Dialogue with Circle CEO Allaire: After stablecoins comes the machine economy

Author: Payment 201

Circle co-founder and CEO Jeremy Allaire and a16z crypto founder Chris Dixon sat down together, and the first question posed by host Robert Hackett was: Why does the world still need a new blockchain?

The conversation began with Circle's entrepreneurial vision from 2013—back then, they wanted to create "an HTTP for money," ultimately choosing to issue USDC on Ethereum; it continued to discuss why, after 2020, they determined that existing networks were insufficient, leading them to build their own L1, Arc, from scratch. The discussion concluded with two directions: who will use this system after stablecoins, and what financial infrastructure will look like when AI starts hiring and paying each other.

Allaire's judgment is that no one in this field has a moat, the speed of software iteration is increasing, and the main participants using this system will increasingly be machines.

They discussed the following core points:

1. Why Circle is building its own chain: What is needed is not a faster chain, but an operating system designed for "real money."

2. Two triggers for institutional entry: USDC proved that the flow of funds could be solved, and regulators began to propose infrastructure-level standards for the chain.

3. Three pillars of the moat: The utility and liquidity flywheel of the stablecoin network, regulated infrastructure, and CCTP interoperability.

4. Arc's positioning: Open, permissionless, open-source, but validators are known, elected, and have operational and legal commitments.

5. The overestimated "overnight" of the mobile era: The real mobile era began with the Palm Pilot and took fifteen years.

6. Why there is currently no moat: The mindshare of developers is still incredibly small compared to traditional platforms.

7. The next big thing always looks like a toy at first because it is based on an exponential curve.

8. The strangest experiment on Arc: The true fly brain of 16MB bytecode.

9. Five years later: The scale of the machine economy and the reality that "bots are already the main entities in online activities."

00:00 Opening

01:00 Circle's initial vision for programmable money

06:22 Why institutions are starting to take stablecoins seriously

11:07 Building a global financial network

14:05 What is the moat of stablecoins?

19:05 Is crypto still waiting for its iPhone moment?

25:35 Adoption beyond speculation

27:18 From improving old things to creating new markets

32:06 The strangest experiment on Arc

36:09 When AI hires AI

37:12 The next big thing always looks like a toy

38:12 Can blockchain hedge against the centralization tendencies of AI?

39:47 How Arc balances openness and accountability

43:32 The machine economy five years from now

1. Why Build a Chain Yourself

Allaire has been discussing this line for twelve years. In 2013, Circle wanted to create "an HTTP for money," initially joining the Bitcoin community, only to find it too closed and rife with ideological disputes; after Ethereum emerged, programmability and asset issuance really took off, so in 2016 they abandoned the layer of moving dollars on Bitcoin and turned to building a protocol on Ethereum, which is the starting point of USDC. By 2020, the issue became that existing networks could not hold up—Ethereum was slow and expensive, with gas fees being a hard constraint. They first collaborated with L1s like Solana, and later with L2s, allowing USDC to run anywhere like Spotify, and then built an interoperability layer. What truly led them to decide to build their own was the realization that there was no network designed for "real money and real securities" in the market.

2. Two Triggers for Institutional Entry

Allaire believes the change comes from two things. One is USDC itself—stablecoins made people feel for the first time that the flow of funds was genuinely solved, leading to growth in non-speculative use. The second is that regulators began to examine: if global financial institutions are going on-chain, then the chain itself must meet the standards of financial infrastructure, which are genuinely enforced by central banks and capital market regulators. These two things happening simultaneously led to the establishment of Arc.

3. The First Pillar of the Moat

The stablecoin network has two flywheels. The utility flywheel is the platform effect: the more applications that connect, the more the next developer will choose it. The liquidity flywheel is unique to currency: Circle mints and redeems $150 billion in a quarter, integrating into the global banking system; there is foundational liquidity in primary markets, secondary markets, on-chain markets, and protocols, leading to higher capital efficiency, attracting products that require the highest capital efficiency.

4. The Second and Third Pillars of the Moat

The second pillar is regulated infrastructure. Allaire says this is no small matter: to become recognized infrastructure in various markets, the engineering effort and compliance costs are very high, and they are basically the first to achieve this. The third pillar is interoperability—CCTP is currently the most widely used cross-chain protocol across all blockchain networks because it carries USDC, the digital currency with the largest cross-chain transfer volume. Allaire states that CCTP accounts for about half of the value of cross-chain transfers, with recent quarters seeing it rise to 70%.

5. CCTP Transformed from a Lane to a Highway

Arc has a design: any asset issued on Arc—tokenized stocks, stablecoins, tokenized credit—qualifies to run on CCTP's payload. Previously, only USDC was on this highway, but now any asset issuer can use the same road, gaining the same interoperability, liquidity, and distribution capabilities.

6. Not a Winner-Takes-All, but Winners Take Most

Allaire's judgment on market structure is very restrained: the stablecoin market is not winner-takes-all, but winners take most of the share. As the market grows from hundreds of billions to trillions, several networks will emerge. Circle's own hypothesis is not to capture 90%. Another piece is CPN (Circle Payments Network), a network built for financial institutions, which is logically no different from Visa or Swift, and currently has nearly two hundred financial institutions on it.

