Panel I Based Rollups and the MEV Gambit | Pragma Istanbul 2023

[Music] okay so this is actually round two for us because we've already done a panel with uh where four of us were there uh on the same topic um and so what I want to do is try and briefly recap some of the learnings of the previous panel and then try and and and and set the context for for for the next panel for this panel so um in terms of some of the learnings one of them was that there's different types of Shar sequencing the shed sequencing with homogeneous domains uh for example the optimism collectives you could have multiple rollups under the same umbrella and shed sequencing here at least in the short term is is much much simpler um and so homogeneous and non-homogeneous shed sequencing should be treated diff differently um another thing is that there seem to be consensus on the panel that eventually um we com um rollups would be forced to have shed sequencing with their competitors just just market demand for that because you have this this composability uh that that that comes out of it um I guess another thing that we seem to agree on in the previous panel uh was that even though Network effects are probably going to be very strong around this composability where multiple rollups kind of act like one rollup uh when they have a shared sequencer these Network effects will take a long time to materialize because there's a lot of technology that needs to be built and so as a consequence what we should expect is in the short term a lot of experimentation on on sequencing but then in the long term quite a bit of consolidation um which is you know good for ux so maybe Best of Both Worlds uh from from from that perspective um another takeaway is that um we have different types of shed sequences right we have um external sequences um one of them is asria uh another one is espresso and there's there's maybe others out there you know you could say Sola might be a shed sequencer for etherum at some point um and then there's like the IAL uh shared sequencer for a given L1 which is the L1 itself and that's called a based sequencer um and you know we were trying to see okay is there a world where both can play a role and you know one possible such world is where the the based sequencer takes precedence for the rollups um but then maybe application specific um sequencing sequencing within the rollups for example for dex's uh will happen uh by uh by more specialized sequences more ad hog sequences that are non- based um and then um one of the things that we kind of seem to open up the can can of worms and we we didn't come to complete complete consensus is interactions between different levels of of sequencing so there there's three levels that we identified level um level zero I guess well lowest level is application uh level sequencing for example a decentralized exchange and then one level up is the rollup level uh sequencing and then another level up is sequencing across l1's across uh rollups on different L ones and it's kind of unclear how these levels will will will um kind of interact with each other okay so that's for my takeaways what I want to do now is briefly introduce the panel and ask them to um share maybe an additional takeaway that that I haven't listed uh here uh Josh do you want to get started okay hi I'm Josh I'm the CEO founder at asria um we're building shared sequer network I guess positioning us here we're primarily focused on the celestial ecosystem so that kind of puts us in a slightly distinct Camp here I think the biggest takeaway from the previous panel and it's maybe not a distinct take but kind of a summary take of what Justin said is we're very early to this design space you know shared sequencers aren't like real real and that like there aren't any running in like a main net capacity we haven't seen what network effects kind of apply there's a lot of design and theory around how you know systems might work how negotiations for rebates or pricing or whatever you know could happen but fundamentally we're like very very early in this design space um and what I'm excited for is to get things kind of over the line so that we can start seeing some of these things in practice um and then you know answer some of these questions wherein we stray a little bit out of the kind of two theoretical and into a two years out it's like an extremely long time in this industry we're early Mark hey so I'm Mark and I'm a contributor to the optimism Collective and my main takeaway from our last panel was uh really this difference between the shared ing between heterogeneous domains and homogeneous domains and for how in the short term it's a lot easier to uh build a shared sequencing system when all of the domains are the same they're all homogeneous when you're building share seening across domains that are different um basically it becomes the security model of the weakest link and it's a lot more complex to reason about and um I do think it is really really important for us to not build walled Gardens right so we want to eventually get to a point where we have a shared sequencer that can you know work across all the different domains all the different types of execution um all the different L2 ecosystems U my concern here is that in practice it's really hard for the l2s to work together because they all have their own kind of business needs they're all working on their own road maps um anytime that they spend standardizing open um you know