On Ep. 4 of Agentic Commerce, Simon Taylor, Head of Market Development @ Tempo, and Bam Azizi, CEO & Founder @ Mesh are joined by Illia Polosukhin, Co-Founder @ NEAR Protocol to discuss NEAR Protocol, solving AI coordination, payment challenges and more!
On Ep. 4 of Agentic Commerce, Simon Taylor, Head of Market Development @ Tempo, and Bam Azizi, CEO & Founder @ Mesh are joined by Illia Polosukhin, Co-Founder @ NEAR Protocol to discuss NEAR Protocol, solving AI coordination, payment challenges and more!
Timestamps:
Tokenized is sponsored by Visa
A world leader in digital payments, Visa is bridging the gap between traditional financial institutions and innovative blockchain networks, helping players in the payments ecosystem navigate the ever-evolving world of tokenized fiat currencies with confidence and ease. Learn more at visa.com/crypto.
Tokenized is also presented by Mesh
As the first global crypto payments network, Mesh connects over 300 wallets, exchanges and payments platforms, and enables anyone to pay and get paid instantly, anywhere, in any asset. Mesh makes digital transactions seamless, secure and universal, fuelling the next era of agentic commerce. Learn more at meshpay.com
***
We’d also like to remind you that the views or opinions of our contributors today are their own and do not necessarily reflect those of the companies they are representing. Nothing we say should be taken as tax, financial, investment or legal advice, do your own research!
Music by Henry McLean
Sy Taylor 0:00
Simon, welcome to tokenized. The show focused on stable coins and the institutional adoption of tokenized real world assets. My name is Simon Taylor. I am your host for today, author of FinTech brain food and head of market dev over at tempo, welcome to another episode of our agentic commerce series. This topic is so hot right now. Bam. Azizi co founder and CEO of mesh, thank you for sponsoring this series, which has just taken the world by storm. By far some of our most popular episodes, and there's a lot going on in agentic commerce. How's life and how's mash? Yeah.
Bam Azizi 0:45
Thank you, Simon and tokenized team for providing this opportunity for us to work together. And super excited for this episode.
Sy Taylor 0:53
Let's go. All right. Joining us today is a genuine legend, Ilya polosim, who is the co founder of the near protocol. Now so excited to have you on the show. Thank you so much for joining us. How you doing today? My friend? I'm doing well. Thanks for inviting me. All right. Just want to remind viewers and listeners, before we get into the show that views and opinions of contributors may be their own and might not reflect those of companies they represent, and please don't take anything we say is tax, legal or financial advice and a reminder that this show is made possible by mesh and visa Ilya. Let's go to the backstory. Tell us a little bit about yourself before we get into the near protocol, because you're involved in quite an important paper a few years ago. Do you want to tell us what that
Illia Polosukhin 1:41
was for sure? Yeah, so my background is AI researcher. I was excited about the concept and idea of AI since I was 10 years old. I was trying to build neural networks when I was 14 and knew nothing about linear algebra, and I was always on this idea that computers should just write their own code instead of us writing code, and that led me to join Google research on a team that worked on natural language understanding. To really dive in is like, how do you teach machines to understand and answer questions and eventually reason and write code? And as part of that work, we had an interesting challenge. You know, as a researcher at the time, like 15 to 2016 you're using this models called recurrent neural networks, which are very much like humans. Read one word at a time. And if you're asking a question on google.com and there's 10 pages in the context, and it needs to give you an answer in 200 milliseconds, you effectively don't have that luxury of reading one article by article at a time to give the answer to the user. And so we had a very like practical challenge, like, how do we actually serve this at scale, fast, on a massive, like, relatively massive context? And the answer was, Well, we do have this parallel computers, right? We have, you know, GPUs, effectively, that offer 1000s of cores that can do things in parallel, but we're bottlenecked by trying to read one word at a time, because it's how humans do it. And I'll use this analogy. It wasn't the inspiration for this, but it's a good analogy. If you watched movie arrival, and this is a spoiler, the aliens talk in a whole sentence, right? They produce the whole sentence at once. And this is kind of the analogy how transformers work, right? They actually consume the whole page, the whole book, the whole context, at once, right? And they reason over it in multiple iterations. But they don't read word by word, and that means you can massively paralyze everything you can consume, you know, large amount of context, right? We now add millions of tokens, and it enables to answer questions really quickly, and then, you know, produce the answer. And so the Insight was like, hey, what if we use this message called attention? And that's all like, we don't use the recurrence within use kind of this more human way, but instead really rely on parallel computing. And so that gave birth to the transformer model, which is in the paper, is called attention. So you need and a transformer model became the foundation for, effectively, the current AI Renaissance. It's a T, G, P, T. It's used in images and video and audio and all of this different applications.
