a16z GP Martin Casado：Anthropic 与 OpenAI 之争，以及为何开源在中国问题上构成国家安全风险

There's only been one sin, and that one sin is zero-sum thinking. We always worry, like, “Oh, is this defensible? Will this layer get margin? Will this layer get value?” The answer has kind of been unilaterally yes. Every layer has gotten value. Every layer has winners.

These markets are so large, and they're growing so fast. We're actually seeing brand effects take place in this phase of model scaling. A lot of the approaches to scaling don't generalize. This gives a ton of room for the application developers to build their own models. I think that right now, open source is most dangerous because China is better at it than we are.

Martin, man, I love our conversations. I was so excited when you said you'd join me again. Thank you so much for doing this, man.

1. Analysis of Current AI Investment Landscape

I freaking hate these “How did you get into venture?” intro questions, so I just want to dive right in. It is a freaking nuts time. Starting off, how do you evaluate where we're at today in the AI investing landscape? Peak hype cycle? Great? Super-excited? Both? How do you evaluate it?

2. Exploring Value Concentration Across Verticals

I'm kind of of two minds. On one hand, I do feel like my intuition doesn't really work like it has for the last 20 years. The future is just very uncertain, and one of the reasons is that this is really the first time software development and software creation are being disrupted. On one hand, I'm like, “I don't really know what to think.”

On the other hand, observationally, there's only been one sin, and that one sin is zero-sum thinking. We always worry, like, “Oh, is this defensible? Will this layer get margin? Will this layer get value?” The answer has kind of been unilaterally yes. Every layer has gotten value. Every layer has winners.

Things that we thought were silly are making money. It's been solved. There are profitable companies. The business case is there, et cetera. I think the one sin is not playing the game.

Do you agree with the playing-the-game-on-the-field sentiment? When we look back at 2021, I remember everyone saying, “Playing the game on the field.” I wish I hadn't played the game on the field, to be transparent, Martin. Do you agree that you have to play the game on the field?

I think behavior should follow business; it shouldn't follow marks. I think in 2021, behavior was following marks, right? The public markets just decided these companies were valued a whole bunch. Tiger Global came in with a ton of money and deployed it a whole bunch. Behavior following investment and marks is a bad idea.

In this case, you have some of the fastest-growing companies we've ever seen by users and by revenue. The amount of value that's shifted to this is so significant, and so I think investors' behavior should follow that. If not, what are we doing?

When you think about shifting value—again, I'm diving right in, but this is not going roundabout—you said there's a disruption of software development. There are a ton of players in the vibe-coding space. They are predominantly all sitting on top of Anthropic. Claude Code is gaining more and more dominance. How do you think about these providers' reliance on a tool that could eventually shut them off?

There are 2 futures for code. In one future, you've got Anthropic as a monopoly. In another future, you have what we'll call an oligopoly, or maybe even more of a market, of these coding models. They're just very different futures, and I think when you answer this question, you have to consider both of them.

I will say that the timing of this conversation we're having right now is pretty soon after Claude 4 launched. That's a major model launch, and these models are so episodic. Every time one launches, everybody's like, “It's the future. Everything's going to happen.” Remember the whole Ghibli OpenAI launch? We were like, “Oh, images are going to change forever.” Then it comes, we're excited, and then it kind of passes. Maybe that'll happen here, maybe it won't. I don't know, but for sure, our perception is colored by that launch.

Let's consider both of these. I'm going to consider the first one. Historically, models don't really keep much of an advantage because they're so easy to distill. Even in the last week, we've seen launches of models—Qwen and, I forgot, [Kimi?]—that came out, and they're great. People like them, and they adopt them.

In that world, where you continue to have new models from different providers, I would never count out Google. Their coding models are fantastic. The rumor is that GPT-5 coding is going to be great. In this world, where you've got lots of models coming out from lots of providers, you need to have a consumption layer that's independent, right?

All of these companies are going to add that consumption-layer value—for example, for nontechnical users, Python users, professional coders, or whatever it is—and that's going to be a very healthy layer.

Mhm.

The other future is, let's assume that Anthropic is just a monopoly on coding models. In that case, you have what you normally have in these situations: they'll decide where it's not profitable for them to enter, or they'll change their business model. Maybe they're like, “Listen, we want to have the consumption layer, but we're never going to be an app-development tool company.” It's just a different sales motion and a different sales team.

Nobody knows where that stops, but they will put pressure on anybody that they view as being in their core focus. They will do whatever they can to either capture that margin or capture that market share.

I think it's the wrong time to have this conversation, right after a major model launch, because, like I said, these models are so episodic. We always assume every time a model launches that it's going to be a monopoly, and it really hasn't been the case.

Going to your zero-sum thinking, if you were to put a bet on which future is more likely, which future do you think is more likely?

Oligopoly. This is how the cloud—well, this is how the cloud played out. I think probably the best analog we have is the cloud, right?

The other companies that are behind the models can subsidize these things arbitrarily. I think about Gemini, and they don't have to do this in a way where they have the same economics as an independent company. If you look at how the cloud played out, AWS was at something like 70% or 80% market share early on. Nobody thought they could ever catch up to AWS. They were the massive market leader that created the category. They had way more dominance than Anthropic has now.

Microsoft and Google said, “This is an important, big market. We have to be in it,” and they just basically spun their way into it. Then you ended up with an oligopoly in the cloud. I see no reason why that won't happen here.

Gemini 2.5 is a great model. If you actually look at price performance, I would say that in many use cases, it's the one I actually use as my standard model. It's better than Anthropic for some use cases if you take price performance into account, and Google can arbitrarily subsidize that, too.

Never count out OpenAI. They started the party. They haven't had a major model release in a while, certainly around code, so that's going to show up. I just feel like the players, the money behind the players, and the fact that these models distill so easily will result in an oligopoly. But I don't know; that's just my guess.

To what extent do you think the large model providers in 10 years' time have already been created, or are they yet to be founded?

I think that you end up with models with different flavors, and there's going to be a lot of new-flavor models that come out. We haven't even seen all of them. Mira Murati and Ilya Sutskever are out there creating models. We've got these very legitimate teams that were some of the pioneers. We're just starting up models for the sciences.

As you get more into RL territory, these models really get a certain flavor. They don't generalize nearly as much. That's going to naturally, from a technical perspective, fragment the models.

For the core base model—for language, search, and code—I think it's still very early. It's very, very early in the supercycle. In previous supercycles, it took 2 or 3 generations for the winners to emerge.

Google was third-generation search. Facebook was third-generation social networking. Remember, there was MySpace, and there was Friendster before that. I think there's a lot of change to come.

I do think that both Anthropic and OpenAI have done a remarkable job with brand, incumbency, and market share, so I suspect they'll continue to be stalwarts in the industry.

Are you in either of them?

We're investors in OpenAI.

3. Are AI Models Actually Terrible Venture Investments?

Got you. My question to you is, fundamentally—there are many elements to it—but do you think models are fundamentally good investments for venture firms when you look at employee stock compensation and the dilution that comes from it, and then the dilutive nature of the businesses?

Yeah, it's a hard sell. If there's one thing I've learned, honestly, for anybody who's listening, this will be worth your time.

There is no one way to think of AI, and there is no one way to think about models. The models themselves are entirely different businesses depending on how you talk about them. So, to even answer that question, we have to tease apart what you mean by “model.”

For example, if you look at diffusion models, like ElevenLabs, Midjourney, Black Forest Labs, and Ideogram, these are wonderful businesses that have great economics because the models are smaller. The ecosystem isn’t subsidized in the same way. Google subsidizes language, code, and video, but not speech.

From an investor perspective, these are clearly great investments if you just look at the metrics alone. On the other hand, the frontier language space is much more complicated because there’s so much subsidization. You have Meta and Google, as well as a bunch of Chinese players, entering it.

For a subset of the players—and this is why it’s a tricky question—you’re like, “Yeah, clearly these are the fastest-growing companies we’ve ever seen. There’s tons of value.” These are very valuable entities, right? Anthropic and OpenAI. But at the same time, even 3 years in, there have already been a number of companies that have had to exit early.

I would say it’s a high-stakes game where the winners really win, but it requires a lot of capital to enter the game. If you’re not one of the leaders, that capital is forfeit.

We do a show every week with Rory O’Driscoll and Jason Lemkin. Rory very aptly, I think, just said, “Listen, with the transition to AI, every investor has just accepted a willingness to go massively up the risk curve on investing.” Do you agree with that?

I think it’s the requirement of the game. These are very capital-intensive companies to build. They have to get the capital from somewhere. They’re also the fastest-growing companies. For the winners, it’s justified.

I don’t think investors are willing to go up. We’d be very happy not to. I know you would, right? It would be great to have great returns with low risk. But the nature of the system and the game we’re playing requires it.

This is the dissonance in all of this, and it’s so important to call out. On one hand, you do have these great businesses that are very fast-growing, and zero-sum thinking has been tremendously wrong. NVIDIA is continuing to grow in value. The hosting providers, which everybody wrote off as non-defensible businesses, continue to grow in value. The model companies—which I can’t tell you how many investors wrote off; this question has been around for 3 years—continue to grow in value. Every layer of the stack continues to grow in value.

On one hand, you’re like, “It’s all working. You should be in the leaders in every layer of the stack.” On the other hand, we’ve seen tons of wipeouts already for the non-leaders. It’s almost this bipolar or paradoxical situation where you have to play, but it’s very, very high risk. If you don’t play, you’re missing one of the fastest growths in value that we’ve seen in 20 years.

Do you think you see the concentration of value in 1 or 2 players across markets, in every market? Whether you look at voice, it’s obviously ElevenLabs; whether you look at it, it’s kind of Replit and Lovable, and OpenAI and Anthropic and Cursor.

This is such a great question. Here’s one thesis. It’s early, so we don’t know, and maybe in a month all of this gets proven wrong. But we actually talk about this a lot internally, and here’s one thesis that I’m attached to: These markets are so large and they’re growing so fast that we’re actually seeing brand effects take place. We haven’t seen that since the internet.

By brand effects, I mean that if you become the household name, you will get the adoption because it doesn’t require a lot of education. It doesn’t require a lot of competitive discussion or competitive positioning in the field. For many of these models, is one better than the other? Yeah, maybe, but they’re pretty close. People know ChatGPT. It’s a household name. My mom knows ChatGPT.

Honestly, why did I do Lovable? For the exact same reason that GPT wins. I thought it was the consumer brand that would win.

100%. I just think these markets are so large that brand effects work.

Let’s talk about Midjourney. Midjourney was the first that got above the quality bar. It’s taken zero investment from institutions, it’s still the market leader, and it continues to do great. Meanwhile, a bunch of other people have entered the market.

I do think it’s not unreasonable to assume that these markets are very large, that the leaders are going to have brand monopolies and brand moats, and that they’ll be able to maintain them until things slow down.

In general, I’ve found that markets do this: Markets tend to expand and then contract. Think about cloud, right? It was kind of this funny thing, and then it became very massive, and then, of course, it slows down. When it slows down, you have the consolidation, and then competitive dynamics come in.

We’re clearly in a massive market expansion phase. It’s very clearly the case, in which case the leaders are going to continue to have a distribution advantage just through brand recognition.

When does that tail off, or does it not tail off? When does the importance of brand and brand recognition dwindle, and product prioritization or product quality trump it?

I think it’s as soon as market growth slows down. Let’s take cloud as an example.

Do you think these are actually tools of market growth, or are they just consumer intrigue? There are a lot of people who want to try building a website on Replit or Lovable or Bolt, or any of them. There are a lot of people who want to try voice with ElevenLabs. To what extent is it market intrigue versus expansion of the market?

I just think the expansion of the market provides the dynamic so that you don’t saturate the user with competing messages. The idea of market expansion is that the frontier continues to expand, and the first thing the frontier hears is the household names. So the household names win.

That’s a natural artifact of expansion. As soon as the expansion slows, that frontier is going to hear both names. Then, all of a sudden, you’re in a discussion of which one to use and which one not to use.

For the longest time, when the cloud market was expanding, everybody knew AWS. It was the leader. It had 70% to 80% market share. As soon as that growth slowed down, market share started to shift dramatically, and it just wasn’t obvious anymore. Do you use GCP? Do you use Azure? And so on.

I would say that’s less an artifact of Google and Microsoft deciding to enter the game, and much more that the market growth itself started to slow down.

So we see market growth slow down, and then we see the dispersion of value across players more?

That’s right. The market slows down, and once that happens, the frontier becomes more saturated, just because we’re not adding people as much. They will get more of the educated message, and they’ll start making more decisions. You can have more of a conversation.

Of course, Anthropic would love to have the same brand as ChatGPT, as a household name. But how do you reach that frontier? If it’s growing that fast, it’s operationally tough to do. The only way you do it is through brand recognition, which is this word-of-mouth-y type thing. It’s on every podcast, and it’s through friends and whatever.

I do think we’re seeing brand effects happen now, and we saw these in the early internet. The brand leader tends to get 80% of the market. It just tends to break out Pareto for a while, and then over time it slows down. These things even out based more on product differentiation.

How do you factor that into your thinking when investing today?

You just try to invest in the leader, and it’s worth paying up for the leader, honestly.

For me, I ask 2 questions. Question number 1 is: For the area that it’s focused on, is it the leader? If it is, it’s definitely worth paying up.

The second question is whether the story has actually been that, in a competitive space, almost everybody just found a new niche-y white space. Let’s take the example of OpenAI. OpenAI was the first to code with GitHub Copilot. They provided the weights, as far as I know, and they lost that. They were the first to image with DALL·E, and they lost that. They were the first to video with Sora, and as far as I can tell, they lost that.

And yet they’re still the massively dominant player in language, and they continue to be so and will be so. Arguably, that was the right thing for them because that’s by far the largest market. OpenAI acted totally rationally and has the largest market.

But that gave Midjourney the ability to take image, or Black Forest Labs to take image. Google seems to have grabbed video with V3. Code—I mean, on the model side, Anthropic has turned that into a wonderful business.

When markets expand, not only do you have these brand effects that we were talking about, they also tend to fracture a bunch. What seems to have been a submarket will emerge as a leading market.

You even see this on the image side. You’ve got a bunch of viable image players that focus on different things. Ideogram is great for designers and the professional design community.

BFL is for the open-source community, especially for developers who use these things in products. Midjourney is for more of the fantasy—also professional designers—but it’s a very stylized, opinionated view. All of these are independent, viable companies. I think we’re going to see fragmentation for quite a while before we see consolidation.

I need the show to be successful because I’m very open with my troubles. I need your advice. Abridge in the US—I’m not sure if you’re invested in it, but I’m sure you know it—is very simple.

There’s a European player that does medical transcription for nurses. They went from 1 million to 8 million in a year, and we’re looking at leading their A. I’m thinking exactly the same thing: you’re going up against Abridge because you’re going to need to compete in the US, as this is going to be a big business. Is that a losing game when you’re a European competitor?

This is a great question. Another very interesting thing that we haven’t seen in a very long time is geographic biases showing up with AI. The regulatory environments are quite Balkanized, and there are language and cultural biases that are also Balkanized, so we’re actually seeing a lot of regional players show up.

4. Why it is BS to Denigrate AI Apps for Having Low Margins

I think it’s very legitimate. The thesis cannot be, “European company X wins the American market,” but I promise that, when it comes to AI, the European market is large enough. I think a very legitimate thesis is that this becomes a regional player in Europe and then maybe a portion of the US market.

Can I ask you: a lot of people denigrate these businesses that we’ve discussed because of their margins? They’re simply pass-through funnels to the large language models. Do you think that is something that changes over time? It’s the same for all great businesses: Uber started off with terrible margins, and now they have better margins. Is it the same?

I just don’t buy that these are endemic to the business model. This is certainly not my experience at all. There’s always this question: if you’re a founder and you get access to relatively cheap private capital, and you could make a trade-off between margins and distribution because it’s land-grab time, what would you do?

The argument is that the incremental user is someone you can monetize forever down the road, and if you don’t get that user during the land grab, you can never monetize them. The rational business decision is to sacrifice margin for distribution. It’s just the rational business decision, and we’ve seen this forever.

The web wasn’t even monetized, right? Literally. At this time, we can actually monetize these things, but forget break-even or negative margins—it was massively negative because we didn’t even have a business model until advertising came along. This is the most rational thing that markets, at least tech markets, have been doing forever. It’s no different this time with AI.

I do think there’s a question of, if you do want to turn on margins, how do you do it? You’ll either have to build a traditional moat, a two-sided marketplace, a brand moat, or the long-tail kind of integration and domain understanding.

Let’s say you’re a healthcare company. If you really crack the European market and understand all the regulation, Anthropic is not going to take the time to do that. There’s clearly pricing power you have on that side. Or you have to do actual technical differentiation.

One thing that we’re learning is that, in this phase of model scaling, a lot of the approaches to scaling don’t generalize. If I want to be much better at coding, I may not be so good at something else. This gives application developers a ton of room to build their own models that service certain areas the large models just aren’t focused on. I think there’s even a ton of room to differentiate at the technical level.

My sense is—and again, I don’t want to talk too much about my portfolio and what I see, just because there’s sensitivity around the number of companies—but in my experience, most of these companies that are, let’s say, at break-even margins, it’s a board-level-specific choice to prioritize distribution, not just because this is systemically something they have to do.

We mentioned sovereignty. I’m intrigued by how you think about safety around AI and models. You’ve had Vinod Khosla say, “We have to lock this down. If this was not locked down, it would be like nuclear secrets being handed out.” I remember then Mark came and was like, “Fuck that. No way.”

Yeah.

How do you feel about the future of safety within this landscape?

It’s crazy to have VCs talking against open source, right? Founders Fund did, too. For me, it’s just wild when pro-innovation sectors of the economy—and academia, too—have decided that open, transparent innovation is somehow an antithesis of safety.

I know that’s not what you asked, but I just want to make the point that we were in very Bizarro land for a while, and it seems like we’re coming out of that now. So let me just draw a bit of a parallel.

Do you think we’re coming out of that? I think we’re moving more and more into that. Meta, Mistral, and xAI are going to town; Llama fully closed.

Great. Let’s go back to that in just one second. I want to answer the question that you actually asked, because you asked a great question about how I view this, and then let’s go to whether we’re coming out of it or not.

So, how do I think about safety? I was actually very close to security during the rise of the internet. I worked for the intelligence community. I worked for Lawrence Livermore National Laboratory, and when I did my PhD, about 50% of my work was in security. I taught a cybersecurity policy course.

The thing with the internet is that you had these very specific examples of new types of attacks that impacted nation-states. Critical infrastructure would go down. You had things like the Morris worm. You had these really significant examples, and that kicked off this large discussion on how to handle it.

It was so significant at the time that, at the nation-state level, we started to think that we actually had to change our doctrine. We had this Cold War-era mutually assured destruction, and we had to change it to this notion of defense asymmetry, which meant that the more we relied on these things, the more vulnerable we were, as opposed to a country that didn’t rely on them because you could be attacked.

Then, of course, there was the whole terrorist information-warfare stuff. The implications were so absolute, and you had so many proof points that you could articulate them incredibly well.

If you look at the AI stuff, every computer system has security considerations, but we’ve got this 30- to 40-year, very robust discourse around this that we can draw from and use. The thing that I don’t understand is how, all of a sudden, we’ve decided that these are not computer systems. They don’t obey the same laws, and we have to throw out everything that we’ve learned and revisit the discourse, even though we don’t have the same proof points.

Nobody can make a strong argument on asymmetry or the need for a shifted doctrine. If they can, let’s go ahead and have that discussion. I still have yet to see the dramatic new attack. It’s going to come for sure, but we haven’t seen it yet.

I just feel like the discourse around this is not in line with the reality. It’s not in line with historical precedents. We should absolutely take these things seriously, but we should draw on the information we’ve learned from in the past and the approaches we’ve taken in the past.

The biggest difference this time is that, in the past, the people creating the technology were pro-tech, and the people selling security solutions were the fearmongers. You’d have somebody create the internet and say, “This is safe, and it’s great for everybody,” but then you’d have somebody create a firewall and say, “Oh, the internet’s dangerous—every sociopath, your next-door neighbor.”

You had the same voices, but in 2 different bodies based on their interests. The interesting thing this time is that they’re in the same body. The person creating the thing is also saying, “This thing is very dangerous.” I don’t recall the last time we had something like that, but it’s created a dynamic that’s just been very confusing for everyone.

Do you not think open source increases the opportunity set for hostile actors like China and Russia to harm us? I think it’s tautologically true. You can say, “Do you believe computers and the availability of computers increase their ability to harm us?” and I’d say absolutely: computers and the availability of computers do.

Specifically, open source over closed source?

I think that right now, open source is most dangerous because China is better at it than we are. As a result, we’re seeing a proliferation of Chinese open-source models everywhere.

Unfortunately, we don’t have control over Chinese regulation. I would say the answer is yes because of China, not because of us. The right way for us to respond is to fuel our open-source efforts against that.

So, let me just be very specific. Chinese open source can be a national security issue, for sure, and any of the software that's produced by a nation-state that we view quasi-adversarially. The way that we combat that is we also are incredibly open, and we also proliferate technology.

5. “Open Source Is a National Security Weapon And We're Losing”

What do you think we can learn from China's regulatory approach that would enable us to have the same or better open-source ecosystem?

I mean, to me, the United States has a long history of being pro-innovation: pro-innovation for national security and pro-innovation for national defense. I think we should be funding this stuff like crazy. I think we should get the national labs involved, and we should get academia involved.

We should make this a national priority, just like China does, and we should give a full-throated endorsement of all of this stuff. I think we should do closed stuff. I think we should do open stuff, and we've done this forever.

My first job out of college, in 1999, was working at Lawrence Livermore National Laboratory in the ASCI program. What were we doing then? The broad program was simulating nuclear weapons. This is what it was.

6. “The Oligopoly Is Coming, Just Like Cloud”

A lot of the concerns we have today were concerns we had then around compute. We actually stopped Saddam Hussein from importing PlayStations because we were worried about using them for simulation. We put export controls on the hardware, and we'd say the same things: “Computers are out there. Computers are going to enable the enemies,” and all sorts of stuff.

This is nuclear weapons. This isn't some abstract AI thing. This is actual, on-the-ground weapons. The posture that we took at the time, and the conclusion, was that we're just going to be the leaders in all of this stuff. We funded academia, we funded the labs, and we won. We were able to control the technical discourse of the planet going forward.

This time, instead, we want to put our head in the sand and let somebody else do it. They're going to learn from our success, and somehow we're not.

Do Trump's cuts to universities and research labs not impact your ability to do what you just said? Are you not actively going against what you should be doing?

I am very pro-investing in academia and in the national labs. I think there's always a political shift in money depending on what they view as being in line with administration politics.

I did my PhD at Stanford. I've done a bunch of NSF grants. I don't remember anybody ever saying, “We like indirect costs.” Every researcher, every professor, every single one was like, “Indirect costs are terrible.”

Obama tried to get rid of indirect costs. He said, “You know what? Universities have tax-exempt status, so why don't we just have them spend 5% of their endowments, like any other tax-exempt organization?” That would cover a lot of indirect costs, and he couldn't get it through.

This is a bipartisan issue that is longstanding. I would say that a change is needed now. To the extent that these things are very hard to implement, I would say concretely, yes, we should invest in these things. Yes, we need a shift in how funding happens.

I do think that indirect costs have gotten way out of hand, and until it was Trump doing it, everybody that I know in academia totally agreed. But yes, of course, changes and shifts in funding will be disruptive, and so I think all things are true.

I just don't want to reduce this to a simple “Trump does bad things,” because I don't think that is the case, and then funding science is arbitrarily good, because I don't think that's the case. I definitely think we should fund as much or more. I definitely think that a shift in funding and a change to the system is needed, and that the right path through that is complex. I don't quite know it.

You very kindly said that I asked a good question on the reversion back to closed source when we mentioned Alex joining Meta and what it meant for Llama. I said, quite zero-sum-wise, to your point, we're clearly seeing a movement back toward closed and away from open.

How do you see that, and do you disagree? Is my statement there on the transition?

No, I agree on the ground 100% that I think we're seeing a movement away from open source, but the rhetoric around open source has shifted, right? We just had America's AI Action Plan, which is a full-throated endorsement of open source. So, discourse-wise, there's more support for open source than ever before.

Ecosystem-wise, I think you're right. I do think it's quite likely that we're going to see less open source. Now, listen, OpenAI has said that they're going to open source. That would be wonderful, and if they do that, I think that would be very, very positive.

Do you think they will?

I have no idea. I hope so. Maybe here's the thing: we say “open source,” but it's such a misnomer when it comes to AI.

The standard model of open-sourcing AI is that you open-source the smaller model and keep the more capable model closed-source. It's a way that you get distribution and brand recognition, but you don't actually erode your business.

This has been very successful as a business model. Unlike actual software, just because you release your model doesn't mean somebody can replicate it. To replicate it, you'd have to recreate the data pipeline and the training pipeline.

I think there's just a lot of concern about investing hundreds of millions of dollars or billions of dollars to train something and then just giving all of that away. But I feel very confident that the business justification is there, and behavior will always follow business.

We're going to continue to see open source be a large part of the ecosystem. Remember, historically, open source has only been about 20% of the total market value. I would say it's much higher than that for AI, so in a way we're doing better than software has historically.

What did you believe about the AI landscape that you now no longer believe?

We've touched on so many different elements. My mindset has changed around so many of them. The one that I've just consistently got wrong is how fast these coding models advance.

This is probably just sunk-cost fallacy. My entire life, I've just been this nerdy programmer. I've been programming since the '90s. It's my happy place, and I never thought that they would advance to the level that they have.

I still develop most evenings. Instead of watching a sitcom, I just goof off, mostly writing old video games or whatever for fun. It's silly stuff, and I'm already at the point that I just couldn't go back to working without them. I've spent 30 years without them.

Their ability to offload all of the shit I didn't want to learn is remarkable. The thing that kept me away from code for a while—I would dabble with it and then drop it—was that you had to learn all of these weird frameworks.

None of the knowledge was foundational. Some random developer came up with some weird way to do something, and you've got to learn some poor design decision to do it. None of it made any sense. It felt like you were wasting your brain space on poor decisions made by random open-source developers.

In the late '90s, programming was: you'd download your IDE, sit down at your computer, program something, and then it would turn into a binary, and you'd run that binary. You could get a lot done just by sitting down and writing code.

By 2015 or so, writing something meant you'd have to download 50 packages. To run it, you had to run some stupid dev server, and to have anybody else use it, you had to learn how to host it. It was a bunch of libraries dealing with incompatibilities for what was a weird fucking platform.

Ninety percent of your time had nothing to do with code. Ninety percent of your time was just dealing with all the environment and platform shit.

What's so nice now is that you can just focus on your code. I use Cursor, and I just have the AI tell me how to host the thing and tell me what package to use, and whatever. I strictly focus on what I want and the logic.

It's almost like it's brought coding back. You can see this across the industry. I've grown up in the industry, and I know a bunch of very strong developers who've been developing for a very long time. They've basically stopped developing because they're running companies now or whatever, and they're all back to programming at night.

I really think that there's the adage of the old man who goes into the garage and makes the train set for nostalgic reasons.

7. Does AI Make 1x Engineers 10x or 10x Becomes 100x

I think the modern version of it is these old systems programmers vibe coding at night just because it’s become pleasant again. And so I know you asked about the thing that’s kind of surprised me the most, but I really think it’s such a marvel what these coding models are able to do, and they add very real value.

Do you think they make 1x engineers 10x, or 10x engineers 100x?

I don’t actually think it’s that. I think they make 10x engineers 2x. I would say every company I work with uses Cursor, right? And then if I actually look at whether that has increased the velocity of the products coming out, I don’t think that much, just because so much—

So what’s changing, then? Developer productivity is going up, so is the quality of the product going up if the product release cadence isn’t—

I just think the things that are hard remain really hard. Let’s just talk about creating a model. Let’s say I’m creating a new model, a new frontier model, right? To create that new frontier model, I’ve got to collect data, I’ve got to run a pipeline, and I’ve got to sit with my Jupyter notebook and look at the loss curves and rerun it.

That’s just a lot of experimentation and so forth. There’s no coding model that’s going to do that for you. But if I wanted to create tests or a test suite, visualization, or write documentation, it’s actually really good at that. I would say that probably, in the long run, having more robust, maintainable code bases with fewer bugs is just as likely to be the impact as feature velocity.

Because in startups, again, I’m an infra guy. This is probably different for the apps. I’ve always thought apps had no technology to begin with. Every time I look at vertical SaaS, I’m like, “Why do we even care about the technical team? It’s fucking CRUD, man.” CRUD is create, read, update, delete. They all do the same thing. They all just kind of look like a web app. They’re all—who cares about the technology?

The technology is simple. These are all kind of go-to-market things and whatever. But infrastructure is different. Infrastructure has very real trade-offs in the design space that only somebody who understands computer science would know. So for infrastructure companies, I think it’s quite unlikely that AI will really help speed that up, because it comes down to something that the developer has to decide on and has to articulate the trade-offs.

But I do think it could really help with the development process, so you have fewer bugs and things like that. I actually view it more as a more robust development methodology that necessarily speeds up the core product.

Given the kind of developer productivity changes that occur because of these tools, how does that impact defensibility within companies today? If time to copy—which Misha said on the show has basically been reduced to nothing—to what extent does that change defensibility for companies?

I still think we should just go back to the split between apps and infrastructure. For apps, how long does it take to copy it, anyway? There are entire companies whose stated purpose is just to copy another company in the app space. It’s just so easy to do.

There is no core technology for a random app. There’s no differentiated technology for a random app. Let’s say that you’re creating, I don’t know, some healthcare vertical SaaS thing. You could contract out the actual app, and you have been able to forever. The business is actually the long tail of understanding that domain, so I just don’t think it changes that paradigm at all.

When it comes to core infrastructure, which is what I focus on—things like databases and foundation models—there’s no way that, right now, models can just copy. The reason there’s no way is not that the models aren’t capable of doing the technology. It’s just that there is a long tail of understanding the trade-offs for the particular use case and domain.

Because it’s a new market, often you understand that through market exploration. I just don’t feel that these models really help with the software development process for non-deeply technical areas like apps. Sure, they can help speed it up, but over time, all of these reduce to a long-tail understanding of the market.

I mean, Aaron [Levie?] said it so beautifully. Do you know what the average PR—pull request—is for a production code base? How many lines of code is the average change that gets accepted, would you guess, for some production enterprise app?

I have no idea.

It’s 2. It’s 2. Yeah, it’s very, very small. It’s actually 2, but let’s say it’s 12, right? And what do those 2 or 12 lines signify? Those 2 or 12 lines signify probably some learning in the field or some understanding of what is needed.

The long tail—the hard thing—is to understand the specific deployment environment and market you’re going to. That’s the hard thing. The hard thing isn’t the 2 lines of code. That’s actually quite easy.

In many ways, I would say AI is getting rid of the middle. They don’t know how to do very new computer science, like models, just because nobody’s done it before, and that’s kind of pushing the state of the art. Then, in the app space, all of the hard stuff is the business anyway, right?

This is why the changes are very small, and you learn everything through go-to-market, which the models don’t know just because you’re exploring a new market. It’s all the shit in the middle that they’re helping us with, and so for me, it’s just kind of net creative.

Do you think computer science holds the same weight as a field of study, as an educational discipline, that it always did, and would you always recommend it? Or does that change in a world that’s, bluntly, more democratized in terms of creation, like we discussed?

I feel very strongly that if you care about building systems out of computers, you have to understand how they work.

What do you think we do today, Martin, that we will look back on in 5 or 10 years’ time and go, “I can’t believe we did that”? It could be prompting. It could be choosing the model that we’re working on. I find it ridiculous that we’re supposed to choose which model. Like, Grok 3, Grok 4, Grok 5, Grok Shopping, Grok Weather—what the fuck? Just figure it out.

I’m just taking it from a programmer’s view. Hopefully, we’ll just stop worrying about frameworks altogether and maybe even languages. Maybe even a proto-language evolves, and we can just focus on logic and fundamental trade-offs.

We’ve gotten into this very backwards world where, these days, programmers think about all the nonfundamental stuff, and they don’t think about the fundamental stuff. Let me give you an example. This is going to be a weird philosophical rant, but I always worried, while I was doing grad school and when I was doing research, that we kind of entered a space where there’s so much research that has been done over the years that you never know if you’re doing something new.

