THE 2028 GLOBAL INTELLIGENCE MIRACLE

The Bull Case On AI Written From the Future

I want to be honest about something before we dive in.

The Citrini “Global Intelligence Crisis” piece is genuinely brilliant. Read it twice. Citrini modeled a scenario where AI exceeds every technological expectation, and that excess breaks the economy. Ghost GDP. The Intelligence Displacement Spiral. White-collar workers flooding the gig economy and pushing down wages for the workers already there. A $13 trillion mortgage market built on income assumptions that no longer exist. It’s thorough, it’s sobering, and it’s the kind of work that makes you rethink your portfolio.

Here’s my constructive critique. It’s half a model.

It assumes AI exceeds expectations in capability and that human adaptability fails to keep up. The technological forecast? I think they’re right. AI will exceed expectations. Where I push back is on the other side of the ledger. The assumption that policy, markets, and human creativity stay static while the technology compounds.

That assumption has been wrong every single time in history. Not sometimes. Every time.

So this is my answer to that piece. Same format. Same June 2028 dateline. Same technological premises, because I think they got that right. Different outcome, because I think they underestimated the adaptive response.

Call it the bull case nobody’s writing.

This is my macro memo from June 2028.

The Number Nobody Expected

The unemployment rate printed 3.8% this morning. Consensus was 4.1%.

The S&P is up 61% from the October 2026 lows, the correction everyone now calls the Great Re-Pricing, back when it felt like the beginning of the end. It wasn’t. It was what corrections are supposed to be: a violent repricing of bad assumptions that clears the deck for the next leg higher.

I know what you’re thinking, because I was thinking it too in early 2026. The layoffs were real. The SaaS implosion was real. The ServiceNow print (14% net new ACV growth, down from 23%, with a 15% workforce reduction) was a genuine gut punch. The long tail of SaaS got obliterated, exactly as the bears predicted. The intermediation economy got hollowed out. Agentic commerce routed around interchange. DoorDash’s moat evaporated practically overnight.

Everything the pessimists said would happen, happened.

And here we are, two and a half years later, with 3.8% unemployment and a market that’s made new all-time highs.

Let me walk you through how.

The Adaptation Nobody Modeled

The bearish scenario has a telling assumption baked into it: that the new jobs won’t arrive fast enough, because AI can immediately do any role humans might transition into.

I spent time thinking hard about this. It sounded airtight in early 2026. It hasn’t aged well.

Here’s what actually happened.

The physical world didn’t get easier to automate just because the digital world did. Someone still has to crawl under your house and re-run the ductwork. Someone still has to set the femur. Someone still has to wire the panel box, fix the HVAC, frame the addition. Claude is extraordinary. Claude cannot do any of those things.

The wage premium on physical skilled trades, which had been grinding lower for thirty years as the cognitive premium expanded, inverted hard. By Q2 2027, licensed electricians in major metros were pulling $140,000 and getting signing bonuses. HVAC techs (a punchline for years among the laptop class) were fielding competing offers. Vocational enrollment surged 34% in 2026 alone. High schools that had gutted their shop programs in the early 2000s scrambled to rebuild them.

I’ve been tracking this through my portfolio. Companies in skilled trades and construction technology have been among the best performers in my book since mid-2026. That wasn’t an accident. It was a direct read on where the labor market was heading.

The second emergent category was accountability. This one took longer to understand.

AI agents proved extraordinary at the execution layer, faster, more thorough, and less error-prone than humans at most cognitive tasks. What they couldn’t do was own the outcome. Boards still need a human who can be sued. Patients still want a doctor who bears moral responsibility for the diagnosis. Clients still retain lawyers who can be disbarred. Teachers still need to be answerable to the parents and students in front of them.

A new class of professional emerged: the person who doesn’t add raw cognitive horsepower, but stands behind the work. They manage the exception. They absorb the liability. They hold the relationship. They make the call at 11pm when the algorithm produces something technically correct but humanly wrong.

The product manager who got laid off didn’t become an Uber driver. She became a Client Intelligence Director, spending 20% of her time managing AI agents and 80% with customers, translating capability into adoption, bearing the flag of the relationship. She made $215,000, up from $180,000.

The third category was the one nobody wanted to admit to. Authentic human creativity became more valuable, not less, when AI made production cheap. The market for things with genuine human provenance (music made by people with something to say, writing that carried a real perspective, craft objects made by actual hands) expanded dramatically when AI made everything else abundant.

