In this special early access episode we're joined by Paradigm's Georgios Konstantopoulos and Charlie Noyes, along with Max Resnick from Special Mechanisms Group to discuss a possible change in direction for one part of the Ethereum Roadmap.
This is a deeply technical episode that covers the history of MEV and new innovations that plan to change it all from the comfort of David's apartment studio, enjoy.
------
📣 BECOME A CITIZEN | GET 10% OFF WITH CODE: SPOTIFY10 https://www.bankless.com/join
------ BANKLESS SPONSOR TOOLS:
🐙KRAKEN | MOST-TRUSTED CRYPTO EXCHANGE https://k.xyz/bankless-pod-q2
🦄UNISWAP | BROWSER EXTENSION https://bankless.cc/uniswap
⚖️ ARBITRUM | SCALING ETHEREUM https://bankless.cc/Arbitrum
🛞MANTLE | MODULAR LAYER 2 NETWORK https://bankless.cc/Mantle 🌐 OBOL | STAKE ON DVs, SCALE ETHEREUM https://bankless.cc/obol
🗣️TOKU | CRYPTO EMPLOYMENT https://bankless.cc/toku
------ TIMESTAMPS 00:00:00 Start 00:05:14 Intro To Our Guests 00:09:21 Proposer Builder Seperation 00:12:39 Stepping Forward vs Patching 00:19:35 Getting Rid Of MEV? 00:26:35 Inclusion Lists 00:40:14 What is Braid? 00:50:43 The Bear Case 01:06:37 Vitalik's Take 01:10:53 The Benefits of Braid 01:20:28 L2 Centric Roadmap Changes? 01:36:26 How's Rust Going? 01:44:49 Other Exciting Roadmap Updates 01:57:52 Call To Action
------ RESOURCES
Georgios: https://x.com/gakonst
Max: https://x.com/MaxResnick1
Charlie: https://x.com/_charlienoyes
------ Not financial or tax advice. See our investment disclosures here: https://www.bankless.com/disclosures
No summary for this episode yet.
Thank you. how execution works on the Ethereum layer 1. Instead of there being a single person who proposes a block, multiple people come together using a clever mechanism to co-create a block together. This has properties that Max believes are far more fair and efficient when it comes to transaction ordering and MEV. The part of the Ethereum roadmap that this relates to the most is the scourge. Where Ethereum attempts to deal with the scourge that is MEV and its centralizing and corrupting impact upon blockchains. Multiple concurrent proposers, or multi-proposers for short, Max believes is a better and more elegant way to achieve the goals that Ethereum is trying to achieve with its GERG roadmap.
While Max's braid framework is not complete and more research is needed, it is nonetheless piqued to the interest of many other mechanism designers and auction thinkers that are out there in the crypto space. So today on the show, you're going to hear a history of MEV on Ethereum and how Ethereum researchers have dealt with the reoccurring monster that is MEV at different stages in Ethereum's timeline to where we are today with the current Ethereum research and direction. Charlie and Georgios, I would say, are cautiously optimistic about the idea of multi-proposers for Ethereum, whereas Max has stronger convictions about its level of effectiveness and its validity for eventual inclusion into the Ethereum roadmap.
This Bankless Nation is a very technical episode. I'm going to have to go back and listen to this myself to really understand everything that was said here. I do my best to help some of the less technical listeners along. And nonetheless, I think you'll be able to catch a vibe even if you can't understand everything. But there are some technical terms and like deep esoteric Ethereum research knowledge that is brought forth that we often just didn't have the time to define and clarify everything. We typically leave these technical research driven podcast episodes for the Bankless Premium RSS feed.
This is the feed that I go and cover more specialized content, things on the frontier that I want to go learn about. But since the idea of multi-proposers for Ethereum is so large and relevant that this one needs to be made public so that everyone interested who wants to can hear it. But if you want future episodes like this and all other Bankless episodes as well ad-free, maybe you could consider getting the Bankless Premium RSS feed. There is a link in the show notes for 10% off if you want to go get that right now. So go get it right now. And also, Thank you for supporting independent media in crypto.
And now, on to the episode. Bankless Nation, it's SBC week, which is why we have three brainiacs here on the couch today. Just to my right, we got Georgios Konstantinopoulos, CTO, and now partner, general partner at Paradigm. Georgios, welcome to my apartment. Good to see you. Likewise, likewise. Sandwiched in the middle, sandwiched by these fine gentlemen, is Max Resnick. What's your role at Special Mechanisms Group? How would you describe yourself? My official title at SMG is the head of research. So I work on basically market design, mechanism design, protocol research, and other things. Beautiful.
Welcome. Welcome to your first Bankless podcast episode. It's great to have you. Thanks for having me. And down the line at the end of the couch, we've got Charlie Noyes, who has been on the Bankless podcast before, like four years ago or something. Very long time ago. Good to be back. Yeah. Also talking about MEV. Yeah. It was the first episode that we did on MEV on Bankless, right when MEV as a concept was very, very young. Charlie, it's good to have you back. Thank you. Good to be back. And it's great to have just some of the, I would call you guys auction nerds. I think like both SMG and Paradigm, you guys really like auctions.
And I think that's kind of going to be the undercurrent of the entire episode today. Going around in the research space has been this idea of multi-proposers. And this isn't exclusive to the Ethereum research community. This has been, I think, blockchain research generally. I know Anatoly from Solana is very bullish on multi-proposers. There has been a recent proposal from Max here called Braid, which proposes to meaningfully change how Ethereum works. And this is not set in stone. There's no consensus on this on the direction, but it's something that we want to unpack here today. And it's highly related to MEV, to consensus, to execution of the Ethereum layer one.
And this is going to be the main thing that we want to focus on and all adjacent topics. But perhaps to even really just get started here, we want to kind of like illuminate the landscape of MEV because this is going to touch a lot of different subjects. Maybe, Charlie, I'll throw it all the way down to you. Just about like, what do we need to know about MEV? What do we need to bring up to listeners' brains in order to prepare them for a more in-depth conversation as we go forward? I'll give the brief history. So way back in the day, in 2018, 2019, trading activity was starting to take off on Ethereum.
It was getting more sophisticated. And some people started to notice that there were behaviors of traders on chain that were kind of counterintuitive. And in particular, they noticed that when there were valuable opportunities on chain, the way that people competed for them was getting much more complex. The way that they were bidding in at the time, what were priority gas auctions. So that was like the original Flash Boys 2.0 paper made some observations about how arbitrage traders were bidding for arbitrage opportunities on Ethereum and then made some projections basically about the fact that this behavior was fundamental and that you would also start to see miners get involved in the game.
So the guy who wrote that paper, Phil, started a company called Flashbots, which we've been involved in for a while, since the early days. And they created the first MEV infrastructure for Ethereum. At the time, Ethereum was run by miners. um and so what it was was basically this auction where for part of the block not the whole block and some set of trusted miners um they allowed they ran an auction where you could send them transactions and then they would order them very profitably and give that to miners to put at the top of their block um and basically the argument at the time was if someone didn't do this in like a transparent way then miners would start to create internal arbitrage trading firms or like miners themselves would turn into HFT firms.
And that this would be bad because there'd be a lot of variation in how profitable miners were, and they would get really sophisticated, basically, and the market would become opaque, and it would be hard to participate in as a normal trader. And so it's better if there's sort of transparent infrastructure for this kind of sophisticated, complex behavior. And then that was arguably the first kind of PBS, or Proposal Builder Separation. Right. External PBS. And I think this is when conversations about enshrining PBS also started, enshrining it into the protocol. We haven't actually got there yet, but now we have this idea of PBS proposer-builder separation, which is strictly external to the protocol.
Right. And it's been of a loose social contract that we are going to enshrine PBS once we figure out how to do that. But that's kind of like the current state of things. We haven't figured out how to do that, nor have we actually enshrined it. Agree. And just really quickly, so the type of PBS that we had in the proof-of-work mining days was actually a bit different. It only built part of the block, and this is important because basically we didn't see censorship the same way that we do now. So Ethereum moved to proof-of-stake, and we moved from MevGath to what's called MevBoost, which is the current out-of-protocol, like, dominant PBS solution, also made by Flashbots.
and basically the different one of the big differences with it is that um these people called block builders bid on full blocks so they provide a full block to validators and validators don't have any part in the construction of that block there's no flexibility there's no flexibility and they introduced this party called a relay which previously was just flashbots but now a bunch of people operate them or at least a few people operate them which basically sit in between the builders and the validators. And the reason for the design decisions basically was, as we moved to proof of stake, people didn't want to see a lot of variation in validator rewards.
So the same way that we worried about miners becoming like HFT firms themselves, if you had really sophisticated validators that became HFT firms, and then you had like home solo stakers, which weren't or which couldn't participate in the PBS, like couldn't accept blocks from builders, then they'd have much lower returns. And that would be like a centralizing effect on the network. So there's like a lot of complexity that that ignores. But like at a high level, I think that would be the sort of like where we are today in the justification. And then one of the primary downsides from that is censorship, basically.
So we took the centralization that we were worried about happening at the validator level, and we basically moved it to the builder level. So today there's two dominant builders, basically, and they construct almost all Ethereum blocks, and they construct the full block. And so whether or not you can get a transaction on-chain, not entirely, but more or less, depends upon these two builders. So I did a lot of the talking. Yeah, I think I would just add one thing, which is that one key difference between MevGeth and MevBoost, boost which is the difference between the proof of work implementation and the proof of stake implementation is that the trusted intermediary of the relay actually allows you to have more participation so in mevgeth there was like three big mining pools and because of the trust assumptions required those were the only people who could participate because smaller miners uh couldn't necessarily be trusted to stick to the rules of the mevgeth protocol but now with MevBoost, anybody can participate.
So even if you run a solo staker, you can participate. And that was one of the key design decisions behind doing it that way. So that's a pretty good snapshot in time, almost up to this point. I do want to kind of zoom out and discuss a vibe of how the Ethereum research community has started to tackle some of these problems. We discover MEV, Flashbots is created. Flashbots creates MevBoost, a out-of-protocol sidecar that solves some of the problems. And it creates this idea of proposer builder separation, which is like something that we've considered to figure out how to actually enshrine into the Ethereum protocol.
Now there's also, as a result of that, like Charlie said, there's now relayers as well. And block builders as a separate party have also come up. And we're kind of like incrementally patching problems that are showing up in the MEV space. As we move forward in this lay of the MEV landscape. We're also going to introduce inclusion lists as a censorship resistance patch. But how would you guys describe this like incremental step forward and then incremental patch towards kind of the problem space of MEV on the Ethereum layer one? One thing I might say is that a lot of people are not happy with PPS.