specs between them um it's like it slows down because you have to gain consensus between all the groups um we can kind of see this sort of thing play out uh many times in like the history of Technology right like for example uh Cuda versus opencl right both of these are kind of like GPU programming uh Frameworks right Cuda was proprietary by Nvidia and it was able to move much faster compared to open CL which was this open uh you know thing that was governed by a Consortium of many companies so when things are built in a closed way they're able to move much faster versus when they're built an open way and you kind of have to go between all the orgs and all the orgs have to make sure that the requirements are fulfilled you have to build consensus so I have some concerns about what building a shared sequencer uh across all the different projects actually looks like in practice Thomas hi everyone I'm Thomas from nethermind and um I think that previous panel started building some intuitions new intuitions for me around the role of um l2s and l1s for each other so while we were talking about short sequencing being this Atomic layer for uh for communication between l2s uh I definitely felt that there is some role of l2s doing maybe nonatomic coordination of of breaching between l1s and I feel that maybe maybe this can lead us to removal of bridges in general because there'll be like two strong components L2 will be serving l1's to communicate while l1s will be serving l2s to communicate um and then we had this longer talk about liquidity how easily will move so we went a bit deeper like where where liquidity stays where does it feel most comfortable if you start feeling that the uh movement of liquidity between different chains is simplified it's it's costs less because of sh sequencing or or rollups being involved Daniel my name is Daniel Daniel Wong I'm the co-founder of Tao I was not part of the previous discussion so I have a lot to catch up um um I'm not a uh I don't have enough knowledge in um share sequencing or uh me uh you know those stuff because Tao is a is a base rollup so basically what we say to ourself is you know we just focus on the minimum scope of work we let other people give up the opportunity uh let other people to use the existing uh M framework you know to extract value from our l two transactions we don't really care because we uh we don't have centralized component so we don't really you know maintain the network right um so basically I'm here to to learn more because although sequencing sh sequencing is is not so important for now uh for Tao but we also have this um new idea called um uh booster rollup which will heavily rely on like shared sequencing uh that will change uh layer one and layer two states simultaneously so that's some idea we have um yeah so quite happy to be here to to learn okay thank you Daniel that that's a great segue because the the the title of the panel is base rollups and the meev Gambit so Bas rollups just to make sure we're on the same page is when the the layer one is doing the the sequencing and what Daniel just explained is that Tao is a base rollup which gives away the m to the layo one um and in giving away the me as income they've made a sacrifice um but you know Daniel was saying that's not a big deal for us because you know we don't really care about this little bit of income the sacrifice that we're making is actually fairly small and that's similar to a game of chess where you have a Gambit whereby you sacrifice a porn for example but then you gain some sort of positional advantage and and you win the game um so I guess the the big overarching question for for this panel is you know do uh rollups want to be doing this Gambit like do they want to be sacrificing uh the Me by giving the sequencing rights to something else maybe a shared sequencer maybe a decentralized sequencer that's that that that's external to to them um and I guess one of the the the first questions that I have here is around the fairness of splitting meev let's say that um you know we don't want to make the Gambit we want to keep all all our points for for for for ourselves what happens when there's a cross domain M so there's me that originates from the interactions of two rollups if we're not going to make the Gambit how do we fairly split me so are we assuming that we have two different sequencers each kind of capturing their own I two sequencers for like two distinct domains so I mean another possibility is that um kind of both rollups have an agreement with a shared sequencer but this agreement mandates that some of the M comes back to them and then how do we figure out all the details of the these these rebates right and there's been some previous discussions around this and I'll say I'm not optimistic on like the viability of doing something that is like kind of like mathematically pure in this and that you do some kind of economic counterfactual calculation of saying like in the absence of rollup a rollup B would only have generated you know X me and when you add rollup a back in you generate you know X plus whatever me and therefore the difference that is worthwhile to like the other rollup is that and you calculate that back and forth I don't think it's feasible there I think this is also a question of like the state of like the number of rollups right right now if we're saying yes it's just roll up a and like roll up B that is like a feasible kind of like governance level discussion