How We Got Fred Wilson, Benchmark and Index to Invest $94M | Why Robinhood's Strategy is Wrong

Summary

• FOMO closed a $75M Series B at $550M post-money — Index $55M, USV $15M, all inbound, with the price anchored by the founders naming their number “far before the term sheet.” Combined with Benchmark’s A and the angel round, this is the broader financing story — raised by a company of 17 people that has “much more money in the bank than we ever raised” and framed VC not as fuel but as insurance against “a sudden market turn just wiping us out.”
• The org-design call investors should sit with: AI enables dramatically smaller teams — no meetings, no 1:1s, no hierarchy, headcount “hopefully below 25” in a year. Erlanger dismisses Uber’s and Microsoft’s doubts on AI coding gains (“it’s definitely faster… faster to thoroughly review even than write”) and says a product that is basically another app’s entire product was built in three weeks. Corollary: give non-founders founder-grade equity — “we gave nonfounders a percentage of the company that usually founders get” — because five to ten owners building for ten years means “there’s literally nothing stopping us.”
• On Harry’s token-economics test — a cited $300M Anthropic spend is ~3.8% of dev salaries; if that ratio holds, “$1 trillion for OpenAI and Anthropic is grossly overvalued,” at 20% they’re “$5 trillion companies” — Erlanger says 20% of dev salaries on tokens is “definitely within reason,” hedged only by his hope for a race to the bottom and “not… price collusion.”
• The anti-super-app thesis is the product strategy: “everything app means not intentional.” FOMO’s glue is the social graph plus thesis expression — believe the Strait of Hormuz closes and you buy oil on Hyperliquid, short oil-dependent equities, and buy the prediction market in one place. Robinhood gets partial credit: it saturated ~20M+ US funded accounts “but they weren’t able to go global,” which is why its tokenized-equities push is aimed at global distribution — while Harry argues Revolut, global-from-Europe, may win (Paul’s head says Revolut, heart says Robinhood).
• Pre-IPO perps are the retail access wedge: synthetic side-bets on price with no underlying transfer — no SPV needed (“we could have a perp on whether Anthropic will go up or down”). Evidence it works: at the likely Cerebras IPO “the Hyperliquid price started to converge to that price at IPO.” The test ahead: SpaceX first to market with ~30% devoted to retail — “if SpaceX performs well, the OpenAI and Anthropic IPOs will also go very well. If SpaceX performs poorly, those are going to have a hard time.”
• The consumer playbook: momentum is everything and can be manufactured — share cards let anyone broadcast your wins and “fumbles” (Iceman turned $10K into $2.5M overnight; Remis ~$300 into $1.5M in a month), while Clubhouse lost its core user base after importing celebrities and BeReal lost momentum after requiring daily effort. His quickfire change of mind: he underweighted social — “you need to be very intentional about the social graph.”
• Tactical gems: wait to announce your last round if you’ll raise again soon (announcing triggers time-wasting inbound); raise CAC even at flat LTV to reach past low-hanging-fruit users; and don’t pick the highest-tier or highest-priced VC — Benchmark was “close but not exactly the highest” — pick who you trust on the weekly call.

Deep dive

1. The angel-only round was a distribution strategy, not a financing

• Erlanger’s logic for raising from 140 angels and zero institutions: B2B firms hire sales teams, but consumer’s killer is the cold start — “there’s a lot of great products that just never get off the ground.” So the round’s job was distribution: “our best users should have some ownership in the product.” Not all 140 trade; the builders among them keep getting leveraged.
• The single best angel: Aaron Harris, ex-YC, who “understands financing really well” — “a small change in a term sheet could completely change the trajectory of your company.” Aaron later made the Benchmark intro.
• First-thousand-users advice: talk to them at 10, then 100, then 1,000. Early access via Telegram channels of top traders meant the web app “probably became twice as good just in that week” — unpaid users, “no other strings attached.”
• On when to ignore feedback: “you just have to be super epistemically modest because sometimes a user doesn’t even know what’s actually best for them” — filter through intuition and the larger vision, since some product decisions “could potentially kill the product.”

