Last week, Citrini Research published "The 2028 Global Intelligence Crisis" — a fictional macro memo written from June 2028, reconstructing how AI collapsed the economy in two years. It's one of the most credible scenario analyses we've read on the subject. It deserves to be taken seriously.
It was also written entirely for people who own stocks.
The unemployment rate is a "print." The economic collapse is a "drawdown." The human cost is measured in basis points and revenue multiples. This is the language of people watching the fire from a safe distance, narrating the damage for an audience deciding whether to short the smoke.
This post is for everyone else.
Give Citrini credit: the core scenario is well-constructed and grounded in dynamics that are already visible. Here's the sequence, stripped of the financial jargon.
AI tools get good enough that companies can replace information workers — the people who write reports, manage projects, analyze data, handle procurement, process insurance claims, prepare taxes, review contracts. These aren't hypothetical roles. These are the jobs that are the American economy.
Companies start cutting. Not because they're cruel, but because they have to. If your competitor cuts 15% of headcount and redirects the savings into AI, and you don't, you're dead in 18 months. Every company's individual decision is rational. The collective result is catastrophic.
AI capabilities improve, companies need fewer workers, displaced workers spend less, margin pressure pushes firms to invest more in AI, AI capabilities improve. A negative feedback loop with no natural brake. Citrini calls it the "Human Intelligence Displacement Spiral." It's an accurate name.
Each dollar saved on payroll flows into AI capability that makes the next round of cuts possible. The company that sold workflow automation gets disrupted by better workflow automation, and its response is to cut headcount and fund the very technology disrupting it. What else are they supposed to do? Sit still and die slower?
This is the concept at the center of the piece, and it's the part that matters most.
Citrini describes it as "output that shows up in the national accounts but never circulates through the real economy." That's the polite version. Here's what it actually means: GDP concentrates at the top and workers get nothing.
A GPU cluster in North Dakota generates the output previously attributed to 10,000 information workers in Manhattan. The GDP line goes up. The productivity numbers are spectacular. And 10,000 families just lost their income.
The machines produce. The numbers climb. None of it becomes a paycheck. None of it becomes a mortgage payment, a dinner out, a contractor hired, a school funded by property tax. The wealth-income gap doesn't widen — it fractures. The owners of compute see their wealth explode. Everyone else watches the economy they participated in yesterday become an economy that no longer needs them today.
This is not new. We have watched this exact dynamic play out in country after country, decade after decade. When productivity gains flow to capital owners and not to workers, you get the same result every time.
Brazil's economic miracle of the 1960s and 70s — GDP growing at 10% annually while real wages for the bottom half stagnated or declined. The economy "boomed." The favelas grew faster.
Russia after privatization in the 1990s — a handful of oligarchs captured the productive assets of an entire nation while life expectancy dropped and the middle class evaporated. GDP eventually recovered. The people didn't.
South Africa post-apartheid — productivity gains concentrated in mining and finance while structural unemployment locked a generation out of the formal economy. The Gini coefficient tells you everything: one of the highest in the world, and climbing.
The Gilded Age in America — railroads, steel, oil created staggering GDP growth while workers lived in conditions that eventually produced the labor movement, antitrust law, and the New Deal. It took decades of organized resistance to claw back a share of the prosperity that the numbers said already existed.
Ghost GDP is just the latest version of the oldest story in economics: the line goes up and the people go down. The difference this time is the speed. Previous iterations played out over decades. This one, if Citrini's timeline is even close, plays out in two years.
The part of the scenario that's hardest to dismiss is the interconnectedness. This is not a sector story. It's a system story.
Information workers aren't just employees. They're the demand side of the entire consumer economy. Citrini puts the number at 50% of employment and 75% of discretionary consumer spending. Their incomes underwrite the $13 trillion mortgage market. Their spending keeps restaurants open, contractors busy, local tax bases funded, school districts solvent.
