MEV Alignment

Hi everybody. My name is Shay. I'm from Flashbots. And today I want to talk about the rollup-centric roadmap, but through the lens of MEV. I think when we talk about the relationship between Ethereum and its rollups, a term that often comes up is alignment, where for rollups, it often means being aligned with Ethereum's values, so building open source, decentralized protocols, and also being aligned economically with either the ETH token or by funding public goods.

But I think there's another element of being part of the Ethereum ecosystem, which I'm going to call MEV alignment. And basically I think Ethereum has taken a pretty unique approach to MEV compared to other L1s. And I think that the tools that Ethereum has developed, the lessons from this actually create a competitive advantage for rollups, which I think will help them reach their goals faster and better than in other ecosystems. So to sort of set the scene, what is Ethereum's philosophy when it comes to MEV? Well, one thing that I think makes Ethereum unique among smart contract chains is its commitment to decentralization, and basically this means we build systems which assume people are rational, acting in their own self-interest, not honest or doing things out of charity.

The reason for this is that asking people to be altruistic is not incentive compatible in the long run. And it also kind of requires a degree of trust and coordination, which you can't decentralize. A consequence of this is that we have basically been doing MEV on hard mode for years. We have assumed from the start that you can't take shortcuts like asking validators not to sandwich people. Instead, we have assumed or even accelerated maximizing MEV extraction at the protocol level.

Because in a protocol that doesn't have a central policeman, this is what will eventually happen. All available MEV will be extracted. So if you want to kind of reduce its impact, you have to actually minimize the amounts that's created in the first place at the application layer. Doing MEV on hard mode has been super painful. So if you have ever been sandwiched in the public mempool, you probably know this.

Maximum extraction at the protocol layer is a pretty brutal experience. But also so is not addressing MEV head on. So other chains that took a different approach have also seen their own issues. So some that have tried to ignore MEV or sort of solve it at the consensus layer have caused some interesting, entertaining and sort of infamous episodes like this one where a bunch of searchers almost took down Avalanche because they were exploiting a sort of quirk in its consensus mechanism that let them build almost 100% of blocks while they were fighting an arbitrage war. Other chains have also tried social presser, so Solana.

But peer pressuring your validators not to censor is not exactly a lasting strategy. And we've also already seen some validators get sort of socially slashed for doing this. So meanwhile, Ethereum's approach on hard mode has actually forced it to develop sound mechanisms from the start. And while painful, I think this is sort of hugely valuable in the long run. So I would actually argue it's created a competitive advantage for the broader Ethereum ecosystem like rollups.

And for the rest of the time, I want to talk about sort of three specific ways that I think this plays out. First, when we talk about MEV, what we are often actually talking about is DEXs. And one thing that often happens when you do MEV on hard mode is that you very quickly find out all the problems with your DEXs. And one thing that often happens when you do a medium hard mode is that you very quickly find out all the problems with your DEXs. Xerox did a study a little while ago where they basically found you are guaranteed to get the worst possible price if you trade in the public mempool.

And if it's possible for a bot to front run you, they will. And this is actually quite useful because it has forced us to make much better DEXs, much faster than other ecosystems. So now we have, you know, things like new versions of Uniswap. We have sort of intent-based batch auctions like Cowswap. And these are all sort of built in the crucible of Ethereum.

And now it can be sort of easily ported around to all of its rollups like Unichain. But DEXs are also just one part of it. Almost every time you transact, you're kind of generating MEV of some kind, whether it's through price movement in your trade or, you know, gas fees that you're creating. And today, by default, all of that MEV goes to validators and maybe searchers, not the people who are generating it. And seeing just how much they have sort of left on the table has created a massive incentive for apps and wallets to go and try to recapture MEV at the app layer.

So in the past two years on L1, we have developed a ton of different what I'll call redistribution products, which try to refund this value back to apps, wallets, and users. So a couple classic types of products here are order flow auctions, which basically auction off the right for a searcher to back run your trade and kind of split the arb with you. Or priority or gas fee refunds, where block builders will lower their bids in the block auction and use that profit to refund fees back to users while it's in apps. And this is actually a fairly large market on Mainnet today. This is just, I think, the last seven days of profit by some top builders.

And it's well over 1,000 ETH a week. And that's all going back to users' wallets and apps, for the most part. These tools are great for rollups, because they can help rollups create much better, cheaper, and easier gas markets for their apps. So even on low fee chains, the MEV really adds up. There are, I think, some games on base that spend about $100,000 a month just in gas fees.

And even if inclusion is cheap with MEV, contention fees can be really high. So if you wanted to, your rollup could actually copy the exact mechanisms that we use on L1. This is the rule we're using at Flashbots right now to sort of redistribute priority fees to users and upgrade Ethereum gas market to be second price, which is very cool. Not even Solana has second price gas markets. But you also, you know, this doesn't have to be one size fit all.

