Building Resilient DeFi: How Protocol Design Shapes User Protection and Performance

Good morning. Hello. Hello guys. Good morning. Um, do you want to start with a brief introduction of yourselves?

Yes. I go first. You go first. Hi everyone. My name is Alio.

I work in uh a labs. I have uh been in a since the beginning 2017 and I am one of the uh lead developers of the protocol. I we were um leading development the development of a v1 v2 I led uh the v3 and now I designed before so we are currently building uh close to release let's say the version four of the of the protocol yeah I'm I'm Martin I'm doing my role is control is speedy but I've been a full stack dev in the past and I even got into smart contract coding and Um uh I'm I'm not developing the new uh landing market but I'm uh yeah have a lot of experience but because I'm like 10 year in this space and I think a lot about our landing market was like isolated lending market um llama land and uh I did setup market and set params and make simulations to find the best params and these kind of things and yeah great so we're going to kick it because you guys are both lending protocols about talking about your collateralization process. Um with AV4, what have you decided to do differently and with uh Llama Land, what what are you guys how did you decide to do what you do for your collateralization? Want to start?

Yeah. So, because our market is in theory it's permissionless. It's just that it's not like Yeah. There's so many parents and so you need to know how it works. So, we did we never did see anyone setting up their own market.

So we help them. So if somebody is like approaching us, what we do is we we we make some simulations with existing data and look at the volatility of the asset basically and then we we we we find out what's the best parameters for this asset and because it's isolated. Yeah. Just look at the one I asset we have. And one thing is like crucial is to see how much liquidity is in the market because in isolated market you can also like list quite exotic um assets and we also had some issues sometimes with assets who then the liquidity has been too thin and then yeah backfired somehow.

So it's it's delicate process but um yeah mo mostly it's inbound. simply ask us we we we advise them and we help them to open the market basically. Yeah. In our case I mean uh the the process kind of starts from the uh the way the protocol works right. So uh this process is also in place in a so there are risk managers that do exactly the same.

So uh they evaluate what kind of collaterals are um high quality enough to be listed on a um but not only that uh there there are other risks uh related to choosing what kind of collaterals you you might want to allow people to collateralize. Centralization risk is one uh um if there are uh entities for example with a large amount of token. So if there are token locks and which might affect uh liquidity over time in the market. So uh you might be at a stage where the uh the circul circulating supply of a token is relatively small. Therefore the market is healthy.

But then if there are huge unlocks over time, this might uh uh progressively deplete the the market or it might open up in uh situations where uh uh large token holders might create nonliquidable positions. So these are also aspects that are kept in consideration. But then also like the a model of a is to is pool model with aggregated liquidity which is very capital efficient but is also very important to keep healthy because like a um bad collateral could potentially spread bad depth to to the rest of the assets. So there are a v variety of levers that the dowo can move to prevent this from happening. Speaking of DAO, you both have different governance models.

curve is famous for VE curve. A is governance for risk parameters. Um do you want to speak more about that and why how those governance models dictates I guess the risk factors your risk management. So on the lending market the DAO is not involved basically. So we don't go over the DA or any DA it's to change things afterwards.

Yeah, we need DS but to initiate the market there's no the DA is not needed basically the parameters are not immutable right in some are some are immutable some are not and we we are currently thinking oh we're doing a new iteration a version two basically so we are now in the process evalate what we experienced in the last since it's out and how we going to iterate on this and uh so we go to have more params to to change because we see yeah we need to govern more we have to need to or change things if yeah volatility of a pair changes or the market changes or instance happened and then we we can more interact with the market and then the D is needed basically. Uh yeah in our case uh uh we do have like two risk uh managers risk providers uh which uh they continuously monitor the the system and validate check and validate the health of the various collaterals and then eventually adjust parameters accordingly. Um in V4 we will have like this process can be tedious sometimes on the on the avid especially when when you have like tens of markets across multiple networks. So if you have a a collateral or an asset that is listed on 12 networks and you want to adjust the risk parameters you might have to adjust like risk parameters on 12 different networks and 12 markets. So in a v4 we will have some improvements on that side.

So some of the parameters can be changed on the fly only for new users which is an occurrence that happens very often. Sometimes like risk managers they might want to check uh and change uh a risk configuration not necessarily for existing users but just for new users and say I don't know until yesterday uh USDC was okay with a collateral factor of 80% but today is not good because the market condition have changed. So today we want 60 right. So in that scenario to with a v3 is not that easy to do because like changing the collateral factor of an asset affects everyone. In a v4 you risk managers will be able to change uh risk configuration only for new users which hopefully should make things more streamlined.

You can offload uh risk management to the risk providers without changing the uh trust assumptions of the protocol. Moving forward on that, excuse me, can we discuss your different mechanisms for liquidations? Um, I think Curve has an interesting methodology and obvious is moving toward that in V4. So, could we just describe what were the thoughts and processes behind the new kind of soft liquidation mechanism? Yes.