7. The Mobile Era Actually Took Fifteen Years

Allaire uses the mobile era as an analogy: people think the mobile era began with the iPhone and Android, but it actually started fifteen to twenty years earlier. Palm Pilot, BlackBerry, Windows Phone, Symbian, Nokia, NTT—countless attempts, hundreds of billions to trillions invested, and countless apps that never turned into businesses, with many infrastructures failing. It wasn't until the right devices, the right operating systems, and the right developer incentives all aligned that Apple and Google reached that point around the same time. "When I look at blockchain networks, the model in my mind is: these are operating systems."

8. There is Currently No Moat

Allaire's judgment is very direct: no one has a clear advantage in developer mindshare; the number of active developers on these networks is incredibly small compared to traditional platforms. His own company is the same—software iteration is extremely fast, and the deliverables are much more, while the number of developers needed may actually be fewer. Therefore, he believes this is a very open space that will be redefined in the next three to five years.

9. Fly Brain: The True Fly Brain of 16MB

When asked about the strangest thing on Arc, Allaire mentioned Fly Brain—someone converted neurological information and computational rules about real fly brains into about 16 megabytes of bytecode, running on Arc, and even used AI for the experience layer. These fly brains interact, evolve, and operate in the environment, and he still doesn't fully understand what it is, even considering it might be a gimmick. But this shows that AI, blockchain, and economic incentives are already being tested in very different ways.

10. The Next Big Thing Looks Like a Toy

Chris Dixon explained: the most important things in technology are all based on an exponential curve. Moore's Law is one, AI's scaling law is another, network effects are one, and composability is also one—blockchain allows applications to interoperate like Lego. Because it is based on an exponential curve, it looks like a toy in the early stages; but as long as it can truly scale, it will grow very quickly. Thus, one of the core policy arguments he made while writing "Read Write Own" is that it would be a pity if policymakers do not allow open-source, decentralized forces to develop in parallel with the centralization tendencies of AI.

11. The Middle Path: Open, but Validators are Known

Arc is open, permissionless, and open-source; anyone can connect, build, or hard fork. But Allaire emphasized the difference: the actual operators validating the infrastructure are known, not anonymous, and are elected based on who they are and the commitments they make at operational and legal levels. The reasoning is very practical—if this system is to be relied upon by financial institutions, businesses, and ultimately households worldwide, governments and central banks need to know what commitments are made around how the network operates. He calls this the middle path, and USDC follows the same logic.

12. Five Years Later: Machines as Main Participants

Allaire said that no one really knows the answer now because the pace of technological acceleration is faster than at any point in human history. But he believes that in the coming years, there will be a shift: a large amount of creative and cognitive work, as well as economic substance and exchange, will primarily be operated by AI machines. He used 1995 as an analogy—at that time, if you said that the volume of communication and information worldwide would be a billion times larger than today, no one would believe it; but it did happen. Chris Dixon added a more specific observation: he vibe-coded a small application deployed on Arc, and a week later it had 15,000 visits a day, all from bots looking for environment variables, SSH credentials, and cryptocurrencies. "The machines have arrived."

Robert Hackett (Host / a16z crypto):

I am Robert Hackett, and this is the A16Z Crypto Show. With me are Circle co-founder and CEO Jeremy Allaire—Circle has just launched its own chain on mainnet, called Arc; and Chris Dixon, founder and general partner of A16Z Crypto. Thank you both for being here.

Jeremy Allaire (Circle co-founder and CEO):

Great, thank you.

Chris Dixon (a16z crypto founder):

Thank you, it's great to be here.

Robert Hackett (Host / a16z crypto):

So Jeremy, you and I are old acquaintances, going back to the years when I was a reporter at Fortune. I have reported on Circle all along, watching it go through several transformations: from a Bitcoin consumer business to a Venmo-like product Circle Pay; you also did trading business and acquired an exchange. Ultimately, you struck gold with USDC—it's one of the most mainstream stablecoins in the world today. Now you have launched Arc. Tell me, what have you learned along the way that made you feel the need to build your own blockchain?

Jeremy Allaire (Co-founder and CEO of Circle):

I think we might need to take a few steps back first. This actually goes back to when we founded Circle. In 2013, to be precise, at the end of 2012 and into 2013, we were just starting our entrepreneurial journey. What attracted me to this field was the idea of programmable money and these networks—back then, people were talking about blockchain networks, or more accurately, the Bitcoin network—becoming computers that could run applications. The vision at the founding of the company was that these networks would become new types of distributed computing engines, well-suited for scenarios that require ensuring the reliability of data, transactions, and computations. This is a problem they could solve that had not been addressed before, and it would be solved in a public environment—you essentially have a public computer on which you can issue things.

You can issue digital dollars. Back then, we always talked about digital currency, never cryptocurrency, but you could indeed have a digital dollar. The initial idea was that you could build a protocol for dollars on the internet, like creating an HTTP for money. And to create this protocol, you needed these computer networks to truly allow you to build protocols—you know, Chris and I have known each other for many years in the area of internet protocols. So you need to be able to build higher-level protocols on this computer, and then it needs to be scalable, user-friendly, and so on.

Robert Hackett (Host / a16z crypto):

Right, you also worked on internet video in your early years, right? I mean, that counts as—

Jeremy Allaire (Co-founder and CEO of Circle):

I mean, I worked on early web protocols, Web 2.0 protocols, video protocols, communication protocols, programming languages, virtual machines; basically, I have touched on the infrastructure of the first and second generations of the internet. So to me, this is the next logical layer of infrastructure for the internet. Our thought at the time was: Bitcoin has a first-mover advantage, it's a good open-source project, so let's join this open-source community and add to it to make it stronger, add more programming language capabilities, and add what we need to realize our vision.

