Benchmark GP、Victor Lazarte：所有顶级创始人都具备的3个特质

Big companies talk about, “Oh no, AI is not replacing people. AI is actually augmenting people’s abilities.” This is bullshit—it’s fully replacing people. But then the first question that we ask is, if models get a lot better, is your company worse or is your company better? Is there a bubble? Are people getting paid for the risk they’re taking? What I’m telling you is, in the next 3 years, I’m sure someone will start a company that’s going to be worth $1 trillion. So I think, on average, buying the winners today will work.

Victor, dude, I have basically stalked the shit out of you for the last 24 hours. I spoke to Peter Fenton, Sarah Tavel, Bruce Dunlevie, Pat Grady, and the Brex founders. I did my work for this one, so thank you for joining me.

1. Lessons Scaling Wildlife Studios to 4BN Downloads

It’s great to be here. Thank you for having me.

When I spoke to Pedro, he was like, “Not many people know, but, respectfully, Victor really came from nothing—like $100—to building a huge $350 million-revenue gaming business.” Can you take me through that business-building journey and 1 or 2 moments that really shaped who you are as a person and what you learned?

So, yeah, I bootstrapped a mobile gaming business. It makes for a good story, but the reality is that we tried really hard to raise money. We didn’t bootstrap by choice.

Why don’t you think you could raise? What did you do wrong, in hindsight?

I was never very business sophisticated. This was 2011 in Brazil, so the venture capital scene there wasn’t as developed, right? But the idea was that I loved mobile games and thought it was going to be a good business, but there wasn’t a very strong, well-thought-out thesis. I guess that’s why the best term sheet we got was someone offering us $50,000 for half the company, right? Which, obviously, we couldn’t take.

That’s a relief. We had Oscar Pierre on the show from Glovo in Europe, and they sold a third for €100,000.

Yeah. There are so many horror stories from investors in Brazil at that time, and I think they were less sophisticated. I wasn’t very sophisticated, but there were people who were even less sophisticated. I had friends who raised money in Brazil and agreed to a 20x liquidation preference, which seems like, “Oh, there’s this clause in the contract, but I don’t really know what it is.” People do that and end up losing their own company—their whole company.

Wow, that’s absolutely nuts. So you’re like, “Fuck, we can’t fundraise. No one’s giving us money. We need to build this business naturally.” What happens then?

I think the core insight—the thing that made us choose to do mobile games—was that I started a company with my brother, and in college we loved video games. Initially, it was, “Hey, let’s build a company with Nintendo,” but that sounded like a really bad business idea. Then we looked around: “Hey, what’s working in the world?”

We saw that app downloads—this was 2010, when we were studying the market—had grown 42x from 2008 to 2009. I thought, “Wow, this growth is insane. We should make mobile apps.” We loved games, so let’s make mobile games. I think that was a great insight, and we started the company. In the first few games, we did everything ourselves; we didn’t have any employees.

Does that shape your thinking on market timing? Because what you basically just referenced—

100%.

So how do you think about that?

2. Why Predicting the Future is Wrong When Starting a Company

I think when people want to start companies, a lot of times they’re trying to predict the future, but it turns out it’s very hard to predict the future. It’s much easier to understand the present. So a good heuristic is, “Hey, what’s working? What’s working right now?”

There’s the famous story that Jeff Bezos was at D. E. Shaw, and he saw the internet was growing 23x a year and said, “Hey, I need to make a business on the internet.” That was the initial insight.

And then, what’s the insight right now? It’s quite clear that LLM usage has grown 100x in the last 2 years. If you measure it in the number of FLOPs, it went from 1x of FLOPs to 100x of FLOPs now. Of course, this data is not very precise, but it’s directionally correct. I think the first thing is understanding what’s happening today, what’s working, and then how can you build on top of that.

Or build things that are inspired by it, right? When you look at the history of the best companies—the most impactful companies—Facebook was not the first social network. When the Facebook founder built Facebook, there were MySpace and Friendster before it, right? But you could see that there were things in that direction that were working.

In 2000, building an internet company was an amazing thing. In 2010, building a mobile gaming company—building any app company—was a great thing. That’s why you have some mobile gaming companies that were successful and other companies that were even more successful, like Uber and Instagram. They were all started around the same time.

Now, with LLM companies, it’s quite clear that it’s the best time in over a decade to start a company because we have this insane phenomenon of adoption. Adoption of LLM inference—we started with, “Okay, what’s the thing in LLMs that’s working?” I think a year and a half ago, there were basically 2 things that were working. There was ChatGPT and Character.AI. I would say Character.AI was one of the things that was really working.

Now I think you can argue that there’s Character.AI, ChatGPT, and Cursor. Those are the things that are working, that people love. There’s a lot of usage—tens of millions of users. What these things have in common, and a lot of times the great opportunities are adjacent to things that are working today, is that they’re built on those existing behaviors. I think a great place to start is just having a profound understanding of what’s working and then building adjacencies to it.

But on that topic, social networking—we now have software that makes us more efficient at connecting with our friends. You have WhatsApp, Facebook, and Instagram. What does an agent for that look like? What’s an agent for your social life?

I mean, is it not, bluntly, the removal of other humans? It’s a 24-hour-a-day constant companion that understands how you feel, responds to you in real time, and is your continuous, on-demand friend. So I think that’s a really powerful idea.

I wouldn’t even make a larger claim. The claim is: What do people want? In the end, people just want to be happy. There’s a lot of study on what makes someone happy, right? It turns out that wealth makes people happier. The more material wealth you have, the happier you are, but it’s not a linear relationship; it’s log-linear, right? If you get to $75,000 a year for a US household, it starts to taper off.

Do you know what astounds me, Victor, honestly? You said beforehand that you wanted something conversational, and I agree with you. I’m terrified about the mental health pandemic that we have today. Young men are more likely to die by suicide than of cancer. That is a terrifying statistic.

