Exploring the decentralized stablecoin design space - Vitalik

Okay, great, thank you. So I think, I mean, the way that I approach the stablecoin question is like, I'm personally not as interested in like getting 7% yield instead of 4% yield. The thing that drives me is not getting minus 100% yield. And there have been plenty of stablecoins that have given people that. So basically, the way that I think about this is if the customer is not even necessarily users but applications, right?

And as an application, you're not able to kind of quickly flick between things based off of changing opinions about which provider is trustworthy, which provider is not trustworthy. You want to be decentralized yourself. And what is a mechanism that you can set down, be sort of minimal, as credible and neutral as possible, and be something that you can make a five-year bet on PolyMarket with, right? So I think basically the way that I think about this is that there's three major types of stablecoins, right? So there's centralized stablecoins.

So we know of USDC, USDT, BUSD and so on. Then you have what I call like DAO governed real world asset backed stable coins and so DAI is the most famous example here, though I think there's definitely room to try to innovate and make more examples. And then there is governance minimized stable coins that are fully decentralized and Rai is the best example there. So I think the second category is interesting, right? And it's interesting because there's a lot of value that you can get out of like actually connecting people directly into real world assets.

And like real world assets are ultimately the productive economy, the point of a financial system should ultimately be to support the productive economy. There's a lot of different things there that have different properties in terms of stability, different properties in terms of generating revenue, different jurisdictional properties. So what principles do I think it makes sense for a Dow-governed real-world assets table going to follow, right? I think to me one big one is total excess in collateral should be greater than the amount held by any single issuer or in any single jurisdiction, right? So this is, like, pretty basic if you care about decentralization, right?

You don't want the thing to break solvency even if like any one specific issuer breaks, right? And I mean, DAI itself, right, actually historically has not really satisfied this well, right? When Black Thursday happened back in 2020, they basically went immediately from being ETH backed to being like basically a front end for USDC, right? And I think to me it's like if you're willing to trust one single actor in order to be solvent, then like you might as well just have a centralized stablecoin. And to me the goal of this is to basically have a governance structure that gives you some kind of like long-term reliability that you have something that like actually accomplishes the core property of decentralization which is basically no single issuer going wrong or no single jurisdiction going wrong should be able to like actually break the core guarantee and the core guarantee is one unit of the stable coin is worth like a dollar or like one of whatever you're tracking, right?

So like that, like basically, you know, rule of thumb in terms of collateral. And that's like a very simple rule of thumb. If you want to like go fancy and mathematically optimal about this, then obviously, yeah, you know, you go into like the entire theory of like how insurance works. And, works. And you start thinking about correlations and uncorrelated things and what kind of collateral ratios are needed based on level of decentralization, how correlated the risk is.

And there's entire textbooks on this, I'm sure, that I haven't read. But if you want to make something perfect, you probably should. But, like, you know, basic rule of thumb start here, right? Then, obviously, standard requirements for a DAO, right? Basically, you want to avoid governance attacks.

And, like, to me, this is still a very unexplored part of the design space, right? Because, like, most people, it feels like we have this effect that I really hate in the crypto space, which is like, you know how there's this meme that nobody got fired for buying IBM? And it's like in the crypto space, the meme is nobody got fired for forking Compound. And we've gotten this meme in our heads that the only way to be decentralized is to have token voting. And token voting is an awful mechanism, right?

Token of voting gets easily, it's like basically one of two things, easily dominated by a very small, highly centralized group, or you get something with a tragedy of the commons effect. And like sometimes you only need a small percent of the supply to take over the thing. And it's like, even in theory, you have this, you have this problem that, okay, I'm trying to figure out if that was a metaphor for something. Like, even in theory, you have this problem, right? That like, if the market cap of your assets under management cannot exceed the market cap of your governance token, because someone could just buy the governance token and then cause all the oracles to go to zero or whatever numbers allow them to steal all the assets under management and then you're screwed.

To me, the only way to break past that barrier is to have some form of non-financialized governance. At the very least, you want some kind of delay mechanism. If you have to have a governance token, you have them vote on oracles, and then you have a rule, you can change at least like maximum one of like, say, 13 oracles every week. And that way, you know that even if a governance attack starts, you still have like six weeks of honest oracles within which you have to like get your money out and switch to something else, right? So avoiding governance attacks, avoiding vote buying, avoiding voter non-participation, also at the same time actually ensuring quality, though you could also make an argument that it's a standard argument that the thing that democracy is worst at is making optimal decisions.

