Look, we’ve been talking about this.
Way back in August of ’25, Jeff hopped on his solo Summer Six-Pack pod, no guest, no agenda, just a guy with a microphone and some things on his mind, and dropped this little nugget about AI:
“The end game of the AI race is massive deflation and an end possibly to the consumer economy… every firm’s racing to automate everything, lower their labor costs, boost their profits… but if everyone does it, are you not automating away the bulk of the jobs, and then who’s left to be a customer?”
At the time, it was one point in a six-pack, sandwiched between a tribute to a legendary Kansas farmer and a rant about Costco gas lines. Nobody was exactly sounding the air raid sirens.
Fast forward six months, and someone turned that cocktail napkin thought experiment into a full-length horror movie.
The Piece That Launched a Thousand FinTwit Threads
Citrini Research, alongside co-author Alap Shah, published “The 2028 Global Intelligence Crisis” on February 22nd, and the thing went absolutely nuclear. Nearly 8,000 likes. 1,600 restacks. The S&P supposedly dipped about 1% last week, and half of FinTwit was pointing at a Substack post as the culprit. A Substack post. We truly live in the weirdest timeline.
The premise? It’s written as a fictional “Macro Memo” from June 2028, looking back at how everything went sideways. Think of it as The Big Short meets Black Mirror, narrated by your smartest friend who’s had three espressos and just discovered what agentic AI can do.
Here’s the CliffsNotes version of their doomsday daisy chain:
- Late 2025: Agentic coding tools take a step-function leap. Suddenly a competent dev can replicate a mid-market SaaS product in weeks. CIOs start asking, “Why are we paying $500K for this renewal?”
- 2026: Layoffs boost margins, earnings beat, stocks rip to S&P 8000. Everyone’s popping champagne. But underneath? Real wage growth is collapsing. They coin the term “Ghost GDP”, output that shows up in the national accounts but never circulates through the real economy. Because, spoiler alert, machines spend zero dollars on discretionary goods.
- 2027: The negative feedback loop kicks in. AI gets better → companies cut headcount → displaced workers spend less → margins tighten → companies buy more AI → AI gets better. A feedback loop with no natural brake. White-collar workers who were making 180KasproductmanagersarenowdrivingUberfor180K as product managers are now driving Uber for 180KasproductmanagersarenowdrivingUberfor45K. The pool of remaining “human jobs” gets flooded, wages compress everywhere.
- Late 2027: Private credit blows up. All those PE-backed SaaS deals underwritten at 25x EBITDA on the assumption that “Annual Recurring Revenue” would keep recurring? Turns out when AI agents handle customer service without generating a ticket, Zendesk’s ARR is just “revenue that hasn’t left yet.” The $5 billion direct lending facility gets marked to 58 cents. Dominos fall. The “permanent capital” that was supposed to make private credit resilient? It was annuity money from Main Street, structured through insurance companies that PE firms had acquired. The locked-up capital that couldn’t run was your neighbor’s retirement savings.
- 2028: The S&P is down 38% from its highs. Unemployment hits 10.2%. Prime mortgages, 780 FICO, 20% down, clean histories, start going delinquent because the borrowers’ jobs no longer exist. The loans were good on day one. The world just changed after they were written.
It’s vivid. It’s specific. It names names, ServiceNow, DoorDash, Mastercard, American Express, Zendesk, Apollo, Blackstone. It traces the transmission mechanisms from agentic coding tools all the way to the $13 trillion mortgage market. And it ends with the kicker that you’re not reading this in 2028. You’re reading it in February 2026. The canary is still alive.
If Jeff’s Six-Pack pod was a guy at a bar saying “hey, has anyone thought about what happens when AI eats everything?”, the Citrini piece is that same guy coming back with 10,000 words, a detailed timeline, fake Bloomberg headlines, and receipts.
Citadel Securities Says: Calm Down, Nerds
Two days later, like clockwork, the adults in the room showed up.
Citadel Securities published a rebuttal — and let’s be honest, when a firm that clears something like 25% of all US equity volume tells you to take a breath, people tend to listen. Their analyst Frank Flight didn’t mince words, noting that the forward path of labor destruction was apparently being “inferred with significant certainty from a hypothetical scenario posted on Substack.”
Shots fired. 🔥
Their argument boils down to a few key punches:
- The data doesn’t support the panic. Job postings for software engineers are up 11% YoY. The St. Louis Fed’s Real Time Population Survey shows AI adoption at work is stable, not accelerating on some terrifying exponential curve. If displacement were imminent, you’d see daily AI usage for work inflecting upward. It’s… not.
