Kyle Harrison
← All podcasts
interview September 18, 2025

“90% of VCs Are Neutral to Negative Value”

Run the Numbers · hosted by CJ Gustafson

Watch on YouTube ↗

Summary

A wide-ranging conversation with CJ Gustafson on the Run the Numbers podcast about the evolving mechanics of venture capital. Kyle lays out his core framing — the industry has bifurcated into capital glomerators (mega-funds optimizing for AUM and fees, comfortable with 1.5–2x returns) and cottage keepers (small, deliberately-capped funds that need 7–10x to work) — and argues the “stupidest game” is playing one game without realizing a competitor is playing a fundamentally different one that disadvantages you.

Key threads:

  • Fund size is strategy. A $5B fund needs a handful of generational outcomes just to clear its losses; that financial gravity warps behavior. The “unholy trinity” of yield-farmer LPs, capital-glomerator funds, and capital-absorbing companies (OpenAI, Anthropic) is a “match made in heaven” for deploying enormous sums.
  • Investing in competitors. With founder-friendliness no longer decisive, the biggest firms back multiple companies chasing the same thing — a16z across OpenAI, Thinking Machines, xAI, Mistral AI, and Safe Superintelligence. When you must deploy $2B, you cannot afford to miss the winner. The sin of omission (missing the next generational outcome) is infinite; the sin of commission is capped at 1x.
  • The founder’s trap is optionality, not dilution. Raising a high-priced round from a big firm means picking up “the other end of the stick” — a liquidation preference that can turn a healthy $50M-revenue business into a non-event for founders and employees.
  • Narrative has tangible weight. Riffing on the greater fool dynamic and Kyla Scanlon’s work, Kyle argues story and vibes have become real economic drivers (Cursor vs. Windsurf, the Sydney Sweeney / American Eagle pop). He closes on a reversal of his own prior belief: he used to think reality always eventually beats narrative — now he thinks narrative carries real weight onto the scale.
  • Unbundling / borrowed credibility. What founders actually hire VCs for: Marc Andreessen’s “borrowed credibility,” offset against Vinod Khosla’s claim that most VCs add zero value and a chunk add negative value — “90% of VCs are probably neutral to negative value, we just don’t know which 10%.”
  • Compounding and FTX. The takeaway from Sam Bankman-Fried: acting immorally forfeits the right to compound. The one unforgivable move is getting knocked out of the game — as soon as you’re out, compounding stops. Coinbase’s survival through crypto winters is the counter-example.

Transcript

Kyle (cold open): What do founders hire VCs for? What is the most important thing that you pay for when you hire a venture capitalist? Marc Andreessen had an AMA a year or two ago where he had a great answer — he said, “I have always thought that the thing founders are hiring VCs for is borrowed credibility.” Let’s say you were an engineer at Databricks for a couple years and now you’re leaving to start a company. Who are you? Maybe you have a little brand rub-off from Databricks, but by raising from a venture firm that other people respect — the firm has built respect, through the companies they’re associated with, whatever — then as a hirer it’s de-risked for me. As a partner, as a customer, I feel like that’s de-risked. The other argument comes from Vinod Khosla, where he says the vast majority of VCs provide zero value, and an unfortunate portion of VCs drive negative value — it’s actually worse to have them along for the ride.

CJ: Welcome back to the Run the Numbers podcast. Today I have my friend Kyle Harrison, general partner at Contrary, one of the sharpest minds thinking and writing on Substack about the evolving game of venture capital — check out Investing 101. Kyle’s explaining the madness to us: why fund size isn’t just a number, it’s a strategy; why some firms are playing for yield instead of upside, and why that changes the game not just for VCs but for the founders they back. We talk about how the biggest funds are warping incentives, how misaligned fund size can sabotage great outcomes, and why the most dangerous trap for founders isn’t dilution — it’s optionality. Is it worse to back a loser or to miss the next Figma? Kyle shares his take on venture’s original sins.

[Sponsor break.]

CJ: Kyle, man, this has been a long time coming. Thanks for joining the pod.

Kyle: I’m excited to jam. You’re a kindred spirit, so I’m sure we’ll enjoy the conversation.

CJ: There’s no greater fan of your writing than me — you were one of the reasons I started on Substack. You found product-market fit, or audience-market fit, before I did.

Kyle: I don’t know about that — you’ve built a true empire.

CJ: That’s the benefit of now making this my full-time job. You’ve had a ton of success before coming to Contrary, and now you’re building this. I want to pick your brain on the trends we’re seeing within venture capital, and reference a couple of the pieces you’ve written. From the top: are you a little grumpy that these big funds are coming in and messing everything up for the cottage keepers?

