Summary

Fund size is your strategy because the power law is a curve, not a slogan: Pareto compounds, so 4% of deals yield 64% of returns — in a 25-investment fund, the single best deal must return 64% of everything. Maples’ pole-vault image: fund size is “the height of the bar that you set that you promised to jump over.” A sub-$100M seed fund works only if it’s “way less than 100” — think $10M writing 100K checks.
Seed is “hard but not complicated”: 5% of first checks at 100x cash-on-cash plus 10-15% at 20x gets you a 10x fund, and the loss ratio is roughly the same between a 3x and a 10x fund — only the magnitude of winners differs. Price therefore matters despite the fashionable “pay any price for a great company” line: at a 25 post entry with ~half dilution, the exit needs to be $5B. “I’ve studied venture returns for the last 50 years and the physics of what a good fund looks like has not changed.”
Follow-on capital is closer to index investing than seed funds admit: Floodgate’s biggest winners included Demandforce, Twitch, Lyft and Okta, with several later followed by major firms. Maples treats top-firm participation as a strong signal and gave one partner, Iris Choy, sole accountability for the 30% reserve pool. His warning shot: if LPs tracked follow-on returns against first-check returns, “there’d be pitchforks and revolts in the street.”
Seed funds are better positioned than multistage to sell — “by a wide margin” — because a $100-200M secondary sale moves a seed fund like an IPO but would neither move a16z’s fund nor survive the signaling. Floodgate’s 2015 “IQ test” Post-it: sell Lyft at ~$25/share (private marks then above today’s value), returning the entire fund. The move is to pre-agree it with the founder — and by round close “everybody becomes pigs” and the founder is begging you to sell more.
From deep-diving 100-baggers of the last 20 years (~100 exited, ~100 not), founder-future-fit is often the most discernible early signal, alongside insight and inflection — Zoom began as consumer “likely SaaSbee” with no obvious insight, but Eric Yuan had lived video conferencing at WebEx/Cisco for a decade. Floodgate’s worst errors were “failures of imagination” — passing Airbnb and Datadog — not bad follow-ons; the $500K minimum check also cost them 11 Labs at 25, “probably the best company coming out of Europe,” now $3B.
Temperament is the current edge: Maples made exactly one investment in all of 2021 (Hadrian) while everyone swung, and Ann wrote 750K into Lyft at $5.5M post in fetal-position 2009 (~250x). Buffett’s no-called-strikes framing governs: “if you’re not finding inefficiencies in the game you ought to be asking yourself what am I doing.” He thinks venture has too much money, exits are cyclical (half of profits land in 18-24-month windows every ~15 years), and multistage funds raised on 2020-22 exit comps that “just aren’t going to happen” again.
2025 calls: Bitcoin is “the thing hidden in plain sight” — a world exists where it’s worth more than gold with a financial ecosystem built on its rails; his year-end guess is 130 versus Reed Hoffman’s 200. SpaceX is his 2024 company of the year and he doesn’t “have to squint too hard” to see it becoming the most valuable company in the world as the platform supplier for space. His other 2024 awards were 20VC as fund of the year after its $400M raise, Elon as founder of the year, and Loom as exit of the year at $975M.

Deep dive

1. Fund Size Sets Strategy

Maples’ core math, stated as physics: Pareto isn’t just 80/20, it’s continuous — 80% squared is 64%, 20% squared is 4%, so 4% of investments yield 64% of returns. In a 25-deal fund, your best investment alone must return 64% of all profits; want a 5x fund and that one deal has to produce 64% of five times the fund. Hence the signature line: fund size is like a pole vaulter’s bar — “the height… you promised to jump over, and if you don’t jump over that height you have a bad fund.”
Can a seed fund under $100M work in 2025? “As long as you’re way less than 100” — 100K checks à la a Tim Ferris angel-plus-brand play, implying a ~$10M fund, not $80M. Floodgate itself went 70-80 → $150M, and Maples concedes “we were better at 150 than we were at 75,” though he attributes the dip to load, not size.
The Pinterest confession, worth keeping verbatim: on 22 boards, “your phone’s blowing up all the time… you’re not as awake to the possibility of what Pinterest could be when you get pitched by Pinterest.” He flew to Yale to tell likely David Swensen the mistakes he thought they were making — and restructured the follow-on model.

