Taylor Brent - Guaranteeing Stability with Permissionless, Self-Healing Stablecoins

baby [Music] all right hi everybody I'm Taylor from reserve and today I'm speaking about permissionless self-healing stable coins um for some background on reserve Reserve is a platform that anyone can use to launch over collateralized uh self-rebalancing asset-backed currencies called R tokens our tokens are not limited to USD stable coins though that is a behavior that's contained as a subset of the overall protocol functionality in general the protocol can be used to create many other types of tokens as well such as yield bearing tokens or tokens that are pegged to a linear combination of any number of Target units but I'll leave that piece for another day and last thing for getting started the structure of this talk will be first half abstract and theoretical second half more empirical and third half cut so that we can get to the Q a so let's Dive In okay so I want to start with my own characterization of where we are in 2023 on the road to stablecoins being used as money now there we are now crypto loves triangles so I had to start with the triangle here um but the thing I want to emphasize is this is not a trilemma this is not one of those cases where we're looking to get two out of three in the best case this is more like uh more of the merrier or virtuous cycle case where a token that excels at one of these three functions will sort of find itself being used inevitably for the other functions and obviously the best money is a combination of all three the edges are places where you might identify properties that play into both of the nodes so you can imagine for example sorry about that um so where are we in the evolution of stablecoins is money I think we are a third of the way there and this is obviously a vast generalization but I think if I had to say where we are today stable coins are used because they succeed uh being a medium of exchange and to the degree that they're used as a store of value or unit of account it's sort of despite their properties on those Dimensions rather than because of them um stable coins you know they inherit the main property of tokens which is to say that they are instantly transferable in one block or whatever the duration of time until you have a finality on the Chain um to epochs and ethereum and transferable from anywhere in the world and this is just such a um out competing of the status quo that stable coins are in fact used uh virtually everywhere the two examples that come to mind are Arbitrage between centralized exchanges and decentralized exchanges and the role that tether has played in China to avoid capital controls if there were a stable coin with stronger store of value or unit of account properties than tether with equal liquidity then I expect that token would be used instead and forgiving and managing notes at the same time as I Advance these slides so the question I want to ask is what property should we add to stable coins to make them better money and in particular I'm going to be looking at Shoring up this far edge of the triangle foreign so this is the token formerly known as UST I believe it's a UC St nowadays um I think we're all pretty familiar with it but to give some background uh UST was actually sale for quite a long time as you can see here um you know it's a bit like the the pig or farm animal that thinks the farmer is its best friend right sort of using it's actually doing a type of science right it's an inductive reasoning every day it gets fed things are going well and you know if you were to extrapolate you might think it's reasonable to conclude that will continue to be the case forever and then of course the Pig gets surprised when all of a sudden it's slaughtered and I think a very similar thing is going on with UST holders where they thought they were in one Paradigm and then very quickly things changed this is the chart of new bits this is a early on stable coin using a hybrid algorithmic design it was algorithmic though not automated so it was uh fractionally backed with BTC and new shares and as you can see it shared a similar pattern where it uh held its Peg for substantial periods of time now you can see the d-peg between 2016 and 2017 and regained its Peg for another two years before falling off a cliff and doing the same thing that USD did and the third and final chart usdc so usdc I'm sure everyone's familiar is a 100 collateralized uh stable coin run by Circle on the x-axis here we are zoomed in so we are just looking at that one week during March where the svb bank run occurred I'm flipping back and forth just so people are aware so do you see it now I'm being glib I don't actually mean to imply that usdc is going to have the same fate as new bits in this case but you can see how usdc's dpeg sort of looks now a new bit steepeg looks you know a major difference is a algorithmic stable coin is going to go much further down to zero when it has troubles whereas usdc you know it's more realistic it might lose a few percent depending where the backing is held but it's pretty unlikely it would full on go to zero right the point I want to make here is not that usdc is going to go to zero in the future um that would be very bad and I don't think that's probably going to happen but I simply want to point out that we don't know what happens next right um there's some probability that usdc will maintain its Peg for a long time and there's some probability that we'll lose some portion of its value over time people can disagree on exactly what the probabilities are but and the degree to which they agree or disagree on that will inform what sorts of economic transactions that they're down to do with the token and to best illustrate this I want to talk about biology's Infamous hyperinflation bet so again to recapitulates though people I'm sure have this pretty fresh in their memory James Medlock said that he doesn't think the US dollar is going to hyperinflates and he would bet anyone a million dollars that it will not apology took him up on this bet as you can see here and the interesting thing here is they first thought about automating this through smart contracts with the use of usdc it's a pretty cool idea I like it but James Medlock quickly pointed out that usdc could lose its Peg now I think he's concerned trolling but he's also right right it could be the case that usdc lose its Peg and we simply don't know um the interesting thing is