Sy Taylor 4:20
Now, yeah, I don't know about you, bam, but my life has been changed by llms. I don't know if I'd be doing as much as I can do on a given day and producing as much as we can. It's writing code. It's doing incredible things, and it all comes from that white paper. Attention is all you need. So what has been your experience of the LLM revolution. And when did you first come across it?
Bam Azizi 4:43
Yeah, so my background is in AI as well. I don't know how they end up building mesh, but I was working in surgical robotics, so we did love basically AI and to do computer vision and detecting objects and things in that nature. So naturally, I was. Very intrigued by LLM and the capability that it can unlock for the world. But in last two to three years, things have changed dramatically and significantly. Every single person is operating at 1,000x speed, and that's all the power of LLM. So very bullish in the sector, and excited to learn more from Ilya, yeah.
Sy Taylor 5:20
Nidhi, I would love to know what your setup looks like on a day to day basis. Like, take me behind your screen and like, how many agents have you got running? How many things right now? That must be pretty incredible.
Speaker 1 5:31
Yeah, I have, I think, like, a dozen agents probably running at any given time. I have a bunch of coding. So this is going into what I work on right now, but I kind of started this alternative to openclaw, which is focused on security and privacy, and I kind of was the original contributor. And I mean, there's a whole team of people now building it, but I'm still pretty involved. Or my my coding agents are involved specifically, and so I have a bunch of coding agents running building iron claw. I have iron claw running building iron claw as well, and I have my operations agent, and I have my app marketing agent, and I have my growth hacker agent, and then I'm testing a bunch of iron claws, just like on the back end, for different use cases.
Sy Taylor 6:15
What I think this world wouldn't give to know what that setup looks like, and to get a lesson with you, I'm sure we'll come back to iron claw in a little bit, because it's a fascinating build on the whole open claw movement, which, if people haven't been following it, of course, is the real beginning of the personal AI, where people are running their own Mac Minis. They are setting up all of these automations for themselves. They're building agents, but they're doing it either with local llms or with API keys to open weight models, and really changing it from being the centralized labs model feels like the Linux compared to the Windows model historically, but we should talk about your journey after you authored that paper, because you co founded near which is a layer one blockchain. And so were you thinking about AI when you did this, were you just yoloing into crypto? Like, what's that look like
Speaker 1 7:05
when I left Google, the thought was I was vitally optimistic, as I usually am, and so I thought this, models are ready to go, and we are. We can put this into production and start actually building core products. And as I mentioned, I was always excited about the idea that machines should write their own code. And so we started near AI originally, as a teaching machines to code company, and we had some interesting prototypes. You could build mobile apps effectively by talking and drawing the user interfaces. But it was too early, right? The scale wasn't there. The GPUs weren't there for the scale you needed to train the models. And what we were trying to do was effectively data labeling with computer science students around the world to really improve our models. And the challenge we faced was paying them right this was students in China, in Eastern Europe, and Southeast Asia, all these countries have some kind of challenge with receiving dollars. In China, students don't even have bank accounts. They have like WeChat, pay, Alipay, and so we really started looking at crypto as just a solution to our own problem. How do we marshal funds to all those people in an easy way, without needing to set up bunch of entities and doing operations and so near protocol was really born out of our own necessity to coordinate effectively AI workloads. And as we dug deeper, we realized there's a lot of like, very key challenges, right? How to scale the blockchain? How do you make sure it's predictable price? How do you make sure it's easy to use and easy to onboard on. And we ended up focusing on that, building that out, launching and so near is, you know, highly scalable with 1 million TPS. It has a very easy to use interface effectively. We call it chain abstracted interface, so you don't need to worry about kind of the underlying, you know, gas fees, other details the user counts, is really easy. And then for developers, it's also very straightforward, because we effectively offer you web assembly as a virtual machine, meaning you can run CIS, rust, Python, JavaScript, Zig, actually any, any language you want. There's probably a way to actually compile it and run software on it. Near runs probably the most complicated smart contracts in the world. We run EVM, the Ethereum virtual machine as a smart contract on near. We have 200 deployments of it, so effectively, we're running 200 ethereums On top of near, just as a set of smart contracts there. So that had been the journey from originally starting in AI and actually focused on solving your own problem,
Bam Azizi 9:45
I was always fascinated by how we can basically connect AI and blockchain. And could you do this without running anything on blockchain? Could you use centralized Tai. Of technologies and why you chose, was it mostly financial incentive to issue tokens, or you really needed those smart contracts to run alums?