2. Follow-Ons Resemble Indexing

Harry’s pushback — his fund does zero follow-ons because “we grossly overestimate our ability to pick our winners.” Maples’ rebuttal: pro-rata is “a right that you have that nobody else has,” and setting reserves to zero means giving up something worth something. Floodgate settled on 70% first checks / 30% reserves, with Iris Choy solely accountable for follow-on returns — “I can’t strong-arm her into trying to protect some investment that’s not working.”
Against the “I know this company better than the market” instinct: if Benchmark, Sequoia, General Catalyst and a16z all pass on your portfolio company, “I used to say, you know things that aren’t so.” But when they chase aggressively, they’re picking from every seed fund’s portfolio — a strong signal, though not blindly followable (Sequoia aggressively followed VarageSale; it didn’t work).
The receipts: fund one’s top two were Demandforce (Gurley/Benchmark followed) and Twitch; fund two’s were Lyft (Mayfield, Founders Fund, a16z) and Okta (a16z, then Sequoia, Greylock). Conclusion: follow-on dollars are best deployed as a subset of where the best firms follow — “if the LPs tracked what’s the return on follow-on checks versus first checks there’d be pitchforks and revolts in the street.”

3. Seed Returns Require Outliers

The whole model in one sentence: “Our business is hard in seed but not complicated — 5% of our checks need to be 100x cash-on-cash on the first check and about 10-15% need to be 20x… you achieve that, you’re a 10x fund.” The loss ratio is about the same between a 3x fund and a 10x-plus fund; only winner magnitude differs.
His scenario planning inverts the optimist’s version: given it’s 85% likely a deal is not in the top 15%, ask “for this to make 100x on the first check, what would have to be true?” That kills the “if the company’s awesome you can pay any price” debate — true only to the extent 100x remains achievable. Entry at 25 post means the exit needs to be 2.5B before dilution — and with dilution “half, or more,” $5B.
Opportunity cost enforces it: a $150M fund gets ~40 shots on goal. Take one shot that can only be 20x — even with a great founder — and “now I have 39 shots on goal and one fewer way to make 100x.” His test for expensive inception rounds: “is the founder Caster Unice?” Applied Intuition at 10 on 40 post was “a real stretch” needing a north-of-$5B outcome; it just raised at 6 billion.
The discipline has a cost he owns: the $500K minimum check (“it needs to move the needle on the fund”) meant passing a 250K allocation in 11 Labs at 25 — now $3B and “probably the best company coming out of Europe.” His deflection, delivered laughing: “that’s why you have partners — they’re like children, you just blame them.”

4. Temperament Creates an Edge

Asked what advantage matters now that didn’t before: “having a temperament advantage makes a big difference.” In 2021, with rounds at 30-35-40, he and Ann just looked at each other — “we don’t have to do that” — and he made exactly one investment all year (Hadrian), telling restless associates “Silicon Valley will make more.”
The Buffett frame governs pacing: “investing is like a game where there’s no called strikes… I don’t see my pitch, I’m just going to wait until a meatball comes over the plate and swing at it with all my might.” The mirror image was 2009 — everyone “in the fetal position” while Ann funded Lyft at $5.5M post with a 750K check, roughly a 250x.
The Munger-derived mechanism: with a defined circle of competence, “if everything’s systematically overpriced you do fewer deals… if everything’s systematically underpriced you do more deals — but that’s the situation you want to be in.”
On Gurley’s “play the game on the field,” a precise splitting: “you have to play the game that’s on the field, but you don’t have to play the way everybody else plays… if you’re not finding inefficiencies in the game you ought to be asking yourself what am I in this for?” To Harry’s jab that Maples sits in the most efficient seed market on earth: the mistake is thinking of startups as a “market” at all — “there will always be 30 or so every year that are great.”

