Summary

The Andrew Tulloch trade defines the new game theory of venture: he left Thinking Machines — the company he co-founded at a $10B post-money seed after raising $2B — for a reported $3.5B from Meta, and Rory O’Driscoll reduces it to pure option pricing: “Would you prefer $2 billion in Thinking Machines unlisted stock with the chance to be amazing or the chance to go bust, or $3.5 billion of liquid Facebook stock over the next five years?” His darker read: “Once you’re not playing a multi-period game… you’re playing a one and done. You’re gonna get bad human behavior” — and no one knows how to underwrite that risk at $10B pre for seven engineers.
Jason Lemkin takes the other side on morals, not math: “If my kid did that, I would not be happy with my kid… You leave the people that brought you to the dance because you see a prettier girl over here?” His seed underwriting has always rested on one implicit covenant — “my downside protection is he won’t quit” — and mega-offers just broke it. Practical takeaway for founders and investors alike: six-year vesting, cliffs, and repurchase rights are now legitimate diligence items.
Goldman’s $665M purchase of Industry Ventures (up to ~$970M with earnout) prices at ~10% of its $7B AUM — right between Carlyle/KKR at ~20% and public asset managers at 1-2%, roughly 10x revenue / 20x earnings on a 50%-margin business. The strategic logic: public asset management earns <10bps on S&P exposure vs ~2,000bps on private capital, so Goldman jams a productized secondaries platform (18% historical IRR) through its ultra-high-net-worth channel. Rory’s structural point: productizable GP businesses are the ones that can be 100% sold — “you can’t sell 100% of Benchmark ‘cause then you don’t have Benchmark.”
SoftBank’s $5B margin loan against Arm to fund OpenAI is “a relatively low-octane Masa move”: he still owns 90% of Arm at a ~$90B market cap, roughly $80B of equity, and Roger figures he “could take $25 billion against the Arm position easily.” Caveat from Rory kept as spoken: in 2002 individual Nasdaq stocks fell 90% from peak, “so it is possible the loan will get called. It’s just unlikely.”
On the AI capex bubble question, the constraint won’t be technology or demand — it will be the marginal capital provider. Rory, off Dwarkesh Patel’s scaling oral history: the smartest people treat “1% of GDP for compute” as a matter-of-fact to-do list — “10,000 computers worked, so we’re gonna buy 100,000… and somewhere along the line we’ll get AGI” — but “every economic phenomenon tends to be diminishing marginal utility,” and at some point capitalism says “you can’t have your $1 trillion dream.” Jason’s counter from the trenches: 8 vibe-coded apps in 100 days, 12 AI agents replacing Sastre’s sales and content teams, and he could use 100x the tokens today — with only 0.1% of Salesforce customers using AI yet.
Polymarket ($2B at $9B) vs Calshi ($5B from a16z/Accel) is “the purest regulatory arbitrage play of all time,” not king-making: Roger’s read is 90% of the business is sports betting escaping the tax and regulatory stack legacy books carry, racing to avoid parallel regulatory scrutiny — with Eric Trump on one board and Trump-family money in the other. Rory’s test for when capital can crown a king: only when money overwhelms a rival or brand-name VCs sway customers — “I don’t think anyone betting on Poly or Calshi gives a damn… who that money came from.”
Portfolio construction is bifurcating by stage as the exit bar moves from $200M to $400M ARR at IPO: Founders Fund concentrating from 31 growth deals to a target of 10 makes sense for quasi-public companies (“the only surprising thing was they weren’t there already”), while early stage demands diversify-then-concentrate — Roger runs 20-25 names with 3-5 taking 75% of capital via follow-ons. Harry’s pushback stands: Clubhouse, Hopin and BeReal looked like winners early and weren’t — but Rory’s data answers it: hit your first two underwritten years of revenue and the probability of >5x jumps from 30% to mid-70s.
Roger’s follow-on discipline is the masterclass: every check judged “independent of the check prior,” because “who cares about ownership? It’s a cash on cash business” — a $3M check into TTD at a $280M post (after four checks totaling ~$2M) returned $40M from a $50M fund; Wise started at $750K on a $5.5M post and ended 13% at IPO. Rory’s distillation: “provided every check you write is money good, in the end you’ll die rich.”

