the underlying skill level and output quality differ substantially.
In 2026, vibe coding is mainstream enough that '138 tools are now on the market' as one comparison notes. The bigger question for builders has shifted from 'should I vibe code?' to 'which tool fits which task?' App builders like Lovable and Bolt.new are best for non-developers shipping MVPs. AI-native IDEs like Cursor and Claude Code are better for developers working on complex or production codebases.