Unchained - The Chopping Block: Why the Once-Taboo MEV Is Now a Core Part of Ethereum - Ep. 449

Episode Date: January 28, 2023

Welcome to “The Chopping Block!” – where crypto insiders Haseeb Qureshi, Tom Schmidt, and Tarun Chitra chop it up about the latest news. This week, Flashbots co-founder Phil Daian joins the show to go deep on what maximal extractable value (MEV) means and how it’s shaping the Ethereum ecosystem. Show highlights: how Phil started working in MEV and why he believes it resembles 'Alice in Wonderland' Phil's definition of MEV and when he realized this was an actual problem how Flashbots was born the story of how Vitalik didn't care about sandwich attacks at the time Uniswap was released whether it's possible to minimize MEV via more competition the never-ending debate about whether MEV should be accepted why Tarun thinks the concept of fair ordering is against nature  how Bitcoin uses social norms to solve its weaknesses  how Google is being called out in court for creating auction to front run its customers how Flashbots worked prior to and after Ethereum's Merge why Flashbots complies with OFAC sanctions  what is the Suave project and how it aims to decentralize Flashbots what the future of MEV looks like Hosts Haseeb Qureshi, managing partner at Dragonfly Capital Tarun Chitra, managing partner at Robot Ventures Tom Schmidt, general partner at Dragonfly Capital Guest: Phil: Twitter Disclosures Links The Block: Flashbots seeks up to $50 million at a billion-dollar valuation CoinDesk: Opinion: Miners, Front-Running-as-a-Service Is Theft CNN: DOJ sues Google over its dominance in online advertising market Unchained: 51% of Ethereum Blocks Are OFAC Censored Given the Sanctions on Tornado Cash, Is Ethereum Censorship Resistant? - Ep. 390 The Cost of Decentralization in 0x and EtherDelta Previous coverage of Unchained on MEV: Why Is Ethereum Trying to Maximize Value From Users? Two Sides Debate Learn more about your ad choices. Visit megaphone.fm/adchoices

I think anyone who's been into MEV for a while kind of has this story about how they stumbled onto it and fell down the rabbit hole. I often compare it to kind of Alice in Wonderland because for some people, once they discover MEV, the entire world just stops making sense and you fall into this kind of like crazy, you know, alternate reality. Not a dividend. It's a tale of two Kwan. Now, your losses are on someone else's balance. Generally speaking, air drops are kind of pointless anyways. I'm in the trading firms who are very involved. I like that eight is the ultimate

Defi protocols are the antidote to this problem. Hello, everybody, and welcome to the chopping block. Every couple weeks, the four of us get together and give the industry insider's perspective on the crypto topics of the day. So intros. First up, we got Tom, the Defy Maven and Master of Memes. Next, we've got Tarun, the Gigabrein and Grand Puba at Conlit.

Joining us today's special guest, we've got Phil Dian, the founder and frontiersman at FlashBats. And then you've got myself, Asib, I'm the head-hymed man of Dragonfly. So we are early stage investors in crypto, but I want to caveat that nothing we say here is investment advice, legal advice, or even life advice, and see choppingblock. xyZ for more information and disclosures. So today we've got Phil Diane from FlashBots. He is one of the premier thinkers in MEV. But before we start the conversation, so today we wanted to do a deep dive into MEV, into FlashBots, into all the things that are going on that are changing the landscape of how we think about blockchains and block space and mempools and transactions.

and all that stuff. Before we get into that, though, I want to caveat a couple of things. The first thing is that both Dragonfly and Robot Ventures are investors into FlashBots. So we have a, you know, be mindful of that as we have this conversation.

Also, it's been reported publicly that FlashBots is in the process of fundraising right now. We're not going to be commenting on that. This conversation is we purely about understanding FlashBats, the history of MEV, and what role MEV plays into crypto and blockchins generally.

So with that said, Tom, I'll kind of let you take it away at structuring the conversation. So we'll start by talking about the history of what MEV is. And then we'll go into sort of a broader discussion about, you know, some of the interesting topics that it brings up. Yeah, I'm really excited to have Phil here, who I sort of think of as like kind of the granddaddy of MEV and sort of the formalization of MEV on Ethereum. I think one thing that's weird about MEV is it feels like this extremely niche deep rabbit

hole where there's a certain subset of crypto people who are all the way down the rabbit hole and love it. But if you are sort of starting from the first time, it's sort of confusing as to what is going on, why we're in this place, all the different actors are and what they're doing. And so I guess, you know, Phil, do you want to give us sort of a little bit of your impression or your background on sort of what was MEV like or how did it sort of get started? Like what do you think of as sort of the first example of MEP in crypto? Yeah, definitely. Hello, everyone. I'm Phil, co-founder of FlashBots and a PhD student at Cornell who works on smart contract security.

What was MEV like for me? It's a super complicated journey. I think anyone who's been into MEV for a while kind of has this story about how they stumbled onto it and fell down the rabbit hole. I often compare it to kind of Alice in Wonderland because for some people, once they discover MEV, the entire world just stops making sense. And you fall into this kind of like crazy, you know, alternate reality. So for me, kind of what that journey was like, kind of started with getting into Bitcoin. I got into Bitcoin on some digital activism work that I was doing. And I wanted to kind of think about how to further the cryptocurrency cause kind of around 2015.

I got into Bitcoin kind of a little before that. But around 2015, I decided, okay, I want to change my career from formal verification, which is what I was working at the time and kind of formalizing car software, formalizing planes, things like that, into helping the cause of cryptocurrency because it's a super important and revolutionary technology. I started thinking a lot about, okay, how do I formalize like what is security in these systems and what are we actually building? So I think if you want to do original security research, you kind of have to like start with a blank slate and ask yourself, what do people actually want out of these systems? What are they trying to do? How do we actually write down

mathematically what properties and what guarantees people expect out of the things they're using? So starting to do that for smart contracts and crypto currencies in general, kind of fell down the Ethereum rabbit hole, fell down the smart contracts rabbit hole, fell down the rabbit hole of people using these things to build financial products at the times. Exchanges, Stablecoins, Dow's, or kind of early applications. And I started thinking about, okay, how do we formalize, like, what guarantees do these exchanges, these Dow's, these other dreams that people have actually give them when you execute them in the code of the system itself? And what is the difference from kind of the way people want these things to work to the way they do work? And one thing that kind of jumped out immediately was the enormous amount of power that miners had in the system. And that kind of led me down this rabbit hole of, okay, well, how do we formalize security here? All security is economic. This is the field we call crypto economics.

We like to think about rational actors and incentives and writing down this model of how people should behave. But because the miners have all this power, the incentives for how they behave is actually changed by the applications people are using by the transactions people do by what people basically build on top. And it's something that I kind of noticed that not a lot of people were thinking about or talking about. A lot of these application developers kind of saw Eith as this like magic black box where like it works like it works. It's like this magic computer that just like executes things in order if you like pay for it. And we can just build amazing things on top. And I think that's a super exciting kind of hype way to think about it. But when the rubber meets the road, you get into

the kind of like thorny issues of economics and power and everything like that. So trying to formalize that is kind of the birth of MEP. There are also many other kind of fun adventures along the way. We ran one of the first MED bots trying to prove, okay, is this real? This was before even PGA's, which is an acronym I haven't explained, but I'm sure we'll talk about later. We also started, you know, kind of created this token, especially for use by MEP bots in the early days called gas token and kind of went from there just studying it, measuring it, formalizing it, building on top and seeing what questions were open. So that was a very long rant, but that was kind of like the abridged TLDR of falling down the MVP rabbit hole for me. I will say one more

thing, which is kind of funds, which is that like for a long time, I wasn't convinced it was real. It was like, okay, miners can do this. They can exploit it. But is it real? And people would always ask me at conferences, like, is this real? Can I write a bot, et cetera? And it was kind of this academic moment of like, you know, imposter syndrome, like dysphoria, whatever you want to call it. I like almost felt like a false prophet saying like, yes, this stuff is real, but like it was also kind of obvious at the same time. So all that to say that like, you know, it's definitely been a journey and pretty insane to even be on a podcast talking about it, to be honest. Yeah, I can definitely see that. I think, I guess, you know, one thing I sort of think about when it comes to this

space that's happened in maybe the past, you know, two years or so has been like the formal definition of MEV in front of the creation of the term MEV. I think to your point earlier, you know, you sort of, there's sort of an assumption with, you know, blockchains, which is mostly true for Bitcoin, right, which is like you price your transaction and then miners rationally should mine transactions based on, you know, how much people are willing to pay. And that's sort of the end of the story, and that's how people, I think everything works. And I think we saw, you know, in the early days, that wasn't necessarily, you know, always true, right? Like, there was sort of very famously status ICO, which was this one of the most hyped

ICOs, I think of 2016-2017. And very famously, you know, F2 pool, which is one of the largest, you know, Ethereum mining pools at the time, front ran a lot of people trying to participate in this ICO and actually doing it themselves, I believe. And of course, there's, you know, people discovering that, you know, there's A&M. front running, so people front running trades or sandwiching trades on bank war. I guess, you know, Phil, how do you sort of formally think about like an MEV? Like, how would you formally define MEV? And then you sort of touched on on PGA's a little bit earlier, like walk us through, you know,

what is a PGA? How did those function? Yeah. So I would love to also hear from Turun here, because I know Tarun has opinions here, but there are many formal definitions of MEV. Maybe I'll give like a simple English one that I like first and then explain why I like it. And then and we can kind of start building from there. Very simply, MEV is called minor extractable value. It's for maximal extractable value now that there are no minors. It's basically the maximum amount of money that the validator, minor, or any kind of privileged actor or set of privileged actors in a system is able to extract if they kind of act maximally in their own self-interest, maximize their own profit. One way to formally define this in ETH is you look at a validator, you look at a blog,

