Stephane Gosselin & Vasiliy Shapovalov: MEV-Boost – Preparing for the Merge #455

Transcript

so welcome to epicenter the show which talks about the technologies process and people driving decentralization and the blockchain revolution i'm brian crane and i'm here with my co-host felix luch so today we're going to speak with stef stefan ghosla and vasily shapovalov they're from the flashballs team and the lido team uh so just a few words on that so flashbots of course is kind of the team that's really pioneer to work around mev which has become one of the most important topics in crypto you know across all blockchains people are worrying and thinking about this topic and flashbacks done you know really pioneered this this field and

focusing on ethereum and how to address this uh on ethereum and then of course the cv is uh from lido which has become uh the leading liquid staking solution and especially on ethereum has you know a huge amount of stake and has also kind of taken on a key role in sort of shaping the future of ethereum and of staking on ethereum now just before we go into the episode uh briefly about our sponsor so our sponsor is uh tally ho so taliho is redefining the wallet as a public good you can think of it like a community owned alternative to metamask with tally ho you can enter the

meadows with a web3 wallet that's fully community owned and operated it's the first wall that's also a dao so and their commitment to community ownership in public goods goes beyond the wallet so for example they became one of the first sponsors of ether js an open source javascript library helping to develop was connected to ethereum and they also pledged some two and a half percent of their tokens through a bitcoin aqueduct so head over to italia t-a-l-l-y dot cash to try out uh their wallet and play with features so with that let's go into uh the episode so thanks so much for joining us guys

uh maybe we can maybe we can just do like brief interest from each of you before we sort of you know get into the meat of the the topic maybe stefan do you want to go first yes uh quick intro hey i'm stefan i'm one of the co-founders over at flashbots i focus on everything that's product product architecture related and yeah one of the big focuses for me over the last year has been figuring out an mev solution for proof-of-stake ethereum so i'm super excited to be here chatting with you all about it hey i'm vasily uh i work with the lido uh as

like basically attack lead or lidar and uh we are doing a liquid staking for uh basically we want to be a dao that builds liquid station for everything was taken currently we are on dream uh polygon solana polka dot kosama and building one for avalanche and on the room we uh uh is the leading liquid sticking protocol cool well let's uh let's just do a little bit of context uh for people uh i mean we've done a whole bunch of

episodes about meb but still like some people it's a complex topic so maybe we can just do a very brief intro like what is meb and like why is it important okay i can take this on so mev is sort of all the value that a privileged actor within a system like a blockchain is able to extract from their position as this privileged actor when it was initially sort of brought to the um through to the community as part of phil diane's research it looked at a minor extractable value so the ability that miners had on ethereum to extract value from inserting reordering reordering or censoring transactions and looking specifically at sort of

arbitrage and and pga wars on on ethereum and the goal was to sort of identify what are the ways in which this value that you know miners have the ability to extract is actually being extracted in practice and which was through the use of uh of you know gas wires on uh and gas wire bots and and try to model what are sort of the negative externalities and the end game of of uh of this value extraction um quickly built up to be sort of a massive problem um or a massive opportunity depending on who you are in the system um you know over the last year over a billion dollars of of mev

has been extracted just atomically on ethereum alone and the mev ecosystem has sort of grown to be much much bigger than this um you know mev pro poses a bunch of different threats to uh to these kinds of decentralized systems um the main one being a centralization risk um and so uh that's why we sort of form flashbacks as an organization to try to understand the problem better do a lot of research and then build solutions that help mitigate them cool yeah so great to have you guys on um also and i wanted to like dive a bit into that right so you're there is this concept that you

introduced stefan or like flashback introduced which is this mev supply chain maybe can you talk a little bit about the mev supply chain and where exactly the centralization risks lie that you just kind of briefly touched upon yeah so the mvv supply chain is a simplification i suppose of how the ecosystem is emerging around mev how the industry is emerging and the different roles that actors are playing so the way that it works is we look at starting from a user right that has certain set of preferences they want to execute some kind of action on top of on top of a chain they want to execute

some sort of of state transition but they need help to do so so really what they have is a set of preferences then what they do is they um they interact with uh an application right whether the application is is adapt like uniswap or a wallet like metamask they inter they interact with some kind of interface that helps them encode their preferences into a transaction it's sort of then the job of that application right to figure out how to take this transaction and actually get it executed and unchained historically it's been pretty simple right they send that transaction to the transaction pool it gets picked up by a miner and it gets included um what we've sort of noticed

is with these mev opportunities but operators picking up transactions from the transaction pool and sort of inserting themselves in the middle of the supply chain adding additional transactions around it right to form sort of a bundle of transactions and then submitting that to to the miners for for inclusion so if you're keeping up right now we have already four different actors on the supply chain the user the application the searchers and the miners finally you know with the the approach of of proof-of-stake we introduce this additional role which which we refer to as the builder and that is really just separating out

the the role that the miner has into two so the role of of taking a bunch of these bundles from search rows and transactions from users and producing a block with it the block builder role um and then the role of taking um you know the the most valuable block and proposing it to the network this is the this is the validator rule so that's the that's the ensemble of uh of the supply chain now now your question here was you know where does centralization emerge in all of this um the risk of mev is initially that bot operators who were very good at extracting these kinds