- Recursive technology ≠ recursive adoption. Just because AI can improve itself doesn’t mean organizations can deploy it at the same pace. Tech diffusion follows an S-curve, not a hockey stick. Integration is costly. Regulation emerges. Organizational change is messy. (Anyone who’s tried t
o get their company to switch from Slack to Teams knows this viscerally.)
- There’s a natural economic brake. If demand for compute explodes, the marginal cost of compute rises. If compute gets more expensive than human labor for a given task, firms won’t substitute. Econ 101 supply and demand, baby.
- The accounting identities don’t allow for the doom scenario. If real GDP rises (which it does in a productivity boom), then by definition some component of demand must also be rising. A world where productivity surges but aggregate demand collapses while output rises is, in their words, a violation of national income accounting.
- Keynes got this wrong 95 years ago. He predicted the 15-hour workweek by now. Instead, we just consumed more. Human wants are elastic. We always find new stuff to spend money on.
They close with a thought that deserves a mic-drop emoji: maybe AI isn’t the thing that breaks the economy. Maybe it’s just enough to offset aging populations, climate change costs, and deglobalization, keeping trend growth at a modest 2% instead of falling below it.
So Who’s Right? (The Reconciliation, Abridged)
Here’s the thing, and this is the part where 25 years in the alternatives space actually comes in handy — they agree on more than you’d think.
Both accept: AI capex is massive. Data centers are booming. The displacement narrative is dominating market psychology. Neither denies AI will transform labor markets. And both are making probabilistic arguments. Citrini literally says “this is a scenario, not a prediction.” Citadel is saying the current data doesn’t support the panic.
The real fight is about speed and brakes.
Citadel’s whole framework rests on the S-curve, adoption is historically gradual, and there are natural economic governors (compute costs, regulation, organizational friction) that prevent runaway displacement. Citrini’s scenario requires a capability discontinuity — a step-function jump that breaks the historical pattern. If Citadel’s S-curve holds, the Citrini scenario is a low-probability tail risk. If a genuine capability breakpoint arrives, Citadel’s reliance on historical adoption curves becomes the weak link.
On demand destruction, Citadel is technically correct that the accounting identities hold, GDP components must add up. But Citrini’s “Ghost GDP” concept isn’t really violating accounting identities. It’s describing a world where output accrues overwhelmingly to capital owners (who save more) while the household sector (which drives 70% of GDP through consumption) gets its income base gutted. The math still works. The economy people actually live in doesn’t.
On historical analogy, Citadel asks a killer question: Was Microsoft Office a complement or substitute for office workers? Answer: complement. But Citrini’s counter is equally sharp, every prior technology automated specific tasks and created new ones requiring human intelligence. AI targets the common ingredient across all of them: cognitive labor itself. If the thing being automated is the general-purpose problem-solving ability that humans previously redeployed to each new wave’s jobs, the historical pattern may not hold.
The honest synthesis? Citadel is probably right about the present. Citrini is asking the right questions about the future. The adoption data, labor market indicators, and new business formation numbers Citadel cites are real evidence that, as of right now, the doom loop hasn’t kicked in. Citrini’s value is in mapping the specific transmission mechanisms, white-collar income concentration, private credit’s software exposure, prime mortgage assumptions, government revenue dependence on payroll taxes, that would light up if things start breaking.
What This Means for Your Portfolio (The Alternatives Angle)
Okay, so here’s where we bring it home to our world.
Whether you think we’re headed for Citrini’s 38% drawdown or Citadel’s “everything’s fine, touch grass” scenario, the case for true portfolio diversification has never been louder.
This is exactly the kind of regime uncertainty, not just “will stocks go up or down” but “will the fundamental structure of the economy change in ways no one has a framework for”, that a boring old 60/40 portfolio is completely unequipped to handle. Your bonds aren’t going to save you in a world of “Ghost GDP” or an inflationary AI capex boom. Those are opposite scenarios and 60/40 gets smoked in both.
You know what does have a shot at navigating a world this uncertain? Strategies that don’t need to know what’s coming, they just need something to happen. Trend following. Managed futures. Volatility strategies. The kind of stuff that surfs the wave regardless of whether it’s a tsunami or a ripple pool.
As Jeff said on the pod, trend followers are surfers. They can’t make the waves. They just need to be ready when they show up. And brother, whether the Citrini doom loop materializes or Citadel’s “steady as she goes” thesis plays out, there are going to be some waves.
The canary might still be alive. But maybe it’s time to make sure your portfolio can handle a world where it’s not.