And you say, like, of all the things the validator can do in the world, how much can the validator increase their eph balance during this block? That's kind of the definition that's in the clockwork finance paper. It's maybe my current working definition, although there are many, many others. Why do I like this? Because I think it kind of indirectly captures any behavior that someone would be willing to kind of pay an eith bribe to this actor to, assuming this actor can like only receive information through the channel of that bribe. That's kind of the technical reason. But basically, it looks at all the things a validator or a minor can do, including reordering, censoring, and inserting their own transactions,

and asks how much can they profit from what they're able to do? Basically, the original MEV definition was kind of a nerd snipe, because I wanted more people to kind of think about application security this way, to question the impact on L1 and the boundary of abstraction between L1 and the DAPs. And also, math people just love formulas. And there are so many, like, brilliant math people working on things like convex optimization and AI and, you know, numerical analysis and formal verification. These are all kind of ways to approach math problems that once you have a formula for

what you're trying to look for, it becomes much easier to build work on that. And so part of, like, coming up with a formula was trying to get all these people interested without them needing to go learn everything about how blockchain works. So I remember in the very early days, like, you know, chatting with a lot of mining pools about the concepts of MEV and front running. Obviously, before they were very popularized. And what was very noticeable is that almost all of the mining pools, they all really, really burned into their brains what happened during the F2 pool, satisfied CEO front run. Right? Sort of going back to MEV prehistory.

Like, why weren't people, why weren't mining pools front running just as a matter of practice? back in 2017, 2018. And the answer, broadly speaking, was that it's kind of like, you know, why do Bitcoin mining pools not go above 50% hash rate? Why do they just magically stay right at that threshold or underneath?

And the answer is that people yell at them. Is that like, if you go high enough, people just start yelling at you like everywhere. Everyone you know, everybody who cares about Bitcoin, prices start to move around. You're like, oh, my God, something bad is happening. And F2 pool just got slapped so hard in the face for them inserting their own transactions

to front run. this ICO, that everybody just learned you never fuck with block ordering. You just don't do it. It sort of became like this gentleman's agreement that if we're all running big mining pools, we're all sort of stewards of block space and of Ethereum, and so we don't mess with it. And so I remember when the Flash Boys paper came out, a lot of people sort of saw you as like kind of a gadfly, like you're sort of bringing up this problem that's not really a problem. And it's like, no, no, no, it's happening. Like there are people doing this stuff in the men pool and it's like, yeah, maybe you can sort of pull it out.

Some weirdos are doing that. But at the time, it felt very peripheral, right? And I remember, you know, you and I, we spoke about, this is back 2018, about, you know, Bank War front-running, and Yvonne and I worked on an article that we wrote sort of demonstrating how to front-run Bank War. But all these, you know, DeFi was so tiny back then, you know, like these Dexas at that time were trading like one to two million dollars a day. And so in absolute terms, all this stuff was really.

small. When do you think that transition really took place when this was less of a, oh, hey, I guess this is happening in some corner of the world to, hey, this is like a really important phenomenon that everybody started paying attention to. Yeah. So a lot to, a lot to kind of unpack there. I think I agree on the status ICO. A few people have mentioned as kind of like a watershed moment. In fact, I was in a kind of lot of MEPV chat groups and like, you know, talking to people about MED at the time, and we were all very paranoid that, like, oh, man, once the miners, like, really figured this out, like, it's going to be this kind of immediate downward spiral to vertical integration of Ethereum validation, where because

you need kind of a lot of expertise to front-run these transactions, you need to stay on top of all these protocols, you need to trade, you need to ARP them, you may even need to take risk, right? And the best party's position to do this are kind of large trading shops. And so the natural thing to do is to have kind of very close first party relationships between validators and these large trading shops. And for me, that was like my worst fear for crypto at the time because I was like, imagine if you have an edge in validating Ethereum, but only if you're a JP Morgan or or jump crypto or something like that. That would be a pretty, pretty sad universe for the dream of like permissionless finance, anyone can come. There's like a level playing field. You can validate from your cell phone or from your raspberry pie or whatever.

So we were super worried about this. And when the status thing happened, it was like the watershed moment of like, holy crap, like, we're screwed. You know, we were probably some of the ones like being really loud that you're saying and like just to try to, you know, figure out what to do about this. So I think, yeah, Flash bots itself as a protocol in many ways is like kind of trying to counter this vertical integration by democratizing the MEP itself and saying to validators like, look, that's even more problem. profitable if you outsource this to a permissionless market than if you have like trust-based contractual relationships or if you take the risk yourself in these markets. I also think that's a lot of the reason that people weren't doing it. So famously when mining pools started to do MEV, there was one pool. I'm pretty sure it was ether mine, but I don't want to misspeak, so I'll caveat.

But started running their own bots doing this vertical integration thing and quickly kind of got rugged by other MEV searchers who found a bug in their body. and use that to steal a few hundred thousand dollars, which at the time was more than they would make off MEP in nine months probably or more. So there are a lot of pitfalls. There is a lot of expertise. I think there's a lot of benefit to the separation of roles. So the market eventually did converge on that. But there was like a holy crap, like our miners about to go full tradfai. And, you know, none of us cool people will have anything to say in this market anymore. I think we should also not forget the more exotic M-AV that I guess we maybe we see a little bit of, but like I think at that time it was shocking was FOMO 3D and like people

buying up entire blocks in 2018 where they like, too, can you describe FOMO 3D? It was like, FOMO 3D was sort of like a you could think of it as a Ponzi scheme game where basically, you know, you put money into a smart contract and your money gets paid to everyone who's put in money before you. But if you're the last person to put money in for a certain number of blocks, so a certain amount of time, and no one comes after you, you get most of the money in the pot. And so the idea is that like if you were at the back of the line, then someone goes behind you, then you like, you know, you add more money and you kind of have this thing until there's just no one to do it. And so one of the most more, I think this is still considered an exotic form with that maybe if we have,

if we have a taxonomy, it would definitely be in its own category of buying up entire blocks, block space just to be last. And they spent, you know, $58,000 to $100,000, which at that time was like an insane amount to spend on this. And I think that was also one that really stuck in people's heads as like, you know, this is actually pretty strong form, a strong form of manipulation and perhaps censorship that, you know, needs to be internalized correctly. But yeah, to, to, to, to, Phil's earlier point, though, I think there is no one definition of MEV. That's my definition of MEV that there is actually like, it's like, it's kind of like ZK, where people say ZK, and they can mean any of the following things. Fully homomorphic encryption, multi-party communication, actual zero

knowledge proofs that are zero knowledge, constructing snarks, constructing starks. Like, the interesting thing I've learned in crypto, and this never happens in other forms of VC or finance or tech, is that there ends up being this thing where invests. need a meme that has a very short acronym that covers many things so that they can like pour money into it without thinking incessantly because like everyone's telling them that's the thing and MEV happens to be now one of those things. ZK is one of those things. Well there was AI long before that. Big data didn't have an acronym but that that was definitely a thing. Yeah yeah. AI. Yeah, yeah. Exactly. Exactly. Like MEV became a meme to encapsulate so many different things now that it like

sort of lost its fundamental disness in some way. I don't think that's only investors. I think that's all humans. People just love their soundbites, basically. In crypto people make the soundbites so that people invest more into those categories. And then they try to do some affiliation marketing. Like, my thing is ZK2.

Or my thing counts as MEV so that you can like span the pie of investable things. Anyway, sorry, those are tiny tangent. pet peeve. Maybe, you know, rewinding a little bit because I feel like we, again, kind of, you know, skipped ahead, like, walk us through a little bit of the journey from publishing Flash Boys 2.0 to then, you know, the creation of Flashbots and then getting all of these miners on board with MED GETH and, you know, so the V1 of Flashbots. And, like, you walk us through that journey. And then I guess also for folks who are maybe new, like, what is Flashbots? Or what was sort of the V1 of flashbots. Yeah, tell us the Genesis story too, because it's a good one. Yeah, I will, I'll tell you a

bunch of stories here. So I'm going to be, I'm going to be ranting for a little bit, but I think it'll be fun. So the first thing we did was kind of this blog post called the cost of decentralization in zero X and ether delta, which is when I believe it was zero X was doing their ICO, I kind of, I guess, fudded, I don't know, found things in the protocol that I didn't like specifically the MEP, and like kind of front-running vulnerable abilities. And I kind of wanted to make a blog post about this. It was this work that I was doing with a bunch of people like Cornell where we would just like look at a new protocol every week

and kind of just mark it up and shoot the shit on like what's wrong with it. And try to kind of be a source of truth that was like less financially incentivized in the kind of crazy world of crypto that was 2016, 2017 time period. So we took a look at zero X and the minor front-running thing, kind of really stuck out and was also kind of unifying with some previous work that I'd done on bug bounties. So it turns out to run a secure bug bounty on E to me to solve front running. We had a project for that called a live submarine. It was part of this paper called the Hydra back in 2015, 2016. And then this kind of came up again in ZeroX. We were like, wow, here's minor

