The Essential Guide to General and App-Specific Order Flow Auctions | Season 6 Episode 4

Transcript

all right everyone welcome back to another episode of bell curve before we jump in quick disclaimer the views expressed by my co-host today are their personal views and they do not represent the views of any organization with which the co-hosts are associated with uh nothing in the episode is construed or relied upon as Financial technical tax legal or other advice you know the deal now let's jump into the episode hey everyone this episode is brought to you by Maverick protocol awe of liquidity tools built around an Innovative amm Maverick helps token projects Dow treasuries LPS or basically anyone in defi shape their liquidity with efficiency and flexibility how you might ask stick around and you're going

to be hearing about them more later now on with the show welcome back to another episode of bell curve today we are going to be talking about orderflow auctions and I am super pumped for this episode it's a very cool sort of on the vend diagram of me and dexes that we've been talking about in the season order flow auctions are squarely right in the middle of that so in this episode we start from a high level just to find what orderflow auctions are then we go through some of the more nitty-gritty designs decisions such as the order types that they're designed for so more generic sort of broad-based orderflow auctions or more appp specific orderflow auctions that might optimize for something like swaps permissioning so who is allowed to participate in the auction the information that's shared is it full info partial info when does that

information get shared and some of the more nitty-gritty design decisions we talk about which components of the OFA should be onchain versus offchain and the centralizing impacts of that we talk about what it takes to run a credibly neutral auction and then finally we move into intents and the Improvement that intents have over transactions and the Implement and the uh the implications for how these ofas are designed so this is a really great episode and without further Ado we'll just jump right in all right everyone welcome back to another episode of bell curve today I'm joined by my co-host Dan um and today we're joined by Robert and shin of Unis swap X and uh flash bot SWA this is going to be a really fun one um and my hope is that we're going to get deep into the weeds a little bit and we're going to be talking

about orderflow auctions or ofas and um I thought we could actually just start sort of with a a 10,000 foot definition of of how you view like what an orderflow auction is um eventually we're going to delve deep into the design space but and maybe we could then sort of segue into where Unis swap X and swab fit uh within that defin definition so maybe Shen I can I can pick on you to to kick us off here order flow auctions to me uh as the name suggests just means when there is order flow and Order flow you know it's just a I guess an array of orders uh which right now typically mean swaps but it can also mean other things transactions

and uh there are participants in the market who assign value or uh positive value to these set of transactions or swaps and they have somehow willingness to pay for the opportunity to interact with these orders and then the orderflow option is just the um the auction for them to kind of bid on the opportunity to interact uh that's that's my understanding of it that sounds right and then ideally bidding um with the proceeds of that auction typically going to the to the user who created the the transaction or intent um or in some cases to the interface that uh that originated it I think um generally I

think we think of is not it's not about something like the Mev boost auction um where the proceeds of the of the Mev are going to ultimately the proposer of the block onchain although I don't know if others would disagree with that um limitation of the definition but that that's typically the kind of auction that that determines what kind of meth happens and I Ico phase is kind of a Escape valve for those yeah I I think that's a The Helpful way to slice it and um there's there's a great uh piece on Frontier research which sort of has a a 2 by two of sort of like um maybe swaps versus more General transactions and then uh execution um uh sort of like block space auctions but Robert maybe I could pick

on you here because Suave um I'd love to get your sense of you know you just you I we were talking a little bit beforehand you sort of described Suave as sort of a a platform for ofas so I think we all sort of get the concept of like an orderflow auction why orderflow is is valuable in the process of auctioning that off and returning that back to users but would love to get a sense because there's already this like plethora of designs for orderflow auctions and they all sort of serve different purposes so we love to just get your sort of sense of what the landscape looks like and where swab fits and then we can go down into some of the examples of some of the ofas that exist today yeah so I thought that uh what should and then laid out is a really

helpful starting point for a discussion um orderflow auctions originally talked about this type of auction which uh auction off the right to back run a user's transaction and where the proceeds to that auction would be pay back to the user and so starting with specifically you know ethereum transactions um it's most easily conceptualized in the Contex context of a trade where the trade generates some M because it wasn't routed properly and that M can be paid back to the user instead of just going to the the minor the validator um instead and in um since the time that that idea of an orderflow auction was generated we

have started to conceptualize other types of orderflow options for more General forms of execution so I think Dan you're referring to um intents at some point where a user isn't making an ethereum transaction specifically but they're saying hey here is uh a signature saying I want to trade from usdc to eth and um please fill this in the best way possible uh there's many ways to do this usually they involve some form of competition among many different parties to give the users the best best uh execution possible which is where the auction part comes through and Flash Bots has created one concrete orderflow auction which works specifically with transactions called Mev share we can get into the details of

that but in the last year we've seen a huge explosion of different types of rlow auctions ones that are more similar to me share so they're working with individual um you know ethereum transactions but also uh also orderflow auctions which work with intense like unisoft X and I'm I'm sure we'll get into the details of that or other you know RFQ based systems other types of intent execution systems and um one of the things that we worry about here at flash Bots is that while we have really helped mitigate Me by introducing these orderflow auctions that pay me back to users and that's a good thing we worry that we are centralizing the

market in new places um largely like in these offchain pieces of infrastructure that run these auctions and our answer to this new form of centralization in the market is suave which is a platform where uh developers can create the orderflow auctions and other types of MV infrastructure as smart contracts on a decentralized blockchain um with built-in credibility and and privacy Happ to get into the details of of how that works but at a high level that's that's kind of how we think about order FL autions at at flash blots and different types and um the motivation behind Suave as a platform for ofas and the problem that we're trying to address

in the market which is moving this you know infrastructure which used to be a runon chain offchain um and introducing a new Vector of centralization yeah I think that's I think that's super helpful and that's that's a really good I'm getting a little uh flashbacks to the m season we did with hosu we used to talk about sort of centralization wack-a-mole which is sort of the game that you end up playing in in me land but I I would love to get a sense maybe because maybe we can start with sort of the most straightforward definition of an orderflow auction where what we're talking about is just swaps right something like that um I know swab has a much more broad Vision Unis swap X is pretty laser focused on swaps would love to get a sense um for you guys maybe just starting there of you know we

talk about orderflow auctions as the ability to aggregate you know preferences you know very non-specific sort of arbit preferences I mean how much of that do you see ultimately like what's the sort of market share uh in between just swaps versus these sort of broad more arbitrary preferences um maybe just to start out and and how do you see that kind of shaking out because I'd love to start digging into this Dynamic of like more general order flow auctions versus app specific ones I think it today the majority of me does come from swaps um a large majority of it is either arbs or or sandwiches there's a small amount which is liquidations and a small amount which is just weird stuff that's happening on chain which is kind of hard to to quantify um a bunch of N Stuff sometimes gaming sometimes just odd things on

chain um and so to that end like a OFA that addresses swaps is going to address the majority of the market space I will say that it is not um immediately clear to me that this is going to stay the the same into the future I would expect swaps to be a very very large and dominant and important important category but I think meev finds way and is often times popping up in unexpected places as an example there's a lot of me on friend. Tech um which like kind of looks like swaps but it it it uh is maybe can be conceptualized a little little bit differently too um and so who knows where it'll come in the future maybe a gaming use case maybe maybe something else but today it is

definitely that the majority of MV is coming from swaps that is for sure was there another part of your question Michael or did that address it yeah no that that makes sense I I would just love to maybe Shin as um sort of a more app specific approach which focuses exclusively on I just love to get a sense of like the the pros and cons for folks out there who are listening maybe to be really overly simplistic the way that I've sort of thought about it is if it's app specific you can really tailor the user experience as opposed to if you make it more General then you your total addressable Market is much much larger right is sort of broadly how I would classify the trade-off but would love to get a sense from the two of you of like you know just as Developers the two