Deep dive

1. Goldman buys Industry Ventures at 10% of AUM — a fair price for a fee machine

The news: Goldman Sachs acquires Industry Ventures for $665M upfront plus a ~$300M performance earnout running to 2030, on $7B under management. Rory’s first reaction is personal: “Good for Hans. He grafted for 25 years” building the secondaries business from just after the 2000 crash. Jason’s: “congratulations on letting the ego walk it back” and not forcing the headline over $1B.
Jason plays the straight man — with $7B AUM, why isn’t the firm worth 20% of that? Rory’s valuation ladder, worth keeping: Carlyle and KKR trade at ~20% of AUM because they own all the economics; a secondaries/fund-of-funds hybrid gets ~10% (in line with StepStone and Hamilton Lane); plain public asset managers sit at 1-2%. Roger, who sold asset managers in financial-institutions M&A early in his career, confirms: as a fund of funds you keep ~10% of the carry pool, so this is roughly 10x revenue, 20x earnings on a 50%-margin business — “a straight on-market deal.”
Goldman’s rationale: S&P exposure costs under 10 basis points; running private capital earns 2,000. “Active management’s going away… getting a platform like this that you can expand just makes a ton of sense.” Roger adds the distribution detail — Goldman’s Apex platform for ultra-high-net-worth clients can now be institutionalized around Industry’s product, a win on both product and distribution.
Jason’s endorsement is from the buyer’s chair: the Morgan Stanley private-equity pitches he gets are “so dumb,” whereas Industry’s ~18% historical IRR “every day in and out” is a baseline he’d take “100 days out of 100.”

2. What kind of venture firm can actually be sold

Rory’s structural insight: 100% sale only works for productizable businesses — secondaries, fund of funds (Greenspring’s sale to StepStone [likely — spoken as “Cepzone”] is the same trade). “You can’t sell 100% of Benchmark ‘cause then you don’t have Benchmark… all you have is the three people, and if you cash them out 100%, then you don’t have anything.”
The continuum runs from brand-and-platform (sellable) to pure judgment (not): a16z “has clearly embarked on the AUM and great investing journey to bigness and maybe an IPO,” General Catalyst same, and even Y Combinator “is the definition of a business… independent of the greatness or not of the current operators.” The needle to Harry: “your media company with a venture fund attached could be monetizable in a way that Roger’s fund or my fund will never be… the only asset in Roger’s new fund is Roger’s IQ as a stock picker, God help us.”
Rory’s honest envy: for 25 years primary investors looked down on secondaries — “that’s not nearly as interesting as the business we’re in” — but “you can’t sell this business. Hans could sell his, and he did. Who’s laughing now?”

3. $3.5B to walk: the Tulloch morality fight

Andrew Tulloch leaves Thinking Machines — which he co-founded and raised $2B for at a $10B post-money — for a reported $3.5B from Meta. Jason is genuinely rattled: “people are checking out of $10 billion seed companies now… when I was a founder, there was no way I would leave, no matter how tough it was.” His sharpest line: “If my kid did that, I would not be happy with my kid. You leave the people that brought you to the dance because you see a prettier girl over here? Something’s broken in the way that we’re evolving as humans if everything ultimately reduces to what’s in it for me.”
Rory’s rebuttal, flagged as “genuinely not meant to be snarky”: Jason never “faced the existential dilemma of being offered $3 billion to quit.” His signature framing of the episode: “In the face of unprecedented wealth, I’m shocked to discover that most people behave badly. The loyalty conversation erodes pretty quickly when you enter the third comma on the check.”
Then the trade itself, put to Roger as a forced choice: “$2 billion in Thinking Machines unlisted stock with the chance to be amazing or the chance to go bust, or $3.5 billion of liquid Facebook stock over the next five years?” Roger concedes “obviously” — while insisting the embedded option value is real: “that could be a $500 billion company.” Rory’s translation: highly fixed $3.5B plus-or-minus 50%, versus two billion that “could be zero, could be 10.”
Rory also complicates the betrayal narrative: for the engineer — roughly 10-14 years at Meta, under a year at OpenAI, under a year at Thinking Machines — returning to the mothership “is kind of not an unusual pattern of behavior.” And how much emotional commitment did the VCs actually earn? “You thrust a bunch of money at people, some of whom you maybe didn’t meet at all, and less than 12 months later… oh, well.” Harry’s line in the sand: “you should be fired if you write $100 million plus check and you don’t meet the co-founders.”

