为什么现在正是应用层的时刻｜为什么创业公司应该 TokenMaxxing｜Mike Mignano，USV

1. Fear of Failure vs Thrill of Winning

Ready to go, Mike. Dude, I am so looking forward to this. We first did our show when you were co-founder of Anchor.

Correct. 10 years ago.

10 years ago. I looked it up. I was underage, and I remember Anchor was the hottest thing in New York and in startups at the time. I was legit 18 or 19, and I was absolutely shitting myself before our show because you were the—it-boy entrepreneur at the time.

For a minute. Like, 5 minutes, most.

Well, now you get 20 seconds, so I would appreciate it. Dude, I want to start with something that I ask a lot of great founders, but given your founding roots, I want to ask it of you: Are you motivated more by the fear of failure or by the thrill of winning?

I think both, to some extent. I think the fear of failure is very, very important. I strongly believe that great things, great companies, great products, and great people come from constraints, and I think that failure is the ultimate form of a constraint.

You talked about Anchor. Somebody just asked me the other day, “At what point in your life did you do your best work?” I thought about it, and the honest answer to that question was that I believe I did my best work when we were 3 months out of cash. When we were about to run out of money and when we were about to fail is when we did our best work. It’s so clarifying to know that you’re going to fail unless you turn the thing around.

I think failure can be an incredible motivator, but I also think that you have to have an insanely ambitious mission that you’re trying to accomplish, right? You want to win. You have this hunger to achieve something really hard and really impossible-sounding. I actually think you need to hold both things in your head to be successful, and I try to do that.

We were talking before about insights that we had years ago. With Anchor, you wanted a very natural, conversational style—record and let people listen in. I was always like, “Nope, we want manufactured.”

You were doing clips way before clips.

We were doing clips. I was absolutely into the airbrushing of audio, as we do today, and we disagreed on that. What are your thoughts on that?

Wow, there’s so much content, and the tools have gotten so much better that, out of the box, you can have a really decently sounding or looking podcast with almost no effort. If you want to break out, you actually have to go above that. The baseline has to be excellence. The baseline has to be incredible editing. The baseline has to be a phrase that you used—I remember 10 years ago—“insights per minute.”

I’m now a huge fan of insanely high production value when it comes to podcasts and YouTube videos. We were talking a little bit before, and you asked me about some of the content I’ve been producing. We’ve tried to do something very different, where we break out of a studio and bring in multiple cameras and multiple mics, and we try to do something very, very high production value because I think that’s the only way to stand out. Anyway, I think you were right.

It took me 10 years to agree with you.

Listen, I was totally fishing for that. I was right.

I do think insights per minute is important. I also think time to value is very important. People ask the most stupid fucking question, which is, “Tell me, where did you grow up?” Very little is often learned in the “Where did you grow up?” question.

What’s your version of that question?

I want to go straight to, “Are you motivated more by fear?” Straight away, I have someone on a hook. You have to create hooks in content pretty early.

I also think, very much like venture, you have to be either really big or really small in content. You have to be very natural, like a teenager putting their phone up on the side and hitting record, or you are Logan Paul or MrBeast, or you’re like us at 20VC, which has studios all over London. It’s much bigger. The mid-tier firm doing it from their conference room office doesn’t work, baby.

Yeah, you’ve got to be on the extremes. I totally agree with you.

2. Why Mike Left Lightspeed for USV

Speaking of being on the extremes and constraints, Lightspeed is a cash-constrained world to be in. It’s tough over at Lightspeed, and you made the move to the extremely well-funded USV. I’m obviously joking. We’re going from the very large to the very small, with USV being constrained in comparison. Why did you decide to make that move, dude?

I decided to go to USV because I am from New York. I’ve always been in or around New York, and as a builder, when I was building Anchor in New York 10 years ago, I got to meet USV. Despite them passing, I had an incredible experience and started a very, very long friendship.

I ultimately became an LP in the funds, and I always appreciated their ability and willingness to be really opinionated and to be very, very thesis-driven, even at times when the market said it made sense to be very consensus-driven. By the way, there have been times in the market, including recently, when being consensus-driven has been extremely valuable and extremely lucrative as a fund.

But the way I like to work, the way I like to build, and the way I like to bet is by being thesis-driven. I recognize that being thesis-driven can also be super risky, right? If you bet on a thesis and you’re wrong, you miss and you fail.

Well, the trouble is that being thesis-driven can also be conventional. What I’m actually worried about is: What if the winners of venture—and what if venture is, from this point out, consensus-driven?

I think we’re coming out of a time in the market where we’ve undergone a massive infrastructure build-out in AI. We’ve had these companies building brand-new technologies, which have required vast amounts of capital to build out the infrastructure. Obviously, I’m talking about the labs. I’m talking about OpenAI, Anthropic, xAI, Scale AI, and so on.

Those required a lot of capital, and they produced magical technology that has generated billions. We’re about to see trillions of dollars in value. But now that infrastructure is built, it reminds me a lot of the early days of the internet, when we built out fiber and broadband. Then we had these new technologies to play with, and the internet came along, and an application layer came along and took advantage of that new technology.

3. We're Past the AI Infrastructure Phase — Now It's About Applications

To me, it feels like we’re back in that mode right now, where we’ve got this infrastructure and now it’s time for the applications to be built. I think there are going to be so many applications and so much software that you’re not really going to be able to make a bet unless you know what you’re looking for. I think that’s what USV has always been great at: knowing what they’re looking for and placing the bet when they see it.

Brendan at Mercor said the other day, “We will see much greater value accrue in the next 24 months at the infrastructure layer than we will at the application layer.” Agree or disagree?

I think there is still value to be created and accrued in the infrastructure layer. We’re not done. The build-out isn’t done. But I also think that now that we’ve got enough new toys at the technology layer, we will see massive value creation in the application layer as well.

Do you think the future of AI is always-on? Sam Altman is suggesting so more and more in his verbiage. God, it makes me excited for NVIDIA. The future of AI is always-on. Do you agree with that?

I do think that we’re entering a world in which context is increasingly valuable for the labs and also for the application layer. I think we will see products and businesses increasingly try to push us further and further along the edge to an always-on world.

I can remember 10 years ago when I got an Alexa in my home and thought to myself, “Wait a minute, this is always listening for the wake word.” Now, I think, to your point, we’ve moved past that. Let’s say you’re in a meeting, in a conference room. More often than not, I’m hearing somebody say, “Hey, I’m Granola-ing this.” I think we are moving more and more toward that world, for better or worse.

Do you think the multistage funds deliver an amazing pre-seed and seed product? Because I—I say this, and I think you’re probably also, and there’s no problem with this, but I think you’re a little more protective of your brand than me. I don’t worry about upsetting people.