front running again, solved with kind of the same general techniques with the same general tradeoffs. Let's see if this is real. So then we ran a bot. This was before PGA's. GA's for people who don't know are priority gas options. So we ran a bot. We were like, can we exploit this vulnerability by just paying more than people who are doing trades on 0X and Ether Delta and going in front of them and then selling right after them, basically doing a sandwiching attacks. So we started doing this and we were like, wow, we're making around a million dollars a year on this. Let's publish a blog post because we're academics. We're not in the business of making money. Publish a blog post. A few hours later, you know,

hundreds of emails coming in. Can you share your bot? I'll buy you. your boss, please send me your code. How did you do this? A million dollars a year. That sounds amazing. And then within three days, there were like four other bots doing ether delta. So then what ended up happening is when you have these four bots that are competing and the miners not playing this game, who's going to win? Well, at the time, the miners were kind of ordering by transaction price per unit of computation, which is called gas, gas price, to try to get kind of more valuable transactions into the top of the block space. That's just how the optimization problem worked. So these bots started forming these auctions where they'd quickly pay more and more gas than

each other in this kind of real-time game. So like I bid $1, you come in, you bid two, because you saw that I bid one, then I bid three. This is the game we kind of broke down and studied and what later became Flash Boys 2.0, the paper, when we kind of deployed nodes all over the world to measure what these bots were doing, collected a ton of data on this market, formalized this into a game theoretic abstraction and studied basically what is the equilibrium in the game theory here. And we found a lot of interesting results that are in that paper. For example, in many cases, it's better to not even look at what other people in the market do and just bid yourself on like your own curve. Why? Because then their network latency to you doesn't matter. There's

like all sorts of counterintuitive conclusions like this. So that was one phase. And at this point, MEV kind of had become a cottage industry where there were like a lot of bots doing PGAs. Uniswap had come out. There was a lot of MED introduced there that kind of made this market real. And there were a lot of things in parallel going on, including gambling games like Turin was talking about. New Dex is coming online that created a lot more MEP. ICOs where people would kind of want to buy the whole block. NFTs started to come into the picture and people wanted to get the hot drops before others. All these things kind of compounded to just increase the incentive to manipulate the block ordering, basically, or to place bids for this block order. So actually,

Phil, let me pause you right there. You mentioned a bunch of sources of MEP that kind of arose in the last couple of years. By far the biggest was Uniswap. Uniswap really just an order of magnitude more MV than pretty much anything else in the space. And certainly more than Ether Delta, which was an order book, base exchange. Can you give the audience an intuition? Why does Uniswap just spit out so much MEV? Because Uniswap basically has the fundamental property of an A,

which is like you want it to be the case that one person can come make a trade at any time without needing a counterparty because the counterparty is this liquidity on the blockchain. Now that being said, that means the blockchain basically has to offer you a price at any time. And that price will never be necessarily fully accurate to the outside world, which is moving in its own kind of very fast-paced separate way. Right. So like this always kind of creates opportunities for people to come in and trade against it because you don't need a counterparty to like agree to your trade. You can just kind of do it at any time. And you're always able to quote some price that depends on the state of the world and the transactions other people are making. So there's like a huge space of arbitrage to do there

if you can control that order. Also, user trades on Uniswap need to specify an approximate price because their price depends on what order they're put in. And so therefore, like, they need to know whether they'll be mined or not. And this, each user trade changes the price that the contract offers because of how it works, which also then creates even more opportunities. So it's kind of this like perfect storm of like on chain primed trades that like are constantly creating these opportunities. So maybe two more tangents. I'll go on just to kind of indulge the myself because I think it's cool and I've never told these stories and I think the audience might find it fun. One of the things that was mentioned before was buying full blocks. Little known fact in creating gas token through

IC3. We actually bought many full blocks and we kind of use this clever exploit in Ethereum where you could store gas. You could buy it when it was cheap, store it, and then kind of redeem it later to get these blocks. And we ended up actually buying two weeks of basically Ethereum block space, if I remember correctly, it was around two weeks for around $60,000 to like basically clog up the whole network, which at the time we were like, wow, that price for the gas seems like super underpriced because there's no like kind of efficient way to buy it. So that was one fun kind of early MEV tangent. Another one relates to uniswap. So, when the Uniswap forum post came on ETH research, I actually sent a message to Vitalik and Hayden

basically saying, like, this is a horrible idea, this is going to create a ton of MEP. It's super exploitable. Users are going to get reordered and like sandwiched and like they're not going to have any guarantees, like not the right way to design an exchange on a blockchain for like all these fundamental reasons. And Vitalik wrote back pretty quickly and he was like, you know, actually I think this is fine because like maybe we can just use these incentives as like another fee payment on the blockchain. And I was so angry and triggered. I was like, what do you mean another fee payment? This is like so hard to optimize. You know, it's like NP like super piece based like star problem. Like what are you talking about like Goldman Sachs is going to come in and fill us all?

But funnily enough, I've like pretty much come around to his his way of thinking. So Vitalik, the profit confirmed. Always always the problem. But yeah, basically always the So getting into flashbots, what is flashbots? So flashbots is basically, it's a, it's a collective, a company, a startup formed to do research and development in this problem space, come up with solutions, come up with open problems, come up with products as well, and kind of solve MED basically, is like the long-term mandate of flashbots. And the first product flashbots kind of was a whole, hold on, Phil.

How did flashbots come together? How did all this start? How did all it start? Well, it started in a telegram group. So it started with me kind of meeting a bunch of people who were working on MEP and us just kind of talking about MED and like being concerned about it and starting just, you know, a network of basically cool people who were working on this. One of them was Tina, Tina the Flashbot's goat, who I met in a Uniswap party in New York, actually funnily enough. But she was running this like traveling hacker house that did basically M.EV extraction using various strategies and was also trying to like run minors and kind of ran based on this collective model where you would bring tribute.

They did do off-chain MEV extraction too by purposely like winning like five million hackathons. And using that to write like MEV related code. It was actually kind of genius. They they had Tina, Tina's house was like one of the craziest things I think of ever seen in this industry. I agree. So I think there should be a separate episode just on Tina's house. And I would love to come reminisce about, because it was called the Pirateship. And that's like to ruin said kind of what it was. It was like place to place extracting tribute. It does feel like one of these things, though, that needs to be like a legal music venue where like everyone has a great memory. But like you never really write about it. We might have to continue that tradition. Yeah, the few grainy photos we have. Exactly. Exactly. Yeah. Yeah. It's all that's left. Yeah.

So, yeah, basically, I knew all these people were kind of working on any of these stuff. When F2 pool stuff kind of front ran the status ICO, we all kind of were getting concerned. I started thinking about, okay, what can we do about this? And I had this kind of idea of like, oh, we should just democratize this and build like an explicit auction that like works around transaction ordering. Because clearly it's a valuable resource and clearly like, you know, I have this whole. post called MED What Do, where I lay out like the logic for why extract MEP instead of kind of trying to circumvent it or leaving it on the table. But kind of built up that philosophy and that led to this product called MEPGAP.

Well, Phil, can you explain the intuition of why, if you run an auction, why that's better than, like, what you described of what's happening in the MEP pool with these priority gas auctions? That is also an auction, right? Why is it better to have a, quote, quote, explicit auction than the auction that was taking place in the MEP pool already? Yeah, so because the demand and like the economic energy is there for the MAB to be extracted, like there will be an auction. You're right. Like someone will win the auction and they will extract the

MEP. This might take the form of spam. It might take the form of priority gas auctions, which we've already talked about where people quickly kind of increase their gas payments. It might take the form in some systems of latency or like, like it is on Wall Street with like a single order book server where people try to build kind of as quick links between various computer systems and as low latency systems as possible. These are all just different ways for this financial energy to kind of like express itself and be extracted. So the downsides of PGAs are, they have a lot of externalities. So first of all, having good latency is a huge advantage in a PGA. This is to me one of the biggest issues with PGAs. Why? Because you can see other people's bids faster. You can make faster

bids to the miner. You can see transactions as they happen on the network faster and update the bids you're making with that advanced knowledge. So we formalize a lot of these effects in Flash Boys 2.0, but basically the TLDR is like latency is huge in that game. And one of the good things about the explicit auction is it reduces the benefit of having all this kind of low latency infrastructure in a way that lets more people participate. And then you have like a more competitive, more democratic, less predatory market. I also think latency in crypto is like one of the underrated memes of like, you know, 20, 23, 2024. because if you have a system that you need to have low latency on,

you end up with geographic centralization, and you end up with regulatory chokepoints, and end up with fragility, just all these bad things. Like, we don't want something that requires low latency, I don't know. So that's one thing, but there are many other externalities. There's also, like, failed transactions still have to pay in priority gas auctions because of the way Ethereum's rules work. So you end up with a lot of failed transactions that take up space

and pay money that they don't. need to be paying this money because they're not doing anything and they don't need to be taking up this space because you know other transactions could go there they're not doing it that's another externality there's also a lot of messages sent on the network on the peer-to-peer network so all these people running nodes on their own homes or running nodes for their business or whatever when there's a priority gas option you're relaying like hundreds of bids that are going on in real time just for this option for maybe one legitimate transaction that the transaction is trying to arbitrage. So that's just super inefficient. And it's an externality on the entire peer-to-peer

network that just makes the system more centralized and makes it more expensive to run it up. And that's not a comprehensive list. So there's many, many externalities of having a bad mechanism of just having this yolo inefficient way to do things. So if I can summarize the philosophy that behind flashbots, it's that MEP is not going away. Like we're going to have MEP. And if we're going to have MEV, we know that people are going to compete for the MEP. That also goes without saying there's going to be competition. But that competition doesn't have to be wasteful. And if you can minimize the waste, whether it's messages sending over the wire or whether

it's failed transactions on chain or whatever else it, or competition over latency, where people are buying hardware that's closer and closer to wherever the center of the network is, if you can reduce that competition, then the competition that does take place can be more fair, more equitable, and there's more room for people to invest in things that are good for Ethereum. rather than that are zero-sum, just people kind of wasting each other's money? I think that's part of it. I would even take it two steps further. And so one step further, I would say, like, the last thing we want to do is create systems

where being a bad person or, like, you know, creating externalities is better for you individually and is more profitable than being a good actor and behaving within the rules of the system. I think this is a big way that TradFi has failed. In TradFi, if you seek to game the edges of the system, that's the most profitable thing you can possibly do, right? And like, it doesn't necessarily need to lead to a social good. Maybe it does. Maybe it doesn't. It doesn't matter.