4. Big boys rules: how investors should reprice the quit option

The actionable layer: founders and investors should treat extended and cliff vesting as core deal terms — “is there six-year not four-year vesting? Are there repurchase rights?… if you leave for a competitor, something really bad happens?” Rory frames it as founder-to-founder game theory: seven co-conspirators must price “how will I feel if one of my seven bails on me?”
Jason’s confession about his own underwriting: “I don’t believe liquidation preferences matter… my liquidation preference has always been knowing the founder would never quit. That’s my protection as a seed investor.” A world where the founder might rationally quit in six months is “a risk I’ve never taken in my history.”
Roger summarizes Rory’s realpolitik with relish: “essentially reduce it all to fuck the big VCs. They’re playing the momentum game. If shit happens, shit happens.” Rory: “Exactly. Big boys rules.” But his closing note is genuinely uneasy: when the core asset is seven minds that a rival will pay billions to poach, “it makes it real how risky those investments are, and I’m not sure what the answer to that is. It’s quite terrifying, really.” Roger’s generalization: collapsing careers from multi-turn games to single-turn games “wildly increases the volatility of potential outcomes.”

5. SoftBank levers Arm for OpenAI — “Masa being Masa”

Roger’s instant read on the $5B margin loan secured by Arm shares: “Masa rules. Nothing new to see here… When he has a feeling, he goes all in. All chips, max risk, personal, financial, everything.” He’s watched Masa through the Nasdaq run-up and crash — “so many existential moments where he’s waking up in the middle of the night, sweat pouring down his face… but he went right to the line, and he’s come out.” Verdict: “a relatively low-octane Masa move.”
The leverage math says more is coming: Rory checked — SoftBank still owns 90% of Arm at a ~$90-odd-billion market cap, roughly $80B of equity. Roger: a $100B position is “leveragable pretty much to $50 billion… he could take $25 billion against the Arm position easily.” Jason adds it’s actually smart treasury management — “rather than pay capital gains,” a margin loan is probably smart, and “that loan’s not gonna get called under any scenario probably.” Rory’s hedge, kept as hedged: in 2002 individual stocks fell 90% from peak, “so it is possible the loan will get called. It would just… it’s just unlikely.”

6. The scaling-law priesthood vs. the marginal capital provider

Jason tees up the bubble question: we’re building more data centers than office buildings — “is this not fundamentally different?” Rory swats the comparison (“who the hell is gonna be building offices? There’s no one in them”) but takes the substance seriously via Dwarkesh Patel’s Stripe Press oral history of scaling, read over the weekend: what struck him was “the matter-of-fact way” the field’s smartest people treat it — the scaling law “has been proven to hold for six, seven years now at a high degree of accuracy,” so “of course we’ll need 1% of GDP to invest in computers, but then we’ll be fine because we’ll have AGI.”
His caricature of the logic, worth quoting whole: “It was kind of like 10,000 computers worked, so we’re gonna buy 100,000. And then we’re gonna buy a million, and somewhere along the line we’ll get AGI. And what’s your point, and why are you even questioning it?”
The real constraint in Rory’s model: not technology, not demand — economics at the margin. “The scaling law might be log linear, but every economic phenomenon tends to be diminishing marginal utility. At some point, capitalism is gonna say, ‘I don’t know how to tell you this, guys, but you can’t have your $1 trillion dream because we just can’t afford it.’” The open question he’s trying to solve: does the economic return arrive quickly enough to warrant the investment?

