AI and the Future

It's the 1920s. New York City.

AT&T — the largest employer in America — is about to automate an entire workforce out of existence. The weapon: mechanical switching. Rotary dials. Automated routing. No humans required.

By 1920, telephone operation was the single largest occupation for young American women under 20. Hundreds of thousands of them sat at switchboards, connecting calls by hand, cord by cord. Think of it as a manual mobile money agent for phone lines — no wireless, no cell towers, just copper wire. You picked up the phone, spoke to a human, and she physically plugged your line into someone else's socket.

Then the machines arrived.

The layoffs were big enough to trigger Department of Labor reports and Congressional hearings. Unions fought. Cities protested. By 1940, mechanical switching had replaced operators across more than half the U.S. network.

Here's the part that rarely makes it into the story: AT&T wasn't just modernizing for fun. Call volume was growing so fast that, without automation, they calculated they'd eventually need to employ every young woman in America just to keep the phones running. It was survival, not vanity.

The machines won. New jobs did eventually appear — but the women who lived through the transition mostly didn't get them. Many never recovered financially.

That was over a hundred years ago. The story hasn't changed. Only the technology has.

We've always built tools to make ourselves lazier — and the labor market has always paid for it.

Horses → engines. Looms → factories. Punch cards → keyboards → the cloud. Mail carriers, taxi drivers, bank tellers. Even chess, supposedly the last bastion of human intuition, fell to Deep Blue in 1997.

The pattern repeats: people resist, unions protest, politicians debate — and the economy moves on anyway.

So where does that leave those of us building software right now?

I won't pretend the anxiety isn't real. But it's worth being honest about what the data actually shows in 2026, because the picture is messier — and more hopeful — than "the bots are coming for everyone equally."

AI coding tools are genuinely good at scaffolding, boilerplate, and well-scoped tasks. That's real, and it's already changing what a junior engineer's day looks like. Entry-level hiring at major tech companies is down sharply from a few years ago, and young engineers in AI-exposed roles are feeling real pressure. That part of the fear is legitimate.

But the wholesale "software engineers are finished" narrative doesn't hold up against the broader evidence. Experienced engineers — the ones who can decide what to build, judge whether AI output is actually correct, and own systems end-to-end — are, if anything, in higher demand, and senior roles now make up most of the growth in engineering job postings. Long-term projections still point to growth in the field, not collapse. The honest read: this is a transformation of the job, hitting the bottom rungs hardest, not a clean replacement of the profession — at least not yet. Reasonable, well-informed people, including some of the researchers who built this technology, disagree sharply about how far and how fast that changes.

And here's what's different for those of us building from Africa.

The car reached Europe in the 1880s and most of Africa a century later. The internet hit the West in the early '90s; meaningful access here took another 20-25 years. Every revolution arrives on a delay, and by the time it lands, the wealth is captured and the standards are set elsewhere.

AI might move faster because mobile has leapfrogged some old barriers. But the majority of Africans still eat at the small hotel down the road, still ride the matatu, still rely on a human to answer the phone or stamp the form. That's not "behind" — it's a different reality. And very little AI research is written with that reality in mind: unreliable power, expensive data, and a generation of young African developers just entering an industry that Silicon Valley papers are already declaring obsolete.

That's not paranoia. That's history repeating with a new machine.

So what do I think we should do?

Embrace the technology — genuinely. Don't be the horse that refused the engine. But receive it critically. Most frameworks published about AI's future were never designed with our power grids, our data costs, or our politics in mind. That absence isn't usually malicious. It's just costly if we don't notice it.

Unreliable power. Expensive data. Complex politics. Those aren't our weaknesses — they're our design constraints. The most dangerous thing an African builder can do right now is mistake someone else's defaults for universal truth.

The technology is coming. That much is certain. The question is whether we shape how it arrives — or just receive it, like we've received everything else.

I'm still figuring that out. But I think the conversation starts here.