Onchain Analysis and Trading in an AI Powered Future

Transcript

Good morning everyone. Uh I am Miguel Morell. I'm the founder and CEO of Arkham which is a blockchain intelligence and analytics company that also has a centralized exchange uh for spot and per attached to it. Today we're going to be talking a little bit about Arkham and uh and trading with intelligence, our our motto. Uh and that's going to be a breakdown of chatting a little bit about onchain analysis. Uh we'll be talking about trading uh in an AI powered future. Um and we'll also be discussing a little bit about how Arkham uses artificial intelligence uh within its technology as well.

So let's talk a little bit about uh AI with Arkham. Um AI has been at the heart of our company since its founding. Um from the very beginning given that our entire business is around data uh and in particular blockchain data as well as off-chain data. We need to be able to at scale take in information, clean it up, and then apply machine learning techniques in order to identify who the owners of the different respective blockchain addresses are. This is the kind of thing that you can't really do by hand. You can't really do it manually. And if you did attempt to do it manually, it would be extremely expensive, which as a as a startup um obviously is one of the the main

priorities, not necessarily overhiring people uh who then go and sort of analyze blockchain addresses one by one, but rather uh hiring one person who can build really amazing machine learning uh data science and uh and AI tools um that you can throw at uh all of the the the large quantities of data that are produced by blockchains on a daily basis from all of their transactions. uh and therefore can produce all of the information that you see on a platform like Arkham. Um so in the early days of the company uh we actually used AI kind of quietly in the background uh in a purely internal capacity uh in order to help us kind of arrange and uh make all of the data that we were ingesting into our software useful. Um but as these

sort of AI products have become more and more popular um we have started surfacing some of these tools on the front end as well uh taking inspiration from a lot of the products that are already out there in order to try to make uh UI uh and UX for our our customers um significantly better and easier for them to understand. Um and so what I'm going to do now is I'm going to go through some of the concrete examples in which we have done this. uh and we'll also do kind of a quick demo of the platform for for people to actually see it all in action.

So let's talk about AI application uh example number one uh which is uh what I mentioned just before this uh the labeling of addresses uh at scale using our internal system which we call Arkham Ultra. Um and so Ultra is essentially a system which we built um that has multiple different components. Um the first of these components is actually uh ingesting uh raw blockchain data and in addition to this uh which is not on the presentation uh also the ingestion of information which is offchain that means uh things that are coming from the broader internet that don't necessarily live within the nodes of uh different blockchain related networks. Um so this means actually taking in every single

transaction that has ever occurred historically on the networks that we do support with Arkham. uh and in addition then also taking in live transactions as they come in in order for us to be able to produce information in real time and insights in real time for our users. Um and then last but not least um all of the information that you can find uh in Google and different social media sites basically anywhere in on the internet uh where people are talking about cryptocurrency and their transactions. The next step is uh actually the uh metadata collection uh and entity classification. So this is using machine learning in order to take all of the different addresses um that exist within our system and then cluster them

together to their respective owners. So um in a situation where for example you collect a sample size of 10 uh different addresses um and you don't know if those are 10 addresses owned by different people uh or 10 different people or if they're uh 10 addresses that are all owned by one person or maybe it's two people with a set of five. Um this is essentially where that kind of entity classification comes in where prior to actually applying the label of who the owner of that address is, you need to be able to know how many people are in the set. Uh and so that's kind of the first step uh prior to the final one uh which is the actual part of deanonymization.

So once you know how many people you're actually working with uh within the sample size then comes actually using the data which you've collected and the metadata which you've collected uh to then associate the clusters of addresses to their respective owners. So actually tagging the company and tagging the individual um to uh to to to their entity profile uh within Arkham. And what this means is that over the past six years, we've managed to build a platform that has 92% of all onchain value uh attributed. So 92% of all dollars that have moved on the networks that we support, Bitcoin, Ethereum, Salana, Tron, etc. um have been deanonymized by Arkham and are searchable on our platform. Uh and this accounts for about $830 trillion of

asset flow um since the inception of uh these different uh protocols. And so um I'll switch over to the to the platform now so that we can kind of go through some of the examples. Um so uh what you have here is the the homepage of Arkham. In the bottom right hand corner we have our uh sort of basic uh block explorer. The really interesting thing about Arkham is that uh we're not a block explorer for a single chain. Right? So the default uh prior to looking at Arkham was that you would go on Ether Scan for Ethereum. uh you would go on uh soul scan for Salana, Tron scan for Tron, etc. um versus with us, you actually just have all of the different protocols uh in a single space uh and so