But progress was not smooth because it became quite a closed community, tightly controlled. The scope of the open-source community and contributors was relatively narrow in terms of goals. There were many ideological struggles around this, with us on one side and others on the other. We don't need to revisit that history. However, Sean and I have always been thinking: should we build such a computer network ourselves? But considering we had other things on our plate—like you just said, we have done a lot along the way—this idea never came to fruition at the right time.

Interestingly, when the second-generation computer network Ethereum came out, many ideas—programmability, virtual machines, asset issuance—really worked out. Of course, you could technically question it. Because if you are a technical purist, you might say: wow, this thing looks like it's held together with chewing gum and wire, I don't even know what Solidity is, it doesn't count as enterprise-grade, however you describe it. But by 2016, we were clear: it was good enough. So as you might remember and perhaps have written about, we abandoned the layer of moving dollars on Bitcoin and said: let's build a protocol for dollars on Ethereum. This led to the development and launch of USDC.

So basically by 2017 and 2018, we felt: okay, this layer can be done now. But once again, we were not sure whether we should build our own computer network and operating system. Then fast forward to 2020 and 2021, it became clear: the infrastructure model of these networks was not sufficient. Ethereum itself was slow and expensive, and gas fees were a big problem. So many people began to say there was an opportunity to build a de novo network from scratch in a clean environment. We also started to say: let's align with these networks because we need them to realize our vision of scalable applications, programmability, and the speed of capital flow. We need them.

So in 2020, we collaborated with the Solana team and other emerging L1s, and later began working with L2s. Our philosophy became: ensure that the digital dollar we create and the protocols supporting it can run anywhere. I often compare it to USDC being like Spotify or Netflix—you buy a device, and it just works. We hope that regardless of what blockchain you have, these digital dollars can be used, and you can integrate them. Then we built an interoperability layer to connect everything, allowing for secure transfers. Throughout this process, our view was that these networks were continuously improving, and you know.

But about two and a half years ago, we started to see a few things happen. One was that some large institutions began to say: hey, we might actually use this technology to handle real money and real securities, and there is indeed a tangible prospect here.

You know, looking back, something like USDC was really crucial. At that time, stablecoins felt like they really solved the problem of capital flow. So people began to think, I can really use this to transfer money better. It was also at that time that we started to see the growth of these non-speculative uses, right? That was important.

Then I think another change was that regulators really began to examine this, saying: if global financial institutions are going to connect to blockchains, then we have certain expectations for these blockchains, and we hope they meet certain standards. Long story short, the infrastructure in the financial system has standards, and these standards are indeed enforced—by central banks, by the institutions that regulate our capital markets. These are all standards at the infrastructure level.

So we started to hear these voices and began to meet more and more with institutions interested in stablecoins and their applications. What we really found was that there was no dedicated blockchain network and computer built for this purpose at that time. We saw a new era emerging, a mainstream era. So we decided to start from scratch and build a new operating system and a new network with Arc.

I think we designed it in a way that inherits a lot from what we learned about how blockchain networks are incentivized, how they grow, and how they are governed. It also inherits many practices from the history of large platform ecosystems, like Google's Android, and a lot from the open-source world. We tried to blend these together, allowing this network to be jointly owned by a very broad group of stakeholders—ultimately including the users on the network—while ensuring that this infrastructure meets the standards such that global central banks would say, okay, you can run real money and real investment contracts on this.

Finally, I want to say one thing, and then we can move on: during the building process, we really saw that the main participants using this new type of operating system in the future will increasingly be machines. We have believed this for a long time. We have always felt that autonomous software running on the internet can do these things. Of course, long ago when we thought about this, we had no idea about generative AI. But it became very clear later: the primary users of the software, the main deployers, and the entities that will primarily run on it in the future are likely to be AI. We started to think about a whole bunch of questions that arise from this, and we are still thinking about them. The transition to an agentic economy is a very core set of design features in what we are doing. In short, that’s the background, a long-winded background.

Robert Hackett (Host / a16z crypto):

So, compared to the early days, what has changed that suddenly makes these institutions willing to participate and see this as a viable path forward?

Robert Hackett (Host / a16z crypto):

I want to dive deeper into some of the design decisions you made. But before that, Chris, you’ve seen a lot of pitches for new public chains, and we’ve invested in quite a few. What makes you believe there’s still room for another chain?

Chris Dixon (Founder of a16z crypto):

I think a big part of the reason for us is that I have known Jeremy for a long time and have always been impressed by what he has created. So when we invest, the most important thing is the team and the team he has built around this. I also resonate with his vision—I have written and talked a lot about this—it’s a new type of computer, a global financial computer, with all sorts of platform possibilities, new applications, and so on.

Blockchain is interesting because on one hand, there is a community, a sense of social cohesion around them, but on the other hand, they are computers. You would hope to have people who think deeply about operating systems, programming language design, and so on. I also feel—maybe Jeremy can talk about this—he has an exceptionally deep understanding of regulatory and international aspects. Money has the most obvious network effects, USDC is extremely popular, and the dollar is very popular worldwide. I think we have the opportunity to make a system like Arc a global monetary layer, which is quite a key idea. So I would love to hear you talk about this, Jeremy, from the international perspective, especially given how complex the situation is with governments and regulators around the world.

