Citrini Lingering Echo

Excellent articles can lead the market to mistake "scenario planning" for "realistic prophecy."

On February 22, 2026, a report titled "The 2028 Global Intelligence Crisis" set social media and the financial markets abuzz, garnering over 27 million views. On the day of the report's release, IBM plummeted by 13%, while companies like DoorDash, American Express, KKR, and others saw their stock prices drop by over 6%.

The report was authored by James van Geelen, the founder of Citrini Research. This 33-year-old researcher boasts over 180,000 followers on X and ranks first among financial writers on Substack. His focus is on equity investment and global macro research, known for his style of cross-asset, lateral thinking. His real investment portfolio has generated a return of over 200% since 2023. The report took the form of scenario planning and imagined a future set in 2028, where AI massively replaced white-collar labor in just two years. This led to consumer shrinkage, software asset defaults, credit tightening, eventually pushing the economy into a state of "technological prosperity" coexisting with "social decline." Geelen noted at the beginning of the article: "This article discusses a possible scenario, not a prophecy." However, the market evidently did not have the patience to distinguish between the two.

However, more worthy of attention than the brief market panic is the widespread discussion sparked by this article in the past few days. From academia to the investment community, from Wall Street to the Chinese internet, over a dozen response articles from different perspectives have emerged. Perhaps, instead of believing in a single extreme conclusion, we can piece together a clearer future from the "divergence and overlap" of various viewpoints.

What Citrini Said

The logical thread in Citrini's article is not complex: the leap in AI capabilities leads to the massive replacement of white-collar jobs → rising unemployment triggers a contraction in consumer spending → structured financial products based on SaaS face a wave of defaults → credit tightening spreads to a broader financial system → the economy falls into a state of "technological prosperity" coexisting with "social decline."

Each link in this chain of causality is not baseless. However, connecting them end-to-end and deducing a crisis requires a series of rather radical assumptions.

There are many ways to break down this chain. We might explore three core sub-arguments—namely, the speed and scale of labor replacement, the transmission mechanism of demand collapse, and the possibility of a financial crisis—and examine what different voices are debating around each link.

Creative Destruction

The starting point of Citrini's deduction is the large-scale replacement of white-collar labor by AI. In his narrative, this process accelerated sharply between 2026 and 2028, with professionals in fields such as law, financial analysis, software development, customer service, among others, being the most affected.

Change in the share of expenditure by companies on AI model suppliers and online labor platforms, grouped by industry AI exposure

There is indeed evidence supporting Citrini's view. An empirical study by Bick, Blandin, and Deming based on enterprise spending data showed that, after the release of ChatGPT, the most AI-exposed companies (i.e., those that previously had the highest share of expenditure on online labor platforms) significantly increased their expenditure on AI model providers while reducing their expenditure on online labor platforms, with a reduction of around 15%. It is worth noting that this substitution is not a one-to-one replacement— for every $1 reduction in labor market expenditure, companies only increased AI expenditure by $0.03 to $0.30. In other words, AI is performing the same amount of work at a much lower cost than human labor.

But Citrini may have overestimated the speed of the transformation. Some critics point to the real estate agent industry in the United States as an example. Despite the long-existing technological capability to significantly reduce the number of real estate agents, this industry still employs over 1.5 million people. The inertia of the system, regulatory barriers, and internal industry dynamics form a much stronger defense line than technology. They believe that Citrini severely underestimated the resistance of "institutional momentum."

There are also critics who cite a study from 1998 by Kimball, Basu, and Fernald, pointing out that technological shocks have historically been a positive stimulus to the supply side—while there may be short-term adjustments in employment structure, the output space it creates in the long term far exceeds the jobs it destroys.

In fact, looking back at the diffusion process of every previous general-purpose technology in history, the journey from the laboratory to widespread adoption has always been much slower than the maturity of the technology itself. It took 30 years for electricity to go from a 5% household adoption rate to 50%, 35 years for the telephone, and even the fastest-diffusing smartphone took 5 years. While the technical capabilities of AI may already be sufficient to disrupt many industries, the gap between technical capability and institutional absorption has never been one that could be bridged by capability alone.

The second key link in the Citrini narrative is a downward spiral on the demand side: unemployment → reduced income → decreased consumption → declining corporate profits → further layoffs.

In this link, Citrini confuses demand-side deflation with supply-side deflation. The former implies a shrinking consumer purchasing power, while the latter is where technological progress lowers production costs — AI-driven price decreases are essentially closer to the latter, similar to the price trajectory of electronic products and communication services over the past few decades. Some analysts believe that the Jevons Paradox will still apply: when AI significantly reduces the cost of services such as legal consulting, medical diagnosis, software development, etc., the demand that was previously excluded by high prices will be unleashed, resulting not in shrinkage but in explosive growth. At the same time, the "Moravec Paradox" will also come into play. For machines, the truly difficult tasks are often not deep logical reasoning or massive data retrieval, but rather human routine physical movements, sensory perception, and emotional communication. This means that jobs requiring physical labor and intricate perception in the service industry may be more resilient than we imagine.