What I find astounding is that the largest mental health company is Calm, respectfully. It’s not a huge business. It’s a $1.5 billion, great, fantastic journey, but it’s not a huge business. If our job is to solve humanity’s biggest problems, where’s the disconnect?

And so making people wealthier actually does not affect their happiness that much. What does technology do? Technology helps us be more productive, which is—hey—it makes us wealthier and helps us achieve the goals that we want better. So it makes us more productive.

But research shows that the biggest predictor of your happiness is the quality and depth of your relationships, right? Technology has not attacked that. But I think that the really cool thing now is that we're able to have a digital friend, and I think that would be fantastic.

I think you look at the studies, and religious people, on average, are happier than nonreligious people. A lot of the difference is explained because religious people have stronger relationships with the community, but also religious people have a relationship with God. God is this friend that you never see. You talk to God, but God doesn't talk back.

So, to me, that existence proves that you can have this entity that you never see that is an important relationship to you, and that makes you that much happier.

I say, actually, that the death of religion is the thesis for CPG and consumer brand investors today, because we've shifted cult-like worship from God, Buddha, and these deity figures to HYROX, CrossFit, Taylor Swift, Lululemon, and SoulCycle. They are, ironically, in the minds of modern consumers, a community that consolidates beliefs and brings unity. It is this feeling of togetherness, which is extraordinary to say, that's replaced God. Maybe it's signifying of society, but I think about that a lot.

Would you not say that social has brought more quality and depth to relationships? Think about your ability to follow your friend's wedding on Instagram.

No, I think social is definitely positive. I first interacted with my wife through Instagram. I met my wife three years ago, and I first interacted with her through Instagram. We had friends in common, so I started following her and commenting on her stories.

Maybe our relationship wouldn't have happened because of Instagram, and she's the biggest source of joy in my life. I'm incredibly grateful that these things exist. But I think this thing just scratches the surface. The way Instagram is set up, it does not help you connect nowhere near the way it would be able to.

When you think about why it's adaptive to have friends, a lot of it is that we need to have a story that helps us understand what's happening with us. We create our own internal narrative, and when you have a friend, you explain your narrative to your friend. Your friend comments on your narrative and helps you shape it, and that makes you better able to understand yourself and where you're going.

Humans have this intense need to feel seen and understood, and this is very adaptive. We're at a point where AI is able to do that for you. AI is able to help you shape, “Okay, what is your personal narrative?” I have this very strong belief that, 5 years from now, for the majority of people, the person that will understand you the most is going to be an AI.

Is that a world that you want your kids to grow up in?

I think that's where I disagree with a lot of people. I think we're always very afraid of technology. Humans are always very afraid of what's coming next. I actually think it's going to be wonderful.

People feel isolated, and being isolated is bad for a bunch of reasons because you feel like you don't understand yourself. If you have a challenge, you have a hard time thinking about that challenge. Imagine you have this hyper-capable friend that helps you understand yourself, knows everything about you, and, whatever situation you have, helps you figure that out. You feel supported. You feel like you're never alone.

Even more than that, a lot of people say, “Oh, this is very dystopic because once you have these AI friends, you won't have regular friends.” I think this is very untrue because, in the end, we will still want to see people in person, and your AI friend not only will help you understand yourself, but it will help you connect with great people.

It will help people find out, “Hey, I know you really well, and it turns out that there are 100 million people that I know really well, and I'm going to connect you guys.” But it's going to function very much in the way that a person introducing you to someone else functions. You get this introduction, but you have a common friend, so you know you have to behave well because, if you don't behave well, I'm going to tell your friends that you're an asshole, and your AI friend is not going to introduce you to anyone anymore.

Speaking of AI friends and the quality and depth of relationships, when I bluntly spoke to so many people about you, there was one relationship in particular that struck me: when I spoke to Pedro at Brex about you. He mentioned a number of things that really highlighted the depth of your friendship. You invested pre-product, pre-revenue. Can you take me to that and what that journey has taught you about investing, company trajectory, or how it shaped you?

The way I met Pedro is that I was building my company in Brazil. As I said, we bootstrapped the business, and we eventually grew it to over a billion users. There were very few people building technology companies in Brazil at the time.

Pedro had a problem with an investor. When he was 16, he created his first payments company in Brazil, and it turns out that he signed a deal that he shouldn't have. He lost control of the company, and the investors made it impossible for him to capture any value.

So I spent a ton of time with him and with lawyers, saying, “Hey, how do we get out of this?” I had no shares. I was just like, “Man, this kid is so smart. This kid is so good. He deserves to succeed.” So I really wanted him to succeed.

After a while, I was like, “Hey, man. I'm selling my mobile gaming company. We sell things in the US, and the US is a much better market. You should just let go of the company that you have and build something similar in the US.”

Eventually, he agreed to it. He moved to the US, and he started Brex. Then, when he started Brex, he said, “Hey, can you invest in my seed round, and can you be my first board member?” I was like, “Yeah, of course. I'd love to do that.”

I think our relationship came from this shared distrust of investors and this idea that, hey, we're going to have each other's back. The other part that I really enjoyed is that, to me, it was just so fun spending time with this young kid who loved business so much.

Frankly, I don't think I would have invested in Mercor if I hadn't invested in Brex. When I met the Mercor guys, at first glance, the businesses were so different, but the company was doing $1 million in annual revenue, and Brandon reminded me so much of the Brexers.

3. Two Traits That All the Best Founders Have?

There are 2 traits that I look for in entrepreneurs that they both have, and the 2 traits rarely come together. I like to invest in founders who are very open-minded but very disagreeable.

Normally, if someone is very open-minded, you say something and they want to learn more and are very curious about what you're saying. You make a whole argument, and then you start feeling, “Okay, I convinced this person because he's just so interested in what I have to say.” But by the end of it, the person says, “No, I actually think that the opposite is true.”

It's this idea that I'm interested in what you have to say, but I have no problem disagreeing with you to the point that it's going to upset you. Both Pedro and Brendan had that to a very high degree.

