Tackling 2 Billion Dollar Problem In Production: OEV Design That Works

Transcript

one one. Okay, let's uh kick off. Uh thanks a lot for coming over. I hope you enjoy FCC even though it's uh pretty hot over here. Uh my name is Marson. I'm co-founder of Redstone, which is the fastest growing Oracle on the market right now. and I'm going to talk about uh tackling the Oracle extractable value problem with a design that works at scale in production. So to kick off giving you a sense where we are right now, uh liquidations are a nice fight. I believe you've never participated in knives fight. Uh if you had there's chances you wouldn't be here today,

right? So as you know there are casualt casualties in nights fight. It's sometimes was needed in medieval ages. We also see conflicts nowadays but in general the principle is usually like less conflict less casualties the better EV for everyone. So it's not necessarily the situation we would like to keep having. Why is it a nice fight? So today's situation is that when there's a protocol and there's a liquidation transaction, usually the bots are fighting to bribe the uh validators to get the transaction included so that they can capture the liquidation and basically it's a bidding war. The result of that is protocol gets nothing from that. There's a liquidation bonus that

the protocol gives. Let's say 10% depending on the design, depending on the asset and so on. gas is wasted because those bots are essentially like fighting to get the slot. So they're wasting the gas as a consequence they spam a network. So if it's a network that is a word computer like for example you know Ethereum strives to be uh then probably it's not necessarily the transactions you would like to have on the network in general and the risk control is outsourced from the protocol to the let's say validators market.

Hence the current situation is not a system and D5 protocols are missing out on billions escalated over a longer period. So we have a better way to handle ME and the proposed solution is okay maybe first how oracles works. So we have two major systems right now in Oracles. One is the pull oracle the other one is the push oracle. The pull oracle means that whenever a user is interacting with a protocol that requires an oracle update, the user himself would update the price on chain. I tend to call it on demand.

Only when there is need a price update, you're going to update that. Uh it's a newer fresher system, less adopted. So the one that is used by a compound, morpho, liquidity and many other protocols is the push oracle where you take the data and you throw it on chain in intervals usually time interval and deviation threshold. Deviation threshold means the maximum deviation that the onchain price should deviate from the freshest price that is available on the market. So imagine for simplistic words, imagine Bitcoin USD uh right now trades at $1,000 and we have a deviation threshold of 1%. So now we update

onchain $1,000 is the price of Bitcoin. If it moves to 100,2 for example, the new update is not going to be triggered. Everyone who is using that price is accepting inherently that it can be wrong by half percent. only when the half percent is you know met from the previous update the new update is going to be delivered and this is the model that right now 90% maybe 85% of D5 uses right now and most importantly the big protocols are using a compound and as I mentioned the blue chips so we propose and majority of OEV uh solutions in the market work that once you have the price updated on chain then

you do the OE action or any action Our approach is that the auction should happen before because then you have no delay. So the auction is happening offchain with no blind guessing, no gas wars and no duplicate transactions clogging the network. So the transaction that is updating the price is already containing the result of the bidding in the auction for having the priority to liquidate a specific position.

And in general, OEV shouldn't be just a revenue tool. It it is a missing piece in DeFi in my opinion because it turns noise into order. It gives the protocol and the wider D5 ecosystem another tool for managing how the system is operating. So some principles that in my opinion are crucial when you're designing any OEV system. First, it has to be fast. Liquidations are very very sensitive part of the uh DeFi um you know workflow flow because if you delay delay a liquidation it can turn out that the position is underwater with uh more that the position is underwater in a

sense that the protocol is accuring bad debt because for example a price of an asset is plummeting so much that you didn't manage to liquidate it within the reasonable um threshold. old and then no one wants to close that position because it's improfitable to close that specific position. So the OEV system should be fast. It should be easy to adopt. So we have battle tested codes like in example a v3 compound v2 and so on so forth. So we do have a lot of systems that work at scale and we know hacks happen like usually I would say every quarter I hear about a bigger hack. So in order to omit them, you should probably try to adjust to what's working and incrementally

improve that. So the s this the obvious system that doesn't require any update from the protocol is a good one. It should be safe. So it shouldn't hinder the security of the protocol and the flow from the get-go. Permissionless and open. So liquidators should be able to join the uh bidding process and it should be an open um process to to do do so and it should be crosschain by default. the majority of OEV is happening on Ethereum nowadays. But as we see uh for example, Robin Hood announced their L2 using arbitrum stack on Monday. Uh there's bass, there's arbitrum, there are many other networks