of opportunities would partner with with miners to get some exclusive access to to block space right exclusive access to the hash rate um and be able to um to to extract uh the mev and get transactions included um using sort of uh one-on-one deals as opposed to an open market for uh for that block space um what would happen in that case is these miners or these integrated miner me researchers would get super linear rewards relative to their hash rate so they would receive more rewards than an other miner would for the same amount of hash rate

this translates of course in the proof of stake world by receiving more reward per stake than another another staker over time this allows them to compound the reward you know take a larger share of of the network so that's the biggest source and the biggest worry with regards to centralization in the supply chain is like seeing some validator who's able to accrue uh an outsized share of the network um using uh mev extraction techniques um this is not to say that this is the only place where centralization emerges maybe we can get into that i think every single step in the supply chain also has a risk of centralization

so we see the like okay you you just gave a little bit of background about lido but maybe before we go kind of into the meat of me uh i don't know if that's premature but like how do you think about uh the significance of med for lido so you can think of lido as a like aggregate protocol for staking people like trust heather to lido to be delegated on multiple uh not operators we have i think 30 right now so they

they want us to to distribute this uh either to stake on multiple operators these operators run the nodes this like this is the validators that actually run the nodes make the blocks my catchy stations and stuff and lido acts as a um as a like as an engine of of a staker to make a good distribution the one that gives a good apr that one that gives low risks that one that prevents the centralizations taken there so the question for lido is what is the best outcome for our stickers like how do we do the best

outcome for stickers here because we are the agent of stakers and as a protocol and uh the the protocol that um uh should be aligned with status interest if we want to have any success at all uh it's pretty clear that what people want for like just taken is for if we remove mvb from from the caution is a good distribution of stake so the stake is not centralized with one entity or something and low risks and decent api with the mvv that gets uh and that's not

controversial like there is nothing to there there is very little uh place to to go for something else like every liquid sticking protocol is designed exactly the same way they optimize for guest best operator set low risk and good api they take different places on like pareto frontier for this but uh the the the direction is the same with mvv it's not so simple because we need to balance uh now apr not being harmful to the network and to the price of token and like centralization risks which are much less

clear like what is the best course of action here is much less clear i'm on the same page personally with stefan here on how the this stuff will play out there is a significant risk of centralization for making blocks and that's why i think uh that we as lido should refrain from ever ever making blocks ever like lido doesn't make blocks um otherwise slider is a risk because if we start actually participating in like maybe extraction actively like by calling the shots of how it should be done uh preventing people from participating

in it like taking some exclusivity here uh it's very slippery slope uh on prioritizing api extraction over like mining extraction over long-term health of the network and that's not good for stakers they want network to be valuable and to be valuable it should be decentralized because like we that that's the value of the um like that's why the um is is great in the use and like because it's uh has a track record of being uh trustless and neutral and decentralized so lido has to keep that cool actually that's uh

it was kind of a question we had for later on but like i think let's just get to that right now because that we were also yeah so for example the the lido stands then would be that node operators uh can choose how to handle mev themselves or or or do you see that like lido would basically have a policy that you know all of the entities operating nodes they have to use something like flashbots so yeah um so we are in process of making this policy at this moment uh everyone can participate in it it's

uh public on forum basically um i really love people to wait in here because i think it's very important right now there hasn't been a lot of uh feedback from uh like anyone from uh that is not like oh not operator of flashbot basically uh from flashbacks basically and i like i would appreciate some some more feedback from larger community here uh what we are thinking right now is that lido has to participate in the mv because like it's the right thing to do if if we don't uh there we will have two negative consequence one is that like we are not competitive versus uh

stake status taking options that participate in me and the other is that like leaving media on the table is not actually good for for the ecosystem uh it it will lead to like again to centralization we like that's the same uh i think that flashbots also had been uh saying for all this time like that's healthy way to do as a validator as long as you don't do actually uh harmful things like blocked organizations etc so that's one thing the other thing is uh uh lido doesn't make blocks like light or dispute stakes to operators who make blocks and uh

we want to keep it this way like we want to don't we do not have a lot of influence here like we want the blocks be made by uh by external entities that are not lido or validators that's like not operators that run with lido uh because otherwise we introduce a significant centralization risk here and we want to select the best technical solution that allows us to do that in a way that like is uh is aligned with what we think is uh the best way uh for mev extraction at this moment and there is uh like

one option right now uh that is like i'd say like maybe 0.8 options right now because uh if boost is is not battle test yet of course because there hasn't been no merge yet and it fully works with us like mvv boost is uh proposal blocks proposal builder separation on the chip basically and it prepares us for the future that will be uh uh in the room probably in like in five years or so and it allows slider not to make blocks but but participate in the mvv and that democratized access to me so like it it's fully working uh

with what we think should be our policy it's not finalized though that's what we are thinking that's what we are putting on the uh on the governance forum if you think we are on uh please make sure uh we know it thanks so i guess a lot a lot to unpack here uh maybe we can go back a little bit into you know how mev boost actually works so stefan you touched upon it a little bit that it separates the block building from the validation part and as far as i understand from vazili basically the validation layer is um kind of lido's part where it stays distributed and then the block building part

is what flashbots or mbv boost delivers now can you talk a little bit about again how does this work and maybe how does flashboards actually decentralize the the block building or like are there multiple people that can run this block builder and um yeah how does it solve this this issue okay cool let's get into it um so i think the way facility framed meth boost is is a great way to think about it it's it's pbs lite so pbs is or you know we could refer to it as enshrine pbs is one of the the road map items for

