EigenLayr: Permissionless Feature Addition to Ethereum with Sreeram Kannan | a16z crypto research

Transcript

today i'm going to talk about one of the products that we're building which is also an idea called eigen layer we call eigen layer the restaking collective but what it does is enable anybody to add new features to ethereum uh as tim introduced i'm associate professor at the university of washington seattle where uh we've been working on blockchains consensus and other stuff over the last four and a half years and over the last year i've been only filling the startup player lab we've done a lot of work on consensus protocols um you know

we had this paper called everything is a race which analyzes conditions under which approval work proof of stake proof of space any of these longest chain type protocols when they're stable when when are they secure uh we built on top of some of that understanding you know we have a paper called prism which is a proof of work protocol which has very low latency we the other work we did is called posat which is you know how do you create a dynamically available proof of stake protocol dynamic availability means your protocol continues to work under variable participation we also did a bunch of other stuff you know for example

when are blockchains accountable uh one heuristic that we have is when you have quorums and signatures then you know if if a co if a group of stakers double sign on a block then though such blockchains are accountable but there is some more subtleties to this for example a protocol like algorand which is also uses quorums and stuff is not accountable because it relies on some other timing assumptions where you can create safety violations by not speaking anything so we did an analysis on on some of those things the two most recent works in this space are being on

multi-resource consensus suppose you want to build a protocol which uses proof of stake proof of space proof of work all combined into one protocol and you want this protocol to work even if you have a majority of proof of work miners being malicious as long as like a very small fraction of proof of stake miners are not malicious so you can and characterizing what are the trade-off regions what sets of adversary ratios can you stabilize across these multiple resources so we did a bunch of work on consensus protocols also on peer-to-peer how do you design peer-to-peer topologies adaptively how do you make sure that you know in a blockchain when you have a

peer-to-peer network the consensus protocol respects the ordering of messages in the peer-to-peer network one of the things that's happening rampantly in blockchains is front running and one way to one type of front running you can call as non-targeted front running you just want to go ahead of everybody else because you have a price advantage in doing so and to prevent non-targeted front running if the blockchain had some native first in first out property messages that came in first into the distributed system need to be executed first how do you define it how do you make sure you have a highly efficient protocol which achieves this we have a recent paper called themis which achieves this on top of the consensus protocol you

have scaling solutions sharding how do you create codes that ensure that you have a scalability how do you do dynamic resource allocation we had a couple of favors coded merkle tree and free to short on that okay so this just gives a background of the type of works we've done and also as since i'm giving a talk in this research group i want some of you who are interested in any of these topics to stop by and talk later okay um one while doing some of this research i mean we did talk to industry a bit and some of our ideas are integrated in some of these protocols like chia for example uses our consensus

protocol in their proof of space uh system but one thing we found as a major friction in blockchain is the rate of innovation at the core layers at consensus charting or peer-to-peer is way lower than the rate of innovation at the application layer because applications are permissionlessly deployable anybody can come and deploy an application on top of an existing blockchain like ethereum whereas core protocol upgrades are permissioned in a very deep sense you can say permissioned or democratic but either way very few innovations get it so this has stalled our space quite a bit and this has kind of bothered me the

most because on the one hand we see there's a massive wave of open and permissionless innovation in the dap layer but all of this is throttled because of the slow rate at which the core layers of the protocol have been able to get upgraded so this project eigen layer is an attempt to resolve this core contradiction at the core of blockchain we want to build permissionless systems but there is no permissionless innovation at the base of these systems how do we solve this that's what we're trying to address here okay so zooming back i'm going to take the story back to 2009 2008 2009 bitcoin pioneer decentralized trust and we all know this and the way this system

is built is you have a layer of decentralized trust in bitcoin it's done through provo work mining on top of this mining there is a consensus protocol for in bitcoin the proof of work longest chain or heaviest chain protocol which then decides what the valid chain is on top of this there is the bitcoin script which sets the execution semantics what do bits that are committed on this blockchain mean oh it means i can send you bitcoin whatever you know transaction types and so on so you know the way we are organizing the blockchain stack there is a trust layer at the base there is a consensus layer on top there is an execution layer on top of it okay so

bitcoin pioneered this model but bitcoin also was an application specific blockchain what do i mean by that bitcoin is specifically designed for one application exchange of bitcoin among clients that's the primary application and if you go back to like 2011 any new application that needed to be built on a blockchain needed its own trust network for example somebody wanted to build a domain name system a decentralized domain name system called namecoin and if you wanted to build namecoin bitcoinscript does not give you enough programmability to start a new system for maintaining domain names and how

they are allotted and so on so you have to create a new scripting layer the name coin scripting layer and but it's not enough for you to create a new scripting layer you have to create a new trust network you have to create a new proof of work mining or use the same mining but let a different group of miners mine and secure your network there was no way to share trust between namecoin and bitcoin okay so this is the story circa 2011 2012. many other examples color coin metacoin you know this was the bitcoin talk era of like 2012.

