Chamath Palihapitiya and Rick Rubin: Poker, Intuition and the Ego Problem

Chamath Palihapitiya tells Rick Rubin how poker taught him to take big bets and absorb losses, and admits that ego and insecurity drove his worst financial mistakes. The nearly two-hour conversation on Tetragrammaton with Rick Rubin covers the accidental start of the All-In podcast, the 2022 comment that got him nearly canceled, the medical loss ratio, poker as a lifelong game against yourself, how he uses AI as a decision filter, the Groq deal with Nvidia, why he thinks crypto is capped, and the regrets he says he carries every day. It is less a business interview than a man auditing his own judgment out loud, with Rubin occasionally pushing back.

TLDW

Palihapitiya describes himself as a gambler who is wrong 80 to 90 percent of the time and lives off the other 10. Poker, he argues, is a single infinite game that trains you to deviate from conventional wisdom and to absorb loss aversion, and he says investing differs only because money can bend the outcome, especially in fringe ideas. He credits his wife with separating his intuition from his insecurity, now runs important decisions through an AI setup built around the Nash equilibrium, and lists his biggest mistakes (leaving Facebook early, a gratuitous Stanford speech, a run of attention-seeking deals, a huge credit line, refusing to sell in 2022) as products of ego. Rubin’s main counterpoint is that great work comes from the making game, not the outcome game, a line that visibly rattles him.

Thoughts

The most revealing thing about this conversation is how consistently Palihapitiya diagnoses his own failures, and how consistently the diagnosis is the same. Leaving Facebook before the job was done, the Stanford speech where he wanted “to inflict the pain of it” on former colleagues, the run of splashy deals that felt “godlike,” the bank line of credit taken because “as much as you’re willing to give me” sounded good, refusing to liquidate when Russia invaded Ukraine because selling would make him look like “a tactical bumpkin,” and scaling his fund down while competitors scaled up by an order of magnitude: every one ends with ego or insecurity overriding a signal he says he could already see. That matters because it reframes the rest of the interview. The poker talk, the taste talk and the AI tooling are not abstract philosophy. They are a set of guardrails a self-described insecure person has built to stop his own worst instinct from making the call. Whether they work is an open question, but the honesty of the inventory is rarer than the advice.

The sharpest exchange belongs to Rubin. When Palihapitiya says he has had hits but has never made “an album” he is proud of, and that he daydreams about building something people would say “we made that together,” Rubin calls it what it is: he is still in the outcome game, and the artist is in the making game. Palihapitiya admits to “an internal panic” at hearing it. It is a real tension rather than a platitude. An investor is paid on outcomes by definition, and everything he says about publishing his returns every year, up or down, is about being measured. Rubin’s claim is that the measurement itself bends the work. His own answer on revisiting records (“never” wanting to change them, because he followed his intuition until he thought the thing was great) is the opposite of Palihapitiya’s habit of mentally diffing old decisions against what he would do today. The two men are describing the same tool, intuition, used under very different scoring rules.

The poker argument is better than the usual “poker teaches risk” cliché because of one specific mechanism. Palihapitiya says 80 to 90 percent of the time the right move is the conventional one, in poker and in life, and the hard part of deviating is loss aversion: you mentally discount everything you would give up and freeze. His less obvious point is what happens to people who flirt with the deviation and then retreat. You cannot live the conventional path the same way afterward, because you now know you chose it only because it felt safer. Pair that with his claim that luck is close to 100 percent in a single hand and mostly gone after about 10,000 hands, and that every session you ever play is one continuous game, and you get a coherent case for poker as practice at making bets you might lose half the time while judging yourself on process rather than results.

His description of how he actually uses AI is more modest, and more credible, than most executive AI talk. He does not claim to be more productive. He says he is “limited to the same number of good decisions” a year, two in his example, but makes them better, with “less turbulence” on the way. The setup he describes is a model loaded with summarized books (Dale Carnegie through psychoanalysis) and a Nash equilibrium framing that asks where a deal, an email or an investment is exploitable, with the explicit job of separating intuition from insecurity and ego. That ties straight back to the regrets: the tool is aimed at the exact failure he keeps naming. The detail that one model ended a conversation on him after he berated it, and that he went back, reread the chat and conceded he was over the line, is a small but telling example of the filter doing its job.

