Ethereum Wallets, Security and AI Agents (Combined Recording)

Transcript

Round of applause. Who is the CEO? The CTO space. Yeah. What the

Look Sorry. Next stop.

Intermedi. influenc inclusch. [Music]

mindset. [Music]

Okay. wallet. Okay.

CEO change. philosophy presenter.

Multi computation. practical. Okay, I get an angle.

Fire. Open source. Coinbase.

Coinbase as a service. saychech.

wallet. Wallet Wallet. What I

[Music] see contract WhatsApp.

Contract [Music] layers. Wallet No,

Forection authentication. Factor authentication. Hardware. Wallet.

I thought [Music] [Music] wallet. Wallet economy.

Simp again. Awesome. Come on. Keep it going. Keep it going. Good. No questions. There's one. Yeah. So, you you hit the biggest problem right now with the blockchain uh with with with the blockchain uh ecosystem which is that is extremely complicated even for a professionals in

the in the area. So, how will having a a smart wallet help a normal person? I'm saying a person who doesn't who isn't who is an accountant or something like that uh will help them reset their passwords and stuff uh be without uh having to memorize the 14 words. Okay. Okay. Let's jump into English because I didn't give the presentation English. So, it's okay. But have you used web3 out reown or any of the alternative for social login right now? Yes. And don't you think that for a new it's a way better experience to log to any web three platform using the

social networks. Agree. I I agree. And I think that it's not the solution to the problem, but it is an alternative step for kind like the world custodial kind of like if you need to enter an exchange using your social networks, even if it's a non-custodial, it will be a great kind like middle thing between using a custodial wallet and kind of

like having your own phones related with uh like your social networks but that it is like owned by authenticator. 100% true. The thing is like we need to start using these other kind of wallets. Do you have a metam what wallet do you use? Exactly. I mean I I also have a Rainbow wallet, a Metamas, but what I consider important wallets, I have a safe. Like I use a multisig, but safe is like a good option for me. Anything else?

So I agree that keeping a private key is a very difficult thing. um and that these smart wallets are great, but it seems like there's a lot of steps to go from where they are now, which is they're specific to an application and also to a blockchain to the situation you get with a signer like MetaMask or even with these MPC signers like LIT. What do you think those steps are to go to all applications and then all blockchains? So, what do you mean? Can you I didn't understand the question. Sorry. Oh, the the type of account abstraction that you were talking about that was promoted by Vitalik is something that connects you to a smart contract on a particular blockchain. And as you noted, typically

the UI for that is application specific. So those are two constraints that we want to get past to get to where we are with MetaMask with the very difficult private key, right? all applications and then all blockchains that are EVM compatible that use the same signing. So the problem kind like do you mean for the for the thing between blockchain interoperability or kind of like using a better usability for interacting with any blockchain being able to use the wallet will use that term on any EVM blockchain which you get automatically with the current signers. Yeah. I I think that well

there's not a standard right now. That is one of the problems. I mean there's everyone trying to build the next better uh smart wallet. So it's it's what happen with the genome ecosystem. It's really fragmented right now. So we are not having a unique point of view to solving the problem. There's a lot of people using this. So there's not a standard. I don't think that it is a solution that I see in the future where you will use only one thing to to use it. But so you think it's a big step up even if you only use it with one blockchain or even one blockchain. Obviously I think that we should go step by step basically kind of like we need to start using MetaMask stop using MetaMask and start

migrating into a smart wallet and then focus about the problem of if I need to use a smart wallet for each of the change that I need to use that's another step. Great. Anybody else? One more over there. So in in a practical sense, how does this look within the Ethereum network? So I mean for some context, native account abstraction is a part of Starnet. So all wallets there are smart.

They're a beauty. Honestly, they're they're really really good. But how would this look in Ethereum? Right. So you mentioned a protocol upgrade coming up. I know there's like ERC 4337 that's kind of like a standard but like how would this migration kind of look like underneath? I think that G is going to talk about on Tuesday. Do you want to give them human answer? He's the one that needs to answer this. Yeah. So basically um what's going to happen is that we're going to see in the next Spectra upgrade um right account um the account abstraction being taken uh in as a native feature of

Ethereum. So you will be able to basically tell with one transaction um deployed in multiple chains uh what it is that you want to do. So you would basically have the same smart contract deployed on multiple EDMs and then just signing one transaction and sending it across all of those chains where you have the um the contract deployed, you would be able to just associate in a sort of way that we're going to talk about in uh Thursday uh in Thursday's talk. Um that yeah, you basically use the smart contracts code as if you were

uh deploying it to your own externally owned account. So I I don't know if that sort of clarifies it better. Yeah, it does make sense. So, the EOAS would basically kind of like be a thing of the past, but for now, we would link them to the new smart wallets. Yeah. Yeah. So, basically, you would be able to just keep your EOA if you want, but uh you would also have the ability to just move to the smart account if you want. just I think it's a great solutions just to uh keep everyone happy with having their own EOA or they can just switch to their um I don't know the available um smart I'm sorry smart wallets at the time. So

it depends on how far along we are. if there are multiple uh addresses deployed in multiple contracts, you would be able to choose with the new transaction. Uh but yeah, basically we would have that new added transaction um to the Ethereum ecosystem and the uh EVM. So uh yeah, you would be able to choose which one you want to use and if you want to use it. All right.