You just couldn’t do the literature search. There’s so much. The entire industry just spent all its time redoing research. It’s like you’re cleaning a room and trying to sweep out the dust, but rather than sweep it out the door, you’re just moving it. You move it to the bed or you move it to the wall, and that’s all you do: you just sweep the dust around, but you never actually get it out of the house.

That was what research felt like to me. It was like we were in this mad delusion. On top of that, it also felt like many of the most important problems were between disciplines. In order to even solve them, you just have to know too many things, and we couldn’t do that.

I felt like the entire scientific-industrial establishment was just redoing the same stuff. In a way, I think AI has the ability to pull us out of this mass craziness, this mass ineffectiveness, because it’s very good at telling you if you’ve done it before. It actually knows all the literature and knows all the history, and it’s also very good at tying different disciplines together. It is an expert in all of these things.

I think we’ve been stuck in this morass, and it’s a bit of a liberator, so we can actually focus on the new problems and know we’re doing new things. I’ve got this very optimistic view of where it’s pulling us. I know it’s more of a philosophical answer to the question that you asked, but in a way, I think it needed to happen to get to the next level of problems that we need to solve.

8. “We’re All Dead Wrong About AI and Job Loss”

In terms of societal implications there, the worst question ever is the job displacement question. But I am intrigued because, on the one hand, I see intense job displacement happening faster than ever, and then I’m also very aware of—Brad Feld wrote a brilliant post where he basically said every single cycle, every time, we’ve always said, “Oh, what are we going to do? Calculators, what are we going to do? Computers, what are we going to do? AI, now what are we going to do?”

To what extent does this actually require the “What are we going to do?” versus another…

For God's sake, don't we see the pattern?

Yeah. Listen, I'm very sympathetic to concerns around job displacement. I think we should take them very seriously as a society. I'm in no way libertarian. I think this is kind of where governments do step in and we help out.

But first, we have to understand it, and it's actually very unclear. So, let me tell you a quick anecdote. My cousins are all pretty—I think “high-end” is the wrong term—but they're pretty established translators. They've been for a long time, working across multiple languages.

They visited recently. This is a husband-and-wife pair, and they're like, “Listen, we have to change jobs because translation is all going to AI.” I asked, “So, the jobs are going away?” They said, “Well, no, they're shifting. Now, instead, we've got to spot-check these AIs, and the only way we can hold them up to our standards is if we rewrite the entire thing, but they won't pay for that.”

These are Italians, so they speak this way, but they're like, “I can't work on something without a soul,” right? I think their dilemma is a good microcosm for the broader dilemma. One thing that's very unique about AI is that it actually requires a human handler today. They're just so unpredictable.

Most of the use cases that we know—all the monetized use cases—have a human on the other side of them, right? In coding, you've got a professional coder. In all the creative stuff, you've got somebody doing all of the creation. These are kind of an enabler and a tool, but the nature of what you do does shift.

That's very different from, for example, electricity, where it doesn't require a human. It's like either you light the fire or there's no fire to light. I think we, as a society, need to understand the level of displacement. We have to understand it. I think it's very important that we do, and I think these are things that governments should get involved in.

I do just have to turn to your venture investing before we do a quickfire. Do you enjoy it as much as you did before? It's a much faster landscape. The money is much bigger. Do you enjoy it as much as you did before? I spoke to many of your founders, and they said they didn't think you enjoyed the administrative work that you now have to do with the size and scale of Andreessen Horowitz.

Oh, well, those are 2 different questions. I love the investing. The investing is great. This is the most exciting time in the industry since the late 90s. It's great to be part of a supercycle. I love it.

Actually, I really like the firm-building side. Frankly, I could do without endless meetings, but I've actually been pretty good at limiting those, too. I think this is the most exciting time to be in the industry and venture.

No, dude. I'm a venture investor, too, so I'm with you, and I say the same to our LPs. Is your price elasticity more on deals because of the supercycle entry point that we're in, or less because of the risk or uncertainty level that we're in?

Philosophically, for me, I just think the market sets the price. I don't have the hubris to think I can somehow outsmart the market or that a single deal is going to bend to my will. Philosophically, that's how we think about investing in general—

Walk away because of price often?

Price? No. Ownership? Yes.

What is the ownership you need?

It all depends on the fund, the market, the size of the market, and the risk. Everything comes down to ownership for us, not price. You just can't make the fund mechanics work if you don't get the ownership.

For very, very large markets and very large checks, we don't care as much. But that tends to be growth territory anyway. For early-stage investments, you need to understand what the median outcome is, and you have to be able to size the median outcome in a way that at least returns, say, a fifth of the fund or half of the fund.

Is that not the joy of being at Andreessen Horowitz? You can take 5% ownership on the first check because you can size up into the next and size up into the next. Is it not my challenge that I have to get as much as possible on the seed or the A?

The way that I view it is a bit different. I think there are 2 legitimate ways of investing that have emerged. One of them is that you're very much a specialist, and you've got your special network, special value, and you understand a special size of the market. You're very much a specialist, and that's kind of how you win deals, get the ownership, keep the ownership, and then make your company successful.

The other one is—and I wouldn't say it's an AUM thing—but you have all of the products so that you can be adaptive in the market. I've been doing this for 10 years, and the strategy that works has shifted this entire time. Sometimes it's early, sometimes it's mid-stage, and sometimes it's collaborating with growth.

If you don't have all of those products—honestly, sometimes it's credit, which we don't have a credit fund for, but I can understand why people do it—the market is competitive and everybody's scrambling for deals. If you don't have the different funds or products to offer, then often that's where people are going to squeeze you out or get alpha, et cetera.

For the game that we play, it's very, very important that you have all of these funds and the ability to enter at all stages for exactly that reason. Again, I don't think it's a you-me thing. I think you play a very different game than we do, because on one side, you have to go very specialized, very focused, very early, whereas for us, we're trying to find out what the right time to enter is to get the ownership that we need.

What's the size of the fund that you primarily invest out of day to day? I know you have flexibility.

$1.2 billion. I run the infrastructure fund, which is a $1.2 billion fund.

My challenge here is that your cost of capital is so much less than mine. Your ability to put in a larger check, bluntly, with much more confidence, is there, because I'm investing out of a $275 million Series A fund and a $125 million C fund. It's just much more meaningful dollars for me than it is for you, which will affect my willingness.

Yeah. My challenge is that we have to live with these investments forever, and conflicts are very, very difficult for us to manage. We don't enter very often at the stage that you do for this reason.

I mean this respectfully: everyone chastises Andreessen Horowitz for their conflicts and for investing in many conflicting companies. Do you think that's unfair?

It's so hard to keep your nose clean on this one, because especially with a shift toward AI, companies pivot all the time after you invest. I don't recall intentionally investing in a conflicting company. In fact, I would say one of the top reasons we don't invest in companies is because of conflicts.

