Unchained - Why EigenLayer May Be the Most Innovative New Ecosystem Since Ethereum - Ep. 618

Starting point is 00:00:00 You know, I'm not sure I've seen a dev ecosystem with so many new ideas come up so quickly since Ethereum itself. Where with Ethereum itself, you know, it unlocked this whole universe of kind of a multi-asset blockchain, smart contracts. Hi, everyone. Welcome to Unchained. You're a no-hype resource for all things crypto. I'm your host, Laura Shin, author of The Cryptopians. I started covering crypto eight years ago and as a senior editor of Forbes was the first Main Tree Media Reporter to cover cryptocurrency of full-time. This is the March 12th, 2024 episode of Unchained. Pocodot is the original and leading layer zero blockchain with over 2,000-plus developers. And the Pocodot 2.0 upgrade will be a massive accelerator for the ecosystem,

Starting point is 00:00:47 making it faster, more secure, and adaptable. Perfect for GameFi and Defi to build, grow, and scale. Join the community at Pocodot.network slash ecosystem slash community. Bet mode activated. The scorebed app here with trusted stats and real-time sports news. Yeah, hey, who should I take in the Boston game? Well, statistically speaking. Nah, no more statistically speaking.

Starting point is 00:01:10 I want hot takes. I want knee-jerk reactions. That's not really what I do. Is that because you don't have any knees? Or... The score bet. Trusted sports content, seamless sports betting. Download today.

Starting point is 00:01:23 19 plus, Ontario only. If you have questions or concerns about your gambling or the gambling of someone close to you, please go to Connixonterio. With Amex platinum, $400 in annual credits for travel and dining means you not only satisfy your travel bug, but your taste buds too. That's the powerful backing of Amex. Conditions apply. Local news is in decline across Canada, and this is bad news for all of us. With less local news, noise, rumors, and misinformation fill the void, and it gets harder to separate truth from fiction. That's why CBC News is putting more journalists in more places across Canada,

Starting point is 00:02:03 reporting on the ground from where you live, telling the stories that matter to all of us, because local news is big news. Choose news, not noise. CBC News. Today's topic is what Eginlayer and restaking mean for Ethereum. Here to discuss are Sri Ram Khanin, founder of Eigenlayer, and Olaf Karlsson Wee, founder and CIO of Polytane Capital.

Starting point is 00:02:26 welcome Sri Rahman Olaf. Hey, thanks for happens. I'm really excited to do this chap, predicting it's Olaf. Yeah, yeah. Olaf and I were chatting a little bit before the show, and he was saying he feels like he hasn't seen an ecosystem sprout up. Eigenlier is one of the fastest ecosystems he's seen that sprout up since Ethereum, which is, in my opinion, quite a high compliment.

Starting point is 00:02:50 So, Svri-R-R-Wen, let's start with you, and let's just make sure the audience is all up to speed on what Eigenlayer is, and what problems it solves for Ethereum. Yeah, absolutely. I didn't use a mechanism that allows any developer to tap into the trust network underlying Ethereum. What is the trust network underlying Ethereum? People lock up a portion of steak, eat, into a contract, and then promise that they're validating the Ethereum blockchain correctly.

Starting point is 00:03:17 And this is the root of trust of any approval state chain. And the reason this is trustworthy when people lock up a bunch of steak is that they are potentially liable to lose their stake is they misbehave. So even if all the stickers on Ethereum come together and try to attack the system, they know that they have a concrete parity in the form of lot of stake, which basically makes the system trustworthy. And what Igon layer does is to let developers who want to access both the stake and the operator network underlying Ethereum to be used for arbitrary use systems. Anybody who wants to build a new decentralized protocol, which requires like full flexibility in contributing and programming what the nodes in the network does,

Starting point is 00:04:00 you can do it through eigenlayer. Aigenlayer is a system of smart contracts on Ethereum, which basically allows people to build new kinds of services we call AVES, actively validated services. Actively validated service is any service that requires a decentralized group of nodes to actually do active work and validation to verify that they're doing the service is working correctly. So you can think of this as, you know, in the cloud world, you have the idea of EPS, application programming interfaces where you can kind of query an API and access the data.

Starting point is 00:04:35 Think of how do I build APIs, but with a trust and a bit decentralized trust tacking them. That's what an EBS is. That's the ecosystem that you're enabling is people coming and building new sets of actively valued sales is on top of the Ethereum trust network via EGEN. Olaf, Polly Chain was a seed investor in Eigenlayer. Why are you so excited about it? And what do you think the launch of restaking on Ethereum will mean? Yeah. So I remember this goes back to the first time I met Shuram. I think we had about 45 minutes calendared off to talk. And we ended up speaking for, I think, about three hours about the implications of EigenLayer. It instantly made a lot of sense to me where these node operators that are running staking nodes and proof of stake systems, they have these fixed costs they've already taken on, right? So if you're an Ethereum staker, you already have high connectivity, high bandwidth,

Starting point is 00:05:35 you already have capital at stake, and you're running this node. You know, incrementally staking for services built in eigenlayer basically just means clicking a button and saying, yeah, I'll stake in this, you know, AVS service. in addition to staking the main Ethereum chain. And so the incremental cost for existing Ethereum Stakers to add an eigenlayer service is extremely low. So, you know, you can think of eigenlayer in a sense as a two-sided marketplace between the application developers that need security and the Ethereum stakers or, you know, or stakers really in any proof of stake system that validate those services. So it felt like that side of the marketplace stood

Starting point is 00:06:22 to benefit. Like, it's just quite rational for them to stake and accrue additional value from what they're already earning staking Ethereum. Then when I think about, okay, what could you build if you could have the economic securities of Ethereum without having to spin up Ethereum 2, right? Which today, in order to, you know, get equivalent security of Ethereum, you literally need to launch a crypto token and have it become worth. half a trillion dollars and get that whole staking network online just for your one application. And it's actually something that I've helped lots of people do or attempt to do. And it's very hard. The idea that you could plug into a set of tools and sort of just jump on to the existing

Starting point is 00:07:11 economic security of Ethereum to add that level of extremely robust security to arbitrary application logic. That side of the marketplace, it also felt like it makes a ton of sense that application developers are going to want to build secure oracles, more flexible oracles, like a whole suite of sort of middleware services that require extremely high security guarantees and unlock new types of application logic without having to spin up their own blockchain. I've since realized it's even further than that, you know, in that eigenlayer, even if you want to spin up your own economic security system with your own native token, eigenlayer also allows you to do that, but initially bootstrap on Ethereum Stakers while your token sort of grows in value. So it's pretty remarkable what eigenlayer unlocks.

Starting point is 00:08:07 It's only expanded in my head as I've watched a real ecosystem emerge here, around this. As we were talking right before the show, Laura, I said, you know, I'm not sure I've seen a dev ecosystem with so many new ideas come up so quickly since Ethereum itself, where with Ethereum itself, you know, it unlocked this whole universe of kind of a multi-asset blockchain, smart contracts. And in those early days, we had things like crypto-collateralized stablecoins, you know, non-crypto-collateral, like offline collateral stable coins, you know, early defy contracts, early NFTs, like all those things were emerging around the Ethereum blockchain. I'm seeing very similar early, extremely promising ecosystem built around eigenlayer today.

