The State and Future of MEV - Hasu, Vitalik Buterin, Phil Daian, et al.

up next we have our medic panel hosted by hasu and um it is on the state and future of mev and we have here are panelists vitalik phil mahimna giorgios and dan so um hasu anytime when you're ready the stage is yours uh yeah thanks tina um yeah i'm uh super excited to moderate this very stacked uh panel with you guys um why don't you start by introducing yourself um oh sorry start my video okay why don't you start by introducing yourselves um with like one sentence starting with your vitalik yeah um hello i'm i don't know i'm vitalik i make uh posts on youth research from time to time dan i'm dan i'm a research partner at paradigm and i have to mention that both for me and georgios it's a as investment firm but none of what we say is investment advice and these are our own views they're representing not those of the fund hi everyone i'm mahima i'm a phd student at cornell and a lot of my research has been focused on fair ordering protocols okay hi everyone i'm phil um i'm a phd student at cornell also and uh co-founder of flashbots and coin the term nab great and finally you judge us hey everyone i'm georgios i also work at fire them i'm a research partner i do a lot of engineering and research work on mev and have worked with flashbots since the very beginning i'm very excited for this discussion thank you and just in terms of structure please i'd like this to be a fluent conversation so um just feel free to interrupt each other and you know follow up on each other's points that would be great um i start with maybe a controversial thesis so um charlie noyes has argued in his first talk um one of our friends from ferdinand that we have entered the era of mev and it's basically only gonna go up from here but with all the amazing innovations that have already been showcased today and some more that we will see in the next block um could the mev era end much sooner than we expect and you know just feel free to take it i don't have any um no go ahead i was just gonna say that would that would be amazing i'd be so happy if that happened uh and you know flashbacks can pivot to building the fairest uh markets out there um but yeah that that is the goal i think it's one that will will take a long time to truly achieve but uh it's definitely my goal i think there's definitely important types of mev that are not going away no matter how hard we try um like one simple example is like just unit swap price arbitrage right like in between one block and black and plus one there's 12 seconds of market movement on like any exchange that's not on the ethereum blockchain itself and so you know you can making the swap trades off of that like you can make trades on every decks on top of that and so i yeah i don't really see how something like that can like even can see uh even conceivably be removed entirely i mean i do i do expect that any of you will be reduced just because there's a fairly big incentive to reduce it but there like there's also other types that seem to be harder to reduce and like i do think that gains from specialization in being a block builder and is something that's going to continue to continue to exist for in the long term future to enough of an extent that we need to worry about a design around it i agree with that and i think not only are there kinds of mev that can't be eliminated but attempts to eliminate them or naive attempts to do so um often cause greater problems and so uh yeah first rate of a block first rate after a price movement the way that this is resolved in traditional markets is just whoever gets their order executed first and while that seems like the goal of a lot of a lot of anti-muv research or a possible goal um what that means is that there's a massive market in co-location and then just sort of and then these arms races to communicate at as close as possible to the speed of light and and have minimizes as much as possible which is totally wasted and so that's where the mev in traditional markets goes is basically toward research into fiber optic cables and co-location fees new york manhattan real estate yeah i think that's exactly right i think i i fundamentally do agree with you dan and uh you know uh as much as i'm excited about reducing mev i think that was my original research agenda um however i think for a large amount of structural opportunities and a large amount of like market design issues it is necessary and it is kind of fundamental in my opinion i have a blog post about this that people disagree um but i think for that kind of remaining med that we do need democratizing if it's really the key yeah one interesting one interesting anecdote here was back from 2017 back when the very first kind of gas market uh shenanigans started happening with icos um like at the beginning people were trying to get in front of everyone else by raising their gas prices but then some projects like really naively bans that by just banning gas prices above 50 way and what that led to was like instead of uh trying to get ahead of everyone by me by sending one transaction with a thousand way people just got ahead of everyone by sending 20 transactions with 50 way to increase the chance one of them would get in first and so you like you ended up replacing competition with competition but that's even more negative externalities so it's definitely a good idea to try to avoid that yeah and i told them that that would happen when they they asked at one of the dev cons whether that was a solution and i was like no people are just going to spam and then it happened so it was really funny and i think that's a good analogy to also other protocols where people are like oh look there's no leader everything's probabilistic there's no mev here like uh but but oftentimes just because you've made something probabilistic doesn't mean it's still not worth playing i think i feel like there's some kind of mvp that will uh exist in the system always and that can be