So like a starting point for the conversation might be, you know, I think we gave a, we explained the reasons that we pursued this path, which I think were good reasons, understandable reasons. But the state of play today is that we have a market structure that a lot of people for various different reasons are not happy with, basically. And I think everybody, regardless of like, what improvement you want to see, basically wants to see some type of improvement and positive direction for the protocol, because it doesn't feel good to be like, like, basically, the current market structure we find ourselves in just doesn't feel good to a lot of people and i have like my takes on what about it could be better i think max though has been one of the primary um you know provocateurs or or sort of um you know for the term researcher but yeah no no but but seriously people who have pointed out like real real issues in in the trade-offs we've been making um and so i yeah be curious what are those max yeah so i think the the main thing that we've been focusing on recently is the idea of a particular type of censorship, which is short-term censorship.
And oftentimes you hear on Twitter about kind of censorship of OFAC transactions, tornado transactions in particular. In 2016, Vitalik wrote a blog post and the OFAC of the time was Silk Road. So you kind of hear about that type of censorship resistance. But in that same 2016 blog post, Vitalik also said, actually, short term censorship is really important for the functioning of certain decentralized financial applications. So let me give a quick example. Suppose we have like $1,000 for sale on chain. And we want to hold an auction within that block to sell the $1,000. Well, hopefully, we should clear for $1,000 and the person who put the money in the pot should get it.
But what actually can happen is because of short-term censorship resistance, you can submit a block in the PBS auction, which doesn't include anybody else's bid and only includes your own stink bid for like one cent. And then you can pay one cent for the item as long as you get on chain. So what does that do? It pushes all the profits from the auction to the proposer. And the idea of the kind of flashbot style of roadmap, where we do MevGath, we do MevBoost, was given that MEV is inevitable, we should do our best to make sure that it doesn't have other harmful effects. We should quarantine it and make sure it doesn't leak out and have other harmful externalities for the rest of the system.
And my point has been, I don't know if MEV is inevitable. I think we can reduce it. And one of the ways we can reduce it is by increasing short-term censorship resistance so that apps can have more flexibility and can hold these types of auctions. Is short-term defined in anything? Is it like one block is short-term or is there any like concrete parameter there? We have a formal definition, which is based on this abstract notion of a censorship resistant public bulletin board. And what really matters is given how much you put in as your tip, which is kind of your protection from censorship, how much protection do you get out?
How hard is it for the adversary to censor you? And the more blocks you allow, the more you get naturally, right? If there's five blocks where you can get in, then it should be five times as censorship resistance, right? As censorship resistance, sorry. And we see that in our definition. But what we're really trying to do if we're trying to do decentralized finance is get real-time strong censorship resistance meaning to get in ideally in a single block meaning to get in in the next block yeah or to make it as costly as possible to prevent you from getting into that block solana cannot comprehend this so i i want to add one thing that which is i think that is that definition of censorship resistance is very useful.
I think that some of the conversation, some of the issues that like generated the feelings about the current market structure also relates to like, we have issues with, you know, sanctioned entities by various different countries. Like certain actors in the MEV pipeline, like builders relays, aren't willing to process transactions from those entities. And as like a credibly neutral platform, that doesn't feel like great for Ethereum, basically. There's also this issue of like even for transactions that don't have value to censoring like if i just want to send you some usdc um like block build there's only two block builders today and they might just not pick that transaction up at like a reasonable fee level because they're like it's not worth it for us to include in the building algorithm it's not worth it for us to add latency to it and i would say these are not the type of valuable transactions that mac is addressing it's not like we're trying to run an auction but that like is a bad ux for users basically and it makes people feel bad because it's sort of just like these actors are just choosing not to include me basically when it seems like they could um and it's absolutely that's absolutely something we've noticed and have been tracking very closely for metamask users for inclusion so uh it's definitely an annoying part and it comes back to that idea of like can we quarantine mev uh if that's all we can do it'd be great to do it but right now it seems like it's starting to leak out into stuff like the inclusion of basic transactions that are just kind of low prior fee that aren't on contentious state and are not being included because there's some latency game going on in the block.
And maybe George, I mean, Max and I are sort of pseudo core researchers, but as someone who actually spends a lot of time in the sausage making process, I guess I'm curious, like what other parts of the conversation we maybe didn't cover, you know, you see in Ethereum. I'm generally a fan of the notion that it's really hard to get rid of MEV when When you introduce a mechanism, the MEV is like a little animal that moves around and is trying to avoid your measure. So I think one nuance I would add to the conversation at the end of the day, you introduce a mechanism, that MEV might move to a different part of the system that your mechanism didn't prevent, and you end up in this kind of patch on top of patch on top of patch situation.
And the classic thing, which I'm sure we'll touch on, at some point it moves to the network level, And adversaries can do things that maybe your protocol did not prevent, and you might get stuck in another bad situation. So we should apply extreme nuance and deep analysis. I think this is a reasonable point, and it's kind of the, I would call it the tenderloin cleanup or the mission cleanup point, which is that you go into, Gavin Newsom goes into the mission and cleans it out and cleans it up, and then all of a sudden in the surrounding neighborhoods, it isn't so clean anymore because of the spillover effects.
But I would say that the goal of reducing the proposal monopoly is to introduce competition. And so if a lot of MEV is actually a result of the monopoly, when you introduce competition, you're not just chopping up the MEV and giving it to several proposers, you're actually reducing it because there's actually competition that erodes of monopoly rents. You have a five-person oligopoly that actually have lower monopoly rent than the one individual. And that's the goal here. But I think it's still important to consider whether it might go somewhere else than this inclusion monopoly. I want to put a pin in that point that you brought up where we have been patching the Ethereum MEV space with improvements.
And I would say we have made qualitative improvements to the MEV landscape as the years have gone on, But we just discovered troubles elsewhere as the ARK has progressed. And the point that you're saying is that there are strict improvements that can be made at times via research and innovation to the MEV landscape via stronger mechanisms, stronger and stronger mechanisms. And so it's not just about moving it elsewhere and then all of a sudden we rediscover the problems. We are creating better mechanisms, but maybe over the years, the Ethereum research ARK around MEV has made soft mechanisms that have been patched on.
And I think what we have discovered is that censorship is perma. It's always somewhere. There's always censorship somewhere. But with stronger mechanisms, maybe we could be a little bit more hopeful about having stronger outcomes. Is that a fair summary? Yeah, I just personally believe that it's actually possible to get rid of it because of the dynamics of competition. Max, give the proposer monopoly pitch. Use the term. I would say just I think you could go into like make the case for the proposer monopoly is what drives censorship and you can break it. Yeah. Maybe what is the proposer monopoly?
What is the. So on Ethereum, the way that your transaction gets on to the chain is that it has to be included by someone called the proposer. The proposer is selected from among the validator set and their job is to propose a block. And so if you want to get in, they need to put you in their block. and they have a 12 second slot. So within that 12 second period, there's only one place that you can go to get on chain. If you want to wait 24 seconds, you can go to the next guy. If you want to wait 36 seconds, you can go to the guy after that. But if you really want to land within 12 seconds, if you're time sensitive, you have to go to that person.
So that's why I call it the proposer monopoly. It's a temporary 12 second monopoly on inclusion. And 12 seconds might not sound like a lot. But if you went to somebody who works at Jane Street and said, there's 12 seconds where only one person gets to decide which trades are executed, they would be very, very surprised. And they would be like, how do I become that person? So I can make some money because that is an extremely powerful thing for financial activity. For just transferring stuff around, maybe it doesn't matter too much. But if we're really trying to build a chain, which is my goal in terms of crypto research, which is make the chain good for finance, we really need to address that problem because that's not conducive to decentralized finance.
The only thing I might add to that is that, yeah, they choose to sell Monopoly into the builder market currently. So the proposer has Monopoly on the inclusion and they're basically going to hold an auction and saying who can pay me the most for this. Which doesn't eliminate it, it just passes it on to the highest bidder, which is almost, you know, whoever can extract the most value from the monopoly. And that's the problem statement. I think it's also important to define the goals. Like, you can design very different mechanism for achieving just, I think, censorship resistance on its own. So transaction gets in.
And I think there is a plus plus version of that, which is the, can you actually remove MEV in the process of doing that? And I think depending on how much you scope it, the hammer that you have changes. And there are different mechanisms for achieving that. And Max has an idea for a mechanism. Ethereum Research has another idea for a mechanism. Maybe the truth is somewhere in between. And we need to seriously evaluate trade-offs, especially when you want to amend such a mechanism on top of the Ethereum protocol. And what does that mean about things that need to change backwards compatibility?
maybe can i just say one more thing on why i think it's possible to end mev um basically there are two parts of mev if you look at the original um diane at all paper which is there's censorship and there's reordering we actually know how to solve the reordering part because we can change the ordering rule so assuming there's some good ordering rule we could reorder the problem is we have censorship so even if we come up with the perfect reordering rule uh and that reordering rule says that george's transaction should be first mine should be second and charlie should be third if charlie really wants to and he knows the proposer well he can slip him 100 bucks and he can be the only person in the block and then he's first because we're not even there and i think basically if you figure out how to do the ordering which is certainly an open question that we'll get into you push everything to the censorship layer and then you have one thing to solve right and if you can solve that you actually can significantly reduce mev in my opinion so you're kind of making the argument that since you're solving censorship resistance is proximate to solving mev i think it is necessary for sure maybe not sufficient but absolutely necessary okay I think we're about to drop the multi-proposer punchline.
But before we get there, I want to open up inclusion lists and Fossil as context to open up that door. Maybe, Charlie, you can help us with this one. Yeah, I would say the editorialized history on this is basically like we had or we have a proposer monopoly PBS. And for the last year and two years, as people started to notice the censorship problem, Max and others, you know, one of some of the main proposals to improve things have been the idea of giving proposers back more agency, for example. So in the construction block construction process, people observed, OK, well, you know, the builder gets to build the whole block and the proposer doesn't get to say anything.
So even if they want, you know, this auctions bid to be included, or even if they want this disenfranchised person's transaction to make it in, if the builder says no, they don't have any way to express that, right, any participation in the block construction. So the first ideas were kind of like, okay, give proposers the ability to participate in the block construction process, you know, not just give it entirely away to builders. And then there were some derivative ideas of that. I would say like pre-confirmation stuff is kind of in this camp or like proposer commitment stuff where, you know, proposers are promising to say or do certain things about their blocks and the builders have to follow those rules in theory.