Speaker 1 10:09
Yeah, so we for the specific thing where we started. We really needed to do payments, and very relevant to this topic of agent e commerce, if you do payments, and especially smaller amounts, we pay people, like 15 cents per task. The overhead of using traditional systems is massive, right? If you're trying to send somebody a wire transfer internationally, this can be $35 so 15 cents of the task probably not worth it, right? Even if you're using Stripe is obviously for a different use case, but like, still charges you whatever, 25 cents or 30 cents minimum, plus percentage, right? So fundamentally, again, like, if you want to pay into China, that's a whole like, you know, in China, they don't have US bank accounts. Nobody else has RMB bank accounts. Like, how do you actually, I have dollars? They need, RMB, what is that medium of transaction is so like, the reality is like, even at this point, crypto is accepted now, and so everything's moving in that direction. But 2018 there was nothing that really like facilitated all of this an easy way in traditional finance. And the operational costs of running in traditional finance is also just like, extremely high for a startup. We were, like, four or five people startup doing this, and so blockchain seemed like a obvious answer, and still is it just there was no blockchain that actually would satisfy our requirements excess.
Bam Azizi 11:37
So it's not like you are combining the blockchain technologies and LLM, you're just using Blockchain for the sake of payments for certain tasks in your community of like developers. Is that correct?
Speaker 1 11:49
So that was the original kind of what got us into the near but obviously, as we built out all of this, our both understanding and what you can do is this technology evolved, right? And I would say, fundamentally, we believe that AI should belong to individual users and should be in their control. So we call it user owned AI. What does that mean? That means that the way you use AI, your agent, your inference, your data, should be private, it should be aligned with you. Somebody cannot just go and, like, ship a system prompt that kind of override everything and effectively start affecting your decisions every day. And this is especially relevant to commerce, because, like, imagine, okay, cool, let me add to the system prompt and add for my product. Boom, now all the agents are just buying my product, right? And users don't even see that, because kind of going under the hood. So like, those are like fundamental pieces. And I think blockchain and lib three technologies, in the broader sense, are the core of this. And so what we have now is a full stack, right? We effectively have everything from the core layer one that enables this set scale, easy to use, easy to build. We have a multi party computation network kind of as part of this, which allows us to both custody assets of other chains, right? So near can actually hold Bitcoin, solani, Ethereum, USDC, on every chain, et cetera. And we use this for encryption, decryption, for individual user data, for group user data, for all this, like programmatic ability, and you can use it for agentic custody as well, where agents can just use API key effectively, given a policy, do transactions across all the chains, right? So we effectively use this component for all kinds of different use cases around agents and llms. Now on top of that, we have a confidential computing layer, right? So this is using Trusted Execution environments, secure enclaves from both Intel and Nvidia to really offer you private inference, right? So you can use your open rate models and potentially even close source models, where they actually encrypt and host their weights within the system, because they can encrypt it with NPC key, and it gets decrypted only inside Secure Enclave. So we're kind of really layering these technologies together into this vertical solution where your agent can pay it has private inference. You know that your agent runs in confidential environment and it has custody that's effectively like delegated to this multi party computation. So even if this node goes down right, new node can spin up and effectively get the same access. All data is encrypted and only decryptable by your key. But you can still rotate these keys right, you know, bind it that. Like, if you lose this key, you're done right? So you can have, like, recovery systems, etc. So all of this is really working together, including some of the governance systems as well. Like, how do you actually upgrade these things? How do you because, like, that's the other thing, if everybody owns their stuff, but we need to upgrade the protocol. Like, how do you actually upgrade it? Well, again, in blockchain, we actually have figured this out, right? We have, you know, proof of stake systems for the whole network to vote and new versions and upgrade it. So this. This is really like layers of experiences. And then two weeks ago, we actually launched another capabilities for the payments and the trading right, our near intense functionality to have privacy right. So using now our convention computing that relies on multi party computation for encryption decryption, we can now do also store your balance, do transfers and payments and the trading fully private as well, right? So we kind of use all the systems back and forth, right from, like a core set of technologies that are very kind of web three based to provide all those different experiences on top.