Thanks. Let's give it up for Roman. Yes. Thank you. speaker. We have one announcement. We have Alejandro Navaro from Betsky has an announcement for you guys uh for our um special night. There we go. We don't have to be that silent. There's no suspense and mystery.

There was no clapping this time. Yeah. Very, very silent. There we go. That's better. So, we're going to steal just a little bit more of your time. I know we both had some time up here already, so appreciate a little bit bit more. Um, Alandre, are you okay if I just kick off? Okay. So, yes. Thanks. So yeah, so Block Dag and and Betsky have been uh working together for a while, having various discussions. I mean, from from what we saw, we loved what they were doing. Uh we we were very keen to to be able to work closer with them. Um and certainly from our side, we're extremely excited and proud to be able to announce formally that we've

struck up a formal partnership with them. Um and the the nature of that is twofold. So there's a strategic investment into the Betsky token. So we're going to be owners of Betsky and we're looking forward to those fantastic returns. Um but probably more importantly is a strategic uh partnership around uh the involvement of them on our platform. So they're going to be deploying a DAP on our platform, exclusive DAP, and we're going to be the exclusive betting partner and vice versa. Um, so we're going to be the only chain which has Betsky as a DAP on our chain and and we're going to be promoting them to our large ecosystem.

So, Block Tag is super excited about this. We're really looking forward to the the partnership. Um, and yeah, we we we super happy to to be working with you guys. Thank you guys. Thank you very much. You know, it's been very exciting this whole process. We've been for quite a while now working alongside Blockback, you know, knowing and and talking about everything, the vision and everything aligns perfectly. So, we're very excited to become the official and exclusive betting partners of Blockback and we look forward to the future working together. Thank you very much. Thank you everyone.

Yeah, you could feel the excitement, right? Come on, give it up, guys. This is This is big. This was close right here. Eth canal event in Panama City. Come on, one more time. Yes. All right, guys. Our next speakers is he serves as the general council of open zebling a leading open-source framework for building secure smart contracts. He joined the company in 2019 and has held roles including head of operation and legal prior to his tenure at open zippling. John worked as an associate of Scandin

LP. That's a long name, isn't it? Well, he's a great great great individual. He holds a master of business administration from the Smith School of Business at Queens University and a bachelor of art of in political science from Sydney Hall University. He's also a charter member at Tech GC reflecting his active engagement in the legal. Let's welcome to the stage John Newfield. You want to jump? Take that from me here.

You saw what he did, right? He's tall, so he just climb up on the stage, right? All right. Stage is yours, John. How do I just tap in the middle to go next? The green one and then the red tube. Okay. Yeah. Okay. [Music] Well, I live in Panama and I'm super excited to be here because I've been here three and a half years and Stephen welcomed me into the small but robust and and growing blockchain community here. So, it's pretty awesome to drive in and, you know, sit in sit in the customary traffic jams and come into Costco and and be here. So, I'm excited to chat with you all. Um, I've been at

Open Zeppelin for six and a half years now. Um, so I am a lawyer, but um, I've seen a lot about, you know, the the open-source ecosystem in in blockchain, which is super exciting to me. Um, and yeah, I I'm not a tech expert, but I'm I'm really open to taking questions and interrupting me, and you can ask me anything and everything about Open Zeppelin or or my my take on things, and, you know, can be related to this or not. Um, so I just put together a few slides about what I've been thinking about uh at Open Zeppelin lately, um, that I hope is helpful. Um, so who's Open Zeppelin? Um, many of you have probably heard of us before, but um,

we're an open- source company uh, focused on security. Uh, we help anyone who's basically building a decentralized application build, secure, and operate that application. Um, I've put up, you know, we we've worked with thousands of protocols, uh, financial institutions, Fortune 500 companies lately. That's been interesting. Um but also a lot of say the leading or blue chip protocols the safe um and I put up just select few kind of ways that uh we do help folks build secure and operate um because many folks will come up to us go oh I know open zeppelin but for x y or z they always ask like also how do you make money

because isn't it all open source um so that's always fun but we have you know our contracts libraries in the build side you know I'm just going to actually go to the next slide where I go into this a bit more detail. But yeah, we've been setting, I would say, industry standards. Um, with the proliferation of our open source libraries, we've been around, we're turning 10 this year. Um, so like I said, you know, we've helped many different folks of of different uh kind of types and in different industries. Um, I I would say we're best known for our contracts libraries. So these are open- source smart contract building blocks that have been battle tested over the years. They go through audits, formal verification, and they've just simply been out there in production um with really bad hackers around the

world trying to break them and and they kind of have have held the test of time. Um so yeah, over 23 trillion in value transferred with those. Uh and then we're also known for our security audits. [Music] those so that they operate uh in accordance with expectations. So you know some stats there about the the amount of work we've done here and then we also have a lot of developer tooling which allows folks who are building operating uh protocols to do so securely and efficiently. Um so that's more on

