The Truth About Vibe Coding
It feels faster than it is
A lot of people come to me saying that they vibe code their whole products now. Even our students. And I’m like, yes, I know, it f***ing sucks!
But that’s because they do it wrong. You can actually use agents to code much more efficiently without breaking stuff.
The creator of Claude Code just admitted that for the last few months, 100% of Claude Code was written by Claude Code itself.
And on the team that builds the tool, roughly 95% of everything they ship is written using Claude Code.
The tool builds itself. Humans review at the end. That is the state of coding in 2026.
The guy saying that is Boris Cherny. He runs Claude Code at Anthropic. And the first time I read that quote, I had to read it twice. Three years ago, a sentence like that would have sounded insane. Even a year ago, it would have still sounded insane. Today, nobody even flinched.
And if you’re an engineer, you already feel it. The thing we used to call coding is now popular as vibe coding.
So how did we get here? How did we go from opening Stack Overflow to solve one bug, to agents writing entire features on their own? That didn’t happen overnight, but in slow progress in about 5 years. And today, I want to walk you through the full story. The timeline, the tools, the tweets, the numbers, and the people, or even agents, who actually built this… Here’s the story behind “vibe coding”.
Watch the video now (or the written article here):
p.s. A video on how WE do “agentic coding” towards AI is coming in two weeks. Make sure to subscribe to the channel here to not miss it!
Anthropic's '95% of shipped code written by Claude Code' figure and the '45% AI-generated code is insecure' counter from the comments belong in the same sentence to be meaningful. If both numbers hold, Anthropic is either accepting elevated security risk as a controlled experiment or has internal review tooling the rest of us don't see. From the AI builder side at theaifounder.substack.com, my hunch is the gap is closed by internal evals not yet published, and outside teams replicating the 95% number will hit a security wall in production. Which review process do you think outside teams need before adopting that ratio, and have you seen any open implementations?
"Humans review at the end."
Except someone asked Anthropic employees if they vibe code everything. All hands went up.
Then he asked how many do NOT review the code.
Many hands remained up.
This when we know - and studies have proven - that AI generated code is messy, unmaintainable (especially if you haven't reviewed the code, you not only don't know how it does what it does, you lose your skill at doing so) and over 45% insecure.
On top of which, more and more companies are finding out it actually costs more than they thought.
And what the AI security crisis caused by vibe coding and agentic AI proves is what I knew all along: the entire edifice of "software engineering" is built on sand - because it's not "engineering". It's a "craft."
And vibe coding is even less so. Unreliable LLMs trained on unreliable human "amateur engineers" are producing crap.
This is a catastrophe waiting to happen.
And "AI influencers' are encouraging it.