A16z Crypto Raised $2.2 Billion for Fund 5. Here's How They Plan to Deploy It

Transcript

Hi everyone, welcome to Unchained, your no hype free source for all things crypto. I'm your host Laura Shin. Thanks for joining this live stream. >> [music] >> Today's guest is Ali Yahya, general partner at Andreessen Horowitz. Welcome Ali. It's great to be here. Today you announced that a16z crypto raised its fifth fund of $2.2 billion. Congratulations. Thank you. >> 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 this smaller round and why there was like a longer gap and what your goal is for this particular fund.

Yeah, happy to give you the 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 for founders and entrepreneurs. And with the passing of genius last year which provides a legislative framework for how to think about stable coins and legitimizes stable coins, we've we've had very significant inroads in adoption for stable coins specifically and for various different kind of application areas for crypto that stem stem from that. And so because of those two things, this 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 uh its its

more kind of niche Cypherpunk um uh kind of origins into becoming more of a mainstream technology uh and touching many more uh many more aspects of of the economy and of of kind of the tech ecosystem. As for the size of the fund, I mean we we always uh we always think about how to size a fund in a kind of bottoms-up way based on the on the kind of number of opportunities that we know are available. And uh one of the things that we wanted to make sure of is that is that we uh went to a shorter uh cycle, shorter time cycle for funds. So, essentially, the to 3-year uh cycle of uh for deployment for the fund as opposed to the longer sort of 4- to 5-year uh cycle that we that we were

in the in the previous fund, which which motivated the the kind of the reasoning behind uh raising a a smaller fund this time around. 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, you know, 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, you know, just your overall broader strategy and how you thought then it would differ from previous funds because we're at this different stage. Yeah, definitely. I think that there are may maybe two broad trends that uh are both very exciting from an investing standpoint uh where we're seeing a lot of the activity and new founders uh be building. Uh the first one is is these 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. Uh the whole idea of the DeFi mullet of having uh an application or 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 uh fintech companies uh that we are now that are a ton of being started now and and are uh now building with this new technology in mind. Um so that's kind of maybe the first the big the first big trend uh we've seen a lot of founders uh who are uh building companies with that playbook in mind. The the the second one the second big one is the intersection between crypto and AI uh which which is also I think it's an it's an initial is it started to kind of ramp up recently and we've we've started to see quite a few companies in that in that category as well. Uh in a particular

around things like agentic commerce where crypto is a 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 um that actually is a first-class participant of the financial system. Um so that's an interesting category as well and there's like the the two kinds of opportunities that we There are many others but those are the kind of the two major ones. Okay. Yeah, I mean this is you know really just this moment where you can sort of see like this is when sort of the real um adoption phase begins and the real competition happens and um 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. Um 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?

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 rule making from the agencies that the guidance the guidance that we've seen from SEC and the CFTC, which already make the space a 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 to ban it. Uh clarity will be a kind of a significant next next step for for the for the crypto world to take. Uh it will do for everything uh outside outside of stable coins in crypto what what genius did for stable coins.

Uh and where stable coins are say maybe 5% of the space, everything else is the 95% that remains uh 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 it will will get it done this this year. I I think as you as you've seen very recently, but as of Friday, there's been significant inroads into uh getting getting clarity over the line and actually getting to a markup ideally in the next few weeks. Um even in the event that it doesn't pass though, uh we again still expect that the uh regulatory agencies through rule making and guidance can make for a for a much better environment uh to build in the space as they already have been

doing. Uh we we expect fully that they'll continue to do that. And uh we are also not not stopping until uh clarity or some form of legislation like clarity finally makes it into law, whether it's this year or in the next in the next few years. 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 the show possible. You know what I'd do with a share of $1 million 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 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 it 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 trading 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? Yes, I do think it I do think it it does. Um I think I mean privacy is is it one feature that everybody agrees is absolutely critical to get uh to get the world's finance on chain, and that is the one feature that almost no blockchain uh has and has it like really at all. Um 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 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. Uh like sort of the most fundamental aspects of the infrastructure were to just be to be able to run computation on chain. Um 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, so projects like Canton, uh many other projects, some of some of which are in our portfolio, that

do have privacy are very well positioned for this for this next era. 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 you know large financial institutions and it is a top 20 chain. There are some other privacy chains but you know none of that there I mean

there are 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. Yeah and I would say it's 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 carved out their own niche. What ends up being on chain finance.

