Protocol economics the Ethereum protocol’s perspective - Vitalik Buterin

So, next we have Vitalik. He will talk about protocol economics protocol economics , uh, from the perspective of the Ethereum protocol. Good. Thanks, guys. Can you hear me well ?

Yes, we hear you. Okay, great. Um, today I'd like to provide a brief overview of what the protocol economy looks like in general , focusing on Ethereum and some of the changes that have occurred in both the Ethereum protocol itself and the Ethereum application ecosystem over time, and also touch on some questions of how this interacts with events that may happen to the Ethereum protocol and application ecosystem in the future. Um, some of the protocol economics issues protocol economics issues that matter are pretty basic. One of them is: what incentives are in place for users?

Second: what incentives do consensus participants face? That is, miners and validators are meant. Um, what is the network behavior, both in normal and extreme cases? And how does network behavior affect incentives at the application level? For example, if an application relies on certain network operations for its incentives, then if the network suddenly starts operating differently due to an attack, it could cause problems for the application.

And how do apps affect any of the incentives? Incentives for regular users, incentives for consensus participants consensus participants , incentives to do things that can break the network, and all that kind of stuff . Um, there are some pretty complex interactions between all of these, are n't there? So, users send transactions to the network. Users read data from the network.

Application state exists within the network itself. And so we can say that this state affects or the network affects the state of the applications. Applications create incentives for users, and applications are the reason why users send transactions with certain data sequences instead of other sequences. Consensus participants maintain the stability of the network, and the network provides certain intentional incentives to consensus participants consensus participants . This includes mining rewards, validator rewards, and various other rewards defined by the protocol .

This also includes fines, and the apps provide additional incentives, right? So there are intra-protocol incentives and also extra-protocol incentives, and ultimately this whole MEV situation this whole MEV situation is about extra-protocol incentives that arise mostly by chance and how they ultimately affect some of the incentives to keep the network running. So, some risks to look out for: look out for: first, for me, and this is what we've been focusing on the longest, are incentives for validators to act correctly, not the other way around; specifically, the incentive to continue developing the chain, rather than trying to carry out a 51% attack or to carry out a 51% attack or , for example, censor the chain, right? And a kind of struggle between the user incentive user incentive , which is simply the transaction fee, and now the priority fee for inclusion in the chain, versus the potential bribes for censorship. And, you know, can censorship bribes happen by accident, or is this something that can happen intentionally?

And trying to set up the economy so that the user incentive has the upper hand and the side incentives don't dominate, right? In many ways, this has been a big part of the discussion around things like resistance to censorship at PBS. But beyond that, there are also economies of scale in incentives, and one example is and one example is if user incentives are complex user incentives are complex , it can lead to wallet centralization, right? So, for example, before EIP-1559, the incentives for users on how to set gas were quite complex, and the user was faced with a rather complex choice curve between quickly enabling a transaction and waiting for an indefinite and variable amount of time. And this caused a huge amount of wasted losses because many users were waiting for their transactions to be included for no good reason.

And that also, I think, made it harder to create wallets and probably contributed to some extent to the fact that new wallets were n't being created. After EIP-1559, incentives for users in general became much simpler. You just set a high enough maximum base fee and then set the priority fee at some constant level, or if there's an active situation going on , you can set it higher using an algorithm, and you'll get significantly fewer penalties, right? But overall, the situation has become simpler for users. We also see economies of scale for consensus participants, right?

And this leads to the centralization of consensus. So, if the strategies that consensus participants see are constantly complex and changing, and especially require access to confidential information and all that, then that will be very bad for solo very bad for solo staking. This will even be very bad for decentralized pools. This will even be very bad for small centralized pools. And that would just lead to the dominance of the big pools, wouldn't it?

And I think we even saw at the end of the proof-of-work era that pool centralization in Ethereum started to become quite a serious problem, right? We got to the point where two mining pools were enough to get over 50%. And, uh get over 50%. And, uh , Bitcoin is a little better than Ethereum's proof-of-work, but not by much. And I think we need to make sure that proof-of-stake also evolves in a healthy direction.

Um, and finally, low latency incentives low latency incentives are also another kind of incentive for centralization, because they are an incentive for geographic concentration. They are also an incentive for cloud computing. And both of these factors actually contribute to jurisdictional risks, which is another very important and big risk to the credible neutrality of blockchain, as we seem to have seen over the last year. Um, so a lot of this started this started , essentially, with the rise of DeFi, right? That is, having in mind these opportunities to receive a large amount of money by being the last participant to create a block.

Many of them, uh, actually existed back in 2017. So there were these pretty famous ICOs that happened in 2017. Yes, the image on the left is a graph of gas prices, they reached an extremely high level, and there was a huge irreversible loss. I think it was during the Basic Attention Token ICO. And if there was some infrastructure to collect MEV back then, it could just be collected by just be collected by , you know, any mining pool advanced enough advanced enough to have that infrastructure, right?

