
I.
I suspect the younger generations must have committed some unforgivable sin in their past lives. And it’s come to haunt them. Otherwise, how do you explain their bad luck: A financial recession we still haven’t recovered from, the COVID pandemic that turned their most valuable years into a forced lockdown, no kids, no sex, no love, no friends, a housing crunch that nudges them to stay with their parents, a rampant epidemic of anxiety and depression enabled by an invention they didn't ask for and can't escape, the uncertainty of AI that dazes and worries them in equal measure, another Trump term—about which I have no comments—and now it turns out that not only do jobs not pay a subsistence income, but there aren’t any.
Derek Thompson, a sharp journalist known for diagnosing the most urgent modern issues—from anti-socialization to abundance to AI—is the one who might offer us a better hypothesis, because if this bleak outlook is not the misdeed of some otherworldly force, then I'm out of ideas. Let’s see how he answers the big, big question: Why has employment among young university graduates in the US dropped so drastically, to the point that they're worse off than the average?
I’m going to skim right past the first two causes he proposes—not because they’re unimportant or wrong, but because I have nothing to add. The first is that the culprits are the Great Recession and COVID: what we’re seeing is the continuation of a trend we’ve been watching for over a decade. The second is that college degrees just aren’t worth as much in the job market anymore; the investment in learning no longer guarantees a solid return. Both hypotheses are plausible.
The third cause—the one we’ll focus on and which, I suspect, Thompson also believes deserves more attention—is that AI is to blame: the world is already being reshaped by this technology (just a little), and it has decided to start, of course, by screwing over fresh graduates.
Blaming AI for all our problems is practically a meme. Tariffs are about to wreck global trade and drag us into a new recession because the interns used ChatGPT to come up with the formula. Yesterday, I fell down the stairs on the fourth floor because I was talking to ChatGPT and he didn't warn me there was a step. Your neighbor, the artist, hasn’t landed a commission in a decade because ChatGPT launched two years ago. Your forever sexless marriage is falling apart too, thanks to the “spiritual fantasies” ChatGPT sparked in your partner.
Now, young people are out of work and suddenly it's AI that's taking their jobs? What happened to “AI won’t steal your job”? But Thompson, data-driven as always, has arguments to spare and an ominous-looking chart to convince us:
The strong interpretation of this graph is that it’s exactly what one would expect to see if firms replaced young workers with machines. As law firms leaned on AI for more paralegal work, and consulting firms realized that five 22-year-olds with ChatGPT could do the work of 20 recent grads, and tech firms turned over their software programming to a handful of superstars working with AI co-pilots, the entry level of America’s white-collar economy would contract.
That reading fits what Harvard economist David Deming tells Thompson:
When you think from first principles about what generative AI can do, and what jobs it can replace, it’s the kind of things that young college grads have done. They read and synthesize information and data. They produce reports and presentations.
Thompson admits that all three hypotheses hold weight, and that the state of youth unemployment among the highly educated is probably a hybrid effect. If he spends more time exploring the AI hypothesis, it’s only because there’s far less written about it (or maybe because he’s a bit partial to it—not that I’d blame him). He also acknowledges, to his credit, that other sources offer a conflicting take: maybe AI is not even one factor among many, but actually not playing any role (yet). There’s no sign in the data of any “supercharged productivity growth” that AI would be driving if the hypothesis were true (I agree, I’ve written about this), and the trend seems to extend back well into the early 2010s.
But I’ll venture to say, playing devil’s advocate against myself, that the reason AI doesn’t show up in the data might come down to two factors that help avoid prematurely discarding the hypothesis. I’ll just name them here. First, AI is as much a productivity tool (better and faster work) as it is a distraction engine (“aww, look how cute is this image that ChatGPT made, I’m going to spend the next 4 hours ghiblifying my gallery”). Maybe AI doesn’t show up in productivity charts because one use case cancels out the other on average, leaving a net zero. Second, innovation takes time to be adopted and diffuse, whether it’s AI, the internet, mobile phones, or computers. There’s a delay between when AI starts replacing workers and when the productivity gains begin to show up in the metrics.
However, Thompson finds, doing his due diligence, that there’s more reason to doubt the hypothesis beyond its absence from economic indicators: “a New York Fed survey of firms released last year found that AI was having a negligible effect on hiring.”
If AI isn’t slowing down hiring, then there’s not much more to say. AI is not the culprit. Of course, there’s still the question of how they reached that conclusion: when was the survey conducted, where, and in which sectors? A survey done in May 2025 in San Francisco in the administration sector might yield very different results from one done in early 2024 in New York’s manufacturing industry. Still, the data doesn’t offer much support, and there’s little use in inventing untestable scenarios. We’ll have to wait.
II.
But since this is an AI blog and I’m still young—and even though I’m loyal enough to the truth not to inflate an idea we haven’t confirmed—I’m going to dig a bit deeper into the effect AI might have on my young fellas. Maybe Thompson hasn’t found enough evidence to confirm his preferred theory, but I think his intuition is solid.
