[CLS] Ethereum Magicians Infinite Endgames: Block construction by Alex Stokes | Devcon SEA

uh slides prepared for this event but I have some from a past talk that I can use to kind of kick off the discussion and I think that's more the framing is just to have like a conversation with the people here in this room uh but yeah I can like tell you a bit about like the state of things now in ethereum with my Boost and PBS and things like that and uh yeah then maybe that will seat off the conversation and we can go from there so yeah I guess uh one question is just like has everyone here heard of Med boost does this mean anything yeah okay there's like maybe a few nods so anyway uh maybe I'll just start from the beginning and end so we have ethereum we all we all know and love ethereum and ethereum is cool because it let you send transactions which can do all sorts of types of things with all these decentralized applications it turns out that uh you can basically uh you know craft the blocks of ethereum in a certain way and and a certain way it turns out that like a builder of the block can get more uh you know revenue or profit compared to other ways we generally call this thing M and point being is this poses a concern for the protocol because um yeah essentially if you're better at M extraction than someone else then you now kind of have this flywheel where you can reinvest in your operations and you just get bigger and bigger uh what this could even look like right now is offering better returns to validators and then you have this flywheel where like all the STI ends up with you and ethereum is no longer centralized so this is very bad and you know I think the way we generally think about this uh in terms of at least protocol research is with PBS PBS here is proposer Builder separation and it's thinking about ways to separate off this building role of these like very valuable execution payloads uh in the protocol from these validators right so what you could imagine is there's this like very uh even light layer of validators that are again sort of attesting to the validity of the network but then you entirely fire allall off these concerns of uh this me stuff and the point here is that even if you do end up with like a very centralized Builder or block Builder world I mean even nowadays we talk about like a transaction supply chain so it's like very much professionalizing and like operationalizing uh even if that does you know become more centralized than we would like uh you still have this like very robust Valor set that attests to like the truth of the chain so is there a question oh sorry uh I'll keep going so what else is there to say so PBS right I guess from PBS then there's a question of I just killed it what happened oh my computer went to sleep sorry so from BBS there's a question of you know how would you actually implement this and there's a lot of ways to think about it you can either do it you know I guess like one high decision point is like is your PBS onchain or offchain and we'll talk about onchain flavors in a bit but given that that you know um like an onchain setting would be a protocol change it's probably going to take a while to even come to an agreement around what the best way to do this and even right now maybe I'll talk about this in a little bit but even right now there's still like big open questions around uh the best way to do PBS in the protocol so what can we do well in the meantime we can essentially do PBS offchain and the implementation we have that uh has the most adoption at this point in time is M boost and I think that'll team me up for these slides so the way you can think about this working is that uh it's almost this like Grand service between again Builders and proposers so these proposers you know as you go to like validate the chain uh as a validator you then also have the opportunity every so often to propose a block this is building like the consistance block and then also this execution payload and now this is where the PBS comes in is you can think about carving off the task of building this execution payload the execution part of the block where all this like jucm is um and handing it off to like someone else and me boost again is a software stack and a protocol that implements this and one way to think about it is yeah you have this like uh this green sort of rectangle over here is like the execution payload the gold is meant to like represent the M around it because it's like this nice money Everyone likes to fight over and then you have this Blue Block on the right and this is the consistence block um that the proposer wants to you know send to the network and the question now is like how do you have like this uh you know secure exchange between these two parties who might not trust each other and in fact we don't want them to trust each other uh because otherwise again you kind of fall back to the status quo I was talking or not the status quo but you know a failure mode that I was talking about a minute ago where you have like a very small Valor set that all works with these like you know one or two Builders and things like this so in any case uh you know we now have to think about how to get uh this block uh the execution payload into the block and make this all work well and at a high level the way to think about meus today is that it's like this commit reveal protocol meaning you like commit to something and then once you've committed to it it you know the person committing now is the proposer but once they committed to it uh then the the Builder is free to like release their block with with no worries and you know diagrammatically you can imagine that the builder then has their like nice green execution payload it has like some juicy M with that gold border and then then uh the like shaded thing on top of it is meant to represent that it's blinded and the reason why you want this is because if the Builder sent the payload and clear then the proposer basically could like copy the Builder's work take all the sort of you know rewards from that work for themselves and the Builder would be unhappy what this means then is that the pro like the mechanism itself is not sound and then no one would use it and it would be bad because proposes would just fall back to like these backroom deals and God knows what so the way to get around this is basically you need some way for the proposer to commit to the block this is a commit step uh and so they get a blinded payload which goes into the block the way this works is that you know sort of like zooming way into the details the proposer signs over this block in a compatible way with what's actually uh the thing that ends up on chain and uh yeah basically they do this and then they can show the Builder kind of as proof that like they've committed to this particular payload say including one that pays the Builder for their work and then once the Builder sees that they can release it to the Wild and this bottom right square is what would go on chain and yeah so the thing to consider here is like this sort of me stealing issue that I was just trying to describe like you can imagine that there's a block with some M the builder takes like some amount of funds to that address and wants to pay the proposer this if there isn't this like commit reveal step baked in to the protocol then the the proposer basically could just take the block say thank you very much for all this hard work they could rewrite the payment transaction for example that sends all the revenue to the proposer and then Builders don't want to use it there's no Builders no proposers use it and again we don't have some sort of PBS to protect the Valor set from pressures of centralization so that's bad okay so that's like kind of a very high level rview of like PBS today offchain and how it works and yeah so this has been live basically since the merge so for a few years now and at this point it's like an even more complicated picture in this cuz now there's like solvers and like bundlers before they do 7 and yeah like every day you know new parties come online that are part of this like transaction supply chain but all being said uh just to give you a sense of like what this looks like right now uh you would have the validator here that's labeled they use M boost to talk to these different relays uh which I'll talk about more in a second the relays then interact with Builders and you have this so I called out this relay actor this one's like very important and pretty much the source of all of our problems with this system today um the relay basically act acts as like a trusted broker between proposers and Builders otherwise there's not like a nice way to like trustless have this exchange that I just laid out with this commit reveal protocol and because of that uh yeah