But like, so I think one thing FlashBots really doesn't want to do is advantage people who are willing to extract MED. Right. In a world where we say, everyone, like, please, let's just be kumbaya, leave this on the table. We don't want it to be the case that if you do do it, you'll be way more profitable. Right. So like that's a big part of creating a level playing field, especially in a geopolitically complicated universe like crypto where we do have bad actors. Like North Korea does Ethereum transactions every day. You know what I mean?

Things like that. So we really want to make sure that like the honest behavior that we are saying to people, this is the happy path is also not like super unprofitable or doesn't penalize them for taking it. That's one step. And the next step is like we do think this energy can be harnessed also for the users. and to subsidize people who secure the network, yes, but also to subsidize transactions on the network, to get people better prices, to give people the execution they want in these systems.

So we think if we design everything correctly, MEP doesn't have to be a bad thing. It can actually be a really, really positive force, and it can be the difference for like why crypto works better than what was there before. Okay, so before we sort of go further in the story of how MEPV has developed, I think this is a good time as I need to actually try to take a crack at that question

because I think it's been it's been a big debate that it still rages in many corners of the crypto industry is should we accept MEV as just like a it's just like an element of nature. It's like the wind. Like you're never going to get rid of it. So you have to learn to live with it, to harness it and to make it harmonious with the rest of what, you know, what your goals are. There was a very famous disagreement between yourself and Ari Jules, who is your PhD. advisor, a very, you know, highly regarded cryptographer in his own right. And he wrote a piece in

Coin desk called Miners Front Running as a Service is theft. And he has a great line at the beginning of this piece where he says, as an analogy to MEV, he says, I'd like to announce a great new idea I've devised to reform the police. Today, cities direct their police forces to prevent and prosecute theft. But crime is a tough problem and policing is costly. What cities should do instead is auction off the right to mug people and burglarize their homes. And using this money that they get from the auction, they can use that to pay people's salaries, offset tax revenue and fund new policing initiatives. Use this as an analogy to say, hey, isn't it kind of fucked up this idea that instead of saying, wow, MEP is terrible, it is hurting what people's

expectations are about how blockchain should work and how people were taught that blockchain actually function. And what your guy's solution is, or what your conclusion is, is that, oh, we should embrace MEV. We should sort of take it all the way to the finish line and then sort of be redistributive or figure out how to minimize the externalities of MEV. But MEP is just going to have a fact of life, so we should just embrace it. Talk us through, like, how you think about that objection. today? Yeah.

I mean, I think it's an intellectual disagreement. I think there are different approaches to MEV. The way I think about it is honestly that it's just an argument over like how to get to the same end goal that we all want, which has nothing to do with like MBV or the definition. It's like building a system that works as well for users as possible that has all these financial apps that is perceived as fair, that people feel as fair, that's not like creating all these exploitative situations or rents or anything like that. Like I think all the camps in the MEP diaspora and thought, including the ones who think

MEP as a service is theft, they share this objective. They like really, really care about fairness. My true belief is like if you take fairness protocols as they call them or like fair ordering protocols or whatever and you iterate them to like improve everything that's like wrong with them, all the externalities that like people trying to extract MEP anyway will create, you make them compatible with crypto as incentive model and secure. I think you actually end up with what FlashB is doing, which is some form of MEV as a service. I think there's like, in my mind, maybe less of a distinction than in other people's. I don't know. I think people get

very emotional about this question and they attach a lot of emotion to like what I consider more technical concepts like the ordering of transactions or, you know, the MEV definition or something like that, sandwiching, like one transaction, then another, than another, like the semantic fact that one transaction comes before another. To me, like, these are all, like, second order things. Like, what matters is what outcome are the users actually getting? Is it good or not? I think ultimately my strong belief is, like, to get a good outcome, you need to, as you

said, acknowledge, you know, physics, nature, whatever, and use the leverage this energy for the user. but it's certainly a debate. Some people believe that if you make it hard enough to harness this and you just have a world where it's like never harnessed, this is a better outcome for the user because no one else can like trade in a privileged way against them because no one has the information. Do you agree with that? Maybe that's true. I mean, I think they get a worse price that way and the users would prefer for people to like give them a better price than to like have the intellectual purity of like no one's seen my information. But I think it's like basically a religious question and both approaches will be tried.

And so we'll see what people prefer. I'm curious to hear Terud's thoughts on this too. I know you've thought of. I'm a much more militant, anti-fair ordering person than Phil. Partially because I've like written a lot of papers on proving these bounds where you can't actually achieve these under different notions of fairness. So one of the biggest sort of like academic areas of the last maybe five years at not in crypto is AI fairness, which has gotten like a ton of, people have worked on research

in that forever. It's like, what does it mean for a model that's like making a decision in a court case or like identifying your face and telling the police to be fair? Like, when is it actually fair? And that field is filled with a ton of impossibility theorems of like there's many competing notions of fairness. And if you satisfy one, you won't satisfy the other. And there's a ton of those types of results. For some reason in crypto, people seem to be unwilling to believe that such similar things will happen. And I think, like, I find that offensive personally, which is why I write a lot of research on this, because it's sort of like ignoring some very basic mathematical facts about what ordering means, the set of ordering, the set of permutations, has a very

concrete definition, has a representation. There has, like, all sorts of things that should be restricting how well you can actually give a notion of fairness. And we know that in voting systems, but for some reason, people don't want to accept that for these systems. And so, yeah, so I basically think that the fair ordering thing is, in a lot of economics, the types of results you show theoretically in mechanism design, or how do you optimize social welfare? Well, one way of optimizing social welfare is charging the winner of the game.

game of the mechanism, the amount of externality that they cause on all the other users. And in some sense, an auction is designed exactly to do that. It's designed to make people compete to charge the externality that they're causing to the users otherwise. And in some ways, fair ordering is trying to say something akin to, there are no externalities. And we're not causing them to you because everything's hidden. And that's sort of like, you know, violating a lot of nature in the sense of like, it's not really possible to actually do that. You are causing externality. Sorry, can you can you can you can you elaborate like why is it against nature for there to be a concept of fair ordering? What do you mean by that? So the idea that you can submit

your transaction in isolation and not impact the overall performance of the network, the overall fees the network has to charge to guarantee you isolation and that those fees are less than the place where it's a competitive market or that the latency is as low as possible is sort of ignoring physical limitations to doing such a thing. This notion of a fee is really another way of measuring your externality of your particular transaction that you are causing on the set of validators

and the set of relay nodes and the set of everyone running infrastructure. And there is a cost of that. And somehow saying, like, I get to be a special snowflake, but I also get to pay less than if it was competitive. It doesn't make any sense. Here, let me, let me offer a counterargument. Okay, so let me first motivate that counterargument. So when I first heard about flashbots,

and I remember being back at the pirate ship events in SF, and, you know, this is like Flashbots was just still sort of a loose affiliation of people that was sort of germinating at that time. I remember getting into some debates at that time that I was very strongly against the concept of Flashbots. And in general, there were a bunch of projects around that time that were called accelerationist, where they sort of said, like, look, there's this thing that's, like, kind of messed up. I remember there's another one at the time that was doing sort of bribery as a service, like governance bribery attacks.

And it's like, well, this is going to happen anyway. So we should just. Yeah, yeah, yeah, yeah. Well, there's EGL. There was also automata was another one. And there were a few others. You know, you know what the promoters of EGL are doing right now being blocked building? Yeah, I'm well aware.

So they participate in the flashbot's auction. Yeah, yeah, look. Let's also observe the empirical evidence. 100%. 100%. Like once, obviously, like today, it's a very different story. And I think once we cross that threshold, like once we moved out of the equilibrium

of nobody's extracting MEP, and we moved into the equilibrium of everyone's extracting MEP, there's no question that now we need flashbots. We need to minimize the externalities from, you know, MV extraction. But the original thesis of flashbots was acceleration. which is that this is the end state,

and so let's move there as fast as possible in the most orderly and efficient way. And there were a bunch of projects that espouse the same general philosophy. And we had a lot of debates about this internally. I was probably the most staunchly anti-accelerationist. And the reason why I say that is because I think that the general philosophy-

Because you're an effective altruis. Not true, not true. A lot of effective altruists are accelerationsists in very, especially, Especially when it comes to AI, like they're very much accelerationists. Different ways. Yeah, yeah, yeah. I'm the opposite.

