chainlink networkIC3 and Chainlink Meetup: UnReasonable Design of Stablecoins

BNB price today, BNB to USD live, marketcap and chart If you would like to know where to buy BNB at the current rate, the top cryptocurrency exchanges for trading in BNB stock are currently Binance, OKX, Cardano Vs. Bitcoin, Their Differences - LinkedIn Apr 10, · Cardano vs Bitcoin Energy Usage Since Bitcoin uses the proof of work consensus mechanism, according to annual statistics, its energy consumption is at 126.09 terawatt-hours in a single year. chainlink network IC3 and Chainlink Meetup: UnReasonable Design of Stablecoins

foreign and uh do you prefer people asking live questions during presentation or do you prefer people asking at the end just asking because with zoom webinar we can enable mics now or we can wait till later just based on your preference oh hmm um hard to say i guess uh im open to questions during it but if its like if its like a particularly targeted question i guess okay so uh attendees you may get a notification from me saying your mic is enabled uh so then youll be able to unmute yourself at your leisure cool and then sir i just dropped the um youtube link as well and then let me let me share this out with everyone great thanks and welcome everyone yeah well get started here in just a few minutes let some more people join us uh really excited to have arya and sarah joining us from ic3 today um sharing their new research that theyve been working on around stable coins if you have questions definitely feel free to drop them into the youtube chat and well be able to share those as we go along throughout the presentation um perfect and aria are are you ready ready to go yeah cool uh well thanks a lot everyone for joining us today uh really excited to have ic3 and arya joining to share his new research i really want to thank sarah for helping uh set this up and all the work and research that everyone is doing over at ic3 if you have questions throughout drop them into the youtube chat if youre listening there if youre in the zoom with us uh feel free to share your q a weve also enabled talking so you can just raise your hand and then we can unmute you um without further ado sarah yeah ill let you uh introduce and share a little bit more about ic3 and um everything that aria is working on okay thanks kenan so hey everyone my name is sarah allen i am the community manager for the initiative for cryptocurrencies and contracts uh ic3 ic3 is an academic blockchain research initiative based at cornell tech with researchers at nine global campuses and chain link is an ic3 partner so thanks today to chain link for hosting this webinar aria clogisment is an applied math phd student at cornell his research brings together economics and computer science to study the design of d5 protocols and economic networks and hes also working on the gyroscope stablecoin and uriah i will turn it over to thanks for the intro sarah yeah im mariah claudesmont im a phd student at cornell and ill run you guys through a little bit of my phd research on what i call the unreasonable design of stable coins reasonable and unreasonable so the past year or so has been really a year of defy defy assets under management has grown basically like 50x or even more in that time and the system of d5 protocols has also become much bigger but also much more complex so its been a year of defy but its also been a year of d5 crisis so in particular about a year ago back in march 2020 uh die there was a large earth crash and uh this led to sort of deleveraging spirals something that well talk about a little a little bit more later as part of my research in the dye system which led to increased volatility and kind of like a short squeeze effect uh on the dye stable coin and it had some side effects then that led to uh sort of like maker auctions being liquidated for zero die and sort of like minor extractable value occurring around these liquidation events theres also been a number of other things over the past year including uh sort of like money loopholes uh ability to steal money from different protocols uh actual attacks on protocols that sort of uh made use of of oracle quirks uh actual other attacks on uh on minor extractive extractable value on stable coins and then uh recently this past winter uh the introduction of some of these new algorithmic stable coins that uh then proceeded to crash such as basis cash empty set dollar there were also other uh stable coin crashes such as around origin dollar and a hack in that system and very recently uh sort of struggled launch of uh the fey stablecoin so that kind of motivates uh my research in this area on stablecoins um and here were thinking of a stablecoin as uh basically a cryptocurrency where theres this added economic structure and the aim of this added structure is to stabilize the price for purchasing power uh and some of the issues we have in this area that are sort of motivated by these uh these past examples are that we lack really good risk-based models that span this design space and help us understand the different trade-offs when were designing these systems and thats kind of where this work comes aims to fill in kind of filling the gap and sort of seeding where the the right sort of questions to ask and models to start using in future stablecoin research and so in this talk im going to go through a little bit of decomposition of how we think about the design space of stablecoins uh then sort of motivate two primary fundamental design questions around stable coins that were going to start to answer and then ants start to answer those using uh models of price dynamics of these systems and then models of what we call governance extractable value and gev or and minor extractable value mev in these systems to sort of help understand are these systems secure and this work is based on these four papers which are all available on archive so if you really like the talk do look these up and then this is with co-authors including dominic harts and louis gudgeon sam werner and daniel perez at imperial college london and then also my advisor andrea minka at cornell so lets start by decomposing the design space of running you through how i like to think of stable coins so one of the first uh distinctions to make i think is between stable coins that are based on custodial mechanisms and stable coins that are non-custodial so this is an example of uh on the custodial side we have tether and on the non-custodial side we have die and on the custodial side uh we have risks around sort of counterparty credit risk uh censorship risk and traditional sort of financial risks and how these uh banking sort of systems are set up and the good thing is that these are well understood uh but the bad thing is that uh this is kind of the risk we were trying to minimize by entering the cryptocurrency space in the first place and then on the non-custodial side we have sort of new risks and attacks that we have to first map out what are the risks and then uh help to solve in terms of like how should we be designing to mitigate these risks and as well see there are risks around sort of deleveraging crises and these uh sort of stable coins uh risks around sort of price feed and governance manipulations and more generally sort of like minor extractable value on the on the base chain because miners have the ability to reorder transactions to their benefit and then also the risks around sort of like smart contract bugs and how these things are actually implemented and whether an attacker could manipulate that uh those implementations and the problem here is that these are not as well understood and thats kind of where this uh this research leads so the