you need the relay the relays exist what this does it adds like another intermediary into things that can have its own issues whether that's being a single point of failure or especially with concerns of censorship and yeah you know these things also have bugs and like can go down and there's just a bunch of headaches that come out of this uh but this is what we have today and yeah if you follow the space at all you've you've heard about these relays I'm sure so yeah again like the system shipped around the merge and yeah uh again these numbers are a little dated but what I can probably do is even just go here right now uh here we go and you can see that they should look about the same so yeah I mean you can see here like there's over 90% of blocks say in the last two weeks have uh gone through the system so people like it people use it um again it like adds extra me uh you know extra Revenue to to your block uh for example you could say the average payment here is like 005 but I think we'll see a bit later the they're are like very outli like there are payments that are very much outliers and they sometimes pay quite a bit um yeah I mean one thing you can see here is that there are like 12 relays uh and 31 Builders registered or at Le like observed which is like pretty nice for decentralization purposes um but yeah I think I'll touch on in a second that the pictures uh not so great once you get into actual details so anyway point being uh people use me boost and it's what we have today and you know it works um which is cool however there are some uh some downsides the downsides here are that uh since you're especially at this like relay actor I was talking about like quickly entered the chat meaning like within you know pretty soon after like the launch of this system uh it became clear that like relays and Builders would be censoring certain transactions and this like directly harms ethereum's values which is not good um again I can actually go grab the web page here to see the latest data I know it's improved a bit uh since here but in any case uh people were very concerned uh about this and still are right I mean it could be the case that like new relays enter the the the market that uh get a lot of adoption and they start censoring and again we have the same problem um yeah so maybe if this isn't self-explanatory uh basically it's saying that uh right basically this like red blob is like sensoring uh the green or not and yeah let's actually go check because it's right here so that's over for the last day we'll get this nice graph right so you can see here okay I mean this doesn't look great uh point being is that yeah I mean CIP is still very much a problem um and yeah so we want to think about ways to improve it another thing that's maybe not quite well I think there's like two points here so like one point here is that you know there's not just a censorship issue with relays there's also very much a concern of um yeah the system not being secure right so just in terms of like this is software running out the world there can be bugs accidents happen sometimes there uh harmless accidents or at least like not malicious um say relay goes down um what this would likely mean especially in the MEO context is if like the relay can't get the final block out in time then like the slots basically missed uh the logic here is that the proposer has already committed to a particular block they want to essentially buy from the Builder and if the relay can't you know facilitate the whole auction or exchange then uh the proposer is just out of luck and then we miss a slot uh in the protocol right so sure it's maybe just one slot but generally this this is bad um so there's like some security things which I think already are like make people uncomfortable with relays and that being said there can even be malicious attacks right and so you might you again if you follow space You' heard about this one the low carb Crusader um L what did I want to call out here I think right so I think the point here with the call outs is that this was an incredibly profitable attack um I don't again I don't know how many people are familiar with this but basically what happened was there was a bug in the relay implementation at the time where as a validator you could basically trick the relay into giving you the block without having to commit to it uh and so that's exactly what these attackers did and and they also did a bunch of other very clever things to then also in turn uh mess with the transactions in the payload if you've heard of like Searchers in this context there are people who are like making different trades to like ARB say Unis swap or different different token pools and um yeah they even like went as so far as like bait the Searchers who were trading and then basically sandwich the sandwiches that makes sense and anyways it was actually a very like it's a very fascinating attack and I would recommend reading up on it uh but that being said this is very bad so um yeah again we just see more and more cases where relays are causing problems and it'd be nice to get rid of them so well okay before I get to here uh maybe like another sort of new update uh since I made these slides is like also the question of like relay sustainability and this is actually quite a big one because again over 90% of blocks are coming through the system uh there are effectively like a very small number of relays I know that the me. piix website should more but you know in terms of market share it's really a small number and I may actually yeah so let's do this just to make the point right so these are looking at relays and you can see here so this is basically breaking them down by market share and really just you can look at the pie chart on the right so we see basically yeah there are like a number here but you know flash Bots only has 5% uh as a relay and then otherwise you have BLX route uh which is here then you have another Blox route relay and then ultrasound so honestly there's like three relays that have pretty much dominated the market um and really you know this one is like honestly just one relay so yeah that's that's not ideal we can do the same thing for Builders if this loads okay yeah so if anything this looks even worse um again so the left and right here the left one is just also including like nonm boost blocks so for the purposes of now you can kind of just ignore it on the right here you see uh for All Med boost blocks you know who's responsible for what and you really see it's just these two Builders here right uh this is this is very bad I will say these two Builders are like quote aligned a lot more than like random Builders could be but again you could imagine a much more unaligned actor comes to the table in the next year or two say like one of these more like traditional like uh institutions like these big trading firms and they could just Dominate and care even less about like ethereum's values and things like this and maybe just to give you another view on this you know you can go to Ether scan these days and then you can see like it it gives you the minor but really they just mean like the the Builder here and yeah it's like Beaver Titan okay someone from Lio this one's fine like Titan Titan Titan you know it just it keeps going beaver beaver beaver TI Titan Titan Titan like it's really not like a great picture so if anything there's now like two levels of concern here there's like relay decentralization and also Builder decentralization and people think about both these parts of the stack um and yeah there honestly just aren't like any silver bullets yet because it's actually like quite hairy once you get into it but yeah so given all of this you might even be wondering why we did this in the first place because it sounds like it's actually pretty bad and you know I won't lie it's not great uh but that being said now the question is well okay are there things we can start to do via the protocol that would uh remove some of these trust assumptions are there like other protocol features that would play into this like you could imagine that we have Builders but then they are subject to like inclusion list or something like this uh and that might fix the censorship problem so there's like a lot of things we could think about and the question then is just what um and once you get into this Rabbit Hole vpbs you find it like is perhaps infinitely deep I don't think anyone's found the bottom yet uh there's like a number of different flavors of design uh there's many different concerns across the design they touch all parts of the protocol from like Fork choice to uh block production to the incentives around you know making blocks and attesting to blocks and yeah just it's it's like a very core part of the protocol and it touches a lot