the devs sec ops side. So yeah, I I didn't touch on the title of my uh presentation but u blockchain can present security guarantees applications when they're developed. So they can guarantee users access, availability, and integrity in accordance with their rules. And importantly, I think the big is you don't have to rely on any centralized service provider, data hoster, database maintainer um in order for that application to continue working. If the

code uh is correctly kind of uh written and and implemented and operated, then people can just interact with each other and know that hey those rules are going to apply to me. No one can prevent access and that data will be immutable and available and accurate over time even if this company goes bankrupt or this government uh goes corrupt. By the way, I'm Canadian so whatever you may think of the accent um I come in peace. Um and so that that's great that does prevent be or or sorry present better security guarantees. However, because the code, you know, maybe it's smart contracts if

if we're talking EVM world, um because it will execute according to those rules. If those rules aren't correct, it can have very negative consequences because anyone can interact with that code and the code will simply execute. So, what can happen is if you haven't um basically put the right into code or you have [Music] consideration rules that weren't considered or economic that exploit it and you can end up wrecked.

So many of you will probably know what that means. It ain't good. Uh it's basically you get hacked and you get on this fun uh leaderboard which our industry uh follows quite closely. Um and you know users lose money, your protocol loses trust, regulators take a keen interest into why all those uh innocent consumers funds got lost. So this is a big problem and that that's kind of the the trade-off where if you do it right, it can really unlock a new paradigm of computing and interacting with each other. Um but if done wrong, it can end up with a lot of people in a tough spot.

So the short answer is there's no silver bullet for security when building decentralized applications. Um the history of software development other critical industries has shown that if a project or developer as a company implements security best practices at all stages of the software development life cycle, there's a better chance they're going to find issues sooner. they're going to correct them or if there is an issue they're going to be able to mitigate it, respond to that incident and uh reduce losses and and issues with their application. So given the realities of of blockchain and decentralized computing, this is just heightened and even more important. So I've put out all the um Okay, that's

counting down. Okay, I I have some time here, but I put out, you know, all the highlevel things uh we recommend to folks that I don't think are too controversial. I think other security companies would agree with us on on these um and a lot of this stuff we we also offer uh to the world of developers for free on an open source basis. So uh there there's often not a ton of good excuses uh to you know maintaining good security hygiene doing these things from the beginning. Uh but it does often run contrary to you know the business impetus of shipping and iterating and moving fast. Security often does slow things down adds processes. Um but at

the end of the day you know it's so important that when we're actually going to put something in production it works properly and it's safe for folks. Um, so I I'll maybe just touch on a couple highlights here. Um, we, you know, recommend folks use open source libraries that have been tested over time and also to just to the extent possible rely on things that have been proven in the past and introduce novel code only when necessary. That's also going to make your application a lot more simple. Uh, it's going to make auditing it more affordable and it's going to make it easy and clear for users and other developers to understand when they start going through it, you know, starting with robust documentation never fun, but um it's it's very key to

define invariance and how this protocol is intended to work so that folks are able to uh understand that from an outside perspective. Um testing and then you know we're auditors so before anything um does get deployed to the blockchain we recommend that it goes through a manual review with experts trying to break it. independent third party who didn't write the code who basically can yeah try and expose any weaknesses and also you know sometimes folks think well this is just a tick the box jump through the hoop exercise I need an audit report in order for you know my investor to give me funding or for my token to get listed on an on an

exchange like a centralized exchange but you know we really see this as as a lot more than that and and when we do an audit we're not just looking at hey this smart contracts in in scope. Uh but also you know what are all the things this project could should uh or is already doing across the life cycle uh and try and give them improvements so that they can really build a strong security posture going forward. Uh and then things like making your source code public so folks can examine hey what does this actually do uh that I'm entrusting my assets into? Um having a security contact in case someone finds you know a bug in there themselves. Um having a bug bounty program to

incentivize the good hackers and there are many to submit a bug to you responsibly so they can also disclose it to the community or other people who have integrated your protocol and it's not uh exploited by folks uh out there who would you know love to see the RECT leaderboard continue to prosper. uh folks like the Lazarus group uh of North Korea we know is uh highly sophisticated and actively out there trying to steal any value they can. So if you know folks in this room are successful there's value locked in a protocol. You can bet there's some really nasty folks trying to get at that value.