different styles like how you both think about it I I might I might flip that around because in fact again I think like more than half of naevi would say is is decks related um uh so in some way in some ways actually I think the it's not necessarily the address Market is a lot larger but actually you do have a lot more power once you go to more General General yeah but I I can give uh thoughts after I curious for Shin's thoughts too uh yeah I mean I think know by definition if you take a more inclusive approach uh uh you yeah by definition the the addressable Market is definitely larger I think um the main reason that we have taken a very focused approach is just we

we believe that's what we specialize in that's what we are good at and um like unof for history has always almost exclusively done swaps only like we never try to Branch out into lending or other like defi stuff no ST no stable coin no anything else just swaps and so uh since right now uh there's so many me and other execution related problems with the way swaps happen on chain I feel like unop X is just a nice way to sort of improve the user experience here um and right now yes the the maybe that's like more than

half of the me problem but uh you never know uh as Robert said like maybe in the future there's there's a ton more activities and and I would love to see that right I think that's a that's also a good scenario for me if in the future uh crypto industry is is way more than people swapping tokens there's uh there's like a lot more activity happening um but I think yeah right now um we're still taking us a somewhat special more specialized approach uh smaller addressable Market I think the the upside is um uh you can just focus on doing one thing really really well um so like there's only a few parameters that we have to think about even though

that's still a hard problem but at least we don't have to consider you know like hundred different Corner cases when it comes to nft minting or lending liquidations or like other kind of activities so that's the upside I would say I think one one other interesting thing here and would be curious how you all think about this this Shen and Dan over on the unisoft set of things is that you have a revealed preference for people to I think participate in some types of ofas which are swap related but not not in the way that you might immediately think and so there like a type of Market participant which wants to back run uh like someone who is adding liquidity for a new token as an example it's like willing to pay for the

rights to do that and that is swap related but it's not the Offa That you design for something something like that is not um a competition just to fill the user at the best price because it's conditional on something else happening on chain or copy trading also so I think in some ways it like this question depends on where you draw the boundary of of swap also like does that address the same type of swap as un swap X probably not right so I was gonna say you know I've been generally a partisan 4 let's try to solve this the simple version the simpler version of the problem where it's just um ordinary retail swappers just swapping with each other and with and with uh market makers um so sort of batch action or something more like like Unis swap X um just

because it's simpler and it's like an example of a of a problem that is very large but I I absolutely agree with Robert that I think one of the reasons you know not not only are there there's a lot of like really longtail stuff then there's a lot of kind of middlet tail uh that is deeply entangled with swapping um you know swapping on on onchain exchanges for something you know um doing that atomically with user orders um is you know where exactly does that fit in terms of like is that is that just pure swapping um and then yeah stuff like liquidation auctions where things you know often doing a lot of other complicated stuff on chain does involve a swap and so yeah so I think it it ends we end up deep in Mev territory you know I know this is this is the deck season and you've

already had a Mev season um that Robert's been on but um so you know I think we can we can focus on the deck parts here but I do think it really bleeds into the rest of Mev for sure it absolutely does and especially the point like a huge part of that season as well was just the idea of returning Mev to users which is Robert as you stated sort of the explicit uh purpose of um of order flow auctions from your perspective hey everyone wanted to take a quick second to shout out this season's partner Maverick protocol now many of you probably know Maverick as an Innovative amm which they are but in reality they're a lot more than that as well Maverick is a suite of tools for defi users and builders that allows them to put liquidity where it will get the most work done since

Maverick launched in March they have been gobbling up market share and at the time of this recording which is the end of September on a trailing 7-Day volume basis Maverick is now a top three decks by volume and they support over 50% of the volume on the L2 ZK sync a chain Maverick enables LPS and token pairs to process higher volume with limited tvl which allows them to support some of the highest levels of capital efficiency for lsts like WAP Ste another very cool feature is something called Maverick boosted positions so that allows protocols looking to bootstrap their token liquidity to Target the shape of liquidity of any token pair with surgical precision Maverick is backed by some of the leading institutions in crypto Founders fund panta coinbase

Ventures Finance labs they are all backing Maverick in their Vision to revolutionize the next generation of dii daps and helping them build their liquidity in all market conditions click the link at the bottom of this episode let them know that I sent you thanks guys I would also be curious just um just because we we've talked about this a little bit in in past episodes but you just mentioned sort of the onchain offchain component i' would be curious one idea that we're sort of uh poking at for dexes this season is which parts are which parts of a deck are important to be completely onchain and you know uh decentralized and which parts are we like ah you know maybe actually this this part of the decks is is more okay to be offchain and would be curious to understand like how you think about that

from the perspective of waterlow auctions like which parts of an Offa like need to be or should be onchain and decentralized from your perspective and Dam how how did you think about what parts of unisoft x to have on chain maybe let's start there uh sure yeah I'm I'm happy to start and I'm sure Dan will have other thoughts after that um so I think unisoft X is um a combination of of a few things the protocol that was mostly described in the white paper is actually mostly on CH and it's mostly about uh sort of the the declining the Dutch order uh that happens uh and then uh for us to like create a nice user

experience and to correctly parameterize these orders you know where to start the Decay how long it decays where to end the Decay uh we have sort of for now um attached a uh offchain piece to the start of it so that I know we we ask a bunch of people what how how can uh what's your best quote for this order and then using those quotes we sort of say okay maybe like this is where we should start this is where we should end um and that's technically not part of the uh the protocol itself um but you know it it's it's part of the bootstrapping and now it's uh it's sort of relied upon by the protocol to to

function um so I think uh there's like an ideal world where everything is on chain and you know everything is permissionless and things goes start to everyone but on the other hand there's also a realistic version where we have to start something and we have to see if things work uh so that's where I would say we are now and we're um definitely things are on road map to to make it more decentralized and more like the version that we envisioned um but yeah right now just for Real Estate concerns we have some offchain pieces there that's right so yeah so I think of usop X is sort of uh well I think of usab X generally not as an orderflow auction protocol or anything but really it's an

it's an order format for signing uh intense it's a it's a format for these orders um as currently implemented I think I see it as this as this amalgamation of two orderflow auctions one of which is a Dutch auction um one of which is a is a seal bid first price auction um run offchain and I do think um one of the reasons they do this this the sealed bid offchain auction is because of just limitations of what you can do on chain because of just the latency of instantaneous somehow um it might be a lot you know might be a lot easier to to get optimal price uh using the Dutch auction as well as about you know some parameterization around like oh how do we actually find price Discovery around how do we find the right starting price for a Dutch auction which is limitation

of uh Dutch auctions have um but yeah I think there are other kinds of auctions that you could do and I think um you could imagine you know a sealed bid batch auction you could imagine a second price auction um and all of these are I think you know ways that you can just generally do an auction here you know with when combined with usof X you can form turn them into an orderflow auction why so why do you need a offchain RFQ in order to choose where the starting price is I mean what's the alternative alternative would be just you you could just pick a you just pick a starting price right um or maybe you could do something uh just based on looking on chain but what like what what's sort what's the alternative for picking where you start a Dutch auction you have to

start it somewhere right yeah you have to start it somewhere why not use um you know the top pool for a given pair on chain or use like you all invested quite a bit at Unis swap in routing as an example so you certainly like have a perspective on like a price that you could execute users order on chain so why not use that yeah and well more or less it does use that but then it you know as as a fallback as ultimately like if that's that's what it sets the uh price to if if nobody um if nobody's able to ride a price Improvement on it right um it shouldn't remind me if that's if that's exact I maybe mischaracterizing a little bit there uh yeah so I think that's uh the whatever you can do on chain that's sort of the worst case right uh and you only

you only need this system if you can do better than what you can already do and uh also the I think the to put things into context we're working with like uh basis point here we don't have you know 100% potential Improvement and we just Decay one point each second uh it's really you know you Decay one or two basis points or maybe less than that and so I think it's it's important to get really accurate really fast and that's why you sort of need this RFQ system otherwise you know you pick one point and you Decay and no one takes your order and the next second or next block you are way past your your potential right so that's uh that's what I see so you so you need some starting point and