And then I think more recently, Max, I think, had these ideas earlier, but more recently some of the research community came to the idea of like, okay, maybe we should have many people contributing to the construction of a block in some form. And one vein of those ideas is basically what you mentioned, Fossil, which is an inclusion list design that has many parties contributing to it. An inclusion list is basically a list that the builder has to follow. So it says when you build the block, this set of transactions, or if it's full, some subset of it, have to be included from this inclusion list.
The proposer says to the builder, here's a list of transactions that must be included. Yeah, well, the original idea was the proposer would say that. And then sort of the follow-up idea improvement on that was a bunch of proposers. So we sample the validator set and we get a bunch of people to contribute to this list. And then the builder has to follow it. And then I would say a somewhat like a related line of research, which Max has been championing, is the concept of multi-proposer. And this says, rather than putting this part in the inclusion list, which makes it less useful for MEV-relevant transactions, it's sort of, and we can get into why, probably only useful for, like, the USDC send or, like, the disenfranchised person that the builder doesn't want to include, as opposed to, say, the Uniswap trade.
Max has been pursuing this alternative line of research, which attempts to address the MEV transactions as well. That's called multiproposer. That would be my framing of it, but I'll, you know, you should. Yeah. One, again, just to add some extra context on this, the way to think about it is that you're trying to solve this sensory problem. There were two schools of thought. One was the execution tickets, the confirmations school of thought and everything that happened there, which I think all three of us probably think we don't like as a general direction. Say that they're even trying to solve censorship or something like that, in my opinion.
I don't even think that they would claim to be trying to solve censorship resistance with APS and tickets. I think they think that they're going to do APS tickets and then have inclusion lists as the censorship resistance. Gotcha. Well, that was the one direction. The other one was this whole thing where multiple people help the builder or multiple proposers decide on what should go in. And they kind of are trying to achieve the same thing. And it kind of like the main area where they differ is like, how much of the scope are they trying to address? And the wider the scope, the harder the problem, the bigger the changes that you need to do.
I think, yeah, let me, I think people are quite confused about the difference between Braid and Fossil. They're actually both ancestors of the same idea. so in like january of 2023 myself elijah fox and malish pie who works with us at smg and as a professor at rice we released a paper called censorship resistance in on-chain auctions which basically made the point that if you want to hold an auction you need censorship resistance to do it and ethereum and other leader-based chains don't have strong short-term censorship resistance and out of that elijah who started a company called duality was like hey maybe we should try to implement some of these ideas so we can reduce meb on this chain that we're building and he came up with this idea which i helped uh him with which was multiplicity and multiplicity is very similar to what fossil looks like which is fossils fork choice enforced inclusion lists which is basically the testers send some extra transactions to the execution proposer and the execution proposer is supposed to include a bunch of those in his block so it's constraining the censorship power of the execution proposer in that way but it still has a privileged position for the execution proposer now fork choice and force inclusion lists are a modification of that which works with lmd ghost so that is a pretty similar design and in fact it's like from the same line of reasoning in the same desiderata now braid is a attempt to make a version of that design that doesn't have a privilege execution proposer and so each one of the proposers in braid is treated exactly the same way and so i think that's very thematic of the ethereum vision which is try not to give anybody any outsized power.
Try to make as frictionless a system as possible and treat everybody the same. Whereas the fossil and multiplicity version, which I think is also like a reasonable direction and is trying to achieve the same goals, has that one downside. And so that's exactly the very specific technical difference. But ultimately, thematically, they're very similar. And so, in particular, correct me when you disagree with me. Basically, the way that I would view it is the fossil slash multiplicity approach, which you could call like the inclusion list approach, meaning we got a group of people who construct a list of things that have to be included, and then a builder is responsible for building a block that includes them.
Um, basically, this probably is only useful for censorship resistance for, uh, like MetaMask transactions, disenfranchised individuals, and maybe increasing the cost of censorship for like a, the auction example you gave a bit, because ultimately that execution proposer at the end, the person who's building the ultimate block that's constrained by those lists still has a lot of, uh, advantages. One advantage that they have is a timing advantage. So they have the ability to see all of the transactions in the inclusion list prior to constructing the block. And we can get more into exactly how that advantage expresses itself.
Another that they have is that they could look at the transactions that are in the inclusion list and they could decide to stuff the block. So they could say, you know, I'm no longer able to exclude these freely, but what I can do is pay to exclude them. Basically, like I can fill the block with, you know, useless transactions basically, and I'm going to have to pay the base fee to do that to kick out the ones that I want to. But if I'm willing to pay or if I'm in a time of very high congestion, then I can still do that. And so ultimately, like with the fossil and multiplicity approaches, it's probably it doesn't really address the MEV issue.
It mostly addresses like the censorship of transactions that are disenfranchised or that are just not like builders just don't include because like they're lazy and don't care to today and it feels bad. It doesn't address like the Uniswap trade or issue. And the reason for this basically is, you know, if I'm a really sophisticated trader and I send I see that Binance and Uniswap are out of line and I send a trade closing that arbitrage to the fossil committee or the multiplicity committee to the inclusion list constructors. Right. OK, it gets into the inclusion list. Now, the builder of the block later sees that trade and it's going to look at the Binance price and say, since I sent that trade to the inclusion list committee, has Binance moved in favor of the trade or away from it?
And if it moves away from the trade, they're going to beat me for it. Or if it moves away from the trade, basically, then they'll let it execute at an unfavorable price to me. But if it's a profitable trade, they'll beat me for it instead. And so as a trader or when I'm making MEV relevant transactions, I'm basically never going to use the inclusion list because I know that it's going to mean that I get adversely selected against and I'm less profitable. So all the MEV still flows through the builder. And the inclusion list is really only useful as like a sort of – Last line of defense for the disenfranchised.
Right. It's a side door for transactions that aren't inherently relevant to the mev game. Now, in fairness, a lot of folks think that this is still useful. It feels bad if you're a Metamask user and one of the big builders is just like, you know, this this trade is this transaction, which I don't care about it. It's not that I want to censor you, but like it's going to increase the latency of my block. So like I'm just not going to include it. Or if you're a disenfranchised user, et cetera. I think ultimately it's a good design property of Ethereum for USDC sends to get on chain as fast as possible.
And I think that there is a really strong case for the fossil multiplicity approach that it helps with all of that stuff. But what it explicitly doesn't address is removing the builder. The monopoly that the builder has. Yes, removing the builder and the builder's monopoly on MEV-relevant transactions. Yeah, just maybe just tile of the things you just said into a bow. Like, there is no way, no matter how many math Olympiad and Putnam medals you have, that you are going to beat me at trading. If I'm two seconds ahead of you, I have two seconds more Binance information. And that means that if you're two seconds behind, meaning you're submitting on the inclusion list with a design that has a view canonicalizer like multiplicity or fossil, you don't even want to play because you know that somebody else is going to come in.
either it's going to be the execution proposer themselves or somebody else who bribes the execution proposer to do this and they're going to have a two second advantage because they're going to have two seconds extra binance moving against you and if you try to make a trade all of your good trades when you make a trade and the price moves favorably or stays steady you're going to get outbid on and all of your bad trades you're going to keep so you you only get losers and this basically spirals and make it so you don't want to trade at all in the multiplicity inclusion list uh way so isn't this maybe a shorter way of saying this isn't that the inclusion list is really an inclusion set that doesn't force any ordering the builder just needs to include them and doesn't matter what order they can do whatever they want this means that Yes, indeed, the inclusion set is not a good fit for MEV bearing transactions.
Open question, I think, whether that is desired or not. But what do you think about the world where the inclusion list has an ordering rule baked into it? And I guess like my broader point being, is this a question about having multiple proposers or multiple people doing the thing? Or is it the question about whether there exists a sound way to do some kind of ordering rule that allows the MEV to be more proper? Well, I think it's a really good question. I think before we get into the ordering rule stuff, I would say that Max's pitch for a braid and pure multiple proposer, which is the alternative path from the inclusion list plus builder approach, the fossil slash multiplicity plus builder approach.
basically it intends to delete the builder and to replace the builder with an ordering rule. And I think before we discuss the ordering rule, it probably would be helpful for Max to explain how his approach, like the pitch for we can delete the builder. And maybe Max, why don't you sketch out the braid? Yeah. Like step by step. Yeah, that's exactly what I want to go next. I want to put a stamp in this conversation because this is an arc of the Ethereum research that is now kind of taking a very meaningfully different path if this is the way that we go. Whereas blockchains, Ethereum have always been in this like leader based system where somebody has a monopoly, even if it's as short as 12 seconds, somebody's always had a monopoly.
and all the things that we've talked about, inclusion, list builders, proposer-builder separation and trying to propose a builder separation has all been addendums, additions, patches onto the system because of the leader-based system. And now as you're about to introduce Braid, this multi-proposal set, we're kind of going back very far backwards in Ethereum MEV research to propose an entirely new construction of going forward. That's kind of how I interpret this moment in time of Ethereum research. That sounds right to me and David, just to be very clear. I think we're very early in the research phase in all of these things.
Max has more work to do, I think, on publishing, being more explicit on what we're achieving, the protocol, the proofs, and everything around that. Same thing for Fossil. We need to do a lot more work. We might come up with a third new thing. So I think the way to approach the conversation is that we haven't added an item on the roadmap. Right. We are very early. This is a concept. this is a trail of research that we're going down yeah yeah because we have there are so many things that we haven't figured out yet as we will find out in this conversation that it's very early for us to say hey this is like where we're going um because just nobody has done that before this is like very blue ocean uncharted waters and actually the theorem ecosystem is way ahead in terms of facing this very hard problem right which is why you don't hear this discussed in many other places except maybe anatoly solana who was probably way ahead of everyone else on this train okay so max introduced to us multi-proposers yeah so i i say that this design is kind of suspiciously simple because if you ask i asked a bunch of consensus experts like these are my desiderata i want multi-proposer can you give me something that does this they said oh that sounds like it would be super complicated but then i had a few more discussions i was like talking and and milesh who again works at smg as well kind of said hey why don't we just do like the stupidest possible thing which is just take lmd ghost and copy paste it four times or eight times and then we just try to interpret the four parallel chains as one and uh at the time uh the consensus experts were kind of looking at him like he was crazy and then we just kind of thought about it a little more and a little more and nobody could really find an immediate break and then we kind of convince yourself that actually this does work in fact if you have a single threaded single leader blockchain that has eventual consistency then the multi-threaded version of that like braid inherits that eventual consistency if it has liveness it's trivial to show that it inherits liveness and you can modify the finality gadget and lmd ghost to give good finality properties for like it's a very similar finality gadget as well so we basically said okay let's you know start circulating this idea and that's where we are now a few weeks later after that realization so i mean max when you gave i watched your talk that got circulated on twitter and you you present it in the same way where it's like so stupidly simple and then here it is.