Starting point is 00:08:56 And, you know, since the system has gone live, I mean, it's really gotten a ton of traction very quickly. You know, I think that there's over $10 billion now at stake in the eigenlayer. system, I tweeted something like a year ago, if eigenlayer works, it should absorb, you know, most Ethereum in existence. What I really meant more precisely was most Ethereum that's being staked in the staking system. I think we're well on track for that to happen because it's quite logical for both sides of this marketplace to participate. Yeah. And, you know, my understanding is only a fraction of both of yours, but the more that I've been learning about it, the more I just keep picturing,

Starting point is 00:09:38 and layer as enabling Ethereum to become this giant stack of decentralized software as service providers that all share the same security. And so it's just kind of like if you take the Web2 internet that we have now and there's all these different services that you can sign up for, it would just be like that but decentralized in on Ethereum or something. So why do we talk a little bit more about these Aveses just because they are so central here. And I was curious to hear your thoughts around, you know, how they should be designed or even like which one should be created or who decides or is it just, you know, let a thousand flowers bloom or, you know, what are your thoughts on on that? I'll give an analogy, I think just playing off the cloud analogy to start with.

Starting point is 00:10:28 The first thing is we're calling these actively validated services rather than chains. When people thought about restaking, they were thinking like, we're going to allow many people to build chains, but chain is one kind of an object, which is its own audited ledger. But really, a service is much more general. And I think the right kind of analogy is like the software and as a service on top of the cloud, right? So one way to tell you this discussion, if you go back in 1995, and if you're a web application developer, what you need to do is you have to build your own server stack, you have to build your own identity, you have to build your own payments, you to build your own database and then build whatever thing that you were building. Imagine your

Starting point is 00:11:09 Amazon building bookstore. You have to do all these other things. And in 2024, that's absolutely not how you design a web application. It'd go to EWS. It'd go plug into like one of the identity service like Oath. You'd go plug into a database service like MongoDB. You'd go plug into a payment service like Stripe and then builds like whatever website you want on top. Even that, maybe you'd do something like Shopify. So you just have like layers and layers. of hyper-specialized software services that have built, which, which, you know, is reminiscent of our own theme, open innovation or permissionless innovation in the crypto space,

Starting point is 00:11:45 the idea that anybody can come and build like a hyper-specialized focused service, and then end-user-facing applications concatenate a set of these services to provide the functionality they want to their users. So that's the illusion of the web application development stack. And we see something analogous, empowered by something. something like I can learn is once you have a framework on top of which people can build like very specialized services, these ABSs, now end user applications can concatenate a bunch of these services and then like provide really interesting user experiences. So you know when people talk about

Starting point is 00:12:22 in the modular blockchain paradigm they say maybe there are like three modules like execution consensus data availability. Our view is no there are going to be thousands of models. Hyper specialized to do like very, very specific things. And when you have like hundreds of modules, and another thing people ask, you know, in the modular ecosystem is, oh, is really an application going to pay a little bit of drip across all these modules?

Starting point is 00:12:48 Yeah, exactly. That's how cloud works. Like each web application uses 20 SaaS services in the backend and pays a drip of Cs. And we have hundreds of unicorns in the SaaS economy, maybe thousands, and not doing the math here. But, you know, it is one of the most probably venture investment sectors over the last 20 years. And the reason it is very profitable is because

Starting point is 00:13:10 it's much easier to find who is the best person to build a NoSQL database for AI than it is to find which end user application is going to win. It's much, much harder to take a bet on that because this is hyper-specialized, the best database guys building this service to offer to everybody else. And to like compare this vision with other visions of application development in the blockchain space. One other vision of application development is the app chain thesis, right, which is just integrate everything into the app chain. The things that I see happen there is, for example, a project like osmosis, which is a highly technical project, which is innovated all across the stack, for example, figured out how to do encrypted mempools as a part of like its whole

Starting point is 00:13:57 you know, osmosis development because they said like, we don't want mb, we want an encrypted maphole. But actually, encrypted mempools, is an awesome service for lots of applications, not just osmosis. So encrypted mempool should be a service running on eigenlier. Now I can just like plug and play across all these different services to get the end user experience that I want. So applications emerge as a consequence of like interlocking across various services. And so that's really what we're super excited about is actively validated services

Starting point is 00:14:30 being these hyper-specialized focused services that you can concatenate. to form rich and use applications. So that's kind of the broad vision. I can also talk a little bit about what categories of these AVSSs are we seeing, what are approximate, what are in the long term. So that's interesting. Yeah, why don't you go ahead and do that, actually, because I was going to ask you for more examples.

Starting point is 00:14:53 I love that example about the encrypted men pole. That's amazing. But, yeah, I would be curious to hear about the universe of ABSs. Absolutely. So, you know, we can put them into like four, five meta categories and inside each meta category, there's a bunch of categories. One meta category is roll-up services. So ATEM has a roll-up scaling roadmap where roll-ups are going to be a lot.

Starting point is 00:15:13 What are the services that require decentralization that can augment the rule-up scaling roadmap? So, you know, some people are confused whether Agen-Layer is like in coalition course with the roll-up roadmap or not. In fact, it is actually turbo-charging, I think, the roll-up roadmap. So one of the things we see is roll-up services that can be built on Agen-Lair. The first service we are building ourselves is eigenDA. When you're a roll-up, you want to write data to Ethereum

Starting point is 00:15:38 to make sure that, you know, anybody has access to the inputs and outputs of your computation so that they can challenge it or continue the computation on if you go away. So that's a data availability layer. So eigen-DA is a data availability layer. We built on top of Igan-Lare. Igan-DA has, you know, is launching with a throughput of 10 megabytes per second that is, you know, compared to like Ethereum's upcoming upgrade called 48444 or Dengoon is up 30 kilobytes per second.

Starting point is 00:16:06 So it's a couple of orders of magnitude out there. So that is an example. And then how Igan deal is basically take the Ethereum roadmap, darn starting KZG, you know, encoding and all that stuff that is already there. One just build it, right? Because there is much more energy and interest in private markets to fund things, fund these kinds of services than Ethereum itself can necessarily do all the development in-house. So that's an example, IGNA,

Starting point is 00:16:34 but lots of other services that we see spring up. Like another example of a roll-up service is bridging. Like, I want to bridge from one, you know, fragmentation of roll-ups is something we all talk about. I want to bridge from one L-2 to another L-2. How should I bridge? If somebody has locked up enough capital and promises that, hey, you know, I have this $10 billion of stake, and I promise that, you know, this is the roll-up state of this optimism or operatrum or ZKSing,

Starting point is 00:16:58 then I can use that to lock the state and move around, like, information across these L-2. that are much, much, you know, lower latency than even Ethereum. So, bridging using crypto economic security. The third one is MEV, right? Like, how do I handle MEV on these roll-ups? Do I need a decentralized sequencer? Do I need some threshold mempool?

Starting point is 00:17:18 What kinds of, like, modulation can I do on MEV? And this is very application-specific. There's no universal way to do MEV management. Sometimes you want first-in-first out. Sometimes you want auction. Sometimes you want threshold encryption. Sometimes you want event-driven actions like limit orders should be automatically executed. All of these are services coming up on AgonLare.

Starting point is 00:17:40 So this is something we're very excited for. Then another category is watchtowers, right? So you have these like execution environments which can have faults. You know, somebody can and, you know, but you need a group of nodes which are watching and can trigger these fault alerts. So they need to be incentivized. they need to be staked, you know, and that can happen on eigenlater. Like, that's watchtowel is a category come. Fast finality.

