LFGHO 2024 Summit - GHO & Decentralized Stablecoins

[Music] so this is the official lfk Summit and I'll quickly give an intro on what we're expecting in the next two hours so you can get a good idea and you can get get right back to hacking so first up we're going to start off with giving you an overview of go and other just how decentralized stable coins work after that we'll be doing a short panel on Dow governance especially on how a community shapes the feature of go we're going to then talk about the RW facilitation with go as our third panel and that come will be the summit so the first up is to make sure everybody has a really good understanding of go and decentralized stable coins want to bring on Emilio to talk about all these things so without further Ado let's welcome Emilia on stage hello everyone I I I can talk from now right okay yeah hello everyone uh so my name is IO I'm not sure if everything is okay because I already have the sh the slides shared full screen I guess so uh so for those who don't know me uh I am like currently VIP for def in avara and I am responsible for uh leading the engineering and Innovation for uh the defi team specifically aabs uh that works with the a protocol as a contributor to the other community and also builds and contributes to other products as well uh so one of the latest things that aabs has built is go which is the main subject of this hackaton uh Soo is a decentralized stable coin integrated with uh the a protocol and uh the nature of the stable coin we will discuss about it later and how uh the basic design behind it the main topic that I want to focus on which is a topic that is particularly important especially uh lately because like uh go since launch has been facing a bit of challenges from the peg aspect is how are we working on uh helping the stability of go and what's cooking in terms of new feature for guaranteeing that U go keeps a stable value over time so um let's go further the first topic I want to talk about is what's stable coin I guess everyone listening is more or less aware of what the stable coin is uh but in general a a general definition of a stable coin that we can give is a digital asset that closely tracks the value of underlying real world asset asset uh through generation Redemption and stability this means that we are mostly accustomed to call stable coins uh USD packed stable coins but that's not the only uh kind of stable coins you can have stable coins packed to the gold value you can have stable coins packed to a stock price you can have stable coins back to eth or to bitcoin right so um the the difference is that the doors are synthetically generated so they are not actually backed by the real asset it could be potentially but then not only but they are also backed through other meaning meaningful mechanism that I we will discuss in the next slide so what what are these uh mechanism like for which stable coins are usually generated I would say that we can classify these mechanisms in three main categories uh one is over collateralized the other one is centrally controlled and the other one uh which is a bit INF famous let's say like let's say like this because like of things that happened in recent years is the algorithmic so over collateralized is the category at which go belongs and it's basically the asset is generated against a certain collateral which is guarantee for the the issuance of the asset so um whenever this asset is generated a certain amount of collateral is locked uh in the within the system through which this asset is generated and this collateral acts as a guarantee for the value of the asset itself the system essentially ensures that a generation and redemp and Redemption at $1 is always possible as long no BB exist and the the the over collateralization of the system is maintained uh the second one is centrally controlled which means that there is usually a centrally controlled entity that uh handles generation a Redemption of the assets against the real assets so typical examp example of these are usdc usdt uh which are usually generated against uh USD supplies so users usually give uh teet or Circle real USD H to be able to use this USD on chain so they receive um um what is basically an IOU of uh the dollars that were deposited in the form of usdc and us DT and the third one is it uses it can use a variety of different mechanisms to algorithmically attempt to track the price of the underlying asset in this case uh dollars uh without maintaining any form of collateralization centrally controlled or or um or or decentralized through decentralized protocols unfortunately until now in no design has been found that is actually able to maintain a reliable Peg and actually survive um using algorithmic approaches but we never know in the future something else might actually appear or might be developed and might actually manage to achieve this goal uh so how does like let's focus for the purpose of this particular talk um with over collateralized stable coins and mentioning as well going forward centrally controlled so how does like over collateral how do over collateral stable coins keep uh their pack so uh we can classify the mechanism to sustain the peg of a stable coin in two category two main categories soft Peg mechanisms and hard Peg mechanisms so uh soft Peg mechanisms are mechanisms that usually indirectly affect uh the the value of the asset and the peg against the underlying the typical example are the borrow rates uh but there is also generational Redemption at a fixed price this is how go actually works and is integrated within the a protocol a go within the a protocol doesn't have an oracle it's like the value of go is fixed to $1 so uh whenever you will generate um whenever you generate new go you basically have the opportunity to get them at $1 whenever you uh uh redeem go for the underlying for the collateral you and you will redeem it at the value also of $1 whatever is the the the value in the market is this basically helps the pack because creates an Arbitrage