some kind of good move like arbitrage but i do believe that most of the other muv that comes from like ordering manipulation can be removed and to the point or uh to at least to the point where it's not worth it to go after that mev and i think that's the goal that we should try to go i'd like um to get your view vitalik and also dance on an earlier point uh that you made about sort of arbitrage dex arbitrage never disappearing from ethereum so i i kind of see two counter arguments to that so for one um the share of rfq orders among old exotics seems to be only going up and this is a sort of decentralized exchange approach that does not generate any on-chain mev as far as i know um so first what do you think about that and second um couldn't you in theory design a decentralized exchange that gives the exchange operator um sort of the first look in every block so the exchange operator has an on chain cemented right to be the one who performs the arbitrage and only after that everyone else can trade i guess my answer to that would be like number one like yes that's true but and that might like reduce mev somewhat but it's not going to reduce if i put it by anywhere close to 100 percent and my answer to two was like yes you can and that's great um but that's not any minimization like that that is anybody extraction and then um yeah like like it could be democratization depending on where that um where the revenue for sort of auctioning off the the the other first dibs ends up going through but like that's not something that's in the minimization cap that's just kind of shoveling the mvv around yeah i think i think one example of that is idex so like when i first gave the flash boys talk a lot of people ask me what to do and i said i think my favorite design right now is like just having a centralized sequencer and yes all the mev is still there but like if you're legally obligated or business incentivized or whatever not to extract it uh that's probably practically the best design for users where they get the least exploited but it's also the least decentralized because in some ways you're centralizing that trust uh so there's also like in my opinion a fundamental trade-off between mev and trust assumptions uh that's kind of deficiency also if you want to add the sort of mvp mitigation measures in your system you reduce overhead so if you wanted to lock let's say the first transaction the certain person you introduce an extra load operation an extra state load operation on every transaction for every user which is not ideal if you're trying to mitigate some certain component or exploiting your system okay yeah um moving on um another point that charlie made in the same presentation was he argued that you know time-banded attacks the the fabled time vendors attacks from your paper um will finally become real and quote-unquote short-term meaning wonderful block re-org frequency may increase what does this group think about that i think so far every time i've underestimated um how sophisticated mev would get an mvp extraction tools um and this is well before we invested in flashbacks but every time um i underestimated him charlie would say like no they're gonna do this crazy sci-fi thing and then they started doing the crazy sci-fi thing so i think in this case um yeah i think i think multi-block mv is only a matter of time i'm kind of surprised or or skeptical if it isn't happening already um i think this is a question that we as flashbots should probably answer for this space in terms of our data efforts but we've seen uncle banded attacks we've seen inadvertent uncles and the way they affect uh bots and kind of block producers and arbitragers so since we've seen these and they're benign and like people are okay with them like who's to say that a few of the benign ones weren't actually intentional there's also a post eth2 world where there's a small change in the protocol where instead of having the block producers being randomly elected now you know all the block producers for the next let's say 12 minutes which means that potentially collusion will be more likely there or as a validator maybe i get to randomly assign two or three or many content uh continuous slots um and uh that would make the extraction easier uh vitalik i think that may have some faults here in particular the example use case i have on my mind is just manipulating unison particles you insert the last transaction in the block and then you insert the first transaction in the block and you do this for many blocks and you manipulate the t-walk um i see it um again i guess that that particular attack would be possible and maybe unesco would have to redesign their oracle i guess i don't know i haven't thought about the details of that um there are other things that the completed switch to proof of stake we'll do it though like one of them is that it'll make uh short re-orgs basically impossible right because unlike the current proof of work design where the fourth choice basically has like one actor per block uh the upper mistake for choice has hundreds of actors that contribute to the fourth choice in parallel and so like unless you control close to a majority there just isn't a way that you can revert someone else's block um so like that's one way in which proof of stake will improve things um there is i mean of course predictability is that of uh proposing is definitely an issue um now of course we are adding i'm going to add a single secret leader election at some point and that will probably come with like penalties for revealing ahead of time when you're going to be the proposer uh but if that happens then the yeah like collusion will get harder but it will still be possible for individual proposers to detect when they're going to have two or three blocks uh in a row and and there's going to be things that i guess are possible as a result but like at the same time there is the question of like well how much is possible that just isn't possible in the status quo