Vinod Khosla and Keith Rabois on Building and Investing in Enduring Companies | Ep. 40

I've invested a lot in AI, so I, yeah, I love your quote about AI and how you'd have done it differently if you hadn't joined Vinod Khosla Venture.

Yeah. I don't hear it. I don't know.

Before I joined KB, I joined KB literally 2 years ago, basically this week. Yeah. I had invested in zero.

Rejoined.

Zero. Rejoined.

Yeah.

Rejoined.

Rejoined.

Yeah, I do. That's not good.

And literally had invested in zero AI companies before. Interesting. Since then, in the last 2 years, I'd say it's about 70% of my investments are AI. Mm-hmm., and had I not rejoined KV, I think I would either miss the whole wave and been completely irrelevant or been reckless.

Keith and Vinod, I'm incredibly excited. I got to do this with each of you individually last year. First thing I wanted to do coming into the new year was ask two of you to come on together. So thanks for doing it.

Pleasure to be with you.

Yeah.

Obviously I follow both of you a lot, you know, sort of on online and watching a lot of podcasts you've done and just knowing you over time and I've never seen you together, um, in this kind of format. And so I was really excited to set this up. One of the first things I want to get into is like how the two of you work together, because I think like it's rare that you have two people who are both super individually accomplished at a venture firm working side by side. You've done it for many years now, and you're obviously very different people, but there's like a lot in common. Like I see you as this like Venn diagram with a lot, you know, in the middle, but you also have your own styles. So I guess just to start, like, what's the like texture of your day-to-day working relationship like? Like how do you guys operate together? How do you communicate with each other? Like, what does it look like just like all the time?

You wanna start?

Sure. Well, we first started working together when, uh, Vinod joined the board of Square. So, and you know, I learned a lot of things from Vinod.

Actually, Slide.

Well, we worked together at Slide, but that was more intermediated through Max. And I would, I would hear about these meetings with Vinod and all these ideas and these grand theories about how we should rewrite the business. But it wasn't like hands-on. I actually worked with another one of our partners, David, at Slide, uh, very directly. He told me my first revenue model was terrible. And so Dave's a little bit of an acquired taste.

Sounds like David.

Yeah.

He was like, this "Your revenue plan is very mediocre," was the exact quote.

That's great.

Which turned out to be right. Um, but in any event, while he was on the board of Square, he taught me a lot of things, including the most important precept of the team he builds, the company he builds. So when I was considering being a VC, it was a very natural fit because a lot of the contributions that Vinod had at Square resonated with me, and I could see the style of KV and how that translated through my brain, and that I felt like it would be a really good pairing.

What was it like for you? So like, you know, when you're starting to work with Keith, like, could you tell immediately that stylistically it was like what you like? Like, how did you know?

For me, it's really one thing. It's first principles thinking. If you can do first principles thinking, it's easy to know where you agree and where you disagree. It isn't this hand baby thing.

Yeah.

A, B, and C, then we can debate those three factors. And it's worked out very smoothly. It's seldom we grossly disagree. 3 will even come down to if X were true, then this is a good decision or a bad decision.

Yeah.

Which is incredibly helpful. Like there's one specific investment I remember last summer where I couldn't actually decide what to do. It was close to the line, wasn't quite sure. And then Vinod said the key attributes of the founder that really matter for this are 1, 2, 3. And then our, our junior colleague John Chu and I were like, well, on that, on those 3 dimensions, this founder's AAA. So it made the decision really easy 'cause he was able to isolate the key variables for that particular company.

Yeah. Do you guys like, um, get into like strong debates over ideas, or have you mind melded so hard at this point that you don't even need to as often? Or— because it's hard for me to imagine either of you shying away from a very— like, you're obviously both going to say whatever you think all of the time.

But I think that helps the direct style, which you've talked about for years.

I, I've always said I prefer hypocritical— yeah, uh, brutal honesty to hypocritical politeness.

I can't imagine you dancing around a topic with each other.

No. And it doesn't matter whether it's internally in a debate on a date, uh, uh, or on Twitter, you know, it doesn't matter. Being very direct saves a lot of hassle. And once you have that culture, nobody's guessing at what you think. And that's worked really well within the partnership. Nobody's ever guessing what you think or why you think that.

And then externally, I think one of the reasons why we pair well with really ambitious founders is they appreciate clear communication. Succinct and direct communication, they process extraordinarily well.

On you two working together, like, if you had to sort of like guess, like, I don't know, like the pie chart of the time that you guys are spending talking to each other, how much of it is about existing investments, new companies, operating the firm, anything personal?

So nobody ever discusses operating the firm very much. You don't discuss the ventures?

Interesting.

You know, It's almost no— I would guess it's far less than 5%.

Is that because it's so clear how it works or because it's like, yeah, maybe around if you include hiring, like then there's an allocation to assessing.

Yeah.

We'll call that separate. Yeah. Out of curiosity, is that because you think that it's like, it's already so clear there's nothing left to discuss or is it just so much relatively less important than the work of investing itself?

Well, first we have a lot more fun investing than managing. And the firm doesn't take much management. Honestly, other than comp once a year, there really isn't a lot of— and hiring that we talked about. There isn't much in terms of— I can't remember when we had a strong policy disagreement.

It's also, you know, at the end of the day, Vinod is energized by investing in the future through technology and founders, and I'm energized by pairing with people who want to change the world, which is roughly similar. And so the management part is a distraction. Some extent from those two core activities, which are exciting.

So what else?

We spend almost no time with LPs.

Yeah, that's nice.

I mean, I—

They have to edit that. Just kidding.

I, I, look, I don't, I don't mind. Uh, I, I spend less time with LPs than almost any other senior partner. And I think that's generally true of all the senior partners at Khosla.

Well, you're at a place where you probably don't need to spend that much time.

And frankly, entrepreneurs are a lot of fun to work with.

Yeah.

So are you spending most of your time talking about new companies, existing companies? Is that basically all of it?

Actually both. I mean, we take the current portfolio very seriously. Every single Monday meeting starts with the current portfolio before we ever talk about a new opportunity because we're in the build a company business, you know, sort of that's what we focus on is we're an investor. We're going to be a partner for 10, 20 years. How do we help the company achieve its highest ambition, highest potential? And so we start literally intentionally that way with the portfolio before looking elsewhere.

So the funny thing is, in 40 years that I've done venture capital, I've not once called myself a venture capitalist or an investor. Always say I'm a venture assistant to entrepreneurs trying to build companies. And that's what our website is focused on. And that's what we talk about mostly internally. How do we help a company change its trajectory if the potential exists to change it?

How different do you see today, like the sort of ethos of like new young investors versus like when you were getting started? Because like right now it is such like a thing to use.