of opportunity so if the value of go is above one on the market like there is incentives for people to actually mint go and sell on the market for a profit and when uh go is below Peg there is an incentive for people to buy from the market and repay their debt or eventually to to liquidate a further profit than the the the so liquidation bonus uh the rates I already mentioned so so usually the rates are be are used as a way to incentivize or disincentivize forther minting of the stable coin so whenever the asset is below Peg interest rates usually rise to dis disincentive ites for their meetings and help the price recovery uh whenever the uh the asset is above bag it means that there is demand on the market so the borrow rates are low to incentivize for their meting for their meeting and allow users to capture the uh opportunity that I mentioned before and then there is another approach that is for example used by liquid that quite peculiar liquidity doesn't have a uh interest rate a borrow borrow rate for LUSD uh so but what they have in exchange is uh generation and Redemption fees so every time like someone meanss a new Deb or redeems the stable coin for the collateral uh a fee is paid and this fee adapts depending on what's the price of the market of of the asset on the market so it it rises or decreases whether or not we want to incentivize or disincentivize forther uh meetings so these are the soft Peg mechanisms the hard Peg mechanisms are uh for example direct redeemability against an underlying asset or a stable coin with a stronger Peg which is what like for example centrally controlled um stable coins do or for example what maker does through the peg stability module and what I will do as well with through with go through the go stability module or another hard pack mechanism is the RM direct redeemability against the most under under collateralized positions so again this is how like another peculiarity of uh liquid USD uh so basically when the asset goes below P there is an incentive for people to actually redeem the stable coin against the least under collateralized position so uh there is incentives to go and repay the debt in uh uh for the least under collateral position in exchange for the collateral for $1 worth of collateral for each L USD that is uh redeemed minus the Redemption fee that I mentioned before uh so uh let's uh let's focus a bit on the borrow rates which is the the topic main topic of this uh talk uh because like we are going to describe what are What what is the the current strategy for go and what are future strategies that we want to adopt uh for go and for helping uh maintaining the peg over time so at the moment uh go uses a governance controlled uh um mechanism for borrow rates in uh in D this is called stability fee so what happens is that the Dow basically manages the the the borrow rates uh and decides what the proper borrow rates are depending on the state of the market so if the asset is under Peg the governance decides to rise acting a bit like um Central Bank or a central organization decides to rise the interest rates for the borrowing making the borrowing more expensive uh and incentivizing repayments or if it's a bull Peg it makes the borrowing cheaper to incentivize further minting and uh helping dropping the price on the market or algorithmic algorithmically controlled um uh so there is an interest rate curve that defines uh the the interest rate of the asset depending on Market condition one example here is um CRV USD uh where in CRV USD this particular formula is applied by using the the P keeper that is another mechanism that they have to help maintaining the peg to understand how what's the situation on the market for crvd and this dynamically adapts increasing or decreasing the interest rate to make it cheaper or more expensive to borrow we have something similar on the other protocol for assets though that are supplied so the the mechanism is similar but the concept is different because like the assets that are supplied it works on a strict demand demand and and Supply um relationship rather than evaluating Market condition but conceptually they are very similar so the the the interest rates fluctuate and are variable with very high variability so as I mentioned go is currently using a governance control mechanism which is the the mechanism that was released uh I mean was active since the release and there are a few problems with governance controlled grades the first problem is that it requires a lot of governance overhead so there is a lot of modeling needed participation for people to vote H entities that are able to decide what actual proper interest rate is and is also very slow uh to enact react because like it's slow usually takes days for a proposal to be approved voted and and executed and it's also slower to react because if there are sudden changes in market conditions uh then it requires again the the the to bootstrap whole process again to actually adapt the rates further and it's also very hard to estimate properly H and and uh like but also by analyzing the only algorithmic uh interest rate approach that there is on the marketed this uh CRV USD what we want is not what CRV USD does which like brings a lot of variability so H there is uncertainty and lack of predictability on what the borrowers are going to pay which is a very similar again to what a offers with over other stable coins we want something that is a tradeoff between the um maintaining the Peg and also predictability so we want something that is actually that at least the borrower can have an idea of where the interest rates are directed if like on they are an upward Trend or downward Trend and like an estimate for example of how much they are going to spend in the in the I don't know one month three months six months one year from the time uh of the of the borrowing so the initial idea that we had was to use uh PS so uh PS like it's um are mathematical structures very very uh common in control theory and they are basically used in general