Kyle: It’s funny — when I jam with people about the stuff I’ve written, I find some people say, “Man, these glomerates are such a fascinating business model, so compelling.” They love it. Then I meet other people who say, “This is disgusting, a bastardization of everything pure and good about venture capital.” I’ve experienced people on two dramatically different extremes — and who each assume that I’m on their extreme. I take that as a compliment: I think I’ve done a pretty good job stating the reality of the situation and trying really hard not to pass judgment on “this is bad venture capital” versus “this is good venture capital.” I frequently pass judgment on what I think is bad behavior — there are plenty of hucksters and charlatans I’m happy to trash. But on the fundamental business models, I go back to a piece I wrote about playing different, stupider games. The stupidest game to play is playing a game not realizing that somebody else is playing a fundamentally different game that disadvantages you. So I like glomerators, I like cottage keepers — to each their own, they can do interesting things. The only thing you have to appreciate is that they’re playing dramatically different games, and everybody needs to adjust their strategy accordingly.

CJ: You’ve described the VC world as divided into these cottage keepers and capital glomerators. In plain terms, what does that encapsulate — with a couple of examples people can lock onto?

Kyle: The clearest illustration is to look from the outside in, because the internal marketing of how a firm talks about itself is very similar. Every venture capitalist talks about themselves as founder-friendly, a champion of the underdog. The marketing of venture capital and the business model of venture capital are very different — don’t get distracted by how people talk about themselves. If you step back and look at different firms, you can clearly see there’s an optimization for a very different number. The number capital glomerators optimize for is AUM — and you boil that down to fees. Every venture firm on average makes 2% of AUM in fees every year, plus 20% of the upside, the carry. When you’re growing massive-AUM firms — 50, 60, 70 billion in AUM — the fee is massive. You’re making hundreds of millions of dollars a year on fees. Then you have firms that keep funds fairly balanced — usually less than two or three hundred million per fund — and it’s clear they’re optimizing for the ability to drive 10x returns. A $200M fund, you can turn into $2B. When you have seven billion dollars of funds, it’s very difficult to turn that into 70 billion of value. So glomerators are generally focused on 1.5–2x-ish returns, whereas cottage keepers keep funds very small deliberately so they can turn that into a 7, 8, 10x-plus fund. Just very different models.

CJ: And there’s this inherent pull toward taking on as much money as possible. But to play it back — it’s harder to find enough great investments to park that capital into.

Kyle: The pull of a bigger fund — there’s a Steve Schwarzman quote, the founder of Blackstone, where he says it’s just as hard to build a small business as a big business, so why not build a big business? There’s validity to that. But there’s a breakaway point — at a certain size, the sheer weight of financial gravity becomes a difficult obstacle. The gravity in building a bigger and bigger venture firm is your point: there are only so many companies that can be truly generational to generate these massive returns. If you want to generate a 3x fund on $5B — I’ve done the math in a couple of my pieces — you need to turn $5B into $15B of value. The power law holds even in larger funds, so 80–90% of that is going to zero, a chunk returns something small, and you’re going to need a handful of companies to generate dramatically more in market cap to justify your losses and still get to that $15B. The number of companies that can do that is very, very small. That’s why in one of my pieces — the unholy trinity of venture capital — you have these yield-farmer LPs who aren’t looking for 10x, they need 4 or 5% IRR yield; they pour that money into glomerators who are racking up massive funds, trying to be everywhere in everything and take up all the oxygen in the room; and then you have these hungry-hungry-hippo capital absorbers — companies like OpenAI, Anthropic — that want to raise tens of billions. Those three are a match made in heaven. You see it — the Anthropic CEO said a couple years ago, “We’ll never take Saudi money, we’re never going to go looking for these ethically questionable sources of capital.” Then recently he said, “Well, it’s expensive to build these safe systems, we’ve got to go.” So now they’re taking Saudi money — of course they are, because that’s the triumvirate they’ve built for themselves.

CJ: The hungry-hungry-hippo analogy hits, because it is a marriage made in heaven when you can take on so much capital. What comes to mind is when a16z invested in Adam Neumann’s new startup — where can I park, I think it was like 300 million?

Kyle: It was $350 million. It was Andreessen’s largest-ever single check.

CJ: And you think, okay, we hope we can 3x this, 5x this — but the management fee on that alone is huge.