4. Is AI "Always On" the Future?

The worst place to be is a $50 million to $100 million seed fund. You will get crushed by the large funds that are able to write large checks and lead rounds, but you’re also not small enough to be collaborative.

Very difficult middle ground to be in. Do you agree?

I think the best funds in the world at doing seed are really, really good about building their networks and meeting amazing people. I think you've been doing a good job recently at empowering a new class of entrepreneurs here in Europe and in London with very, very early-stage seed checks.

So, I think you have to build a great network. I also think to do seed well, you have to put your ideas out there into the world. You have to show the market what you're looking for. You have to be willing to take a stand and take a position and say, “I just put out a post yesterday about the Rebel Alliance,” saying, “Hey, USV, we're really interested in the Rebel Alliance: this new universe of open-weight models, open-source harnesses, distributed compute, and agents—human-aligned agents.”

When you put things out into the world, you're sending up a bat signal that's telling these early-stage founders that maybe haven't gone out into the world yet with a product, “Oh, this is the VC, or this is the person, or this is the team I should be talking to.” So, I think you need a great network, you need to ship your ideas, and you need to be willing to take bets on people, which, again, I think you've done a good job of with your program.

Can I ask you on the shipping ideas? This is why I love asking advice, Mike: what if the ideas aren't fully formed? What if I'm just wrong, and then I get a deluge of companies, and actually the thesis wasn't what I thought it would be?

I think that's totally okay. Much like building a startup or building a product, if you're investing, you also have to be willing to put yourself out there, take a risk, and place a bet.

You can hedge that bet a little bit. You can place a counterposition bet on something else, maybe to offset the risk of that. But I think you have to be willing to put yourself out there if you want to succeed at anything in technology, whether it's a startup or venture capital.

5. The Rebel Alliance: USV's Thesis on Open Weights & Human-Aligned Agents

Open weights, open harnesses, open models. Ding, ding, ding, ding. We're going to do a prediction round. Fun. I told you we had this truth-or-donate round. This is a new prediction round.

The model landscape in 5 years' time: what does that more mature market look like in terms of composition, the players, and their weightings? What does that look like?

One future is that the labs, especially the ones that are starting to approach something that feels like superintelligence, once they reach it—if they reach it—will have some form of recursive self-improvement, such that we hit this exponential growth of intelligence.

The thinking goes that whoever gets there first will run away with it. Once you reach recursive self-improvement, you're gone. No one can catch up. I think that's not necessarily the future that I am hoping for.

For everyone who doesn't know, what is recursive self-improvement?

This is self-improving AI. This is superintelligent AI that can do its own AI research. Once you have it, once you deploy it, it can just continually improve itself exponentially forever, until it plateaus or until it self-limits in some way.

By the way, we don't know what that self-limiting factor may be. It may be that, for some reason, these architectures—the Transformer architecture, let's say—don't scale past a certain point. We don't know if that's going to happen. It could be that the model needs some amount of data that it can no longer access. There's a limit to the amount of data.

It may mean that we can't keep up with the hardware infrastructure that we need to keep recursively self-improving. In the past, in the history of new technology adoption, if you study it, these technologies all follow S-curves. They have a slow takeoff, then they have what feels like an exponential ramp-up, and then they plateau for some reason, like we just discussed.

That's potentially the second future for AI: we reach some plateau in terms of intelligence. If that happens, then the technology might start to look somewhat more like a commodity. If it plateaus, we're at a point where all the other labs can catch up and they can all have the same technology.

If they all have the same technology, then we're going to have a lot of competition in terms of price, product experience, and the various components that make up the new intelligence stack.

6. What Happens If We Hit Recursive Self-Improvement?

If we go down route one—recursive self-improvement, leading to this kind of exponential advancement—is that done by OpenAI or Anthropic, or a leading model provider today? Is that done by one of the many others going after it? I was literally for 2 hours in Shoreditch yesterday with one. Or is that done by an entirely new company focused on recursive self-improvement?

I think it's probably done by one of the labs that already has all of these advantages, these vast compute advantages. They already have models that are at the frontier, they have the infrastructure, and they have the chips. So, I think it's probably done by one of them.

That said, there are a bunch of amazing researchers working on incredible new architectures that maybe haven't broken through yet, but maybe they can. Maybe they're building something that is better than the Transformer. There are a number of these labs out there: Thinking Machines and Safe Superintelligence from Ilia and Daniel.

If I were to bet, I would say the labs that already have the advantages. But that doesn't mean there isn't a new architecture being worked on in a lab somewhere that could leapfrog all of it.

Okay. Number two is that, actually, we have less of this kind of exponential growth and more of a linear style. When you look at that more mature market on an S-curve basis, what does that composition look like in terms of the market between open and closed, Anthropic, and OpenAI?

7. The S-Curve Future: AI Plateaus & the Market Commoditises

I think in that world, enterprises and individuals are probably optimizing for a couple of things. They're probably optimizing for cost. They're starting to think about the trade-offs between intelligence and spend, and how they can optimize their token spend.

What does that mean? It probably means leveraging things like open-weight and open-source models. It probably means leaning into things like the routing layer to optimize token usage toward the model that maybe gives you the most bang for your buck, rather than always just token-maxxing and picking the most powerful model.

I think the other thing it looks like is choosing tools, products, and—hate to use the new buzzword of the past 6 weeks—harnesses that are maybe a little more human-aligned.

Can you help me? What the fuck is a harness? No, seriously. I feel so bad. On X, I really want to do “Grok, what is a harness?” but I know everyone will be like, “Oh—”

I think of the harness as the application that tightly, tightly couples with the model. In the case of Claude, it's the Claude desktop app, and I would say specifically Claude Code and Claude Co-work, which are having this loop, this flywheel of product-and-model engagement.

Some other harnesses that I think people are pretty excited about are things like Hermes. I consider Hermes a harness that takes over your Mac mini. I don't know if you've set it up yet. Pi from Arendelle is another great harness. It's a European company.

Someone legitimately called a harness after one of the greatest fashion brands.

What's that?

Hermès is a fashion brand.

Oh, right. Well, isn't that Hermes?

Well, it's spelled the same, dude.

Good point. So, anyway, I think having products, agents, and harnesses that are human-aligned and aligned with your own goals and incentives is going to be really, really important.

8. Who Is Your Agent Actually Working For?

I think people are going to be a little more self-conscious about the incentives of the model they're using. Obviously, the incentives of the models that come out of the big labs are to make the labs' models smarter, better, and faster.

You might think to yourself, “Hey, if I'm outsourcing all of my agency and all of my personal information, these goals, my credit cards, and all these things to an agent, I want that thing to work for me. I want that thing to be completely aligned with my incentives.”