You know, they, the, this young and old doesn't matter as much. What matters is have you built companies and earned the right to advise an entrepreneur? I think most people who advise entrepreneurs haven't earned the right to advise an entrepreneur. Almost all the senior partners in our firm have earned the right by helping build a company, being inside a company, having empathy for the founder. So I think that's a pretty distinctive feature of how we think of our role. A lot of firms just want to be nice to founders and it hurts the founder. 'Cause yeah, it's like I, I say, it's like saying yes to your kids all the time, no matter what they want. You, you want about, they know you love them, but you're trying to get them to be the best they can be.

Yeah.

That includes pushing them to be the best they can be. And I think we both agree our business is much more about helping the entrepreneur build a successful company than about investing.

I'm newer to this, obviously, but it seems like there was some moment in time where like the dominant marketing strategy for VC firms flipped towards just like extreme founder-friendly, whatever that means. I don't know if it was 10 years ago or 15, something like that though. You guys probably experienced this going through being on one side of it and then the other. I kind of always lived in that world, but like, was there like a moment or like a period of a few years where it just like became the strategy for some reason for VCs to just go full?

You know, our goal has always been what's good for the company, not what sounds good or get, uh, what will get us more referrals for the founder next time they need to refer. So I think this hypocritical politeness, which is pervasive in our business, is really bad for founders. And when a founder selects for that, they're selecting, they're generally a weak founder. Strong founders almost always select for the best feedback they can get. And also know how to say, "No, thank you. I disagree with you." I think that's a really, really important characteristic. One of my favorites is this. Some founders, I don't know the names, there are two founders did something called founderschoicevc.org or .com.

Some survey of founders.

Some survey of founders. And what I love is they wanted to avoid the hypocritical politeness. So they said, only the VCs that you've worked with can vote on you, on the VC firm. And you can't say, we have these 3 investors, they're all great. They forced them to rate them head to head, like a chess Elo rating. And I'm very proud, we're sort of at the top of that list among 400 VC firms and hundreds and hundreds of votes. I think it's the most important survey for me to know that in retrospect, our founders prefer us.

Mm-hmm.

And we prefer backing the same founders again and again. Like our LP decks always have a huge component of how many repeat founders have come back to us to work with us.

Keith, I'm actually really curious, 'cause like obviously Founders Fund is an amazing firm as well as Khosla, and the ethos is the same in the sense of like we want to do what's right for the company, but I think like the pathway to get there is like the opposite.

Yeah, I think the goal of finding bold, ambitious founders and companies is very similar. I think the way we actually practice what we do is very different. Our craft is to be the partner in building the company. Like, I look at my role as being the consigliere to the founder, and like sometimes the consigliere tells you that's a bad idea, and sometimes they tell you that's a great idea. They help the principal, and they're not confused about who's the CEO and who's the consigliere. That's what— how I think of my role. Founders Fund thinks their role is to provide the capital and get out of the way. And if you have an idea where they might be helpful, please call.

Yeah.

And we'll do everything we can to try to help. Right. Proactive versus reactive.

Totally.

That's exactly right. Very, very similar goals in betting on really bold ideas, whether they're popular or not. Whether they're on trend or not. They've done an incredible job of that. And we like to think we do that also.

Yeah.

It's almost like you can imagine the conversation would play out like, you know, on one side it's, well, we should back entrepreneurs that like, don't need help. And then the other side of the conversation would be like, well, yeah, but even amazing people can still be helped.

Of course.

Like, you mean, like, even the best baseball players have, you know, coaches. I knew I was going to sports for you.

Or, you know, they have shooting coaches even, like, they're very dedicated, or fitness coaches. Yeah. Like, there's almost no profession where the best at what they do doesn't have an advisor, coach, mentor.

Yes.

Yeah. Look, take somebody, a really strong founder like Max Levchin. I've never been on the board. The company's gone public. We've distributed. Great outcome. We are the first investors. To this day, Max and I do quarterly phone calls because he wants my help and advice and second opinion or something he's thinking about or always looking for areas I'll prod him in to think harder. Like you're ignoring this or that.

I want to talk about like, so what, this clicking into the source of like great founders, great companies, because obviously that's like the center of like the work. I posted this, you know, that I was going to have you guys on the podcast. Somebody said, you know, we want to hear about this, but Keith, you can't say the thing about comparative advantages. And, you know, you're like, but it's true. And I think it is true. But like, I want to try to click in as closely as possible to like what makes a great founder. And I think one of the things that stuck with me was you both have said that you guys are basically always aligned on like the read of the founder. You might disagree on a market or if there's some business you want to be in, in, but like more or less you think the same thing about if a person is a great founder.

Very rarely is there significant divergence on the assessment of the founder. I can probably name 2, 3 examples in like 8 years.

Yeah.

What specifics can you sort of like describe?

Well, I'll give you my formula and yeah, feel free to, you know, edit. So my, mine is one of two traits. Either I meet a founder and on some dimension they're the best I've ever met in my life. It can be different. They can be the smartest person. They can be the most tenacious person. They can be the best assessor of people. They can be the most strategic. They can, they can, it's just like, oh my God, top 1 basis point on some dimension. And because what I'm trying to do is when you're, we do mostly first institutional capital. We want to be as bold and as early as possible. What I'm trying to find out is, is there a non-zero chance that this person can change a vertical or the world? Like, that's really it.

Is it one of those two things?

Either the world or like at least illegal industry.

99% of humanity is not going to change and reinvent an entire industry, let alone the world. And so it's like, what's— is there some probability? Usually the people who succeeded at that have some trait that's just like, oh my God, your ears like perk. The other exception is they have a Venn diagram overlap of traits that you don't see in common. So for example, Max Levchin, I've talked about this before, but literally when I met him in December 2000, Reid Hoffman came up to me and said, you're getting ready for your first one-on-one with Max. Max is a first-rate technologist and a first-rate business mind. There's less than 5 people in all of Silicon Valley that are that. Reid was dead on. 25 years later, it's still true. There's less than 5 people and Max is one of them.

Yeah.

And that's led to his trajectory. So if I see these traits, Jack Dorsey, who we've both worked with, is actually a pretty damn good design mind.

Yeah.

Pretty good technologist and a very good business strategist. He has 3, which is also why he's been very successful.

So, okay. So you meet, you meet somebody that you trust, refers somebody, you, great, I'm going to meet them. You're in the meeting with them for an hour. Are you trying to pull that out in that meeting? Is this work happening outside the meeting?

I'm usually pulling it, usually if it's so strong. It usually shows up in 3 minutes. Like, literally, you meet someone who's the smartest person ever. You did, like, like, you just feel this energy.

Aren't there some less obvious traits than, like, super smart though?

Yes, there are.

Like grit or something like that.

Exactly. But, like, so, for example, one of my favorite gritty founders told me the story of when he was working at Uber, his team in this foreign country that he just joined, like, because he was a launcher, was going to run a marathon on Saturday, and this is like Thursday. And he's like, I wanna be part of this team. Like, I wanna fit in. I'm gonna do it. I haven't trained at all. Didn't own a bike. So what'd he do? He rented a city bike and ran the tri— and did the whatever triathlon on a city bike.

