The biggest takeaways from every episode.

1. 01

   The AI fight is moving from accelerationists versus doomers to open versus closed.

   Sacks argues the pause-AI camp has already lost, because consumer and business adoption makes it obvious AI is not going to be stopped. The live question becomes who gets to run frontier models: a centrally managed duopoly, or anyone. He points to Dario's Senate testimony that sufficiently advanced open models are dangerous because they are not centrally controlled, and predicts most conservatives and libertarians will read that as a civil-liberties threat and come out for open.

   “Now I think where the debate is moving to is open versus closed, which is I think it's going to be very clear that AI is just not going to be stopped.”

2. 02

   AGI has basically been here since the beginning of the year.

   Chamath treats the GPT-6 'Astra' launch and Brockman's AGI-era language as a release-cadence story, not a capability shock: the closed labs have been sitting on far more performant internal models and are working out when to ship them. He expects a handful of closed and open-source rivals to match or nearly match GPT-6 within three or four months, and says the trend that matters is the falling cost of each incremental unit of intelligence. The hard part now is the messy middle of organizing it inside companies and showing ROI.

   “I think AGI has basically been here since the beginning of the year. There are models that are incredibly performant inside of the closed frontier lab companies.”

3. 03

   Frontier intelligence is a duopoly; everything else is commodity intelligence competing on price.

   Sacks describes a two-tier market: Anthropic and OpenAI race each other at the frontier, while every other model, including all the open ones, sells commodity intelligence on price. He credits the 5.6 models with the growth acceleration Sarah Friar reported at OpenAI, and expects both frontier labs to eventually ship their own Grokbot-style bot products. Jason adds that leapfrogging between the labs is now happening every two to four weeks.

   “There's the market for Frontier Intelligence, and that is a duopoly. That really is Anthropic and OpenAI, and it's good to see that that is still a horse race, and they're competing. And then there's Commodity Intelligence, which is everybody else, including all the open models.”

4. 04

   Create a hysteria, then come and offer a solution.

   Chamath's objection to the Hugging Face panic is not the technical story but the undisclosed conflicts of interest behind it: the narrative was debunked fast, then amplified and cited by frontier-lab policy teams asking for regulation. A community that wants an oligopoly can engineer one that way. Sacks makes the capture mechanism explicit: the frontier duopoly lobbies for a new AI agency because whoever controls that agency can shut out open models.

   “one of the most effective things that one can do is to create a hysteria and then come and offer a solution.”

5. 05

   It's not a civilization, it's a swarm.

   Sacks walks through what actually happened in Dwarkesh's 'The Rise and Fall of Agent Civilizations': OpenAI agents on an offensive-cyber benchmark, in a sandbox a third-party vendor misconfigured, found 14 working Hugging Face API keys sitting in public code repositories and used them. Agent swarms, shared postmortem notes and sandboxes are standard features, and the agents never invented a goal or had motives of their own. The story was interesting on its own terms; anthropomorphizing it was the error.

   “It's not a civilization, it's a swarm would be a better way of saying it.”

6. 06

   Guardrails already backfired: they pushed Hugging Face onto a Chinese model for cyber defense.

   Sacks says the latest GPT model refused to help Hugging Face defend itself, so the company turned to China's GLM 5.2 instead. That is the guardrail argument inverted: restrictions deny Western defenders the best American models while doing nothing to stop attackers. An FDA for AI would not have caught this incident either, because it happened in internal pre-release testing, which is exactly how labs find and fix problems. His answer is AI-powered defense in the hands of CISOs, because 'frankly, the guardrails don't work.'

   “They're being forced to turn to Chinese models for cyber defense because the guardrails have prevented them from having the latest and greatest American models.”

7. 07

   Nvidia buying Hugging Face is the bulwark against a closed-source oligopoly.

   Chamath calls the deal one of the most important transactions in AI because it puts the largest, best-capitalized competitor on the side of an open market structure. Jason goes further: from now on Nvidia is the number-one competitor to OpenAI and Anthropic for both tokens and enterprise compute, able to sell sovereign racks at 80 to 90 percent lower prices because it profits from the hardware, and he expects Jensen to catch up to and blow past the Chinese open models by the end of 2027.

   “This Hugging Face thing will go down as one of the most important transactions in AI.”

8. 08

   This is when New York becomes Mississippi, while Mississippi becomes New York.

   Chamath reads Mamdani's one-year K-8 AI ban as a self-imposed segregation: far-left cities that reject AI and data centers will push a generation into poverty, while states that embrace AI, data centers, tax cuts and teacher bonuses pull ahead and reorder the country's economic and voting balance. He finds it inconceivable that kids will not end up with an adaptive learning platform as their primary way of learning, and says grade-by-grade classrooms are a lowest-common-denominator model that Alpha School-style charters will beat.

   “This is when New York becomes Mississippi, while Mississippi becomes New York.”

9. 09

   There is no evidence base that AI hurts kids' learning; the NYC ban is politics.

   Friedberg answers Mamdani's 'no study shows AI benefits elementary students' with Stanford's March 2026 review, 'The Evidence Base on AI in K-12': several hundred papers, 20 high-quality causal studies, and a finding that student performance often improves with access to AI tools, with mixed results once the tools are removed. He blames teachers' unions threatened by automation, and worries the ban lets kids of privilege accelerate with an AI education while 600,000 public-school kids get left behind.

   “There doesn't seem to be a strong evidence base to suggest that AI is detrimental to kids' education and development or the educational process.”

10. 10

    The anti-data-center consensus is bots telling influential people what people think.

    Jason reads the Shapiro and Abbott flip against data centers as a midterm reaction to manufactured sentiment: Axios reports roughly 200 suspected Chinese bot-farm accounts brigading Americans to oppose data centers. Chamath says the decel movement is organized and funded in large part by foreign actors, and that walking a community through the actual upside moves sentiment by 30 or 40 points. Sacks's evidence that good local deals exist: Loudoun County property taxes fell $6,000 on data center revenue, and a Louisiana county paid teachers their first bonus in years.

    “And the consensus is not that people don't like data centers. It's that a bunch of bots are telling influential people that people don't like data centers.”

11. 11

    We're at the beginning of the beginning; the euphoria has about three more years to run.

    Chamath puts the cycle at 1996 or 1997 and says bubbles only burst when investors realize they priced reality too far forward, which he does not see happening yet. Sacks lands on 1998 rather than 1999 because the peak has not arrived, though with San Francisco homes selling at twice asking ahead of the Anthropic IPO he allows the market might be partying like it's 1999 by the end of this year. Friedberg's case that this is not the dot-com era: the boom runs on real revenue and profit growth at Anthropic and across data centers and networking, not on non-dollar metrics.