that could be had with you know three parties right rollup a shared sequence or rollup B if we're at 100 rollups that is kind of infeasible if we're at 10,000 rollups it's just not going to happen there's going to have to be something there so I think as we see the structure of the market evolve we'll find out more but in the short term I'd assume this is going to be some kind of governance level action between a relatively small number of parties so I guess what you're saying is that there there's two complications complication number one is that if we're going to do some sort of mathematical splitting the math is going to be like really complicated and then problem number two is that as soon as there's more than just two or three or four parties the governance just becomes way too high overhead yeah exactly and so if I were to read between the lines what you're saying is okay maybe rollups should just accept and just do the Gambit because that's the the most practical and simple thing yeah I think at some point we'll see you know kind of in the question of like to like Mark's point on like like homogeneous domains right that may be the kind of breaking point where you see a ma domain split off from some shared sequence way say okay when there were like three of us negotiating this we had a fair enough deal there's now a bunch of other people I am like deprioritized within this broader shared space I think it is better for me to kind of like Splinter off into my own domain for like my entities over here because I can't negotiate in that kind of computationally like overly complex negotiation um and we'll just have to see kind of where the power and like the ownership and kind of the you know what is the stickiness of Any Given rollup is it going to be you know subservient to a shared sequencer and you know the flows or the liquidity or whatever goes there or is there going to be a rollup that says okay you know up front if I get a 50/50 split between two sequencers fair enough larger piie I get a cut and then once it's 100 rollups and I'm getting you know 1% of that no thank you I think I have stronger pricing power in the market I will Go My Own Way um that's generally how I think is going out which is like someone unsatisfying answer and they say like we'll see how things happen but that's my best guess I I see and I guess as a consequence of this with all this short-term experimentation before the network effects really kick in and and force a window take all you're saying there has to be governance right because you want to keep this this uh this potential pressure and advantage that you have over others to opt out and and and and for yourself and also just last point is like pricing power right now is going to go to the rollups right the shared sequencers are going to have like dramatically lower negotiating power they're going to say oh theoretically you get Network effect in the future and they're like I have flow now like what do you have right yeah okay interesting Thomas um so even in simple scenarios I think that whenever we were thinking of the designs for the cross domain M that that split was not trivial uh so almost feeling like the other way around that the calculations were trivial but the the decision of split was nontrivial so that governance part would make sense but is it that complex system maybe not necessarily they I think if you want to build a simple intuition as like being a merchant between two countries and taking advantage that one country is producing something that you can sell in another country and they buying it for a higher price than in original country then uh if you want to later pay back the money that you made from being a merchant to those two countries how do you split it is it is it benefiting the the one country because it produces cheaper or the one country that pays more um it seems like you can cated but deciding really how it works in the in a more complex system of of multiple different potential trades as uh it's nontrivial and and usually that was where where we were stopping with many of those designs we were thinking like okay uh it's cross domain but we can only leave it somewhere else we cannot pay it back because resolution of that calculation is is requires some governance some decision Mak to toas you mentioned that maybe the calculation is actually quite simple Le the one question I have for you is what about the input to that calculation because sometimes you know me is like very like nebulous like there could be some uninformed flow you know that just means that you can do some sex Dex Arbitrage but like all the centralized exchange data is kind of the chain is not aware of it and so how do you even gather the input on chain to feed to your mathematical formula we we did this exercise when talking about the cross domain m in the first place like when when flash BOS was publishing the the paper on Cross domain M um and we said to ourselves maybe not always announce it that there is some vision of the the one price that if you look at the the entirety of everything that you can assume that there is one true price of every single asset so for every single chain you can design how how far away from the true reconciling price it's a bit like you would create a single auction where everyone what agree everything that they want to exchange what the price that leads to the the greatest benefit for all and the and and you can check how each of those chains