of things one thing we could start to do is get rid of this relay um and the way to think about this is like you can imagine the relay itself becomes the protocol uh somehow like yeah this protocol is now going to like be this like broker between Builders and proposers and you want to do this in like a safe way uh there is this bit about optimistic relay I we can talk about this if you guys are curious I think I'm going to skip this for now but it just is referring to a mode of operation of the relays today and yeah this might be more interesting to think about for today essentially like yeah flavors of PPS so we C out here EPS PTC uh there's this payload timeliness committee and essentially yeah there's now an EIP for this EIP 7732 that is a proposal to implement this design um the way to think about it is yeah you just kind of remove the relay uh and now effectively the protocol act as a relay in some sense uh a key bit here is you need to have like trustless Builder payments uh because otherwise again you could imagine the proposed like I was telling you a few slides ago the proposer could just steal the Builder's work and so the the protocol work in a way that you basically have this money locked up and then it could be used to paid out regardless of what the Builder does or doesn't do or the proposer does or doesn't do so that's pretty cool and what else do I want to say about PTC at the moment um yeah I mean interesting part of the design lately is that uh you could even imagine like the the builders and proposers directly talk to each other because again the way the the protocol works is you get a lot more of these like assurances in the protocol itself and so then uh yeah you can very much imagine relays going away that being said there are a number of things relays I think still would even do in this world around either providing essentially low latency as a service or like a number of other like uh nice to have features like cancellations are another one and there's more we can probably think of um so yeah I think it's very much an open question to what extent uh this particular design is actually going to like solve this pain point of the relay and yeah so this now I think that might be everything that I want to call out at the moment this is some stuff yeah I mean there's always an infinite Buffet of things to talk about um yeah maybe I'll just throw up this slide which is kind of nice and yeah so I guess from here uh maybe we can open it up I don't know if anyone has questions or concerns anything they'd like to discuss you could even come up here and grab a mic okay I don't know how to access the Q&A is it in some chat somewhere um oh here we go here there's questions up here um okay so removing Rel layers each validator will have to discover and manage the builders they connect to will it's just not cause disconnect to the two biggest ones and forget about it entrenching them yeah so again I think there's like a number of questions here on this and let me grab up so maybe to add some context this question is thinking about the latest uh EPS design that I mentioned with EIP 7732 and the idea here is it's perhaps even pitched as like a benefit of the CIP that uh you don't need the relays because again the protocol handles all the like trusted bits for you and so then a proposer can just directly connect to the Builder um one issue I think you could immediately see with this is just like privacy considerations so now the problem is that you know if I have my validator let's say I'm even just like a home Staker and I don't have a sophisticated way to like off office skate all of this if I disc connect to a builder then like they have my IP and they can start doing bad things possibly for for example if it's myp they now have a Target to like Doss and so you could basically just like get a list of these things and just wait until someone comes up that you want to Doss and then what this would look like is suddenly the chain goes down so that's not great uh this question then is in that context asking right in a 7732 world if you get rid of them in this way uh and just yeah have validators connect to the people they want to um right would you basically just have like the same entrenchment going on so I think this is another good point [Music] um I mean that I agree with I mean just I guess to be transparent I'm not like sold on 7732 but that being said the concern here is that yeah even today it's really hard to get new validators to connect or maybe I'll frame it this way it's really hard for new relays to get adoption by validators like it's almost this BD effort for them to go out and like track down all these validators because again it's like ethereum is actually decentralized so like it's very hard to even like know who to talk to to like let them know you exist um and so because of that yeah you could see this sort of effect where even if we have this world like validators still may just go to like the names they know simply because that's what they've been doing and yeah I mean if anything uh I think there's some good parts of 732 but I do have some concerns around um yeah especially this point of like removing the relay I think it might provide an opportunity for relays to be used less but I don't think they'll go away and yeah that is what it is uh thank you so next question in early discussion of Max EB it was said that collapsing the set was a prerequisite for epbs and ssf is that still considered to be the case um I'm not sure about epvs I mean yeah if someone else in the audience can think of something we're all yours um I don't think so much about ups but definitely so ssf so uh ssf again is maybe a bit off topic but the idea is basically uh like right now the way the protocol works is there's a notion of finality once enough validators have essentially voted on like a block in the chain and right now again because we want ethereum to be decentralized we have the protocol designed in a way where like many validators can come to the table this is kind of at odds with like getting this like a very fast finality because they need time to basically share their messages and run the protocol and all agree and ssf is short for like single slot finality and again it's like a research direction that wants to say okay can we like again have finality in a single slot rather than like the current uh protocol of say like an Epoch or so and yeah so in this case so max CB sorry there's like a lot of context here so max CB uh this is an EAP I believe it's 7251 coming into uh Petra the next hard fork and basically this does like a number of things but one thing it does do is allow a validator to be represented uh with like much higher stake and so what that means is today in the protocol um beam chain question interesting uh in the protocol you basically have every validator that has like a fixed stake is the way to think about it they all kind of look the same and this was done because it's much easier to reason about especially like a large number of validators if they all kind of look the same and there's no like special you know jewelry hanging off of any of them but uh this also kind of artificially inflates devel set and is at odds with things like ussf so it would be much nicer for even if you are still a single entity you know say you come with like you know more than 32 eth uh you know it's nice to just go ahead and like have that in the protocol be represented as one validator versus like you know say you have 64 wath rather than two validators you can now be one all that is to say that now with this feature there's a another part of this that lets you consolidate validators and then you start to see how we get to fewer validators in the protocol it makes ssf easier for like some other technical reasons and We're Off to the Races so cool I guess I'll go to this next question thoughts let me go to the next question simple cue what happens internally within relays um do they just relay the best block okay so relays do a lot uh this is actually a fun question so relays do a lot and if anything they keep doing more and more over time so one way to think about relays is that okay so they're running an auction and it's kind of this like two-sided Marketplace is way thinking about it again between proposers and Builders and maybe just kind of walking sequentially through the flow the way you could think about this is that Builders build blocks they also need to be watching the chain to know like the latest head to know like what the next valid block could be but on top of that they don't just want the next valid block they want the next valid most valuable block so they do their thing you know sorts in order flow from various places to make a very sophisticated block uh that pays them a lot of money and so