world where i know sometimes you do just have like one single block for a span of 45 seconds i don't know so i think even with like single secret leader election you might be able to collude off band and prove to someone that you are chosen yeah you can right yeah you can so this is why i suggested adding this uh what i call an anti-pre-revelation game which is like i mean if the people who are including really trust each other that's not going to solve it but that'll still reduce the risk like basically if someone else tells you when they're going to or when they're going to create a block then you can sort of anonymously predict that and if you predict correctly that you can penalize them but we'll well yeah and i think like still it will be possible to get especially for a kind of single pulse to do mvv extraction off of a couple of blocks in a row that definitely will exist phil i don't think it's possible today for a searcher and flashbots to you know propose uh to a minor to perform a reorg attack what would it take for this to become possible um and where do you stand on sort of allowing or disallowing you know the proposal of such blocks i don't know i think hop on the next roast and let's talk about it um i don't want to claim to speak for flashbots here since flashbots is kind of a collective and we we have to exercise the collective in making any decisions um that being said um we have no current plans to support anything like multi-block reorgs um it is an incentive in the system however and like our general philosophy as an organization is that like in order to build systems that are robust uh we should kind of either mitigate or work around any incentives that might have negative consequences to the system uh so that is one of them so a few approaches that could be possible one could be just like letting it burn like trying to incentivize reorgs as much as possible and like gaining robustness from like statistical evidence that like the incentive isn't strong enough for the reorder to happen um i think the big risk there is especially around black swan events uh reorg's become much more attractive another option could be a more active kind of economic kind of metagame design in which uh you you create a system in which reorganic and attacking the system is less profitable than doing the right thing and i think that's more of the direction that we've been thinking in but yes definitely want to consider it in our security model and build tools around it because it is a real possibility you once mentioned to me that you will allow miners or stakers to express preferences for the sort of bundles or block templates that they can buy would this be one of those preferences possibly i mean i think uh it's it's hard to say so i think the unit the the mission of flashbots is to allow anyone to express any programmatic preference uh that is ethical to do so in like a sustainable way so the question is can we allow that preference to be expressed ethically i think if we have like broader medications and we're using that in a principled way as part of a broader system i don't see anything wrong with adding that feature i think just adding it like willy-nilly probably will destabilize the system needlessly and like break social norms so like we don't really want to do that either i see um and i have a last question for you about flashbot so um could you briefly walk us through where you see flashbots in in two years and uh what are the biggest hurdles to getting there um it's really hard to say like i said flashbots is very much like a research collective so we're trying to answer in this next phase we're kind of transitioning from phase zero which was proving the concept proving that there is a need to solve these questions and a strong market need and i think we've done that very kind of well and now we're transitioning into a phase where we're thinking okay now flashbots looks like it's going to be around for a long time what is its role in the ecosystem uh what does the organization uh kind of look like so i think those are all questions that are still being answered we welcome people to participate my personal vision is like uh i would like to build kind of a turing complete market for any kind of uh bot style preference um and and kind of an associated game around that much in the way we've been doing on top of beef today so very much want to keep that momentum going and keep building community and uh also keep doing fundamental research and community events like this one let's see thank you um so what was really interesting to me in the last block was the entire section on fair ordering um so um now you earlier defined um a sort of quote-unquote sent ordering as ordering transactions in the order that they were sent and um let me describe this as fair um in your opinion why does this not uh just succumb to you know the high frequency trading that you know this sort of picks up from you know what dan said earlier but i think it's a very interesting angle to you know to explore so i was thinking about send ordering in sort of an ideal world where you have uh user uh trusted timestamps at the user side and you have some kind of synchronous network so for send order in particular there's no impact of the actual network delay so it doesn't actually become high frequency trading what you might be referring to probably is like receive order uh fairness where you're thinking about sending the transaction and you're timestamping it at the time in which each node is receiving the transaction which technically um has the potential to be similar to high frequency trading if not done correctly so in our um in our paper we distinguish between like two different networks like an external network and an internal network so the external network is this network between the users and the protocol nodes and if you can basically control this entire network so you're this global adversary that can control the access for all users to that protocol