custodial side we can kind of map out uh along the lines of sort of traditional finance and traditional financial models i wont really focus on this in this talk but ill give just a basic overview theres more information in our papers but basically you can think of these as uh one type being basically a fully reserved fund and this is uh basically like usdc busd that would fit in here and it kind of acts like like an etf sort of model as in you can mint and redeem uh essentially shares or tokens of these state on these coins uh in return for the underlying collateral through the issuer then there are other types that are kind of more parallel to like traditional banks or sort of like money market funds and these are sort of like fractional reserve funds that are kind of like fully collateralized but uh sort of like fractionally liquidly liquid collateralized and then you can similarly like interact with these custodians to create and redeem the supply but they have sort of also added threshold risks akin to like uh how banks work and then on the other side you also have sort of like central banks that could come out with central bank digital currencies um but right uh were going to sort of focus on the other side through the rest of this talk on the non-custodial uh design of stable coins and so here i like to break it down first as a first distinction in terms of like where are these stable coins getting their value what sort of collateral are backing these systems and in one case we have like in dye exogenous collateral so something like eth where theres sort of uh enough scope to this collateral that its uh it moves relatively independently of the stable coin system itself and these have uh what well see as market deleveraging risks uh through the rest of this talk theres also a type of endogenous collateral and this is where theres been essentially a new collateral asset thats been created uh for the specific reason of being collateral in these systems but that also means that its very uh the value of this uh this collateral is very tied to sort of like expectations in this particular system and thats why its an endogenous and well see that it has sort of like amplified feedback effects as opposed to the exogenous collateral case and then theres also a type of what we call implicit collateral this is kind of the basis to type stablecoin design and well see that this essentially works by trying to set up uh that theres like an incentive to absorb risk in these systems during a crisis but theres no sort of like forward-looking obligation from this point in time it really comes down to speculators uh becoming choosing to absorb the risk in the actual uh crisis and well go over a little bit of this in more detail in a coming slide as well so right this is the area were going to focus on in the rest of this talk so to sort of motivate how these systems work uh lets go through kind of like how a cdo type structure sort of works and this is how most stable coins today actually work so in a cdo design we have a portfolio of underlying assets and we split this portfolio into two tranches uh one tranche is a junior tranche which is uh sort of more risky and then theres a senior tranche and the idea is that this is a less risky and then if theres a loss thats incurred this is first born by the junior tranche and the senior tranche is protected and the idea with the with stable coins then is that theres a risk absorber which kind of takes the place of this uh junior tranche and a stablecoin holder uh the stablecoin holders are sort of fulfilling this senior tranche with trying to be protected if there are downfalls and then theres an added mechanism where if there is enough sort of like of a crisis event uh theres a deleveraging process that uh dynamically adjusts the supply so that uh its kind of like a continual system so that essentially the cdo structure can continue into the future uh just in like a downsized form the space of stable coins is a bit more expansive than just this though um and so ill try to like motivate some several different areas here weve already touched on this like idea of primary value or sort of the collateral backing the system uh and right this is sort of spans from exogenous collateral to endogenous collateral and what we term implicit collateral and then theres another sort of function in the system of uh who actually is absorbing the risk through this sort of like primary value and in one case we have sort of like uh equity risk absorption where theres some sort of like uh uniform equity token in the system and holders of this token are basically all absorbing uh supposed to absorb the risk in the system as opposed to there are other systems where there are agents who are making their own sort of custom decisions about uh about how to absorb risk in these systems um and then there are newer sorts of systems where the protocol is kind of designed to absorb risk itself as kind of like a reserve or insurance based sort of mechanism so lets uh sort of like go through a little bit of how this how this works the main type that i was alluding to earlier is this leveraged based sort of design this is like the cdo model that we just stepped through and here we can have it can be based on exogenous or endogenous collateral um and seniority shares is a term is kind of thrown around too to describe some of these systems and its kind of a sub case of this leverage based system so essentially here we have a market engineered shares theres like a market cap of endogenous equity shares and the idea is that these equity shares absorb uh are meant to absorb the volatility in the system so this is looking back at uh at the diagram here where you have sort of like endogenous collateral but also equity risk absorption in that case and then theres another type uh the basis sort of design or what we term implicit collateral and in here speculators are meant to maintain the peg by betting on future supply expansion uh something thats akin to like betting on leverage and this implicit collateral kind of like backing the system during a crisis and the idea here is that theres no pre-committed collateral as there would be in these leverage-based systems before but instead essentially like the speculators are coming in and either deciding its profitable or not profitable to reduce the supply voluntarily um and in doing so they kind of have to like in judging if its profitable they essentially have to bet that the spy is going to expand beyond this like pre-crisis sort of level uh or else the system kind of like uh this this uh this basis type design uh doesnt really work and thats kind of also like what weve been experiencing uh in the past winter with the the rollout of these these basis type designs and then theres another type which is uh basically reserve backed and in here the idea is that the protocol maintains its own balance sheet and uses this balance sheet uh to market make around the peg um and there are some key questions around here around like what actually makes up this balance sheet and sort of like various attacks to whether you can exhaust that balance sheet and sort of break the peg uh but its another type of design thats also being rolled out these days there are also various types of meta-stable coins that are basically combinations of all of these different uh sort of lower level stable coins so moving on to sort of the rest of the uh the types of functions in these systems weve just covered sort of different types of risk absorption but this is also intimately connected with uh with an issuance function and here uh issuance can be agent-based uh which it kind of is in die essentially if youre opening a vault and