The term “vibe coding” has taken the internet since Andrej Kaparthy posted about it online:

There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper so I barely even touch the keyboard. I ask for the dumbest things like "decrease the padding on the sidebar by half" because I'm too lazy to find it. I "Accept All" always, I don't read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it. The code grows beyond my usual comprehension, I'd have to really read through it for a while. Sometimes the LLMs can't fix a bug so I just work around it or ask for random changes until it goes away. It's not too bad for throwaway weekend projects, but still quite amusing. I'm building a project or webapp, but it's not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.

Vibe coding is writing code guided primarily by intuition, aesthetics, or creative flow, rather than strict structure, best practices, or detailed planning. "It's amazing to see someone as smart as you embrace AI on this level. And yet, I have seen a few takes from devs who think it's a badge of honor to write everything themselves and not use AI, unaware someone using AI will replace them first and soon," someone posted in a reply to the former Open AI founding member. It happened to Johannes Gutenberg's moveable type printing press and it will happen to a VS Code fork that is AI-native. When technology or capability longly held by experts, the democratization arc has three stages:

1. “Non-experts” get access to new technology or capability

2. They begin building things they never could before, and “experts” rationalize why it is a bad idea

3. The non-expert new adopters begin hitting the ceiling of the capability, and experts say, “I told you so.”

We are at stage three for vibe coding.

A large portion of making great software is not related to code, and vibe version control, vibe refactoring, and vibe CI don't have the same joy as vibe coding. A response to this may be to stop the vibe coders and leave it to the professionals. There is too much bad code out there we don’t need more. The second response is to abstract everything away; and let superintelligence handle it. Finally, there is the third approach that I believe, is to teach people conceptual models that evolve over generations of builders.

The vibe coders are too legit to git. They are a new breed of builders, and many are our future. Have them learn the concept of version control systems before Git (or Subversion, Mercurial, etc). Evolving the conception model of version control systems before defaulting to Git may unlock new interaction paradigms in making software. It takes time for the medium to naturalize and evolve with the technology.

In his 1983 talk at the International Design Conference in Aspen, Steve Jobs spoke about how content takes time to evolve alongside new media. He pointed out that when television was first invented, the earliest programs were essentially just radio shows filmed with a camera. It took years before creators began to understand and embrace what television could uniquely do—how it could tell stories visually, how pacing and editing could shape narratives in ways radio never could. Jobs emphasized that this pattern—where new mediums at first imitate the old before maturing into something original—is a natural part of technological and creative evolution.

This insight offers a useful lens for looking at modern developer tools. Much of today’s development workflow still leans heavily on paradigms like Git and the command line—technologies born in an era of text-based terminals and minimal graphical interfaces. While incredibly functional, they often feel like early TV equivalents: tools designed around the constraints of a previous generation. Just as television had to grow out of radio’s shadow to become its medium, perhaps it’s time for developer tools to do the same—reimagining the workflow in ways that feel native to the capabilities and expectations of today.

Re-imagining version control

What does distributed version control look like for vibe coders? Git is a distributed version control system created in 2005 to manage changes to source code. At a high level, it allows multiple people to collaborate on the same project by tracking and managing every change made to the codebase. Git provides a robust system for branching, merging, and maintaining the histories of a project, which is essential for collaboration, experimentation, and rollback when things break. It’s the backbone of modern software development and underpins services like GitHub, GitLab, and Bitbucket. Fundamentally, Git helps developers move fast without breaking things—by creating a secure, traceable record of how software evolves.

For beginners—or even experienced developers coming from more visual or collaborative disciplines—Git can feel like a foreign language. The core concepts (like commits, branches, merges, and rebates) are abstract, and the mental model it impose is often rooted in a low-level, file-based, command-line-driven workflow. This can be intimidating. Worse, when errors occur (e.g., merge conflicts, detached HEAD states), the feedback is cryptic and unforgiving. Git assumes a level of cognitive overhead and fluency with terminal-based thinking that doesn’t match how most people are taught or how they naturally think about creation and iteration. For vibe coders, Git can feel like an unnecessary gatekeeper rather than a helpful collaborator.