4. Why You Should Always Ask What a Founder Does in Their Free Time?

The other thing is that I think a great way to understand founders is to ask them how they spend their free time. If you understand how founders spend their free time, you get a very deep window into who they are.

Can you unpack that? When you ask that, how do the best display themselves? What are the variants of answers?

To be honest, this is a trick that I stole from Yuri Milner. I was having breakfast with Yuri, and I asked him, “What makes you a good investor?” He was telling me, “When I meet a founder, I ask him, ‘What's your day like?’”

But it's not, “Tell me 2 or 3 things.” No: What time do you wake up? What is the first thing you do? Walk through the entire day. What time do you go to bed, and what do you do in bed? The choices that people make in their free time are the ones that really tell you who they are.

For Pedro, it's, “Hey, in my free time, I love sitting in front of my computer and tracking the packets that iOS—the iPhone operating system—sends to servers because I really understand how that architecture is done. If I pay enough attention, then I get a chance to hack it.” That's how he found one of the jailbreaks back in the day.

5. Why Knowledge Work Will Be Destroyed and What Happens Then?

He was the first person to find one of the jailbreaks for the iPhone, right? But he did that for no reason. He didn't make money out of it; it's just, “Okay, I like going very deep on this technical thing.”

And then with Brandon, our first conversation was—okay, the guy's 21 years old, but you're talking with him about business and it feels like you're talking to your peer about business. He was telling me, “What is it like to work with Bill Gurley? I've read all the blog posts.” I thought, “Why would you read all of Bill Gurley's blog posts from years and years ago?”

He said, “I've heard all the podcasts.” There wasn't a specific reason for him to do that; he was just like, “I love studying businesses and how businesses work.” When you have very young people who spend a ton of time on something that's an intrinsic passion for them, that compounds really well over time.

To me, it's like, “Hey, really smart kid in a very promising space.” But he's not just smart; he's very, very special because of this almost obsessiveness with understanding businesses, right?

Little did I know that Mercor and Brex would have something else in common, which is that Brex went from $0 to $100 million in revenue in 18 months after launch, and Mercor went from about $1 million in revenue when I invested to over $100 million in 11 months.

I want to talk about the revenue scale, but I do want to talk about emotional maturity. I look back to myself when I was 21. I work with 21-year-olds, and as brilliant and insightful as they are, in my mind, you gain scar tissue, nuance, and just a little bit of wisdom with age. You can plant that with great mentors, but do you find you have to change the type of investor, board member, or coach that you are with a younger founder?

6. Why If You Start a Company in SF You are 1,000x More Likely to be Successful?

Yes, 100%. I think a lot about the role of a board member. I think it's related to what's special about Silicon Valley. I was trying to come up with a number, but if you start a company on the West Coast of the United States, you're 1,000 times more likely to build a tech giant on the West Coast of the United States than you are anywhere else in the world.

The raw math is that 70% of the largest tech companies in the world were started on the West Coast of the United States— the Bay Area and Seattle—and I think 0.1% of people live there. You do the math, and it's roughly 1,000 times more likely.

Why is that? People are just as smart in Silicon Valley as they are in Europe or Brazil. But the big difference is that there's so much knowledge about how companies are built. When someone starts a company, they're able to access all that knowledge.

When Zuckerberg started Facebook, he hired a bunch of people from Google, and he had board members who had been at other places. I think a lot of the role of a board member is: How do you catalyze that? How do you help a founder access all the knowledge of what has worked before?

In Brandon's case, it's like, “Okay, he's a 21-year-old running a company doing $100 million a year.” Very few people have been in that place. It turns out that Pedro has been in that place—being in his early 20s, running a company doing hundreds of millions—and there are a bunch of common patterns that emerge.

Then it's like, “Hey, Pedro, what were the things that you didn't know that you wish you did?” I connect the two of them, they spend time together, and those types of things are very valuable.

I think the way a board member adds value is a function of 2 things. One is how exposed you've been to hypergrowth in the past, and the other is how much time you put in to actually understand the context of that company so that you're able to draw from the right lessons.

Do you like being a board member?

I love being a board member. I wish that was the job.

How many companies do you meet a week, on average, that are net new?

I probably meet 1 or 2 companies a day.

Okay, so 5 to 10 a week. How many of those are raw inbound?

1 a week.

That's pretty good. So 1 inbound and 9 outbound. Is that good? Do you do active outbound, where you're like—

No. The majority of what I do is outbound.

Frankly, everyone thinks you join Benchmark or Sequoia and deals just come to you. Deals just come. Does that happen?

A lot of things come to you, but the question is: Are the deals that you want to do coming to you or not?

As you build your network—and I'm not from Silicon Valley, and frankly, I've been a founder for 13 years and I've been an investor for about a year and a half—a lot of people don't think of me as an investor. Very little of what I want to do just comes inbound, with someone saying, “Hey, go meet this company.”

Most of the time, you're talking to really high-quality people and asking them, “What has impressed you most?” Or you get super interested in a sector and say, “Okay, what are all the companies in this sector?” You go to someone and say, “What are all the companies in this sector?” You do a little bit of homework, and then it's like, “Oh, okay, this is an interesting company.”

I've done 4 companies since I started at Benchmark. 3 have been this way, and then the fourth one—I actually am a co-founder. I started the company with someone I knew for a long time.

You started a company?

Yes, we're still in stealth. We're not talking too much about it, but maybe a year from now I'll come back and tell you all about it.

I absolutely love that. We have that. Can we go back to Mercor and Brex, as you said? Revenue scaling: I think Brex took 18 months to reach $100 million, and Mercor took 11 months. It's insane.

7. Why Spreadsheet SaaS Investing is Dead

The challenge I have is: Have revenue rules broken in terms of scaling expectations? Have we lied to a generation of founders with “triple, triple, double, double,” when that's just not the case anymore?

I do think SaaS was a moment in time when rule-based investing worked really well. You got to $10 million in revenue, and you were a category winner. I don't think that's true anymore.

