The Centralized Risk of DeFi - Guillaume Felley

so uh we're gonna shift a little bit we're gonna keep it into in d5 um but we're gonna talk about something that um so basically if you listen to your typical bitcoin maxi they will tell you that d5 is actually just decentralization theater of course this is complete um there are many important elements of d5 that are in fact sufficiently trust minimized to make them credibly neutral permissionless immutable all that good stuff that we've come to love with you know ethereum and of itself however that's not to say that there there is not um risks with with centralization um you know this could include uh you know problems with uh vulnerabilities in terms of smart contract upgrades um you know admin keys things like that and even that down the tech stack in terms of other elements of of the daf so you know d5 is going to get more decentralized over time our next speaker guillaume has some great insights into uh sort of the centralization uh risks perhaps and especially a first-hand perspective working on a uh a collateral uh uh token or stable coin for for d5 so with that i'll give it a uh give it to you all right thanks for the intro okay so yeah my name is girom and um i'm with reflexolabs and thanks for coming to my talk and uh i will present the centralized risk of d5 all right so um the agenda so the first we're going to speak about the state of the phi particle governance then the rise of the custodial risk within defy and uh and last the our problems with the our dependency on the us dollar all right so first a bit of history so they used to be back in the days around 2017 2018 a very hot debate um about on-chain versus off-chain governance so at the time there was this new protocol called teasers and they were doing something pretty new that was um coin voting so on chain voting for protocol upgrades and um the the ethereum community and the ethereum leadership at the time was was rather skeptical of those and um of the approach and uh yeah so like those two are two blog posts from from italy at the time where he explained some of the issue with uh with uh unchained on chain voting orange and governance if you call it and uh yeah not only mentions uh yeah entrance voting is vulnerable to economic attacks and uh it's uh fundamentally a plutocratic system that is governed that will ultimately be governed by the rich so yet a few years later in in 2021 uh this debate is not really going on anymore and pretty much every protocol is now doing on-chain governance but the issue that were mentioned back in the days are are still are still true today right okay so um i want to look at some of the issues with governance and the first i want to look at is governance as an attack vector right and for that we need to understand what can what can the governance which is usually implemented as a smart contract can do so in many protocol um governance can completely update the logic of the smart contract and yeah like do whatever they want they can control the the oracles and do things like setting the price to zero or setting the price to infinity uh in some protocol again you can like governance can cease to all the totality of user collateral you can mean an unlimited amount of governance token you can exacerbate protocol fees you can shut down the system you can and yeah and this is uh this is an example of um of such governance attacks so earlier this year in in october someone took a flash loan mkr flash loan on avi for seven million dollars and uh voted on the proposal within that flash loan made the proposal pass and repay the flash loan so um i think in this specific example like fortunately the i think the attacker more wanted to prove to prove a point than really like still still uses fans and yeah so it shows that this attack can happen right um i think since that maker took some measure to to mitigate the this kind of issues but yeah let's not forget that like this protocol handle billions of dollars right and that you need to you need to only one mistake in your in your design and all of this can be gone all right so not another category of issues with ancient governance is what i call governance fatigues so often you have low participation like less than 10 is is pretty common and um moreover you have like delegate schemes where you end up very few voters on on proposal right like the delegates have a lot of a lot of votes and and yeah like you can never really really know how much collusion there is under the hood right whales and insider my control of my my controlled vote and um in the long run this kind of thing tends to to centralize uh it lowers social scalability so this is an essay from quite a bit old they say from nick sabo which is still super super relevant today and um yeah in this say he explains that the faith in machine machines scout better than the fate in human and that's really for him like the big breakthrough behind bitcoin it's not the technology but is the fact that it's immutable you can't change it and and yeah like governance goes against that uh that that trend and um yeah so typically example like it will lead in internal conflict within the community and um one one example that that i have is that uh yeah like for example uh with maker dao when they made the decision to add collaterals such as usdc uh they had very good reason to do so but um this this led this um became a frustration within a part of the community and uh yeah in the last point it's controlled as a liability so uh governance as i showed in my previous slide is an attack vector for um that can be targeted for to to be exploited but it's also a point of control for regulation right if a regulator comes in the easiest way is to control governance so what um i want to discuss what really needs governance in defy right and uh it comes down to what really really needs human input um oracle is a good one so this is a very critical part of a protocol if your protocol needs an oracle you will need some form or governance to monitor it or take action if this oracle gets gets malicious and there is no really ways around that and more broadly you need governance for external dependencies right if your protocol relies on another protocol something might go wrong on that hand and you you might need you might want the governance to keep a bit of control over over those dependencies protocol kill switch so basically a function that can stop enhance your protocol um it's fine to have that early on but i believe that in the long run uh like the governance should revoke such kind of abilities the community treasury so if you want to give out grant to develop i don't know tooling around your protocol or front ends this is fine but yeah even with community like treasury if you have to allocate funds to for example to some i don't know keeper network you can have some scheme where it is incentivized and you have automatic payouts and the cash flow protocol is automatically allocated so protocol parameters like collateralization factors or liquidity provider fee and ideally those should be immutable as well and what you might want to consider governance for is ipear parameters so parameters that set the parameters and this would this is like a higher level where that you would have to touch less often upgradeability so i think governance a lot of protocols have governance over some credibility of the protocol i i believe like this is not necessary and um what what i prefer and i think the best way to go is like opt-in