how much they differ from the true price and then then you can start Distributing according to those differences um but it's not necessarily by governance that that the difference between the two price and the price on the chain is what LEDs to the to the Wi by introducing this connection the bridging between the chains and the and as Josh said with more more participants it becomes trickier uh but that part of the calculation we say okay one price and how far you how far away you say that there is something nebulous maybe even if you introduce centralized exchanges to this then then it still seems simple in this context it may become less simple when you think that blockchains move at this like discrete stages so like every 12 seconds every one second while centralized exchanges keep fluctuating so then uh if they have different final some some differences between settlement then you take some risk yourself so then you're asking actually if I pay back the meev do I pay back two chains or actually it was all the reward for my risk how much of my risk was there um yeah Ro to yeah so this is just making me think about how uh there actually might be social scalability kind of limit to um the number of domains that a sequencer can you know sequence as part of shared sequencing and um just because if all of the different domains kind of have their own preferences in the way they want to um you know if they want to earn some of the money back um it's going to become really difficult and complicated um it seems like a very obvious solution is if they're all aligned and all this money goes to like a shared kind of place and as as a full group they're able to allocate that Capital back into the ecosystem or something that seems like an easy way to solve it um but if they're not aligned then um it's making me think that there will probably be many um shared sequencers and not necessarily one you know great shared sequencer for everyone so I think it can the conversation can happen in like two different ways like from a technical perspective maybe you can have have a standard say okay if this transaction is involved in the mbv then based on some um some properties of the transaction you split a share right so that's like technical but I think most likely people will talk about business like what's your like if you use a shared sequencer uh what do you get right the the service level like agreement like you want to um sequence as many transactions uh you know from this ler tool or you feel free to uh sequence as few as as you want right so if you have this agreement for like uh like the um how to say that um the the m is it's only part of the whole agreement so may maybe the the rollup will be willing to give up all the M opportunity and have no income at all in return you know it gets the really fast uh transaction confirmation right so that's that can also happen I feel like this may be the case I mean from Technic it's really hard to Define how you split percentage based on the transaction it's it's super hard and it has also like like liquidity you know uh dependency right who suppli the money and as rollups can I also you know give money to the S sequencer for MV right if that's another Factor then this is too complicated to build a standard okay so I guess Daniel what you're saying is that um if a rollup kind of Ops in to a shed sequencer like they get all sorts of benefits they get pre- confirmations they get maybe finality they get um you know synchronous composability um and so you know maybe it's kind of an obvious decision like a no-brainer like all the advantages combined are just greater than the sacrifice you're making with the MV right um and like even if like the M was significant like the Practical considerations around getting all the input running it through all the formulas and then negotiating everything with the social are is just just too much work um maybe maybe to build additional intuition if we have like three chains and one is selling for eight the other is buying for 10 and these are the only rollups that are talking to each other via shirt sequencer there is some massive liquidity place somewhere else that buys for 20 uh and you don't take it as an input then theoretically you should you should move all the muv to the one who sells for eight because they're still so far away from that 20 price that is dominant price on the market right uh so but because you're not taking into account all the inputs as a shared sequence or you're not participating all the market you don't really know how to split it but they socially people looking from the outside think like oh okay this opportunity you're just getting some liquidity and someone else maybe will start arranging the the trade between you and the the big market with the 20 price so the calculation is simple like people can look at all the chains but you don't have all the inputs connected in the sequencers yeah and if you want inputs now you potentially have an oracle problem like how do you get this true price because one of the problems with with sequencing is that you can manipulate the the Oracle so that's a very simple example is um as a as a proposer I can I can create a block which has 10,000 e of me uh but it's all fake me and the reason is that it it's me that pays myself and so you can make an arbitrarily kind of big Distortion to the MV Oracle um I guess just trying to shift gears um another possible uh way to resolve this this tension