That's crazy.

That shows so much grit. Like, it's like, that's all you need to hear. And then, and then his co-founder, like, wouldn't, uh, finish second in the spelling bee when he was in high school and passed out, like, due to, like, stress. So he didn't quit his senior year. He went back and tried to win it. Wow. Like, I was like, there's a dimension there that you're like, you don't hear very often.

So to get some of these like ones that don't show up in a, like a live meeting, are you, do you have certain like things that you're prodding towards? Like, are you asking more about life than business?

Like, no, I actually don't do the— Daniel Gross style thing where tell me about your history and your sibling and you know, that it does work, you know, for the people, it definitely does. I just like, tell me about your company and like, why are you doing this company? And it just, It just shows up somewhere. There's just a spark. They can't help themselves.

Have you tried the Doug Leone thing and it just like didn't work for you?

I never really tried it. The mechanical—

Give it a shot.

It's like when you assess people, Vinod's interviewed executive candidates for, you know, literally 40 years for companies. There's a mechanical way of assessing or doing an interview where you go, you know, experience by experience. Like, why did you leave? Yeah. Why'd you leave? What'd you accomplish? What was the biggest challenge? What are people going to say about you? And then there's different ways when you interview people that's a little bit more freeform. I'm definitely in the freeform version of interviewing.

Do you do a founder meeting similar to that, or is it a whole different thing for a founder?

Um, you know, every situation's different. I want to add a couple of other things to what Keith said. I think the most important thing is exceptionality in some dimension, whether they're going to be a good CEO or not. But related to that is two things. In the areas where they're not good, perfectly good to back a founder if they don't know an area. Often it's a professor or something. That's where venture assistance comes in that I was talking about, where we can help them be a complete founder. But also the, the, the thing I want to emphasize is to me, what's key is the learning rate of the founder, how open-minded they are to new ideas. How good they are at rejecting bad ideas. This, so let's say too many founders just take every idea and try and execute. If a person listens to me all the time, I'll almost never invest with them. I know they're not critically examining. I often take positions I don't believe in just to test how the founder's thinking about something. I have a document I put out, this public document on how to do an interview. It is not, it is very much like it's free form, but it's, everybody knows what answers to give in an interview. So how do you get past that? I have an internal document I give to only our founders on how to interpret the answers they get. If I put them out publicly, then every candidate would understand how I'm interpreting. So I can't put that publicly, but the first half I made it public, like Here's how I assess somebody. So I have a pretty clear style, and it's usually about putting people in a situation they haven't been in. It's not soft, tell me about your life history, because people know how to storytell, and the best storytellers may not be the best candidates. So you have to get past the obvious answers, like I did this in that company, or I opposed that in retrospect. Mm-hmm. Uh, I think it's very nuanced. So that's a great example and ties back to the team you build is the company you build, not the plan you make.

Mm-hmm.

Because the right team will evolve the plan to the right thing. And your initial plan is seldom the right plan. But if you can help pick the right team in a new context, which is you're trying to do something bold and different. Mm-hmm. Uh, That's, that's a pretty critical part of it. So both part of what we can do to help a founder and why most people are not qualified to even interview candidates. I've seen such bad specs for what the company needs. Most of the time they're wrong on what a company needs to hire.

Well, you have a great example in marketing specifically where you talk a lot about like the zero to end marketing versus—

Yeah.

I don't know if you want to amplify that.

Yeah, I'm curious.

For example, most startups are trying to create a brand. If you've been marketing at Cisco, you know nothing about creating a brand. You know how to incrementally sustain a brand. That's a very different skill set.

I was going to say something even worse than that, but yeah. You know how to like make this quarter's number look a certain way so that your manager—

I got in trouble for saying if you're at Cisco more than 10 years, you're not qualified for a real job in the entrepreneurial world.

Yeah, we don't— no, it's, it's generally true.

Yeah, it's hard.

They're just so different. They have like trying to figure out like they have nothing to do with each other.

Mention that leads to success for what the startup needs.

Yeah.

And then how do you find that trait or that proven ability?

Yeah.

And make sure there's a Venn diagram overlap there, which first is diagnosis and then it's like assessment.

Mm-hmm.

And you know, this is the kind of place where venture firms are pretty different. We'll agree if somebody can advise somebody on marketing. But if you look at a spec a board will put out, it'll be cookie cutter, then X business to get somebody from X related businesses. It's just a terrible idea.

It's actually a source of a lot of like management mistakes, which is like, it has to like look good on LinkedIn for the board. And that's a disaster.

Oh my God, the worst thing ever.

Yeah.

And it's really hard if you're like a young founder to like stand up to that.

This is the world where you need to get advisor or board member.

Sometimes just giving the founder confidence.

Yeah, permission to trust their own gut.

You're not wrong. Just saying that when they're getting a lot of pressure from somewhere else, like especially the first-time founder, just saying like, no, you're probably more likely right, actually. And then, then all their instincts kick in and they have enough confidence to just say no.

Outside of this idea of like exceptionalism and like finding this dimension of special, like, you know, basically what you're saying is like if somebody's like B+ at everything That's not a good investment for you.

Usually an A+ in either incomplete— like Vinod likes the term incomplete.

I like that.

It's a work in progress. Really a really good formulation. The rate of growth is hard to tell in your first meeting. Yeah. That is one of the hardest things because you, when you have one dot, you know, geometry is like, yeah, so what do you do? You put a number of lines through one dot. What do you do about that? So you need—

do you try to go historically?

You know someone over time. Like the reality is—

so the easiest, uh, let me interrupt with my favorite example. For YC founders, the most important question I can ask the partner who's working with the company is how much have they learned in the last 3 months? 3 months is enough to tell if they have a high learning rate or not.

Totally. So I guess in a, in a fast-moving process, all you can do is basically try to get a data point from the past.

Yeah. And then once in a while you can try to get data points from external people, but most people, you know, have the wrong, the wrong prism. They don't. Unless you ask the question perfectly and unless you can really retrain their eyes, it doesn't help.

But, but, you know, there's other ways. Like to say, imagine you were in this business that you're not familiar with. Mm-hmm. How would you go about learning X or Y?

Yeah.

Yeah.

Let's even see how they would start thinking.

Putting them out of context. Yeah. And having them think through, very easy to tell. One of my favorite questions, um, is it doesn't matter whether they're doing a startup or not. Let's say they're a candidate for a marketing job. If I gave you a seed amount of funding to investigate 3 ideas, which 3 would you pick? How would you go about evaluating them over the next 6 months?

Yeah, that's good.

It's a pretty simple test. You can tell a lot about a person, how they, just based on how they answer that question.

Yeah, that's good. Are there any things that a founder can't be completely deficient in?

Ethics.

Ethics?

No question.

How do you figure that one out?

Mostly through references.

Even on that point, a lot of people that seem not so friendly are actually very ethical. And like, um, I've noticed that a lot of the like highly disagreeable, you know, intense founders actually have a very strong moral compass.