I actually think that there are a lot of things in crypto and in blockchains that are basically not enforced by software, but are enforced by essentially norms. And I think it is very, very tempting to believe that all the norms should be ripped out and replaced with sort of pure software-based equilibrium. And I think that's very dangerous. I think there's like a sort of Chesterton's fence thing that I very intuitively, that very intuitively appeals to me. That's like, this is such an AI safety argument. This is an AI safety argument, but for a blockchain user.

This is an effective altruist argument. It's very simply the idea of like, look, the same thing is true for, you know, for mining pools, right, with respect to Bitcoin, put a theorem aside. For Bitcoin mining pools, there is an economy of scale. The more, the more hash rate you have, the more profitable. your mining pool is going to be, the more smooth your returns are going to be. And so it really makes sense to give all of your hash rate to a single mining pool and for one mining pool to control all of Bitcoin. Now, that doesn't happen.

Why doesn't it happen? It doesn't happen because of norm enforcement. We've decided that, hey, the system itself has this weakness, which is that, I mean, there's a ton of weaknesses in Bitcoin. People have talked about them forever, like feather forking and all this other stuff. Like, there's a fundamental instability to the protocol. But we sort of use culture and our social. norms to kind of paper over the weaknesses in the protocol itself. And you can look at that and say,

oh, well, that's a mistake. Like, really, we should get rid of the norm enforcement and we should find a software-based way to solve that problem. And you could, but you might make the protocol on the whole fall into a worse equilibrium. That was my belief about MEV. Is that like, okay, you could do it, that you could certainly try to say, look, we're going to take a totally software-based solution. But at the end of the day, people have a certain assumption about blockchains work. Like the way they were taught, the way that I was taught how blockchains work back in 2016, 2017, right? This very simple notion of, oh, it's a computer in the sky, and you can, you know, use your time in the computer in the sky to do whatever you want,

and then the next person gets to go, right? That is the way that people think about how blockchains work. And when you tell them, no, no, no, no, that's totally wrong. It's very cute that you think that. But in reality, this is like some giant auction and people are going to steal your transactions and reorder them and mess with you and, you know, do all this stuff. That's how blockchains really work, welcome to reality. Like that does kind of suck. And you do probably lose something that's hard to quantify. It's hard to point at exactly what is the cost that you're paying

because there's some externalities you're alleviating, but there's some other costs you're paying that are very difficult to quantify. And when you're paying, just as a general thing, right, when you're paying a very, when you're receiving a quantifiable return, but you're paying an unquantifiable or sort of, you know, deep uncertainty about what the cost is you're paying, you should be very, very careful to make that trade-off. That was my argument for why I was not in favor of the FlashBots approach to MEV.

Now, look, we've crossed that chasm years ago. So at this point, there's no question that FlashBots is necessary. But that was my take about why I was anti- Flashbots in the beginning. In a world where you're competing with human U.X requirements versus of many automated systems also generating transactions and replying the transactions in the same system, you want to make something as resistant as possible so that if the algorithmic agents are causing huge externalities

on the humans, they're paying them for it. And that's like that in my mind is the goal long run of keeping in auction versus forcing the human to actually have to compete with the algorithmic entities on the same level. Because the auction is a fair way to charge the algorithm. I think maybe actually this is a good segue because I realize we're running a little bit low on time to, you know, kind of this other topic that I wanted to get to recently, sort of around norm enforcement and sort of social consensus around blockchains, which is sort of this transition to E2.2.0 to proposer builder separation. And then, of course, to MEP boost and sort of, you know, OFax sanction compliance and sort of relay censorship in general. So I guess Phil kind of going back to the story a little bit, walk us through, you know, the transition to each 2.0 from the

flashbot side. And then I kind of want to get into, you know, some of the politics of running these relays. Definitely. Let me just quickly address Haseeb's long question because I want to at least say something. I agree with you, actually, in some cases. I think it's complicated. I think there's like a very, it's very hard to have any one size fits all ethical rules for me. I'm kind of very like subjective and I use use my gut. So I wouldn't necessarily advocate for like, you know, kind of indiscriminate accelerationism like all the time because I think that just kind of redirects human energy into like a darker place than necessarily it needs to be. That being said, I think it is sometimes useful and sometimes warranted. I think the the key with when Flashbot

started is like the norm was already breaking down in many ways. And so like there weren't that many steps left and there wasn't that big of a window of time to change the norm. And we wanted to move it towards maybe a norm that was more reflective of reality. I also think saying like, okay, it's this magical computer thing, which is what you said, Peseed, which is also what I was ranting about earlier when like I started working on NB, which is what everyone taught and yes, like how these things were taught to everyone. I don't know. I think that's also dangerous because it kind of sweeps the economic and the power realities under the rug a little bit. So I think we, we need like a perspective that's like both technologically and socially aware, a lot of these

questions being like kind of above my pay grade. Sure. Well, so actually, before we get to the history, sorry, before we, before we get to ETH two, one question maybe to kind of point at the crux of the disagreement, if there were, okay, let's just say, there were a simple algorithm that would allow the blockchain to work the way that we sort of were described at the blockchain work. Obviously, all these fair ordering algorithms, they have very big data. downsides. There's slow latency. There's all this, you know, kind of you can fudge things at the margins. But let's say there were a magical algorithm that were just... There's security assumptions are weaker. Let's say that there were a magical algorithm that you

could just send your transactions to God and God would guarantee that they get ordered in the way that, you know, a single core would order them. Would you consider that preferable to a world with MEV extraction? And Turin, same question to you. No single core you buy will guarantee you deterministic execution in 2023. Like all of the Intel microcode instructions, you have no code that guy. Okay, it doesn't have to be deterministic. This has to be fair. Does that be deterministic?

Just has to be fair. I don't think that necessarily makes it fair depending on how it's implemented because then your incentive is to like be close to God. So this is how Tradfai works with like a single threaded order book that like locks memory and like modify state. Is that the way I would build a financial system? Probably not given like our knowledge today. Like I think actually harnessing NVV is a better way.

Worth noting also, a lot of roll-ups do work this way. Like, there's God, aka the Arbitrum Sequencer, and you talk to God, and God has a single-threaded processor, and God processes you on their nondeterministic Intel machine and tells you whether you got it or not. There's also a lot of externalities there. So I think it's complicated, but like, let's say you were to have a magic protocol that does meet all my needs. It's economically efficient. It gives the user the best price. It maximizes the kind of social welfare of the system.

and there's no incentive to play weird latency games or like edge kind of rents. There's no geographical centralization. And you can't maybe extract MEPB in the same way as the flashbots auction. I would definitely prefer that magical system. I just think there's like definitely impossibility results there. If you want to get into super esoteric ones, eventually you even get into like special relativity effects of like two people sending transactions at the same time from like other sides of the world. You know, so like there's not a clear, there's no clear objective God here, unfortunately. This is partially what I meant by physics.

Physics matters. Sure, sure. Nature does deem that we have these fundamental lower bounds. But, Drew, do you basically agree, like, look, if it, even if it reverts to being latency wars, you're like, that is not preferable to a flashbots type MVP auction? Yeah, I think auctions are interestingly one of the most. funny ways of parameterizing information aggregation because they have, because how you design the auction, how you choose the auction actually can give you very, very different

equilibrium in a way that like latency, you can't control as much. You just get the equilibrium that you get from optimizing for latency. With the auction part, at least there's a notion of like some intelligent design. Now, a really great example of auction design in the news recently was the DOJ yesterday. sued Google for antitrust. And they explicitly pointed out Google spamming the second price auction to increase the floor price. So basically in a second price auction, you have end people who bid.

The first place highest bidder pays what the second place person get. Now, such auctions have like nice properties for bidders. Like bidders are truthful. They don't need to come up with a strategy. But they're actually horrible for the auctioneer because the auctioneer can take advantage. of you because what they do is they insert fake bids and force the first place person to pay much higher than what the actual second price is. And the DOJ has this in the report. It's actually amazing. They actually went to the level of being like, you designed an auction to frontrun your

customers. And that's part of their antitrust complaint. So my point is auctions are actually a very nice mechanism for actually really being able to dictate the properties you want of your equilibrium and then design the auction around that, in a way that latency is not. That's just allowing full chaos. I will also say, like, I agree with all of that, and that's a great point. And I think there's literally like a thousand different angles you can like kind of cut down latency wars from. Another one is just very practical.

It's like in practice, if you have a latency system, all your nodes will be either in the US or Europe and like you're under like a single regulatory regime and like you iterate to trotify eventually. Like I think that's a very practical. practical reason you just can't have like latency sensitivity in crypto. And I do think like the auction does give you a lot of knobs to like figure out how you actually want this value to work for you. So agree with through.

Got it. Okay. So let's let's get back from that detour and walk us through. Okay. So in Ethereum 1.0, FlashBots sort of was up and running. It was this kind of, you know, sidecar auction where you were auctioning off block space and all these people are competing. Describe to us first how that worked.

Like who are the different players that are involved in this process and how it actually plugged into Ethereum. And then we'll move on to what happens with Ethereum 2.0. Yeah. So how it worked originally is we released basically an MVP of this ordering abstraction, letting people bid for ordering and inclusion and all these other validator powers, basically letting the validators run an auction to outsource these and letting people bid on them. How did this work?

