A Tale of Two Job Markets

Software engineering job postings have surged since mid-2025, particularly in AI-exposed industries, according to a Citadel Securities report, while a Stanford study shows actual employment among junior software engineers aged 22-25 has fallen roughly 16% relative to less exposed roles since late 2022. The divergence reflects that job postings are a forward-looking indicator of hiring intent, while employment data captures current workforce numbers, with both trends potentially coexisting as the pipeline of new openings fills even as the existing junior workforce shrinks. The decline in junior employment may stem from a mix of AI disruption, COVID-era overhiring corrections, and interest rate changes, with attribution to AI alone remaining unclear.

Vivek Haldar / A Tale of Two Job Markets It was the best of times. It was the worst of times. That, I think, is how you’d have to summarize the state of the software engineering job market right now. Everyone keeps saying AI is eating the entry-level rungs of the ladder. I said it back in early 2024 https://vivekhaldar.com/articles/llms-eat-entry-level-swes/ And to show just how confusing things are, I want to put two seemingly contradictory pieces of data side by side. Two Graphs, Two Narratives The first graph comes from a report from Citadel Securities https://www.citadelsecurities.com/news-and-insights/the-economics-of-intelligence/ that came out about a week ago. The author’s argument is that AI-exposed industries are not seeing headcount contraction. Software engineering job postings have been spiking since around mid-2025. He looks at another heavily AI-exposed industry — customer service — and sees the same upward trend in postings. The thesis: AI is creating jobs and accelerating new business formation, especially in industries exposed to AI. The second graph is from a paper by Brynjolfsson et al. https://digitaleconomy.stanford.edu/news/ai-and-labor-markets-what-we-know-and-dont-know/ , cited in the recently released Stanford AI Index 2026 https://hai.stanford.edu/ai-index/2026-ai-index-report — specifically Chapter 4 on the economy https://hai.stanford.edu/assets/files/ai index report 2026 chapter 4 economy.pdf . The Brynjolfsson paper looks at the number of people actually employed in these roles. And the picture is starkly different: among workers ages 22 to 25, employment in the most AI-exposed occupations has fallen roughly 16% relative to the least exposed. Junior software engineers, in particular, have fallen off a cliff since around late 2022 — with no comparable decline for mid-career and senior engineers. So which is it? Is AI creating jobs, or destroying entry-level jobs? The simplest reconciliation is that these two graphs are measuring different things. The Citadel chart counts job postings . That’s a forward-looking signal: how many roles are open, how many companies are willing to hire, what the demand side intends to do over the next few months. It’s not a snapshot of who is currently employed. The Brynjolfsson chart counts employed people . That’s a backward-looking signal: who actually has a job right now. The two can diverge for long stretches, especially in industries going through structural change. So both can be true at the same time. The pipeline of new openings can be filling up while the existing population of junior engineers has shrunk. One is the leading indicator. The other is the lagging one. The Causality Is Messy There’s a wrinkle, though. The Brynjolfsson authors issued an update to their paper https://digitaleconomy.stanford.edu/news/canaries-interest-rates-and-timinga-more-on-recent-drivers-of-employment-changes-for-young-workers/ where, after controlling for various confounders, they found the timing of the employment decline shifts to 2024 rather than late 2022 . And the causality gets a lot messier. You can’t pin it just on AI. You can’t rule out the usual culprits — ZIRP and massive hiring during COVID. The honest read on the data is: yes, junior software engineering employment is down. No, we don’t yet have a clean attribution to AI. Most likely it’s a stew — AI plus the COVID overhiring correction plus the rate environment. Ground Truth I talk to a lot of folks who are close to graduation, or have recently graduated, or are trying to break in. The vibe matches the Brynjolfsson chart, not the Citadel chart. The entry-level software engineering market just feels really tough right now. But the Citadel chart is forward-looking. So if you take it seriously, the tide may have already turned. The worst may be behind us. The Old Scripts Are Torn Up The framing I keep coming back to when I talk to people about this: there is plenty of opportunity. But all the old scripts have been torn up. For more than the last two decades — through the internet boom and into the cloud era — there were fairly stable scripts for a software engineering career. Stable job expectations. A stable ladder defining how to grow your skills and climb through the levels. A well-known script for what to study, what to major in, how to interview. AI, especially at the junior end, is completely rewriting what that ladder means. The well-understood scripts that an entire generation of new grads, engineering managers, and leaders have been running on no longer work. They have to be rewritten, by all of us, while we’re already running. So if you’re trying to start a career in software right now, or trying to figure out how to manage one, the right mental model is to… figure it out yourself, because nobody else has any idea either. There is plenty of opportunity. The old scripts just don’t work anymore. Write your own. All the best