Jeremy Allaire (Co-founder and CEO of Circle):

Clearly, what’s different about Arc is that to operate the infrastructure of this network, ultimately to run staking validation nodes on the network, you actually have to meet very high operational standards—operational, information security, compliance, and so on. And because we know that the people doing these things have met and are adhering to these standards, with service level agreements and such, this part allows the network to provide what we call certainty of settlement finality in less than a second. This is extremely important for real money.

But another thing is that when we think about an economic operating system running on the internet, it needs to support a whole vocabulary of economic activities—far beyond payments, including various financial contracts, economic contracts, and governance decisions.

If you want it to reach an internet-scale level, it must possess diversity in geopolitical and geoeconomic aspects. For the stakeholders who operate and ultimately govern this infrastructure, it must be trusted and neutral. This is one of the commitments we have made, and it is reflected in its performance along the way. A considerable number of the world's largest participants have already joined—capital markets, asset management, the payment sector, and many global banks covering a wide range of international markets, including Asia, Europe, and the United States. So we are just getting started; this is really very early, just trying to find a way to bring together this diverse group of infrastructure stakeholders so that if I am an entity in a market somewhere in the world, I can understand how this system operates.

We hope to go much further. We want every major geoeconomic block in the world to operate this infrastructure, to have substantial economic interests in it, allowing leading companies in their local markets to use this infrastructure. This is crucial. Without these, if it is just something from the United States or elsewhere, achieving internet scale would be very, very difficult.

So blockchain networks and distributed governance provide us with some interesting tools to do this. This is a significant driving force. Again, we are still very early in this matter—when Mainnet launched a few weeks ago, the first batch was about 10 to 11 companies that started with us. This scale will obviously expand. Ultimately, there could be about 40 seats on the network. We are indeed considering it this way. Eventually, those staking validators and operators will vote through elections to decide the expansion of global infrastructure partner members. But I feel we have been working to guide it from the very beginning, ensuring there is a solid foundation behind it.

Robert Hackett (Host / a16z crypto):

Jeremy, what are your thoughts on your moat and the defensive points you have?

Chris Dixon (Founder of a16z crypto):

The current situation can be viewed from two directions. On one hand, AI has made it far too easy to write software from scratch; you can create products that compete with existing vendor software. On the other hand, traditional financial institutions are also entering the scene with their own new products. We see some alliances forming around their own stablecoins, and Citi and Goldman may also launch their own products next year, along with many other banks and partners. What do you think about the defensiveness of what Circle is doing—not just USDC, but also the new infrastructure you are building?

Jeremy Allaire (Co-founder and CEO of Circle):

What we are doing has about three pillars. One is the stablecoin network, which includes our USDC, our Euro stablecoin, and other digital assets we issue. We have a tokenized Bitcoin product, a Digital Euro, and I believe currently the largest tokenized government bonds and money market products. So that is a whole family, but USDC is clearly the core role within this group.

The stablecoin network itself, its platform effects and network effects are the business. The more applications that connect to the network, the greater the utility of the network, so the next developer will think, hey, which network should I support? Oh, I should support this one because everyone else is supporting it. You have this very powerful utility flywheel that exists. You also have a very strong liquidity flywheel, which is unique to currency and, of course, unique to the digital asset market and market infrastructure—liquidity is a big deal. Primary liquidity, which can be smoothly mobilized in global markets and global financial centers, is crucial. We mint and redeem $150 billion each quarter, and it connects to all these global banking systems.

The liquidity that exists in primary markets, secondary markets, on-chain markets, and on-chain protocols is very powerful. Binance, Coinbase, Hyperliquid—just to name a few—have such deep foundational liquidity that it means higher capital efficiency, allowing other products that require the most efficient capital to benefit from it. So there are both utility network effects and liquidity network effects.

Then there are these—what I broadly refer to as infrastructure moats. That is to say, we have built regulated infrastructure around the world, and that is no small feat. Becoming that recognized infrastructure in these markets is a significant undertaking. And overall, we are the first to achieve this. Then there is real infrastructure, such as interoperability infrastructure. Currently, the most widely used interoperability protocol across all blockchain networks and assets worldwide is Circle's Cross-Chain Transfer Protocol, CCTP. The reason this protocol is the most widely used is that USDC is the digital currency with the highest transfer volume between these networks.

Robert Hackett (Host / a16z crypto):

Let me confirm, it accounts for about half of all cross-chain transfer value.

Jeremy Allaire (Co-founder and CEO of Circle):

Yes, that number fluctuates. It was as high as 70% in recent quarters. So that is important, right?

And one point—this actually ties back to Arc. One major thing we are doing on Arc is that if you issue an asset on Arc, whether it is tokenized stock, stablecoin, tokenized credit products, whatever it is, your asset is actually eligible to run on the CCTP's payload. So in a sense, we have built the highway for transferring value across all those networks. Before this, only USDC could run on that highway. Now, anyone issuing assets—if you are an asset issuer—you can run on the same highway and gain that interoperability, liquidity, and distribution. This is very powerful. So the stablecoin network has these characteristics. What we often say is that the stablecoin market is somewhat like a winner-takes-most scenario.