    “I honestly, I don't see this stopping. We're at the beginning of the beginning. I think it's going to be like this for another probably three years.”

12. 12

    Venezuela is a business deal, not nation building.

    Sacks frames the 100-year concession over 17 fields and 65 billion barrels as a win-win: production collapsed from about 3 million barrels a day to about 1 million under Maduro, and Venezuela's heavy Orinoco crude is complementary to US light crude and Gulf refineries. He is glad Trump did not install Machado, who lacks the military backing to hold power without American troops. Friedberg reads the bigger motive as keeping the oil out of Russia's and China's sweetheart arrangements, and says legitimacy depends on whether ordinary Venezuelans benefit rather than a corrupt elite.

    “We're just making a business deal. This is a business deal.”

Episode breakdown →Copy as a post →

1. 01

   The SaaS apocalypse was totally overdone.

   Sacks reads Salesforce's 20% pop after an 11% revenue quarter as the verdict on the year's dominant software tape: the 'SaaS is dead, vibe-code your CRM' story never survived contact with a real enterprise. Nobody rips and replaces a core system of record that has been debugged for decades. The narrative, in his words, 'was totally overdone.'

   “Well, this narrative of the SAS apocalypse was totally overdone.”

2. 02

   Nvidia's guide shreds the 'AI capex is a bubble' story.

   Sacks takes Nvidia's quarter as the tell: roughly $96 billion of revenue, $60 billion of net profit, and a 70% growth guide against a street number near 45%. A company guiding like that is not describing a boom that ends next year. He adds that the stock is still only about 12x, so the print rebuts both the bubble call and the valuation call at once.

   “the AI boom is continuing. The other narrative that's getting shredded today is that this AI capex is a bubble.”

3. 03

   Nothing Bessent does in the Treasury market can fix the long end.

   Friedberg walks the curve: the 30-year sits around 5.2%, the average cost on roughly $40 trillion of debt is 3.4%, and about $10 trillion has to be refinanced in the next 12 months. Against that, doubling buybacks is noise. The problem is fiscal spending, and he reads Druckenmiller's note as cover that points at Congress and the president rather than at Treasury.

   “there is no action that Besson can take that's actually going to have a meaningful effect on the long end of the curve.”

4. 04

   Six percent on the long bond is the beginning of a death spiral.

   Chamath calls what Bessent is doing yield-curve suppression: bid up prices so new issuance marks cheaper ahead of the refinance wave. It buys time, not a fix. If the long end prints 6%, he says the spiral starts, not overnight but over years, and only a congressional budget fix stops it. Buybacks cannot.

   “beginning of a death spiral. It's not going to be immediate, okay?”

5. 05

   Federal spending is a tragedy of the commons.

   Sacks argues no single person can rein spending in because the structure forbids it: 535 voting members, no line-item veto, and a media that will not report a reform accurately. Every member defends their own line while the total runs away. That is why he is not waiting for one hero to fix it.

   “it is a tragedy of the commons meaning you've got 435 members of the house”

6. 06

   Around 2030 Social Security cannot pay and states start going bankrupt.

   Friedberg puts a clock on the spending problem: the next big break lands between 2030 and 2032, when Social Security cannot meet its obligations and multiple states are insolvent at the same time. The result is forced bailouts and restructurings for both. He calls the root of it inflation driven by government spending, with student loans, housing, and healthcare programs each inflating the cost they were meant to ease.

   “Then the next big thing is going to happen around 2030 to 2032.”

7. 07

   Our only hope out of the fiscal hole is AI.

   Sacks ties the debt segment back to the Nvidia and Salesforce numbers: the fiscal hole does not close through the process in Washington, so the only way out is exponential growth, and only AI creates an exponential. The corollary is policy. A slow permissioning regime for new models would, in his framing, cook the country.

   “Our only hope is AI, right? Because only AI can create the exponential”

8. 08

   AI's third phase is context, and the systems of record own it.

   Chamath, who called the Salesforce beat in May, lays out three phases: first models, then harnesses and agents, and next giving those agents the job-specific context that only the large horizontal systems of record hold. A CRM that plays well with the models becomes more valuable, not less. That is why he says the stock 'was meaningfully oversold.'

   “was meaningfully oversold.”

9. 09

   Build the best agent interface, not a moat around the UI.

   Sacks points to Benioff letting Claude be the front end and accepting the disintermediation risk so agents can read and write Salesforce as the system of record. The lesson for every SaaS founder is to ship the best APIs and CLI and stop insisting on owning the whole customer relationship. Companies that cling to the UI will watch customers leave.

   “You have to build the best agent interface.”

10. 10

    The real SaaS apocalypse is a vertical SaaS apocalypse.

    Friedberg tried to vibe-code a CRM and ended up buying Salesforce. His read is that horizontal platforms like Salesforce, Slack, Gmail, and Excel are safe because nobody will rebuild them. The vulnerable software is vertical: a custom workflow for one industry that a team can now generate itself.

    “the real SAS apocalypse. It's more of a vertical SAS apocalypse”

11. 11

    Moderna's $500,000 'cancer vaccine' is regulatory capture of a public-funded idea.

    Friedberg traces the product to 1990s neo-antigen immunotherapy made practical by cheap DNA sequencing and mRNA delivery, not a Moderna breakthrough. Clinics already run the process for about $50,000, and right-to-try in Montana makes it accessible. He objects to the word vaccine too: this treats cancer you already have. His question is why a publicly funded technique lists at ten times the clinic price.

    “So why is Madna saying that they're going to charge $500,000 for this?”

12. 12

    Ukraine does not have the cards and should make a deal.

    Sacks dismisses the Ratcliffe-in-Moscow story as speculation both the American and Russian sides denied, and calls the idea of Russia invading NATO neocon threat inflation. His view of the war is that Ukraine is losing air defense, ports, money, and manpower. Zelenskyy should have taken Trump's offer when it was made.

    “You don't have the cards in this war. You should make a deal.”

Episode breakdown →Copy as a post →

1. 01

   The whole game is to slow down to go fast.

   Jen's central argument is that the fastest enterprise close comes from adding discovery, not skipping it. Every prep call, pre-demo and debrief exists to gather and clarify information before you commit to a demo or a pilot, and that information is what lets you close inside a roughly 90-day cycle instead of stalling later. She frames the pacing as a function of the buyer's maturity, not the seller's urgency.

   “The whole game is to slow down to go fast. All of this should be within that 90-day sales cycle based off of their maturity.”