7. Jason’s testimony from the token mines: demand is not the problem

The ground-truth counterpoint: Jason has vibe-coded eight apps in 100 days and runs 12 AI agents at Sastre that “replaced almost all of our sales team and our whole content team.” His conversion moment: “I wouldn’t have believed this 90 days ago, but folks like Amjad or Replit are saying, ‘You’ve got it backwards. Everyone will consume every available token.’” Today he could use 100x the tokens — “I gotta wait 20 minutes to build one feature… it don’t work at Google speed.”
The multiplier on the multiplier: it’s Dreamforce week, and per Mark, only 0.1% of Salesforce customers are really using AI yet — “so it’s, like, 100 times 100 times something. We’re not remotely servicing the demand that exists today.”
Roger probes the one deflationary vector — could step-change efficiency gains in processing shrink the infrastructure bill? Jason: “I don’t think so. I think we burn more tokens.” Engineers shipping 50% faster don’t take the afternoon off; they build another feature. “The better that gets, the more tokens you’ll consume.”

8. The investable window has compressed to half an hour

The consequence for seed: “so many of these companies are born almost instantly today… that company probably didn’t exist seven days ago, or being less facetious, 30. When we all started, startups were never good 30 days in.” The old tells are gone — “Aaron and Dylan built a folder you could put a file in. I’m in. Those days are long gone.”
Exhibit A: Lovable’s first anniversary at over $170M ARR. Rory’s framing of the squeeze: what every investor really wants is “that wonderful period where you know, but it’s not obvious… it turns out that period may have declined to, like, a half an hour. That sweet spot is vanishingly small, and therefore you’re left with the choice of do you invest into acute uncertainty or do you invest into two billion pre?”
Roger’s answer — and his new fund’s thesis: acute uncertainty “does not trouble me in the least when it’s expressing a deeply held, well-researched thesis,” and he’s deliberately hunting legal-and-regulatory-complex spaces (financial infrastructure, media rights, IP) where the contest isn’t just “do I have better or faster code.” Rory grafts on Aaron Levie’s concept: diffusion rates differ by market — Lovable’s market is “done and dusted in six months,” regulated verticals take two years to the first lighthouse customer, “but then it’s bowling pin and you get the other five in six months.”

9. Polymarket vs Calshi: regulatory arbitrage, and why king-making fails here

A week after Polymarket raised $2B at $9B, Calshi raises at $5B from Andreessen and Accel. Roger doesn’t dress it up: “the purest regulatory arbitrage play of all time. You can look at the cumulative market cap of regulated sports betting and how it has dropped in response” — value transferring to venues exempt from the rules legacy books have carried since PASPA, sprinting “to avoid parallel regulatory scrutiny.” Rory strips the last veil: only ~10% of the business is political prediction; “90% of their business is sports betting, but we’re not calling it that.”
On Harry’s king-making thesis from last week, Rory rules this a counterexample. Capital crowns a king only when (1) one company gets enough money to overwhelm the other, or (2) brand-name VCs cause customers to default to you — real in enterprise software (“Sierra’s amazing, we do want to take them on”), irrelevant here: “I don’t think anyone betting on Poly or Calshi gives a damn how much money they have, provided they can pay their bet.” Roger’s softer definition survives: funding an oligopoly’s marketing and distribution war chest is its own kind of king-making, as long as LTV to CAC holds.
Harry names what nobody else quite does: Eric Trump on one board, another Trump investing in the other, Howard Lutnick’s son running the fastest-growing investment bank — “an awful lot of coincidences in one go.” Rory’s zoom-out is ideological: regulate as little as possible, because “the minute something is regulated, people have an economic incentive to incentivize the regulators” — the only novelty is today’s generation does it “at scale… just give me 5% of the company.” Roger adds the fiscal spiral: as states like Illinois jack up gaming taxes, regulated handle shrinks and players drain offshore to Bovada, Crypto.com and Stake.

10. Founders Fund concentrates to ten — and why early stage can’t copy it

The Thiel news: Founders Fund shifts from caution to concentrated AI bets. Rory read the article and was struck backwards: growth fund one had 31 investments, fund two mid-high teens, fund three is aiming for ten — “the only surprising thing was they weren’t there already,” given their SpaceX non-diversification. The math is bloodless: “diversification reduces your upside. It also reduces your downside. It’s the central limit theorem… the more certain you are that you can call the shots, the more focused you should be” — and Thiel is ~40% of Founders Fund’s capital.
Jason’s read on the opposite camp — GC, Lightspeed, DST indexing Mistral-Anthropic-OpenAI: “being too diversified from investing in AI today is biding time. It’s not knowing… I think plan B is to make a lot of bets” — better than sitting out and sneering at the rounds.
Rory’s stage-bifurcation resolves the debate: concentration-to-ten fits growth funds “effectively investing in what should be public companies but are just private.” At the will-this-even-work stage you must diversify then concentrate — and the goalposts moved: “an exit is now $400 million in ARR at an IPO, not $200 million… the finish line has receded another two or three years,” so Rory says his fund is expanding from under 20 deals per fund toward 25 even as journeys stretch from six-seven years to ten.