Why revenue? I think it's because of the nature of the revenue. A lot of AI revenue is experimental. The reality is, it's now incredibly easy to take ChatGPT and say, “What are companies using ChatGPT for?”

Let's say lawyers are using ChatGPT to write demand letters. I'm going to make this very thin workflow around ChatGPT, and then I'm going to go after every law firm that writes demand letters and say, “Buy my tool.” You're going to get millions of dollars in revenue.

But if that's all you're doing, there's just no enterprise value, because the reality is that as the models get better, the workflow you're building becomes less valuable.

One test that we always do is this: A company comes in and pitches, and the first thing they have is revenue growth. Revenue growth is still worth something. It's not worth what it was worth before, but it's like, “Hey, you have revenue growth. Okay, we should pay attention.”

The first question that we ask is: If the models get a lot better, is your company worse or is your company better? If your company is worse, then it's going to be very hard to back it. But if your company is like, “Hey, as the models get better, my company gets better,” then it's like, “Oh, that's a great place to be.”

For example, what got me excited about Mercor is that one thing I think is super exciting right now is replacing people. It sounds really bad when you say it this way, but I actually think it's the most exciting opportunity in venture right now, and it's going to be fantastic for humanity.

Figure out the profile of a person who does knowledge work that you can replace with a model. Replacing a recruiter—I'm like, “Oh, that's pretty cool.” I think models are going to be better at interviewing than people, and the recruiting process is so inefficient that if you have a model interviewing, that's going to make it a lot better.

There are a few things that make it super interesting. Interviewing someone is a very hard problem. You talk to the best founders in the world—even Elon Musk and Jeff Bezos—and they all spend a ton of time interviewing because figuring out who's the right person is a problem where you continue to have gains in quality.

This is a problem where we already have value now, but as you get better, it continues to be valuable. Investing a lot of money in solving that specific problem is valuable. If you get specific data—you hire people, see how they perform, and adjust your model—having that model is incredibly valuable.

The other thing is that you have this network effect: There's a platform, everyone is coming to the platform, and you have this marketplace of people. To me, it was like, at the time, they didn't have a lot of revenue. They had $1 million, but I thought, “This is a place where AI was going to do a much better job than humans, and you have a lot of defensibility.”

We spoke about the quality of revenue. Applying that to Mercor—you've had them on the show, and I love the guys—people were saying, “It's not revenue. It's not ARR.” I had smart people in my DMs saying, “Harry, that's not revenue.”

Is it revenue? Is it not revenue? Was I wrong? No, it’s revenue. It’s revenue for sure. And then on the defense, is it ARR? No, it’s not. You don’t have an annual contract that people are signing. It’s a run rate, right?

Yeah. Some people criticize it, and the criticism that I hear is, “Okay, you have this recruiter, and you’re building this platform, and a lot of the growth is coming from labs.” There were 3 stages of language models. I think the first stage was that you just took text from the internet, trained your model, and had GPT-3, which was very cool, but people didn’t want to use it.

Then you had RLHF, where models produce a bunch of answers. You take your base model, produce a bunch of answers, have humans pick what they like best, and then build a model that gives answers humans like a lot more. Then you had ChatGPT. That was the big ChatGPT innovation: RLHF.

Now we’re getting to this third moment in models, which is pure RL. I think this is the most interesting thing happening right now in AI. If you want to get a model to be better at something, you find a deep expert in that subject, and the expert creates a question that the model is not able to answer.

Instead of giving the model the answer, the expert creates a rubric to tell the model, “For any answer to that question, here’s how you’re going to rate it.” Once that’s done, the model produces a very large number of answers, and then you have this very large number of pairs of questions and answers that are graded. You use that to train your model, right?

This is the most interesting thing happening in RL, and it’s a big deal in AI because what this means is that we’re getting to a point where, if you’re able to create a benchmark for any task, then through reinforcement learning and just through throwing a lot of compute at it, you’re able to create a model that surpasses that benchmark, right?

So where does Mercor come in? It turns out that, to create questions that the models are not able to answer and to find people who are able to create these rubrics and evaluation criteria, it’s very hard. You’ve got to be an amazing interviewer to find these people. That’s the interviewing that Mercor does.

Mercor helps all the labs get a lot of the really best people. The criticism is, “Hey, is this recurring?” It’s not an annual contract, but the reality is that, for as long as humans are better than computers at any knowledge task, you’re going to need people to create the evals so that you can create these RL environments that will make better models, right?

Completely understand and agree. My question to you is: am I getting paid for the risk that I’m taking? When you have a reduction in revenue quality—revenue run rate versus enterprise ARR—and I’m just like, take HeyGen: a lot of it, like other people in the space, is experimental revenue. It’s not super-sticky ServiceNow revenue.

You need to be paid for the risks that you’re taking. Do you think we’re getting paid for the risks that we’re taking, given the price premiums that AI companies are demanding?

I think, in the case of HeyGen and Mercor, they were very cheap rounds, and I feel so lucky that I was able to—

Do you think $2 billion was cheap?

I was talking about the round that I did.

Ah, right.

But then, just as a framework to talk about the rounds in general, right? I think, as an asset class, the question is: is there a bubble? Are people getting paid for the risk they’re taking?

You think about Anthropic at $60 billion, or Perplexity at $15 billion. I think the reality is that AI is a big shift. I think AI is much bigger than mobile, maybe on the order of magnitude of the internet.

If you take a basket of the winners, it’s going to work out, because what I’m telling you is that, in the next 3 years, I’m sure someone will start a company that is going to be worth $1 trillion. How do you make sure you’re in this company? It’s like, okay, where’s the best place to find these companies? What are the interesting spaces that have great teams?

It’s working because there are these $1 trillion companies out there, and the early winners are not that many, right? I think smart investors are saying, “Hey, if software spend in the US is $1 trillion, but labor spend is $10 trillion, and now the AI companies are going after the labor spend, then the outcomes are going to be insanely larger than they were before.”

