In February 2025, AI researcher Andrej Karpathy described a new way of writing software: describe what you want in plain language, let an AI model generate the code, and largely stop reading it line by line. He called it "vibe coding." A little over a year later, it's not a novelty — it's close to the default way software gets built.
Where the Term Came From
Karpathy's original framing was casual almost to the point of being a joke: a workflow where a developer fully gives in to the vibes, embraces exponentials, and forgets that the code even exists. The phrase caught on fast. Collins Dictionary named it the Word of the Year for 2025, and Google Trends showed searches for the term jump roughly 2,400% since Karpathy first used it.
How Big This Actually Got
The adoption numbers by early 2026 are hard to overstate. 92% of US developers now use AI coding tools daily, and 82% do so weekly worldwide. GitHub reports that 46% of all new code is now AI-generated, and among Y Combinator's Winter 2025 startup cohort, 21% of companies have codebases that are 91% or more AI-generated. Even at a company the scale of Google, roughly a quarter of code is already AI-assisted. Forecasters expect the share to keep climbing — Gartner projects 60% of all new software code will be AI-generated by the end of 2026.
The money has followed the usage. The AI coding tools market is growing at roughly a 38% compound annual rate, well ahead of traditional developer tools at around 16%, and Cursor reportedly went from $1 billion to $2 billion in annualized revenue in just three months. Average deal size for AI coding startups jumped more than 70-fold between 2022 and the end of 2025.
What "Vibe Coding" Actually Means Today
The term has drifted a fair amount from Karpathy's original, looser definition. Most simply: you describe what you want in natural language, AI generates the code, and you accept it without fully reviewing every line — iterating by prompting rather than by typing code yourself.
But by 2026, that casual version has largely given way to something more disciplined. Vibe coding now more often means a developer writes natural-language specs and AI generates code under structured human oversight, with persistent project context and layered validation, rather than a developer surrendering all control and accepting output as-is. Developers describe it less as abandoning oversight entirely and more as a shift in role — the developer stays involved in specification, architectural review, and quality control, but expresses intent through precise prompts and conversational refinement rather than writing implementation code line by line.
One writer's description of how disciplined practitioners actually work captures this middle ground well: most developers who vibe code today aren't blindly accepting every line of output — they treat the AI like a very fast junior developer, giving it direction, letting it execute quickly, and checking the work afterward.
The Uncomfortable Gap: Speed Without Trust
The most striking statistic in the current data isn't the adoption rate — it's the trust gap sitting right next to it. While 92% of US developers use AI coding tools daily, only 29% actually trust the code those tools produce. That's not a contradiction so much as a description of where the industry actually is: teams have adopted the speed of vibe coding faster than they've built the process to verify what it produces.
One analysis put the underlying dynamic bluntly: testing practices haven't kept pace with how fast code generation has been automated — the differentiator in 2026 isn't whether a team uses AI tools, since virtually everyone does, but whether they actually test and verify what the AI produces before it reaches users.
Why This Matters Beyond Individual Developers
Vibe coding hasn't just accelerated how existing developers work — it's lowered the barrier to building software at all. The approach has opened software development to non-technical users, letting entrepreneurs go from an idea to a working application through prompting alone, without prior coding knowledge.
That accessibility has reshaped the competitive landscape for startups. Gartner has found that more than 70% of new SaaS categories now have 50 or more competitors — a direct consequence of the barrier to shipping software dropping so far. In that kind of crowded environment, the advantage shifts toward teams that can execute reliably, govern AI-assisted development, and keep their engineering process portable across tools — rather than teams that simply generated code the fastest.
The Risk Side of the Ledger
The core risk of vibe coding isn't that AI writes bad code — it's that unreviewed code, generated fast and often deployed straight to production, tends to accumulate the kinds of problems that only show up under real usage: hardcoded secrets, unhandled edge cases, and architecture that wasn't planned so much as arrived at. The tension isn't going away as adoption climbs — it scales with it, which is exactly why testing and review discipline is emerging as the real differentiator between teams that vibe code well and teams that just vibe code fast.
Top comments (0)