… It is not a winner-takes-all market structure, but rather a structure where the winner takes most of the share. And as this market grows from hundreds of billions to the hoped-for trillions, several networks will emerge, right? That is our assumption; we do not assume this is a market that captures 90% of the share, right? It is a winner-takes-most scenario. Another piece is our CPN, which is Circle's payment network. It is also a network model designed for financial institutions. It is no different from networks like Visa or Swift, meaning financial institutions join in and can use the stablecoin track and blockchain as the coordination mechanism for all these value exchanges to transfer value globally. This network is growing very quickly, and we announce its growth every quarter. There are now nearly 200 financial institutions on it. This, in turn, supports the demand for the stablecoin network.

But on the Arc level, what I want to say is related to what I mentioned at the beginning, which is that we believe this field is entering a new chapter. In my view—I think both of you would resonate with this—the development of the internet has gone through several eras. One of those eras is the mobile era. When we mention the mobile era, we think of the iPhone and Android phones. But the reality is that the mobile era actually started 15 or 20 years before that. There were countless attempts to create mobile operating systems and platforms that we would all use. It attracted hundreds of billions of dollars in investment, from open-source developers to the largest companies in the world, like Microsoft and Compaq.

So we had Palm Pilots, BlackBerrys, Windows phones, Samsung phones, Symbian, Nokia, NTT, and many, many others. At that time, we all thought, wow, mobile is huge. Look at these things. You can make an app on your phone. But the reality was that all of them were terrible and could not run for mainstream users. So only early adopters rushed in. A lot of money was invested in making apps, but those apps never turned into real businesses. A lot of infrastructure was built, and many of them died, turning into cardboard boxes and junkyards around the world, along with those physical products, etc. But that was a long journey, about 15 years. Later, some things came together and converged to truly make this possible—creating the right kind of devices, the right software operating systems and experiences, the right developer incentives and economic incentives to really get it running. Apple and Google reached that point around the same time; you could argue about who pushed it more. But when I look at blockchain networks, the model in my mind is: these are the operating systems. They attract developers and assets; that is their economic operating system.

And I feel we are in that 15-year journey. Today, it is hard to say, oh, this is the iOS of that field, and that is the Android of that field; it is hard to see. And I think most of us would agree that the applications running on these networks are mostly not usable for mainstream users yet. We are just starting to see those applications emerge, just beginning, like FOMO-type applications, where you might not even realize you are using crypto while using them. The same goes for stablecoins; those financial apps, those neobank apps—you don’t even know you are using crypto while using them. And AI makes this easier. So my view is that moving from early adopters to mainstream scale, from hundreds of millions of hardcore users—who are trading and making money or believe in this—to billions of people, making this the ordinary infrastructure that underpins economic activity, all of this is still ahead of us.

So my view is that this is a completely open space. No one has a clear advantage in developer mindshare. The number of active developers on these networks is incredibly small compared to the scale of developer activation on traditional platforms. So it is really very early. So no one has a moat; that is what I mean. And I think you are absolutely right that the speed of software iteration is increasing at an incredible pace. We, as a company, are also experiencing this speed of software iteration; we are delivering much more software, while the number of developers needed may not be that many. It is astonishing; what we are seeing is truly remarkable. So I believe this is a very open space right now.

And I think Circle, as the driving force behind this network, has advantages such as institutional credibility, the stablecoin network we bring, the design of the computation layer and gas layer centered around stablecoins, and other aspects, as well as the design for such a future space. But we also designed it so that ownership and governance are basically distributed, all of which is public and out in the open. We have the Arc token, and the white paper for the Arc token is available publicly, with many large institutions being stakeholders. This situation will continue and will become increasingly decentralized. This is also why I feel a bit like looking at Google's Android. Android is an open-source operating system, and Google is the main author. They package the applications that most people will install, but others don't necessarily have to use them. Then device manufacturers, hardware makers, telecom operators, retailers, and others all have roles to play and can derive value from this ecosystem. It occupies about 85% of the market, roughly like that. I believe it is still early, and I believe that in the next three to five years, this will truly be redefined.

And this brings us back to another statement: the main participants in all of this will be AI, with AI creating software. Perhaps there are people behind AI, but it will also be AI creating its own software. AI will act as an intermediary, will interact, and will execute economic work. In fact, cognitive labor is economic work. So cognitive…

…labor itself will also be executed, embodied, coordinated, and exchanged on these networks. Things will get very strange.

Robert Hackett (Host / a16z crypto):

And as you said, you are drawing an analogy with what happened in the history of the internet, especially the mobile revolution. Chris, I know one point you like to mention is that when the iPhone was first released, it was actually quite rough; the first generation model itself wasn't very user-friendly. This is something people often forget. They think that when the iPhone came out, everything changed overnight. But in fact, the journey to the mobile-first world we take for granted today involved a long period of stop-and-go development.

Chris Dixon (Founder of a16z crypto):

Yeah, when the iPhone first came out, I carried two phones because it dropped calls too often. A BlackBerry or any feature phone would do. And I remember the iPhone was in 2007, but it wasn't until 2012 that Facebook made its mobile transformation, which is quite symbolic, indicating that Silicon Valley finally realized this was a big deal.

Robert Hackett (Host / a16z crypto):

At that time, many people were still skeptical about whether they could succeed.