die you are deciding sort of how much stable coins to issue from that uh from that vault and so deciding uh in aggregate what should the uh the supply of dye be and theres also different sorts of algorithmic issues uh which is where these algorithmic sort of designs come in and theres also this deleveraging process that we alluded to earlier which is sort of behind the scenes in all of these different designs which is also essentially algorithmic and then there are stable coin holders who are sticking seeking stability uh through holding this asset and then sort of like behind the scenes theres also like governance that needs to update the system and similarly price feeds that need to say what these assets are actually worth and then further behind the scenes there are miners who always have the ability to reorder transactions to their benefit so we can sort of map out how these different stable coins today uh fall along several different axes so this this is showing like how they fall within three particularly particular dimensions one being what type of risk absorber is there another type what what is where is the primary value coming from and then in the sort of like third dimension uh how is issuance happening whether its agent-based or algorithmic um ill sort of put out the slides later if you want to peek into this in more detail uh but just very briefly uh sort of sketching out some of the issues thats come up here in black thursday we saw that there were sort of deleveraging crises in in some of these designs like dye weve also seen that other sort of implicit collateral designs uh fall apart that the market for these fall apart uh heres an example of new bits and recently this past winter sort of both basis cash and empty set dollar and these senior shares types designs also have problems and so whenever were making new designs we should have this in mind as as essentially like a case study and so steam dollars ran into quite a lot of trouble back in 2019 2020 and then very recently the rollout of fey uh led to several sort of liquidity problems uh in how the mark how the protocol is actually market making uh using its balance sheet i also then briefly want to sketch out kind of like how do we think about different types of algorithmic stable coins and sort of approaching this whenever a new user is coming into the system essentially they are paying if if the systems working that new user is paying a dollar for each newly minted stablecoin and one key question here is where does that dollar actually go and we sort of map out a spectrum upon which it can go uh sort of on one extreme it goes to sort of the pockets of uh of stakeholders in the system uh and on another extreme it goes completely to the uh the protocols balance sheet and then theres a room for in between where it can go partially to both and then the next sort of key question is what happens in a crisis in these different sort of design choices and in the case in in one extreme where its going sort of like to to stakeholders that means that the system is not retaining extra value essentially to handle crises uh and its solely relying on sort of speculators to jump in in these situations and if they dont find it profitable to jump in in those situations essentially if a sort of future demand growth becomes less credible then the system doesnt work and then somewhere in between where its like partially going to the reserve partially going to stakeholders in this case the reserve is kind of smaller than it could have been but its there it means its potentially less stabilizing but it also means its prone to sort of like bankrupt style risks and this needs to be like a critical concern when designing these systems and the other case is when 100 of it goes to the reserve in this case its basically strictly stronger as if theres more value to work with in a crisis um and theres theres still worries about it uh but uh but its strictly stronger than uh than some of these other designs and were sort of seeing this with the roll out of algorithmic stable coin so far so weve seen sort of empty set dollar basis cash and sort of similar designs like dynamics at dollar all essentially failing to maintain their peg and so sort of the extreme on this side uh definitely seems to be not working weve also seen similar things uh historically with new bits which is kind of in between but further to the side of all the way to the uh to the left and then there are some several ongoing experiments around sort of like fractional reserve designs and then there are also some experiments here on the the other side uh and so were going to keep like kind of a keen eye on these going forward um but there are also like several other dimensions then theres just this one dimension to think about um so here for instance theres sort of like a its very important to think about the composition of what are the actual assets in these reserves what are their risks and basically are those assets valuable when you need them in a crisis and then a follow-up question of how does the protocol actually maintain liquidity in the system and how liquid uh how liquid is that of a system does that lead to and this sort of like contributed to some of the uh the problems in the fae launch and so these are sort of like three primary dimensions i like to think about in terms of analyzing different uh algorithmic stable coins another uh sort of area of difference i want to point out before leading into sort of the main models and results is sort of like parallels and differences between uh the governance sort of the setups of these systems and sort of like uh how do how do these compare to traditional money and so here im going to compare sort of like the die system to the traditional money system but similar other designs can also be compared sort of on the top level we have uh on the die side maker governance and this is kind of parallel to how the central bank is acting in traditional money but a key difference here is that we can usually assume that a proper central bank is a stability seeking kind of for its own right its targeting economic stability in a sense whereas maker governance is really designed to be profit optimizing and so we have to take that into account when were sort of thinking about these systems then a level lower uh on the die side we have uh dive vaults or cdps and this is kind of parallel to how commercial banks operate in the traditional money system essentially vaults are here to absorb risk and sort of determine the issuance of an endogenous stable asset or the die stable coin itself this is kind of parallel to what commercial banks are essentially doing by making loans um except that on the commercial bank side we can basically assume that the asset thats being generated is stable uh sort of as backed by the central bank whereas in the vault case we really have to sort of be worried about that this is an endogenous asset and its working its either stable or not stable based on how we set up the system and then sort of at the bottom level we have die holders which are kind of uh parallel to depositors uh at the actual commercial banks and traditional money and as were sort of seeing here because of these sort of key differences uh we sort of need nude models to think about uh about the non-custodial setting and so this sort of motivates two fundamental questions that are then uh sort of were trying to answer with the rest of our our research in this area so the first question is around incentive security and so basically this is asking is there mutually profitable continued participation across all of the parties in these systems or basically is it is it profitable for everyone for everybody