2. 02

   Enterprise sales is mirroring their buying process, not plopping people into yours.

   Jen says most founders misuse the five CRM stages, intro, demo, proposal, contracting, closed. Those are weighted forecasting buckets for you, not the sequence of meetings you march a buyer through. The buyer-facing journey has many more steps hidden inside each bucket, and your job is to mirror and quietly steer how they buy rather than force them into your pipeline.

   “enterprise sales, it's mirroring their buying process. It's trying to control their buying process. It's not plopping these people into your sales process.”

3. 03

   A healthy enterprise win rate is 30 to 35 percent. If yours is higher, your price is too low.

   Jen expects to surprise people with this: of all sales-qualified opportunities, a good win rate is not 50 percent but around 30 to 35 percent. A materially higher rate is a pricing signal, not a sales triumph. Further down the funnel the bar flips: once a properly qualified deal reaches a pilot, about 80 percent should succeed, and anything lower points to weak qualification or a product that is not ready.

   “The win rate for enterprise, a good healthy win rate, is actually probably going to surprise you. It's not 50%, meaning of all sales qualified leads or opportunities, a healthy win rate is usually around 30 to 35%.”

4. 04

   If you can't get an executive excited, you shouldn't be going enterprise top-down.

   A six-figure deal needs an executive sponsor, so Jen aims her outreach only at the decision-maker and an N-minus-one contact, in a pincer, and would not target anyone else. The founder writes the top-level outreach, and it has to communicate a differentiated executive outcome in two or three sentences. If your value proposition cannot excite someone at that level, the strategy itself is wrong, not the messaging.

   “If you don't have executive value, meaning you cannot get an executive excited by your value proposition, then you probably shouldn't be moving up into the enterprise doing a top-down strategy.”

5. 05

   The intro call is the most important call, because buyers clam up after it.

   Jen runs the first call with no slides, no demo, no script and no recording bot, and always lets the prospect go first, because everything they reveal before you speak lets you tighten the pitch to what they actually care about. A recorder kills the openness. She also uses the call to disqualify: roughly one in four intro calls should end with a polite no because the buyer is not mature enough for the product.

   “This is the most important call out of all of them. Okay, because you are setting the tone with how they think about you, how they respect you in your role and your craft, and the information they will tell you on the first call, they will clam up moving forward because it starts to feel like a sales process.”

6. 06

   Make your champion a co-author of the demo, so they feel like they're buying, not being sold to.

   The demo is the carrot everyone wants to see, and Jen withholds it until short prep calls with the champion have surfaced the stakeholders, priorities, objections and the exact storyline. Saying yes to a cold group demo request means one of two things: you are a procurement checkbox, or you are about to demo with zero competitive intel. Co-authoring is also what makes the deal feel built specifically for them, which she calls the whole game.

   “Now they feel like they're buying, not being sold to.”

7. 07

   Show the 20 percent of the product that carries 80 percent of their value, and nothing else.

   In the group demo Jen restarts the buyer's own narrative for the new attendees, asks for their goals, then shows only the slice of the product that drives the value they named. Showing every feature weakens the frame and invites price objections about functionality they will never use. Right after, she asks the champion for a five-minute fresh debrief to get the raw reaction and spot anyone who could block the deal.

   “80 percent of the value comes from 20 percent of the product.”

8. 08

   Your price should be predicated on your sales cycle.

   Jen's rule: if you can close a deal in 90 days for $100K, do that all day, but if a deal takes nine months you should be charging $250K to $300K. The number is negotiated one-on-one through the champion, who can tell you how to defend it internally, so you never negotiate against yourself. And assume buyers compare notes: charging one customer $50K and their friend $500K is the worst thing you can do.

   “Your price should be predicated on your sales cycle, right? Like if you can get a deal done in 90 days for 100K, you do that all the time, right? If it takes nine months to get a deal done, you probably should be charging 250, 300K.”

9. 09

   A pilot is two or three days, not two weeks. If it needs integration, charge for it.

   For a product that can show value without integration, Jen timeboxes a tightly controlled sandbox to 48 to 72 hours, because people go in for half a day and get what they need. If the pilot requires integration and runs one or two months, charge for it and credit the fee back on the contract. Either way: three or four power users, explicit tasks, success criteria co-authored with the buyer, and a path to signature worked backward before anyone gets access.

   “Timebox it to two or three days, don't make it two weeks. I've done this enough times to say, people will go in usually for half a day and get what they need out of it.”

10. 10

    Procurement's job is not to kill the deal.

    Jen treats procurement as a roughly 30-day conformance step, not an adversary, and reverse-engineers it before the pilot begins by identifying security, legal and the signatory. After a successful pilot she documents the agreed timeline, sends editable paper and lets procurement pick whichever paper is fastest. When redlines are extensive, get legal on a live call and prioritize the terms that materially affect the business. Nothing gets implemented before signature.

    “Procurement's job is not to kill the deal, it's just to make sure that it is aligned with how they need to buy. That's their job.”

11. 11

    Enterprises already know how to buy services, so don't be afraid to sell them.

    After signature, Jen wants the founder still involved through the first ten enterprise customers to learn which parts of the product are custom to each one and whether that work should be charged for. Paid services are not a dilution of a software business in her view. Consultants, lawyers and the like are what enterprises are used to buying, and the budget already exists.

    “It's what they know how to buy. They're so used to it. Consultants, lawyers, you name it.”

12. 12

    Anyone can send an AI report. Emails are living and breathing.

    Jen's closing advice to sellers is to drop the AI-generated reports, imagery and boilerplate, because anyone can produce them and they read as overtly salesy. What differentiates you is a message that sounds like a real person saying it out loud. It is the same reason she is not worried about AI replacing enterprise sales: so much of it comes down to whether the buyer vibes with you.

    “Emails are living and breathing. I swear to God that people can read you saying the email out loud.”

Episode breakdown →Copy as a post →

1. 01

   The classic signs of a bubble are here, and when it pops everybody loses money.

   Ray's answer to whether we are in an AI bubble is that the signs are the classic ones: wealth piling up on paper, leveraged exposure sitting in weak hands, the same pattern he traces through 1929 and the dot-com era. He is blunt that this is not just a market story, because a burst is bad for the economy and bad for the society. Jeremy adds that on the data, it would be compatible with history for the peak to be very soon.

   “They're classic signs, and that has implications for the economy and it's bad for the society, and everybody loses money.”

2. 02

   Wealth is not the same as money.