The thing that democracy is best at is making mid-dec, but avoiding terrible ones. And maybe that's like exactly what a stable coin requires, right? Because it's like, if I have a stable coin, I want to frigging not lose my money. If I'm okay losing all my money for the sake of like better returns, I'm just going to hold ETH, right? So, right.

And then there's like interesting ideas about futarchy like voting dynamics, right? Which is basically that, you know, if you vote for a delegate who vote for, who like makes some decision that's upstream of a collateral asset failing, then your tokens get slashed, right? And the nice thing is that that's something that you can like kind of see and kind of enforce in the protocol, which is interesting. So let's get into the fully decentralized stablecoin world. So what I always found interesting about Rai is it really tries to do the minimized governance thing.

The goal of Rai is basically that the governance does exactly one thing, and that exactly one thing is managing a price oracle, right? And Rai has this very simple model where there is a target price, and that's basically a target price of Rai. So there is a target price of Rai denominated in US dollars. And then the Oracle tells you the USD ETH price. And then you can look up the Rai ETH price on Uniswap.

So like the app itself can just go in and do that. And then you just like multiply one by the other and you can see whether the actual price is lower or higher than the target. And what you also have is you have like a built in interest rates, right? So the target price adjusts by like some percent per year. And if the actual price goes too high, then the interest rate starts going lower.

If the target price goes low, then the interest rate starts going higher again, right? So that's like a pretty simple mechanism, right? It's how Rai works. And the basic idea is basically that what you want is you want to find an equilibrium, right? You want to find the equilibrium interest rate at which the number of people willing to hold positive units of your stablecoin equals the number of people willing to hold negative units of your stablecoin, right?

And Rai has been running for a couple of years and it seems like it actually does that. So I wrote this post about two years ago on thought experiments for evaluating decentralized stable coins. Right. And I had two thought experiments. Right.

So one is, is a safe wind down possible? Right. Is it possible for a system to safely wind down to zero users in such a way that no one loses money? And obviously, Luna is like one big example of this not being true, right? So who here remembers Luna?

From the moon! So, you know, let's basically, you know, first we space the way that Luna worked is it was this like system where it's like actually kind of similar to the senior shares paper by Robert Sams back from 2014 where you have two coins, you have a small tile coin and the stable coin and you can like constantly trade one for the other. But the key property is that like the total quantity of stable coins that gets issued is positive and those tokens get backed by the volatile coin right and if the total issued quantity of positive of stable coins ends up actually exceeding the market cap of the volatile coin, then actually you're screwed. And the system has this potential to basically flip and completely crash. And the problem is that Luna basically did everything that they could to exacerbate this risk because they started adding these Ponzi dynamics, they started adding this mechanism where you can stake your tokens and you can get 19% APRs.

And if someone offers you 19% APRs, there are exactly two possibilities. One is that they're scamming you, and the other is that it's denominated in a fiat currency that's scamming you. But a lot of people were more optimistic, and then they ended up being very wrong, and people lost like 30 billion dollars right and and because of this a lot of people disel ended up like being afraid of algorithmic stable coins as a concept but to me that's the wrong lesson right to me it's uh that you want to be afraid of a specific type of design of algorithmic stable coins and the rule that would have prevented this situation completely is that if you have a decentralized stablecoin, the total quantity of stablecoins issued has to be zero. What that means is for every positive one you issue, you have to issue a negative one, right? And this is actually something that DAI does, it's something that Rai does.

And basically, you keep the amount issued and you keep the debt exactly the same. And that way you guarantee that you can pass the wind down test because during a wind down, you basically just like match every positive with a negative and then everyone gets back the value that they had and then it all comes out of collateral. So that's thought experiment one. Thought experiment two, which is like, what happens if demand for holding the EKS stablecoin exceeds demand for holding negative amounts of the stablecoin, even at a zero interest rate, right? And this was actually the thing that caused DAI to like pretty chaotically flip over from being ETH-backed to being a USDC front-end back in 2020, which is basically that there was not enough demand to hold negative units of DAI that were backed by ETH.