So I will end this post with a reflection: Our young graduates face not one but two AI-related handicaps that, unfortunately, amplify one another, which deepens one of the great tragedies of the modern world: the disempowerment of the young generations. In a world where labor might eventually be fully replaceable by capital, it’s sad to see these promising young lads lose the little power of negotiation they may have had.
Since I don’t want to prod that wound by speaking to you through your parents or teachers or employers, I’ll speak to you directly. I believe you're more than capable of listening, reasoning, and acting accordingly.
The first handicap is that many of you, who in the past would’ve put in the effort to learn diligently (whether willingly or not), are now handing that work over to AI.
It’s easy to blame you or hold you accountable—and I get why your teachers do—but I think that’s only partly fair. What would we have done, we self-righteous adults, in your place? Are we saints, somehow? Didn’t we also, whenever we could, try to dodge homework? You’ve got other worries, and given the declining value of a degree (yes, they still matter, don’t kill me!), maybe those worries are more urgent in a way. It’s not ideal that your education gets spoiled, but it’s just as wrong for society to use a double standard just because the problem at hand is twice as serious.
So, as a matter of principle, I’m not opposed to you choosing your own path. Society should trust its adult students as autonomous beings capable of making their own decisions, right or wrong. But I do think it’s worth analyzing—if not outright addressing—the extent to which each of you might be developing an unhealthy dependency on this tool. Like any other, it can be incredibly useful, but also a source of drawbacks disguised as shortcuts. Leaving gaps in your basic knowledge is a key ingredient in the recipe for professional failure. It’s not a life sentence, but it is a long-term handicap.
If your teachers keep sounding the alarm, it’s because it’s hard to remain impassive while an entire generation in your care decides not to become contributing players in this collective game we call society. You have rights, but you also have duties.
Let’s get to the second handicap and I’ll wrap up (you’ve got studying to do): You will run into an AI that’s already mature, already capable of doing exactly the kind of low-level work that you, still without job experience, would normally be doing.
If a task is easy and economically worthwhile, AI will go for it first, pushed by the forces of the market and a tendency to grab the low-hanging fruit. That’s what Thompson fears, and what Deming thinks is inevitable. Maybe no AI can handle all the tasks (they’re unreliable, non-corporeal, etc.), but that remaining small percentage could be done by someone who already holds a stable place in the professional world.
That’s been a long-time worry of mine: what will companies choose, to hire the same number of interns and boost productivity, or to hire fewer and keep it steady? In sectors where demand can’t be scaled at will, the second option makes more sense. (In time, automation will win, new jobs will likely appear, and old ones will be redefined, closing this hiring gap—but I’m deeply concerned about the transition.)
It’s fair to point out that the first handicap is up to you, and that, in fact, if you use AI wisely—to learn more instead of less—you could turn that unfavorable imbalance into an advantage. (Again: are you using o3 to explore topics you don’t understand, or 4o to make Ghibli pictures?) But the truth is, the social forces tossing you around usually operate far beyond your still-forming sense of the world. I think back to myself at 18: a clueless kid worried about popularity and girls, with no clearer view of the future than “I’ll do whatever seems hardest.” Working life was just something that would eventually arrive, no rush, and I didn’t feel like I had any real say in it. “Use AI or not?” I would’ve thought. “I’ll do whatever everyone else does.”
I know paternalism won’t help, but neither will leaving you to fend for yourselves. What does help is proper instruction aligned with the world you’re about to step into. Every high school should implement an introduction to AI and a guide on how to navigate the job market. It’s hard—the lessons that matter shift every month, there’s no regularly updated manual, and, well, it’s something we’re all figuring out at the same time—but it being a formidable challenge doesn’t make it any less important.
That’s why I don’t say this is self-harm (although some want to frame it that way), but a tragedy. I can’t judge you by a higher standard than the one we old people held ourselves to. So don’t take this as a sermon or a scolding. It’s a warning. A call to action. For those of you waiting on a deus ex machina, know this: the world hasn’t been fully automated, nor has a universal basic income been put in place—so I’m afraid no one’s coming down to save you.
I feel for you, but I also trust that you’ll find your way.
Still, I can’t help but wonder what unforgivable sin someone must have committed in some past life to have such bad luck in this one. On behalf of a society that welcomes you with such insufficient hospitality: I’m sorry.
So it's the old Cath-22, supercharged by AI: You can't get a job without experience and you can't get experience without a job. The new twist is that AI can do the entry job that you would have started on to get the experience to get the "proper" job. There is no way around this gap :-( . AI will take away the bottom rungs on the ladder. Eventually more humans will be needed to replace those retiring, but they will be taken from the PhDs - not because a PhD is necessary but just because they are the best available.
Nice article. I think of this in terms of time vs tokens: more business owners and hiring managers will push tasks to (AI) tokens instead of paying for people's time.
For those who skill levels will justify having their time paid for - might end up being a good few years until AI agents are universal...