Frontier AI models are now performing at the 90th percentile or higher on a bunch of tasks that form the backbone of what white-collar workers do: reading comprehension, coding, legal research, medical diagnosis, quantitative reasoning, analytical writing. And they do it dirt cheap compared to hiring a human.
Here's the uncomfortable truth that nobody likes to say out loud: most people, by definition, have average cognitive abilities. And a huge portion of professional ("white collar", "elite", etc.) employment has never actually required exceptional intelligence. Dilbert, Q.E.D. It required a combination of average-to-moderately-above-average ability, reliability, and the ability to be productive when working with others and to not make others less productive. The college degree was mainly a signal that you cleared that bar.
This created a pretty stable economic arrangement that worked for decades: credentials restricted labor supply, and the tasks themselves were difficult enough to need a human brain but routine enough that most people with degrees could do them adequately. It was a nice equilibrium while it lasted.
That equilibrium is now under pressure from AI (as well as from cost inflation beyond the value conferred on the person seeking a degree, but that's a story for another day...)
If a cheap tool can do the thinking part of your job really well, then "I'm reasonably smart and I have a degree" stops being worth as much. The stuff that's left over - accountability, showing up, trust, knowing where the bodies are buried, human judgment when things get weird - that stuff still matters. But is it enough to keep paying people what they're getting paid now? I doubt it.
Here's my main point: today, a lot of white-collar jobs don't actually require exceptional intelligence, critical thinking, or problem-solving ability. These are the jobs at highest risk.
Your value in the future won't be about your title or your credentials anymore; it'll increasingly be about your ability to deliver observable results in short time horizons.
AI is showing us where human brains fall short - laziness and manipulability.
Leaning too hard on AI tools likely will make people worse at thinking. They're outsourcing their mental work to the machine, and studies are starting to show what you'd expect: weaker critical thinking, worse memory, and less ability to solve problems on your own. The more you let the AI do your thinking, the worse you get at thinking.
Surveys of tech experts are predicting mostly negative impacts on stuff like empathy, creativity, and moral reasoning over the next decade. A lot of people see AI as chipping away at the things that make us human agents rather than just consumers of algorithmic slop.
Speaking of algorithmic slop, AI is magnifying the ability to sway opinions via subtle media manipulation- not just deepfakes, but algorithmic (or intentional) bias towards or away from certain positions on certain issues, framing, selection bias, etc. AI can pretty reliably automate large-scale or targeted manipulation.
All this stacks the deck pretty heavily against humans.
The silver lining (because I'm not a complete doomer about this): many roles in business, law, medicine, and other fields require exactly the kind of critical thinking, problem-solving, and nuanced judgment that AI doesn't handle well yet. Complex situations where you need to weigh ambiguous information, navigate human relationships, and make calls where there isn't a clear right answer- AI is still pretty bad at that stuff, and might be for a while.
These roles will continue to need highly skilled humans. But I suspect larger pools of humans will be competing for fewer of these jobs, so if you want to stay employed, you should probably be figuring out how to be one of the people who can actually do the hard stuff that AI can't.
How do you both become good at using generative AI, for the productivity boost it gives you, while at the same time not outsourcing your unique value to the AI and letting it atrophy? You're not going to like this - you have to use some of your productivity to keep your skills honed, and you have to develop skills complementary to AI. For technical professionals I suggest that you do a lot of agentic coding to develop your skills in specification using natural language.
I've spent over a thousand hours of my personal time in 2025 doing agent-assisted coding. I have taken every opportunity to increase my skills as I went along - I started focusing much less on writing code, but much more on writing specifications and prompts. Every time an opportunity for a side quest came up, I took it and invested a few hours learning the basics - this year I learned about MCP, typography, CSS, JavaScript test frameworks, OpenAPI, API test automation, fuzz testing, web application frameworks, web theming, color blindness, accessibility, localization, and various other things, while still keeping up with my skills in information security and AI and trying to develop my soft skills more. I tried to automate just a little of what I learned in each skill, to put it into practice rather than just reading about it.
YMMV, but as for me, I'm trying to maximize my continuing employability by maximizing my value.
There doesn't seem to be much pressure in the near-to-mid-term against skilled trades. Electricians, plumbers, mechanics, etc. - their jobs require physical presence, adaptation to unpredictable environments, and human interaction in ways that are hard to automate, at least until Elon starts selling Optimus robots.
I think that there will still be lots of service jobs - restaurant workers, hotel staff, customer service, etc. - for all but the lowest end service establishments - I think that service by a human will become a luxury good, sought after by affluent people. And in 2026 tips start getting excluded from income tax!
Just my opinion. No refunds.