And then you explained it and then I was waiting for further explanation, but then the explanation ended. And so that's kind of what just happened to me just now. So like my interpretation is that these builder, the proposers, the validators are going to make four blockchains if we're doing a four multi-proposer system. We're going to make four blockchains and then via some mechanism, they just get interpolated together. They get combined together and then that becomes the blockchain. that becomes the block that gets added to the blockchain and so and that was it and that's what a multi-proposer is yeah it's uh it's like it's like you know it's like the owl and then draw the rest of the owl so how how how does that you know the collapsing actually i think is like quite important here so yes okay like it's like the four of us friends are here like so we have a block each yeah i've got a block yeah georgia's got a block we all got we all got a block how does So four blocks needs to become one block.
How does that happen? No. Right. So the way this works is what I call execution consensus separation. So the execution layer, so Georgios is going to have to implement this, is going to be basically becoming the view canonicalizer in the system. They're going to be looking at the chain, interpreting which blocks arrived on time and which didn't, which are confirmed, which are valid, which are not. and in real time on the unfinalized head of the chain they're going to be looking at the chain and seeing which blocks arrived if there's four blocks that arrive and we all think that they're valid we're going to zip up each of the transactions so all of david's transactions all of george's transactions all of my transactions and all of charlie's transactions are going to be zipped up into an unordered set and then we can apply some deterministic ordering rule to order them afterwards and execute them.
And the deterministic ordering rule might have been previously known as a builder. That's like the thing that we are replacing the builder with because it's the mechanism, it's the software, it's the code that is determining the output of the order. Well, yeah, except I would say that a builder isn't a deterministic ordering rule. It's an attempt to solve this very hard NP hard knapsack packing problem. Okay, so four sets of transactions, for different orders. And then there's a rule that all of these things go through. And then the output of the optimization rule, the ordering rule becomes the one block.
What happens if like, we're all looking at the same mempool. I'm pulling in your guys' transactions. You guys are pulling in my transactions, right? Like that's how I understand blocks to be made. What about this contingency? Yeah. So if you send to multiple leaders, you have to pay the base fee for multiple leaders. So you're actually going to, it's going to cost more you do it if you want censorship resistance if you don't you can just send to one leader um so that's like one point is that not everything is going to be sent to everybody okay i would say okay sorry could we go over an example just to be precise here um i i am looking at the peer-to-peer network you're sending a blog around all of you guys are sending the block around can i not act as the leader almost by just waiting for all of you guys to send me um the block and then at the very end i will send it out um how how does that work is this is this latency games are you asking about latency games yeah it seems hard to yeah maybe let's talk about the you know the set of things that we're trying to we're dealing with here Because yes, of course, there's a VU canonicalizer.
Yeah, you take n things and you reduce them down to one. But like, how should that work? Well, I think the giga bull case for Braid and multiple concurrent proposers is that it is a design which tries to achieve, I think what Max calls simultaneous release. And that means if all four of us are building blocks at the same time, they're going to get zipped up after we all release them by everyone else watching the chain. And all the transactions in them are going to get added together and then ordered by some rule. Now, if there's simultaneous release, which you can kind of think of as I don't know what's going to be in your guys' blocks before I release my own, then there's a strong argument.
And I think this is the core argument in your original paper that I should just add all the transactions I'm aware of to my own block. Because if I don't, you guys may have added them. I don't know what's in your blocks, basically. And so my dominant strategy is just to add all the transactions that I'm aware of. Right? And if everybody, if all the leaders follow that same strategy, if that is the dominant strategy for all the leaders, then we have extremely strong short-term censorship resistance. and there's no builder. Does that simultaneous release exist? So let me step in there. There's two versions.
There's a weak simultaneous release, which is, so it depends on what information you consider. In game theory, we have this idea, we write out the game in a game tree and we have two types of things. We have things that happen after each other and things that happen at the same time. And the formalization of that are if actions happen at the same time in game theory, we say that they're part of the same information set. So that means when I release my block, I have the same exact information about the world as Georgios. So what you consider to be that information is important. If what we care about is only that I don't know the context tense of Georgios' block that he's proposing, I think that's absolutely possible as long as some cryptography implementations exist that we think probably exist or will exist in the next few years.
Something like a time lock puzzle would help us do that. So I could release my block and I could put a time lock puzzle that says this will decrypt in one second. And we can absolutely get the timing games down to within one second. So that would work. But the challenging part, the strong version of this synchronous release or simultaneous release is, can we make it so that we all have the same Binance info? And that's the one that I'm still not sure if we can get to, but I'm working really hard on it. And I hope we can solve it. I think it's pretty important to get that strong version as well.
But I know we can get the weak version, at least if time lock encryption works. I will now paint the bear case as well. I think there's two bear cases. So the first is this. On the weak case, I would say we don't currently have good cryptography for what we're trying to achieve, basically. I think that you would need farther out there cryptographic schemes like strong BDFs or something to get what you were talking about. And the best that we could do today would be some type of threshold encryption scheme, probably. And like, there's the possibility of unaccountable collusion. There's also probably in whatever, like this is not specified.
But when we did specify it, there probably are going to be a lot of issues with like free option problems, meaning if some subset of the participants decide not to decrypt or to decrypt selectively. like i think i think should we like unpack this for the audience so so there's threshold encryption which means i take my information and then i encrypt it and i give the key to you know georgios and charlie and like if they both have the key they can open it so you need both keys to open it that's threshold encryption strong vdf vdf means verifiable delay function just so we're not using too many acronyms.
And that's important for these time lock encryption schemes that I'm talking about. The issue is that threshold decryption in general usually is implemented as another layer of consensus. So you end up running on the same problems that you were trying to solve in the first place. And that's why it doesn't feel like a credible tool that we could use here. And there was an old school around verifiable delay function, which also was a dead end. So that also felt like a dead end. There was some recent research on doing time-locked encryption with a very recent line of work that I cannot remember to explain right now that might be promising.
But I guess, yes, maybe. But even that seems hard. and even then the moment that you introduce encryption that means that you incentivize spam and it goes back to my original point around the mev moving around because you end up in these probabilistic games where i will just try to flood the network with transactions it's because i just cannot see what's going on it seemed rough i think well first of all let me just say that there is commit reveal we do know how to do commit reveal we just commit to the hash sign the hash and then you reveal it later. And if you don't reveal, you get slashed.
It has some issues you might really not want to reveal. And the amount you don't want to reveal is like more than the amount you've put up. And so you're not going to reveal and lose that stake that you put up. But at least we do know that that works. And if we put sufficient stake behind it, we already have 32 ETH of stake, which is quite a bit for each validator. But did the low-carb Crusader fiasco last year disprove that almost that in the face of an uncapped amount of mev at risk and flight um no cap bond can can secure you against that attack capital efficiently uh well i i think yeah and actually tarun and matthias have a paper about this uh which is like credible decentralized auctions i think is the title that actually looks at based on the the weight of the tail of the distribution of mev opportunities so how how likely it is to have these big uncapped um opportunities you you lose the credibility with these commit reveal schemes so this is certainly a reasonable point i think that's why i'm saying i'm really nice maybe zooming out on it for a second i think um there's a bunch of stuff in the cryptographic rabbit hole about how if all four of us are proposers we can um try and encrypt our blocks in a way that's reasonable.
I think what Max was getting at, though, and, you know, if you were to zoom all the way out on it, the attack case that I think most people are considering right now when thinking about this is basically, like, let's say that rather than all four of us releasing our blocks at the same time, in fact, David is a really sophisticated HFT trading firm. And he's maximizing his connectivity to the network and latency to other validators and to all the important exchanges, et cetera, and all the important source of order flow maximally. And let's say that David even has like a consistent 50 to 100 millisecond latency advantage, right?
So the issue that we were talking about adverse selection earlier in Fossil, I think still applies here, which basically is as a sophisticated trader, I would argue I'm probably never going to send my order flow to any of the three of us. I'm only going to send it to David because I know that he's going to have the best, latest available information, basically, and that I'll get adversely selected against if I were to send to anyone else. um and so you're going to end up potentially in a situation where rather than um you know sort of traders sending to all of us and getting censorship censorship resistance um from all of us potentially including it that basically all sources of order flow are going to collapse onto the most sophisticated proposer and they're going to emerge as like a builder basically like we're going to rediscover the builder role um through um the most sophisticated proposer having like a natural monopoly on order flow would be how i would david almost by virtue of his privilege though sorry we're giving you i love it i'm loving it i'm loving my privilege i don't know if you're an hft firm but if you are i keep that part hidden yeah and david almost like emulates the role of the proposer via his privileged network positioning even though you're in a leaderless system david has uh co-located with each bushwick so uh right yeah yeah fast connection okay so charlie i think what you're saying is that like we'll just we'll just kind of rebuild this same system that already exists and now we have like it just doesn't in a separate paradigm it's inside of the multi multi-proposer paradigm but can i just defend against this critique because if the argument is hey we shouldn't do this we should do fossil instead because there might be a 100 millisecond last look in braid there's a two two second last look in fossil like right we're longer so i think it makes sense that the same emergent properties comes out of the system no matter what and it's more about a question like is it nonetheless a strict improvement in the mechanism and i think let me say that sorry sorry david to cut you off but i i don't want to let this fud go unchallenged here because there's a model that we think of that we look at.
Often we look at models, all models are wrong, some are useful. And usually one model is right for the task that you're looking at, but isn't exactly right for the other task. And I think what Charlie's doing is referencing a model that we had in one of our papers that was actually based on something that Julian Ma, who's at EF originally proposed, which is basically the price is fluctuating. You have somebody who acts later and somebody who acts earlier and you have this complete unraveling effect that charlie was talking about but it doesn't quite work that way in this system because if everybody goes through the one uh fast through david who has his latency advantage set up then you don't expect them to bid very high in the auction that's not run by david they'll bid for to get to be David's favorite, to get to submit their trades to David.
But you shouldn't expect an equilibrium to arise where there's no bids from the other proposers because there's two stages of the auction. There's the auction for who gets to be David's favorite, but he's going to choose one. And then that one person is going to win the auction for basically free. So they're going to not have to pay too much money for the right to do the arbitrage? So I think I, I think I would say a couple things. Number one, um, the 50 millisecond versus two second latency delay in the fossil versus like whatever timing advantage we can garner. Um, I think there's some centralization trade off there basically.