to participate with these uh these terms and this has to take into account sort of like the attack potential of these systems the sort of like equilibrium interest rates that uh they need to fund these systems um and two key things were worried about is sort of what we call governance extractable value because there are governance systems uh as part of these systems that are sort of updating either parameters or basically even changing the systems over time upgrading the contracts directly and they could do these in sort of like uh best interests of the protocol and thats the aim but they also have sort of like other avenues that may be more profitable and sort of the design of these systems has to uh understand that and take that into account and then theres similar problems around the lines of a minor extractable value where basically uh miners are if theyre even more abstract sort of like governance process behind the scenes of the base layer and deciding the ordering of transactions in these systems and then the second question is basically after we know that theres participation in these systems we can then ask are uh are the incentives actually leading to stable outcomes and this is what we call economic stability in these systems and so to address these uh these two questions were developing three different types of models one around understanding price dynamics of these systems basically how do the issuance incentives either lead to stability or instability and then two types of models to help understand governance extractable value and minor extractable value and so were going to touch on the price dynamic models first and then also touch a little bit on the other ones in a little bit sort of set up the landscape with these price dynamic models uh in the traditional financial literature we usually start from a setting where we assume that theres like a stable asset and this asset is borrowed against some collateral say and then there may be feedback effects in the system but its sort of on the collateral asset liquidity side um as opposed to on the actual stable asset side and this is one of the key areas where non-custodial stable coins are different because this stable asset essentially has an endogenous price and participation and this needs to be factored into these models and so we sort of need expanded models that help us understand this and so as part of our work weve been setting up sort of stochastic models to understand these sort of endogenous stablecoin prices as part of two papers and this leads to sort of this effect we term as deleveraging spirals which are essentially short squeeze like effects that amplify sort of collateral drawdown in liquidity crises and this also leads to sort of like regions we can describe uh as stable and other regions we can just describe as unstable for these stable coins so to illustrate a little bit of what we mean by these models although i wont go through the fine details um we set up sort of a system with two different types of agents one is a stable coin holder and these are people who are seeking stability and have also like an imperfectly elastic demand for these stable coins and another type of agent thats a speculator who is essentially deciding the supply behind these stable coins uh secured by a collateral position and so this is essentially modeling the sort of uh leveraged base stable coins were describing earlier and then there are two types of assets in this system theres the the collateral asset for eth itself which is a risky asset which were assuming has an exogenous price although you could start to change these systems to build in an endogenous price as well and sort of like map out what happens there and then the other asset is the stable coin with an endogenous price thats sort of determined through market interactions between these uh different types of agents and this stable coin is over collateralized beneath but again its uh its actual stability depends on this stablecoin market thats clearing sort of demand and supply so lets illustrate a little bit about like what the speculator uh how theyre making decisions in this model so they have a collateral constraint and the idea here is that the protocol is basically enforcing a degree of over-collateralization uh within the protocol itself and so this is sort of set up by this equation but basically the idea is that uh the amount of eth that the the speculator locks into the system the value of that eth has to be greater than or equal to some collateral factor times how many stable coins theyve actually generated through the system and then theyre making a decision uh thats deciding how much they want to change the stablecoin supply to sort of maximize their expected returns into the future uh taking into account this constraint and this is what were calling sort of like the honest behavior because if the this is how speculators are intended to to behave in these systems and so we set up an optimization problem that essentially does this the important thing is that theres a value function that sort of defines uh how profitable this this is into the future for these uh speculators and this has to take into account this liquidation effect where basically uh either the speculator is like self voluntarily liquidating to like preemptively handle uh sort of like their leverage or the protocol can also like directly liquidate their positions if it breaches this threshold and that comes with extra costs and market effects and this the speculator has to take this into account in their decision making there are some other assumptions that also go into this to make it more attractible and well touch on one of these later sort of in the discussion and so then some this this model leads to some very interesting results uh one of one of them is sort of in the area of uh sometimes we can really think of these systems as very stable actually um and so just kind of like at a high level describe these results the first result is that uh these are there theres sort of we can bound the probability of large deviations in these systems within within a certain region and this is sort of describing the uh the stable region of these stable points uh and a technical idea sort of how we get there is using dubs inequality but you dont have to focus on that um you can see the paper sort of for more details the second result then is sort of bounding uh in addition to sort of just like bounding large price deviations we can bound actually this like idea of a large quadratic variation or essentially describing like how variable the process is how much it moves around uh in this similar regime and so both of these sort of mean that uh that the stable coin is has quite contained price ranges uh within this like stable region and the technical idea behind uh the second result is using burkholders inequality uh then sort of on the flip side theres also uh another sort of distinct region where we can think of this as being an unstable system and this is pretty important to sort of like see the difference here so in the first result here in this different region uh we characterize how the stable coin experiences essentially a short squeeze or what we call a deleveraging spiral or very formally uh the stablecoin has what we call submartingale prices that are sort of like increasing or expected to increase and sort of the idea here i can kind of illustrate it quickly um is that in equilibrium the stable coin sort of uh has demand equaling supply and then theres uh theres this like excess of collateral sort of backing the supply and now if there is a liquidation essentially that means that some of the collateral is being used to reduce the