Starting point is 00:18:06 Like, right now, a roll-up sequencer makes a kind of pinky promise that, hey, you know, it's final. Like, you know, it will get finalized on Ethereum, but I promise that I'm going to post this to Ethereum. But instead, you could have, like, a committee which is staked, which makes that, you know, promise with an economic credibility rather than just reputational credibility, which is what sequences are doing today. So examples of categories on just. roll-up services and I would say like we didn't think of more than just the first one

Starting point is 00:18:35 I can hear ourselves the rest is permissionless innovations just coming out of like the market these kinds of ideas so this is a meta category roll-up services another category that we're seeing is what I call a co-processor a co-processor is different from a roll-up and that it's not a chain like there is nowhere you go to a block explorer and see the order of transactions off that that system a co-processor is something that is like a service that is not, you know, I'm on Ethereum. I call a service, you know, let's say I want a AI like inference protocol. The inference protocol runs somewhere and then the answer comes back with a certain amount of trust. And you know, you can get trust from

Starting point is 00:19:14 two kinds of things. One is zero knowledge proves that yes, this is executed correctly. Or it could come from crypto economic assertion that, here, I have like $500 million backing this claim. If this claim was wrong, you will be able to whoever is receiving this. claim will be able to flash and redistribute $500 million. That becomes very powerful. So Aguilier helps the second category of co-processors. We call them crypto-economic co-processes, a co-proscer that comes with a certain amount of economic integrity attached to them. And this could be, I want to do AI inference. I want to do SQL databases. I want to run arbitrary Linux programs. I want to verify ZK proves, not on chain, but off-chain. So these are the categories of things we're seeing in

Starting point is 00:19:58 co-process. I think this is one of the most interesting and exciting sectors, because for the first time, we call these smart contracts. And I think Olaf was alluding to this a little bit before when we started. And if you think about how smart are they, they're rigid contracts, really. Like, you know, when you enter into a smart contract, they're very rigid. They're not that smart. Like, this simple curve that, you know, Uly Swaps started with X, Y, equals K, right? Like, it's a very simple equation. It is not a huge amount of intelligence. It doesn't have access to the history of all the data. It's not using some complex machine learning to adjust prices. But that's how contracts have to evolve, if they have to actually bring smartness. Whether that is databases, that is

Starting point is 00:20:41 machine learning AI protocols, whether that is running general purpose computation, like a general Linux box on top of it, all of these are really needed for us to go to systems that are actually smart and rigid means if they say something that is actually right. And one way we bring rigidity to the system is that each of these claims are packed with an amount of economic security. So you know that either the claim is right or you will be able to redistribute a certain amount of money. And that gives you very strong rigidity. So as long as you don't move more economic value than what the stake is, you actually have like unconditional correctness. So that's a really powerful rigid mechanics by which you can now bring an arbitrary amount of computational intelligence

Starting point is 00:21:29 to bear on the smart contracts. So that's a second category. I'll stop before going to the other one. And let me also jump on to add a very specific example that I think is pretty compelling of how this could work. So with this, you can have smart contracts execute based on the secure output of a machine learning model. And when I say secure, I mean like Ethereum level security, right, where you run a consensus

Starting point is 00:21:59 algorithm to decide, you know, the output of this AI model, you know, we all agree that it's correct and it can be used to trigger financial logic. You can't just connect to private APIs and say, okay, you know, this smart contract is going to manage a billion dollars based on how this private API, you know, tells it to. You really need this sort of economic security where if you're managing a billion dollars in that smart contract, you need the output of a machine learning system to be validated and you need it to be secure in the way that Ethereum is secure. So a very specific way you could use this, we all know that the interface layer to, you know, talk to blockchains and smart contracts is not very good. We've all used Metamask and fumbled around trying to figure out

Starting point is 00:22:46 how to interact with a smart contract, why did it fail, this gas market, you know, like it's, it's all quite complex for your average person. Imagine instead if the end user could interact with their wallet using natural language, like a large language model. So let's say you have a bunch of assets on the, you know, optimism roll up and you want to buy an NFT on the Arbitrum Rollup. you could say, hey, you know, I want to buy this asset at this price, you know, get it done. And basically, you know, eigenlayer, the security afforded by eigenlayer allows a service to interpret that language that the user wants and actually put together a series of smart contract signatures that gets that done. Right. So you could imagine the interface layer that we've all been using for years to interact with blockchain systems could,

Starting point is 00:23:44 in theory, be reduced to natural language models. So the way you would interact with smart contracts is the way you interact with chat GPT, where you just in normal language, layman's terms, say, I'm interested in buying this asset or making this trade. And in the background, there might be five or six different signatures that need to happen or attestations. And we can get, you know, very, very high economic security that those things are happening in the correct way that the user needs.

Starting point is 00:24:14 intended. So the potential here is, you know, this is just one example, but I think just this example is a really big deal. It's sort of replacing the entire way that users interact with these systems with natural language models. And this is, this to me would be a huge breakthrough in usability for your average person who understands what it means to type, you know, hey, I want to buy this asset and I want to use the Bitcoin in my wallet. And the asset, and the asset might be on Ethereum and you might have to then bridge your Bitcoin to Ethereum, sell it for ether, you know, go it onto a roll-up and then go buy the asset, right? And the large language model, you know, can interpret your instructions and come up with a series of smart contract

Starting point is 00:24:58 signatures needed to execute that and have it all validated by one of these eigenlayer ABS services. So that's just one example that I think would be like a, you know, seismic shift in the way people interact with blockchains and genuinely be like an order of magnitude improvement for the whole ecosystem. Yeah. And what I find what I like about that is that it addresses the issue of, you know, like a lot of people, if they're not in crypto, they're not even going to know the difference between arbitram and optimism. They're not going to understand why can't you use your Bitcoin to buy that NFT that's an optimist. Like they won't even, you know, so hopefully in that future, people won't even know like where things are located or whatnot. But I did want to ask because, you know, there's a lot of

Starting point is 00:25:42 talk about how now with all these different roll-ups, there's also a lot of fragmentation. So do you guys see this as a way to address that problem? Yeah, I think I was mentioning earlier that bridging across these rollobs would be done by some kind of economically secured bridges. And we're seeing many, many of those bridges come up. And what I think Oloff's adding here is those bridges could be triggered from natural language. Right. and somebody says, hey, do this, you know, for me.

Starting point is 00:26:11 And then, like, those instructions are interpreted in a rigid way that, you know, you're not just trusting some third party, but actually having consensus on, like, you know, the output relationships and then actually have that happen. So I think in some sense, the way I think about it is these will be different services. And then there is a kind of like a service that triggers other services, a U.S. service, which could be this NLP models, which will then. like go and trigger these. Now trigger the bridging, trigger this action, you know, that'd be like super fun to see. And LPS in like natural language processing or something?

Starting point is 00:26:47 Yeah, and LPS national language. Exactly. And then you, I think you had like another category you were going to talk about. Yeah, we have three more categories to talk about like. Okay. Typography isn't a category. You want to bring like more advanced cryptography into blockchains and why? Like because right now, blockchains are either verifiable and transparent or you know, but don't allow for privacy. So you want rich privacy expressions in blockchain. And while also preserving very fabability, you need to have more complex cryptography, like secure multi-party computation where you're distributing a secret across many, many nodes,

Starting point is 00:27:25 and they all do computation, portions of the computation, and arrive at the answer without no one actually having access to the actual private information. You can have trusted execution environments, which are special hardware zones, which are built with cryptographic integrity by Intel and so on. And you can actually require that validators or not operators actually have these networks of trusted execution environments, which augments which augments other kinds of trust. You can have fully homomorphic encryption where you encrypt the data and then all the computation is done inside the encrypted space. So again, like amazing breakthrough technology.

Starting point is 00:28:11 These kinds of things we are seeing many, many projects building on idyllar with all of these types of fundamental technologies. And they used to exist in some way or the other in other chains. For example, Oasis was a trusted execution environment chain on cosmos. The power of these services becomes much, much larger when they become composable across all the other kinds of things we are talking about. So people, you know, we've gone, the average of the Agile ecosystem has been really interesting to watch where we are already seeing people build an AVS that will serve other AVSs or are only possible because other AVS will start

Starting point is 00:28:52 interacting with them. And, you know, it's, of course, all of these are simultaneously or co-emergent, but it's exciting to see that because now, oh, I can build a service that will help these other services and now if I put these services together, it's a whole, it is something that will actually like, you know, 10x the infra landscape of crypto. So this is another category, cryptography. I would say another category is proofs, you know, broadly laid out. Normally we think of proofs as zero knowledge proofs or computation proves, but there are other kinds of proves you need to assertions that are verified, right? For example, right now I'm doing this interview from the University of Washington and Seattle. And how do I prove to a block

Starting point is 00:29:34 that I'm actually in Seattle. This is called a proof of location. Until now, we talk a lot about blockchain being decentralized and geography is a core aspect of decentralization, whether it's regulation or other reasons. But how do we measure a decentralization from a blockchain?