so the picture is any new application needed to create its own decentralized trust network decentralized trust is an exotic variety of trust it rarely is able to happen like there should be very few decentralized trust networks whereas each application having to create its own decentralized trust network is clearly not the correct model and many people realize this around this time most prominently vitalik and friends who pioneered the notion of what i call modular blockchains the core idea that was built by uh by ethereum was the idea of decoupling trust and

innovation what do i mean by that what they actually did was something quite you know simple in a very basic sense they just took the bitcoin scripting layer and replaced it with a general purpose touring complete programming layer the ethereum virtual machine okay so this was just a small technical upgrade in some sense but what it did is created modularity of trust okay what do i mean by that now that you have this ethereum system you anybody can come and build distributed applications decentralized applications on top of such a system so the person building that innovation

the person who built ens the naming system that is now used in ethereum did not have anything to do with this trust network anybody can come and deploy dapps on top of this virtual machine that's modularizing trust so the trust of ethereum network became a module that can be supplied to any distributed application that wants to build on top this has actually led to a massive acceleration of the pseudonymous economy why do i say that i say that because anybody creating any of these applications they are not themselves trusted they're just bringing innovation you come up

with an idea for how to build a new swap which is just a decentralized exchange and you can be a nobody you don't need to be trusted you just write your code you put it up on ethereum everybody trusts ethereum will continue to execute the conditions as stated as long as you can understand the code you don't need to know who the who created this distributed application so this has led to a massive error of what we can call open innovation where anybody can come and contribute their innovation plug it on top of this massive truss network and deploy it for anybody to use so

one way of modeling this is you think about the base layers the truss network consensus and the virtual machine together bundled into a trust network and producing trust so essentially the ethereum blockchain is a producer of trust decentralized trust and the distributed applications are a consumer of trust they consume trust and in return pay fees back this is the value exchange the value exchange is the dapps get trust from ethereum and in return they p pay fees back to ethereum this was the model by creating this new economy where trust and innovation got decoupled many more things became possible that

was simply impossible in the previous era prior to ethereum if you create if you created unit swap you have to build your own trust network and you know how many people mine on that trust network how big is it depends on your own particular economy whereas the size of this trust network is universal it's just shared and pooled across all the all the applications that run out of ethereum okay so just one high-level analogy just like venture capital was a massive innovation it was the decoupling of capital and innovation ethereum decoupled trust and innovation i don't need to be trusted i just bring the innovation

ethereum brings the trust and together we can create a product so it's a it's a serious upgrade but the barriers to open innovation continues to persist okay for example you have this stack and i have an idea how to upgrade the ethereum consensus protocol okay let's say uh it's 2019 and i came up with this avalanche consensus protocol and i want to just deploy it onto our ethereum there is no way to do it right so you look at it and you say oh if i wanted to create a different consensus protocol because it has different properties it has different latency different assumptions

i'm simply unable to create it on top of this existing trust network so what do i do i go and create my own whole world the era of alternative layer one blockchains different blockchains have different consensus protocols different virtual machines and but each of them have to build their own new trust networks this picture looks exactly like the 2011 picture of bitcoin and name coin and color coin except now innovations that are at the level of dapps can simply build onto our ethereum but innovations that go deeper and touch the heart of the stack have to actually segment and create fragmented trusted

ecosystems okay so this leads to a double coincidence if you want to start one of these ecosystems it's not sufficient for you to have great technology you also need to be able to bootstrap a decentralized trust network which is you have to be an evangelist and all this there is another layer at which we see this barrier to open innovation okay so if you just abstract out ethereum as basically supplying trust to the dapps what we see is ethereum does not supply trust to the dapps for anything that the dap wants it only supplies trust in block making what is block making

block making is transaction ordering and transaction execution that's all that block making is so the apps are able to borrow trust from ethereum on this limited access of transaction ordering and transaction execution if they wanted trust on anything else i want to read data from the internet i want to read data from another blockchain i want to run a different execution engine i want to run a gaming engine i want to run an authentication system any of these things you have to create your own new trust network a simple example is something like chain link chain link is an oracle protocol which helps fetch data from the internet into the blockchain

but chain link cannot be built on top of ethereum natively chain link has its own trust network it has a set of smart contracts from which it brings the data into ethereum but the trust network is not trust is not borrowed from ethereum or stakers trust is its own group of new middleware stickers so each middleware creates has to create their own decentralized trust network any other dimension that the dapps need in which they want to consume trust other than block making they need other trust networks to supplant those actions so the

the i highlighted the bootstrapping problem that when you want to create these new middleware for example you not only need to have innovation but you also need to have your own new trust network and it's very difficult and just take the example of oracle's chain link is a market leader and you know it's worth whatever 20 billion or something and the second largest article is like 100 as big as chain link and part of the problem is you cannot bootstrap a new article because you have to build your new trust network it is not simply a matter of getting adoption it is also a matter of bootstrapping your trust network so one of the things i started seeing is that even though we think blockchains are creating an error of permissionless open innovation