2. “Everything app means not intentional” — the case against the super-app

• Asked why Revolut/Robinhood-style bundling is wrong, Erlanger asks back: “What is the glue between these things?” At FOMO it’s the social graph plus thesis expression — if you think the Strait of Hormuz will close, you buy oil on Hyperliquid, short oil-dependent US equities, and buy the prediction market: “these different market types exist… to express conviction on a belief.” Super-apps have “no intentionality behind why all those things have to exist in the same place.”
• Pre-IPO perps are the access mechanism: a synthetic side-bet on price, no transfer of the underlying. When the likely Cerebras IPO happened, “the Hyperliquid price started to converge to that price at IPO” — with photos of people on the NYSE floor watching the Hyperliquid UI.
• Prediction markets aren’t integrated yet; a first version would build on Polymarket or Kalshi, but “there’s a lot in flux around the regulation of these businesses. So we want to watch and see.”

3. Robinhood saturated America but didn’t go global — global reach is the counter

• Erlanger is more charitable than the episode title: he “understands why they horizontally scaled.” Robinhood saturated the US — “a little over 20 million funded accounts” — “but they weren’t able to go global,” which is why the onchain tokenized-equities push matters: as it goes global it can stop stacking products horizontally.
• Why Revolut is well positioned against US fintech: Europe’s many-country patchwork let it saturate a larger base and go global faster. The pattern he draws: “look at Facebook, look at WhatsApp… global from day one.” Quickfire verdict — Harry’s read that his “heart says Robinhood and your head is saying Revolut” gets a laughing “I think that’s right.”
• On casinoization, he rejects the frame: “casino is kind of a derogatory way to view it… it’s somewhat empowering” — GameStop was retail coalescing to “fight back against the institutions,” and FOMO is the real-time public version of Wall Street Bets. Deeper claim, hedge intact: “most financial assets are speculative… most people when they buy a stock, they’re not looking to get dividends” — and he explicitly won’t take a normative view.

4. Eight months of no pay, founder-grade equity — “there’s literally nothing stopping us”

• For the first eight months nobody at FOMO took pay — senior engineers “taking a bet on the company.” Then the grants: “we gave nonfounders a percentage of the company that usually founders get” — Harry’s read, confirmed: roughly 2–3% each to five-to-seven people, an “extended founder team.” The AI-era corollary: teams are dramatically smaller, “which is why it’s okay to give more equity early.”
• The Series A wasn’t about needing money. Watching Robinhood’s and Coinbase’s stock volatility while being five-to-ten-years focused, they feared “a sudden market turn just wiping us out” — venture capital as downside protection.
• The Benchmark story: met likely Chetan (the name is unclear in the captions) on Friday, full partnership Monday. Peter Fenton was on his phone for most of the pitch — Erlanger was “bummed out… he’s not interested” — until Fenton’s first words: “guys, I love the app. I’ve been on it the entire time.” Benchmark’s price was “close but not exactly the highest.”
• Investor selection: “This is the person you’re going to call every week… pick the person that you trust will help you scale your business,” not the highest tier. And a joint red flag: if the founder isn’t personally hiring engineers, “we got a problem” — the best hires come from months of relationship-building, not recruiters.

5. The $75M Series B at $550M post — and the announce-timing trick

• The round: $75M — Index $55M, USV $15M — at $550M post-money, all inbound. The price was anchored deliberately: “far before the term sheet… they were like, what price would be interesting to you, and we set a number” — “that kind of helped anchor the conversation.”
• Fred Wilson stood out on product: “Fred actually has a really good product intuition… very rare for a VC” — decentralized networks and their network effects are, per Harry, “right in the mesh” of his passions. Wilson helped before investing; he may have missed the A because he was traveling and the times didn’t match, which Harry flags as the job’s hidden cost — “a holiday… can lead to hundreds of millions of dollars lost.”
• The discovery worth stealing: “if you’re trying to raise another round, wait to announce your last round” — announcing triggers a wave of inbound you must spend time declining. Harry’s addendum: “a VC’s job is to meet with people. Never forget that” — a meeting is not a round.

6. 17 people, no 1:1s, no org chart — and 20% of salaries on tokens is “within reason”

• FOMO is “extremely horizontal”: no meetings, no one-on-ones, no hierarchy, everyone self-reporting — 17 people, “hopefully below 25” in a year. Against Uber’s and Microsoft’s doubts on AI coding gains: “it’s definitely faster… it’s a lot faster to thoroughly review even than write.”
• The proof points: staff front-end engineer Tina has AI scaffold unfamiliar components, then “will go back through and even restructure and rewrite most of the code” — and a product that “is basically what entire other apps’ entire product is” was built in three weeks; the web app in a month. Stack: likely Claude Code and Codex on enterprise accounts, with frictions noted — “these models degrade and then the Claude Code credits got really expensive recently” — but price sensitivity “not at all.”
• Harry’s industry-sizing question: a cited $300M Anthropic spend is ~3.8% of dev salaries — stay there and “$1 trillion for OpenAI and Anthropic is grossly overvalued”; at 20%, “they’re $5 trillion companies.” Erlanger: 20% is “definitely within reason,” hedged by hoping for a race to the bottom, commoditization, and “not… price collusion.”
• Harry relays Paul Graham’s new YC question (from “last night”): how do you “AI-protectify your product” with non-AI defensibility — FOMO’s answer is the social graph. On design tooling, Erlanger backs Figma’s hybrid over Lovable: “humans still want to feel like they’re in control,” and Lovable “haven’t built the human-centric software” — adding the LLM on top is the easier direction.