when we want to stabilize a certain parameter of a system around a certain value you know so what's what happens is is you very usually um applied in industrial machines and applications so uh what happens is that we have we usually have given un certain error E T we have one component that is proportional to the error one component that is the integral that basically tracks the historical of the error and another component that is derivative that basically tracks the trend of of the error so how quickly the error is moving from the uh from the current value to uh the next value and of the value the target value that we are actually aiming for so I initially we plan to do to use this this like would have required an oracle which would have given like a a t up of the past uh prices to basically estimate the the historical error and also use a derivative approximation to basically approximate from the last n prices what the uh how is the trend the acceleration of the of the variability uh we tried to apply like the P module it felt very limiting like were not completely satisfied uh with the the results that we were getting because it's the PID uh module are very limited like as I mentioned again you have the only V variable that you can check is the error against the the target uh value and then like the only components that you can actually uh you can actually play with to try and approximate the the value the best way that you can is just the proportional derivative and integral uh parts of the equation so it felt very limiting and didn't perform uh the way we wanted we wanted something that actually would be able to uh give a not an actual rate but a a rate of increase or decrease of the of the borrow rate so it would basically tell how much far how fast we need to write the interest rate to bring the asset back to the Peg and how fast we need to eventually decrease if the asset I mean increase if the asset is below Peg and decrease um if the asset is above Peg so rather than outputting directly a a borrow rate we want something that would output a a rate of increase of decrease of the of the borrow rate and the system would up date the borrow rate in like continuously depending on this rate of increase and decrease this way we give we give the borrowers the possibility of understanding what's the trend in the interest rates and eventually uh how fast this rate is changing so that they can have an idea of how much they are spending so uh this is because we basically with go we want to the to to have the a protocol offering a different protocol than uh other stable coins that are already listed on a that are based on full variab Full variable interest rates so the interest rates can be 3% today in 64 tomorrow there is really no guarantee or what or capacity to predict what the interest rate will be so uh we have been evaluating other options uh rather than PS and uh what we realized is that we actually needed more control inputs not not just the error but we also want the ability to embed in mechanism like the the uh an an evaluation of the current market conditions an evalu evaluation possibly of the state of the GSM so what's the ratio between uh uh the assets that are supplied in the GSM because this tells us how uh much we can expect to sustain the current P for example so we decide to use another um technique that is very very common in control logic that is called f in control theory that is called FY logic so what's FY logic um I would say that it's very novel application of control theory to blockchain because I'm not aware of anyone using FY logic for um interest rate controllers so um it's a form like you basically have a logic that it's not just zero or one so not just true or false but it's based basically on degrees of truth so the idea is that you have input values that are hard values called crisp values and then you map those crisp values in sets that set that Define the state of the system or the the the belonging to a certain set depending on on on the value itself so these sets usually go from zero to one so you can let's imagine for example temperature you can define a temperature as high very high medium or low so you have I don't know 20 degrees for example has a certain membership of 0.5 uh within the medium uh set and a very low membership on the on the high and on the low which means that basically uh this allows to approximate a a bit like the way humans Reas reason right for us temperature is never like first of all is subjective so it's never uh super cold or super hot for everyone but also we can we usually approximate uh values in a way that are actually like not hard hard true or old false but there are usually degrees of uh of membership and evaluation so the idea with go is to use this particular uh technique to be able to uh approximate the the the Peg and the way that the interest rates need to react so how does f logic work we have basically three ma major components of of with with this particular technology one is falsification then the definition of the fuzzy sets and the rules and the third one is the fuzzification so once you have once you apply the you fify your values so you pass from hard readings uh to um values on mem membership within the sets and you apply the rules uh then you basically get the output of the rules from the output of the rules you def fify the the the value to get an hard number in the case of go what we are going to get is the rate of increase or decrease of the of the current borrow rate so uh again falsification so the main the simplest value that we can use to understand fication is the the current price right of go so let's say the current price of go is 098 uh we Define a set which defines the peg as very much of Peg average off Peg ineg uh average off Peg on the below Peg and very much off Peg below Peg and with these sets the current price will belong to certain sets so 98 is we can say very much uh below Peg with um membership one so we can say for example is the lowest um value that we are willing to