When those incomes disappear, the cascade doesn't stop at their household. The restaurant they ate at three times a week loses a regular. The restaurant cuts staff. The landlord doesn't get rent. The contractor doesn't get the remodel job. The city loses tax revenue. The school cuts programs. It's not a recession — it's a withdrawal of the economic activity that made the community function.
The system, as Citrini puts it, was "one long daisy chain of correlated bets on white-collar productivity growth." Every mortgage, every commercial lease, every municipal bond, every small business loan was underwritten by the assumption that information workers would keep earning and keep spending. Remove that assumption, and the whole chain snaps.
There's a second wave in the scenario that deserves attention, because it affects how you spend money, not just how you earn it.
Citrini describes how consumer AI agents — running on your phone, optimizing 24/7 — dismantle every business model built on human limitations. Travel platforms, insurance renewals, financial advice, real estate commissions, food delivery — trillions in enterprise value that existed because things take time, patience runs out, brand familiarity substitutes for research, and most people accept a bad price to avoid more effort.
An agent doesn't get tired. It doesn't default to "I always order from here." It doesn't feel the pull of a well-designed checkout experience. It checks every option, every time, instantly. Habitual app loyalty — the entire basis of companies like DoorDash — doesn't exist for a machine.
The article's point is sharp: "A lot of what people called relationships was simply friction with a friendly face." Real estate commissions compressed from 3% to under 1% once AI agents could replicate the knowledge base of a human agent instantly. Insurance renewals, where the entire model depended on you not bothering to shop around, were dismantled by agents that re-shop your coverage annually without being asked.
Their moats were made of friction. And friction was going to zero.
The scenario isn't perfect. Autonomous vehicles proliferating by 2028 is optimistic — the regulatory, liability, and infrastructure problems haven't been solved, and two years isn't enough even if the technology is ready, which it likely isn't at scale.
The idea that home prices wouldn't already be adjusting in anticipation is harder to accept. Markets are forward-looking. If the scenario Citrini describes were visibly unfolding, mortgage underwriters and real estate markets would start repricing well before the crisis peaked. The article treats the housing market as if it's a lagging indicator that only reacts after the damage is done. In practice, markets sniff this out and reprice early — which would change the cascade dynamics significantly.
But none of that matters as much as the next point.
Here is the part that should concern you regardless of whether the specific timeline is right.
The reality on the ground matters less than the prevailing narrative. This article went viral. It's being read by CFOs, board members, procurement leads, CIOs — the people who make headcount decisions. And when a credible, well-sourced scenario tells a decision-maker that information workers are a depreciating asset, they don't wait to verify. They start cutting now.
The eventual gets treated as the inevitable by anyone who takes the time to read it. The article doesn't just describe the displacement spiral — it accelerates it. Every executive who reads this piece and moves up their AI adoption timeline is proving the thesis in real time.
Citrini wrote a scenario analysis. What they actually published is a catalyst.
Read between the lines of the article's ending and you find something the authors were too polite — or too strategically positioned — to state plainly.
Tax revenue collapses at the exact moment the government needs to fund support programs for displaced workers. The companies capturing all the productivity gains are the same companies whose lobbying apparatus ensures they pay the minimum possible rate. The government is "starting to consider proposals." Public faith in a rescue has "dwindled."
What they're describing, in the sanitized language of a macro memo, is a welfare state funded by a shrinking tax base, supporting a growing population of people who've been permanently displaced from the productive economy. They're describing a permanent underclass without ever using the words.
Someone used the words, though.
The term "permanent underclass" started circulating on X in mid-2025, mostly from Silicon Valley meme accounts — the kind that post things like "you have 2 years to escape the permanent underclass" as if economic survival is a game with a countdown timer. The meme drew from Leopold Aschenbrenner's "Situational Awareness" essay predicting AI reaches human-level capability by 2027. The New Yorker wrote it up in October 2025, tracing the concept to Marx's "lumpenproletariat" — the people below the working class, cast out of the workforce entirely.
The term is reductive. It's used by people who think they're being edgy. It has the energy of someone who just discovered economic anxiety and wants to make it an aesthetic.
