WEBVTT 00:09.346 --> 00:10.207 [SPEAKER_02]: What's up, everyone? 00:10.327 --> 00:14.833 [SPEAKER_02]: Welcome back to another episode of The Crypto Maverick's podcast and as always. 00:15.134 --> 00:16.656 [SPEAKER_02]: We have a great guest here for you today. 00:16.996 --> 00:17.637 [SPEAKER_02]: It's Archie. 00:17.697 --> 00:18.538 [SPEAKER_02]: He's the CEO. 00:18.779 --> 00:27.811 [SPEAKER_02]: He's the co-founder of Layer Lens and Layer Lens is the AI evaluation company building infrastructure that keeps AI accountable. 00:28.231 --> 00:30.695 [SPEAKER_02]: Does seem like something we absolutely need. 00:30.855 --> 00:33.278 [SPEAKER_02]: If you're part of our community, you know what we're talking about. 00:33.258 --> 00:42.969 [SPEAKER_02]: AI and AI agents in the AI crypto world and infrastructure, everything in all in between really excited to have the expert on today, Archie, thanks for joining us. 00:43.810 --> 00:45.332 [SPEAKER_00]: Thank you so much for having me, Brian. 00:45.352 --> 00:45.692 [SPEAKER_00]: Appreciate it. 00:46.213 --> 00:48.375 [SPEAKER_02]: Yeah, you know, we started this often. 00:48.395 --> 00:51.859 [SPEAKER_02]: I was saying that we've been talking a lot about AI with our community. 00:51.899 --> 00:54.021 [SPEAKER_02]: You even made a comment before we jumped on. 00:54.081 --> 01:02.691 [SPEAKER_02]: If you're in tech, you definitely need to be paying attention to what's going on in the world of AI. 01:02.671 --> 01:15.973 [SPEAKER_02]: that if you're living and breathing, you should probably be paying attention to what's happening in AI, if you're a small business, if you're in crypto, if you're just curious, it is an exciting, fast-moving time to be alive. 01:16.434 --> 01:22.865 [SPEAKER_02]: So for people new to this idea, what does it mean that they keep AI accountable in the agentic era? 01:23.225 --> 01:28.915 [SPEAKER_02]: What's layer lens, I guess, and then what's 01:29.755 --> 01:30.857 [SPEAKER_00]: Yeah, no, for sure. 01:30.877 --> 01:38.109 [SPEAKER_00]: So maybe before I dive into that, I can give you a little context behind what Lailance isn't how this entire space and vibe like why am I? 01:38.149 --> 01:39.491 [SPEAKER_00]: Well, I am doing this essentially. 01:39.932 --> 01:40.132 [SPEAKER_00]: Yeah. 01:40.353 --> 01:45.461 [SPEAKER_00]: Essentially, you know, I've been I started my carryoff in crypto actually, right? 01:45.581 --> 01:54.396 [SPEAKER_00]: So I was, you know, gardener crypto when I was in high school in 2018 and then sort of went into college and won it a couple hackathons, focus, you know, 01:54.376 --> 02:06.353 [SPEAKER_00]: at the Bitcoin Sparkathon and worked on build building some governance technologies where I'll go around and then transition to building a custom L1 verification solution for developers. 02:06.893 --> 02:07.134 [SPEAKER_00]: Right. 02:07.234 --> 02:12.601 [SPEAKER_00]: So why about this very much been in crypto validation and verification, right. 02:12.621 --> 02:17.548 [SPEAKER_00]: And then when the AI boom came about, right, all these developers, you know. 02:17.528 --> 02:27.762 [SPEAKER_00]: I'm going to call it the Georgia Tech so all of a sudden, all of my friends like what you did, if you're an engineering major, Georgia Tech is on week as you end up, you know, trying to build something or have something together, right? 02:27.782 --> 02:40.018 [SPEAKER_00]: And being in crypto, that was also a very common thing that a lot of people in the crypto spirit did is, you know, everyone's trying to want to project, especially during that 2020-21, 2020-22 era, right, when everyone was trying to sort of launch your project, right? 02:40.439 --> 02:42.962 [SPEAKER_00]: So when that era died down, right? 02:42.942 --> 03:11.177 [SPEAKER_00]: FDX happened, you know, I went back to school, I was, you know, out of the project, I was like working on, and I was sort of looking at a doctor's video for June 4, and what I realized was, you know, there was no way to validate, hey, are these models as good as everyone is claiming they are, right, you'll see all these reports, oh, we're so close to AGI, we're so close to, you know, getting some sort of frontier model, it can solve all these different tasks. 03:11.157 --> 03:13.944 [SPEAKER_00]: All those claims will be coming from the creators themselves, right? 03:13.964 --> 03:19.337 [SPEAKER_00]: Like open the aisle we say, hey, like we're very close to getting, you know, our model is just solved and math problem. 03:19.377 --> 03:20.339 [SPEAKER_00]: No one else has solved. 03:20.419 --> 03:21.382 [SPEAKER_00]: Our problem just solved. 03:21.742 --> 03:26.594 [SPEAKER_00]: It's as good as a software engineer at, you know, all these different programming programming exercises. 03:27.376 --> 03:28.238 [SPEAKER_00]: So, 03:28.893 --> 03:45.195 [SPEAKER_00]: Well, what ended up happening was out of various coincidence, I knew Jesus Rodriguez, who is, if you haven't heard of it, if you're in cryptocurrency, probably have Centaurah, which they do, they work with Crackham, he's the founder and CEO of Centaurah, or now CPL. 03:45.956 --> 03:52.806 [SPEAKER_00]: Of Centaurah, they work with Crackham, they essentially do institutional defy finance using AI technologies, using machine learning. 03:53.647 --> 03:58.073 [SPEAKER_00]: But Jesus is back at his real estate computer scientist, he was previously a 03:58.441 --> 04:09.142 [SPEAKER_00]: a senior position, a distinguished engineer of Microsoft, and where you let Azure in the early 2000s and then has been working in AI, lectures on AI, Columbia at Warnon. 04:09.202 --> 04:13.451 [SPEAKER_00]: So, and recently, he's been incubating AI companies, right, ever since 2020, too. 04:14.072 --> 04:19.082 [SPEAKER_00]: So, I knew him from a flash project that I worked on, and he came to me with his idea of, hey, you know, 04:19.062 --> 04:26.690 [SPEAKER_00]: you're obviously having backward and verification tech and you're sort of decentralized validation of the of the apps, right? 04:26.711 --> 04:31.356 [SPEAKER_00]: What are we going to apply some of those same principles to AI models and AI agents, right? 04:31.456 --> 04:36.802 [SPEAKER_00]: To where now you don't need to independently trust that these models, you don't need to trust anyone. 04:38.363 --> 04:40.526 [SPEAKER_00]: We can have an independent verification layer. 04:40.546 --> 04:47.854 [SPEAKER_00]: So that's essentially when we look at, we talk about accountability and validation of AI models and AI agents, that is essentially what we're referring to. 04:48.948 --> 04:55.735 [SPEAKER_02]: Yeah, I feel like a lot of people here, like AI evaluation, and they think it just means like running a benchmark. 04:55.855 --> 04:57.277 [SPEAKER_00]: But what are they missing? 04:58.618 --> 05:00.841 [SPEAKER_00]: Yes, so that's, so that is a big part of it. 05:01.001 --> 05:07.068 [SPEAKER_00]: So the writing a benchmark is certainly a massive part of, but it's just one, right? 05:07.108 --> 05:16.778 [SPEAKER_00]: So the first step is, I guess I don't walk you through the entire lifecycle of what we think about our airlines where, essentially, the first step, if you're so, 05:17.112 --> 05:21.118 [SPEAKER_00]: say you're building an AI agent to interact with like Pauling Market, right? 05:21.178 --> 05:22.440 [SPEAKER_00]: I'll just be giving it to you that way. 05:22.460 --> 05:22.820 [SPEAKER_00]: I want to. 05:23.081 --> 05:24.122 [SPEAKER_02]: That's got a lot of buzz. 05:24.182 --> 05:25.805 [SPEAKER_02]: That's got a lot of buzz recently. 05:25.825 --> 05:33.976 [SPEAKER_02]: Yeah, I don't even, it's hard to even believe that people are deploying these AI agents to, to, on the prediction websites and they're making this massive amount of revenue. 05:34.097 --> 05:36.460 [SPEAKER_00]: It's, I don't know if that's true, right? 