you're choosing an offchain RFQ as a way to like to get a very accurate price which you're worried about you need that because you need to Decay that price over time and you want to just start at a place which is like within reason correct yeah yeah ultimately in order to get the best latency you can um yeah get the best price while still also preserving laty you could just do a really slow batch auction uh Dutch auction with like with a low Decay to get a pretty good rate although you're still going to have some imprecision due to the due to block latency um but yeah and then so you're using the S chain RFQ to get a price and the only thing that happens on chain is essentially like the settlement of the users intent is that right all liquidity

offchain or it's like Amalgamated by Searchers onchain well there's still the Dodge auction that happens right um it's just uh there's two auctions the first auction decides where the Dodge auction is going to start and end and then the Dodge auction happens and whoever fills it first fills yeah I I guess my point being that this like this doesn't leverage just Unis swap um LP pools like it'll either take whatever the offchain RFQ liquidity is and settle that way or it'll take whatever your solvers find as the the best route on onchain in the case that is not filled through the offchain or fq that's right it's not yeah it's not it's not limited or partial to um to any

particular pool or UNT to on unisoft liquidity right but just just to be clear in either case uh people are free to use Unis liquidity right as long as they're competitive it's always in fact best to use it yeah um how did you how did you all think about the decision to put your flow through an offchain RFQ first because it's I think a pretty significant change to the the market structure that you have a handful of market makers which get as I understand the first right to execute users trades now right they can go through the GDA and if someone getes a price Improvement if I think a 1% correct me if I'm wrong um then they can execute that but it is

initially permission to like a smaller set so can you walk me through like what your considerations were and how you thought about moving to this Market structure uh I think you know first of all it's it's definitely still in beta so uh as we as we try to roll this out we're still like uh on boarding uh these market makers but on the other hand um like as you mentioned there's a handful of them uh these handful of them are practically you know 100% of them uh if you look at the volume they process and the in the competitiveness there so I would say if you if your if your

best interest or if your intent uh as the market structure designer is to get the best execution uh for your uh for your users then what you really need is the competitiveness not the numbers right and so you you have the best NBA team by putting together five players not to you know just increase number of players but to have the the five best players and I think you know as I said like this is not exactly where we want to be long term but I think it's a it's a pretty good bootstrapping strategy just because it it kind of suffices to have a handful of them that's very very good and then we can think about how we can kind of

open the gate to to everyone who has uh the interest to compete in this system yeah the the way the way I'd put it and I've talked about this with Mike in previous episodes is that I think the essential imperative in my view is to move from transaction space to intens space to move from um orders that are from the coming from the unof interface are meant to be executed on a specific pool using a specific route and are encoded as an ethereum transaction because an ethereum transaction as Robert as Robert uh will tell you is like is it not a very flexible thing it is hard to program around an ethereum transaction um it can't be like bundled together with a bunch of other ones you can't do a lot of math necessarily on it you can't execute it easily executed atomically with other things in a within

an ethereum transaction um or at least you know add a lot of complex to try to try to do these things whereas intents are are much more flexible and you order order of Express intent you can do a lot more things with it um and I think that's just essential for I think it's essential for the future of decks I think it's essential for something like unisoft V4 where a lot of pools on unisoft V4 I think are actually going to to a lot of ideas for how to do lever reducing hooks for example um I think are going to depend on having uh orders expresses and that AR that aren't necessarily just like locked to a particular pool but in fact like can be executed on on any of these pools for example um now again I think I think the current design is is you know one one

way that you can I think get potential Improvement on orders using intense but again really like since the order format is not tied to that particular one I think it's actually that's the more important piece of it to me sorry to hijack the conversation a little bit Michael no you're fine man no I do want to hear your perspective though I'm very curious and then I I do I do want to um just ask you guys some of my basic uh questions about just how to construction auction or like how you think about it but I'd be curious get your take Robert before we get into some of the more nitty-gritty yeah I think with with all these things um there's a general trend of taking more and more work that used to happen on chain and moving it off chain in order to give users some sort

of me mitigation and we're concerned about that in Flash spots and centralizing the M Supply Chain centralizing critical parts of of crypto and so I think the question that we need to ask ourselves the three of us four of us on on this podcast today but the entire industry is how we get to that point where we're giving users best execution without accidentally centralizing all of uh all of the critical parts of the MV supply chain along the way um there's a broad design space and would be be curious also and and Shan and Dan how you all are thinking about the roadmap for x and decentralization and what you all have um you know on the horizon here um I have maybe somewhat of a

random thought thought um but uh I guess you know as a question I I do wonder how Robert you think about um things like private rpcs are they sort of uh would you consider them more of a exclusive setting uh one thing we try to uh this separate research that I was doing I was I was wondering if if everything switched to uh a private RPC or if some you know some varying percentage of swaps uh switch to a private RPC whether uh it will be more centralizing or less centralizing in a way uh and the conclusion I think I I came to was that uh it will be less centralizing in the sense that uh there

will be less extractable value assuming people honor their sort of social contract in the private RPC setting so there's less to extract and so naturally you know the the even the best guys cannot extract more than the the worst guys and so you don't have this like Winner Takes all situation uh but it will be more centralizing in the in the sense that uh now you have a barrier to entry right and so now suppose there's a new guy who's much better than everyone else he cannot get any flow uh because he's not connected to any of these private rpcs and there's no deals between them and yeah so so I do wonder you know how or have you have you

thought about it and if so what was your uh what was your um thoughts on private rpcs and uh and and do you think that's that's a that's a direction that we should or will move towards it's a good question and I appreciate you asking it Shen um so are private rpcs more centralizing than uh like essentially centralized intent infrastructure in general is how I'm taking the question and I think it like like I think it's hard to say which is more centralizing than the other I think in practice they're both very centralizing if I wanted to come to some um some numerical amount that would tell

me which was more centralizing than another I'd probably look at like the economic value that flows through these things um and to that end I think there could be a I think right now uh I would expect these things to be about equivalent I would worry more in the future that um the centralized intent Based Services end up becoming more centralizing because they start at a point in supply chain which is before private rpcs too and so private rpcs May obviated in a year if everyone is using intents because the primary way that users transact on the Chain will not be through ethereum transactions it'll be by expressing your with a signature that someone else executes on your behalf and

in that world you don't have much of a need for private rpcs well you have like a sorry you for whoever is filling and whoever's submitting the transaction that includes those intents right like so that the intents can't be like plucked out of their transaction and sent and done shared somewhere else like somehow the block still needs to get built out of other people's transactions maybe unless unless the Builder is just is just operating directly on intents yeah that that is a good point um that is a good point so private rpcs will still be around I think the the caveat to that may be that like the extractable value assuming that your private RPC actually holds like is credibly neutral and holds up to the properties that it says will be a lot

lot lower at the private RPC Point than it would at the like intent centralized service point or whatever we're calling it the intent point of the supply chain um both of them concern me a lot for for what it was worth so I don't want to like minimize any concerns around private RPC also and I would like all of that infrastructure to also be built on sve too all right um guys I I want to get into almost like treat obviously you when we talk about something like auctions I think it's very simple if you're not in the weeds be like oh yeah you run an auction and you know the highest bidder gets you know XY Z thing that you're auctioning off but there's uh just almost like infinite sort of permutations right of different um if different uh auctions that exist and I

would love to get a sense of you know when you're going through actually constructing an auction like what are some of the design decisions that end up making their way into the final product what are some of the different levers that you pull I've been waiting for four episodes to ask Dan why he loves Dutch auction so much and why they why they find their way into virtually everything that this man designs but uh yeah I I would just be very curious like almost you know treat me like a 5-year-old and I'm looking at something like Unis swap X or one of the like maybe Mev share uh for instance from from flash Bots and it's like all right what are the different um you know what are the different types of auctions that exist and what are some of the design decisions that that end up running

