Last Sunday (around the time we were sending our update), a Substack newsletter did something most hedge fund managers only dream of. It started a reaction that would move the market (which may have been part of the point, read here).
The post, titled “The 2028 Global Intelligence Crisis,” was written by the team at Citrini Research along with Alap Shah. By Monday’s open, the piece had racked up tens of thousands of reposts, on its way to millions of views, and equities were selling off.
The thesis was written as a fictional “macro memo” dated June 2028. The first line, “What if our AI bullishness continues to be right…and what if that’s actually bearish?,” set the tone. In this thought experiment, AI had done exactly what its boosters promised. It got incredibly good, incredibly fast. The problem was that “getting good” meant replacing a staggering number of white-collar workers. And those workers, the ones who buy the houses, lease the cars, and fund the 529 plans, are the backbone of the American consumer economy. Remove them, and you don’t just lose some jobs. You lose the demand that keeps the system spinning.
What made the piece land so hard wasn’t just the logic; it was the specificity. The authors walked through a daisy chain of consequences. Starting with SaaS companies losing revenue as clients cut headcount, to private credit funds sitting on leveraged buyouts that assumed recurring revenue would keep recurring, and ultimately, the $13 trillion mortgage market facing a crisis not of bad lending, but of good loans written against a future that no longer exists. It was a compelling read, and the market treated it accordingly.
The Response
Within 48 hours, multiple major rebuttals emerged, each attacking the bear case from a different angle.
The most data-driven response came from Citadel Securities. Their macro strategist argued that Citrini conflated AI’s theoretical capabilities with its actual pace of adoption. Using data from the St. Louis Fed, they showed that daily AI use for work has been surprisingly stable, with little evidence of an exponential takeover. They argued that technology adoption has historically followed an S-curve, not a vertical line. And they also noted that software engineering job postings are actually rising, up 11% year-over-year, which is a hard number to square with the imminent displacement narrative.
Citadel also introduced a useful concept about the natural economic boundary of computing costs. If companies rush to automate, demand for computing power rises, which pushes up its cost. At some point, the AI agent becomes more expensive than the human, and substitution stops. It’s the laws of supply and demand applied to AI adoption, and it’s a check that the bear case largely ignores.
The Kobeissi Letter took an even more contrarian stance, which was evidenced by its opening line. It went something like, the stock market just erased $800 billion in market cap because AI “taking over the world” is becoming the consensus view, and the obvious trade never wins. They argued that the bear case makes a critical error by assuming demand is fixed. History, they pointed out, suggests otherwise. When the cost of producing something collapses, demand rarely stays flat; it tends to expand. When personal computing costs fell, we didn’t consume the same amount of computing more cheaply. We consumed orders of magnitude more and built entirely new industries on top of it.
Kobeissi also reframed the disruption as a price story rather than a jobs story. The services sector makes up nearly 80% of U.S. GDP, and those services are expensive primarily because they require trained human attention. AI reduces that cost. And if the cost of running a business falls, small businesses become more attainable. If the cost of accessing services falls, more households participate. In their framing, AI functions as an invisible tax cut for the entire economy. The companies whose margins depend on cognitive labor may suffer, but the broader economy benefits from lower service costs and higher real purchasing power.
Moreover, others made similar arguments, noting that the bear case rests on a “fixed pie” assumption, or the idea that there is a finite amount of cognitive work to be done, and that AI simply does it more cheaply. That assumption has been wrong after every major technological shift in history, largely because cheaper production unlocks demand that was previously priced out entirely.
Is either side right?
Having read as many of these pieces as possible, as is typical, I think the truth will be somewhere in between. Both sides seem to be partially right, and the gap between their positions is probably narrower than the divergent scenarios suggest.