Yes, that is absolutely true because they believe in their principles.

And it like is the source of both. That's interesting.

You've seen this in the Valley. We were talking about this earlier. People have changed their political affiliation at a whim, not stayed with what they really believe in.

Yep. We're definitely coming back to politics in a little bit. I have a lot there. Is there anything around like ability to like recruit? Anything around like, yeah, ability to recruit is a very important dimension.

Can you envision this person at Mark? It sounds like they gotta hire 1,000 people at some point. Yeah. But can they hire the first 10? Maybe 100 is like really critical because those people are going to replicate themselves. You know, Patrick Hallahan talks about this at length, like your first 10 are going to multiply by 10.

'Cause you could imagine somebody who's really exceptional in some dimension, but you just don't think they could recruit well.

Right. But then the question is, can they parlay whatever unique assets they have into convincing people to work with them? And sometimes they can actually, even if they're very uneven, it's like, oh my God. Like, yeah, they can still, or you can help them. You can help them communicate like, why is this ambition worth chasing? What kind of people do you want? But their energy or their special secret sauce like comes through. Like, let's say I set you up to interview with one of these great founders. I think you would pick up this person's kind of unique. Even, even if you couldn't articulate the exact reason.

Yeah. You just come out feeling that way.

You come out feeling like, wow, that was interesting. At a minimum, that was interesting.

Yeah. I mean, one of the things I often think about is like somebody who is good at selling anybody is kind of on some level good at selling everybody. Like it's the same person who can convince candidates and investors and customers and kind of all of it. Like it all kind of does bundle together.

Yeah. You definitely have to be able to tell a story, right? Like at the end of the day, you have to convince people that you haven't proven almost anything usually in the beginning.

Right.

And you've got to convince people to come along on the journey with you.

Yeah.

Those are investors, early beta, alpha customers, employees. You have to retain your employees in a hot market. Like you're, you're doing some version of that constantly.

When you look back at like a bunch of like the greatest investments over the last, let's say 15 years to go back, like a, but like, and like an Airbnb or, you know, whatever. There's like a bunch of companies that were like not hot early and like everybody passed. Is that still the case? Do you think, or do you think that it has become more like, has the consensus become more accurate or is it still the case that like it's random signal?

I don't think it's random at all, by the way.

Yeah, go ahead.

I mean, I think for example, like let's talk about like Airbnb. To me it was obvious 3 minutes into Brian's monologue. I was like, this is the coolest thing since YouTube. Literally told him this, told him exactly why it was so obvious. I needed to meet him.

Yeah.

And I didn't like the Airbed and Breakfast name that they was using at the time. So I kept deferring the meeting, which was a very costly decision for me.

Yeah.

That's at the end of the day. If you met Brian for 3 minutes, there was no way of missing this. This team was very special and he was able to convey a couple, 3 key things he said in the 3 minutes that I was like, oh my God, this, this is really, really amazing.

But you saw it, but most people didn't.

Yeah, but I'm saying it's not random. Yeah. Like, or like you saw OpenAI, like it's some company we all know. The key investment decision, literally, like if you read our memo,, which, um, he sent, you know, to the LPs because it was so much of an outlier at the time, was this outside Google and DeepMind specifically, this was the only critical density of research-grade people that could possibly pull off AI.

That was the investment hypothesis in, in like, in one sentence, given all the other stuff, like not, it was a nonprofit, you know, there was no product plan, no revenue plan, just this AI capability. It's the only time in the history of 20-year history of Khosla Ventures, we sent an apology letter to our LPs when we made the investment because it was twice the largest investment we'd ever— initial investment we'd ever made. We sent an apology letter. I know it makes no sense, but we are doing it anyway, not asking for permission. It's really— that letter is now part of our fundraising deck. Wow. But we sent it in 2018.

Yep. And then here's a good example too of rate of growth. Like it did help a lot that David, me, and Vinod all knew Sam for a long time.

Yeah.

For a sustained period of time. So you could see certain lines there and the critical density of talent that he'd assembled, plus certain traits about Sam specifically led to that investment.

Do you think that as time has gone on and like more people are in tech and there's more founders and more investors. Do you think that there's any increasing, not from, you know, people who can really see it, but from the average, you know, reception to investors? Do you think that the consensus hot deal is becoming any more or less likely over time? You know what I'm kind of—

At seed, I don't think so.

You think it's still the case that like these hot seeds are no better than the non-consensus seeds?

Personal opinion? I agree totally with Because it's like, we're talking about like Rocket Lab, which is one of the best investments anywhere. No, no. I mean, I don't think they would've raised money from anybody.

Like, well, it's worth $40 billion now, but we bought a third of the company for $5 million.

Yeah.

And it's because nobody wanted to invest in space.

Or Commonwealth Fusion, probably.

Or Commonwealth Fusion, or even OpenAI.

Mm-hmm. Mm-hmm.

Right. It wasn't, we were the only venture investor. In 2018 that committed because it didn't make sense. I think at the seed level, consensus hotbeds are generally around people, not around the idea.

Yeah, it's like, you know, if you have two people who left, you know, pick your super hot current, you know, two people leave Cursor, you know, it's like an engineer and a, you know, designer leave Cursor tomorrow. Like what's that going to go at?

A lot, but it's— but a seed investment around, let's say, undiscovered people. Yeah, I think the consensus ones aren't going to outperform the outliers at all.

Maybe just talking about, like, themes that you guys are interested in. I know that you're mostly focused on people, but, you know, you also care a lot about markets and technologies. Maybe to, like, start with AI and then we can go outside of AI. So obviously AI is, like, in the midst of playing out. I don't know what inning it's in, but it's one of the early ones, probably. You know, like, I think there have been, like, a lot of changes to, like, what you know, the labs say that they're focused on, and maybe like the types of companies that are getting funded is adjusting a little bit if you sort of like just like kind of track every 6 months or so. I'm curious what you guys see as the lay of the land for like outside the labs, where you think most of the opportunity lives, what you're most interested in, what you think is going to be like a big deal in the next year or two. It's—

that's such a broad surface.

Of course. Sorry.

Um, I think I was counting this, 30-some startups in our portfolio that are building an AI worker of some sort. Yeah. An AI oncologist, an AI mental health therapist, an AI chip designer, AI structural engineer. As many professions as there are, there's that many opportunities.

Basically to fully do the work.

Do the work.

Yeah.

So one thing we decided a couple of years ago, probably 3 years ago, we wouldn't do a lot of copilot. Copilot had just come out and we said, Copilot, humans get in the way, let's just do the work. So we love really people doing the work as opposed to helping a human do the work. Now there's exceptions to that, but mostly that's true. That's a big category. So we haven't invested in any competitor to OpenAI, obviously. And we should come back and talk about it. We are fiercely loyal to our companies. Uh, we, we, we can talk about loyalty. Well, OpenAI, yeah, I think I was the very first one when Sam got fired to come out and say, we'll fund Sam for whatever he wants to do next.

