Every few months, someone writes a viral post about how AI will destroy the economy. It looks like that’s just our lives now...

This month it’s Citrini Research’s “THE 2028 GLOBAL INTELLIGENCE CRISIS” (yes, all caps, so you know it’s serious). The post imagines a 2028 scenario: AI automates white-collar work, companies collapse, private credit blows up, mortgages default, unemployment hits 10%.

Stocks fell earlier this week. People panicked, based on a random blog post.

If you’re terminally online, you’re probably sick of hearing about this one already. I don’t want to go through point by point. Way too many people have already done that (and I’ll link some throughout.)

But if there’s a niche here, Economic Forces is about taking that step back and thinking through the economics, through the price theory.

So why do these types of arguments keep happening? How do they keep fooling smart people, but at the same time basically none of the economists fall for this line of thinking? More importantly, how can we avoid these pitfalls?

As usual, basic economics can help us a lot.

Because the Citrini post is not a new type of error. It’s the same error we see reading all sorts of news stories, this time dressed up in new clothes as fantasy. No economist or economics-adjacent person was blown away by it.

We have a few basic principles that rule it out before the conversation starts. As usual, we don’t need to invent new theories every time a new technology comes along. The old ones work fine.

Basic economics cuts against Citrini

The piece is a narrative around a few companies. That’s fine. The company-level stuff is worth thinking through. ServiceNow mechanics? Well observed. DoorDash losing its moat because AI agents don’t have a home screen? Maybe. Whether any particular company survives AI disruption is a fine question. Maybe the scenario is right for some of these firms.

But Citrini doesn’t stop at companies. They leap from “these businesses will be disrupted” to “the economy will enter a crisis.” And that’s a different kind of claim. Firm analysis is about what happens to one company.

That leap requires economics. Economics is about what happens in markets: where prices adjust, resources get reallocated, and one person’s lost revenue becomes someone else’s cost savings.

And every basic principle we have cuts against their narrative. Mankiw lists ten principles in chapter one of the intro textbook. They’re a little hokey. But they’re in the textbook because people keep violating them.

Let’s go through the red flags.

Ghost GDP (accounting identities)

Right at the start, Citrini introduces a concept called “Ghost GDP”: output that “shows up in the national accounts but never circulates through the real economy.”

Ummm… yea, that’s not a thing.

GDP is not a number someone estimates and hopes is roughly right. I mean they do estimate it, but that’s not what matters here. It’s an accounting identity. Every dollar of output is, by definition, a dollar of income to someone. There is no output that “doesn’t circulate.” If a GPU cluster in North Dakota does the work of 10,000 white-collar workers, someone owns that output. Someone earned that revenue. The money went somewhere.

Where does the money go? That’s the question Citrini never asks.

If labor’s share of income falls from 56% to 46%, that’s a massive shift. But it’s a shift to capital owners, not a shift into thin air. Capital owners do things with that income: they consume some, save the rest, and those savings finance investment. Citrini treats income accruing to capital as though it disappears from the economy. Their own numbers show it doesn’t. Hyperscalers are spending hundreds of billions on data center capex. That is enormous investment demand absorbing savings, which is the opposite of a demand collapse.

The essay needs the consumer economy to collapse and the investment economy to boom at the same time. It never reconciles the tension because it can’t. You cannot have GDP going up while every component of spending is going down. The identity won’t let you. As Steinberg puts it, “it only makes sense to invest in AI if there is income to buy the things the AI is generating.”

The spiral with no brake (supply and demand)

The same section describes a “human intelligence displacement spiral”: AI improves, fewer workers needed, layoffs increase, displaced workers spend less, more AI investment, repeat. “a negative feedback loop with no natural brake.”

There is a natural brake. It’s called prices.

This is the most basic idea in economics. Before accounting identities, before GDP, before any of that. Supply goes up, price falls. Price falls, quantity demanded rises. The system has built-in dampening feedback. That’s what equilibrium means. That’s what markets do. More basic, that’s what incentives do.

Are there exceptions? Sure. Sometimes feedback reinforces instead of dampens. Economic growth can build on itself: more capital, more productivity, more savings, more capital. Supply curves can slope downward. But that’s growth begetting growth.

Citrini is describing the opposite, a doom spiral where disruption begets disruption with no floor. Does that sometimes happen? Sure. We have panics and negative spirals, but the dampening forces are the default. You need a specific story to override them, and Citrini never provides one.

Citrini’s scenario freezes the economy at step one. Companies cut workers. Workers lose income. Spending falls. Recession. Full stop. They never ask what happens to the price of the services those workers used to provide, or what happens to demand when intelligence gets radically cheaper. It’s a simple point, but people forget one side of the market all the time. Citrini forgot the demand side entirely.

From a company to the economy (micro =/= macro)

Then the company stories begin. In “How It Started,” ServiceNow loses seats because its Fortune 500 clients are cutting headcount. “Each company’s individual response was rational. The collective result was catastrophic.”

What’s true for one company is not a prediction about the economy. There’s all the feedback we talked about above.

Also, the fact that the economy is huge. 160 million workers. 6 million firms. $28 trillion in output. Citrini tells stories about a handful of SaaS companies and leaps to the whole thing collapsing.