So there were a few components. number one, we added this thing to the client that miners were using to make blocks on the blockchain called profit switching, which was basically this endpoint coming in where you could provide alternative blocks to the ones they were going to kind of build anyway with their own sets of rules and offer, you know, profit to the validator for this in a way that was like trustless. They could verify on the client that they'd basically get this profit. And the logic was like, if you let people do this, they'll kind of switch in between what is. the most profitable block to mine. So the other component was basically this middleman kind of

service. So the problem is if you open up this endpoint for the miners, right? Everyone's going to be sending them stuff and to actually verify if they're getting paid, they need to run this whole block and check it, which is really an expensive thing to do. So if they're getting like thousands of these per per minute or something like that, it'll essentially be a DOS on their system. So another piece you need it is basically a party in the middle, which in this case was the flashbots, what is now the flashbots relay, but we didn't call it a relay back then, which is kind of a centralized service that just filters out, only the most profitable blocks coming into these minor endpoints, basically. And at the time, we were kind of forwarding these whole blocks to the miners,

so the miners could actually verify, okay, I'm getting paid this much money. And at the time, there were mining pools. So each mining pool would kind of sign up to this endpoint, plug it into their client, and be able to start accepting these permissionless bids on our network. But it was somewhat centralized in that you were relying on this flashbots middleman for spam control and for validating that bids were profitable, basically. What ended up happening in the merge is we saw a change in the actual Ethereum landscape itself. So rather than having a few large pools that every miner kind of connected to one of a small set of pools,

the Ethereum Foundation wanted to enable people to run their own separate infrastructure and do solo validation, solo staking, let's say, at their house. So there were going to be many, many more people on the network, many validators signing up that were smaller. And these validators had kind of less of a reputation and less of an ability for the centralized FlashBoss service to kind of, you know, enforce that they were kind of following norms on this network that we built. And so that led to a system redesign, which basically decentralized the role that FlashBots

was playing, where now anyone can be a relay, any of the way. can play this role of spam control and checking profitable transactions for the validators in proof of stake. But the validators now have to trust any relay they select in this kind of decentralized network of relays because they no longer see that whole block. Why? Because if you send them the whole block, they can use that to misbehave in ways that would be very difficult for anyone to police or catch because there are many of these one-shot

validators that just like have one block and you never see them again. So because of this architecture change, we had to kind of both decentralized and centralize the market at the same time. Or validators have a little bit less control now because they don't see the block. But there's many, many different relays and choices of these parties where before it used to be kind of only flashbots playing this role. So talk us through then. Okay.

So going from the kind of decentralization of relays, one of the big flashpoints within the history of flashbots has been within the flashbots relay, the enforcement of OFAC restrictions. So for those of you who are not totally aware, so OFAC is the Office of Foreign Asset Control. Basically, it enforces sanctions. So when we discover that, you know, the Lazarus group from North Korea has docks itself, basically for some address on the blockchain, that generally gets sanctioned. And if you are a, you know, kind of legally compliant entity with some jurisdiction, if the U.S. has some jurisdiction over you, then you're legally required to enforce those sanctions,

meaning that you don't touch that address or that particular contract. There was a big color blue about FlashBOT's relay enforcing OFAC restrictions. Talk us through that and how you saw that from your perspective. Yeah. So basically, as you said, the OFAC list is this kind of list that's published by the Office of Foreign Asset Control, I believe it is, that kind of deems people that the U.S. considers really like high level enough that, like, they cannot do business. No one can do business with this person or they face kind of the highest level.

of national security kind of sanctions possible. So that's what it is. Flashbots currently does on its relay, its builder, any centralized infrastructure it runs. We do filter out transactions that are kind of on the list that's published by this entity. I think this kind of caused people to be afraid that maybe all of Ethereum or all of cryptocurrency would end up censored, especially because around the merge flashbots kind of very quickly started growing in dominance. immediately after the merge happened. So I think what people saw was this exponential curve of like flashbots building more and more blocks through the centralized infrastructure it was running, plus the centralized infrastructure

enforcing these restrictions and we're kind of very afraid that this censorship-resistant property people care a lot about in these systems, the global property of the system that kind of goes beyond flashbots, beyond any one actor, would kind of be violated. I think my view on it is that it's very very, very important. difficult to ever expect, especially in a proof of stake system where a lot of the actors are like large regulated parties, that you will ever be able to like really almost coerce actors into following norms. So we can have a debate about like, you know, what is kind of the legal tradeoff calculation around like OFAC versus no OFAC. But I think what's like hard for me to imagine is like, you know, crypto people and crypto developers imposing their preferences on like what other people. will do in the financial world outside of themselves. I think what we can and should do is build systems that are like globally robust, no matter how any one person has like their utility functions or their legal analyses, et cetera. Like if the system depends on flashboss legal analysis,

it's probably broken. So I think what we try to do is build competitive systems is optimized for things like geographic decentralization and privacy, which ultimately I think are the source of like the anti-fragility and the diversity in like crypto norms. So that's kind of how we see it. And I think people's concern is very valid. I'm not trying to dismiss that. I think their fear and concern is very valid. I think they also have a question about like ethical alignment,

whether FlashBots is like, you know, here to kind of impose like norms and its own preferences on all of crypto. I think we're all like crypto natives who the last thing we want to do is like ruin the censorship resistance of cryptocurrency. I think that being said, we're going to have to prove that to people and like decentralizing infrastructure so that it's meaningfully outside of ourselves is like really the way to go. It's a tough spot that you guys are in because you're very unique in being a venture-backed startup that is also in the business of being core infrastructure to a public blockchain, the second largest public blockchain in existence. And I sort of

it didn't occur to me before, but I realize now that you guys are actually, in many ways, very similar to Blockstream, in that there are a lot of really thorny, like, kind of philosophical and cultural issues that you guys are constantly brushing up against. Talk about a way to insult someone. It's, really, that was a... I think it's actually a surprisingly apt analogy. For how much you guys take shit for, like, you know, sort of micro missteps in terms of of norms, but also like how carefully you have to navigate this, this tightrope of we've raised a lot

of venture money. We have to somehow make money because we have investors and like, you know, you like that's just economic reality. But you're also trying to both create public goods, but also, you know, sort of do so in a, in a fair, transparent, open, legitimate way that obeys the norms of the blockchain that you guys are working on. That's really tricky. And you guys have taking a lot of shit for it. How do you, how do you pull that off? Like, do you think that it is doable for a venture back company to play that role? I think blockchain, like, shows it's really fraught. I don't think it's easy. That's for sure. And I think this is one of the, like, questions that comes up again and again, like you said, in many kind of different micro ways.

And I think it's a very valid question. The way I think I see it and like the leadership at FlashBoss, I won't speak for everyone. But the way I think that we all, see it is like there actually isn't a conflict there. I think we are all about like looking for nuance in places where people think there are obvious conflicts. For example, we think minimizing and maximizing MUB is the same thing. Ari Jules would disagree, right? But like there's a nuanced argument there. And I think it's the same as true for a venture back startup in this particular piece of critical infrastructure and public goods. And I think that's true for a few reasons. Number one, I think the financial incentives in MED are like huge.

And if you look at any kind of system that gets deployed, it does end up being affected by these financial incentives. So I think in kind of playing in this arena, it's very hard to imagine that like a small nonprofit effort or something that sustains itself on grants will A, be able to avoid these incentive complications that we deal with as flashbots. And B, be able to be effective. when it's coming up against like billions or trillions of dollars.

So I think to be effective, to build public goods, to build a better system, to build a better network, we need to be sustainable, we need to be scalable, and we need to be on the same playing field as like, you know, people who are, who want to be like mean and don't care about public goods and are just playing zero-sum games. I think there's a huge temptation when there's a zero-sum game to be played, which the MEV formula literally is, to just play the game. I think we believe, that like the bigger game is like designing the game. And to design the game, you need to be credible, you need to be reputable, you need to be legitimate, you need to make choices that people respect that aren't just for your own interest or for the zero sum game. So I think if you take

this time horizon, the two can actually be harmonized quite nicely, where you can have a sustainable, scalable thing. And it is, you know, a profitable company because you've grown the rest of the world so much that like it's hard to imagine it not being like I think crypto should eat all of tradfi I think every Google ad auction should run on a flashbot's option etc etc etc right and like things will just work better so if you actually make things work better right like you will be sustainable and you can still make things better for everyone else and I don't think there's actually a conflict there it's just a different time horizon but I do understand why people see the pressure and see the norms. I think what we've tried to do to counter that is a few things. Number one, we're

like super rigorous in how we design our incentives and our structures, our cap table, our equity, etc. to the point where like all our investors hate us and like, you know, not necessarily hate us, but you know, it's not like necessarily like, okay, let's just do the obvious thing. We think really deeply about like the power dynamics of what we're creating, does it work for what we're trying to solve, et cetera. And the other thing is choosing partners that are super aligned and that understand that building an Ethereum that works is more important than FlashBots becoming, you know, a high valuation enterprise or IPOing or anything like that. We really don't care about that and we don't want to succeed that way. We want to succeed by making ETH succeed and like having had a role in that that deserves to kind of be sustainable.

And I think our partners get that. They're like crypto-native people who build crypto companies. and we intentionally structure our conversations with everyone who may or may not ever, you know, have a business conversation with us to scare away people who are here for the zero-sum game. So I think me saying this also is not going to give people like the ultimate faith. I think people will still have the question and like I'm glad they do. I think ultimately actions over the next 10 years is like, well, it will matter. Don't build a saddle. Why not?

