What is Vibe Coding? It’s Not About Turning Off Your Brain

The following is an excerpt from the forthcoming book Vibe Coding: Building Production-Grade Software With GenAI, Chat, Agents, and Beyond by Gene Kim and Steve Yegge.

Dr. Erik Meijer, a visionary Dutch computer scientist with a lifelong penchant for tie-dyed shirts, is one of the most influential figures in programming language development. Across decades, his lifetime of contributions have shaped how millions of developers write code every day, from his groundbreaking work on Visual Basic to his work on C#, Haskell, LINQ, and Hack. Few people on Earth can claim such deep expertise in language design and implementation. And yet, in 2024, Dr. Meijer gleefully made this striking and startling declaration:

The days of writing code by hand are coming to an end.

When we heard Dr. Meijer make this claim, we were both excited. It was one of the most important and validating confirmations of something we had started to suspect over the last year—that coding is changing right underneath us. So, why would such a prominent programming language pioneer make such a polarizing claim, one that implies that much of his life’s work would soon become obsolete? Because he sees what we see: AI shifts how humans create software.

We’re witnessing this transformation happen across the industry. At Adidas, seven hundred developers using AI coding tools reported a 50% increase in what they call “Happy Time”—hours spent on creative work they enjoy, rather than wrestling with brittle tests or debugging trivial errors. High-performing teams now spend 70% of their time directly coding, compared to 30% for teams using traditional methods.

Even more telling are the stories from developers who had left programming. A former machine learning engineer who hadn’t written code in nearly twenty years successfully built a calendar synchronization tool in her first session with AI assistance. Even Kent Beck, creator of Extreme Programming, excitedly shared how he’s “coding at 3am for the first time in decades!”

For decades, programming has meant laboriously typing code by hand, hunting down syntax errors, and spending countless hours on Stack Overflow. That era is ending. We are living through a fundamental shift in software development that is redefining how we code, who can code, and what is possible to build.

What we and Dr. Meijer saw now has a name: vibe coding. It was coined by the legendary Dr. Andrej Karpathy, who has been at the forefront of AI research for a decade, to describe a new way of programming. 

When we say vibe coding, we mean that you have AI write your code—you are no longer typing in code by hand (like a photographer going into a darkroom to manually develop their film.)

Although the most visible and glamorous part is code generation, AI helps with the whole software life cycle. AI becomes your partner in brainstorming architecture, researching solutions, implementing features, crafting tests, and hardening security. Vibe coding happens whenever you’re directing rather than typing and allow AI to shoulder the implementation while you focus on vision and verification. 

Let’s Be Precise: What is Vibe Coding?

Like any newfangled term, there’s a lot of disagreement and misinformation about what vibe coding is. Plenty of people and the media have painted it as “turning off your brain.” This is far from how the rest of the professional world is using it, however. Before we go any further, let’s get precise and define what we mean when we talk about vibe coding, agents, etc. 

When we refer to manual coding or traditional coding, we’re talking about pre-AI style software development, where you type in code by hand.

In 2021, we saw AI-generated code completions, where the IDE (integrated developer environment) would auto-complete code based on what you had typed (like your phone auto suggesting words as you text). GitHub Copilot pioneered this capability, and it is in almost every coding assistant product on the market today. This was shown to speed up some coding tasks by 50%, but coding is still labor-intensive work. 

Chat coding is one of the successors to code completions. Beginning in 2023, you could ask AI to examine and modify code or generate new code, and it would emit an answer. It may seem quaint now, but you had to copy the answer back into your IDE by hand. Over time, the tooling has become faster and more fluid, but chat is still a back-and-forth interaction. Whenever we say “chat,” we mean a conversation with AI unfolding one turn at a time. Many first discovered this style of coding with the release of OpenAI’s ChatGPT-4o in May 2024.

Agentic coding (where AI autonomously generates, refines, and manages code) appeared in early 2025, and is a game-changing step up from chat. In this workflow, coding agents act like real developers and actively solve problems using the tools and the environment. Agentic coding is increasingly predicted to replace a significant portion of coding by the end of 2026. 

Agentic coding had been long conjectured, and many of us were first exposed to it with the announcement from Cognition AI’s Devin, an autonomous AI assistant designed to collaborate with humans on software development tasks, in March 2024.8 However, it wasn’t until early 2025, with the release of Claude Code from Anthropic, that agentic coding took the developer world by storm. Claude Code is a terminal application that you interact with. You tell it what you want it to do, and it modifies files to implement. It can even run tests and execute programs (including mini utilities it builds for itself). 

With agentic coding, instead of AI telling you what to type, the agent makes the changes and uses the tools itself. This speeds the development cycle far more than you would expect.

If you’re in development today, you’ve probably been using AI and coding assistants already or have at least dabbled. The list of players in the space is long and includes a spectrum of offerings from chat to limited coding agents to extremely powerful autonomous coding agents (e.g., Aider.code, Augment Code, Anthropic Claude Code, Bolt.new, Cline, Amazon Q, Cursor, GitHub Copilot, Google Cloud Code Jules, JetBrains Junie, Loveable, OpenAI Codex, Repl.it, RooCode, Sourcegraph AMP, Tabnine, and Windsurf, to name a few).

