Loading summary
Transcript53 lines
- [00:02]
A
Hi, everyone. Welcome to Unchained, your no hype resource for all things crypto. I'm your host, Laura Shin. Thanks for joining this live stream. Today's guest is Ali Yahya, general partner at Andreessen Horowitz. Welcome, Ali.
- [00:15]
B
It's great to be here.
- [00:17]
A
Today you announced that a 16Z crypto raised its fifth fund of $2.2 billion. Congratulations. The last one you raised was in 2022 for $4.5 billion. The one before that was in 2021 and also 2.2 billion. I was curious why you chose to raise a smaller round and why there was like a longer gap and what your goal is for this particular fund.
- [00:42]
B
Yeah, happy to give you the rundown. So there are two major things that happened in the last few years that have really shifted the landscape for crypto. The first one is that the infrastructure has matured to a point where it can now actually support hundreds of millions of users, if not billions of users. Pretty soon we went from something like the 10 to 14 transactions per second that Ethereum and Bitcoin are able to provide to being able to send any amount of money from anywhere in the world to anywhere else for less than a penny. And in under a second, the infrastructure is great. So that's the first thing. The second thing is the regulatory landscape has become a much clearer landscape for builders, for founders and entrepreneurs. And with the passing of genius last year, which provides a legislative framework for how to think about stablecoins and legitimizes stablecoins, we've had very significant inroads in adoption for stablecoins specifically and for various different kind of application areas for crypto that stem from that. And so because of those two things, this feels like a perfect time to raise a new crypto fund to focus on this next chapter, which is one where we expect to see crypto grow up from its more kind of niche, cypherpunk kind of origins into becoming more of a mainstream technology and touching many more aspects of the economy and of kind of the tech ecosystem. As for the size of the fund, we always think about how to size a fund in a of bottoms up way based on the, on the kind of number of opportunities that we know are available. And one of the things that we wanted to make sure of is that, is that we went to a shorter cycle, shorter time cycle for funds. So essentially the two to three year cycle for deployment for the fund, as opposed to the longer sort of four to five year cycle that we, that we were in the previous fund, which, which motivated the kind of the reasoning behind raising a smaller fund this time around.
- [02:51]
A
Yeah. To your point about how we're in such a different stage now. You know, I was thinking about how in previous cycles it wasn't always apparent that crypto was going to become a thing. But now it's not only so apparent, but you know, there's this kind of crazy competition going on between crypto companies, fintech companies, you know, tradfi companies. And I wondered, because this is sort of that first moment where now everybody sees it and everybody is actively competing on the same turf. And I wondered how that informed how you guys will think about the types of projects and companies want to invest in this time around. Just your overall broader strategy and how you thought then it would differ from previous funds because we're at this different stage.
- [03:35]
B
Yeah, definitely. I think that there are maybe two broad trends that are both very exciting from an investing standpoint where we're seeing a lot of the activity and new founders be building. The first one is, is this intersection between crypto and fintech. I think you could actually go as far as to say that there will be no more fintech without crypto. That new, new fintech is actually now entirely crypto powered in the back end. The whole idea of the defi mullet of having an application or a service look very much like a fintech company in the front, but really be entirely driven by defi in the back end, I think has become the playbook for a lot of fintech companies that we are now, that are kind of being started now and are now building with this new technology in mind. So that's kind of maybe the first big trend. We've seen a lot of founders who are building companies with that playbook in mind. The second one, the second big one is the intersection between crypto and AI, which is also, I think it's an initial. It started to kind of ramp up recently and we've started to see quite a few companies in that category as well. And in particular around things like agentic commerce, where crypto is the technology that can help convert an AI agent from what normally would be a tool that a human uses into something that's more like an economic actor that actually is a first class participant of the financial system. So that's an interesting category as well. And the two kinds of opportunities that we. There are many others, but those are the kind of the two major ones.
- [05:17]
A
Okay. Yeah. I mean this is really just this moment where you can sort of see like this is when sort of the real Adoption phase begins and the real competition happens, and sort of the real winners and users will be made. And I think what's interesting, too, is, like, this is happening at the same time that it looks like Clarity will possibly get passed. You know, Bitcoin has sort of been in the clear a little bit, so there's kind of a lot more activity around that. But sort of the rest of crypto for a long time has had this sort of regulatory overhang. And I wondered how much of the thesis of the fund sort of depends on Clarity getting passed. And, you know, if it doesn't get passed, like, does that change how you invest? Like, would you, you know, would it be more companies that maybe aren't going to be available to Americans or, like, how is that part affecting how you're thinking about this?