Realistically speaking, there weren't enough cases like that at the time to justify the fixed costs of trying to create the infrastructure to capture all MEV. But then in 2019-2020 we had the "DeFi Summer", and at that time we had essentially the beginning of permanent MEV. Yes. Which happened on almost every block, and even bigger and wilder spikes than we saw in 2017, right? And that created economies of scale for consensus participants consensus participants , which eventually led to the need for things like PBS to try to separate the role burdened by economies of scale so that the rest of the role didn't depend on it so much.

And the graph on the right shows the daily gross profit from MEV mined, demonstrating how the underlying MEV is stable, but at the same time highly volatile. EIP1559 has significantly reduced waiting times for users, and also significantly reduced the incentive to use complex gas payment strategies, which, as I said, lowers the barriers to entry for wallets. The Merge wallets. The Merge actually reduced the wait time for users even more, and I think there will be some research coming out soon on this topic, and you know, it's something that will improve over time . Fusion actually does a lot of weird things at once, does n't it?

So, first of all, the merger completely changes the class of participants supporting the consensus, right? This means that the fringe incentives of people who support the consensus change greatly. It used to be miners, now it's ETH stakers. Miners were generally a group that stayed away and were often not even very loyal to Ethereum very loyal to Ethereum , as they switched between Ethereum and other protocols. Of course, there were miners who were very loyal to Ethereum, but there were others who were not.

These were people who had access to the equipment and whose primary expertise was in the equipment. Meanwhile, ETH stakers are a subset of ETH holders who are typically less tech- savvy, but they are also, you know, typically more integrated into the Ethereum ecosystem in other ways. And you know, I think there are complex implications of changing exactly who is mining mining versus who is staking ETH on average, right? I know that a lot of miners eventually became ETH stakers, so a lot of participants switched from one side to the other, but, you know, on average, the composition changed a lot . Mergers have also made reorg attacks much more difficult , so there was an article by Giorgio and I almost two years ago about how post- merger reorgs will become much more difficult much more difficult .

Also, there's a completely different landscape of incentives around things like centralization, joining a staking pool, what types of pools are even possible, and all of these issues, right? So, the Merger changed a lot of things. And I think the Merger is one of those things that even in 10 years, if someone were to ask, " if someone were to ask, " What were the consequences of the Merger?" You know, as they say in a famous quote about the French Revolution French Revolution , the correct answer would be: "It's too early to draw conclusions." So, that's about the current changes.

Now about the upcoming changes: withdrawals will be activated in the next hard fork next hard fork . Next, regarding prospects: secret leader elections, different types of intra-protocol PPS, and finally, single-slot finality. So, withdrawing funds: this has two main consequences, right? People can withdraw funds, and so they will do so, which will reduce the total amount of ETH in staking. But on the other hand, people can withdraw funds, so they become more confident that they will be able to get their money back, which will lead to an increase in staking volume.

Which of these factors will prevail—we have no idea; it can be different. Although, ultimately, I believe that the main practical consequence will be a change in the total amount of staking, and perhaps a change in the composition of stakers. I hope that conditions will become better for solo stakers, and the ability to withdraw funds will slightly reduce the advantage of pools. Because pools now have the advantage because they are the only way to get liquidity get liquidity , and with withdrawals you can withdraw your money in a few days. So it will be interesting to see what consequences this will lead to.

But secret leader elections are interesting: only the provider will know that he is the provider before the block is created, although he can voluntarily disclose this information and even prove it. The goal of all this is all this is to reduce the risk of DoS attacks on providers. Right on providers. Right ? If no one knows who you are before the block is created, then it is impossible to subject anyone to a DoS impossible to subject anyone to a DoS attack.

And once you reveal yourself by creating a block, it will already be online. So even if you are attacked, the block has already been created and it will be added to the chain. But an interesting research question arises: are there any unintended effects, particularly in combination with PBS, such as whether there will be incentives for providers to disclose their identity early? I'll skip the PBS topic because I feel like everyone will be talking about it, and in a way it will be a major topic for years to come. But, you know, it's a fascinating rabbit hole .

What about finality for one slot? Single-slot finality means that instead of completing blocks after two epochs, each slot will instantly finalize a block, except in special cases with an inactivity leak. A one- slot final means even more reorganizations more reorganizations . This could make Ethereum more bridge-friendly. But, as we will see with second-level pre-confirmations, the finality per slot de facto will not be that for many users.

And I think that will be an interesting nuance. So, finally, the transition to the second layer and the economics of the protocol. How does the growth of the second level affect the incentives for protocol participants at the first level? A lot depends on how the whole story develops with the second level. First, it's the sequencing style.

At ETH Research, Justin Drake wrote a post about what he calls "based roll-ups." These are rollups that allow the first level to directly engage in sequencing, as opposed to sequencing being controlled by the second level through auctions. And all of this leads to the question: what percentage of the first level will move on to the second levels? If the answer is zero, then the second levels have no impact. If 100%, then it all depends on all depends on how the second levels work.

But the answer could be anywhere between zero and 100%. Fraud proof windows and censorship. This is a pretty simple question. Pre- simple question. Pre- confirmations, which we will talk about later.