Yeah.

Is, is the loyalty, does that stem from sort of like a business decision or idea, or does it stem from just like a, this is how it ought to be, like a more ethics kind of thing?

I think it's, this is how it should be. It's about ethics. Yep. And if I disagree with the entrepreneur, then my job is to sit down, tell them why, and agree that we agree or disagree before taking the right position publicly.

Yeah. If we believe the, on the principles, like to the point, if you have principles and you believe you're on the right side, then we're willing to use our, you know, brand capital audience to help. Yeah.

But along those, the AI question you were asking, we've invested a bunch in all their other approaches than transformer models that make sense. And we probably have 4 or 5 different efforts that are, they don't have to replace transformer models, but they're different because I think the big labs are doing transformer models really well. And then doing some other things.

So what else do you think is promising?

Look, it's too early to tell. So I think any of us who pretends we know this technique or that neurosymbolic technique, we have bet on category theory in math, we have bet on interpretability that leads to different models, we have diffusion models, there's plenty of others.

Do you think—

Very excited in real-world models.

Yeah, I was just going to ask you that.

So I think that game hasn't played out yet. It's not clear at all who will win.

Yeah.

It's very clear what the major labs will do in that area. And at least the major labs all have efforts, but I think it's completely up for grabs.

And you have no doubt it's going to work. It's just a matter of how and when.

Oh, I have zero doubt it will work and work increasingly well.

And it'll be embodied robots that are—

It'll be embodied and more than that, understanding the physical world.

Yeah.

Let me give you an example. This is a public clip we showed. Intuition is a very big deal. And I don't think current models embody intuition. So we have something called General Intuition based on gaming data. And I saw a clip of of this, a live feed of Ukrainian soldiers trying to escape Russian attackers. And then we gave the AI half the clip and said, replicate the other half. It replicated the intuition of the Ukrainian soldiers trying to escape almost identically. That's intuition. There's many dimensions like that. But there's many obvious ones. You know, we all talk about the fact that frontier models hallucinate and there's probably not a good way to avoid them. There's ways to minimize them and the labs will do a pretty good job. But you see the hot agent startups like Sierra and Decagon, they're completely, I think, missing the point. We've focused on customer support agents that do not hallucinate. If you're a bank, if you're Visa or MasterCard or insurance, you can't afford to hallucinate an answer, even if it's low percentage. We've heard a lot about mental health therapy where bots take people down towards paths. Those are all examples. So I think the winning customer support thing will be something that does not hallucinate for a lot of class of applications. Some applications hallucinate, some hallucination is fine.

And are you saying that you think that like a company can't just increment its way from hallucinating sometimes to hallucinating never and you just need to, you can't just, those companies can't just improve their way to no hallucination?

You know, it's possible, but I think if you architect for low hallucination, know when to use no hallucination when you can afford to, giving somebody's bank balance, you'd better not ever hallucinate.

Yeah.

I think it's pretty important to look beyond normal. Lots of people will do the normal extensions of LLMs to lots of applications, massive market. Lots of people will do it. Every area will have 10 startups.

Keith, are you interested in hard tech as much, or do you invest more in like B2B typically.

Well, I've invested a lot in AI, so yeah, I love your quote about AI and how you'd have done it differently if you hadn't joined Vinod Khosla Ventures.

Before I joined KV, I joined KV literally 2 years ago, basically this week. Yeah. I had invested in zero— Rejoined. Rejoined. Yeah. Rejoined. Rejoined.

Yeah, I do.

That's not good. And literally had invested in zero AI companies before.

Interesting.

Since then, in the last 2 years, I'd say it's about 70% of my investments are AI.

Mm-hmm.

And had I not rejoined KV, I think I would either miss the whole wave and been completely irrelevant or been reckless. Neither one's good because what I was able to do when I joined is learn by osmosis. So I'd sit in partner meetings and 80% of the companies that we would discuss in the portfolio, new opportunities would be AI-based and I would listen and learn. And then B, as I started meeting founders who were interested in AI that fit my normal like standards. I could send like the deck to, uh, actually 3 people here, uh, Vinod Khosla and John Chu, and sometimes actually Kanu too.

Yeah.

Get feedback and then actually often have some combination, maybe all 4 meet the team. And so when I first started making AI investments, I felt like I had this air cover of like a lot. They, they, you know, could understand how good is the team, B, how smart is their approach, how differentiated is their approach. And then B, is there anything else in the landscape that's even better? And then that allowed me to start making, you know, new investments. And then you develop some taste and some ideas about what works. And then you join the board of the companies and you learn like how these companies are built and what works, what doesn't, what are problems, what are not. And so now I basically do almost virtually all AI.

Yeah. I'm actually curious on the topic of how the companies are built. And obviously you're super involved with like company building. You've built companies yourself a bunch and It's just like something that you talk about a lot. Do you think AI companies are built in any substantially different way?

I think fundamentally different. First of all, they, they are growing at rates that are unprecedented in the history of technology. And it's a little bit like, to me, it's like running the 4-minute mile. Once you see someone runs a 4-minute mile, then no company should have an excuse for not growing rapidly. It's like, you see all these enterprise companies going from $0 to $50 million, you start asking questions. Well, why, why can't you? At least what are the limiting factors? And there sometimes are real reasons, but you start with the question of why, why not? You know, why not versus like, oh, that's impossible. Like if you had said we're gonna start a company and have $10 million of revenue in year 1 from launch, you would've said, all of us would've said 10 years ago, that's impossible. Yeah. That is not, definitely not the right answer. Now doesn't mean every company should do that, but it is an open question. Then the question is, well, what do you do about that? I think the idea of them borrowing from Peter Fenton here, the idea of PMs does not make sense in a rapidly emerging technology field. What do PMs do? They go talk to customers and they create a sequential roadmap over the next 12 months. Well, if the field's evolving and the capabilities are evolving, Like literally every month there's papers published. You probably read papers every week, let alone things that, you know, are launched live. You can't have a 12-month roadmap. That makes no sense. And so you have to rethink that. Then also, how does sales work with your research team? Like OpenAI, you know, actually pairs, as far as I can tell, like the people doing customer acquisition with the research team. And that's a completely different model than how most technology companies were built.

Yeah.

Compensation, this is in the public domain, completely different. Completely different. And people haven't even thought through the implications. Yes, there are some big companies like Meta that can afford because of how they mint or Google because of how they made money to pay things that only professional athletes could aspire to when we were growing up. However, if you're a startup, how you can afford that level of cash comp while you're not minting money?

Well, you can't, right?

Well, this is, but then you still have to compete with people who can.

So you have to think, doesn't that go to like your diamonds in the rough type of idea either?