- [06:13]
B
Yeah, well, I think the first thing to point out is that we've come a very long way in terms of regulatory clarity with things like Genius and with all of the rulemaking from the agencies that the guidance. The guidance that we've seen from the SEC and the cftc, which already make the space a much better environment to build new companies than it was in the previous administration, which was overtly hostile towards the space and actively trying to ban it. Clarity will be kind of a significant next step for the crypto world to take. It will do for everything outside of stablecoins in crypto, what, what genius did for stablecoins and where stablecoins are, say, maybe 5% of the space. Everything else is the 95% that remains will be governed by Clarity. And so it is a very important, an important piece of legislation to get through. We're still very optimistic that it will. That it will get it done this, this year. I think, I think, as you, as you've seen very recently and as of Friday, there's been significant inroads into getting getting Clarity over the line and actually getting to a markup ideally in the next few weeks. Even in the event that it doesn't pass, though, we again, still expect that the regulatory agencies, through rulemaking and guidance, can make for a much better environment to build in the space, as they already have been doing. We expect fully that they'll continue to do that. And we are also not stopping until Clarity or some form of legislation like clarity finally makes it into law, whether it's this year or in the next. The next few years.
- [07:51]
A
All right, so in a moment, we're going to talk about one of Ali's biggest theses around privacy as a moat. But first, we're going to take a quick word from the sponsors who make this show possible. You know what I'd do with a share of $1,000,000? Honestly, I'd probably buy a one way ticket to an island, bring a good book and invest the rest. And if I were sitting VIP at the World Football Final this summer, I'd be dancing in my seat the the whole time. And both are possible. This Coinbase One member month hit a 14 day streak to split $1 million in Bitcoin and the first to finish gets a VIP trip to the World Football Final. Make at least a $25 trade, stake, deposit or spend on your Coinbase One card. Any of these count and do up for 14 days this month. If you want that VIP trip. That means you have to start today and be the first to complete. Coinbase One is the ultimate membership to make the most of your money. Zero fees on thousands of crypto assets, 3.5% APY on USDC boosted staking and lending rewards, and up to 4% Bitcoin back with the Coinbase One card. If you trade crypto regularly, the basic annual membership can pay for itself. Hit a 14 day streak to get your share of $1 million. Plus you still get 20% off the first year of Coinbase One annual plans and a $50 Bitcoin bonus when you spend $100 on a new Coinbase One card in the first 30 days. Don't miss your share of $1 million at coinbase.com Unchained Again, that's coinbase.com Unchained. Back to my conversation with Ali. You wrote at the beginning of the year that you believe, quote, privacy will be the most important moat in crypto this year. And I'm going to say something a bit controversial, so I know it might sound like a joke question, but this is actually a real question. Is that why Canton is a top 20 blockchain at the moment? Like, do you, do you think Canton exemplifies your thesis?
- [09:54]
B
Yes, I do think it. I do think it does. I think. I mean, privacy is one feature that everybody agrees is absolutely critical to get to get the world's finance on chain. And that is also the one feature that almost no blockchain has really at all. And it's essentially been an afterthought up until now. There are of course many efforts to add privacy to existing blockchains or to build new blockchains that have privacy. But by and large most of the efforts of the space up until now have been focused on scalability and developer Experience and not so much on privacy, and maybe rightly so, because those were the things that needed to happen first. Like sort of the most fundamental aspects of the infrastructure were to just be. To be able to run computation on chain. And privacy has always been more aspirational. But now that we are at a point where crypto is starting to touch institutions, institutions are interested in adopting crypto in a real way. It's becoming painfully obvious that privacy is essential and that it's completely missing. So projects like Canton, many other projects, some of which are in our portfolio that do have privacy, are very well positioned for this next era.
- [11:09]
A
So something that was really interesting to me was I was listening to a podcast you did about this topic, and in it you said that you feel like once a company, or really just any user starts using a privacy blockchain, then that becomes a huge moat because the cost of switching to a different chain raise the potential of data leakage. And it made me wonder if then you feel like now it sort of looks like Canton has an edge because it already has a number of large financial institutions and it is a top 20 chain. There are some other privacy chains, but you know, none of that. I mean, they're, they're all sort of different, you know, some of them are more like just pure payments. So yeah, just tell me a little bit about like how you think the privacy blockchain competition is shaping up.
- [11:58]
B
Yeah, I would say it's still likely not a winner take all type of dynamic where you have only one chain that wins everything. There are many different kinds of use cases and the trade off space is very broad. So I suspect there will be different kinds of chains, all of which maybe have privacy as a feature that specialize in the various different regions of the trade off space and all kind of carve out their own niche. What ends up being on chain finance or. I think that the argument that I was making in that post that you mentioned is that block space is essentially becoming commoditized and performance by itself is no longer enough of a differentiator. And the reason for that is largely that it's become very easy to move between chains, given that we now have very good interoperability solutions and cross chain asset bridging solutions. And so therefore, if you are just a public blockchain that has high performance, your fees are ultimately likely to be compressed down to zero because it's very easy for people to migrate to other chains. And block space, high performance block space ends up being the same everywhere. It's no longer really Enough of an edge that any one chain might have. And so as a result, other features kind of like, I mean, privacy being the best example, maybe the key thing that is needed for a layer one blockchain to be defensible. And privacy has this unique. One of the unique things about privacy is that as you alluded to, private state is much harder to migrate between chains. So once you have decided to use a particular chain and to adopt it and to make it part of your infrastructure, it becomes much harder to switch to a different chain than it is if you're building on top of a public blockchain. And part of the reason for that is that whenever you move from one privacy zone to another, if you have encrypted state, you have to decrypt that state and then re encrypt it and move it into another private zone. And that whole process risks leaking information and undermining privacy. And so this one shift, this one change that privacy introduces, creates stronger network effects that make privacy chains more of a winner take most type of dynamic that are more defensible than just public chains. So yes, I think Canton's in a good position. Other privacy chains will be in a good position too. And I think that it will be winner take most type of situation where we end up with a handful of winning chains, all of which have privacy on which most of the world's finance, as it moves on, chain ends up running.
