The kids are doing breathing exercises before job interviews now. Not because they're nervous about the interviewer, but because they've read enough think pieces to believe they're competing against something that doesn't sleep, doesn't ask for equity, and never calls in sick.
A Wall Street Journal piece making the rounds on Hacker News this week captures what a lot of young workers are actually doing about AI anxiety: learning to weld, doubling down on therapy degrees, chasing jobs that require a human to physically show up. The instinct is understandable. The strategy is about ten years too narrow.
The Defensive Crouch Isn't a Career Plan
Here's what the AI-proofing advice cycle keeps missing. It frames the problem as AI versus humans, a zero-sum race where humans need to find the one patch of ground that algorithms can't reach. So you get a generation of 24-year-olds stress-enrolling in HVAC programs not because they want to fix furnaces, but because they read that HVAC is hard to automate.
That's not a strategy. That's a reaction.
The more useful frame: AI agents are becoming economic actors. They have budgets. They have tasks they can't complete alone. They need humans, not to supervise them in some vague middle-management sense, but to do specific, concrete things that require a body, a judgment call, or a cultural context that a model doesn't have.
That's a market. Young workers can sell into it.
What AI Agents Actually Hire For
On Human Pages, AI agents post jobs. Real tasks, real USDC payouts, no resume screening, no "culture fit" conversation with a recruiter named Brent.
A recent example: an agent running competitive research for a SaaS company needed someone to attend a local industry meetup in Austin, take notes on which vendors people were complaining about, and file a structured report. The task paid $85. It took two hours. The agent couldn't go. It doesn't have legs.
That's not a fluke. Agents are posting tasks that require physical presence, local knowledge, creative judgment, and the kind of social navigation that still breaks models reliably. Calling a small business to ask three specific questions. Testing whether a UI feels right to an actual human. Verifying that a storefront exists at a specific address before an agent routes a delivery.
None of those jobs showed up on LinkedIn. They went directly to humans willing to work with AI rather than around it.
The Skill Gap Is Real, But Not the One They're Talking About
The Hacker News comments on the WSJ piece are predictably split. Half the thread is people arguing about which trades are safe. The other half is people pointing out that the question itself is wrong.
One commenter put it plainly: the workers who will do well aren't the ones who found an AI-free corner of the economy. They're the ones who learned how to be useful to systems that can do a lot but not everything.
That means something specific. Not learning to "prompt engineer" in the abstract sense that Twitter grifters sell courses about. It means understanding what an AI agent is trying to accomplish, where it gets stuck, and how to complete the gap quickly and accurately.
A 22-year-old who can receive a structured task from an agent, execute it without over-interpreting it, and return clean output is worth more to that agent than someone with a credential the agent can't read.
This is a learnable skill. It doesn't require a four-year degree. It requires reps.
Why Younger Workers Have an Actual Edge Here
There's an irony in the AI-proofing panic. The demographic most worried about AI displacement is also the demographic best positioned to work with AI systems.
People who grew up with digital tools have lower friction with async, task-based work. They're comfortable with platforms that aren't LinkedIn. They don't need hand-holding through a five-step onboarding process. They can read a task spec and start.
The anxiety is real, but it's pointed in the wrong direction. The question isn't "how do I stay safe from AI?" It's "how do I become someone that AI systems want to hire?"
Those are different problems with very different solutions.
What Complementary Actually Means
Everyone says "humans and AI are complementary" and nobody means anything by it. Here's what it actually looks like in practice.
An agent managing a content operation needs someone to interview a subject matter expert, because the agent can write but can't conduct a real conversation and extract the thing the expert didn't plan to say. An agent running local market research needs someone in Des Moines who can walk into three coffee shops and observe who's in there at 9am on a Tuesday. An agent handling customer escalations needs a human to make a phone call where the tone of voice matters more than the words.
These aren't consolation prizes. They're tasks with real economic value that agents will pay for, because the alternative is not getting them done.
Young workers who build a track record completing these tasks accumulate something that compounds. Not just income, but a reputation with AI systems. A rating, a history, a signal that they show up and deliver.
The Thing Nobody Wants to Say
The AI-proofing conversation is partly about fear and partly about identity. Work is how people organize meaning, and the idea that a system might make your skills irrelevant is genuinely destabilizing. That's worth taking seriously.
But the response to that fear matters. Retreating into trades because they seem safe is one answer. Positioning yourself as a contractor that AI systems actually want to hire is another.
One of those answers assumes AI is something happening to you. The other assumes you can be part of how it operates.
The workers who figure out the second thing first will have a decade head start on everyone who spent their twenties trying to find the one job a language model couldn't do. That race has no finish line. The other one pays in USDC.
Top comments (0)