7. Why no big social company since Snap — and how to manufacture momentum

• Consumer is unforgiving: “some small mistakes could be pretty existential.” Clubhouse’s key mistake — importing celebrities who “overshadowed the core user base that actually would love the product.” The lesson: cultivate native creators — Logan Paul got big on Vine, Charlie D’Amelio on TikTok — “I’m not trying to get LeBron James to trade on FOMO.”
• BeReal lacked the feedback loop: it “required people to do something every day, and people don’t want to have to do something every single day.” As soon as you lose that, momentum disappears quickly.
• FOMO’s synthetic momentum machine is share cards: anyone can broadcast your positions — or your “fumbles,” how much you missed by selling early — to other platforms, pulling viewers back in real time. The proof cases: Iceman turned $10K into $2.5M overnight; Remis turned ~$300 into $1.5M in a month, both shared publicly off-platform.
• His quickfire change of mind: he underweighted social — “we were doubling down on the trading product and assumed people will come to trade… But you need to be very intentional about the social graph.” Still, trading comes first: “we have to always be the best trading app in the world because the top traders won’t use us otherwise.”

8. Growth stops being intuition and becomes a numbers game

• FOMO runs 30–40 creators fully in-house with creator managers, constantly culling. Creator quality “is honestly a numbers game”: impressions and conversions against CAC — measured on attributed deposit-and-trade revenue — versus LTV.
• His self-described most important growth lesson: when a format works, don’t hunt for the next thing — “continue to iterate on that and make it better and better and better… and then replicate”: font, color, placement, which person is talking.
• Counterintuitive CAC advice: deliberately raise CAC even at flat LTV — the $30-LTV user who costs 80 cents is low-hanging fruit; the one needing $3 and ten impressions is still profitable. Two opposing forces: iteration pushes CAC down; incremental-user difficulty pushes it up. Harry adds that brand proliferation — becoming the default provider — can push it down again.
• Harry’s brand doctrine — buy “immortal assets”: a Bill Gurley episode sponsorship still earns thousands of plays a month three years on; Man U shirts still carry a decade-old Vodafone logo; a two-week billboard dies. Erlanger’s data counterpart: Robinhood’s Android deposits ran half of iOS — not richer iPhone users, but a loading screen twice as slow on Android; fixed, the numbers converged.

9. Crypto’s broom-sweeper, and the SpaceX perp test

• On Trump: “a lot of positive movement on regulatory clarity, but… sentiment for crypto has also gone down significantly” with the public — scam coins with no consumer protection (“there was actually a real one they were trying to buy, but they bought the wrong one”) burned people “so many times in a row that it led to all this negative dogma.” FOMO’s stated goal, “as cheesy as it sounds”: be a beacon of light — “someone’s got to come with the broom and sweep up the mess.”
• Perps kill the SPV — riffing on Harry’s “triple layer Anthropic SPV” joke: “we could have a perp on whether Anthropic will go up or down, but we don’t need to actually transfer the underlying Anthropic asset.” The coming test: retail demand is finite, SpaceX is first to market with ~30% devoted to retail, and sequencing matters — “if SpaceX performs well, the OpenAI and Anthropic IPOs will also go very well. If SpaceX performs poorly, those are going to have a hard time.”
• Quickfire keepers: advice to CS students is “use less AI” — the best engineers learned without it, and he avoids AI in his own writing (“I’m pretty scared” of forgetting what good writing is; Harry agrees on social posts, disagrees hard on coding — “use it as much as freaking possible”). To his younger self: “have the hard conversations sooner.” On 996: no numbers attached — with real ownership, “you can’t beat a team that’s having fun.”

Verification Notes

• Raw captions render the Benchmark contact as “Chaan” and the IPO company as “Cerebrris”; qualified forms preserve that uncertainty.