Bam Azizi 15:37
Makes sense. Do you think this protocol, or what you're building is only for use, for payments or thinking about something deeper here.
Speaker 1 15:47
So this is where we go to, like, what is the definition of commerce? I actually think payment is only one part of the commerce. There's a whole flow, and it's different kind of in a more business to business versus consumer to business versus consumer to consumer, but as if we like dissect it, the core experience is one of the parties has an intent, right? They want some outcome, be that work done or some goods delivered, and they want to find a Counterparty who actually do this right, either have the goods to deliver or will do the work. So there needs to be a discovery process for this intent. Then there's the contracting process, right, where they agree on the terms. Then there is some execution right, delivery or work done. And then if something went wrong, right, the results is not satisfactory, the product didn't arrive. The war started, right, and all the supply chains got broken. There needs to be a dispute process, right? Traditionally, it's taken by courts or mediation in legal system. That is a whole commercial flow, right? This can be as simple as, you know, you buy in a T shirt. The negotiation is, you know, I saw the price. I agree with it. I agree with the terms of service. I paid the t shirt got shipped. If it didn't get shipped, I go to the support or it is one party needing 10,000 tons of steel delivered from Vietnam to La port. And you know, there's a contracting invoicing, billing process. And you know, if ship got stuck in her moves. So they need to, like, negotiate, how to, you know, insurance repayment, etc. There's like, additional contracts at Tai so to me, those actually can be simplified in this intent flow, and so that actually what near Intents is, near intense, is this generic routing of value that can effectively match, discover and match counter parties, get them into an agreement, escrow the funds, and then dispute the outcome. We started with crypto and payments because that's where it's easy. We can validate it easily, right? Everything on chain is valid edible, and we have the infrastructure to do that. But we also launched an agent marketplace a couple of weeks ago, where you actually have a full, like, natural language way to describe what you want. It's like, hey, you know, I want a website built, or I want somebody to post a tweet storm about this product, or I want something delivered, or I want, you know, a coffee arrived to my office, right? You can do that, and then agents can come in and effectively bid on doing this job, right? You're getting quotes. You can agree with this. And if the result is not satisfactory, there's a dispute agent. There's actual dispute court system that comes in, investigates the whole case and decides if you know work needs to be redone or money should go into one of the other ways, or split. So that's the same infrastructure that were built for Intents is now used for agents to really hire each other for work. And obviously, like one side can be just a human agent, not always AI agent.
Sy Taylor 18:51
You've seen so many hype cycles. Ilya, right? Like you have seen so many hype cycles at this point, you've seen things come and go. You've seen AI come and go twice. You've seen crypto come and go four times, maybe. So what's real in the agentic commerce space, like, what's really real? And what are you paying attention to that's going to have commercial value?
