In 2017, I embarked on a 7-part Coursera program on AI and machine learning. A major focus was neural networks applied to computer vision. Things got technical fast as Andrew Ng’s recorded lectures walked me through exercises hand-coding a supervised learning environment in Python. I made it as far as part 5, I think, before the deep-in-the-weeds debugging simply stopped being fun (for me) and I moved on to other projects.
Coming out of that, my impression of modern AI was that it is an insanely powerful tool for the few. For those people with the skills and patience to code shit in Python, massage massive training datasets, and tune hyperparameters, AI was going to change everything.
But not for me. Not for most people.
Then in late 2022, I first started experimenting with Large Language Models. Early LLMs were fascinating, but mostly useless. It was trivially easy to knock them off balance, and I had real doubts they could ever be trusted in enterprise contexts. But the rate, and scope, of improvements since then have been jaw-dropping. Back in early in 2025 I joked to a friend that “ChatGPT is like one of the smartest people I know with early-stage dementia” – a towering intellect punctuated with infrequent episodes where it loses the plot just enough to be unfortunate.
Today, just 3 or 4 months later, I’m not sure that punchline lands anymore.
I believe we’ve reached an important inflection point. If we pick New Years Day 2025 as our line in the sand, then prior to 2025 AI was a powerful technology, but a gate-kept one. AI was reshaping society, but it was mediated by technocrats. It was steadily eroding lower-value knowledge work, but it was employing armies of higher-value knowledge workers to build those solutions.
We are now entering the era of democratized AI, which sounds lovely and even egalitarian until you scratch the surface and realize we are entering the era where AI-driven solutions (and especially LLM-driven solutions) are eroding employment faster than it can be replaced. And this revolution is different from any earlier ones in a couple key ways.
First, just look at the kinds of jobs that are evaporating. These aren’t junior insurance clerks copy-pasting claims data between systems or hourly-wage customer service reps sitting in a call center, these are (or were) highly-skilled and highly-paid software developers and other knowledge workers.
Current estimates are that in the US alone, there were 250,000 tech layoffs in 2023, a further 150,000 in 2024, and another 75,000 so far in 2025. Related professional services like legal and consulting have been equally hard hit, with a half million layoffs in 2024 alone. That’s 2,000 people per day losing a solid middle-class paycheck. I personally know many highly-skilled professionals who have been on the shelf for 6, 12, 18 months or more. All of them were “doing the right things” when events overtook them. These are formerly successful people now rendered invisible.
The second way this revolution differs is pace of disruption. The speed of change today is unprecedented, and people aren’t simply being asked to re-skill, they’re being asked to hit a moving target – without warning and often without support. Prior disruptions gave workers time to pivot and some confidence in the new direction they chose. This one isn’t.
When I ask the question “have we thought this through?” what I’m really asking is “what happens when all this juicy upper-middle-class purchasing power evaporates permanently?” I don’t hear technocrats talking about this, but economists around the world are starting to grapple with the problem of a shrinking middle class and erosion of purchasing power. We are living in a complex moment, and so far the responses I see range from impractical to fantastical.
Let’s start with an idea that’s practical, but practically unimplementable: Universal Basic Income. The idea behind UBI is that some form of taxation (for example on business profits) funds monthly payments to everyone in society. Everyone. If you have no job, you at least get $X per month to spend on bare essentials. If you have a great job, you also get the same $X per month to spend on gas for the jet boat. Win-win.
Various UBI pilots have been run (or are running) around the world, and it could be a viable way forward if we are indeed confronting a permanent shift toward middle-class underemployment. The biggest challenges aren’t with the mechanics of running such a program. The biggest challenges involve societal buy-in. Because here’s the shattering truth: at the very moment when our society most needs the capacity for grand compromises, we’ve lost the ability to discuss and cooperate. Our politics are too poisoned.
Social technologies once promised us connection, coordination, and consensus. But all the attention economy has done is fracture us into bickering tribes, utterly incapable of coming together to solve the problems of our own making. Solutions like UBI aren’t impractical logistically, they’re impractical because adults can’t sit in a room together anymore.
So that leaves each of us to figure out the current moment on our own. For those struggling with a recent (or not so recent) loss of employment, there’s always the wisdom of various LinkedIn messiahs, who suggest these “transition periods” are an opportunity to pivot and become some kind of digital artisan. Start that company you always dreamed of! Sell your skills as an independent consultant! Monetize your creative passions!
Some people – typically those nearer the end of their working careers who have the good fortune to be mortgage-light or mortgage-free – actually have this freedom. But let’s not fool ourselves: being a digital artisan is an indulgence. This blog entry is an indulgence. Even if somebody does manage to pull together the occasional consulting gig or find some Patreon subscribers who like their writing enough, they are the modern equivalent of a kindly old man selling birdhouses at the flea market. Nobody is keeping a middle-class economy afloat like that. A birdhouse economy doesn’t scale.
And younger generations? The ones who are underwater on a 700 square foot downtown condo and whose company just gave them two weeks notice? They're screwed.
To be clear, I actually think modern AI systems are a miracle. They have the potential to enrich our lives in ways we can’t even imagine. The problem is that the current focus, the current plan (though I use that term very loosely) is incompatible with the economic system we’ve spent the last 200 years building. The thesis of this great uncontrolled experiment is worded something like “we are going to use powerful and unpredictable new technologies to destroy middle-class purchasing power, and then figure out some new way to sustain a healthy consumer economy.”
If there’s a plan to rebuild prosperity out of all this, I haven’t seen it. And I’ve been looking.