- [14:38]
A
So when we look at the different privacy chains that are currently competing with each other, it's sort of interesting because I see Monero and Zcash and Canton all close together between number 17 and 20 on the list of biggest market cap chains. And then I don't see another for a very long time. And I know you've invested in some chains like ZK Sync or Alio, and I was just wondering, you know, how you think this competition will keep going, because two of those chains are just old privacy chains that have been around forever. They're sort of benefiting from this recent uptick in interest in privacy. And then Canton, you know, has certain specific things about it. It's got all, you know, like Goldman Sachs and JP Morgan and, you know, London Stock Exchange kind of doing things on it. So how do you expect that this competition will play out?
- [15:30]
B
Yeah, well, I think that it's worth teasing out the various different approaches that exist for privacy. So there are the very pragmatic approaches which just involve trusting a centralized party to handle privacy for you, where users send transactions to a centralized sequencer who can See the transactions, so the transactions are not private from that sequencer. And then via something like a ZK proof potentially, or even just based on the trust in that centralized party, things eventually get settled on chain, but without all of the detail about all of those transactions becoming public. So that's like the most centralized, most pragmatic approach that lets you have privacy today and allows you to then cater to institutions who absolutely require privacy and want a very high degree of control.
- [16:15]
A
So you're like referencing Canton, there's Canta
- [16:19]
B
and there's a number of other players. And by the way, many of these projects are in different stages of their roadmap because many of them do want to be more decentralized and fully credibly neutral. And so depending on the progress that they've made internally or the degree to which they've changed their architecture to something that provides privacy in a different way, they may no longer fall in that category as time changes continues. But then the, maybe the intermediate category is one that relies on trusted hardware, where you're essentially trusting chip manufacturers like intel who build in trusted enclaves into their chips, which provide you some protection from the people who interact with the computer as it processes transactions. It becomes much more difficult to be able to hack and exfiltrate information from a computer that's running transactions inside of a tee, because you really, in order to really do that, you actually need physical access and physical control to the computer. So that's another level of protection. And you can provide privacy in that way where there's still trust in the system, but it's like a far higher tier in terms of the guarantees that you get. And then the final approach that is maybe the holy grail would be to provide privacy in a way that's entirely cryptographic and based on math, where no one, you're essentially not trusting anyone for the privacy of the system. And as maybe the most advanced and most difficult to get right and leverages things like zero knowledge cryptography to provide those guarantees. So we've been involved at every level. I think we actually have portfolio companies, some of which are not announced for every approach. So the ones that you are referencing are some of the ones that are leveraging cryptography for this purpose. Again, those I think are the, they're the holy grail and the hardest to get right. I think ultimately we will get there, especially as security, computer security becomes ever more important with AI agents being the source of, of more hacks. As, as we've seen, end to end, cryptographic security will become more Important as that continues as we, as we realize that computer security is essential. So I do feel like in the long arc of time, the end game will be to incorporate some of the most cutting edge cryptographic solutions to provide privacy.
- [18:39]
A
Okay, so as you just referenced, there have been so many hacks recently, and even just like having this discussion about privacy almost feels nerve wracking to me in this context. And I was wondering if you could describe for me in a world where privacy chains are more dominant and by the way, that's obviously different from a private blockchain for people who might remember. Yeah, those, those days. But anyway, so you know, do you think, like, what happens to the potential for hacks in a world of privacy blockchains? Does it vastly increase or does the fact that it's a privacy chain also make it harder to perpetrate a hack or, you know, only. Or does it only make it harder to detect a hack? Like how, how does it affect security if we go to a world that's predominantly privacy chains?