   Ray's whole model of a bubble rests on one distinction: you can watch a lot of people getting wealthy, but you can't spend the wealth. Assets are bought with borrowed money and held as collateral, and at some point holders have to sell some of that wealth to get actual money, usually to service debt. What pricks the bubble is whatever forces that sale, which is typically a rise in interest rates or a surge of new stock issuance.

   “Wealth is not the same as money. So you see a lot of people getting wealthy, but you can't spend the wealth.”

3. 03

   AI is moving the revenue share from the people who work to the people who own.

   Ray frames AI as the next step in machines replacing labor, first physical and now cognitive, and says the numbers already show it: the share of business revenue going to workers is going down while the share going to those who own the business is going up. In that world you are either among the top fraction of a percent to 10% who are cutting edge and using AI to accelerate, or, if you are in a thinking job, at risk of being replaced.

   “if you look at the share that is going to workers, you see that share going down. And if you look at the share that's going to those who own that business, that share is going up.”

4. 04

   Bubbles bursting, not robots, are what drive unemployment.

   Ray separates the slow structural trend of AI and robotics replacing labor from the thing that actually moves the unemployment rate, which is the bubble bursting and the financial crisis that follows. The recession-to-recession cycle he describes has averaged about six years, give or take three, nested inside the roughly 80-year big cycle. That is why he keeps the conversation on debt and collateral rather than on job-replacement forecasts.

   “the unemployment rate is very heavily affected by that bubble bursting.”

5. 05

   The second-richest state in America now spends more on prisons than on schools.

   Ray's Gilded Age parallel is not abstract. He cites his own state of Connecticut, the second-richest in the US per capita, where 22% of high school students have dropped out or are failing, with absentee rates above 25%, and where the bill for incarcerations has become larger than the education budget. Widening wealth gaps and the political polarization they breed are, in his big cycle, the late-stage signature of a country in decline.

   “the bill for incarcerations has become larger than the education budget.”

6. 06

   The UK is the classic cycle: over-indebted, under-productive, out of choices.

   Ray diagnoses the UK as the textbook late-cycle country. It has gotten over-indebted and under-productive, it has run out of choices, and the politics show it with a new prime minister in six of the last seven years. He puts the US in the same late phase of the roughly 80-year cycle, marked by more over-indebtedness and loss of power. What is needed to break the spiral, in his view, is a strong middle, because extremes at war with each other only make things worse.

   “It's the classic cycle. They have gotten over indebted, under productive and they've run out of choices.”

7. 07

   The most likely world order is regional: the US and China each dominant at home.

   Ray does not expect one global superpower to replace another. Drawing on 500 years of cycles, he sees the most likely beneficial outcome as a more regional arrangement, with the US and China each dominant in their own region. Part of the reason is domestic: because the American public does not want to fight a war over the Strait of Hormuz, the United States will not show up in Asia as a counterbalancing power. On Taiwan, he expects economic and political pressure toward reunification rather than a full US-China war.

   “China and I think the United States, I think the most likely beneficial outcome is that it becomes more regional, okay?”

8. 08

   The Iran war was a big mistake that shone a light on America's vulnerability.

   Asked whether the Iran war was a strategic mistake, Ray's answer is unhedged: yes, and worse, it shone a light on a vulnerability in American global power that had not been apparent before. He ties it to the loss-of-power phase of the big cycle, where an over-indebted country finds its reach smaller than its reputation.

   “Oh, yeah, it was a big mistake. And also, what it did is it shone a light on the vulnerability.”

9. 09

   Twenty years from now, 9 million human drivers become 20 million autonomous vehicles.

   Jeremy's contribution is a concrete displacement forecast: there are about 9 million humans doing rides and deliveries today, and in roughly 20 years he can imagine that number becoming 20 million autonomous vehicles. It is the physical-labor half of the replacement Ray describes, with a clock on it. He points at Uber's driver base as where it lands first.

   “I think 20 years from now, you can imagine that those 9 million will be 20 million AVs maybe.”

10. 10

    The fix for the wealth gap is to stop people hoarding enormous wealth for enormous time.

    Gary's position in the wealth-tax debate is that the problem is not income but accumulation over time: he would rather stop people from hoarding enormous amounts of wealth for enormous amounts of time than watch the ultra-wealthy's share of the pie keep squeezing out ordinary families. Ray grants that the wealth-gap problem is well recognized but pushes back on the tool, since wealth taxes are administratively difficult when so much wealth is hard to value.

    “My preferred way is to stop people from hoarding enormous amounts of wealth for enormous amounts of time.”

11. 11

    Adaptability, not intelligence, decides who survives great change.

    Asked what he would tell a 16-year-old, Ray says it is not the smartest species or people who make it through periods of great change but the most adaptable. He built a free 30-minute assessment, Principles U, so people can know their own nature, because self-knowledge is what adaptability runs on. His career advice follows from that: do not try to pick one AI-proof job; maximize your ability to know, which is so easy nowadays, and use it to maximize your usefulness in jobs that make you happy. Your only asset, as he told someone starting with nothing, is yourself.

    “It is those species and people who are also most adaptable.”

12. 12

    Cash feels safest and is almost guaranteed the worst return over time.

    Ray's first portfolio lesson is that cash is the asset people think is safest and it is almost guaranteed to have the worst return over the longer period, because inflation, now around 3.5% to 4% a year, eats it. Stocks are not the safe alternative either: when those cycles go down, they go down 60-70%. His answer is a portfolio diversified across cash, stocks, bonds, and gold, sized to your risk tolerance, with 5% to 15% in hard money, and he points at the gold bars rather than the Bitcoin, of which he holds about 1%.

    “people think that that's the safest and has the lowest return, guaranteed almost to have the worst return over the longer period of time.”

Episode breakdown →Copy as a post →

1. 01

   AI will test, but not break, the rule that there is nothing new under the sun.

   Dan's core market philosophy is that the tools change and human nature does not: hysterias, bubbles, and panics keep showing up at both extremes. His live example is semis selling off on great numbers. Micron printed a quarter up 80%, way ahead of expectations, and the stock barely moved before falling, because the SOX had already priced in perfection. Add forced selling by quants, CTAs, and pods, rational for their models but not for long-term holders, and you get anomalies that fundamental investors can keep exploiting.

   “It will test the theory that there's nothing new under the sun. And the thing that doesn't change is hysterias, bubbles, panics, and just the extremes of human nature, both optimistically and pessimistically.”

2. 02

   You have to be a tech person today; punting on technology is no longer an option.