I published something a few weeks ago on X called “Who Is Your Agent Working For?” That was the big idea we were trying to get across with that. As you become more and more comfortable handing over the keys to an agent, you're going to want to know that it's working for you.

So, I think that's a little bit of what that second future looks like. I have to say, I think people will get incredibly comfortable handing over the keys in the same way that we got incredibly comfortable putting our credit cards online, using Apple Pay, and finding our husband or wife online. I think that will be de facto in 5 years.

I think there's a chance you're right. In fact, I don't think anyone gives a fuck about privacy other than hardcore technologists. I know that's blunt. I'm going to get killed for that.

No, no, no. You're not. I think there's a good chance you're right. If we're just thinking in terms of bets and probabilities, the history of the internet and personal computing suggests that you are right, Harry.

But two things. First, I think we have never handed over so much of ourselves to a technology as we're about to do with agents. Yes, you're right, we browsed the web through Chrome and we used—

But it didn't do shit for us.

It didn't do shit for us, right? We still got to control it. Yes, maybe we let these companies learn about our interests and the things we care about, but it wasn't out buying stuff for us. It wasn't out sending very personal messages to family members and loved ones.

It wasn't a second self, right? It wasn't us. I think now we're going to be doing stuff like that, and it might just make us think a little bit more about the incentives of the models that are doing those things for us. That's number 1.

Number 2, I'm not sure that every model and every agent needs to be superhuman-aligned. I'm not sure every person needs to care about it. I just think enough people have to care about it such that 1 player or a few players keep the others in check, right? I think we need market forces where a couple of good actors keep the other actors in check. So I think it's possible.

Now, is it going to happen? I don't know. History says it won't, but I think it's possible.

9. Engineering Teams Are Getting Smaller

Do you think Dario has had a massive own goal in terms of marketing and messaging? He's been saying that about labor displacement for years. Everyone's going to be unemployed, and telling your customers they're going to be unemployed is always a bold statement.

What I see in Anthropic is an extremely mission-driven company. I believe that mission-driven companies can be very, very successful. Anthropic, love them or hate them, there's no doubt they've been extremely successful. I think at the time of us recording this, they're the most valuable privately held company in the world.

I think core to whether they will continue the exponential revenue growth that they've had is the percentage of spend, in terms of developer salaries, on tokens. Mark Benoff said that he spent $300 million on Anthropic, specifically for the dev team, which equates to 3.8% of dev salaries spent on tokens.

Now, if that goes to 20%, Anthropic's grossly undervalued. That exponential revenue will continue, and well done to everyone who's invested and all their employees. Great. If it goes to 100%, holy shit, this is just so much bigger than we ever thought it could be. But if it stays, or if we see migration to open models, it's a very different game.

I think it's a risk. I think there are 2 schools of thought playing out right now with regard to token spend. One school of thought, which we're seeing and which probably applies more to the incumbents and the big companies—the Salesforces of the world rather than the startups—is that if every employee is just spending like crazy on tokens, the fundamentals of these businesses are going to be in trouble.

They're just too big, right? There are too many employees at a Salesforce or, I don't know, a Microsoft for every employee to have an unlimited token spend budget. So I think we're going to see a lot of those companies start to constrain their spend, and we've seen mentions of things like these.

You see Meta, Uber, and Microsoft. Yeah.

I think if you're a startup, the calculation probably looks a little bit different. Your organization is smaller, right? So you're not going to have a situation where you have 5,000, 10,000, or 50,000 employees just spending out of control. You can more tightly control the spend.

But also, I think as a startup, you need every advantage you can get right now. So if I were the CEO of a startup right now, I would still be pounding the table to maximize token spend on the right things, definitely with coding. Maybe not with other things. For simple tasks like summarization or operations-type work, I'd probably be leveraging Claude Sonnet or something. But if I'm at a startup and I'm coding, I want to be using the frontier. I want every advantage I can get against Salesforce.

So if you are the best dev, are you going to go to a big incumbent that's going to give you a budget and really constrain your abilities in terms of model usage, or are you going to go to a startup where, fundamentally, they say, “Hey, it's a free-for-all. Be your best self”?

I think you're going to go to a startup, especially the startups right now that are mission-driven, which, by the way, I don't know that we've really been in that mode for the past couple of years. Yes, Anthropic's been mission-driven, but there's so much money and capital sloshing around the Valley in this ecosystem. I think a lot of companies haven't needed to be mission-driven. I think now they will, and it's going to work to their favor.

Do you not think we enter a world now where Fable will come back in some rejuvenated form, one way or another, where FAANG comes back and newer and better models come back, and you just have the 100x engineer? You pay for that token spend, which is enormous, with the replacement of 10 other mid-level engineers.

I do think engineering organizations are going to evolve. I don't know exactly what it looks like yet, but I think you will increasingly have somewhat smaller teams of higher-caliber and higher-quality engineers, because I think a lot of the lower-level tasks increasingly can be delegated away to agents.

In terms of the open ecosystem, when we look at it today, what percentage of enterprise workflows do you think can be done with open models?

10. 80% of Enterprise Tasks Don't Need Frontier Models

80% of non-coding tasks in the enterprise can be done with models that are not at the frontier. I think if you're coding, you probably want to be leveraging frontier models. But things like summarization or the generation of docs and briefs, I don't think you need to be at the frontier. You probably can leverage open-source models.

I think a lot of the open-source models we're seeing are catching up faster than they were previously. So I don't know that you need to be at the frontier if you're not coding.

China has the best open-source models. It's unbelievable to see the rate and the evolution of their open-source ecosystem. Is that concerning, do you think?

11. China's Open Source Lead

I think startups and teams go where the incentives are. As we shift to a world in which this rebel alliance might actually have a fighting chance, we're going to see a lot of smart teams going more and more toward open.

12. Is There a $50B Company to Be Built in the Routing Layer?

I'm really intrigued that you said routing layer. Do you fundamentally think there will be $10 billion to $50 billion companies built in the routing layer, or do you think, candidly, your inference providers like Fireworks or Baseten eat into that? Or, candidly, if you want to go a step above that, you've got Nebius, which sits on top.

When I had the founder of Nebius on, he was like—you need to eat all of the stack. I'm not going to put on his accent because I'll sound like a Bond villain, then he'll fucking kill me.

I do think routing is interesting and important right now, right? Again, as enterprises are trying to optimize their token spend, they want to make sure that the model they're using is the right model for the job—not only in terms of its capability and what it can get done, but its cost.

So you're going to have companies that are singularly focused on this. Companies like Open Router out of New York are doing some really interesting work. You're also going to have companies that are leveraging other parts of the stack for their products, building their own routing layer as an opportunity to monetize.