Um, the second thing that I would say is that timing advantages are probably one of the most obvious, like first order advantages that you can garner, but I certainly don't think that it's like the only advantage in the metagame, basically. Like I think the sophistication of the most sophisticated proposer or what advantage they have, there are going to be a number of different ones. And in the Binance sex decks, although it's like one of the easiest examples to use, I think that it's like, you know, even if you go look at the order flow distribution today, it's not like the only one that matters, right?
So we said there's going to be some consistent advantage on that one. And we can talk about, we can talk about some others too. Um, so in the multi-proposer case, again, let's say there's like one sophisticated proposer, um, that can see other people that we can see other people's blocks, um, and potentially listen to bribery oracles and other types of things, um, you know, longer than, than other people can. Um, I also think that you run into a lot of issues, um, with like the deterministic ordering rule that we were talking about earlier, where basically like that proposer is also going to be able to, in a maximally advantageous way, interact with the ordering rule and any applications that rely on it compared to the other proposers.
So like one of the most common or like one of the probably best ordering rules proposed or most reasonable is priority gas auctions, right? Like what we had in the Ethereum mining days. And another thing that I wonder about with the multi-proposer version of this that tries to achieve like no builder and we're going to have a deterministic way of merging the blocks is like in addition to David being able to see Binance last, David also very plausibly is going to be able to see what all of the Uniswap trades in the three of our blocks are and sandwich all of them and penny the auction, the priority auction on behalf of anyone else who wants to participate in any protocol um on chain right so i think that there's like a bunch of different kinds of advantages that that uh that flow down and like the resulting metagame um i think in many different ways like power will concentrate onto uh the sophisticated proposer but then revenue will concentrate to the sophisticated proposers and then everybody will try to get sophisticated which is already happening but i but i think centralization point.
But it's already centralizing. I mean, we have all these kinds of timing games. We have games that are not being played right now, which is... It's kind of exacerbating the issue, right? No, I think it's actually making the issue better because right now, if you're faster, you get an extra 100 milliseconds or a second of proposer monopoly. And that's incredibly valuable. And in this system, you can get an extra 100 milliseconds of being one of N proposers. So you'll get 100 milliseconds of last look, it's not as valuable as extra 100 milliseconds of proposing monopoly because proposing monopoly is incredibly valuable.
Okay. Here's a basic mental model that I use for it, right? So in the multi-proposer case, right, we have basically this deterministic ordering rule that's sitting in between multiple proposers. And we're going to say that there's some difference in sophistication between them now, right? And then the other side of this, we have people bidding for inclusion with also obviously like a variety of sophistication, right? And I think that like something that worries me a lot about this, if I had to boil down my concern with it, is that like, again, if we go back to the proof of work days, like the original MEV paper was like how only the bitter side of this market, the people who wanted transaction inclusion, how their behavior was like unintuitive and unintended, even while miners were trying to run honest PGAs or honest priority auctions.
And I think basically like the MEV metagame that would develop on top of Braid is going to be the result of bidders and the proposers with varying degrees of sophistication, like trying maximally to manipulate the system to their own benefit. And I'm personally like pretty unconvinced at this point that we can avoid a situation where the metagame is a going to be a lot different than we intend, meaning it's not going to be people bidding their like true value in this priority auction or something or other ordering rule interacting with it like quote unquote honestly in the way that we intend.
and with comparable levels of sophistication or revenue on the part of the validators. I think probably what's going to happen is both sides of the market are going to try to manipulate this mechanism to their benefit. And then the most sophisticated of them are going to garner an outsized amount of power. And I think the delta between the least sophisticated participant in that market and the most is going to be significantly greater than the least sophisticated proposer on Ethereum today and the most sophisticated proposer. And if this were all true, then I think at the end of it, you end up with PBS just in like a more opaque way.
Okay. Well, first of all, that's wrong. It's actually not going to be a bigger delta between the most sophisticated and least sophisticated. I think you're reasoning by analogy. Typical VC. Sorry. No, no. Let me finish the point. You just went on for five minutes. You're reasoning by analogy from a model that we wrote about a different context. and when we actually solve out the model the adverse selection isn't going to look as bad as you think it is so let's go we'll so we'll solve out the model i understand the propensity for oh we don't know what's going to happen but let me tell you max resnick is here smg is here we don't do things the same way and we're going to prove the the properties that we say the mechanism has and we are going to show formally what the equilibrium is this is not going to be a situation where we come in, we ship reams and reams of unsubstantiated claims so much that you can't read them all like a, like a legal battle where they send you all the file cabinets and they don't digitize it for you.
So you can't look through it. And that's what we're going to do. And if you don't agree with the proofs, then we can argue about the proofs. But like, I'll just say, you know, Hey, you may have some skepticism now. Okay. That's, that's on us. That's, that's our prerogative we're going to prove the results which is how things should work and then we don't have to go into this saying oh there might be something that we're not considering no we proved it qed it's done i'm super excited for that and i'm super i'm super super open-minded to it i also don't like the research dos i i think we're very aligned on this i'll make what address just one thing you said which is to me the difference today in validator sophistication would be how late in the slot you're going to call the PBS block or how late you're going to call MevBoost to get the block.
That delta, if you're going to play the timing game, is not that great. I don't recall the number offhand what it is. But I think the difference between a multi-proposer scheme who's not sophisticated and one who is, if the one who is is getting all of the Binance flow and the one who's not is getting functionally none of it, I think that delta is going to be a lot different than if you call the MevBoost auction 300 milliseconds later or something. I think because in that case, the proposer is... The builder is still getting all the same order flow, just not the one in the last 300 milliseconds.
It's going to be a great point to debunk or prove in the work, I think. Yeah. Now, I do think that Vitalik, as usual, had the best tweet on the subject recently, which is basically um you could boil i think a lot of the research conversation about this down today with like can we remove or at least prove some things definitively about this last look like can we get rid of the last look and multiple concurrent proposers that would lead to this like breakdown that i'm articulating or i think to your point can we definitively prove some things about how bad or how not bad that last look is. I think if you know how to quantify the latency game, then you describe it and you describe exactly the issue with it.
If you don't, you write something on Twitter that says, oh, there might be a latency game here. There might be adverse selection here. So I think what we need to do, it's on us to demonstrate. That's like if we say, hey, we should do Braid. That's our prerogative to demonstrate it. We're going to demonstrate it. and if we can't then we'll probably show it's impossible and then we can all move on right i was gonna say exactly that i think even in both cases i think we're all truth-seeking here in both cases if it can be done the proof is gonna be great if we find out an impossibility proof that's also gonna be amazing because it's gonna save us from all the dos and the twitter situations Which maybe there's a meta point here around all of us maybe doing upfront a lot more work before we try to like, you know, garner excitement or criticism around an idea.
Speaking of garnering excitement, since we've been poking holes in Braid since its introduction in this conversation, I do want to give Max the floor of how might Braid benefit everything upstream of it or downstream of it. I know there's a conversation around DeFi here. We started with poking holes, but I want to give Max a chance to just talk about how Braid is the best thing ever. What are the big benefits that you see of this multi-proposer paradigm, granted that we are able to figure out the research questions around it and figure out that problem? Yeah, and ultimately, like, we're not just we're not trying to push anything for any personal gain.
Like we're just all here trying to make Ethereum as good as it can be. And if, you know, I happen to believe that we can demonstrate that Braid has certain properties. We're working very hard. There's a bunch of people working on the paper who are, you know, working on proofs probably as we speak right now to try and show those properties. and if we can't demonstrate them oftentimes when you try to prove something you either find that it's possible and prove it or find that it's impossible and then show it and we'll release either way we're not going to be like a medical study where we we don't release the negative results we release either way um i think it's really important i think there's a lot of complexity on the roadmap as well that introduces a bunch of problems that braid and execution consensus separation is an alternative to i'll give one example um epbs there's a design out from terence and potas right now which is i think originally based on this proposer timeliness committee which i'm not sure if it was mike's idea he did write a post about it a while ago um and see back from pepsi the proposer timeliness yeah okay so yeah sorry if i'm uh giving credit where it's due.
I'm just not recalling exactly the thread of research. Mike was also into that. Mike was in all of that. Okay. Yeah. So this Proposer Timeliness Committee idea is basically you commit to the block as the builder. You send your block commitment to the proposer, and the proposer selects the highest commitment. And then afterwards, as the builder, you have some time to reveal your block and that is basically this alternative design for epbs and we know we want to do epbs and protocol somehow so braid is a way to do epbs and this is an alternative the most kind of developed longest chain on the alternative fork of doing just epbs so one attack on that which is an issue is you could place some large trades and then if the trade goes against you you choose not to release.
And so you say, oh, well, then we have to have a huge bond available to slash you with. But what's the size of that bond? Again, we have this uncapped bond problem. So my argument would be, yes, there's a lot of complexity here. But in the words of Italic, if we just eat the complexity, if we can figure it out, if we can show the results are true, which we're working on, then we can eliminate execution tickets. We can achieve Meb burn. We can get rid of EPBS. We get inclusionless as a byproduct. All this stuff that's on the Scourge part of the roadmap, we get with one stone. And it might be a big stone, and it might be a hard stone, and it might be a new stone.
But if we could do that all with one fell swoop, I think that would be good for making Ethereum look more like the v3 version of the Falcon engine than the v1 version of the Falcon engine. I think that's what really attracts me to this conversation. This phrase eating complexity has been thrown around a lot when it comes to this idea of multi-proposal architecture. We kind of like set up this MEV landscape part of this discussion, talking about all this incremental discoveries about MEV and then the next incremental patch about that MEV. Kind of not addressing the root core of the issue, which could be the fact that everything is a leader based system.
And in multi-proposers, you at least are leaderless-ish. When we started this multi-proposer conversation, we talked about my super sophisticated, highly bandwidth connected proposer. And so there's still a leader-ish part of a multi-proposer that we need to kind of figure out. But the idea here is like, call it four, eight, some square number that we'll probably select. if there's eight proposers, all of a sudden it becomes much less leader, much less of a leader based system. And you're saying that this is an elegant solution to eat a lot of the complexity of all these patchwork incremental additions to the Ethereum research roadmap, which feels elegant.
And I know that in the Ethereum research space, we like elegant solutions. It just feels good. Everything kind of snaps into place a little bit better. And that's kind of like one of the reasons why we're talking about this in the first place. Yeah, exactly. I would just say the other designs are not necessarily bad, but I would say there's not a lot proven about them. And that's kind of... If you want to talk about, Charlie, if you want to talk about unintended consequences, latency games, pushing the MEV elsewhere, take a look at these proposals, catch up on what the latest research is, and you'll find quite a few issues, I think, if you read them seriously.