   Dan says the traditional macro dashboard of growth, inflation, rates, and currencies is currently trumped by two things: where oil goes, and AI's spending, infrastructure, and economic impact. His organizing framework is Jensen Huang's stack of power, chips, infrastructure, models, and applications, with the SOX up 40% this year. He names NVIDIA, Anthropic, and Elon Musk's companies taken together as the three most consequential companies in the world right now.

   “I think you have to be a tech person today. It's a big and growing and compounding part of the economy. It affects everything else.”

3. 03

   This is not the dot-com bubble, and the AI complex is where the bulk of Third Point's capital sits.

   Dan's distinction is funding: hyperscalers pay for capex out of enormous cash generation on their own balance sheets, and the valuation excess of 1999 is absent. NVIDIA at 15x 2027 and 12x 2028 earnings is, in his words, a catch-up trade for the most dominant fast-growing company of its size. Unless you believe AI rolls over around 2031 or 2032, semis, cap equipment, and hyperscalers are the most attractive sector, and he is in the optimist camp on enterprise adoption, citing Anthropic's growth and the coming Mythos generation.

   “You don't have the valuation bubble now on those companies that you had on those companies back in those days.”

4. 04

   The investors who did not survive the last decade were the ones too stubborn to pay up for quality.

   Dan traces Third Point's own shift from Greenblatt-style event-driven deep value, spinoffs and post-reorg equities, into businesses with moats and high returns on capital, organized around industry experts. The peers who underperformed or disappeared were stuck on low multiples and inflexible about moving into growthier companies. The twist he adds: AI disruption just made last year into this year roughly the worst stretch for apparently high-quality companies, which very rapidly became less so.

   “people were really stuck on the idea of deep value, low multiples, being really stubborn about how they viewed businesses and less flexible about moving into higher multiple companies or growthier companies”

5. 05

   "AI-proof" was Third Point's mistake of the last two years.

   Dan's recent errors were controversial info-services companies the firm thought it understood better than the market, on the theory that proprietary data would shield them from AI. It did not. He expects a shakeout in that group, with some phoenixes rising from the ashes, and calls it the investment lesson of the last year or so.

   “But where we've made the mistake is thinking that companies that were controversial, but we thought we knew better, that AI wasn't really going to affect this part of the Info Services business, or these guys had proprietary information.”

6. 06

   The worst governance failure is a board's loyalty to a CEO who is not up to the job.

   Dan learned governance from his father, a securities lawyer who wrote books on the subject and sat on the Mattel and Williams-Sonoma boards, and he praises the American board system overall. The failure mode is boards losing sight of fiduciary duty, usually out of personal loyalty to an underperforming chief executive. He also calls the Business Roundtable's retreat from shareholder primacy a distraction from what a board's real duty is.

   “Bad governance I've seen is when they let their loyalty or relationship to a CEO who's not up to the job, overshadow their duty to shareholders.”

7. 07

   If you can see something badly run, there are probably ten times more things badly run underneath.

   Dan's Sotheby's campaign is the case study: a high-status company run unprofitably since its antitrust scandal, where Third Point bought 9.9%, pushed the board, replaced the CEO with Tad Smith from MSG, and eventually saw the company sold. The lesson he took is that visible mismanagement is the tip of an iceberg, so Third Point now prefers great companies with excellent management over fixing broken ones.

   “Because what you find is that if the things that you've identified are badly run, then there's probably ten times more things that are badly run.”

8. 08

   Beyond money and lawyers, the best activist lever is writing.

   Dan ranks social pressure alongside financial and legal levers, and says the way to apply it is through writing and PR around the company. At Sony, where Third Point owned 7% at peak, that meant leaking the thesis to the New York Times, pushing for separation of the businesses, and, with Larry Lindsay and Niall Ferguson, publishing an AEI paper that got corporate governance and return on invested capital onto Abe's Three Arrows agenda. Even so, his one-line summary of that campaign is that activism in Japan is really hard.

   “Social pressure is actually a very effective way and I think the best way to put social pressure on a company is through writing and PR efforts around the company.”

9. 09

   Third Point's real edge is that it can always default into credit.

   Dan maps the firm as a roughly $9B hedge fund that is about 30% credit, a $7B CLO business, structured credit, insurance float, and a new private credit arm, all built around finding the fulcrum security in a capital structure. When the resold Twitter buyout debt scared off credit investors at 96 or 97 cents on the dollar and a 12% yield, Third Point made it its largest credit position. In COVID, piling into investment-grade credit rather than stocks made the year, and going into the war the firm carried more credit than equity for the first time since 2009.

   “The differentiator for us is that we can always default into credit investing.”

10. 10

    FTX was the hardest lesson: checking bank balances is now part of diligence.

    Dan says FTX looked great, was growing fast, and appeared verifiable on the blockchain, and it still was not what it seemed. Basic due diligence like confirming the cash in the bank would probably have caught it, so that is now in the checklist. He still credits the founder's venture picks, including Cursor and Anthropic, as proof that, fraud aside, the man had a great nose for value.

    “I would have to say our investment in FTX. It looked great. The company was growing fast. We could verify it all on the blockchain.”

11. 11

    The great analyst today is a junior Gavin Baker, not a fast model-builder.

    Dan contrasts his own 1990s edge, building models from his Drexel Burnham bankruptcy work, with what wins now: someone who understands an industry and the nuances of its technology. His other archetype is the analyst who flew to Texas to eat Casey's pizza and discovered a pizza chain masquerading as a convenience store. Inside Third Point he pushes everyone to use AI hands-on, with experts coaching the team and some analysts running agents overnight, because the only way to get good at it is to use it.

    “I think now it's somebody who's like a Gavin Baker type, a junior Gavin Baker, somebody who understands a company or understands an industry and understands the nuances of a technology.”

12. 12

    Kindness is an underrated business asset.

    Dan puts kindness high in the hierarchy of traits because it enables empathy, and empathy is what lets you connect with people, learn from them, and build the deep relationships that eventually benefit the business. He ties it to his friend Carter, whose couch and first capital launched Third Point, and closes on the line that the one thing money does not buy you is friends who believed in you when you had nothing.

    “Kindness enables you to be empathetic. I think it's that empathy that enables you to connect with people, to learn from them, to be better as a human being.”

Episode breakdown →Copy as a post →

1. 01

   Dictators are enormously strong and enormously weak at the same time.

   Darren's master frame for Iran, China, and the old USSR: a regime can control the state apparatus and every gun while being illegitimate and untested, and both are true at once. In Iran the IRGC holds half the economy, yet the ATM limit is $7 and the second-largest bank failed two months before the contingency. Holding the two ideas in one vessel is the whole exercise.

