AI Twitter's Favourite Lie: Everyone Wants to Be a Developer

Twitter's latest consensus on inevitability: now that large language models can write code, everyone will become a software developer. People, you see, have problems, and software solves problems, and AI removes the barrier between people and software, therefore everyone will build their own software.

It's a syllogism, after a fashion, but its premise is so wildly disconnected from how actual humans behave that it borders on fantasy. Because the average punter does not want to build software. They don't want to prompt software. They don't want to describe software. They don't particularly want to think about software.

They want to tap, swipe and scroll with zero friction and next-to-zero cognitive input. They want their problems to go away, and they would very much prefer if that happened without them having to open a terminal, a chat window, or anything else that reminds them of work.

This is damn-near universally applicable.

MOST PEOPLE ARE NOT TINKERERS

There's a deep assumption embedded in the "everyone will build" thesis, that most people are latent creators held back only by technical barriers. Remove the barriers, and creation floods forth.

But we've run this before. Desktop publishing tools became accessible in the 1980s with the Macintosh and PageMaker. Did everyone start designing their own newsletters? A handful did, and the rest continued to hire designers or, more commonly, didn't make newsletters at all.

WordPress has made it trivially easy to build a website for over twenty years now, and the vast majority of small business owners still pay someone else to do it, or they use a template and never touch it again.

The people excited about vibe coding are, almost by definition, people who were already interested in building things, and they're projecting their own enthusiasm onto a general population that has repeatedly demonstrated a preference for buying solutions over building them.

And why wouldn't they prefer that? Building things is cognitively expensive, whether or not it's financially viable. And even when the technical barrier falls to zero, the conceptualisation barrier remains. You still have to know what you want, specify it clearly, evaluate whether what you got is what you wanted, and iterate if it's not. That's work—effort—and for most people it is accompanied by functionally zero dopamine.

THE SPEC PROBLEM NOBODY TALKS ABOUT

An old joke: the hardest part of building software is figuring out what the software should do. This has been true for decades, and AI hasn't changed it. If anything, AI has made the problem more visible.

When the bottleneck was writing code, you could blame the difficulty of programming for why your project never got off the ground. Now that an AI can write code in seconds, the bottleneck is clearly, embarrassingly, you—me—us.

This is the part that the AI manics keep skating past. They demo an app built in ten minutes and declare that software development has been democratized. But the demo is always something with a clear spec: a to-do list, a calculator, a simple game with obvious rules.

The rest of the world's problems don't come pre-decomposed into clean specifications. The rest of the world may not even be able to fully articulate what's broken and what they want fixed.

WHAT PEOPLE ACTUALLY WANT

Most folks don't want to build a custom CRM. I do! I might! I couldn't be more excited about what this era unlocks. But I am not most people. They want to sign up for one that works. They don't want to create their own budgeting app. They want Mint or YNAB to do the job.

The entire SaaS economy exists as proof that people will pay monthly fees to avoid having to build or even configure things themselves. And is there anything wrong with that preference? The division of labor exists for good reasons, and Adam Smith figured this out in 1776 and he was a good deal smarter than a good many of us.

What people will actually do with AI is use AI-enhanced versions of existing products, with smarter search and better autocomplete inside the tools they already have. The revolution won't look like a hundred million people vibe coding custom apps. It'll look like existing software getting better at understanding what users want and doing it for them, which is what good software has always tried to do.

The tech industry has a long history of confusing what power users want with what everyone wants. The folks on AI Twitter who are building apps every weekend with Claude and GPT are having a great time, and the tools they're using are the same ones I'm obsessing over most of my waking hours. But we are a self-selected sample of tinkerers and builders, and the conclusions they're drawing about the general population say more about their own relationship with technology than about anyone else's.

Most people, given a magic wand, would not wish for the ability to write software. They'd wish for their software to work properly without them having to do fuck-all.