But the Jevons Paradox may also fail.University of Chicago economics professor Alex Imas has suggested that if AI automates the vast majority of labor and the share of labor income in total income sharply declines, then who will be able to purchase the goods and services produced efficiently? This touches on the distribution mechanism itself. When output capacity tends toward infinity while effective demand tends to concentrate, what we may face is not a recession, but an imbalance that economics textbooks have not fully discussed — material abundance that is out of reach.

A Glimpse Behind the Curtain

In Citrini's extrapolation, the most significant part of the scenario is the transmission from employment shock to financial crisis. In his narrative, structured finance products backed by SaaS revenue (which he refers to as "Software-Backed Securities") faced widespread defaults during the AI transformation wave, triggering a credit crunch similar to 2008.

However, commentators note that, compared to 2008, the current leverage of the U.S. corporate sector is much healthier, and the banking system is far more robust after experiencing Dodd-Frank reforms and multiple rounds of stress tests.

Compared to the eve of the 2008 financial crisis, various resilience indicators of the current U.S. financial system have significantly improved: the bank tier 1 capital adequacy ratio has increased from 8.1% to 13.7%, the household debt-to-disposable income ratio has decreased from 130% to 97%, and the non-performing loan ratio has dropped from 1.4% to 0.7%.

Even if some SaaS companies do face revenue decline, the scale is not enough to trigger a systemic credit crisis. Former Bloomberg columnist Nick Smith believes Citrini made a common mistake at this point: linearly extrapolating micro-level industry shocks to macro-level systemic risks. For demand collapse, Smith's answer is fiscal policy. If unemployment indeed rises significantly, the government has the capacity and willingness to shore up demand through large-scale fiscal stimulus.

The institutional responsiveness also appears to have been underestimated, as evidenced by the policy response during the COVID period. For instance, on March 11, 2020, when the WHO declared a pandemic, just 16 days later, the $2.2 trillion CARES Act was signed into effect. In the following year, the U.S. introduced a cumulative $5.68 trillion in fiscal stimulus, equivalent to about 25% of the 2020 GDP.

If AI-driven unemployment does materialize at the speed and scale described by Citrini, policy intervention is unlikely to be absent.

Some commentators have raised doubts from a more fundamental level. Technological doomsday scenarios often stem from a lack of faith in the humanities. Citrini's extrapolation views the market as an unmanned machine, allowing "causality" to unfold until collapse. However, the real-world economic system does not operate in this manner. Law, institutions, politics, culture, and ideology profoundly shape how the real world absorbs technological shocks.

Consensus and Dissent

Perhaps we can try to annotate some consensus and dissent.

AI is currently and will continue to alter the demand structure of white-collar labor almost beyond dispute; the dissent lies only in the pace and scale of change. Moreover, the pains of transition are very real and should not be obscured by excessive optimism. Additionally, the quality and speed of policy responses will greatly determine the outcome.

Dissent lies in a more fundamental level of logic. Some believe that this current technological shock may surpass historical precedents in both speed and breadth, thus limiting the relevance of historical analogies; while others have more trust in institutional adaptability and historical repetition.

Heads Up

Citrini's article presents several issues, with overly tight logical connections, systemic underestimation of institutional responses, and a lack of sufficient intermediate arguments from micro-industry impacts to macro-systemic risks. But its most fundamental issue may lie in an underestimation of human society: it assumes a static institutional environment in which technology crushes everything at an almost unstoppable pace. There has been no shortage of doomsday scenarios in the history of technology, often unassailable in technological logic, yet almost uniformly overlooking the variable of "humans." The complexity of human society, its friction, its redundancy, its seemingly inefficient institutional arrangements, precisely constitute a powerful, distributed resistance to shocks. We have ample time to avert those extrapolated doomsdays, provided we are not intimidated by the extrapolations themselves.

What about the optimistic narratives? The "Jevons Paradox" is an observation about long-term trends. The "Moravec Paradox" tells us that physical labor is temporarily safe but does not tell us where those displaced white-collar workers should go. Historical analogies are enlightening, but history never repeats exactly; it only rhymes. Optimistic narratives need time to be tested, and we are currently at the starting point of that test.

Doomsday scenarios are produced, and the anxious pay the price. Forge your judgment, take risks, manage positions, instead of indulging in those "see into the future at a glance" articles.