11. Roger’s playbook: 20-25 names, then pile into 3-5 — because every check is a new option

The construction: 20-25 portfolio companies as “the farm team” with significant ownership per check, then deep concentration on second and third checks — historically 3-5 companies taking 75% of deployed capital. And the entry prices still exist: “we just wrote a one point five check at a ten post — fifteen percent ownership” in an analytics company with multiple six-figure ACVs. Rory, narrating Harry’s face: “what you’re seeing in Harry’s eyes is the wide-eyed look — can such things even exist? Yes, they can.”
Harry’s pushback on the whole premise — worth keeping: from his own fund-one portfolio reviews, the eventual winners weren’t obvious early and the early screamers (Clubhouse, Hopin, BeReal) “did not signify enterprise value… if you think you can pick your winners early, I think you are wrong.” Roger’s answer is that winners take radically different paths: TTD had “multiple near-death experiences, multiple bridges, no product-market fit for a year and a half — then once it hit, it hit,” while Wise was “as close to up-and-to-the-right as I’ve ever been involved with” — $750K first check at a $5.5M post, piling in with Valar at $20M then $160M, ending 13% at IPO; Datadog returned the fund at just 2.2% at IPO because it became a $40B company.
Rory arms the rebuttal with data: you don’t always know, but you know more than an outsider — at Rory’s stage, if a company hits the first two underwritten years of revenue, the probability of a >5x jumps from 30% to the mid-70s. “It’s not flip of a coin… you owe it to yourself to use that information.” Jason’s contrasting model: 8% of a fund into nearly every first check means “half of them have to work,” so he must turn away the Clubhouses and maybe the Datadogs — “it’s a stupid model… you gotta just find the Wises and go all in.”

12. Follow-on religion: “every check independent of the check prior”

The TTD sequence is the case study: four checks totaling ~$2M (pre-seed, bridge, bridge, a barely-Series-A at a $16M post), then an air gap, then a $3M check at a $280M post out of a $50M fund that turned into $40M. Same with DigitalOcean: $3M first check, then $7M more when Andreessen led the $37M Series A. Jason’s objection — today the next round is at $300M or $500M “because the AI kids come in,” so the second check barely moves ownership. Roger’s retort, verbatim: “Who cares about ownership? It’s a cash on cash business. If it goes from 300 to 10 billion and that check’s at 30X, that certainly impacts my personal economics.”
Rory names Jason’s real worry — if every follow-on round is mispriced against exit value, does the strategy die? — and answers it: then you simply don’t write those checks. “The worst thing that happens is my initial check gets marked up and I don’t need to chase the money… provided every check you write is money good, in the end you’ll die rich.” He pairs it with the Thiel heuristic: when a reputable outside investor marks up a deal you’re already in, “do everything you can in it” — adjust your scales upward, don’t anchor at your entry.
Jason’s honest regret closes the loop: he copied Founders Fund’s 10%-of-fund threshold for winners, but with three or four breakouts “you could exhaust 30 or 40 million of a $100 million fund in a year… I have come to regret some of my third checks.” Roger’s structural fix: parallel LPs across funds enabling conflict-free cross-fund investing (turning a $100M fund into a $260M one), plus recycling to 110-120% invested — and a refusal to “optimize my asset allocation because of the potential of uncomfortable conversations down the road.” His worldview in one line, the episode’s opener: “Everything in life you can price as an option. I walk through life, everything looks like the Greeks.”

China vs America: The Battle for Global Dominance Explained | Dan Wang interview Ben Horowitz and Ali Ghodsi: How to Run a $100 Billion Business