You can’t narrate the aggregate from the most-exposed slice. Even granting Citrini’s claim that white-collar workers are 50% of employment, the other 50% — healthcare, construction, government — operates on entirely different timescales. A hospital can’t replace nurses over a mid-year budget review. I just had a converation with someone in healthcare that was talking about the new tech of SHAREPOINT!

You can’t get to 10% unemployment by telling stories about software companies.

Even within those stories, though, look at what’s actually happening. ServiceNow’s revenue falls, but its customers’ costs fall too. The price of routine cognitive work drops. The price of human judgment in areas AI handles poorly? That barely moves. Some prices fall, others rise, and relative prices shift all over the place. That’s a market adjusting. And when relative prices change, new comparative advantages show up.

“For every new role, dozens obsolete” (comparative advantage)

Then comes the labor section. Citrini’s implicit logic: AI is better than humans at cognitive work, so humans lose those jobs. “For every new role AI created, though, it rendered dozens obsolete. The new roles paid a fraction of what the old ones did.”

We’ve heard this a bunch of times.

Again, let’s put on our Econ 101 hats. The assumption buried in there is that being worse at something means you don’t get to do it. That’s absolute advantage thinking. It ignores the most powerful idea in economics: comparative advantage.

Even if AI has absolute advantage in every cognitive task, humans still have comparative advantage in something. That’s a mathematical necessity, not a hopeful guess.

Here’s how it works. Suppose AI is 100x better than humans at coding and 10x better at in-person care. The cost to AI of doing one hour of care is 10 hours of coding foregone. The cost to humans of doing one hour of care is only 1 hour of coding foregone. Humans are the cheaper option for care, even though AI is better at both.

Comparative advantage is about relative costs, not absolute capability. Wages may fall in those tasks. But lower wages are the price adjustment that makes humans competitive, not a sign that humans are useless.

In the here and now, and I’m willing to bet in 2028, there is labor. And so comparative advantage is real.

The pie isn’t fixed (lower costs, more demand)

Comparative advantage tells you humans aren’t useless. But it goes further. The total amount of work to be done is not fixed. It has never been fixed.

And Citrini’s own essay shows this. In “When Friction Went to Zero,” they describe AI agents handling travel booking, re-shopping insurance, assembling itineraries. Token consumption rising 10x in a year. Stablecoin-based machine-to-machine commerce. Agentic shoppers running in the background.

They’re describing massive new economic activity while arguing the economy is stagnating.

Automation lowers costs, which lowers prices, which gives consumers more purchasing power, which creates demand for things that didn’t used to clear the market.

The ATM was supposed to eliminate bank tellers. The number of bank tellers increased after ATMs because cheaper branch operations meant more branches, and tellers shifted to customer service and sales. Every round of automation has done some version of this. (I’ll address the horse example in another post; I promise.)

The question is always “what new jobs will emerge when these tasks become cheap?” The essay doesn’t touch it.

An economy generating new forms of consumption is the opposite of a stagnating one. The essay assumes maximal disruption on the supply side and zero creation on the demand side. Its own content contradicts that assumption.

Again, if I wanted to sum up, its the price theory principle that markets are connected, and the seen/unseen asymmetry. The Citrini post is a very detailed, very vivid first link in the chain. It never gets to the second. And in economics, the second link is where the action is.

Nothin’ new under the sun

This error happens all the time. It’s a pattern that becomes easy to spot (with varied repitition). Different details, same structure. Someone observes something true about a company and leaps to a conclusion about the entire economy, without thinking through how prices adjust along the way.

During the post-Covid inflation: PepsiCo executives told investors they were raising prices. Kroger admitted to pricing above costs. Therefore, corporate greed caused inflation. But asking companies why prices rose told you nothing about inflation.

Kroger raising prices is the seen. The unseen is all the other markets. Those are relative prices: Kroger got more expensive relative to everything else, and that changes behavior everywhere. We need to remember to trace that out. traced how one price rising meant others adjusting.

Price theory as antidote

Economists make mistakes. I’m sure I’ll make one someday. But we’re relatively immune to this type of mistake, and the responses to Citrini show why. Noah Smith, Alex Imas, a month earlier, and Steinberg all applied standard tools, and the scenario fell apart. These responses are routine. That is precisely the point.

Yes, AI is genuinely new. What it can do is impressive and a little unsettling. I understand the impulse to treat it as something that requires a whole new economics.

It doesn’t!

The economy has absorbed general-purpose technologies before. The transitions were painful for real people, and this one will be too. But the analytical tools that help us think through those transitions are the same tools we’ve always had. Ben Thompson made a similar observation from the business side: the Citrini post is “grounded in a fundamental lack of belief in dynamism, human choice, and markets.” He’s right. But believing in dynamism isn’t enough. You have to reason through the logic of dynamism, choice, and markets.

You don’t need to spin stories out of thin air or write sci-fi from 2028. Supply and demand. Accounting identities. “And then what?” That’s the whole toolkit.

As Josh put it, price theory is the antidote to bad arguments. It was the antidote during Covid. It was the antidote for trade deficits. It is the antidote now. It will be the antidote next time.