Substack subscriptions grew 8x in two years. Independent music revenues exceeded major label revenues for the first time in recorded history. The number of working craftspeople doubled. It turns out that when everything becomes cheaply producible, scarcity reasserts itself. Now it’s the scarcity of genuine human origin, not cognitive throughput.

Human scarcity got more valuable when AI abundance arrived. Not less.

What Happened to Healthcare (And Why Nobody’s Talking About It Enough)

I want to spend real time here because this is the part of the story that the financial media has badly undercovered, probably because it doesn’t have a clear ticker to trade.

The healthcare transformation between 2026 and 2028 is, without exaggeration, the most significant development in medicine since the discovery of antibiotics. I don’t say that lightly.

In late 2026, AI-assisted drug discovery stopped being a research curiosity and became an operational reality at scale. AlphaFold had already cracked protein structure. What happened next was the application of similar reasoning to drug-target interaction, molecular design, and clinical trial optimization simultaneously.

The numbers are staggering. The average drug development cycle, historically 12-15 years from discovery to approval, compressed to under 5 years for several drug classes. The cost of bringing a novel therapy to market dropped by roughly 60%. Pfizer, Eli Lilly, and a dozen biotechs I’ve been watching closely posted pipeline acceleration that analysts simply hadn’t modeled.

More importantly, things that had been scientifically intractable started falling. A working Alzheimer’s therapeutic (not a slowing agent, an actual therapeutic) completed Phase III trials in Q1 2028. If you want to understand how emotionally significant this is, talk to anyone who has watched a parent disappear to that disease. The financial market for Alzheimer’s treatment globally is over $15 billion annually, and that’s before you account for the caregiver cost reduction, which runs into hundreds of billions.

A functional cure for Type 1 diabetes entered late-stage trials. Personalized cancer immunotherapy, matched to a patient’s specific tumor genomics by AI analysis, moved from experimental to standard of care in three major cancer types.

The diagnostic side was equally transformative. AI reading of medical imaging across radiology, pathology, and cardiology reached a point in 2027 where routine diagnostic reads were largely automated, with human radiologists focusing on edge cases and complex interpretation. What this means in practice: a patient in rural Georgia now gets the same quality of diagnostic reading as a patient at Mass General. The geographic lottery of American healthcare quality began, for the first time, to equalize.

On costs (and I know that’s what most of my readers actually care about) the compression has been real. Administrative costs in healthcare, which had ballooned to approximately 34% of total spending, began declining sharply when AI automated prior authorizations, claims processing, coding, and scheduling. One mid-size hospital system I looked at closely cut administrative overhead by 40% in 18 months. That’s not a rounding error. That’s the difference between a hospital that operates at a loss and one that doesn’t.

Health insurance premiums, which had risen for a decade straight, declined in 2027 for the first time since 1999.

None of this is fully priced into the market yet, in my view. Healthcare is where I’ve been quietly building positions.

What Happened to Crime (And Why This Should Be Part of Every AI Conversation)

This doesn’t get covered because it doesn’t fit neatly into a portfolio thesis. But it matters more than most of what we talk about.

Human trafficking is a $150 billion annual criminal enterprise. It’s the second-largest criminal industry in the world, behind only drug trafficking. It’s also extraordinarily dependent on the same things that make any supply chain work: communication, coordination, financial flows, and information asymmetry between operators and authorities.

AI destroyed that information asymmetry.

Pattern recognition across financial transaction data (wire transfers, crypto flows, prepaid card usage) that would have taken a team of human analysts months now takes minutes. By Q3 2027, the National Center for Missing and Exploited Children reported a 67% improvement in time-to-identification for trafficking networks, driven almost entirely by AI-assisted financial surveillance tools deployed by Homeland Security and partnered NGOs.

The Polaris Project, which runs the National Human Trafficking Hotline, integrated an AI triage system in mid-2026. The system screens incoming contacts, identifies high-probability trafficking situations, and routes them to the appropriate response within minutes instead of hours. Victim identification rates increased 40% in the first year.

This is one area where I’ll be direct about the geopolitical angle. Our biggest adversaries run state-adjacent cybercrime operations that had, for years, outpaced U.S. investigative capacity. The AI capability advantage flipped that equation. American intelligence and law enforcement agencies now operate with analytical capability that adversaries haven’t matched. I’ve seen nothing classified (I want to be clear about that) but the open-source signals from indictment rates, network disruption announcements, and asset seizures tell a clear story.

Drug trafficking networks faced a similar disruption. AI modeling of precursor chemical flows, the raw materials that become fentanyl, allowed DEA and partner agencies to interdict supply chains rather than just street-level product. Fentanyl seizures at the border increased 58% in 2027 while intelligence analysts estimated supply to street markets declined. That’s a different kind of result than the seizure numbers alone would suggest.

Financial fraud, the more mundane crime that hits ordinary people, saw equally dramatic improvements. Real-time AI transaction monitoring at major banks reduced fraud losses by roughly $40 billion annually in 2027. Elderly Americans, who had been disproportionately targeted by sophisticated phone and email scams, benefited most from AI call screening tools that identified scam patterns in real time.

This isn’t just humanitarian. It’s economic. Fraud, trafficking, and drug distribution impose costs on legitimate businesses and government systems that ultimately get passed through to consumers and taxpayers. Reducing those costs is a real economic tailwind.

The Friction Economy Was Always a Tax

The agentic commerce revolution happened exactly the way the bears predicted. Qwen’s open-source shopping agent dropped, every major assistant integrated commerce features, and by early 2027 the median American was running agents that renegotiated subscriptions, re-shopped insurance, routed around card interchange, and assembled travel itineraries with no human platform in the loop.

DoorDash’s moat evaporated. Mastercard’s volume growth decelerated sharply. Insurance renewal economics compressed. Real estate commissions collapsed to under 1% in major metros.

Here’s the thing the bears missed: that money didn’t disappear. It moved.

The 2-3% interchange that disappeared from Mastercard’s income statement went into consumer wallets. The 15-20% of insurance premiums that funded passive renewal inertia stayed with policyholders. The 5-6% real estate commissions that compressed stayed with buyers and sellers.

Aggregate it across the economy: somewhere between $800 billion and $1.2 trillion of annual value transferred from intermediaries to households. GDP looked weaker because it counts the intermediary’s fee as output. Household welfare improved substantially.

I wrote in early 2026 that the biggest risk to my portfolio was companies whose moats were built on habitual friction rather than genuine value creation. I sold positions in several names on that thesis. The intermediaries I kept were the ones that created genuine value and would survive a world where consumers had perfect price information and zero switching costs. That call has held up well.

The Policy Response (The Part That Surprised Me Most)

I’ll be honest: I was skeptical.

The scenario I was running in my head in mid-2026 included a version where congressional gridlock, incumbent lobbying, and partisan warfare produced a policy response too slow and too small to prevent the negative feedback loop from locking in. I’ve followed Washington long enough to be cynical about its ability to move at the pace that technology demands.

I was wrong. Not completely, because the process was ugly and slower than it should have been. But it worked.

In Q2 2027, Congress passed the American Worker Transition Act in a 71-29 Senate vote. The bill extended unemployment insurance to 24 months for documented AI-displaced workers, funded retraining at scale through community colleges and apprenticeship programs, and established what got called the National Productivity Dividend, a quarterly direct payment to every American household funded by a 1.5% levy on commercial-scale AI inference compute.

The Productivity Dividend averaged $620 per household per quarter by mid-2028. That’s not transformative on its own. Combined with the retraining programs and the lower consumer prices from the friction economy’s collapse, it provided enough of a consumption floor to prevent the demand spiral that the bear case required.

The other piece that worked was the mortgage intervention. The FHFA implemented a Structural Income Bridge for households with documented AI-related income displacement, a federally-backed modification program that allowed them to restructure without triggering default or credit damage, in exchange for equity sharing on eventual upside.

The $13 trillion mortgage market didn’t crack. Delinquencies in tech-employment ZIP codes rose, peaked in Q3 2027, and began declining as retraining graduates entered the trades and other shortage occupations.

The irony that the technology threatening the mortgage market was deployed to administer the program protecting it was not lost on anyone paying attention.

The Private Credit System Bent, it Didn’t Break

PE-backed software LBOs that were underwritten at 25x EBITDA against ARR assumptions that couldn’t survive the agentic coding era were impaired, and the marks that came through 2026-2027 were painful. The Zendesk restructuring was real. So were a dozen others.

What the pessimists got wrong was conflating impairment with systemic failure. The financial architecture of private credit (closed-end vehicles with locked-up capital, no depositors to run, no repo lines to pull) proved to be exactly the shock absorber it was designed to be.

The difference between painful and catastrophic came down to one number: loss rates of 15% on the software book versus the 40% the bear case required. That difference came from the operational response of the portfolio companies themselves.

Companies most threatened by AI became AI’s most aggressive adopters. The bears got that part right. What they underestimated was how many of those companies emerged from restructuring as leaner, AI-native businesses that could actually service their debt at compressed multiples.

The PE firms took hits. None of them failed. The life insurance scaffolding absorbed stress. The regulators tightened capital treatment for private credit at insurers, which was real and necessary, but the system processed it without a crisis.

I’ve owned Blackstone since before this started. The drawdown was brutal and I held. It’s been one of the better performers in my portfolio.

The Geopolitical Piece

India’s IT services sector contracted hard in 2026-2027. That part of the bear thesis landed. TCS and Infosys lost contracts. The rupee came under pressure.

India’s response was faster than most emerging market skeptics would have predicted. They repositioned toward what I’d call AI-managed physical world services: remote industrial supervision, medical diagnostics review, infrastructure monitoring, AI system auditing. English fluency, technical training, and cost competitiveness remained real advantages in roles where human judgment and accountability still mattered.

The current account surplus contracted, then rebuilt on a different foundation. India ended up in a stronger negotiating position at the G20 on AI governance because they had a credible story about managing the transition.

Taiwan and Korea, whose economies were essentially pure convex plays on AI infrastructure spending, outperformed dramatically. That was the easiest call I made in this whole cycle. Back the infrastructure build, and find the most concentrated beneficiaries of the hyperscaler capex cycle.

What 3.8% Unemployment Actually Means

Real median household income grew 6.2% in 2027. Fastest rate since the late 1990s tech boom.

The components are not what you’d expect. Wage income grew 2.1%, below the historical average. The Productivity Dividend contributed 1.8 percentage points. The deflation in friction-intensive categories (insurance, financial services, subscription software, travel, consumer goods) contributed the equivalent of 2.3 percentage points in real purchasing power.

The economy produced more with less human labor. Households captured a meaningful share of those gains through direct redistribution, through lower prices, and through new jobs in sectors where human presence retained genuine premium value.

Labor’s share of GDP is lower than it was in 2024. That compression raises legitimate long-run questions I don’t want to dismiss. The transition was genuinely painful for millions of workers who got displaced faster than anyone wanted. Private credit absorbed real losses.

But the S&P is at 9,200, not 3,500. Unemployment is 3.8%, not 10.2%. A working Alzheimer’s therapeutic is in Phase III. Trafficking networks are getting dismantled at a pace that would have seemed fictional five years ago.

This is what the abundance looks like.

The Portfolio Takeaway

I want to be concrete about what this scenario means for where I’m putting money.

The companies that win in the abundance scenario are not the companies with the most AI. They’re the companies that create genuine value in a world where cognitive throughput is cheap and human presence, physical capability, accountability, and authentic creativity command real premiums.

Healthcare, specifically drug discovery infrastructure, diagnostic platforms, and companies that help health systems absorb and deploy AI capability, is where I’ve built my largest new positions over the past 18 months. The tailwinds are generational and the market has been slow to price them.

Infrastructure, the physical layer that the intelligence economy runs on, remains the most straightforward bet. NVDA, TSM, IREN, NBIS, and the hyperscalers themselves have continued printing despite everything, because the buildout hasn’t stopped. I still own all of them.

The trades and physical skilled labor economy is a theme I’ve been playing through smaller names in construction technology, vocational training platforms, and field service management software. Less glamorous than the AI hardware names. Better risk/reward than most people realize.

The intermediaries I avoided, I’m still avoiding. Companies whose moats were built on human friction (passive inertia, information asymmetry, habituation) are structurally impaired regardless of how this scenario plays out. That’s not a cyclical view. It’s a permanent one.

The Last Thing

You’re not reading this in June 2028. You’re reading it, most likely, in early 2026.

The layoffs haven’t really started in earnest yet. The policy architecture doesn’t exist. The Productivity Dividend hasn’t been proposed. The Alzheimer’s therapeutic is still years from Phase III. The trafficking networks are still intact.

Which version of 2028 we inhabit is not determined by the technology. The technology is going to do what it’s going to do, and I think the bulls have that right.

It’s determined by what we build around the technology in the next 18 to 24 months. The policy response, the retraining infrastructure, the safety net design, the willingness of the financial system to absorb losses without seizing up, and the human creativity and adaptability that have outperformed every dire prediction in every prior technological transition.

I’m long the adaptive response. I’ve been long it my entire investing career. Every time I’ve bet against human adaptability, I’ve been wrong.

The Citrini piece ends by saying the canary is still alive. I agree.

My view is that the canary is going to be fine.

Position accordingly.

Nothing here is investment advice. Do your own work.

This was written as a mirror image of the Citrini Research “Global Intelligence Crisis” scenario published February 2026. Same technological assumptions. Same timeline. Different human response.