AI + ML
It’s the end of the world as we know it, but AI feels okay
Human economists Maxim Masenkov and Peter McCrory report that AI will not result in as many job losses as experts predicted.
CEO Dario Amodei is one such expert. In January 2026, Amodei revisited and expanded on his 2025 prediction that “half of entry-level white-collar jobs could be replaced by AI in the next one to five years.”
“Eventually AI will be able to do everything. We need to work on it,” he said in an epic post.
One day, robots may be folding clothes and transporting bored citizens trying to use up their government-allocated basic income at pop-up stores in depopulated areas. But today is different.
For now, AI appears to have little impact on people whose jobs are considered “exposed” to automation.
“We do not find a systematic increase in the unemployment rate for highly exposed workers after late 2022. However, we find suggestive evidence of a slowdown in employment among young workers in exposed occupations,” Masenkov and McCrory wrote in their report. [PDF] The title is “AI’s Impact on the Labor Market: New Measures and Early Evidence.”
There are also recent examples of layoffs that can be attributed at least in part to the impact of AI, such as Jack Dorsey’s decision to lay off about 4,000 people, or about 40% of Block’s workforce. Other factors may be at play, given Dorsey’s failure to generate revenue during his early years running Twitter and Bullock’s missed revenue in November 2025.
Indeed, now that sprawling software projects can be reformulated and relicensed with a little provocation, it seems necessary to rethink the economics of software.
But overall, the Anthropic business community is not ready to sound the alarm on AI. As they state in the blog post announcing their work, “The track record of past approaches is that [to AI labor impact forecasting] It gives us a reason for humility. ”
Masenkov and McCrory are proposing a new measure that they say helps clarify things: observed exposure.
This aims to measure how AI is used in practice, rather than how it is used in theory. They admit that “AI is far from its theoretical capabilities.”
About a year ago, economists investigated the actual economic impact of AI on workers in Denmark and found that it had no effect on jobs or wages. Practitioners of the dark science of antropics can see the needle moving a little if they look closely.
But if we use this new standard of actual AI usage rather than theoretical AI, the result is… pretty much the same.
Based on data from the U.S. Bureau of Labor Statistics, Anthropic researchers predict that occupations with high observed rates of exposure to AI will grow at a slower rate than other occupations through 2034. And when that happens, the most exposed roles are expected to go to older, better-educated, higher-income female workers. But we are not there yet.
The average change in the unemployment gap between highly exposed workers and those more insulated from AI since the release of ChatGPT is “small and insignificant, suggesting a slight increase in the unemployment rate for the more exposed group, but that the effect is indistinguishable from zero,” the researchers observed.
The exception is younger workers, whose employment has slowed among exposed occupations. But even this was rated “fair” by Anthropic researchers, who noted that the estimated average 14% drop in job application rates from ChatGPT’s introduction in 2022 to now is “barely statistically significant.” ®