Builders are doing their things they eventually at some point in time have a payload uh there's sort of like a I mean there's a spec but if anything there's like an implicit agreement that relay support very like a specific sets of ways to like have Builders pay proposers and so then the Builder makes the block out of whatever Revenue they happen to find they then make a payment in a particular way which usually you know a way to think about this is they just staple a final transaction on that pays the proposer uh some amount that amount then is basically their bid and the auction for that slot so then the Builder submits the block to the relay the relay gets this they then do a bunch of stuff one thing they do is they take the block they validate it locally because you know a builder again Builders are it's like an open set of Builders as well a builder could just like be spamming a relay with like garbage so they have to validate the block they validate the block with respect to the protocol but then also this like sort of subset property that uh it it sort of follows again this like PBS protocol of my Boost uh and so a key thing here is like uh this payment this payment that the proposer actually receives some payment so they do all that then from there they can extract out what looks like to the proposer as a bid the bid says like some metadata about uh the block maybe like the block header and then also the value question of the of the proposer uh so yeah right uh so there's a step before all this which is that uh as you start your node let's say you're validator you start your validator your CL and El also run by Boost when it does this you tell your consistent client that you want to use uh use this external network and as part of that then it basically sends these like registration messages to the relays you've configured via me boost um right now it actually happens like once in Epoch or something so like pretty frequently but in that message you say hey I'd like to register like I'd like to be in your your you know set of validators that Builders could build for and as part of that message there's like a proposer payment address so then separately this is again why Builders need to be watching the chain because they need to see you know say they have like an Epoch of look ahead on the next who the next proposers are when they say oh I want to build a block for you know the millionth slot they can go look and see ahead of time who the millionth proposer will be and then they can go to the relay and get the payment address there and that's who they pay yeah okay so we in the relay they validated the Builder side of things they now have a bid and the bids can be served and there's like some apis that Meo knows how to call so then when it's time for the proposer to Now sort of uh let's say accept a bid out of this set of bids uh this is all time such that again because we registered the CL knows to call the relays at least it knows to call me boost the way to think about me boost is like this little slide car software that's essentially a multiplexer across multiple relays and again like you don't want to work with just one party it would be nice to be able to work with like many of them and in practice that's what people do so now the way it looks as a proposer is as you're about to propose for your slot you're like oh hey I want to use this external you know building Service uh so then you know a little before my slot I start talking to them and then you call all the relays you configured uh uh they then respond with these bids so you know on this other side of it actually maybe if I had a whiteboard this would be easier but anyways just follow along so if uh we got to this place where Builders are putting blocks into the auction relays are exposing bids the proposer then VM boost is calling multiple relays for bids they now have like a pile of bids and they can pick whatever they want I mean the way the protocol works right now is that they pick the highest bid because you know we assume people want to be maxing Revenue so then out of the set of all bids you get for your slot say you pick the highest one from there we get to this like commit reveal game that I was talking about where as part of the bid the validator has the block header that can go in a compatible way uh into their consensus block they can then sign that and this is the same signature that would go on chain so that's like a nice property that makes this work a lot better once the proposer has this signed message um that commits the proposer to this exact Builder's payload they send that to the relay the relay then can validate that and that's actually the bug I was talking about with the low carb crator that the relay at the time did not validate that signature correctly which was a huge problem uh it's been fixed now but in any case you validate the block and then if it all looks good you can really at that point like in relays do this as well they should send the payload like the Builder's payload to the proposer but they really don't need to like at that point the relay has all the day they need to make the next Block in the chain and they can just gossip it out uh and you do both like relays gossip things they send it back to the proposer and they gossip it themselves as well so okay that was a lot the story keeps going I don't know how much we want to get into this but um yeah I would say that's like the basic flow and then from here even in like the last year year and a half we've seen more optimizations I guess this is the next point to make is that time is money and what this means is that there's like again this like very sophisticated you know pipeline or like supply chain of actors and the way to think about this is because of the way MV Works uh basically yeah every even like millisecond is like some money to some you know in practice these Builders usually end up being like very sophisticated Financial entities they're like uh are being like centralized exchanges against decentralized exchanges and all sorts of things like this and in this case like it VAR as much the case that you know the market can move against you and so suddenly like what you thought of the world as stale and you basically want to be able to react to that what this means is like now this pipeline's become incredibly latency optimized like the top relays basically do like very very aggressive latency optimizations this can look like things like having Builders collocate with relays uh which is also not good for like vertical integration reasons but it is what it is um and yeah then even from there I mentioned this point of like relay sustainability so right now relay well let's say when the system launched the claim was um essentially relays would be public goods in the sense that they would just perform the service uh I could think well you know I can say the concern was essentially that if you had a relay that um basically could charge more than other people then you still have the same sort of like centralizing flywheel that could play out so in any case I think there's like pros or cons to both models we have sort of this public goods model and that's a problem because it turns out public goods funding is really hard and a lot of the relays I think are quite strange right now with respect to like sustainability so one way Rel have thought about this is like trying to monetize somehow and there's even like more sophisticated ideas around I mean we can get into if you want but basically the way to think about it is if I'm a relay I can look at other relays because these are all public apis I can look at other relays I can based on the other relay I can uh basically charge like the second price across the entire auction set and then capture the Delta and maybe share half with the builder half of myself but then now I have like some Revenue coming in and there's some like sustainability argument you can make off of that that's going on there's also this notion of timing games which is like uh very concerning with respect to at leaste the protocol and the way to think about this is again time is money and you kind of have like a lot of room right now to play with when you actually release your block to the network you should propose at the start of your slot uh this is like the honest Behavior but you really have up until like this 4C deadline uh where the rest of the Valor set would sort of observe your block and attest to it so there is quite a like a lot of play here and you can be more or less aggressive with respect to like how much of that time you take up uh again because you're trying to maximize profit so that's not great but point being is like relays I think are even thinking about um essentially offering this as a service where they would do all