These products make different choices about what to offer you, and where to offer it. Some are still mostly completions or chat. Some have limited agents. Some offer full-featured, semi-autonomous agentic coding assistants. Some support running many agents together. Some coding assistants live in your IDE, some are standalone IDEs themselves, and some are command-line tools. Some support complex enterprise environments, while others are geared more toward casual coders. Many coding assistants support multiple models, but some align themselves to a single model family for performance, reliability, or cost reasons.

So, in this mixed landscape of manual coding, chat coding, and agentic coding, let’s examine what vibe coding is and where it fits. 

You don’t have to “turn your brain off”—as many have wrongly implied. You will often be an active participant. Instead of writing the code yourself, you’re overseeing your AI-assistant doing it for you and critiquing its results.

We and many others have felt that, at times, you can be 10x more productive with vibe coding compared to manual coding. We know this sounds like hype—we were skeptical too. In Chapter 1, we’ll walk you through a detailed, real-world example of how Gene wrote over 4,000 lines of production code in just two days to help this book make its deadline.

And as Gene did early in the DevOps movement, we are both working on research to quantify the impacts of AI on development and on the conditions required for AI to create value, jointly working with Google’s DORA research group. We’ll talk more about this in Part 4 of this book. But it’s clear that vibe coding will be reshaping our work for decades to come.

So, What Are the Benefits of Vibe Coding?

Vibe coding lets you build things faster, be more ambitious about what you can build, build things more autonomously, have more fun, and explore more options. This is what we’re calling FAAFO (or sometimes “the good FAAFO,” to contrast it with certain other kinds). Let’s look at each in turn. 

First, vibe coding helps you write code faster. Tasks that once took months or weeks can now be done in a day. And tasks that took days can now be completed in hours. This acceleration comes not only from code generation but also from having AI help with debugging, testing, and documentation. Projects that have been sitting on the back burner for years due to lack of time can finally see the light of day.

Second, vibe coding enables you to be more ambitious about what you can build. It expands both ends of your project spectrum. It brings seemingly impossible projects within reach, while simultaneously making small tasks with marginal ROI easier to take on as well. This is due to the speed, vast knowledge, and capabilities of AI. Vibe coding reshapes your approach to development, eliminating many of the painful trade-offs that have always constrained what gets built.

Third, vibe coding allows you to do work autonomously, often being able to complete things that previously required multiple people or teams. That’s a bigger deal than it might sound. Features that once demanded specialists from multiple disciplines can now be handled by a single non-specialist developer with AI assistance. Being able to work autonomously or alone on a task or project eliminates two expensive taxes. It reduces the coordination costs (scheduling meetings, aligning priorities, waiting for availability) and the communication challenges (where teammates cannot read each other’s minds but must still create a shared goal and vision of what to build and how). Working more autonomously or alone with AI significantly reduces or removes these obstacles.

Fourth, vibe coding makes programming more fun. You’re spared from the least enjoyable parts of programming, such as debugging syntax errors, wrestling with unfamiliar libraries, or switching test infrastructure for the nth time. Instead, you can focus on solving user problems, building cool stuff, and getting things done. Working with AI is also strangely addictive, an aspect we explore in the book. You might be tempted to discount the fun dimension, but we think it’s one of the most valuable, because it’s bringing people out of retirement, attracting non-programmers, and encouraging leaders to take on more programming work. That’s a deep societal change in the works.

Finally—and this is possibly the most important and transformative dimension of all—vibe coding increases your ability to explore options, either to find a solution or to mitigate risks. Instead of committing to a single approach early on, you can rapidly prototype multiple ways to solve the problem and evaluate their trade-offs. We’ll revisit this topic often, so that when you recognize a problem where exploration will help, you’ll reflexively spin up parallel investigations. FAAFO!

Stay tuned for more exclusive excerpts from the upcoming book Vibe Coding: Building Production-Grade Software With GenAI, Chat, Agents, and Beyond by Gene Kim and Steve Yegge on this blog or by signing up for the IT Revolution newsletter.

- About The Authors

Gene Kim

Gene Kim has been studying high-performing technology organizations since 1999. He was the founder and CTO of Tripwire, Inc., an enterprise security software company, where he served for 13 years. His books have sold over 1 million copies—he is the WSJ bestselling author of Wiring the Winning Organization, The Unicorn Project, and co-author of The Phoenix Project, The DevOps Handbook, and the Shingo Publication Award-winning Accelerate. Since 2014, he has been the organizer of DevOps Enterprise Summit (now Enterprise Technology Leadership Summit), studying the technology transformations of large, complex organizations.

Follow Gene on Social Media

• 
• 

Steve Yegge

Steve Yegge is an American computer programmer and blogger known for writing about programming languages, productivity, and software culture for two decades. He has spent over thirty years in the industry, split evenly between dev and leadership roles, including nineteen years combined at Google and Amazon. Steve has written over a million lines of production code in a dozen languages, has helped build and launch many large production systems at big tech companies, has led multiple teams of up to 150 people, and has spent much of his career relentlessly focused on making himself and other developers faster and better. He is currently an Engineer at Sourcegraph working on AI coding assistants.

Follow Steve on Social Media

•