through your head as the implementor I can answer on Dutch auctions if that's I want to know yeah I gotta know yeah well for me yeah I think I think whenever when you're reaching for something for for Price Discovery particularly um uh for things that are IL liquid um or don't where something doesn't an oracle doesn't exist on chain um what I what I generally like about I generally like auctions just as a method of price Discovery and I think Dutch auctions are uniquely Well Suited as a just off-the-shelf solution um for Price Discovery um on a blockchain because ultimately you know if you have an English auction for example you have to have uh uh censorship resistance basically over the over this you have a long enough period of sensorship resistance um to get enough to get

enough bids in um which you also need on on Dutch auctions but there's sort of a limit to the amount of value that could be extracted by a sensor for a period there um and then you also need an English auction altim the number of bids could be a lot higher the number bids actually H to the hit the chain was an a nice thing about a Dutch auction just gas efficiency wise is that because prices start High and um uh you only actually have a have it finished when the auction uh when transaction goes on um you don't have as many transa it's it's a lot cheaper in terms of gas onchain and then finally um for onchain auctions why I like Dutch auctions is once the once the auction is uh the bid also happens at the exact time that the auction completes so you can like do a

lot of other things potentially at the moment of that that uh uh the bit the bidder at the time that they that the auction completes I can do a lot of other things so if you think like a uniso x order that has gone to a Dutch auction um at the time like that that uh bidder actually gets a place on chain they can also ARB something on chain atomically or something like that as opposed when if you had like a lot of bids that were had to be Bon chain um and then potentially canceled as new developments happened so yeah so I think I think it's just uniquely well suited for the onchain environment I would actually say I'm not a big fan of Dutch auctions outside of outside of onchain I think like actually um English auctions are sealed but second price auctions are

better in uh from an auction Theory standpoint in a lot of in a lot of uh situations but onchain and Dutch auctions are equivalent to a um uh to a to a seal B first price auction rather than a second price auction so um uh in theory but uh but yeah but I think from an auction Theory perspective I don't love them as much but on just in terms of onchain implementation which I think is just really important so that's my that's my pitch for Dutch auctions um uh as an off Toth shelf uh price Discovery solution and I should say you know I think Dutch auctions for um uh uh Mev as as as a way to implement an OFA aren't original to to Unis swap X or us I think um uh there was there's an idea generally for um like an escalator fee

algorithm um that I know flash Bots had been uh thinking about and some others in in the ethereum space and been thinking about including as a core protocol feature um and I think it's cool because it you know you could build it's a it was an alternative to basically the 1559 auctions that end uh or uh second price auction basically ended happening for uh um as a way to to handle gas price but um but yeah but it would do it in a way that potentially could actually return more of the Mev to users that they created you know it's it's the first um again it's the first version We intend to iterate on it so when we were sort of thinking about the design we we didn't really you know exhaustively survey every single uh auction format and like do a do a comparison or anything I think the the

things in mind are first of all when you have a um repeated game uh a lot of these auctions kind of just give you the same equilibrium results and what you want is sort of ease of integration and implementation um which I think you know um like seal bit sealed bit first price auctions are pretty easy to do uh and the Dutch auctions are as D mentioned uh very suitable for blockchain environment because minimizes sort of needed communication and communication on blockchains are are super expensive and all that so we sort of just pick the the easiest

format um and we kind of analyzed the the potential strategies people can can use based on these formats and and we sort of went over the corner cases and we think okay maybe we we covered the ground more or less um I think there's there's a lot that we can potentially do on top of what we have right now like you can imagine uh allowing sort of batching of of bits or batching of orders or um I think you can allow different type of bits um like conditional bits or or so on so uh yeah a lot to be uh experimented and and discovered in the future but

yeah I think to start we just pick something that's easy to use I'll jump in real quick with the considerations we had with be share when we were thinking about Dev in M share we talked to a bunch of wallets this was quite some time ago and essentially all of them told us that they had no appetite to implement uh you know intent-based execution essentially like they wanted any OFA that we built to be uh backwards compatible with regular user transactions and uh you know at the time there was much less appetite from our good friends at the application layer to use private rpcs to um is how I would frame it to ask users to switch to private rpcs and so it didn't seem like

at at the time developing Ana which is face on signatures was really really compelling in the absence of us um since you know flash Bots doesn't operate a front end like like uh unisoft does as an example so we needed it to be backwards compatible we needed it to support MVB redistribution for regular transactions interestingly you can use M share for other types of things um for it's compatible with other types of orders like unx orders that's an aside um we also wanted it to be permissionless for Searchers so anyone in the world should be able to search on um the flow that goes through M share and we didn't want to enshrine a single block Builder um there are variants of

orderflow auctions that are a little bit that could be closer to a builder um but we wanted to make sure that you could use multiple different block Builders and have a more healthy decentralized Builder Market with me share and if you take those as your constraints in designing an Offa I think you come to roughly M share um although you could do it with different privacy settings too so that's how we thought through the design of of me share and how we ended up coming to the design that we did yeah it it's just fun even you know when when doing a little bit of research for this episode there was some sort of counterintuitive lessons and I don't know when when you hear the word auction maybe you think in your head sort of this like canonical guy with like a gavel and it's like do I hear 500 or 600

or whatever but but there are tons of different types of auctions that just occur in the real real world that have very different parameters based on how much information is known at the time the time in between your bids and you know when you know whatever gets released and uh like payment for order flow would be in trafi is a really good example of a very different type of auction than the one that might be happening in your head where bids are submitted well ahead of time and exact information isn't necessarily known and even when I was looking into sort of the OFA design space within um you know within our world within within crypto here there was like even just for something like permissioning right like who should be allowed to participate in an auction my immediate you know gut take was that the more people the better

right the more competitive the pricing is going to be but then you also end up there's kind of like this this concept of winners curse and suffering from adverse selection uh because you know the people that win that that auction might be doing it they might be pricing it incorrectly and sort of the more optimistically and sometimes unintelligently you price it the more disproportionately you will you will sort of win um so I don't know if if uh you know all the three of you having thought through many different you know types of auctions and scenarios if there are other sort of funny lessons that you've taken away either from permissioning or when information gets revealed to participants or like sort of the strategy for um or sort of like bid selection or settlement or whatever it is but would love to I don't know if

there were any like sort of surpr something that was sort of surprising to you from a design perspective in ofas or we just sort of proceeded into we intense because we've talked about the the limitations of the the transaction on ethereum but I would love to love to get into intense but I don't know if there are any sort of surprising lessons that you've taken from designing auctions over the years I let Dan I shouldn't talk to things like Windows curse and ERS collection and some of the more hardcore auction Theory stuff I will say that one one anecdote that I have which was surprising to me at the time is how even like a relatively large sized pool of competitors can collude to keep prices down in auctions so the anecdote is is that uh when uniso V3

launched almost two years ago it it was very challenging for Searchers to integrate uh and you all also launched it without that much of a lead time for the market and so it kind of just dropped a bunch of liquidity migrated to it and a lot of people then were trading on unisoft V3 whereas a month before the primary um the primary amount of volume on chain was going to V2 and things like old curve pools and and whatever and that meant that for a period of time there was an extremely small pool of Searchers who were able to AR unisoft B3 Pairs and if you were a Searcher at the time the amount that you had to bid in the auction uh for unisoft v2 arbs was

something like 95 to 99% um because it was a really competitive market that was spread out across all the entire world and for unisoft B3 even when you had something like a dozen Searchers who could ARB unisoft B3 they managed to arbitrarily keep the price down in the auction to pay only something like 60% of the value of the ARB to the minor in meth and that was quite surprising to me how long a collusion ring could go on with um how large of a pool of of bits were in that collusion ring too and it was within their incentive um the funny part about this that I'll I'll drop as an anecdote really quick is what would happen is a

new bot would come online uh and start arbing unisy 2 and people would go to the flash Bots Discord and like just yell at addresses say like hey zerx you know babe or dead you are aring a swap P3 like if you don't come and join my collusion ring I am G to grip trigger you push your profits to zero you you don't need to be bidding all of this to the Val to the to the minor at the time you know we can we can work together here something like that and people would find each other this way they would like check the Discord and this actually worked for some time so um there was wasn't there something like there was there were people who would who would alternate even an odd blocks or they'd alternate like every like uh