Starting point is 00:29:51 And, you know, it's shocking to know, like there is no locus of decentralization today. The first server is measuring decentralizations coming up on eigener called Witness Chain, which is a proof of location protocol. which basically what it does is it has a decentralized network, already the large decentralized network underpinning Ethereum, they send network pings back and forth to verify that you're close enough to like other nodes, which are also in the same zone and concordance among the latencies

Starting point is 00:30:19 from all the different measurement zones to actually establish, okay, yes, you know, Sri Ram is right now in Seattle. And this is a really powerful service. So as an example of approves, another example of a proof would be a proof of missionhood. Like, I want to prove to you that I have a unique iPhone device, right? It's not the same as any other device. And these devices come up with this route certification. Like, these can be verified on, like, on chain and then, like, present it to other services. So this is another example of, like, proof of Mesh and Hurt.

Starting point is 00:30:54 A third example would be, I have, like, some private data which is locked across many servers. Imagine, like, Olaf's building, like, a decentralized marketplace, which wants to compete with Amazon. And he wants to offer a discount for anybody who has used, you know, at least $10,000 a month over the last 12 months on Amazon. Like, that's the customer that he wants to attract. How do I verify to him that that's actually the case, that I have made these transactions with Amazon. I can log into the Amazon Web server and I have like a, you know, encrypted link with Amazon using this thing called TLS to do that. But how do I certify this to a decentralized network? Can I, you know, there are these like really brilliant

Starting point is 00:31:34 schemes called tierless notaries coming up where, you know, when I opened my encrypted connection to Amazon, some node sits in the middle that I give authority to that kind of like verifies the encrypted traffic and then later I can do a proof against that traffic that actually I had like $1,000 of balance last month on Amazon. Another example of a really cool service coming up on Ionle. So this is another category of proofs. And finally, a category which is very special and unique to restaking an ideal layer is all these other services that I mentioned could have been built by anybody staking enough money and having enough node operators. The last category is what we call Ethereum Inclusion Guarantee, which is, it's about

Starting point is 00:32:18 particularly the Ethereum L1 blockchain. I want to modulate, you know, when you're validating the Ethereum L1 blockchain, you have to make sure to the transactions you put in are all valid. But there is nothing that decides what transactions you put in and what transactions you don't. And there's no way for you to make assertions or promises about specifically what transactions you do put in. For example, if Ethereum had this ability to a validator could say, hey, my blocks coming up in three slots. And, you know, if you give me an additional fee, I'll promise you that your L2, like blob or whatever thing, will absolutely be intruded three blocks down the line. I'll make sure of that.

Starting point is 00:32:57 And I can make a promise like that about an Ethereum L1 block. So this is what we call L1 inclusion guarantees or Ethereum inclusion guarantees. Lots of interesting applications. Imagine I want to build a stable coin liquidation protocol. I have to make sure that liquidations clear as fast as possible. But if the block proposals are already promised to take liquidations and enforce it, otherwise they'll get slashed. You will have a high, rigid guarantee that actually those things will happen.

Starting point is 00:33:24 So lots of interesting applications. to modulate Ethereum L1 itself. So that's another kind of meta-calidate. So these are all the types of things we are seeing. It's a kind of explosion of ideas. We're so excited about it. Yeah, it's super exciting. There's just, yeah, as we've discussed,

Starting point is 00:33:41 a lot of innovation that's happening. So in a moment, we're going to talk about potential risks, but first a quick word from the sponsors who make this show possible. Pocod is the original and largest layer zero blockchain with over 2,000-plus developers. and the anticipated Pocodot 2.0 upgrade will be a massive accelerator for the ecosystem, upgrading the infrastructure with eight times higher transaction throughput and twice as fast block times, perfectly tailored core time for the needs of every protocol,

Starting point is 00:34:09 trustless bridges internally and into Ethereum, Cosmos, Near, finance smart chain, and revised tokenomics and the implementation of a token burn to reduce inflation. Perfect for GameFi and DeFi to build, grow, and scale with one of the most active crypto communities in this space. Pocodot recently announced a partnership with mythical games, bringing top games like NFT rivals with over 650,000 players and 43 million transactions to pave the way for GameFi and the Pocod ecosystem. Get your Web3 ideas to market fast with economics that work for you.

Starting point is 00:34:43 Think big, bills bigger with Pocod. Join the community at Pocodot.network slash ecosystem slash community. When McDonald's partnered with Franks Redheart, They said they could put that shit on everything. So that's exactly what McDonald's did. They put it on your McChrispy. They put it in your hot honey macnuckets dip. They even put it in the creamy garlic sauce on your McMuffin.

Starting point is 00:35:06 The McDonald's Franks Red Hot menu. They put that shit on everything. Breakfast available until 11 a.m. At participating Canadian restaurants for a limited time. Franks Redhot is a registered trademark of the French's food company LLC. Investing is all about the future. So what do you think is going to have? Bitcoin is sort of inevitable at this point.

Starting point is 00:35:26 I think it would come down to precious metals. I hope we don't go cashless. I would say land is a safe investment. Technology companies. Solar energy. Robotic pollinators might be a thing. A wrestler to face a robot? That will have to happen.

Starting point is 00:35:41 So whatever you think is going to happen in the future, you can invest in it at WealthSimple. Start now at WealthSimple.com. Amazon presents Jamal versus the Shih Tshitsu. Descending from the gray wolf, shih Tzu's live by their own untamed primal code of not giving a single shitsu. But Jamal shopped on Amazon and bought dog treats, chew toys, and 32 ounces of carpet cleaner. Hey, Jamal, you've been promoted to pack leader. Save the everyday with deals from Amazon.

Starting point is 00:36:19 Back to my conversation with Sri Rama and Olaf. So, you know, this has come up time and again. But one of the potential risks that people commonly call out is possibly re-hypothecation. More, I guess, you know, it really relates to the security and then these like cascading risk that can happen due to the restaking. But interestingly, I even see people saying things like this crypto analyst, Miles Deutsche, said, quote, I see restaking as the next version of the DFI's. So, you know, I'd be interested from both of you, what your response is.

Starting point is 00:36:51 to these concerns. I'm sort of assuming that the underlying criticism there is that basically note operators now are staking in, you know, potentially hundreds of different proof of stake services simultaneously. And, you know, in theory, there could be slashing risk across all 100 services. Again, in theory, because each service decides its own rules. It's an open marketplace, right? So as a staker, you get to decide what to opt into. And likely, I think the end state of this is that staking nodes select the types of risks

Starting point is 00:37:31 they're willing to take on for AVS's and sort of define those risks very precisely and sort of auto-join every AVS that lands within that risk profile. That I think is the kind of the end state of this marketplace. you know, I do think there is risk every time, you know, a staking node is staking on a new system. I also, though, think that there's a lot of work being done on sort of preventing slashing from operator error. So, you know, slashing is in proof of stake systems is meant to deter bad behavior. and when I say bad, I mean somebody actively trying to hurt the system and sort of attack the consensus protocol and come up with an output that's different than what it should be.