- [19:33]
B
Yeah, I think the privacy mode is a harder mode to get security in because you don't have full transparency and it is harder to detect attacks that may have happened. One of the benefits that we have with public blockchains is that when an attack happens, it's very easy to tell. And there are countless people who are observing the chain at any point in time who do tend to see such attacks and flag them. And therefore the response to vulnerabilities can be much, much faster. With privacy, that becomes harder. But this is why some of these more advanced technological solutions to the privacy problem are so important. Cryptographic solutions like zero knowledge proofs allow you to create a mathematical proof of the correctness of a system that anyone can verify and give you some of the same properties of transparency that you have on public blockchains, in that you can convince yourself that nothing went wrong by just verifying a zero knowledge proof. Whereas the more pragmatic centralized approaches are more vulnerable to an attack that might steal money or steal funds in a way that really is undetectable, because again, you're trusting a single party. And if that party gets hacked, and because everything is private, we may not find out that something went wrong for a very long time. So this is one of the motivations for moving over to these more advanced solutions. Now, one counter argument that we often hear is that even the cryptography itself for these advanced solutions may itself have bugs. So you can have a zero knowledge system that is buggy, such that an attacker can actually generate a Proof that looks correct and verifies from the perspective of an outsider, but actually manage to somehow subvert the integrity of the system and steal money. And again, maybe the good approach to deal with that kind of thing is to always use redundant systems. So some of the advice that we've been giving some of our projects in this category is that you could always pair two different proof systems at the same time for any kind of transaction or all activity that through completely different code paths, provide the same kind of guarantee so that if one of them has a problem and has a bug, the other one is always there to kind of as a fallback to kind of catch any issues.
- [21:57]
A
Yeah. This reminds me of when ethereum experienced those DDoS attacks in 2016. And the fact that they had multiple clients is what enabled the chain to keep functioning because some of the attacks only targeted, you know, like Geth, and some only targeted parody. And. Yeah, so it. Yeah, it's just a resiliency thing.
- [22:17]
B
Exactly.
- [22:18]
A
So then I wanted to also ask. Just about. So sorry, I lost my train of thought. Oh, I remember. So, as you mentioned, you felt like that the cryptography was kind of one of the safest ways to, you know, have prophecy on a chain. What about something like a trusted execution environment where it happens in hardware? Would the attack vector for that kind of require something similar to like the wrench attacks that we're seeing today or like. Yeah, because I know there's like different ways to, you know, for this to occur.
- [22:52]
B
Yeah. Well, so there are two potential attack vectors. One of them would be getting physical control of the machine that is running the secure enclave, because none of these tees or trusted execution environments are resistant to an attacker who is able to take physical control of the machine and then use tools to extract the confidential state of the tee. And so then the attack would be someone if the machine is running outside of a data center. If it's like, so like it could be if the machine is run by a startup and there are employees who have access to that computer, then they could potentially subvert the privacy of the system. But that is less of a concern if the machine is running inside of a data center, where it seems far less likely that an employee would be able to break into a Google data center or an AWS data center in that same way. The other attack vector may be just bugs in the system. It could be that the way that the TE is implemented exposes, is exposed to, say, side channels. There's many kind of side channel attacks that have been documented over the Years where you can actually exfiltrate data from the tee just in software, without having to have physical access to the machine. But I think as time goes on, people find the vulnerabilities and fix them. And it is definitely a higher level of guarantee than one where you're just relying on a centralized party to behave correctly.
- [24:22]
A
And then so on the other side, like once a hack happens, how does being in a privacy chain world affect the ability to track down those hackers?
- [24:36]
B
Well, again, I think this is where zero knowledge cryptography can really help because it can allow you to prove various different aspects of the system without necessarily unveiling or unshielding people's privacy and undermining people's privacy fully. It gives you that kind of like a granular lens into what actually happened. That is the best of both worlds. You have some degree of transparency, some guarantees as to what actually happened in the system without having full visibility into everyone's transactions. That's kind of the ideal case. In other cases, the only answer would be to, to have to audit all of the transactions and to have maybe trusted parties look through everything that happened to be able to determine whether something went wrong. Which is, by the way, the way that it works in, in traditional finance where we have no such guarantees. We, we have full trust in banks and full trust in financial institutions that everything is going as it's supposed to be going. And whenever something goes wrong, it's this insane forensic effort where just humans have to get involved and look at this Excel spreadsheets and comb through every transaction to figure out how it is that an attack happened or where it is that the money went.
- [25:54]
A
Yeah, and as far as I understand, I don't think it's even that efficient. I think it's literally the prosecutors have to send requests to banks. The banks might take months or even a year to respond with the information. And all you get is just one hop in the, in the transactions. And then you, once you, you know, realize where that hop went, then you have to send another legal request to another bank. So anyway, so yes, point is. Yeah, so, but, but would that, then that would be done with like viewing keys or something? Is that how that would go?
- [26:25]
B
Okay, yeah, exactly. I mean, I think the beauty of some of these, some of the cutting edge privacy chains is that they're fully programmable. So you have, you have arbitrary flexibility to build a system that has the right safeguards and the right mechanisms for getting information about what's happening in the system. So that when something happens, if something Happens you have the tools to be able to track down the attacker and also so that you can comply with laws and regulations. It may be the case that if you get subpoenaed you may need some visibility into the system. Or you for example, might need the ability to freeze a particular address if it happens to be sanctioned by ofac, if it happens to be associated with North Korea or something like that. All of those things need to work in a system that also preserves users privacy. And the only real way to do that is with a programmable system that lets you express complex rules and requirements to be able to satisfy all of the constraints whilst maximizing the kind of what users would want, which is their own privacy.