notes then you can still arbitrarily reorder how much ever you want but in practice this should not be the case the other point i try to make is other sort of networking layer solutions can be used for that layer in particular if you're ordering more so deals with um once the transactions have been received how do you classify fairness in terms of the ordering that were received in and other sort of network layer solutions can be used for the first stage of sending the transactions from the users to the protocol nodes so you can use things like hiding the um transaction data so you can't front run based on it in that layer or other like orthogonal techniques can you follow up on that guys how do you feel about um fair ordering basically requiring that you call up the set that creates the ordering rather than rather than just having one party in in a way it does not seem to me that you're solving the problem rather you're you're you're increasing the cost to attack by by a constant so it's not actually a constant because you're not corrupting the set it's sort of uh so let's say if you are in like an l2 permission network right so you're taking the uh proposals from most of the um nodes in the network so not like a constant but if you have like n nodes and you're taking them from like two-thirds of the nodes so um you you do require like a substantial more amount of corruption power so it's not just a constant number that you're corrupting it's uh basically almost all nodes that you need to corrupt it seems to me that even if you have perfect send ordering you're just incentivizing an arms race and essentially like processing and responding to to news right you st there's still some kind of race that uh where this value ends up going to wherever runs the fastest one of that and that seems like a wasteful use of resources right so send order as i said like doesn't actually be affected by like the network latency so it's not like not the network latency literally your transaction sending latency like something happens in the world and you react to it your reaction speed reflexes right are are what determines it right right i mean that you can argue like provide some kind of unfairness but i would say it's an open problem to whether that can even be solved and that's something that even exists in any solution that you can find i think i have one thesis which is that applications shouldn't try to have a heartbeat that is faster than the heartbeat of ethereum itself so the 12 second block time and so you know one consequence of that would be that you should use sort of batched auctions for example in the decks you know um and wouldn't this be just fundamentally even more fair than even the fairest ordering solution so i would like to yes that's a good question like i'd like to talk a little bit about batch auctions um so when you're talking about batch auctions you're really what you're saying is there's some l1 ordering service and any transactions that are posted to this contract that or smart contract that does this um amm they will it will take the transactions from the l1 and the transactions that are relevant to it it'll batch them according to how they're posted on l1 but that still hides the fact that the person ordering on l1 can take advantage of their power on l1 so they can exclude transactions they can reorder even arbitrarily far right so batch auctions when you're considering it like a smart contract setting is not really batch auctions because of the fact that or it's not really fair because of the fact that um whoever controls the l1 ordering can control the quote-unquote batches in layer 2. now i would like to draw this comparison between batch auctions and fair ordering where you can think of batch auctions in a centralized fashion so suppose there's only like one server for the exchange and you send it to that server now you can consider a batch auction based on transactions that were received by that server in the in that batch time right and what fair ordering tries to do is tries to distribute this over a network of servers so now you don't have just one server that's receiving transactions but you have n servers that's receiving transactions or potentially like a large global network now the question is how do you uh do batches which is not obvious right you need to change the l1 protocol which is exactly what fair ordering does how do you see the trade-off between um having memphis privacy and fair ordering so when you say memphis privacy doesn't mean hiding particular transactions in the memphis so you know they exist but you don't know what they are or the entire mempool is dark i guess that's a good question unless i understand it um i mean is there is there even a way to make the entire memphis dark right because the the miner can always see their own transaction i don't know i mean um so in the case of when you're adding privacy to the mempo i think there's still problems suppose you um encrypt transactions that are in a mempool and you choose some random ordering within the mempool that still leads to a problem of like a flooding attack where you can just send like thousands of transaction and because the block space is limited you can uh have a higher probability of your transaction being selected in the block right on the other hand like with fair ordering what we're trying to do is um have like a definitive way of ordering transactions that get processed into the network so basically in order to that you can't do the same kind of flooding attack in a fair ordering protocol because you would still need to send those transactions at the same time like in a random ordering protocol uh you can send those like much later than a user transaction you're trying to front run and if you send like thousands of them with high probability uh at least one of them will be like sequence before in a random ordering i see um i have one for georges and then vitalik um so why do you think that fair ordering protocol is often