Incentives are his master key, and they are useful, but some of his numbers deserve a second look. His account of the Obamacare medical loss ratio is a known critique: if an insurer’s administration and profit are capped as a share of premiums, the easiest way to earn more dollars is for total spending to grow. But the rule caps that share at 15 to 20 percent for insurers, not for hospitals and drug makers, who set many of the prices he blames. His claim that 60 to 70 percent of research money goes to overhead also seems to read a negotiated indirect cost rate, which is charged on top of direct costs, as a share of the total. A 60 percent rate works out to roughly 37 cents of each grant dollar, still large but not two thirds. The broader argument, that perverse funding incentives push researchers toward watered-down proposals, friendly peer review and work that does not reproduce, holds up without the inflated figure.

Key Takeaways

All-In started during COVID because a group that played poker every Thursday at Palihapitiya’s house was locked down and began recording themselves. He says the show recently hit number one in the US.
He describes the four hosts as archetypes: Jason Calacanis the loyal, infuriating Brooklynite, David Sacks the strategist, David Friedberg the earnest first-principles thinker, and himself the “insecure immigrant type.”
His Uyghur comment in January 2022 led the Golden State Warriors, where he was a co-owner, to distance themselves, and drew a White House response. A friend’s advice to never apologize if he believed it shaped how he responded.
He calls that episode the best thing the podcast ever gave him, because knowing he survived it makes him less afraid to say what he thinks.
He argues social and AI ranking systems are “fame weights” built around attention, and that fear of the fringes stops people from discussing the 99 percent of issues in the middle.
He blames the Obamacare medical loss ratio for rising US healthcare prices and contrasts it with Canada, where he says his 85-year-old mother now waits a year and a half for an MRI.
He estimates $100 to $200 billion of roughly $500 billion in annual US government healthcare payments is waste, with fraud closer to $25 to $50 billion.
He considers prediction markets both a sign of societal decay and better forecasters than the New York Times, but for his own views he uses X, several AI models, Google, then a short written draft he rereads after a day or two.
He says his wife, raised in Milan, taught him taste and separated his intuition from his insecurity, including talking him down from a $200 million investment to $25 million.
He estimates his unconventional bets are wrong 80 to 90 percent of the time, and that luck dominates poker until roughly 10,000 hands.
Investing differs from gambling in one way: money can shape the outcome, and the more fringe the idea, the more a small amount of capital can move it, as with SpaceX reaching viability for what he estimates was under $2 billion.
He thinks AI models predict the next word, are not conscious, and differ most on subjective questions because their founding “constitution” documents differ. He backs open models as the antidote to a “modern form of feudalism.”
His big wins include buying Bitcoin heavily around 2010 to 2012, the Warriors stake as a hedge, and co-founding the chip company Groq in 2016 with the engineer behind Google’s chip, which he describes as sold to Nvidia last year for about $20 billion.
He thinks all crypto is capped by poor fungibility (central banks will not buy a traceable asset), leaky custody and the long-term risk of quantum attacks, and he says he personally owns no crypto now.
He argues money printing slows progress by keeping bad ideas alive longer, and that only the bond vigilantes can force governments to stop.

Chapters

0:17 How a Thursday Poker Game Became All-In

Palihapitiya traces the podcast to a long-running Thursday poker game with Sacks and Calacanis, later joined by Friedberg. When COVID separated them, they recorded their conversations and he decided to just post them. Rubin says the show felt like overhearing the highest level of insider information. Palihapitiya argues it works because people can tell when they are being sold something, and the hosts are not selling.

8:53 The Uyghur Comment, the Warriors and the Fame Machine

He recounts the clip of him telling Calacanis to stop virtue signaling about the Uyghurs, which went viral over the Martin Luther King weekend. The Warriors distanced themselves and the White House commented. He stood his ground on a friend’s advice and now calls it the best thing the show gave him. He then argues that the algorithms shaping public perception are attention machines, and that fear of the extremes chills ordinary debate.

