You are not alone

Transcript

Okay. Hearing me? Uh I'm the principal engineer at Cambrian. Uh the next 10 minutes I will be trying to explain what we like from Ethereum. And then the thing that's like like is a very tiny design choice in the EVM and that exposure turns into a multi-billion dollar ecosystem. But uh I believe the potential of that tiny design choice uh is not realized yet. There's a huge gap and we call it the intelligence gap.

So, at Cambrian we are we trying to solve that problem, the intelligence gap. Uh I believe I can click this, right? So, I will start by the idea, the tiny EVM choice. So, uh the control theory like basically explains what is observability. Says that if you can uh get the most of the system state uh out of its outputs, actually it is called observability. So, I believe uh Ethereum didn't just do the execution part, optimizing the execution part. Actually, it's it's also optimized the uh exposure.

So, what is it that basically we have some really low-level opcodes called logs, okay? Log zero to log four, uh which is base for the solidity events. It's like whenever you do a swap, whenever you change a liquidity, actually you uh emit an event in the solidity contracts. But it makes it so easy that easily other part uh other other parts like indexers, analytics, aggregators can easily watch it and collect the data.

There is another uh example. Uh I'm not saying that Solana is bad, but the the the there's a difference because actually if you want to have that state, the like how much swap has done, what is the state of the liquidity pool, you need to go really in deep. You need to find out the instructions. Uh there are some specialized log strings. Even I have seen ZK compressed uh data states inside the Solana. So, you need to go deeper and it's not easy to integrate.

So, that difference really matters because when the integration is easy, the other ecosystems can uh basically do what you cannot do by yourself. Uh Okay. So, there's a proof of it, by the way. With numbers. So, probably most of you know Uniswap V3. Uh basically Uniswap built the liquidity pool idea. If you ask me, like it's the greatest thing that's done in EVM.

The idea is you basically lock some tokens and people come and do the swap. But it seems like basic, but Uniswap also designed their exposure really well. Their state really well. So, others can came in and build what is not built inside the Uniswap system. And one of them is which I worked for a very few weeks is an aggregator, Oddos. Uh they just watch the state, know which liquidity pools have enough liquidity to do the swap and create a I'm from an industrial uh engineering background, create a really great solution to traveling salesman problem. Actually, you you try to find the best path. Okay?

Uh usually if you go for other aggregators, they have one hop or two hop. Uh by hop I mean you go for a one liquidity pool to another, then you finish the swap. But uh due to the exposure of the liquidity pools, Oddos has the capability to do it like 10 or 12 times. And it generated 40 billion dollar volume in the first year over Uniswap. So, as you see, this observability bringing a huge exposure and that exposure turn into a huge volume.

But there is a problem. There's a button it. Why? Because this data, which we call event data, is not enough for generating intelligence. By intelligence, your protocol cannot decide on it. Why? Because it needs historic data. Like whenever you want to have intelligence like analytics, it's nearly impossible to execute it. So, it creates a huge gap. If you ask me, it's we call it like basically intelligence gap. So, I will try to explain it in in a very simple example. So, in the left side, you see that someone borrowed 10 ETH, okay? So, this data can be used by liquidity pools, okay?

But whenever you want to have something like is the market safe to do some let's say another uh lending or thing based on the idea, you can't do it because why? If you want to do it on chain, uh you will spend thousands of gas. Basically, you will lost in it. It will kill it. And if you want to go off chain, which is oracles, yes, now you lost the trust idea because the oracles you basically looking for an AWS lambda or a centralized database.

So, in Cambrian we really want to solve this problem. Uh we have been working for a year. Uh so, I will try to explain what we have finished. I feel like everybody's liking animations. So, the idea is like we don't want to introduce a new oracle. Basically, we want to reinvent the stack as much as we can. Uh I'm coming from like I'm an industrial engineer, but I worked for more than two decades on databases, how quorums work, how uh you have a cluster of data. So, we understood that we need to have something called verifiable data network. So, each node is like collecting the intelligence, the data,

and agree on that they have the same data. This is the basic idea a blockchain has. So, we introduced that trust back to the into the to the analytics. So, the next thing is we understood that the the idea about certainty, the trust is differ from application to application. If you're doing swaps in few dollars or let's say your volume is small, you don't need to run all the network like thousands of nodes to get get into a quorum for that for a number. You basically request let's say two or three is more than enough. This creates a cost efficiency. So, it is not like you ask the network a data and then you

basically charge you thousands. You can easily say, oh okay, I'm happy with two two or three or I'm happy with like we also introduce slashing saying that, okay, uh I want to run this uh I I can't I don't want to say query. I don't want to say uh data. I basically it's intelligence. I want to have this intelligence uh by slashing, for example. The next thing is uh it is not here, but the idea is we understood that like the protocols is not just on chain. They have off-chain components.