Well, that's one possible solution. Or you've got to hire people that don't care about short-term cash comp and have a different missionary zeal or different orientation. Or you've got to take costs somewhere else, a lot of money from somewhere else that you're not paying for. Like, but no one's really— yeah, very few people have thought through company building from scratch. If you need research-grade talent and you have to compete with a market, how do you reorient the entire P&L of a company?

Yeah, do you have like—

well, I think it depends what you're trying to do, like depending on what market you're in and who you're competing with and how much cutting-edge AI talent you have. Do you need one person? Do you need a team?

10?

Yeah.

Those are all very different, you know, financial— there's a lot of financial consequences to those.

Yeah. How different do you think it is for like, let's say like a, I don't know, somebody who's building like a new like AI-centric system of record or something like that. Like how different is like the first 30 hires look like? What are like, yeah, could you just sort of like guess what you think that might approximately look like?

I think it's very different.

Yeah.

First, the whole idea of systems of record is going to change. Pretty dramatically, and it may be the old system of record don't go away, but the operating substrate is wholly different. In fact, most likely that'll be the case. Take ERP, a hot area. It's really becoming unbundled. So there's procurement, and then there's finance, and then you can go through the various modules in an ERP system. If you don't have the right substrate to have it operate under an agentic architecture, you're not going to have huge success. Used to be in an ERP system, it's what features does it have? I'm a manufacturer. Do I have manufacturing features? I think now it's about how do I reduce the number of people I need in accounting or supply chain or others.

Yeah.

So one of my favorite examples, we invest in dual entry. They have a client called Slash, small business lending, complex business. It's both regulatory complex and credit scoring and all our complex areas. $150 million ARR company with only one person in accounting. That was my reference. When we invested. Why? Because the architecture's right. And by the way, these aren't founders who are credentialed. I think they're from Venezuela. Really great people. Love the entrepreneurs. That was number one agreement. We love the entrepreneurs. Two, love the architecture. Three, love the impact. And the benefit is very different than a feature list.

Another difference is, like, so for example, traditionally we built an ERP-ish system of record or whatever. You think about like your defensibility would be around the number of integrations you would do. Like think Rippling, like we have all of these integrations, blah, blah, blah, and creates a big moat. With things like Cognition, Devin, doing 100 integrations is something you could actually feasibly do in a month. Yeah. At very low marginal cost.

Totally. Like a company like MuleSoft makes no sense now.

Yeah.

Well, exactly. Like, so a lot of these incumbents are much more vulnerable, I suspect.

So given that you've got like all these different intentions for the company, the way success is measured. It's like the building, the way you build a company, like the old, like playbook to whatever extent those ever were any good or like definitely no good now. Have you found that retraining, for lack of a better word, like experienced execs who like came up in pre-AI ports over well? Do you have to be more cautious with that as you like bringing—

I think it's fun and challenging. Like, so I'll give you an example. One of the best companies like ever is going to be Ramp. But RAMP started in the pre-AI era and we're very AI forward. Like we talk about this publicly, there's stats about it. We leaning in, we're hiring AI native, you know, people constantly. We have incredible, we have the best intern pool on the planet for the last 3 or 4 years in a row. But we actually have to rethink the company because we were really started in 2019. And so we have a lot of things that were based on doing the best possible version of a technology company in 2019. And that's changing.

Mm-hmm.

And it's an interesting board level conversation that we have of like, Wow, should we rip up everything we've learned or which pieces should we rip up so that we can be the best company in the next 10 years? And if RAMP's doing that, imagine what every other company should be thinking.

You know, my way of explaining this, most experts are experts in a previous version of the world, not the one you're trying to create. And so fast learning, and I come back to that, is much more important than lots of experience in this AI world. Even how you do anything, computer architecture, system integrations, marketing, customer support, all that is so radically different. You want rapid learners, whether they're experienced or fresh.

Yeah.

Let me give you a mundane example. So like, think Lattice, back in the Lattice days. Right now in enterprise, because of the hype, uh, because of the hype of AI, top-down sales can work extremely well. The CEOs feel pressure from their board to be AI forward. Their executives feel pressure to be leaning into AI. Historically, not the best way to build a company is to depend upon top-down CEO sales. Yeah, but it actually does work in certain verticals right now extremely well.

Yes.

So the whole go-to-market playbook, you have to rethink too.

What's funny, like in legal or like a couple other categories, it's like they just want to buy AI, like more than what the specific solution is. Like, we want— we have a big budget for AI.

In the short term, they have the budget. They may not care about the impact long-term. It'll harmonize. You have to produce results ultimately. We care about that. We evaluate when we look at application-level companies. What is the actual impact you're driving for your customers is a key, key input for us.

But I assume if you have a founder you love and there's this crazy market pull, even if they haven't worked out all the pieces later, like that's a good enough starting point to bet.

If they really understand that they need to versus just chase revenue.

Yeah. Outside of AI, are you guys, I mean, you said 70% AI, so like there's other stuff you guys are interested in. I know we've talked about like robotics. Last year.

We've talked about— well, robotics is AI.

Yeah, of course.

But pretty big on robotics.

Yeah. Well, one thing people know less about us, but we've been consistently excellent at across the history of the fund, is in financial services. We have this point we make to LPs that in every single fund we've had, we've returned the fund solely on one, just on one financial services investment.

That's a crazy—

so we love things like, you know, Square, Stripe, Affirm, just examples in the modern stuff.

First investors in all of those except Stripe, we were second.

Uh, like we have Avin, which is a very excellent company. Um, you know, like, yeah, it'll be the next surprise strike. Upstart was incredibly successful. We did some of the seed, led the A.

Has finance gotten like less AI-ified than other areas?

People are using it, I'd say maybe not as strategically yet.

You think it's because people are like a little scared, like, or, or like rightfully scared? Because, because I do feel that the fi— the fintech companies of the last era seem more they seem more insulated from the AI wave and threats. And like, I haven't seen as many like—

Well, there may be some hallucinations. Maybe some good example of—

well, some of them with hallucination concerns.

Yeah.

But Avan is a good example of a company that's deeply using AI in every aspect of it. That's why you can get a home equity line of credit and consolidate all your credit cards onto a new credit card that you get. That has 10 points lower interest rate.

And then you have a credit— and then they have your credit card.

And well, they issue a credit card based on your home equity line of credit. They can get it done in an hour. And normally that's weeks and weeks.

And AI is what makes it possible to go so fast, which makes it fit in the right way into like the process.

And Ramp's using AI, you know, very aggressively. But I think it hasn't been quite as transformative across the broad—

versus like, well, you know, the thing is there aren't too many great, uh, fintech companies. Uh, I would venture to guess Ramp and Avan will end up being two of the best of the new breed started in the last 5 years.

Mm-hmm.

Yeah, I agree with that. And so that may be true. Like once every so often there's an amazing fintech opportunity, at least in the United States. There's been like Nubank and company in Germany traded publicly that are awesome, Revolut. But in the United States, it's once every 2 or 3 years at most that you have a true iconic company.

Yeah.

