I would like to point out two further research studies which look at different countries and also use different methodology, complementing Stanford's paper quite well: https://www.rivista.ai/wp-content/uploads/2025/10/ssrn-5425555.pdf and https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5516798 The latter paper I helped to work with looks at UK data and we use a triple difference estimator which shows that if you are holding the firm constant, even then, more exposed shops have seen lower hiring. I think I know that interest rates affect firms, not individuals, so we can discard the interest rate evidence, at least for the UK. I think HUMLUM et al. also only covers data until 2024. I also summarized the state of evidence regarding AI's impact on young workers in my newsletter. Maybe this might be of interest to you: https://windfalltrust.substack.com/p/brief-4-ais-2025-labor-market-impacts
"Autor and Thompson (2025) show that an occupation being made more productive by an emerging technology need not automatically suffer job losses." This is an important point. Much of the current discussion about AI's impact on employment reminds me of similar discussions about personal computers back in the late 1970s and early 1980s. While some types of jobs declined, others were simply transformed, as work was conducted using new tools and methods.
Excellent analysis and alternative theory for observed stalling in the labor market. That being said, some of the arguments you put forth do not rely solely on interest rate cycles. For instance, you present some claims regarding labor market tightness and impacts on early career unemployment during standard business cycles; however, GDP growth in the US has remained relatively robust during this period as well. How do you reconcile that observation against your theory?
Fantastic work challenging the AI displacement narrative with actuall timing evidence. The point about ChatGPT API only launching in March 2023 while job postings started dropping in early 2022 is exactly the kind of detail that gets lost in hot takes. What's interesting is how 'AI exposure' ends up being a proxy for interest rate sensitivity since those occupations cluster in tech and finance. Makes you wonder how many other macro patterns we'll misattribute to AI before really understanding what's happening.
I would like to point out two further research studies which look at different countries and also use different methodology, complementing Stanford's paper quite well: https://www.rivista.ai/wp-content/uploads/2025/10/ssrn-5425555.pdf and https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5516798 The latter paper I helped to work with looks at UK data and we use a triple difference estimator which shows that if you are holding the firm constant, even then, more exposed shops have seen lower hiring. I think I know that interest rates affect firms, not individuals, so we can discard the interest rate evidence, at least for the UK. I think HUMLUM et al. also only covers data until 2024. I also summarized the state of evidence regarding AI's impact on young workers in my newsletter. Maybe this might be of interest to you: https://windfalltrust.substack.com/p/brief-4-ais-2025-labor-market-impacts
"Autor and Thompson (2025) show that an occupation being made more productive by an emerging technology need not automatically suffer job losses." This is an important point. Much of the current discussion about AI's impact on employment reminds me of similar discussions about personal computers back in the late 1970s and early 1980s. While some types of jobs declined, others were simply transformed, as work was conducted using new tools and methods.
Check my claims on AI affecting job market:
https://open.substack.com/pub/thevcpivot/p/why-ai-isnt-the-real-reason-you-cant?r=77fc23&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
Excellent analysis and alternative theory for observed stalling in the labor market. That being said, some of the arguments you put forth do not rely solely on interest rate cycles. For instance, you present some claims regarding labor market tightness and impacts on early career unemployment during standard business cycles; however, GDP growth in the US has remained relatively robust during this period as well. How do you reconcile that observation against your theory?
Fantastic work challenging the AI displacement narrative with actuall timing evidence. The point about ChatGPT API only launching in March 2023 while job postings started dropping in early 2022 is exactly the kind of detail that gets lost in hot takes. What's interesting is how 'AI exposure' ends up being a proxy for interest rate sensitivity since those occupations cluster in tech and finance. Makes you wonder how many other macro patterns we'll misattribute to AI before really understanding what's happening.