upgradability and a typical example of that i have is the way uniswap transitioned from v1 to v2 v1 was not shut down or the logic was not replaced like it was up to the liquidity provider to migrate from v1 from v2 if they want it and v1 is still around and if anyone didn't like v2 it could stay it could have stayed in v1 i mean you can still stay and provide liquidity on v1 if you want and this way it's much more socially scalable last uh if you have some very experimental module or concepts um you might want to keep some kind of governance um just for for safety in the first in the first phase but like in the long run it should be it should be immutable and you can change it um the last the one quick note like um the price of governance tokens has been growing like like crazy over the last year and over the last and there is such a strong market demand for for this kind of token that it's a lot of pressure on on product on default teams to to issue a token that they might have investors at once or simply just the community that wants to to to invest into the token so that's that's a reason why all those governance scheme are not always well justified your thought through i think so yeah so switching gears a little bit and going to the second part of my talk about the the rise of the counterparty risk in d5 this is something that we're all familiar of stable coin supply is rising uh enormously and the tesla and usdc is leading the way so those are custodian and centralized tokenized u.s dollars basically so they're built on top of a bank and to me this is not really defy um like um we have all ops and i mean i still have big hopes and assets such as die will succeed so and and yeah so this is from november last year they die and from mecca just crossed one billion and a and i believe now they are even closer to two billion so it has been going very very fast as well but uh we have to put things in perspective that it's still a minor share of the stablecoin market around six percent of the supply it's stabilized since september so yeah it's not it's not taking over and like centralized stablecoin are not slowing down and to that i would add that like now part of the supply of dye is by by the same competitor that i mean that dye is competing against such as usdc which is a custodian collateral okay so um the next part of the next part is about i want to quickly talk about the dependency on the us dollar and to me this is something that that is a bit weird or disturbing like when i started to get interested in d5 you had all this idea about self-sovereignty this intermediation code is low and all this this concept and nowadays i feel like what basically what defy runs on is centralized and custodial us dollars and not only is custodial but then why why us dollars right like uh yeah we it's uh it's it's it's surprising to me and um one one i mean the obvious definition are like yeah it's the default unit of account for many of us there is a strong need from from the market right for for a low volatility uncorrelated asset when you are in the wild west of cryptocurrencies and there is simply no better alternative and um yeah and what what are what is the problem with that right like this is something not new right uh the the valuation of the us dollar um like it's kind of back to where we're where we started with bitcoin right where um yeah it's not this this is not going to stop it's here to stay the the us dollar will keep being devaluated and not and um second big problem is regulation right so the for example the american government is increasingly going after uh stable coins or anything that is pegged to the us dollar they want to classify them as securities and they want to make it harder for those issuer to to operate so it might be in the future that uh we run into issues and like stablecon operators are not able to operate as easily as they can as they are doing now so yeah the the stable act recently is uh an example of that and i believe we will see more of of this kind of move in the future so um the rise of custodial assets um the the need for governance and the problem of the us dollar like this comes to the part of my talk where i want to speak about reflex labs obviously and what we're building so uh we are we are building rye uh rye is at its origin a fork of maker dao mcd so the fork of die basically and um it is free floating um it is not pegged to one us dollar and it's just low volatility but it will not be pegged to the us dollar this price might drift over time and it might change but the goal is to have it low volatility we put a lot of emphasis on having it governance minimized meaning that um it will be we want it to be as immutable as possible it will be very hard to change or like this like the parameters over which governance will have control will be greatly reduced and one example is that for for example for setting the stability fee which is one of the core mechanism to to regulate the pre the price in mecca dial which is set manually for us it will be set algorithmically by your controller um i will explain a bit about what is that later and uh yeah it's it's meant to be the the unit of account for d5 so used as collateral maybe in compound or even even even for maker a strong decentralized and decentralized collateral that is that is not native to to defy and uh it will only be back uh by ether and the the reason is that we want the purest and the more decentralized as possible collateral backing rai in order to achieve this governance minimized aspects so maintaining stability um we apply um something called control theory which is a subfield of uh engineering that is about uh algorithm and yeah algorithm to to to control dynamical systems uh it's everywhere it's all around us uh cruise control of cars heating system industrial systems and yeah plenty of things maybe power supply and and many more and uh we are pioneering that bri applying those concepts to to crypto economic systems and more um more precisely to set the the interest rate of our within our system uh so if you study engineering this kind of diagram will be familiar to you this is the high level description of a pid controller in our case the the system will be rye itself the sensor is the is the market so what is the price of the asset what is the price of fry and by with that you you feed that into the controller that will choose an interest rate accordingly uh in order to achieve its goal of having low low volatility and this loop goes on goes on forever all right so ry we successfully run a test that we call pry on on e3 mainnet we peaked at 600 k tvl and the pid controller successfully stabilized the price the price of pry during the test net that's what you can see on on the chart at the bottom right um so the blue the blue curve is the is the market price and the orange one is the is the target the target price or redemption price and we tested global settlement on it so we wish like triggered the emergency shutdown and uh everybody got his fair share of um of collateral against pry the real rye uh will launch this month so um yeah join uh join join the community and uh we are on discord twitter and this is the link to our docs and uh yeah please get in touch with me or anyone from the team on discord or twitch thank you very much thank you yum appreciate it um super cool project yeah guys look at their info um i think this is going to be something that's going to be increasingly used in d5 in the nearest room here and thanks again have a good one thanks