Rollups are a great place to experiment. And we've also started seeing things, people do interesting things on rollups that are not direct copy-paste, but inspired what we do on L1. So Unichain as an example is using priority fee ordering, which can let you do redistribution through things like MEV taxes. Also Blast is doing app-level priority fee sharing with its apps, which I think is pretty cool. Why this is exciting is that by mitigating and redistributing MEV at the application layer, you can, as a rollup, become the best place to develop crypto apps.

So a good redistribution model helps you retain and support existing apps by giving them a path to monetize. But it also creates an incentive for those app developers to stay on your platform and keep building new apps, which creates a nice flywheel that helps rollups attract and retain apps that give their users a reason to do things on chain. But second, I think if we want to talk about bringing the world on chain, we also really have to talk about one thing, which is censorship resistance. So personally, what I think censorship resistance really means is, or really comes down to, is involving many parties in building blocks. Real robustness comes from meaningful global participation and power distribution.

A block that incorporates input from a bunch of different people around the world is going to be a lot more inclusive than a block that was built in US East One. And on L1, we have sort of been working towards this goal for years. During the merge, Ethereum introduced a form of proposer-builder separation called MetBoost, which basically lets proposers outsource block building to an out-of-protocol market. And this was really successful in sort of quarantining block building and the centralizing influence of MEV to this external party party called a block builder Which helped, you know prevent centralization censorship and other problems from the Ethereum validator set But the same time PBS just sort of moved the centralization risk to another external block building market And so over the past two years we have been doing the hard work of figuring out how to take the next step Which is basically how to meaningfully distribute power and prevent censorship within block building. So this has inspired a sort of large and growing body of work.

At the protocol level, there are designs like inclusion lists or multiple concurrent proposers, which basically are trying to get proposers to play a larger role in choosing the context of their blocks. There's also work on improving censorship from censorship resistance from within the block building market. So the designs like Suave that sort of aim to create secure, performant, and globally distributed platforms for many parties to collaborate in building blocks and redistribute their MEV. I personally am actually quite excited about the progress that we're making on L1, and I also think rollups should be excited about this. Rollups can use all of this giant body of knowledge from L1 to inform their own sequencing designs, make them more robust.

And second, especially with decentralized block builders like Suave, you might even just be able to plug directly into that decentralized block building network Suave, you might even just be able to plug directly into that decentralized block building network as a rollup and inherit its resilience without actually having to recreate this on your own, which I'll say a little bit more about later. But I also wanted to note that censorship resistance is not just protocol level value. It actually really impacts your user experience. So kind of back to the earlier point on redistribution, some of the redistribution methods that rollups are trying today really heavily rely on strong censorship resistance. So as one example, since it is uni day, censorship is really bad in priority fee ordering.

That is the sequencing rule that Unichain uses. And so having good CR is actually really, really important as a design goal for L2s. More generally, involving many people around the world in block building is important for just making crypto apps work. One of the things that makes DeFi DeFi is that it's decentralized. So it's not beholden to one person's preferences or one person's compliance needs.

And this is what makes it kind of the most equal platform to trade on. And to realize this goal, I think rollup sequencers need to be global. We need to find ways to let lots of parties participate in running them. And fortunately for rollups, that is what we have been spending the last many years doing on L1. Okay.

So finally, the last thing I want to look at is how we actually scale rollups to support bringing the world on chain. How do they make their block space as abundant, as resilient, and as useful as possible? So we had the same question actually on L1 not too long ago. Back in 2020, Ethereum had this sort of problem where a small number of sophisticated traders, some of whom are in the audience, were basically doing HFT on chain. And they were spamming Ethereum a bunch to try to outbid each other at the last second.

And this just congested the chain, made it really hard for regular people to kind of get included. This was actually one of the things that inspired the introduction of Flashbots, because spam had a bunch of really negative externalities for UX. So it sort of wastes gas that regular people could have used. It wastes storage and compute. It's a bunch of extra load for your sequencer, also for P2P.

And we realized, oh, there might actually be a way to solve this, which is if we create an out of protocol auction that gives all of these traders much better outcomes, i.e. more profit, better prices, than the spam games they were playing, we could actually attract them from the public mempool and limit the negative externalities they were creating for everybody else. So it would be basically much faster and cheaper for everyone, including the traders, to get included. And in fact, we've actually found this to be universally true.

So if you do not make an efficient off-chain auction, then all the competition from MEV is going to move on-chain and hurt your user experience, either in the form of spam or maybe latency wars as some other roll-ups have seen. So we have spent a lot of time on L1 trying to figure out how do you design the most efficient off-chain auction that can kind of effectively resolve all this competition for MEV. And now roll-ups get to benefit from that. So, first of all, because of how competitive MEV is on L1, we've had to develop really performant block builders that can handle spam and low latency use cases like statistical arbitrage. And Unichain is actually using a bunch of those features today conveniently.