- [27:28]
A
Okay, yeah, all of this is super interesting because I am very interested in financial privacy. People might remember, or maybe just because I'm older, but I don't remember what year this was, but there was this New York Times magazine article that came out about how stores were using the shopping data to predict what your next purchase will be. And a father got very angry when his daughter received a, his teenage daughter received a marketing flyer for, you know, something regarding babies basically. And she was, you know, a teenager and it turned out actually she was pregnant. And you know, the fact that Target had figured known this before the father was very disturbing to him. And anyway, the, the reason why I'm talking about all this is just that, you know, your financial data really says a lot about like what is happening in your life. So I think privacy in, you know, financial transactions is so important and yet at the same time I'm not a fan of North Korea. I am ethnically Korean and you know, just think they're very bad actors. And so it is very disheartening to see that this world that I'm so enthralled with is also, you know, the source of a lot of funding for North Korea's nuclear weapons. Anyway, so I did then now want to ask a little bit more about the privacy and maybe this is just me not knowing the technicals about this, but in this world where privacy chains become more dominant, you know, we're used to right now seeing a lot of kind of like composability in defi. You know, unfortunately recently we saw the downside of that. But, but this money Legos concept, like what happens to that in a world of privacy chains, does that go away in some fashion? Because you know, I just think about what you said about how one of the moes is the fact that when you switch to different chains there is like that data leakage that could potentially, you know, like make some part of it unprivate or public rather. So yeah, what happens to this notion of defi composability in a world like that?
- [29:44]
B
Yes, well, I think this is why chains that have privacy will have stronger network effects. The reason being that it is easier to provide composability within a single private chain than it will be across chains. And so within a, a particular private chain, if it's still, still a unified system, you may still have the ability to compose different smart contracts and different systems that are built within that chain or within different zones of that chain, if it's all kind of a unified architecture. But then I think in order to provide for composability across chains, then you'd have this issue where you have to re encrypt state from one chain over to the other chain. And whilst that will be possible, I do think it'll be less likely to happen because of the fact that potentially leaks a lot of metadata that could, that could actually undermine privacy. And so I don't think the composability dream is dead in that you can still have composability within privacy zones and cross zone composability will also continue to exist. It'll just have this additional hurdle that you'd have to cross in order to move from one chain to another.
- [30:51]
A
Oh, okay. So basically what's fascinating to me is it almost feels like it is something like a cosmos vision of like kind of the base layer and then app chains. Is that how to think about it
- [31:09]
B
for certain, for certain projects? Yes. Where I think the, the only real way to scale is to have different shards that can operate independently of one another so that the entire chain can scale horizontally. And in that world it, it looks a little bit like, like kind of like the cosmos vision, but if you have like tight integrations between all of those zones so that they can still, still interoperate, you preserve the benefits of composability that you have if you were to just stay on a single chain.
- [31:41]
A
Okay, and so this is just very interesting to me because if I kind of keep playing out what you're saying, then it almost feels like, and tell me if I'm completely wrong, but it almost feels like then the natural kind of conclusion is maybe that at least at this moment in time, Ethereum is sort of best positioned in this way because they're making privacy a priority. Like they already have like, or you tell me like. Yeah, I'm curious what, what you think about that.
- [32:11]
B
Well, it's hard for me to make statements about any specific project that we have to be. I'm sort of careful about investment advice. I do think that the, I mean the architecture inevitably will have to be one that, that scales horizontally. And I think a single chain where every validator has to process every transaction and, and confirm that the work that every other validator has done is correct, is one that just inherently doesn't scale. And it's okay to go up to say something like 10,000 transactions per second with that architecture, but in a world where maybe we will need something like millions of transactions per second as agents begin to transact, and as most of the financial transactions in the world are actually driven by agents as opposed to humans, I suspect we're going to need a far, far higher number of transactions per second provided by our infrastructure to really meet the challenge of what's coming. And the only way to do that is again through horizontal scaling. It's a system where you can add compute and rather than have that compute, just have to redo the work of all of the other computers that are already in the system, have those new computers process new work, process work that's unrelated so that it can happen in parallel and you can continue to scale the system horizontally in that way. So whatever architecture wins, whether it's ethereum via its L2 roadmap, or whether it's other chains that are following this architecture, I mean we. Another example of this is layer 0 with the recent announcement of 0, the 0 blockchain, which has the same property. I guess Canton also has different zones, all of which can operate independently from one another. Tempo and Arc have similar visions for how they will scale down the line. That is just inevitable in my mind. In order to scale a computer system you need parallelism and you need hierarchy. And the single monolithic chain can't really scale indefinitely. It'll cap out at some point.
- [34:05]
A
Yeah, yeah, yeah. And I agree with you that even though I just said what I said, we're still so early that yeah, the competition is definitely not going to be decided for a while. Okay, let's talk about the AI future because your background was in AI, so I'm just going to ask you the most open ended question about this because I just want to know what you're thinking as you're watching, you know, kind of your, your past interest in line of work take off with crypto. Yeah, I love your top line thoughts.