this optimization for you and basically just give you like the the best block they can so yeah this was a short answer short question with like a very long answer they do a lot and they're doing more over time and yeah it's quite crazy I'll go to the next question then after I grab some water I don't know if there are any questions from what I just said that was like quite a lot okay well thanks for the questions these are good so let's see moving right some Builders have their own Searchers or something right what's the additional value okay uh yeah so this is basically a question around like exclusive order flow I think is sort of like the buzz phrase that you would hear about this and yeah like the idea here is that you know the way you should the way at least that you know I would want this to work is that these are all like uh independent entities that you know maybe are in competition in some sense because then that drives usually better to centralization um then from there uh they all have like say open apis like any of these Services these like auctions we're talking about and things they should be like independent like open in some sense because then now it like lowers the playe field or lowers the barrier's entry to level the playe field so it turns out though that if you deviate from this nice ideal model you can make more money and so this is what we see um you know you can have I mean like yeah sort of like a very pathological instance you can imagine is like I'm like a big wallet or you're like a big telegram bot uh so I have maybe already a lot of valuable order flow and then from there I could even like solid direct L to some Searcher there's some Builder and then yeah you end up with this like very sort of um perhaps collusive uh arrangement in the world and that's not great but ultimately yeah it just turns out that like yeah with these types of like uh let's say agreements then yeah you can usually make more money uh I think this is like sort of in direct conflict with like what we want to see out of this part of the world but that's what we have uh it's very hard to like prevent these things in protocol or like with any uh anything like that but it is an interesting research question so if you think about things there could be some fancy mechanism design there could be fancy cryptography there's there are ways to think about this that I think start to remove the ability or even like desire for these sort of like back room deals what about moving from the English auction to the first price field auction so yeah I'll talk a bit about this this was kind of what I was saying with this like relay sustainability sort of uh experiment that has been playing out so the way to think about it sort of from the beginning was that Builders would bid uh you know hopefully the true value of their block but even already again these are all public endpoints so like Builders have like sort of a PVP mentality where they could look at other relays to figure out what other blocks there were and they could try to have re uh responsive strategies and and you know in light of that so in any case the idea is that uh I'm a builder I like basically say okay you know if this block pays me X I'm going to pay like x minus Epsilon to the proposer I keep Epsilon for myself that's sort of like my tip for the service and yeah then you to have this bid that goes into this this auction um and this is what we mean when we say like uh English auction yeah let me think about this right so English auction here is referring to the fact that like the bids would keep going up and I don't even know if I would say this is an English auction actually um if anything I think we'd call it yeah maybe I don't understand that question but I I think what we would call this is a first price open auction if there are any auction theorist you can correct me but in any case um what I think this is asking about is then moving from that setting where basically you're just bidding in the open um and the bids are yeah just kind of they can be whatever um but again the key point is that whatever you bid is what you pay there are fancier auction designs where maybe you have a set of bids from people who are like you know trying to like sell a good or something and then you could imagine doing more sophisticated things to like score the outcome of like the whole set of bids and this is now what this like first price versus versus second price is referring to where um again there's a result in auction Theory where if you have a set of bids sort of like the best auction you can think of ever having is one where everyone just bids and what you do is you basically charge the second price um because that's basically like the only thing you would have to pay if that makes sense uh and the reason you do this is because there's now like no strategies around like different bidding games you can play that become pathological and and all this so in any case uh we now have this thing I was describing where uh the relays some of the relays have exper exped with having like um yeah different different ways to process the bids and the idea with this one is that okay you have a relay who observes all their relays and let's say like you know the highest bit from some other relay is like X and the builder then is bidding like X over two um let's see how would this work it's hard to describe out loud so you only need the let's see it does it go the other way so I think then it might be the case that you have the Builder who is going to bid yeah I'd have to think about this to give you a nice example but the point is is that uh if you have a builder locally who's overpaying then uh there's no reason for them to overpay relative to someone else because otherwise if they had not bid then the next Builder would have won and they would have paid much less and so that's where you think get this like rebate function yeah actually this might help if I can just draw this out I think this works I just need to scribble so I think this will be clear so like you can imagine there's like relay a and relay B and so the idea then is like you know Builder one is going to bid X Builder two is going to bid I don't know let's say 2X and so the idea then is that um right let me flip this 3x okay so the idea then is that if you had this set up where the Builder here is going to pay 3x and the Builder here was going to pay 2x if Builder one did not bid imagine this is not here then the auction would have been one by 2X and X I don't know I can make these concrete numbers but say x is just like an eth so say this is two e and 3 e and the idea then is that you know if you want to like draw this out two versus three something like this if this was not here because this was the second highest bid then this bid would have won the auction if this Builder was not here when the Builder is here they're obviously going to win the auction but this in some sense sets like a floor or a ceiling depending on how you look at it for the auction itself and so what this means is that Builder one could have paid 2.1 eth let's say and one right and so that's the intuition here is that this relay then can basically rebate the Builder some amount here say now there's like you know what is that 1.9 e uh that doesn't really need to be paid to the proposer and could just be sort of retained within the relay Builder system and that's what this sort of uh I don't even have like a nice name for this but there's like a rebate function Builder rebates is probably the best way to describe it there's a a builder rebate thing here where now the Relay can say Okay rather than you paying all of this to the proposer we can keep some and then you know split this up between ourselves it could be like 50 to the relay 50% to relay 50% to the Builder could be other things but now you go from having absolutely no Revenue in the relay to now having some I think that's all I'm going to say about that question so yeah so yeah the the um yeah the proposer is only interacting with one relay meaning that um the builders don't are not aware of what the other Builder are um bidding uh about the other bits meaning that we can increase the Winner's curse and mainly for uh integrated Builders so because we don't really know how much are like what the what other margins with the uh six arbitrages uh yeah sorry what was the question exactly I is it make sense like to to switch to first price s auction to uh increase uh the rewards because in the English auction like the builders are looking at each others and they will just outbid at the last you know at the end so if we are moving to a first price SE auction we have to suppose that there's only one relay because I guess it doesn't make sense to have different auctions types between relays so we have to be