Major tech companies like Meta and Microsoft are cutting entry level software positions to fund AI development, with developers aged 22 25 down nearly 20% from their 2022 peak. While AI tools like GitHub Copilot increase junior developer productivity by 26%, they simultaneously…
On April 24, Meta told 8,000 employees they would be gone by May 20. Microsoft confirmed it was offering voluntary buyouts to about 7 percent of its U.S. workforce — a first for the 51-year-old company. Both announcements cited the same imperative: fund the AI build-out by cutting the cost base that AI is making redundant. The first roles AI is making redundant are the entry-level software jobs that have historically trained the next generation of senior engineers. CNBC
A 2026 Motion Recruitment study found that AI adoption is slowing hiring for entry-level and generalized IT roles while demand for AI-specific positions grows. The numbers behind that finding are three years old but have not improved: software developers aged 22 to 25 are down nearly 20 percent from their late-2022 peak, while developers 30 and older in the same roles grew 6 to 12 percent over the same period.
The paradox runs deeper than simple replacement. A controlled field experiment at MIT tested GitHub Copilot across developers at Microsoft, Accenture, and one additional firm. Developers given access to Copilot completed 26.08 percent more tasks than a control group. The gains were largest for the least experienced developers. If junior developers are simultaneously more productive and less employed, the economics are doing something more complicated than displacement. One interpretation: AI tools are making senior developers so capable that one senior developer with Copilot can now do work that previously required a junior, a mid-level, and a senior reviewing both. The junior rung does not disappear because juniors are less productive. It disappears because the unit economics of the team change when the senior slot becomes a force multiplier.
On February 17, Microsoft Azure CTO Mark Russinovich, writing with VP Scott Hanselman in Communications of the ACM, published a peer-reviewed analysis arguing the profession is in the early stages of a structural crisis of its own making. The paper cited the employment data as evidence the damage is already underway. It also documented examples from their own work with frontier coding agents: an agent handling a race condition by inserting a sleep call, a fix that masks the underlying bug rather than solving it. An experienced engineer catches this immediately. A junior developer might not. Agents that claim success despite significant code bugs, duplicate logic across codebases, dismiss crashes as irrelevant, and implement special-case hacks that pass tests but fail in production. The judgment to catch these failures is exactly what early-career developers are supposed to develop through hands-on production work. Russinovich and Hanselman call this quality systems taste: the acquired instinct for how systems actually behave under load. If AI handles the struggle, the junior developer ships clean code today and enters a mid-level role without the underlying knowledge that makes them reliably self-sufficient tomorrow.
A two-year longitudinal study conducted by researchers at SINTEF and the University of Oslo at NAV IT, a large Norwegian public-sector organization, tracked 26,317 non-merge commits across 703 repositories. They compared developers who adopted Copilot with those who did not. Their finding: no statistically significant change in commit-based activity for Copilot users after adoption. Developers reported feeling more productive. Their actual output, measured by commits, did not change. The gap between perception and measurement may reflect something genuine: AI tools reduce friction, boost confidence, and make unfamiliar tasks feel manageable. That is not the same as more code shipped.
Whether the junior developer pipeline problem has a solution is an open question. Russinovich and Hanselman propose a preceptor model borrowed from medical education: pair early-career developers with experienced mentors in real product teams, with learning as an explicit organizational goal rather than a byproduct of shipping. Russinovich confirmed Microsoft is piloting such programs internally. Whether corporate incentive structures will sustain them is another question. Charity Majors, CTO of Honeycomb, noted on X that every place she has seen start hiring junior engineers in recent years, that decision was led and lobbied for by senior engineers rather than driven by bottom-line economics.
The second-order question surfaces over years rather than months. Today's cohort of 22-to-25-year-old developers who are not getting hired are not just losing three years of salary. They are losing the production code, mentorship, failure modes, and institutional knowledge that historically turned a computer science graduate into a senior engineer. If that cohort is structurally smaller, the pipeline for senior engineers five to eight years from now is also structurally smaller.
Whether that gap matters depends on a question nobody has answered yet: whether AI tools are also changing what senior engineers need to know, or whether the tacit knowledge junior developers historically accumulated through struggle is still necessary, just accumulated differently. If the answer is the former, the disruption is transitional. If the answer is the latter, the industry is drawing down on a knowledge base it has not yet learned to replenish.
One finding from the Stanford team is stranger than it first appears. Erik Brynjolfsson, who directed the Digital Economy Lab research, and his co-authors used AI to process and clean the ADP data and to search literature. They used AI to help with some of the writing, as co-author Bharat Chandar put it in a TIME interview. The researchers who documented AI's displacement of entry-level workers used AI to do their own entry-level work.
The tool that made junior developers look like seniors has not yet answered whether those juniors are becoming seniors.