The bears are correct that this does seem structurally different. When the ATM was invented, there was fear that bank tellers would lose their jobs. However, not only did that not happen, but the number of tellers increased dramatically. What that anecdote misses, though, is that ATMs didn’t learn to become better every six months. AI is a general-purpose intelligence that continually improves (at least for now) at the very tasks humans study and train their whole lives for. That’s new, and discounting it with “technology always creates more jobs” ignores the way in which this particular technology operates.
But the bulls are correct that the doom loop requires an extraordinary number of assumptions to all break the wrong way simultaneously. Near-total labor substitution, no fiscal response, no cost-of-living deflation reaching consumers, and unconstrained scaling of compute. That’s a bet against the adaptability of the world’s most dynamic economy (which we discussed in last week’s update), which tends not to go well.
There’s also a practical buffer that deserves more attention. There’s a concept in robotics and AI that the tasks that feel sophisticated to humans, like legal analysis and code generation, are comparatively easy to automate, while the tasks that feel trivially easy, like walking or pulling wire through a wall, remain extraordinarily hard for machines. AI will help design the next generation of infrastructure. But because the robots aren’t ready to build it, we’ll need human hands, and the demand for them is growing.
What This Means for You
For investors, the practical takeaway is less obvious than either side would have you believe. Yes, capital is rotating. Money is flowing toward compute infrastructure, energy, and the platforms that sell intelligence itself. That rotation is real and worth paying attention to.
But the narrative that every SaaS company is a dead man walking deserves more scrutiny. The bear case assumes that AI-generated software can simply replace battle-tested enterprise platforms. And anyone who has worked in the corporate world knows the time required to test and ensure you have a reliable, secure, scalable system is enormous. Spinning up a workable alternative is not the same as running mission-critical software for a Fortune 500. With some of these names down 30, 40, or even 50 percent, the market may be pricing in a worst-case scenario that overstates how quickly enterprises will abandon proven tools for AI-generated alternatives. That gap between sentiment and reality is where opportunities tend to live.
For the average long-term investor, the most important lesson from this week isn’t which side is right. It’s that the debate reveals the extent of the uncertainty. When a Substack post can move the S&P, it tells you something about how fragile market narratives have become. The appropriate response isn’t to pick a side and go all-in. It’s to make sure your portfolio can survive either outcome. Diversify across asset classes and across assumptions. Own the infrastructure powering the transition, and maintain enough liquidity to weather the turbulence.
Listening to the Canary
The Citrini piece ended with a line worth repeating: “The canary is still alive.” The rebuttals didn’t really dispute the direction of travel. They disputed the speed.
The premium on human intelligence is narrowing. Whether the repricing is violent or gradual, whether it creates a crisis or an abundance, depends on variables that no one can forecast with precision. What we can do is prepare, stay diversified, stay disciplined, and resist the temptation to mistake a compelling narrative, no matter how well-written, for a certainty.
The canary is still singing. The question is whether we’re listening to the song or just waiting for it to stop.
Markets / Economy
- Market volatility has remained high, with concerns about the impact of AI filtering through prices. The S&P finished the week down -0.4%, the Nasdaq down -1.0%, and the small-cap Russell 2000 down -1.2%.
- Producer prices increased 0.5% MoM in January, following a downwardly revised 0.4% rise in December. This was above forecasts of 0.3%. Prices of services increased 0.8%, the most since July, led by a 14.4% jump in margins for professional and commercial equipment wholesaling.
- Core producer prices, which exclude food and energy, climbed by 0.8% from the previous month, following a downwardly revised 0.6% increase in December and surprising analysts, who had expected a 0.3% rise. This marked the steepest increase in core producer prices since July 2025.
Stocks
- U.S. equities were in negative territory. Financials and Technology led the decline, while Utilities and Consumer Staples outperformed. Value stocks led growth stocks, and large caps beat small caps.
- International equities closed higher for the week. Developed markets fared better than emerging markets.
Bonds
- The 10-year Treasury bond yield decreased 12 basis points to 3.96% during the week.
- Global bond markets were in positive territory this week.
- Government bonds led for the week, followed by corporate bonds and high-yield bonds.
