Good Things Are Good

A viral memo, proposing a 2028 market crash because of too much AI-driven productivity growth, makes little sense.

Dear readers,

If you follow the stock market, you’ve probably seen the viral memo that Citrini Research published Sunday, laying out a scenario of a market crash by 2028 driven by AI — not due to some sci-fi future calamity which the AI engines become our tyrannical overlords, but in a more mundane financial-crash way, with AI productivity sharply reducing the demand for human labor, leading to elevated unemployment, defaults in private credit and then in the mortgage market, bankruptcies by firms that act primarily as intermediaries, etcetera. The memo has been surprisingly influential, with stocks named in it — a diverse group including DoorDash, ServiceNow, and Blackstone — falling especially sharply on Monday.

I think the memo makes little sense. And I don’t think you need AI-specific expertise to reach that conclusion — the problem with the memo is that its proposed mechanism of financial-crash-from-positive-productivity-shock is implausible, whether that positive productivity shock comes from AI or anything else.

To understand why, imagine that you and your spouse are two of the workers who do not lose your jobs in the Citrini scenario. (That would be most workers: The memo, dated June 30, 2028, starts by saying the unemployment rate has hit 10.2%, meaning the vast majority of people are still in work.) If you’re in healthcare, you might lose your job to a computer if you are a radiologist or a medical biller. But fortunately, you and your spouse are both nurses, and your services remain as in demand as ever.¹ Not only do you still have your jobs — you’re enjoying a wide variety of benefits as a consumer, because of the productivity gains that underlie the economic scenario Citrini is describing.

What sort of benefits? Well, Citrini describes how you now pay less for all kinds of insurance:

Insurance renewals, where the entire renewal model depended on policyholder inertia, were reformed. Agents that re-shop your coverage annually dismantled the 15-20% of premiums that insurers earned from passive renewals.

You get better deals on your vacations:

Travel booking platforms were an early casualty, because they were the simplest. By Q4 2026, our agents could assemble a complete itinerary (flights, hotels, ground transport, loyalty optimization, budget constraints, refunds) faster and cheaper than any platform.

But really, it’s a lot broader than that. You become a more efficient shopper for basically everything, because the machines handle all the comparison shopping for you:

Consumer agents began to change how nearly all consumer transactions worked. Humans don’t really have the time to price-match across five competing platforms before buying a box of protein bars. Machines do.

And there are entire categories of professionals you can stop paying. Nobody needs a buyer’s agent when shopping for a house anymore, and that shaves 2-3% off your cost to buy a home:

Even places we thought insulated by the value of human relationships proved fragile. Real estate, where buyers had tolerated 5-6% commissions for decades because of information asymmetry between agent and consumer, crumbled once AI agents equipped with MLS access and decades of transaction data could replicate the knowledge base instantly. A sell-side piece from March 2027 titled it “agent on agent violence”. The median buy-side commission in major metros had compressed from 2.5-3% to under 1%, and a growing share of transactions were closing with no human agent on the buy side at all.

The Citrini memo focuses on the carnage this creates in certain industries: the realtors who lose their jobs and the travel agencies that lose their clients. But what does this mean for two married nurses who remain gainfully employed? It means your real income has gone way up. And presumably, that means you go out and buy more stuff.

And while the lost commercial activity has an obvious nexus to AI, the newly-created commercial activity will often have no obvious AI link at all, even though it is in fact downstream of AI-driven productivity gains. Vacations offer better value, and with more disposable income, you may go on more and nicer vacations. Maybe you renovate your home. Maybe you send your kids to six weeks of summer camp instead of three, and maybe you buy a new car every three years instead of every five.

Not only are those changes that make your life better. They’re also changes that create jobs — just not necessarily in the same parts of the economy where people are losing jobs due to AI.

Citrini also discusses sharp changes in business-to-business transactions with rough implications for specific industries. For example, firms start finding it practical to vibe code their own software, so they spend less on software-as-a-service, or they demand (and receive) sharp discounts from their SaaS providers. Obviously, this is bad news for the SaaS companies. But the companies that use SaaS now have improved profit margins — so they have better opportunities to hire and expand.

Or consider the memo’s description of the implosion of DoorDash:

Coding agents had collapsed the barrier to entry for launching a delivery app. A competent developer could deploy a functional competitor in weeks, and dozens did, enticing drivers away from DoorDash and Uber Eats by passing 90-95% of the delivery fee through to the driver. Multi-app dashboards let gig workers track incoming jobs from twenty or thirty platforms at once, eliminating the lock-in that the incumbents depended on. The market fragmented overnight and margins compressed to nearly nothing.

OK, but if DoorDash’s margins get competed away to nothing, someone has to be enjoying the savings. In practice, the savings would be shared in some fraction across consumers (lower delivery prices), workers (higher wages), and restaurant owners (higher profits). Those savings all mean higher real income for people, who can turn around and consume more, though again, it won’t be obvious which specific consumption is the result of no longer having to pay DoorDash fees. And to the extent the change means higher restaurant profits, that also means it’s easier to open and operate a restaurant — i.e., more job creation.

The Citrini memo implicitly denies these observations of broadly shared benefit by saying that, in their scenario, the share of GDP that goes to labor income falls sharply, and that on-paper GDP growth is largely a function of outsized AI-firm profits, leading to a political crisis over exactly how to tax those specific firms to stabilize the macroeconomy. But that is hard to square with the fact that their memo describes all sorts of apparent benefits flowing outward to consumers and to non-AI firms: The reason AI is bankrupting the SaaS firms, for example, is that firms can replace their expensive SaaS contracts with cheaper AI-driven solutions that work just as well. That sounds like an economic gain that’s flowing widely through the economy to all kinds of firms, not one that’s being captured by AI firm shareholders. Indeed, the entire premise of the memo — AI solutions prove irresistible to firms and consumers because they save so much money — is impossible to square with the idea that the gains are concentrated in the hands of the few.

Now, it is true that even a positive change can entail transition costs. An economic transition that increases demand for electricians and personal trainers while reducing demand for lawyers and radiologists could produce a skills mismatch. But the transition costs have to be weighed against the direct benefits of higher productivity, which, again, the memo indicates are compelling in a way we’ve never seen in our lifetimes. There’s no logical explanation in here of why the sweeping productivity gains that raise real incomes for the vast majority of consumers somehow lead to plummeting consumption and a liquidity trap. It just doesn’t make sense that all this good news is, in fact, bad news.

None of this is to say that AI couldn’t cause a catastrophe for other reasons. Maybe the superintelligent machines will kill us all — it strikes me as far-fetched, but it’s not my area of expertise. The more modest thesis of the Citrini memo — that AI could cause an economic calamity merely by being very useful, even without achieving any sort of singularity — is easier to reject. Many things worry me about the economy, but “productivity growth will get too high” is not among them.

Very seriously,

Josh

1. You can imagine versions of this higher and lower on the income scale, too — you could be a neurosurgeon or an orderly, and you’d still be employed.