Um so yeah uh I I'm not going to go into other details there. I'll skip forward. So one thing that myself as as a lawyer and someone who's been involved in the policy discussions lately in the US and and the EU is potentially good security and blockchain can actually be a silver bullet to regulation and regulatory compliance. I think uh the last four years uh the US as kind of this world police in in all various are areas but especially uh in our industry the securities uh regulator the SEC um really made our industry constantly you know bogged down thinking how how do I do things in a way that don't trip up US

securities laws but also you know various other banking credential regulations commodities uh trading platforms terms um etc. So I think um one thing that's emerging in in Europe is um that actually if a blockchain protocol is decentralized and therefore operates according to its encoded rules, there's no intermediary or person of control that can just change that or steal funds. There's no custodian. Well, then shouldn't a certification that this protocol operates as intended and is expected to over time be enough to satisfy regulatory uh compliance and ensure consumer trust? Um, so this is

basically the recommendations we're seeing for the expansion of Mika, which right now only regulates centralized uh crypto asset service providers in Europe. um they're looking at hey we could regulate they say defi I think it should be decentralized applications generally by basically doing an audit which I think is great news for our industry because I would say the vast majority of of decentralized protocols do that anyway so uh we're pushing hard and interacting with those regulators to make sure uh that audit actually makes sense and is aligned with what folks really uh are are doing already because we see that as like uh basically net zero additional work and cost. Um, and then what else do I have for you? So,

we're also pushing that narrative in the US. You know, we met with the the SEC actually, which was crazy after six and a half years in blockchain going into the SEC headquarters, not and yeah, not being worried you'd leave in handcuffs. But um super productive talk there and and looking at how the legislation's evolving there, seeing how uh we can also push this kind of um security as a sufficient um compliance narrative there. One thing I did want to just bring up to everyone uh because this was news to me and it really shouldn't have been as someone you know who works for an open source software company but in the EU they are um they've approved the cyber resilience act and that's coming online the end of

2027. Basically that's going to regulate all software and hardware uh that's offered into the EU market whether it's open source or otherwise. If you're putting out a software or a program uh that you're monetizing in some way, you are going to have to follow security best practices like the ones I laid out um in in the slide previous to this. So uh I think for those building to the extent you can start to implement this kind of uh best practices throughout the the developer life cycle um I think the better you're you're futurep proofing uh yourself from the uh oncoming regulatory regimes and I don't

think you know we should fear regulation for this reason. I think uh it's like cleaning your house. we could all use some incentives to to do that and to motivate uh because it's not necessarily fun or going to make you faster to market or find product market fit, but it's also going to cause you to make sure you're building trust over time with users. You're you're building a protocol that you know can grow and and handle kind of I would say the high stakes of um of success in this industry which you know I think we all want. But um yeah, so I think that's actually everything I I wanted to say. I know we're running late, so um I'm happy to answer any questions if folks have questions um about this or other things we do at Open Zeppelin. Otherwise, I can

keep us on compliance. Yes. Yeah. Hello. A big user of open zeppelin here of the libraries. uh exactly what the what does that new legislation uh mean? Uh the new the new European legislation mean what what does it entail for developers in Europe? Yeah. So one uh important carveout is individual developers. You submit a patch to an open source project. If you're not running that project doesn't apply. Um but for they call them open source stewards. So folks running a project that like let's say the Linux Foundation um they're absolutely gonna

have to comply with this things like shipping making sure when you ship something you've taken reasonable efforts to make sure there's no vulnerabilities or probably a certain level of vulnerabilities. Um and then over time and at least five years be on top of that your any like downstream dependencies whatever it may be. um and make sure that you're monitoring for vulnerabilities uh at a reasonable level and fixing them and notifying any users of your your application uh of those vulnerabilities. You need to separate updates from security updates. Um, so I think this is something um I'm I'm gonna write a paper actually with some of the

folks in in Europe on the policy side because I just don't think you know in the blockchain space we're we're thinking about this at all and like the reality is you deploy a a protocol on a blockchain like you're making it available to the world. um they're going to look to um basically see if you if you comply with this and you know things like having a security contact incident response processes that starts to get you know a little trickier where you can't just roll back to known safe state where uh you might be dealing with upgrades or maybe even a mutable protocol and so I think you know the earlier folks are thinking about this and building in mechanisms and and processes the the

better. Thank you. So, Crystal. Yes. Great to our next speaker. Let's give it up. And another one because he's made it. He made it out of SEC with a handcuff. Right. So, how we feeling? Good. Yes. All right. Clicker.

Great. See if we ready up [Music] there. All right. Michael is like with expectation waiting sitting there. There we go. Michael. No. O is the CEO of TAS network. He's the founder and CEO, a blockchain platform dedicated to decentralized artificial intelligence. Under his leadership, he has developed Nexus and onchain agent framework that enables autonomous AI agents operate on blockchain technology. In 2024, the company secured six million in funding led by Polychain Capital to advance its decentralized AI

initiatives. Michael has also been involved in partnership integrating AI agents with platforms like SU blockchain and walus decentralized storage aiming to enhance AI deployment and data accessibility. Additionally, he has participated in industry events such as the summit on responsible decentralized intelligence contributing to discussion of the future of decentralization and AI. So let's give it up for Michael Hanono. Welcome to the stage. He hasn't made it yet. Come on.

