The Canaries

Researchers at Stanford published a paper last month with a title that has stayed with me: “Canaries in the Coal Mine?” Using payroll data from millions of workers, they looked at what’s been happening to employment in jobs most exposed to AI.

The finding, in short: for experienced workers in those jobs, employment has held steady or grown. For young workers - people in their early twenties, just starting out - it has dropped noticeably. Among the youngest software developers, by their estimate, employment has fallen by roughly a fifth from its peak in late 2022.

Last year I wrote about the K-shaped developer market: the top arm rising, the bottom arm falling. I wrote it as an observation. This paper reads like a measurement. And measurements have a way of making things feel less abstract.

The view from the top of the K

I’m on the top arm. I’ve been programming for twenty-five years, and these tools have made my work more productive and, honestly, more enjoyable than it’s been in a long time. Most of the senior developers I know say the same.

It would be easy to stop there. To say: the market rewards experience, experience is what I have, the tools amplify it, good for us.

But I keep thinking about how I got the experience. Someone hired me when I didn’t know much. Someone reviewed my clumsy code and explained, patiently, why it was clumsy. Someone gave me a project that was a little too big for me and trusted me to grow into it. None of that was efficient. All of it was generous.

The question this paper raises, for people like me, isn’t economic. It’s moral. What do we owe the people who are where we once were?

What the canaries are telling us

Canaries were carried into mines because they were more sensitive than miners. When the canary struggled, it was a warning about the air everyone was breathing.

Young workers may be the canaries here. What happens to them now is a preview of something that affects everyone: an industry that stops creating experienced people eventually runs out of them. The seniors of 2035 are the juniors of today. If we don’t make room for them, the top arm of the K will eventually have no one to hold it up.

I wrote about this last year as a practical problem for companies. I now think it’s more than that. It’s a question of what kind of profession we want to be.

What I’m trying to do

I don’t have a grand solution. But I’ve been trying to act on a few small things, and I’ll share them in case they’re useful:

Make time for people starting out. A conversation, a code review, an introduction. It costs an hour and can change a trajectory.

When I’m in a position to influence hiring, argue for at least one early-career person on the team, with real mentorship attached.

Teach the slow skills, not just the fast ones. How to read a system. How to debug without guessing. How to ask what problem you’re really solving. These are what AI doesn’t hand you, and what turns a junior into a senior.

Be honest with young developers about the market, without crushing them. It’s hard right now. It’s not their fault. And the people who learn deeply, care about their craft, and find good mentors will still find their way.

A thought from tradition

Pirkei Avot says: in a place where there are no people, strive to be a person. I’ve always read it as a call to step up when others won’t.

The market may not reward hiring and training young people right now. The incentives point elsewhere. That’s exactly when it matters that some of us do it anyway - not because a spreadsheet says so, but because someone once did it for us.

The canary is not the problem. It’s the warning.