Vibe coding is programming with a language model without reading the code: you describe in natural language what you want, accept what the model writes and, if something breaks, paste in the error message and try again. The term was coined by Andrej Karpathy, OpenAI cofounder and former head of AI at Tesla, in a post of 2 February 2025: a new kind of coding in which you "fully give in to the vibes, embrace exponentials, and forget that the code even exists" (Karpathy, 2025). He presented it as fit for throwaway weekend projects, accepting every change without reading the diffs. Collins Dictionary made it its word of the year for 2025.
The phrase caught on because it named a new experience: people who cannot program build working software by talking, and expert programmers hand whole parts of their work to coding agents that use tools (tool use). For the hyperos (doomer) it is the democratization of programming; for many engineers, a fast new way of accumulating technical debt.
Critics distinguish between programming with AI assistance and abdicating understanding of the code. Simon Willison warned that vibe coding one's way into a production codebase is clearly risky, because most engineering work consists in evolving existing systems whose quality and understandability matter. A randomized controlled trial by METR with experienced open-source developers found that, with early-2025 AI tools, they took 19% longer, even though they believed they had been 20% faster (Becker et al., 2025). As Santelli (2026) summarizes, Narayanan and Kapoor argue that the core of engineering, deciding what to do and delivering products, is not automated even when execution speeds up. The underlying question concerns knowledge and responsibility: what does it mean to be the author of, and answer for, an artifact nobody has read?