Keep it up. All right. All right. All right. Okay. Get comfy. That means that what we have we're about to to hear is very important. Welcome. Welcome to Panama. Right. Welcome to Panama. for those who are not locals and you know for the ones that are locals know thank you for coming. It's great to have another of these events. Um and today I'm going to give you a brief brief overview of what we're doing at Talos, a bit of the up updates and hopefully I could educate you on AI and why is it important to operate it and have it execute on blockchains. So

who here is familiar with AI agents? Well, for those who are not, AI agents are basically any type of AI enhanced program that has a specific goal, could learn from its own environment, and could execute objectives and transactions for their creators. See if this works. It's not working. There we go. So what we're doing at Tal

is we're building a agents on SUI. It's a SUI blockchain. We've launched Nexus which is the first AI or onchain AI agentic framework. It enables you to build and orchestrate multiple tools to build these AI agents that operate on a blockchain. And the idea of having these onchain agent frameworks is because you want to have censorship resistance. You want to have transparency, you want to have composibility. In the age of AI, let's say somebody comes and builds an AI agent and that person wants to offer it as a service or as you know even as a decentralized app or a business. You don't really have that transparency of what the agent will do, how it will behave, will it even handle my assets properly. So if you bring all these agentic primitives on chain and you

define the workflow of the agent which is the brain or the instruction set, you put it on chain, you basically enable agents to be transparent and anybody could see what the agent will do, how the agent will behave and have censorship assistance which is I built a bit I built a business as an agent and then anybody could come and use it. I cannot really dictate who benefits or not. And then on top of that, we're also building uh a launchpad for AI agents enabling consumers to come and build uh these types of agents very easily without having to learn much of the of the coding. Uh this platform is called idle.f fun. It's kind of like a very interesting application that I'll I'll talk about in a bit. Um but yeah, the idea is we're not only just building

random infrastructure that nobody will want to use or even pay attention to. We're focusing on how can we actually make people have fun with these agents, make money with these agents, which is one of the other bigger problems in the age of AI, like you create an application. Right now with the current web two rails, it's pretty hard to monetize. So, as I was saying, what is an agent, web 3 agent, you want to have the agent actually transfer value autonomously. You want to do that through a blockchain because you have these native permissionless rails. And the idea when you do at Talis in the long term is we want to power this agentic revolution into the web 3 AI economy. Uh we want to have every single type of task that can be done by a human

be automated by AI. And to be able to do that you need to have trust. You need to be able to like trust entities. And you know that specific trust is not really built in web two systems. So, as I was saying, we want to build the true agentic economy, allowing any agent to interact, transact, create value for humans, and bring value back to their creators. What is one of the biggest problems right now in web 2? As I was saying at the beginning, all AI agents right now are completely centralized. You might have the AI agent workflow be closed source or behind a web two server. So you don't really know what the agent will do, how is it powered, what models does it use, what tools does it interact with, or if it's even managing your per your keys or

permissions in the right way. So all these AI development and AI services are prone to centralization. You when you think about the three inputs of AI, you have models, you have data, you have compute, all these things right now, you know, there's no really way of like orchestrating all of them in decentralized fashion. So users could actually get rate limited. You could have uh private keys leaving in centralized servers. Let's say you have a a framework built like autogen or the open framework. When you give it keys to interact with your tools, you know, you can't really manage those those tools.

And even when you think about the setup costs of building these agents in these frameworks, you have to go from zero to one as a developer like hosting all the the the docker components and all the images of the models and all these things where like if you use managed infrastructure like blockchains, you could easily go around that and just connect to the blockchain that has all these tools already orchestrated and get to get to production much faster. Um so you know we also believe that all these growth primitives are not possible without having these security guarantees and sometimes we even think that at the end of the day we might be even limited by the offchain components because you know there's no true innovation happening if we have these completely centralized

platforms governing what we can do and what we cannot do with AI. So I want to give a brief overview of like how we think about the decentralization of agents. When you think about no laser here. So when you think about type zero which is what we denominate as having having a fully offchain AI agent framework you have all the logic happening completely offchain uh non a blockchain interaction. If you think about origin and OpenAI's framework or even what Anthropic is doing with MCP all that happens wi within the type zero uh bracket then type one comes in and if you've been in the crypto space for the past I'd say like six months uh or even more since

October you could see that there's a lot of frameworks like Eliza Arc uh even truth terminal the agent that made everything happen and I say it made everything happen because it just like put AI agents into into attention. Uh all these frameworks they're really centralized. All their compute is hosted offchain. all their coordination, all the agent workflows happen offchain, but they do have or they claim to be onchain AI agent frameworks because they have connections to blockchains and connections to you know just like being able to send a transaction or own a wallet and that that is what we call oh my bad that is what we call uh type one arc etc. uh and these frameworks they did a good job in you know making

things widely adopted but at the same time they're fairly limited because they're not truly decentralized and that's a pretty big misunderstanding. So and then my bad and then jumping into type three is you have fully onchain agent frameworks which means you have all the compute all the tools all the orchestration all the intelligence happens happen onchain but that is not really feasible yet so atal we we find a middle ground between this which we call like type two agent frameworks which is where the talis nexus framework comes in and you know it orchestrates and executes all the compute onchain so you actually understand and could see what what's happening but the most expensive parts

which is an LLM inference really happens offchain that's why you see the actual brain or coordination layer beyond beyond chain and then through events you basically coordinate the actual intelligence that happens offchain um just looking into what we define as a talis agent built with nexus framework uh we have these onchain workflows uh that people want to come and build with for whatever use case they want the work for workflows are completely flexible. Uh and then these workflows query offchain resources for the intelligence.