I think that the argument that I was making in that post that you're 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 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 and block space high performance block space ends up being

the same everywhere it's no longer really enough of an edge that anyone chain might have. As well as a result um other features kind of like I mean privacy being the best example maybe the the key thing that is needed for for a layer one blockchain to be defensible. And privacy has this unique one of the unique things about privacy is that as you 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 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 more they make privacy chains more of a winner winner take most type of dynamic that are more defensible than than just public public chains. So yes, I think Canton's in a good position. Other other privacy chains will be in a good position too.

And I think that it will be a winner take most type of situation where we end up with a handful of winning chains, all of all of which have privacy on which most of the world's finances it moves on chain ends up ends up running. 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 like 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 Aleo, and I was just wondering, you know, how you think this competition will keep going cuz two of those chains were 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 J.P. Morgan and, you know, London Stock Exchange kind of doing things on it. So, how do you expect that this will uh you know, this competition will play out?

Yeah, well, I think that uh 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 uh 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. So, you're you're like referencing Canton there? Canton, there's a number of other another other players and by the way, many of these projects are in in different stages of their road map because many of them do want to be more decentralized and fully credibly neutral. Um 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, um you know, they may no longer fall in that category as time continues.

But then then the maybe the intermediate category is 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 uh provide you some protection from uh 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 the computer that's running transactions inside of a TE uh 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. I would there still trust in the system, but but it's it's like a far higher for higher tier in terms of

the guarantees that you get. And then the final 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 where no one you're essentially not trusting anyone for the privacy of the system. And as it may be the most advanced and most difficult to get right and leverage things like zero knowledge cryptography to to provide to provide those guarantees. So, we've been involved at every at every level. I think we we actually have portfolio companies, some of which are not announced at every 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. Uh end-to-end cryptographic security will become more important as that continues as we as we realize that computer security [clears throat] 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. Okay, so as you just referenced, um there have been so many hacks recently and even just like having this discussion about privacy almost feels nerve-racking 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 and by the way that's obviously different from a private blockchain for people who uh might remember yeah those those days. Uh 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 um or does the fact that it's a privacy chain also make it harder to perpetrate a hack um 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?

Yeah. I mean I think do the privacy mode is the is is a harder mode to get security in because you don't have full transparency and that 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 and there are countless people who are observing the chain at any point in time who who do who do tend to see such attacks and flag them and therefore the response to uh vulnerabilities can be much much faster. Uh with privacy that becomes harder, but this is why some of these more advanced technological solutions to the privacy problem uh are so important.

Cryptographic solutions like zero-knowledge proofs allow you to create a mathematical proof of the correctness of a system uh that anyone can verify and give you some of the same properties of transparency that you have on public blockchains and 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 uh 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 uh and because everything is private, we may not find out that something went wrong for a very long time. This is one of the motivations for moving over to these

more advanced solutions. Now, one counterargument that we often hear is that even the the cryptography itself for these advanced solutions may itself have bugs. So, you can have a zero-knowledge system uh that is buggy such that an attacker can actually generate a proof that looks correct and and verifies from the perspective of an outsider, but actually, um, manage to somehow subvert the integrity of the system and steal money. Uh, and the the good I I got maybe the good approach to deal with that kind of thing is to always use redundant systems.

So, one of the one of the advice some of the advice that we've been giving some of our, um, projects in this category is that you can always pair two different proof systems at the same time for any kind of transaction or all activity, um, that, uh, it through completely different code paths was, uh, provide the same 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 to kind of as a fallback to kind of catch catch any issues. Yeah, that 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 Parity and

yeah, so it yeah, it's just a resiliency thing. Exactly. So, then I wanted to also ask, um, just about, uh, so sorry, I lost my train of thought. Oh, I remember. Um, so as you mentioned, you felt like that the cryptography was kind of one of the safest ways to, uh, you know, have privacy 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 how yeah, cuz I I know there's like different ways to, Yeah. you know, for this to occur.

