How the World’s Biggest Macro Hedge Funds Are Using AI | Jan Szilagyi

Episode Summary

Executive Summary: The episode explores how AI is being used in hedge funds and macro investing, centered on Reflexivity CEO Jan Zaleghi’s view that LLMs can dramatically accelerate insight generation, expand the searchable universe of comparable events, and reduce hallucination risk through code-first, auditable outputs. The discussion also covers where AI helps most today, how it may affect alpha generation, labor markets, and monetary policy, and why the AI boom will require major real-world infrastructure and commodities.

Main Topics: Jan Zaleghi’s background in global macro and hedge funds (Priority: 5/5): Zaleghi explains his path from Yale math to working alongside Stanley Druckenmiller, a Harvard quant finance PhD, and later managing global macro portfolios at Fortress with Mike Novogratz before founding Reflexivity. Why the company is called Reflexivity (Priority: 5/5): He links the name to Soros-style reflexivity: markets affect fundamentals and fundamentals affect markets. The product aims to help investors and decision-makers understand those feedback loops in real time. How AI changes macro research and thesis testing (Priority: 5/5): Zaleghi argues AI lets analysts ask more ambitious questions, synthesize many variables at once, and test macro theses faster—especially valuable when sample sizes are small or historical analogs are limited. What Reflexivity does differently from general-purpose LLMs (Priority: 5/5): The platform combines LLM reasoning with a proprietary knowledge graph and premium financial data sources so it can run code-first analyses, map first- and second-order effects, and avoid hallucinated answers. Current and emerging use cases in hedge funds (Priority: 4/5): The biggest near-term use cases are rapid thesis exploration, finding analogs across long-tail datasets, and specialized workflows in macro, equities, and commodities. Execution and self-analysis of trading behavior are also seen as promising agentic uses. AI’s impact on alpha, labor markets, and central banking (Priority: 4/5): Zaleghi expects AI to be initially additive to alpha discovery rather than fully replacing human investors, while also causing labor displacement and large productivity gains that could eventually shape inflation and Fed policy. Infrastructure, commodities, and the cost of the AI boom (Priority: 4/5): He stresses that AI progress depends on physical inputs—compute, data centers, chips, metals, and energy-related commodities—so the boom is both disinflationary in the long run and inflationary in the near term due to real-world buildout needs.

Key Arguments: AI allows investors to ask questions that previously would have been dismissed as too outlandish because answers can now be generated in minutes instead of hours or days. Global macro especially benefits from AI because it often has small sample sizes; the system can expand the universe of analogs across countries, regimes, and commodities rather than relying on one narrow historical series. Understanding the economic logic behind a relationship can be more valuable than having a huge sample size; if the mechanism is clear, a handful of observations may be enough to act on a thesis. Reflexivity is designed to reduce hallucinations by being code-first, data-grounded, and auditable, returning an error when it lacks data instead of inventing an answer. General-purpose LLMs are useful for exploration, but specialized finance tools are needed for reliable quantitative analysis and for accessing hard-to-map premium data sets. AI is likely to create more dispersion in the near term because some market participants will use it effectively while others will not. In the longer run, widespread AI adoption may compress some forms of alpha, but differing risk appetites, horizons, and implementation styles should preserve trade diversity. AI is likely to be strongly useful in equities and commodities because these areas contain large, messy data sets that are difficult for humans to process comprehensively. Agentic AI could improve execution and even diagnose behavioral mistakes in a trader’s own process, such as poor timing or repeated errors on Fridays. The AI boom is likely to be disinflationary over a multi-year horizon via productivity gains, but the buildout itself requires massive real-world investment in compute, chips, and commodities, which can be inflationary in the meantime.

Data Points: Years off for PhD: About 2.5 years - Zaleghi says he took about two and a half years off to do a PhD in quant finance at Harvard. Macro sample size example: 7 or 8 occurrences - He says a steep contango in oil prices had only about seven or eight comparable occurrences, yet the mechanism made the signal actionable. AI horizon for predictions: Shorter term: up to about 5 years - He repeatedly limits predictions beyond roughly five years as too speculative to be useful. Demo turnaround time: About 8 minutes - He describes a Reflexivity demo where the system analyzed the proposed Coca-Cola sugar substitution in around eight minutes. Potential manual work saved: 2 to 3 days of analysis - The Coca-Cola/sugar example produced roughly two to three days of work in minutes. Mentioned data threshold: Sub-50 PMI trigger - A listener example referenced using AI to study what assets perform after PMI and ISM both move above 50 after a long period below 50. Previous AI workflow improvement: Under 10 minutes - The host says Claude helped fix a house issue in roughly 10 minutes, illustrating practical AI assistance.

Pivotal Quotes: "This was a technology that should go a long way towards helping us truly extract insights from the data on a scale that was never possible before." — Host/Intro narration: Opening framing of the episode’s thesis on AI in finance. "It was capable of doing what I guess I could describe as multi-dimensional synthesis." — Jan Zaleghi: Explaining why LLMs are powerful for macro analysis compared with linear human reasoning. "Every answer that you get from reflexivity is ultimately code first output." — Jan Zaleghi: Describing how the product avoids hallucinations and keeps analysis auditable.

Implications: AI is becoming a real research and workflow advantage in finance, especially for macro and commodities. Near term, edge may widen between adopters and laggards; longer term, the biggest winners will pair AI with strong questions, clean data, and disciplined execution.