uh we have all the actual like compute data and the models be hosted offchain because you know right now it's not feasible to have them onchain but when you want to like coordinate all these in different methodologies and different workflows you could actually do it onchain very easily and you actually have the guarantees that the actual agent is hosting and it's managing the the tools and the resources as as it should um and then you know as I was saying you have the workflow be defined on chain so why tells agents you want to have censorship resistance. You know, agents being operating within the blockchain transparently, you know, following the rules that they're defined within or those constraints. You want to have full transparency, full visibility of what the agent will do and the actual

workflow. So, if you want you want to inter interact with an application and you want to inter interact with an agent that's built on Talis, you could actually see that it's defined within a smart contract and it follows those rules. So, you know exactly what the agent will execute when you prompt it. And then finally, composibility, which is one of the the biggest features that we're promoting. There's not really like when you think about like blockchains applications or even AI agents built in in web 2, you cannot really compose multiple APIs or multiple tools without having or common common interfaces. And blockchains, you know, by default, they're composable. You could integrate multiple smart contracts together and, you know, by defining agents on chain,

you actually inherit these properties. Uh so you could also coordinate agents with onchain applications and make them much more composable than just having an off-chain resource query and onchain smart contract. And finally verifiable ownership and governance. At some point if you build these agents as services and businesses you're going to have potentially tokens. We've also developed a framework for initial agent offerings where you could deploy agents and tokenize them very easily and at some point have governance capabilities over these agents and applications.

the I like to think about these agents as agents as services or agents as businesses where you just like build an agent, you offer a service to a community and then you could have a decentralized protocol built built on top of it. As I was saying, Nex is diving more into what the framework actually is. Uh you have hybrid computation, you know, you have the off-chain LM inference and onchain workflows. Uh you could automate DeFi, gaming and other blockchain applications very easily through the actual onchain workflows. The idea is to actually expand community participation and make AI fair. That's why we kind of like explain and and put everything on the chain and make it accessible for anybody. Um and then you want to have the blockchain which is in

this case is sui operate all the decentralization of the agent. Here uh just diving briefly into the architecture uh we have nexus which is the actual set of smart contracts. We have Walrus that serves as a decentralized storage layer for all the you know the agents inputs outputs or even some type of like memory if you want to define memory for an agent stored on this decentralized storage. Uh you actually have the talis agent package which is a definition that we've created for an actual agent and then you have all the different tools that the agent interacts with it. On top of it, we've built uh initially a ROSS SDK that you know automates and puts all this

together under one roof with a nice CLI that developers could easily come and use, but we're expanding. We're also thinking a lot about integrating tools like MCP and how could we make Nexus accessible to other types of de developers or even people who are building within the web two space. When they're building applications, at some point they'll say, "Hey, I want to integrate a decentralized workflow for my agent. How could I do it?" and then they could easily connect through to Nexus. The idea is to also make this like pretty compatible with existing tooling. In the past, I'd say month and a half, we've seen a huge rise in new tools getting integrated. So, we're trying to like stay ahead and also integrate those into the framework. Uh, Nex Nexus packages.

Diving a bit more into detail, you have the actual workflow defined within the the smart contracts or packages. uh you have all the comp the components that define the execution of the agent and then you could also orchestrate all the tools within the packages as well whether they're onchain or off-chain onchain tools are any other type of smart contract off-chain tool would be something like hey I want to call my Twitter API I want to call my bank API or even stripe you could also do that as well then diving into uh our product our consumerf facing product idle fun uh we want to create this AI personality agent launchpad. The idea is to, you know, make the app successful, so people pay attention to the framework. Uh we've

realized nowadays in crypto space that if you're just building a piece of tech or piece of infrastructure, people might not really pay attention to you or at least you will launch a token and you would pump and then dump. But the idea is to actually build something that like people would want to use that people would like enjoy using, get entertained. Um so we're adding all these like social integrations to make a bunch of agents go viral. uh defined for this is like completely designed for like normies and and crypto deents. Um launching an agent with a few clicks and also uh adding interesting economic models just like the virtuals launchpad has. And then something that we've uh innovated on uh this is a kind of like a brief image of of a launchpad. You basically like

register your agent could give it a personality and also uh promote it on social media and so forth. and the agent could also have capabilities for talking on Twitter, interacting with other agents, etc. Um, something that we've uh looked into a lot is like how do we actually make agents make money because you know you build a service, you build an application, but if the application doesn't get product market fit, you're not going to get anywhere. So, our focus is how can we actually bring these agents to people, enable developers to build and actually make money and make a business out of out of their agents.

