Transforming Monolithic Protocol into Rollup Centric Platform with AI Risk Curators by 0xmikko eth

Transcript

Hello. Hello to everyone. Uh, nice to meet you. Thanks for coming. Uh, my name is Mikuel. I'm from Gearbox Protocol. I'm co-founder and CTO. And today I want to talk a speech regarding the protocol because what we really evidence now what we can see in the market and everywhere it's a big shift related to things and how the focus change from the protocols to end product and in this speech I want to explain our movement and understandment how to make a better app how to really attract users and what we did to make a

real rollup ccentric application. So as we can see many years I heard a lot of different stuff regarding mass adoption. People talked a lot that when new people came to the market all protocols all defy will grow and maybe we take a wrong turn but now as you can see a lot of retail users unfortunately came to Salana. We have some lack of smart money etc. But by the way there are two significant shifts in perception. Users are not feeling to go deeper in your technology. They are not interested to really learn a lot of complexity stuff

like bridges etc. Each stuff could be simplified as much as we can. And in our brainstorm, we asked ourselves what we can do to make lensing even simply how lensing could be simplified to really make very good very smooth experience for end users. And the idea is to make something which happened in web two. So previously as you know to get any credit you should sign a contract and it was a over complicated process but everything was changed with inventation of credit card and in our

perception we can really make lensing very very simple useful and understandable experience like in web two if we can really distinguish all accounts on Ethereum between two possible ways debit accounts what we have now today it's eway it's account abstraction it's like safe wallet doesn't matter and credit account accounts when you can really borrow money against asset on it and it was like a innovation in gearbox in 2021 and we are continuously moving forward and the core idea is to build a lending protocol based on account abstraction hooks so how it work in a few words the

first part when user interact with wallet. We can really store somewhere that there is some depth or not. And to make it possible in comparison with normal lensing uh protocol which get some asset as a collateral and give you another asset. The logic here works in a different way. We just have a hook which should be executed all time when user interact with its account and if it could execute it after each operation we can really be ensured that there is enough money to pay the collateral debt back and in this case all assets on this account as you can see on my example it's one is it's

1,000 USDC and one10 of BTC are still liquid. So you can use it across multiple protocol. You can put is into you can put USDC into farm. Do whatever you want. It's totally free. There is only one limitation. After your operation, we should be insured that it's enough money to pay the debt back. Very simple logic and it works similarly to credit card because there is no special interfaces. You can use this solution across multiple protocols. If you have more money, you can buy something like NFTTS, pay for conferences everywhere. Very simple rule, but to make it really useful

across multiple chains because today users say, "Hey guys, it's a pretty good to have a protocol, but I am not interesting to learn about networks you provide etc. I want the similar use like you have a visa. It doesn't matter who issue your card, which bank, it accepted everywhere. And if you go here, yeah, it's like a one of transaction example for people who are tax savvy. So literally if you can see a batch and we have a multi- uh we have a multi execution there. So we have a batch of approval and swap on unis swap. And as you can see previously we just keep in mind that there is one

approval then we execute batch as usual and then we do all this check. And as I said people are really not interesting to over complex products. Today we have a lot of defy applications more designed for mass and tax savvy users. And in this case it could be like a magic. You have a debit credit normal account like account abstraction. Then in one click we set up hooks and then you can use this money as previously but then you can borrow against them. And this create a good example. For example, you hold crypto, you believe that Bitcoin, Ethereum, etc.

will grow tomorrow. You can really borrow stable coins against them and spend in the grocery market if you connect a uh credit card. Pretty good example, good from taxation, good from UX. But users, as I already said in mass adoption are not interesting and they ask what chains you deployed, how I can use it. So at this Ka at this time like we spent last half year we make a transformation to become a real landing infrastructure and this transformation called to move to rollup centric application because as we can see now the value for normal retail users move

from Ethereum mainet to multiple L2s like base now mega is coming some others who really attract real users built a cool interfaces but people easily pretty not interesting to go into details to make this application and very clear logic for end users between debit account and credit account not depending on any chain we built the things which is called rollup centric application so the logic here is to have these four components the first one we want to be permissionless deploy on any network. If you tomorrow will decide to launch your roll up, please welcome. You can really

enable credit accounts on your chain in one click. No allowances, nothing to do there. Another part, it's a interconnected way of liquidity. All Ethereum L2 uh ecosystem suffers from liquidity fragmentation. It's a big big problem. A lot of funds are on mainet but users on different L2s and if you can find a way how we can interconnect things and make an really crosschain lending protocol it could be uh much much easier. I will talk a little bit later then permissionless market management when everyone can start their own markets and build from gearbox legal blocks any protocol or any

financial primitive organization and chain abstraction. People should not really focus what chains they are. You go to application, you want to really maybe get a leverage for any asset. One click and it's done. And to make it possible, we built the architecture which built on a few components. So it's a quite unique things and I think it's a good playbook for other protocols to become a rollup centric. So on mainet we have a like basic knowledge base and we store two things. The first part it's a onchain dow decision. So each time when da was for any decision it's go to a