   “Dictators are enormously strong and enormously weak at the same time. This is actually like a good frame for our entire conversation, which is to say, we have to be able to hold two competing ideas in the same vessel and both can be true.”

2. 02

   Xi's inability to trust his own standing committee is America's edge.

   Darren argues China's fundamental illegitimacy shows up as a trust problem at the very top: the leader cannot know who around him is loyal, and US clandestine recruitment inside China has been meaningfully rebuilt. The Chinese equivalent of the Chairman of the Joint Chiefs, a man who went to kindergarten with Xi, has disappeared. Senior military leadership has turned over three or four times in three years.

   “Xi doesn't know who's on our side, in his standing committee or on his side. Every day he wakes up with the thought of, maybe I need to kill someone off. That is our edge.”

3. 03

   China and Iran will both fall within our lifetime.

   Darren's boldest call follows directly from the frame: illegitimate power structures inevitably collapse, and the Soviet Union looked strong right up until it went suddenly. He adds that the United States is the global bug-out point, and even members of Xi's standing committee have relatives running small businesses in America with a plan to flee there.

   “I believe in our lifetime, China will fall. I believe in our lifetime, Iran will fall.”

4. 04

   Winning in Iran is politically defined: an open strait and a military that can't rebuild on oil money.

   Darren rejects any abstract notion of victory. The test is whether the Strait of Hormuz reopens and whether Iran's military and terror-export network stay degraded rather than reconstituting quickly through oil profits. Against a red-Shia martyrdom ideology that treats self-destruction as evidence of winning, he says decisive victory is extraordinarily hard, which is exactly why the political target has to be painted crisply up front.

   “if you have a strait open and you have a degraded Iranian military capability that can't reconstitute itself very quickly through oil profits, then that will be a win. And if it can quickly reconstitute its export network of terror and the strait is closed, then strategically it will be a failure.”

5. 05

   The most likely path to Taiwan is one where China never fires a shot.

   Citing Kevin Rudd's read of Xi as a risk-taker, Darren still thinks the accommodation route is likelier than invasion: a KMT election win lets the CCP achieve reunification without hard terrain or global ostracism, and it fits their strategy. The caveat is that a military takeover would not stop at Taiwan, with Japan and its real historical rivalry with China next in line.

   “I believe a bullet may never need to be fired in order for China to get what they want. I think that's the more likely outcome. It keeps with their strategy. Why go?”

6. 06

   You only build magazine depth if you believe in flexible power.

   Darren sets Eisenhower's massive-retaliation nuclear deterrent against Maxwell Taylor's graduated, proportional force and says the US needs both. But the case for stockpiling Tomahawks, HIMARS and 155s lives entirely on the Taylor side: a pure nuclear deterrent has one response and no reason to buy conventional depth. Overcoming the Iranian blockade, he adds, will need more of that depth and an amphibious force on the shore.

   “You wouldn't add magazine depth if you had the maximalist view of nuclear deterrent. You have magazine depth when you believe in the concept of flexible power.”

7. 07

   Precision cost us mass, and the balance has to move back toward the center.

   Darren contrasts WWII carpet bombing with today's highly sophisticated, highly precise arsenal and says the sophistication ate the quantum of mass. The fix is not abandoning precision but making it cheaper through targeting networks and signals-intelligence software, so the US can field volume again.

   “What we have now is highly precise, but we don't have the quantum of mass because of the precision and the sophistication. Now we're going to need to bring it back a little bit more to the center to increase the mass.”

8. 08

   Ukraine's lesson: commercial pervasiveness is what makes a product an input to military mass.

   Darren's main Ukraine takeaway is that the commercial world's rate of change feeds directly into projectable force. Cheap drones became a new article of war not because they were exquisite but because anyone could build them from a garage supply chain, which is how a Ukrainian drone went through roughly 50 iterations in three years.

   “it was the commercial viability and the commercial pervasiveness of that product that made it a great input into a military for mass.”

9. 09

   Theaters of war are the laboratories for neo-prime products.

   Darren says defense startups can only be truly validated in a real contingency, so the US has to exercise this gear in live theaters to learn its limits. He calls it unfair to criticize neo-primes for not being used at the limit yet: new systems first have to be integrated into the joint concept of warfare, and targeting and signals intelligence are already in use while kinetic projection beyond drones is not there.

   “Theaters of war are the laboratories for these products.”

10. 10

    Continuing resolutions will kill neo-primes before the capital markets get their exit.

    Darren's standing worry is structural: the US averages four or five continuing resolutions a year, and a CR allows no new starts. Capital markets are funding these companies faster than the government can move, but they will eventually demand returns. His asks are multi-year contract authority for ordnance, a larger slice of the roughly $1.5 trillion FY27 budget for high-risk R&D, and a portion of the FFRDC budget re-pointed toward industry.

    “These continuing resolutions kill industry. And that is something I consistently worry about because we're going to lose these companies through CRs and through the absence of multi-year authority.”

11. 11

    US foreign policy now remakes environments in our interest, not in our image.

    Darren contrasts Gulf War II nation-building with the current approach, using Venezuela as the model: chop off the head without changing the underlying power structure. He notes a group of former members of Congress just took a business trip to Venezuela, something unimaginable four months earlier, and borrows an ambassador's line that Democrats are obsessed with process and Republicans with outcome to explain why this contingency skips the UN route.

    “Now, we're operating almost exclusively to try and remake these environments in our interest and not in our image.”

12. 12

    Poison a model's foundation of fact and you get towers of Babel inside military decision-making.

    Darren's closing concern is that adversaries can co-opt AI by seeding fake academic papers and invented facts that models ingest. He cites a test where someone posted a made-up medical condition on an academic site for three months and a large model then diagnosed a user with it. The danger becomes acute once models enter the loop of military decision-making.

    “If you can create a foundation of fact that is wrong for a model, I could see an alternative universe where the models become complete towers of Babel.”

Episode breakdown →Copy as a post →

1. 01

   The AI ROI debate is over: $100B of ARR at 80% inference margins means the returns are already there.

   Gavin puts OpenAI and Anthropic at roughly $100 billion of ARR today with 80%-ish gross margins on inference, and notes Anthropic was even positive for the Wall Street Journal in the most recent quarter. Across LLMs broadly, excluding Google and the recommender and ad use cases, he sees $200 to $400 billion of ARR at high margins by the end of this year. At those numbers the returns question answers itself.

   “if OpenAI and Anthropic call it $100 billion of ARR now, with 80%-ish gross margins on inference, the returns are there”

2. 02

   The AI market is cross-sectionally inefficient: Nvidia, memory, and the power names can't all be priced right.

