Decentralized Decisioning: The Next Frontier of Being Right

Transcript

I can tell that I feel like um a lot of the parties were last night as well as um the emptiness of the conference. But I'm I'm super excited to be here. I always like coming to ETHC. I've been here for multiple years in a row now and it feels really special to be in con. So beautiful. Uh it's very warm, but here we are in a nice AC seed room. Um I'm really excited to talk about something that our team has been working on for quite a while now. It's called decentralized decisioning. And I'll get into what that actually means. Uh, but my name is Mike Isagawa. I'm the co-founder and CEO of WebC or DD.xyz. And I'll also talk briefly about what we do there. But we only have 20 minutes, so we can dive into it. So in crypto, we always say we're so early, we're so

early, and being early in crypto is kind of the main thing that we we think about as a team because that's when you get the most upside. Uh, but when you're early in crypto, the only thing that really matters is making the right choices, betting on the right tokens, and being right about the choices that we're making. That's really everything when it comes to what you're looking at. And within the decision-m space of being right, there's a lot of choices that we make in our lives, uh, both conscious and unconscious. Uh, the the psychology of decision-m is something that we could do an entire like lecture series on, and it's very interesting. like there's many um choices that we make on a daily basis that we're probably unaware of. I think uh it was a Harvard study or something that did a uh like a metric of how many

decisions we actually make during the day. And it's something more than 20,000 decisions that we make as human beings every day. And a lot of those decisions are shaped by things probably out of our control. Things like where you were born, how you were raised, what you what language you speak or what you ate in the morning. Uh it's actually quite deep and very interesting. But in the context of my talk today, this is more about decision-making when it comes to having information, especially with blockchain information. So decentralized decisioning is a term that we created as a team when it came to understanding that the way that our tech works is by aggregating data from across different points, across different kinds of risk, putting that all together, running intelligence and models on it, and then coming up with a decision of whether

something is safe or not. So this image is kind of meant to meant to show that. So if you things if you see things like volatility, you see wash trading, you see sanctions, there's a lot of uh categories of risk. You could have risk about the legality of someone connecting to your platform. It can also have risk of a token that you're about to trade. Who are the holders? Uh who actually created the token in the first place? How was that wallet funded and was it funded by a sanctioned address? So there's a lot of depth when it comes to sanctions and risk and so on. And at Webby, we take all of this information are able to turn this into an API and actually serve other customers about the kind of data of addresses, tokens, and so on that they're using on their platform that they're letting connect to

their smart contracts, even the smart smart contracts themselves. And so this concept here is to showcase that you actually just take multiple different parts of this decision-making process and you get something more intelligent if you combine them and put them together. So we are um we're seeing a combination of decentralized infrastructure and all of the advancements in AI is a really really great use case for something like this and we'll talk about why uh and you can actually generate some really cool autonomous decision engines as well not just agents which we will talk about as a use case but um there's a there's a lot of really cool things you can actually do. So what problem are we actually solving? Uh crypto is really chaotic. So if you were to send your

friend or your family members to crypto, they're going to be hit with a lot of different information. I even have trouble explaining what crypto really is and encompasses simply. So the top left you see an AI agent kind of researcher on Twitter. You've probably seen it or maybe follow it yourself. On the right uh hand side is Pump Fun. I always get a headache looking at this website, but it's really chaotic. And then we have Dex Screener, which is a great source of information on new tokens, but it also requires a lot of intelligent filtering. And for a novice it might be uh intimidating to look at. And then even throughout all of this the platforms that we are told are safe are told that we should uh be able to utilize uh still get hacked at the end of the day. And these are majority uh they're audited by

normal like companies that are like reputable and do a really good job. But all of the hacks even from the last six months have been based on audited smart contracts. And so that just goes to show that we have a lot of growth to do as a company on the security side. But crypto has this huge trust issue. It's transparent, but it's hard to understand without better systems to make all of this data understandable. And so this is where decentralized decisioning actually applies very well to a problem like crypto. So starting from the lefth hand side, there's three main parts that I think you need for a good decisioning model. And the first one is data providence. So you need good sources of data. You need to be able to rely on them. You need to make sure they're accurate. And one thing we do at

Webby is we aggregate data across many different types of platforms. And one thing we see is discrepancy between data providers about the validity of a token, the information of a token and so on, which is kind of scary because if you're using just one provider, you're might not actually be getting accurate data. So the way we do it, if we have four different data sources, if one of them is incorrect and the three others are correct, you're going to actually like raise uh raise the quality of the response you're getting by aggregating this data together. And so that's that's one piece of the puzzle. The center uh column is about uh decentralized computation. So how do you actually compute decentralized computation also gets into AI if you're doing like federated learning or there's ways to do more privacy related um like activities

with this type of model but for this particular use case for us this means taking this in uh taking the variance of information aggregating it and seeing the insights on top of that. And then the last piece, AI is very hyped up. Um, and I come from a machine learning background and so I I it's a little bit different like I see the value of utilizing it for doing things like improving false positive rates. Uh, but truly the AI systems that we see today uh they they are restricted by the data and the things that we can think about to feed it and train it on. And so I I gave a talk last year about the limitations of AI for cyber security and smart contract auditing is that it's very hard for AI systems to discover new hacks and new vulnerabilities because the data sets are trained on all of the

previous hacks. And so we're we are seeing new improvements to this. But for now, we see a really good use case for AI on things like improving the system performance of new solutions and new technologies that we're building. So I I want to showcase like two use cases that are actually applying decentralized decisioning. Uh because none of this really matters unless it's useful to us as builders and consumers. Uh so address poisoning is one. Has anyone heard of address poisoning before? This is a very common hack. It's been going on for years, but it's becoming more popular now because of how frequently it's being utilized mostly on uh Bitcoin blockchain, but now we're seeing it on Ethereum, Salana, all kinds of chains.