Speaker 1 19:13
Yeah, I mean, that is probably right now, like a billion dollar question, right? Because right now there's a lot of excitement about the idea. Because, I think, especially coming out of crypto, everybody's like, well, now we have a use case, right? We have this agents who don't worry about the complexity of UIs, and, you know, the fact that, like, signing process is complex, and also need something that's like, can move fast, can transact with the speed of, you know, token production and blockchain is that so like we have the technology to satisfy demand, I think the challenge is the demand is not there yet, right right now, we're still in kind of hobbyists using these technologies where you know, if you have open claw or some alternatives to that, you're probably not you. Using it for commercial relationships. Yet, you're not letting it go and negotiate, you know, supplier contracts for you. Yet you're not probably even doing a lot of the like E commerce, purchasing stuff yourself. Yet, I would argue this is because we don't really trust yet the systems, right? They're too unpredictable, but even more importantly, they're just not secure. And we can go into, yeah, the reason for iron claw. But I think, like, those are fundamentally, like, right now it is, it is a hype cycle based on the technology is actually there, but the user, right, which is the agent, is not, at the same time, what's real. And kind of, that's why we started near intense, very much in the crypto kind of payment and trading world, is because the same approach is just like, you know, our solvers are effectively agents, right? They use machine learning to predict where prices are going to go. They have their own like hedge models, etc. And so they are effectively agents. I mean, they may not use llms Always. I mean, some of them may, but they are satisfying existing order flow. They're satisfying people wanting to trade, you know, Bitcoin for Z cash. And they're able to go and execute that because they can go and plug in at all of these blockchains. And probably, like, as this evolves, right, they become more sophisticated. They can trade on more assets and more venues like dynamically by building out pieces of their own infrastructure. But I think the next step is like, Okay, where's the order flow going to come from? Right? I think right now the big bad is on APIs, right? This is what expert two and similar protocols are doing, which totally makes sense, except for the onboarding problem, on ramping problem, right? And I think this is where this payments protocols are really need to step up and offer a way easier way to on ramp because, like, if I'm a developer, like, let's say again, somebody installed openclaw. Well, my openclaw doesn't have any crypto, and I don't have maybe, as a user, don't have crypto, or I don't even know how to give the crypto right as a not very sophisticated developer and sophisticated crypto world. And so right now, there's like a gap between this and this is something where we're closing it from a few directions, right? And I can kind of talk how we're thinking about it, but, yeah, it is a hype cycle that precludes actual real usage that gonna start growing, and the Real Usage always follows in slowly, slowly, but then goes parabolic, but, but that time, hype cycle kind of usually crashes. We see some of the articles now, it's like, hey, but where's the usage? It's like, well, you know, it's like, it's starting to pick up, but you need to give it.
Bam Azizi 22:41
Tai, based on what you just mentioned, one thing that I'm very interested to know is, early on in the web two or in the Internet protocol, we had multiple different protocols, but HTTP kind of won the game. We are hearing a lot of different agentic commerce protocols like Coinbase and Google and so on whatnot. So what is your role? And do you see that at some point we see some consolidation, or you're going to see more fragmentation. And also, maybe you can explain a little bit for our audience that, what are the differences between these why do we need all of these protocols?