Starting point is 00:38:24 Now, that, interestingly, though, the vast majority of slashing, and I would say like 99.9% of cases where somebody gets slashed are, in fact, the intentionality is not really attacking the system. It's operator error of some kind, right? they go offline or their software is misconfigured or something like that. I do think there's a kind of big push that it's going to be accelerated because of eigenlayer to get software in the hands of node operators that can basically detect if they are going to run the software as intended and can better tell the sort of quote intentionality behind those staking nodes. I do think this is going to be a very important part of like the growth of the eigenlayer ecosystem

Starting point is 00:39:13 is staking nodes feeling confident that they can stake on arbitrary Aveses that they've never heard of without risk that there's going to be an attack in that system where all of a sudden they're out of consensus and therefore slashed, even though they had no intention of hurting that mechanism or like being out of consensus in that mechanism. I really think it's quite different than just endless re-hypothication, right? So, you know, like, okay, I'm going to stake ether, receive a bunch of tokens that represent that staked ether, and then stake that, receive a bunch of tokens that represent that, you know, then you're at a derivative of a derivative, and then I'm going to stake that. You know, it's not really how this works. You're not sort of re-hypothicating a derivative over, and over and over and over. You're sort of just re-hypothicating that core collateral, you know, over and over. It's sort of like, you know, I put up my house for a mortgage and take a loan. It's not like I then take that loan to buy a new house with a new mortgage, and I take that loan to buy a new house with a new mortgage, and I'm basically leveraged now. So I think that

Starting point is 00:40:27 in that particular quote, you said, Laura, it's more about re-hypothecation leading to leverage, right, and that being at systemic risk, I don't view this as that at all. I think that there is risk, like I described, with staking nodes like staking in, you know, 100 different systems. But I think the way to mitigate it, it's just a very different type of risk than like a derivative of a derivative of a derivative adding leverage to a financial contract. One question before, and maybe Sruwami, you can answer this is just in that scenario, like if you do get slashed, it sort of feels like, well, so then I don't really know what happens. Like, let's say you are staked in, you know, 10 different things and you get slashed. Then how does it affect the security of the other systems

Starting point is 00:41:12 that you're staked in? And then the other thing that I asked, I wanted to know about that is if it's like every ABS is setting its own standards, then is the security of the overall system determined by like kind of the lowest common denominator ABS? Like, yeah, these are questions that came up for me. Yeah, no, this is useful questions. One way to start from where, OLAF, left off, which is I think the type of risk, why is it fundamentally different is flashing risk is endogenous. You control it as a validator. If you're not bad, you will not get slashed. As long as the system holds that property, that if you're not malicious, you will never get slashed. That's a technical problem to solve. How do you build a system which is actually the risk? This is the same problem

Starting point is 00:41:54 for Ethereum itself, or any proof of stake system. How do you make sure that you're not getting slashed for just running the get node? Like, this was a big thing that, you know, the whole community was discussing, like a few months back. This is a very important problem, but it's a technical problem. So fundamentally, it's a different nature. And it's, the way to think about it is slashing is an endogenous. You control that you will get slashed or not. Because you're opting into these objective smart contracts on these various, you know,

Starting point is 00:42:25 aviases, which specify what the slashing condition is. So you control whether you'll get slashed or not. as opposed to, if you actually take 100x leverage instead of opting into 100 services, you're taking exogenous risk, something you don't control. Like, the market price moves 1% and then you will get liquidated.

Starting point is 00:42:44 That's absolutely not the case on Agenle. Igenlead is a validation platform and it's only meant to slash you if you're actively malicious. Why is there non-zero yield in this system? It's due to an information asymmetry. If I'm a validator, I know I'm honest, so I'm not going to get slashed.

Starting point is 00:43:00 You don't know that I'm honest. So you're paying for this, you know, covering these, eacemetry of this information. But I know I'm honest, so I'm actually willing to back it with the portion of capital. Now, going back to, like, Nora's question, like, how do AVS do know, like, how much, you know, security or whatever they're getting? I think the right way to think about it, re-staking became a popular word, and, like, it's a very, it invokes a certain D-Gen, like, D-5-type feeling to it, and it sounds like we have

Starting point is 00:43:29 So people like conflate that. But really, I can layer is Shad security. And Shad security is one way to think about it is, imagine there are 100 protocols. And each of which can afford for the fee that they're paying a $100 million of stake. Okay, 100 protocols, each of which can offer $100 million of state. This is one world where like they're all separate. They have separate pools of stake. They're maintaining and then they have these protocols.

Starting point is 00:43:54 The second world is all of them pulled their stake together into a $10 billion pool. which is staked across 100 services. Now, the power restaked across 100 services, I argue that the second world is actually really much better when the reason is to attack any one protocol. Now, it's the same fee across the two worlds because, you know, the same amount of stake and same amount of fees.

Starting point is 00:44:20 But in the second world, to attack any one service, you need $10 billion of capital and you know that you will lose $5 billion of it. So there is a kind of like hard, hardening of security at scale. Like as you get more and more security bundled together, you're just like much stronger. It's the same reason why nation states have armies,

Starting point is 00:44:38 not city states, and even across nation states, you pull together to form alliances. So that's the spirit. But one thing you lose when you're aggregating security like that is you're losing attributable. Because maybe in the first world, one of them is Coinbase.

Starting point is 00:44:55 And Coinbase says, you know, these other services maybe have 100 million stake, but I have 3 billion stake because I'm on Coinbase and I'm running bits. Now, this was missing in what we call the first simple model of security, which is pool security, where everybody is just pulling together. There's no attribution. So we solve this by doing more complex accounting on eigenler. We have something called attributable security. Attributable security is you have this common pool. Everybody's validating everything together, but among this pool, I'm going to give a given service an attribution which says that no matter how many other services

Starting point is 00:45:34 triggers slashing simultaneously, you will be able to, this you as an ABS will be able to slash and redistribute this much amount of money. There's an auction which actually sells these rights. So this means in the second world as Coinbase, you could still take out $3 billion of economic security while attributable economic security while still getting $10 billion of pool security. So you're simultaneously getting the benefit of attribution while also getting the benefit of pooling. So this solves this problem kind of completely because now it doesn't matter what's happening to these other services and it's their problem to deal with. I have the eigenlayer like interface to an ABS is you tell me how much attributable security you want and I'll give you that and I can layer

Starting point is 00:46:22 to now maintain, do much more work, to maintain this, what we call proof of solvency. It has to be solvent if all the EBSS simultaneously draw their claims, but that is what we maintain. Eigerlil remains solvent even if all of the attributions are simultaneously requested. So you're now completely buffered across all these services requesting, you know, attributions separately. So this is a kind of like a breakthrough that we made in the last six months on how actually eigenlayer attributions work. So the other thing that builds on top of this is now imagine you're building a bridge. A bridge may require anywhere between let's say $10 million VT volume to $1 billion

Starting point is 00:47:04 VT volume. You know, it's crypto, there is a hundred X variance between a bull market and a par market and a good week and a pat week. So, you know, say you may have like a variance which is from 10 million to one billion. So if you were just on your own, you would have to provision for the most demand that you would experience or miss out on that demand. If you only provision security for 100 million, like any week that there is volume more than 100 million, you're out. Like, you don't have security to be with it. In eigenlayer, in the pool security model, you can buy attribution on the go. Like you'd say, man, I'm getting more demand. So I'm going to actually go and pull more from the pool security by more attribution claims. I can actually now have no

Starting point is 00:47:48 1 billion for this week and only 10 million for the next week. So this is, you know, we call this elastic scaling of security, which is exactly what like for compute Amazon did is elastic scaling of compute, which is what the EC2 is elastic cloud computing. And this is elastic scaling of security, you get as much security as you want. And by having a common pool, you can kind of smooth out all the random variations because, you know, when BTC is going up, Saul is not. And when ether's going up, something else is not, but having a common pool means you can average out across all these fluctuations. Each one buys the amount of security that they want. That's a really, really interesting dynamic. There are other like further more complex dynamics when imagine, you know,