And so basically demand for positive DAI was high, but then demand for putting in collateral was low. And if they had done nothing, the price of DAI was high, but then demand for putting in collateral was low. And if they had done nothing, the price of DAI would have started going up and DAI would have entered a speculative bubble. And so just basically quickly scramble and throw in huge amounts of USDC as collateral. And I think right now this is one of those things that's easy to forget because if right now we're in this relatively high interest rate regime but macroeconomic conditions can change lots of things can happen maybe in 2027 we're gonna be back in negative interest rate lands like actually yeah people love making like all kinds of prognostications that are like super confident but like really nobody has a friggin clue right so I think any protocol that exists now, like it has to, you have to keep in mind the possibility that you are going to have a sustained negative interest rate regime at least for some portion of the time during which the protocol is operational.

And you basically have two possibilities, right? One is you accept the possibility of your stable coin having a negative interest rate. And so in that case, you accept a floating peg. And like you could change it around on the UI level, right? The other thing you can do is you can have like an ERC20 where the negative interest rate is kind of baked in.

So basically, every time you do ascend, the negative interest rate gets calculated first and then you lose extra amounts, but then it's like basically, you know, you do end up having to suffer one sort of UX issue over the other. And the other possibility is like, just accept the risk that your stablecoin is gonna turn into a bubble or turn into something else that's bad in this situation, right? And like personally out of this, I prefer option one. So that's like another part of the stablecoin question. I think another thing that's really interesting is this question of what index do you follow, right?

So basically, if we want to accept the user experience and the accounting downsides of not just perfectly following USD, then there is an argument that maybe you want to actually go further and we want to actually be less skeuomorphic with these things, right? So skeuomorphic is this term for UX design land that's basically making computer user experience items, things that really try to look like physical things that they're modeled after, right? And basically the equivalent here is that skeuomorphism in stablecoins is giving people US dollars because people are used to US dollars. Now the argument here is, well, what if we don't be skeuomorphic and what if we take a first principles approach to a stablecoin, right? If I hold a stablecoin, what do I like actually want?

Like if I want stability, what do I actually want? And the answer basically is that I want to guarantee that if I have k times n units, I can pay for n days of my expenses. I think that's ultimately what stability really means from a very practical perspective. The conclusion of this, well, is you make an index that actually tracks people's expenses. Decentralized CPI is a meme and it's starting to become an actual thing.

I've talked to even people here working on decentralized CPIs. I've seen people elsewhere. I mean, Frax is, I mean, they're just like basically have a pointer to the Federal Reserve CPI. So, a lot of different people working on different indices. And, like, today, I personally think that Federal Reserve numbers are reliable for inflation.

So sorry to any ShadowStats fans in the room. But, you know, there are tail risk situations where that might end up actually changing quickly. And there is an argument that I think you want to internationalize and you want to minimize dependency on the economic situation of any one country. And so if you do that, then maybe these are arguments that we actually do want some kind of crypto-native decentralized CPI in some sense. And what would that mean?

That could mean you have a DAO that manages oracles. But then those, instead of those oracles being dollars, those oracles like actually are a basket of all kinds of different expenses that people have. Right. So I think that's interesting. I mean, I, I think there's also another radical idea, which is like, you just create this like crazy multidimensional market where each user chooses their own index out of thousands of, basically literally everything that has any kind of price feed anywhere.

And then you have this multidimensional market that just balances interest rates across all of these and you make UI wrappers that make it really easy for people to make their own CPI out of whatever they want. And then this completely breaks fungibility. Basically everyone has their own coin, but then what you get out of... And then the way that anyone actually pays anyone else is you just basically shove every payment through a Uniswap or whatever. And the theory is that this could actually work relatively fine with mature enough infrastructure, probably aided by AI or something, something, and hackathon people figure it out.

So I think it's another interesting topic. So another topic is what's the collateral, right? So I think historically, we've seen like basically a couple of options. Really trying hard to be governance minimized. And so Rai had ETH as its collateral, right?