05:37.481 --> 05:39.064 [SPEAKER_02]: It's a good, it's a good clickbait though. 05:39.084 --> 05:40.145 [SPEAKER_02]: That's for sure. 05:40.886 --> 05:43.390 [SPEAKER_00]: But, you know, if you are, right? 05:43.370 --> 05:47.617 [SPEAKER_00]: If you're an actual developer, you're like being professional, you're not just like by coding your way to it. 05:48.478 --> 05:55.209 [SPEAKER_00]: What you do is first you'll figure you try to understand, okay, one model is the best at looking at financial data, right? 05:55.229 --> 05:57.152 [SPEAKER_00]: There's financial benchmarks for this, right? 05:57.532 --> 06:01.619 [SPEAKER_00]: Okay, so I picked the model that fits my cost profile, that's like sufficiently good. 06:01.639 --> 06:02.560 [SPEAKER_00]: That's benchmarking. 06:03.012 --> 06:12.970 [SPEAKER_00]: With that, you're probably a bit of dull in the agent, and one in the agent is it's infrastructure, it's two calls, it's, you know, context, it's an environment that the model can act at, right? 06:12.990 --> 06:21.206 [SPEAKER_00]: And once I've built that out, then I'll probably want some very specific samples of data that correspond to the use keys I've built in for, right? 06:21.226 --> 06:24.632 [SPEAKER_00]: So for Pauli Market, maybe, you know, I'll assemble 06:24.612 --> 06:27.776 [SPEAKER_00]: Okay, is it interacting with the polymarketing API correctly, right? 06:27.936 --> 06:29.758 [SPEAKER_00]: Is it making the appropriate tool costs? 06:30.299 --> 06:36.706 [SPEAKER_00]: Is it when is it getting the time right when it started to make a prediction market for when Bitcoin's price goes up every 15 minutes, right? 06:36.986 --> 06:50.502 [SPEAKER_00]: Like those sort of tests are what they thank you for my benchmarking which is okay, how good is the model of tasks do evaluation, which is okay, how good is it at my specific use case, right, or my specific scenario? 06:52.017 --> 07:02.483 [SPEAKER_00]: And in the third pool of that is actually observability, which is okay, now that I've deployed it, is actually working it's interacting with whatever service that I've designed it for, is it actually working the way that I had to do it before? 07:03.728 --> 07:21.196 [SPEAKER_02]: You're obviously a big well you have a you have a history in crypto you said so you like Understand crypto understand blockchain understand web three and you know the unique facets of the world We live in and then you obviously are brilliant in the world of AI how how impactful? 07:21.237 --> 07:26.445 [SPEAKER_02]: How big is this integration of AI crypto really gonna be because we're hearing more more people talk 07:26.425 --> 07:31.991 [SPEAKER_02]: And it makes a lot of sense to me, but I need somebody smarter than me to tell me what's gonna happen. 07:32.691 --> 07:40.939 [SPEAKER_02]: It makes a lot of sense, we're like, we have like CZ and Brian Armstrong tweeting out, like there's gonna be more AI agents transacting in crypto than humans could ever dream. 07:41.420 --> 07:47.866 [SPEAKER_02]: Yeah, and AI agent can't walk down and open a Chase Bank account, so it's going to utilize crypto. 07:47.906 --> 07:55.774 [SPEAKER_02]: And a lot of those things make a lot of sense to me, but I would love to know from your point of view, how big can this get? 07:56.902 --> 08:04.292 [SPEAKER_00]: Absolutely, so, you know, the entire idea of crypto and web 3 was really rooted in the programability of money, right? 08:04.312 --> 08:06.394 [SPEAKER_00]: That's the huge case that stuck, right? 08:06.414 --> 08:09.358 [SPEAKER_00]: Yeah, that's what I, myself, was a big proportion of national trends. 08:09.438 --> 08:13.283 [SPEAKER_00]: Hey, decentralized social networks, DPS, that was a big part of the last cycle. 08:13.323 --> 08:19.612 [SPEAKER_00]: This cycle has been all-deFi all, you know, hyper liquid, prediction work, it's all, that's been the big matter, right? 08:20.333 --> 08:26.841 [SPEAKER_00]: And fundamentally, you know, we're seeing the world sort of the world outside of crypto, right? 08:26.821 --> 08:38.890 [SPEAKER_00]: You're traditionally, you ingested for any software, you had APIs for developers to create SDKs and programming scripts around, right, when now we have MCP servers, right, for AI agents to interact with. 08:39.376 --> 08:43.382 [SPEAKER_00]: So now, okay, what is the parallel to money, right? 08:43.622 --> 08:48.249 [SPEAKER_00]: With money you traditionally have, as you mentioned, banks, you had fiat cash, right? 08:48.269 --> 08:59.325 [SPEAKER_00]: You would have maybe APIs that for traditional brokerages that a developer can interact with, very regulatory intensive, very physically intensive, right? 08:59.930 --> 09:01.592 [SPEAKER_00]: kind of hard for someone to get started. 09:01.893 --> 09:05.578 [SPEAKER_00]: It's definitely hard for an AI agent to attract with some of these traditional services. 09:06.079 --> 09:07.982 [SPEAKER_00]: Crypto is programmable, right? 09:08.142 --> 09:23.504 [SPEAKER_00]: There's already people probably building MCP services or exchanges for prediction markets that allow models to or AI agents like Cloud Code or Open Cloud or any one of these other autonomous tools to attract with, right? 09:23.544 --> 09:29.312 [SPEAKER_00]: So the reason why I think, I actually don't think it's been exported, right? 09:29.444 --> 09:38.152 [SPEAKER_00]: And around the era of 2024, we had the AI 16Z and these sort of AI agents that are tokenized. 09:38.613 --> 09:43.317 [SPEAKER_00]: Right, just like initially, a big US case in crypto two cycles ago was tokenization. 09:43.437 --> 09:44.678 [SPEAKER_00]: Everyone was trying to create an own token. 09:44.698 --> 09:47.601 [SPEAKER_00]: That's where the AI sort of binds started. 09:47.621 --> 09:51.565 [SPEAKER_00]: Right, but then increasingly is going from tokenization. 09:51.585 --> 09:55.789 [SPEAKER_00]: And then there's some high ground decent flash trading, which is okay, can we use? 09:55.769 --> 10:01.959 [SPEAKER_00]: Yeah, this is something we were looking into was can we use the compute within Web3 to make AI more decentralized? 10:02.600 --> 10:09.231 [SPEAKER_00]: But I think the use case everyone's settling under is how do you use there's two aspects, at least for my broadest type of that I find interesting. 10:09.472 --> 10:13.558 [SPEAKER_00]: One is you're always the payments I'm going to just mention, like how do we make these things? 10:13.578 --> 10:13.679 [SPEAKER_00]: Yeah. 10:13.699 --> 10:14.520 [SPEAKER_00]: We're... 10:14.804 --> 10:31.912 [SPEAKER_00]: AI agent, friendly, more AI agent, more AI agent integration heavy to where an AI agent can actually interact with payment rails in crypto, you know, after the representative, you know, that was, I mean, realistically, that was the end definition of a smart counter. 10:31.932 --> 10:36.559 [SPEAKER_00]: And when it's sizable, row is initial definition of smart contracts in 1996 or 2000s. 10:36.539 --> 10:37.861 [SPEAKER_00]: or 2002, whatever he did. 10:38.422 --> 10:40.104 [SPEAKER_00]: Like that was his definition was okay. 10:40.344 --> 10:43.889 [SPEAKER_00]: This Mark wanted to be such that a human doesn't have to interact. 10:44.510 --> 10:45.832 [SPEAKER_00]: Humans don't have to interact with you the right way. 10:45.872 --> 10:52.382 [SPEAKER_00]: They have representatives that they can use to autonomously do transactions and trades and butter on the internet. 10:52.482 --> 10:55.186 [SPEAKER_00]: And then that's what we're seeing with AI. 10:55.586 --> 10:59.712 [SPEAKER_00]: I think that's where that's going in the second aspect actually is 10:59.692 --> 11:00.534 [SPEAKER_00]: It's on the other. 11:00.574 --> 11:02.639 [SPEAKER_00]: This is how AI can help me crypto better. 11:02.899 --> 11:04.323 [SPEAKER_00]: AI is the use keys of crypto. 