there are the anything that requires its own trust network gets entrenched very deeply and it's very difficult to upgrade it and add on new innovations on top of this okay so but i want to highlight what the micro economic problem is on each of these different players in the current architecture so the microeconomic problem is if you're running a middleware let us say i'm running a middleware which is data storage claims that you know you give me data i'm just going to store it and i'm going to have some cryptoeconomic game to make sure that you're storing data so i want to create a system like this for what i have to do is i have to

usually create my own staking mechanism okay so i create this like data storage token and let people get this data storage token and stake it and by staking it they're agreeing to provide service correctly if they don't do provide service correctly their staked tokens will be slashed or taken away okay when you want to create a system like this and you want high economic security you want a lot of capital state and when you have a lot of capital state you have the opportunity cost of capital and so for example you want 10 billion staked in your data storage layer because you know you want to have comparable security to ethereum now

you have to pay a 10 or a 5 or something annual rate on that capital state in order to sustain that staking group in a non-speculative world today people are you know three months back people may have been buying these tokens taking them because the token number go up but you know in a non-speculative world you need to account for the cost of capital so there is a massive cost problem in running the middleware the cost problem is you are not bounded by the operational cost of storing data on end you're bounded by the cost of feeding a massive economic capital base that's the dominant cost

so one thing as we were designing some of these protocols is i started finding that most of the rail systems are not at all capped by operational costs they're all capped by capital cost and one way to know it in reality is you look at any proof of stake ecosystem you see how much reward goes to the person who provides operations of staking versus how much reward goes to the person who holds the capital 96 or 97 percent goes to the or 94 goes to the person who holds the capital and only six person goes to the person who actually does the operations so it's all cost of capital so even if you come up with a breakthrough idea for reducing the operational cost like 10x it's still like the 94 remains unchanged

so this is a massive problem so when you're designing middleware your cost structure is actually capped by cost of capital okay when you're if you're adapt the microeconomic problem you're facing is you're paying a very high fee to a large trust network like ethereum so that you're getting trust from them but you're limited by the weakest trust that you're depending on if you had an oracle which is not as trusted if you had a bridge that is not as trusted you could get rugged there so your secure trust is always the least common denominator so you even though you're paying a high fee the security premium for a better blockchain you're not getting trust on all the dimensions that you want

finally as far as the core blockchain is concerned there is also an economic problem the economic problem is if the core value proposition of the blockchain is to provision decentralized trust and make revenue on it ethereum is only able to provision decentralized trust on block making not on all the other things that may be required to run a decentralized service so what is happening is there are islands of decentralized trust that are being created by other middlewares and instead of the revenue aligning and creating a massive trust network revenue gets fragmented into smaller islands fragmenting trust for the ecosystem as well as losing to a value

uh value misalignment to the core ecosystem okay it's actually a ridiculously simple idea that solves these problems all at one go okay the solution is called eigen layer um it is a mechanism to leverage an existing trust network to do other things i want to leverage an existing trust network to do other things it was not intended to do okay that's it that's the core idea how do we do it i'll get into that okay so ethereum supplies trust on ordering and execution as we said earlier

eigen layer is a series of smart contracts on ethereum okay and what it lets you do the core operative word here is restaking that's the new idea restaking what is restaking re-staking is so when we're talking about ethereum we're talking about proof-of-stake ethereum in my mind already we are in the proof-of-stake world right it needs to happen but already for example several tens of billions of dollars maybe 30 or 40 billion dollars is already staked in the ethereum proof of stake network called the beacon chain which is just a run-in to actually start running the proof-of-stake network so

there is already a lot of stake locked up in ethereum so what is eigen layer eigen layer is a mechanism by which stakers stakers re-stake what is restake re-stake is they put the same capital at additional risk same capital additional risk okay so they lock the stake in ethereum the same stake they commit to additional slashing conditions slashing is a mechanism by which your stake can be taken away but you add additional reasons by which you can be penalized your steak can be penalized that's what you commit on top of these smart contracts on eigen layer we do not need

liquid staking solutions i'll come to address that point but right now i'm stating the property i'll state the mechanism next the property that we want is the same stake it takes on additional risk and additional risk on what on providing any new services that have been built on top of eigen layer any new services somebody wants to come and build an oracle somebody wants to come and build a bridge somebody wants to build a data availability layer somebody wants to build a new consensus protocol any of these can be built onto our wagon layer okay if you're a staker and you are opting in to eigen layer you also specify which subset of services you're opting in to

provide trust for you may say oh i like the oracle project because i'm getting x revenue and need to do x work and that's good enough for me so i'm opting into the article but i'm not opting into whatever this fair ordering or some other service so as a staker you can specify which subset of services you are opting into and that therefore gaining revenue and also potentially taking additional slashing risk on okay so now let's see how this aligns everybody in the ecosystem correctly first