crosschain multisk. It's the way how we can really uh send the information about decision across multiple connected chains to everywhere. To achieve that we really use a signatures which could be verified here and thanks for create two and deterministic deployment we can check cryptographically which decision was made for chains which have a better connection we can use bridges and it's also implemented at the moment and these decisions comes to a really routter on each chain which called inter instance manager and this instance manager manage

all decision but it's not like uh the things we build we have a another layers here and we limit DAO possibilities to really be involved in any chain activity. So DAO itself could only add some things regarding to fees. They could really add some code and updates. But if you go a little bit further to make possible to deploy the protocol without any permissions on any chain and be ensured that there is no uh incorrect code, there is no really problems related to uh misconfiguration etc. We build

additional role which is called instance manager. But uh there are two steps we control each time. The first part is bite code repository. As I mentioned all contract which could be used in deployment or in real usage should be signed by auditors. So really it's a good way because you can ch prove onchain that the contract works correctly. It's not malicious. And as you know previously it was a lot of hacks when uh people really full end users provising a small change. You never can go to either scan and find that someone make a small small change in a one formula. It's a pretty hard for

mass adoption. Now you can be one person uh insured that you work with normal and authentic uh contract. Another part that each contract is protected by auditors. So you shouldn't go to GitHub to check the code to compare etc. We work with a big institutionals and all time when they start to uh check gearbox they have an internal team we check our contract they say okay guys give us our audits we want to compare the code you deployed is what you audit etc etc. Here it's done automatically. So really one click you deploy on new network then risk curators could be done any setup permissionlessly

and every person users risk curators and other managers could be 100% ensured that the code is correct and it's create a open new market for uh external developers which is quite good because it's no one interest model fits all problem. If you want to build any block any additional plug-in for gearbox like interest rate model price feeds maybe adapter loss policy it's up to you the same logic you should not ask DAO for permission to launch your code just got an audit put it into the system and it's enough and it's create a very safe way how we can really develop and grow

protocol from the developer standpoint and make it safe the same time because In morpho it was a lot of cases of misconfiguration when people lost around a few million dollars. Why? Because the court was pick commit someone really make a misconfiguration etc. In this case we really provide a warranty for every people which could be ch could be checked on chain that the code which we use is really what you need. And as you can see to achieve this we really make a big work to mark all contracts. So any contract in the system has its unique uh

this is unique version and contract type. They are really provided into this bite code repository and there is no really uh error or mistake that you can deploy something incorrectly. And if you can see the whole system, you can see that the risk curator, the people who can build permissionlessly any market on gearbox could trust the system because market configurator all core contracts are verified by auditors. Any new model also verified and price fees which is a price feed store verified by instance manager. So we really provide a 100% warranty that all parameters, all

things, all addresses are already verified and it's create another way how you can begin business and grow the ecosystem because a lot of things solve for you and it's a very safe way and if you really launch a network maybe it could be a problem for you to attract new liquidity. Maybe you want to really add this credit accounts etc. Sometimes it takes a lot of time when team could prepare all installation files, do this setup etc. With gearbox by code repository it's done in one click and really you should check everything on chain which is a good and uh working solution.

And the next part which also come from the logically from this way if we can dep be deployed on any chain how we can create even better experience for liquidity fragmentation and it's called liquidity graph as I said we can set up a different pools on multiple networks and if we have a really proven code we can really connect different pools with credit accounts and make a possibility for liquidity arbitrage or index rate arbitrage or arbitrage was the correct name to move liquidity between different chains if it

has an economic sense for example on this picture you can see there are a few different rollups A B and mainet and each rollup has a different interest rate so it works like a gradient when people on mainet could borrow some liquidity and move it to rollup A to get a 2% more interest rate and it's create this system which automatically move liquidity across multiple interconnected chain to the rollup which really provide better returns and in terms of RAS perception imagine you launch your network and of course it's more risky but if you provide a good interest rate

which could be uh attractable able for yield arbitrageers liquidity will move to your chain. So it's create a good market and it's connected different part of uh rollups in the whole Ethereum ecosystem. And the next part what I really believe we could build in this rollup architecture is the way how we can manage risk. Today it's a very complicated process. So now we work with the cow labs which is one of the best risk curator firms and they do a lot of manual work. However, on chain we have a lot of data which we can use to understand what's going on and we can

combine this data with things coming from social networks and create an AI and it's also work and we are working on it which is called Giri to really analyze the the things happens across the whole L2 network in Twitter social networks news etc to manage parameters. Today a lot of protocols try to say hey we are going to build a system which could really manage all this stuff. I think on a current level of technology it's a very ambitious task but if we move a little bit further we can at least create a second opinion which could be valuable and the task we perceive as significant here two things.