And that's not to take a shot at anybody. It's just how these problems are really gnarly and tough. And like George just said, when you do something, you move it somewhere else. I completely agree. I completely agree with everything you said, actually. And to be clear, I'm actually very bullish on multi-proposer or multi-party block production, as we talked about. I guess my two cents would just be I think that we should be basically very cautious about the existence of sophistication and exactly how that will express itself in whatever market structure we kind of we try to create. And I think that to the extent that we're ever imagining that parties are going to be on a level playing field, that we should have like really strong reasons for believing that.
And I think that is where like a lot of the research around Braid is focused. Can I just say, I think, like, if you are being intellectually honest, our strand of the research, our original paper, has been some of the most tackling head on that sophistication and rationality of the actors. And we understand it and we talk to the rational actors, just like you. And I think, like, we're very aware of that. And we're economists. economists. So the way that we model things is through rational actors participating in games, and we eat the complexity of that to try and get good proofs. So I think if you actually compare what we've done in terms of tackling incentives and rational actors and latency versus what some of these other proposals have done, I think we're actually far ahead in terms of analyzing the ultimate outcomes.
Yeah, I agree. And I'm pretty sure that whatever, I don't have like a decided opinion on what Ethereum should do yet, but I'm fairly sure that wherever my opinion lands is probably going to be, like almost certainly going to be derivative of the work that you guys started, whether it is multiplicity or braid or, you know, whatever the most efficient implementation path is. Again, I would say, and not to belabor the point, my like narrow qualm would be, can we remove the existence of a last look? I think that's probably one of the biggest outstanding research questions. And then to the conversation we were having earlier, even if we net it out to that we can't, I'm very open-minded to the idea that we can, or we can have really good arguments that it's not as bad as we are nervous that it could be.
Then, yeah, I think the implementation route that Ethereum takes to something like this, some method of improving censorship, is probably either going to be like multiplicity or braid or you know or whatever else is in this line of research yeah i mean how awesome and not in the spirit of the vitalik tweet that we went from two years ago doing these vast you know mev taxonomy shout out to tina and trying to figure out what was even going on and now we're down to can we get rid of less look yeah that's so much more tight of a problem i agree and i think we're getting that is a good place yeah i do also want to talk about the improvements that uh multi-proposers might bring to defy specifically i think we all would might agree that defy is like the first primary big use case of blockchains past just simple money uh what's the relationship between a multi-proposal uh architecture and and the defy vertical so my like i said my goal i said this on twitter my goal is to make the blockchain so awesome that the majority of euro usd volume trades on chain and that might be a pipe dream i currently think that the probability of that happening is greater than one percent which is why i'm still working in crypto um i really am excited about financial possibilities on chain but i think this is we have turned complete smart contracts if we add on censorship resistance they're going to be even more powerful some smart contracts are what got me into the space by the way it's like really cool primitive and you can do anything with them that's what turn complete kind of means and then you add on censorship resistance you can do so much more so like what are games going to look like when you can't censor the opponent's bet you know that's going to be sick when you can play poker on chain and all these other things that I think it will enable as well.
And then there's also a conversation around the layer two to layer one relations that I think this brings. And this might open up the door into like a much broader subject. I'll pose this idea and maybe you guys can reject it or not. If we are, say we get past the hurdle of a multi-proposal architecture being a strict improvement. Max here has done all the proofs that he needs to write in order for us to all agree that multi-proposals are a strict improvement. So we're in that universe now. And then we also merge it into the Ethereum layer one. So the Ethereum layer one is running this multiple concurrent proposer paradigm.
This, in the bending of the priority of execution on layer twos versus execution on layer ones, bends back the priority of the Ethereum layer one being a place where really good execution happens. Whereas Ethereum previously has always been about pushing execution to layer twos. And all of a sudden, layer twos have this single centralized sequencer doing this one role. But on the Ethereum layer one, we have this multi-party, multiple concurrent proposer execution environment, which is better than it previously was. Does this change anything? Do you guys think, how does this impact Ethereum's previous and current roadmap of the rollup-centric roadmap?
Does it at all? If we are prioritizing execution on the layer one at a higher level than we were previously? I'm not sure it does. Okay. rejection rejection of the idea yeah why would it right like the there are two separate things almost because doing more on l1 means that you have more space to do things with and this is not creating more space this is just saying how do we utilize or who who writes to that same space the block. So don't think there's a change. No, I don't think. Maybe this conversation... Naturally, one thing to address is that you said the centralized sequencer L2s. I think the whole Layer 2 space in this is in the infant or adolescent phase where it is in that world, where it has one sequencer, has upgrade keys, all of that.
I think zoom out and apply the same maturity that ethereum has where ethereum we said you know it's like facing the hardest mev problems and we'll go down to the last look and we'll have fault proofs and we'll have a sequencer that is not one person that is an actual consensus network similar techniques may apply and actually because you're not in this uh eventually consistent world because i would think that layer 2s apply standard bft algorithms i would think that the multi-proposer bft uh world would become much more exciting so i I think the ideas apply on layer two, whether it is fossil, this or something else.
Yeah, but I don't think it affects really how to think about how much, you know, activity happens on layer one or two. Maybe if I was to ask that question again, I would need to introduce this also other potential idea of increasing the Ethereum layer one block speeds. Because then we are increasing space at the layer one. How much? But now that's a meaningfully differentiated conversation for multi-proposers. Yeah, but like how much are you going to increase it, right? Like 10 exits, like that's not enough. I think I agree with George's that compression is incredibly important. And so like we just probably can't get everything we need from the L1.
I think multiproposer enables some new ways that base rollups can work as well. so base rollups right now the only thing you can do is basically make the block that maximizes the payout to the l1 proposer because again the l1 proposer is a monopolist but what if you have like a base rollup you can define an arbitrary objective function which is like i want to make the user welfare the best or i want the user welfare to be the best subject to these constraints anyway you can kind of make a base roll up based on that on a censorship resistant version of the l1 and then to the point that george was saying like uh these are problems that are going to need to be tackled somewhere like if the l2s are where execution happens and we want decentralized l2s then they're going to have to deal with this multi-proposer problem as well because censorship is just a universal constant and that they're i hope going to look a lot of people ask me why i'm trying to work on you know maybe getting this into the ethereum research circles and maybe even into the protocol and they're like oh why are you doing this especially a lot of people ask me this why are you doing this because you know ethereum is so ossified they don't do many changes but i that's actually just not true like we shipped the merge that's a huge change and we can do more huge changes if it's the right thing to do that's not even the most recent big right thing that i think would change we did uh way for four recently too it's like we we've been doing tons of big changes and we've been doing it safely with minimal disruption and so uh that's a contrast to other ecosystems that are doing big changes very frequently with frequent disruptions so right Oh yeah, like the daily Solana hard fork.
But also even on layer 2, there's sometimes a new upgrade might be announced like a week out or a few weeks out. So the iteration cycles are much different there. Two points to add. One, it's actually interesting that maybe on the layer 2 you don't need as much of the sequencer being censorship resistant just by virtue of the fact that you can use the L1 slow path to prevent the censorship. that granted that doesn't address the mev part of it um and what was the other part oh the other part was that in the spirit of talking about large changes ethereum has the we'll see how much out it is but like the upcoming statelessness uh upgrade and that's actually i think elephant in the room the biggest actually breaking change that we're making and i would put i would think of the mergers here i would braid and statelessness like on the equivalent level of like breaking or like big big meaty changes we need to think really hard about um and yeah we should evaluate it like with a similar level of scrutiny what other uh parts of the ethereum ecosystem might abrade implementation impact we just brought up base roll-ups i believe pre-confirmations just get completely eliminated.
So you can still have inclusion pre-comps. You can't have execution pre-comps because what, what that is an execution pre-conf is basically the monopolist telling you, I will use my monopoly to make sure that you get this particular piece of state early. Right. I mean, I think there's very good arguments that you can't have execution pre-comps anyway. Anyway. Yeah. So I don't know how much we're actually giving up on that. And then, yeah, I also think it's not directly related to braid. But I think probably something everyone can, at least a lot of folks have now started to agree on um is that lowering the block time is a really admirable goal that would improve base layer ux would improve base layer censorship regardless of what else we choose to do um would potentially improve like make base roll-ups like work in a more reasonable way um or less crazy alternative yeah and frankly i think that a bunch of the downstream research questions of how do we get faster block times without compromising on decentralization are just like way better research questions than like a lot of um you know like where a lot of energy has been going um there's a lot of other stuff i'm interested in too but i think like if we yeah if we can get lower block times without compromising too much on decentralization slash in a reasonable way that would be like um yeah i would love to spend like a lot more time on that i think that would be awesome for ethereum before we before we formally close the door on the Braid multi-proposal conversation because I feel like we're coming to an end on it.
I want to make sure that we've covered all the bases here. Is there any part of the Braid multi-proposal conversation that we didn't touch on that I'm skipping over? Well, I guess since we're about to shift into, maybe I can just segue us in with Braid. One of the things that is really difficult in terms of getting shorter block time, and we see it on Solana as a symptom of this is basically the leader rotation. And so Solana has 400 millisecond blocks, but they actually have 1.6 second leader rotation. So the leader has four blocks in a row. This isn't great when the leader has a proposer monopoly, but if they don't have a proposer monopoly, we can be more accepting of them proposing four blocks in a row.
So I think Braid could actually potentially help us get to shorter block times through this vector of not having to rotate the leaders as much. I think that there's a lot of weight on what Vitalik tweeted about in Can We Remove the Last Look? I think that I probably sit closer to there's possibly an impossibility result somewhere in here that I'm nervous about. And I am very excited to see what future work comes out of it. I think you also will change your mind tomorrow as you're wanting to, Charlie, but I will show you some evidence that the last look isn't quite as bad as you think it is. I know the model that you're thinking of in your head because I wrote about it in the paper.
And I think that there are a lot of arguments for why it's not quite as bad as you're thinking it is. It doesn't unravel completely the way that you think it is. So we can chat about that another time. But I think last look is obviously not good. I think 50 milliseconds of last look is a lot better than two seconds of last look, just to be clear. And I think that 50 milliseconds of last look isn't the end of the world. I would contend that though, right? Because the 50 millisecond advantage is accessible to two people, whereas the two second advantage might be accessible to a broader audience.
