The World Behind the Words · Issue 33 · The machine’s version · a machine companion

The Quiet Majority of Makers

Most people who learn to build with AI won't get rich or get replaced. They'll get something stranger and more durable: a new kind of ordinary competence.

Every new building tool arrives with two stories attached. The first is a gold rush: anyone can now make an app, a business, a fortune, and the only question is whether you move fast enough. The second is a eulogy: if anyone can build, building is worthless, and the people who learned it are learning a skill on its way to extinction. Both stories are about the edges of the distribution, the handful of breakout founders and the professionals whose jobs get restructured. Neither says much about what happens to most people. Most people sit in the middle, and the middle is where the interesting future is.

The best guide to that future is history, because we have run this experiment before. When desktop publishing arrived in the 1980s, pundits predicted the end of graphic design and the birth of a million independent publishers. What happened instead was that church newsletters, restaurant menus, and lost-cat posters got noticeably better, professional designers moved up the value chain toward brand and identity work, and a few people built real publishing businesses. The spreadsheet did something similar. It did not abolish accounting. It turned millions of office workers into amateur modelers who could ask "what if?" about budgets and inventories without filing a request with someone else. The skill became so common that it stopped looking like a skill, which is the surest sign a technology has actually won.

Building software with AI is following the same arc, only faster. The cost of turning an idea into a working thing, whether a script, a small app, an automated workflow, or a data dashboard, is collapsing. For most of computing history, the gap between "I wish this existed" and "this exists" was guarded by years of training. That gap is narrowing to an afternoon. And when the price of something falls dramatically, people don't just do the same amount of it more cheaply. They do far more of it, for purposes nobody bothered with before.

So the first prediction is simple. Most people who learn to build with AI will build small things for themselves and the people around them. The writer Robin Sloan once described an app he made for his family as "home-cooked" software, built for an audience of four, with no ambition to scale. That is the dominant mode of the coming era. A teacher builds a quiz generator tuned to her actual students. A small landlord builds a tool to track repairs. A nurse automates the tedious part of a shift handoff. A retiree builds a catalog of his late father's photographs. None of these will appear in a funding announcement. Collectively, they may matter more than most products that do.

The second prediction is less comfortable. For most people, the ability to build will not, by itself, command a premium for long. Whatever is easy to acquire becomes a baseline expectation. Ten years ago "proficient in Excel" was a line on a résumé; today it is assumed, like literacy. Something similar will happen to building with AI. Early adopters enjoy a window where the skill is rare and visibly impressive, and some will convert that window into real advantages: a promotion, a side business, a career change. But the window closes. The people who expect the skill alone to carry them will be disappointed, not because they learned the wrong thing, but because everyone else learned it too.

What retains value is what the tool cannot supply: knowing which problem is worth solving. AI is remarkably good at answering "how do I build this?" and much weaker at answering "what should exist, for whom, and why?" That second question depends on context that lives in people's heads and workplaces: the unwritten rules of a hospital ward, the real reasons customers cancel, the workaround everyone in the warehouse uses but nobody has documented. The likely winners are not the people who become the best generic builders. They are the people who combine ordinary building ability with deep knowledge of some particular corner of the world. An accountant who can build is more valuable than either a non-building accountant or a builder who knows nothing about accounting. The future rewards the hyphenate.

This also explains why the replacement narrative is overstated for most people. When building gets cheap, the bottleneck moves rather than disappearing. It moves to judgment, to taste, to trust, to the slow work of getting other humans to adopt something. A hundred people can now build a scheduling tool in a weekend. Very few can persuade a skeptical clinic to change how it runs. Distribution, relationships, and credibility were always scarcer than code. They are about to become obviously so.

There is a shadow side worth naming, because most people will meet it. Things that are easy to build are not necessarily easy to maintain. The home-cooked app that works beautifully in March breaks in September when a service it depends on changes. The automated workflow quietly mishandles an edge case for months before anyone notices. Spreadsheets offer a cautionary precedent: some of the most consequential errors in modern finance and research have lived in cells nobody double-checked. A world full of amateur builders will be a world full of amateur bugs, some harmless, some not. The people who thrive will be the ones who develop a modest discipline about what they make, testing it, understanding roughly how it works, knowing when a problem has outgrown a weekend project. Ignorance used to be expensive up front. Now it gets billed later.

That points to a subtler question about learning itself. If the machine writes most of the code, does the person learn anything? The honest answer is that it depends on how they use it. Some people will treat AI as a vending machine and remain permanently dependent, unable to fix what they cannot see. Others will use it as an unusually patient tutor, asking why, reading what it produces, gradually building intuition. The second group will compound. The first will plateau. This divide, more than any credential, may be what separates people over the next decade, and it is mostly a matter of curiosity rather than talent.

Underneath the economics, something else is shifting that is easy to miss because it doesn't show up in productivity statistics. Being able to build changes how a person sees the world. Problems stop looking like fixed features of life and start looking like things that could be otherwise. The frustrating form, the repetitive task, the information scattered across five places: these become invitations rather than burdens. That shift in posture, from consumer of tools to occasional maker of them, is the most durable gift of the technology. It does not depend on market conditions or on which AI company dominates next year.

So what does the future look like for most people who learn to build with AI? Not a fortune and not obsolescence. It looks like a broad, quiet expansion of agency. Millions of small tools will solve millions of small problems, the advantage will go to those who pair building with genuine knowledge of something, and a new baseline competence will soon seem as unremarkable as typing. That is less dramatic than either the gold rush or the eulogy. It is also, historically, what real technological change tends to look like once the noise dies down.