Workers aged 22 to 25 in the jobs most exposed to AI now have 19% fewer jobs than if they had kept pace with less-exposed peers, according to the August 2026 update of a Stanford Digital Economy Lab paper. The same measure stood at 15% when the study first came out, on July 2025 data.
The paper went viral again on Sunday. A post by Unusual Whales said employment of software developers aged 22 to 25 has fallen about 20% since November 2022, per the lab.
The authors call their findings “early, descriptive indicators”, not proof that AI caused the drop. Our own read of official payroll data shows the same direction, more mildly, as hiring data did earlier this month.
Business Pill · AI Agents
An evergreen Business Engineer explainer. It teaches the general idea only and does not discuss this report.
The key insight: Stanford’s August 2026 update puts workers aged 22-25 in AI-exposed jobs 19% behind less-exposed peers, up from 15% on July 2025 data, through less hiring rather than layoffs; BLS data shows computer systems design jobs down 5.2% since November 2022 against +3.3% for all US payrolls, by our arithmetic.
What the Update Says
The researchers, Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, use payroll data from ADP covering millions of US workers, now through June 2026.
Their first finding is that there is “no evidence of widespread, economy-wide job displacement.” The change is concentrated among young workers in exposed jobs.
In plain numbers, employment of 22- to 25-year-olds in the two most exposed groups of jobs fell about 11% from November 2022 to June 2026. In the three least exposed groups it grew about 10%.

From 15% to 19%
The first version, in August 2025, headlined a 13% relative decline. The authors now prefer a simpler measure that needs no modelling choices.
On that measure, the shortfall was 15% on July 2025 data and 19% on June 2026 data. That is a 4-point widening in about a year, by our arithmetic.
In the authors’ words, the divergence “has widened steadily since we first documented it in August 2025.”
Fewer Doors, Not More Layoffs
The paper says the gap comes from hiring, not firing. Separation rates fell for young workers, and fell at least as much in the most exposed jobs.
The authors say that is “the opposite of what displacement would imply.” The gap in hiring rates opened after 2022.
The decline shows up where AI is mostly used to do tasks instead of people. Where AI mostly helps workers, employment is flat or rising, the paper says.
The Software Developer Numbers
The paper shows software developers by age in an appendix chart, normalised to November 2022. Its text gives no age-by-age figures.
Unusual Whales’ post read the drop for ages 22 to 25 as about 20%. It put ages 41 to 49 at plus 20%, a 40-point gap by its own subtraction.
What Official Payroll Data Shows
Government data counts jobs by industry, not by age. Jobs in computer systems design fell from 2,482.9 thousand in November 2022 to 2,354.1 thousand in October 2026, per Bureau of Labor Statistics data.

That is about 129,000 fewer jobs, down 5.2%, by our arithmetic. Total US payrolls rose 3.3% over the same period, an 8.5-point gap.
Jobs at software publishers fell 1.6% from November 2022 to September 2026. The October figures are preliminary and may be revised.
The Structural Read
As we read it, the entry door is narrowing rather than workers being pushed out: the paper finds the gap in hiring, not in separations.
As we read it, the update is a measured signal, not a collapse: the descriptive gap widened while one headline estimate weakened.
Brynjolfsson, Chandar and Chen, Stanford Digital Economy Lab, August 2026
“early, descriptive indicators—canaries in the coal mine—rather than causal estimates”
Three Implications
FOR GRADUATES Stanford’s update finds the gap for AI-exposed workers aged 22-25 comes from reduced hiring, not more layoffs.
FOR EMPLOYERS The paper finds declines where AI substitutes for tasks and flat or rising employment where it complements workers.
FOR POLICYMAKERS BLS data shows computer systems design jobs down 5.2% since November 2022 while all US payrolls rose 3.3%, by our arithmetic.
What Is Not Established
The authors say the patterns weaken when controlling for education, and show “some divergent trends predating generative AI.” They are also stronger in the ADP sample than in national surveys.
One headline regression estimate weakened in the update and is “no longer statistically significant,” the appendix says. The descriptive gap is what widened.
Payroll data by industry cannot show age, and other forces such as interest rates also hit tech hiring. The paper says its results hold when controlling for rate exposure and remote work. Anthropic chief Dario Amodei has forecast far bigger effects, as we reported today.
The Bottom Line
Stanford’s update puts young workers in AI-exposed jobs 19% behind their peers, up from 15% a year ago, driven by less hiring rather than layoffs. Official data shows computer systems design jobs down 5.2% since late 2022 while all US jobs rose 3.3%. What to watch is whether the next update widens the gap again.
97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. The 19%, 15%, 13%, 11% and 10% figures and all quotes are from the August 2026 version of the Stanford Digital Economy Lab paper, which uses ADP payroll data through June 2026. The software developer figures by age are Unusual Whales’ reading of the paper’s appendix chart; the paper’s text does not state them. Industry job counts are from BLS Current Employment Statistics, seasonally adjusted, with September and October 2026 preliminary. The 4-point widening, 5.2%, 1.6%, 3.3%, about 129,000 and 8.5 points are our arithmetic. The paper describes its results as descriptive, not causal.
Sources: Stanford Digital Economy Lab, Brynjolfsson, Chandar and Chen, ‘Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence’, August 2026 version · Unusual Whales post on X, 22 Nov 2026 · US Bureau of Labor Statistics, Current Employment Statistics series CES6054150001, CES5051320001, CES0000000001 (seasonally adjusted), read 22 Nov 2026 · FourWeekMBA, US software engineer hiring, 10 Nov 2026 · FourWeekMBA, Amodei WSJ profile, 22 Nov 2026