Chris Dixon (Founder of a16z crypto):

Yes, there was a lot of skepticism. I think people were still stuck in the framework of "the phone is a secondary device, not a primary device," and it took a long time for them to turn around.

Yes, I agree with everything Jeremy said. I think it's a very interesting moment for crypto right now: on one hand, some OGs seem a bit down, and some more ideological people may not be as excited; but on the other hand, from the actual data, the traction and use cases now are much greater than ever before. I mean, the level of adoption—Jeremy, I don't know how much you can talk about, but for example, the level of adoption by financial institutions. A large part of this is regulatory; the passage of the GENIUS Act, I think, almost overnight normalized stablecoins in the tech world. You can also see companies like Stripe, who are very pragmatic and will choose the best technology, and they are now using blockchain. So this situation can be seen at all levels.

And I think, looking at these things together: crypto has been around for a while, gone through several cycles, and now all the attention is on AI, which in my view is actually an important trend that is being overlooked. And as Jeremy said, it will converge with AI. I think we are still relatively early in the AI revolution. Looking at the data, enterprise-level adoption is only about 2%, and the adoption rate of things like agents on the consumer side is still very low so far. Clearly, this will grow. This will squeeze existing internet business models; for example, then you will need new models. So I think there will be a period of 10 to 20 years where the entire digital economy landscape will be reconfigured. And from a timing perspective, this aligns very well with the rise of high-performance blockchains like Arc.

I've also been thinking, at first what you see is—I like to use the framework of "physical" and "native"—at first, when breakthrough technologies emerge, you see people bringing existing use cases over, like transactions, payments, and personal payments, onto the platform. But as time goes on, what really excites me about this wave is when developers start to get creative. Jeremy certainly remembers that the internet in the 90s was all physical; people were making directories, basically just brochures. It wasn't until the 2000s that there was a whole wave of creative social media, etc. So I don't know when that will happen. I guess it will take a few more years. Jeremy, I don't know what you think. We will first see more of the existing things being made more high-performance, more globalized, and more programmable. But I think at some point, people will really dive in and start creating.

Jeremy Allaire (Co-founder and CEO of Circle):

I completely agree. I often say that what you need to do is create something that allows developers and entrepreneurs to tinker on top of it. When you build these new open platforms using new primitives—in this context, currency primitives, financial contract primitives, and other software primitives on top—ultimately, it is the creativity of developers that will guide us. So people will ask, what will the most important applications be? It is easy for people to say based on their priors, oh, it will be better for cross-border money movement, or you can more easily pledge stocks and borrow money in more places. It's easy to say, yes, it will be better, faster, cheaper, just done here. That is probably true.

I use credit as an example. Credit, when you add these new network operating systems—blockchain network operating systems—along with this new foundational layer, which is fully reserved digital dollar currency, plus AI, these things together will allow us to create entirely new types of credit tools: those providing funds, those needing funds, and machines providing and needing funds. We will create a credit market that looks more like AdWords rather than someone packaging a bunch of credit into bonds and selling them. So ultimately, we will get a very, very different market structure, unit economics, and distribution models, applying the long-tail theory we are all very familiar with to financial markets, which has never really happened before.

We have already seen some early signals and early forms, such as the tokenization of RWA, making stocks accessible to more people, and allowing you to do more with those stocks. This is an early glimpse of the long tail.

Chris Dixon (Founder of a16z crypto):

That is just an early form of the long tail—the market participants of the long tail, rather than the issuers of the long tail. So what if we really see long-tail credit, equity, token issuers, and these economic tools?

Jeremy Allaire (Co-founder and CEO of Circle):

My view is, yes, I very much agree with your assessment, and I think it will be very, very interesting. The timeline you mentioned, I think is probably correct.

My view is that several things are happening simultaneously: technologies like Arc, stablecoins becoming a legitimate form of currency that people can rely on and truly depend on. What really surprises me is that I have personally felt this, because I am now using AI extensively for fun. Software infrastructure and APIs, whether blockchain computers or software applications, client applications, are all deterministic. So AI can learn and implement at an astonishing speed.

This means that what people can create on software will explode wildly next. Just last week, I saw some very interesting things starting to be released on Arc, like some crazy games and simulations. They have economic incentives and stablecoin funding. The experience of using them is simply that you just need to bring USDC, and nothing else is required. I could never have imagined these things before. I don't know, will this become a killer application? We don't know. But I feel that the simplicity with which people create experiences, deploy them, and make them usable is now arriving, and AI and AI software engineering are accelerating it.

So I really expect that looking towards 2027—just to use a round number—I really expect to see a large number of new applications appearing on these computers that will break that rule, as Chris said, escaping from physicality. They will start doing things that make you think, wow, I never thought this was even possible. So from this perspective, this is an exciting moment.

Robert Hackett (Host / a16z crypto):

Jeremy, what is the strangest thing you have seen on Arc so far?