It blocks through. Bring it back to Blockstream. I will say, look, for you guys, it's a great example, as you mentioned, that, you know, very often like, okay, we could be like a small fledsling nonprofit and try to do this in this, like, kind of more traditional way. But the reality is that if you look at the changes that have actually been made to Ethereum, most of them have not been made by small nonprofits or like sort of, you know, little kind of research collectives. It's been groups like flashbots. It's been groups like optimism, which used to be the plasma research group, that have actually made the largest changes to how these protocols work. And so I think the incentives to actually change things and to actually have a real impact and actually ship things is so much stronger in the form of private companies than it has been just demonstrably from pure research organizations. And so I definitely feel that.

But you're totally right. Like there is a very important line that you have to tread carefully. But you guys have done a really incredible job, at least from my perspective, of doing that. Of all the things that are really close to the beating heart of Ethereum, you guys are just surprisingly beloved. I mean, obviously you guys got a lot of shit for the OFAC stuff, which we also gave you on the show. But I've been impressed that your guys' commitment to transparency and openness from the beginning about flashbots and MEV extraction has been very high. highly regarded by most of the Ethereum community. Yeah, I think when I think about flashbots,

it's sort of this constant sort of yo-yoing and tension between, like, decentralization and centralization, right? It's sort of like you start out with sort of this, you know, platonic ideal of what mining is supposed to be. And then it sort of centralizes into like, you know, MEDGath and the initial auction. And then it sort of decentralizes again with sort of the open sourcing of like the relay and boost.

But obviously it centralizes a little bit again in the fact that like, yeah, I mean, the majority of validators now, or almost every single validator on E2 runs MED boost, and the majority of those boost blocks are, you know, come from FlashBot's own instance of MEV relay. And I think, you know, there's inherent centralization effects or incentives in different markets. And I think you can look at some other blockchains

that also have their own sort of MEP markets as like almost an alternate reality of what could have been with Ethereum where you have one single entity that is running the clearinghouse and extracting profit. And so you can sort of see that, okay, flashbots didn't choose this route. They chose a different route. But like the market is the market to a certain extent. I guess, Phil, I mean, how do you think about like decentralizing sort of flashbots

in the long term or maybe this sort of process overall? That's a great question. That's like exactly what we're working on now. So we have this thing called the Swave Project, which we're going to be releasing like more and more information on, which basically aims to decentralize the existing flashbots in where we have and get, you know, everyone who is participating, including even our competitors, including L1 validators, including searchers and users, to have a share in like running this.

So it's not only flashbots. Of course, we took kind of a baby step in MEP boots, but this will be kind of like a full technical, architectural decentralization step. I also think it's, it is interesting. So like the non, I think the nonprofit for profit also plays into like this centralization decentralization thing. So I think one of the things that's interesting is like people give us a lot of shit for being a for-profit company. And then they say, like, look at the EF. It's like so non-profit,

so aligned. Partially true in charter, but like also like look at the EF in the broader blockchain world and in like the world of actually building the best financial system. It's a nonprofit, but it's incentivized by like this massive token treasury of like, which was launched actually by a different for-profit entity, which represents like essentially shares in the network they're building. Right. So like the incentives aren't as simple or clean as like for profit, nonprofit because thinking like the token price doesn't affect the EF incentives to me would be like a silly, silly argument, right? I don't know. I think maybe we need a more nuanced framework for this whole question of centralization and decentralization, how this maps onto entities, how this maps onto systems. Certainly the way I think about it is like we want to be more decentralized if we can at every step, unless we have like really convincing reasons not to,

that stand up to public scrutiny the way we did when we released MEP Geph and we said, like, look, this is like headed this way and we need to do this now. Let's talk to everyone about it. But in general, what we want to do and what the SWOP project is doing is decentralized at every step. And what we want to do is kind of build the gold standard of decentralized MEP auctions, build the best possible auction from a theory standpoint, from a design standpoint, from a product standpoint, user experience standpoint.

that's like our next goal. And I do think it needs to be decentralized to achieve those things. And I think this is part of like the wait and see is like, like you said, look at other chains that haven't implemented flashbots like mechanisms, many of which either now have flashbots clones or are like looking to do so. And I think as we like iterate another four or five steps in this market, you know, it'll kind of hopefully if we're right about our research and the state of the world and etc. will kind of just keep building what we consider to be the best possible market and

this will like kind of speak for itself. So one one question that I've had for quite a while and I think this is one probably big misprediction that I had about how MV markets are going to play out was that I sort of assumed that okay there's going to be the sort of flashbots or like the systems of relays and you know maybe it's one or maybe it's a bunch of them but these are you know you can't necessarily guarantee when you are entered into a flashbot's auction that your transaction is going to remain private. You can sort of, you can sort of force it through norms and say, okay, well, if you leak this stuff or you try to grab a transaction and go do something with it, we're going to slap you and you don't get to play in this game anymore. But if you really want privacy, let's say,

for example, you're about to hack a protocol. So you have a very nefarious reason why you want privacy. You're about to hack a protocol. There's no real way to protect it from somebody else seeing that hack and front-running you, or maybe it takes some time for you to perform that, whatever. There's some reason why you want to make sure nobody else sees this transaction before it gets included in a block. My assumption was going to be that there's going to be sort of these two competing forces.

One is going to be the open auctions with something like flashbots. And the second is going to be private sort of side channels with validators or minors where you can directly send a transaction, pay some, you know, fix fee or some higher fee, or maybe even a percentage of the, of the profits in some way that they can measure but there will be these two competing markets, right? One is for private transactions that nobody else gets to take a look at, and the other will be for these open auction mechanisms. And it seems like in the early days we had that with Tai Chi

and there were kind of a few other things. Sparkpool was, I think, one of the early actors there. But it seems like overwhelmingly that second thing hasn't happened. There aren't a lot of these very active side channels. There's really just the MEV auctions and the MEV auctions and the MEV auctions kind of enforce their own form of privacy, but they still engage in this auction format. Why do you think that is, and do you expect that to continue to be true going into the future? How do you think about the equilibrium there? Yeah, so, I mean, I think what we aim to do at

flashbots is build systems that are like optimal to use that you would use even if you also use the private channel. And I think in today's MEP boost, as long as there's one relay on the flashbot system that you trust that also connects to that validator, there's like very little. difference economically between that than using a private channel. In fact, like, if there's more than one validator that connects to any set of relays you trust or any subset, then you should always just like use flashbots and there's no point of like resending the same message to the validator that already is on flashbots. So those are like the kind of guarantees and equilibrium we aim to make. That being said, a lot of this is because Flashbots is not currently redistributing any MEV

and like or taking profit from the relay or the builders that it's running. So this is exactly the same for the validator. Maybe in a world where like the economics of the auction work differently, there is more of an incentive to like sidestep it. It's kind of important to us in general to design obviously auctions with as little incentive to sidestep it as possible because they're much harder to reason about, much harder to secure once there's like an incentive to go outside.

I think you can look at ad auction literature and theory and practice. I know it's Harun loves to tweet about this, but if you want to see all the ways people gain, like weirdly designed auctions, it's super fun. I mean, we're about to see the largest antitrust case of the last 20 years, which is the one against Google, built upon this. So it's going to be very exciting.

I mean, yeah, that's also why Google's kind of like, Google's in a very weird spot over this, because it's a very technical case. It's not like the Microsoft one of like, oh, you couldn't install Netscape. It was like it is about actually like auction front running and like financial manipulation, which is like a very, you know, first time I think we've seen a government make a technical claim against a tech company rather than the more emotional or political claim against them. Yeah, I think we may see a ton of like auction theory experts inside the government very soon,

which would be super funny to see. Cool. Okay. So I think you've bringing us, you brought us up to kind of modern day, what flashbots is working on now. You've kind of hinted at suave and, you know, the sort of, the, the attempt to decentralize flashbots. Describe, okay, let's say two years from now. Years 2025. What does MV look like in 2025? Oh, man, I'm so bad at making predictions. Well, but this is your roadmap. Like this, you guys should be the ones able to tell us this. Okay, yeah, sure. What do we want?

I think what do we want it to look like and what it will look like in different predictions. So what I would love MED to look like in two years, number one, I would love there to be like a much more decentralized infrastructure market than we have today. So I think we touched on this. People rely on the relays in MFboost today for two functions. Number one is privacy. Why do you need privacy? Because this is an auction. If you can't have private data, a private bid, you know, you basically degrade back to a PGA.

And we already talked about the problems with that. So the relay by keeping these bundles private from the validator and securing this market provides basically this negotiation, this new marketplace, this new option that's more efficient. So privacy is a key feature and right now is provided because you trust whatever set of relays you select. I would love to see that be much more decentralized. So many more options for how to achieve privacy and options that are competitive and that none of them kind of degrade the core properties of the system in the same way the centralized choices do today. So I think that's a big part of the swap push. Another thing you're relying on the relays for is spam control. So basically creating this more incentive compatible market where like you don't need a centralized party to pull spam, something I would love to see.

Another thing is redistribution. So I wouldn't necessarily even call it redistribution. I just think user benefit from MEB. Like in two years, I would like MEV discourse to be very different. I'd like it to be like a much less dirty word. My troll prediction is people will love being front run. and sandwich because they will see the benefits of this directly. And if not, then they're just leaving economic power of their own transaction on the table.

I don't think that's something you could ever imagine in its like heresy in Tratfy or in social contract-based finance. But I think in MEV, it actually works fairly well. I don't know. Robin Hood users, you could argue that they're happy they got front run because they're not paying any fees. For a very different reason, for a very different reason, for a very different reason. For a different, no, no, I agree.