between keeping or giving away the me is that there's a possibility that me will tend to zero over over time and there's there's two basically uh Technologies for that and kind of two types of M that I see like M type number one is whereby you have a user that makes a transaction and that transaction itself generates uh me and we have Technologies like encrypted menol that make sure that it's very difficult to um to extract me from this encrypted transactions and we also have Technologies like me share where if you're willing to disclose information about your transaction well the me comes back to you as a rebate so for user generated me there's very little that will ultimately end up to the sequencer and then there's another type of me which I guess you could say is passive me for example LPS they're not actively making a transaction they're already on chain and now they could be hit by toxic order flow um that creates the sexex Arbitrage and I think the there's this new crop of of decentralized exchanges which are uh which has basically have an external sequencer at the application Level um so just a brief yeah and and and and then in in these designs um which are me smart or M aware the me is internal ized through some sort of auction and rebated back to the LPS and so again the doesn't go to the to the sequencer or the or the proposer and so I guess my question to the panel here is do you agree that as we have just better uh tools for both users that are actively hitting the chain and LPS that are you know just passively providing liquidity um that the the me Gambit won't won't be a problem because because me will tend to zero I think that it does go to zero in the sense that uh me arrives from approximations right either we approximate time by the fact that we are on this discrete uh blockchain and and we just update pricing every 12 seconds we cannot do that immediately after any event after any signal from the offchain world we also do approximations of calculations right we cannot introduce the advanced pricing models on chain for all the type of financial instruments or other type events uh so we just run approximations like I don't know un need swaps amm instead of detailed pricing models we don't have access to all the data so we approximate based on the most available data and as we as we introduce more and more computation on chain and more and more data on chain and as the liquidity becomes um dominant on chain instead of offchain then we no longer have to approximate and the M becomes nonex existent over time it doesn't disappear but this just keeps going towards zero for sure I guess what you're saying is that sexex Arbitrage goes the away once we don't have Sexes anymore and like everything is on everything's a DEX if if we have 90% of liquidity on the centralized exchange is 10% on chain then uh then the information from centralized exchanges is not really available it's easily on chain but if you 90% on chain that you have lots of information about the pricing that is uh that is at The Source on chain available to you our vision is no sex in crypto anyone else yeah so um basically this idea of meev going to zero um my understanding is that it would be the me uh is no longer going to the proposers right so um I wonder if someplace else in the stack there will be like opportunity like if there's say like an intent kind of based system like or for uh 4337 um there's kind of entities out there that can like make a bunch of money by you know figuring out what the users like want to do and you know uh sending the transactions on behalf of the users so um I wonder if instead of me going to zero it's more like me no longer going to the proposers but instead being you know in a different you know part of the system yeah I guess Rel say I think it's useful to think about like different types of MV I think like gen one from like flash Bots like 3 EV framework of like there's like molok Mafia and monarchy meev where like monarchy would be like there's just one proposer right the mafia is like there are like entities that have like a dominant position um you know in the stack and then there's like um like mooc is like economic inefficiencies I think we want to remove the molok me which is like market inefficiencies right now it'll definitely Trend towards like a lower number than it is now because I don't think crypto markets are that efficient especially relative to like larger you know centralized like you know trafi markets um but to some degree on like the me capture I do think there will be always some like unsophisticated flow whether that is unsophisticated user flow you know you can think about people who click on like the Unis swap front end have accepted up front like I'm fine losing like 15 bips to like the front end I'm not going to go take one more click to go to a different front end that doesn't charge 15 bips so okay you're clearly comfortable for like a ux preference over like a pricing preference and there's probably going to be other layers in the stack where you know that value is extracted is that me or what are we calling that same on the back end right you know is there going to be you know essentially this like unsophisticated liquidity that may be using you know an older amm for whatever Network effects and it's not using something where all of the kind of you know um with the loss versus rebalancing is captured and rebated to them um but directionally yeah I do think we want to see like