We've talked to a number of companies building at various different levels of the stack that, when you talk to them about their business model, talk about building a routing layer and monetizing it. So I think it's a really interesting area that has a lot of opportunity, and a lot of different companies are going to go after it because they see the potential.

How do you monetize a routing layer without becoming just a commodity pipe?

It's a good question. A lot of companies right now are thinking about charging a small margin on top of it. To your point, that might not be a great business model.

I put out a piece on X about this a few weeks ago, and a number of people chimed in with some of their ideas. One interesting idea I heard was this notion of almost a bounty model at the routing layer, where the routing layer gets rewarded for choosing the right model. If you choose the most efficient model or the best model for a certain task, that's when they collect a fee. I thought that was kind of an interesting idea. I haven't necessarily seen it built out yet, but I thought it was a pretty interesting idea.

I think it's hard to see that $50 billion company built in routing alone, I have to say.

Maybe, on the other hand, we've seen time and time again in the enterprise—in infrastructure specifically—that these infrastructure companies become deeply embedded in developer workflows and in developer ecosystems, and you just can't rip them out because they become the gold standard. Maybe somebody will build that routing layer that every developer adopts and never wants to remove.

Would you say it's unfair of me to say that USV missed the model game?

I think USV was playing a different game, and that's actually been the history of USV.

Is the game not unbelievable returns in generational-defining companies?

So USV, again, I think has never been afraid to take a position. USV maybe did not invest in any of the big model-layer companies.

What USV has now been doing for a number of years—and I'm a big believer that we haven't seen the full circle of this narrative play out—is betting on energy. They've been betting on energy since 2021.

13. USV's Long Bet on Energy Since 2021

The idea has been for a long time now that, no matter what model wins, if you believe in AI and if you believe in intelligence, there's going to be an energy layer underneath that is going to be needed to power these things. Over the years since 2001, we've learned that we need even more energy than we thought, and we need more portability of energy than we thought.

Energy is a big theme at USV, starting in 2021, and it still is to this day in 2026. Did I say 2001? I meant 2021.

How do you think about investing in energy given the capex-intensive nature of it? We're invested in a company called Fuse Energy, which I think is incredible. I'm so happy, but it's a capex-intensive business in terms of the energy space.

It is. But I think there's also a lot of innovation happening, and where there's innovation, you can find early teams that are doing science experiments before anyone else is thinking about them. USV, a few years ago, invested in a great company called Radiant, which is building small nuclear reactors that literally come off a factory line. It's going to be one of the first companies in the world to test in the DOME in the United States for nuclear energy.

I think there are a lot of interesting models and ideas happening on the edge of innovation. Those are the best places to bet as a venture capitalist because, in the earliest days, they're actually not that capital-intensive.

You know, one story I love is Boom Supersonic.

Crazy.

Yeah. Don't get me wrong, we all love Concorde, and I love flying too, dude. Yeah, game on. But there's this massive business in turbines for AI.

Yeah. What a wonderful world.

We have a bet in a company called Rune. Amazing company. Have you heard of Rune?

No.

They're building micro data centers that sit directly next to generators, wind farms, and things like that to solve the portability issue. How do you get the energy as close as possible, as quickly as possible, to the compute and the data centers?

I think we're on the lookout for things like this that are innovating the model of energy and the portability of energy.

Isn't this just the joy of capitalism? I met Panthalassa, this company that does data centers at sea.

And it's just fundamentally Adam Smith's invisible hand. The market solves itself.

Yeah. We've got an energy crisis, and then you get Rune and Panthalassa and Elon building data centers in space.

It's crazy.

It's awesome. It's a cool time. It's a fun time to be investing.

It is. This is why I get really pissed off with a lot of European entrepreneurs who are still building SMB accounting solutions. I think there's an opportunity for automation in the enterprise. Personally, I would much rather build products and companies that obliterate. That's actually something we talk about at USV all the time: don't automate, obliterate. But people want to automate.

14. "Don't Automate - Obliterate": How USV Picks What to Invest In

What does that mean, “don't automate”?

You just talked about automation, and I think you were talking about SaaS in a sense. There have been amazing companies built that automate existing workflows and existing processes. But at USV, we like to bet on businesses that obliterate—businesses that literally obliterate markets and existing business models. We don't do a lot of enterprise investing for this exact reason.

Enterprise investing often requires selling to a middleman, selling to customers, or building things that just make existing businesses faster. We want to invest in businesses that literally reinvent the way something is done. Take Doctronic as an example. Are you familiar with Doctronic?

USV led the seed in Doctronic a couple of years ago. At the time, it seemed like a crazy idea. The idea was AI putting a doctor in literally everyone's pocket. We could have invested in AI that makes medical practices faster or more efficient, or helps you with insurance claims, but that's what it did.

We can put doctors in everyone's pocket and totally reinvent the model with AI. I think that's what we're talking about when we talk about “obliterate, don't automate”: Doctronic, doctor in your pocket, da-da-da. It reminds me of education. Sorry, I didn't mean that disparagingly. It's awesome, and—

But it reminds me—sorry—education kind of reminds me of the same thing, actually. Do you know who the biggest winners are in both education and, in most cases, medicine?

What was that?

It's a horizontal platform. It's YouTube.

And my worry is especially with health—

ChatGPT's second-most-used thing—

Maybe.

15. How Abridge Built a Healthcare Moat Over 8 Years

Health. And so my question to that—it’s not tied to Doctronic—is how do you think about where model providers will go and eat into the application layers and not kill our lunch? Maybe we just talked about Abridge. I think you had Shiv on here recently.

Abridge has been working on this problem for close to 10 years now, and it's given them a massive advantage. It turns out that in highly regulated industries like healthcare, you can't just walk in the door and say, “Hey, we're going to do healthcare now.” You have to actually build relationships, build partnerships, and clear regulatory hurdles, and that ends up being a form of a moat.

USV invested in Abridge, I think, in 2018—8 years ago now. It took that company 5, 6, 7 years to hit this inflection point where the product is being used by so many doctors in so many different healthcare systems because they were able to clear those hurdles.

I think we can say that we live in a world where the labs and the hyperscalers can take over any product and any market, but the reality is that's been the story of technology and innovation since the beginning of time. There are always going to be startups and founders and entrepreneurs that specialize and do the really hard things before anyone else has thought of them that end up prevailing.

Just take Spotify, which we talked about at the beginning. Spotify went up against a behemoth. It's a classic David-and-Goliath story, and it could have been killed 50 different ways along the way, but it specialized and did the hard thing early and often, and now it's obviously the winner.