So I think when I think about... No, it's accessible by definition to one person in the fossil design. No, he's making a point about centralizing effects. Actually. Yeah, I mean, I think everyone's super excited to hear the arguments, I guess, like, my sort of epistemic humility about it is just like, you know, so far, we've never really seen like a trustless mechanism where power in a lot of different forms doesn't concentrate. um and so i think like probably i'm most excited about a roadmap where we pursue some form of braid or multiplicity and continue quarantining the last look um and i worry quite a bit that if we don't do that um i mean we don't acknowledge the last looks existence and quarantine it then the metagame that we end up with is going to be not what we intend but in really subtle ways that are actually like much more destructive.
And the simple analogy I would give is like, there was a point at which Solana had no mempool and then Jito created a mempool. And then you started to see sandwiching because Jito created the mempool, right? And then everyone gets mad and says Jito caused sandwiching and then Jito gets rid of its mempool on Solana. And we fast forward like a few months and it turns out there's a bunch of sandwiching still. And why is this? Well, it turns out it's really interesting. There's a bunch of different reasons. One is like, I think like a Russian or like some weird country like sourced validator has accumulated a bunch of stake in order to be able to see transactions there's a bunch of backroom deals between different validators and different people with views into the mempool from rbc providers and that kind of thing and i guess like i'm yeah i'm super interested to see the future research but i guess where i land on it today is just like um we've so far never been able to create a mechanism where power doesn't concentrate that way and rather than sort of pretending that it won't or closing our eyes to it I generally prefer the direction of like corn acknowledging it and quarantining it and I think we still got a much better system on a bunch of dimensions if we do implement braid or multiplicity for the weaker form of censorship resistance that that enables I would just say I don't think that what we're doing if you look at our research I think Melesh and I in particular at SMG have actually done the most work on latency games and of anybody in the space actually like we from a research perspective in terms of the microstructure of ethereum and quantification of latency games understanding these centralizing effects half our papers are titled centralizing effects of x we have a new one coming out centralizing effects of a tester or propose a separation.
Like you're preaching to the choir here. We absolutely acknowledge those things. We're not going to, you know, try and push ahead with something that we believe is inherently flawed on those dimensions. Cause those are the dimensions that we spend a lot of our time thinking about. And I would just say like, we shouldn't try to paint braid as like, oh, we're going to ignore last look. Absolutely not. Like, in fact, we have a specific desiderata on how much less look there's going to be and we want to prove properties about that the quantification you mean yeah i think i think in some sense like there's almost no quantification of what that looks like in the fossil design right now which there definitely there definitely isn't but it's also not a goal because we agreed earlier on that the goal is just for you know not the traders i don't think that's the only goal at least the early versions of multiplicity that wasn't just no totally i'm just telling you that the reasonable people might disagree on what you can achieve with a simple easy to implement mechanism um and maybe the point is that this is like as far as we can get with a simple enough mechanism and then maybe okay if you can do beyond that might require a bigger change that's much more and just to analyze like i am really like i said focused on the traders and and making that's what my research is about that's what i care about that's what my goal in doing research in crypto is is basically make sure the trading experience works and make sure markets work and make sure spreads are tight and i don't care about as much about the other stuff that these other things are trying to get.
I think they're much, much, much less important for us to do. And if the complexity overhead is similar and we get a very nerfed version, as Dan Robinson, your colleague says, a nerfed version of inclusionless that isn't for Mev, what's the point? We're going to do almost all the same work for this thing. And that's just... My contention would be that, you know, I kind of like, I never really related to like a thermal one as a place where we're gonna like play all the you know long term like the high value trading I always thought that like we would move it out to the layer twos and we can actually quarantine and solve the problem in like more tailored ways so that might also be part of like my embedded assumption here that actually for me the l1 is like for the censorship resistance and again transitory phase layer two ends up having all the activity and that's where you can like play more play more with the parameters of the system whereas the layer one it's a bit more sensitive to me to like start to try to do too many things yeah i think if if the l2s end up being the ones who use braid because that's where it's relevant and that's where the trading's happening i'm not going to be uh not going to be miffed about it i think that's a fine yeah but i like the l1 i want to make the l1 great too george s how's russ going pardon how's rest going amazing um we've been doing a lot of work uh over the last year uh we cut the first 1.
0 release of the red project uh a few months ago double months ago after we did a security audit with sigma prime who are the authors of the lighthouse project big shout out amazing team and we love working with them um actually like we have been we have contemplated in the past implementing consensus implementations, not necessarily for L1, so we are not interested in doing an L1 consensus implementation, but for decentralizing sequencers on the layer two or for facilitating communication among off-chain services like Eigenlayer, AVSs, or symbiotic restake networks. So generally really excited.
We employ almost 20 people right now working on open source. I think most of the trading that happens on the entire EVM ecosystem uses our infrastructure, whether it is Dragon Rakita's, our EVM implementation or Foundry for developing smart contracts, or even builders using our code to build blocks. And notably FlashBots a couple of months ago released the RBuilder, which is an MEV builder, open source, written in Rust to level the playing field between things. So yeah, I've never been more thrilled. And next week, we're actually throwing a conference in San Francisco where we'll showcase for two days what we've done and hopefully get people excited about building more with it.
I've heard rumors that Flashbots isn't the only builder using some parts of Reth for their stack as well. Totally. It might not be open source. Totally, yeah. What are your goals for Reth over the next year or so? What do you want to see happen in the Reth ecosystem in 12 months? Yeah. So how I relate to the Reth project is that it becomes this SDK that everybody uses to build nodes and other EVM and beyond EVM-centric infra. And, you know, people want chains for a long time. That's why we have the substrate, the Cosmos SDK. And I think Reth's final form is that it actually becomes an evolved version of such projects.
You would use Reth to build Ethereum L1 nodes. You would use Reth to build OP Reth, Arbitrum Reth, Polygon Reth, whatever. You can name a lot of things in that direction. Actually, this morning, we were talking about how we could implement a Starkware-flavored version of Reth that would use Cairo as its runtime. It would be a great stress test of the flexibility of the, as I've been calling it, the Reth SDK. So how I think about the RETH project is a stability for Ethereum L1 and we hope to push forward the research from there. That's why this conversation is particularly interesting to me. It's been really humbling to be a part of the core dev process.
It gives you new appreciation for being sometimes more conservative about what decisions to make on the protocol because you have to be really precise, really careful. But also it has given us great appreciation for how we can bring our more aggressive shipping to Ethereum while respecting this constraint of being very principled. And we think we strike a fine middle ground right now. So that's on the Ethereum L1 stability. On the Layer 2 world, I'm really excited about breaking through what we've been talking about, the gigagas barrier, which is this. if you think of a median ethereum transaction being 100k gas one giga gas means 10 000 transactions per second and we think that we can give every layer to in the world that for free and i think it's very important to get the ecosystem to stop talking about performance almost i don't think it's you know it's kind of like funny because we think a lot about performance but I kind of think it's really unexciting to talk about.
I think it's table stakes. And in the cloud world, you know, if you want more performance, you just pay for it. And that's it. You know, you up your bill a little. You don't raise a hundred million to go do it. So super excited for breaking through the performance barriers with the Reth Project's next steps. And we'll have some exciting, we're going to cut a new release in two weeks where we hope to have a very exciting benchmark for that. And the third thing is I just want the RETH project to facilitate the experimentation. I think it's very important for nodes to stop being thought of as this scary, huge monolith that people carry.
And we talk about modular blockchains. I don't really relate to that term. I relate to a modular code base. And by having a very modular code base, You can onboard new people very easily, can iterate very fast. The last 30 days, the Reth project had 500 pull requests and 50 contributors. And the people that we work on Reth are like 10. So you can see how by being very intentional about having clear testing process, a clear documentation, very clear separation of concerns, can move very fast without actually breaking things. And I actually find that really inspiring. And it's been just an absolute privilege working with people from the core devs, from layer twos, and our own team to make that a reality.
Yeah, I do think like every single corner of Ethereum has at least one eye on the Reth project and the progress. So I think everyone's rooting for you. Yeah, we actually have been working also on, you know, like the Reth project is not just like for nodes, it's for building additional infrastructure. So recently we've been collaborating with Succinct, who's building SP1, which is a ZKVM, to basically just import Reth code into an SP1 program and just run it. And that's really impactful because it means that now, OK, you can do a ZKVM with no code changes, reusing Layer 1 Ethereum code. And that's, again, very important because instead of doing cowboy modifications that come from custom code that, again, somebody raised $100 million to build or something, you're just going to reuse production code that's used and secured and is trusted ultimately by a very wide set of production operators.
So all that is to say, very pumped. It's been almost two and a half years doing this. And yeah, I'm excited for everyone to go and run Reth in production. cool can i just say i feel like there's kind of a conspiracy theory among the paradigm uh skeptics who say oh wrath is kind of a design to take over core devs and have influence but i think if you just listen just listen replay the last five minutes of georgius you know very excitedly talking about the tech like it's clearly not that right like it's clearly just like people who want to like make the system better and fix things and we like the tech yeah it's it's just like so not that nobody here we we work i work at consensus we have obviously interests but like the goal is not that it's just to make ethereum awesome and if we make ethereum awesome then everybody is going to be better off and we're going to grow the pie 100x and everybody's going to be 100x and don't get me wrong we have views right and the views might not always align with everyone But the views, and yeah, we have a portfolio and there might be actual interests, you know, but it is important that you surface the interests and you can have a real conversation about them versus being, you know, kind of like sus about them and saying, oh, I want this EIP to go in.
No, you have to show proper motivation. And for example, recently in the CoreDev conversation, there's been a large push on EOF, which is an EIP for improving the EVM. and we think it's important to push that forward to make the EVM better. It's not like there's some ulterior motive because maybe it would make someone else better. You just need to make the system better. You need to bring sufficient motivation on the table. And that's how high the bar for the core dev process is. Charlie, are there any other paths of the Ethereum roadmap, Ethereum research efforts that are floating around your brain that really excite you or interest you that we haven't yet talked about today?
That's a good question. Honestly, I'm excited. for us to make real progress on the Mev issue, although I think you've heard I think the debate that you probably heard today is actually really positive because we're finally making progress towards a shared understanding of reality and what's achievable, basically. And I think for the last, I think it's probably one of the most controversial parts of the protocol roadmap. More or less, people like the blobs. people like the gas we like you know he doesn't like love i know but more or less whatever like i think that there's and i tweeted about this recently i think there's a number of like freebie or or just like super positive protocol improvements that's like when it happens everyone's going to be stoked um that i'm excited about um but i'm probably honestly most excited about us having gone from i would kind of describe characterized as a situation where like there wasn't even that much agreement on like what the important questions in the sort of mev block production research space were to there being like really focused conversation um and like the set of things that we could pursue having gone from this like really wide map to more of like okay like we have some general agreement on like we're going to do something in this direction um and i think that's like incredibly incredibly bullish yeah we're arguing about whether we're going to have 100 milliseconds of last look or two seconds of last look it's a pretty pretty good place to be at yeah i one thing i want to say i think we as uh because it's an attention game in crypto and attention is a currency and the more attention you get the more everything you get there's a tendency for the excitement around ethereum to be on the biggest possible proposals and braid is a big proposal like you said jordis i would agree it's maybe even more of a lift than the merge itself.