the quantity of me go down there is some kind of floor price and that like me is just part of like the overall pricing of these systems right now right you know get into like your Bitcoin fee debates versus like block issue and stuff um but right now like proposers are making a good amount of their money running meev and collecting that Revenue you know does that need to be made up for somewhere else in the stack if it's no longer coming from me is the system economically kind of sound without that from a number of nodes you want to be incentivized to run the network uh a lot of modeling you could probably do on that but um generally yeah like we want the mo i to go like down but there will probably still be some in the system just due to unsophisticated actors okay so my theory is um maybe intuition is MV will never go away I mean they will never go down to zero somehow I don't have proof you know just a some itation I feel like if um I don't familiar with this um encrypted mampo idea but I guess it will have some like operator owner right so if you have multiple those uh encrypted mind pools will the operator become the only party who can do mvv so basically you shut set the door close the door for other people to you know get grabbed opportunity but those operators will still be able to even grab a a bigger opportunity to extract value so I'm not sure whether there even in theory there's a there's a solution to minimize a zero the M uh DME take okay okay interesting so there seems to be consensus that me won't go to the sequencer instead it will go in various places it could either go you know to these uh like 4367 type me share type um uh kind of providers uh it might be extracted by by wallets you know kind of even higher up uh in in the flow there will be this dep specific kind of sequencing which will protect LPS and even if we have encrypted m pool like maybe some sort of operator uh of the menol like uh might might extract and it's definitely true today um so for example flashboard protect um is kind of an encrypted men pool which is centralized which has a which has flashboards as the operator and my understanding is that this flow is like one of the main reasons why that Builder has you know has about 10% dominance so even though they're not profiting directly from like Financial from from this encrypted M Pool uh they are profiting I guess uh kind of reputationally or you on the leaderboard that they have a fairly High position because of that uh inced mol but I would push back a little bit Daniel on the um on the mol necessarily having an operator like we have some forms of of of uh cryptography like threshold cryptography where you have a committee instead of a single operator so that's kind of in the right direction and then there's something even more powerful called delay encryption where there's no operator whatsoever you just have pure cryptography um and the technology is not there yet uh but eventually we will remove that that that operator but how how do they got incentivized I mean who are going to pay those people who got you know going to maintain the encrypted pools yeah so it would be like publicly maintained like peerto PA gossip it would be a public M Pool just like we have a me pool today and it would be I guess um maintained partially altruistically by people running nodes but also partially kind of rationally by uh because of the tips right this every transaction has a little tip and so like the Searchers and the builders are incentivized to listen on this public men pool and I guess mure proposes won't truly go to zero there's at a minimum the tips right but the nice thing about the tips is that that that they're non-toxic U that they're they're fair for everyone and then encrypted mles only remove the remove the me related to reordering and sensoring but it doesn't remove the M from injection because even if you are a builder that operates only on the encrypted mle then it can always inject at every every block at the top of the block I can inject your transactions to delay uh me extraction and inject them encrypted but they're not encrypted for years yeah so you prevents kind of some form of injection like like sandwiching where you inject before and after but you don't prevent uh like sexx Arbitrage but sexx arbites there seems to be some sort of consensus that that's going to go away for some other mechanism maybe um now we have only a few minutes left so I just want to kind of wrap up with one final question which is related to me and that's the transition from the status quo for most rollups which is centralized sequencing to decentralized sequencing and like one of the features of centralized sequencing today is that uh you know is amazing ux for users first of all we have pre-confirmation but the point I want to touch on is specifically that there's no sandwiching right like it's it's it's effectively a first come first serve and the centralized sequencer just for reputational reasons and for business reasons they don't want to screw on their users and so if we're going to shift to a decentralized sequencer we need to fix this sanding issue somehow and the only way that I really know how to do it is with encrypted me pools and so my question here is are are encrypted me pools a prerequisite for for for decentralized and shared sequencing and if so uh what kind of timelines are we talking about so I guess I can give like a practical example like I'm not sure like strictly