With AI-native capabilities, what if version control wasn’t steps of snapshots, but a living and breathing sacred timeline of branches? You could rewind, remix, or fork ideas the way a musician samples beats or a writer edits drafts. Past branches could “breathe” into the present—offering suggestions or insights based on your previous experiments, not just preserving history but informing the future.

In this reimagined world, version control might look more like Marvel’s Sacred Timeline: a dynamic, canonical history that branches off when major ideas or experiments are pursued, but with a sense of narrative continuity. Some timelines might collapse or merge. Others might become “variants” you revisit later. The UI could be visual and spatial, showing how your project has evolved in a nonlinear, intuitive way—less like a file system log, and more like a creative journal or a memory palace. It would feel alive, collaborative, and forgiving—perfect for vibe coders and creative beginners who want to build with flow, not friction.

Progressive comprehension

I've worked across the spectrum of high abstraction (no-code) and no abstraction (pure code). The best software tools are auto-didactic. The best approach is to progressively increase their comprehension before they give up. Replit started with a mission to enable a billion software creators. In the beginning, that was building the most seamless online dev environment for people to begin learning to code. AI has advanced, but the majority of their users don't even code using the agent. However, people are still learning through the osmosis of making. Ask AI chat and look at the audit logs, people unknowingly learn concepts to advance what they can build. This was the same at Webflow. The users would know code as they no-code, whether they realized it or not.

When I first started coding, my friend Adam taught me Git in the Terminal. It wasn’t until several weeks later when I got the hang of the concept he told me about Git Tower, the GUI desktop version of Git. He wanted me to learn the conceptual model before using tools as a mystery vs. knowing how they worked. In today's world, it may look more like AI agents explaining the concept in a learnable way.

Vibe coding gives people better chances to make progress in their output before completely failing. These failures might be frustrations with getting their development environment set up, knowing how to implement the rest of the code, or deploying it. Amjad Masad at Replit calls this people falling into, “the pit of success.” Through vibe coding, they’re able to make it to the checkpoint to continue building without even knowing how they achieved it.

The majority of people quit making software when they can’t overcome bottlenecks. Vibe coding gives people more shots on goals to pursue their ideas, and that’s worth cultivating. I recall the moment I quit pursuing software engineering. It was when I learned the concept of unit testing. I refused to write code for my code just to get my code to work. I wanted to ship software to bring an idea to life, and that's what a lot of people want.

The craft of software is in a multi-generational shift right now. As we look back, we’ll laugh at the way we used to code. Why are we still talking to super-powerful laptops the same way as communicating to a giant mainframe computer and acting like we’re timesharing computing? The vibe coders are too legit to git.

To recap:

• Vibe coding encourages non-experts to enter a space they fearfully avoided, and that is positive.

• Code is only one aspect of shipping software. People must learn other concepts to ship secure and scalable software.

• Teach people conceptual models and anticipate what tooling looks like for the future, not the past.

• The atomic unit of software making is the workflow. Find the workflows and how it can be evolved with AI and tooling.

Making software is about maintaining a flow state to make your ideas come to life. Vibe coding fosters that. We should be keepers of the flow state for people to make their ideas, not keepers of archaic ways of doing things. Vibe coders want flow state to make their ideas, and they are too legit to git.

Hyperlinks + notes

A collection of references for this post, updates, and weekly reads.

• Steve Jobs: The Objects Of Our Life (1983)

• Double click: When coding becomes conversation by Carly Ayres

• When everyone can build anything

• The Opportunity Ahead of the AI App Builders

• Beyond Chat by Nikunj Kothari

• Play to Xcode: Congrats on the Play team shipping this incredible feature

• The role of comprehension in AI workflows