Did you know the people most certain AI will take their jobs are the ones building it?
Anthropic's own survey says its staff expects 95% of AI model-building to be automated by 2030. Maybe that tells us more about the people in the room than about the economy.
Anthropic just published a survey of 1,334 of its employees. Most expect AI to handle at least 80% of their jobs by 2030. Only about 5% of American workers agree. The company's chief economist says this is a "leading indicator" because Anthropic runs about a year and a half ahead of the economy on AI adoption.
Here's the contrarian read: that's exactly the problem.
Did you know the sample is a bubble? Anthropic staff have bigger tool budgets and often get early access to unreleased models. They spend their days watching the technology improve in ways most people never will. Asking them how fast AI will progress is like asking a city's real estate agents whether housing prices will keep rising. Their confidence is real, but it comes from proximity, and proximity is a poor forecasting tool.
Did you know the company has been predicting this for over a year? In March 2025, the CEO said AI would write 90% of code within six months. That's a prediction, not a result, and the headlines it generated were a big part of why it spread. The same CEO later said AI could push unemployment to 10% or 20%. Those are bold forecasts, and forecasts are cheap when you're the one selling the thing being forecast.
Did you know the independent evidence is murkier? A Yale team recently found that AI has mainly changed jobs rather than eliminating them, with no clear connection to shifts in unemployment. The unemployment rate is 4.2%. The survey's employees expect it to top 12% by 2030. Most American workers think it'll stay around 8% or below. Workers outside the lab may be reading the data more accurately than the people who built the thing.
Did you know the GDP predictions are wild? Anthropic employees guessed median growth of 17.1% in 2030. The median American guessed 5.4%. The economy grew 2.2% last year. Even the more modest number is a huge jump. A group that expects world-changing growth has every incentive to believe its product is world-changing.
None of this means AI won't disrupt work. Entry-level jobs, customer support, and parts of white-collar work are already under pressure. But the most dramatic version of the story is coming from the people with the most at stake in it, and they're the least representative sample you could pick.
So before you panic or celebrate, ask who's making the prediction, what they're selling, and whether they've ever been wrong about the timeline.