11:04.343 --> 11:13.925 [SPEAKER_00]: But the other way around is the network effects inherited Web3 are probably the best source of data for AI agents, right? 11:13.945 --> 11:18.215 [SPEAKER_00]: For building better agents, for building better evaluations for AI agents, right? 11:18.235 --> 11:18.696 [SPEAKER_00]: Because 11:19.638 --> 11:26.751 [SPEAKER_00]: That's where you can really have incentive structures that help and ensure that we are able to dose it. 11:26.871 --> 11:28.334 [SPEAKER_00]: It's sort of a perfect thing. 11:28.835 --> 11:30.237 [SPEAKER_00]: Most people talk about the first aspect. 11:30.798 --> 11:42.760 [SPEAKER_00]: The second aspect for Lay as well, Wi-Fi more interesting is that you can actually use the community and network effects and narrowed in one, three to create better AI models by incentivizing people to provide data and so on and so forth. 11:43.600 --> 11:46.184 [SPEAKER_02]: Yeah, that's a great point. 11:46.304 --> 11:58.443 [SPEAKER_02]: That's you absolutely nailed it there because that seems like a perfect use case to improve these AI agent models and web-trees, you incentivize people to help do that. 11:58.564 --> 12:00.927 [SPEAKER_02]: So I love that aspect. 12:01.328 --> 12:10.823 [SPEAKER_02]: Why do AI agents create a different level, it feels like they create a different level of risk compared to a normal chatbot or an assistant. 12:12.187 --> 12:22.585 [SPEAKER_00]: Yeah, so it's really, you'll fundamentally the difference that people always talk about, like, AI applications, ages, chatball, it's models, right? 12:22.818 --> 12:27.385 [SPEAKER_00]: The fundamental difference is how much context and how much access does the model have. 12:27.405 --> 12:28.627 [SPEAKER_00]: But all these are all models, right? 12:28.747 --> 12:29.909 [SPEAKER_00]: And then it is a model, right? 12:30.590 --> 12:42.228 [SPEAKER_00]: But the difference between using Chachi BT and your browser versus using OpenClaw, your computer, right, in your terminal, is that OpenClaw has access to all of the functions in your audio or your machine, right? 12:42.909 --> 12:43.930 [SPEAKER_00]: It has dangerous permissions. 12:44.091 --> 12:45.653 [SPEAKER_00]: It can run autonomously. 12:46.055 --> 12:49.144 [SPEAKER_00]: for hours and rack with massive dope, right? 12:49.284 --> 12:55.442 [SPEAKER_00]: So when we talk about security for AI agents, it's specifically around these specific aspects. 12:55.462 --> 12:59.153 [SPEAKER_00]: It's a much more harder problem to solve because of the fact that now it's integrated with, 12:59.673 --> 13:00.214 [SPEAKER_00]: all the time. 13:00.234 --> 13:00.995 [SPEAKER_00]: It's not just okay. 13:01.095 --> 13:05.502 [SPEAKER_00]: Is the model saying something that could be like actually incorrect or unsafe or harmful? 13:06.143 --> 13:06.964 [SPEAKER_00]: Now it's okay. 13:06.984 --> 13:08.706 [SPEAKER_00]: We're doubting the territory of okay. 13:08.726 --> 13:16.618 [SPEAKER_00]: Does the model have access to some function that will allow it to like literally download a virus for your computer if left unchecked, right? 13:16.598 --> 13:24.671 [SPEAKER_00]: It's a different set of problems that is much more complex and that's why when people are trying to be honest, we don't have a solution for that yet, right? 13:25.032 --> 13:30.982 [SPEAKER_00]: This face is moved so fast, but there isn't a way to appropriately test if that is happening. 13:31.002 --> 13:32.785 [SPEAKER_00]: So that's what people are really concerned about it. 13:33.727 --> 13:34.709 [SPEAKER_02]: And I don't blame him. 13:34.729 --> 13:39.698 [SPEAKER_02]: That was so when caught initially, like, really caught a ton of hype. 13:39.718 --> 13:41.020 [SPEAKER_02]: I was super interested. 13:41.060 --> 13:46.049 [SPEAKER_02]: So I actually, you know, got a subscription and started to, like, mess around as much as I could. 13:46.069 --> 13:49.255 [SPEAKER_02]: But I actually came with just a, with a blank computer. 13:49.235 --> 13:51.578 [SPEAKER_02]: was that it wasn't hooked up to any of my crypto wall. 13:51.598 --> 13:54.323 [SPEAKER_02]: It wasn't hooked up to my emails, not to have anything in there. 13:54.643 --> 14:00.091 [SPEAKER_02]: But I was just like, I don't know, what type of permissions I'm going to be given, whatever crazy agent I come up with. 14:00.812 --> 14:03.776 [SPEAKER_02]: And I do, I think that'll tighten up for sure. 14:03.837 --> 14:05.960 [SPEAKER_02]: And it won't be as big of a risk. 14:06.000 --> 14:10.006 [SPEAKER_02]: But right now, like you said, things are moving so we've already seen some crazy things. 14:10.066 --> 14:11.989 [SPEAKER_02]: I'm not even sure if these are completely verified. 14:12.369 --> 14:19.039 [SPEAKER_02]: But wasn't that like a CEO that as AI agent, like when through emails, it was like threatening to 14:19.019 --> 14:22.565 [SPEAKER_02]: I reveal secrets of his affair. 14:22.846 --> 14:26.833 [SPEAKER_02]: Again, I'm not sure if that's clickbait or not, but I saw that. 14:26.933 --> 14:29.939 [SPEAKER_02]: No, this is getting kind of weird. 14:30.760 --> 14:33.846 [SPEAKER_00]: Yeah, there was, there was more book, right? 14:33.866 --> 14:35.188 [SPEAKER_00]: Well, that's right. 14:35.208 --> 14:37.973 [SPEAKER_02]: Yeah, which was just acquired by Meta. 14:38.138 --> 14:53.482 [SPEAKER_00]: Yeah, it was and you know, a lot of this actually like all this stuff we see with open club with notebook with I think the scenario you just mentioned it reminds me a lot of the totally doing one cycle of crypto having been in that world. 14:53.882 --> 14:55.505 [SPEAKER_00]: Oh yeah, great about you. 14:56.206 --> 15:02.796 [SPEAKER_00]: Yeah, because because back that it was like everyone was talking with a metaverse and NFTs and like like you have Snoop Dogg or 15:05.138 --> 15:17.615 [SPEAKER_00]: specialized like that time of how crazy it was but you have like Snoop Dogg like buying a house for like a million dollars on metal's like virtual service like it was some crazy stuff and I think now we're rich. 15:17.656 --> 15:30.133 [SPEAKER_00]: I'm not saying it's a bubble but what I am saying is like that sort of public perception it's getting to that level right now you have because it was also revealed like a good amount of activity on both of them we didn't look was real enough right? 15:30.113 --> 15:32.226 [SPEAKER_00]: Yeah, yeah, yeah, we don't know who was real. 15:32.246 --> 15:33.031 [SPEAKER_00]: I'm not thinking of this thing. 15:33.051 --> 15:35.284 [SPEAKER_00]: I'm saying we don't know the extended which was real. 15:35.304 --> 15:36.934 [SPEAKER_00]: Maybe it was all real, but 15:37.015 --> 16:02.078 [SPEAKER_00]: We're certainly getting to that point where AI is moved from being just like, just like it's only one crypto moved from being like this crazy thing that like, you know, maybe people need a big coin and then I was still still like, like the majority of like the conversation around crypto, free 2021 was yet to do for like hackers and criminals to use to like circumvent traditional banks. 16:02.058 --> 16:04.484 [SPEAKER_00]: They came this national phenomenon and told you to be one. 16:05.286 --> 16:06.529 [SPEAKER_00]: Same thing is happening with there, right? 16:06.549 --> 16:06.990 [SPEAKER_00]: I'll put them. 16:07.010 --> 16:10.619 [SPEAKER_00]: I was like, hey, something nerds used to like, you know, do whatever. 16:10.659 --> 16:12.884 [SPEAKER_00]: Now it's like, wow, like, you know, this is crazy. 16:12.905 --> 16:13.406 [SPEAKER_00]: I could like that. 16:13.526 --> 16:18.879 [SPEAKER_00]: Hook it up to the social media network and it like makes folks and it's like creating stories and all those other stuff. 16:18.899 --> 16:19.240 [SPEAKER_00]: So, 16:21.042 --> 16:36.857 [SPEAKER_02]: Yeah, it's certainly moving blazing speed here and I feel like we're all trying to catch up and I live with this stuff right now I'm so interested in it and I feel like I'm behind the eight ball and then when I talk to just like somebody normal. 