ethereum of creating the separation between the entity that's in charge of taking transactions and inserting them into blocks which is sort of a highly specialized activity of muv searching of scalable simulation of merging algorithms and optimizing the value of block separating that from the role of the validator which we want to be a role that's as simple as possible of proposing a block to the chain validating that the block that someone else proposed is valid and sort of you know contributing votes to to uh to to keeping their chain moving forward um so pbs aims to separate these two to avoid this this uh this problem of

centralization at at the validator level however there's uh there's some challenges to getting and trying pbs in in the protocol um it's still at the research phase so there's some designs for how to deal with uh with the separation using sort of this commit reveal system over multiple blocks um as well as doing some modification to the the fork choice rule but those have not been sort of implemented or or um or fully researched yet so it's on the roadmap not ready yet the question is well there will be mev at the merge right there will be mev on proof of stake ethereum so what is the

mev solution that we want to to use at that point do we want to continue using the same solution as we have today on ethereum or is there a way to go one step beyond and implement something that's closer to to to full and shrine pbs mev boost is the closest thing we could design to enshrine pbs without doing any consensus modifications so it attempts at reusing all of the uh design objectives of um of entry and pbs uh while uh uh having this sort of set up outside of of consensus um uh to to achieve them um you know practically

what what what is mev boost it's it's a side car it's a it's a piece of software that any uh any node operator can run as part of their client so right now right node operators the full stack is sort of three different pieces of software there's the validator client the beacon client and the execution client now what mv boost is it's a plug-in for the uh the beacon node for the beacon client that um that allows it to query for for outsource execution um mev boost connects to a network of uh of block builders to uh

to request uh blocks from them and then it selects whichever is the one that that produces the the highest value for uh for the the staker and just uh asking on this so you said like okay it connects to a network of block builders so then like how is that some sort of gossip network or like how does that network function so to start um the way that it works is the node operator defines a set of relays that it accepts blocks from so what a relay is it's a sort of a trusted counterparty that a node operator says

i uh i trust you to aggregate blocks from a bunch of different block builders to measure the value of it and verify that this block is accurate pays me and then custody this block until it is ready to be revealed to the network so the node operator connects to these relayers and then these relayers aggregate blocks from from several different block builders in the future the goal is to transition that to being using a gossip protocol so it is possible and there's sort of specs for this it's not currently doing this because using the gossip protocol requires additional

changes to the gossip network of of the clients and the goal for the initial launch is to keep things as as simple as possible so you know one one thing that's important to note here right facilities said you know mev boost is like 0.8 of a of a solution the reason for that is you know mvv boost is being developed side by side with uh ethereum and and proof-of-stake ethereum um and so what will be launched at uh at the merge is only sort of step one of the mev boost road map and it will continue to sort of iterate and ship new features

that bring it closer and closer to uh to full pbs and so this gossiping is one of those features so yeah maybe like a interesting question now the way you put it is that essentially like flashbots is this intermediate solution until full pbs is there which is like a protocol feature but does that mean like once pbs is there flashbots becomes kind of irrelevant does it replace this and and or like what's the vision for flashboards at that point are there like other things in the works that flashbots would work on or problems that you see that aren't fully addressed by full pbs maybe

there are um so so we see the relationship between block builders and validators as only being one piece of the of the mev puzzle right it's only a subset of the mev problem um just and you know the way to think about it is the supply chain this is only one of the relationships in the mav supply chain that we that we discussed earlier um so once this relationship is sort of embedded into the protocol um then that's great right like flashboss doesn't need to do or build anything there anymore we'll continue to provide research uh but there's no um there's no product to operate specifically right

what intranpbs removes from mev boost is the concept of these relayers um so you know there would be no longer a need for for flash bots to operate a relay once uh once i tried pbs is there however the relationships aren't figured out yet is the relationship between applications and searchers and between searchers and builders so you know we've we've posited uh two different futures for mev right the dystopic future and the utopic future the dystopic one is is one that we call centralized block building right it's it's a future where there's a very limited few actors out there that are able to

build the most profitable blocks across all different blockchains and they become sort of the gatekeepers of of blockchains because they all the transactions sort of go through them if they wish to be extra included in the chain and then the more utopic future is is one of decentralized block building so there's block building is much more commoditized and there's a wider set of actors that are able to do it this grants you better censorship resistance for example and this is you know this distinction between centralized block building and decentralized block building is not one that uh that pbs solves um so you know the focus for for flashbots once once pbs is fully enshrined is how do we

figure out how to build a decentralized block building for for ethereum and but for other trains as well so you mentioned this relayer right so basically i guess the issue is right that you said okay uh the the relay of custodies the block uh until it's revealed later because i guess if you give the block directly to the miner then you know they could basically it's the reason they could just like uh make the same transactions themselves and sort of not use the the work from the searchers

and so now of course but then at the same time that relay or of course could do something similar right and that oh they know their content of the block and now they're trying to basically sort of like you know front on what's there uh and and then is the idea that flashbot runs this relayer or like how can you talk a little bit more about sort of the the trust assumptions that like exist right now and the sort of risks that come with that yeah for sure um so so the idea behind pbs is that a block builder is able to construct a block uh create a proof that says this block is worth x dollars so if or x eth right