supply essentially by buying back on the on the open market but this means that demand is now out of sync with supply which means then that in an inelastic market for these stable coins the price of the stable coin has to increase uh and then that ends up sort of resetting demand and supply but it also means that if theres a second round of liquidations because theres a higher price to the stable coin theres a short squeeze type effect uh more collateral actually has to be used uh in this second round of liquidations uh to do the same sort of decrease to the supply which in turn makes the price actually increase further so this is this short squeeze type effect um which brings the system back into balance but sort of like this can keep happening and it means that theres faster collateral drawdown increasing prices but also like even though its increasing prices its a drastically riskier system because of this uh sort of dynamic with collateral we then sort of like can characterize also like that even though theres this like short squeeze effect um this is actually distinctly uh higher variance thats happening in this setting as well compared to the stable setting so in in one result we have sort of like in this different regime uh this uh in this different regime theres we can characterize a variance approximation essentially telling like how much the uh the the price is wiggling around and that this approximation increases by uh by quite a large amount uh as the ether return uh sort of shock scales or also as like the initial collateralization decreases um so this means then that like this variance approximation the variance is much higher in this unstable region as opposed to the stable region which helps us solidify that yes this is actually less stable and the technical idea behind this is using an implicit function theorem and then we can go even further uh as instead of using the variance approximation itself we can actually use sort of like a forward-looking idea of variance and show that indeed in the stable and unstable regimes this is a good interpretation because forward-looking variation variance is much higher in in the unstable regime as opposed to the stable regime and this builds on some technical ideas around inequalities on variances of convex functions of random variables we also can back this up with uh with some simulations and in a simpler agent-based model that we constructed in a different paper and essentially this histogram is kind of showing this in certain regions we can uh like at this orange uh sort of region the uh the returns of the stable corn are much more constrained than uh in this unstable sort of region or blue region where its much greater uh variation in in the uh the returns that are realized and then this is essentially what happened on black thursday uh in in march 2020 when we had like a very big heath price crash and it led to essentially this short squeeze like effect that you can see and die on this uh this plot in the right hand side which is showing you sort of like liquidation uh the prices of uh of dex trades as liquidations are happening uh over over several days here and this sort of a short squeeze effect persisted for quite a while um and uh actually several changes to the maker system had to be built in to sort of like bring it back toward peg um and this sort of leads to some interesting complications which we can also understand in the context of this model so one is interesting point is that uh one of the assumptions we had to make to get these nice results uh nice analytic results is that uh there are essentially like sub martingale prices uh in the collateral asset which is meaning that uh in simpler terms that speculators are thinking that this uh this collateral asset theyre expecting it to increase in price or at least stay the same um and intuitively this makes sense uh because um no ones no ones going to participate in these systems if they dont think theyre going to make a profit and they have to go leverage long in these systems to uh to participate so they essentially have to bet on this happening and this means that theres theres no results about a stable regime if this uh this sort of assumption falls apart which realistically it does fall apart sometimes uh and then an interesting sort of side effect of this is kind of a seeming contradiction in the design of these systems so the idea here is that our goal with a non-custodial or decentralized stablecoin is that we want something thats fully decentralized but we can only really fully stabilize it if we have like many uncorrelated assets that are backing it um which today realistically those uncorrelated assets are currently custodial and so we cant really do this without bringing in some sort of custodial asset or sort of like different designs into the mix and theres been sort of several solutions to try to address this so ill give like a brief overview of this and sort of like ideas we have toward addressing it maker since black thursday has essentially tethered dye to usdc which imports custodial risks from usdc and the idea here is that they set up this peg stability module or psm where this this psm is maintaining direct exchange ability and in the pro with usdc and in the process essentially building up like a usdc uh sort of in these crisis situations and then there are other approaches such as what rye is doing which is instituting negative rates during crisis um but this leads to sort of a a question of like do do people in equilibrium participate in these systems given that they could experience negative rates and this is sort of like also depend on how long those negative rates have to happen and then there are other sort of designs around like liquidity buffers which are essentially dedicated liquidity pools for for sort of helping out in crises um and so we suggested a sort of form of uh vault insurance and maker and liquidity today is also doing something rather similar in some of their uh something like liquidity liquidity buffer sort of handle uh liquidations as they come up and then theres sort of a generalization of what maker was doing with the psm which is more general reserve backed mechanisms to sort of like help maintain exchangeability with not necessarily a dollars worth of usdc but a dollars worth of assets um and this is like a generalization of what baker is doing so so far weve sort of like addressed the second question around economic stability um and well sort of shift tack now to sort of think about the first question of of incentive security so again the idea here was sort of like um how do we understand governance extractable value minor extractable value in these systems and given sort of the risks and sort of like different attacks that can happen do people choose to participate in this in these systems compared to alternatives and to sort of motivate the thinking about these questions um there are several events over the last year such as uh when there was uh pointed out sort of like a flash attack that was possible in maker where governance if they wanted to could have chosen to just uh upgrade the system and essentially uh be able to mint infinite dye and sort of like exhaust the uh the collateral in the system and then also many instances of essentially rug pulls in these systems so we think of governance extractable value as a more general case of of just an explicit run poll where where if youre rug pulling the system this is governance just directly choosing i want to steal the collateral in the system but they could also do this indirectly and its sort of like hard to pick out what is the difference between like a rug pull and uh sort of just like natural outcomes of the