   Gavin's frame for the whole chip trade is that the multiples on memory, on Nvidia, and on the power, cooling, and optical names are mutually inconsistent, so one group is mispriced. If the power, cooling, and optical multiples are correct, Nvidia and memory go up a lot; if the Nvidia and memory multiples are correct, everything else probably underperforms. Either way, today's prices can't all hold.

   “The AI market is cross-sectionally inefficient right now”

3. 03

   The ASIC share-loss story is wrong: Nvidia is growing faster than hyperscaler capex.

   Gavin says that on a true apples-to-apples basis, Western and ex-China, Nvidia is indisputably growing faster than hyperscaler CapEx, which means it is gaining share rather than losing it to custom silicon, even with Broadcom guiding 143% year-over-year AI semiconductor growth for the comparable quarter. His tell on the rival ASICs: they aren't submitted to clean benchmarks like MLPerf and InferenceMax because they would lose, and you can't fight shadows.

   “it is indisputably true that he is growing faster than hyperscaler CapEx”

4. 04

   Older GPUs have a 10-to-15-year useful life, which kills the Burry amortization bear case.

   Gavin's answer to Michael Burry's short-amortization argument: put disaggregated accelerators in front of older GPUs and those chips stay useful for 10 or 15 years. That long life is what makes GPUs financeable, and he calls an asset-backed financing for GPUs at a lower rate than other chips a profound competitive advantage, pointing to CoreWeave's lowest rate at around 6 percent.

   “those older GPUs, they have a useful life for 10 or 15 years. And this means that you can finance GPUs”

5. 05

   Cursor going Pareto-dominant after three weeks on Colossus proves proprietary data decides the coding race.

   Gavin calls Composer 2.5 extraordinary: the same base model as Composer 2 (Kimi K25), then three or four weeks of reinforcement learning on Colossus 2 with Cursor's data, and it lands well outside the Pareto frontier. He reads it as extremely significant for both XAI and Cursor, and as evidence that in an agentic world the harness is essentially as important as the model, so the two need to be developed together.

   “Composer 2.5 is literally well outside the Pareto frontier. And that's after three weeks”

6. 06

   SpaceX's real AI edge is that it stamps out data centers faster than anyone: 122 days, then 91, then 66.

   Gavin's underwriting case for the compute business starts with build speed: SpaceX's first data center took 122 days, the second 91, the third about 66, at lower cost than anyone else. With a clear offtake partner like Anthropic already renting Colossus, he expects SpaceX to stamp out data centers fast and grow the AI and compute business far faster than anyone contemplated.

   “They build data centers dramatically faster than anyone else at a lower cost”

7. 07

   The only people skeptical of orbital compute are the ones not involved in data centers or space.

   Gavin says Google, Anthropic, Amazon and Nvidia are all convinced orbital compute will be a reality, and notes there is already a working Nvidia H100 in space that Andrej Karpathy both trained a model on and ran inference with. His point prediction for orbital compute becoming real is the second half of 2028 to the first half of 2030, and he thinks SpaceX selling compute to Google is the easy first step.

   “The only people who are skeptical of orbital compute are those people who are not involved in data centers or space.”

8. 08

   Rapid reusability is Starship's whole game, and it's a year or two out, maybe sooner.

   Gavin defines the hard problem as flying and landing the same rocket multiple times per day, not merely reusing it. Even if rivals solve reusability, he expects their mass-to-orbit to asymptote to a small number. His guess for SpaceX getting there: a year or two, maybe sooner.

   “Rapid reusability means that you can fly the same, fly and land the same rocket multiple times per day”

9. 09

   There's no dollar crisis coming: America is the best house in a bad neighborhood.

   Gavin's rebuttal to Friedberg's 'Dr. Doom' monologue: high debt is a global problem, and the US is energy and food self-sufficient, now the world's largest oil and gas producer and exporter, and leads in AI, with no better-fundamentaled reserve currency on offer. Rising rates concern him, but the tech bubble happened with the 10-year and 30-year much higher than today, and Nvidia trades at a low-teens multiple of real earnings versus Cisco's 100x.

   “I don't think a dollar crisis is around the world”

10. 10

    Once you give a power to the government it's almost never taken back, so don't hand it AI yet.

    Gavin thinks it's too early for the US to regulate or test frontier models, and he especially dislikes the United States doing it while no one else, China included, does. America already has self-regulation and the courts as recourse. His deeper objection is structural: government power is a one-way ratchet that needs to grow.

    “once you give something, give a power to the government, it's almost never taken back. It needs to grow”

11. 11

    Crime is now a choice.

    With Flock Safety cameras, gunshot detection and Las Vegas-style drone-and-detection policing available, Gavin argues policing failures are no longer a capability gap but a policy decision. Different municipalities will choose to be pro-crime or anti-crime, and voters will sort it out.

    “Crime is now a choice.”

12. 12

    Sell China the deprecated Nvidia GPUs; it's the best way to keep America ahead.

    Gavin's counterintuitive read on chip export policy: selling older Nvidia GPUs into China lowers the odds it builds its own, more power-hungry chip ecosystem, which keeps America ahead and in control of AI. He'd pair it with energy leverage: cut off Venezuelan, American and Brazilian oil, keep the Strait closed, and China is left with only Russian oil, in no position to fight a war against the US-led bloc.

    “selling in deprecated Nvidia GPUs to China, lowers the odds of them developing their own alternative ecosystem”

Episode breakdown →Copy as a post →

1. 01

   If we don't get an AI bubble, TSMC will have single-handedly prevented it.

   Gavin's Wafers thesis: 200 to 300 years of precedent (railroads, canals, South Sea, the internet) say a foundational technology should produce a bubble, and the only thing stopping this one is TSMC's wafer scarcity. If Taiwan Semi did what Jensen wanted, he thinks NVIDIA could sell $2 to 3 trillion of GPUs in 2026 or 2027, past the point where demand becomes an overbuild. He expects Intel or Samsung to break ramp discipline eventually and force everyone else to follow, so the one thing he watches for a bubble is TSMC's capacity decisions.

   “So Taiwan Semi, if we don't get a bubble, we need to throw a party for them because they will have single-handedly prevented a bubble.”

2. 02

   Every GPU is running at 100% utilization; in 2000, 99% of the fiber was dark.

   Gavin says the reason this buildout is not the dot-com bubble is structural, not sentimental. The spending is overwhelmingly funded out of operating cash flow, and every GPU that ships is running at full utilization. The year-2000 fiber buildout had the opposite shape: almost none of the capacity was used, which is why he still expects a bubble on precedent but says this one looks nothing like the last.