5. Seed Funds Can Sell

Two ways to make money: entry pricing inefficiency and exit price inefficiency — and “seed funds are actually better positioned to make money on the sale than anybody… by a wide margin.” a16z couldn’t sell Lyft: it would send a signal, and a couple hundred million doesn’t move their fund. For a seed fund it’s an “initial liquidity event” — Iris’s coinage — “an event that has the same impact on fund economics as an IPO,” not “10 million here, 15 million there.”
The Lyft case as told: 2015, private stock ~$25/share — worth more than the company is today — sitting behind a $1.5B preference stack, in at half-a-million post, competing with likely Travis Kalanick, “a freaking maniac who I respect a lot.” Ann put a Post-it reading “IQ test” on her monitor in January; in 2015 she’d sold enough to return the entire fund.
The founder-alignment move: agree in advance that if the next round clears a threshold, you’ll make room for Fidelity-type holders. “What ends up happening in reality is by the time the round comes together the founder’s coming to you saying, dude, you got to do me a solid… I need you to sell more, because everybody becomes pigs.” He sold some Applied Intuition the same way — “in full cooperation with [Caster],” never the day the round closes.
He keeps the counterargument intact: Brian Singerman’s “value of the next double,” and Bessemer selling Shopify at $2-3B — “probably the worst financial decision ever… they will say the same.” Lyft later traded to 75 and Floodgate exited around there post-lockup, so yes, “you never make a trade that you don’t somewhat regret — unless you sell at the top.” Still “the right risk-adjusted decision.” On Avi of Entrée’s mechanical thirds structure: case by case — but decide “when we were sober,” pre-committing conditions rather than improvising in the euphoria.

6. Failures Reflect Imagination

His worst mistakes weren’t bad follow-ons: “our biggest failures have been failures of imagination” — passing Airbnb (“it was a nuts idea, he wasn’t from a blue-chip company”) and Datadog. The response isn’t self-flagellation but system-building: “is there a set of frameworks that we embrace today that would have caused us to say yes?”
Hence the 100-bagger deep dives (Marquetta, Zoom, others): he tracks a little over 100 exited and 100 non-exited hundred-baggers of the last 20 years “like a trainspotter.” Floodgate’s own tally is three or four: Twitter a little over 300x, Lyft ~205x, Twitch 94x (“close but not quite”), Applied Intuition “encroaching on 100x.”
The three frameworks: did they have an insight; did it harness an inflection (Lyft rode the iPhone 4S getting a GPS chip); and founder-future-fit. The dives reconstruct a “time capsule” — what was actually knowable at the seed round — and the repeated finding is that founder-future-fit is “the most discernible way to figure out if the founder’s likely to figure this out.” Zoom is the proof case: it started as consumer “likely SaaSbee,” the initial product vision “was just wrong,” inflection and insight both “hard to argue” — but likely Eric Yuan had lived video conferencing at Cisco/WebEx for ten years.

7. Future-Fit Founders Matter

The definition, via William Gibson’s “the future is already here, it’s just not evenly distributed”: great startups come from “a founder who’s living in the future and who notices what’s missing in the future and builds what’s missing in the future” — obsessed trainspotters, like Newton’s answer on gravity: “it’s because I was thinking about it all the time.”
Okta as the canonical pitch: Todd McKinnon, VP of engineering at Salesforce, saw cloud adopters accumulating identity problems — “if anybody can do it, he can do it, and if anybody knows, he knows.” Living in the future makes founders both more likely to know what to build and more credible to early believers; the closing question is “is this team the most likely team in the world to make this future real the quickest?”
It doesn’t exclude first-timers: likely Marc Andreessen at Illinois “didn’t know what markets were,” just built what the internet was missing — he was “living in a time machine,” on tomorrow’s machines and protocols, while everyone assumed AT&T, Time Warner or AOL would build a top-down digital superhighway. “Mark’s advantage was not born of his experience in business, it was born of his experience with the future.”
Assessment candor: his stated biggest weakness is being “too optimistic about whether people can pull it off — exceptionalism is so rare.” Liking the founder is irrelevant: “a breakthrough startup is a provocative act… quite often these founders are disagreeable people, because the present will fight back, and it won’t fight back fair.” When he loses faith: “detach with love” — step back without recrimination, door open.