presented as sort of the diametrical opposite to what flashbots is trying to do i mean firstly i don't think anything is is a solution that solves everything so basically i think that instead of having them as diametric opposites we should have them as more like as you did in your talk as more synergistic approaches we use one to minimize the surface where possible and where where whatever remains we try to auction it off in a in another way um compared to the normal one um so honestly that narrative seems a bit misplaced to me and i i would prefer that it was more synergistic than not i would agree with that um i think there's a lot of mischaracterization of uh the flashbots narrative uh which is possibly expected because i think you know mevgeth was a relatively opinionated software release and it had a relatively like widespread effect on the ecosystem and it's a very accelerationist piece of software that being said flashbots does not start and end at mevgeth like has a number of other research streams activities projects uh that i encourage you to take a look at and get involved with and fair ordering has been there since day one so we're definitely interested if anyone wants to work on fair ordering please reach out we're happy to fund that or help you do it in whatever way we can um because i think i personally see it exactly the way georgios does where like if it can be minimized then you get the same effect like why why not do that that's like a no-brainer that being said i do still think it's it's somewhat fundamental which is an open question an open research question as well and just to add on that phil um minimizing does not mean that the total number will not be large like the attack the surface the mav surface is so large that i don't think that any minimization let's say techniques would get us to a point where searchers are not making money or miners are not making enough money from flashbacks bundles yeah yeah i think that's true because fundamentally like it's searchers provide a service to the network like they're the ones that liquidate transactions and do the arbs and like every single flashback transaction is doing some service for adapt otherwise the there wouldn't be any money in it except the ones that are hacking or doing other kind of unexpected things um but like the vast vast majority of them and uh yeah i think the the the bots need to get paid enough to survive because that's like an existential thing for those applications like they need those bots so in the long-term equilibrium i do agree with georgios that there will be some very tight and efficient but still kind of uh profitable marketplace dan from you i'd like to hear a bit about um your experience in building unisop so how much did you think about mev when designing unison b3 or even prior versions so a little and and for unicef v3 specifically i think that a little obviously about how it affects the oracle the big uh discussion was about this weird um sandwich attack that becomes more feasible in b3 than v2 um and that we've actually started to see uh to see now the strange variant on on a sandwich attack uh yeah this is this was going to be my next question so maybe to give the listener some context so over the last weeks we have seen a new type of sandwich attack right that doesn't target traders anymore but lps and the way this works is the frontrunner deploys tens or even hundreds of millions of liquidity to unison pool just in time to to front run an incoming trade and then and capture the fee and then withdraw their liquidity right after and as a result they capture the fee without any of them permanent dos so what what are what are your thoughts on that and um does it mean that market makers and miners are actually converging right now so we thought about that a lot i think there are some ways that you could maybe you could prevent it by like charging a fee on depositing liquidity with trying but that all those have trade-offs and hurt other things i think ultimately what this is is helping the user you're improving the price that the that the trader gets when they trade and this is only profitable to do when um the user's getting a worse trade than some you know centralized market maker would be able to give them so yes you are sort of allowing someone to essentially use unit swap in this in the sort of hackish way uh as um you know use it to provide price improvement to users and therefore intercept some of the fees yourself i think you know and this was always possible in unit swap but uh i think in v3 makes it easier because it requires less capital um ultimately it's an inefficient way to do this and so if it's actually because you don't do two transactions and you've got to buy the mev if you're gonna do this i think uh that means there must be like a big opportunity and so ultimately um i think those are gonna be competed away anyway so i think like this it's you know right now it's probably better that it happened within uniswap than out than outside of it and if you don't allow it like this then it's then a big opportunity like that certainly is going to get seized somehow by dex aggregator by um by someone else so ultimately i didn't think it was a big problem we'll see though experimentally we'll find out whether it's a big problem yeah and i'm i mean i'd also maybe point out that concentrated liquidity does a lot to mitigate mev actually right so um i covered this in my own presentation earlier but sandwich attacks are a function of price impact that you cause with your trade um back running opportunities are a function of price impact that you cause with your trade and the lower the price effect of your trade is going to be hence the less med you generate so i think even sub-pc does more than meets the eye in that regard that's right um yeah so um if you have no further comments on the dex section then let's move on to the last blog vitalik if you had the chance to completely redesign ethereum and i guess you'd do a little