“Why are white-collar jobs so resilient , even with AI?”

Every year brings new predictions of AI rendering large swaths of white-collar jobs obsolete. Or falling salaries for white-collar workers due to AI. Or a video will go viral of a business mogul–be it Tesla CEO Elon Musk, Anthropic CEO Dario Amodei, or Ford CEO Jim Farley–predicting mass-layoffs of white-collar jobs, or how automation will destroy jobs. Viral posts predicting job loss, such as below, are typical:

And with every passing year, those predictions keep being wrong. In January 2025, I said in no uncertain terms that white-collar jobs would continue to thrive despite AI. So far, this prediction appears to be correct. And a similar post, “Why AI will not free people from work or lead to social equality,” also predicting that AI would not free people from work. The awaited or hoped ‘white-collar job crunch’ hasn’t come to be.

College grads from top schools continue to be hired by the largest, the most prestigious, and rapidly-growing tech and finance companies, and keep earning huge salaries with no end in sight despite AI, whether it’s Open AI or Jane Street. For example, Open AI in October went viral by offering prospects mouth-watering $1.5 million pay packages. Top hedge funds such as Jane Street and Citadel are similarly offering huge compensation. “But just do the trades, bro! AI will make your office job obsolete!,” say the masses and the media.

True, Silicon Valley has been in something of a job slump since around 2022. Google’s headcount fell from a peak of 190k in 2022 to 180k in 2025. But this largely attributed to the post-COVID hiring boom and subsequent recession. It’s premature to say that AI has played a large role in this. But otherwise, we’re led to believe that there is also an imminent white-collar job apocalypse, so what can explain this apparent disconnect between the popular narrative of AI obsoleting white-collar jobs, and the reality of such jobs being much more impervious than commonly assumed?

For one, popular narratives tend to be wrong more often than not. Popularity is a poor predictor of correctness. All too often, something is popular not because it accurately reflects reality, but because people want to believe it’s true. Perhaps there is a sort of delight in seeing white-collar workers, who thrived from crisis to crisis unscathed from 2008-2024, finally having to experience actual adversity with the rise of AI.

An example is the popularity of the theory of multiple intelligences, which has not withstood rigorous scientific scrutiny and lacks strong empirical support. A closely related idea is the claim that there are distinct learning styles (such as visual, auditory, or kinesthetic learners) and that instruction should be tailored accordingly. Large-scale reviews consistently find no credible evidence that teaching to supposed ‘learning styles’ improves learning outcomes. Despite the absence of solid supporting evidence, both ideas remain popular largely because they feel intuitive and have been repeatedly reinforced in education and promoted by the media.

Or I am reminded in late 2024 and early 2025 of how it was widely assumed Trump would make Bitcoin a top priority, specifically the creation of a Bitcoin reserve. After some initial hype following his inauguration and the rollout of a ‘shitcoin’ in Trump’s likeness that same day, which has since lost 90 percent of its value from its peak, Bitcoin has been pushed to the back burner, with scarcely even a mention by administration officials. The entire ‘Bitcoin president’ narrative basically just died, and along with it Bitcoin’s price by the second half of 2025, as it became clear that nothing was going to happen.

Other examples of popular but misguided narratives include so-called systemic racism and “critical race theory” (CRT), which peaked around 2022, along with the broader wave of wokeness. This is not to say these ideas have been entirely discredited–there are still holdouts in academia and on the Left who subscribe to CRT–but the performative displays have largely disappeared. There are no more kneel-ins in the Capitol chamber or on NFL sidelines. Many people, even on the Left, came to view these gestures as cringe and superficial shows of solidarity that, if anything, distracted from and even undermined meaningful social reform.

To answer the question, “Why are white-collar jobs so resilient, even with AI?” consider that US real GDP from Q2 2025 to Q3 2025 came in at a stronger than expected 4.34 percent annualized, on top of a strong 3.8 percent in Q2. It stands to reason that as the US economy booms, thanks in part due to a tailwind from AI, that salaries will also increase and the labor force will grow.

I have argued that AI increases productivity of already smart people, so they pull further ahead. AI, if anything, accentuates innate differences of human ability. People who are smarter will use AI more effectively to improve their productivity, as a sort of force-multiplier effect. I’ve often observed, such as on Twitter, that people who are already strong at coding or math tend to be much better at using AI as well. Those who can quickly build high-quality AI programs are usually talented to begin with.

Again, an example is the rise of the personal computer, which only saw wealth inequality and white-collar wages surge-the exact opposite of a leveling effect. People who were already smart and talented saw a major productivity boost with computers, and the overall labor market grew, too. Same for the rise of the World Wide Web. If I had to wager, AI is just the latest iteration of this.

Finally, a bigger and more interconnected global and US economy means greater returns to outlier talent and solving coordination problems or adversarial problems. These are situations that require feedback and updating faster than is possible with LLMs, which use old data for training purposes. When you got trillion-dollar tech companies competing in high-stakes product launches, or huge quant firms competing in high-frequency trading, it makes sense to pay top dollar for outlier talent when so much is at stake, than trying to save money with AI and being at a disadvantage.