You know, tens of billions or $50 billion or $100 billion market cap. That's not that often. And so I do think the next generation will use AI in a very significant way. Yeah. So fintech is a really interesting area. We still do a lot of sustainability stuff. I am really bullish on energy. We are We are very, very bullish that that area will keep sustaining.

Actually, so maybe this is a good moment to talk about this.

By the way, manufacturing, another area we haven't talked about. Oh yeah. Huge interest for us. Defense is another huge interest.

What in manufacturing?

I think applying AI to completely change the paradigm of how manufacturing is done. As part of that, you can onshore stuff that was offshore. So it's these two trends colliding.

Is robotics part of that too?

It is part of that. It's not the most essential part.

What's essential?

Essentially reducing labor costs in other ways, not by a robot do the job, but running a system. In a way, an iPhone assembly line would have a few thousand manufacturing engineers. If you can do that function with AI, then you're, you have a pretty large advantage manufacturing onshore.

How much of the opportunity is that like the create, like the points of creation of goods versus like the operations and logistics around a manufacturing company? Both.

Both. We are seeing both.

Yeah. Yeah.

Supply chain was sort of a minor part of whole ERP. We talked about that. There's going to be lots of opportunities to replace supply chain software with new AI software.

Yeah.

Yeah.

And then in defense, I guess, like, obviously with like Anduril, SpaceX, like, you know, obviously like now there's like huge inroads here. Is your guys' sense that there's a lot more opportunity for those flavor of companies to be built? Will those companies dominate in their markets? Like, how do you think about what those markets will play out like?

I think there's room for lots of new startups. I mean, uh, we have big investors in Hermeus. Now, that was not started in the current fashion. It was started 5 years ago to do supersonic aircraft. I think that'll be an important part of national defense, become even more prominent because Russia has used supersonic aircraft missiles in Ukraine and we don't have any.

Yeah.

Uh, so that's an important area we did a while ago. Um, there's many other areas. Keith, you can talk about MACH and some of the others.

Yeah. So we've been in that— well, we've been concerned like politically, geopolitically about the threat posed by our adversaries, the CCP, et cetera. Vinod got involved in Hill Valley Forum before it was cool, one of the first 20 people, like when, you know, setting up to alert— by the way, under Democratic Yeah, under a Democratic administration, because it was common concern, because the country needs to take advantage of the best and brightest in technology and cutting-edge technologies, or we are going to sacrifice our way of living to people who do. And we have started— we started investing in things like that ahead of the curve. There's now more interest among VCs because some companies are perceived to be doing quite well. VCs, you know, are always like a herd. Um, but we invested in Varda, which has a significant, you know, Defense component, mock technologies is a very high potential rocket lab many, many years ago. Yeah, all of these were before was kind of quote cool. Now also the country needs to take advantage of technology. The country has more threats and has more potential adversaries to worry about. It has to do with less money. It has to survive or thrive with less money. Technology is a great magic wand. It is, it's been for consumers. You know, for 40 years, it was like magic wand, you get more for less. The government needs to embrace that, and this administration is, and putting people like Emil in place hopefully to catalyze a new world order where we take advantage of technology.

Yeah.

And make America better.

Yeah. I guess on politics, like, both of you are like pretty willing to like get into politics stuff on Twitter and stuff, X, and you have very different politics obviously. But I guess my first question is like, um, you both are like willing to just sort of like get into, I don't want to say fights, but like fights on X about politics and stuff like that, clearly it comes from like strongly believing in what you think. Why spend the cycles on it? Like, why do you—

I'll give you my answer. I don't actually know yours. Mine was like, you know, through technology I developed a following and I woke up one day and said like, look, I don't want to die one day and regret that I didn't use my audience to proselytize about ideas and things that I find important. So if I can surface new ideas or new or rebut bad ideas, I want to, you know, finish my life thinking I did the best I could to have influence and I have a platform. So I started using it that way, particularly on like topics related to politics. Um, and so I was just like, I don't want to regret having not tried to change people's opinion.

Are you stressed at all when, like, when you're like getting into it on X with somebody, are you like feeling anything about it? Or are you just like, oh, I'm just saying what I think, but I'm—

well, there's times when I wish I had more time to like research. Like if I was doing nothing else, if I didn't have like a day job, you know, there's times when like I know how to construct an argument. I know where all the evidence is. I don't have time to go do that. Um, like my friend David Sacks back in the day hired a research assistant secretly before, you know, to help him. I don't, I don't have time to do that.

Yeah.

So there's times when I happen to know the answer. I used to be very involved in politics. I, I know a lot of details, but I do wish sometimes that if I wasn't doing a real job, I would be spending more, more care in marshaling more evidence and probably be more effective.

Actually, one other bit that I'm now curious to ask, cause I think it's kind of funny, is you're like kind of harsh on Twitter, but you're like really friendly in person. Is that just like how you write or like, do you mean to, like, are you doing it for the clicks?

I also had this crazy idea. This is a dumb idea. Like a decade ago, I was like, I'm going to combat every bad idea on the internet. It was so fun. Like, I was like, well, if someone puts a bad idea or something wrong or false, nobody responds. People think it's true.

Yeah.

Like LLMs are going to pick it up and think it's true. I was like, I'm just going to respond so that at least there's a written record. But like, there's so many bad ideas. There's so many dumb ideas in the world that this is like the worst idea I've ever had in my life.

Yeah.

But you get addicted to trying to fix every mistake on the internet.

I like the ones where you don't even like explain what's the problem, that you're just like, wrong.

Yeah, well, that was a funny Square joke. That was old Square. We had this critic at Square named Rakesh, and he used to just like constantly complain about like Square, Square was never going to be successful, and he knew payments for 20 years. He's a canonical expert in payments. And so sometimes I'd engage and write back like substantively, but sometimes he'd be like so far off, just so wrong. So it became like this internal Square joke to Rakesh and it became like a meme.

That's good.

That's where it started.

That's good. Why are you motivated to like get into it online?

You know, I don't spend very much time. All of social media, I probably spend less than an hour a week.

That's good.

So if I incidentally run into something that is blatantly wrong, then I'll express an opinion, but I won't even have the time to read the replies I get.

So like if you're like fighting with somebody on X about whatever and you're getting all these replies, are you just like, ah, I can't, Like you're not like bothered.

I'm not bothered.

Yeah.

But if something's an important idea or something, you know, usually it's some principle somebody's violating. Like this weekend I happened to run into a tweet by Bill Ackman.

Mm-hmm.

That was sucking up to Trump on capping interest rates at 10%. You know, said, uh, he's sort of recommending that idea. And then saying, well, maybe there's a market approach. Hedging the truth is clearly a market person. So I replied to him pretty bluntly. I like Bill. He's a good guy, know him. But I just couldn't let that pass of sucking up to Trump, which on a truly bad idea of capping interest rates. Almost like Trump and Kamala Harris would have the same idea. Price controls. Yeah, right. It's like, come on, speak up. Yeah, don't be dishonest about your opinion. And he was being dishonest. So when it bothers me, I reply, but I don't spend a lot of time, and I don't have a lot of time otherwise.