Kyle: I have no inside knowledge of how a16z thinks about their funds, but my sense is they don’t — maybe for plain vanilla math they benchmark to 3x, but they’re definitely not thinking about 5x. And candidly, even if their funds return 1.5, maybe 2x, generally their LPs are fine with that, because if you’re an LP parking $200M a pop into these funds, you don’t need it to generate 10x. That’s why I call them yield farmers — when you’re CalPERS managing 100-plus billion, you don’t need 10x-es, you need 4 or 5% yield. The Neumann example is perfect: when it happened people said, “You can’t deny he’s built a big business.” Right — maybe. But you can also not deny he built a bad business. The thing he’s good at is raising a ton of capital, telling a massive story, chasing a huge vision that will justify pouring hundreds of millions, if not billions, into that story. From a16z’s perspective: most stuff fails, so yeah things will fail, but if this works we can’t afford not to be in it. It’s also one of the reasons a lot of these firms have started investing in competitors — they say, “We don’t really care anymore about founder-friendliness. We just have to be in the winner, because if we’re not in the winner we cannot justify the volume of capital we’re deploying.”

[Sponsor break.]

CJ: We’ve got to touch on that. I was listening to Jason Lemkin on 20VC the other day, and he was on the idea that when you back a founder in one space you can’t back their competitors — people have a visceral reaction: you believed in me, and now you’re doubling down on my direct competitor. A lot of funds say “we’re founder-friendly,” but it sounds like a lot of that’s going out the window. Their downside is they lose 1x, but what happens if they miss out on an OpenAI or Anthropic and the upside was 100x?

Kyle: 100%. I wrote a piece a couple months ago called the horse, the jockey, or the whole race, based on a conversation with Eric Newcomer about a tweet — there’s this anonymous account, Endowment Eddie, who’s amazing, you should follow him. He tweeted, “Wait, so is investing in competitors on the table now for VCs?” And I posted a picture I made super quickly: Andreessen is invested in OpenAI, Thinking Machines — one of OpenAI’s former co-founders’ new company — xAI, Mistral AI, and Safe Superintelligence. Five companies right there basically all chasing the same thing. A number of firms are invested in OpenAI and Anthropic, xAI and Anthropic and OpenAI. It’s very incestuous when you have the largest firms and the largest AI companies. If you looked at the Uber-versus-Lyft battle years ago, there was some overlap, but that was a baby version. AI is that times ten on steroids. There are a few exceptions — firms like Thrive and I think Founders Fund who’ve been monogamous, committed to investing in one company. But if I’m a16z: why would I invest in multiple companies chasing the same thing? Because founder-friendliness no longer matters as much when a firm has so much capital, so much power, so much halo effect that it becomes a black hole of influence you can’t ignore. Even if you don’t like them, it’s really hard to ignore them because they’re so powerful. Maybe they justify it — “OpenAI is focused on consumer AI, Thinking Machines is more focused on X, Safe Superintelligence on a specific kind of model development.” That’s lip service. These are all companies chasing superintelligence, whatever that means, through building the highest-quality models and deploying them in the highest-volume use cases to get the volume of data and feedback. If you’re a16z or Sequoia or Lightspeed, your thesis is: we have to deploy $2 billion to justify the existence of a $2 billion fund. If OpenAI becomes the breakout — already a $200 billion outcome — but if it becomes the $1 trillion outcome, you cannot miss that winner.

CJ: What blows my mind is how much times have changed. Five years ago, if I said a fund would make 14 or 15 investments and they’re investing in both Brex and Ramp, I’d say, “Isn’t concentration supposed to be a feature, not a bug?”

Kyle: The biggest problem these days is the volume of capital. A lot of people say it’s a fundraising winter, it’s really difficult for funds to raise. That’s true — but it’s not that capital isn’t being raised. On average most venture firms are having a hard time, but the top five are raising 70–80% of the capital raised in a given quarter. It’s the exacerbation of the power law and the hollowing-out of the middle. That’s true in companies, in funds, everywhere — it’s really difficult to build a middle-ground company that’s big but not too big, and it’s hard to build a venture firm that’s big but not too big. Once you raise a billion dollars, to deploy that capital when there are so few truly generational businesses, you have to compete with people who have $5 billion funds. To compete with a $5 billion fund, a billion feels like not enough — we need to write a $200 million check comfortably. At a billion, that’s 20% of our fund; uncomfortable. But at $3 billion, that’s not as uncomfortable. So you get pulled into having to compete with the bigger firms. It’s really difficult for anybody to exist in the middle ground.

CJ: You’ve said your fund size is your strategy. Can you give an example of when that disconnect between fund and founder became a problem?

Kyle: If I had to articulate the number-one thing that drives me to write, it’s the volume of conversations I have with founders who lack a fundamental understanding of how venture capital works as an industry. That scares me. These are people making very big decisions with implications for their entire business and career, and they’re not taking into account the underlying psychology and economics driving these venture firms — through no fault of their own. Venture is opaque and difficult, and it’s changed dramatically even in the last five or six years. So I try to pull back the sheet and expose how the psychology works. When you think about the mismatch: a lot of these AI hyperscalers throw a rock in the equation — if you can not raise a ton and also get to $200M-plus of ARR, that’s compelling. But in general, as you build a business, you start with an idea, get product-market fit, get a few customers and revenue, maybe you’re at four or five million of ARR growing 100% — nice business. You’ve raised a seed at, say, a $20–30M valuation. Founders and employees probably still own 60–80% of the company. Say you grow, you’re profitable, you get to $20–30M of revenue, and somebody slaps a 10x multiple and acquires you — a $300M outcome. That’s a win for everybody. Investors got their 10x, you get a ton of money, life-changing for all the employees.