- [34:42]
B
Well, I love seeing all of the progress in that world and, and I have to admit I have this deep affliction, which is that I'm addicted to vibe coding and to spending a lot of my nights and weekends just playing with the available AI tools and letting the AI psychosis wash over me. And it's incredible too just how much you can do with AI tools today. I mean, I'm just kind of reflecting as to my productivity level as a full time software engineer at Google for three years and the amount of code that I produced throughout that time and then the amount of code that I'm producing now, very much just nights and weekends only. It's very part time, it's a couple of hours every week. And I think that I'm actually more productive as an engineer today than I was then, which is absolutely staggering. It's also really cool to see some of the intersections between crypto and AI just because of the fact that agents today are very single player in that they are very much a tool that a human uses to get a job done. But it's very clear that where this is going is that agents will become economic actors, they will become first class members of the financial system, they will be able to pay for things, they'll be able to get paid for things, potentially, they'll be able to raise money. There are all sorts of financial behaviors that they'll be able to engage in that they currently cannot. And it's inconceivable to imagine that all of that is going to happen on ACH and wire transfers and the traditional sort of payments rails that we rely on today. Crypto, on the other hand, is Internet native, it's fully programmable, it's instantly settled, it's global from day one. It's the perfect fit from a technological standpoint for AI agents as a way of bringing agents into the financial world. So it's exciting to see all of that play out. There are many companies that are running after this opportunity, some of which already are already in our portfolio. So couldn't be more excited about that whole trend. The other big one too is, is the way in which crypto can help combat the big problem of AI deepfakes and kind of helping us reason about the provenance of media and information. Crypto is purpose built for that problem. And so in that way, crypto and AI are also really good counterweights for one another.
- [37:05]
A
Oh, okay. Can you talk a little bit more about that? Does that mean that something like a world coin you feel like is kind of going to be a crucial piece of this new world that we're moving to?
- [37:19]
B
Yeah, I think that increasingly is going to become very difficult to know whether the people you're interacting with online are actually people. And I think this is already an issue in places like sort of say, dating websites or just general social networks. And it's only going to become worse as the AI models become better and as people realize that actually if it's quite easy to, to pretend to be a human now, now that you have as powerful a model as, as people have today, crypto is a perfect solution to, is like a, is a perfect technology to, to, to combat those problems, because I think world is a great example. They've built a technology that allows you to prove humanity, to prove that the, the person who's actually interacting with you online is, is actually a human. And they also done that in a way that preserves that person's privacy without actually taking any of the biometric information and having it be stored anywhere. You guys have seen the orb, the world orb, none of the biometric information ever leaves that orb. And instead, again, we use zero knowledge cryptography in order to create a proof of humanity that you can then use in a dating site to prove that you're actually a real human.
- [38:36]
A
Yeah, yeah, I think like, that is, you know, something that I'm noticing more and more. You know, yesterday on X, I was fed like a, a tweet about some very successful OnlyFans account. And it shows the guy making the video and then the AI image of the woman, and he made, you know, healthy five figures in the first month. Yeah, well, okay, so I did want to ask, you know, like, we are in this era, as I mentioned, even just earlier on in the show, before we were talking about AI agents, about how we're at this moment where clearly, you know, every, every entity in finance is converging on the same turf. And part of that competition we're seeing is in these different payment protocols for agents. And you know, I don't know how deeply you've looked into all of these. There's, there's actually quite a number of them already. X402 by Coinbase is well known. AP2 by Google, there's MVP. It's just I, I was curious to, to hear you talk about, like, what you think will be the different structures of, of those payment protocols that will, you know, be advantages or disadvantages and like, how you think that competition will play out there.
- [39:49]
B
Yeah, I mean, I largely, largely, especially MPP and S402, they're essentially the same thing. They're, they're very similar. I think that naturally the Reason that multiple standards exist is because there are sort of political allegiances between the various different companies that are competing against one another. I think that the way that this plays out will depend strongly on just the level of traction that each of these companies ends up getting. And ultimately the thing that ends up getting the most traction most quickly will begin to accrue network effects. And then that'll just become. That'll become the standard that everybody rallies around, whether it's better or not. So I think this is the thing with standards. They have very strong network effects, even though they themselves are not networks. And once a standard becomes standard, it becomes very hard to dislodge simply because if you want to interoperate with everyone who's already using the standard, that's the only option you have.
- [40:44]
A
Okay, so that is more like a go to market competition, you're saying.
- [40:48]
B
Yeah, I think driven by product, driven by go to market. Whoever wins at actually gaining adoption will have the privilege of defining the standard.
- [41:00]
A
Okay, so I also wanted to hear you talk about, when you think about what a mature agentic AI economy will look like, whether that's a few years from now or even five or ten years from now, describe what our everyday lives will look like.
