A post on a developers' forum went viral last month with a title blunt enough to stop the scroll: brutally honest advice to freshers. A senior engineer described handing the kind of work that used to fill an intern's whole sprint — the small tickets, the test fixes, the tidy-ups — to an AI assistant, and watching it come back in about an hour. The conclusion he offered new graduates was two words long: there's no demand for you.

The internet argued about it for days, as it does. Defenders of the market pointed out that juniors are still cheaper than seniors, that new grads bring energy and current skills, that every technology panic in living memory has ended with more jobs, not fewer. All reasonable. But the interesting thing about the thread isn't whether one senior developer is right or exaggerating. It's that the largest datasets we have say the same thing he said — just in the language of payroll records instead of forum posts.

The Sector Isn't Shrinking. The Door Is.

The biggest payroll-data study yet run on this question — millions of workers, tracked month by month through the AI adoption wave — found no mass displacement of jobs overall. What it found instead is narrower and stranger: workers aged roughly twenty-two to twenty-five in the occupations most exposed to AI are now nearly a fifth below where their employment would be if it had merely kept pace with their peers. And the gap is still widening, quarter after quarter.

Here's the detail that matters most: it isn't layoffs. Young workers aren't being pushed out — they're not being let in. The entire effect runs through hiring. Experienced workers in those same occupations, sitting next to the same AI tools, show no comparable decline at all. The machine isn't taking the jobs. It's standing in the doorway where the junior jobs used to be.

The study has one more distinction worth carrying around, because it predicts where the gaps appear. The decline concentrates in work where AI is used to substitute for a human task. Where AI is used to complement a worker — to make someone experienced faster rather than to replace the task entirely — employment is flat or growing, and the growth favours the experienced. Same technology, two doors: one closes, one opens. Which side of that line a role sits on has become the most important fact about it.

Zoom into a city and the picture sharpens. In New York — America's biggest tech market — job postings for entry-level tech roles have fallen by almost half in three years, the steepest drop of any career family. Total tech employment in the same city? Flat. The industry is the same size. The on-ramp is half the size.

And the squeeze compounds at the bottom. Training providers who spent years building routes into tech for people without degrees report the employers who used to hire their graduates by the dozen have cut junior intake to a near standstill. Where a junior posting does survive, it now attracts Ivy League graduates and laid-off workers with two or three years of experience bidding down into it. The competition for the remaining rungs is fiercer precisely because there are fewer of them — and because everyone displaced from somewhere else is standing on the same step.

The Rung Was Always Made of Junior Work

Look at what an entry-level role actually was, and the pattern stops being mysterious. Junior positions existed to do the small, bounded, learnable tasks — and doing them was how you stopped being junior. The work was the training. Unfortunately, "small, bounded, well-specified tasks" is also a precise description of what current AI does best. The bottom rung of the ladder was made of exactly the material the machines learned to eat first.

The forum argument surfaced the uncomfortable arithmetic underneath. Companies didn't hire juniors because juniors were productive — they hired them as an investment in future seniors. If the justification for the investment was "someone has to do the small tasks," and the small tasks now take a senior an hour, the investment case has to be rebuilt from scratch. As one commenter in the thread pointed out, an industry that stops hiring juniors is eating its own seed corn — there are no senior engineers in ten years if nobody becomes one now. Everyone seems to know this. Nobody's balance sheet seems to care.

This is worth sitting with, because it reframes whose problem it is. For the individual graduate it's a closed door. For the industry it's a slow-motion shortage: the senior talent everyone is competing for in 2036 is the junior nobody hired in 2026. Companies are making a rational choice quarter by quarter that adds up to an irrational one decade by decade. But "the industry will regret this" is cold comfort when you're the one holding the degree and the empty inbox.

The Rung Moved — It Didn't Vanish

Buried in the same data is the part that almost never makes the headlines. While entry-level postings collapsed in the most AI-exposed occupations, postings for entry-level roles with minimal AI exposure actually grew. And across the market, junior listings that specifically ask for AI skills are up by a fifth since the decline began. Demand for new workers hasn't died — it changed shape.

Read that again, because it inverts the doom headline. The employers cutting junior roles fastest are the same ones posting new junior roles with different requirements. They're not exiting the market for young workers; they're repricing what a young worker needs to walk in with. The bar moved up and sideways at once — toward exactly the fluency the doom posts tell graduates is pointless to build.

So the honest picture is neither "everything's fine" nor "it's hopeless." The old transaction — show up with potential, learn on the small tasks, climb — is genuinely broken. What's replaced it is a market that still hires juniors, but increasingly wants them to arrive already carrying the evidence that used to be accumulated on the job. The rung is still there. It's just been moved to a different wall, and nobody handed out a memo.

There's a quieter injustice folded into this that deserves naming. The generation now hitting the wall did everything in the agreed order — the degree, the projects, the portfolio, sometimes the loans the forum post specifically warned against. The deal they were sold assumed the bottom rung existed. Nobody told them it was load-bearing on task work that a chatbot now drafts in minutes. They're not behind. They're early to a transition the institutions haven't admitted to yet.

That's the strategic problem every graduate and career-changer now faces: how do you present senior-shaped proof when the junior roles that used to create it are half gone? It's not impossible — people are doing it — but it isn't guessable either. It means knowing which parts of the market are still hiring, what the surviving listings are actually asking for, and how to build a candidacy that answers a job description written for someone more experienced than you officially are.

It also means unlearning the default playbook at the exact moment everyone else is running it harder. When the front door narrows, the crowd at the front door grows — thousands of near-identical graduate CVs chasing a halved pool of listings, each optimised by the same advice and increasingly by the same AI. The candidates who get through aren't necessarily better qualified. They're the ones who stopped queueing at the narrowest door and found the entrances the crowd hadn't noticed.

Climbing a Ladder That Moved

The viral thread's author was right about the mechanism and wrong about the conclusion. There isn't "zero demand" — there's demand that moved somewhere harder to see, with requirements nobody explains to newcomers. Navigating that shift is now a skill in itself, and it's learnable like any other.

The mistake worth avoiding is the passive one: waiting for entry-level to recover as if this were a normal downturn. The data doesn't show a downturn — total employment in the affected sectors is stable or growing. What changed is structural: the tasks that justified junior roles are automated, so the roles were redesigned rather than paused. Waiting for them to come back in their old shape is waiting for the pre-AI market to return, and nothing in the trend lines suggests it will.

Our Jobsearch Strategies for the Digital Age course is built for exactly this market: seven days, one focused lesson a day in your inbox, on how hiring actually works now — the ghost listings, the screening layers, the changed entry door — and how to build a search that produces interviews instead of silence. It doesn't promise the old ladder back. It shows you the current one.

The ladder didn't break. The bottom rung was sawn off and reattached somewhere else — and the map to where it went isn't secret, it's just unevenly distributed. The people getting hired are the ones who found out where.

Learn where the on-ramp moved →