If you're early-career and it feels harder to get hired than it should, there's now a well-documented economic study behind that feeling, not just anecdotes.
Stanford's Digital Economy Lab, led by economist Erik Brynjolfsson, has been tracking this since 2025 in a paper nicknamed "Canaries in the Coal Mine." The August 2026 update found employment for workers aged 22-25 in the most AI-exposed occupations is now 19% below comparable less-exposed fields — up from a 13% gap measured just a year earlier. The gap isn't closing. It's widening, and accelerating.
What's actually driving it
The researchers are specific about the mechanism, and it matters: this isn't mass layoffs. It's a collapse in new hiring. Companies aren't firing junior employees in these roles — they're just not opening as many junior positions in the first place, because AI tools are increasingly capable of doing the "automative" version of entry-level work directly, rather than merely assisting a human doing it.
That distinction — automation versus augmentation — turned out to predict the pattern well. Occupations where AI mostly automates a task (software development among them) show the sharpest entry-level declines. Occupations where AI mostly augments a human's existing work show a much murkier, less consistent picture.
A separate data point adds texture: U.S. Bureau of Labor Statistics figures show narrowly-classified "computer programmer" roles down 27.5% between 2023 and 2025, while the broader "software developer" category — which includes more senior, judgment-heavy work — barely moved, down just 0.3% over the same period. That's not the industry shrinking. It's the entry rung of the ladder getting narrower while the rest holds roughly steady.
The honest caveats
The Stanford authors are careful about their own claims, and worth being equally careful repeating them. Their language is explicit: the results are "consistent with the hypothesis" that generative AI is affecting entry-level employment — correlational, not a proven causal chain, and they acknowledge other factors (interest rates, post-pandemic hiring corrections) plausibly explain part of the earlier data, with the AI-attributable signal becoming clearly significant mainly from 2024 onward. This is a serious, replicated, stress-tested study — not a hot take — but "consistent with" and "proven to cause" are different claims, and the researchers themselves don't blur that line.
What this means if you're trying to break in right now
The skills gap isn't closing the door — it's raising what gets you through it. The entry-level work most exposed to automation (routine implementation of well-specified tasks) is precisely the kind of work AI already handles reasonably well. What's becoming scarcer is demand for juniors who only do that. What's still in demand: junior developers who can verify, debug, and catch what an agent gets wrong — the exact skill that's easiest to skip building if you lean on AI from day one.
See where your own habits land, before the market decides for you →
Source: Brynjolfsson, Chandar & Chen, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence," Stanford Digital Economy Lab, updated August 2026; U.S. Bureau of Labor Statistics occupational employment data, 2023-2025.
Related reading: Junior Developers and AI Dependency · Are PMs Replacing Developers?