So, I think I did like think about this panel preparation. I think what's what what's important for the users basically to have time to react. It's not about how that's the goal basically give people time to act and our solution basically is that you have like collage a uh a zone where you basically tap in and then you getting partially liquidated but if the price recovers it it dequidates you and it's like a zone you are in and it if you enter this zone it gets quite expensive quite fast because every time if there's a lot of volatility you pay a lot but you still can basically save your position. If you just uh top it, you you pay back part of the depth and you have crease. In theory, you can go under that.

You can be fully soft liquidated and still have the position. And it's also it could also be seen as some kind of stop-loss basically or it's it's a it's a stop-loss um where you still have the position and if he recovers basically you you get back to the same place you are but you lose some of your health basically. So that's the feature but that cop this is basic give people time to react that's the main purpose and I think uh maker had like this one hour um time till the oracle updates and this is in a way similar because the goal is to have a solution for this. Yeah, in in a uh the liquidation engine is pretty much unchanged between V1, V2 and V3. Uh it proved to be very resilient.

Uh so the way it works is that in a there is this value that is called health factor and it tells you how healthy is your position. If your position is uh uh health factor of your position is above one, you're fine. If it goes below one, you will cry essentially. And uh uh so uh the moment that the the health factor drops below one like liquidators can uh perform liquidation. They can take up to 50% of the collateral of the user and perform liquidation.

This ours worked very well like he's gone through every possible market scenario that we have we have had from 2019 until now and um it worked very efficiently but it's uh we think there is margin for improvements. So in before we did apply some improvements to this uh so first improvement is that um uh liquidators will not be able to liquidate up to 50% of the position but only up to only be able to restore the health factor to a threshold decided by the governance. So if the governance says okay you know what let's not uh let liquidators to liquidate up to 50%. But only until health factor is one right or that governance can say can set 105 depending this is like will be a governance decision that will model it by risk uh managers. A second improvement is that the liquidation bonus is going to be like with a reverse Dutch auction.

So uh instead of starting it will start at the minimum when the health factor is one and then will progressively increase as the health factor drops. So if the position doesn't get liquidated immediately the uh the uh liquidation bonus increases to incentivize more liquidators to to liquidate the the position. And then like there would be also optimizations for liquidating multiple positions at the same time with the same pair. uh uh asset being liquidated and collateral plus a variety of minor other improvements. That's really interesting because Flashbots is uh working with chain link on their SVR on the OEV capture.

Um just just a plug for Flashbots. Um moving on to I think the crosschain aspect which I think is really interesting. You know we have a lot of fragmentation going on in EVM space. How does your chain sort of deal with that in terms of helping the users like access the different liquidity on various chains? Is this something that you consider?

Um and what are your plans for the future on this? Yeah, I think we we are quite mainetentric even if you have a lot of installation on other EVM chains because it's still the case that most revenue is made on mainet. Basically at the end uh we we have crosschain governance much more but it's not like yeah we have a crosschain governance because the in if we do fresh installation the installation is hold by is hold by the DAO basically but you know like unifying the liquidity crosschain we currently don't work on this are you planning to I think you know if you're like if you have product market fit you you you're cautious so basically you much more look what's is proven to work because otherwise you have to put a lot of uh resourcing something who is not like yet like a standard or anything like this. So yeah you don't have to move that fast so you can move slower than somebody just tries to penetrate the market right now basically. So we look what's going to be established and then yeah I guess we integrate at some point some things.

Yeah, in our case like we have this model of having a presence on uh every network was traction and uh it worked really well so far. So one of the things that we were thinking over time would have picked up traction is like crosschain lending borrowing. So I have my collateral on my network and then I want to borrow on another network. it really didn't or at least like we didn't observe any major uh demand for this feature. So in our v3 we have actually we were planning to do something similar.

We had like a feature was called portals which theoretically would allow us to do so. We never did it because first of all like the demand was not there. uh and like the one market per network model served the users pretty well already and also there are some security concerns related to doing so because like the moment that for example you access you allow I don't know a proof of stake uh like um validators based network to access uh Ethereum liquidity for example uh and that network has inherently lower uh security guarantees than Ethereum, then you are basically lowering the security of the Ethereum market exposing it to the risk of the of the of the network that is accessing liquidity. So there are some security concerns. Aido I think it got burned once with um uh I don't remember the name of the network.

One network was at the bridge was hacked. Yeah. And the and which one was that? F uh FTM. What's like Phantom?

Phantom. Phantom maybe. No, it was not Phantom. Yeah, because BF Phantom Phantom Harmony. Yeah, it was Harmony.

Yes. And I think that I lost uh a few millions because of that just because like the the bridge was hacked and not all the assets that were in a were bridged. So people managed to borrow using bridge assets very basically valueless. So that kind of left a mark in the DAO and it was not pursued like the this initiative of crosschain lending and borrowing V4 will still allow to do so eventually. We will see if it actually will come ever come to life or not.