bit with you too um with mev in mind then what would you change um i mean i would definitely just look into some of these ideas that uh like allow you to separate block building from block proposing more um aside from that it's not super clear how much can actually be done at protocol uh ad protocol layer um i get a positive just like find ways to design the protocol around the yeah i uh like around the expectation that kind of transaction selection and memphis and memphis logic is going to become more specialized like if you're if you're going to accept that there's a lot of other games that you can get right like for example you can make account abstraction much easier uh so i guess like think about a about and like do some of all of those things yeah so um i think rollups are great place to experiment with some of the protocol layer approaches to mev that have been discussed today um are there any that you're particularly looking forward to well you know i think some of the kind of thought leadership that optimism has been doing around this idea of like hey let's uh auction off sequencer rights and then take that revenue and you know use that revenue to fund public goods so i guess they've been talking about you know for at least over a year now and that is uh potentially really interesting um like i kind of like this uh idea that you know from it like basically ethereum can do some of these uh things at a like outlay or two and so you can preserve the economic simplicity of the base layer but then you know if you have public goods funding that way or two and then you can kind of get back some of the benefits of these you know platforms that have like unchanged treasuries of different kinds but you get those benefits without actually yeah like breaking um interfering with the monetary properties of the base lawyer itself um so that's interesting um but the the one thing that i think one still needs to be done is like you don't just want a democratized extraction like you do also i think once at least some minimization and like basically ways to kind of detect uh sequencers that are doing um that are doing a bad job and like potentially voting them out or something like that i yeah like these like i i basically i do think that there's an opportunity to be a sort of a little bit more governance activist there because uh the costs of governance breaking are are lower than if there's cost for uh we're at layer one because ultimately you know you still have your fallbacks you're still guaranteed safety you're still guaranteed censorship resistance uh after a delay and so forth um the other interesting thing that rollups potentially could do right is they could add like some notion of transaction pre-selection and transaction free confirmation sorry like they could even like i think that this was this wasn't uh or this was the omg proposal back then i'm not sure how much this omg is actually implemented right but they wanted to have like some kind of a more centralized uh like delegated proof of stake style consensus that would make off-chain agreements on transactions so then they would give users fast free confirmations and then they would submit badges to chain once every couple of minutes [Music] with um with fifth two i guess uh providing free confirmation because you know the validator ahead of time i also think that solana does a variant of that um where if each block let's say it takes 400 milliseconds to get reduced um you can anchor your transaction into let's say sub slot times um yeah so we we could try doing something like that to improve latency on our system um yeah i think that stuff's interesting too the other thing that ethereum sharding is considering is uh staggering the publication of the shards so there's always going to be one chart that publishes something within uh within one second and so if he wants to get data in quickly you can you can just select the shard that does that and then the theory would be that rollups would have a weight of kind of building around that but i guess we'll see how that goes because like i get like making uh experienced vlog times faster is itself a type of mvp minimization right because uh like you know if you have uh like two end transactions then the amount of reorderings is like 2n in brackets factorial but then if you split them in half then you have n factorial squared and it's like strictly smaller so there's also the argument though that uh shorter block times make it easier to mask short reorgs um which would be very possible yeah well i mean i guess in the if in this particular case like one of the benefits of like the ethereum sharding is that there's just lots of different actors involved but there's lots of different blog proposals there's these large committees of attesters so it is hard to do any of that stuff until you get fairly close to the 50 mark but you know so far this is all conjecture so we'll see how it goes what we haven't really seen so far but i'm sure we will see a lot of in the future is cross-chain md um georges could you maybe give an example of how that could look like in practice maybe we have um chain a and chain b let's say one is optimal optimism the other is arbitrary or maybe two are optimum chains doesn't really matter if you want to send a message from one chain to another um you need to somehow relay the message after you've committed to it on the sending chain and relate to the receiving chain and if there's many messages um and there's no way to commit on the ordering on the sending side firstly the relayer could just reorder the messages as they relate them on the receiving chain or um they could simply if you know if ordering is committed maybe they can simply choose to delay and just not uh publish a message on the receiving chain and an example would be if you're doing a cross-chain arbitrage where your transaction needs to get somewhere within some time because that's the window when it's economically profitable but maybe the relay