if you are a middleware okay let's say i'm running an oracle and a staker who's already staked in ethereum has to opt in to the also provide services on this article they do not have an additional cost of capital they'll already staked on ethereum and earning apr or whatever on top of it and you cannot do much else with that stake so by opting into eigen layer the marginal cost of capital that you're incurring is either very small or you can say theoretically zero of course there is a slashing risk that you have to account for but if you know that whatever middleware that you're

opting into if you're an honest node you will never get slashed then that slashing risk is either zero or minimized based on whatever smart contract risk and other platform risks that may be there but the marginal cost dynamic of a staker is fundamentally hey what is the operational cost of running this middleware and am i getting enough revenue for doing it if the marginal operational cost is justified by the revenue that i'm getting on that middleware i will opt in it's a very simple equation so the cost structure of middleware suddenly transforms from a capital limited to an operational cost limited so this is extremely valuable if you

want to come and build a new oracle you know you don't have to worry about how am i going to compete with chain link who has a 20 billion dollar at stake that is 40 billion dollars take on ethereum and you can get them to opt into your oracle and start providing services okay as far as the dapps are concerned especially services which are very popular which many stakers will opt into you're fundamentally getting the same trust as ethereum itself if all stakers potentially opt in so you could get the core ethereum trust on services that were not natively built into ethereum and finally its extremely value aligned to the core ecosystem to ethereum itself

why because not only is it giving additional yield opportunities for the stickers people who staked on ethereum get block rewards and transaction fees but they can also get oracle fees and data availability fees and ordering fees and all these other things that were unavailable to them earlier so the fact that you have potential additional sources of revenue for staking e increases the value of the token itself so by kind of providing a mechanism by which you can supply this trust to other services and other people can create these services and it doesn't have to be created by one

monolithic group and does not need uh approval from anybody basically creates permissionless innovation on top of a common layer so eigen layer is a two-sided marketplace one side is takers opting in to into this eigen layer the other side is middlewares services that are built on top of wagon layer opting in to use these stickers is this system assuming honest majority or is this system operating on crypto economic assumptions which is you can get slashed i think that depends on the particular details of the middleware

as long as you can write an on-chain slashing contract i'll give an example of where you can write a simple on-chain slashing contract for example you want to run a new blockchain on top of the same ethereum trust network let's say i want to run a solana on top of the ethereum trust network the solana blockchain the c level virtual machine and whatever other consensus protocols i want to run this on our ethereum one of the core slashing condition in a protocol like this is if i double sign a block at the same depth of a block i assign two different stickers signs on two different blocks this is a very easily on chain verifiable slashing condition then that set of nodes gets slashed so

we do not have to make an assumption that people are going to behave honestly all we need to make all the statements the type of statements we are shooting for if they did something wrong on your protocol x dollars gonna get burnt so that's the type of crypto economic statements we want to make if they did something bad to you x amount of dollars getting burnt at stake so so that's the kind of assumptions we are shooting for we do not require the transference of majority honest assumptions there are certain things for which you cannot write on chain slashable contracts this goes back to the discussion we had earlier in the morning here

for example threshold signing or threshold revealing and stuff like that there is no clear way to do slashing for those things you may be happy with saying that hey you know it's the ethereum stakers that are doing it and i'm happy to get a majority trust from them but many many use cases actually you can write sharp on-chain slashing contracts for example if i'm writing an oracle oracle is one of the harder cases but in when you want to build an oracle you and i may reasonably agree that we will get slash if i bring in a price feed which differs significantly from a coin base and a binance and ftx or whatever price feeds whereas that kind of an assumption cannot be done at the base layer of a

protocol but as an opt-in layer you and i may agree that that's a slashing condition that we've opted into which becomes on-chain verifiable so there are certain things which are on chain verifiable there are certain certain things which are not and each middleware has to design for this correctly so if to the extent that your feature is on chain verifiable to that extent you get crypto economic guarantees rather than majority trust guarantees and you're getting the potentially if all each stake is opt-in like i've come up with let's say a much better protocol than all these consensus protocols than a virtual machine i don't have to go like figure out how to like shill a new token and make it worth 40 billion for the system to be useful i just need to convince the these each stakers to opt in and run this new

chain and as i run this new chain it does it's not security limited anymore it's adoption limited and it transforms the dynamics of creating new distributed systems so that's that's the idea so other than that there is no tethering to ethereum but because it is ethereum adjacent for example slashing contracts and other things are there on ethereum you can actually start interoperating with ethereum much more fluidly than if you are a separate chain transaction fees for example goes you could you could do many things you could create your own new token and this token is used for like be it is the fee payment mechanism in your system and you use the same tokens to also reward the ethereum restakers in