Interesting. Um a theme of I think this conference has been that the institutions are coming. um how has a new customer base affected your decision-m or are you planning on creating features for them or you know are they part of your like profile currently? Yeah, if if you go more on the AMM side like basically it's we have quite a simple product based on the stable coin side. So it's not that we at the core I don't see anything we can basically improve for like institutions because yeah that's how it works and it works well in this test basically but what we can do is basically help them to see more what the pools are doing in these kind of things that build on top and help them to understand better what's going on basically and what we also think about what we see that we and now we also planning to now still we govern pool parents we change sometimes because if it's more liquidity here then we change make it more dense liquidity and now it's everything has to go over DAO but now we think about having dedicated people where we can basically assign and then they can uh change params on pool so we can have like some have pool creators basically which I guess should help us to to scale better basically and it would be also faster because now yeah D needs seven day and so I think it will open up the ways to to yeah make things more efficient basically that's yeah um yeah in our case like what we noticed is that there is demand for at the institutional level for a variety of things that defy today doesn't uh really serve that well uh one is um so compared to a today's a model like institutions really like risk diversification.

We were talking about this like uh behind the scenes. So uh really like yield and risk diversification be being able to uh position themselves in the a different place on the yield risk curve. Right? So this a today it allows it to do so but not that efficiently. One of the reasons is that it's because of the model because I think it's proven at this point that like a model with completely segregated liquidity and extreme heavy on risk uh diversification doesn't really scale like I we have seen many protocols in including the latest ones like morpho they really struggle in scaling while a is still printing all-time high uh even though they have liquidity mining and they uh have like deals and everything like they still struggle with scaling and the problem I think is the model so it's because like it just doesn't scale and uh but at the same time uh institutions like this particular um the possibility of diversifying across different risk tires.

So, AB V4 will still m maintain the same model as V3 but like uh it will create the possibility of diversifying risk by having uh pools of liquidity with different asset classes that can interle each other. So we can provide liquidity that is uh in a um in a pool and we can allow uh another pool to access this liquidity up to a certain ceiling to bootstrap the market and let it grow by itself. That is something that today we cannot do because whenever in a you instantiate a new market then you have to start from zero. So you have to find like a way of provision liquidity uh either with incentives or other other strategies and tomorrow we will be able to do this very easily and uh without constraints uh thanks to thanks to a v4 and this particular feature and also it will be because a v4 is going to be much more streamlined completely new design has nothing to do with v3 it's going to be much easier also to allow to do this permissionlessly compared to today where like a v3 is quite big. It's huge.

So it's hard to deploy it for example permissionlessly to create a permission infrastructure to allow access to a technology. So we're about under four minutes and we're at the stable summit. So I want to ask a stable coin question. We should get at least one in. Um, how has your stable coin in your protocol fit into I think your design mechanisms and what do you plan to do with it going forward?

Yes. So yeah, we have a stable account is is in a way a specialized lending protocol basically and we have curve USD and it's also where basically it was the third iteration of the llama land because llama land is just a yeah a similar product basically with few changes. So, so curfew is is for us very important basically because I think yeah it helps us it's about moneyess at the end you know if you can own your own stable coin and you have like uh basically as an exchange you can partly uh yeah nudge uh people to open pool with your own curved coin so you get uh some kind of um uh how say medium of exchange properties you can not force but you can bring into the your user and uh yeah that's and um now I lost track so what's the well how does it fit into your future road map yeah you know now basically yesterday Mitch did present YB and YB is like the next step for curve USD basically and it's kind of uh yeah I don't go into details But the idea is that you pay basically you pair um bitcoin or bluechip asset with curve USD and then leverage up and then you can offset the impermanent loss over this. So basically also there curious is is an important part and the main reason is curve USD is generating quite some revenue because all the minting right is going to the uh the uh vcurve holdership. So it's at the core of the strategy is curve USD and because we anyway our biggest market is stable coins.

So that's um yeah uh yeah in our case uh with go one of the things that we wanted to uh one of the reasons why uh go was launched is that uh we wanted to have a product that is uh behaves differently compared to other stable coins in a mainly in relation to predictability of the interest rates because like interest rates on a are very volatile. So you can end up having uh paying 18% today and after 2 minutes you pay three depending on where you are positioned on the interest rate uh curve. Uh while uh with go we purposely decide to have like governance uh decided interest rates. So like people using go and minting go they could have a certain degree of predictability of where the interest rates are and where they are going to be shortterm because like they are changes through governance process. So that was the main reason why some people say you know it's just obsolete because you don't know how to do variable interest rates.

No, we don't care. It's not that the reason like the the reason is I do want to have a different product otherwise like why would people mint go instead of just borrowing USDC right so uh but like the idea with go is much bigger and the plan is much more long-term because in reality the design of go allows like uh is very flexible allows not only to be minted on a but allows on a variety of strategies that can be built around it And like recently we we created a proposal on the unis swap dow to allow people to mean go against uh unis swap lp positions and that's only one of the ideas that we have in mind for the future. So scaling a stable coin is really hard like really really hard and it's one of the main uh uh it is a very long-term process. Um that's it. I think we're about out of time.

So, I just wanted to thank the panelists. Thanks. Thank you guys. Thanks.