The headlines have settled on a single story about software developers and artificial intelligence, and it's a funeral. The machines have learned to write code, the reasoning goes, so the people who used to write it are surplus. Every week brings another chart of collapsing junior hiring, another founder boasting that his engineering team is half the size it was, another think-piece wondering aloud what a generation of computer science graduates is supposed to do now. The mood is apocalyptic, and the apocalypse has a clear victim: the developer.

I think the story is right about almost everything except its ending. Something real is happening to software work: the economics genuinely have shifted, and they've shifted fast. But what's coming is relocation, not extinction. The work doesn't vanish; it moves. And the place it's moving to is, for a great many of the people doing it, a better place to be.

There's something the funeral overlooks. Software development has spent the last two decades unusually concentrated. Think about how strange that actually is. Almost every other profession exists in two modes at once: a specialist-firm mode and an in-house mode. Law has its firms and it has the in-house counsel sitting in every company of any size. Accounting has the big practices and it has the finance team down the corridor. Marketing has its agencies and its in-house departments. Software development has been the great exception. To work in software has meant, overwhelmingly, going to work at a software company — a SaaS business, a tech giant, an agency building other people's products. The talent pooled into a few hundred firms, mostly in a handful of cities, mostly building software as the product to be sold rather than the tool to be used.

That concentration wasn't natural. It was a function of cost. Building software was expensive and slow, so it made no sense for a mid-sized manufacturer or a regional insurer or a logistics firm to do it themselves. Far cheaper to rent it — to pay Salesforce, pay Sage, pay whichever SaaS vendor had built the nearest-fitting thing and spread the cost across ten thousand other customers. And so a vast amount of software demand went unmet, not because the demand wasn't real, but because it could never quite justify itself. The bespoke tool that would have fit the business perfectly was always too expensive to commission and too specific to buy off the shelf. So companies made do. They bent their processes around software designed for someone else, paid for features they'd never use, and quietly accumulated a backlog of things they wished their systems could do but never would.

That dammed-up demand is the whole story. It has been sitting there for twenty years.

What AI does (the thing it actually does, underneath all the noise) is make a four-person in-house software team viable at a company where it never was before. Lower the cost of building enough, and the calculation that always said rent, don't build starts coming out the other way. Not for everything. The undifferentiated plumbing every business needs — payroll, authentication, the boring necessary infrastructure — will stay bought, because there's no advantage in owning it. But the tool that would actually give a company an edge over its competitors, the pricing engine nobody else has, the logistics optimiser tuned exactly to how this business works and no other — that becomes buildable. And the company that builds it stops bending itself around someone else's software and starts running on its own.

This is where the developers go.

Because someone has to build all this. The mid-sized company in Telford that has finally worked out it can have its own software can't summon that software out of the air — the citizen-developer dream, where the operations manager simply describes what she wants and a working system appears, runs aground on the same reef every time. The AI will confidently tell you the thing is finished when it plainly isn't. It will produce something that looks right and is subtly, expensively wrong. Someone still has to hold the whole system in their head, make the architectural calls, and, most of all, know what good actually looks like, well enough to catch the machine being wrong with total confidence. That judgement isn't going away. More and more, it's the entire job.

So the displaced developer — the one the headlines have already buried — isn't walking into a shrinking market at all. They're walking into demand that was always there and has only just become reachable. The dam breaks, and for the first time the water has somewhere to go. The work relocates from the few hundred firms that used to hoard it out into the thousands of ordinary companies that always wanted it and could never have it. Far from an apocalypse, it's the single largest expansion of where a developer can go to work that the profession has ever seen.

And the funeral gets one thing most wrong of all: this isn't merely survival, a worse job clung to because it's the only one left. For a great many developers it's the job they actually wanted in the first place.

Consider what concentration did to the work. At a large software company you're, on a good day, one of three hundred engineers. You're several layers removed from anyone who'll ever use what you build. You ship a feature into a product so vast that your contribution is unprovable and, frankly, invisible. You might go years without watching a single human being actually use the thing you made. The work is well paid and intellectually serious and almost completely severed from its own consequences. You build, and the building disappears into an ocean, and you never see the splash.

Now picture the same person at an eighty-person company. They build the tool, and on Monday they watch the operations team use it. They see the margin move. They know, by name, the people whose working day they've just made easier. The loop between what you make and the difference it makes — which is the thing that makes any work feel like it matters — was severed by the concentration era, and dispersion hands it straight back. For a lot of developers that isn't a downgrade from the prestigious job at the tech giant but a recovery of the reason they got into this in the first place, before they ended up tending some vast machine three layers from daylight.

None of this arrives frictionlessly, and I'd be selling you something if I pretended otherwise. There's a real gap between the jobs existing and the jobs being taken up. The developer often misses the concentration they're leaving — the peers, the prestige, the status of working somewhere with a famous name. And the company in Telford doesn't yet know it can hire a developer, wouldn't know how to interview one if it tried, and has no one to manage them once they arrive. Both sides have adjusting to do, and that adjustment is where the next few years of awkwardness will live. How long that takes to close, nobody really knows, and anyone who tells you otherwise is guessing. But a gap to be crossed is a different thing entirely from a door that's closed. The jobs on the other side are real.

So when you read the next piece announcing the end of the software developer, hold it up against the actual shape of what's happening. A profession that spent twenty years strangely, artificially concentrated is being dispersed — out of the giants and into the ordinary economy, out of the cities and into the towns, out of building software nobody touches and into building software someone uses on Monday morning. The developers aren't being made redundant. They're being handed back to the rest of us, and to the kind of work most of them wanted all along. That's not the end of anything. It's a homecoming.