sure that there is only one relay um if it's a good way to increase this Windows C because I I I think that some market makers are really want some blocks and they are willing to pay more that what they're actually paying at the moment yeah it's a good question and like I guess like the first thing to say is there's again like even just making a small change like this it like has like pretty big impacts for again bidding strategies and how everyone like shows up to this game so things need to be changed very thoughtfully um I guess one thing to say here is yeah like I I think the like key difference that we're talking about right now just be the fact of whether the bids are like sealed or not so rather basically they're private and like one way to think about this implementation wise is so the bids basically from the Builder like the blocks the Builder sends are all sealed in this sense like other builders can't see that from the relay assuming the relay is acting well um but then they expose them all to like anyone and instead you could imagine now that the API is like authenticated or something in a way that only the proposer can access them so already because of what I just said uh you can imagine like Builders cut deals with proposers to like Sid step this fact like you still need to have the relay talk to the proposer somehow and you could still then have a side Channel like a proposer could be malicious in the sense of like still sharing like order flow somehow um you can scope it more tightly to like just the proposer from like this particular slot and that helps some but there was a concern there at least I know that's been discussed before um another concern is just general observability of the system like if we do this then it's like much harder to know like what Builders are even bidding and like how they're acting so right now there's actually like pretty rich data apis that all the relays support and because of that then they can uh you know not they but us everyone can actually look quite deeply uh into what's happening and uh yeah things like this would get in the way of that and it might be you know good I'm not saying necess is but it could be it's it's just more that it's an open question and there is like there are a lot of like Dimensions to this yeah question like Network performance of epbs uh I suppose these are happening on the P2P layer right sending the uh nothing right now no no not the right now on epbs well not even yeah so that that's a question like I think if anything epbs is like somewhat agnostic oh okay so let's talk about 7732 the IP um so that is if anything agnostic to like how bids are produced uh you could imagine there's a gossip layer but the concern then is that uh again time is money and so what will happen is like there's some overhead to gossip and so bids will propagate more slowly and then because of that Builders just won't use the system they'll kind of you know we could imagine they very directly go to some more like pipe or point-to-point Communications and this is where you have like proposers directly connecting to builders um again just for latency reasons so so yeah um the only case I would see P2P being relevant for epbs is the case where um it's now like a an even lower barrier entry for new Builders to enter the scene and the way this would work is that uh rather than needing to convince a validator to connect to me directly I can just kind of have like a almost like open Channel where like anyone can send bids and it's then you know you have DOS protections and things in place but then now you can imagine it's much easier for like new Builders to enter the enter the network uh then the question is like yeah how valuable are these new players and how does that all shake out with the other ones is there any evidence that the Builder relays are collocated um good question so colocation refers to this fact that um right again because of how important Lany is here this was referring to the fact that rather than have your Builder like halfway around the world from the relay like let's say there's a relay somewhere in Europe the Builder could be somewhere in like I don't know Tokyo uh there's not quite like a substantial incy between these two even on the internet and then the question is like yeah well if if time is so important we could just have them right next door like even literally like there's a computer and then there's a computer right below it and they're directly connected together right that's very low latency so uh when we say Colo this is what we mean that they are directly connected in this way um is there any evidence I think there might have been some research that tries to get into this um I'm not sure if there's anything public because I think it's kind of hard to to assess just from uh yeah public data I mean one way you could maybe get about this is like you know both Builders and relays have apis EXP exposed and you could maybe just like Ping them directly and try to like geolocate them in a sense yeah pus exactly which servers they are they have two different servers from two different continents and they are collated and they are under one but relays or Builders like I don't know I think they build are cating with them and the communications today is under one MC there's a bit more to that because obviously the relay wants to be close to the proposer so if the relay's closer to the proposer you can delay things a little bit more as well so what the Rel wants to do is to be in lots of locations that are close to proposers and the reality is that they want to be close the Li and and other validators but yeah I mean I have seen relay designs where they have this like Geo distribution baked in just again to reduce Laten Z to to validators do relays get paid for doing their task if no what are the advantages for someone to run a relay I'm also going to see how we are on time okay about 30 minutes so do Rel get paid for their task if no what are the advantages for someone to run a relay if yes do block Builders pay them and how uh the short answer is no other than again I think the most like public thing we've seen is just uh this like um Builder rebate mechanism I was trying to describe and otherwise were the advantages yeah I mean I think the initial idea was you know there are altruistic actors in the space who care about ethereum and want to like provide this as almost like a public good and so that's what they do um there have been a number of like funding efforts in the space like PBS Foundation is one uh that tries to like help the situation and like provide at least some like tools for sustainability and yeah otherwise do block Builders pay them yeah so um yeah I'm trying to avoid getting into to maybe sensitive things but ultimately let's yeah let's just say you you might imagine that Builders are paying four things uh for special treatment which if anything motivates epbs right because you know we talked about this with the Searchers a second ago like you guys you know we aren't dumb like we can imagine that there are like things happening uh with these actors and if anything like the way we remove that is by sort of enforcing more things in the protocol either with like cryptography code economics all of our usual tools uh and that really does motivate Solutions like you know different players of epbs uh trust assumptions okay what are the trust assumptions between Builders and Searchers yeah so this is an interesting question maybe I will say just a little bit about Searchers because I think I just threw the term out [Music] um there is this Vision that there is like a actually well okay I have some slides for this but it's not a big deal so ultimately um there was this Vision that you would have again the like many actors and a very like diverse setup for this transaction supply chain which is basically saying like Okay uh these days we think about intents so you can imagine I even have an intent you know I like want some meme coins and then I just like yell this at someone on the internet and then it somehow turns into like a final transaction on some blockchain and then we ask how does it go from like you know the thought in my head to uh actually landing on some chain like a concrete State transition in one of these systems so uh there's like a lot of ways to think about this and again like I said a bit ago like every day there's like some new actor entity that like kind of shows up to the table but you could imagine that uh sort of simplified model that's helpful to think about is that there's you know myself as a user making a transaction for that I like use either an application or like a wallet they then would make a transaction that goes into say them impol from there could have these like specialized entities called Searchers