Vibe coding has become a dirty word

For most of 2022, we did not know what ChatGPT was. Nobody in our circles was talking about it. Today it is how we build.

Somewhere in between, "vibe coding" went from a phrase to an insult. In the builder spaces we spend time in, and especially among experienced developers, calling something vibe coded has become a way of saying it should not be taken seriously. We tag our own posts with it, so we wanted to think about why.

Some of it is earned

When the phrase first caught on, it meant building something by prompting a model and accepting what comes back without really looking at the code. For a weekend project, that is fine. For anything a person depends on, it is exactly what a seasoned developer has spent a career learning not to do, and they are right to say so.

We have also seen what earns the reaction: apps that look finished and fall over the first time a real person uses them, launches that never mention what was tested, and confidence that runs well ahead of the work. When a developer sees enough of that, the word starts to mean all of it.

Some of it is something else

Not all of the pushback is about quality. A lot of it reads like the reaction any profession has when the hardest-won part of the job gets cheaper. Someone spends years learning to do something by hand, a newcomer gets something that runs in an afternoon, and the natural response is to say the newcomer's version does not count. Sometimes that is true. Sometimes it is just uncomfortable.

We do not think this is extinction. Nothing we have seen says the need for people who understand how software actually works is going away. It looks a lot more like the resistance that shows up in any field when a tool changes how the work is done. The people pushing back are often right about the flaws. Where they go wrong is on whether it is coming.

What we think happens to the ones who opt out

This is not a blanket statement about traditional developers. Plenty of them have taken this on and are using it well, and some of the most careful thinking we have read about building with AI comes from them. This is about the ones who have not, why they might not have, and what we think happens if they keep choosing not to.

Today, if you are not actively investing in your own understanding of AI, and we do not just mean using it to get through your inbox faster, we think you are going to miss the boat, whatever you do for a living. That goes for operations, sales and design, and it goes for developers too. The developers who fight it will, at some point, miss that boat.

That is the unfortunate part, because they are the people with the most to bring.

Why what they know matters more now, not less

What makes an app built with AI hold up is not the prompt. It is the checks around it. In our own process, each step's work has to pass its checks before the next step starts. A part only goes into our library of reusable code after it has run in a live app and passed its tests. And the mistakes we have caught so far were caught by checks, not by luck.

Every one of those checks is something a seasoned developer already knows how to do: testing, review, knowing where a system breaks under real use, knowing what not to ship. We had to learn them. They already have them. Put that knowledge together with a tool that can write the first draft of almost anything, and that person is not at risk. That person is the one setting the standard.

Where we stand

We are builders, not traditional developers, and we do not claim to know what it feels like to watch a skill you spent years on get cheaper. What we do know is that a few years ago we had not heard of any of this, and today it is how we build everything. The developers who learn it on top of what they already know are going to be very hard to compete with.

If you are a developer who started skeptical and changed your mind, or who looked hard at it and decided against it, we would like to hear what did it.