eigen layer it's a free market i can let so the way eigen layer works is you come in and say here is a new service and here is the reward risk dynamic of this service and stakers opt in whether they like it or not they're just voting yes or no on this free market so you can specify what the reward mechanism is reward mechanism could be something different it you don't need a new token you just create transaction fees are paid to the e3 stakers perfectly fine if you're on if you're building an oracle you may say that anytime anybody queries the article that fee directly goes to the e3 stakers so you can create your own like economic models on top of this but even just that one thing the fact that you can pay in arbitrary tokens rather

than using only ethereum is itself a value proposition to run like interesting things on top so how eigenair works it does not require something like liquid staking what you do is you go and deposit your eat into the ethereum smart contracts and in the ethereum smart contracts you have something called withdrawal credentials who has the power to withdraw the stake from the native staking and we said so if you want to participate in eigen layer you set the withdrawal credential to the eigen layer smart contracts and by doing this i can layer as withdrawal power to withdraw your stake whenever and therefore when it withdraws the stake you may not get back your entire

stake if you did something bad in eigen layer so and to the extent that everything is written on chain and verifiable all slashing conditions are return on chain and verifiable you're not trusting me or anybody for the correctness of these slashing conditions so you opt into iron layer you agree to these slashing contracts and your is stake being over leveraged and how do i know if it's being over leveraged i think that's the question and to give a very concrete example let's say there are 10 different apps which are running their own chains now there is a kind of trend people are thinking about each dapp may have its own chain because it doesn't have to suffer free

congestion from other things happening like recently there was this ape coin and bayc had this you know created some externalities to the rest of the system so each tap could potentially be running its own chain and let's say you have a dap and you think that hey i have like one million dollar in like value in this and i'm happy if like two million dollar stake in this system and but it turns out ten of these daps each of them have one million dollar in value rely on the same two million dollar quorum that's this stake became over leveraged so this is the scenario of concern and we're very much of course worried about this scenario and this is why eigen layer itself is also the risk management layer so what

do we do we can actually model this problem graph theoretically you each staker is a node and each service depends on a bunch of stakers and has a condition that if more than 60 of the stakers cheat or whatever that service kind of collapses and there is a profit from corrupting that service each service declares what this profit from corruption is you have this two sided graph there's a profit from corruption for each service and there is how much is take by each of these takers and then you can calculate cuts on this graph which say that if you if you remove some bunch of stakers out they collude together then this much of stake gets burnt

and this is the tv you know this is the uh profit from corruption total profit from corruption and we want to keep this graph stable it's a beautiful combinatorics problem i think you guys will like it a lot we don't have an exact solution for it but we have sufficient conditions and actually uh mechanisms by which we make sure that the system is never over leveraged one thing is you know we haven't analyzed this scenario this gaming scenario of tvl itself because then the neutral assumption is that tvl or total value locked in any of these contracts is kind of verifiable right but if somebody randomly under declares the tbl so over

declaring tv you're just like more conservative but if you under declare your tvl then there could be some issues we have not analyzed whether that leads to a contagion or contained attacks only on your system the idea would be that only your system gets attacked right okay so having said all this the really useful use case for eigen layer is when almost everybody can run everything okay this is actually you know one would think that this is the most over leveraged everybody running everything right but this is actually the best case scenario that you actually want in fact what it does is it massively deliverages systemic risk as

of what is existing today let me explain on the left side what you see is you have daps and there are like three middlewares each of these middlewares have one billion dollar at stake and there is a layer one which is has 10 billion dollar state just some hypothetical examples and every dap doesn't depends on all these middlewares some oracle some bridge something right and on if if you had this picture and all these middlewares were mission critical which means if you attack them you can kind of tank the dap then what is the cost of corrupting this left side ecosystem

is just corrupt one of these middleweights the cost of corruption on this system is the min of all these numbers okay let's see what happens in an eigen layer world in an eigen layer world let's assume that all stakers provide services on all these middlewares the fully over leveraged example what happens in this case is if everybody's providing service on every one of these middlewares now to attack the dap you need to attack a majority of all the stakers okay and the interesting thing is how much staking will be supported on this summed up system is actually the sum of all the staking

that was supported on all these systems why because each of them were enough to sustain a 5 or 10 apr on the fees that it was getting on each of these systems so what happens is you're moving from the min norm the minimum of all the things that you're depending on to the sum of all the things that you're depending on if every staker can run every service that's really huge so it massively reduces systemic risk from being the least common denominator to all services being commonly provisioned by one ecosystem when is this feasible this is feasible when middleware are

designed to be scalable horizontally scalable as you get more and more nodes the per node's workload goes down then actually everybody will opt into everything because the operational cost of opting in is very low and you're actually just uniformly dividing it by endnotes so when middlewares have scale out economics then actually we expect basically every staker will opt in to everything okay so all right so eigen layer is the restaking collective a mechanism by which you opt in it's kind of become like a layer directly