takes too long to relate and as a result um it doesn't work so probably what will happen with cross-chain transactions will be that we'll have some form of liquidity providers and channels to make transaction latency lower um or some other protocol like uh cellos optics by james first which which would be some alternative to having lower latency message message passing while preserving some notion of ordering hey thank you that makes uh that makes sense and also maybe like in the cosmo somewhere where that thesis will be stress tested i think will be probably the cosmos ecosystem with ibc actually and this is the point i made in my talk as well that if you have um two sort of cross chains that um each have some kind of fair ordering protocol running on it uh the cross chain muv might not be actually large as you would expect because those even if they're different definitions of fairness their uh orderings might actually not be that different this is just a conjecture but i think it's an interesting open problem um yeah i'm actually out of questions since we skipped a few um so if we don't have any questions from the audience so we can check the chat um the zoom chat i have one while we wait uh with that without knowing anything and it's two years from now you guys need to make a massive trade are you going to go on the fair sequence roll up or are you going to go on the meb option roller why not both because okay both at the same time you'll do half your trade on each one see here's the thing phil like you dollar cost average georgios is that what you do here here's the thing phil like everyone here except me has like vested financial interest in what are these solutions i literally have no vested interest in chain link no vested interest in arbitrary no vested interest in flash pods so you know like you know follow the money and you'll reach your answers everyone i don't think flashbots has a business plan so other than dan and georgios you know burning their vc money under our auspices uh i i i don't tell you know a flashback was funded by paradigm yeah i mean obviously we do want to be sustainable but i think at least my goal is not to make money my goal is for cryptocurrency to succeed and uh you know uh yeah i think i think my track record on that front kind of speaks for itself i would trade it on the on the fair sequencing chain i don't i don't know why you wouldn't i think the argument that mev is inevitable is different from the argument that you should just walk right into an mvp wall if you if you have another choice yeah i'm actually with dan here like as a user why shouldn't i prefer um in that case the ordering chain right like it's it's fine that you know the operator or some public funding source gets the mev that's extracted on that chain but doesn't mean that i get it so that is true but i also didn't say the liquidity or usage is equal so you know i think your assumption that you'll get better execution is what i would call into question uh making that making the personal decision myself but uh but i i see your point and you know a totally reasonable decision i don't disagree necessarily i just want to control yeah certainly successful so you would argue that basically the um the mav chain would get more adoption and more liquidity and that makes it better i don't know i think optimism versus arbitrary is the first round of this testing this hypothesis so i'm open to whatever happens um you know either universe is i think fine i mean i would argue with how many users would actually use uh you know like selling to the highest bidder if they knew the true cost so just to kind of give a very weird example um like robin hood right it offers free trades but users actually don't know the cost of the free trades that they're giving like there was this filing from the sec last december that users if they just use some other brokerage that didn't offer free trades they would have made like 34 million dollars more even if they paid for those brokerage trading fees then they would just make like using robinhood yeah yeah but that's a terrible argument but like it sells order flow and it actually uh users get a worse price they would have made way more money even if they just paid for transaction fees and another brokerage i agree with you although i think doing that study like without thinking about the the market impact of like the order flow activity is probably how it was done and it's like a little bit of a flawed methodology but on that i agree with you but my counter argument is that like payment for overflow is like the number one model right now as is like same as ad revenue on the internet right like they're both very sub-optimal models but like very clearly what the market wants and like every single brokerage is moving towards it because they don't have a choice because like people don't want to pay for trades anymore because they're just not used to it no it's because they realize they can make more money that way the brokerage can make more money that way like robin hood made 300 million dollars of they make more money because people want to trade there and like pay the higher implicit fee versus the higher explicit fee right like if they charge 20 bucks a trade people wanted to pay that go the real cost of what uh their transactions are actually costing them you know would they want to do something like that would they actually want free trades who knows i think it's an open question like would users want to pay for facebook to not have their data sold i think is another version of the same question and i think like yeah yeah as much as i wish the answer was different in both cases like to me the market has like very clearly spoken in both so yeah great guys uh who wouldn't have guessed that the last six minutes after i ran out of question would be the most entertaining one um so thanks vitalik dan georges mahinda and phil for joining this panel and thanks everyone for listening have a great night thanks everyone