accept within the sets right so we can apply uh with after that we defined these rules that essentially gives us how the system should react based on the on the fuz values that we get so one rule for example could be if the distance from the peg value is high like in the example before 098 and the value is negative the rate of acceleration of the interest rate is very high so that that is a rule that defines how the acceleration of the interest rate should behave depending on on the state of the P so the third step is so for example inex that I did before 098 is very much of Peg so this rule would be satisfied with um with value one which is essentially uh the maximum this gives the maximum rate of acceleration and the last step is the fific so translating the values the outputs given by the formulas to actual values of acceleration so again we Define some sets that map the results of the rules and from these sets they will uh we essentially obtain actual values that we apply to the borrow rates so uh there are a variety we can discuss a lot like it would require way more time to actually describe the FY logic in detail but let's say that uh this should be enough to give a basic understanding of how we plan to apply uh this mechanism to uh the go borrow rates so the interesting thing uh compared to P is that while in P you have AIDS you have only one input which is the the current price with uh fzy logic you can apply a variety of inputs that all contribute at the end to actually decide what the rate of acceleration of the pag should be so for example the first three that you read here distance from the peg historical of the distance from the peg a rate of acceleration these three will actually map exactly what the pad was mapping before but in addition to that we can add for example balances of the GSM uh state of the secondary Market or really anything that we can POS possibly bring on chain so for example um an oracle that gives us the state of the market liquidity of go across centralizer exchange if we have an oracle that gives that we can write a rule that actually that will actually contribute together with all the other rules in uh defining what the rate should be this G gives us essentially the flexibility to cover anything that we need uh to B to basically manage uh the the the interest rates depending on the state of the p uh the rules will be defined on chain so the contract will actually have all the rules uh that will be and those rules will be extend uh will be possible to be extended by the governance so the governance will be able to add more rules with eventually more input values and uh the approaches to the falsification so there are a variety of mechanism that you can translate the output of the rules to actual values also this will be C it will be possible to customize and the governance can decide eventually by monitoring the system if other methods of fific can actually be uh better suited for the the the the the state of the market and the current like usage of go so uh what are the next steps uh we are currently building a model in Python uh to sim to simulate this rate mechanism uh we will run this model in Python against the historical of the of the peg soon the GSM will also go uh live unfortunately will not not help at the beginning because like the state of go is currently uh below Peg so the GSM really helps when the asset is above Peg uh so uh after that there will be a stage of validation uh by the governance uh to understand if the governance accept this solution and uh after that after that and after running again the simulation we will uh execute the onchain uh proposal to activate the new interest rate model and then monitor the performance in production to essentially be able to evaluate if the initial set of rules was uh targeted properly if the deification that was chosen was targeted properly uh therefore we will monitor the performance of the default implementation of the of the strategy to adjust it accordingly depending on how it performs uh so in conclusion um stability is a very very uh complicated topic as demonstrated by um many events in the past um it it is very hard for a stable coin to maintain the pag of all times we should accept we should accept that uh stable coins can lose the pack from time to time it basically happened to all the stable coins even the centrally controlled ones at a certain phase of of their life this is part of uh of the life of the stable of a stable coin the automated mechanisms usually provide faster response times uh compared to centralized and or governance controlled mechanisms and that this of course will help go uh reach and maintain the peg but this also cannot be uh the only way the only mechanism really Li on to actually uh reliably maintain Peg maintain go on Peg use cases and and growth are always very important to incentivize people to participate in and using the asset and holding the asset and make sure that the asset is actually um the asset can scale uh um to the full of its capabilities so uh that's it I don't really know if I'm on time over time because there is no timer here but uh that covers everything and I would say thanks for everyone thanks everyone I don't know if I can accept questions otherwise yeah that's it um ailia that was actually very amazing and and detailed um there actually are a lot of questions but we are over time so my request would be for you to go to E global. TV and just answer those questions in the chat because there's a lot of questions about how the the peg works and how you actually think about the the mechanism for um getting all of the the redemptions um and just kind of how the facilitators and Pad control works so um there's like really good four or five questions if you can just go to E global. TV and reply in the chat sure will appreciate that so thanks again um really amazing to kind of have a deeper understanding of how the protocol Works which will give all the 800 plus hackers is a good idea for what they can do appreciate it thanks guys [Music]