A Fictional AI Doomsday Memo Just Spooked Wall Street - Here Is What It Actually Says

A Substack post written as a fictional memo from 2028 has racked up 16 million views on X, triggered a sell-off that wiped billions off US-listed firms, and prompted Citadel Securities to publish a point-by-point rebuttal. The post contains zero new data. It's a thought experiment. And Wall Street treated it like an earnings miss.

Claim / Our Take

The claim: AI will hollow out white-collar employment so fast that the US hits 10% unemployment and a 38% S&P 500 crash by mid-2028
Our take: The scenario's distributional argument is its strongest point, but the macro model is thin - no formal framework, no sensitivity analysis, and current labor data contradicts the timeline
The signal: Markets are so anxious about AI disruption that a well-written blog post can move billions - that anxiety itself is the real story

What They Showed

Citrini Research CEO James Van Geelen and AI entrepreneur Alap Shah published "The 2028 Global Intelligence Crisis" on February 23 as a fictional memo dated June 2028. The format is deliberate - it reads like a post-mortem written after the catastrophe already happened, which makes the narrative feel inevitable rather than speculative.

The Scenario Timeline

The memo lays out a cascading collapse:

Timeline	Event
Late 2025	Agentic coding tools achieve major capability jump
October 2026	Job openings fall below 5.5M; S&P approaches 8,000
Q3 2027	Initial jobless claims surge to 487,000; recession begins
November 2027	Market crash begins - S&P drawdown rivals 2008
June 2028	Unemployment at 10.2%; S&P down 38% from highs

The Core Mechanism

The memo introduces "Ghost GDP" - output that shows up in national accounts but never circulates through the workers who used to earn it. Companies adopt AI agents to stay competitive, cut headcount, and boost margins. Individually rational. Collectively catastrophic, the authors argue, because the displaced workers were also the consumers.

The specifics are detailed enough to feel real: ServiceNow announces 15% layoffs in Q3 2026 after AI erodes new contract growth. Mastercard purchase volume slows as AI agents route around card interchange fees. San Francisco home values drop 11% year-over-year. The median American consumes 400,000 tokens daily by March 2027.

"You're reading this in February 2026. The S&P is near all-time highs," the authors write in their closing. They frame the exercise as stress-testing portfolio assumptions, not making predictions.

What We Tried

You can't "reproduce" a thought experiment, but you can check whether its starting conditions match reality. We looked at three testable premises from the memo.

Premise	Citrini's Assumption	Current Data
Software engineer demand collapsing	Mass tech layoffs accelerating through 2026	Indeed data: software engineer postings up 11% YoY (Citadel Securities)
White-collar job postings in freefall	Job openings below 5.5M by October 2026	Postings stabilized after 2+ years of weakness (22V Research)
AI agent autonomy replacing humans at scale	Agents running continuously without user awareness by late 2026	Current agents still require significant human oversight for complex tasks

The memo's technological trajectory is plausible - AI agents are advancing fast. But the employment timeline assumes a speed of adoption that has no historical precedent. As we noted in our coverage of AI labs' own mental health crisis, even the companies building these systems are struggling to scale their own operations.

The Gap

Citadel Securities' Rebuttal

Citadel published the sharpest counter-analysis. Their core arguments:

The Recursive Technology Fallacy. Citrini assumes AI's code-writing abilities create compound effects that accelerate vastly. Citadel argues technological diffusion follows historical S-curves with natural plateaus. Displacing white-collar work at scale "would require orders of magnitude more compute intensity than current utilization." Rising compute costs create a natural economic boundary where human labor becomes cheaper than automation.

Basic supply-shock economics. Productivity shocks are positive supply shocks that lower costs, expand output, and increase real income. Profits get reinvested, distributed, taxed, or spent. Citadel invokes Keynes's 1930 prediction of 15-hour workweeks by 2000 - productivity rose dramatically, but jobs persisted because humans shifted consumption toward higher-quality goods and novel services.

The Economists Weigh In

Gerard MacDonell of Front Harbor Macro Research called the piece "allegorical" rather than data-driven, invoking Say's Law: production generates income that must accrue to someone who's likely to spend it. Ed Yardeni maintained his S&P 500 target of 10,000 by decade's end, arguing AI augments rather than eliminates workers.

Noah Smith was more blunt, calling it "just a scary bedtime story" and noting the memo uses no explicit macroeconomic model, making it unclear how the authors think the economy actually works. He questioned whether DoorDash - one of the companies the memo flags as vulnerable - would really be disrupted by AI when its competitive advantage comes from network effects, not code.

"I have seen this market exhibit incredible resilience in the face of actual negative news. Now, a literal work of fiction sends it into a tailspin," said Michael O'Rourke, Jonestrading's chief market strategist.

Where the Memo Has a Point

The strongest element is the distributional argument. Even if GDP grows, the gains may concentrate among capital owners and a shrinking pool of AI-skilled workers while everyone else sees real wages fall. This isn't science fiction - it's a pattern that has been playing out since the 1970s, when labor's share of GDP began its decline from 64% to roughly 56% today. The memo projects it hitting 46% by 2028.

Indian IT stocks lost roughly $50 billion in market capitalization in February alone on AI-disruption fears. That's not a reaction to a thought experiment - it's a market pricing in a plausible shift in how outsourced knowledge work gets done. NVIDIA's record $68 billion quarter exists because companies are spending real money to automate exactly the kind of work the memo describes.

The AI agent market is already projected at $7.6 billion and growing. The question isn't whether AI will reshape the labor market. It's how fast, how unevenly, and whether institutions adapt in time.

Verdict

The Citrini memo is a well-crafted piece of scenario planning dressed up as a dispatch from the future. It contains zero new data, no formal economic model, and a timeline that current labor indicators don't support. It moved markets anyway because it gave form to an anxiety that has no good outlet - the feeling that AI is changing the economy faster than anyone can measure.

The critics are right that the macro framework is thin. Citadel is right that compute costs create natural braking forces. The economists are right that production generates income.

But the memo's authors are right about one thing: nobody has a good model for what happens when the most productive new technology in decades hits an economy where income distribution is already stretched to its limits. The gap between "AI raises GDP" and "AI raises your paycheck" is where the real crisis lives - and that gap isn't fictional at all.

Sources: