In this edition of Author Talks, McKinsey Global Publishing’s Raju Narisetti chats with Chris Dixon, general partner at Andreessen Horowitz, about his new book, Read Write Own: Building the Next Era of the Internet (Random House, January 2024). Dixon chronicles the evolution of the internet and argues that a decentralized framework could promote innovation and growth. An edited version of the conversation follows.
Why did you choose Read Write Own as the title for your book?
The reason the book is called Read Write Own is because the internet is fundamentally a participatory medium, a network that was designed to be owned and controlled by the users and the communities that build software on it and use it.
I see the internet as having gone through multiple eras, with each era deepening the level of participation. In the first modern era, when the internet started to go mainstream in the ‘90s, most people were consuming information—you’d go to Google or some other search engine, type something, and read something.
Then in the 2000s, the internet became much more of a two-way medium, where there was the rise of social media, where people could not only consume information but also publish information. That’s where democratized publishing and the first era of consuming information began.
So the first year you could read things; the second year you could also read and write; and the emerging new era of the internet is one in which users can go even deeper into participating in the digital services they use—by actually owning parts of those services.
That’s where blockchains come in. Bitcoin was invented in 2008, so it’s been around for 15 years. We’ve seen, throughout that period, a lot of evidence that [users are excited about] owning the digital services they use.
Take bitcoin as an example. Bitcoin is controversial. Some people love it. Some people hate it. But, fundamentally, bitcoin is a financial network that’s owned and controlled by the users. This idea has excited a lot of people.
Since then, we’ve seen other digital services created with the same architecture—most notably, Ethereum, and then a whole bunch of other platforms that take that kind of architecture—and expand those ideas of a network that’s user-owned into many other different kinds of use cases.
That’s really what my book’s about. It goes somewhat into the history of the people, the early phases of the internet, which I was part of, and I go deep into the history of that. But then, a lot of the book is about how to take all this energy people have for this new architecture, for being owners on the internet, and how to harness that, channel that energy into productive, positive use cases that benefit the internet as a whole.
Take all this energy people have for this new architecture, for being owners on the internet, and harness that, channel that energy into productive, positive use cases that benefit the internet as a whole.
What did we get wrong when we first envisioned how the internet would work?
One miracle of the internet is that it came out of, of course, governments and academia and was partly designed to resist nuclear war and for a whole bunch of other reasons, so it was built in a fully decentralized way.
It wasn’t owned by a company in the way TV channels and newspapers were owned. It was an open, permissionless network: anyone could build a website; anyone could build a start-up.
And you truly owned that—you truly owned what you built. If you put up a website—you know, you’re Mark Zuckerberg building Facebook and Facebook.com or Jeff Bezos building Amazon—you built that, and you owned it. You didn’t pay a tax or fees or ask permission from some intermediary. It was a truly decentralized network.
This led to a golden period of innovation through the ‘90s and 2000s, because it provided incentive for people to build on top. What happened in the 2000s, with the rise of social media and then iPhones, is that a lot of what was the decentralized network became de facto centralized.
Today, the top five big tech companies now account for 50 percent of the Nasdaq market cap, up from 25 percent a decade ago, in almost all categories of meaningful internet services: social media, search, e-commerce. You have a very high concentration of market share among the top two to five players.
What this means is that users and start-ups and creative people have to go through these large services, these intermediaries, to reach their audiences. And these intermediaries have very high “take rates.” The take rate is the percentage of revenue they take that passes through the network.
They exert their control in a variety of different ways. They basically have figured out how to set up the system so that all the profits and control go to these companies.
What I really focus on is how that affects innovation and how you no longer have the same kind of incentives you had in the ‘90s, where entrepreneurs and creators could build services and go directly to their audiences.
As a result, we’re potentially setting up a situation where we significantly stifle innovation and creativity. I’m very pro AI. I think generally it will have many positive effects on the world, but it is a very centralizing technology.
I’m very pro AI. I think generally it will have many positive effects on the world, but it is a very centralizing technology.
It’s a technology that rewards companies with deep pools of capital and data and, left unchecked, will continue to accelerate this trend toward centralization. Blockchains and tokens and the things I discuss are a natural counter-lever to that.
They are decentralizing technologies that shift power back out to the edges—power, control, and money—and become even more important as new technologies like AI come online. Self-driving cars and VR [virtual reality] and all the other major technologies outside of blockchains that are on the horizon tend to be centralizing technologies. It’s very important to have these countervailing forces.
Isn’t it true that a lot of the recent headlines around blockchain are about fraud?
There are obviously frauds, like FTX [or Futures Exchange, a defunct cryptocurrency exchange platform]. I do think this is true of a lot of emerging technology. You have bad actors who come in and take advantage of it. Notably, none of the issues you read about or that have ever happened have actually been at the blockchain level. For example, you hear a lot about crypto hacks and things like this, but Bitcoin and Ethereum and all the major blockchains have never been hacked, not once.
There are people and organizations that use some of the language and appropriate some of the language of blockchains, like FTX, and commit fraud. With the right policy and regulatory approach, however, they can be handled.