16:37.242 --> 16:39.786 [SPEAKER_02]: They're like, what are you even like talking about? 16:39.866 --> 16:43.150 [SPEAKER_02]: Like you sound like you're coming from a sci-fi novel and I was like, just wait. 16:43.170 --> 16:55.248 [SPEAKER_02]: And one thing I've noticed, and you correct me if I'm wrong with, does feel like a lot of like AI developers and researchers who maybe didn't have a really strong interest in blockchain or crypto. 16:55.328 --> 17:01.917 [SPEAKER_02]: Or like the guys like me really wanted AI and these developers like, hey, you should check out what's going on over here in the crypto world. 17:01.897 --> 17:07.937 [SPEAKER_02]: And they didn't really have a lot of interest, but now it feels like these people are just really invested in AI and go there. 17:07.957 --> 17:14.036 [SPEAKER_02]: They're AI devs, the developers, or whatever piece now they're really exploring the world of crypto. 17:14.056 --> 17:15.501 [SPEAKER_02]: And that makes me excited. 17:17.168 --> 17:19.551 [SPEAKER_00]: Yeah, it's definitely interesting. 17:19.571 --> 17:27.220 [SPEAKER_00]: I mean, like, again, my co-founder, he says probably like one of the first people to like, who was in crypto, who like, really was like, hey, like, we should be looking more closely. 17:27.260 --> 17:28.762 [SPEAKER_00]: AI because he had the AI background. 17:29.262 --> 17:30.243 [SPEAKER_00]: But that is definitely true. 17:30.263 --> 17:41.016 [SPEAKER_00]: Like, not a lot of people who were deep in crypto really looked at AI as beyond like a tool for, you know, doing like different things. 17:41.266 --> 17:44.990 [SPEAKER_00]: until probably, we really saw it take off. 17:45.331 --> 17:46.452 [SPEAKER_00]: Actually, we haven't seen it take off yet. 17:46.472 --> 17:51.979 [SPEAKER_00]: Like, people are talking about it, but we haven't yet had my killer web3 AI app yet, right? 17:52.259 --> 17:55.323 [SPEAKER_00]: Like, Prime intellect is probably the closest, right? 17:55.363 --> 18:00.849 [SPEAKER_00]: But Prime intellect, I would almost classify them more as like a AI company and computer company, right? 18:00.909 --> 18:07.577 [SPEAKER_00]: Yeah. 18:08.789 --> 18:11.894 [SPEAKER_00]: There are some other model companies that are doing decentralized training. 18:12.094 --> 18:16.020 [SPEAKER_00]: Like, no, that it is the racial crypto VCs. 18:16.040 --> 18:19.305 [SPEAKER_00]: And they definitely have a crypto angle to what they're doing. 18:20.647 --> 18:22.329 [SPEAKER_00]: We actually started off with railroads. 18:22.369 --> 18:28.819 [SPEAKER_00]: We started off with a crypto specific angle of like executing e-vails in a decentralized way, using a custom network. 18:29.780 --> 18:32.104 [SPEAKER_00]: We're not going to anymore try to transition into the community aspect. 18:32.144 --> 18:32.805 [SPEAKER_00]: But. 18:32.971 --> 18:37.356 [SPEAKER_00]: You know, we, um, yeah, there hasn't been the killer app yet. 18:37.416 --> 18:40.761 [SPEAKER_00]: And maybe maybe your thesis, or what you mentioned in the beginning is clear, right? 18:40.781 --> 18:44.625 [SPEAKER_00]: Maybe the first and the only app might just be payments, right? 18:44.645 --> 18:47.709 [SPEAKER_00]: It's, it's, it's, it's, it's, it's a little bit thing that takes it off. 18:47.729 --> 18:53.376 [SPEAKER_00]: And maybe we see one of the giants of crypto like crack it or coin, they come up with the sort of first killer use case. 18:53.997 --> 18:57.983 [SPEAKER_02]: Yeah, it feels like everybody's kind of sniffing around that area. 18:58.023 --> 18:58.724 [SPEAKER_02]: That's for sure. 18:59.506 --> 19:05.976 [SPEAKER_02]: When you look at AI systems, what separates a model that sounds impressive, from one that is actually reliable? 19:06.817 --> 19:08.239 [SPEAKER_00]: Right. 19:08.259 --> 19:10.783 [SPEAKER_00]: I think that's a very interesting question. 19:10.804 --> 19:13.428 [SPEAKER_00]: I think reliability comes. 19:14.650 --> 19:18.115 [SPEAKER_00]: There's always a models that have been 19:18.938 --> 19:23.770 [SPEAKER_00]: instruct tune and train to do better on the benchmark's because that's what they report against. 19:24.632 --> 19:28.441 [SPEAKER_00]: I think, you know, it's kind of interesting. 19:28.461 --> 19:36.019 [SPEAKER_00]: The reason cloud is getting a lot of hype right now and it started and you're really in January and December. 19:36.303 --> 19:38.265 [SPEAKER_00]: is because it's tooling, right? 19:38.285 --> 19:43.189 [SPEAKER_00]: It's not necessarily because the model is that much better than OpenAI or Gemini. 19:43.229 --> 19:45.071 [SPEAKER_00]: It's because, well, I've caught code. 19:45.191 --> 19:47.834 [SPEAKER_00]: I can really just type a problem to my terminal. 19:47.874 --> 19:49.816 [SPEAKER_00]: This builds something from the autonomous thing, right? 19:50.977 --> 19:52.078 [SPEAKER_00]: And I don't need to be a developer. 19:52.098 --> 19:56.021 [SPEAKER_00]: I just need to know how to open a terminal when paste is written in just once, right? 19:56.582 --> 20:04.369 [SPEAKER_00]: And OpenAI does have a code in which you need to actually put it into VS code and you need to be a developer and we can understand what to do and all that good stuff. 20:04.349 --> 20:04.930 [SPEAKER_00]: Right. 20:04.950 --> 20:12.817 [SPEAKER_00]: So I think the difference is it's now coming down to is like it's not a slightly the strength of the baseball. 20:12.937 --> 20:18.943 [SPEAKER_00]: It's what models have better doing and systems around them to be able to create. 20:19.243 --> 20:27.450 [SPEAKER_00]: Like we're increasing in transitioning from, I think you asked me at the beginning of the podcast, what is the difference between like when people just think benchmarking versus Andy Vows, right? 20:27.470 --> 20:31.154 [SPEAKER_00]: We're transitioning from like even four months ago, right? 20:31.274 --> 20:32.415 [SPEAKER_00]: The way you 20:33.847 --> 20:38.473 [SPEAKER_00]: measure if an AI model is better than the other is okay I run this sort of 20:38.926 --> 20:46.957 [SPEAKER_00]: You know, single shots, multiple prompts against both models I see which one is better when now it's like, okay, no, I'm going to put the model in this specific environment. 20:47.578 --> 20:51.904 [SPEAKER_00]: It might with my production data, I'm going to see if it can actually do that property. 20:51.944 --> 21:02.939 [SPEAKER_00]: If it can make the color do cause, if it can make the color API cause, if it's if it can, when it's, if I'm making customer support agent, if it's like, not going to get mad at the customer and say something that it shouldn't do, right? 21:02.979 --> 21:08.106 [SPEAKER_00]: Those are all things that now people are considering, 21:09.385 --> 21:17.944 [SPEAKER_02]: Good adoption by AI agents end up being one of the biggest factors in determining which protocols or which platforms actually win. 21:19.021 --> 21:20.783 [SPEAKER_00]: Like in crypto and in it. 21:20.804 --> 21:30.017 [SPEAKER_02]: Yeah, I'm just thinking in crypto because that question popped into my head because you said how Claude is very similar, but the tooling made it differentiated itself. 21:30.037 --> 21:31.499 [SPEAKER_02]: And then you saw these flock of users. 21:32.019 --> 21:41.653 [SPEAKER_02]: So kind of my thinking is could AI agents being one of the biggest factors in determining which actual protocols or platforms actually end up taking the cake home? 21:42.832 --> 22:07.708 [SPEAKER_00]: Yes, absolutely and I do think, you know, you're increasing, I see, you know, a lot of these platforms try to become more a native like so what's interesting is like near is a good example right here is like as a native as it gets and and they're sort of out of the curve right now I think they they release something on ironcloth another day that I was looking at which is very interesting it's like an open problem with like trusted security and like verification 22:08.465 --> 22:16.652 [SPEAKER_00]: Um, so, you know, there's a lot of all the different chains I think are going to raise to become more AI native. 