Starting point is 00:48:34 there is an application like Farcaster using, let's say, five ABSs. You know, it uses an Oracle, it uses the data availability. It uses like AI co-processing. And a forecaster is doing some kind of weekly volume of $1 billion. If it had these as separate pools of security, it has to go and buy, you know, this attributable security of $1 billion separately from each of these services. Because any one of those services, the Oracle, the data availability, the AI, any one of those services screw up, their money could be lost. So now, you know, in Eigerlir, because there is a lot of shards staking, the same group of nodes are sticking across many services, you only need to buy 1x of that attributable security and say that any one of these services fails, as long as you can

Starting point is 00:49:24 assure me that any one of those secure services fails, you will pay me $1 billion. Now you're suddenly like saving your insurance or attribution fee by 5x as you're bundling many, many services. So there is an economy of scale of attribution that is like elastic scaling of security, that is the power of pool security. All of these are like foundational features that we've thought about when you're building the first universal crypto economic system of security. So instead of, I think a lot of people have this view that it's some kind of like a DGEN platform, it's the exact opposite of being a pure DGEN platform.

Starting point is 00:49:59 It is the sharpest system of economic security available today. Well, you know, that's going to be disappointing news to the DGens who are following your points program and probably eagerly anticipating an airdrop. So, you know, I got to ask that. like will Eigenlayer do a token? You guys have a points program. So just curious to hear, you know, what your thoughts were when you went to design it. Like, why do you have a points program? You know, why did you design it the way you did? And what are you thinking when it comes to a token?

Starting point is 00:50:30 And either of you can answer because I know, OLAF, you're an investor. I'm sure you guys discuss these things. I'll take this and let Olaf add something on top. The points represent the amount of eat you have state and how many hours. or how many weeks. So it's basically eight hours is the kind of fundamental unit of, it's just a measure of participation in the eigenlared system. And unlike many other like point systems which are arbitrary, like it's just a very simple, clear measurement of, you know, how much staking contribution you've done into the protocol. So that's, that's the eigenlare points. There's nothing mysterious about how these points are created and are allotted. It's a simple

Starting point is 00:51:14 measurement. And while we have the points program, you know, as we build this protocol, you know, when services come in and they want to demand, you know, stakers and operators, they want people who have committed, you know, are committed deeply into the ecosystem. So that's a simple measurement, for example, AVS is. If there's over-provisioning of security, they might choose only those who were around in the system for long enough. There's a simple measurement for it. And as we aim to decentralize the protocol in the longer run, you know, obviously there is governance rights that is accorded to people who are actually like, you know, participated in the protocol. And sticking is one element of participation in the protocol. So obviously there will be, you know, stakeholders in the

Starting point is 00:52:01 governance of the protocol as we go on. But, you know, we have not thought far enough to actually talk about it. Oh, really? Interesting. Olaf to you. Yeah, because we'll actually, let's this in there as well. I'm sure you know Kelpdow recently began allowing people to take their eigenlayer points and get liquidity out of them. So in a way, that is kind of right now the closest thing that we have to an eigenlayer token. So, you know, I was kind of curious for your thoughts on that. And yeah, like one of the things I like to actively discourage is people taking leverage on top of these LRTs. You know, I said there's no leverage or, you know, in the Iger layer protocol. But you actually can take leverage outside of the eigenerlare protocol, go and, like, you know, deposit your LRT into like a lending pool and then borrow against it and then like loop this several times.

Starting point is 00:52:55 Assuming. Right. Liquid restaking token for people who, yeah, are wondering what that stands for LRT. Yeah. And, you know, there are many protocols like Kelp and Etherfi and Puffer and Renzho and Rio, you know, which are building these kinds of protocols, which are very, very useful actually. because they outsource management of how do I choose operators, how do I choose aviases, how do I do risk management, like, you know, have a consistent decentralized governance protocol.

Starting point is 00:53:24 Each of them have their own things, so you should do your own research before engaging with any of them. They are external to eigenler, but, you know, it is a useful service. But while doing it, people are now, you know, the Dgen version of it is to actually go and take leverage on top of it, which is you take a liquid restaking token, which is, for it represents your ETH deposited in the protocol, and then, like, you know, go and borrow against it, ETH, and then go and do this again and again. And if EIT to the liquid restaking token moves like from 1 is to 1 to like 90 cents to 1, they'll get liquidated in and lose their capital.

Starting point is 00:54:00 Like, you know, Laura, you were asking me earlier, like, what do you think about people doing, you know, expecting this to be like a new DFI, you know, yield? And, you know, is there any, Ponzi nature to it. I can layer it is designed to have zero principal risk. If you actually staked and the operators ran correctly, you don't take any principal risk. But, you know, the way our space is somebody figures out how to take principles is we're actually going and taking leverage outside the protocol

Starting point is 00:54:31 and going and doing this thing. This can be very dangerous. You know, this is a very, very risky thing to do because there's all sorts of reasons why like a liquid restaked heat will not be worth 1-Eath in the short term, even if it becomes equal to 1-Eth in the long term, there is liquidity crunches.

Starting point is 00:54:50 There is, you know, let's say there's a hack somewhere in their code, which they can even like have an upgrade council and upgrade, but who knows, that takes a week. And within that, the thing will defect all kinds of reasons why like a liquid restaking token will not

Starting point is 00:55:06 be equal to 1-8 in the short term. So doing anything with this leverage on this is extremely risky. And when I'm saying AgenLair doesn't have leverage, it's in the AgerLair protocol when you stick across multiple services. You know, it's the only risk is the operator risk that OLAF already are loaded to. But people are doing this stuff. This is very risky. We advise people not to do it. In terms of like, you know, just a representation for rewards, the points, it is not something we are encouraging, but by itself, it doesn't create any leverage.

Starting point is 00:55:40 or anything crazy in the protocol. It is, of course, these points may be worth nothing. That is the risk that you're taking when you're going and buying a point. But that's a different kind of risk than actually taking like these kinds of leverage risk, which you're thinking that you're not taking any risk and actually you make it lose all your capital. Yeah, the only other thing I'll add is I know that sometimes in our space, people get confused about product market fit and sort of economic incentives. So they, you know, like, let's say you're building a social platform and it's post to earn, right?

Starting point is 00:56:23 Like, you might see a lot of people posting and you might think you have product market fit, right? But in reality, you've added an economic incentive that is perhaps not sustainable or like short-term oriented, where you in fact don't have product market fit. And I think it's very important to understand that a lot of these incentive models can increase the rate of growth of something that already has product market fit. This is like Uber saying, get your first ride free, right? Uber has product market fit. That first free ride is not the only ride people take on Uber. After that, they like continue to use Uber because they like the service. I do think sometimes in crypto I see like first ride free and then it's confusing to people

Starting point is 00:57:20 whether like anyone wants the second ride or not. Now you can, so you can subsidize through these sort of incentive systems the growth of things that already have product market fit. However, these subsidies do not replace product market fit. And the one thing I'll say about eigenlayer is it's very, very clear that there's product market fit without any sort of subsidy system. Whereas like in a post to earn, you know, social protocol like, you know, you might see the behavior that's more like, say, an axi infinity, right?

Starting point is 00:57:57 where it's like, wait a second, are we subsidizing the growth of players in an awesome video game community? Or are people just kind of grabbing that first free ride and the moment their free ride is over, like they leave the system? Right. So it's very clear to me that eigenlayer is not reliant on any sort of incentive outside of just the usefulness of the service. So one quick question that I did want to ask is you kind of hinted, I feel like it was Sri Romp. Maybe you sort of hinted that potentially the points at some point would translate into being used for governance. Did you say that? Yeah, it could normally be used by like AVSers to choose like what stakeholders to opt in for a given service. It could be used as a mesh of participation in governance.