8. Product-Market Fit Matters

The categorical claim, kept categorical: “I’ve never worked with a company that got product-market fit that wasn’t wildly successful.” Harry’s direct challenge — Clubhouse had millions of users engaging for hours daily — draws the distinction: “I don’t think they [had] product-market fit… they were like a solar flare.” Harry’s alternative framing: PMF is chapters in a book you must keep re-earning; Maples cites a Robinhood co-founder: “when we got it at Robinhood I was like, oh, that’s what product-market fit looks like.”
What a seed investor can actually do about it: not much — “addition by subtraction.” Channel Lombardi: “product-market fit isn’t everything, it’s the only thing — eliminate distractions,” and keep asking “last time we talked you said this is the bottleneck… is that still the case?”
The unvarnished aside: “being a founder is not a fun job… almost like being an artist, sometimes more of a curse than a blessing… it’s a hard freaking job.”

9. Capital Constraints Matter

Harry’s worry, which Maples endorses: growth funds assuming dollar-efficiency is stage-invariant, dropping $100M “two years ahead” into companies — which then “do 10 other things” and destroy the likelihood of the big exit. Worse is raising big before legitimate PMF: hire ahead of it and companies “become culturally broken… they never develop any muscle memory for what an attractive customer is.”
His verdict on the $100M+-round, $1B+-price, no-PMF cohort: “most of them won’t clear their preference stack” — but with five-to-seven years of runway “GPs just keep going and telling LPs it’s fine,” because “everybody in the game has an incentive to keep the plates spinning.”
The seed-round first-principles rebuild: its purpose is “slightly more than the minimum viable amount of money” to kill the single biggest risk — proving a non-consensus insight true — at the stage where capital is most expensive because “you never get more diluted than in the seed round.” Raising $4M “because that’s what it takes to dilute 20%” is “just stupid,” and it hurts founders more than VCs: “the one thing you never get back as a founder is your time.” Harry agrees-but-disagrees — you have to re-prove the insight at every stage — and cites likely Klaviyo, likely UiPath, and ServiceTitan as constraint-built winners; Maples: “constraints are the thing that allows you to understand what the true laws of physics are for your company.”
The macro overlay: exits are cyclical — every ~15 years “close to half of the exit profits are made in an 18-month-to-two-year window” (crediting likely Horsley Bridge’s analysis), so the game is having good companies in flight when the window opens. Multistage funds “raised money on exits predicated by 2020-2022… there’s a lot of evidence that’s just not going to happen” — they’ll rationalize fund sizes “slowly and quote-unquote deliberately.”

10. 2024 Review and 2025 Calls

Company of the year: SpaceX — “I don’t have to squint too hard to see a world where they’re the most valuable company in the world,” the platform-dominant supplier for space, launching everything from arbitrary global broadband (his contrast: a $40B federal broadband bill that built nothing versus satellites over hurricane-hit North Carolina) to likely Baiju Bhatt’s solar-beaming satellites. Investor of the year, as homage: Charlie Munger. likely Howard Marks is the other formative influence — Pattern Breakers is second-level thinking radicalized: “only by being radically different can you make a radical difference.”
The other 2024 awards were 20VC as fund of the year after its $400M raise, Elon as founder of the year, and Loom as exit of the year at $975M.
The Bitcoin call, hedges intact: smart crypto VCs focused on Ethereum and likely Solana, but “I look at Bitcoin and it just feels to me like it’s the thing hidden in plain sight… there’s a world where Bitcoin becomes more valuable than gold and then some,” with a startup ecosystem building rails around it. Year-end 2025 guess: 130; Reed Hoffman, asked the same morning, went 200.
On DOGE: success means changing “the cultural norms of what’s acceptable from an accountability standpoint” — today’s government is the worst company you’ve ever seen, where “the worst departments get the most money.” He flags Joe Lonsdale’s project Cicero as the same fight at the local level. (Separately, an enterprise-AI portfolio company he also calls Cicero is “doing really well” — same name as spoken, apparently distinct.)
The closing register: Christianity’s philosophical gifts — forward arrow of time, inalienable human rights, unconditional love (“boundaries are different from conditions”). And from his father — whom Harry described as a very early Microsoft employee — via Adam Smith’s comparative advantage: “do your best, don’t be the best” — likely Peter Thiel wins the macro-history game, “but if it’s who’s the better philosopher-king about seed… I think I can win that game against anybody.”

University Students on Relying on and Adapting to AI Volodymyr Zelenskyy: Ukraine, War, Peace, Putin, Trump, NATO, and Freedom | Lex Fridman Podcast #456