So through Rollup Boost, which is here, Unichain is outsourcing block building to the new FlashBots Rust builder, which we completely rewrote in the last year for performance. Uniswap is also adopting, I think someone mentioned a talk earlier, Unichain is adopting an OG sort of feature that Flashbots introduced many years ago, which is to prevent failed transactions from landing on chain, which is obviously really important for reducing searcher spam. And as we make new improvements on L1, well, Rollups can just copy all of this, either directly through Rollup Boost or by kind of just extending our open source software. And this is great, because it helps Rollups make the most use of their block space for real users, not spam from a bunch of bots, and that's obviously essential to scaling. Second, kind of going back to the point on censorship resistance, we have been investing very heavily on L1 in figuring out how do we involve more parties in block building rather than relying on one centralized entity that could have faults or even act maliciously towards some of its users.

And this one is really hard. It is really hard to build an auction that is fast, private, and decentralized. Basically no technology has been able to do this so far and all we really have today are these partial solutions. I like to call this the MEV trilemma and there's a longer talk on it I think from ECC this year if you're interested. But this year we have had somewhat of a breakthrough in this trilemma, which if you have talked to But this year we have had somewhat of a breakthrough in this trilemma, which if you have talked to anybody from Flashbots, you have probably heard a lot about trusted execution environments or TEEs, which the TLDR basically is hardware manufacturers like Intel will make these machines that have really strong privacy and integrity guarantees.

So they give you a couple things. You can encrypt data to an enclave on the machine that nobody, not even the operator, can see. You can also attest that a very specific piece of public code and only that code is running. And together this means that sort of anyone can run software like an MEV auction without compromising security. And this solves the MEV trilemma.

It kind of finally makes it possible to safely distribute block building across a lot of operators around the world. And it actually works in practice. So Flashbots, we have been running builders in TEEs like TDX at native speed for almost six months now. And it makes for a really great off-chain auction that's fast, private, and decentralized, which is great news for L1, and also great news for roll-ups. So it is super important for roll-ups that their sequencers are robust and available.

The decentralized network of TEEs that we're making conveniently is designed to be super resilient against censorship and faults. And so you could actually quite easily take that same network and use it to distribute the sequencer for an L2 around the world. It even actually makes it possible for other teams to run your sequencer, so you're not even, you know, you're resilient to faults in your own infrastructure. And we have a pretty exciting release related to this coming out later this week, so stay tuned. And it also brings me to the final way that I think rollups can benefit from an efficient decentralized block auction on L1, which is to help them reach their own decentralization goals in a way in particular that preserves the things they care about like speed and redistribution.

In one of Vitalik's recent posts about rollups, I think he commented that it's so hard to decentralize your sequencer on L2 today that you may as well be building a whole new L1. And teams that have really prioritized decentralization from the start have also found this to be true. So what if you didn't have to? What if the L1 could just solve this for you? So far, one of the leading proposals on this front is an architecture called based rollups, which essentially just says, hey, what if rollups can use the L1 proposer set as their L2 sequencer?

But there are a ton of drawbacks. Some L2s don't want to forego their MEV revenue to the L1. L1 block times are notoriously very slow. And all the solutions we thought of so far to make them faster like precomps are potentially quite centralizing. So what if you don't want your rollup to be based?

What if you want to control key features like speed or redistribution while accelerating your decentralization roadmap? What do you do? A modest proposal today. What if you could just plug your rollup into the same decentralized off-chain block building network that powers the L1? This is what Rollup Boost does.

This is what's under the hood. The same fast, private, decentralized L1 block building network with some extensions and integrations for rollups. And I'll call this idea built rollups, L2s that are sequenced by a decentralized off-chain block building market that also powers the L1. If rollups can very easily adopt this, as I think they can, without having to migrate to entirely new architectures like Ming-based, they can move a lot faster and scale safer to reach their goals. And by the way, if you do this, you also start to get kind of some nice things, maybe like shared sequencing, or at least the foundations for it just for free.

But that comes out as a bonus, not something that rollups have to sort of contort themselves to be compatible with, or that requires us to change really fundamental properties of the L1, because it all happens through an efficient off-chain auction that has been sort of hardened and refined over the years that we've been Doing MEV on hardware mode on one I think this is what we should be thinking about when we talk about alignment Not just how rollups can extend the values of aetherium But how they can use the tools and work we've done on a as a competitive advantage to kind of advance their goals. I think this is how we maximally leverage the base layer to create a more vibrant ecosystem through its rollups. And I think the future is bright. So thank you.