after some massaging he actually became okay with the idea right and as he says here in the parenthetical he was more worried he was worried about the d-peg risk but ultimately it's a small risk and he'd rather make the deal move forward quickly and I think this speaks to the superpower of tokens there's also legal differences that I think are playing an important role so I want to step back and ask uh you know what what is money money is a solution to the world's largest coordination problem right we have a coordination problem that includes everybody this isn't something that some people partake well there are people who opt out of the system but for the most part this is the largest coordination problem that uh we deal with we live in a world where disagreement is the norm I don't think we should seek to eliminate disagreement I think we should build systems that flourish when we have disagreement good money can be used by anyone good money can cut through disagreements and allow two people to coordinate productively even if they don't see eye to eye or even if they think the world is wildly different the strongest form of this statement is oops money is for enemies I don't think you have a money if your money can only be used among friends right that's not money that's Trust and the last thing before I wrap up this section um I want to reframe the argument I just made about the problem with non-healing stable coins as a subset of the overall problem with volatile tokens as money because I think that's how it's best viewed so what we're looking at here is a Monte Carlo simulation of uh just a random walk nothing fancy um you know to A first order approximation this is how stocks are treated and generally you know volatile tokens like BTC and eth though they have higher variants are this kind of thing or at least one way to view this is every line here could sort of represent a different view that a person might have about the fate of a volatile token as a money over time and its value so people disagree about the future and you know imagine denominating a mortgage in eth right the bank has to make money in expectation over all of its customers so it's going to ask for a very high interest rate in order to justify making the loan denominated in eth um and what happens in fact well I'm just not going to agree to that relationship to the bank because the interest rate that they're charging is too high and we're never going to end up doing the economic transaction we'll probably fall back to using US Dollars um and there's sort of a much more stable interest rate that they can charge me around that overall this results in less economic efficiency and makes this all worse off so to recapitulate the points here volatile tokens like BTC or eth have the problem in its most severe form non-self-healing stable coins have the same kind of problem but less so and none of these are going to be optimal money so next let's dive into the empirical section the dock okay okay so forgive me this is the one slide where I'm going to read a bunch of text um so some background on eusd eosd is an R token created by mobilecoin on the reserve platform its name is the electronic dollar it's an erc20 token on ethereum it's a bridge to the mobile coin chain mobile coin is a privacy chain with its own volatile crypto asset called mobe remob eusd is the largest R token to date it was deployed in February this year pegged to one dollar so it's non-yield bearing and it's backed by yield bearing usdc and usdt it has permissionless issuance and Redemption just like all our tokens and is backed one to one it's also over collateralized uh just like all our tokens it also has collateral default detection built into it so the protocol actively monitors the state of the tokens that are backing it and if something unexpected were to happen there's an algorithm for swapping it out and replacing it with better collateral this is done via an autonomous trading slash rebalancing which is where the exciting stuff happens and it's behind an eight-day decentralized governance cycle okay so this is a visualization of eusd on Friday March 10th before the news that usdc was having issues with its backing so as you can see on the right eusd is backed by equal four parts of Ave usdc compound usdc Ave usdt and compound usdt on the left you can see the colored Circle representing the direct backing and the black circle representing the rsr stake that over collateralizes the r token now I just want to point out that 128 is not a normal overlateralization um this was a long time ago this this back in February there were mainly a set of Enthusiast rsr stakers that kind of were just down to stake no matter what and that's how we ended up in this absurd situation where the rsr was 228 percent collateralized every R token also has a set of emergency collateral that the deployer gets to enumerate and can be updated through governance over time here you can see that the five emergency collateral that were chosen were USD 3C through dye and you can see the diversity Factor basically what that means is if there's a default of one of the tokens in the basket how many of these backup tokens should be used in order to take its place so if the diversity Factor were two and half the basket defaulted then that half of the basket would be replaced by two tokens splitting it into a 25 and 25 share um and you'll note that diversity Factor greater than one probably makes more sense but in this case that I walk through it's going to uh it was set by uh the deployer a mobile coin to A diversity factor of one's only one token is going to be used all right so this is the tweet that started all uh this is circle admitting on about Friday nights depending on what time zone you're in that about eight percent of the backing of usdc was trapped in Silicon Valley Bank the other 92 percent we just don't know and it seems like the market did think that some of that may have gone missing or be in danger as well oh and last thing this is on a Friday so there was no Redemption over the weekend and nine to five Bank hours meant that people were going to have to wait over 48 hours in order to begin redeeming their usdc and for the peg to begin being defended again yeah so now we're looking at the same visualization but one day later Saturday after usdc depegged the eusd Redemption price has fallen as you can see here in the top left down to 95 cents um I do on a flag this doesn't mean that