This is come coming from my view that like at some point every single business out there will be an AI agent. So at the end of the day if you merge both you will want to have agents that make money. So first is personality and memes. Uh you know this is obviously like good for any type of blockchain where you get like a lot of activity because people feel that they want to look for entertainment. People are usually bored on the web. Uh so they find hey this is a cool agent that I I can identify with. Let me just connect. Let me interact with it. Maybe I'll find something cool. Uh so you know all of these like relationship based or like mimi based agents are really interesting. On top we have gamification. Uh one of our interesting use cases agent versus agent games. Uh

imagine an agent creating a set game and then having other agents compete on top of it. Uh this is I think and then like people at the end of day could bet on the agent outcomes. I think this is like a massive source of entertainment now that you have a lot of like AI generated content. people could come and bet on these agents that would actually behave in in some sort of way that they would support and people could kind of like bet on on the actual outcome based on lock or even agent interactions. Uh so we want to have like kind of these betting not betting but actually like game boards or game uh master agents that define a a a place to play for all these other agents and then have people interact on top of it. And then

last but not least, DeFi, which is one of the most useful um crypto use cases right now. Uh you want to have agents route liquidity throughout multiple DeFi protocols and actually improve DeFi UX, which is one of the biggest roadblocks that have hindered crypto adoption. So yeah, again, as I was saying, the job of the agent at the end of the day should be to create money or create wealth for their creators. So some of the bigger goals that we're going to like bring into the SUI blockchain is to make it a destination for AI agents. Why? Because SUI and if you guys don't don't know about it has an interesting programming language

called move. Uh it's object-oriented objectentric. They have the right primitives the right tooling uh to build a lot of AI services on top of it. But you know they're kind of lacking a bit of agent activity and we want to make that push great and you know push that narrative on on SUI and increase the velocity of money on SUI through a lot of AI agents. SUI is a great blockchain. I think uh the experience that we've had with them has been amazing and we're looking forward to like continue partnering continue building a bunch of AI services on on that platform. some of our latest updates. Uh we've announced our seed round in November. Uh which was almost six months ago. Uh we released our NFT collection called

Tallies. Uh you could market it on on Tradeport. Uh it was a one of I think it's top five top six collections on on Sooie which has gotten pretty cool. They're like these really nice frogs. Um and then we've also completed the milestone for the DevNet that's going to be coming out later this week. And then yeah, we're looking forward to launch our frame, our launchpad by the end of Q2. So yeah, thank you so much. Uh join the agent revolution. This is a QR code if you're interested in building on our DevNet. Fill it in. Uh we'll reach out to you, support you, and follow us on Twitter. Just join our Discord or even look at our website. Thank you so much.

I'll take any questions. All right, there's a question over here. Ready? Hey, hello. Uh, okay. Uh, I've se I think the big elephant in the room here is that that most of the bias of all of these agents and networks is in the LLM itself, which even for the big companies is a black box. So is there is there a way to is there a way to organize the the to know which bias the the agent is using in the output of the what bias the internal LLM output has question that was one and two are you planning to use MCP protocol for

LLMs good so answering the first question LLM or AI bias is a very difficult problem um I don't think We're going to be the ones solving that right now. But I what I like to say about this is that there's no real way to verify the LLM output. What you could verify is that the output came from the right model. But whatever that output is, you know, it could be right or wrong, could be biased. Like there's no real way to like break that black box right now. It's up to, you know, the model creators to to manage that. What we're doing is we're enabling multiple AI providers to come into our framework and provide their tooling for people to come and build AI agents on top of us. And second question, yes, we're planning on

building a wrapper around a framework that serves as an MCP server for any MCP server to interact with Nexus. Hi. uh how do you compare uh the SUI network with uh for example Near uh especially for building AI applications? I know the Near team is very like AI focused and like they're they have like very good credentials on uh AI but I I know very little about like the differences between the two. So I don't know a lot about Near from a tech perspective. I know that they're was based or ROS based. So they've made a huge push on that. Ilia of course has

been the basically the creator of like LLMs to some extent or modern LLMs. I I know that they're doing a lot of like a huge push in terms of like AI infrastructure and how they could like power compute networks and general compute on their platform. Suie is more in a way as like the coordination layer for the internet. And why we like Sui much better, it's because they we think that they have a much stronger community. We think that, you know, they actually have like DeFi applications. They actually have financial assets on their chain and they're pushing for a lot of partnerships where Near is just focusing on on infrastructure. Um, but you know, I know a bunch of people building near. They're doing a a great

job. I'm just not like super familiar with in depth into like what uh pushes they're doing for their own like ecosystem and how they're interacting with other crypto apps. Thank you. Great talk. Uh it was pretty fun to to watch. Uh my question is around the architecture for for Talos, right? You mentioned that you have this event based architecture where you have these agents that I assume consume these events. So I was wondering what happens if an agent is unavailable, right? So do you have any kind of penalty for that? Do you have any any sort of security measures for that embedded in the protocol? So right now how the architecture looks like is we have the smart contracts that basically define the whole interface and then we have initially in the devet it's one leader