and if i submit that over to a validator and the validator signs the block header and includes in the chain they are guaranteed to receive that uh that x amount of eth regardless of what the block does even if the block is not a valid block they still receive this uh this value that's that's the design behind uh and trying pbs um what uh my boost does is because we don't have the ability to make those unconditional payments um at the at the consensus level yet it uses this relay to accept the block score it determine how much value it's it's

paying to the to the validator withhold the block body only submit the header with a promise that says hey you know this block includes x amount of eth if you sign it then it will pay you um and then the the validator signs it the block is then revealed to the network and and the payment goes through so this this relay replaces um replaces these sort of consensus level guarantees with trust guarantees the block builder trusts the relay that the relay is not going to leak the information about the block construction and is going to accurately score and prioritize their blocks and then on the other side the validator trusts the relay that the relay is not

lying about the uh the value of the block that it's producing it's not lying about the fact that the block is valid and is going to reveal the block to the network when when it comes to time the really important part about this is because this relay role is trusted any misbehavior from the relay can sort of be identified by the network right so other relays in the network or any of the validators in the network can notice if a relay starts misbehaving and disconnect from that relay um so it's not just that you know they're a fully trusted third party that's like unaudible no like the behavior of the relay becomes uh auditable and it's easy to say okay this really is not behaving

anymore we'll no longer be accepting blocks from uh from this relay and this is the the key uh the key property that makes this the system work is that while the relay is trusted on a single block basis over multiple different blocks any sort of misbehavior is attributable and therefore you can you can recover the system to to to default operation mode you had a question about will will flashballs operate a relay so yes flash plus will operate both a block builder and um and a relay this is mostly for bootstrapping reasons for bootstrapping the network ultimately what we see is there being many many different types of block builders

emerging and competing on on producing the most valuable block likewise we see emerging many different relays competing on providing you know the lowest latency highest availability services for for block builders and node operators so you you mentioned right that the misbehavior of the relay can be kind of detected by the rest of the network is this also true for i guess node operators or the other parties and and what would happen to them if let's say node operator starts to not follow the convention or like build a block himself will he be like excluded from the relay or

how how would that work in practice so really there this is designed in such a way that there isn't really a way for um for node operators to misbehave um the the reason why you need this is because you want to be able to support solo stakers right to be able to support 100 of of the network you need to be able to to support node operators are completely untrusted and so there isn't really a way for for node operators to abuse and and grief the system um what you can uh uh what what the where the misbehavior sort of occurs or has the potential to occur is is on the relay um so there's three specific

ways in which a relay can misbehave with regards to their relationship with the node operators they can lie about the payment that um that they are making for the inclusion of this block they can produce a block that's just invalid right it has a transaction in there that violates some consensus rules and therefore the bloc just gets thrown out or it can withhold the block so it can just not reveal the block to the network when when it has to so each of these different misbehaviors have specific mitigations that mean that that node operators can can observe and identify uh when when these take place um and then sort of terminate the relationship with that with that relay

maybe a question to vasili kind of on on this now you kind of mentioned that the payment is received once a blog is revealed to the node operators now if there were a lido policy kind of how would that go to the would that kind of accrue to the sde folders somehow uh can you kind of explain your thoughts on that uh yeah we we've got uh already an upgrade out like literally is ready for uh for the merge and like after merge things we expect operators to set up the payment address as light address the other accrued on this address

will be staked like in proportionally accrued to stick tether holders so it's a bit of a bump on apr for for other whole stick data holders so proportionally mean like the same kind of uh commission structure that applies to the main if staking rewards will also apply to them ev yeah same same same same fee structure so like uh out of uh 100 percent of amy rewards like 90 goes to stickers and 5 to note operators and 5 percent to leiden

it has one fun side effect of automatically smoothing the mvv rewards so in light operators refuse get the average mev rewards per day basically and in proportion to how many nodes they're on so there is no outsized winners and like and losers in the uh on mev extraction like it's like all smoothed out um yeah thanks that's helpful maybe we can talk a little bit about sort of the

the markets here right because you mentioned a bunch of different parties right so you basically have you know the person making the transaction and there is like sort of the application let's say a wallet or something similar then there is three layer the block builder i guess and then the the validator uh i guess in the lightest case there's even you know like the dao as like an additional party that like uh that's in there but can we talk a little bit about you know what what are what are the sort of markets that exist and where do you see the

the economic value that derives from value and cost right that derives from mev where do you see that accrue and i guess flashbacks is another party that's uh in there right yeah so so we sort of see ourselves as as the game designers here right like our goal is to design a game that is um incentive aligned with uh the you know the interest of all the the stakeholders in the system and also aligned with the objective of the of the chain that that we're designing a game for um and what we sort of see as being a healthy game with regards to the supply

chain is one in which the value accrues to um to the edges so it accrues to the user who's um who's trying to to interact with the chain and to the uh the stakers at the other end who are ultimately providing the the capital and the security for um for the chain um so what uh what a healthy supply chain looks like is a lot of competition at each intermediary step between the users and the stakers such that you know each role takes a minimum minimum fee a minimal rent extraction the

you know the market should be the one that that sets exactly how how that that how big that value is but in a competitive market it should trend towards um towards zero right or like as close as as possible to it so really that's that's that's how we see the the the utopic game evolving um now economic incentives are such that any actor in the system want to accrue outsize rent for themselves they want to monopolize and reduce the competition for their role to be able to increase the percentage of the economic benefit that they can capture themselves and