you don't uh necessarily have to go through different websites in order to get the information. Instead, it's actually just all uh on a single platform and I can kind of search for whatever info I want across uh all of these uh different blockchains. Uh and and so to to kind of prove the point um if I were to go to let's say Etherscan here uh and look at some of the transactions, I can look at Etherscan's uh latest transactions here. Um but really the only information that they show is the the transaction hash. Uh and then you see the from and the two, which is essentially just unlabeled uh Ethereum addresses. This could, you know, really be anyone. Um, however, if I go back to to Arkham, uh, and then I filter just for the Ethereum transactions, um, then I actually surface significantly more information,

right? So, these are the same sets of transactions, these are just the most recent Ethereum transactions that have happened. Um, but instead of me just seeing a bunch of random addresses, I can actually see, okay, well, these is a gate.io uh, transaction, this is a curve transaction, this is a meme uh, memeland transaction, unis swap, 1 in, Coinbase, etc., etc. And so you you kind of get the basic point uh of of how we actually surface the information to the end customer and why uh it would actually be useful. Um and so the cool thing about being able to label all of these wallets and addresses across all these different networks is um you start seeing patterns that the same people own multiple sets of of wallets, right? So exchanges have all of their hot wallets, cold wallets, there are OTC desks um that rotate their

wallets as well. Uh and so what we can do is we can actually create entities which are essentially uh uh profiles that have aggregated uh addresses across uh all of the different networks but within a single profile. So say I wanted to look for one of the you know largest uh crypto market makers in the world. So let's let's look at Wintermute. Um can search here click and now I can see all of Wintermute's crypto activity across uh all of the different chains in which they operate. I know that they have at least $579 million. Uh here's the breakdown of their AUMUum uh from what they're holding. Here are all of the different exchanges that they've used in a pie chart breakdown of how

much they've traded on each. It looks like they have about a trillion dollars of deposits and withdrawals, which is uh pretty expensive. Um their first uh transaction was February 5th, 2019. And obviously we have data going back from then all the way until today. Um, and then here is a list of all of the transactions u that that they are making across uh the different networks. We can see here some arbitum. Uh we can see some BMBB chain um so on and so forth. We we've got some information about their balances, P&L, token balances um etc. So this is kind of how we actually use um the aggregation of this uh information.

Next we have uh on demand technical analysis with the Arkham AI analyzer. Um so this is essentially when you are digging deeper into a given transaction. Say you you find something that's really interesting. You you go okay well I'm looking at this transaction on Ethereum. It looks like this address did this other thing to this other address. This looks interesting. I want to learn more uh about it. Um but maybe you don't actually understand what's going on behind the scenes. So how do you actually try try try to attempt to understand what's going on within this blockchain transaction? Well um for something like this you can actually use uh the Arkham uh AI analyzer which we can find here. So here I have a block explorer uh again uh looking at a

particular transaction. Um and what I can do here is I can actually try to understand uh what's actually gone on. So here I see that there's been a transaction from the Oiler finance exploiter uh to the Oiler Finance Exploiter. Okay, I have no idea what that means. Um so let me try to understand so I can click the analyze button. Hopefully my internet is good enough for this. Okay, great. We're we're working. Um and then the uh AI process uh happens in the background where it sort of analyzes every single part of the transaction one by one.

So rather than you needing to actually dig into what's happening with the transaction, um it it's actually just analyzing it for you. So you don't actually need to understand all of the functions. Once it goes through and parses all of the information within the transaction, it then gives you a sort of very basic natural language explanation for what actually happened, which is in this case, this is a liquidation. So this is a case where um there was a liquidation between this entity um and this was a case of seizing $85 uh million of uh WBTC to repay 569 million of uh DWBTC. Um so this is kind of an example of a way in which you can uh very easily understand a transaction in a way that wasn't previously possible um

without the use of uh of the actual AI. Uh and then the last uh AI application we have here um is uh application number three. So we call this surfacing alpha from the noise uh with Arkham insights. Uh and so this is our new product which actually isn't released uh just yet um but will be coming at some point next week. Um, and so this is where we actually use our uh AI engine uh to combine and then synthesize all of our database um and then all of the information that we take on across exchanges uh global macros, sentiment, news data and then of course onchain analysis. Um we sort of take all of this into one place and uh and then actually

try to produce signals in real time uh that are uh useful uh for traders for example so that they don't have to just dig through as much blockchain information as they can find but instead the AI will just produce whatever it believes to be uh meaningful. So this is tailored analysis of whale flows and transaction patterns um identification of material market events and dynamics. um the ability to filter by token, time frame, entity, tags. You can sort of curate what kind of news you actually receive and alerts that you actually receive. Um and then of course uh it gives you optionality uh for your desired level of uh of granularity uh for the kind of info that you're actually looking for. So let's skip to that. So here we actually have the