Uh and if you don't know what it is, uh typically when we send an address, uh you can see the chart here, but on the left hand side there's a victim. You sent money somewhere. So maybe you sent it to a wallet that you know, the attacker will sp uh will spam or dust a wallet or make it look like transactions occurred that did not with an address that looks very similar to an address you've sent money to before. And so the victim will accidentally copy paste that address and send money to the wrong location. Um I've done this in general before but with copy paste too just because of the way that uh addresses are displayed. It's just a poor UX uh happening in crypto. And this is a screenshot of a wallet uh very recently.

So this is the current UI of one of the wallets that I use. And if you look at the addresses, you truncate the address. So you actually can't see the full thing. So it's impossible really unless you're really digging deep into the transaction and checking it to see if this is the accurate uh address that you're trying to send it to. And then even then as a human we're probably not going to compare every letter. you're looking at the first four, the last four, and we think that's good enough and that's what the hackers actually use uh to take money from you. And then if you look at the data, so if you were going to actually check if the transactions accurate, you see something like this. And for a human, this just doesn't make sense. But for a machine learning model, this is a really great

training use case. And so we've actually um developed a solution for this. So there are solutions on the market for things like address poisoning. And now as consumers and customers and users of crypto, we just have to demand the platforms actually integrate this. And so this is something that we're really proud of, but it's also a good exh example of decentralized decisioning where we're taking data from not only the addresses that are interacting, but also more in depth. So where where has that address interacted with? Has that address actually done an address poisoning attack before? So you can get really in-depth about that and connecting those dots make this model even more accurate. Uh and then if we look at the actual products themselves, I like to always mention this, but uh we we need as builders to be building products that are actually more useful

for consumers. And on the UI side, there's so much space on this UI. And this this screenshot just kind of goes to show like there's a lot of information that's not useful as a user and there's not there's a lot of white space and visual space that we could utilize to be more uh visually indicative of what is going on within the transaction. Some teams are doing this, but it's uh definitely has room for improvement. The second use case is AI agents. We've saw this as the narrative winner for probably the last six months. Now it's about tokenized stocks, which is also kind of fun. Uh but a few of these so on the left hand side AO is a trading platform. So it's AI agents that actually trade money on your behalf. Uh you've probably seen AIXPTt another screenshot of him. And then we have entire protocols like

virtuals, Eliza and so on that allowed builders to have AI agents and do a bunch of things with them. But more and more we're seeing money associated with these agents and real consumer funds at risk. And the issue with this is that there's a lot of lack of checks and balances and also decision-making when it comes to agents. If you've ever tried to build an agent before, looking at the reasoning behind an agent decision agents decisions is something that is can be tricky uh but also should be controlled because if you don't include them, you could go to jail. So there have been instances where AI agents are allowing fun flow from nefarious activity and who is responsible for that right now. It's ultimately probably going to land on the developer. Uh but there's ways to check for this. You just

put an OFAC sanctions checklist to make sure they don't send money in there or interact with them. That's the baseline level. There should be a lot more on top of that and that's something we do at Webby. Uh but those are things that developers of AI agents are not doing. You could also lose all of your customer funds or your own funds. So if your AI agent is trading tokens uh that are poor choices in the first place, like a highly centralized token or an obvious honeypot token, uh you're likely going to be at risk of getting dumped on as a user or a buyer. And this also happens with AI trading agents that don't have the proper controls to make better purchases of tokens and so on. So we've seen a few use cases of builders using our endpoints to make better purchases and have their AI trade on better tokens

which is pretty cool to see. Uh so if you use us at least you can mitigate some of these things at least not go to jail. Uh and there's a lot of different use cases for decentralized decisioning overall too that you can check things like is this address allowed to even use my protocol in the jurisdiction that I'm in? Am I sending money to the wrong place? Is it even safe to interact with? Is illegal? Uh the the list goes on and on of things that you can utilize by leveraging something like decentralized decisioning. This is just for safety and security, but there's all kinds of use cases you can do when you think about aggregating data and having better insights. It's kind of like 1 plus 1 equals 3. So um we we actually really believe that um the the future of the next couple companies is not just

reading the data and showing the data from blockchain. It's having intelligence on top of it and being able to act intelligently on top of that. And we're seeing this today. So these are companies you probably actually know. Top left is a company called Mintify. Big trading platform. Uh they they were using us originally for flagging smart contracts that had issues with it. So any smart contracts that were obvious drainers to the end users were completely removed. But still in crypto you want the choice to be able to trade tokens like trade at your own risk. You don't want to be limited. So they did showcase some uh smart contracts and NFT collections that had issues but weren't exactly drainers and malicious to a user. and they're able to visually indicate that and at least give their users some sort of decision-making uh abilities. This bottom one, this is