system and this is where really thinking about governance extractable value is important so we have then two uh two types of models that that help us think about these things and the first is helping to think about governance extractable value using what we call capital structure models and these models were originally a type of model to describe ipo incentives in traditional finance and were essentially taking those models and expanding them to think about uh stablecoin incentives and attacks in these new types of non-custodial systems and so here to set up to set it up we have three types of assets a collateral asset uh like like eth before a stablecoin asset and then a new governance token um which is uh sort of saying like if you own this governance token you have a voting power in changing the system through through the governance process and then there are three types of agents theres as before a risk absorber or sort of like the vault in a hidden maker a stable coin holder and then an outside governance holder who is actually like determining the governance process and holding these these governance tokens and then uh you can consider some further variations of this too and we do that in the paper but you can you can check that for for full details so we set up first a very simple case of this where there are no attack vectors and so this is essentially just like directly adapting these capital structure models to this setting um and i wont go through the details of the particular setup but the general idea is that we have sort of a game theory problem thats set up where you have a governance optimization problem and in this problem governance is deciding an interest rate to sort of maximize their their revenue but taking into account how the vault uh is going to respond to this uh to this interest rate decision by deciding the issuance in the system and then the vault on the flip side is uh is is deciding the issuance in the system to maximize their expected returns um subject to several constraints so the original collateral constraint if were considering the maker type system a participation constraint uh essentially saying like is it profitable enough to participate in this compared to some alternative and then the stable coin market pricing sort of like how that market works out itself as another constraint and then we sort of build from this very simple model into adding various governance attacks uh and modeling how how this changes the system and the idea here is that in this problem too we might have a fraction of governors who can steal a fraction of collateral at some expense of their like share of the governance tokens essentially like if governance tokens value went to zero plus some outside cost to attack alpha which is kind of a parameter describing how much centralized recourse we have sort of like through traditional legal means so to say and here now that the problems are changing a little bit because in the governance problem theyre deciding an interest rate but now theyre also deciding do they want to attack the system all right and then the vault problem is changing by uh sort of factoring in this uh this attack possibility and we also set up like even more complicated ones where theres more complicated uh sort of collusion attacks that could happen but essentially like agents could collude to restrict the exit of other agents um and indirectly steal collateral and then in this case uh the other agents in the system may essentially want to bid up the price of the governance token or essentially issue like bribes to governors to make sure that this doesnt happen that they dont choose to attack the system uh and so the problem changes in other sort of like subtle ways and adds in sort of like a stable coin holder uh optimization problem as well so it becomes much more complicated but there are several sort of interesting takeaways coming out of this uh which ill ill go over a little bit more detail the idea is that the governance token fundamental value is kind of related to the the future discounted fees that are coming into the system that they might hope to uh to accrue and this is sort of the idea of like the honest governance fees uh the value to honest governance and if this is small relative to sort of like the actual value or collateral in the system its essentially guaranteed then that you need a high alpha parameter to actually have a secure system and this leads to what we call as a price of anarchy which is kind of this extra cost to secure a decentralized system compared to a centralized system uh which is reflected in this like high value of alpha so we based on this conjecture that a lot of very flexible designs that are fully decentralized are actually uh impossible to realize as in like nobody chooses to participate in the long run and the the idea behind this is comes with an analogy so basically if theres a bank and its unsecured equity is less than worth less than two times the actual like deposits at the bank it doesnt really make sense for depositors to participate in that system and basically up to this point a lot of defy systems have been essentially arguing that uh that their governance token ought to be worth this because of this attack potential but that also like has a flip side as were saying that uh whos gonna choose to participate in such systems we then also introduce a potential solution to this and were uh sort of further investigating this uh today which we call optimistic approval and the idea here is that were essentially giving users the option to veto governance changes um uh if theres like a malicious change thats proposed and the idea is that this helps to sort of align the vision of governance with the actual vision of the actual users who are using stable coins in these systems and so this is an area of sort of like uh uh ongoing research but were quite optimistic that this can help solve some of these challenges this then leads to sort of like minor extractable values sort of concerns and i see were sort of coming up on time so ill sort of uh sort of cruise through this a little bit faster um but a basic idea is that uh there are several components that make attacking a stablecoin uh kind of different from attacking traditional currencies uh and its kind of about manipulating the interaction of the agents involved in these systems as opposed to sort of like attacking the willingness of a central bank to sort of maintain whatever target it has in mind and the primitives were sort of developing previously in our research kind of leads to like these new types of attacks so the deleveraging spirals lead to sort of like arbitrage-like trades around liquidations and sort of more generally liquidations uh basically automate these arbitrage opportunities and then uh sort of behind the scenes we have to keep in mind that miners can censor and reorder transactions and as well see that can sort of lead to extraction of value and sort of attacks to these systems so we sketch out a few different attacks uh that uh that could happen here um ill leave the first one sort of like more to understanding in the paper and sort of like sketch out what i mean about this like second attack uh because there are some visuals here but basically in one of these attacks you could consider that uh the attacker sort of knows the past history uh maybe the past several hours and in this history theres been a large uh price decline of the collateral assets say and if they want to and then this has led to sort of liquidation events that the the attacker can see and say the attacker thinks that they can they can that it might be profitable for them to capture these liquidation events they can