every few blocks or something yeah yeah yeah yeah for they would do something like there was some probabilistic way I forget how exactly it was done but some probabilistic way that like you would have an even distribution of the ARs at a certain point of profit um for V3 I didn't hear the even odd blocks thing but I wouldn't be surprised if there were two participants if if they did that um anyway takeway here just was for me that collusion finds a way and uh it works with a larger set of people than you might expect man that's a great anecdote do you do you ever any like theories for how they were coordinating was it just over the flash Bots Discord was like other Michael I don't need a theory

they're doing it on Discord in broad daylight people would post addresses and transaction hashes and say like please DM me like let's keep bribes down blah blah blah oh my God that's so funny man that's really interesting um Shen I don't know if you have any anything or we can go into intense oh I was gonna say uh interesting social coordination uh use case um I don't know if you guys were following the eth research um Forum but I think someone posted something about uh Searcher Searcher cooperation or something I wonder um it looks like a cool technical post but I wonder it will evolve into weird use cases at at some

points maybe I'm misunderstanding that post but uh it definitely you know uh it rings a bell now that you you mention this use case um but uh in terms of surprises I I don't think there's a ton I think maybe as you said yeah like as you the like the optimal number of participants in auction if your objective is to maximize the proceeds of the auction is maybe not Infinity is uh like a unconfirmed surprise to me like that that being a

POS possibility to start uh you know intuitively I just thought you know the more the Maria but then the more I think about it the more like I'm like maybe maybe it's not Infinity maybe it's you know 100 or something uh or not even 100 but there's also argument for you know maybe best execution is not the only thing that we're trying to optimize for uh and in that case you know the that that optimal number will will change I I've got sorry two more questions for you here and then Dan I'll turn it over to you to talk about intense but something I was I was curious about um is H how does the when the information gets revealed impact the

pricing of that information and and what what I'm referencing here is like you know I think about the the EV the sort of expected value of this information there's like the there's like an ordering component and then there's a signal component so like if I have this information then that's valuable to me and those are kind of the two components into how much you might bid on a given you know block or or bit of order flow or or whatever it is um and I would imagine like if that gets revealed to all the participants suddenly the signal portion of that order flow is like well this is now available to all the participants of this auction and that isn't super valuable to me as opposed to you know if that information wasn't necessarily revealed um then you'd sort of be factoring that in so I I would

just love to get a sense of I I have the intuitive feeling that like when information gets revealed in an auction is very important to the outcome but I'd be curious from your perspective whether or not that's actually true I think as you said right there's there's there are auctions where um bids are placed before anything is is revealed or bids are placed before the the thing even happens right then there are some some auctions where bids are placed when the the transaction is signed and you send a executable into the M Pool and then people will send bits uh I think yeah I think those are very different and uh depending on exactly who you send the actual intent to uh they will assign

different value to the um to the thing like uh suppose you there's two cases Case Case one is you uh you give the exact same uh set of order flow to someone and they have the exclusive right to like you know um put bundles around the the swaps or uh or or do whatever they want with them on chain um but before you do that you you tell everyone else that you're going to do it and the second case is you you only give it to them and you you don't tell anyone else and they they are responsible for like constructing the orders and and do what they say uh what they see fit uh and obviously the second one is going to be worth more because

there are a lot of actions that people can take out of outside of directly interacting with the orders right they can they can do things on centralized exchanges or um try to like front run back r or do other things so yeah in general I think that's like a pretty helpful um I don't know if it's a ex exhaustive dichotomy but it's a pretty helpful framework to think about these things again um you know call back to the me season we did you know there was this term popped up a couple of times like auction extractable value actually as opposed to maximum extra actable value and I guess like one of one of the problems that that like I've sort of thought about as well just in terms of

an orderflow auction outside of the strict design is how do you host one that's in credibly neutral and in a permissionless environment which many of these orderflow auctions are like really basic question of how do I make sure the auctioneer isn't actually participating in the auction as well in addition to many other um sort of concerns that participants might have so I'd just be curious you know for uh maybe all of you like how do you think about hosting like incredibly neutral auctions especially in a more permissionless environment I think I think Robert could could talk a lot about especially especially the trusted execution environments but I want to point out I think this is it's maybe early for this in orderflow auctions but I think Robert's also seeing a lot in terms of what relay what is possible in the relay Market or what relays can could do in um

uh in the Mev boost auction so I'd be curious for his thoughts on that as well yeah so there's a lot here in the me boost auction um perous if you're referring to something in particular Dan I think like what immediately comes to mind is that uh and you can tell me if you're you're thinking something else is that relays have sort of a privileged position to be able to say what block wins um or it act on the information within blocks that Builders are sending them to like update their bids faster or what are you specifically refering yeah I'm thinking of just the ability of a of a relay in the a dishonest relay to um Penny like another re another bidder's bid right they know exactly what the

other biders are bidding so they know exactly how much they'd have to bid to win yeah exactly um so the the relay right now is uh unfortunately sort of like a second price auction um in in some ways which is why Searcher Builders are as powerful as they they one of the reasons why Searcher Builders are as powerful as they are I think luckily I think we've gotten to the point where the majority uh relays or blocks that are going through are from neutral relays that don't run Builders or aren't integrated as a builder but it is a worry that I have where you have a more integrated Builder relay that um you know it's accepting external blocks and then using those external blocks to immediately Penny on top of them and win

the meth boost auction um that's one of the reasons why we want to move the auction and M boost uh on chain not on chain but within the ethereum protocol too to obviate the role of relays and like obviate the ways that you could Penny or mess with other other bids um at the very least leveling the playing field in in some way and ideally implementing a seal bit auction instead of this public auction that exists today um and they did do you want to talk as well about ways to ways to implement credibly neutral action infrastructure yeah so that that's one thing we think about here at at flws a lot um you know we've been concerned with decentralizing flws ever since we decided to build this offchain

centralized infrastructure to CH Point earlier today and ever since we saw uh this you know Trend coming of offchain infrastructure is going to become increasingly important to to ethereum and have increasing amount of economic value going through it the way that we're thinking about decentralizing today and and how you run these credibly neutral auctions is um uh creating a platform which we call sve where you can Define uh the rules of your auction as a smart contract and the execution of that smart contract contract happens within what is called a trusted execution environment um so this is a a piece of Hardware which is designed in such a way

that even the operator the person who physically has the hardware can't tamper with or see what's going on inside of it so you know that a piece of code um is running as intended within a trusted execution environment and you know that you have privacy with them too so developers can Define their orderflow auctions or like a me boost relay or a builder you know in some sort of intent service within a smart contract on top of sve and the when um you know a user wants to use that uh what the smart contracts execution dictates will be loaded into an sgx and and executed within that SX and so that gives you the

credibility of a trusted execution environment it gives you privacy so no one can see the orders and critically what it also allows is for anyone in the world to be able to run these trusted execution environments and and execute users orders instead of relying on you know a flesh Bots or some other trusted third party to do this type of execution um so that's how we're trying to provide like General infrastructure for credibly neutral and private auctions and remove the trust that you need today and trusted third parties in order to run these things um because to your point Michael I think it's very hard sometimes to tell uh if you know auction extractable value is being extracted I think it's it's

really interesting it's not in the domain of of crypto but it took the Department of Justice years of collecting data in order to come to the conclusion that Google and their at auction platform was messing with their auctions and you know the contention is is that they were extracting auction extractable value um so it can be very difficult I think to to tell whether auctioneers are putting their finger on the scales in some way it's funny you brought up the Google example I think that was the very first thing that we talked about in that Meb season was Google and and their ADW auctions so um I think this would be I'd love to close on on the idea of intense so I know Dan that was something that that you were very interested in so maybe if you could just sort of preface us with sort of the