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. Cuz none of these none of these are TEEs or trusted execution environments are uh resistant to an attacker who who's able to take physical control of the machine and then use tools to extract the confidential um state of the TEE. Um And so then the attack would be someone someone um if the if the machine is running outside of a data center of like if like a So like it could be um 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 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 TEE is implemented exposes uh 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 uh the TEE just in software without having to have physical access to the machine. Um but I think as time goes on people 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. And then so um on the other side like once an a hack happens um how does being uh in a privacy chain world affect the ability to track down those hackers? Well again I I think this is where zero knowledge crypto will be really helpful because it can it can allow you to prove various different aspects of the system without necessarily unveiling um unve or or unshielding people's privacy and and undermining people's privacy fully. Um it gives you that kind of like a granular

lens into into what actually happened um that is the best of both worlds. You have some degree of transparency, some guarantees as to as to what actually happened in the system without having full visibility into everyone's transactions. Um so that's kind of the ideal case. Uh 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 uh 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. 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 point is yeah. Um so but but with that then that would be done with like

viewing keys or something. Is that how that would go? Okay. Yeah, exactly. I mean, I think the beauty of some of these some of the cutting-edge uh 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 uh something happens, if something happens, you have the tools to be able to track down the attacker. And also so that you can comply uh with with laws and regulations. Uh it may be the case that if you get some peanut, you may need some visibility into the system. Or you, for example, might need might need the ability to freeze a particular address if it happens to be sanctioned

uh by OFAC. If it happens to be associated with North Korea or something like that. All of those things need to 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 and requirements um to be able to satisfy all of the constraints whilst maximizing the kind of what we users would want, which is their own privacy. Okay. Yeah, all of this is super interesting because I uh am very interested in financial privacy. Um people might remember or or maybe just cuz I'm older, but I remember what year this was, but there was this um New York Times Magazine article that came out about how um

stores were like using the shopping data to kind of 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 um you know something regarding babies, basically. And she was, you know, a teenager. And um it turned out actually she was pregnant. And you know, the fact that Target had known this before the father was very disturbing to him. And um anyway, the the reason why I'm talking about all this is just that um you know, your financial data really says a lot about like what is happening

in your life. So, um I think privacy in, you know, financial transactions is so important. And yet at the same time, um I'm not a fan of North Korea. I'm ethnically Korean and um 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. Um, 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, um, you know, we're used to right now seeing a lot of kind of like composability in DeFi. Um, you know, unfortunately recently we saw the downside of that, but um, 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 moats is the fact that when you switch, um, to different chains, there is like that data leakage that could potentially, um, 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?

Yes. Uh, 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 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 if it's all kind of a unified architecture. Um, but then I think in order to to provide for composability across chains, then you'd have this issue where you have to re- re-encrypt state from one chain over to the other other chain. And whilst that will be possible, I do think it'll be

less less likely to happen because of the fact that that leaks potentially leaks a lot of metadata that could that could actually undermine privacy. Um, and so, I don't think the composability dream is dead in that you you can still have com- composability within privacy zones and 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. Oh, okay. So, basically, um I 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?

For certain for certain projects, yes. Where I think the the only real way to scale is to have different shards. That uh can operate independently of one another so that the entire chain can scale horizontally. Um and in that world, it looks a little bit like kind of like the Cosmos vision. Uh but if you have like tight integrations between all of those zones so that they can still still interoperate, you preserve uh the benefits of composability that you have if you were to just stay on a single chain.

Okay. And so, this is just very interesting to me cuz if I kind of keep playing out what you're saying, then it almost feels like And 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. Well, it's it's hard for me to make statements about any specific project.

Um that I we have to be I was sort of careful about investment advice. But I do think with 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,

um I suspect we're going to need a far far higher number of transactions per second provided by our infrastructure uh to really meet the challenge of of what's coming. Um 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 uh or whether it's um other chains that are following this architecture. I mean we another example of this is Layer Zero with their their recent announcement of Zero the Zero blockchain, which has the same property. I guess Canton also has different zones um all of which can operate independently from one another. Tempo and Arc has similar similar visions for how they will scale down the line. Um that is just inevitable in my mind. You in order to scale a computer system, you need parallelism and you need hierarchy and a single monolithic chain can't can't really scale indefinitely. It'll cap out at some point.

Yeah. Yeah. Um yeah, and I I agree with you that even though I just said what I said, um we're still so early that yeah, the the competition is is definitely not going to be decided for a while. Um okay, let's talk about the AI future because your background was in AI. So I'm just I'm just going to ask you the most open-ended question about this cuz I just want to know like what you're thinking as you're watching, you know, kind of your your past um interest in in line of work take off with crypto. Yeah, I I'd love your top-line thoughts.