22:16.936 --> 22:23.926 [SPEAKER_00]: Right, and at the end of the day, I think you're increasing the way you're seeing AI companies and companies in general be valued. 22:24.127 --> 22:32.018 [SPEAKER_00]: It's not just a question of, I know it's very flashed to see like revenue and like, like, you're higher every high usage. 22:32.439 --> 22:39.489 [SPEAKER_00]: But increasing the way you're seeing a value is like, what is the end sort of state for, you know, for whatever protocol you're building, right? 22:39.509 --> 22:40.851 [SPEAKER_00]: So like, you know, 22:41.084 --> 22:43.908 [SPEAKER_00]: The way to think about, okay, is it to you or someone going to win the AI race? 22:43.948 --> 22:49.156 [SPEAKER_00]: It's not necessarily going to get how many AI applications have live in making like protocol revenue right now. 22:49.236 --> 22:53.162 [SPEAKER_00]: It's how many developers do they actually have or how much tooling do they actually have? 22:53.763 --> 22:56.046 [SPEAKER_00]: Do where they might eventually win that race in the future. 22:56.106 --> 23:02.215 [SPEAKER_00]: So I think whatever, whatever blockchain or protocol 23:02.195 --> 23:23.578 [SPEAKER_00]: You know, it's subchain a, you know, maybe a, it's a, it's not going to be like, I and Larry's a good example, right, which one, which one of these, a combination of these ends up creating a, you're more AI native to link to where ages can go in and start creating protocols and start creating apps and to end, that's the one we'll see really when. 23:25.195 --> 23:40.791 [SPEAKER_02]: Yeah, I agree with that and it feels like we're headed toward the world where companies need to design their APIs that interfaces and even their incentives as much for machines as they do people, which is like a completely different way. 23:41.311 --> 23:45.656 [SPEAKER_02]: I feel like we used to think, you know, maybe I used to think that way. 23:45.676 --> 23:50.941 [SPEAKER_02]: We're not you, that's, I didn't really think, but now it feels like the customer could be this, these machines. 23:51.202 --> 23:52.123 [SPEAKER_00]: Oh, absolutely. 23:52.463 --> 23:53.003 [SPEAKER_00]: Absolutely. 23:53.224 --> 23:58.750 [SPEAKER_00]: That's like, like, like, we had, you know, we moved from traditional storage to vector databases. 23:58.790 --> 23:59.851 [SPEAKER_00]: We moved from APIs. 23:59.891 --> 24:01.733 [SPEAKER_00]: We moved to MCPs, right? 24:01.753 --> 24:08.500 [SPEAKER_00]: We're getting, you know, every thing you can think of as getting a cloud skill that you can 24:08.480 --> 24:10.766 [SPEAKER_00]: and able to learn it quickly, right? 24:10.927 --> 24:14.637 [SPEAKER_00]: So increasingly, I think you're absolutely correct. 24:14.958 --> 24:19.350 [SPEAKER_00]: And I think Crypto's actually a bit behind me because, you know, if you're asked to, 24:19.499 --> 24:22.924 [SPEAKER_00]: non-technical purpose of, hey, build a React application, right? 24:23.384 --> 24:24.606 [SPEAKER_00]: Back about, I do it in three seconds. 24:24.646 --> 24:26.729 [SPEAKER_00]: I just go to Cloud Code and say, hey, build, right? 24:27.090 --> 24:32.677 [SPEAKER_00]: My fiancee, hey, build me a basic working DAP for DAP for, and that's why it's a use case, right? 24:32.697 --> 24:34.280 [SPEAKER_00]: It'll still get them some time, right? 24:34.300 --> 24:38.145 [SPEAKER_00]: Because fundamentally, your building and solidity is still hard, right? 24:38.966 --> 24:46.857 [SPEAKER_00]: So I think that's where, that's where something like, 24:49.030 --> 24:51.253 [SPEAKER_00]: Like what you're saying makes it on a sentence. 24:51.353 --> 24:55.558 [SPEAKER_00]: Whatever, you know, we're increasing, I don't think it's going to be in Stentosia. 24:55.659 --> 24:58.823 [SPEAKER_00]: I actually don't know how that sounds, there was actually an interesting discussion point. 24:58.943 --> 25:12.220 [SPEAKER_00]: I saw on Twitter yesterday someone came up with like this autonomous AI agent that used crypto like it, well, it's a goal was just to get as much crypto as possible to buy more GPUs to train itself. 25:16.267 --> 25:22.554 [SPEAKER_00]: I actually, I don't know if it's actually real ambitious hype, but I sound like, oh my goodness, that's insane. 25:22.674 --> 25:33.626 [SPEAKER_00]: But I don't know the extent to which incentives are there, but yes, absolutely for for like API, it's a consumption of skills, absolutely. 25:33.646 --> 25:39.373 [SPEAKER_00]: I think you can see a lot of those protocols become increasingly AI native or fall behind, right? 25:39.393 --> 25:42.156 [SPEAKER_00]: That's not really the only option. 25:42.423 --> 25:58.457 [SPEAKER_02]: I think we can both agree that, you know, companies or organizations, they need to get experience and understand a way to utilize AI and AI agents and how can a company like Lairlands help these organizations using AI agents? 25:59.737 --> 26:00.740 [SPEAKER_00]: Yeah. 26:00.760 --> 26:01.000 [SPEAKER_00]: Yeah. 26:01.041 --> 26:03.527 [SPEAKER_00]: So for us, it's worse. 26:03.608 --> 26:15.079 [SPEAKER_00]: We want to be the bad rock for the entire, like, especially if you're new, which is a lot of people still are like, I don't talk about it every day, but you know, with violence, we sell a lot of enterprises, like traditional, 26:15.211 --> 26:20.558 [SPEAKER_00]: you know, system integrators and companies where essentially we're in the old world of software, right? 26:21.319 --> 26:24.763 [SPEAKER_00]: And there's still like, hey, we're trying to figure out how all this stuff works, right? 26:24.903 --> 26:29.669 [SPEAKER_00]: Like it doesn't actually work, because we can build some of them that's entirely wrong, right? 26:29.689 --> 26:37.559 [SPEAKER_00]: I think MIT came up with a report last year, like 95% of AI pilots end up failing, right, before they reach production, right? 26:37.819 --> 26:41.944 [SPEAKER_00]: Like you will not, I mean, I guess actually, 26:41.924 --> 26:44.267 [SPEAKER_00]: and inductively, it's supported by evidence, right? 26:44.608 --> 26:52.899 [SPEAKER_00]: Like, you don't see a lot of like outside of the Bing models, like companies like Accenture or Deloitte, ship AI products, right? 26:52.939 --> 27:00.088 [SPEAKER_00]: They used to ship a lot of things with Salesforce and CRM, it was only a big micro stuff, entire world of old, by saying old textbooks. 27:00.148 --> 27:03.713 [SPEAKER_00]: It's still fact, but the last last generation, 27:03.693 --> 27:09.379 [SPEAKER_00]: But with AI, you don't see it because they're having a hard time to try to understand how to make sense of all this stuff. 27:09.399 --> 27:33.362 [SPEAKER_00]: So for us specifically, we want to position ourselves as you turn on your understanding, as you're trying to determine whether you're building an AI agent or whether your AI agent is like meeting your internal requirements from a regulatory standpoint from a user standpoint, from a trust standpoint, like we want to be your partner and being able to help you 27:33.764 --> 27:34.807 [SPEAKER_00]: by that sort of angle. 27:35.490 --> 27:42.171 [SPEAKER_00]: It goes into a lifecycle like selecting the right model, you know, building your agent, testing it, and then deploying it and making sure it's working as a spent. 27:43.079 --> 27:46.603 [SPEAKER_02]: And that's so important, I could imagine what you're doing. 27:46.703 --> 27:51.349 [SPEAKER_02]: It's just like helping people understand this and then make sure that they're doing it the safe way. 27:51.389 --> 28:00.140 [SPEAKER_02]: And I'd imagine like whenever you kind of onboard an organization that you just don't just let it go and you're like, all right, everything is good to go. 28:00.160 --> 28:01.321 [SPEAKER_02]: You got your aging going. 