the seismic shift um that we're seeing in terms of the sort of ethereum transaction to a more intense shaped world and like where that belongs and fits into this whole word of La auction discussion yeah so I think orderflow auctions are actually a perfect example of the kind of thing that I think intents are really useful for so um in a transaction-based world um if a if a user is trying you know sends a signs and sends an ethereum transaction uh to trade on say a particular Unis swap pool um they create some potentially some Mev uh from doing that maybe some Mev from being possibly front run maybe some Mev from from just being back run um on it but the way to capture you know if you try to capture that meev you have to basically create a new transaction back

on that other transaction um and that's a very lossy process you you you have a lot of transaction fees on um extra transaction fees being done you have extra you know extra gas being paid for this um uh and you have fees being paid to LPS for like this this trade being done in both directions on the pool that actually sort of Nets out to neutral um when ultimately economically what's what's supposed to be happening there is that a the swapper is supposed to be swapping is is basically swapping with a market maker they're swapping with somebody who's who's back run them who's offering to backr them um at least in the case of yeah of like backr backround Mev and so that's what uh uh how existing ofas that have to work around the transaction um uh uh model um so Mev share flash Mev share is a great example

and we're g talk more about how that works but it's an auction to basically be to be the back runner in this case but my point is I think this it's a fairly leaky way to do this um and it it means you actually miss out on potentially a lot of of uh of possible pric Improvement that you could get if you actually had an intent model where um all I'm doing is I I'm saying I want I want to trade and the auction is actually just the right to basically submit a transaction that fills my intent um uh and and you can you could just trade against me or you could fill it on chain um but basically we don't actually have to go through this multi-step process because uh is inefficient um and that would done on chain um because you can do it in

intense space um yeah Robert I know do you want to Robert and I have been going back and forth on this for a while but I think he I think he sort of has a similar has a simar similar view in in general but yeah I could talk more about M share no I think I think you summarized it well Dan um I think you do want to move to in intense space I think there's a lot more you can do with that I think it's probably better better ux for for most people I think like again the trick with this is just moving to intense space in a in a way that doesn't end up centralizing um the the MV supply chain and don't have anything more to add than that I think he summarized it well and yeah I think it totally makes sense by

of course that like m share another another uh o actions that were designed around the way that Unis swap I think it's Unis sw's fault more or less that transaction um uh transactions work this way where you it's actually hard to hard to sell the me back to the user or it's a very wasteful process if you try to do it um but yeah but I think that that was like a key motivation for me at least in and being in going to something like unisoft X yeah I think you can do some some things in order to get um get around the gas cost so in most Med boost blocks you have the Builder paying a transaction to the validator already so you're already incurring like 2100 gas times base fee worth of of cost and turning that into a smart contract I think has the overhead

of like 11,000 extra gas or something there's like a point at which it makes sense in order to or rather after um there are actually cost savings from turning it into a smart contract and you can like amize that over many people um maybe there are other things you can do so you don't need to pay the user immediately you can have them claim it's like a there's a broad design space but I agree that it is it can be kind kind of lossy and yeah I'm cleaning up your design choices Dan you know so maybe we should talk a little bit more about future design choices clearly clearly that's very magnanimous of you Robert very magnanimous um so I I guess like maybe just then this will lead in being selfish question here because

it'll lead into Dan in my next interview but you know when you think about a world of intense um and espe there being maybe front ends right that sort of collect uh these intents with entities like market makers on the back that manage some of the complexity I mean does that you know what's the immediate unlock there obviously there's a concern around centralization but do we also solve some of the the cross-chain problems that we've been struggling with for so long in this space I mean the bridges seem to be I I don't want to say they're they're broken they're definitely a a part of crypto that drastically needs to improve right some crazy amount of the losses or exploits that happen in defire due to bridge risk and it does feel like the architecture

of something like Unis swap X um obviates some of the risk from a from a user perspective so I I would just be curious like how you all sort of see if if we if intents to get adopted and things like um you know uh sve and the whole platform of uh different ofas that operate there or Unis swap X or whatever it is like how does that impact the sort of cross-chain interop which has been a problem for a long period of time I think ideally what you want is is users to just express what they want you know I have asset a on chain one I want asset B on chain two and then have a Marketplace decide what the best path is and take the risk take on risk on behalf of the user for you know managing things

like Bridges managing things like reorgs and so to that end I think it could really simplify the um you know the user experience I think like beyond that one thing that is not int may not be intuitive to readers is that it is oftentimes uh as someone who's trying who has has tried to create an RPC product within a wallet it can be challenging to offer the kind of experience you want with the way that wallets usually work um and Uma I think had a great presentation on trying to create a quote unquote Universal RPC um at uh dba's research day specifically what that was is um she was trying to create a way for a

user to have an eth deposit on one chain and have that eth deposit show up in someone's account balance on a wallet on on another chain and it actually turns out that even if you use an RPC it is really challenging to fake a user's eth balance to their wallet uh or adap even if um the balance isn't really fake it's just on another chain in the rpc's being smart and if you move to intense you have a much broader scope uh and it is much easier to offer interesting product experiences to users like doing this uh this product that Uma was trying to make where you have eth on one chain that is paying uh that is used to pay for a user transactions on another um so domain

abstraction in some way and getting out of the constraints that developing within wallets and like normal depths uh uh put on you today um well guys maybe maybe we can I know we're we're winding down here maybe we can leave it there um Robert and shin you both been super uh um super generous with your time um for folks who want to find out more about either two of you follow you or find out what you're building um at flash Bots or or Unis swap what's the best way to do that for both of you bery Miller on Twitter flash.net if you're interested in reading more about Flash Bots yeah I'm shin1 wa on Twitter and then um some of our paper has got uh put onto our blog

thanks well this was a ton of fun and thank you both again all right Dan great episode great episode what an episode yeah yeah two you know I was really excited about this one in particular we we referred back to it a whole bunch the but the season with hu ended one of the the big questions that we were asking ourselves was General versus application specific orderflow auctions so it was really fun to get shin and Robert on and represent sort of both sides there um I I I thought I mean I I don't know if you had any specific takeaways or or where you kind of wanted to start but I I thought where where we started just in terms of um sort of the market size of a of a general order flow auction and sort

of this aggregator of more arbitrary broad-based sort of intense versus something that's a little bit more swap specific love to maybe like get into I had some takeaways about the specific pros and cons of each approach but wanted to get your take when you think about it uh you know ethereum runs an order flow Au ethereum runs an auction for um uh you know transactions getting included um uh and they pick you know they for a while it was this it was this priority gas one um I think with with sort of the rise of of proposer Builder separation um and dap 1559 you've added other elements into it but there are you know ultimately you do have to order all these transactions and um incl decide which ones get included and so uh you know I I often talk about how and in

this episode talked a little about how generalized um auctions are just Inc with generalized computations just incredibly hard to do unfortunately we kind of have to do one if we have a generalized uh platform if we have ethereum so right now you know and we we we started with sort of the uh sort of more naive uh versions that miners were running about just what what transactions to include and then this got more it got more sophisticated with the flash BR auction and with um uh and with proposer Builder separation but like ultimately you're going to have some kind of some kind of way of deciding what what transactions gets included so really what it's about is how do we how do we return the um the Mev more to the to the users but that said you know I think uh I think that's

one direction in which to come at this problem is sort of from the protocol up and think about like we've got to build these blocks how we're going to build them the other way is from the application down and say like we've got you know Unis has this has this transaction um what are we going to what are we going to do with it they have this user intent um they want to give the user the best possible price and there again I think you get actually do get a lot of Advantage from just getting very specific and saying we're going to build this application around the IDE a that we we we might want to run an auction for this particular piece of uh this order flow and that's why again I think unop X is designed in order to make this uh the user intent a little

more tractable as an object for for that kind of auction yeah I tend to agree I there it's it's almost like uh it's like auctions all the way down or that Meme where it's like it's all auctions always has been there's a you start to see them everywhere I I I do think as well you know there are kind returning to this theme that we've had throughout the season of reducing leakage outside of the Dex ecosystem there has really been one auction in town for a long time which has been the Mev boost right or let's just call it like the block space or execution level auction and as such a lot of the value has been getting um has been getting distributed to eth validators but now what we're sort of essentially saying is there's going to be this this second auction that's run