Well, I 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 live 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. Um and it's incredible, too uh just how much you can do with AI tools today. I mean, I'm I'm just kind of reflecting as to my productivity level as a full-time software engineer at Google for 3 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 in in nights just nights and weekends only up it's very part-time. It's a couple of hours every week and I think that I'm 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 um agents today are very single player in that they are very much a tool that a human uses to get a job done. Um 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 also some 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 that 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 is the perfect fit from a technological standpoint for AI agents as a way of bringing agents into the financial world. Um, so it's exciting to see all of that play out. There are many companies uh that are running after this opportunity, some of which already are already in our portfolio. Um so couldn't be more excited about that whole trend. The other big one, too, is is the way in which crypto can help uh combat the the big problem uh of AI deep fakes and and kind of helping us reason

about the provenance of media and information. Crypto is is purpose-built for that problem. Um and so in in that way it it crypto and AI are also really good counterweights for one another. Oh, okay. Can you talk a little bit more about that? Like, you know, does that mean that something like a Worldcoin you feel like is kind of going to be a crucial piece of this new world that we're moving to? Yeah, I think that increasingly it's going to become very difficult to know whether the people you're interacting with online are actually people.

Um and I think this is already an issue in places like sort of say dating dating websites or just general social networks. And it's only going to become worse as as the AI models become better and as people realize that actually it's quite easy to to pretend to be a human um now now that you have uh as powerful a model as as people have today. Um 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 built a technology that allows you to prove uh humanity, 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 and having it be stored anywhere. If you have you guys have seen the orb the World or 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. Yeah. Yeah, I think um 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 um some very successful OnlyFans account, and it shows Yeah. the guy making the video, and then the AI image

of the woman, and he made, you know, um healthy five figures in the first month. Yeah. Um well, okay. So, I did want to ask, you know, like we are in this era, as I mentioned even just earlier on the show before we were talking about AI agents, about how we're this moment where clearly, you know, every uh every entity in finance is converging on the same turf, and part of that competition we're seeing is in um these different payment protocols for agents. And um 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. XRO2 uh by Coinbase is well known. Um AP2 by Google. There's MVP. Um 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. Yeah, I mean, I largely largely, especially MVP and SRO2, they're essentially the same thing. They're they're very similar. Um I I think that 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. Uh 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 ends up getting.

Uh and ultimately, the thing that ends up getting the most traction most quickly will begin to accrue network effects and then that that'll just become that will 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 the standard becomes standard, it becomes very hard to dislodge simply because and if you want to interoperate with everyone who's already using the standard, that's the only option you have. Okay, so that is more like a go-to-market competition you're saying.

Yeah, I think driven by product, driven by go-to-market. Whoever wins uh at actually gaining adoption will have the privilege of defining the standard. Okay. So, um I also wanted to hear you talk about when you think about what a mature agentic AI economy will look like, you know, whether that's, you know, a few years from now or even 5 or 10 years from now, you know, describe, you know, what our everyday lives will look like. Yeah, I actually really liked uh John Collison's um five layers or five levels of autonomy that he talked about in in the in the annual annual Stripe letter. I have my

own levels of autonomy that are sort of somewhat simplified um and for me the three there's three levels of autonomy in my mind. There's number one is uh having AI completely eliminate all of the tedium that's involved in buying things online. Uh and that today is 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 of these things and that's the process by which by which we still buy things online in 2026. That all that can go away. It could just be entirely driven uh or automated by AI and that's maybe the first layer of autonomy.

The second layer would be one of um 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 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." Uh and then here's the kind of the list of of overall things that need to be bought and then the agent just goes forth and and does it does it all for you.

Um and then the third level is is what I would call more like agentic finance where 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, um it can buy its own compute, it can buy data uh in order to sustain itself and it can become a a sort of 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. Um 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 and that if you ask

even ChatGPT 5.5 or um even Mythos uh or or Opus 4.7 to go forth and generate value and uh not run out of money, it probably does not work. They probably will get prompt injected and ultimately run out of money. But I think given the exponential trend that we're on, give it another year, give it another 2 years, it may be very possible that you have essentially companies whose founder is an AI um as opposed to necessarily a human. Yeah. Yeah, no, I I definitely feel that way. Um but actually that leads me to my next question because this is something I've been thinking about a lot. Um So obviously we, you know, you saw what

happened with that Citronelle article. Like there are these panics happening about how AI will affect jobs in particular white collar jobs. And I'm a big fan of how A16Z for years now has been talking about how blockchain technology will change the way creators earn on the internet. And when like Chris Dixon and his seed Karachi had their little, you know, disagreement on X earlier this year about whether crypto is only for financial applications or whether it can be used for these other things you know, to upend creator business models and things like that. I was team Chris.