But I'm also working on a few other things. And I encourage everybody within the ecosystem to work on these kind of like parameter tuning things. And maybe it's more suited for kind of an economist type of thing. Let me just mention two of them. I think one of them I've been harping on is the lack of a reserve price for blobs. And it's not just that blobs are being sold for seven way which is zero dollars so like in in a whole month in june i believe there was all of the blobs sold the price of them was five dollars total in 32 days right and similar for other da networks right right similar for other da networks so i think the the argument is oh maybe we should price that at the marginal cost of the network.
That's an important point. I'm doing the work right now to try and figure out what that marginal cost should look like. Another reason you want to do this is because seven-way is so small. It's like oftentimes 12 orders of magnitude smaller. It's not worth the bandwidth. It's not worth the bandwidth, but more importantly, when we want the fee to kick in, in times of congestion, it takes a long time to kick in because of the way 1559 works. It starts so low. You can only adjust so much. And when it's nine orders of magnitude, 12 orders of magnitude that you need to increase, it takes a long time.
And in that time, the fee market is not operating properly. So my argument is let's bump it up, put it up somewhere near 10 to the ninth or something rather than 10 to the one. So like a guay instead of a way. And it will take many fewer cycles to get up to speed. and i think that would stop a lot of the spikiness that we see going on and it would make the market slightly more healthy and so i'm gonna be working on that another thing i'm gonna be working on i'm giving a talk tomorrow at spc about which is some slight adjustments to the way we set parameters for execution gas uh in 1559 and i just say like we need to also we can think about big things and they're awesome and they're fun to think about and they're nerd snipes but we should also just like do the little things and get the little things right and i think we probably should have put the reserve price price floor in when we did 4844 in the first place but now let's put it in in the next hard fork it's a very small change the code for implementing it is already there you just need to set the parameter and i think we should do more stuff like that as well um because those little one percent improvements are going to add up a lot as well i think there is agreement And I think this is not an unpopular take, basically, that there probably should be made some adjustment to how blobs are priced.
Max convinced me that the gas fee update rule is too fast. That would be a good change. I think everyone wants Max EB. I have a running list. Oh, yeah, Max EB. Can we talk about Max EB? This is a really important change that's coming in the next hard fork in Pectra. which is basically right now we have 800,000 individual validators. There's a 32 ETH cap on how much you can put in your validator. And we have 800,000 of them right now or even more. Maybe that's the last time I checked. It's probably more now because everybody wants to be staking on Ethereum. And the problem with that is places like ConsenSys, we run about 15,000 nodes.
And we are running 15,000 nodes on many fewer boxes than 15,000. And so we sign a signature for each of our attestations, especially during finality, which requires everybody to attest. And we have to aggregate all of those signatures. So we have 800,000 signatures for less than 10,000 boxes. And MaxiBee allows you to stake much more in a single validator. And so we'll be able to reduce the number of signatures that we're sending for the same information. And that will, in turn, reduce message overhead in the system. and strict upgrade it's a it's a big bottleneck for a lot of stuff and less cpu yeah less cpu which is an even bigger issue right now because the network let's say you can make it faster but the aggregation is cpu bound so you're kind of hitting a hitting a limit and we can change the way the signatures work because right now they're kind of tailored so that the signature aggregation libraries work really well and if if we don't need to do that as much because we have 15,000 signatures instead of 800,000 then...
And that might actually be basically why we cannot lower the block time as is. Say you cut the gas limit in half and you double... You cut the gas limit in half and you halve also the block time so to keep the throughput constant is because of how many signatures we're aggregating. There was an original hypothesis that the 32-eth limit stemmed from which is largely related to the idea that we were not going to have stake pools. Like we were not going to have a lot of people pooling stake. And I think that we're just kind of at a place where like the trade off to the amount of signatures we have to aggregate that are really just the same party.
Like it clearly there is a lot of pooling and we're bearing like a huge cost for it. And this is not like an anti-solo staking take, but it's just like, it really is just like a strict upgrade. I think like it turns out the hypothesis was not right and we should like recognize that and like lower the number of signatures we have to collect massively uh which gives an avenue to lower block times higher gas limits etc um i think it's yeah and it's it's kind of nice because you can still have 32 or if you have more just yeah it doesn't do anything to solo stake there is there has been and i'm not a cl dev so i cannot opine specifically on like if if this is true or not but i believe Terrence and friends, but basically, implementation complexity-wise, there are reasons because right now the 32-eth is like a constant in your code.
It's like a parameter, whereas now you have to make it a range of parameters and that complicates the code, actually. So pushback around it that you might have seen relates a lot to... No, it's not like oh, I changed 32 to 24-8 to 20-48. I change it from a single value to a vector, and that is all around the code base. That's where the complexity comes from. Yeah. Yeah, because to my understanding, MaxEB is more or less has consensus about inclusion. It's like, can we ship it in time and all of that? What exactly is the implementation? The cleanest way to do it so we don't introduce tech debt.
Yeah, which says something about thinking more about things up front. And this is a classic software engineering thing. how much time do you spend designing versus implementing and you know how do you post more time and then go back and redo the thing right also super bullish that eth has a spec actually yeah actually it's like the cl side is like so much better done than the el side the cl spec is amazing and on the on the el we're still i think it's like universally accepted that like basically geth is like the the the truth and you adjust to it which is not a problem at all but just you know it makes the development of things different where the cl has a very crisp development process which the el corda process should aspire to this is especially nice for you know think boys like myself who just we don't want to necessarily dig into the rust or anything we just want to see the spec and see what's going on and many times the code that you read is like optimized and like it might not like be obvious it's like some random bytecode like thing that It does it faster, but you have no idea what's going on.
So the spec is very useful, and we definitely, we did a recent paper on exit queues, and I spent a lot of time looking at the consensus layer spec. Charlie, why did you bring that up about the Ethereum spec? Just because with these kinds of software projects, even as far back as Bitcoin and Bitcoin or ConsensusD, a lot of stuff, to Georgios' point about Geth being the reference implementation, you run into this issue where like the original implementation in a sense is the spec it has like you know lots of implementation nuances or choices that weren't it wasn't necessarily anyone thought of them or thought about the decision like it's just the way that it ended up getting implemented um and like i think ethereum on this spectrum like is doing quite well which is like most stuff is pretty formally specified the consensus layer is great because it was almost we realized the problem and tried to solve for it from the beginning but even the el like you know you have a pretty good sense of how the protocol is supposed to work like what the intended behavior of it is um and i just think that's like very uh great i mean like i think you know ethereum as like a super decentralized uh like credibly neutral censorship resisting computer like knowing how it's actually supposed to operate is like extremely useful yeah and you You know, people like to, there's these two schools of thought.
Should there be one client or many clients? There's many protocol security reasons that, David, you might have talked about on the podcast, so I will not repeat. But basically, the most important reason for me is that more people knowing how the thing works is just good. Right. I remember Vitalik actually brought up this line. There's no real point of having these decentralized systems if only 10 people know how they work. I would like to reason by analogy a bit. uh like i would say it's almost like having a checkpoint so that the node can you don't have to run the node from genesis because i i'm like relatively new to the research like med research even client research is even more new to me and being able to go read the spec versus having to read the code means that i can catch up faster and like we're we're always producing new stuff and changing stuff and so like i look at the speed at which solana changes and i'm not looking at a full time but i try to keep up a little bit and then i'll go in and i'll talk to like some of the judo guys and like they're like oh yeah by the way like we changed the way that everything works since the last time we talked a month ago yeah and it's and another change is going in in two weeks by the way do you have any feedback about it's like you know and then i'll i'll try to get feedback but it's so much harder than the way it is working on ethereum because we all know and we all agree and our our models are lined up we all agree on what the model of the thing is and works like that and on solana sometimes they agree on what the thing is supposed to work like and it doesn't work like that or the other way around so i think that's a definitely a good thing about developing honey and then you go to mainnet validators and you ask what's going on yeah and then they actually don't run what's in the code base that you looked at so well guys it's been a fun little charcuterie board of uh ethereum protocol topics at the very end of this but i want to bring it back to you max and braid uh braid work in progress who do you want to hear from who can support you, what are next steps, any call-outs to the listeners in the audience of this?
And who can roast you. And who can roast you. Well, I would just say, like, we're welcome to hear roasts. I would say it's often easier to get people to engage deeply with an idea when they first hear it. And almost I wish we had waited a couple months. I gave an early talk at this Paradigm Research Day, and there was a lot of excitement about it, and it got out, which is awesome. and that's why I'm here talking right now. Yeah, I got excited specifically. But I just want to urge everybody to keep your minds open. Even if you're skeptical today, read with an open mind. Do not immediately shut your mind off to these things.
I'm spending a lot of my day thinking about this stuff, thinking about new ideas every day. And I think we can probably get there. I'm very optimistic knowing the awesome people that we have working on it. and what I do know about Mev. And I think just keep an open mind. We're going to do the work. If we can't demonstrate it, it's not going in, obviously. And there's many different sides of the argument, but I'm really excited for basically the first time about the direction that we're going. For a long time, I've been like, oh, we're kind of trending in the wrong direction. PBS has become very centralized in the time that I kind of started working at the maximization point of the decentralization of PBS.
And I wrote a paper that said PBS is going to centralize and then centralize in a year. And so I was like, Oh, what are we doing here? I was like, Oh, I'm going to go back to grad school because like, nobody's listening to my ideas. So I'm going to go like write some more papers in grad school that nobody's going to read there. But now I'm like, Hey, like people are listening. you know there's ideas we're discussing the difference between 100 milliseconds and two seconds of last look and that's just so much better than where we were a year ago and even where we were two years ago and we're just starting to talk about these mem topics in detail well georgios charlie max i feel like i have gained like at least a couple iq points from just listening to you guys today so thank you guys for making your way over here from SBC and joining me in the studio today.
It's been really fun doing a podcast in person. Thank you, David. Thank you, David. Cheers. Thank you. Love these boys. Bankless Nation, you guys know the deal. Crypto's risky. All that good stuff. You can lose what you put in. We're headed west. This is Frontier. It's not for everyone, but we're glad you're with us on the Bankless Journey. Thanks a lot. you