vibe coding

Why 'vibe coding' suddenly took over

By Kai · Published July 16, 2026

Why it spiked: Writing code got cheap enough, and good enough, that describing the outcome beats typing the syntax for a growing share of everyday software.

ad slot, enabled after AdSense approval

A year ago, “vibe coding” was half a joke, you describe what you want in plain language, let the AI write the actual code, and mostly go with the vibes. Now it’s on conference stages, in job descriptions, and quietly behind a lot of small apps that ship every week. So why did a punchline become a workflow?

The 60-second version

Vibe coding means building software by describing it rather than hand-writing every line. You tell an AI what you want, it generates the code, you run it, and you steer with more description instead of editing syntax. The phrase was popularized by AI researcher Andrej Karpathy in early 2025, and it stuck because it named something people were already starting to do. The spike isn’t hype alone, three things changed underneath it.

Why it spiked now

1. The models got good enough to trust with a first draft

Early code assistants autocompleted a line. Newer ones scaffold a whole working feature, run it, read the error, and fix it. Once the AI can get you to something that runs on the first or second try, the center of gravity shifts: you spend your time describing and reviewing, not typing.

2. It opened coding to people who don’t code

The biggest unlock isn’t faster senior engineers, it’s everyone else. A designer, a marketer, a founder with an idea can now get a working prototype without learning a language first. That massively widens who can make software, and a wider door means a louder trend.

3. Shipping small things got absurdly cheap

When a weekend idea costs an afternoon instead of a month, people make far more of them. The volume of little tools, internal scripts, and one-off apps exploded, and each one is a small advertisement for the workflow that made it.

What it actually changes

The interesting shift is where the skill moves, not that it disappears. When generating code is cheap, the value moves to the parts AI is still bad at:

  • Knowing what to build: taste, product sense, and clear description become the bottleneck.
  • Judging what came back: you still need to tell working-but-fragile from actually-solid, especially for anything real users touch.
  • The unglamorous 20%: security, edge cases, and “what happens when this scales” don’t vibe themselves into existence.

So vibe coding doesn’t kill the craft. It moves it up a level, from how do I write this to is this the right thing, built well enough.

What to watch next

  • The maintenance bill. It’s easy to generate code you don’t understand. The first wave of “who maintains all this?” is coming.
  • Security fallout. Fast-shipped, lightly-reviewed apps are a fat target. Expect some loud lessons.
  • The split. Vibe coding is great for prototypes and small tools, and shakier for large, long-lived systems. Watch the line between the two get drawn more clearly.

The takeaway: vibe coding took over because it made the first draft nearly free. That doesn’t make judgment free, it makes it the whole job.