essentially try to make a new timeline uh by forking the blockchain uh forking the last several hours essentially a box and they directly inherit the oracle price feed because those transactions are already signed and then in this alternative timeline they can inherit the current liquidations but they can actually do worse they can actually add liquidation events by uh changing the order of transactions or censoring sort of transactions and one concrete idea here is that like if you are a vault in maker and you see a liquidation sort of coming up you want to top up your position or sort of like scale down your position to sort of avoid that liquidation but if youre a miner you can just censor that transaction or include it after youve actually realized the liquidation and its in your interest to do that and so uh you can actually cause more liquidations if youre sort of forking uh the blockchain to try to do that and we actually see sort of like variations on this that actually occurred on black thursday um and this was sort of coming from uh from the from uh liquidation auctions that were able to be carried out at zero die prices and sort of after the fact looking through forensic investigation of the of what happened on that day it turned out that this was uh at least partially the result of mempool manipulation uh causing these uh causing other orders to really like not be able to get in and the attackers to be able to liquidate prices at very low prices liquidate uh the uh the auctions are very low prices and so this is actually like a realistic concern as well so that we uh we also set up these like forking models that like will help us to sort of understand the incentives in these systems um but i think ill sort of leave this to uh to the paper to sort of like fully explain but very generally we sort of like try to do this in a tractable way by considering two separate models one for the base blockchain model where the the miners are deciding whether or not to fork and then an application layer model thats basically building in uh our previous sorts of models and then encoding a specific interaction between these models thats sort of saying like uh theres this sort of like success probability of an mav bribe which then sort of like influences the level of mev which then feeds back into influencing kind of whats happening in these different models and we think this can lead to a very tractable way of of understanding these incentives but in general its quite quite difficult to so that brings us to the end of uh the end of this talk the papers are again available in archives if youve found this interesting and basically weve sort of in the course of this research seeded uh different stablecoin design questions and models and have a couple sort of main takeaways that uh that are worth remembering and one of these takeaways is that uh primary stable coins today are primarily uh leverage-based mechanisms like dye and these mechanisms weve seen uh they need sort of add-on mechanisms to help combat deleveraging spirals and we talked a little bit about this and a few ideas toward this but they uh they can involve sort of importing custodial risks which need to then be carefully controlled or sort of like importing uh participation in different participation incentives in terms of of including negative rates um and itll sort of remain to be seen like how these different things work out and then as a second point sort of there are amplified risks in these endogenous and implicit collateral type stable coins and weve seen these sort of like work out in practice uh in the past winter with the with the failure of several of these basis type stable coins and then as a last point these uh these concerns about like governance and minor extractable value uh are quite critical to sort of like the long term incentive security of these systems and we sort of sketch out at least the first models of how to start addressing these and the right things to be starting to think about uh in those regards and then this has led to what we see as a design gap that uh around the around sort of like robust reserve backed stable coins that are sort of designed with liquidity in mind and also have sort of like robust governance processes that help to maintain incentives and this has led us to sort of design sort of our next project which is the gyroscope stablecoin so im not going to focus on that in this presentation but if its interesting please do check that out and then i think we have a few minutes for for questions so happy to open it up to that thanks sarah so if anyone has questions you can either raise your hand or unmute yourself um i can start off with one so you mentioned that a lot of these protocols have updated after black thursday so with those in mind do you have a sense of what the perfect storm i guess youd call it for stable coin instability would be at this point yeah thats a great question um it depends on which design path they chose for updating the designs but for instance if youre considering how maker chose to update their design which is in introducing this peg stability module or exchangeability with usdc this means that theyre really tethered to sort of usdc risks and if theres like a regulatory clampdown on usdc that means that uh youre back to kind of like in in the best case youre back to the old maker design but in the worst case makers the maker system itself is uh sort of under collateralized if usdc fails so thats kind of like a primary concern with how you how how maker is doing it but then uh there are other attempts at it as well so like like rye is introducing negative rates um and then it comes down to how our user is going to respond to that and rai has kind of just come out over the last couple months so theres not like a clear question to to that answer but itll be very interesting to see how it unfolds over the next next several months and how stable it can be uh sure and then i see a question in the chat about the dye attack uh so about whether it could have been presented or how ah i dont see that in the chat but can you read it out fully uh so its just how could the die attack have been prevented okay okay um the die attack im taking to mean kind of this uh this minor extractable value idea or mempool manipulation um and that is in general like hard to completely mitigate but it comes down to really making uh sort of like can you make auction markets i think it comes out like how do you design these auction markets and how do you uh how do you make it how do you increase the costs to essentially like uh censoring transactions in the blockchain affecting the mempool over a long period of time but if you design these auctions for instance in a way where theres uh one way might be sort of like introducing a reserve price as in like the minute a starting bid um thats not too low or sort of like uh grouping many auctions together so that there arent sort of like thousands of different smaller auctions that uh that uh the liquidators have to be looking at at the same time uh then you can make these auction markets much more efficient potentially um and if its stretched over long sorts of time periods then it becomes much more costly to sort of like manipulate the underlying blockchain or mempool but its not necessarily that its it doesnt make it impossible to do these attacks though you can only sort of like increase the costs and theres also a lot of works on like uh flashbots toward like uh toward toward making the the mev sort of angle of these more efficient and that could have like interesting complications here as well i dont think theyve actually gotten to uh automating maker liquidation events yet but it would be