We do it. I just did it recently. I can't say the name of the company, but we didn't invest because it was a hard conflict. Even though, by the way, the portfolio company was not doing the thing, it was on the roadmap, and the founder called me. He said, “Martin, you just can't invest in this company.” I said, “Okay.”

Okay. Sorry, sorry. Just to push back on you there: if it's not on the roadmap, I'm really sorry, founder. I have as much faith and conviction in you as possible, but if it's not on the roadmap, I'm not having you tell me how to do my job.

Here's my talk track, and it's evolved over the years. I stole this from Chris Dixon. I say, “Listen, you have 1 mortal enemy, and you choose whoever that mortal enemy is. Whoever it is, I'm with you. We're going to go kill that mortal enemy together, but you get 1. You don't get an arbitrary number of mortal enemies.”

In this case, I'm like, “Listen, is this it? Is this your 1 mortal enemy?” The founder said, “Yes, this is the 1 mortal enemy.” I'm like, “All right, fuck them. Let's go kill them.” That's it.

We have a number of companies where they pivot midstream and start competing after we've invested. It happens all the time. We also have the venture and growth funds. We try to minimize conflicts there, but sometimes they happen.

They're companies at very different stages, with very different teams working on them. I would say that we try very hard to steer away from conflicts.

Given the nature of, as you said there, the volume of pivots that occur today, and given your entry point, I always advocate wholeheartedly for being 98% founder. Then you have wonderfully smart people like Elad Gil who wholeheartedly advocate for being market-first. How does the pivot frequency and the experiences you've had impact your prioritization mechanism around where you spend time?

I don't want to speak for Elad, but that's not my experience working with Elad, and I've done many deals with him.

Elad is very focused on the founder. I think the one thing I would say is that he's very good with founder-market fit, maybe the best in the industry. I have huge respect for how Elad invests.

Unpack that. Why, and how does he do founder-market fit better than anyone?

He will find a market that he really likes, and sometimes it's even a fast-follow market, right? Then he will find who he thinks is a great founder for that market. He's very good at this kind of boy-band construction based on the market.

The primary point I want to make is that, in his investment cycle, the founders have always mattered in all of this. He's followed on deals I've done. I've followed on deals he's done. We've done a bunch of deals together. I've never gotten the impression—I mean, I've actually always gotten the impression—that the founder is the primary decision once he's chosen the market. So I would say it's a primary concern for him.

9. “The Only Sin in Venture: Backing the Wrong Winner”

When you have misjudged a founder, what did you not see that you should have seen?

Can I ask you a previous question? You're like, “Okay, so how do we think about it?” We think about it very simply. The only sin in investing—and I've sinned so much—the only sin in investing is missing the winner. There's no—it's fine to invest in a category that doesn't work. It's fine to lose money, but if you choose the wrong company, that's not okay. And listen, it's just so hard to get it right all of the time.

Someone said to me the other day—I'm so sorry to interrupt you—that Andreessen and you get killed for choosing the wrong company but being right about the space. You won't get killed if you were just wrong about a space.

Correct. That's exactly right.

The view is that there's basically no amount of work you can do to determine if a space is going to work or not. I mean, that's just like weather prediction. But given a set of companies, you can actually do the work to understand which one of those is the best. Now, we've got it wrong constantly.

Do you think you can?

The question is, can you beat the market with that strategy? Yes, I think you can beat the market. No, I do not think that you can unequivocally tell the best. Can you beat the expectation of the market by running this strategy? I would say yes. Can you specifically pick the winner every time? Absolutely not. Clearly not.

When did you most poignantly pick the market but pick the wrong horse?

I just don't want to call out any specific company.

Fair enough. When you think about—you mentioned sins there—what's a big sin that comes to your mind?

Well, I can answer the opposite. There's a bunch of markets that just haven't really worked. The entire streaming market has been very, very tough—the data-streaming market. It's just turned out to be a subset of the batch analytics market. Maybe ClickHouse—Aaron Katz is doing phenomenally with it, and I'm not an investor, but he's doing phenomenally—but that may be the one breakout since Confluent.

Whether you're at the dashboard layer, the transformation layer, or the feature-store layer, there have been entire spaces where we played multiple bets and it just didn't work out. Many, many times we'll invest in a space where none of them work.

I'll tell you, there's definitely been companies we invested in where, at the time, the company was the very clear leader, and then something happened—some macro shift, something else happened. I think that's just how the game goes. You've probably heard this, but the thing with actually having a strategy like that is, if you're trying to scale a venture firm, you just need something that you can articulate and teach other people.

I just find it hard that if you pick the right market and the wrong horse, bad Martin, but if you don't pick the right market, fine. To me, some points need to be given for the insightfulness to pick the right market, and some forgiveness should be shown for that. It's fucking hard to pick the horse. Almost I'd fire the one who picked the wrong market entirely. Where was your insight, at least?

Yeah. And this is why you run your own venture firm and you can have whatever strategy you want.

Is that moronic?

No, no, it's not. I just think it's philosophically different in the approach, right? I actually don't believe you can predict the future of technology adoption. It's a very tough thing. You don't know what a big company is going to do that can wipe out an entire market. You don't know when an innovation will wipe out entire markets. This happens all the time.

You could argue that AI is really invalidating tons of markets, and I don't think anybody could have seen that happen. But if you have, say, 10 companies that have some traction, and you can talk to the founders, diligence the teams, diligence the market, diligence the product, and diligence the technical approach, I think you can say something a lot more concrete than whether some future innovation is going to wipe out an entire market.

Do you think it's paradoxical or opposing to believe that both AGI will be dominant and present in a set time period, and at the same time be investing in enterprise SaaS?

I don't know. I would say humans are AGI, and we still invest in enterprise SaaS. This is the problem: everybody somehow thinks that AGI just means unlimited power, and anything I want to disappear in the future disappears. Come on, you're AGI.

I'm AGI.

We invested in process.

10. Quick-Fire Round

I think, to be honest, Sam Altman sets the definition of what AGI is, so whatever he and Microsoft decide is AGI will be AGI, dude. I want to do a quick-fire round. I give you a short statement, and you give me your immediate thoughts.

Yeah.

What's one of the most overhyped AI categories today?

ASI.

What's one of the worst VC takes on AI you've heard recently?

“Open source is bad for national security.”

What one founder would you back in any category? Whatever they did, I just want to wire them the money.

Michael [Truell?].

Why specifically?

I've worked with him for a year. He's just remarkable. It's so rare that I've found a founder who knows what he wants. He's got an intuition that's impeccable, and he listens incredibly well and gathers information. That's a very potent combination. Then, of course, he's incredibly smart and he's got great product taste.

