The AI Metagame

Episode 2 | Controller | Over-Optimization: Why efficiency kills resilience.A perfectly optimized system is just a fragile one waiting to break. Our guest explains why they stopped optimizing for margins and started optimizing for survival, and what that means when AI moves faster than your org can adapt.We've spent decades chasing efficiency. Removing friction, maximizing output, trimming every edge. But what happens when the system has zero slack left and the ground shifts overnight?This episode unpacks the Controller archetype: the leader who builds tight, optimized machines that work beautifully until they don't. We challenge the Explainer instinct to rationalize away uncertainty, and explore why the smartest move right now might be building in inefficiency on purpose.

Episode 1: Builder | Controller | Premature Abstraction"I built for scale before I understood the problem"In a world where tools allow us to scale instantly, the temptation to build big is overwhelming. But what happens when you scale the wrong thing?The AI Metagame is a twelve-episode series that looks at four archetypes in AI — Builder, Explainer, Interpreter, Narrator — to map where AI creates leverage and where it creates traps. For every great advantage of AI, there are disadvantages lurking. Each episode explores both sides.In our premiere episode, we explore the Builder's most dangerous trap: Premature Abstraction. Modern AI leverage is a double-edged sword — it amplifies errors faster than ever before. True control now requires staying small enough to pivot until the problem itself is fully understood.

Alex Hutchinson (@sweatscience) is my favorite fitness journalist. This podcast was an absolute pleasure. We had the opportunity to explore his outstanding new book, The Explorer's Gene. We also discuss, in detail, his 2018 NY Times bestseller Endure: Mind, Body, and the Curiously Elastic Limits of Human Performance. I rate this book among my top ten favorite works of non-fiction.

Jared Dillian (@dailydirtnap) publishes daily market commentary (The Daily Dirtnap, www.jareddillian.com). He hosts a twice weekly podcast called Be Smart. He has a popular Substack (https://substack.com/profile/450570-jared-dillian), and he's the author of Street Freak and All the Evil of This World. His latest book, Those Bastards: 69 Essays on Life, Creativity, and Meaning, was published in late March.

Scott Fawcett (@scottfawcett) is the creator of the DECADE golf system and he has coached Will Zalatoris, Stewart Cink, Keith Mitchell, and many others. In the first hour, we do a deep dive into golf analytics. He describes his career in professional golf, his early discovery of the PGA Tour's Shotlink Data and Mark Broadie's Strokes Gained Metric, his development of the DECADE system, and his experience coaching PGA players. At the one hour and three minute mark, we are joined by David Epstein (@DavidEpstein), author of the bestselling books Range and The Sports Gene. The three of us discuss talent in sports, similarities between golf and poker, "tilt" in golf and poker, the history of game theory in poker, and the emotional make-up of poker players.

Daniel Negreanu (@realkidpoker) is fresh off a $3.3 million win in the Super High Roller Bowl. Perhaps the most famous poker player on the planet, he is ranked #3 on the all-time money list. We start with a discussion of the new era in poker ("the Solver era") and some recent adjustments he's made in his game. We move on to some deep history - his move from Canada to Vegas; his early years with Phil Ivey, Allen Cunningham, and John Juanda; some very big gambling in the early years; and the "pump-up years" in the poker economy after WSOP 2003, Rounders, and the Andy Beal games. He tells some great golf-gambling stories from the late 2000s, and he explains how his competitive goals shifted from cash games to tournaments. Daniel goes through his WSOP routine each year, which includes a prep period, followed by extreme intensity during the WSOP (where he often gains ten to fifteen pounds), and then a long rest period.

David Benefield (@DWBenefield) is a poker legend with an incredible story. He started poker at age 16 after watching the movie Rounders. He dominated the biggest games online from ages 21-26. By his own estimate, he worked 100 hours a week during those years. In 2010, he went back to school at St John's University and then Columbia University (transfer after sophomore year). After his first semester at Columbia, he took a year and a half off to play the biggest poker game in the world in Macau. He also made the Main Event final table during this sabbatical. He ultimately finished his degree at Columbia, and then became increasingly interested in the cryptocurrency sphere. Today he employs eight people in a cryptocurrency-focused fund, based out of Dallas.

Brad Delong is an economic historian at UC-Berkeley. A former Deputy Assistant Secretary of the Treasury in the Clinton Administration, he frequently writes about policy issues on Twitter as @delong, where he has over 80,000 followers, and on his popular blog Grasping Reality (https://braddelong.substack.com). He is the author of a new book, Slouching Towards Utopia, that has been widely lauded as the most important book in economic history in many years.

Shaun Deeb is, far and away, the most successful player at the World Series of Poker over the past decade (when evaluated based on WSOP leaderboard points). We discuss, among other things: his daily routine, WSOP fantasy poker, his view on tournament edges, his rise in poker (playing at Turning Stone in NY as an 18yr old) to his near fall (playing in Mexico post Black Friday), mixing family and poker, and his philosophies about learning in poker.

Eric Olson (@IPlayedD1) is the co-founder of Consensus, a leading start-up in natural language processing. I was interested in learning about the specific topic of AI/machine learning in the social sciences, and Eric was a brilliant guide in this area. Eric is a former football player at Northwestern who graduated with a master's in predictive analytics. With co-founder Christian Salem, he build Consensus, a search engine that uses AI to extract findings from scientific research.