quite interesting to see how they handle that also and then i see a couple other questions in the chat um jim or frank would either be like to unmute yourself or ask your question sure i can i can uh mute um uh thanks for the talk uh what do you think about the difference between like the term stable coin versus pegged coin i.e like rye is designated around like it uses like a pid controller to kind of stable around this arbitrary value which they started at uh pi 3.14 and then depending on the protocol it it deviates or fluctuates while dye is pegged uh to one dollar like a dollar system what do you think about the difference in like the terminology yeah thats an interesting question i so when i think of rye i know they have this narrative about it being like not pegged to the dollar but it kind of i dont know i kind of see it as pegged to the dollar its just like a more its less short term peg doesnt you have this pid controller but its kind of like a over adapted in some sense and so youre theres much more variation around the peg but it does build in this discrete target of uh well its not a dollar its 3.14 instead but this discrete target of like what the uh sort of long-term uh value should be and um then it comes down to just like what are the the actual mechanics to like uh moving around the peg and i guess in some if you want to distinguish its a good just distinguish me a thing to distinguish i guess between like this one has more variation just because thats how the design is made and this other one is meant to be quite tight um and then itll come down to like how how people what what people want to use as a product i suppose and my guess is that like awry will be evolving over time to sort of like adapt what is the right form of the pid controller although im quite optimistic on like pid controllers overall in terms of being useful um but i think the fundamental question then is what is the target not just like how much very variation is there around the target in terms of what what the trading price is but what is the actual target of these systems and right now it seems like everybody wants the target to be essentially dollars if not like being one dollar being in the arbitrary like 3.14 instead but like touching on the value moving around the value of a dollar but in the long term you may not want to be actually targeting a dollar you may want kind of like a a crypto native uh currency where its targeting kind of like metrics about the uh the cryptocurrency ecosystem itself kind of like targeting stability in the crypto economic sense as opposed to just like pushing that off to like uh the u.s federal reserve to determine what economic stability means and i would interpret that as changing then what are the targets of these systems is that does that make sense is that like how you might think of it too yeah yeah thanks uh thanks thanks for that yeah um yeah to totally make sense appreciate it yeah im kind of out of time but i i just wanted to get your input on um what youve how you see the current market now um which coins if any are best positioned to survive the current bull market um and i promise i wont go out and buy or sell whatever you uh if you get specific but really more interested in generally what whats your take on the current landscape yeah a great question uh preface it by not giving investment advice obviously um i am quite well i find most interesting these non-custodial designs um but that said like if you are from a good sort of like a jurisdiction and you have like good sort of like access to banking services um i think i think usdc sort of like staple coins custodial stable coins like can make sense you should just be conscious of sort of like the risks that youre taking on but it allows you to access some of these uh these on-chain uh lending markets which are quite profitable um but you have to take into account kind of like the right the risks these custodial risks and regulatory sort of risks um but kind of one idea of like the worst case scenario for for like usdc would be like the government comes in and says this needs to be unwound and theres like a process to unwinding it but as long as you are like sort of well banked youll be able to unwind your position um thats not guaranteed to happen but thats like kind of what i have in the back of my mind but it doesnt it also means that theres like complications then for for people who are actually under banked which is kind of like the main target of cryptocurrencies and uh and non-custodial stable coins in the first place and so its kind of unfortunate that they would be left hanging potentially in that situation and so thats kind of why im most interested in like non-custodial stable coins just because like the real promise of crypto again was like removing minimizing trust assumptions and counterparty risks and if we can find sort of robust ways to do that i find that quite exciting and i to be honest i really like kind of like these leverage based mechanisms um i its just become apparent that like they cant work in isolation and so like i really like die uh thats why ive been doing this research to try to make it stronger basically and um they theyve been moving toward like uh sort of integrating usdc but i also kind of wonder like are there alternative ways that we could have uh could have solved the same sort of deleveraging problems um but still maintaining sort of like this uh this independence from the traditional financial system and thats led to uh kind of what im aiming to do after the phd so im finishing up the phd right now and the idea is sort of migrating into this this gyroscope project which is motivated by by all of this research that ive been doing and i kind of see the gyroscope mechanism as a way to help combat these deleveraging sort of effects while maintaining full decentralization and maybe even like integrating with uh with systems like maker to essentially like provide that extra mechanism to handle these deleveraging events but uh still like being able to they can be like a back and forth essentially with the extra security of these uh these uh leverage based over collateralized systems thats very exciting uh good luck in uh your research in launching that that new coin thanks for the answer and thank you thank you awesome well thank you so much aria for for sharing your research that was fascinating and thank you everyone for the great questions as well and for the participation uh without further ado uh arya is there a place where maybe people in the audience or those that will listen later can find the slides to this presentation yeah i will put them up cool and then i think i think this is also recorded maybe or will be available to you grab those from from sarah and then we can just put them in the description of the of the youtube video as well and then we can share the video out later sounds good yeah ill provide your link yeah well well good luck in the future and uh congratulations on all this research and really appreciate you taking the time and sarah thank you so much for helping organize this and everything that all the research team and everyone at ic3 is doing um its really fascinating stuff and i look forward to to working together more in the future thanks so much keenan and thanks to chain link for hosting this webinar and for their continued involvement in ic3 cool well thanks a lot everyone and have a good rest of your year afternoons and days Chainlink is a decentralized oracle network that enables smart contracts to securely access off-chain data feeds, web APIs, and traditional bank payments. Learn more about Chainlink: Website: Docs: Twitter: Discord: Newsletter: Telegram: Talk to an expert: