Weekly Dispatch on AI, Geopolitics and Markets #16

President Trump declared victory in the conflict with Iran in Kentucky yesterday, saying the US had “virtually destroyed Iran” and destroyed 58 Iranian naval ships. He added that the first phase of the war was “over,” pointing to U.S. air superiority and the heavy damage inflicted on Iranian forces. But the declaration came with a caveat. He also said U.S. forces would remain in the fight to “finish the job,” warning that withdrawing too quickly could force the United States to return to the conflict later.

On the ground, the situation looks far less settled. Iran has used missile and drone attacks to disrupt shipping through the Strait of Hormuz, temporarily blocking much of the traffic through one of the world’s most critical energy chokepoints. Several Gulf states have already curtailed oil and gas production as a precaution. The disruption is beginning to ripple through global markets. Energy prices have surged, and stock markets have weakened worldwide

This is classic Trump messaging: “We won… but we also have to keep fighting.” Those two things don’t add up. If Iran’s navy and air force are destroyed and the war is “over,” why are U.S. forces still staying?

What Iran demonstrated in this conflict is the region’s real leverage: oil. The war is asymmetrical. Tehran doesn’t need to defeat the U.S. military. It only needs to make the Strait of Hormuz dangerous enough that tankers stop moving. Once oil prices spike, the global economy feels it instantly, and Trump suddenly finds himself dealing with rising inflation and gasoline prices as the midterm elections approach.

That’s the pressure point. Trump cannot afford a sustained surge in oil prices. Between economic pressure and pushback from parts of his own MAGA base that oppose another Middle East war (or any costly war), Iran’s strategy may already be working.

This shouldn’t be surprising. Iran has centuries of statecraft behind it. Persians invented chess for a reason. On the geopolitical chessboard, you shouldn’t expect Iran to lose quickly in a war the regime has been preparing for decades.

Now the United States is stuck in the classic Middle East trap. If the war stops, Iran’s regime survives still angry, still armed with missiles, and still capable of threatening the world’s most important oil chokepoint. But if the U.S. escalates, the only real path to “total victory” runs through ground troops and regime change. That road leads straight back to Iraq.

Trump also created a political commitment problem. He openly encouraged anti-regime protests inside Iran. What happens now if the United States stops short of regime change? What do you say to people who risked their lives believing America would finish the job?

I saw hundreds of Iranian diaspora members in Georgetown, Washington DC the day the attacks began, waving U.S. and Trump flags, celebrating what they believed was the beginning of the end for the regime.

What happens when the U.S. eventually pulls back, the regime remains intact, perhaps even stronger, and many Iranians have already died protesting or under the bombs?

The legitimacy battle matters too. Even if the United States wins militarily but loses economically, Iran could still win the global narrative. Many NATO allies are already uneasy about the war’s legitimacy. Images of civilian casualties spread far faster than strategic explanations. After Iraq, Washington spent years and billions trying to counter anti-American sentiment. This conflict risks fueling it again.

Meanwhile, Russia and China are watching carefully. Beijing is studying whether the United States can control escalation and stabilize conflicts or whether it gets dragged into another costly geopolitical drain. Some analysts already argue that the so-called Davidson Window—the period when China might move on Taiwan before 2028—becomes more plausible if Washington is distracted elsewhere.

In that sense, the consequences of this war extend far beyond Iran. They shape how safe the world will be in the years ahead. And if this proves to be another strategic miscalculation, the cost will be measured not only in American lives, taxpayer dollars, and geopolitical credibility. It could also erode the global stability, legitimacy, and order that have underpinned international peace for decades.

Fukuyama’s “End of History” thesis—arguing that humanity had reached its political endpoint with the triumph of Western liberal democracy —was a dominant framework when I was studying international relations in college.

Later in my studies, I had the privilege of meeting him and attending his classes at SAIS in DC. Last week, Dr. Fukuyama shared the following article on X, and reading it made me even more pessimistic, especially if this reflects how his thinking has evolved over the past two decades.

A new academic paper led by Nobel Prize–winning economist Daron Acemoglu warns that artificial intelligence could unintentionally weaken what economists call the “knowledge commons” — the shared pool of publicly available solutions, explanations, and technical insights that fuel learning and innovation. Historically, engineers, researchers, and developers have documented their problem-solving in public spaces such as Stack Overflow, academic journals, and technical blogs. Those contributions accumulate into a collective knowledge base that others can study, reuse, and improve.

AI-assisted workflows could alter that pattern. In the example highlighted by the authors, a software engineer might diagnose a bug with an AI assistant and ship the fix immediately. The problem gets solved, but the reasoning process never appears in public forums. Over time, fewer documented solutions contribute to the knowledge commons that historically supported innovation.

The paper raises a broader question about how AI could reshape the production, distribution, and access of knowledge. While AI may significantly boost individual productivity, it could also move more problem-solving into private human–machine interactions rather than public collaborative spaces.

I think Dr. Acemoglu is asking the right question. But I also think the framing misses the most important issue. Yes, AI compresses the visible problem-solving process. In the past, when people had a question, they would post it publicly on forums, blogs, mailing lists, and debate solutions with other humans. That created the internet’s “paper trail” of knowledge.