Now take the same business on a similar trajectory. You get to $20M and think, “From 20 to 200, we’re going to crush it.” You burn a ton, add a bunch of people, build a whole new go-to-market to chase larger enterprises, build features to close six-figure contracts. You raise a Series A or B at a $150–250M valuation. As soon as you sign the doc that says your valuation is 250, the liquidation preference of your investors is significant enough that anything below that is very difficult to make the math work. Even in the seed case, if investors put in at 20–30 and you get bought for 100, maybe it’s not the return they wanted, but everybody makes money. If you raise at 150 or 250, do your absolute best, burn a ton, and don’t get to $100–200M of revenue — you build a nice business, 40–50M of revenue, growing 20–30% a year, profitable, good retention. Anybody who owned 100% of that business would be thrilled, it’s spitting off cash. But the odds of clearing that $150–250M valuation are basically zero. So if you sell, your investors might just get their money back and everybody else makes chump change. You could spend years building an interesting business, but because you picked up the stick of raising a high-priced round with a big firm, you’ve picked up the other end of the stick: you need a massive outcome, and your investors are never going to be incentivized once it’s clear you won’t drive a big outcome for them — you become a lower priority, less likely to get their help. And you might still work for five years and maybe break even. I have founders who tell me it’s really hard to motivate themselves — “I’m going to build for five years and maybe break even, and I’m not going to make nearly as much as I thought, nor as much as I feel I should for building a $50M-revenue business.”

CJ: It’s funny — this 200, 250, 300 range. I was talking to an M&A lawyer about the same thing last week. This is the gray zone where there’s the most consternation between a founder having a life-changing exit — walking away with 20, 30, 40 million — and a VC saying, “No, we have to keep going, you picked up that stick.”

Kyle: The biggest problem is people not appreciating the implications of the decisions they make. When you raise at a certain valuation from a certain firm with a certain check size, you’re picking up a bag of implications you have to either live up to or disappoint people over — and potentially get screwed in the process. Even if it’s not the founders who get screwed, it’s potentially the employees.

CJ: I want to dig into the incentives behind the game. How do you reconcile the sins of omission versus the sins of commission in venture? Is it worse to back the wrong company or to miss the one that becomes generational?

Kyle: That’s a fundamental dynamic. It’s better to be making investments, because you already know 90% of them are going to zero. There was some story — one of the partners at Greylock, I think — who has never lost money on a single deal, and has had some really big outcomes. But is he taking enough swings? That’s the immediate question. To some extent, if he’s making money, he’s making money — it’s working. But if you’re so protective of your swings that you miss the next Google, Facebook, Shopify, OpenAI, the loss is infinite. The loss on a sin of commission is 1x — you can’t lose more than one times your money. The loss on a sin of omission is infinite; there’s no telling how much you could have made on the winners you missed. The thesis is that no investor is good enough to get a great hit rate, so you need to be taking those risks to get the upside. This is also an element of the bifurcation between capital glomerators and cottage keepers: when you have a multi-billion-dollar fund, you have to deploy it. When you find highly capital-intensive opportunities, all the better — that’s one reason some big firms have invested in a lot of American dynamism-style industrial stuff. Those are highly capital-intensive but can still be big outcomes. When you’re a large firm, you feel even more pressure to deploy; when you’re smaller, you can be more selective. You still take risks, still assume a high loss ratio, but you can be very selective. In an efficient market where our north star is technological progress, you ask, “Why are we investing in a derivative of a derivative of a derivative, or stuff that doesn’t need to exist?” When you have so much capital looking for a home, the odds of it getting deployed into suboptimal places are much higher.

CJ: The ghost in the background here is dilution. As an operator I think about it a lot — managing the cap table, employees have incentives in it. How should operators at companies gobbling up all this capital think about it? Is there a disincentive, or do the funds not care about all the dilution?