ETH, credibly neutral asset. You know, I don't know, I guess ETH used to be my ETH's ticker now apparently. But credibly neutral asset in the base of the Ethereum ecosystem, and that's great, but just holding ETH is uncompetitive if you miss out on all the yield, and I think this ended up being a big reason for like your rise underlying unpopularity. Right. And like even though, you know, like I definitely don't believe in, you know, like risky yield maxing for stable coins.

I do think that you have to be reasonably yield competitive. And the basic reason is like there's a lot of applications that break if you're not yield competitive. Right. So like if you like from if you think about prediction markets, for example, right? So in prediction markets, like about, so like literally today, right?

You can go on Polymarket and you can make a bet. You can bet that there will not be a civil war in the UK in the next two months and one day. And if you make this bet then the price the price of that is 0.975 and so you can make two and a half percent right so two and a half percent in two months 15% yield minus you know risk of an actual civil war which I think is like counts for like 1% and then minus the risk of errors which is on 1% right so it's like you know they're like you can kind of make like 13% yield but then there's a lot of other questions that are like that, candidates just that like actually stand no chance of winning, just have like lingering percentage rates of like 1%. And then a lot of people just get convinced that like, oh, this guy actually has a chance when like actually they don't, right?

And so like to me, it's like, I think the over-weighting small probabilities and under-weighting probabilities that are close to one is like, like basically everyone agrees that it's like the biggest systematic weakness of prediction markets. And the problem, of course, is that if arbitraging against it requires you to like accept very crappy or like zero yield, then people are not going to arbitrage against it. And so the crazy people are going to set the price. Right. So the you know, you do want an asset that where like if the intended use case is Polymarket, you wanna use an asset where one, the asset is the governance and like all of the different properties around it are robust enough that people feel comfortable sticking money into it in 2024 and getting it back in 2029.

But two, you do want the yield to be competitive enough so that if you put the money in now in 2024, like as an arbitrage strategy, then basically the amount that you lose between 2024 versus 2029 versus an optimal portfolio is like relatively small so that arbitrage strategies are actually possible across long time horizons right so this is I mean oh basically the problem that right right into and so now I mean they're doing high and there's other stable coins and they're getting into staked if the problem with staked ETH is of course like what is the governance minimized that ETH. The problem with staked ETH is, of course, what is the governance-minimized ETH staking solution, right? And one approach is to just pick one, right? Everything into RocketPool or everything into Lido or everything into something else. And there's centralization risks of this.

And you could make plausible argument that like maybe one of these things should just create its own staking pool that's reasonably decentralized enough and that way like if you participate like you're not actually going to be increasing Ethereum centralization on some axis but the problem like whatever you do once you get into the staked ETH game it's like either you are contributing at least somewhat to Ethereum staking centralization or your thing is not governance minimal at all, right? And like that's one of the problems. Now, recently there's been a lot of interest in like eigenlayer and AVSs and basically use it, like what that actually works out to is you're basically saying you accept anyone's staked ETH as collateral. The problem with accepting anyone's staked ETH as collateral is like there's this tail risk situation where what you're actually doing is you're like enabling attacks on Ethereum that are cheaper because someone could go in like 51% attack and then if they 51% attack they're not going to 51% attack with your ETH instead they are basically going to like put up their validators and then they're going to put their validators up into this thing and then they're going to 51% attack and then they're like actually going to get most of their money back because the like their the ETH that they have in that in that stake is not theirs it's like actually like lent out through these protocols right and so that's uh like basically the risk that you're incurring if you just uh do this kind of let anyone participate approach right and then option three is um of course just like plain eth but then wait for the possibility that ethereum will incorporate some kind of stake capping which would realistically realistically mean much lower staking rates. And then at the same time, thanks to the Verge, you get much lower difficulty of staking.

And so instead of staking becoming this kind of high difficulty, relatively high yield thing, it just becomes a low difficulty, low yield thing. And then some people who choose to participate participate can people who don't choose to participate including these protocols don't have to unlike if the penalty you incur by holding plain ETH versus staked ETH is like 1% then like maybe for five-year polymarket questions that's kind of okay so those are like some of the nuances around collateral so So conclusion, right? So optimal open areas to explore, and this is especially if you're really designing stable coins for the stability use case. So one of them is governance mechanisms, and I think a key question that's really important to think about is how do you achieve security even above crypto economic balance? Right.