Starting point is 00:58:47 We have, we don't have like a concrete road map on exactly all the ways that this could be used. So I don't know if you remember back in the day people got concerned that Lido's dominance could lead the LDO token to kind of be quasi-governance token for Ethereum. So do you think a similar situation could happen for an eigenlayer token? Yeah. So I think there is a fundamental difference between something like Lido and something like IgerN. Lido decides who the operator set is that actually can run Ethereum nodes. So imagine 100% of Ethereum stake went through Lyon. Then the Lido operator set is identical to the Ethereum operator set.

Starting point is 00:59:28 eigenlator has no such, you know, subjective judgment. Igonleger just says, anybody who wants to stick and be an operator can be an operator. This is a very conscious decision that we took, that the Agonilor Protocol only, like, makes objective judgments and not subjective, you know, assignments to operators. So that is, you know, a fundamental difference. if 100% of each stake went through eigenlid, you don't have the same problem that, like, Lido has, which is that they are deciding the operator set, whereas Igelat is not designing the operator set. An operator set decision is critical to, you know, decentralization, censorship resistance,

Starting point is 01:00:07 in other properties of the Ethereum protocol. And it is, you know, it's one of those things where the Ethereum protocol is not yet complete. And I say it's not yet complete because, you know, the protocol needs censorship resistance. but has no way of enforcing it inside the protocol. It is reliant on like, you know, us shouting on Twitter and like just canvassing people to actually like get these properties. But I would say like of that kind of a spectrum, Ethereum is the most sophisticated in the sense that

Starting point is 01:00:38 it is actually not reliant on the operator's set for safety at all. Because if you violate safety, you will get slashed. It is written in protocol. It's as legit as it gets. So we're not trusting the operator set to maintain, safety, which means like don't double sign a block or don't sign an invalid block. You don't have to go on. Nobody comes on Twitter and requests this because you know if you do it, you're going to lose your money. So nobody's going to venture to do it. But there are some aspects

Starting point is 01:01:05 which have not been internalized yet, you know, because this is a, you know, new emerging area. And, you know, we have a new proposal for this called Rivier, which is a mechanism to penalize for censorship in Ethereum itself. But, you know, these are the things that we are, actually like we don't like this social herding that is necessary to actually make the theorem protocol work. Because, you know, is it a protocol or is it like some, you know, philosophy that we all have to buy to actually, you know, even though I'm like very aligned on the philosophy, but we don't build a system assuming the participants are going to subscribe to a philosophy.

Starting point is 01:01:42 Like, that's not a rigid system, a rigid system enforces its roles internally and self-consistently that I don't need to know this. So the way we think about it is first is like Aguilir does not impose any objective, subjective considerations on the operators and, you know, operator side, which means like we don't have the same problems that something like Lido has, but also recognizing that Ethereum itself can actually internalize its, you know, constraints much better and like actually proposing. I gave a talk on this protocol called Rivier in the censorship dot WTF workshop. but, you know, how do we make Ethereum better? So that's our two answers to that question. In fact, one of our stated goals is the internalization of the eigenlayer protocol into Ethereum. Yes, okay. You know, eventually, Igelayer should be like a core part of the Ethereum protocol.

Starting point is 01:02:35 It doesn't need to be something external. Of course, you know, there's a lot of details that we are figuring out, like all the things I mentioned today about attributable security, how do you account for this correctly? All of these need to be ironed out in the open kind of market. where, you know, it's like an arm's length from Ethereum where these experiments are on. And eventually, as the results are solid and rigid, this is one of the reasons why many eigenlayers, like, oh, it's not really, really staking or anything, like I think Olaf mentioned

Starting point is 01:03:02 briefly before. The reason is you need a single place which has the full accounting. That's the right architecture. And right now, like, eigenlayer has full accounting. Igen layer has minimal information interfaces to Ethereum because Ethereum is not built with the idea. that something like Agenlare will exist. And now, like, we're trying to kind of, like, get Ethereum to kind of co-evolve, or maybe even just internalize it, because that can be, like, a single place

Starting point is 01:03:28 of accounting for everything. So that's our long-term stated goal that something like Agen-Lor, if Agen-Lair becomes something like enshrined. I'll be super interested to follow that. So with our time remaining, I definitely want to talk about AI because I noticed, you know,

Starting point is 01:03:45 Polychain has invested in things that are working with the eigenlayer, and there's a number of different announcements of eigenlayer working with different AI projects such as Altlayer, Ritual. You guys can kind of talk about any of this that you want, but yeah, we have limited time, but I did also just want to throw in there, you know, stream, I mean, you have such a background in AI as well. So it would be curious to hear kind of on any of the topics that I just mentioned from either of you.

Starting point is 01:04:12 Yeah, yeah. I'll quickly jump in to just say, you know, we have invested. in, I think, about a dozen projects building on eigenlayer already. I also anticipate that many of the projects in our portfolio that did not have his stated goal when we made the investment to build on eigenlayer, will realize that eigenlayer is probably the best home for the service or technology that they're building. We are also, you know, I'm looking for more projects building here because as I said, I really think the ecosystem growing here is a sort of step function improvement on what exists today. It's

Starting point is 01:04:53 a real expansion on the types of things you can build, you know, just on Ethereum today. And, you know, when it comes to more specifically, Laura, your question of this kind of AI, you know, call it AI co-processer more specifically category. I talked a little bit about this replacing like a wallet interface with the language model interface so that it just abstracts away all of this kind of. complexity that the end user has to deal with today, we've talked about this abstraction for years in crypto. Like for over 10 years, everyone knows these wallets are kind of a disaster. We finally have, I think, a possible path, right? And I'm not saying it's easy or it doesn't come with its own gotchas, but a possible path to a sort of natural language model, you know,

Starting point is 01:05:39 where like if you call your banker, you can say, I want to buy five shares of this company. and they say, okay, great, it's done, right? It's like a language model thing. In the background, there's all sorts of mechanisms that allow that to be possible, but the end user doesn't have to know what a market maker is or what exchange they bought it on or where the custodian is, you know,

Starting point is 01:06:04 and they don't have to ever interact with any of the low-level systems. So I do think that AI for abstraction is a big one. I think another one, though, is this fully homomorphic, encryption. So, like, you can imagine encrypting data and then sending it to an AI model to sort of, you know, compute on it and send you back, you know, the output. And interestingly, the AI model never knows what you were trying to figure out. It can sort of do math, for example, on encrypted

Starting point is 01:06:40 data without you revealing the data to the model. So there's, you know, there's lots of interesting applications here, like let's say you have a huge data set that you think will be predictive of prices. But you obviously don't want to reveal the data set because it's sort of your edge or alpha in the market. You can encrypt the data set, send it to, you know, third party to execute predictive computation on it and send it back to you. You pay them for the service, but also they don't get the data. So they can't like take advantage of you sharing it in some way. I think that smart contracts just triggering based on the outputs of AI algorithms is going to be a huge category. And this language wallet, you know, language model wallet, I think is like the first thing I sort of thought of.

Starting point is 01:07:27 But I think people will come up with many, many more things would be built here. And it's sort of like a feeling I had with early Ethereum, which is I couldn't articulate, you know, in 2014 when I read the Ethereum white paper. okay, well, it's going to lead to defy, Dow's, NFTs in this order and on these dates, right? But I did know that it enabled developers to express themselves in ways that weren't possible before. I do think that that intuition or sort of spighty sense is tingling for sure. And we're seeing like these early application use cases. But I really think a lot will get invented over the next year that I have not thought of. One of the, like, quotes from Olaf from way back, I think maybe 2017, was like, I think he said something like, the world's best investor in like 20 years, maybe an old Tesla, you know, sitting and like, you know, doing trades.