Kyle: The math is always that in the outlier cases most of these problems go away. Raising at too high a valuation, taking on too much dilution, raising too quickly, burning too quickly — all of that gets papered over in the exceptional cases. When you’re OpenAI or Anthropic, people see these eye-watering figures — “we’re going to burn $15 billion this year and make $15–20 billion in revenue.” The volume of capital consuming all that revenue is enormous, but the fact that the business can grow so quickly, hire such great people — the fringe questions of “did we take too much dilution, did we burn too much” only matter when you’re not the exceptional case. Venture used to be that 10% drove 90% of returns; now it’s more like 1% drives 90-plus percent of returns. Truly a handful of companies drive the entire asset class. If you’re in that 1%, you don’t care about burn or dilution, because you’re going to be a crazy exceptional business. But the biggest risk is that the vast majority of companies are not going to be in that top 1%, so you leave not just a graveyard but a battlefield of wounded and dying companies who tried to be the 1% and failed — and then all those negative things come back to bite them.

CJ: It sounds like you’re saying the magnitude of outcomes still surprises us. Is it possible we still underestimate how big some of these outcomes can be?

Kyle: The biggest thing is that companies continue to be larger and larger businesses. You can write off Robinhood — their IPO doesn’t go well, they trade down — and then a year or two later it’s a $30 billion company. Is that logical? You look at Robinhood, Palantir, these companies can get massive; you look at Shopify. The biggest problem with venture is that they have a number of case studies they can point to and say, “If this can be even remotely like that, I don’t care what we pay, we have to be in it.” That gives them a lot of anchor bias. What there’s a shortage of is venture firms willing to do the work to drill in and understand the fundamental economic equation of a business and whether it can be very large in the long term. And the reason isn’t just that VCs are stupid — which is often true — it’s that when VCs have tried to do that, they’re usually wrong. Either they spend too much time thinking about it and the company’s a zero anyway, or they’re so dramatically wrong. How big can Google be? It’s a glorified internet index. Or Facebook — it’s for college kids, maybe. There was no way to see the size of the general-population market, the addiction of Instagram, the plug-in of WhatsApp, or with Google things like YouTube and GCP. You can’t underwrite that, so you might as well not try — roll the dice and see how it works.

CJ: If we think about valuation dynamics during a hype cycle — what do you make of the recent rounds for Cursor, who went one way, and Windsurf, who stepped off the track?

Kyle: The underlying question in a lot of AI is how defensible these things are long-term. The debate has gone back and forth — first “GPT wrappers are commoditized, value accrues to the model,” then an open-source model pops off and it’s “everybody’s competing on model quality, you can swap models in and out, what matters is the application value.” With AI coding tools, the most difficult question is: every business has a general sense that as it grows it adds market, product capabilities, revenue streams — it keeps growing. Think about CrowdStrike — I don’t remember the exact numbers, but directionally: when CrowdStrike went public around 2018, consensus revenue estimates for 2024, five years out, were that it would be just shy of a billion of revenue. That’s the cadence the smartest people spending all their time on it decided. In 2024 CrowdStrike had just surpassed $2 billion — off by a billion dollars. So wrong, but so what: the thing they got wrong was the slope, not the direction of the curve. We all thought it was a pretty good business that could add more products. The problem with this AI stuff is they may not have those sustainable curves. It may not just keep going up. You look at OpenAI and Anthropic — they’ve scaled to very large scale very quickly, so we think every business can do that. The problem is I have yet to see a ton of concrete examples of actual business value creation. Everybody experiments with it; a lot of this revenue is people saying “the promise of AI is very compelling.” One of the great quotes I heard recently is Palmer Luckey on artificial superintelligence — the biggest obstacle to ASI or AGI adoption is not which company builds it, it’s going to be human inertia. Look at who’s made the most revenue from AI — right near the top is Accenture, helping people figure out what the crap they should do with this thing. There’s not a ton of examples of the rubber hitting the road. There’s a ton of “we need more compute” — Amazon, Google, Meta pouring 80, 90, 100 billion into data centers; the chip companies popping off; the model companies raising tons to spin out better models; applications trying to do all this stuff; and then people trying to figure out how to use it effectively. We’re in that massive scale-up and distribution curve, but what we have yet to see are concrete multi-billion-dollar examples of “this company used AI, trimmed these costs, saved hundreds of millions, or drove billions in revenue.” It’s still picks and shovels — the chips, the data infrastructure, the models, the applications — for people trying to get a job done: the lawyers, architects, coders. There’s a lot of “it’s driven up productivity among our engineering team,” and that’s great, but is it going to drive a trillion dollars of productivity gains at the rubber-hits-the-road level? Thus far the curve of hype is massive, the curve of picks and shovels is sizable, and the curve of business value creation — I have yet to see discernible evidence that it’s materially following suit.

CJ: Behind all the hype, what scares me — and you’ve written about this — is there has to be a belief that someone else down the road will think this is worth more than you thought. You’re always doing the back-of-the-envelope math on the greater fool theory: who’s going to think this is bigger, because when you step into an investment you have to think about who’s going to buy it from you.