and that is the the key metric to look out for how much of the value is being accrued to uh to the edges to users and stakers versus to to intermediaries who are who are collecting rent do you have any like indication or like how much that currently is or is there any kind of looking into that form on the mbv side and on proof work or do you think it will improve with proof of stake and an mvv boost versus proof-of-work it's like very difficult to quantify um so there's um there's a few goals with with

flashbots so first is to democratize access to mev prevent it from becoming sort of these exclusive games this also means like maximize competition the second one is provide transparency so make sure that in all these markets the price discovery mechanism is transparent and this allows everyone to be well educated about what happens in the system it's not a black box it's like a very transparent box and then the third one is redis reading the benefits so so that is the equivalent of saying okay the value gets pushed to to the edges to the users and to um to the to the stakers um now i think where we are at in the the sophistication of the system is still very early i'd say right now the

majority of the value is sort of accruing to bot operators and and miners um without flashbacks there have been a lot more towards about operators you know the way that the flashback system works today uh with the auction the majority of the value gets sent to to the miners i do think there is still a missing a lot of missing innovation with regards to mev and specifically with regards to application design wallet design to help expose less mev to the network so help users express their preferences in a way that doesn't expose as much value to bot operators and stakers this is sort of the the frontier of of mev that that

hasn't yet been uh been invaded on or explored by um by our industry one thing i'm curious about uh so let's say i'm i'm somebody who's uh i want to make a trade on unisrop and i you know basically put this in and and you know put some slippage tolerance in and of course the the kind of uh did this you know typical um meb use case that people talk about a lot is this sandwiching idea right we're like okay somebody then goes and like

basically moves the price like against me to the worst possible price that i still accept and then my trade goes through and then they move it back and basically i i mean my trade got through right i was okay with that but i also got sort of like you know the worst possible price that i was okay with and of course what flashbacks now done right is that it's kind of creating this market around this okay who can do this thing and move some of the profits to the miners i mean another another another way one could think about this potentially is i as a user you know i want to put

this trade in you know with some slippage tolerance and uh i'm creating sort of like an opportunity to be sandwiched so wouldn't it be nice if i can auction it off and then somebody can basically buy for me sort of the the ability or right to sandwich this and then i get paid for that and then again that should actually kind of go to the place where you know i get most of the sandwich opportunity back and i don't have to worry too much about like you know what slippage to put in and what parameters and because right now you maybe want to put a narrow slippage so that like i don't get sanity so much but then maybe my transaction fails and pay too much gas

so i'm curious is do you see in the future is is that kind of a scenario too where like maybe users or i don't know if that would be the applications to users use kind of create some auction to like sell the med opportunity that's intrinsic in transactions um yes this is like frontier right mev um system design is the question is how do we deal with with order flow so maybe i'll i'll try to frame frame the problem in a way that ties back into the discussion that we've had so far so one of the one of the areas of research that that i mentioned we're

looking at is block builder centralization right and what are the different different sources of this this centralization the biggest one that we see is order flow so as a block builder i'm sort of incentivized to try to aggregate transactions uh and and order flow from sources that are exclusive to me if i'm able to you know get all of uniswap transactions to be sent exclusively to my block builder it's pretty much guaranteed that my block builder is going to be the most valuable right or is going to be able to produce the most valuable blocks and therefore there's there'll be very limited competition uh for uh for for this block building um so order flow and access to exclusive

order flow becomes sort of a a very difficult uh or a very prominent centralization risk when when it comes to um to block building and is why it's it's it's a problem that's that's super important to to solve so what does solving the order flow problem look like um i think it has the same properties that um that i've discussed previously right you want some kind of marketplace that is democratic that is able for anyone to start accessing this order flow if they want to you want it to have a transparent price discovery and then you want it to redistribute the value to um to the end users there isn't yet a generic design for an order flow

marketplace that i've seen um that sort of fulfills all uh three of uh of these requirements a lot of the designs that have been explored so far sort of focused on um providing some permissioned access uh to order flow and sort of compromise on uh on on the democratic side of it um or um or some other uh aspects of it the other thing that i want to mention is um there is sort of some some pretty fundamental challenges when it comes to um to uh to thinking of mev as a fixed value so really the best mitigation for mev is not to sell the mev to someone who can

extract it it's to create applications that don't expose as much nav that's always going to be the best way to to protect a user and make sure that the user gets the most amount of value back i think applications that explore this are ones that realize actually slippage is equivalent to increasing the gas price on a transaction right so as a user if i set sort of a gas price of zero when i send my transaction and set a very high slippage it's equivalent to setting a slippage of zero and setting a really high gas price the reason why is because it allows for for basically prioritizing the execution

or incentivizing bot operators to move the market in such a way that my transaction ends up getting getting included so you know the solutions out there that are looking at this i think are doing so in an interesting way they're playing with the fact that these two different uh values actually mean the same thing to try to come up with an expression of of user preferences that gives them execution uh within sort of a reasonable amount of time and a reasonable price um now you know we've been talking about about swapping as a use case here but this exact same sort of uh preference expression question exists with

a lot of different a lot of different applications cool i i guess one thing that seems so somewhat uh related to this topic and i'm curious about it so so a product that you guys also announced uh some while ago was flashback protect where i think a user can send a transaction sort of to this you know private mempool it doesn't go into normal mempool that's sort of managed by flashbots and then uh you know can kind of kind of like prevent being front run uh is is this something that's gonna exist in you know after the merge as well and