Starting point is 01:08:30 Taking a view on that, actually one of the things I'm super excited about is sovereign. AI, which is sovereign means like self-sovereign. It's an AI that is like not running on behalf of anybody, but it's running on behalf of itself becomes possible on blockchains in a way that's simply not possible elsewhere. Like, you know, for these things to exist by themselves and to actually undergo Darwinian evolution, you know, they need a few things. They need a fitness function. What's a fitness function on a blockchain is how much wallet balance do you have? And to be able to use the wallet balance to actually, say, trigger an AI computation or protocol that actually goes and does a trade, does an MavV action, does like some other thing,

Starting point is 01:09:23 and then like increase the wallet balance. So we're absolutely going to see, I think, sovereign digital AI, this is not robots like, you know, self-constructing like progeny, but it's just like a wallet and an agent attached to a wallet that is sovereign by itself. This has to happen. And my prediction is this is two years up. In two years, we'll see this. And some of this is going to run absolute rampage by like going and attacking wallets and like extracting money and doing all kinds of crazy things. So we are going to need to do AI sanctions before we know it.

Starting point is 01:10:00 So sovereign digital AI is going to be a category. they're going to just be these agents that sit on blockchains and go. Some of them will provide useful services. Others will try to like hack to make money. And the thing is we will lose control really quickly because these blockchains have all the properties needed for these AIs to actually to live. And like not because you have sensitive resistance.

Starting point is 01:10:24 You have immutability. Like all these properties that make it possible for the store rides as its own category. But you know, more fun, more practical, more nice things also will arise. And one of my favorite things is how do we trigger more open innovation in AI? You know, those big fight that's going on at the center of AI is, you know, open source versus close source. And, you know, there are big problems on both sides.

Starting point is 01:10:52 Open source, of course, you know, there's one dimension of safety, which I think actually being open means you are more safe in a very fundamental sense. Imagine you go out to battle with like this opponent, but you know the opponent's entire like battle plan. Why? Because it's open source. Like, you know, if an open source AI goes broke, you know everything about it. And like, you know, can you defend against something about which you know everything? Yeah, I think it's much, much easier. So, but the fundamental problems I see are basically open source has no fundamental incentives. Close source as incentives. Open source has incentives. Open source has permissionless. packing in. Close source has no permissionless innovation. Can we create a new model which has both

Starting point is 01:11:38 permissionless innovation and value accrual to the creator? Because my kind of like highest principle in all of my kind of principles is, is the system have a balanced karma. Like you do something, you contribute to the commons, is the commons giving something proportional back to you? Then you will have an incentive to go and like contribute more and more to the commons. But you know, if I pollute the commons, am I getting slashed? Like, am I actually getting like the negative feedback back from the commons? So that's the two things that are needed to actually make this work. One mechanism we have thought about a lot is a new licensing mechanism, which which is called PDL, permissionless derivatives license. And the idea is, you know, unlike any other license, which

Starting point is 01:12:29 which is usually talks about the creator's rights and what rights are afforded to the commons, the permissionless derivative's license attaches itself to a platform, particularly a blockchain platform. It says, hey, I'm creating this like new software, but I'm putting it up on a PDL on eigenlayer or on like a system on our eigenlayer, for example.

Starting point is 01:12:52 And what it means is now anybody else can come and create. So the PDL says anybody can come and create derivation. permissionlessly, but they also have to deploy the derivative on the same platform and attribute it. Okay, it's not enforced by like, you know, in protocol. Eventually, maybe with FHE and other things, you can actually enforce it in protocol, but initially it can just be enforced by law that this is what the license is, that anybody can create derivatives and post it on the chain. Now what happens is the original, so the value now, you know, the blockchain monitors,

Starting point is 01:13:28 is how much like usage of these models and like value accrued, and it distributes it among both the source and the derivative. And, you know, is it static? Is it dynamic? Can we do complex value allocation across chains? All of these are interesting problems. We have some ideas for. But the core thing is this license, the PDL license, the permissionless derivatives license,

Starting point is 01:13:49 allows, for example, AI creators to come and deploy models and other people to create derivative models on top and have value accrual, like, pass through this sequence of derivations rather than only to, you know, one guy or so it's basically like ties the two things correctly. You have value accrual and you have permissionless derivations. So it's a new model and very examined about and we are going to kind of like popularize this a lot. But this is an idea that could be like really powerful for AI where I think permissionless innovation and value accrual to creators is going to completely

Starting point is 01:14:28 outrun, you know, close source like, oh, we have this brain trust at open AI. Yeah, it's true. But always, the amount of innovation and, like, ideas outside your organization far exceeds the amount of ideas you can capture inside any organization. So that's, that's one direction I'm, like, super excited about is, you know, can we actually turbocharge permissionless innovation for AI on, you know, platforms like IAW there. Wait, and just to understand, like when you talked about that I saw so many different ways that that could work. Like, it could work like for, yeah, just open source, I guess. Yeah, AI is like developers working on things.

Starting point is 01:15:08 But then, you know, you were talking about creators. So I didn't know if you also meant they're like literally like artists or whatever. So it could really be the full range of all these different types of creative. Specifically talking about like the AI model creators, right? People who create and train these models, they're deploying it. But exactly the same ideas hold for all kinds of other things. like media and, you know, even inputs into the models, right? So if you think about a model, the model is like the algorithm plus data and plus compute.

Starting point is 01:15:38 Compute can be hosted on blockchain and accounted for, and, you know, we see many platforms starting to do it. Data should be hosted on blockchains and accounted for and attributed and, like, you know, create, you know, a mechanism for them to be owners of the models, and then model algorithms need to come together. So we can coordinate all these sites on a common, like a legible substrate, which is of course blockchains. And then you see like the system seems to have a lot of power.

Starting point is 01:16:07 Yeah. Get ready, you guys, for tokens that allow you to own a part of these sovereign AIs. You know, instead of owning like a share of a company that controls an algorithm, you have much clearer sense of exactly what you own if you own a token that represents a share, you know, so to speak, of one of these. sovereign AIs and in fact could potentially govern the AI as well. So it's another, I totally agree to Riham, this is going to be, I think, a very big category over the next couple of years and something that lots of very unusual, I think, outcomes will occur. Oh my gosh, you guys,

Starting point is 01:16:47 I'm so excited. I feel like we barely scratched the surface. I had to ditch a whole bunch of questions. But you guys, this was so fun. Where can people learn more about each of you and your work. You know, I think the Polychane website, you know, has all the information you could ever want to know about, you know, our organization. So you can just check that out. From our end, you know, at Eigan layer on Twitter is the EIG and L-A-Y-E-R, is the right place. And also my name's straight on Twitter at Sri Ram Khanan. Both of these are good places to follow and track of what we're doing. Awesome. Well, it's been a pleasure having you both on Unchained. Thank you so much, Laura. Thank you, all, Laura.

Starting point is 01:17:26 Thank you so much for joining us today. To learn more about Sri Ram, Olaf, and restaking, and eigenlier, check out the show notes for this episode. Unchained is produced by me, Laura Shin, with help from Nelson Wong, Matt Pilcher, Juan Aranovich, Megan Gavis, Shashank, and Market Korea. Thanks for listening. Unchained is now a part of the Coin Desk Podcast Network. For the latest in digital assets, check out markets daily five days a week with host, Noel Atchison. Follow the CoinDesk Podcast Network for some of the best shows in Crypto.

Starting point is 01:18:01 video.

Unchained - Why EigenLayer May Be the Most Innovative New Ecosystem Since Ethereum - Ep. 618

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.