The Room Where Juniors Used to Watch

For most of the twentieth century, a university degree bought you one specific, reliable thing: your odds of being unemployed dropped below the national average. That was the entire economic argument for the debt and the years spent studying. It was not a guarantee of a good job. It was a guarantee that you would be less exposed than the average worker to not having one at all.

According to the Federal Reserve Bank of New York, that stopped being true. Unemployment among Americans aged 22 to 27 with a bachelor's degree sat at about 5.6 percent through the first half of 2026, above the national unemployment rate of roughly 4.3 percent. Underemployment for the same group, meaning graduates working jobs that do not require a degree, sat near 42 percent. For the first time in the relationship most people were sold on, the degree is now correlated with worse odds, not better ones.

The instinct is to blame this entirely on artificial intelligence, and there is real evidence pointing that direction. But a study published in May 2026 by researchers at the London School of Economics found that remote work predicts the collapse in entry-level hiring better than AI adoption does. That detail should slow you down, because it points somewhere more specific than "robots took the jobs." It points at what remote work removed, which was never really about location.

An entry-level job was never just a job in the ordinary sense of a task performed for a wage. It was a subsidized seat in a room. A company accepted output from a junior employee that was, for a while, genuinely worth less than what it paid for, in exchange for the junior slowly becoming worth more. The mechanism that made that exchange work was proximity. You sat near someone experienced. You watched them make a decision, watched them catch their own mistake before it shipped, overheard the reasoning behind a choice that would never have made it into any document. None of that was ever written down as training. It was ambient. It required a room.

Remote work did not eliminate entry-level jobs directly. It eliminated the room. A junior employee working from their bedroom is not absent from any meeting they are formally invited to, and is absent from everything they were never invited to but used to learn from anyway. Once that ambient transfer stops happening, the economic case for tolerating a slow, error-prone new employee starts to look different, because the thing that used to justify the cost, the promise that they would become someone who no longer needs supervision, has a much weaker mechanism behind it now. And that is exactly the moment cheap, competent AI tools arrive to make the remaining, narrower case even weaker: if a junior's early output can be approximated for a few cents by a tool that never needs a room in the first place, the seat that used to be subsidized on the promise of future value gets a lot harder to justify keeping open.

What is left in place of the rung is what researchers have started calling seniorization: entry-level job postings that still say entry-level, but now list the judgment, the stakeholder management, and the independent decision-making that used to be exactly what the job was supposed to teach you. You are being asked to already have graduated from a room that no longer exists. Nobody designed this outcome on purpose. It fell out of two unrelated shifts, remote work and cheap AI output, landing on the same generation at the same time and removing, from two different directions, the one thing that used to make a junior worth the risk of hiring at all.

Healthcare, cybersecurity, and the skilled trades have mostly kept hiring new graduates through this, and the reason is instructive rather than incidental. Those fields still require someone to be physically present while a more experienced person works, because the work itself cannot be fully remote and cannot yet be fully automated. The room survived wherever the nature of the job made it survive by accident. Everywhere else, it had to be rebuilt on purpose, and almost nobody has bothered, because rebuilding it does not show up on a quarterly earnings call the way a headcount reduction does.

I am part of that generation, and I do not think the honest response is despair, because despair does not get you a room back. It gets you an inventory of what a room did, so you can go build a substitute for yourself rather than wait for someone to reopen the one that closed. Look closely at my own résumé and you will find the same symptom the labor data describes, dressed up to look like progress. One of my current roles carries the title Chief Strategy Officer. It is unpaid. I gave myself, or rather the small platform I volunteer for gave me, a senior-sounding title because a senior-sounding title is now one of the few substitutes available for a paid, structured entry rung that no longer reliably exists. I do not think I am unusual in this. I think an entire generation is manufacturing titles, without saying so, to compensate for the absence of the room those titles used to be earned inside.

If the room will not reopen on its own, the honest move is to build a private version of it, and to be precise about what needs replacing. It is not the paycheck. It is the proximity to someone else's live reasoning. Reading a senior engineer's finished pull request teaches you what they decided. Watching them review someone else's, in real time, arguing about a tradeoff before it ships, teaches you how they decided, and that second thing is almost never for sale, but it is often available for free to anyone willing to sit in on it: open code reviews, public design critiques, recorded postmortems, the comment threads underneath a decision rather than just the decision itself. Treat every one of those you can find as the seat the market stopped subsidizing, and be explicit with yourself about what you are there to absorb, because nobody is going to hand it to you ambiently anymore.

The generation before mine got years of unstructured proximity to people who already knew what they were doing, and mostly did not notice they were being given something they could not have paid for, because it looked like nothing more than an ordinary office. We do not get the ambient version. We have to go find it on purpose, which is a worse deal in almost every way except one: at least now we know exactly what we are looking for.