21:30 Healthcare, the Loss Ratio and One Shot on Goal

His father’s 30 years of chronic illness made him a fan of Canadian universal healthcare, but he says waits have since become unacceptable. He blames the medical loss ratio for US price inflation and explains why it never gets fixed: each new president gets one big legislative shot, and an esoteric fix does not win midterms. The discussion moves to incentives, AI-driven determinism in claims coding and policing, and prediction markets as truth machines that drift into degenerate gambling.

37:30 Intuition, Insecurity and a Wife From Milan

He describes his research routine and says he trusts his intuition, but used to wrap it in insecurity. He credits his wife with teaching him taste, from dropping logo clothes and a stylist to sizing investments sensibly. He defines taste as where intuition, emotion, selectivity and judgment meet, and says real taste is about recognizing what is beautiful rather than denigrating what is not.

45:42 Poker as One Infinite Game

He admits to the gambler’s strange comfort in losing, then makes the case for poker as training for nonlinear self-improvement. Luck dominates a single hand and fades after about 10,000. He argues the real skill is managing loss aversion when deviating from conventional wisdom, which he says is right 80 to 90 percent of the time, and he warns that retreating after flirting with a deviation changes how you live the safe path.

55:06 Rubin’s Making Game Versus the Outcome Game

Palihapitiya says he is unconventional and wrong most of the time, and that he has had hits but no album he is proud of. Rubin says he never revisits finished work wanting changes, and that great work comes from making it for yourself rather than aiming at outcomes. Palihapitiya admits that this gives him an internal panic and that he still cares what people think.

57:57 Fringe Bets and AI’s Founding Documents

He explains why money can bend outcomes in investing, using a hypothetical cancerless cigarette and SpaceX as examples, and argues the US faces a massive electricity shortfall that someone has to fix. He then says AI models differ on subjective questions because of their founding documents, that they predict the next word rather than think, and that the danger comes from misdirection, much as with any other powerful technology.

1:09:46 Three Regrets and a Path to God

He names the mistakes he carries: leaving Facebook early, a Stanford speech he made gratuitously painful for former colleagues, and a run of attention-seeking deals. He describes a ritualistic Sri Lankan Buddhist childhood, then a turn in his late 30s through a book about Jesus and René Girard’s mimetic theory. He now believes in a God who loves him and puts him through trials so he can grow.

1:22:39 Open Models, the AI Exoskeleton and Broken Science

He argues open models are the antidote to AI feudalism, and describes AI as an exoskeleton for thinking, with a Nash equilibrium setup that flags where his plans are exploitable. He talks about reading articles over books, his weekly newsletter, and the Long Hill Wagyu ranch he backed after tasting its beef. He blames irreproducible research on overhead-heavy grants and back-scratching peer review, then covers effective altruism, staying in California for his kids, and why he distrusts multinational bodies like the WHO.

1:43:40 Bitcoin, Groq, a Credit Line and the Bond Vigilantes

His calls that worked: Bitcoin after a friend from Argentina explained it, the Warriors as an inflation hedge, and Groq. His misses: misreading Obamacare, taking on heavy debt, refusing to sell in 2022 and scaling down his fund. He explains why he thinks crypto is capped, argues money printing prolongs bad ideas, and closes by saying that the thing he believes now and did not as a kid is that he is a really good person.

Notable Quotes

“Taste is the part where you’re like, “That’s amazing.” So I can hone in on the thing that’s beautiful, that’s right, at the exclusion of this stuff without denigrating that stuff.”
Chamath Palihapitiya, on what taste actually is

“Poker is not many many many games. It’s one infinite game.”
Chamath Palihapitiya, on why luck fades over a lifetime of hands

“I think I like what I like. And of that I would say I’m probably 80 to 90% wrong.”
Chamath Palihapitiya, on his unconventional bets

“the artist perspective is in the making game. It’s not in the outcome game. the outcome happens.”
Rick Rubin, pushing back on Palihapitiya’s wish for a proud album

“I didn’t do it because I thought it was right. I did it because I like the sound of how it felt when I said it.”
Chamath Palihapitiya, on his Stanford speech about Facebook

“I think I’m limited to the same number of good decisions but I make them better”
Chamath Palihapitiya, on what AI actually changed for him

“I convinced myself of an alternative future and it was just wrong. And in the end, it was mostly because of insecurity or ego.”
Chamath Palihapitiya, on his biggest misses

Watch the full conversation here.