like what do you see as the role of this in the future yeah it'll it'll exist after the merge as well um so what flashbacks protect does as you mentioned it allows you to route transactions directly to uh today to a minor um right um in the proof of stake world what it does it allows you to route the transactions directly to a block builder um without sort of going through the public uh transaction pool now there's there's some trade-offs involved in this right the the public transaction pool the the beauty of it is it provides um provides some good censorship resistance guarantees because it is distributed to as many different parties as possible

and and in the public so transaction is replicated across the entire ethereum network when you route these transactions through a system like flash boss protect you are hiding it from from bot operators that can for example sandwich it but you're also sort of only routing it to uh to a single actor for them to um to include um on the chain um so flashballs protect as as a product is sort of trying to target this relationship between applications and uh searchers or block builders and you know the future roadmap for that is to increase the the decentralization of it figure out how to route these

transactions to uh to more than just a single block builder and a single searcher and allow for sort of users to specify preferences over how their transactions can be used right we wanted to like also talk a little bit about going beyond ethereum in a way so right like the road map of ethereum um contains a lot this this strategy of having like l2s or like roll ups to scale and of course by having that right like some of the defy activity or i guess in general activity will move um from the l1 to the l2s um and with that the mev opportunities too

and will kind of spread out around us how is flashbots kind of thinking about this space is there are you working on solutions to kind of address like let's say crosstrain mev uh or you know how what is the impact of this in general on the in the amerispace so cross chain mev i think we haven't seen really people exploring this this a lot uh except in the most uh like naive and simple way by arbitrage and the uh the dex prices between chains and even that had been sporadic and slow

i'm for apart from being like in lido i'm also working p2p value data we run some bridges invalid data and we see it like first hand cross chain maybe is slow and efficient these days i think there is a real uh real chance that with the like multi-chain future and a lot of economic activities between between chains uh and like not like if we we should count also exchanges um in in this equation like it's not a chain but it's also a place where economic activity or on blockchain

happens the analog to hft traders uh will emerge that will have like multiple uh extremities on different blockchains like one on like on either one on salon one on finance one on coinbase or whatever and uh they exploit the the different economic state and different uh rates uh rate of updates and like the basically uh border uh border related movie and the best of them will have outsized returns compared to the worst

just like it is with hft and [Music] given that a lot of this economic space are centralized like it changes a lot of these success will depend on good relationship with them with the exchanges and like being able to collocate being able to uh get the best deal flow like transaction flow [Music] and that will lead to like a few hyper competent and well positioned firms to have to have an outsized advantage um i'm not quite sure what what to do with it right now it's not the case

there is not like that much money in it and there is not that much uh like it's it's not super active right now how do you how would you know even right because if you think of like arbitrage where like okay you know i'm selling something on unit swap and then i'm making some content rate on i don't know ftx or finance like we wouldn't even you wouldn't even able to see if that's happening right now or not no i i like it's it's it's not that hard to uh to to read on chain basically like if we take a look at the dream like that's uh pretty visible the the the arbitrage

trading bots are pretty visible they they do like if you take accounts that have three hundred thousand transaction that's basically them that's arbitrage trading bots like you you can see how they trade they you you can like you can't really see but you can like more or less estimate how much they make how competitive they are et cetera et cetera it's like the blockchain is too transparent you can't really uh find out like cross cross exchange arbitrage uh how it works but um so like what we what like what we see uh from like very competent firm engaging this version of the future is that a lot of

professional market makers uh like not not uh some of the professional market makes actually actively like explore this by running arbitrage bots by building cross-chain solutions like wormhole for example by building like price feeds uh et cetera et cetera by getting ties with wallets to get transaction flow that's a conjunction but i think it's happening um i haven't seen this public yet but like i'm pretty sure that it's going to happen so there is uh an unexplored space which people are preparing different like different ways and uh it's not that big right now so we

don't like i i o at least i can't draw the line from now to like the future that will be there it's not fully understood by me all right i agree with almost everything i agree with the analysis i disagree that it's not big already i do think it's already very big in fact i think it's probably one of the biggest sources of mev today is arbitrage between centralized exchange and uh decentralized exchange um you know over i think sixty percent of the volume on unit swap comes from these arbitrage bots they're just taking the price that's set on binance coinbase wherever else and then reflecting that into the price of of of the amms and so a lot a lot of the trading

activity on the percentage-wise on d5 comes from this sort of cross-domain mev i think what what is small right maybe in this is this is what facility was referring to is specifically between two different blockchains um so because there is like limited amount of price setting between uh on blockchains themselves there's limited amount of sort of user interactions that are actually doing the volume on these chains the arbitrage opportunity doesn't sit necessarily between two different blockchains it sits always between like a blockchain and um and a centralized

exchange um i maybe i can i can expand on on the problem of um of cross domain mev so it is probably the largest unsolved med problem alongside with the the order flow issue um we've sort of done some some research and alex obadia on on our team has has published this paper on um on cross domain mev that attempts to formalize some of the uh the properties of uh of this and the consequences that that it has for um for blockchains um basically the um the the takeaway from the research is the existence of cross-domain mev means

that there is a centralization pressure that spreads from a centralized system that has a lot of economic activity to a more decentralized system that has a bit less economic activity that's because anyone who has an advantage on the centralized system will have sort of outsized ability to to capture opportunities and can then execute that power to get more uh influence on on a decentralized system let's take an example here so we have a binance or ftx that uses sort of fee tiers right for um for for giving rebates to to market makers and arbitragers so a large market maker