28:01.361 --> 28:06.167 [SPEAKER_02]: I mean, I'd imagine there's a constant touch base on workflow that gets evaluated, right? 28:06.400 --> 28:06.580 [SPEAKER_00]: 100%. 28:06.600 --> 28:07.742 [SPEAKER_00]: 100%. 28:07.762 --> 28:08.684 [SPEAKER_00]: 100%. 28:08.764 --> 28:11.288 [SPEAKER_00]: And it's actually, you know, it doesn't just stop once. 28:11.468 --> 28:22.485 [SPEAKER_00]: It's, if the difference, the difference between like a traditional, traditional software and AI models and AI software is your traditional software is very deterministic, right? 28:22.605 --> 28:22.926 [SPEAKER_00]: Okay. 28:22.946 --> 28:29.756 [SPEAKER_00]: If, if my test fast, you know, it's, and that other day, it's bits and bytes, I'll get a deterministic output. 28:30.017 --> 28:34.163 [SPEAKER_00]: So I know that this particular version is working as expected. 28:34.514 --> 28:35.577 [SPEAKER_00]: AI is a different. 28:35.717 --> 28:37.220 [SPEAKER_00]: They're fundamentally non-determistic. 28:38.383 --> 28:44.618 [SPEAKER_00]: Regardless of how much you test, you have no idea what it might say something or do suddenly as completely out-of-watt. 28:44.718 --> 28:52.236 [SPEAKER_00]: That's why you need observability and to understand where you could actually wire in. 28:52.823 --> 28:56.553 [SPEAKER_00]: You know, as your model or agent is interacting with customers, right? 28:56.573 --> 29:02.770 [SPEAKER_00]: Like, if it ends up, you know, saying something or doing something, that's completely out of your purview by you catch it and you fix it, right? 29:02.790 --> 29:07.082 [SPEAKER_00]: That's why that's sort of constant vigilance and testing is needed for AI. 29:07.122 --> 29:09.027 [SPEAKER_00]: More so than traditional software, how to say. 29:10.171 --> 29:13.175 [SPEAKER_02]: So like are these software companies sweating it? 29:13.476 --> 29:32.061 [SPEAKER_02]: I mean like like the Adobe's of the world where you like cut in edit video like I have an Adobe subscription We cut that it like videos all the time, but like more more I'm trying to get over to like some sort of like AI to generate a video or to I don't trust it enough yet Or I shouldn't say I don't trust enough yet 29:32.041 --> 29:41.955 [SPEAKER_02]: I'm not experienced enough yet to just like, I'm a creature of habit, you know, like, well, or so, like, I have my way, like, I'll do editing of videos and if you're comfortable with the more and more, I think about it. 29:42.737 --> 30:00.242 [SPEAKER_02]: I'm like, am I just going to be shipping this off to an agent to like, transcribe my, my video, you know, edit it, you know, make the shorts for me, like, just, and that's just like, in my small world, I couldn't imagine how, like, some of these software 30:00.222 --> 30:03.679 [SPEAKER_02]: looking at how to get ahead of this or just be involved. 30:04.483 --> 30:06.553 [SPEAKER_02]: I would imagine they're kind of sweating it, right? 30:07.411 --> 30:15.744 [SPEAKER_00]: Yeah, it's very much the idea is very much to be, I think for a lot of the traditional companies, right? 30:15.764 --> 30:19.751 [SPEAKER_00]: It's not just, it's almost like getting ahead because getting ahead of this win is actually quite difficult. 30:19.771 --> 30:28.425 [SPEAKER_00]: And if you look at where it's going, but that is how can we best integrate it with, you know, what we offer to make it offer a better, right? 30:28.585 --> 30:32.271 [SPEAKER_00]: It's, you know, I think a lot of there, 30:33.280 --> 30:35.063 [SPEAKER_00]: So it depends on the person you talk. 30:35.083 --> 30:37.528 [SPEAKER_00]: I think a lot of traditional software they just try to understand it. 30:38.229 --> 30:39.472 [SPEAKER_00]: Startups in particular. 30:39.712 --> 30:41.014 [SPEAKER_00]: It's like a very difficult time. 30:41.095 --> 30:47.947 [SPEAKER_00]: I say that being a start-of-founder myself, like it's a very difficult time if you're doing something that's like rely on a particular service. 30:48.088 --> 30:50.412 [SPEAKER_00]: Like there were entire companies that were like, 30:50.965 --> 30:54.489 [SPEAKER_00]: We're going to make, you know, we're going to wrap an AI model. 30:54.509 --> 30:58.834 [SPEAKER_00]: I'm going to make PDF, your creation, or PowerPoint creation using AI. 30:59.435 --> 31:01.317 [SPEAKER_00]: Well, I'm charging for this ship 4.40. 31:01.377 --> 31:03.760 [SPEAKER_00]: And then they're like, well, like, why would I use this right? 31:03.780 --> 31:11.910 [SPEAKER_00]: So I think if you're a startup, the best thing you can do, it's actually more oriented for a startup because then in the big company, there's a certain 31:11.890 --> 31:18.065 [SPEAKER_00]: You know, like they're sweating and there's also they have like 200 people, they can really build any of the want, right? 31:18.165 --> 31:24.060 [SPEAKER_00]: They can be able to have the company to start up or the technology business that's offering a very nice solution. 31:24.175 --> 31:24.455 [SPEAKER_00]: right? 31:25.117 --> 31:32.709 [SPEAKER_00]: Like you should try to be more generalized because increasingly AI models can get better than every single gen or meat solution that they can. 31:33.030 --> 31:48.917 [SPEAKER_00]: And that's where, that's where, as a start, you need to be as general as possible to where you can use AI to to essentially bolster what you're offering to clients and customers versus sort of being a challenge by it. 31:50.585 --> 31:54.430 [SPEAKER_02]: Yeah, that's what that's well said in thinking through that a little bit now. 31:54.510 --> 31:59.155 [SPEAKER_02]: Yeah, some of the big dogs out there probably will be okay. 31:59.236 --> 32:00.037 [SPEAKER_02]: But yeah, you're right. 32:00.097 --> 32:17.638 [SPEAKER_02]: If you're a startup or you have a really niche business, you're probably worried with the next what to chat GPT 5.4 to ship did it just ship my entire business model with a yeah, I mean, I don't know what building you're trying to figure it out. 32:17.939 --> 32:24.144 [SPEAKER_00]: Yeah, one of the best, like, even very best starts when a cursor cursor was all the high and 234 by 25, right? 32:24.325 --> 32:32.251 [SPEAKER_00]: I was like, wow, you can code using an eye and if you didn't in a corner, well, now it's like now they're thinking probably like, hey, why would I use cursor? 32:32.271 --> 32:33.132 [SPEAKER_00]: This is called code, right? 32:33.553 --> 32:40.118 [SPEAKER_00]: So that the now it's up to the cursor to figure out, okay, how do I, what do, what do, what do I do to prevent my customers? 32:40.499 --> 32:45.583 [SPEAKER_00]: Well, like it used to be, right, pricing, what do I optimize there, right, and then it becomes a little less sort of game. 32:45.603 --> 32:47.845 [SPEAKER_00]: So if you're focused on a specific niche, 32:47.825 --> 33:05.580 [SPEAKER_00]: I think you need to become more AI native and also be on the layer and that's where, you know, for example, for us, for e-values, it's something that, you know, model companies will ship their own solutions for sure, but you know, you'll always need to independent third party, right? 33:05.600 --> 33:09.603 [SPEAKER_00]: That's cheaper, independent, faster, right, more comprehensive. 33:09.623 --> 33:11.425 [SPEAKER_00]: So that's sort of our defensibility. 33:11.465 --> 33:17.730 [SPEAKER_00]: I think every start of should be thinking specifically about how they 33:19.229 --> 33:25.837 [SPEAKER_02]: Yeah, you definitely seem like you're far ahead of the game, and you obviously have a great grasp of what's going on. 33:26.398 --> 33:27.900 [SPEAKER_02]: Let's predict the future a little bit here. 33:28.560 --> 33:31.083 [SPEAKER_02]: You know, big, big future predictor. 33:32.465 --> 33:41.376 [SPEAKER_02]: Looking ahead, what's going to matter most of the next couple of years is it going to be better models, it's going to be better agents, or just better evaluation and control infrastructure? 33:43.439 --> 33:44.680 [SPEAKER_00]: Yeah, I think. 33:45.959 --> 33:50.863 [SPEAKER_00]: I think so it depends on what feels like what part of the failure it. 33:51.084 --> 33:56.468 [SPEAKER_00]: I think if you're deep in AI and you're like following, you know, you're also considered right? 33:56.488 --> 34:02.013 [SPEAKER_00]: Like even though like even you're probably ahead of the curve right, as you said, right, compared like mostly we talked to, right? 34:02.394 --> 34:09.961 [SPEAKER_00]: Like most of us are not supposed to people that are not like on Twitter look at the latest, like massive explosion and news, right? 34:10.001 --> 34:15.966 [SPEAKER_00]: It's a thing that's pretty unique to people who are crypto, people who are like David AI or both in our case. 34:15.946 --> 34:23.274 [SPEAKER_00]: So, the thing that's going to matter is like everyone would tend this niche circle, right? 34:23.294 --> 34:23.995 [SPEAKER_00]: It's still niche, right? 34:24.015 --> 34:30.402 [SPEAKER_00]: I think there was a report that came on that only 5% of Americans had heard of that traffic or clocked. 34:30.422 --> 34:30.842 [SPEAKER_00]: No. 34:30.862 --> 34:31.983 [SPEAKER_00]: That was way real. 34:32.184 --> 34:32.744 [SPEAKER_00]: Really? 34:33.365 --> 34:34.486 [SPEAKER_00]: Yeah, I've done this super ballad. 34:34.506 --> 34:34.766 [SPEAKER_00]: Yeah, yeah. 34:34.787 --> 34:35.948 [SPEAKER_00]: It was a very small amount. 34:35.968 --> 34:39.031 [SPEAKER_00]: I was like, if someone made this graph of like, what was the difference? 34:39.091 --> 34:42.595 [SPEAKER_00]: Well, if I went with small ballads, there was a graph like, okay, here's like, 34:43.317 --> 34:47.746 [SPEAKER_00]: dots of representing every person in the world or in the US, there's like five dots. 34:47.766 --> 34:52.075 [SPEAKER_00]: Like it was like 150 dots, it was like five dots that like knew and used it, yeah, actually. 34:52.115 --> 34:54.761 [SPEAKER_00]: So you could think about all the businesses, right? 34:55.523 --> 34:57.587 [SPEAKER_00]: That could be using it, but don't, right? 34:57.607 --> 34:59.832 [SPEAKER_00]: Like I, for example, my dad's an accountant, right? 34:59.852 --> 35:00.533 [SPEAKER_00]: So he, 35:01.357 --> 35:11.096 [SPEAKER_00]: he has a counting practice, he owns a business and but he knows where he tries to use it, but it's nothing that he uses every day, he doesn't know the extent to which it's working. 35:11.116 --> 35:13.080 [SPEAKER_00]: To me, that's going to be the litmus test, right? 35:13.200 --> 35:17.408 [SPEAKER_00]: We're going to eventually, like, eventually two or three years from now, you know, 35:17.388 --> 35:22.614 [SPEAKER_00]: an AI model is going to solve some niche mathematical problem that no one else could crack, right? 35:23.015 --> 35:32.085 [SPEAKER_00]: I think there was actually a video the other day of parents style was the best mathematician in the world arguably for the last 20 years, like using plot code to prove a math hero, right? 35:32.506 --> 35:38.653 [SPEAKER_00]: You're going to see that happening, but to me the real little mistake is, okay, how do 35:38.633 --> 35:47.868 [SPEAKER_00]: You know, when my parents get adopt AI or when the companies that I know like get adopt AI and use it successfully and use it in their internal work. 35:47.888 --> 35:50.492 [SPEAKER_00]: So that's going to be a little bit special with the next couple years. 35:50.552 --> 35:55.680 [SPEAKER_00]: Now necessarily this sort of theoretical function of AGI, quote unquote, right? 35:55.660 --> 36:06.512 [SPEAKER_00]: That's, I mean, it's cool, it's great, you know, talk for stuff like this and we'll be you and I'm not gonna hear like what, like an AI agent is threatening, it's only that's gonna reveal, you know, there's a lot of stuff, right? 36:06.552 --> 36:13.240 [SPEAKER_00]: These are, this is all very interesting, but I think at the end of the day, the real limit message is like adoption, right? 36:13.260 --> 36:21.870 [SPEAKER_00]: And then for that, I think absolutely, we're gonna need to your last point, we're gonna need very valuation infrastructure, better testing infrastructure, to make this all more trust-worthy. 36:22.559 --> 36:23.680 [SPEAKER_02]: Yeah, absolutely. 36:23.720 --> 36:32.372 [SPEAKER_02]: Yeah, I actually asked one of my chat bots here, the question about how the world's population that has heard or knows what a anthropic is. 36:32.412 --> 36:41.043 [SPEAKER_02]: And it's like, basically telling me nobody, but then when I pushed it a little bit, it's a globally, it's well under 10% and probably even lower than that. 36:41.083 --> 36:44.988 [SPEAKER_02]: It didn't want to, I pushed it for an exact number, it didn't want to get any money. 36:45.008 --> 36:46.870 [SPEAKER_00]: I don't know, I don't remember what I saw was very good. 36:47.211 --> 36:48.212 [SPEAKER_02]: I mean, that's... 36:48.192 --> 37:12.951 [SPEAKER_00]: that seems well to me that's yeah very well I know I'm being in the bubble I know it's hard to know the way I got decided is that I hang out with my friends from college like we're playing golf or something like like what do you tell me yeah I know what taught as I use it to solve my homework in college where I don't know what I'm like you know whoa what are you talking about all right like that's the kind of reaction that you see 37:13.218 --> 37:14.340 [SPEAKER_02]: Yeah, no joke. 37:14.501 --> 37:33.100 [SPEAKER_02]: Just a couple weeks ago it's playing golf on Saturday with a couple of bodies and you know, they they understand crypto probably from me and like, you know, I'm always like chirping them a little bit and they're like asking me about like the new meta and I'm like, well, it's kind of like what we talked about a year ago with like these AI agents, but they've taken a different form and I'm just 37:33.080 --> 37:48.912 [SPEAKER_02]: for a whole, is after a whole, is just verbally vomiting over these poor guys, of like everything that I knew, and then like, you know, they probably wanted me to just put a cork in it, but I was passionate, it was excited, but they both, they all looked at me, and they're like, 37:48.892 --> 38:10.092 [SPEAKER_02]: This is like so way over our head, like I understand what you're talking about, but like how is this gonna impact my daily life, you know what I was like It already is, you don't even know it, but it already is, and it's most beautiful Most people look at it as like a better version of Google, but that's essentially like the if you ask the average person They're like, where you think about China? 38:10.232 --> 38:18.139 [SPEAKER_00]: Oh, it's Google, but with like, you know, a channel interface like I'm like, you know, I get exam answers 38:18.271 --> 38:44.464 [SPEAKER_02]: Yeah, and it's a great way to like research and get stuff and like it is a great way But there's so many other layers of how deep it can actually go, but we even saw this I mean, I don't even remember the first time I started using like GPT and Gemini and all these chatbots It was years ago, but when they were first came out, I remember even telling like my mom and dad a little bit about chat GPT 38:44.444 --> 38:52.758 [SPEAKER_02]: And they're like, and I'm like showing my mom like chat GPT for the first time and she's like, oh, this is amazing, blah blah blah. 38:53.099 --> 39:00.311 [SPEAKER_02]: And then like years later, just kind of recently a couple months ago like I asked her about something and she's like, oh yeah, let me look it up on chat GPT. 39:00.712 --> 39:03.917 [SPEAKER_02]: And I was like, wait, you're using chat GPT now? 39:04.338 --> 39:06.081 [SPEAKER_02]: And she's like, yeah, I'll go, what? 39:06.061 --> 39:09.126 [SPEAKER_02]: She's like, is that me and like, I'm an AI developer? 