>> [laughter] >> 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 and for anybody who didn't go to Stanford you could probably figure out what those means. 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 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, um, certain types of software / IT type jobs, and certain types of financial jobs like accountants, um, you know, transcriptionist would be another. Like 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, you know, when you think about sort of those two um kind of trends that are intersecting. Like let's say you guys have had the one theory for the long time about how crypto and blockchain can 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 uh are a little bit uh 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.

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 in most cases I think it will actually lead to more employment rather than less as is the case with after engineering and we can talk about why. I think as for the 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 to make that happen, to make that happen for themselves and to reinvent themselves and teach them to teach themselves new skills and new new uh 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 everyone of those three believe that they can do the job of the other two and they're in Mark Mark Andreessen actually referred to this as the Mexican standoff between the three. Um and it is very true and 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 judgement and design taste uh and product vision that a human at least today can only can only

provide. Um so I do believe that there will be big shifts. Uh this will just result in companies being able to do much much more than they were before. Uh productivity skyrocketing case of software engineers actually this leads to greater employment of software engineers not 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 for for every engineer that you hire. Uh 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're seeing. It's interesting to

see that actually there's a there's an uptick in hiring for software engineers as opposed to the opposite which you would you would expect if you believe that AI would just replace all software engineering. Yeah, yeah. I don't think that it will replace software engineers. I think it will replace like certain types. Um but okay, okay. So we only have 5 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 like looking at this quantum threat and they're wondering what's going to happen. I think you know 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 you know smart contract chains really also have quite a large surface area for attack 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. 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 not to say that we shouldn't be thinking about this now and planning for the transition. It is possible >> are you saying even with the recent um breakthrough that Google Oh, okay. Yes, even with that. I mean, to have a cryptographically relevant quantum

computer that can break something like SHA-256 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 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 or the algorithms that you would run on a computer once it exists. Um 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 there again, there could be bugs in the cryptographic systems that we migrate to. Uh 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 these two uh 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 makes sure that we make the transition happen safely and that we and that we are we're also thoughtful about the kind the the cryptographic schemes that we that we ultimately migrate over to.

Um So, some of the conversations on Twitter are like, this has to happen immediately. It has to happen tomorrow. Like, the quantum threat is imminent. And then there are people on the other side who are like, no, this not a non-issue at all. And I suspect is the answer the real answer is somewhere in the middle. We have to be thinking about this now. It's 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. Um and it's just a matter of being thoughtful about the balance. Okay, last question. Um prediction markets have been talked about um kind of everywhere. And one of the more recent cases um 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 um what should have happened. I have heard 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 raid 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} "insiders" which that's the more colloquial definition not like a legal definition to trade or should that be prohibited?

Well, I mean I'm not a lawyer, but there there is quite a bit of legal um detail as to how insider trading is treated. Um and it's not a think of black and white black It is not a It is a fairly black and black and white thing for a lot of these cases where for example, if the person who is trading had a duty um to a shareholder or to a third party and they're essentially breaching that duty by acting on confidential information. That's clearly illegal. That's clearly insider information. But on the other hand, if you are a hedge fund and you buy satellite data to figure out whether uh say Walmart Walmart will have higher earnings or lower 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. Um and that's exactly the kind of thing that you should be able to trade on. Um 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. Uh and I think a lot of that applies uh to prediction markets, and especially if prediction markets end to end up being regulated as as commodities markets. Um and I think uh maybe there are cases where there's a gray area, but but uh as a non-lawyer I I I wouldn't know how to exactly how to exactly think about those cases.

Okay. Well, Ali, it has been such a pleasure talking with you, and congrats again on a16z fund five. Thank you so much. Thank you for having me. And thanks to everyone for joining this livestream. We will catch you tomorrow. Bye everyone. >> [music] >> Nothing you hear on Unchained is

investment advice. This show is for informational and entertainment purposes only, and my guests and I may hold assets discussed on the show. For more disclosures, visit unchainedcrypto.com.