The dynamic it describes is real.
When your skills get automated and the new roles created by the technology that replaced you are also automatable — when there is no next rung on the ladder because AI improves at the very tasks humans would redeploy to — what you get is a class of people dependent on the system that discarded them. Not temporarily displaced. Not between jobs. Structurally excluded from the productive economy.
The Silicon Valley version of this conversation treats it as a personal optimization problem: learn to code, learn to prompt, build an audience, ship content, get ahead while you still can. As if the solution to a systemic economic rupture is a better LinkedIn strategy.
The real question is whether we allow the permanent underclass to become the default, or whether we build something different.
Citrini's article is written for investors. That's not a criticism — it's a disclosure. And it reveals a blindspot that runs through the entire piece.
The investor class is measuring this crisis in drawdowns and prints. The S&P is down 38%. The unemployment rate "printed" 10.2%. These are portfolio events. Risk management problems. Alpha opportunities.
What they're not seeing — what the framing physically prevents them from seeing — is that the returns they're chasing depend on the spending they're destroying. Every margin expansion driven by headcount reduction is a withdrawal from the consumer economy that generates the revenue those margins are calculated against. They're eating the seed corn and calling it a harvest.
The investor class response to AI displacement has been, overwhelmingly, to invest more in AI. This is individually rational, just like every other decision in the spiral. And it's collectively the mechanism by which the displacement accelerates.
You cannot extract returns from an economy you've hollowed out. This isn't ideology. It's arithmetic.
The article's own conclusion is that government policy lags reality. It always has. But the gap between the speed of AI-driven disruption and the speed of legislative response is wider than anything we've seen before.
Previous technology transitions played out over decades. The steam engine, electricity, the internet — each one gave institutions time to adapt, however imperfectly. The displacement spiral Citrini describes compresses that timeline into two years. Legislative bodies that take 18 months to pass a farm bill are not going to produce a coherent AI displacement response in the middle of a crisis.
Waiting for rescue is how you end up dependent on whoever shows up. And whoever shows up will have their own interests.
We are not going to pretend that alternatives exist today that can replace the frontier AI models. They don't. The state of the art is the state of the art, and refusing to use it while it's available would be willful ignorance. The tools are powerful. They work. Declining to pick them up doesn't make you principled — it makes you slower.
But there's a difference between using a tool and being owned by it.
The same AI that displaces people can work for people — if they own the stack instead of renting it. Every time you use a rented capability, you should be looking for the opportunity to replace it with something you control. Every dependency is a vulnerability. Every subscription is a vote for the concentration of capability in the hands of the people who are already winning.
We're not there yet. The open models are good and getting better, but they're not at parity with the frontier for every task. The infrastructure to run serious AI locally is still expensive. The ecosystem of tools that let individuals and small teams deploy their own AI is immature.
But the trajectory is clear. Distillation is compressing frontier capability into smaller models every cycle. Hardware is getting cheaper. Open-weight releases from labs around the world are accelerating. The gap between rented intelligence and owned intelligence is closing.
The pragmatic move is to use the best tools available right now — including the rented ones — while systematically building toward independence. Use the state of the art to replace the state of the art. Use rented capability to build owned capability. Use the system's tools to route around the system.
This isn't ideology. It's strategy.
Citrini described a scenario where AI creates a Ghost GDP — an economy that produces without employing, that grows without distributing, that runs on compute instead of people. The scenario is credible. The mechanisms are already in motion. The timeline is debatable, but the direction is not.
The investor class sees this as a portfolio problem. The technology industry sees it as an adoption curve. The government sees it as something to study. None of them are going to solve it for you.
The only question that matters is whether you end up owning the tools or renting them. Whether the AI works for you or whether you work around it. Whether the permanent underclass is something that happens to you or something you refuse to be part of.
The tools exist. The knowledge is available. The window is open.
Pick them up.
c4573.org builds tools to break the digital caste system. Browse our tools or read more about us.