- [41:20]
B
Yeah, I actually really liked John Collison's five layers or five levels of autonomy that he talked about in the annual stripe letter. I have my own levels of autonomy that are sort of somewhat simplified. And for me, there's three levels of autonomy in my mind. Number one is having AI completely eliminate all of the tedium that's involved in buying things online now. Today it's still a process of filling in forms in which you add a credit card and you add your address and your zip code and all these things. And that's the process by which we still buy things online in 2026. All that can go away. It could just be entirely driven or automated by AI. And that's maybe the first layer of autonomy. The second layer would be one of descriptive search and delegation where you just tell your agent the kind of thing that you're looking for, and then the agent is capable of going and finding the thing that you want to buy. Maybe negotiating prices or comparing prices across different potential merchants, finding you the best price and then actually executing on the purchase and getting you the thing that you want. And it can even be high level. It could be like, well, I need to buy supplies for school for my kid. And then here's the kind of the list of overall things that need to be bought and Then the agent just goes forth and does it all for you. And then the third level is what I would call more like agentic finance, where agents do become first class members of the financial system, where you have an agent that has control over a crypto wallet. It can pay for things, it can get paid for things, it could potentially raise money, it can buy its own compute, it can buy data in order to sustain itself, and it can become a self sustaining sovereign entity in the world that can generate value and can build services or products in order to continue to survive, quote, unquote. And that's maybe the most futuristic example of what may be possible with the intersection of crypto and AI. And it may be impossible today in that if you ask even ChatGPT 5.5 or even Mythos or Opus 4.7 to go forth and generate value and not run out of money, it probably does not work. It probably will get prompt, injected and ultimately run out of money. But I think given the exponential trend on which we're on, give it another year, give it another two years, it may be very possible that you have essentially companies whose founder is an AI, as opposed to necessarily a human.
- [43:57]
A
Yeah, yeah, no, I definitely feel that way. But actually that leads me to my next question because this is something I've been thinking about a lot. So obviously you saw what happened with that Citrini article. There are these panics happening about how AI will affect jobs, in particular white collar jobs. And I'm a big fan of how a 16C for years now has been talking about how blockchain technology will change the way creators earn on the Internet. And when Chris Dixon and Hessey Qureshi had their little disagreement on X earlier this year. About what? Whether crypto is only for financial applications or whether it can be used for these other things like, you know, to upend creator business models and things like that. I was Team Chris and you know, I sort of look at where we are with this AI moment and I feel like we're at this place where basically a lot of we, we've like. So I, I know that you also went to Stanford like me, so you may remember from college there was this dichotomy between like the fuzzies and the techies. Yeah, yeah. So I'm sure you were probably a techie. I'm a fuzzy. But, you know, and for anybody who didn't go to Stanford, you could probably figure out what those mean. So anyway, you know, at the risk of offending some members of the audience, I'm going To just name some jobs that I personally think will probably not really be as lucrative or will be largely replaced by AI, or they'll just be jobs where you're largely using AI to do the job. Some of these are already being disrupted. So lawyers, certain types of software, slash IT type jobs, certain types of financial jobs, like accountants, transcriptionists would be another. Basically anything where you get paid a lot of money to do something that most people find somewhat boring. That is the category that I feel like is going away. And as a creative person, I sort of look at this moment in time and I'm like, oh, wow, this is the first time, at least in my lifetime where creative types like me and my friends have an edge. I would also put, like, entrepreneurial people in the creative bucket. So I wondered when you think about sort of those two kind of trends that are intersecting, like, so you guys have had the one theory for the long time about how crypto and blockchain could upend the creator business model. But then we have this other thing coming where the AI is interrupting these jobs, like I said, that maybe pay a lot for things that are a little bit more repetitive or something like that. I was wondering, like, how you thought this would play out in terms of that traditional power balance we've had between creative types and then people who have, you know, traditionally earned more for certain professions that may not be able to going forward.
- [46:49]
B
Yeah, I do. I do think that AI will dramatically change the landscape. And it will actually. It will change the landscape for the better and that ultimately it'll lead to far greater productivity in most cases. I think it will actually lead to more employment rather than less, as is the case with software engineering. And we can talk about why. I think, as for the delineation between roles, I think you are right that AI will shift the advantage to people who are generalists and who are capable of reinventing themselves. In particular, people who are able to use AI itself to make that happen, to make that happen for themselves and to reinvent themselves and to teach themselves new skills and new capabilities to be able to continue to contribute value in our companies. One thing that we're seeing is the blending of the roles of the product manager, the designer, and the engineer, where every one of those three believe that they can do the job of the other two. And Mark Andreessen actually referred to this as the Mexican standoff between the three. And it is very true in that a very resourceful engineer can now do a lot of product management and a lot of design if they're Just willing to do the work and use AI in the. In the sort of most powerful way possible to. To be able to kind of reposition themselves and do. And do that work. And the same is true for the other two roles. And a lot of what now is most prized and rare is, Is the. Is the level of judgment and design taste and product vision that a human, at least today, can only. Can only provide. So I do believe that there will be big shifts. This will just result in companies being able to do much, much more than they were before. Productivity skyrocketing. In the case of software engineers, actually this leads to greater employment of software engineers, not less, because every software engineer can now generate that much more code, which means you can move that much faster and you get that much more bang for your buck for every engineer that you hire. And it turns out the demand for software is basically infinite. And so you might as well generate 10 times more software instead of spending 10 times less money on fewer engineers, which is what we were seeing. It's interesting to see that actually there's an uptick in hiring for software engineers as opposed to the opposite, which you would expect if you believe that AI will just replace all software engineering.