So how do we achieve security even in a situation where your assets under management are greater than the market cap of any protocol token that you have? And potentially going even above that, like just plain buying 50 percent of the tokens is one kind of attack. Another kind of attack is like vote bribing, and so how do you try to achieve some degree of security even under those kinds of situations? So that's one question. Another question is like indices, right?

So tracking the dollar, at least for like decentralized stablecoins, is like one thing you could do and it's like kind of obvious, but then if you're willing to incur the costs of like not being a perfect dollar then like actually there's a lot of other things that it makes sense to track that are not dollars right and so this could mean cpi and potentially some kind of decentralized cpi it could even mean like you have different indices right so uh you know like you one type of expense that you could try to track is, you know, things like food and commodities. Another type of index that you could try to track is like rent prices in specific cities. You could try to track a lot of different things, right? And then finally, collateral types, right? And for collateral types, like that kind of splits up into two sections, right?

One One section is the very governance minimalistic approach, in which case basically the most governance-minimized collateral with close to competitive yield. Is that ETH? Is that some specific version of staked ETH? Is that some kind of basket approach of different types of ETH? If so, how is that being managed?

And then the other approach is this kind of MakerDAO style approach where you have very active management that puts money into things, which could be on-chain DeFi protocols, and it could also be like off-chain, like real world, you know, government bonds, corporate bonds, whatever. And then the question actually becomes, how do you choose these? Which ones of these do you exist? What kind of formula do you try to use? How do you balance between yield and resilience to tail risks?

And how do you ensure that the governance of that whole thing actually really stays viable in the long term? So, yeah, I think these are interesting questions. And I look forward to more people trying to solve them. like 10 mil you know maybe of like rye high and grye and so and then total market cap of sable coins is like 170 bill and so like that differentiation you know how do we get more people giving a fuck about decentralized sable coins is it just about yield which i think like you know part of this presentation is like the yield is not competitive but like yeah i mean i think like the real and one very very just like realist answer is like yield being like at least greater than zero is part of the answer, right? Because like it does seem like there's like legal barriers to USDC types offering greater than zero yield and like Rai types actually do, right?

So that's, I think that's one part of the answer. Another part of the answer is like, it's like realistically if USDC, if the yield question did not exist, right? So if in the long run we're on average in a very low interest rate environment, then like there is an argument that, you know, if USDC proves to be trustworthy over 100 years, what's the point of having anything other than USDC? And so the contrapositive of that is basically that any argument for things other than USDC has to connect to some risk that people actually are concerned about that actually is connected to USDC. And like I think political tail risks are definitely a valid argument, right?

And like, you know, there's risks to USDC as a whole. And then there's also obviously, you know, like risks in terms of, you know, like what holders will USDC continue being friendly to under what conditions right so that and then with usdt it's like you know basically the same argument though i think for usdt there's uh you know like it's definitely less transparent than uh than usdc right and so there's more uh arguments that like you know we actually want something where it's like very very clear like where every part of the collateral is coming from so it is possible that the answer just is that like you you have projects that are going to keep basically doing their thing with relatively small market caps until some kind of event happens that suddenly makes them interesting for people. And then the other answer is like, you know, you don't need a maximum yield, but like actually creating a stable structure that gives people a couple percents that they would not have if they just held USDC and USDT is also like that's something that would attract users today. Building off of James' point that decentralized stable coins only have like, what, like bips of a percentage of the like total circulating supply of the centralized, of the more centralized stable coins. Do you think that it's better to like, going off of what you were saying earlier lean way more heavily into kind of like moving away from the skeuomorphism so that like at the end of the day we all recognize that these are basically like power user instruments and when you lean into that and think about it as kind of like I don't know like a Vanguard retirement you know 20 and think about it as kind of like, I don't know, like a Vanguard retirement, you know, 2065 fund, then we're kind of able to create a more like elegant solution.