There are going to be huge winners, so we’ve got to make sure that we’re there. We’ve got to make sure that we’re in these companies, right? I think, on average, buying the winners today will work out. Are there some winners that are overpriced? In retrospect, the answer will be yes. Absolutely.

Listen, Bill Gurley has spoken before about passing on Google on price and his lessons from that, and I always think of that. The question is, if you see this shift in budget from labor to technology spend, something that I just worry inherently about is the increasing chasm between rich and poor and the inequality that comes as a result.

Even if there are 5,000 people in that company that’s worth $1 trillion, it could replace $1 trillion of spend on labor. Do you share my concern, and is that just an inevitable progression?

100%. What percentage of knowledge work as we know it today will exist in 10 years? If you define knowledge work as work you do behind a computer, I think it’s probably 1%.

The thing that I worry most about is, if you’re in college, if you’re someone coming out of college, what are the things that you can do that a model won’t be able to do in 3 years? There aren’t going to be that many things.

At the same time, companies will be more valuable because they’re going to reduce costs so much. People that own companies and people that own shares will get richer. Founders will get way richer. You’re going to have these $1 trillion companies being built by very small teams.

8. Why Replacing Humans is the Most Exciting Opportunity in AI

That’s a very destabilizing force, and that is something to worry about. All these things are controversial. You see these big leaders, the CEOs of the big companies, talking about, “Oh, no, AI is not replacing people. AI is actually augmenting people’s abilities.”

This is bullshit. It’s fully replacing people.

My favorite line that we get on the show is, “We’re just not hiring new people, but we’re keeping everything as is.” And you’re like, “Really?”

Yeah, you know. But net-net, I think this would be tremendous for society because I think companies are humanity’s best invention. If we need fewer people to create companies, we’re going to have more companies, right?

I grew up in Brazil, and Brazil is not a great place to grow up in. Then I moved to the US and the Bay Area. It turns out it’s a fantastic place to live. The principal difference between Brazil and the Bay Area is the quality of the companies, right?

Overall, it’s going to be fantastic for society. But then there’s a huge problem: how do you distribute that wealth? I don’t buy UBI, by the way. I think that is fundamentally lacking.

No, humanity requires purpose. If you add UBI, you remove purpose. People say, “Oh, we’re going to fish and we’re going to paint.” I promise you, that is not how human psychology works. That’s why we have addiction. It’s why we have drugs, gambling, and prostitution. Sorry, this is the dark side of humanity. We will not paint and write poetry.

Yeah. Well, I think we either have UBI or it’s a standard democracy. Imagine this society where there’s a lot of abundance and a lot of wealth, but the majority of people are not accessing that wealth. That wealth is very concentrated.

But it’s a democracy. Is that not what we have today?

It’s going to get way more extreme. The other thing is, I think it’s going to be much harder for politicians not to do what’s best for the entire population because the entire population will have a very powerful AI system in their pockets.

They’ll know. They’ll just ask that assistant, “Hey, this politician—the stuff that he voted on, is that good for the general population or not?” Then someone will come in with an agenda that is like, “Hey, big redistribution of wealth,” and it’s a very populist agenda.

I think people, if they very selfishly ask their AI, “Hey, is this politician better for me or not? Is this politician going to get me more material wealth or not?” then I think people who want to redistribute a lot will get votes, right?

I think that’s very destabilizing for democracy. A lot of people who are in control will say, “Hey, right now, you have people that are in control. Are they going to just cede control? Are they going to be less powerful? Are they really going to let go of all the influence that they have?”

I think AI is a destabilizing force because it makes the entire population way more informed and aware of what’s best for them. At the same time, it concentrates wealth, which makes the scenario more prone to the population wanting to get more.

But I think China is actually a very stabilizing force because I think the US right now has this external threat. The US cannot spend a lot of time in internal conflict because we’re at a very important moment in time where someone will get to powerful AI very soon.

It needs to be the US because we want to live in a world where powerful AI is controlled by the US. If we spend too much time in internal conflict, China will do it.

Right.

9. China vs. US: The AI Race

So, I think China is actually a very stabilizing force for the US. The best way to unite is to have a common enemy. Exactly. So, I agree with you.

That said, I don’t feel the US is winning the race against China. I think China is absolutely crushing the US. Europe’s not even in the AI race. When you look at the investments that China has made into its education infrastructure systems, the depth of its talent, respectfully, and the work ethic that they have, it just completely destroys anything we have elsewhere on the planet.

Yeah. I think China has a better work ethic, and they have more people. So, they have more people working harder. But I’m still so bullish on the US because there’s so much knowledge here.

I think our advantage is that the true network effect is a city, right? We have Silicon Valley. You say, “I think China is winning.” I think, by all measures, the US is winning. If you look at all the AI products, ChatGPT is the most-used AI assistant in the world.

Sure, you can talk about DeepSeek and this and that, but the reality is that ChatGPT is still the best product out there.

Yeah. I mean, what’s better than ChatGPT? Listen, I completely agree with what you say. I do not trust or believe anything that China says, and I think they have an armory of AI-related weapons that they do not show the world, that they have ready to go.

I think TikTok is a weapon of consumer data aggregation and acquisition. I think they are so strategic and smart in how they’ve leveraged Shein and Temu as well. Their infrastructure investments in Africa are something I think people don’t spend enough time on.

It’s a shit answer for me because I’m not really giving you one. I’m just saying I don’t underestimate them, and I know that they are doing strategic things which I don’t even know about, which I think we’re not doing.

No, I agree. It’s a very serious competitor. I 100% agree with you. I think we’re still in the lead, but we cannot afford to waste time on anything else. We cannot afford to waste time on internal disagreements and stuff like that.

Can I ask you? I think it’s very easy for us to get excited by transitions, by improvements in human lives, and how we won’t be doing mundane tasks that we all hate doing anyway. But then it’s very easy to get caught up in the short term.