Uh like web UIs, agents, even they have their own REST APIs. So, we bring in the standard API, the web two, the old, but also it helps you to get fast data from the system without quorum. And lastly, uh at day zero we understood that we shouldn't be doing anything for the let's say uh developers to learn everything inside this network. The how to create this pipeline. So, we try to uh achieve this with agents.

So, the next two minutes I'm going to show this as a demo that I recorded. Uh So, this is the VS code with Claude. So, I asked it to implement a full undercollateralized lending protocol. The problem about undercollateralized lending protocol is like you basically you have uh less collateral for the same amount of money that's borrowed. And you you need to create some risk calculation out of it based on whatever you have. And funny enough, actually if you ask like this, it will create the uh risk calculation based on the amount you

borrow, which is like overcollateralized. So, it's useless. It's static. It's like 80 No, 80 percentage of the LTV or something. So, I come up with another idea. Oh, it goes really fast. Uh I come up with a new formula, which is called pulse, which is based on historic data. Uh I need to look at this It's based on I I'm not the risk actor, so it's based on frequency, severity, velocity, and concentration. So, I asked it to be implemented at Cambrian. So, the agent, our Cambrian agent, which is which can understand the formulation and generate a pipeline, data pipeline for it, and create the required changes inside the protocol.

Uh the last one is like as you see, the quorum, like especially when you implement this is initially implemented by one. So I wanted to have at least two data nodes to execute this query or data or intelligence. So it updates and finished it. This is the language we try to build that is that can extract that information. And finally actually I want to bind this to the web UI a user can get in and do the lending. So this is the output the result we we basically copied from Morpho. Sorry for that Morpho guys. So

whenever you connect your wallet actually as you see the pulse risk calculation is directly there. Yep. So this is the end. And lastly thank you. The idea about you are not alone. We want to show you the the exposure you create in your protocols can turn into a multi-billion dollar projects. So you are not alone. Please design with structure and keep in mind that that exposure will help you gather more growth.

And the next one is till now the intelligence part is missing in on-chain execution. So you're not alone. We basically bring it with Cambrian. Thank you. That's it. Wow. Any questions? Probably there will be any questions. >> [laughter] >> Yep. Um does it work? Oh okay. Uh yeah thank you for the presentation. Actually I wanted to ask what do you think of verifiable logs ERC which seems to be achieving pretty much the same goal just on the like on a base layer of the network.

Yep. Uh so verifiable logs is like basically while I was working in graph actually I seen them. Uh it is some part of it basically but when you build over it that's the most important part. You need to execute it. You need to calculate the numbers out of that log even you you keep it in your protocol. The problem here is if you calculate a huge average weighted average. Have you ever seen the numbers actually let's say let's go for a Uniswap USDC BTC pool. Okay. It's like millions. So even calculated per log will cost too much.

Uh another idea is sometimes you can't do the M1 plus state. That is something like for example summation is something like this. So you got the sum till now. So then the new log will introduce another sum. Okay you can optimize this gas but you can't optimize uh weighted average or standard deviation. And also the intelligence about like this like usually you also introduce some external factors. This is not just EVM events. This is something like states instructions and maybe an exchange rate that you trust in your protocol. So you need to introduce it. Yeah it will also

act like an oracle. So I hope it answers your question. Uh probably the missing key I while I was when I gave access to chat GPT or Gemini like if you're an investor or a developer what would be your question is the black box. I didn't explain this. The black box is like basically we materialize the data. The difference is usually when you go for an oracle they will execute the query which will create a huge delay. The summation the average. So we materialize it per block. That is the idea. So your data will be always available. It is not based on your query. It's it's not based on your request.

That's it. Thank you.