- [49:06]
A
Yeah, yeah, I don't think that it will replace software engineers. I think it will replace certain types of. But, okay, okay, so we only have five minutes left. So I definitely want to try to hit maybe two more topics, but at least one which is, I'm sure, you know, everybody's looking at this quantum threat and they're wondering what's going to happen. I think obviously bitcoin has been taking most of the headlines just because of the kind of very particular vulnerabilities there, like around satoshi's coins and things like that. But smart contract chains really also have quite a large surface area for ATT and ck. And I wondered, you know, what you, like, thought would be the best approach for the crypto community to take. And it could be either one of those, like Bitcoin or just the, you know, the rest.
- [49:50]
B
Yeah, I mean, the smartest people I know in the world of cryptography and quantum computing believe that we are still 10 to 15 years out from having a cryptographically relevant quantum computer. So there is still time. That's not, that's. That's not to say that we shouldn't be thinking about this now and planning for the transition. It is possible.
- [50:08]
A
But are you saying even with the recent breakthrough, that.
- [50:12]
B
Okay, yes, even with that. I mean, to have a cryptographically relevant quantum computer that can break something like sha256 or any of our like elliptic curve cryptography. That's still very far out. It could be that, that there are more breakthroughs on the like down that are in the pipeline, especially with AI. And so then again, it could be that those timelines compress. But kind of the latest estimate from where we are now, we're still very far away. And a lot of it is actually the construction of a quantum computer that is that powerful as opposed to sort of the knowledge of how to do that or the algorithms that you would run on a computer once it exists. So given that we're still far away, I think that the trade off to navigate is that on one hand you want to prepare, you want to transition everything to a new system and make sure that we are secure from a quantum perspective. But on the other hand you don't want to prematurely shift because again, there could be bugs in the cryptographic systems that we migrate to, or it could be that we migrate to the wrong thing and we don't do it thoughtfully enough and then we have to do it again down the line. And so there is this tension between these two kind of optimization functions and we want to find the one that like the balance between the two, that simultaneously keeps everyone safe, but also make sure that we make the transition happen safely and that we're also thoughtful about the kind the cryptographic schemes that we, that we ultimately migrate over to. So some of the conversations on Twitter are like, this has to happen immediately, has to happen tomorrow, like quantum threat is imminent. And then there are people on the other side who are like, no, this is not non issue at all. And I suspect just the answer, the real answer is somewhere in the middle. We have to be thinking about this now is not like a, like a kind of P0, like we have to get this done tomorrow in the next few years we should have an answer for how we transition our critical cryptography over to quantum secure systems. And it's just a matter of being thoughtful about the balance.
- [52:05]
A
Okay, last question. Prediction markets have been talked about kind of everywhere. And one of the more recent cases was something that, you know, I saw a lot of people kind of agreeing on the fact that they felt that the soldier who was arrested for bidding on the Maduro raid, like they feel like that was kind of what should have happened. I have heard some, some like other opinions though. You know, some people saying, well, he didn't know if they were going to, if he was actually going to make the money because he didn't know if the rate was going to be successful. And those people, you know, say things like, well, the point of prediction markets is to have insiders who help make the markets more efficient. So I was kind of curious to hear your take on that. You know, do you think prediction markets should allow these quote unquote insiders, which, that's the more colloquial definition, not like a legal definition to trade or should that be prohibited?
- [52:56]
B
Well, I mean, I'm not a lawyer, but there, there is quite a bit of legal detail as to how insider trading is treated. And it's not, I think a black and white black. It is not a, it is a fairly black black and white thing for a lot of these cases where for example, if the person who was trading had a duty to a shareholder or to a third party and they're essentially breaching that duty by acting on confidential information, that's clearly illegal, it's clearly insider information. But on the other hand, if you are a hedge fund and you buy satellite data to figure out whether say Walmart, Walmart will have high earnings or low earnings by monitoring patterns of traffic in their parking lot, like that is not you breaching any kind of duty to anyone. And that is information that is essentially available to the general public. And that's exactly the kind of thing that you should be able to trade on. And so there are, I think, frameworks for how to think about what counts as insider trading in both securities, securities law and also commodities law. And I think a lot of that applies to prediction markets and especially if prediction markets end up being regulated as commodities markets. And I think maybe there are cases where there's a gray area, but as a non lawyer, I wouldn't know how to exactly think about those cases.
- [54:17]
A
Okay, well Ali, it has been such a pleasure talking with you and congrats again on a 16Z's fun five.
- [54:24]
B
Thank you so much. Thank you for having me and thanks
- [54:27]
A
to everyone for joining this live stream. We will catch you tomorrow. Bye everyone. Sa.