uh who then are looking at them in pool and these are the ones now that know these like very sophisticated strategies for MV extraction uh like a typical one to think about is like sandwiching where basically if I have a lot of slippage on like a trade I want to make on some decks which is basically referring to like my uh tolerance let's say for like execution meaning um I would accept a worse price or maybe a better price than like what I exactly think I'm getting and in any case um Searchers then can basically work around this slippage and kind of take out the bits that uh you left on the table in terms of having this flexibility of your trade against say what's on chain because then as you realize there's like markets that are moving other places in the world the chain then like follows them eventually and point being is there are ways to distract me Searchers are the entity that would do this now then if you have a bunch of Searchers in the world they might want to specialize exactly in these like very sophisticated strategies they might not be the best at like building blocks themselves because they are kind of separate task where a builder might have like a bunch of uh usually like bundles of transactions from Searchers they would then uh put them all into like a block that's valid with respect to the protocol and then the builder then would work with proposers via PBS okay so that was longwinded but ultimately uh there's notion of Searcher and if anything what we've probably seen evolve over the years that is not good but we've essentially seen like vertical integration and kind of the dissolution of these different roles um you know in the limit like we're talking so much about why latency is so bad like you could imagine they're like okay well why even why even have a builder separate from a relay right like they could just be the same thing and if anything that also gets to like this col location question like this is exactly what happens so I think similarly you see various arrangements with search and Builders uh there is a lot of research I know that has gone into this question if you just search like probably like yeah Builder Searcher vertical integration or something uh you'll find plenty of research there but ultimately yeah um in an ideal world you could imagine that it would make sense to have them be separate entities at separate parts of this transaction supply chain and they might be mutually distrustful um and in that case yeah then you need to think about like okay how do I make sure that it's really the same set of problems again just between Builder and Searcher like how do I make sure that the Builder is not going to like either search themselves against my orderflow they're going to like you know if I have a bundle of transactions that pays me as a Searcher some amount I want to make sure the Builder doesn't like um mess with that in any way and like respects the you know one way to think about this is it is like sensitive data in this context and so like you just have to be careful who who you share things with and yeah point being is I think in practice you see like all sorts of trust assumptions all the way from like I'm a builder and you can just send me flow and you know ultimately it's sort of a just trust me bro Arrangement and if anything that means then that like you're relying on reputation of the Builder uh because you know where does get around uh otherwise yeah I think you even see things on the other end of the spectrum where like Builders basically are searching themselves uh and being like very active players uh Beyond just like constructing blocks so yeah there's so it's a it's a rabbit hole in itself um let's see are there any other questions I think okay we still have some time um there's one here so what are your thoughts on 7732 do you think it will be implemented okay so 7732 is the CIP I was talking about earlier that's like a particular construction of a flavor of epbs uh there are other types of flavors of vpbs you might have heard of like APS which is like a a tester proposer separation and you know I think there's been more thoughts since even like the earlier PBS days around like how we might want this to actually look in the protocol and because of that uh well yeah basically because of that there are different ideas um and yeah so I don't think it's like super settled that like this is the right uh or like the only direction to go in that being said uh there is in this in this sort of bucket of things of having this like PTC the payload tness committee that called out way at the beginning of my slides uh there has I think been the most work on like formally specifying uh this design and it has gotten to us gotten us to the place of an EIP 7732 and yeah so do I think it will be implemented well it already has been uh pus is back there and has been working on it quite a bit so maybe talk to him after uh if you if you want to hear more about that um but ultimately yeah so like there is an EIP which is you know a very big step very exciting to see and yeah I mean there's a number of things in there um maybe the question is asking like if I think it'll be you know implemented or included on Main net and yeah I mean with 7732 in particular like I think it does start to like if anything it iterates us towards these like maybe more even more forward-looking designs around like APS you might have heard of like execution auctions or execution tickets again there's like a whole research zoo of things there but ultimately um yeah so that's something nice about it is that it sort of sets us up to start moving in that direction with respect to the protocol um another nice thing that I like this is like kind of zooming in just one part of it but it essentially adds this notion of like pipelining to the protocol that uh actually makes a lot of things more resilient um right now sort of the execution layer and the consist layer are coupled together uh in a way and uh what this means is for example if relays go offline uh they can't essentially complete the auction exchange and then there's no there's no block for that slot this is bad and one nice thing with 7732 is it does add this like pipelining which uncouples the execution in the consensus layer blocks uh so you'd at least be like somewhat tolerant to this fault with relays which would be nice um what else about 7732 yeah I mean again I think pus and I could have a bit of a conversation about this but I'm not as convinced that it will necessarily get rid of the relays today I do think think it will like lower barriers entry and we might see new actors enter the market uh which is very good but yeah I think given all these like latency concerns and just like entrench info that we have today it's like not as clear that even if we have something in the protocol it's going to necessarily uh you know get rid of or move people away from using the relays they're using today question sure um yeah I'm just wondering if you have any insight on to like uh if there's been any changes on like how daps or uh let's say intent networks or other application applications are like sending their transactions to builders and like having their transactions show up on chain like are are uh you know exchanges just still sending all of their transactions to like the men pool or do they do orderful auctions or there like other um ways that like daps are evolving and like sending transactions to builders yeah I mean I think it's all the above and it's like hard to say just because again there's like new ideas people come up with every day so like it's hard to keep track of it all um you know even one thing recently has been like 4337 and this notion of like bundlers and so now there's like you know you can imagine this whole sort of pipeline is now not just for transactions but also for like user Ops and it just kind of like broadens the scope of everything um then yeah lately then we've seen like the rise of like intense and solvers and like again just like either a prefix or even like another way to think about like getting into actual like State transitions on chain uh which is like quite interesting and yeah I think maybe you touched a bit or like we're pointing at this notion of like private order flow so like this is something I think we've seen like a pretty huge growth of um I should have numbers I don't have exact numbers on the latest data but basically yeah there's been like very much a rise in private order flow meaning like transactions that never touch the public pole um and you know if you play this out one way you could think about it is that we actually get to a place where like there is no use for like the public men pool which is like very bad for like CR type things sship resistance but it could be the case that and yeah so I