I hear you that the labs can go into any application layer, but not only do I think that's not true, I also just think it's not really fun. What fun can we have as venture capitalists if we don't believe that startups can take down Goliath?

I'm also not sure if it's as binary, even if they do. You saw Claude and Anthropic dedicate an entire team to design and go up against Figma. Figma is still a business that does multiple billions in revenue, and there are still multiple players in the space.

Great product.

Great product. It's not just a trusted brand.

Trusted.

You know, my biggest mistake, especially on the developer landscape, was that I thought there would be a runaway winner. Of course, Claude has run away and done amazingly well, but Lovable's at $500 million ARR. Cognition, Replit, and Anthropic's Claude Code are crushing it. They've all crushed it.

There usually isn't a market winner that takes 100% of the market or 80% of the market. Usually, what you see is that the market winner takes something like 30% of the market, right? Which means the remaining 70% is totally up for grabs. If the market's big enough, that can produce some really large winners.

Can I ask you, do you see different deals at USV than the types you saw at Lightspeed? I'm not saying bad or worse. Do you just see different?

I think that, going back to what we talked about earlier with seed, if you put yourself out there in the world, you're going to see deals that match those themes and ideas. Not to say I didn't do that at Lightspeed—we did that at Lightspeed as well.

But to the point earlier about being very focused on energy, right, or now being very focused on this Rebel Alliance, we're going to see a lot of those companies come to us. I think when you're small, you kind of have to pick your spots. You have to have constraints and focus on a small number of things.

16. The Model Provider Threat to the Application Layer

To the extent that we see things differently, it's because we focus really, really deeply, and for a period of time we may only look at a couple of different areas that we have theses on.

What are you unsure of right now that is worrying you? To my point earlier, I think a lot of people are concerned by the intrusion from model providers into the application layer. We're investors in Legora. Anthropic is very openly suggesting it wants to come into legal. I think that's moronic for multiple reasons. When you're chasing AGI, to be like, “Ah, we're going to go after your lunch, Clifford Chance or Slaughter and May,” doesn't make much sense.

Look, as much as I say that there's always going to be an opportunity for startups, and that one model or one company can't do everything, of course you still worry about competitive threats.

As you know, I partnered closely with Granola when I was at Lightseed, and Suno as well. Those are 2 startups going up against massive, massive incumbents. Granola, in particular, had OpenAI launch a directly competitive product, right? Notion, a very well-funded startup, launched a directly competitive product.

I do worry about that, but again, that's always been the worry when you're building startups. Not only do I think that the labs going into every application isn't true, I also just think it's not really fun. What fun can we have as venture capitalists if we don't believe that startups can take down Goliath?

Is the worry for Granola not just breaking into enterprise—traditional enterprise? Notion has found it incredibly hard to break into traditional enterprise. These are startup and venture-funded products, which is amazing, and you can get to $500 million in revenue, but you need to be $5 billion in revenue.

I think the nice thing about Granola, what it really has going for it, and one of the things that I believe has helped that company with enterprise penetration, is being really focused, right? They're just doing notes. I think there are other companies—startups selling into the enterprise—that try to do many different things.

But dude, I think you’ve got to do more.

Maybe, but I think you want to get your foot in the door. When you want to get your foot in the door, you want to sell one thing. Once you're in, then you can expand horizontally to different categories.

17. Why Being First & Moving Fast Is the AI Product Playbook

But I think if you come in expanded really, really wide, now you have to convince the enterprise to give up all these different things, right? You’ve got to give up your Gmail and your Google Docs and your Google Sheets and your calendar, and this and that. Whereas, if you're just doing meeting notes, it's like, “Hey, we're just going to be your second brain for your company, right? We're not going to do anything else. We're just going to be your second brain. Oh, and by the way, we're the best at it.”

Do you not worry about the enduring wave of Microsoft and bundling, whether it's Teams that crushes or whether it's Copilot? I mean, Microsoft has built a business on doing 65% to 70% meh products, but with the power of bundling, beating the competitors.

I do worry about that. Of course, absolutely. We saw this play out with Slack. By the way, Slack is still around. It's still a massively valuable product.

$27 billion. I love the way we say this. It's like, “$27 billion? I'll take it.”

Yeah, I do worry about it. What I've also learned about AI products and Granola is that context is extremely valuable. If Granola gets inside of your organization and everyone in the organization starts using it and producing and accumulating all of this amazingly rich and valuable context, that's not something you, as an enterprise, want to give up.

You don't want to give up that context. You've built out this incredibly rich history of information that actually helps you work better. I think this era of building AI products is, in many ways, about being first and about moving really, really fast.

You know what worries me? It's just the math of such large outcomes. Again, this all sounds incredibly disparaging, and it's not meant at all because I'm a podcaster, so—an incredible disclaimer—

And a VC.

But my point being, when you have a SpaceX at $3 trillion, Elon Musk made more money overnight than Warren Buffett has done in his entire career.

Credit to him. Well done. God bless capitalism.

Amazing entrepreneur.

Amazing entrepreneur. God bless capitalism.

But my point is, when you have the outcome sizes of Anthropic and OpenAI and the trillion-dollar companies, the kind of company that does $500 million in revenue—

I think it all goes back to: What are you optimizing for? What's your mission? What are your constraints? We talked about VC early on, before we stepped into the studio here. USV has always had small funds. It's the game we like to play, and you can be very successful in terms of what you're trying to accomplish as a venture firm with a small fund.

Now, maybe that same model can't work if you have a $10 billion fund, but we're playing a different game. I think you're playing a different game, too.

I would not do venture without media. I said to my LPs the other day two things that really shocked them. I said, “You can't do Series A unless you have a $400 million fund today. Absolutely no chance.”

18. Series A Valuations Are Now $80–150M Post

Well, Series A rounds are expensive now. They're $100 million post. They're $80 million post. They're occasionally $150 million post. So, yeah, if you want to write a check of around that size and get decent ownership for your fund, you have to have a slightly bigger fund than you had 3 years ago.

Are you being more elastic on ownership?

I think it depends on stage. The way we think about it at USV is that at the seed and Series A stage, ownership is important, and the reason is the outcomes are bigger right now. You talked about it: The companies that end up doing really, really well raise a lot of capital very, very quickly, and the valuations get very high.

In those later rounds, it's actually very, very hard to buy ownership, especially if you're a small fund. So you have to get that ownership early on, and seed and Series A are the rounds where you can do that.

Once a company crosses that chasm, maybe into the Series B or the Series C, and you can identify them as being market leaders and potential generational companies, then I think it matters less. Then it's just about getting capital into those companies and underwriting it as a cash-on-cash type outcome, forgetting about percentages and just imagining, “Okay, if we put $25 million into this, what can the multiple be on that? Is it a 10x? Is it 100x? Is it 1,000x?” If it's something really big like that, then you don't really need to think about ownership.

