Cutting junior developers is a loan against 2036, and the bill is coming

Who will be the senior engineer in 2036 if nobody is allowed to be a junior in 2026? It's an uncomfortable question, and most hiring plans quietly avoid it.

The trigger is a t3n piece arguing that as AI use in software development grows, companies hire fewer career starters, and that this will become a problem. The data behind that worry is not new. A Stanford study of US payroll records, as summarized by Verve Consulting, found that employment of 22- to 25-year-olds in highly AI-exposed jobs such as software development has fallen by up to 20 percent since late 2022. Another write-up of the updated study puts the gap at 19 percent.

The saving is real, and that's what makes it dangerous

Let's be fair to the CFOs. If a coding assistant handles the boilerplate, the tickets and the small fixes that once justified a junior's salary, cutting that role looks rational. Each quarter, the spreadsheet agrees.

But a junior was never only a cheap pair of hands. The role is how a company manufactures the people who can review architecture, smell a bad design and debug a production outage at 3 a.m. Those skills come from years of writing mediocre code, getting it reviewed and slowly becoming less mediocre. Skip the first step and you don't get seniors who were trained faster. You get no seniors.

This is a loan. The company books the savings now, and the repayment arrives around 2036, when it will need experienced engineers and find that the market, and its own teams, never grew them. We think that interest rate will be steep.

"Let someone else train them" is a free-rider strategy

The usual rebuttal is that experienced developers will always exist, so why pay for training? Because that logic only works if some other firm keeps hiring juniors. When everyone reasons this way, nobody does. It's a classic tragedy of the commons, and the pasture is the talent pool.

One write-up, from unternehmer.de, puts it bluntly: experienced developers keep their jobs while one in five newcomers aged 22 to 25 loses theirs. That is a loose reading of employment data, but the direction is clear. And it's a snapshot of today. It tells us nothing about who replaces those experienced developers when they retire, move into management or burn out. Our reading is that the current senior market probably looks healthy because it is still living off hires made before the entry-level squeeze began.

There's also a quality argument. AI-generated code still needs someone who can judge whether it's right. Review is a skill built by writing. If juniors never write, who will be competent enough to supervise the machine? Overseeing AI is arguably the most valuable senior skill of the next decade, and it depends on a pipeline that is being shut down.

What a sane company does instead

Not every firm should hire the way it used to. The junior job is changing, and pretending otherwise is nostalgia. But the answer isn't zero; it's redesign.

• Hire fewer juniors, but hire some, on purpose, with named mentors.
• Make the apprenticeship about reading, reviewing and debugging AI output, not just producing code.
• Treat junior headcount as an investment line with a ten-year horizon, not a cost line judged quarterly.

For aspiring developers, the advice is less comfortable: the entry door is narrower, so show you understand what the code does, not just that you can get a model to produce it. We'd also be cautious about reading too much into any single study. The figures cover the US, they measure employment in exposed occupations rather than proving AI is the sole cause, and forecasts about 2036 are exactly that, forecasts.

Verdict

The companies most proud of their AI-driven efficiency may be the ones most exposed in a decade. Cutting every entry-level role is a bet that tomorrow's seniors will appear from somewhere else. We wouldn't take that bet. Keep a pipeline, even a small one, and treat it as infrastructure. The firms that do will be the ones with people who understand their own software in 2036.