Thread claims CS grads FTE down to 55%

A viral X thread on May 18 revived a fight over whether generative AI is already cutting off entry-level tech jobs. The post said the share of U.S. computer science graduates landing full-time work had fallen from about 70% before ChatGPT to 55% after its release, and framed the drop as evidence that AI was “killing jobs.” Other users circulated the figures, challenged the methodology and posted competing data. Andrew Ng, the AI entrepreneur and Stanford adjunct professor, pushed back in a reply shared widely on X, arguing that software-engineering demand remained strong and that the broader U.S. unemployment rate was about 4.3%. ### Where did the 70% and 55% figures come from? The May 18 thread presented the numbers as a degree-to-job conversion rate for U.S. computer science graduates, but the post did not point readers to a standard federal series that publishes that exact before-and-after ChatGPT comparison. That became the central dispute online: users repeated the headline percentages faster than they established the underlying dataset. The Federal Reserve Bank of New York does publish regularly updated labor-market data for recent college graduates and annual outcomes by major, but its public dashboard tracks unemployment, underemployment and wages rather than a single “full-time employment conversion” metric for computer science graduates. The New York Fed says labor-market conditions for recent college graduates “continued to be challenging” at the start of 2026, with unemployment at about 5.7% in the first quarter and underemployment at 41.5%. ### What do the best public numbers say about computer science majors? New York Fed data show computer science majors with a 6.992% unemployment rate and a 19.127% underemployment rate in the latest annual outcomes-by-major file. The same file lists a median early-career wage of $87,000 for computer science majors. (newyorkfed.org) Those figures support one narrow point made by people amplifying the thread: computer science graduates are facing a weaker market than many students expected a few years ago. But they do not, on their own, verify the specific claim that full-time employment fell from 70% to 55% because of ChatGPT. The New York Fed’s major-level data are annual and do not assign causation to AI. (newyorkfed.org) ### Did Andrew Ng’s rebuttal line up with official data? Andrew Ng’s post cited a U.S. unemployment rate of about 4.3%. Federal Reserve Economic Data, which republishes Bureau of Labor Statistics figures, shows the national unemployment rate at 4.3% in April 2026. Ng’s broader point — that software-engineering demand has not collapsed — also has some support in recent New York Fed research, though not as a blanket all-clear. (newyorkfed.org) In a May 14 Liberty Street Economics post, New York Fed researchers said that while overall hiring has slowed since ChatGPT’s late-2022 release, job-posting data showed “little indication of a distinct AI-driven decline in labor demand” for AI-exposed occupations. (fred.stlouisfed.org) ### So is AI already wiping out entry-level tech jobs? The New York Fed’s May 14 analysis says AI may be contributing to labor-market changes, but “is not the main driver of the slowdown in hiring,” based on job-posting evidence. That is narrower than saying AI has had no effect. The same post says overall hiring has slowed since 2022, and the separate recent-graduates dashboard says conditions remained challenging at the start of 2026. (libertystreeteconomics.newyorkfed.org) The online argument turned on that distinction. One side treated a weak market for new graduates as proof of AI displacement. The other side, including Ng, argued that a softer hiring environment does not establish that generative AI is the primary cause. ### What should readers watch next? The Bureau of Labor Statistics is scheduled to release the next U.S. unemployment report on June 5, 2026, according to FRED. The New York Fed’s recent-college-graduate dashboard and its outcomes-by-major data are the clearest public benchmarks to watch for any further deterioration — or improvement — in computer science employment outcomes. (libertystreeteconomics.newyorkfed.org) (fred.stlouisfed.org)