on top of the staking layer and this restaking collective enables you to pass this trust beyond the limited confines of protocol upgrades basically now you can build any new blockchain that you want a new virtual machine new consensus protocols new data availability layers all of these things into uh on top of eigen layer rather than having to go and build separate blockchains for each of them so instead of having having open innovation only at the dap layer you have open innovation across all layers of the ecosystem where you're sharing a common trust network and that therefore reusing this for any new innovative services that you're

building and just uh note here that we do not need access to things like liquid staking derivatives even though we can use them and finally it lets each node permissionlessly opt in to a subset of services that they are most interested in okay so i think this is repeated um i'll just briefly touch upon for some of you who may have been familiar with uh the evolution of this idea is origins is called merge mining what is merge mining so let's look at approval work system in a proof of work system if you want to run the system there are two different costs one is the mining cost and the other is the

validation cost and the mining cost which is the cost of solving the provo work puzzle far dominates the validation cost and one idea is can i amortize this mining cost across many many different systems where you're validating things and indeed people tried this so that the per blockchain you know mining cost actually goes down but there is a significant problem is that the crypto economics the economic incentives do not transfer to the merge mind systems you can if you merge mind and altcoin along with bitcoin even if 100 of bitcoin miners opted into

the altcoin mining they can attack the altcoin without taking any significant damage because the bitcoin price is not affected and their mining equipment most definitely is not affected so while merge mining was a good idea for amortization it failed in the economic incentives that that accompanied our core understanding is merge staking merge taking completely transforms this economics in a proof of stake system the less appreciated fact is the dominant cost is not validation cost it is the capital cost of staking and this is really dominant and if you

want to amortize this capital cost what you're doing is you basically use the same stake merge stake across many many systems by merge staking do you suffer the same problem that merge mining suffered no it turns out it's the opposite you actually get perfect crypto economic transfer what do i mean by perfect crypto economic transfer if you have five percent of ethereum's takers operating on your protocol you have the economic security of that five percent which is significantly better than uh than the crypto economics of merge mining also the other fact is if only four percent of bitcoin miners mined

your system you have no economic security whereas if four percent of stakers just opt in there's only one your system which is merge taking or not on top of ethereum you get economic security which is a transformational idea it's actually that's why i said this is really like a really really simple idea but it's just not been thought about before i think partly because of the path dependence and people got burnt with merge mining more than solving the economic problems uh eigen layer also solves some of the ecosystem problems one of the problems is the question of block space okay

so what is block space block space is basically each blockchain because it has finite resources has you know only a certain maximum size of the block that it can accommodate the size could be measured in you know bytes or it could be measured in some computational units but today we we're starting to see for example ethereum block space is starting to command a lot of premium it is one of the most valuable commodities in blockchain so the economics of block space it is determined by its relationship to block limit what is the block limit block limit is the maximum size that a block can be and all of these systems have self adjusting economics by which as your

block size gets closer and closer to the block limit your price starts exploring so that the system is doesn't go over capacity okay so that's the basics of block space if block space is and having more block space is in general good because you can build more and more interesting applications even today morning we were talking about auctions and you know whether it's 2.5 million gas or 5 million gas it's only 300 milliseconds on or like 3 milliseconds on a computer we should be able to run it and it's not possible today i'll explain precisely why and how either completely transforms the economics here okay

you look at the notional block space or the maximal block space that a node can run okay one way to think of it in very concrete terms is what is the bandwidth that a node has okay and that determines that puts a limit on block space the x-axis is replicas or nodes different nodes have different potential amounts of processing power or bandwidth let's say different nodes have different bandwidths the x the y axis is just block space you can think of it now measured in bandwidth okay there are some nodes which have very low bandwidth and some nodes which have much higher bandwidth okay and how is the block limit

determined how do you determine the maximum bandwidth that you should consume as a blockchain what you do is you set an entry restriction called block limit you say that i want to at least i want to admit for example ethereum's philosophy is i want to admit a home validator in venezuela that's what i want so one megabyte per second may be okay so you go and say that's the block limit i'm going to set my block limit so that a one megabyte per second guy can participate in the system so that's how you set block limit so block limit is set by the weakest nodes infrastructure the weakest nodes infrastructure

to participate in the ecosystem now what we realize is all of us using ethereum are paying because of this guy the one guy in venezuela we want to accommodate even though this guy has one megabyte per second all of the guys stakers running on ethereum or on amazon web services or all nodes are korus one or figment any of the staking services actually have 10 gigabits per second connection that's like a thousand or ten thousand x difference in the computational resources between the the powerful guys here and the last guy here and block limit is determined by the weakest note you want to admit

so part of this is the paradigm of blockchains today where it is a completely homogeneous system it doesn't matter whether you're talking about ethereum or you're talking about a solana in solana the entry limit instead of being kept at one megabyte per second is kept at 50 megabytes per second or 100 megabytes per second but it's still if instead of having 100 megabytes per second you had 10 gigabytes per second you can do nothing more on the salon system so all blockchains today are determined or are designed for a homogeneous ecosystem homogeneous stake or ecosystem whereas in reality stakers are highly heterogeneous okay how do we solve this problem