EVM, nonem that are growing and this problem is present in all of them. So the solution that you create should be crosschain by default. It's not like optionality in my opinion. So, how the OV works is that um Redstone detects the price update that could trigger a liquidation offchain. So, imagine again this Bitcoin traded at $1,000 and right now it trades at 900 um 92 and the update deviation is 1%. So, as I was telling you, a regular push oracle would not update that on chain even though the position is already

underwater. Why? Because Oracle without OUV is not aware that this new price would cause the liquidation in as it is of today. Any other solution thinks in the way that only the onchain environment detects the liquidations, right? And here imagine there is a price of 992. So the one that could cause a liquidation but in regular system wouldn't do so would wait until it goes to 9990. Right? So Redstone detects that price offchain.

Usually it's on Binance, OKX, Bybit, any other centralized exchanges because that's the most of the volume. Like they are usually the quickest to have this kind of a price discovery for very liquid assets. Then we have a auction that runs offchain with fast lane which is a specialist in me. They've been very long in the polygon ecosystem. Now they're also in Monat. and they created a product called Atlas that allows for this atomic uh auction and the Atlas product gives you an ordered list of bits of the liquidators that want to take part in this process.

So the liquidators are bidding for getting a priority pass and access to liquidate this particular position and then onchain redstone relayer which is this technology that throws data on chain uh would deliver the price update together with this list of bits and the atlas execution uses try catch loop to proet proceed beats in order on chain and then the liquidation happens. And now the kicker. I believe this is the most important slide in the whole presentation.

The whole process takes maximum 300 milliseconds. And to give you um like comparison h in in one second we have 1,00 milliseconds, right? So it's 0.3 of a second. So for a man manrain I think it's already impossible to comprehend. I think we have like 300 uh frames per second or something like that that we can detect as a human eye. So it's extremely fast and for majority of the ecosystems this is faster than a block time. We created also redstone bolt that is the fastest oracle on the market for mega that updates price feeds on chain up to every 2.4 millisecond. over there we need to decrease the latency over here but for

majority of the networks it's faster than the block time and if there is no valid bits in the process of this 300 milliseconds the system gracefully falls back to regular price update so imagine there are no bits or no bits was like worthy of the liquidation after 0.3 seconds a regular update would happen onchain and in the example I gave you this 992 that's the moment the auction would happen and if there is a beater that wants to liquidate that and pays for the price update priority then the update on chain happens so a byproduct of this OU design is that

you have more accurate data on chain you found a new person who is willing to pay the gas for updating transaction on chain and making the onchain data even more accurate because the biggest problem of push oracles is who pays for the gas right because it's public I mean once it's on chain everyone can use that so here we found a new player like a liquidator that is incentivized to update data on chain the moment there is a profitable liquidation to happen if there is no one like imagine this liquidation is like very small like a dollar then probably no one would you know take part in this bid process So an update will not happen and then if Bitcoin keeps on dropping at 990 the

regular oracle update would happen and then regular process of the uh MEV on onchain uh would happen right so um the system I'm design describing over here has already gone through Venus protocol uh I think top four landing protocols in the D5 ecosystem it's the biggest one on BNB chain We've run it for 30 days to observe like the results and uh we recapture directly for the protocol $6,000 in total liquidation volume of 165,000 and 90 liquidations in total and zero downtime because the the

auctions were happening faster than the block time. So there was like literally zero downtime in the whole process. it was invisible almost for the uh lending protocol of the users that it's actually implemented which is very important in designing such a system and literally yesterday no what's this it's Wednesday sorry two days ago if you go to Venus forum they voted in favor of enabling redstone ov on every market of theirs on uni chain and if that gets like proper numbers we're going to roll it out on all of their markets on our chains and we are in talks with all the other serious lending markets on the market.