We just did a deal together. By the time this comes out, it'll be announced. I actually got a message from the founder overnight, which is why I'm able to talk about it: Paul at Fomo.

Oh, nice.

Yeah, and you guys did that round—

Later for us.

Yeah, and that was why I asked that question.

But you saw the size of the market. For people who don't know, it's a next-generation trading app in many respects.

I guess you saw that multiple expansion at the entry price you came in at.

That's a perfect example. We saw where it was, and we said to ourselves, “How big can we see this getting?” We think it can be a lot bigger than it is now. So for this one, we're not going to be ownership-focused. We're just going to put a check in.

We may not be able to bet every round from here on out because, again, we have constraints to the size of our funds. But we think there's enough opportunity and enough upside from here that it's worth it.

Should you do it if you don't think there's a mega outcome?

I know that sounds strange. I'm not sure, for us—for USV—at the later stages, if we don't think that there's a gigantic outcome, it probably doesn't make sense for us to bet.

But should you do it early if it's not a gigantic outcome, is my question?

Look, I think in general you always want to believe that the things you're betting on are playing in large markets with very, very large service areas, and we're going to look at every company that way.

But I do think that because our funds are smaller, similar to yours, there's probably more downside protection for us in that model than, say, at a larger fund where you have to multiply billions of dollars.

For us, the fund that we're investing out of now, the core fund, is a $275 million core fund. You can multiply that a lot easier than multiplying, say, several billion dollars, especially if you get the same type of ownership that a larger fund is getting.

I think it will be the best time ever, though, for the very large platform funds, and I think they'll be able to deliver venture-like returns because of the expansion in outcome sizes. Do you think that's true or not? You look at Thrive.

Yeah. I mean, their numbers are better than 90% of seed fund numbers on their growth vehicle.

Yeah. When you look at them, I mean, you know, they made a precursor phenomenal. The OpenAI fund is obviously incredible. They are getting venture outcomes on multibillion-dollar investments.

We think we're coming out of a period where there has been a massive infrastructure build-out and very, very capital-intensive companies. I think large firms that were able to invest in those very capital-intensive companies have done phenomenally well.

But again, I think we're also now playing in a world where there will be a proliferation of startups and apps at the application layer that are less capital-intensive. For those, again, I think you're going to need to know what you're looking for, and I think you can get into those companies at more modest check sizes.

I don't think you can have a fund that's in the middle to do that. I think you need to either be on the large end and play the consensus game, or be really small and really opinionated and multiply a much smaller fund.

Are you elastic on price? You know, candidly, we've lost 2 deals in the last year.

Mm-hmm.

And listen, we got outbid. I don't know if we would have won if we'd paid the same price. I hate people who say that. But by 2x or 3x: We bid $150 million on one, and on another, we got outbid at $300 million. On another, we got outbid at $400 million. Should I have just paid?

I talked to my partner, Fred Wilson—shout-out to Fred, one of the GOATs—and I asked him a question. I said, “What is the biggest lesson you've learned in VC?” Actually, I asked him what a couple of the biggest lessons were, and one of them was, “Never pass on price.”

So Fred says, “Never pass on price.” I think in reality there's probably some nuance. Just to bring it back to the USV strategy, for us, if we're investing at a slightly later stage, we need to really believe that it's a market winner and the multiple can be very, very, very high. In that case, I think we can be a little bit more elastic on price.

But I think at the earliest stages, especially when we're investing out of our small funds, I do think there will always be a limit for us on price, just because otherwise we can't make the fund math work.

You know why my job is just incredible? It's because I really am a student of this business. By the way, you said, “Shout-out to Fred.” I emailed Fred when I was 17 years old, and he and Brad Feld both responded to me within 30 minutes.

Wow.

That is the testament to giving time to people when, candidly, I had nothing.

I told a story on that podcast where, when Anchor first launched, I had never met Fred. The Anchor app accidentally auto-tweeted for Fred, and he had a big follower count on Twitter at the time—now it's X, obviously—and he was not happy about that. He started lighting me up on Twitter. This was the day after we launched Anchor, and obviously that was uncomfortable for me as a first-time founder.

But to your point, I cold-emailed him right after the fact, apologized, and offered to troubleshoot it with him. He responded immediately, and he was super gracious, forgiving, and understanding. I had a similar experience.

I think all publicity is good publicity.

Yes, true. Actually, you probably—

We got a lot of publicity from that.

Yeah, you probably do.

Never pass on price. Where I was going with the student-of-venture style, I think it actually goes to something that Peter Fenton told me, which is: use price as a litmus test for your conviction.

There you go.

There are founders like Alan Chang at Fuse or Jack at Air Wallace where you could treble the price, and I'd still pay it today.

Yeah.

You have founders like this where, if you doubled the price, you're like, “I'd still pay it.” I'd—

Yeah. Yeah, absolutely. I can think of some of the best deals I've done, and even though I'm happy with where the price landed, in hindsight I'd probably pay double the price if I could go back and do it again.

19. Biggest Investing Lesson: Never Project Your Own Ideas on Founders

If I were to ask you what your biggest lessons are now, having been in venture, in the same way that you asked Fred, what would you say they are?

You can't project your own ideas onto the founder. Sure, you can offer ideas, of course, if the founder wants them, but especially as a former operator, you often think, “Oh, I know how this business should be built, or I know how this product should work.” I think that can really get you into trouble.

With one of my former partners, Jeremy Lou at Lightseed, I made this mistake a few times, and he called me up one day and said, “Hey, I think you're projecting as a former founder.” He said, “In my experience, that's very, very dangerous and very risky.” I think it was a great lesson.

The reality is, even if you have great ideas and you think you know the answer—even if you're right, by the way—the reality is that it's the founder's company, and they may want to build it a completely different way. Not only that, if you project your ideas onto a team that you're evaluating, you might get it completely wrong.

You might look at a business and say, “I know what this company needs to do. It needs to do A, B, C, and D, and then it's going to win.” But if you make your bet based on that evaluation and that judgment, and that team doesn't do those 4 things, you made a bad bet. So don't project. Don't project. You have to really believe in the team, the team's judgment, and the team's ability to execute on their own plan. That was a big lesson for me as a former CEO.

20. Founder, Market, Product

Founder, market, product—greater than 1 through 3. When I first started this job, I would have said product, market, founder. I've completely flipped that. Now I'm founder, market, product.

I think the founder is the most important thing. At the end of the day, most startups, especially early-stage startups, are going to pivot in some form or another. What's most important is: are they resilient? Can they execute? Can they adapt to change?