Today, many of those conversations are happening privately between humans and chatbots or other AI systems. From the outside, it can look like knowledge sharing is shrinking. But I don’t think the real story is the disappearance of the knowledge commons. I think it’s the transformation of how knowledge is created and aggregated.

The old model of the commons was fragmented. Valuable insights existed, but they were scattered across blog posts, forum threads, PDFs, and academic papers. The information was there, but synthesizing it was messy and slow. Until the internet revolution, much of this knowledge wasn’t even accessible. It was limited to those who knew where and how to find it, which was itself a major constraint. AI opens the door to something very different: structured collective intelligence.

When you combine AI systems with expert human judgment, you can build networks where insights are captured, synthesized, and redistributed far more effectively than the old internet ever allowed. Instead of isolated answers scattered across thousands of pages, you get living knowledge systems that constantly update and improve.

That’s actually the philosophy behind what I’m trying to do with “Centaur Strategy”. The idea is simple: humans plus AI working together to produce better analysis, better forecasts, and ultimately better decisions. In that world, the knowledge commons doesn’t disappear—it evolves.

We need to design systems in which human expertise and AI capabilities reinforce each other—and in which the knowledge produced flows back into the public domain. If we get that right, AI won’t weaken the knowledge commons. It will expand it dramatically.

In my opinion, the real risk isn’t that AI shrinks the commons. The real risk is that knowledge generated by AI remains locked within private systems and corporate platforms.

A new analysis from forecasting firm Good Judgment finds that professional “Superforecasters” outperform prediction markets when forecasting complex economic decisions. The study compared expert forecasts with predictions from the crypto-based platform Polymarket across 25 recent central bank meetings involving the Federal Reserve, the European Central Bank, the Bank of England, and the Bank of Japan.

Using the Brier score—a standard metric for probabilistic forecasts—researchers found Superforecasters averaged 0.102, while Polymarket averaged 0.126, making the market predictions about 24% less accurate.

The findings challenge claims that decentralized prediction markets are the most accurate forecasting tools, suggesting that structured expert analysis may perform better when predicting complex policy decisions such as central bank actions.

I’ve always trusted subject-matter experts more than pure crowd sentiment, and this study reinforces that instinct. (Full disclosure: I’m biased. I’m the founder and CEO of an AI-powered expert network, Enquire AI, so I naturally believe more in experts than in gamblers placing bets on markets.)

I had shared a Fed report last week. Prediction markets aggregate opinions, and sometimes that works. But many participants are traders chasing profit, not analysts carefully evaluating macro data and central bank behavior. Forecasting interest rate decisions isn’t a popularity contest.

It requires probabilistic thinking, institutional knowledge, and structured analysis, exactly what professional forecasters are trained to do. But results like this point to a broader lesson: when deep expertise matters, structured analysis can outperform the raw wisdom of markets.

Market sentiments have been shifting with remarkable speed. Just weeks ago, the dominant narrative was resilience. The U.S. economy looked strong, growth was holding up, and AI investment was seen as a powerful tailwind. Now the conversation is drifting toward something very different: the possibility that AI could drive mass unemployment.

Another concern quietly emerging is the risk of a K-shaped economy. AI may boost productivity and profits in capital-intensive sectors while displacing workers in routine or knowledge-based roles. That kind of divergence could leave parts of the economy thriving while others face rising unemployment and wage pressure. In that scenario, headline growth could remain solid even as labor markets weaken beneath the surface—a dynamic that would make the Fed’s policy decisions far more complicated.

Adding to the uncertainty are geopolitical risks in energy markets. Tensions in the Middle East and potential disruptions around the Strait of Hormuz are beginning to raise concerns about another oil-driven inflation shock. In other words, markets are starting to price in a complex set of converging risks: AI-driven labor disruption, uneven growth, geopolitical tensions, and lingering inflation risks. Investors increasingly seem to realize that this future could arrive sooner than expected.

🧭 Category > Company: The Real Alpha in Venture Capital
An interesting take on venture investing argues that the biggest edge isn’t picking the best startup, it’s picking the right category. If you correctly identify where the next major wave will happen, many companies in that category can succeed. https://x.com/nidmarti/status/2030664111286042808

🏗️ Services Are Becoming the New Software
Sequoia recently published a report arguing that the next generation of startups may look less like traditional SaaS and more like service businesses powered by technology and AI. Ironically, just a couple of years ago, every service company tried to present itself as a “tech platform.” Now the dynamic is reversing: being a high-quality service company augmented by software and AI may actually be the differentiator. https://sequoiacap.com/article/services-the-new-software/

🤖 How AI Is Actually Affecting Jobs
Anthropic released a data-driven study examining how AI tools are being used across occupations. The results show a more nuanced picture than many headlines suggest: AI is mostly augmenting tasks rather than fully replacing jobs. But its adoption is spreading rapidly across knowledge work, gradually reshaping how work gets done rather than eliminating roles overnight.
https://www.anthropic.com/research/labor-market-impacts

🔐 Could Quantum Computing Break Bitcoin?
ARK Invest explores whether advances in quantum computing could eventually threaten Bitcoin’s cryptographic security. The analysis examines the technical barriers, potential timelines, and why the risk may be further away than some headlines imply. Still, as quantum capabilities evolve, it’s an area investors and technologists will want to monitor closely.