Etherscan. Etherscan has cards now, which is pretty cool. You can integrate uh there's many companies integrated where one of them that give more information to the user about that address. Uh which is really great because this is kind of also a form of decentralized decisioning more more on the UI side is you can look at all of these cards and see kind of like a collective intelligence of what is happening with this address. And then on the top right, uh this is a trading bot. more and more trading bots. If you trade yourself, you probably see things like mintable, freezable, these are smart contract checks, uh even holder uh distribution checks, too. And these are baseline. Uh but some of the more intelligent traders are now starting to demand things like okay, like what else is going on with this token? What else can I actually hear and figure out about

the holder analysis, the connections of the wallets, kind of a bubble maps type uh aggre aggregation of that data. So, as we wrap up, u there's three main call outs I want to make depending on what kind of person you are in the room. So for builders, I want to encourage uh building better UI product products for your customers. Uh the solutions are out there and it just increases that flywheel of customers coming back. We've seen a lot of teams get drained or have bad UX and the customers never come back and they end up dying and that's because they don't prioritize things like safety, things like security and things like better user experience. Um the for researchers, so this is something we come up with a lot because we have some really great researchers on our team. uh there's a lot of intelligent AI progress

and machine learning progress, cryptography progress that just stays in paperwork and never gets applied. And so I'm a big proponent of applying tech uh applying research to something actionable or something people can use. So if you are doing any research, productize it. If if even if it fails, you kind of learn something from it and at least it gets out there. And for everybody else, um I'm hearing throughout the conference a lot of vibe uh vibe trading, vibe betting, vibe coding. Uh but in general uh do your DD or at least separate your assets and the things at risk from the wallets that you're actually doing crazy stuff with on crypto. And so I'll close there. I'll leave a few minutes for questions. Um but you can contact us at my webcy or infoy.com uh about what we're doing and would love to hear about what you're

working on too. Thank you. We have a minute or two if anyone has a question. small room. Yeah, thanks for joining us, guys. I know it's early. No, I mean I don't need a mic, but um yeah, great talk. Uh I actually work with the EF and we have this new trillion dollar security uh initiative and my question is um I mean security is like a forest of edge cases, right? So how are how do you think about aggregating the discoveries that different people make um into one common place so that everybody can benefit from it in the

context of decentralized decisioning? Yeah, that's a great question. So one thing that we really like to focus on is real-time decision-making. So if you think about things like for example TSA pre-check, if you are an American and you get get TSA pre-check, they check you once and then you're good for the next 5 years. But no one's to say that you're not going to commit a crime in these next five years. And so it's kind of an odd system. And it's very much the same with things like security and auditing and our traditional security in crypto. You get checked once your project and your smart contracts still change uh and then you don't get audited again and you don't get audited regularly. So one thing we do is the the real time uh querying of the security around things. So the tokens change from block to block. You need to query from

block to block and reassess every single time. And so to your question of aggregation, we do a lot of the aggregation. Uh but a lot of the teams like for example Ledger's new uh blind uh clear signing if you've seen this before their new future signing uh this aggregates multiple security providers because what they've been finding is that uh they were originally going to assess security providers to just pick one but some security providers that better are better at some things and miss other things and so it's actually an aggregate of them that provides a better end user experience. Um, so to your answer, there's no like one central location that really has an aggregate like that, but if something like the Ethereum Foundation was able to create that, that'd be really great uh for any kind of company. And then we try to do

that internally as well because we aggregate across data sources from everything from indexers all the way to like hype scores across the board. Quick follow up. Yeah. Yeah. Yeah. So, what how do you think just a couple of thoughts on what would be the best way to achieve a good aggregation like what do you what would it be? Is it like you know awesome you know security GitHub repo or or what do you or is it a tool um just as you know as one of the players and if you're aggregating across multiple players how would you like the aggregation system to sort of be architected just rough idea yeah I think um my my solution here I'm going to just like wave a magic wand and pretend everything is possible because as a

builder like we can just maybe build for that um I would want to be able to like query the Ethereum Foundation security stack and then have everything that's relevant to the question I'm asking to be pinged with all of the vendors and like come up with the the aggregate answer of all of that which is kind of what ledger is doing in a way and kind of what we do at Webby as well. Uh but then I'd want every security provider to actually participate in that. So I want all of the auditors and all of their like auditing tools to be a part of it and to be offered and so as a developer I can say hey here's my smart contract. I want every single auditor and all of the machine learning models to look at this so I have like every possible check. Uh but then it's a lot of

information. So um that would be my wish list especially as like a developer myself. I would want to be able to do that. Uh but then we have to work on things like latency. I would want this to be immediate. Yeah. Make the impossible possible please. All right. Thank you all. [Applause]