Do you think we overestimate adoption in the short term and underestimate it in the long, or do you think this will happen much quicker than we think?

Yeah, I think there’s a lot of wisdom in saying, “We overestimate what we can do in a year and underestimate what we can do in 10.” I think there’s a lot of truth to that.

I still think, in AI, when you look at how much has happened—three years ago, we didn’t have ChatGPT, and now there are close to a billion people that use it every month, and they have a meaningfully different experience. I personally think that, in the next five years, lives will be meaningfully changed. I think AI is going to happen way faster than people believe.

What crazy thing, or what thing, do we do today that we will look back on and go, “That’s crazy”? Do you remember when people said, “Hey, you’ll find your loved one on a dating app”? Shit, you won’t. “Hey, you put your credit card details into the internet.” No way.

I think the thing that will happen quite soon is that we’re going to have an app, and you’re going to wake up, look at your app, read what it says, and just do it. The app will tell you, “Go do this.” You’re not going to understand, but you’re just going to do it, and it turns out that you’re going to be happier after you do this thing.

Over time, you just learn to trust the app. It’s going to be that complete reversal. You’re not going to apps to make it easier to do the things you want to do. For example, you go on Instacart to get your groceries. You know what you want to do; you go there, and it just makes it easier for that to happen.

It’s like you go to the app for the app to tell you what to do, and you just blindly trust it and do whatever the app says.

How far away do you think that is?

We see glimpses of this, right? There are companion apps, like Finch, or these self-care pets. It’s like, “I’m going to help you access—I’m going to be your accountability buddy. I’m going to help you exercise every day, make sure you’re drinking water, and taking your meds.”

I think it starts with these simple things. You open up your app. For example, I try to work out every morning for 20 minutes, but a lot of times I wake up and I just don’t have the energy.

Imagine you wake up and it says, “Hey, it’s your 20 minutes. Remember, if you do that for X days, you’re going to be better.” Then you just get into these routines. These apps are helping you form routines. They’re telling you what to do, and it will get to a point where AIs are going to be smarter than us.

AI is going to be smarter than us on most things, and we’re just going to learn to trust its judgment, right? It has context on you. It knows what your long-term goals are. So, you wake up and it says, “Go spend an hour reading this book,” and you just do it. “Go meet with that person,” and you just do it.

10. Why All Students Today Should Study Computer Science

It’s funny. I got asked the other day, “What advice do you have for me? I’m leaving school, about to go into university.” I said, “Very simple: become a brain surgeon or a dentist. I promise you’ll be safe for about 20 years minimum.” What advice would you have?

I have the opposite view. I think you should go study computer science.

Really?

Yes. To me, it’s completely antithetical to what everyone is saying.

That’s a fascinating piece of advice.

Yeah. They say, “Hey, go into computer science.” To me, it’s very—I feel very strongly about that.

Before we had calculators, we had human calculators, right? If you saw the Hidden Figures movie, there are typically women, and they’re calculating these tables. They’re people who are just doing math.

Then calculators were invented, and when calculators were invented, it wasn’t like, “People shouldn’t study math.” No, people should study math. If you study math, the reality is that you’re studying math in high school and you’re studying math in college. The first few years of college, you’re doing stuff that computers can do, but it doesn’t matter because, first, you’re understanding the technology, and, two, there’s so much transfer learning.

I think when we train models to code, they get better at a lot of other stuff. So, I think the reason I advise people, “Hey, go into computer science,” is that the transformation right now is a technology transformation. It’s an AI transformation.

Go study computer science because even if you’re not going to be a developer—and you’re probably not going to be a developer—just having the basis to understand how that’s done is going to be extremely helpful. Then the transfer learning of just being very logical, being able to take a big task and break it down into small tasks, and being very rigorous—I think that will continue to be very valuable.

When you look at the AI companies of the last 18 months, you’re in several absolute bangers. Which one are you not in that you had the chance to invest in and you just missed?

We participated in one round at Cursor. I think the coding space is very interesting. I like that. That’s an interesting one.

You participated in one round. I’m just interested: Benchmark are famed for leading the Series A. How do you think about the partnership now, about flexibility around stage and whether you have to lead? I obviously know Howie well, and he did Airtable. I think the GPs have flexibility around how you think about the rigidity of, “No, we have to lead the A.”

I think rigidity is nonsense, and frankly, what defines Benchmark is that we want to be a partner to the most important companies being created, which is similar to most venture firms out there.

I think the difference is that we want to be the first call to every entrepreneur that we work with. In order to do that, the easiest way to—

Can I be a dick as a friend? Every VC says that. I don’t even want to be your first call. Your mom should be your first call.

No, I think that’s fair.

I want to be your worst call. When the shit really hits the fan, call me.

Yeah.

It may never happen, but at that terrible time, I’m the person who’s going to help you.

All else being equal, I think everyone wants to be the closest partner. But I think the part that changes is the “all else being equal,” right?

I think it’s, “Hey, we’re willing to take a lot of trade-offs in order to be a very close partner.” So, why does Benchmark make very few investments? It’s because we spend more time with our companies. And you’re saying that all investors want to be their first call.

I don't think it's true, because the reality is, if you're a financial investor, you don't want to be the first call. You've made your investments, and you want to find the next company. There are a lot of investors who are really smart investors who say, “Hey, I hate the board work. I want to write the check.” And there are some investors who say, “That's not the product that I have. I put my check in, and that's it.”

Do you think an investor can move the needle for a founder? Keith Rabois says that the best founders don't need you. If you call the people that I work with—and you should call them—I think you did. I think you can, right?

One thing that made me very happy is that a few months back, Peter came to me and said, “Man, I'm so grateful that you've been a partner on this, and you're the closest thing that I have to a co-founder outside of the company.”

So, can you make a difference?

I think it's actually very hard to make a significant difference from the outside. But the way you do it is that founders need a thought partner, and founders need a way to access all the knowledge in the ecosystem. You can catalyze that.