In terms of the positive use cases, I’ll give you a couple of examples. I think NFTs [nonfungible tokens] are very interesting. A lot of people heard about them in 2021 or 2022 and think of them as people speculating, buying art, and other kinds of things.
What’s really powerful about NFTs is they are a way for creative people to sell things and interact directly with their audiences. For example, let’s say you’re a musician. Today, Spotify, by its own stats, has about eight million musicians on the platform. Of that, only 14,000 make $50,000 a year or more, which is roughly the average American salary.
So 99 percent plus don’t make a living through the platform. Why is that? Because Spotify takes 30 percent. The labels take another 60 percent plus. You ask anyone who does streaming, they get a couple hundred dollars for a million streams.
This is true throughout all of the modern internet services. You put stuff on Facebook, and they make a lot of money. They make tens of billions of dollars a year on advertising. They have what’s called a 100 percent take rate. They take all the money.
TikTok takes all the money. YouTube is the only large network, large internet service, that actually shares with the users. The share is 45 percent. For example, we have a few investments that are platforms that will let musicians sell digital collectibles, NFTs, directly to their audiences. So it’s like selling album covers, art, digital merchandise, and things like this.
The notable feature is that artists take 97.5 percent of the money or something close to that. Essentially, because of the architecture of blockchains, the artists truly own the things they’re selling in the same way they would own an offline good. Why do artists, musicians, sell merchandise when they go on tour and make a lot of money that way? It’s because there’s no intermediary. In the same way, when they sell NFTs to their audience, there’s no intermediary.
In this downturn, people think NFTs are dead. There were many billions of dollars of NFTs sold by creators to their audiences. And the vast majority of that money went directly to the creators. So it’s a meaningful income source for creative people on the internet today, even though NFTs are at a relatively early stage in their development.
I’m generally very excited. I have a bunch of sections in the book about creators—a full section on what I call collaborative storytelling and a number of sections on how you can use blockchains to fight deep fakes and other new kinds of issues that arise with AI.
Another area is payment. This has been talked about for a long time. But with the rise of what are called stablecoins, we’re seeing this really happen at scale.
In 2022, there were $6.8 trillion in stablecoin transactions. These transactions happened on platforms like Ethereum, where people are sending dollars, USDC [or USD Coin, a digital stablecoin], and other kinds of dollar-backed assets. A lot of this happens in the developing world, which is why audiences in the US might not be as familiar with it. It’s a really interesting kind of growing use case.
Why can’t regulation help solve for the internet’s lack of an effective reputation system?
I think regulation, with respect to the internet and technology, plays a role. The right regulation starts with a deep understanding of the technology, what it’s capable of, what the good things are, what the bad things are, and how we prescribe a policy solution that minimizes the bad and allows for the good.
You need to think very carefully about the second-order consequences of these regulations. For example, there’s a lot of talk about AI regulation today, and there’s talk about having certification systems that make people certify frontier AI models.
The risk with things like this is if you start to add in really expensive, labor-intensive certification processes, you tend to favor large tech incumbents. So GDPR [General Data Protection Regulation], which is a European privacy regulatory framework, was a huge benefit to Facebook, Google, and the large tech companies that have thousands of people whose job is to comply with that regulation. It’s very, very hard for start-ups to comply with a lot of these things.
In the US, you often have to deal with five to ten different regulatory entities. There are state-level entities. It’s very complex. So I think regulation plays a role, but it has to be done in a thoughtful way that doesn’t stifle start-up innovation.
If you look at the last 50 years of tech innovation, so much of it has come from start-ups. I also think that in many cases, you can solve technology problems with other technology. For example, one big question with AI is how creators are compensated when their material is used as inputs to training AI systems.
One interesting application of blockchains that we see people building are systems where creators can come together and effectively put their content together as a group and collectively bargain with these platforms. They ensure that if some AI content distributor wants to use their content, they need to comply with certain kinds of financial and other usage terms.
Another good example are deep fakes and bots. As AI gets more and more sophisticated, it’s going to be very easy for people to create sophisticated bots that simulate people. They could be on social media. They could be used for phishing. How do you counter that?
Fundamentally, I think the best solution is to have a way to cryptographically guarantee that somebody is an actual person, that they are who they say they are, that a piece of media came from who it claims it came from. This is a very natural application for blockchains, because blockchains provide cryptographic proof that something is what it says it is and the complete transparent audit trail that shows you the provenance of a person or a piece of content.
My view is generally that regulation plays a role, of course, but a lot of problems can be solved through technology and innovation, along with regulation. Regulation needs to be thought of in a balanced way that protects consumers but also encourages start-up innovation.
A lot of problems can be solved through technology and innovation, along with regulation. Regulation needs to be thought of in a balanced way that protects consumers but also encourages start-up innovation.
Can blockchain better mediate the economic relationship between AI and content creators?
