The era of personal software

Here we go again with the prophecy. "RIP Software Engineers." That half the tech industry will shut down because building software will stop being hard. It's announced, with funeral faces and fairground enthusiasm, by the same PowerPoint undertakers who every five years bury something—email, the office, cash, programming itself—and then stand staring at the corpse when it gets up and orders a latte with extra milk. Well, I have a feeling they're about to be wrong again.

The reasoning, I'll admit, has a deceptive elegance. If making software gets cheaper, fewer hands will be needed to make it. On top of that, if AI does the whole process, well, you tell me. Claude, I woke up feeling a bit lazy today, make me an app that makes my bed and brushes my teeth. Boom, done. Obviously.

Except for one inconvenient detail: every time this same question was asked of the history of technology, it answered exactly the opposite.

Put yourself in 1865. An English economist named William Stanley Jevons publishes a book with a disaster-novel title—The Coal Question—and inside he drops an observation that left everyone scratching their heads. James Watt's steam engines burned coal far better than the old Newcomen contraptions: more work per shovelful. The reasonable expectation was that England would use less coal. But it used far more.

Why? you might be wondering. Because by making useful coal cheaper, steam crept into places it couldn't fit before. More factories, more ships, more trains, more mines, a thousand places where until then it wasn't worth it. Each efficiency improvement didn't extinguish demand: it ignited it. We call this phenomenon Jevons' paradox today, and it's been proving itself right for a century and a half with a punctuality that's almost frightening.

I think the same thing will happen with software. We'd spent decades treating it like England treated coal before Watt: expensive, heavy, rationed. Building a system required large teams, months ahead, and a fortune. And what costs a fortune gets reused to exhaustion; I already told in the first part how they sawed us all down to fit the same iron bed. No need to repeat it.

What's new is that AI is doing to software what Watt did to steam: making it cheaper. And the prophets' conclusion—less software, fewer people—is, word for word, the one Jevons dismantled a century and a half ago. But I don't think there will be less software. There will be a flood. A tsunami. Demand won't shrink: it will overflow into every corner where until yesterday it wasn't worth writing a single line.

And there's the interesting part, the question that really matters: what gets built when building is cheap? Not more of the same. The opposite. Cheap coal didn't give us one giant steam engine for everyone; it gave us a thousand small machines, one for each workshop. Cheap software won't spawn an even bigger, more generic SaaS. It will give us what we could never afford: software tailored to each one of us.

To a company. To a team. To a single person. Software written for the specific case, for the specific quirk, for that workflow that only exists in this head and no other. What wasn't worth coding before now gets coded in an afternoon.

That's what I mean by the era of personal software. Personal like the personal computer was, and the echo is no coincidence: that took computing out of the refrigerated basement of four corporations and put it on anyone's desk. This will do the same, but with the act of programming. Software stops being a product you buy ready-made and becomes something woven around you. Like a suit. Like a habit. Today you can buy a sewing machine and with a few notions make yourself a suit. It might work for you, but it doesn't kill the tailor or the textile industry. Quite the opposite. I think you see where I'm going.

So no, don't prepare for a world with less software and fewer people making it. Prepare for the opposite: for more software than anyone has ever seen, more small software, close, tailored to measure. The prophets of the end will be proven wrong again, like every five years. Jevons already knew it in 1865: making a tool cheaper doesn't put out the fire. It fans it.