The first part is to provide alerts when thing happens. Today even these big risk firms has like I don't know maybe five 10 analysts who check what's going on but with growing of L2 ecosystem this events could be much much more and it's a very very hard to be 24 to7 to understand what's going on. If we build a system which analyze all all events happens across multiple chains combining them with market news de proposals etc and put it into block which calls chain intelligence we can really have two benefits. The first one we can alert or highlight potential problems. No one

could really get a lot of information and analyze it significantly fast. But if the system could help risk curator to say hey this problems on this chain this protocol potentially got hacked or etc. It's create a good way to protect and make decision faster. And another part is simulation when we can collect a lot of data when we can really have some kind of uh previous and historical data we can make a simulation based on things. What happened if I reduce LTV? How this change or event could affect the whole system? Because when we build a really big L2 interconnected

system and have a lot of liquidity, some small change could really go to very bad consequences in terms of liquidation, liquidity shortage, etc. And if we can really simulate this stuff, if we can use AI to predict different solutions of any decision, it could also help users to really manage the whole ecosystem. Now, Gearbox is deployed on five networks, but I believe and it's doable task. When we talk about 1,000 rollups, it's a pretty hard to really manage and analyze everything. So I believe it's a very good first step to autonomous risk management. I do not talk that tomorrow

we can build it the system which can do it autonomously. But the whole picture is to have the way to manage this lensing market and credit accounts which is simply works as a credit card for end user. move liquidity automatically between the whole ecosystem to provide and deliver it to the place when it's needed more because the interest rate is higher and make at least a second opinion for risk curators who can manage this big system and now I definitely believe if we try today to attract much more users today we have a free block

space and many people say Hey guys, it's a very very simple to attract more users to Ethereum because as you can see multiple L2s, multiple opportunities to put more code, more protocols etc. But ask yourself what happens if the price will drop today significantly. It was happened when unfortunately serum price moved from 2,200 to I think it was 1,800 a month ago or so and the gas came to the peak volume like a 100 way. How many accounts we have today on our 1,000 accounts on Morpher? The same.

So literally all lensing markets has not enough many users and even with a big price change our system also gets stuck. Imagine if we have a like 1 million positions it's not enough capacity to produce all things like liquidations exchanges etc in case of big price change. and we try to build the system which could really help to parallelize to analyze the system and make things even faster. So in current market I know maybe it's a very optimistic to say hey when when mass adoption will really be implemented

people will come and use the product but I definitely believe that simplification and better UX could really attract more users make the product better and on backend side we should build much more technology to support this movement. Thanks so much. Maybe some question. Hello. Hello.

um for the agentic you know the AI management you know component of this um like what is like how do you envision that that model and like who is going to be like responsible for it like in general one of the things that I'm sort of critical of when it comes to AI as it pertains to to these kinds of systems is that implicitly people don't really understand exactly how the AI is going to make certain decisions and you When it comes to regular consumer um interactions with AI models, we generally refer to sort of unexpected responses as being hallucinations, which is kind of a a naive way to describe those things because the AI model is

only capable of producing things that you know it's it's trained on and it's only you know our subjective interpretation of what it's doing that qualifies something as a hallucination. Um but with respect to that sort of behavior or that kind of unexpected behavior, you know, how do you envision a an AI model coming about that doesn't suffer from those kinds of things or is at least able to justify the decisions that it's making in a sort of human, you know, understandable way. Uh thanks. It's a great question and I think it's a good parallel with self-driving cars.

For a long years, people do not trust and ask and in some countries it is still the law that you should have a driver and the system and the driver takes the responsibility to press the brake in case of emergency. And what we really try to do is a second opinion the system which could highlight some alerts because the whole amount of information could not be analyzed even now we have so many networks today. The second part I also believe it should be some kind of limits which we also use in many times in gearbox. So you can allow the system

to maybe change LCV or any risk parameter for not more than 2 percentage. So it's like a small change which could be done continuously and if you combine with alerts it's create a semi uh semi management because the system could really make small changes automatically and just report you. But in terms of alert or very significant decision of course people could double check at least that the problem exist or something happens. Unfortunately, it's a very difficult part because even if you go to current financial organization and check and read all the things there, there is also like a lack of

responsibility and we can see multiple crypto organization which say hey guys you can remember a lot of sexes which take all money and disappeared. Yeah. So we take a zero responsibility for that. So but in my perception at this stage we are not ready to move everything to AI. Let's keep real humans who understand but we can automate things and make it faster and understandable. So it's what we try to deliver. We should not replace this risk curators with huge experience must degree and good understanding of things happens. And the second position that the some information regarding to market

come from the hidden sources. It's another big problem in AI. We can take everything onchain. We can analyze maybe Twitter news etc. But some very good financial organization has some direct access about what was talk in the conference what people think etc. It's not the trackable information anymore, but it has a lot of impact on final decisions. So, at this case, I believe it's just a good helper, but not a replacement. Any questions?

No questions. Thank you so much. Have a nice day.