The req that quietly disappeared
This spring I sat across from an engineering leader who had just closed his hiring plan for the year, and he was pleased with it. Senior engineers, staff engineers, a platform architect — and not a single junior. When I asked about it, the answer came without hesitation: the agents do that work now. He is not an outlier, and he is not being lazy. I build hiring software for a living, so this is the question I watch more closely than any other in the field right now, and his hiring plan is the field’s consensus answer. I think the consensus is walking into the most predictable talent crisis of the decade.
The same report proves both halves
Start with the good news, because it is real. TechCrunch ran the headline in June: engineering jobs are the most resilient in tech. The underlying data, from the VC firm SignalFire’s talent dataset, shows engineering hiring at the large tech companies down only 11 percent against 2019 while hiring overall fell 25 percent, and engineers now make up 55 percent of new hires at the majors, up from 46 percent. AI was supposed to shrink the profession; instead the profession absorbed AI and kept hiring.
Now read the same report’s other finding, the one that did not make the headline. New-grad hiring is down roughly 65 percent at those same companies against 2019, and 76 percent at early-stage startups. SignalFire’s own analysts put it more bluntly than I would dare: by eliminating its new-grad pipeline to optimize current balance sheets, the industry “could face a severe leadership vacuum over the next decade.” The Federal Reserve Bank of New York’s data adds the human side: recent computer science graduates run 7.0 percent unemployment, computer engineering 7.8 — among the worst outcomes of any major, in the profession everyone says is thriving.
So both stories are true at once. Engineering as a function is doing fine. The entry ramp into it has been closed, and the people standing at the bottom of it are the evidence.
The canaries, read honestly
The sharpest attempt to isolate what AI is doing here is Stanford’s Canaries in the Coal Mine study, built on payroll data from millions of workers: employment for 22-to-25-year-olds in the most AI-exposed occupations, software developers included, fell 13 percent relative to their older colleagues — a figure the November revision moved to 16 percent on data through October 2025. I want to be precise about its weight. It is a working paper, not peer-reviewed, and it is contested — critics point at interest rates and timing, and the authors’ own follow-up concedes the effect only turns statistically significant from 2024 under the strictest controls. Nobody can yet prove how much of the junior collapse is AI and how much is the hangover from the hiring binge of 2021.
Here is why I do not think the causation debate matters much for anyone running a team: Stanford’s payroll data, SignalFire’s hiring data, and the Fed’s graduate outcomes are three independent datasets pointing the same direction. Whether AI is the whole cause or a third of it, the pipeline is closing on your watch, and the response available to you is identical either way. Even Dario Amodei’s much-quoted prediction — that AI could eliminate half of entry-level white-collar jobs within five years — is best read not as a forecast to believe but as a claim from an interested party that your hiring plan is currently treating as settled fact.
The curriculum, not the chores
The mistake underneath all of this is a misreading of what junior work was for. A working paper from the Atlanta Fed, reviving Kenneth Arrow’s sixty-year-old economics of learning-by-doing, states it exactly: entry-level tasks are not merely low-value work — they are the curriculum through which workers build the capital that makes them productive later. The unglamorous tickets, the small bug fixes, the test coverage nobody wanted to write — that was never just cheap output. It was the syllabus that turned graduates into the seniors everyone is now fighting over.
AI genuinely did eat that syllabus, and each individual company’s response is locally rational. No single firm captures the return on training a junior who may leave in three years, so no single firm feels the cost of not training one. But an industry where every firm reasons this way is eating its seed corn, and the harvest it is spending was never this quarter’s — it was 2031’s. Charity Majors said it in one line: by not hiring and training junior engineers, we are cannibalizing our own future.
Junior is a role, not a skill level
The good news is that this is a choice, and some companies are visibly making the other one. Shopify grew its internship program from around a hundred to over a thousand people a year on the explicit thesis that graduates who went through school with AI are the first genuinely AI-native cohort. Coinbase rebuilt its entire interview loop to evaluate AI fluency at every stage. Neither company is running a charity. They noticed that AI removed the tasks juniors used to do, not the functions juniors serve — succession, fresh assumptions, an early bet on someone’s trajectory — and they redesigned the role around supervising and directing AI instead of deleting it.
I have argued before that in an AI-enabled organization a human owns every merge. That rule has a supply chain. The judgment doing the owning has to be trained somewhere, on real systems, with real stakes, under someone senior enough to correct it — and every one of those training grounds is a line in a hiring plan someone is currently cutting. So treat the apprenticeship the way disciplined teams treat technical debt: a boring, protected budget that survives good quarters and bad ones, with a named owner and a ratio you do not quietly trade away when the roadmap gets loud.
Everyone keeps asking what AI means for junior developers. I think that is the wrong direction to read the question. The junior developer question was never really about the juniors — it is about whether your organization still has a mechanism for producing the one thing AI has not replaced. The companies rehiring juniors in 2026 are not being sentimental. They are making sure that in 2031, the seniors are in their building.