Arkham Insights product. Um I I have a previewer uh version of Arkham, so I get to see the stuff that's not actually released just yet, but this will be out next week. Um, and so here we have an example where there was a news uh uh event that actually occurred uh which was the resupply uh hack of 9.5 uh million. Put that back. There we go. Um the hack for 9.5 million. Um and so in this case um we actually get the news in real time, right? So the the event actually occurs on chain uh which is detected uh by by our systems. Um and then of course there's uh all of the subsequent uh price action uh and uh and and material

happenings on on the exchange and market side which it then also pulls in. Right? So from these same uh things that are tagged as resupply. And then of course uh CVX and YFI um also having a connection to it and an association to it from a from a machine learning perspective. Um it can then create an insight for you which you can receive a notification on which basically explains that this protocol was hacked for X amount. It can show you the kind of price uh over time uh and then it will associate uh different uh sort of events that have occurred uh and then the subsequent sort of reaction in the price. So in this case uh it will sort of detect um that

there has been a news event in this case uh the hack um and then it will uh label it against a uh time graph and then as you can see after the news item uh actually happens there was a very material kind of drop in price on the core kind of protocol token um but then additionally on some of the other associated tokens as well. Um and so in this case, this would be an example of you being able as a trader to actually understand uh what what has occurred uh and and then take some action in the market, be it shorting, hedging, um going long on the token once you think it's bottomed. Um these kinds of things are are are all sort of um very important from a trading perspective.

And now um you can have it sort of fed to you on a silver platter rather than you needing to do all of the research and uh and due diligence yourself. Great. Thank you very much. Uh, do we have time for questions? Actually, I might be able to do it. Five minutes. Okay. Does anyone have uh any questions that I could help answer? Is there a mic for him?

One second. Yeah. Yeah. Go ahead. Go ahead. just use AI to get things like sentiment or other sort of tagging. Yeah, absolutely. Um, so basically we um can sometimes rely on third party uh data

sources that are able to collect uh information from places that would have crypto relevant information. Right? There are plenty of places where people essentially chatter about blockchain addresses or chatter about tokens or chatter about price action. Um, and so it's very important to be able to collect that data. Um, you gave a pretty good example which is sentiment. Um, which is something that you can really only detect at scale once you have enough information from a bunch of the different uh places online where people could uh be providing information about what they believe about a particular token if they're bullish, if they're if they're bearish. Um and so yes, we do collect a lot of that information and then subsequently use it um for our own kind of uh insights and analysis. And uh

any plans on Farcaster on on which Farcaster? Farcaster. No, I don't think we're doing anything with uh with Farcaster right now, but I mean could be could could be interesting. Maybe we should reach out all of the chains that we support. Not Yeah. So can you tell me a little bit more by what do you mean by the anonymization? Yes. So um essentially if you take all of the raw kind of blockchain information uh that exists across all these different networks um you're just going to have a raw list of addresses right it's just every you know you're going to have every 0x um every 1B or something from Bitcoin etc. Um and so if

you just took that entire set of addresses um that has ever kind of existed and every transaction they've ever made and every dollar that has ever been sent on chain um we have managed to attribute 92% of it. So 92 of of $100 that have been transferred on chain we've managed to associate to their respective sender or receiver. And so that's what actually causes the the deanmization. Now obviously um with this it's you you kind of have to think a little bit critically about that number um because it's paro distributed right and so there's an extreme distribution where there are a very few number of players such as exchanges market makers and OTC desks who can transfer you know billions and billions of dollars on a daily basis whereas the smaller fish you

know the other 99% actually move a significantly lower quantity of money and so once you have kind of the initial you know 50 60 70% of the addresses which come from these major whales, then the rest is kind of just a war of attrition of attempting to to get a lot of the smaller ones. But of course, there are diminishing returns. We might never get the last 10% or 8%. Um because they're just a bunch of small fish sending small transactions that nobody ever cares about or let alone would pay for that data, right? Um and so that's kind of how we um h how we operate and where that number actually comes from.

Do I have time for one more? Okay, last question. Thank you. Can you link um users together? For example, can I have a profile on Bitcoin on Ethern identity? Yeah, that's right. Um, and so one of the cool things about Arkham is once you have managed to identify the same entity on different networks, we do actually associate them all under one profile. Um, and so you can go to Arkham and without needing to go to separate block explorers or separate profiles, see that Xperson, X company, X market maker has all of these wallets across

these 10 different networks. Uh, and then you can associate all of the transactions to one another. Um and in some cases you can see them kind of bridge even between uh the different networks and then uh make the actual connection to it. So we do all of those associations across different chains in a way that kind of makes it a bit of like an omni chain explorer rather than just like a single one. Yeah. So that is a core part of the business. Yeah. Great. All right. Thank you all very much.