on ftx or binance that's able to approach them and get some special deal that says i'm going to pay a lot less fees to the exchange because i bring a lot of volume sort of has an entrenched advantage in this centralized domain that means that they are able to out-compete anyone else that's doing this arbitrage between centralized exchange and decentralized exchange this means that they are going to be able to successfully capture the arbitrage opportunities when it comes to executing on on ethereum or or the other chains they can then use this ability to like reliably uh bid more and win the opportunities to sort of start producing blocks to start integrating up and down the sort of supply chain on on these uh

these distributed uh more decentralized systems so actually right the the existence of cross domain mev and the fact that um that exchanges offer fee tiers that are that are very exclusive and not transparent sort of leads to more centralization in the mev supply chain on uh on ethereum and and other chains themselves yeah that's a great point uh i did not sufficiently i think kind of make that link before but yeah that makes perfect sense maybe stefan the the l2 thing so yeah we talked about

this now the cross chain thing and especially as it relates to uh centralized exchanges which i think is a great point but what about l2s they're roll-ups what is the impact they have so um layered two's roll ups um you know in the cross domain mev framing are different domains um so you know there are different execution environments that carry state and and have sort of uh the the problem to solve of how do i order transactions internally and then how does you know the finalization of these transactions relate to the finalization on um on other on other domains now one way to frame what is the

solution to cross domain mev um the the solution is how do you minimize the um the coordination cost between validators on different domains to provide finalization without it being the same logical entity right so let's say that you have a centralized sequencer on a layer 2 and the sequencer decides to begin doing uh mev extraction right they capture one hundred percent of of the mev because they control 100 of the block space um what they can then start offering is saying okay i'm going to start buying validators on ethereum perhaps right and

start controlling more of the block space on ethereum and then offer some neat services that says hey i am able to offer atomic execution across ethereum and this layer too because i control 100 of this block the blocks on this layer too and i control some of the blocks on ethereum whenever i have a block that matches between the two i'll say here's an atomic execution between the two what you want to solve in the cross domain setting is say you can offer this with 100 of the validators on ethereum and any of the layer twos in a way that these entities don't need to be the um the same entity they can be distinct entities that just

coordinate with minimum cost at being able to provide this this service so this is sort of on a green field sort of research still um and flashbots is looking at it and prototyping some things but there isn't a sort of an active solution that that provides this kind of service yet and i guess yeah maybe what's interesting right is that obviously lido is um an entity that is also on the different chain so even if we think about okay of course the validators themselves or node operators are operating those on the different chains but lido as a like aggregator is also on the different things could is lido playing a role in this

system or could lido kind of take part in in allowing to do that to happen is that um something you you are exploring uh so right now we we don't do anything about that uh our light on different uh blockchains is self-contained mostly uh it has the like the overall brand and uh but no and some technical and expertise and stuff like that but there is no like overarching lidar that does things cross chain um between each other i think uh

uh lidar being big on like multiple blockchains is pretty good set up to make the uh cross-chain mvv uh less centralized because like with our domain being explicitly a stake distribution and liquid staking and not immediate extraction we can and we should and that's like mostly follows through from all policies that are not finished yet but that's what they are the the policy drafts basically uh it follows through that we would want to have a decentralized market for cross chain maybe just like much the

same way we we want this market to exist for like in in blockchain as well being focused on one thing will allow us to uh make the like mev stuff less less contained in like one big operator or something and i i guess when it comes to flashbacks i mean right now uh so far you've been fully focused or you you've you know just put out some products for ethereum seem to be very theorem focused but do you also think of building you know flash you know basically solutions for other blockchains yes uh thanks sir thanks for asking so um

the way that we've been thinking is first we have finite attention and resources so we can't build solutions for every chain um and so we have to pick sort of one place that we want to to experiment and and and build these things we pick ethereum because it's the most active smart contract uh blockchain um and it's also the most decentralized and so if there's solutions to be built that will help decentralization it's the right place to um to test them out and our hope is that these abstractions can then be ported over to these other chains but porting over these abstractions is not um is not an easy job there's a lot of intricacies

and design details that go with each different chain and that mean that the solution will have to look um will have to look somewhat different so our approach to to you know cross domain is um to work with uh and collaborate with partners that are more integrated into these other ecosystems um that share the same the shame beliefs and ideologies around how a healthy mev ecosystem looks like and we want to support them in sort of integrating into this this cross-domain mev future in a way that um that provides sort of a similar experience um yeah the other place that we're looking

for for collaboration with partners is is on this um this order flow challenge so both the cross domain and the order flow um problems of mev are ones where i think um will end up being much bigger than flashbacks and solving and we want to collaborate with with third parties so maybe another topic uh i did want to touch on so the the focus of flashbacks right is as we talked about you know making amv transparent making it democratized like kind of spreading the value to the maybe at the end but of course another a very different sort of

philosophical approach maybe can also be oh can we basically try to remove maybe can we try to minimize enemy uh you know for example you have you definitely have some application specific chains like let's say osmosis you know that's like thinking very much uh of doing you know going in that direction i know there's also another kind of cosmos related project and noma that's like kind of thinking of stuff like that i think i also saw there's something on ethereum on east research right this shutterized beacon chain or like some idea that like so what are your thoughts on that do you think like