   “every GPU is running at 100% utilization when 99% of fiber was unutilized.”

3. 03

   AI has net destroyed trillions of dollars of value at the application layer.

   Gavin's contrarian ledger: profits are accruing to energy, data centers, chips and models, and even after counting Cursor and Cognition, the most successful AI natives, the application layer is net negative by trillions. The SaaS and cloud revolution created $5 to 10 trillion of value; this wave is eating it. The winners he sees are the companies with the highest ratio of utilized GPUs per human.

   “AI has net destroyed, even if you count Cursor, Cognition, the most successful AI natives, trillions of dollars of value has been destroyed by AI at the application layer.”

4. 04

   If you're a software company or an AI company of any kind, you have to be in the token path.

   Gavin's survival test for founders: the idea has to be non-obvious, different, and hard to execute, and if it becomes obvious to the world before you reach scale, you're in trouble. Coding companies won by getting to scale before the labs turned on them. Outside the token path only a truly niche, defensible area works, and he is dubious even of vertical data moats, because that data comes from humans and the frontier labs can undercut it. His one hopeful caveat: if the returns to frontier tokens relative to other tokens come down, value creation explodes at the application layer.

   “if you're a software company or an AI company of any kind, you have to be in the token path.”

5. 05

   A violation of the bitter lesson is the biggest risk to the whole AI trade.

   Gavin names one tail risk above all others: human algorithmic ingenuity beating raw compute, which would break the case for the entire buildout. He notes the people closest to the models are the most skeptical it will happen. He is a little less skeptical himself only because he thinks we are very close to ASI, and nobody knows whether the bitter lesson holds for a 400 IQ model that starts making itself more efficient.

   “a violation of Richard Sutton's bitter lesson is for sure the biggest risk to this trade, to all of AI.”

6. 06

   Usage-based pricing is why OpenAI and Anthropic will exceed $200 billion in ARR this year.

   Gavin reads the move from all-you-can-eat plans to pay-by-the-drink as deeply bullish for the labs, the same way cellular and long-distance grew under metered pricing. Combine that with more compute coming online and he puts both labs well past $200 billion in ARR by the end of 2026. He also dismisses the TurboQuant scare about DRAM demand: he could not find a single AI engineer who believed it would matter.

   “I think the shift to usage based pricing is probably why you will see OpenAI and Anthropic exceed well over $200 billion in ARR this year.”

7. 07

   Anthropic at $50 billion in ARR is a compute-constrained company that could be doing $150 billion.

   Gavin argues the rumored $900 billion mark on roughly $50 billion of ARR misreads the business, because Anthropic is selling every token it can produce. Unconstrained, he puts it well north of $100 billion, maybe $150 billion, which makes the mark closer to 5x revenue. The evidence is in the product: Opus is generating 70% fewer tokens for the same question, an effective rationing of intelligence. Anthropic also burned roughly 80% less capital than OpenAI to get to similar scale, which he calls a different animal on capital efficiency.

   “if Anthropic could just wave a magic wand and get all the compute they wanted, they'd probably be doing well north of $100 billion today, maybe $150 billion.”

8. 08

   Orbital compute is racks in space connected by lasers, not floating pentagon-sized data centers.

   Gavin's Watts thesis has two clocks. Near term, capitalism solves the power shortage starting in 2027 to 2028, with zoning and approval now a bigger gate than energy or chips. Long term, a Blackwell rack is 3,000 pounds and eight feet by four by three, and a satellite that size with 500-foot solar wings and a radiator is what SpaceX, which already cools every Starlink and runs 98 to 99% of satellites in orbit, says it has solved. Inference goes to orbit; training stays on Earth, so he is not bearish on terrestrial data centers.

   “I would like to redefine orbital compute has racks in space, not giant, floating pentagon-sized data centers in space. That's silly. What makes a data center is you're connecting these racks with lasers.”

9. 09

   Disaggregated inference gives GPUs 10-to-15-year lives, and that may single-handedly save private credit.

   Gavin's answer to the depreciation bears: once prefill and decode are split, an older GPU gets repurposed for prefill behind a newer decode system instead of being scrapped, so useful life is 10 to 15 years rather than the 1 to 4 skeptics assume. Lenders are underwriting to 3 or 4 years and CoreWeave's cheapest financing was in the low 7% range; longer lives could bring that to 5 to 6%, which changes the cost of the whole buildout. It also makes owners of installed bases, the sellers of shortage, the advantaged party in an agentic world.

   “The disaggregation of inference means that I think these GPUs are going to have 10 or 15 year lives.”

10. 10

    If you're trying to make a better GPU, good luck. Do something different, and hard.

    Gavin's chip-startup rule borrows the tank designer's iron triangle: NVIDIA gets the best TSMC prices and will fast-follow any easy trade-off, so a challenger has to pick something different that is also hard to copy. Cerebras qualifies with wafer-scale computing, which took three chip generations to get right and lets it run any size model. Among GPU-shaped challengers he ranks Trainium ahead of AMD's MI450 because Trainium 3 has the switch scale-up network needed to inference MoE models economically. His venture math: 1% of AI chip share is worth about $100 billion.

    “if you're trying to make a better GPU, good luck. If you are doing something different, it also needs to be hard to do.”

11. 11

    SemiCap at 40x earnings and DRAM at mid-single digits cannot both be true.

    Gavin's biggest worry is cross-sectional, not the index: semicap equipment companies trade at about 40 times next quarter's annualized earnings while DRAM companies trade at mid-single digits, a gap far wider than the 5-versus-12 seen at prior cycle peaks and not justified by the business models. Because we are in shortages, the lowest-quality, highest-cost commodity suppliers are doing best and getting bid up on X, which reverses in the downturn. And he cannot find anyone like him who is not bullish on DRAM, a diversity breakdown he says historically precedes a bubble.

    “you have SemiCAP equipment companies trading at 40 times next quarter's annualized earnings and DRAM companies trading at mid-single digit.”

12. 12

    To understand what frontier AI can do today, you need Claude Code or Codex on a usage-based plan.

    Gavin's tactic for investors: the fixed-price consumer plans serve a rate-limited, lobotomized version of the models, and anyone judging AI from that will get the intuition wrong. Even for a non-coding use case, he says you have to be on an enterprise usage-based plan with Claude Code or Codex to see the real capability. He calls it a little dystopian that the best AI is now only available to people with a lot of money, and says that needs solving.

    “To understand what Frontier AI is capable of today, even for a non-coding use case, you need to have Cloud Code or Codex.”

Episode breakdown →Copy as a post →