we don't need to do anything more eigen automatically solves this problem by creating a free market by which these same stakers can lend out their additional block space for other services built onto our wagon layer for example in this system somebody will come and build another chain which instead of having 15 million gas per block will have 15 giga gas per block because you know these takers can opt in and get a different security you don't get the entire ethereum security you get like a 60 of ethereum security and that's already good enough and i just want to make one comment that there is it's not all about economic security there is a regime where you need decentralization that's for censorship

resistance because whether your transaction got censored or not is not a slashable offense so what we really like is you have a decentralized base station like ethereum which provides a layer of censorship resistance on top of which you can build any other service and you can use ethereum as a backstop for censorship resistance if a transaction is returned on ethereum you're forced to include it in your sidechain or whatever other service so by having this separation that ethereum becomes a censorship resistance layer and therefore keeping the node restrictions very low and on top of which you can construct high power systems which can still get a lot of the economic security and have insane computational liabilities this is a

transformation of block space okay staker heterogeneity extends beyond computational abilities stakers are heterogeneous not only in how much block space they have they're also highly heterogeneous in their risk and reward preferences you and i may agree that we'll get slashed if we differ from a coin base you know api up api output whereas some for somebody else that will be completely unacceptable so stakers are highly heterogeneous in their risk-taking abilities and in their beliefs on the world and therefore this can never be normalized into a core protocol but can be externalized into an

opt-in layer so stakers are heterogeneous and risk-taking abilities stakers are also heterogeneous in reward preferences for example beyond so in ethereum block is thought of as a colorless quantity all transactions are equal and the only signal you use to distinguish block transactions is price whereas this creates all kinds of interesting problems like for example it is very difficult to build a social network on top of ethereum even if ethereum had much more block space because every social network transaction is competing with a d5 transaction which is much more profit and loss on a transaction by by transaction basis but on a 10-year horizon the social network may win out

so there is no ability for stakers to express their coloring preferences that they would much rather bet on a social network than on a financial system so our solution is they opt into different sub chains different layers on top of the core layer in which they have different reward preferences you get paid in this new social token and you're betting that this social network is going to grow so eigen layer maximizes taker heterogeneity and creates a free market instead of having a minimum common denominator market okay so finally it solves the one problem that has been

bugging me for four and a half years which is how do you design a blockchain which is both democratic and agile in its way of innovation so here is a plot the two axis is you have x-axis is governance is governance democratic or autocratic and the y-axis is is the rate of innovation agile or stable or slow right something like ethereum is very democratically governed but is also very very slow to respond to new requirements in the market all protocols today i'll claim are

making a trade-off between agility and democratic governance there's a trade-off between governance being democratic versus the rate of innovation and this is a significant problem for blockchains if we want to take the thesis of permissionless innovation ethereum plus eigen layer actually gets the best of both worlds you have a base layer which is democratic which is slowly updated which is stable over long time frames on top of which on ion layer people can build many new innovations which respond very quickly to market demands in a completely permissionless innovation actually i would even say that ethereum plus eigen layer pushes

this all the way up here so this is the rate of democratic innovation this is the rate of monopolistic innovation this is the rate of permissionless innovation because the upgrade on bsc needs to be shipped by the cto at bsc or something whereas an upgrade on eigen layer can be shipped by anybody who's interested in shipping an upgrade so okay so i think i'll conclude the talk with this um it's i just gave a hint of what kind of things can be built we are exploring building bridges building event driven automation far ordering services side chains and all kinds of other interesting

things like how do you prevent prevent or integrate mev into the blockchains all of these systems can be stopped built on eigen layer eigen is already live in our internal test nets you can deposit things like liquid staking derivatives you can deposit into the into the ethereum beacon chain and opt into eigen layer and not only that we have actually already built the first use case for eigen layer which is a hyper scale data availability layer for ethereum uh i will not be able to talk about it now but i can tell you that it is basically a data availability layer this is one of the biggest problems in ethereum scaling today you have rollups

which write which offload computation and uh memory but cannot offload uh network bandwidth all of the data has to be committed on top of ethereum in order to make sure that they are secure and instead we have an e3 stake layer called data layer which actually does all of it but incorporates some of the best ideas in erasure coding polynomial commitments to actually get a hyper scale data availability layer it already gets you a rate at which you can write data into this layer is on our test net 12.4 megabytes per second which is 10x larger than ethereum 2 which is scheduled to ship in two years so i'll stop the presentation of this uh

happy taking questions so like if i like to point item there and i was like happy with whatever risk like your graph like calculates right then say like 10 more things deployed and that can be just all good but it seems like maybe so the way you would solve the problem if you're over leveraged is by increasing your fee more people will opt in into your layer and you'll get under leveraged again so that's the dynamics that we're anticipating i'm network