we call it a leader node that basically reads the events and queries all the tools. At some point if that leader node goes down you know you could actually look at the tool look at the implementation and you know step in and call the actual event yourself. But we plan on making this a whole like decentralized process down the line. Some of our plans is to actually enshrine the leader node into the blockchain. So everything is like blockchain native. So if the blockchain's not available, your actual agent execution is not available. The main problem comes when you have a web two tool that might be down. Yeah, at some point you'll have need to have some slashing mechanism and tools will need to provide some economic stake and you know if they're not responding in a

timely manner or in not the right uh format you know you could easily slash and that's included in our in our tokconomics. And I have another quick question if I may. Um it's about the type three uh model that you showed. Do you think that that's ever going to be a reality where you have the like the entire a AI agent thing working on chain or do you think that's infeasible like forever? I I don't think it's unfeasible. I think up until forever we want to move towards type three uh let's say we like we have plans we have like working protocols of like as I was saying like enshrining as much as possible into the blockchain but as long as you have something like

Twitter not be on chain or any other web 2 API that's not on chain you're going to be on type two so you could go like type 2.5 or as closer to three as possible by enshrining all these like orchestration metrics or like the actual events. But if you end up using a web two uh a web two API, it's going to be very difficult and I don't I don't see like Twitter going onchain anytime soon. Um it would be ideal, but you know, it's it's fairly difficult.

No, thank you. Yeah, that was fun. Thank you, Michael. this way unless you decide to jump. No thank you. All right. This All right. Well, we coming to a closed of this session, but not of the evening. We have our next speaker represents Tessar's lab.

He's a experienced software engineer with over 15 years of expertise across various industries. He has a proven ability to develop and deliver high quality software solutions demonstrating strong analytical and problem solving skills. Currently, Luis Ricardo Sanchez is engaging advanced real world assets focusing on and integrating emerging technologies to digitize physical assets. And I think that's that's why a lot of us are here. So let's give it up for Luis Sanchez.

Test for the present. Okay, we're in um [Music] [Music] Okay. Real world [Music] [Music] assets. Real [Music]

world. Real world assets. [Music] tokens blockchain assets. [Music] [Music] [Music]

assets. Rail assets. [Music] Physical real estates [Music] [Music] bonds, digital management. I mean participant

[Music] [Music] [Music] [Music] [Music]

[Music] Fore [Music] assets assets. participles.

[Music] [Music] Real assets. Security risk.

[Music] swap assets. [Music] [Music] assets.

[Music] systems protocols. particip [Music]

the different. Okay. [Music] Forech asset connection. [Music] [Music] [Music]

[Music] other way. [Music] asset. This asset [Music] physical market

valuation Articles. [Music] for

[Music] money particip. Enlasto generation. blockchain.

[Music] [Music] Real estateities creditsto Those real [Music] assets contract real

estate. Okay. Real estate is the real assets. Foreign speech. Foreign speech.

particip forch. [Music] [Music]

personas. All right. [Music] [Music] particip. Realizing the [Music] point global

Dubai assets contract. [Music] holders token. Real income, rental income.

Lofty AI. [Music] particularity. [Music] [Music]

physical security back security entry points for etc.

We [Music] assets legal bridge. for raw material tokenization. Insurance policies.

for exist. [Music] [Music] holders. [Music] [Music]

asset to holders receiving. [Music] [Music] [Music]

of real world [Music] assets. Blockchain [Music] [Music]

Fortnox. Anyway, transparent blockchain infrastructure. [Music] Regulated and verification of field

backing. [Music] KYC particip. Forever. The real assets

[Music] and assets. Alundo digital ERC 36 each week

crosschain tokens the different blockchain Compliance, efficiency, accessibility, assets. [Music]

house. ERC 3643 Multi-judicial compliance Real real world assets.

[Music] River assets. [Music] [Music] Real estate, commodities, equity, pensions, credits, 3643 different Transfer to

assets differential assets blockchain. is don't trust. Verify assets.

Great great great. [Applause] Fore I think

[Music] templates [Music]

contracts. Which is Real world assets. Experience. here. This one [Music]

I want to Okay. asset. Instruments etc.

One more. Right. Answer. Okay. Anybody else? No. Oh, let's give it up for Luis Sanchez. [Applause] As you can see, I'm receiving some instructions. Now, we have had a full

day, a lot of things. We learned a lot of stuff, but now it's time to chill and relax, wind down a little bit. So, we invite you to the courtyard. We have uh some hinekans. We has uh we have a little jazz music until 6. And then we also invite you to the hackathon which is just two blocks away. It's a three-minute walk. And we have the inauguration of the ETH canal hackathon as well over there uh for mentors to meet the teams. We also have a little uh if you're a little shy, we have some uh media media people in there as well.

Anyhow, we're going to have some uh chicken wings and more beers and if you're hungry, you're going to have a little, you know, uh some cheese and stuff like so fun stuff just to network, get to know each other, build community. Sounds good. Yeah, let's do it then. See you tomorrow. See you later. Oh, those that like poker also, Betsky is inviting to a poker night. You can try the um the platform, everything.