minimizing mev is like should be purely the sort of the problem of the application developer or like you know the smart contracts that are being written the the amm or does it make sense for a blockchain to try to you know kind of um actively minimize mev it's a it's a difficult question to answer right and it is sort of the um the perennial question of mmv how do we deal with it what is the like right way to approach the the problem um i think hassu gives a really good presentation on this which is called why your blockchain needs an mev solution um and

and in that presentation he looks at all these different approaches to mitigating um mev right from random ordering to first in first out ordering to uh try to do like threshold encryption and sort of obfuscation and identifies what are the externalities of those solutions i think that's really how blockchain developers right need to think about mev is what what are the characteristics of the system that they want to build what is the nature of the the mev that that they expose and what are the externalities that they are willing to tolerate um are they okay with their chain getting spammed

whenever there's like an mev opportunity that occurs or is that too much of a negative impact on user experience or does it like have the risk of ddosing the chain down completely like we've seen on on several chains happen and so you know maybe that's not the best design to follow um are they okay with uh having sort of co-location be heavily incentivized which obviously leads to then more validator centralization right so a lot of these first in first out systems uh that's that's the the end result of those systems is you incentivize but operators to be tightly co-located with um with the validators to be able to be

the first to extract the the opportunity and so it creates more centralization at the validator level maybe for some chains this is okay maybe for others it isn't um so it is it is about thinking it is about thinking about trade-offs uh figuring out what are the properties that the the developer wants their chain to have and then select the solutions that um that support that i think you know looking just at the the ethereum ecosystem the best way for ethereum to mitigate um sort of mev and and to reduce it is for application developers to put more effort and time into thinking about how their applications expose muv to think about

you know what service are they providing to their users and is sort of what they're doing sort of abusive in some ways towards um to their users i think there's a big risk here because a lot of um i think applications on wallets are seeing mev as this sort of hidden revenue generation model right um so the way the way that you know wallets can start thinking about media say okay i have all this transaction flow from users that has some value i'm able to sell this transaction flow to um to some bot operators and and receive some some fee for this that now gives me a way to monetize

but that is that is very very dangerous because um you know it's not transparent to the user how much uh value is being captured from them uh by the wall by the service um it's not just like a transparent fee that the the user is is paying it's uh it's a hidden fee that could be arbitrarily large and can grow over time without without the user knowing yeah i think that was a super interesting discussion i think i wanted to like kind of take it back to like the practical timeline dimension vasily you mentioned lido is ready uh for the merge maybe we can talk a little bit about you know where is

mev boost right now um you know how are the preparations for the merge going and maybe here's some estimates of when the merge is happening from you guys uh yeah so like lido is merged ready uh whether the mvv boost on israeli or not so like it's uh it doesn't uh have uh mvv boost as a critical dependency we want to to start running with it it's not like if merch was happening today it wouldn't be possible for i think

like a decent part of our operators because it's not supported by all the clients yet it's not uh like if it was happening like tomorrow uh we would have to uh run a lot of stake in light or without maybe boost whether we wanted to not because it's not fully ready yet we don't need to be fully ready yet to to be much ready we can just accept the vanilla algorithms of [Music] uh for the prioritization transaction fees uh the fees auction which is not like exactly great but works for us and like works for idiom and like we used to

run like multiple years on that without flashbots and it wasn't great but it was like it was working so if we have to fall back to that we can so that's what i do it the other side is we run test nets we ask operators to run test nets we ask operators to uh to be merged ready to uh to try out mvv boost where whether they can and the policy draft we have is basically stating that i mean boost is important um the blockers are like the general raw ish state of amiboost

the not having support from all the clients yet for me boost and the relay list that we we have to to start with relay can be a malicious thing we can't do permissions work with relays in lido we have to create the list and that also has to happen before we can adopt uh boostfully no yeah i think yeah stefan if you have anything to add on on your end yeah i think the the important thing to to mention here is um you know i mean v boost is a piece of

software that's designed to work with an extremely large set of stakeholders right and we have all the different client teams that need to integrate it there's all these different node operators everywhere from single solo stakers to to to large node operators service providers to the staking pools and decentralized taking polls they'll have different sort of requirements for how this how the system works so what we've done is we've put together this page called boost.flashbots.net i mean it's the best place to keep track of of the readiness of all these different actors uh or stakeholders with regards to the mev boost software um it's also

the best ways to keep track of sort of active development uh of mov boost so you know if the merge was tomorrow it would be possible to um to run mev boost it just wouldn't necessarily have all the features that all of these stakeholders care about and so there's sort of a development roadmap for me boost that that includes iteratively uh shipping more of these um more of these features that will make it uh better suited for for the needs of all the stakeholders and all those updates are posted on that on that landing page so yeah it is a it is a work in progress just as ethereum is a work in progress and so um over

over time it becomes um it becomes more and more feature complete um but it is uh sort of ready to uh to operate um at the merch cool well thanks so much uh guys for joining us today it was really i really enjoyed this discussion i think it was super helpful to understand uh you know where things are going it's obviously an area of enormous complexity and we're just seeing like the complex is only getting bigger when we talk about you know these other parts of the med supply chain and having different markets there and here so i think this will be a topic where we can continue doing podcasts on for many years to come and uh yeah thanks so much stefan and

vasily for joining us today amazing thank you so much this is a great discussion thank you and uh thanks so much for listeners for tuning in if you want to support the show make sure to leave us an itunes review or let us know your comments the episode on twitter youtube and we look forward to being back next week