don't think we'll get to a place where there's like no one using the M Pool but you could get to a case where like 90% of transactions just basically use other methods um and if anything the the mol serves as like a sort of inclusion pathway of Last Resort say for like censorship resistance um and a little followup uh I'm just kind of curious like how do uh solvers on like in temp net works like cow swap or socket or whatever um like do they when they're like creating blocks or partial blocks like are they sending them directly to builders or they send them to Searchers or like where like where where do those intents go like from uh like in various intent networks yeah I think it depends and like even when we say intent networks I think there's ideas around like even having things that look like rollups or like other blockchains that are like providing some like infrastructure for intense uh but even without thinking about like something like that uh you could just have like you know a bunch of s solvers who are like say like coww for example is a good idea or a good example where uh they're taking in things that look like intense they do some work to figure out how to like make the best blocks uh and then yeah like yeah I mean they would all literally go to builders today I mean yeah they don't have to but it would be a huge lift for uh well no I mean they really do have to because otherwise you would just gossip like let's say you have let's like take C swap as an example and you like uh have this nice like C swap partial block or like you know say it becomes a full block you can't put this in the public men pole because then like someone could just steal it right um so yeah they really do rely on these like PBS Pathways um yeah thank you that's really helpful okay I don't see any other questions and I've been talking for a while okay pus has a question yeah I'm curious what are your thoughts on uh things that are off protocol today and that they're not uh Envision to be on protocol like uh prec confirmations and and prec confirmation for L1 prec confirmation for L2 and that we the the whole Builder system is like building towards that and I wonder if uh if we act late we might lose the chance of implementing the endgame auction either execution tickets or execution auctions or whatever research come comes up if we late yeah I don't think it'll ever be too late like I don't know if it is too late then I would say it's also too late even with like PBS today and 7732 for example and like I just don't really believe at least I don't know I would hope that that's not the case um but it's a good point yeah it could be the case that we do have offchain systems that basically get so entrenched it's like very hard to even you know yes we can change the protocol or like we can you know as a community change the protocol but then the question is like uh would it matter right and it's just like being sensitive to the fact that like you know history is real we we live in a world and uh yeah this is definitely part of the consideration when thinking about different priorities around like Research into different topics uh and things like this so you mentioned pre- confirmations uh this is a very interesting one um yeah maybe just like a quick tear for the crowd if you don't know what this is the idea is that basically um like if you go to use a L2 today you basically are working directly with like a centralized sequencer and so you know you're like hey I want to make this trade on you know this Unis swap pool on whatever L2 they can basically immediately say yes uh you're good to go you're at this part in the block and tell you that immediately rather than waiting for the entire L2 block to be ready to be produced and that's more like what it looks like on L1 so pre-com here basically are a way to think about this like ux or this set of features at L1 um and there's been some designs there's like a number of like pre-confirmation teams that are working on this to like add this this D one to be honest like yeah I guess I have mixed feelings because like most of the designs I've seen are basically just copying a lot of this like uh they have different labels but one way to think about this is having like uh this Builder relay proposer architecture but then there's just like different functions they perform in the pre-com context the problem here is that this is all very like very much centralized right like um you would hope to have like a yeah if anything you start to decentralize like the relay or the Builder in this context and with pre- comps they call this like a Gateway that you can think of as like a relay um yeah I don't I don't love it uh people are building it ethereum is a permissionless system so it is what it is the claim is that we will eventually sort of do this like Progressive decentralization over time so it's not that it's like you know this terrible thing that's just going to happen uh but definitely I think it's something to keep an eye on and be very uh yeah just be very mindful of hi uh thank you for doing this um maybe a more specific question so looking back like one year and how uh blocks have been constructed back then uh the proposal payment was usually the last transaction in the block and now looking over the last year especially with Titan Builder growing uh we saw a change in pattern there where now they like send in like a bunch of uh internal transactions to combine this payment um do you have any idea why they do this and is this something why they're so successful and the second part is do we actually have like a spec of what's what what kind of things can be included to pay the proposer could it be in the next block or whatever like uh do we have that uh yeah these are good questions um so the first question is yeah like the way that uh at least the implementation started with this whole thing was that they expect they they were basically expecting the last transaction in the block to pay the proposer and uh yeah but there are other ways you can think about paying the proposer as a builder uh and we've seen some of them crop up over time so you know one way that I know people like is actually just uh having the proposer be the fee recipient in the block because you can just call the coinbase off code and then pay coinbase and yeah there's a number of reasons to do this one of them is it's actually just cheaper like the coinbase off code to load and then pay and then yeah I mean I guess we should double check this but I I would imagine that like you know coinbase in the evm call in the evm is going to be cheaper than this like uh having a whole other transaction so it's basically a way to just have like uh yeah more efficient payments um you could also then imagine you have other arrangements maybe with Searchers or different things where you might want a little more flexibility than just this like one fixed transaction um so yeah there's a lot of reasons to do it and yeah there's a number of ways to structure it in terms of like a spec uh there is one actually uh I believe let's see I think it was just like the last hard Fork uh this question came up in the context of withdrawals you can imagine then that if I'm a proposer so part of the the relay working here is that they get blocks from Builders they work with the one proposer to sort of have this auction happen uh part of this R is like verifying the payment is made correctly to the proposer there's an issue though because now we have this notion of withdrawals uh you know out of the cons layer into the execution layer and that was causing some problems with implementations because they just didn't really have logic to account for withdraws and it would lead to like weird reporting there's like Twitter Bots and things that would like flag High m blocks and it would lead to like weird things um so in any case uh yeah uh I worked with people on like a spec for exactly this uh I think well okay I might have lost slide ability which is okay but anyway um in any case uh there's like a GitHub repo so GitHub ethereum Builder specs like Builder Das specs and there's essentially like payment schemes specified there so right now I believe it's just the like last transaction and then the coinbase route if there's others that should be there uh let me know um if anything the idea is like this specifies like honest behavior and so like uh one like very popular and sort of like foundational implementation of the relay right now is like under the flash Bots org and so uh yeah like their relay follows the spec and you know we could add things but it's should be a bit of a conversation before just like adding a bunch of wild stuff um does that I think that answered your questions yeah cool okay I think we have have few more minutes but if there are no questions