almost Upstream of that and a lot of that is going to redirect value another way right and there's kind of an auction for for swaps where price discover happens but you can also just a way to redirect redirect value and keep it within the deck ecosystem that's right now now as a mechanism designer I do think there's a lot of Advantage at I I find it easier to start from the from really the simplest um uh application and I this is why I think I've I've tended in thinking about these auctions to think about them as and primarily from a DEX lens and think how do we make a DEX that works like this because building an auction that's like that supports uh arbitrary computation um and arbitrary intents and desires to you know front run and back run these transactions Etc it's just incredibly

hard to express all those all those kinds of preferences and actually be able to meaningfully aggregate across many different people's um preferences includ in this you know infinite dimensional computational space whereas um a an intent to trade on Dex while while very complex and there are whole textbooks written about um how difficult it is to actually design this this system properly um it is it is certainly much much more compact of a thing and more and more tractable of a thing I think than the general problem and still has all the same you you'll end up hitting all those all those same problems if you build the more General solution anyway so that's why i' I've tended to to try to think about these from Dex from a DEX lens that's also just where I'm coming from I have this framework that I inherited from my days

during Consulting but it's this guy who owned this company used to preach this it was like Focus Simplicity speed it was like business focus process Simplicity uh speed and I I actually remember there there was a point where I was paying a lot of attention to the a uh strategy versus compound and it seemed like for a while they they had this sort of Divergence where the compound strategy was to focus on instead of being like all right actually to explain a the a perspective is kind of like we want this to be a marketplace where you can pledge any kind of collateral like Borland any type of collateral spin that up whereas the compound strategy was people just want to borrow and lendon usdc look at what look at what's been happening with stable coins and we're going to be the most Capital efficient venue for borrow

lendering of stable coins and I do wonder if there's a little bit of an element there with what we're discussing with the these these uh these auction sort of venues where it's like look it could be like a broader you know display of things but if it's mostly swaps maybe we we start and focus with and we we've talked about this before in the context of like Dex design that you have some number of complexity points you can add maybe every year and you have to be very just stingy with them um and that's why and you know you saw V1 started with an incredibly simple design um V2 only makes it slightly more complicated um and then some steps up certainly steps up in in V3 and V4 but I think you don't try to go go all the way to like you know we're going to be as complex as

current um centralized exchange uh uh infrastructure is today um that said I I do think there's a lot of uh value so conceptually often in generalizing or abstracting really because when you when you look at a problem at the at the specifics of a problem sometimes um it can help to say like okay what what would be like the fully General solution fully ver General version of this problem and then what's a fully General solution to that and often it it will then when you once you've generalized it you will actually see this is more similar to a problem I've encountered in a different domain um so I do think there's again here here we're sort of getting into maybe tactics and just in trying to like invent invent mechanisms like this um but sometimes it can be useful I think to to just go as hard

toward generalization as you can early on just to explore and see oh does that actually unlock something that you wouldn't have noticed in the in the individual simple case yeah the I think the other the other bull case for a more General design like a like a Suave would be theoretically more sources of order flow right like if it's a sort of neutral platform where you can spin these these auctions up then You' theoretically have more access to orderflow you could build more profitable blocks and there'd be a fly wheel that that goes in that direction as well yeah absolutely and and you have access to a lot of more liquidity because all that liquidity a lot of liquidity is like locked up in complex things on chain or offchain that potentially could be expressed in a Suave auction in a way it may not be in

a simple like batch Dex auction yeah I agree with that what did you think about I mean there was a lot of discussion yeah little little little tense discussion too but we we love that about the the onchain versus offchain component and the centralizing and I know you've I'm sure you've thought about this quite a bit and I to be honest I do empathize with Robert's perspective a bit where it's like look if the answer is just to do this offchain then then what are we really doing here on the other hand you know I think to myself I mean is does it really make sense like when we have trillions and trillions of dollars on chain and I that's the vision that I see here and that's where I want crypto to go does it really make sense that the people that are participating in these auctions or

like retail people like do we have to be designing this in such a way that it's like retail people that are participating in these incredibly sophisticated very high stakes Finance auctions and kind of like I don't think so I got inition would tell me my mom's not gonna be participating in that auction so you know I guess that's where I come down I well yeah well so I will say I think um there's a lot of questions wrapped up in there and one one distinction I want to draw is that onchain doesn't necessarily mean more decentralized um what what goes on chain well in for example in a protocol like ethereum right now what goes on chain is whatever the proposer wants to go on chain so certainly an auction for example suppose you were to hold an English auction on chain where everybody put their bids in a one block English

auction on chain that is not a that's not a secure auction and it's not a decentralized auction because ultimately the proposer has will censor everyone else's bids and they're they'll be the only ones to get included right um in some ways I think a lot of offchain infrastructure can actually be more decentralized because it doesn't actually depend on doesn't have this dictatorial control by a current by a current proposer so I want I want to draw some distinctions there where like onchain doesn't necessarily mean more more decentralized um but yeah I I I also think you know this uh the what you said there is is true as well I think this goes back to some of the tension between professionalization and um and amateurs um that we were talking about where uh yeah like some some kinds of um uh designs may may depend more on the

existence of or benefit people who are uh uh you know more professional parties and I think that you know that's true in Mev it's true in Dex um and I think it does uh introduces certainly uh vectors for um uh for centralization uh potentially but also it I think it introduces um you know in addition in addition to a lot of efficiency gains which mentioned it actually introduces sometimes more decentralization because often you can actually get a lot of the work done that that you need to get done sort of bu this by this abstract Market um a market in some is can be very can be very decentralized um whereas I think in in other systems you sort of when you're depending on uh you the

alternative often is like you end up with like one one party that can do it again I think Mev is a perfect example we have a single proposer for a particular block yeah and there there are all these examples Finance is a little bit like I mean it's like other regulated like Healthcare in a weird way it's it's it's analogous because there it's very sophisticated the work that gets done there and it's very uh crucial to people's lives right like if you mess up something in healthcare people die and if your banking system goes kaput then there enormous cons and immediate consequences to that so you there there is a required level of sophistication that does get a little bit moreor you know I'm not defending non- permissionless I'm just stating it's a complicated it's a complicated problem now what did you think about the the

question of just how do you run a a credibly neutral auction I've heard I've heard you talk about this before I know some of the people in The Flash spots Community like to discuss it there's a uh in Cosmos there's there's this idea of vote extensions which sort of allow you to onchain run this credibly neutral auction because you can't be censored by the Whoever has the Monopoly on constructing the next block so I mean I'd love to get like your your sense of like how you think about hosting a credibly neutral option what goes into that I think the I believe the families of solutions that I've seen um I think it's relevant for the for the Mev boost auction it's relevant for orderflow auctions like unop X um it's relevant for um yeah for for a lot of for batch auctions for a lot of things um I think

the fames of solutions I've seen um often involved so what one family would be involving uh some kind of consensus process right and I think um uh you know that that might involve it might include like threshold encryption it might or it might just be something like vote extensions and Cosmos where instead of just having one proposer contributing to a block you have a lot of different people and you need you know two-thirds of them or something to to cons uh so missions from two-thirds of them to construct a single block and that can be used for example to in a yeah in vote extensions in Cosmos is a mechanism for ting that's that's that's one way to try to say okay uh you know a message will only be censored if like two-thirds of uh validators of this whole set are

censoring so it's sort of like going from a single Sig to a multisig in terms of security um uh or censorship resistance on the single block level um so that's that's that's one family there and some I think you can improve that sometimes by adding like special encryption where um also people can't see the orders that they're get that they're including um and there there are other other ideas about how to how to do that but that's one family another is to use trusted execution environments which is one that that Robert was talking about um and I think there are limitations as well on what you can actually guarantee with that but that's you know trusted Hardware Intel sgx something like that um and then the third one which isn't always possible is just after the fact detection and and um punishment of someone and again I think