When you get founder reads wrong, what do you not see that you wish you'd seen?

I often see communication as a big hurdle for founders. I think one of the hardest things you can do as a leader is communicate effectively, because communication touches every part of company building.

It's part of recruiting, which, as we know, is one of the most important things you can do. You have to be able to communicate your mission, your values, why you exist, and what you're trying to accomplish. You have to communicate to investors, right? If you're not a great communicator, you're not going to be able to raise capital or put venture dollars into the business.

You need to be able to communicate your product vision to your team. How can you align the team around building the perfect product if you can't communicate it? You have to be able to tell your story to the market. When I think about some of the mistakes I've made in evaluation, it's maybe been on communication. That's probably been one of the biggest.

What do you think the single biggest mistake in venture you've made in terms of a miss is? I would say mine probably is one that you gave me—well, 2 of yours.

I turned down Suno when David Franklin at Founder Collective introduced me at the seed stage. The lesson was really interesting, and I hope Mikey's okay with me saying this. I'm sure he is. It was a $250K check or a $200K check, and so the ownership was 1%. I was like, “No, I need high ownership. My LPs tell me high ownership.”

Granola, honestly, was a lack of imagination. Chris, I saw, was great, but it was a very unformed idea, and I was like, “I can't.”

Founder first, founder only. When you have the founders, the other 2 don't exist.

Terrible mistakes there for both. Ironically, you did very well on both. Congrats.

21. Suno: Thesis-Driven Bet vs Granola

2 different motions, by the way.

Sure.

Granola for me was a pure founder bet. Pure founder bet. I knew Chris. I've known Chris now for 15-plus years.

He's also not an amazing seller. I'm just going to give it to you bluntly.

But I knew Chris up close. When I was building Anchor, the Socratic office—Socratic was his previous startup—was right behind ours. We did a few meetups together. We got our teams together for a few happy hours, and I would talk to him as another founder. We would share our struggles and get advice from each other, and I was like, “This guy, I know he can do it. I've seen him. I've seen him do it.” Pure founder bet.

Suno was something very different. Suno was a pure thesis-driven bet for me. The thesis came from the days of Anchor and seeing what it meant for a product to democratize a creative medium. Music had never really been fully democratized before AI.

So I went out and met with every team building AI for music. I hunted down every team, and then, of course, I met Mikey and the team. I was like, “Wow, this is an incredible team.” Then the founder lens came on, and I was like, “This is an incredible team.”

How obvious is it when you meet a founder that you want to invest?

With Mikey and the Suno founders, it was immediate. I met Mikey.

Was it normally immediate?

Not always. Not always. With Suno, Mikey and I sat down at a restaurant near my home in Hoboken, New Jersey. His family's in Jersey, so we met up in Hoboken, and we immediately hit it off—just instant, instant connection.

I immediately recognized not only how smart he and the team were about the models they were building and their vision for the world, but also how passionate they were about the problem. Mikey is a former musician, so I felt like he could deeply, deeply connect to this idea: how do you make music creation, expressing yourself, and creativity easier?

When you're doing Suno at $5 billion, what are you underwriting that to?

We believe that has—I'm using this term loosely—unlimited upside potential. If you look back at the history of media on the internet and think about the generational platforms that truly democratized a medium—YouTube, TikTok, Twitter, Facebook to some extent with writing and publishing—I mean, these are insanely valuable platforms, and you don't really think about the upside of that.

Does Suno not have to move from a tool to a platform, then?

22. Suno at $5B

I probably would have said yes when I invested. First of all, I do think it is a platform, but I think you're asking about something different. You're asking—

Yeah, you're talking about whether you need a creator side and a consumer side, similar to what we saw play out on Spotify, as an example.

I think when I originally invested, we thought that it did. Now I'm not so sure that we understood what Suno could be back then.

What's happening now in Suno is what Mikey and the team have been saying for a while and what I now believe: it's literally a new behavior. The team calls it creative entertainment. People are coming into Suno and making music for the pure joy and entertainment value of making music.

Previously, let's say on Anchor, you would make a podcast for the purpose of getting distribution and eventually monetizing it. Right now, AI has made music creation so much fun that you're making music to make music.

I think it's actually not too dissimilar to what we see with Claude Code or Midjourney, or some platforms that are making it easy to make games with AI. You have no greater aspiration for this content, right? You're not trying to become a hit game developer.

You're just having a lot of fun making games on your computer. And we see that a lot in Suno.

What are your biggest misses?

You're brilliant, but you should be a politician.

I asked you what you missed.

I'll tell you about my 2 greatest misses.

Thanks, dude.

One company that I would have loved to invest in a long time ago, which we didn't at Lightseed—we met them a few times. I think we could never really wrap our heads around the price and, at the time, where they fit in the market.

23. Is Traditional Media Dead?

Now, with the benefit of hindsight, I really regret not ever having had the opportunity to invest in Substack. I really believe in Substack. I realize it's not your traditional sort of AI company, but over the past 8 years—and I think the trend is continuing—we're moving more and more and more toward a world of self-publishing, with people controlling their own destiny in terms of media, what they create, what they publish, and how they monetize it. I think no company, maybe other than X, has done a better job of that than Substack. That's a company where I love the founders, I love the purpose behind the company, and I love the product and the platform. Unfortunately, I never got a chance to invest.

Is traditional media dead?

I think traditional media, in many respects, is dead.

I go to these large houses—the Timeses of the world, your Skys—and, dude, they have these insane production places for podcasts that I know no one listens to, and the teams are ridiculous. It's all like, guys, you don't get where this game has gone.

I probably underestimated just how big independent media could become. When we built Anchor, we had this vision of democratizing audio and giving everyone a voice. By the time I left Spotify in 2022, I'll admit I kind of thought the opportunity was baked. I was like, “Okay, it's done. It's done. We've got YouTube, we've got Spotify, we've got whatever, we've got X.”

It's gotten so much bigger since then, and that was only 4 years ago. Every traditional TV show or media personality—everything—has made the leap over to self-publishing or independent media. Huge names that were on some of these large media organizations just made the leap, and now they just have a YouTube show or they just have a show on Spotify. I don't even think we've reached the peak yet. I think television is still in the process of being massively unbundled.

When TechCrunch Europe shut down and now Wired Europe has shut down, I coalesced the best writers in Europe, and I had $10 million from very rich billionaires to bring together the next generation. Then I sat down with my investing team, and they're like, “Why the fuck are we doing this? We have no editorial control over them, just like OpenAI doesn't own TBPN. We can't influence it. What's the point?”

Yeah. And to your point, editorial freedom is everything.