I just mean like there is there is some analogies. Yeah, yeah. So what I always say, I think the idea of Robin Hood is great. The execution is a little more questionable to me of just like the auction theory, I guess. So that's kind of, you know, yeah, I agree. Like it does somewhat exist in Tradfi. Obviously, the idea is there. We can get into like a whole other episode about like, you know, do these kinds of incentives

like degrade the general markets in ways that like affect everyone. What is the rent who's getting it? is it like a fair auction, you know, etc. I think these are all super interesting questions. I will say Citadel's profitability leads me to believe like it's probably not very fair. But that's just my gut feeling. Again, I'm like an intuitionist on all this stuff.

But fair enough, fair enough. I think it is a great idea. But the different, no, but to be clear, the difference between what Robin Hood does, which is called payment for order flow, where basically, you know, retail traders who are just doing random nonsense on Robin Hood, they auction off their flow to a bunch of market makers and those market makers give them better prices than they would get, you know, just fulfilling an order on an exchange is generally speaking the opposite of what happens through MEV. MEV sometimes results in price improvement through like just-in-time liquidity.

But most of the time, MV usually means that you have price degradation, right? You're getting a worse price than what it looks like on screen. I actually think that's not an empirical fact, empirically true fact. Like, the routing quality is actually a lot better for multi-hop transactions too than when you have single-hop transactions due to MV because your searchers actually like push you along certain routes. I actually think like your claim is like maybe if I only like look at particular types of Dex Arm plus Sandwiching is true. But if I actually look at some of the aggregation and larger scale interactions like some of the ones from Macha, it's not, the MV is actually can be

quite positive for the end user. And then in weird ways. That's the beauty. It's quite complex. It's definitely not to like, there's no binary question. Like, there's, there's billions of, roughly a billion, right, in total MEPV that you guys are monitoring on your dashboards. I don't think that billion dollars in MEPV in aggregate is coming from price improvement for macho trades. I think it is. A lot of it is from sex, dex arbitrage or dex arbitrage, and that is literally provision.

And sex tax arbitrage is like, oh, it's purely price improves. Oh, true. Okay, fine. So I, assuming, again, what? Well, maybe, again, there's another argument, which is similar to the PIPF argument, that you're disincentivizing passive LPs more and therefore get worse prices, like, model, et cetera, which I think raises a good research question, which maybe is a good thing to highlight here.

It's like, we need ways to measure this and actually study this. Because I agree with Tauru, and my intuition is today, MEV actually has crossed the horizon of, like, now it's benefiting more people than it's harming, not to say that some people don't get worse prices because of MED. But I think that's mostly because they're, like, actually, you know, just not efficiently expressing their transaction. And I would hope in two years, like, the unsophisticated user can have enough sophistication in how they express their MEP transaction that, like, they have faith that is price improvement,

even if it is a front run. And I think today people have just kind of been, you know, this is like maybe a more controversial take, but people have just kind of been like gaslit by technological determinists into thinking, like, if someone went in front of me, I've been cheated and wronged, when really that's like not necessarily the case. Yeah, I agree with, I think this is kind of my main, my main gripping with like Ari's kind of quote or article from way back in the day, which is like, the space is not static. It's dynamic. And protocols and users and the tools that we use are evolving over time to, you know,

mitigate any VV or redirect it in a beneficial way. It's not like we have Uniswap V2 forever. Dexes are improving and the way people use dexas are improving. And the way their trades and their orders get routed is also improving. And so it's like, you know, it's not like we have this is one system kind of forever. Yeah, I agree with that. And obviously there's a lot of MEV that is just totally innocuous to the end user, like with liquidations and with backgrounds. Front runs as a category, I think, are the ones that are certainly most contentious.

And it's the one that people intuitively, like, okay, so I don't know any of the data about how much price improvement there is on front runs. And certainly just in time liquidity is the prime example of how front runs can actually be better for users than what most people are imagining. But they're still like the perception. And perception and intuition is actually really important. And it's very, very difficult to overcome the sense that MEV just kind of sounds evil, especially when it comes to front running.

Like the word front running just obviously has this stank on it that's very difficult to shake off no matter how academically we sort of coat our observations about it. what are your thoughts there and do you think like is there a rebrand in store where we're going to start calling these different names in order to make it more palatable? Or like how do we how do we make people feel better about this if the economics actually starts to turn in their favor? Yeah, I think that's a great question. I think it covers actually a lot of like what flashbots tried to do in the beginning,

which was like start these conversations. So I think two really amazing things like to reflect on here. One of them is to Tom's point about like the equilibrium not being static. Right now you look at any single protocol in ETH or all of crypto and they're basically all doing MEPAware protocol design. They're all thinking about like, okay, what are bots on our protocol going to do? How can we model these things? How can we try to predict how these things will play out?

How can we design smart equilibriums? How do we monitor these things in ways that work for our users, et cetera? A year and a half ago, two years ago, that was not the case at all. It was just like magic computer. We want to build a lending market. You know, hire someone to write the solidity for what we've explained to them that a lending market is. Get a rubber stamp on it.

Put it on the blockchain. One billion TVL launch a token. That was the roadmap. Very simple. So that's not the case anymore. And I think that's awesome. Like the amount of protocols that actually are really thinking hard and like having conversations with us.

And we encourage more and more of them, of course, to come to our forum post there. And like the more we can help with this, the better. it. So I think that is kind of amazing. And I think the second thing is like choosing the word front running and the gut reactions people have to it. I think we also thought there was like a dirtiness there and like, you know, this this seems wrong. You know what I mean? And so like I called it front running in Flash Boys 2.0 very intentionally because I wanted to trigger lawmakers and I wanted to trigger all these tradfai people to see it through a certain box and then put the onus on us as a community to like really convince them that like, no, this is the way it should be. Because I wasn't sure, you know what I mean?

Like, I don't know. And I think the question is still actually open, although now I'm convinced that like we will succeed in the rebrand. So I think actually we shouldn't rebrand it. We should still call it that and we should like basically force people into a more nuanced conversation about why the power norms in Tradfai are different, why that requires a different type of regulation and and why that doesn't exist in crypto. So yeah, I think that's what we need to do is really think hard about power dynamics, economics, social norms. And the fact that, like, people are having this gut reaction is cool, I think. Of course, it is actually somewhat, like you said, Haseeb, it's complicated because you don't want to go too agro on this.

And then, like, you just get some, like, super ill-informed legislation that, like, causes, like, ridiculous externalities because people's, like, feelings are hurt or something like this. and that is the risk you run by like taking this strategy. It's one that I at least kind of like chose intentionally and I still think, you know, there's a lot of conversation to have here. There's been a few people actually in the regulatory side who are like, oh, I'm writing a regulatory paper. Let me. Here's my knowledge of like, oh, this is front running.

This is horrible. And then we have the more nuanced conversation with them and they actually come around to the viewpoint. So I think it's an opportunity. Yeah, there was this law. postdoccer professor who wrote like a pretty long form legal that I guess posted this week. And it was actually quite, quite good because I think a very nuanced take, read a lot of research from different people, kind of like synthesize some sort of like notion of, hey, like there probably

is a sense in which like the current law legal structure doesn't understand how to handle it's always bad. But yeah, I think the problem is like people like really in 20. 2020 and 2021, maybe made it sound like this evil ogre that's like beat your metamask on its head. And like I just like, I've always like found that to be kind of a little bit sad because it makes people like not see that there's actually a lot of beauty in the structure of this stuff. And it's actually quite nuanced and detailed. And it's not a black and white like type of thing in the same way that like AI safety is not like some type of like bullion function that I'm going to like be able to tell you. yes or no to. And then, you know, I think MEV as a subject deserves that respect and nuance as opposed to like beating something on the head with a blunt force instrument. This is also why I call the blunt

force thing like a technological determinist kind of thing as a joke because you're making like a social value judgment on like this technical expression. And what that really means is you want like technology to judge your social outcomes. Like I'm sure the AI safety people would love to have like a function is safe, parentheses, AI model. Or like, same with the MEV, like, is ethical, you know, like transaction list. But, like, you're trying to project, like, these social norms, which are super important. And, like, we need to have nuanced conversations about, like, do our systems achieve them. On to, like, this very specific, like, semantic execution and ordering that's, like, more general purpose than what you're trying to reason about in a way where you just, like, lose a lot of, like, fidelity in your, like, ability to actually build things.

So I think that's, like, a natural impulse for people. I do understand it. It's also worth noting one of our first challenges at FlashBats on the research side was can we come up with an actual taxonomy for like the ethics of MED? And I can't tell you the number of people. All the other research problems we've had like amazing constructive progress down, I can't tell you the number of people that are like, yeah, we want to work on this, give us money.

And then two months later, oh, wait, this is like really hard, you know, to actually come up with this Boolean function. So, yeah. Well, it's a super important conversation. I think we've gone a little bit over time, so we have to wrap it up. But Phil, look, M.V is something that's going to continue to be an important conversation within Ethereum and within all blockchains. We really appreciate you coming on and explaining the history as well as what you guys are doing with FlashBots. When are you going to finish your PhD?

Because I know Ari is hurting your PhD advisor. Aren't you like seven years into your PhD now? something like that. I actually stopped coming. My website hasn't been updated since I was a second year student. So I still have people Twitter message me and they're like, wow, you've done so much for your second year. And I'm like, if only you know. But soon TM is the only answer I'll give. I mean, why not just never graduate? That seems to be the approach that he's taking. That seems to be the approach. Well, Phil, thanks so much for coming on and appreciate you talking us through everything. For now, signing off. See everybody. Thank you all. See ya.

Yeah.