Jeremy Allaire (Co-founder and CEO of Circle):

I literally just saw something the other day called Fly Brain, as in the fly. Yes, yes. I don't know who the team is; they just made it directly. They basically converted all the neurological information about fly brains—the real, actual computational rules of fly brains—into about 16 megabytes of bytecode running on Arc. They clearly used AI to create the user experience layer, allowing these fly brains to be released. They are really evolving, interacting, and behaving in this environment. I haven't looked at the token incentives for those things; I don't know exactly what the "deal" is, but I think this is an experiment in computational biology combined with neuroscience, taking real-world things and now creating simulations. They gave you some way, I guess, to engage with the world of flies. But to be honest, I don't know; I was just thinking…

I was just thinking, what is this? I asked on Twitter, what is this for? Then there was a long explanation about what this means for scientific discovery. I thought, I actually didn't fully understand. Maybe it's just a gimmick; I really don't know. But it illustrates one thing: AI combined with these computer networks, along with economics, is leading people to try very different things.

So we released an application called Arc Studio for vibe coding, which can basically deploy smart contracts safely. Deploying smart contracts is difficult because security issues are real. So we added something like a "soft package" for smart contract deployment. This is also very interesting, with high developer participation. People are doing various things on it, with obvious financial applications, as well as games and other things with economic systems. So it looks very interesting.

It's actually quite shocking that despite the amazing progress in AI, with so many crazy papers coming out and capabilities rapidly improving, many smart contracts still hold up.

Chris Dixon (founder of a16z crypto):

The Fly Brain network on Arc gives a new meaning to the term "swarms"—groups of agents.

Robert Hackett (host / a16z crypto):

This naturally reminds me of you, Chris, and a phrase you often say: the next big thing often looks like a toy at first. I'm not saying Fly Brain is the next big thing, but you start to see these strange experiments taking shape, and you never know which direction they will go or which ones will have real substance, while people often give up on them because they seem too strange.

Jeremy Allaire (co-founder and CEO of Circle):

I completely agree. I think Chris has written a lot about this, and it’s also the basis of your read-write-own argument: when you can create economic incentives for participants and applications, and these can be globally distributed, the application itself becomes an ecosystem with economic incentives. We are really starting to see people bring this to life. And it’s not just those who manage wallets themselves, needing to click five signatures on a crypto wallet to vote—think about what DAOs were like back then—now, due to improvements in software user experience, you can really fine-tune it. So I’m very interested in those doing truly novel things in governance, participation, and economic contracts, those much more complex economic arrangements running on these deterministic systems. I think because of these advancements, we are entering an era where creativity in this field will truly flourish. Let me give you another example, a project coming to Arc that essentially does arbitration—

Chris Dixon (founder of a16z crypto):

This mechanism is for arbitration of work done by AI. It can start with AI jurors and AI arbitrators, and then scale up to human arbitrators, with its own security and staking model. But it is essentially addressing a reality: AI will hire other AIs to do work, and humans will hire AIs to do work, and in between, you need arbitration, around money and lending mechanisms. All these things are now in place, and you can really start doing these things.

So automated arbitration—arbitrating economic work, creative work, and intellectual property-related work—was completely unimaginable a few years ago. Now it’s not only completely feasible but could also have significant economies of scale, unlocking some very strong potential for the labor economy and the entire work structure. So when I look at these kinds of things, I think, wow, here we go again: entrepreneurs are starting to solve problems that were previously unsolvable, and now they are really getting to work.

The observation that "the next big thing looks like a toy" is connected to another observation: the most important things in technology often rest on an exponential curve. Historically, people have talked about Moore's Law, which is an exponential curve, the improvement of chip performance and cost-effectiveness. Now everyone is very focused on AI's scaling law, which is of course very important. But there are other exponential processes, like network effects, which blockchain has; and composability—the idea that blockchain allows applications to interoperate, stacking together like Lego blocks.

It is precisely because of these exponential curves that things look like toys in the early stages of the curve—think about early AI systems. It’s the same here: it may start with just a group of people coming together to do some kind of economic governance. But if it really works and can scale, it can grow very, very quickly because it has composability and network effects as these exponential forces underneath.

Robert Hackett (host / a16z crypto):

Chris, you wrote the book "Read Write Own" a few years ago, when AI was still very early and just getting started. How has your view of the future changed since then? Is there anything you would do differently? Or how would you say your argument has evolved over these years?

Chris Dixon (founder of a16z crypto):

Actually, I talked a lot about AI in the book, treating it as one of the core areas. Part of the positioning back then was—you see, the scaling law was already visible, which I had written about before—it looked like it would become much better and much more important. Part of my argument was that, in a sense, blockchain could be an antidote to the centralizing tendencies of AI.

I think AI is amazing; I really like AI. But there is indeed an effect where the companies with the most capital and resources have the advantage. So when we do policy arguments, part of the viewpoint is: first, open source is very important and can be used as a hedge. Open source is the most democratizing trend in the history of technology. Look at Linux, then look at blockchain—blockchain is like open source; they are open services that provide a fair competitive environment, low costs, global accessibility, etc., equally for all developers.

So one of my core arguments, especially at the policy level: it would be a shame if policymakers do not allow this countervailing, decentralized force to develop in parallel with the centralizing tendencies that AI may bring.

Robert Hackett (host / a16z crypto):

Jeremy, I want to ask you a question about networks like Arc. How do you balance the degree of centralization and decentralization? Clearly, there are some completely open blockchains, like Ethereum, Solana, etc. And now there is a shift towards more permissioned directions. How do you see Arc's positioning—where it is now, and where it might go in the future? How open can this network be, or how open should it be?

