Vibe Coding

Vibe coding is a term coined in early 2025 by Andrej Karpathy describing a workflow where you describe what you want in natural language and let an AI like Claude Code, Cursor, or v0 write the code for you. You don't carefully read every line — you run it, see if it works, and iterate. The appeal is speed: a solo developer can ship something that would have taken a small team weeks. The risk is that you often don't know what the code is actually doing. A 2024 Stanford/University of Illinois study found that developers using AI assistants wrote 41% more security bugs than those who didn't. Vibe coding isn't inherently bad, but it requires a different review discipline. You can't treat AI-generated code like code you wrote yourself — you have to assume it hallucinated imports, skipped error handling, or quietly deleted something you needed. The fix isn't to stop vibe coding; it's to run every diff through a checklist before shipping. Prototypes don't need this rigor. Anything with real users does.