So key takeaways, the system worked out out of the box. No code changes for Venus. Venus did almost zero work. Like they had to look into the design naturally, but the workload was on ours. Liquidators didn't need to change their infrastructure. Redstone handled sequencing execution fallback mechanism seamlessly. And uh this was real value captured safely and predictably. And the money that was got over here can be redistributed as Venus wants. So it can go back for example as incentives for borrowers because usually this is what lending market would like to incentivize or it can go to their treasury or any other activation they would like. So we redistributed the flow the the money that was like leaking out

of the system and integration is that simple. The heavy lift is on us on Redstone. So the only thing you have to do is actually use Redstone price feeds on the lending market or a D5 protocol and give us a heads up that we should turn on the OV on a particular feed on a particular network or or your feeds if you if you prefer. So and then we handle all the leg work out there. So who wins first? The protocol wins because they have better liquidations as I told you. They don't have to wait for this deviation threshold. We sometimes tell them, hey, we created a push oracle with zero deviation threshold. It's simplistic because it doesn't mean that we always

update with 0%. But from the perspective of a lending or a D5 protocol, what is important is that it liquidates as fast as it's possible the moment it's present uh on chain. So they capture additional value with increasing the security of the protocol. The users are more happy because there is uh less slippage and fewer surprises. And usually when we talk with those landing protocols, they perceive it as a new stream of value. But they don't want to take it to treasury. Majority of them, they're like, "Hey, let's make it more competitive. So we want to stream all the value for example to increase uh the attractiveness of borrowing or lending or any other activation in the system

that is happening. And the liquidators are also happier because they don't have to go into the me bidding wars because over there at least on Ethereum it's many times like some deals and you know cooperating with the large players that are having the majority of the me infrastructure on a given network. Over here it's like open access no white list no collocation. It's like a just a bidding. So it's more of a process of understanding like what is the highest bid I'm willing to pay to still make money out of getting the priority for liquidating that specific position right and as mentioned it's already uh

available on B&B chain and now we are rolling it out as we speak our engineers are working on it uh on uni chain soon uh we already have signed deals with um hypervm so the uh hyperlquid chain base, katana and couple of more that I cannot reveal yet but many of them perceive it as a new system that increases the efficiency of the market. Any EVM compatible chain for us is seamless like a rollout would take between two two days to two weeks depending on how the EVM actually works.

Um and liquidator boards are open source so anyone can join. And another interesting uh positive externality is that Redstone right now works a lot with uh tokenization platforms RWAS we are the major oracle for securityize. So we provide black rock vidd Apollo acret van vibbil and many other tokenized uh products and for them the big problem is the liquidity on chain because these tokenized products don't have onchain liquidity and via making the liquidation system more efficient you can also experiment right now we're talking about that with uler about making the capital efficiency higher so that you need less capital for

liquidation because it's more predictable you don't have to fight on the MEV auction you know you have just better understanding how is going to be the outcome it's still somewhat pro probabilistic but it's getting more and more deterministic in that sense yeah so this is all uh on the OU part um if you have any questions we still have two minutes and I believe soon Vitalik is speaking on the main stage so don't want to keep anyone too long but thank you very much and yeah if there are any questions happy to take [Applause] Okay.

The the question is whether we take a cut. Yes. I mean at the end of the day it's an infrastructure we've built and we are running. So there is some maintenance cost to all those actions and whatsoever. Every single deal with the protocol is one by one assessed because the liquidations on Venus with couple of billions of dollars on BNB chain is going to be a different deal than on uni chain where they have right now I don't know like 100 million roughly. So it's a one by one conversation. uh the um current let's say equilibrium what we observe on the market is roughly 65% to the protocol and whatever and the rest going to pay out the infrastructure and part going to the oracle and then

naturally it can go um usually higher like 70% to the protocol and whatsoever. So it really depends on step by step because for from the protocol perspective the change is hey now I get nothing and out of the blue I can get 65% of this redistributed flow over here right so this is like their perspective um yeah okay I'm not sure I fully understood the architecture so you're you're an Oracle company right so you've got your own nodes because at some point you said there's no collocation But you're collocating presumably your own notes.

Um collocating to what? To the like Binance to get closer to the price updates. Um well, we do collocate for specific chains like for example Mega, we collocated our nodes in their sequencer so that we can update every block. But when you're talking about I don't know Ethereum or even BNB chain where you know it's let's call it 300 millisecond you don't necessarily have to collocate next to Binance to get that accuracy and speed but sometimes we do collocate but do you do you plan to let anyone run nodes redstone nodes whatever that would be yeah yeah right now uh Kao uh I think Kirro uh Alchemy and couple of other data providers are running redstone

nodes we have a running guard program. So if you type in internet redstone guard, the first batch was announced I believe in March and uh I think the uh on rolling for the next batch is right now open. So yeah. All righty. Thanks a lot guys and hope you enjoy the rest of UCCC. Cheers.