Kyle: One of my favorite pieces is the institutionalized belief in the greater fool. It reminds me of an idea — I think it’s a Buffett quote — that when we think about an efficient market, you’re not trying to guess what the market’s going to do, you’re trying to guess what other people guess the market is going to do. You’re doing this second-layer psychological training. The problem is our psychology is all over the place. Every time you see a meme stock — and “meme stock” is hard to pin down because every stock can be a meme stock at different times — Sydney Sweeney does an American Eagle commercial and the stock pops 18% in one day. Or it’s negative — the Bud Light situation, the stock plummets. Psychology drives markets. Kyla Scanlon is probably the person who’s written some of the best stuff on this — the idea that narrative, story, and vibes have become tangible economic drivers, which didn’t used to be the case. Not because we haven’t always been stupid and excited about stories, but because we weren’t nearly as good at telling stories, getting them to everybody, and letting everybody react financially. Now with the internet and social media, we can get stories out very quickly and everybody has an app on their phone that can move hundreds of billions, if not trillions, based on that story. That’s never happened, and we don’t know how to deal with it. To bring it back to Cursor and Windsurf: Cursor believed they could raise at, call it, $9 billion because someone out there thinks they’ll be worth $28–30 billion — that’s what the animal spirits are saying. With Windsurf, they exit for around $2.5 billion, which means the narrative of who the next greater fool would be — there isn’t somebody out there right now who thinks it’s going to be worth seven.

Kyle: There’s this Peter Thiel-ism where he says fundamentally any competitive round is an underpriced round. When Stripe goes out and raises at $90 billion — if you asked every single person in the world with money to deploy whether they’d invest at higher than $90 billion, the answer is yes; there’s probably at least $100 million that would have invested at a higher valuation. It’s literally a supply-and-demand curve with an unoptimized amount of supply, because there’s more than the executed volume of demand. So why wouldn’t Stripe just raise at the absolute highest, with fewer investors? A lot of that depends on the type of business you’re building and the investor base you want. On Cursor versus Windsurf: Cursor is seeing their business grow at a pace where the underlying assumption in raising versus selling is “we can keep growing, we want to own more of this business over time, all the signs are pointing to this going great, let’s keep raising.” The reason Windsurf would sell — assuming a logical process — is they’re seeing something in their business making them more worried about their ability to keep growing. Maybe “our gross margins are so low because paying for models and inference is really difficult, and that only gets worse as we scale. But if we sell to an OpenAI or a Google who have access to the models or the compute, that improves the equation, and we can be a better business with them than standalone.” You have to assume that’s the math. But for some founders it’s just that they’ve got that dog — “I’m going to build the most important company of this generation, come hell or high water.” There’s an element of Elon Musk-style aspiration — his willingness to drive in, tell a compelling story, raise a ton of money, be super competitive, push through obstacles. Buyback Capital said something great: of all the things Elon Musk has built, the most impressive might be the investor base he’s built for Tesla, because the kinds of investments he can make and the business performance he can put up without material changes in the stock — that’s an investor base any public-company CEO would kill for. He’s almost inoculated his investor base to be comfortable with massive swings, because he balances it — “this robotaxi thing could be huge, the humanoid-robotics thing could be the largest industry we’ve ever had.” They keep getting promised things they get excited about, so they’ll accept the day-to-day volatility. There’s something to be said about people like the Cursor CEO who are telling themselves and their investors that story, whether or not they all truly believe it — clearly they’re making decisions based on that belief.

CJ: We’ve touched on the psychology behind portfolio construction and taking on money. Something we bonded over about two years back was the unbundling of venture capital — because if you distill it, VCs are selling a commodity: cash. You can get cash from a lot of places. Can you touch on how VC firms have evolved, and the different services they offer to differentiate?

Kyle: There’s a piece I’ve wanted to write for a long time and put off. When I first started writing, it was because I had two pieces in my head: the unbundling of venture capital and the productization of venture capital. The productization piece was about “what is the product we’re offering.” The piece I want to write now — a couple years of hindsight later — the working title is “building a product-led venture firm.” Not enough people ask: what do founders hire VCs for? What’s the job to be done? As a founder I have a job to be done, so I hire a product that satisfies it.

CJ: Help me hire people, help me get customers.