And the final question: will the transition of activity to the second level make the first level economy more boring, insulating it from side effects, or, conversely, more interesting due to more complex interactions? Therefore. Based rolaps, right? When I wrote my post about rollups about two years ago, I focused on the issue: rollups have packages that collect a lot of transactions and commit them at the first level. The question is, who has the right to submit such a package?

Option one—complete anarchy. Anyone can submit a package at any time. I noted that this is the simplest approach, but it has significant drawbacks. In particular, there is a risk that several participants will try to submit their packages at the same time. And only one of them will be included.

This leads to wasted effort in creating proofs, as well as wasted gas in publishing packets to the network. Thanks to PBS, many of these costs are significantly reduced, especially gas costs, because you can make the gas payment for a package dependent on whether this package is included and processed correctly. And this is something you will have to implement through the builder ecosystem . But if the developer exists, then you can do it. And then suddenly the approach that I call " that I call " complete anarchy," or what Justin calls " based rollup," suddenly becomes viable.

But there are other approaches too, right? Centralized sequencers. You can use sequencer auctions sequencer auctions . You can use DPOS voting. You can use random selection from a set.

You can combine sequencer auctions with DPOS voting to eliminate a sequencer if it misbehaves, which is not noticed by the first level, but which is still very important to users, right to users, right ? So, there are many options that are not " options that are not " based". An interesting question arises: do the Rolaps want to be " the Rolaps want to be " based" or not? Can the second tier absorb the complex MEV of the first tier? If the rollaps are "based", then they can't, right?

As the second-tier MEV becomes the first-tier MEV, because the first-tier PBS system simply performs the MEV sequencing for the second tier. If most of the first-tier DeFi remains at the first tier, then neither will it. Um, if there are other complex interaction effects. For example, if fast inclusion of transactions in the first level helps second level arbitrage, and whoever has their transaction first in the block becomes the first to successfully perform a certain second level arbitrage, or if there are other interaction effects, then maybe not; otherwise, you know, maybe, right? And this is one of those interesting research questions research questions .

Fraud and censorship: pretty easy censorship: pretty easy . If you can censor a chain for a week and that chain doesn't perform a social reorganization rollback, then you can steal funds from optimistic rollups. But I would say that this is not very interesting because there are much shorter-term censorship attacks that are more realistic and could seriously harm DeFi. Isn't that right ? If you can censor for just an hour, you can mine a huge amount of MEV from all kinds of DeFi projects like kinds of DeFi projects like rollups or systems like Uniswap, potentially even liquidations.

You can even forcibly eliminate people and prevent them from replenishing their CDPs, and you can do a lot of different things, right? And that's also something to think about, but it's not a problem specific to just the second level, right second level, right ? This exists at level 1 or level 2 pre-confirmations. So users usually want faster confirmations than even 12 seconds. And many rollovers don't even do 12 seconds because they want to save on commissions.

And also the finality of one slot could eventually increase the slot time, perhaps from 12 to 32 seconds. Uh, and so what a lot of rollaps are thinking about is doing these pre-confirmations, right? So you have an off-chain preconfirmation protocol that preconfirms the level 2 blocks, and then eventually you have a package that feeds the rollup blocks, and then maybe some compact proof of preconfirmation into the level 1 block. All of that gets verified, and then that actually commits the block to level 1, right? And so with that, even if you have one slot finality at level 1, the user experience is: step one— is: step one— their transaction is accepted into a level 2 block.

Step two—the level 2 block. Step two—the level 2 block gets preconfirmed, and finally, step three finally, step three —it gets confirmed at level 1. Uh, so I guess that's an argument in favor of SSF, because if everyone is using level 2 and level 2 has preconfirmation, then it's normal for level 1 to have longer slot times, and longer slot times usually make SSF easier. But still you have two levels of confirmation, and I think those two levels of confirmation can be fundamental because level 1 doesn't want to centralize as much as level 2; therefore, level 2 confirmations will always be faster. There will always be some kind of curve where a user's knowledge of how confirmed their transaction is doesn't jump from zero to 100 jump from zero to 100 % instantly.

It grows over time, so you have a certain notion of preconfirmation that is inevitable. Note that level 2 pre-confirmations are incompatible with based rollups, right? A rollup cannot be based if it has preconfirmation, because the submission protocol must force the system to respect the preconfirmation mechanism, and a set of level 2 blocks that do not respect preconfirmation, it simply has to reject, right? And ultimately, it's Level 2 and the pre-approval system that will collect most of the MEV instead of Level 1, right? So, based rollups can't preconfirm, which I think is an important point, unless you make some kind of universal level 1 preconfirmation system, which could be interesting.

Will level 1 become more boring or less boring? I mean, my hope, what I would like to see I would like to see is for the first level to be more boring, but that's not guaranteed, is it? The second layer can take over functionality and risks, or the second layer can create complex interaction effects between the first and second layers. And I think ultimately, you know, we need more analysis of a world where the second level is fully implemented. And I think, you know, mechaeconomics with second- level considerations—it's a really exciting research topic that, you know, we're probably going to have a lot of people thinking about for many years to come.

So, thank you. Perfectly. Thank you Vitalik for the wonderful report. And this brings us nicely to Hu's next talk, which will talk about decentralization of sequencers and PBS in the context of rollups.