It's amazing. It's incredible, which is why I love Substack and why I like X, YouTube, and Spotify. I think these platforms are amazing.

24. Quick Fire Round

Are you ready for a quick-fire round?

Sure.

Single best first founder meeting you've ever had and why?

Bren Putnham, CEO of Bor. Have you seen Bōrd? It's a tabletop gaming console. They just announced the Series A a few weeks ago from USV.

When I met Brynn, I was blown away. She's a force-of-nature founder with such a clear vision and such deep domain expertise. Prior to Bōrd, which is a beautiful tabletop display, Brynn was the founder and CEO of Mirror, the workout device, which she sold to Lululemon. She took that domain expertise and said, “I know how to build a tabletop gaming console now,” with the same hardware and the same supply chains, and just crushed it. When I met her, I walked out of that meeting and thought to myself, “I have to invest in this company.”

One of the single greatest exits. Also, you have to give her credit.

She's brilliant.

You can invest in 1 seed fund. Which seed fund would you invest in?

I'm a big fan of what Matt Hartman is doing with Factorial. Factorial is a fund. Do you know Matt?

Dude, I met Matt years ago when he was at Betaworks with John, and I went to Betaworks. I remember they had a company called Poncho. Do you remember Poncho? You mentioned chief weather officer earlier; it was the original chief weather officer.

Yeah, and so I remember them doing better. Matt's brilliant.

Matt has innovated on a completely new model of venture where he arms angel investors. These are not scouts; these are angel investors. He arms them with additional funds on top of their own capital, which Factorial then obviously monetizes through carry and whatnot.

The brilliant thing about that model is that now he's got all these great angel investors as part of the Factorial network. These are people like Clem from Hugging Face—real angel investors. I think that's brilliant.

My second answer would be, tied for my favorite seed fund, Haystack. I'm just a huge fan of Samil and Divia and the work they do. It's an incredible fund. Obviously, they've produced some generational outcomes.

Who, when they send you a deal, is the highest signal of quality?

I was really fortunate at Lightspeed to collaborate closely with Nat Friedman and Daniel Gross before they went over to Meta. When Nat would send me a deal, which he did on multiple occasions—and, by the way, that's actually how I met Mikey from Suno—I always knew to pay attention. If Nat was sending me a deal, I had to pay attention to that.

That is the greatest shame in venture—the fact that they didn't invest. They were just so good.

So good.

So good. I mean, those guys—oh my God. What is your favorite growth fund if you were to invest in 1?

I think the Lightseed Growth Fund is phenomenal. Again, they've done Anthropic, they've done xAI, they've done SpaceX, and they've done amazing investments over the history of the firm. But really, over the past 4 years of AI, they've been phenomenal.

What's your biggest lesson from working with Fred during the short time you did?

I would say one of the biggest lessons I've learned from Fred is that there is really nothing more important than your relationship with the founders. That's probably the best thing you can do to help a founder: just be there for them. That's what founders need. Founders are lonely. I could tell you as a former CEO, it's a lonely, punishing job. Just having somebody you can actually trust and actually feel aligned with is very, very important.

But secondly, selfishly, I think reputation is everything in this business. You need a great reputation, and the only way you're going to have a good reputation is if you have great relationships with founders.

What's your biggest parenting advice? I heard that you're an incredible papa. You're also a great investor. What's the biggest parenting advice?

I feel I was extremely fortunate growing up because I had parents who were extremely supportive of me and my interests, but who also did not pressure me in any one direction.

I'll give you an example. I grew up in a home where baseball—an American sport that I don't think is very popular here—was a very big deal. Baseball is a very big deal. I played some baseball growing up, but I didn't continue with it. I think there are a lot of homes like that in the United States where, if you're not playing the sport, going to the university, or pursuing the career in the thing that the family really cares about, you're a failure, or you get pressured into sticking with it.

My dad, who was a former coach and baseball player, was just like, “Great, what do you want to do next?” I think that is so important. It's definitely something that I really try to remember and try to instill in the values of my family with my kids.

I really want my kids to do what interests them, but I never want to pressure them to do it. I want to support them. I want to be there for them. I want to drive them to practice. I want to do everything I can to empower them, but I never want to pressure them too much.

I remember calling my mom and telling her I was dropping out of law school to do this podcast that made no money, having worked so hard for years to go to law school. She was like, “Cool. If that's what you think is best.”

Yeah. She might pressure-test you, right? “Hey, are you sure? Talk me through it.” But once you—

No, not even. Okay. I said, “Listen, I fell asleep in the first law lecture, and I skipped it, and that is for you.” She was like, “I think law's not for you.”

But penultimate one for you: what have you changed your mind on most in the last 12 months?

I do think that there was a moment not that long ago—and I know I said the opposite a few minutes ago—where I thought the model providers could do everything. I thought we were headed toward this world where there would be 1 company or 5 companies that would just do everything.

Fortunately, what I've been reminded of over the past year or so, or 2 years, is that companies can't do everything. They can't do everything.

Even the biggest companies can’t do everything. I remember 10 years ago, 15 years ago, we thought the same thing about Google. We thought the same thing about Apple. We thought these companies were going to just take everything.

The reality is they can’t, and they don’t. I’ve been reminded of that recently, and it’s been a surprise. I’m glad for it.

You work with Fred, who’s obviously been around the block a few times. I don’t work with him, but my papa is Roryo Driscoll from scale, who likewise has been around the block a few times. He got into venture the same year I was born, and he reminds me that Microsoft, at one point, were looking at becoming a bank.

Totally. It works sometimes, right? AWS is a great example, but they can’t do everything.

Final one for you. When you look forward to the next 5 to 10 years, what are you most excited for? I like optimism. What do you like? It’s going to be amazing.

My partners, Rebecca and Nick, and I asked each other the other day, “What do you want to do at USV?” I said three things.

The first thing is, I want to show up to work each day and collaborate with people I love being around—people I enjoy collaborating with and people I have fun with. That’s actually number one for me.

Number two is, I want to partner with some incredible founders. Similarly, I want it to be about that relationship and that dynamic, and I want these founders who are going to change the world. Number three is, I want to produce generational returns in terms of funds—in that order.

Number one for me is to show up to work and have fun with the people I work with. Number two is to partner with founders whom I love and who I believe can produce generational companies. Then number three is to produce generational returns as a fund, in that order.

I think that’s a good recipe. I think if you optimize for fun and enjoying your day-to-day, good things will happen.

Mike, I so appreciate you taking the time out on the London trip. I so appreciate you putting up with my bold stances as you recoil back and go, “Don’t attribute it to me. Don’t attribute it to me.”

I really appreciate the decade-long friendship now. You are a star, dude. Thank you so much.