Kyle: It’s a multifaceted question with a lot of answers, and how do you prioritize? What’s the most important thing you pay for? Marc Andreessen had an AMA where he said, “I’ve always thought the thing founders are hiring VCs for is borrowed credibility.” If you were an engineer at Databricks and now you’re starting a company — who are you? Maybe a little brand rub-off, but are you somebody customers, vendors, and partners fundamentally trust? Probably not in huge volumes. But by raising from a venture firm others respect, when somebody sees “they’re backed by so-and-so,” as a hirer, partner, or customer it’s de-risked for me. So the first thing he puts front and center is borrowed credibility. The other argument comes from Vinod Khosla, who says the vast majority of VCs provide zero value and an unfortunate portion drive negative value — it’s actually worse to have them along for the ride. The way I think about it is that quote: “I know that 50% of my marketing budget is wasted, I just don’t know which half.” Same thing — 90% of VCs are probably neutral-to-negative value, we just don’t know which 10% aren’t. And maybe it’s not always the same 10%, because there are firms that have grown up over the last few years that you wouldn’t have put in that 10% ten years ago but would now. So every venture firm is fighting to articulate a product good enough to put them in that 10% that can actually drive meaningful value. It’s not that everybody should have the same product — it’s that everybody needs a well-articulated product. Why should I hire you versus them?

CJ: I’ve always thought it was funny that Tiger’s product is essentially “we’ll leave you alone.”

Kyle: Tiger’s product in 2021 was “we’ll leave you alone.” Tiger played a thesis and played it poorly, and I think Andreessen is playing Tiger’s game much smarter. Tiger was the velocity and volume — the size of the round, the speed they’d move, the valuation they could offer, and then leave you alone. There’s a fundamental truth underlying that, but the truth was not “free capital, just take money and go away.” Maybe some founders were like, “I’m sick of it, I’ll take that,” but I think most regret it now, because you take it and then, “Bummer, I wish I had a board to help me figure this stuff out.” The underlying first principle that is true is that people want somebody in their corner to throw around their weight — to move fast or be really loud. The ability to throw your weight around is valuable. The danger is if that weight gets thrown against you, it’s very unfortunate and dangerous. When people hire a firm like a16z or General Catalyst, if you peel back the onion, the first layer — this is what I wrote in the unbundling piece — I maybe leaned too heavily into the idea that individuals would become more powerful, that there’d be a lot of Eli Gils. There haven’t been nearly as many as I thought. But the unbundling has a principle of truth: people care way more about the individual partner. It’s not just “I can take a16z’s money” — if I take it from so-and-so, that’s good; from someone else, not as good. There’s relative value in the individual partners. So there’s the personal relationship first, but then you’re also picking up a firm’s style. You could take it from a firm with a16z or General Catalyst-style weight, or from a firm that acts more like Benchmark — even Accel, which is not a small firm, acts a lot more like Benchmark than like a16z. You’re picking up a style. You’re not going to get loudness from Benchmark. It’s up to you to decide what’s most advantageous for your business.

CJ: But there’s a Kyle brand and a Contrary brand you get when someone takes money from you — they rhyme but aren’t identical.

Kyle: That’s the unbundling — it’s a looser unbundling than I thought. I thought people would feel so good about their brand at a16z or Lightspeed that they’d leave and start their own firm as an Eli Gil. You’ve seen a handful leave and start firms, but it took the shape of how other venture firms get built — they’re not spinning out to be solo capitalists, they’re spinning out to be a traditional venture firm that’s younger, scrappier, less well-known. When I was at Index, four people who were there with me left and started their own firms — folks like Mark Goldberg, who started Chemistry. That style is much more common; very few go and become solo Elad Gil-style capitalists. The other thing founders don’t think enough about: raising from a16z can be good or bad depending on who drives the investment. Sometimes we have conversations where a founder says, “I’m talking to so-and-so at big firm XYZ,” and we say, “We’ve heard some not-great things about that partner.” Or they name someone else and we say, “They’re great, super sharp, and they can go to bat for you internally at such a big firm” — that’s important. If you’re getting a very junior person, or someone with a not-great reputation, there’s a fundamental difference. But regardless of the partner, there are still implications inherent in raising from that firm that wouldn’t be true raising from a different type of firm.

CJ: So you get the power dynamic — can they throw their weight around, get you in the room, knock down political or governance barriers — and the element of the partner themselves. There’s another box you’ve touched on: one benefit of taking money from Insight, if you’re a cybersecurity company, is they have 450 portfolio companies you can hawk your wares into. What are some of the other benefits founders or CFOs should stack-rank when raising?

Kyle: Every firm is different and every company is different. When you’re a first-time founder who’s never hired anybody, tapping into a talent network to benchmark what good looks like can be valuable. If you’re a third-time founder who’s going to hire a lot of the people who’ve come with you, that’s way less valuable. There’s not enough conversation around the variability — different founders in different spaces benefit from different things. Another example: there are companies selling into government increasingly — defense, aerospace, industrial — with a lot more government work than a traditional SaaS company. Some firms have built experience to cater to those people. Founders Fund is a great firm to raise from if you’re selling into defense and government. You also have folks like Construct Capital in DC — if I remember correctly, one of the founders was Uber’s general counsel for years and navigated a bunch of complex government-regulation questions, so Construct is great if you want access to that expertise. Different people get built up around different things, so it depends on what you as a founder feel is most valuable. That’s a benefit of raising multiple rounds — you put together an investor base: “I get this from so-and-so early on, then over time I need more of that, and in later rounds I need folks who can just write big checks.” That’s why you see people raise from firms like Wellington or T. Rowe Price — massive asset managers who can write sizable checks even quicker than a16z, and can double down if you IPO.