I talk about the history of search engines in the book. In the ‘90s, essentially what happened was Google rose up using fair use exemptions and copyright law—meaning it would go and index a website and then use snippets of it. It would use that content in its index but then also use snippets on the search results. Basically, it said, “This was fair use because we’re only using a snippet. We’re not using the full content.” At first, some people protested. But, for the most part, it seemed fine. They had a system where you could opt out of it.
Today, sites like Google have become so powerful that you have to be listed in Google, and you can’t really opt out, practically because you lose all your business. They basically set all the terms. You go to a search engine, and, at least today, if you give them your content, you have some promise that you’ll get traffic back.
That’s sort of the covenant, as I call it in the book, that you have between content and distribution on the internet. For example, I’m the New York Times. I let Google take my content and show snippets of it, but they’ll send me some amount of traffic back. Everyone who runs a content site is used to that.
In the AI world, it’s very likely that they won’t send traffic back; they’ll just give you, the user, their answer. You’ll just say, “What’s going on in XYZ?” And it’ll just tell you. There won’t be links, and you won’t go back [anywhere]. It’ll use your content to make those systems smarter, but they won’t send traffic back.
Now one outcome is people put up paywalls, which is already happening, by the way. That means people block their content from being searched, and artists take down their content, and all these other kinds of things, because they don’t want it training generative AI algorithms.
I think this would be a shame. You are going to end up with a siloed internet. That’s where we’re headed now if we don’t think this through. Another outcome is—and we’re already seeing this also—AI providers going to content forums for content because the content’s getting blocked off by these services.
So they’re going off and paying people who are offshore in an office somewhere saying, “Hey, give me a bunch of pictures that look like this, and feed it into the algorithm.” I think both of those outcomes, a siloed internet and offshore content providers for AI, are both depressing outcomes.
A much better outcome would be to construct a new covenant, a new relationship between content providers, AI, distribution, and search engines. What is the natural way to do that? Get large groups of people to coordinate on the internet and create networks. That’s what blockchains are. It’s the technology for creating networks that are open and governed by the participants.
There are a bunch of entrepreneurs working on things like this, where you basically think of them as economic networks, where content providers can join, AI providers can join. They can come and agree on terms and collectively decide on what the optimal relationship is for everybody to have the best outcome. We didn’t really have systems like this on the internet before, where you could come together and have large groups of people collectively decide on what kind of economic relationship they’d have.
You could build a website in the old model. But then you have this website in the middle that’s an intermediary, that has a different set of incentives and ends up slanting the playing field to its own advantage. Blockchains are a great way to do this in a very neutral way. One of the key features of a blockchain is that you can make strong commitments about future behavior. This is really, in some ways, the unique technological capability of a blockchain.
So the blockchain can say, “I’m going to mediate this relationship. And there will be no fees. I, the blockchain, am not going to charge any of the participants in the network any fees.” The blockchain can make that commitment and promise it in code in a way that can never change. This is what I call: “Can’t be evil, not, don’t be evil.”
Which means people can join and commit their economic future to a system built on a blockchain. They can trust that it will stay the same in the future in a way they can’t in a traditional internet service architecture.
Why do you believe blockchain won’t go the way of the internet?
I started my career in the ‘90s, and one of the main reasons I chose the internet for my career is that I bought into the ideals of the internet. It’s fair to say that a lot of the early ideals of the internet that I bought into related to the idea that it would be this open system that would democratize information and media.
I do think it took a wrong turn at some point. That’s why, in the book, in the first two chapters, I tried to go pretty deep into diagnosing what happened, and then into diagnosing what’s wrong with the new wave of networks—what I call corporate networks—why it led to “disconsolidation.”
Things went off the rails, somewhat, with the development of the internet. Step one is understanding exactly why they went off the rails, which I try to do in the book. Step two is saying, “What are the ways to put it back on the rails?”
There are only two kinds of credible approaches to that; one is regulatory approaches. I haven’t really seen significant momentum there for anything meaningful that would re-decentralize the internet, but that’s, in theory, one way to do it.
The other is technological approaches. I think blockchains are, by far and frankly, the only serious contender for doing that. I spent a lot of the book explaining exactly why that is, specifically around the technology. There is a very important difference between blockchains and, as I call them, corporate networks, like Twitter. Blockchains are open source. And there simply is no company and no owner behind them.
Before you get involved in something like Ethereum or any of these other blockchains, you can go look at the code. It’s all open. It’s all community-owned.
There’s no need to rely on trusting some other person. You can go and evaluate Linux or hire someone if you’re not technical yourself, but it’s all open. And that’s a very important difference. In terms of the international questions, as China has shown, for example, countries can pretty much overwhelm any technology if they want to. They can block anything they want in the end, and some countries have chosen to.
I think the best antidote to this is to create global internet services so compelling that people feel like the pros outweigh the cons, and they want to participate in them. But I do think there will be some risk in the internet becoming more fragmented over time. But sort of like a global financial system, ultimately, you want to make it so that there’s so much upside to be included in the internet economy and the internet—the knowledge and money and all the other things that one would benefit from participating in—that it’s more attractive than blocking it.
Watch the full interview
