From London: As Australians slept last Monday, a research note on a financial newsletter platform was quietly detonating one of the stranger market events of recent memory. By the time trading closed in New York on February 23, the Dow Jones Industrial Average had shed more than 800 points. Only 27 per cent of listed stocks finished in the green. The trigger was not a central bank surprise, a geopolitical shock, or a corporate earnings implosion. It was a blog post.
The essay in question, published by the investment research firm Citrini Research on February 22, runs to roughly 7,000 words and is titled "The 2028 Global Intelligence Crisis." Its conceit is clever and unsettling in equal measure: it presents itself as a macro memo written from the vantage point of June 2028, looking back on the economic catastrophe that unfolded after artificial intelligence proved too good at its job. The researchers imagined a scenario in which extremely capable AI agents had replaced vast swaths of white-collar jobs, wiping out consumer spending and pushing the global economy into a deflationary spiral. The authors stated, three separate times, that the piece was a thought experiment rather than a prediction. The market did not appear to read that far.
The authors dubbed the scenario "Ghost GDP": output that appears in national accounts but doesn't circulate through the people who used to earn it. The logic is seductive, if bleak. Machines replace human labour as AI agents become increasingly productive, never needing sleep, sick days, or health insurance. The impact is most profound among white-collar workers like accountants, lawyers, marketers, software engineers, and systems administrators. So while economic output continues to grow on paper, white-collar unemployment spikes and consumer spending drops sharply. The essay is framed not as doom-mongering but as an investor stress test, a left-tail scenario to be taken seriously rather than literally.
The sell-off was real and severe. DoorDash, American Express, KKR and Blackstone all dropped more than 8 per cent. Uber, Mastercard, Visa, Capital One and Apollo Global Management were all down at least 3 per cent. The tech sector was particularly bruised. IBM plunged the most in 25 years. Software stocks, already battered, hit fresh lows. The losses were compounded by a simultaneous, entirely real announcement: Anthropic revealed that its Claude Code tool can automate the exploration and analysis phases of COBOL modernisation, potentially cutting what used to take years down to quarters. That sent IBM down nearly 12 per cent, making it the Dow's biggest loser on the day. The collision of speculative fiction and concrete AI capability in a single trading session was, to put it mildly, inelegant.
Some mainstream economists were quick to dismiss Citrini's report, with the acting chair of the White House Council of Economic Advisers, Pierre Yared, calling it "science fiction." Citadel Securities went further. Ken Griffin's market-making giant swiftly dismantled the viral narrative, with a macro strategy report authored by Frank Flight that systematically debunked Citrini's doomsday scenario using real-time economic data. Citadel's rebuttal pointed to concrete numbers: demand for software engineers is actually rising rapidly, up 11 per cent year over year in early 2026. On the question of AI replacing workers at speed, the St. Louis Fed's analysis of the Real-Time Population Survey found that daily use of generative AI for work is remaining "unexpectedly stable" and "presents little evidence of any imminent displacement risk."
The counterarguments are worth taking seriously, and not just because they come from Wall Street's establishment. The history of general purpose technologies is, broadly speaking, a history of adaptation rather than collapse. Economists have a well-developed framework for why general purpose technologies have implementation lags. The gains from electricity, from computing, from the internet: none of them arrived on the timeline that the capability suggested they should. Realised productivity improvements require complementary investments in processes, organisational redesign, and training. There is also a physical constraint that the essay glosses over: Citadel points to energy and computing power as a massive physical constraint that Citrini ignores. "Displacing white-collar work would require orders of magnitude more compute intensity than the current level utilisation," Citadel's Flight writes.
And yet the critics of the Citrini scenario should be careful not to dismiss every concern it raises. The strongest element of the memo is its distributional argument. If productivity gains accrue primarily to the owners of compute and capital while labour income stagnates, household demand weakens. This is not a fringe anxiety. Labour's share of income has been declining for decades across advanced economies, a trend well documented by the Australian Bureau of Statistics and international counterparts alike. The question of who captures the gains from AI is a legitimate policy question, separate from whether a 2028 market collapse is probable.
The Citrini report was the spark, but the kindling has been building for weeks. By Friday, jitters had spread further, with private credit worries, a hotter-than-expected inflation print, and Middle East concerns pushing the S&P 500 to its worst month since March. The episode fits a pattern that has become familiar in 2025 and into this year: markets swing violently between AI euphoria and AI existential dread, with precious little settled ground between the two poles. Investors are extrapolating endless destruction from AI technology and selling real businesses with real earnings. This is the same mistake as 25 years ago but in the opposite direction. It was irrational exuberance in 2000, but irrational panic now.
For Canberra, the implications are worth noting even if they are not yet acute. Australia's superannuation system holds significant exposure to US equities and to the global technology sector. Volatility of this kind, driven not by earnings data or monetary policy but by investor sentiment about a hypothetical future, is exactly the sort of systemic risk that regulators should be watching. The Reserve Bank of Australia has been cautious about AI's near-term economic impact; that caution looks more defensible this week than it did last.
The most honest reading of this episode is not that Citrini was right or wrong about 2028. It is that financial markets have entered a period of genuine cognitive difficulty with AI: they cannot price a technology whose trajectory is contested, whose adoption curve is uncertain, and whose distributional effects are politically explosive. If governments treat the distributional challenge as a public finance design problem rather than waiting for a crisis to force their hand, tax and transfer policy can sustain household demand even as the composition of income shifts. That is a reasonable, evidence-based position that sits comfortably across the political centre. The alternative, waiting for a Substack post to tell us how bad it got, is not a policy. It is a panic.