Yeah, yeah, I do mostly like— I mean, an Uber, you know, you have these moments where it's really hard to be effective. Yeah, like, and I just use it as like snack while I'm like in an Uber ride, I'm not going to be able to schedule a call while I'm in an Uber ride, you know, etc. So I kind of like get on there and dunk or something. I do it when I'm eating, which is a bad habit, truthfully. Like I'm having breakfast, I'm like on Twitter or something. But I started, you know where I started? It was actually a business reason. Um, I was kind of famous back in the day at Square. I read every single tweet every single day about Square for years because occasionally you'd see a gem, like either a great story they could reshare, like a positive experience, like you helped my life and you get this great anecdote. Or you'd see like a customer complaint and you could route it, you know, directly to someone, or you'd see a product feature. So I literally read every day, started every single day, but then that gets you in this train habit of like, you have to read everything. And, you know, maybe not the best.

You were like 10 years ago, maybe let's say you were like sort of like one of the first sort of like vocal conservatives, maybe the only one, maybe the only, and I think now you're maybe like one of the only vocal, like at least like liberal leaning? Maybe that's like too strong.

I'm an independent.

You're an independent. For the record. Yeah.

I've never been a Democrat.

Oh yeah. So what was the political journey?

I used to be a lifelong Republican over fiscal issues and switched to independent over climate issues. And I've stuck with that pretty consistently. And then principles matter a lot., you know, my fight against Trump is his values. He has none. Yeah. Um, we can disagree, or he tries to go to extremes to get his constituency rallied around him. Either way, I don't like the idea that people don't mind lying about things.

Yeah. When you watch that and like, so it's like violating your principles, you're seeing like everybody around you, like what's your, what is your like assessment or read on the situation? Cause like obviously. 10 years ago, this wouldn't have been the dynamic and there was no flip.

See, many people have changed their affiliation to be for convenience.

Mm-hmm.

Doing things for convenience is a bad idea. Now I realize some CEOs have responsibilities to others than themselves, but some are just, will change political affiliations and 'If the next president is Bernie Sanders, they'll become liberal again.' That I hate.

So I agree, by the way. I don't like the convenience. I believe in principles. I think a reasonable fraction of people, I think that's true. And then I think there are some who like watch the evidence of just like, there's a lot of things, let's say, the last administration did badly and may jeopardize the country in some ways. And then some things that Trump has done have clearly worked or seem to be working, at You know, et cetera. And some people, it is fine to change your mind based upon evidence. Mm-hmm. I would say there's an element of convenience too. There's probably a mix. It's hard to sometimes, you know, who's who and what camp and all that stuff. And then there are some who have customers and employees that they have to represent. Absolutely. And you know, that's an important consideration. Like, so for example, if you have a significant government contracting revenue, you know, you have, you do have to think about like you're fueling your families, your comp, your employees' families. Like, yeah, you know, there, there are real responsibilities.

No, there, there, there are responsibilities, uh, you CM from a number of companies where they clearly fundamentally haven't changed their principles, but they have to take the responsibility of the role or get out of the role.

It's also interesting cuz it's like, it's new for tech to even care, for it to even be reasonable for tech to care about politics. Like it just didn't matter in the past. And also it was like not possible, like tech was much more monoculture on the left before. And then also there just like weren't like government related tech tech startups.

Some of this, the government is obviously, you know, found it much more interesting what tech— variety of good and bad reasons, probably. And, you know, you could argue that his best, best thing ever was the tech was built mostly on the West Coast, far away from Washington, allowed tech to thrive and, you know, invent itself without a lot of scrutiny. Because generally speaking, government scrutiny early in emerging technologies is not a good thing.

This is why I'm like afraid of like AI regulation. I'm like, what are the odds of getting it right?, it just seems hard to do.

Well, I think that's a real risk with emerging technologies. The government is less likely to be right than wrong on something that's rapidly emerging. Maybe when something stabilizes, the government might have a better, you know, predictive accuracy record or something.

Yeah. I mean, you obviously need regulation at some point in these journeys.

But then you also have to think about the rest of the world. And, you know, Vinod Khosla, even though he's probably a little bit more pro-regulation, generally speaking, than I am, has been very concerned about, like, losing AI to China and what that would mean to this country.

Yeah.

And if you have a threat, you have to think very carefully about regulating a new technology if you know a very serious adversary is putting the pedal down and doesn't, you know, isn't restraining itself.

They're really good at robotics. They're really good at manufacturing.

They're already really good at robotics.

Look, it's very clear. And 3 years, 4 years ago at the Hill Valley Forum, I talked about we are in a techno-economic battle with China. We must do everything to win. And everything we can to disadvantage them. It's just the truth.

What does that look like? What does need to happen there?

So I think too much regulation in AI would be a really bad thing. State-level regulation is a horrendous idea.

Just generally?

Because they don't understand the global implications. You know, not everybody abides by American rules. The Chinese for sure don't. Want. So I think we have to be realistic and pragmatic about what battle we are in over the next 10, 15 years. Economic superiority is up for grabs, and we gotta win, or we'll live under President Xi's rules.

Yeah, yeah.

I mean, obviously you've talked about this a lot too, the, like, you know, it's a bipartisan effort back, you know, 4 or 5 years ago. Um, and fortunately, um, it was very effective. It started changing people's minds. Like, there was a lot of people that were very naive about the threat. Yeah. Um, 5 years ago. And I think if we'd waited too long— and there's still debate about, like, should we export this chip or that chip?

Yeah.

You know, what should the restrictions be on different technologies? Like, there's a lot of nuance to this, but at least the central idea that we just cannot lose this race—

yeah—

for AI, I think this is pretty mainstream now.

Yeah. It's part of why I like a lot of the, like, America-centric companies and investment firms, like it's really resonant for people. Like people want to work at those, you know, like it's like there is a growing patriotism out of necessity. I think probably I have final question for you, Vinod. Did you see the report that Keith went to Barry's 2,000 times last year?

That's an exaggeration.

Too many times.

Yeah.

Is it too much?

It's only 430 years.

They're just trying to get ahold of Keith.

He's just like in, in the interesting thing about it is though, um, literally this morning I went to Barry's 7:10 AM and a founder comes up to me after class.. And he says, you know, I run— nice to meet you. I was like, what do you do? He's like, I run an AI detecting cancer funded by A16z. And I was like, that's cool. It's interesting. Who knows if, you know, who knows if one day we'll invest or whatever. But he said, thank you so much. And I said, what are you talking about? He's like, I started getting myself in shape by going to Barry's because I read about you doing it. And so you change someone's life indirectly through that. Like, that's actually really rewarding.

It's not that much, but it's good.

That was just this morning. That's great.

All right, you guys, thank you very much for doing this. I had a great time. Time.

Really appreciate it. Thanks a lot.