The good founders have access to a bunch of people. The good founders don't work in isolation; they have access to a bunch of people. But the difference—the thing that you can do as a board member that's very hard for other people to do—is that I've been on the Brex board for, I don't know, 7 years, and for every important hire that he made, I discussed it with him. I've had so many close calls with candidates that, when he has a problem, he calls me just because I have way more context than other people do.

So, how do you add value? What's the way to add value? It's like, “Hey, you have a lot of context,” but the thing is, it's very costly to have context. That's why we do very few deals: getting context on companies takes a lot of time.

If you're flexible on your insertion point, maybe participating in a crossover round or doing a later round, do you have the ability to add context, even though you're probably not on the board?

Not at all. When we're participating, we don't have that ability, and that's why we don't do it. Normally, the reason we do it sometimes is, “Hey, maybe this is a way to start building a relationship with the founder, and over time we'll grow our position and build a relationship.” But the reality is that if you write a small check and you're one of many, you don't make any difference. It doesn't matter how smart you are.

Did you feel the weight of the Benchmark GP role? When you come into any investing role, it's a lot of responsibility. Respectfully, you're a GP at Benchmark. As your first investing role, I know you're a billion-dollar founder, but it's still a steep position to come into.

The thing that I love about Benchmark is that I came in and the partners said, “Hey, you're coming here to co-found the firm. There's a lot of tradition, but a venture firm that is sticking to its tradition is soon going to be irrelevant.”

We believe a lot in creative destruction. You almost feel like, “Oh, okay, really?” Then I met with the founding partners at Benchmark, and they said, “The only way we continue to be a great firm is if we have no attachment to just the way things were done.”

I actually think this is a big advantage for Benchmark. Because we don't have any process and we're a small team, it's easier for us to change things.

How do you think venture is changing? Think about a couple of decades ago: the role of a board member was governance. Capital was so scarce that, once you put capital into a company, a lot of your job was managing the downside. It was, “Hey, make sure that this guy isn't going to lose all your money.”

The reality is that's understandable, because there had been fewer big outcomes in tech. So people who invested were kind of afraid. Fast-forward to today: everyone knows investing in technology is a great thing. People should do that, right? It turns out that it's more important to maximize the upside than it is to manage your losses.

The way this changes the role of a board member is that you're not as concerned that some of your investments are going to lose money. Your role is not governance. Your role isn't to say, “Founder, you can do this or you can't.” Your role is to maximize what the founder wants to do.

So it goes from oversight to catalyzing and amplifying the founder's ambition, really. I agree with you, but officially, the job of a board member is to have a fiduciary duty to shareholders. Fiduciary duty to shareholders isn't always the same as maximizing a founder's ambition. How do you think about that?

If you're taking a lot of decisions in isolation—“In this case, I believe that if we do this, it's going to be better for the shareholders”—but you look at it in aggregate, your policy is, “I'm going to choose the best founders, and I'm going to trust them.” My fiduciary duty comes in at the moment where I write the check. I'm underwriting that person: “Do I think this guy will be a good CEO?”

Once you're there, I'm just here to enable this guy. Even when I disagree with him, I'm just going to help him do what he wants to do.

Have you ever lost faith in a founder? Describe what “lost faith” means.

It's when you no longer believe that they're the right person to drive the company, and you no longer trust that the decisions they make are the best decisions.

That's the thing I find hardest.

The reality is that I haven't done a ton of investments. I've done a couple of angel investments where things didn't go well. One of the things that I've learned after joining Benchmark is that I work a lot with Peter. Peter sits on 2 of my boards, and I sit on 2 of his boards, and you see these companies go through the ups and downs.

The thing that I learned is that we don't give up. As long as the founder wants to do it, even if it's completely uneconomical, we're there. I was talking with Peter about a company that wasn't doing well, and it was like, “I know what we have to do.” A lot of venture capitalists would say, “Just cut it loose. Forget about this thing.” We should double the frequency of the board meetings, because we're here for as long as the founder wants to try.

You made a commitment to the founder, and from that point on, you try to be true to this unconditional belief in the founder.

You mentioned Peter there. What would you most change about the current process Benchmark has for investment decision-making?

I think our process is a very lightweight process, which is very good. The way we work is that you find a company, and if you like it, you need to get your partner's opinion, but you don't need to get your partner's approval.

Do you bring them into early meetings with the team and founders? Do you hunt as a solo investor? Do you tag-team?

It's all optional. The one thing that's not optional is that you have to present the idea to the other partners. Everything else is optional.

The way it typically works is that you meet a founder, and then you do a second meeting where you bring another partner in so you can have a discussion. If you know you want to do it and you like it, typically you bring the company in, everyone talks to the company, and then people vote. But it doesn't matter what the vote is: if you really want to do it, you can do it.

Why vote?

The vote is just a temperature gauge. The vote is to disambiguate what people actually think. You get a clear signal, because you're going to bring a company in. It doesn't matter what the company is; there's always a ton of reasons to do it and a ton of reasons not to do it. Then people argue both sides.

You force them to give it a number, and that number carries a lot of information. We vote from 1 to 10, and you can't vote 5.

I asked Peter, “What should I ask him?” He said, “The one question I'd really want to know from Victor is: If you look back in 10 years, in what ways do you want to have meaningfully changed Benchmark to prevent its extinction?”

I don't know if “prevent its extinction” is the function we're trying to maximize. I think the function we're trying to maximize is that there are going to be 10 trillion-dollar companies created in the next decade. How do we make it so that we're in a large number of them and we have a deep relationship with those founders?

I think that's a maximizing function. To me, it just works backward from: How are you obviously the right choice for the founders that have a ton of choices?

And right now, at Benchmark, when we want to do something, we typically do it. Typically, we have the chance to partner with the companies we want to partner with. But because we’re a small firm, there’s a lot of stuff that we don’t even see, and there are a lot of rounds that people are raising that don’t look like a Benchmark round. I think that is the thing that is most important for us to change.