CJ: Gearing toward a close. You made a spicy point that if we’d just left Sam Bankman-Fried alone, a few of those deals might have paid off that whole FTX crater. Is there a truth buried in there?

Kyle: My undergraduate degree is in accounting, and I remember learning about the fraud triangle — means, motive, and opportunity. If there’s a financial opportunity and there aren’t controls in place, and there’s motive. The obstacle is that everybody committing fraud thinks they can get away with it, or cover it up: “I need this money, I’ll take it but I’ll make up for it by selling more next month.” You try to paper that loss over, and it gets bigger and bigger, and that’s where you get Ponzi schemes or Enron cover-ups. It’s very dangerous when you start justifying immoral or illegal behavior upfront with the potential for future coverage. What’s painful in this specific instance is that they brought in the Enron guy to FTX, and to cover the hole they sold positions — I’ve never looked deeply into the details, but I think SBF had both a seed investment in Anthropic and maybe a seed investment in Cursor, and they sold both years ago. If they’d held those, they wouldn’t have — so that’s painful. But the takeaway for me is not “you should have just let the guy cook and it would’ve been fine,” because he shouldn’t have been doing illegal things, covering up, robbing Peter to pay Paul. It’s the laxness of financial institutions — that’s what FTX was, a financial institution — letting people playing League of Legends with one hand and taking money out of users’ bank accounts with the other, without telling them. When you decide to act immorally or tell stories out of turn, you lose the opportunity to compound. That’s the takeaway. It’s a Buffett-Munger-ism — the market can stay irrational longer than you can stay liquid. You will lose and you will win; the only thing you can’t do is get knocked out of the game, because as soon as you’re out, compounding ceases to work. SBF knocked himself out of the game — he flew too close to the sun. If he hadn’t, even if FTX had to see massive drawdowns and big problems, by doing things that were impossible to fix he took himself out, and then compounding ceases. We’re working on a piece for Contrary Research right now looking at a bunch of neobanks — going into the post-crypto-winter, post-fintech-winter era, a lot of fintechs got left for dead. But there are exceptional neobanks built over that fintech winter that haven’t gotten a lot of attention — good enough businesses that they compounded and grew and became the default in certain countries. They just needed time to compound, sometimes through winters. Same thing — the Coinbase guys talk about how when the 2017 crypto winter hit, their reaction was “thank goodness, we need this washout, there’s so much hype and hucksterism we need to wash it out,” and they had relief because they’d built their business to weather the storm and keep compounding. When they went public and the stock — if I remember correctly, that’s a 10-bagger in the public markets; their low was $9 billion and they’re now north of $100 billion. That’s building a business that can withstand volatility in a way the majority of businesses in crypto have not. A lot of AI companies are not building themselves to withstand volatility either. If the AI tide turned right now — “just kidding, this isn’t proving out, we’re turning off the infrastructure investments, there won’t be more data centers, your margins go down because inference gets more expensive” — there’s a plethora of AI companies that would get wiped out tomorrow if their gross margins dropped by 20%. They die. Those are companies not building themselves to give themselves the advantage of compounding through volatility.

CJ: To put a bow on it: fast-forward 20 years — what’s one thing about today’s venture landscape we’ll either laugh about or wish we’d paid more attention to?

Kyle: The thing likely to become more supercharged, not less: people assume we’re in a period of heightened hype. I don’t believe that, because I’ve been through this three times — I’m not a weathered gray-haired veteran, I’ve been doing this 11 years — 2017, 2021, 2023, 2024. What’s stayed consistent is that the value of a narrative is only increasing. It’s never once gone down. Sometimes the story gets harder to tell when the stock or token plummets or customers leave, but the value of a good story is going up. What we’re dramatically underappreciating is that we are fundamentally a society built around storytelling and shared narratives. We assume reality is reality and these fuzzy stories people tell each other are fugazi. But I used to think reality would eventually win out over narrative every time — the Buffett line, the market is a voting machine in the short run but eventually a weighing machine. The problem is we assume narrative has no tangible weight — but it does. When it gets to that weighing machine, the story is a tangible part of weighing the company. So increasingly the hype, the narrative, the storytelling, the apparent distance from reality — that’s only going to become a bigger part of the story, not a blip in time where we all lost our minds.

CJ: Can’t think of a better way to end it. Kyle, thanks so much for being generous with your time today.

Kyle: Super fun to jam. Thanks for having me.

Connections