Jeremy Allaire (co-founder and CEO of Circle):

A few points. First, Arc is a public, permissionless network. Anyone in the world can connect to it, build software on it, and deploy software. It is completely open, public, and permissionless; these attributes are very, very critical. And obviously, it is also fully open source, you can study it, hard fork it, do whatever you want, so it meets this set of standards.

But I think the key difference is that the actual operators of the validating infrastructure are known, not anonymous. They are elected and selected based on who they are and the commitments they make at the operational and legal levels. So that’s different.

Therefore, our view is: if what you are trying to do is a global financial institution, a business, ultimately an economic operating system that families will rely on, then these operational commitments are actually very important. There is now a purist view that says no, operational commitments should be implicit in the network's performance; anonymity and who runs the network are key—we need to resist government censorship, so we need to be anonymous, public, etc., only using crypto to prove it.

But in reality, as a layer one, what we are building has a broader set of commitments, with known infrastructure participants, which I think is necessary because governments and central banks around the world need this—because I do care about them. I want mainstream society to run on this, I want the mainstream financial system to run on this, I want large companies and big public companies to put their contracts and governance on the internet. I want them to do this. And they will want commitments around how the network operates; they will want to understand what that is, and they will want to understand that election mechanism.

That is exactly what we are building. So it is distributed, not completely decentralized, but it is open, open source, transparent, and anyone can develop and use it without permission. What we have been looking for—just like USDC—is that sweet spot in the middle. Some might say USDC is centralized, as the issuer is a centralized entity. But if you want a safe, truly liquid digital dollar that connects to the real-world financial system, it can only be done this way. So we have always been on this middle path.

Even that Bitcoin product you remember, Robert, it was compliant and did KYC. We even created an interface to make it feel like dollars, rather than shoving Bitcoin directly in users' faces. Because we really believe that what people want is the benefits that this technology brings, but they want to price it in dollars, and it has to be compliant, etc. So I think this is the next stage—the middle path. What is the middle path? We are working to find that balance and maintain it.

Robert Hackett (host / a16z crypto):

Paint me a picture of what the world looks like five years from now. With the integration of these new technologies—like Arc, along with swarms of agents—what changes will occur in the global financial system?

Jeremy Allaire (co-founder and CEO of Circle):

Honestly, no one really knows the answer right now, because in my view, the speed of technological acceleration is faster than at any point in human history. If you believe in the compounding laws happening in the AI field—I do believe—then the capabilities we will see are truly unimaginable. When you talk about ASI, it’s actually hard for the human brain to extrapolate, because the reasoning and coordination capabilities of ASI, by definition, exceed the reasoning capacity of the human brain. So maybe we’ve become like flies in our thinking, I guess. But the reality is, it’s already there and it’s happening. As for what the outcomes will be, it’s very, very hard to say. But I do believe that in the coming years, we will experience a transformation: a large amount of work—creative work, cognitive work, all economic substance and exchange—will primarily be operated by these AI machines. And this likely means, just like in 1995 or 1998, I like to draw parallels with the early internet.

Because of the compounding nature of the internet, Metcalfe's Law, network effects, and better bandwidth, if you had said at that time that the global communication volume would be a billion times greater than today, and the amount of information published worldwide would be a billion times greater than now, most people would have thought you were crazy—this is impossible, it makes no sense at all. It is indeed hard to imagine. But that is exactly what happened later. The speed of information, the speed of communication, and the output of information all follow that astonishing curve. So my view is that the speed of economic activity, the speed of value exchange, and the speed at which financial machines are used will all be several orders of magnitude higher.

So how will this be reflected? I think it is complementary to the machine economy and this set of infrastructure. It should be that the creative output of an economy is far greater than today's world, and the total amount of goods and services widely available around the world should also be several orders of magnitude larger. This is what I envision five years from now.

Robert Hackett (Host / a16z crypto):

This vision is very compelling. Chris, please share your thoughts as well. I've heard you use the term "machine age network" to describe what is coming.

Chris Dixon (Founder of a16z crypto):

As Jeremy just articulated very well, I think some of these things will be hard to imagine, but looking at these trends, the direction is quite clear. And I agree with Jeremy's point: the vast majority of participants in the internet are likely to be machines, which seems quite natural. It is also quite natural that these machines have an economic component in what they are doing. So this will be a fascinating and interesting new world.

I completely agree. Most of the online activities are already bots. So the proportion of humans is continuously shrinking. I have a quick little story. I created an application using vibe coding to be deployed on Arc, but I won't specify what it is or the domain name because I'm just playing around and don't want to cause trouble. I made this thing and shared it internally at the company. About a week later, I went to check and asked if it could analyze the traffic. Of course, it might just exist in some whois database or something. The result was that it had 15,000 visits a day, all from bots, looking for environment variable files, SSH credentials, and cryptocurrencies. It was basically saying, "Hey, you have a whole bunch of bot swarms here, just scanning for vulnerabilities and looking for coins." I said okay, we probably only had one testnet user try this thing. So that's it—machines have arrived, and they are scouring every tiny crevice to find where they can take action, which is actually quite scary. They are here, and we need to find something for them to do.

Robert Hackett (Host / a16z crypto):

Jeremy, it's great to chat with you and hear about the journey Circle has taken; it's always rewarding. Thank you very much for being a guest.

Jeremy Allaire (Co-founder and CEO of Circle):

My pleasure, thank you.

Robert Hackett (Host / a16z crypto):

Thank you, Jeremy. Thank you.