like a prerequisite right again there is some amount of like unsophisticated flow that just like doesn't give a damn and we'll get you know take like 50 bips off of their you know execution for you know whatever reason um but we do think they're like very important in generally I think like me should be like resolved at this like order flow level you know at Astria we're like actively looking at and kind of our transaction flow is that an end user is interacting with like a rollup RPC node and then there is like another hop to like the shared sequencing Network and then another hop to the DA layer and so actually where this would happen is like you kind of inject your um essentially like your your RP see that is presumably some kind of encrypted mem pool at that layer where the user is interacting and we've actively you know um worked on some designs to use you know Suave as you know it stands now it's like essentially a form of um me share to intermediate between this rollup node and the shared sequencer um from a like how long you know it's a little bit speculative on like what assumptions do you want to make right we're still going to be in very much like a trusted environment here but an implementation that is structurally equivalent to my understanding of like the flash Bots protect RPC today um you know that's in the order of weeks or months of work to kind of integrate into you know at least like an evm based rollup um how long it takes to get sgx or you know um delayed encryption um you know M Pools like that might be longer but I don't think shared sequencing as like an architectural design makes it much more difficult if you could implement this on an L1 it should be a pretty trivial lift to move it to an L2 running through a shared sequencer okay so I guess you're making two points like one is that there's some dents out there that don't really care about stand wiing uh and maybe this is another real issue for a certain segment of the users and then the other point that you're saying is that you know there's some like fairly practical forms of encrypted me pools that are may be like pretty good and specifically suav which is sgx based it's not perfect because you have the stressed enclaves but you know it's practical and pragmatic in the term yeah and I think you know it might even be you know if we want to be like super pantic it's probably calling like a trusted me pool rather than necessar like an encrypted M Pool but the idea is that like a shared sequencing architecture doesn't necessarily like you can have guarantees about like the sequencing of transactions being like decentralized and having the benefits of not having like one um kind of persistent like leader and like the problems that may come from that and you can still have you know rpcs that are specific to like how the me pool and transaction flow go that don't necessarily leak this me with the whole varying range of um trust assumptions that you may get out of different designs Mark so I think decentralized sequencing kind of means different things to different people so maybe we can break it down and a few different things right so there's um you know leader election that's one aspect of decentralized sequencing and what leader election hello okay cool it died for a second or something okay so what leader election gives you is basically um different qualities of execution right because you've got different you know entities that are building the blocks in different ways they've got different block building algorithms so the transactions might be ordered differently then you have um you know permissionless entry into the set of possible leaders right and this is a very very important uh property for uh ethereum L1 because basically this is how you kind of get censorship resistance right um because if everyone is censoring then more entities can join the validator set and then you know these entities won't censor because uh transaction fees like the tips are very high on the transactions that are being censored now there's this like natural incentive for someone to come in and you know include those transactions so um at least uh in optimism we've kind of committed to this idea of proof of governance right where um we're not going to have you know permissionless entry to the set of sequencers and we've released a blog post about this we've talked about it a bunch um John charb you know came to the same idea um but the claim here is that we don't necessarily um need to have permissionless entry into the sequencer set as an L2 and we kind of use this um reputation and this um completely like we're not completely credible neutral credibly neutral in this sense and we try to use this to our advantage and we think of it as like an iterated game where reputation helps to uh decrease costs in other places of the system and we want to use this as an advantage and so you know I think that encrypted muls will be important in this world where there is permissionless entry to the set of validators because you're not going to trust each other um but it is if it is permissioned um it's less important understood so great point so basically you're saying that there's this intermediate step potentially where you can have like these pseudo you know decentralized uh sequencing and you still don't need uh ENC mles because there's some form of reputation I think we're over over time already thank you so much uh pales for your time and thank you thank [Applause]