Not ignoring, but like accepting the fact that this isn't necessarily something that people are supposed to be using for everyday payments. But it's something of, as you were just saying, like a, youand-forget type yield accumulation vehicle? Yeah, I think that's definitely one type of good strategy that you can have. I think one kind of nice complement to that on the payment side is that I think, at least from a UX perspective, we do need to be moving toward a payments world where it just becomes really easy to pay in whatever currency you have and then it just automatically gets routed into whatever currency the recipient wants to receive and i think some of the discussion around like cross layer two stuff and like chain specific addresses and incorporating milk 7683 into wallets and like that whole direction is Already going to get us about halfway there right because it already is yeah, like at least sets the standard that Ascend is an operation that could be executing complex logic involving like exchange and liquidity providers in the back end and then once you're At that point. That's like relatively a smaller step until I can extend it to say, well, actually if someone asks you for $100 and you don't have dollars, then like the wallet should automatically offer to let you pay in like ETH or RY or CPI tokens or whatever else you happen to have, right?

And if we have a world where like which like currency network effects are less relevant from a payment perspective then that gets us into a world that's like practically speaking more open for people to have different types of stable or approximately yes approximately stable assets. And then, like, I mean, hopefully, you know, yes, like, we will have more room for, like, more custom things like that to exist, right? And I think the precedence there is just to think about is how, globally speaking, we already have a kind of ecosystem where you have lots of approximately stable assets and people can pay with whatever and receive whatever because that's how fiat and the credit card system works already. I think if you compare it to that like the crypto world and even with exchange spreads and all that is already very competitive. Thanks for the insightful talk and answers.

My question is regarding regulations of stablecoins. After the Terra Luna fell from the moon, regulators really activated in terms of, especially in Europe, but I guess in other parts of the world as well and my question is is it the time to really build decentralized stable coins right now when when regulators are so active in trying to capture this field thank you I mean I am my own impression is that like there's definitely yeah definitely some level of regulatory hawkishness on this, but it's definitely far from people trying to ban the category. there are things that the crypto space can do to try to like front run the situation and like prevent you know like another Luna from happening and making the that whole situation worse and I think the thing that we should do is like basically actually try to do the self-regulation thing and like basically yeah like actually yeah make the information more accessible for stable coins of like, what are the trust properties of a stable coin? What are the assumptions that you're making? What are things that might break on the governance side?

What are things that might break on the Oracle side? And the, so like there is one sort of group, I mean, Blue Chip, that's starting to do that. And I think it also probably makes sense to either work with them or have another that's and basically also try to cover the more kind of defi-native and the more decentralized end of the spectrum. And if we have our own pre-existing frameworks where there's clear sites that you can go to and you can see these are the relatively safe stablecoins and these are the crazy ones, then if one of the crazy ones breaks, then our community will have a much stronger position to respond to that situation from. So that's one thing that I think we could do and actually get started on more now to try to make sure things don't get worse in that regard.

I think we have one more question. Time for one more question. Thanks. Might be a little dumb question, but I find the argument that decentralized stablecoin adoption is low because yield is low is sort of a very supply-side focused argument. For me, when I use these stable coins, sometimes I just think of them as like a substitute for a dollar, but no chain.

And in that aspect, there's no notion of yield. Like when I use that, it's just something that I use. So I wonder what or how should I think about, suppose something is very convenient to use and there's a lot of consumer demand for like the circulating ones. Does that create pressure to generate more adoption on the supply side or are they completely separate? Like I haven't thought about this, so I'm not sure what to think about.

I see. So I guess the question is, can you motivate people to hold these stablecoins with convenience rather than with yield? I think it's a valid point, though I think you also want to kind of game out what this will look like five years from now or 10 years from now. And the one outcome I want to think about is the possibility that the user paying with one thing and the recipient getting another thing is something that's going to be as frictionless in the Ethereum world as it is with credit cards and with 10 times lower spreads. So that's one thing that might change.

And in that case, convenience would not really be, the degree to which convenience is a property of an asset would decrease, because it's like every asset would have similar convenience properties. The, yeah, I guess, you know, when I, I mean, realistically, like, when I think about why, like, I hold, hold USDC, it's like basically two reasons, right? One is just laziness, like, I receive it, and then I know that I'm gonna have future expenses in it. And then the, and like if there was infrastructure to let me swap it into like dollars that I could trust that would have an expected yield greater than zero, I probably would. And then the other reason is just that like, USDC is what Polymarket accepts, right?

Thank you so much.