that like currently they're in this great situation where they're like really under leverage it's so much capital and the amount of damage you can do on ethereum with all that capital it's not hold on so i'm wondering sort of like as like like even if there's hundreds of people who are already in a stable step but every new person that gets had it like do i need to be concerned that like a new person who gets added did something done in their protocol analysis it used to be securing me and used to be stable as now yeah i think one way of thinking about it is if you go back like three years the ethereum stake supported like a maybe a hundred million dollar dap ecosystem and today the dap ecosystem has gone

like thousand eggs and the ethereum stake has grown thousand x is exactly what we expect will happen to this ecosystem as more and more services start opting in the yield opportunities go up and as the yield opportunities are going up basically more more capital gets locked in instead of like five percent of each getting locked into stake you may have like fifty percent of you getting logged in to to provide services so i follow that but like what if you know like what happened with like the dow or whatever that it was just like a huge wildly costly mistake like if someone you know like it's it's one thing that was completely separate from the consensus player right but like it's another thing if um you know like yeah yeah so there are

like i think it's a it's an important problem and the problem is okay let's start from the point of view of this taker the point of view of this taker is somebody came up with a new middleware right and it's giving me exild do i opt in or not and okay if it's just yield and not much risk then you opt in that's the you know now that all d5 yield is drying up people will opt in when there is additional actual like genuine way to earn more yield but the risk that is very hard for them to understand today is is this middleware have like smart contract errors distributed system issues that basically could slash an honest note that's where i think all of that

problem if you just want to remove abstract it i think one way of expressing that problem is that i think what we need is uh just like there is a smart contract audit ecosystem we need these like middleware audit ecosystems and my general thesis for it is smart contract audit is serving users who are supposed to like know nothing and middleware audit is serving stakers who are supposed to know something if we can't get middleware audits to work we should not really be trusting smart contract audits so that's my high level of why i think it's okay but you know it's not an answer to this problem the problem is fundamentally if all stake

the the extreme example of your case would be all stake opted into an eigen layer uh system where you can get slashed like even without actually doing anything bad and then you got slashed and now the whole protocol is at risk it's possible it's possible and so whatever these new middlewares are need to be like more and more and i think you know in these scenarios there is a tragedy of the commons but there is also the stakers are actually the ones who are losing their money so they should have been more careful in actually opting in and how do we make it easy for them to be more careful that's all that we are focusing on using l1 block space it depends on what you're doing so the scaling part is not the same as moving to l1 so let me give

a very concrete example i just mentioned the data availability layer but basically what do i mean by scalable middleware the data availability is a great example of a scalable middleware where each node so if i'm downloading if i want to store a file size of size f instead of each node storing and downloading the file that's what happens today in ethereum what you would do is each node only stores roughly one over n of the file size in an array recorded thing they receive like a proof that this chunk corresponds correctly to some polynomial commitment and they all sign off on that commitment and if enough people sign up on that commitment you think that that data is available

so this has a self-scaling economics because the total cost of storing a file is just a constant it's actually two it's like as the two nodes are storing a file because each row is only storing two over end of the file so the total cost of storing a file does not depend on the number of nodes that opted in but the price you can charge for it depends on the number of nodes because you're giving more economic security so what happens is there is a self-scaling economics by which more and more nodes will opt in because you know they can charge a security premium without increasing the cost of operations the cost is constant so basically what is happening is the

erasure code breaks all trade off between scalability and decentralization you can get full decentralization and full scalability simultaneously because as you have you have n nodes you get the security of n nodes but you get the cost of only storing on two nodes so that's like full scalability and full decentralization simultaneously and by using layers like this you can actually like solve it without requiring each node to like provision more and more bandwidth or more and more storage space so that's one way of scaling another way of scaling is i just want to take the salon away right i just want to run like you know at a much higher like block space than what ethereum is comfortable with by that i will cut out the home

validator in venezuela yes but you know the thirty percent of people who can actually opt in are enough and they provide enough economic security for me and as long as for them the op the opt-in cost or operational cost is less than what uh revenue they're making they're happy to opt in so these are different types of scaling and we are just a layer on which you can build arbitrary distributed systems on ethereum that's you want to build your secure multi-party computation you want to build your fully homomorphic system you want to build some other crazy thing that you have you can just leverage ethereum trust to actually do it rather than having to go and bootstrap your own trust and the now what happens is there's a competition between scalable middleware

that's built on top and non-scalable middleware and because of the operational cost advantages of scalable middleware we expect they will win out completely or non-scalable so where is this group of state [Music] that's one use case of eigen layer you can also build things that are simply not a or additional services on on ethereum like articles right oracle's not a side chain so the difference between having a side chain and an e3

staked side chain an eigen layer side chain is you don't need a new token of value you don't need to pay to maintain the cost of capital of that token of value you're just like amortizing that over all these other things that other people [Applause]