What's your favorite trait in yourself that has been most impactful to your own success?

Deep-seated anxiety from being poor.

Unpack. Seriously, I agree, but—

I grew up—you name it: food stamps, dirt road. I mean, I come from Montana. It's so funny. People hear the name Martin and they're like, “Oh, he must be some—” I was actually born in Spain, so I'm a Spanish citizen. They're like, “He must be some sophisticated European.” I'm like, “Fuck, dude. I grew up on a dirt road in Montana.” When there was hunting season, my school shut down. I'm a Western country boy.

I had a great family. I didn't have any of those hardships. I had a wonderful, educated family, and we kind of muddled our way through. You go through that and you see how hard your parents work, and you just don't take anything for granted.

I sold a very successful company, and I could have retired on that day. I still have not taken a day off where I haven't worked since basically forever. No, listen, I'll take a week off while I have a job, but I've never not had a job in, what, 20 years. It's just—

Did that day feel fucking awesome, coming from a dirt track and no money? You can retire today. I know you didn't, but did it feel as good as you thought it would?

You know, it's kind of an interesting thing. No, it was very bittersweet. I think actually selling companies is very bittersweet for any founder, right? It's a death in a way. You spend so much time with something, and then it shifts.

Here's the interesting thing, and maybe this is advice to other founders: you always think about that thing you'll do when you make $100 million or whatever. You're like, “I'm going to go do that thing,” but you only think about that thing in the most stressful times.

My thing was—my cousin's a movie director. His name is Vincenzo Natali, a pretty legit guy. I was like, “You know what I'm going to do? As soon as the money hits the bank, I'm going to drive down to Hollywood, and I'm going to help him make movies and be an actor and just kind of be one of those people.”

And so it happened. The wire hit, and I was driving down the 5, and I was like, “What the fuck am I doing? I love technology.”

I love my job. I hate Hollywood. I have nothing in common with these people. I probably got 2 hours out of town and just turned my car around and came right back because I was like, you only have those visions at the most stressful time. When you're not stressed, you realize that there's something that brought you to this place, and it's genuine interest and genuine love of it.

So my only advice to other people going through this is: don't use those dreams that you concocted when you were really in the pressure cooker—not sleeping, your relationships were falling apart, that whole thing. That's not the thing that, in a steady state, you're going to want to do. You're probably where you are because of the love of it, and letting that go tends to be pretty disastrous to some people.

Was making money, or having money, what you thought it would be?

I had to play all of these tricks. I actually borrowed one, which was very helpful. I just had a hard time spending money because, literally, when I got into the Stanford PhD program—this is so embarrassing—we always thought $20 was a lot of money growing up. We'd call it the Yuppie food stamp because it was 20 bucks, and I remember I was going to go to Bytes Café and pay with $20, with a $20 bill, because that's kind of some stamp of having money.

I was just so naive to all of these things. It was very hard for me to spend money. Once I made enough—once I made generational wealth—to do it, I talked to a friend of mine who went through a similar thing. He said, “You know what I did? I came up with—let's call him Brad—I came up with a Bradcoin.”

The Bradcoin—let's say I'm worth 10 times more than an average rich person, so my Bradcoin is worth 10 times more. I buy things in Bradcoins. If it's, let's say, a business-class flight, that's $10,000, but in Bradcoins it's only $1,000, and $1,000 sounds a lot better than $10,000, so I feel good. I actually had to adopt a lot of these mechanisms where I'll make a Martin coin, and it's worth this much money.

What got worse with money? This is something I have to deal with all the time, but, man, my wife forces me to keep it real. She just won't abide by any of the shit. I got 3 fucking dogs that are crazy. She doesn't like help in the house.

I drive a fucking Volkswagen. We have 3 chickens in the back. I'm fucking schlepping the kid all the time. Listen, man, if it were me, I would be living your life, man. I'd be like, 100 percent, being in New York in a penthouse with a private jet. Instead, I'm in a fucking Volkswagen with 3 dogs in a messy house and no help.

Dude, you're so whipped.

You know, it's not even that, right? It's like, this is what marriage is, man. What are your biggest lessons on marriage? I'm 29. I have a great relationship, but I'm not quite there yet. What would you tell me about greatness in marriage that I should know?

Well, listen, I got it wrong once. I'm not sure I'm the right guy to ask here. My startup was really tough. It was really tough, and I think that burned through my first marriage, and she was great. Yeah, fuck, dude—I'm the wrong guy to ask. I'm really the wrong guy to ask.

I will say something, which is a different question than you asked, but I think it's important. I have found that men in particular who have stable relationships just do a much better job at work. They're just much more stable. I think the best founders I have tend to have families, et cetera.

I do think, again, I don't want to make it a gender thing. Maybe it's not. Maybe it's just my observation. I work with a lot of men, and families are really, really, really good for men, even though they can be a pain in the ass. I just think the only high-level view is that these things are super important, and whoever you have and are working with, it's important that it really is keeping you grounded.

I mean, in my case, listen, I mean—

You've got chickens, baby.

I mean, it's like the—what does Zorba the Greek say?—it's the full catastrophe. I know it's the only way I can do what I do. There's no other way, right? The level of pressure, the amount of work that I do—I probably work, all in, 80 to 100 hours a week. I've been doing it for 10 years.

The amount of demands—I mean, it's very, very hard to do without support and grounding. In my case, listen, I'm not the right person to answer how you treat your wife. I just, whatever—I'm a fucking autistic nerd. I have no idea. But I do know that these things are incredibly important for us, and you should value them and treat them as such.

If you think about Andreessen in 10 years' time, where do you think Andreessen will be then? What does the 10-year Andreessen look like? When you remember it 10 years ago, it was a fucking different firm—amazing and innovative in its own time—but from where it is now, night and day.

Where is the 10-year Andreessen in 2035?

The most remarkable thing about the firm, in my opinion, is that it's able to evolve and adapt very aggressively because of the way it's structured. Marc and Ben really are the top of the firm. They really are, and I think it's a feature, not a bug.

I think it's a historical quirk that VC was created around a partnership model. That's the same thing you use for a dentist's office or a law firm. There are positives in that—there are a bunch of different agendas that sit at the same level—but for decision velocity and disruptive change, it's death.

I think that's a massive benefit to the firm. I'm just delighted this is the way it is because they can make these big, aggressive changes. I don't know what it's going to look like in 10 years. I guarantee it's going to look different as it evolves with the landscape.

Martin, I so appreciate you, dude. You are fantastic. You're open. You're honest. I love the last 15 minutes there, but I really appreciate you, man.

Yeah, likewise, Harry. Always a pleasure. You're the best.