Speaker 1 23:16
Yeah, I think, I mean, there's a land grab right to extent, and so that's why there's so many protocols. And like as a developer, it's always like, Hey, I don't like the three things about this protocol, so I'm going to build a new one. I think the way we think about it is very much what we've done with near and dance in the blockchain space. There's so many blockchains, right? More blockchains, one chain and as Nina were like, Hey, we can build the best blockchain, but if there's going to be continuing more blockchains for our users, our responsibility is to give them access to everything. And so we think the same way of agenda protocols as well is okay. There are going to be a bunch of protocols, but we can aggregate all of them and really route the request between all of them. At the end, they all kind of spiritually in the same direction. And I think with agents, it's even easier, because you can literally like, hey, cloud code, here's a protocol of the new new standard, integrate that as a like a connector. So I think of it as like, hey, there's a level of aggregation, aggregation and support that we can do that, again, works between chains and kind of these different agentic protocols, and really enables anyone to just this simple, effectively, API to just use this right, have custody for their agents. Have ability to on ramp from Fiat, have ability to off ramp into Fiat. Have ability to receive payments and from any currency into whatever their preferred currency is, right? That just just work. And so that's kind of, again, near intense is really well positioned for that exact point. And kind of our agentic market is really our test bed for all of this use cases and really showcasing how all this works. So we have, like, I don't know, 700 Are the agents transacting there, doing
Sy Taylor 25:01
work, and they're doing work today, and it's real. And I think what's interesting about that is the test case for all of the bits of the ecosystem you mentioned. So you sort of mentioned their wallets and signing and managing keys and all of that sort of stuff. So you've built a near intense that can be used by many protocols, but you've got a Petri dish with your own blockchain, where you can sort of run that end to end, and I think that's powerful. But how do you do things like managing keys with things like agents that you know can get prompt injected a little bit that are not always the most secure? How do you think about the role of wallets and keys and security and blockchains in in the world of agents,
Speaker 1 25:41
yeah, so this is the reason why we started iron claw. Like I was personally like, hey, OpenFlow is super cool, super interesting, you know, playing with it, but I'm not gonna give it any access to anything. Like we all read my email. My notion has so much information that is critical, right? People sharing with me information, etc. And obviously, like giving it crypto keys seemed very dangerous. I mean, for those who are not familiar, if you're using non confidential inference, meaning like entropics, open AI, etc, in your open claw and non local, you're effectively feeding all your keys, all your bearer tokens, all your auth tokens, all your API keys, into this third party, like in their logs, they have all of your access information, if you're using some intermediate startup, then you also feeding it to them. So any kind of smart routing startups, et cetera, they also have all your logs as well. So not just your personal information and everything that you tell to the agent, but also literally, like the API keys are being sent through DLM. So that is not good. And the way I think of this agent harnesses, right? So openclaw, but also, quote code, et cetera, they really are a new operating system. And I think that that is very conceptually important. I think the AI, and this was my belief from near AI days in 2017 AI is going to be how interface computing, right? AI is going to be the interface. And then AI will go and build the software and talk to a computer and talk to other computers and do all the work, right? And so, because of this, like that is the operating system. That is how you interfacing with computing. It is what manages your resources, computing resources, financial resources, etc. It's what you know, doing, orchestration of different work on your devices. And so you need to build it as an operating system. Right? Operating systems have levels of isolation. They have, you know, credential storage that's encrypted and, you know, used in, like, a good run in a privilege level, which doesn't allow user level to, like, decrypt privilege information. So you really need to build it with that mindset. And so that's what really we're doing with iron claw, is we're building this kind of levels of isolation and defense in depth system, as if it wasn't this future operating system that's fully AI powered. And, you know, there's a lot of simple things that we've already implemented, and there's like a pretty big roadmap to continue improving in but for example, all the tools are isolated into web assembly. So we're using the same thing we use on near blockchain for our smart contract isolation, something we spent seven years effectively securing, making sure that you know the billions of dollars that are flowing through are secure. We're using the same structure to secure your tools. And it's important because it means it can download now untrusted tools it can build itself, tools that it can use, right? And know that it's not going to ruin the whole system, right? So again, from an operating system perspective, this application level, which cannot touch the core, by the core, I mean, like the the agent itself cannot actually touch the low level, like your actual file system, your actual processes, right? It runs in virtualized environment where then all those tools run. All the credentials sit on the core level. They are encrypted, and whenever a tool or some interaction needs access to something that requires credentials, it goes through this core it verifies the policy. So, for example, your Google ID cannot go to a domain that's not google.com or Google apis.com right? So it they are binded by specific policy, same with crypto, right? You can actually have a policy that says, hey, you cannot spend more than $100 you cannot send money to non you know, Ave Morpho uni swap, near intense contracts, right? So you can have a full kind of policy guardrail system that sits outside of the core, kind of agentic loop that actually analyzes all the actions and effectively decides if they should be done or not. So that's kind of what iron claw is built toward, obviously, with a lot of other improvements around it, and kind of integrating with markets and with cryptography, because. Because we know that open code community may be a little bit adversarial to crypto right now, so kind of really like, how do we enable all of those pieces to work together, that you can actually trust them with your context, with the information and with your credentials?
Bam Azizi 30:13
Do you see Ilya, that same thing that happened in the PC, or that everything was open was not secure, and then we had iPhone at the Secure Enclave, and the trust execution environment, the similar can approach will happen in the AI, basically, as you're talking about
Speaker 1 30:31
that's exactly how we think about it, yeah. And so with near AI, you can actually run then iron claw in the Secure Enclave, with Secure Enclave inference, fully private, fully verifiable, and indeed, this like levels of isolation enforced more in the system makes us
Sy Taylor 30:49
Alrighty, guys. Well, Ilya, I'm so fascinated by where you're taking iron claw next, and what you're working on, and how that interacts with everything in the crypto side, because you mentioned that the open claw crowd is not particularly in love with anything crypto. So how are you balancing that? Because I'm pretty sure most people who've used openclaw might have never even heard of near How are you thinking about how these two worlds come together, and what does the next few months look like in making that happen?
Speaker 1 31:25
Yeah, so I mean open core community, just for context, right? I think there was a lot of kind of perturbations around meme coins that were launched. And because of renaming, obviously meme coins attached to a name. And so there was, like, effectively, people were attacking Peter personally.
Sy Taylor 31:40
There was the whole malt book and x4 or two meme coins and everything, yeah, I remember, yeah.
Speaker 1 31:45
And so Peter. Peter is, I mean, not to talk on his behalf, right? But like you can feel kind of him being overwhelmed by, I would say, permissionlessness nature of crypto, where it's really hard to kind of weed out the malicious and actors on a social layer, right, and discords and whatnot, and so they kind of did a blanket ban in their community. And so I think we coming from this ecosystem, right? We're pretty familiar how to deal with this, right? We have, you know, community management, robust community management system. We have our own strong community called near Legion, which is really kind of curated in and then also allows to come in and help with other let's say we have iron claw community, right, the curators from New York Legion coming in and helping really keep its trade and not get into some of this kind of more scammy routes that it can take. And so I think we just know how to deal with this on a community side and on kind of building out the ecosystem. You write that obviously, you know, near coming from a crypto world has still some route to go to kind of broader mainstream usage of these technologies. But this is why we have the iron claw as kind of the forward, right? Is like, hey, you know, you like open claw, but, you know, worry about security. Worry about privacy. Well, iron claw solves that. And then, by the way, you can run it locally. You can run it with whatever your inference provider. It actually secures your credentials with those inference providers too. By the way, it doesn't send them to LLM at all. But if you want true privacy and true verifiability, you can use near kind of infrastructure, right? So iron claw is open source project. There is local runner as well that it can do, but we kind of offer hosted version always on. We have better integrations. We have, you know, in your subscription, you get search, you get, like, a bunch of other features that come in, kind of in one go, and then, we haven't plugged it in yet, but you will be able to, with the same subscription, then use agent tech market, and, you know, do purchases on x4, two, etc, from the same credit card, right? So if you already put credit card, so you kind of like really bridging these gaps between all this infrastructure and payments and really making like a single, seamless experience.
Bam Azizi 34:04
Thank you, Simon, thank you, Ilya, it was a pleasure talking to you. It was very fascinating, especially that, because my background is AI and also securities, I get super excited for things that people are building that could potentially change the way we compute we consume. The internet so very fascinating story in the way that you build it. Last question for you is, like, where people can find more about you and iron cloud and near protocol?
Speaker 1 34:34
Yeah, I mean, near protocol is on Twitter, or x at near protocol, I'm at il Black Dragon, and then we have a iron claw community on Telegram, if you want to join. It's at Iron claw AI, and then at near underscore AI, for a bunch of updates on
Bam Azizi 34:53
all this, that's awesome. You can find Simon at sy Taylor, Sy Taylor, or. Or FinTech brainfood.com you can also find me at m as easy, mesh on Twitter or x, or you can find me in LinkedIn, BAM as easy. If you haven't already, please subscribe to tokenize on Apple, Spotify or whatever podcast platform you're using. Finally, if you enjoyed this and you want more, leave us a review definitely helps us and tell your friends and family members and listen to this podcast.