A lot of times, people think, “Oh, this is a Benchmark round,” but this is not a Benchmark round. I think what it means is that it needs to be clear to everyone: what’s a Benchmark round? A Benchmark round is when there’s a founder who is very capable, going after a very big market. As long as you have that, that’s a Benchmark founder. It doesn’t matter if you’re raising $5 million or $200 million.

Is fund size now a constraint?

The constraint always was: be useful to everyone. That was the craft of Benchmark. In today’s AI landscape, the rounds we see are so big that you need $25 million to $50 million to start. If you want 15% ownership, you’re putting 10% of the fund out on the first check.

Yeah.

So, my first check at Benchmark was a $55 million check into HeyGen, right? This was my inaugural $55 million. Yeah, this was my first check—a $55 million check. And, yeah, because then, like, hey, we’ve got to be true to—

No, I know. But, dude, that’s 10.5% of the fund if it’s a $500 million fund.

It’s actually a $600 million fund.

11. Why Portfolio Construction is BS

That’s a lot. I mean, that’s amazing balls, dude.

And, you know, I think a lot of funds think about portfolio construction and this and that, but the reality is that our LPs are the same LPs in all funds. If you want to do one $3 billion fund or five $600 million funds, for our LPs, it’s actually the same. So I think fund size has never been a constraint.

I think where we can make a difference is in the beginning. We’re early, and we work best with founders who want to have a relationship. Some founders are like, “Hey, I do think it’s valuable to have a sounding board that is going to have a lot of context on the business.” If you believe in that, then Benchmark is a good fit.

Most early rounds are less than $50 million. But we’ve done things that are later—we’ve done larger checks—and sometimes, if we’re going to do a larger check, we do an SPV. The check size is not a constraint if you’re talking about inception rounds. It’s never been a constraint.

Does Brett Taylor come in and pitch to the partnership like that?

He’s a special dude. Sierra is a special company. In Brett Taylor’s case, I think he had dinner with the partners. It was more of a conversation than a pitch.

Brett Taylor is a good example. The guy is so accomplished that he could raise from anyone and do whatever he wanted. But I think he valued the relationship with Peter and said, “Hey, I think it’s going to be valuable to have you along this way.” Because it’s an equal partnership, these people are going to be his partners in the company. So it was much more about getting to know each other than, “Hey, I’m pitching you.”

That is an insane one. Brett could fund it himself for the rest of eternity, in many respects, but he values Peter so much that he brings him in at that stage. That’s pretty special.

12. Quick-Fire Round

Yeah, and frankly, I think it’s a smart choice. I think the best people want to maximize for, “Hey, how do I get my conditions—my initial conditions—right?” How do I get the smartest people around the table? Because, in the end, the initial conditions affect so much of what you can actually do.

Dude, I could talk to you all day. Do you mind if we do a quick fire round and then I’ll let you get on with some proper work? What do you believe that most people around you disbelieve?

I believe that pretty soon, this thing that we talked about is going to happen: we’re going to have an app, and we’re just going to wake up and do whatever the app tells us to do. In a way, we’re going to be obedient to machines, and we’re going to love it.

13. What Makes Peter Fenton One of the Best Ever

What’s the single biggest lesson from working with Peter Fenton?

I think the thing Peter does better than anyone I’ve met is that he understands people and can get people to open up. A lot of times, before we decide to work with a founder, Peter takes people to dinner. I’d introduce people to Peter and take Peter to dinner with someone, and even so—for example, I took a COO to have dinner with Peter, someone I’ve known for many years. At the end of that 2.5-hour dinner, I felt like I knew the person a lot better than I did before.

Peter is able to make people open up in a way that has a real effect. People leave the dinner and say, “Wow, I just had a really special dinner.” I think it’s a superpower, and I think it’s kind of known that this thing happens.

A lot of times, if I want to spend time with someone and it’s hard to make it happen, I’ll say, “Hey, do you want to go have dinner with Peter and me?” Most times, people say yes. They have a fantastic experience, and then they tell more people: if you get the invite, you take it.

I love the fact that no one’s like, “Nah, sorry, I’m watching Netflix.” I’m sure it’ll happen at some point, but that hasn’t happened yet.

14. Why Duolingo Will Be One of the Most Valuable Companies in the World

No, I don’t think it’s going to happen. Don’t worry. That’s just too funny.

You can buy and hold 1 public company stock for 10 years. What do you buy and hold?

Duolingo. AI is going to change people’s lives in really important ways. One really important thing is that the way we’re going to learn is going to be through AI teachers, and Duolingo is very sneaky—in a good way. They’re starting with this language app, but they’re making your AI friend that teaches languages. Then they’re going to teach more stuff.

Luis, the founder, is a total genius. I think he’s going to make a free AI tutor for everyone. That’s going to be incredibly important to humanity, and it’s going to be a very valuable company.

You can invest in 1 seed fund and 1 growth fund. I’m not choosing one because that’s tough for you. Who do you choose?

There are so many great ones. I’m actually an investor, so there’s an objective answer to that. I’m an investor in Green Oaks. I think they’re pretty good.

Neil Mehta, just take me already. God, this man is brilliant.

Yeah, he’s amazing. No, I really like him. On the early side, I’ll go with Conviction because I think Sarah’s great.

Final one: when you think about the next 10 years for you, if we do this in 2034, where do you want to be then? What do you want to have achieved? What does Victor Lazarte look like then?

I want to be an important part of the most important companies being created. That can be as an early board member, and that can be as a co-founder, but I want to look back and say, “Hey, AI is going to solve so many of humanity’s biggest problems,” from education to security and companionship. These companies are being created now, and I want to be an important part of their story.

Dude, I’m so glad that we had this schedule that I completely ignored. I think you saw that it was a much more free-flowing discussion. I so appreciate you rolling with the free-flowing discussion. You’ve been fantastic to have on. Thank you so much for joining me, man.

Thank you, man. This was a lot of fun.