the price of eusdc is 95 cents on markets given the rsr over collateralization the market is free to price it however he wants but it sorry uh I'll explain that uh okay I'll explain that um let's see what else and the last thing I want to raise is the how recently the r token was deployed you know this our token was deployed on February 23rd it's the first R token and Reserve is stated we're not going to deploy any R tokens so the first mover here is mobile coin is taking a significant amount of risk right two and a half weeks after deployment this happened um I think we and others we built the emergency system because we thought it might come into play at a point far off in the future nobody expected this was going to happen in the first two and a half weeks of the r tokens life cycle so in some sense this is actually like the perfect storm in a positive way for Reserve protocol the market cap of eosd was small it was crazy over collateralized and there was a relatively small d-pag of usdc so now we're looking at the same chart as earlier but just zoomed in to the narrow range of the weekend so as you can see usdc fell down to depths of about 87 cents about a 12 discount and slowly recovered its price over time with some volatility what I've marked here with the lines and points in time is where the protocol starts to act differently so you can see that uh slightly before March 11th the protocol detected default in usdc as it was trading below 1.25 percent less than the expected price a threshold set by governance and this default period lasts for 24 hours at the end of the 24 hours the basket is changed right and I would distinguish between the basket and the token balances the token balances begin being traded uh give Whenever there is a difference between the basket and the currently held token balances so at that point USD selling begins took a few hours and uh completed a few hours later at an average clearing price of about 95.5 cents yeah there's some details here about issuance and Redemption but I'm just going to simplify that and say the Dynamics of redemption of an R token during this kind of situation are such that the Redemption value is monotonically increasing over time and this is important because this creates the opposite of a bank run right holders are incentivized to wait for the rsr to kick in rather than redeem early if they redeem early they'll sustain a loss okay so this is it post uh basket change and rebalancing so as you can see on the right um I'm going if you that emergency collateral slide that we looked at was usdc followed by usdt since usdc was already defaulted the protocol skips over it and chooses to use usdt instead swaps out the ausdc and cusdc for Pure non-yield bearing usdt um let's see after selling all the usdc you can see that the protocol was uh 98 backed so you know he was trading at 95 cents and sustained a loss uh sorry the price of usdc also changes during this time but basically uh it sustained a loss of about two and a half cents after all this is said and done now I think there's some value and honesty for game theoretic reasons um and I want to be straightforward here that not everything went as perfectly as planned there was an issue where a trading circuit breaker had a false positive and what this caused is the stake rsr to not be sold in short the protocol has some mechanisms that allow it to pause itself when things are going uh quite poorly or it thinks something has gone wrong this doesn't require a human to do anything this is just you know the result of some trading was slightly different than expected so in this case and this may answer the question that you're getting at even though the rsr was not sold yet the rsr is locked inside until the arc token regains its Peg so that 120 percent staked rsr was going to stay in there until governance re-enabled trading and one week later you can see that it gets sold for the missing basket tokens and the backing percentage goes back to 100 percent foreign that this is all tether on the right side and again that was an outcome of the way the emergency collateral was chosen and didn't necessarily have to be the case shortly after this governance acted to vote to replace the usdt with the original cusdc and ausdc since usdc had regained its Peg there's another point that the protocol lost money by selling usdc right it got a clearing price of 95.5 cents on average usdc regained its Peg and you could imagine okay I need to speed up and you could imagine that uh if it awaited it actually would have been better off so well I think this is true this is sort of zooming in too narrowly on this one case right you know reality is highly uncertain and if we played the scenario over again a thousand times and you longed usdc every single time with all your money you would eventually be wiped out right this is what the Kelly Criterion says it says when you're playing an iterated game you should be risk-averse so the conclusion I want to leave you with is that self-healing stable coins uh are better money on the uoa front they reduce disagreement about the future value of the token and on the store value front uh it's kind of obvious token is going to hold its value better over time and last thing I want to mention is the definition of self-healing here so the idea is not new a stable coin that's able to recollateralize itself maker has a dilution backstop Ave has a rainy day fund but in both these cases you require governance to take an action and I think this is not a stable equilibrium this ends up looking a lot like the pig in the farmer where you know given small losses they sort of play along and recapitalize the thing and then the thing goes wrong enough and then maybe governance decides not to do it so when I say self-healing I'm strictly trying to talk about stable coins that heal themselves on their own without governance taking the actions all right well I was hoping to leave more time for Q a um but that's my 20 minutes the third section just in case people are interested is on what we learned around the challenges with implementing a legible autonomous decentralized rebalancing mechanism given that people can look into the code and try to attack it but yeah I'll leave it there [Applause] questions we got time for one happy to talk about the trading if that's people are interested in looks like you answer all the questions because you got one during the talk there we go all right thank you everybody