AI will kill programming. SaaS is dead. We're all getting replaced. I've spent months, years even, reading the same old refrain, almost always signed by people who never wrote a line of code or set foot in the backroom of a company at eight in the morning, when the ERP won't boot and the invoices still have to go out. And no. Listen: I don't think that's how it goes.

I think AI won't kill software. I think it'll turn it inside out: how it's built, how it's sold, where it lives. And along the way it'll hand us back something we lost on the road, the trade. These days, with a Cloudflare tunnel and a few cheap boxes, the plain old datacenter makes sense again. The centralized cloud is starting to shed ballast. And generic software, the kind stitched for thousands and fitting almost no one, is about to give way to software made to measure.

Let me start with an idea that's been drifting through this industry in a low voice for two decades, so baked into how we work that we barely see it anymore. I was taught to abstract. To generalize, to hunt the pattern, to build the one piece that solves many problems at once. It sounded like a virtue. Mostly, it was the old arithmetic.

The thing looks like a no-brainer. A company has a problem and buys software. Managing customers, a CRM. Internal processes a shambles, an ERP. Support running slow, a platform. So obvious nobody stops to watch what happens right after. And right after, a silent negotiation begins. Not between the company and the vendor: between the company and itself.

Because software promises to solve a problem for you, yes, but it hands you a bill that isn't in the budget: learn to run the way I expect you to run. We christen it with catalog names (rollout, digital transformation, integration), but underneath, the deal was always the same. For decades, the company adapted to the software far more than the software adapted to the company.

The CRM brings its sales logic, and sales contorts to fit it. The ERP takes certain processes for granted, and operations reshuffles its flows around the machine. The booking tool assumes a way of treating the customer, and the business ends up treating them that way, even after forty years of doing it differently, and doing it well.

There's an old Greek myth that fits like a glove. Procrustes had an iron bed and a peculiar hospitality: the guest who didn't measure exactly the length of the bed, he'd stretch on the rack or saw off whatever stuck out. Enterprise software was, for twenty years, Procrustes' bed. And we, the engineers, were the smiths who forged it. The company went in whole and came out cut to the size of the cot.

And the worst part is it made sense. For a long time, all the sense in the world. Building software was expensive in a way that's hard to remember today: years of work, pricey teams, heavy infrastructure, endless cycles. This trade built a whole liturgy to wring the knowledge out of people and pack it into repeatable processes. And out of that scarcity came a conclusion hard as the bed itself: if building software costs a fortune, that software had better serve many. Thousands.

And so SaaS was born, and with it a whole era. One product, thousands of companies, all signing the same quiet surrender under their breath: this wasn't made for me, but it's what there is. We got so used to the workaround we stopped arguing about it. Whole industries bloomed to help us swallow it: consultancies, integrators, implementers, configuration experts, people whose trade was, look closely, teaching companies to resemble their software. From a step back, it's a very strange idea. We let the tool dictate how the organization should run. Not the other way round.

And this is where the bed starts to creak.

For the first time in ages, building software is ceasing to be expensive. I don't mean servers or clouds; that got cheap years ago. I mean the act itself: turning a concrete need into working code. That, the truly expensive part, the thing that justified the whole liturgy, is exactly what AI is dragging down in price. And when the cost of building falls, all the old arithmetic that ordered the industry falls with it.

So the question changes. For decades, generic software wasn't an aesthetic choice: it was an economic necessity, because tailoring a bespoke suit for every business was an unthinkable luxury. What if it stops being one? What happens the day we can afford to sew, again, to the exact measure of each body?

Because no two companies are alike, however much they look it from afar. Watch one long enough, the corner bar or a multinational, doesn't matter, and something surfaces: an operating personality, an internal language, quirks nobody ever wrote down, exceptions, decisions that on paper are nonsense and yet have worked beautifully for twenty years. All the stuff that makes that company that company and no other. Its fingerprint. Its signature.

Software always treated those oddities as what an engineer means by default: friction. Inconsistency. Noise to standardize. It took me years to understand, but those quirks were almost never inefficiencies. They were the business. Its concrete way of existing, its inner craft, its signature. And we spent two decades sawing it off to fit the bed.

We flipped the prize, too. Software rewarded whoever looked like everyone else: the more your business fit the mold, the better the tech treated you. That deal is over. Now it's possible to capture nuance, learn the exceptions, hold on to what's unique about a place and mold around it. And what used to stop you automating, your oddities, is turning into the opposite: an advantage, an operating identity.

Don't bury SaaS yet, it's not time. What's dying is the alibi. I think we're going to see the biggest software explosion in history, but it'll be a different beast. There'll still be generic product for what's genuinely generic, because nobody's going to hand-sew their own email. What's new is that, alongside it, there'll be plenty of room for software that's close, small, cut to the measure of the particularities that make each business unique. The tailor's suit, not the one off the rack.

And then, for the first time in this industry's very short history, technology will stop asking us to file down our bones to fit inside it, and it'll do what a good tailor does: take measurements, watch how we move, sew around us. And each tailor gets to leave their own mark, widening the range of what the end user can have.

At autarqui.co we're up to our elbows in that move. We're not after a future where every company looks a little more like the next; we defend the opposite. Software that keeps the useful oddity of each house (that quirk that shows up on no org chart and yet makes payroll at the end of the month), and that makes it, along the way, more productive and more its own master.

For too long, software charged us the same toll: you change, because I won't. That toll stops being paid. It's time to melt down Procrustes' bed. And with that iron, while we're at it, we'll sew a suit made to measure.