many times as Robert pointed out I think that's just not feasible and and people can actually and Shen Shen mentioned uh you you can sometimes get away with just abusing um auctions like that um the auctioneer can with it's tough to detect but I think that's you know it's it's another tool in the toolbox is if you know you sort of just like detect look after the fact to see if this if these auctions were executing in a fair price or something um then you know and then punish the or or you know disqualify move on from the uh auctioner if it doesn't work you know what I was just laughing about that anecdote that Robert told us was so great because you know there's so many brilliant people in this space you know yourself researchers the people people who really focus on mechanism design and there all there's

all this um emphasis on how can we avoid coordinating parties on chain all that stuff and here people are just hopping into Discord you know like openly colluding it was just oh yeah collusion resistant infrastructure is kind of impossible I think it's it's it's that's that's just sort of unsolvable unfortunately um and I think I think he's I think the the main thing to do is to try to is to lower the barried entry and ultimately I think that from what I understand once that uh uh I don't know if I'll call it a cartel but is it a cartel maybe it's a cartel once that group of coordinating uh Searchers got large enough I think at some from from what I've heard at some point it collapsed and and um now you know the the ARB is a lot more efficient so ultimately I think what matters is let's

just make the markets more and more efficient so that it's not possible to um there aren't such a small number that they can actually collude in that way but yeah I agree it's it's um uh it is troubling it's tough It's a tough problem and uh and I think I think that's a challenge for auction like big challenge for auction based uh Mev Solutions or auction based sort of anything yeah it was maybe to just described by we talked about Winner's curse and adverse selection but it it's funny that Robert immediately launched into this anecdote because in a common value auction where you have winners curse the ideal uh strategy to deal with that is collusion um and and winners curs this idea just be like very explicit about what this is right when you're when there's a common value

auction where like a swap right is the same value for everyone everyone's kind of running these different calculations on what that's worth the reason there's adverse selection there is because you know someone who just might not be as smart as pricing at pricing might think you know they might just bid something that's kind of a stupid price they're actually bidding something that's more exceeds the value of the swap but they'll continue to win and win and win and and that's the the sort of adverse selection there so the the other the other solution there is to shade your bed um and I think I think you know ultim it's just to lower your bid because you know you might be wrong um but yes I think that's that's uh is a um yeah I it comes up all the time also once you start looking at things in in

winners curse uh terms yes well the the history of it is there were it was in 1971 uh some it was like a oil and gas yeah it was oil and gas I think yeah some economists ran this study on why consistently uh in this area like they these companies just kept going bankrupt and they won these deals it was like oh well they weren't bidding in a particularly sophisticated way so they kept winning the bids but then they'd go bankrupt and have to capitulate the yeah but right if they if they collude to to C their beds then yeah then they might um well they'll probably face some antitrust problems but they also yeah it could potentially Escape that where they where they just share information with each other ahead of time um well I I

also just one funny irony is one thing we're building here is coordination infrastructure and coordination done by done by your your enemies is collusion so yeah one people can use blockchains often in order to actually coordinate this kind of uh um uh action so I think that it's it's an irony here that you're building the the same tools often can be used in order to uh to cheat the mechanisms you're trying to build okay let me ask you I've never asked your opinion on this there was a there was a great I forget someone tweeted this out but like this the it was paraphrasing here but it's like the solution to meev is that you know if you do illegal things you go to jail I was like oh boy yeah I mean there probably is some element of that that's true and I I

guess I just you know in hearing you describ like collusion is is always going to occur especially in these permissionless systems if it's not illegal people just do it right out in the open on like an official Discord Channel I mean when if how does the law you know start to get factured into the facted into these systems like as a mechanism designer how do you think about that I think there I think there are things that come up before the law and I've been thinking about this recently a lot um like reputation um and uh and some degrees of trust where yeah I think I think some some things that just if you try to design them as a game that would work in with just totally Anonymous one in a one shot game with totally Anonymous parties it's sort of just impossible like there are um you could

like prove that actually there's no um uh way to make this this system actually work um sometimes you it takes a very small sprinkling of some other uh uh incentives like reputation or or trust or or legal incentives um to just shift it into an entirely different equilibrium and actually and actually there is a there is a solution and so I I do think part of what blockchains are is you want to leverage some amount of trust and I think we call this often like the social layer right it's it's some amount of of subjective like um you know we have to make some as as a society you know a group what we G have to make one like decision uh every few years about like do we hard forar the blockchain or not something like that uh you make you leverage that small amount

of social coordination and Trust which is very expensive and we don't want to have too much of it to get like a lot more um uh all this like decentralized stuff and I think that's that's just generally true is sometimes um you can get sometimes you do want to fall back ultimately on some kind of some sanction like that like um reputation or trust or again again maybe even maybe legal ones as well um but the The Leverage on it is quite high and I think one one example of that I think is like what it would be really imagine you could you know we end up with like trillions of dollars on ethereum right like what H what would happen if somebody tried to like you know to 51% attack it or you know 67% attack and and like they could like steal everything on ethereum right like

the answer is like probably they they wouldn't get away with it like ultimately you you'd be able to steal something but you would steal all of that and so ethereum doesn't necessarily have to be worth like two-thirds of all the value secured on it or whatever to be for the whole system to work because ultimately like we have these fallback mechanisms and I I see law as one of those it's like it's a it's enough of a deterrent that it shouldn't actually have to be used I think that's a good point I I I would push back just a tiny bit I'd be curious to get your on like the reputation because I don't know I mean there are all these examples in history where that has sort of fallen down and it was like like to just to give you an example of like

organizations that cheat solution not everyone in the organization has the same incentives this happens over and over in finance and banking right where you have like hot shot Trader Hot Shot Trader takes too much risk because the calculation for Hot Shot Trader is hey if I have two like bang up years I could pull you know if you're a Solomon Brothers in late 80s you could pull like 1020 million doll out of it and then I could leave but for Solomon Brothers it's like if you put me in bad enough positions this could bankrupt the whole firm different sets of incentives within an organization so I hear you on the reputation but I've always like push back on push back I fully agree and and that's why I work in crypto it's it's a big part of why I work focused specifically on decentralized Exchange

because I think one one such party that has betrayed their trust often or type of party is is centralized crypto exchanges and so I think I think it's I think it's an incredibly important um thing to to minimize uh trust there but again I think the uh ultimately sometimes you do need you need like a drop of this in some in some place or another ideally at areas where it's not like fatal to the system um necessarily and ideally at areas where can be detected um to get more leverage yeah so maybe to just wind down here and get into introduce your tease the idea of our next episode a little bit but one thing that I was curious about on something like even like Unis swap X which uses leverages a little bit more offchain infrastructure today is there a path to maybe like when you have

something like anoma or something like that or a more general purpose framework and blockchain that supports uh that sort of intense first so to speak like do you sort of see a future where maybe the stuff that offchain today becomes onchain in the future and there is sort of this pattern in history that you bring up centralized exchanges okay well in the beginning a decentralized exchange made absolutely no sense because ethereum didn't exist and we had no infrastructure for that over time you know centralized infrastructure paved the way but now dexes are making Headway against even centralized exchange volumes um so there's sort of this pattern of like okay in the beginning you probably need a centralized organization which is a little bit more um command and control more structured like an executive branch um to to make

some of these decisions but then over time it does tend to decentralize in crypto do you see a path anything that you can share about that do you see that happening for UNIS swap X too I think the most in my view like one one of the most decentralized systems is a market if we can design a market with where people are that's actually competitive then I think like you actually end up with a lot often a lot of good outcomes um and so I I see yeah exactly you sort of want to move from maybe a um a system a system where you need to depend on on more trust um on certain parties to act um to one where yeah exactly like every sort of this house of cards like won't fall down because everybody's actually holding each other together and there's and there's a deficient Market that's

been that's been bootstrapped out of it this next episode uh Dan and I are GNA go a little bit more into intense and also just how cross chain Dexter can work and that'll be a really fun one but uh for now this was a ton of fun day and I will see you uh seal next [Music] week