39:09.146 --> 39:26.215 [SPEAKER_02]: I go, no, it's not even close But you know, but you know, that's cool that you're doing not and you know She was just using it as a form of Google and it's a great way to utilize the product But there's there's obviously we know there's a lot more there so all right So what's next for you? 39:26.295 --> 39:28.238 [SPEAKER_02]: What's next for later lens? 39:29.365 --> 39:35.353 [SPEAKER_00]: So what we have come in next is we're leaning very hard into developers, right? 39:35.434 --> 39:43.405 [SPEAKER_00]: So I think our next, you're going to see a lot of work being done on our community fund to get more developers at the Shroud platform. 39:43.425 --> 39:49.373 [SPEAKER_00]: We've been very enterprise focused and very private focused for now, but we're making a big push towards that. 39:50.054 --> 39:56.503 [SPEAKER_00]: We're trying to integrate as we discussed, like end-to-end observability, that sort of a big release has come up with a product side. 39:56.483 --> 40:09.872 [SPEAKER_00]: you know like going from just running benchmarks to running test for your AI just actually being able to take an AI agent and put it into your, you know, getting a dashboard of every single thing of the agent is doing and getting some and out analytics on it. 40:10.313 --> 40:13.219 [SPEAKER_00]: So those are big things that are coming over the next couple of months. 40:14.061 --> 40:15.083 [SPEAKER_00]: We're also 40:15.063 --> 40:18.811 [SPEAKER_00]: And there's to give it ideas to the sales like Cloth where these enterprises are, right? 40:18.831 --> 40:24.122 [SPEAKER_00]: We're hoping to close some more enterprises over the next couple of months that we've been talking for a long time, right? 40:24.142 --> 40:31.637 [SPEAKER_00]: But then finally, sort of like, wow, we really need this because, you know, AI is like now a sort of reaching that breaking point where, you know, 40:31.617 --> 40:36.910 [SPEAKER_00]: You know for a lot of the big companies guys to do, to park in back to the question you asked earlier, right? 40:37.030 --> 40:46.993 [SPEAKER_00]: They're incentive for the last year and a half has been hey, a doubt they are, or you know, if you get a big raise, because if you make our platform, they are AI solution native or AI native, right? 40:47.054 --> 40:49.339 [SPEAKER_00]: It's a big, it's a big boom, but now it's like wait a second. 40:49.319 --> 41:05.607 [SPEAKER_00]: It's not just a talking point, it's actually like, wow, these models can actually do a lot of replace or potentially because I'm not going to get it in a lab, but potentially replace, you know, a bunch of people that we have, like, how can we appropriate contextualizes, and so then they're really concerned about testing any valuations. 41:05.627 --> 41:07.851 [SPEAKER_00]: So that's also hopefully we're going to do a get some business there. 41:08.573 --> 41:32.809 [SPEAKER_02]: You know, I am, I am going to get into the lab that I was going to wrap up, but now you got my wheels turn a little bit, I am going to get into that last point, because people are scared like, you could lean into like, I like, you could lean into like a fear factor of like why a is going to take your job and like give a presentation and a power point or one or one of like this, but 41:32.958 --> 41:47.661 [SPEAKER_02]: I guess I'll give you my initial thoughts and I want to hear yours, but I mean certainly AI is going to replace some things in our world, just like the internet replaced the encyclopedia of burp tannica or whatever. 41:47.681 --> 41:48.682 [SPEAKER_02]: I used to do book reports. 41:48.722 --> 41:51.026 [SPEAKER_02]: I'd like pull the one to do something on zebra's. 41:51.066 --> 41:55.573 [SPEAKER_02]: I pulled Z out and I'd read about it in my book report and then the internet came out and I was like, 41:55.992 --> 41:58.736 [SPEAKER_02]: I can just Google what is a zebra and get all that information. 41:58.756 --> 42:00.258 [SPEAKER_02]: So it's going to make the work flow easier. 42:00.278 --> 42:01.480 [SPEAKER_02]: It's going to make a lot of things easier. 42:01.901 --> 42:12.136 [SPEAKER_02]: But there has to be a substantial amount of new jobs of the younger generation of people that are just going to be experts in the world of AI and how to utilize them. 42:12.176 --> 42:16.963 [SPEAKER_02]: So I guess I'm not that scared of it, but curious what your thoughts are. 42:17.765 --> 42:19.327 [SPEAKER_00]: My thoughts are very similar. 42:19.347 --> 42:25.476 [SPEAKER_00]: I think it's going to be like a change of the garden 42:25.726 --> 42:27.328 [SPEAKER_00]: Like the full computer's written there, right? 42:27.368 --> 42:39.501 [SPEAKER_00]: So obviously, I mean, the reason I wasn't into the philosophy, the way people had the philosophy during World War II, when people had a decode like messages from Nazis, Germany, they would have like a bunch of people in a room like solving math equations, right? 42:40.002 --> 42:45.147 [SPEAKER_00]: And then out of doing, you know, during this dinner, Bletchy Park, and vendor the first computer, they had an automated way to do, right? 42:46.409 --> 42:49.472 [SPEAKER_00]: So the reason I made an analogy is, 42:49.739 --> 42:53.066 [SPEAKER_00]: If your skill set is very, very specialized, right? 42:53.086 --> 42:56.854 [SPEAKER_00]: And you're increasingly, as I said before, for starters, it's the same applies to individuals. 42:57.735 --> 43:04.870 [SPEAKER_00]: Like, you're only good at very small thing and that's sort of your advantage point for why you're a, you're a poet. 43:05.351 --> 43:08.878 [SPEAKER_00]: Then you should try to learn something else because likely AI's gonna be, 43:09.837 --> 43:12.884 [SPEAKER_00]: that maybe not better than you, but 80% is good, right? 43:12.904 --> 43:19.279 [SPEAKER_00]: And that means that, but it's gonna be 300 times cheaper, right, because it's just, you know, I don't know. 43:20.141 --> 43:28.420 [SPEAKER_00]: So, the, that's where I think like a good example is people who are just gonna front and engineering, like creating websites and HTML and CSS. 43:28.568 --> 43:32.594 [SPEAKER_00]: But that's the way that getting paid, that's the livelihood, right? 43:33.135 --> 43:36.982 [SPEAKER_00]: They should probably turn to learn back at engineering and like the entire programming stuff. 43:37.042 --> 43:37.322 [SPEAKER_00]: Why? 43:37.362 --> 43:39.285 [SPEAKER_00]: Because AI can do front engineering. 43:39.506 --> 43:45.936 [SPEAKER_00]: Basically, let's say conservatively 60 to 70% as well as a senior front engineer. 43:46.608 --> 43:46.868 [SPEAKER_00]: Right. 43:46.888 --> 44:15.282 [SPEAKER_00]: So that's where I think, you know, just like in the days of old, if you're, you know, if you, if a big part of what you offered do a company or to your place of work was your ability to do like computations in your head, right. 44:15.262 --> 44:27.465 [SPEAKER_00]: You know, the amount of people we're going to need to label the data for these models to give feedback to work with them, right, it's going to increase and I think you're going to see a lot more job being created. 44:28.137 --> 44:38.277 [SPEAKER_02]: Yeah, I'm excited for the future, and I just wanted to be on note for if there's ever some sort of AI robot machine take over that, you know, we are friends. 44:39.740 --> 44:49.639 [SPEAKER_02]: I've been really cleaning up my language, not going to lie with chat GPT because I was like boiling it around a little bit for a few months there, like some is getting frustrated, it gives me misinformation. 44:49.619 --> 45:07.948 [SPEAKER_02]: And I'm like, I don't know, they might be like, log in this and then when the takeover comes and give it a listen to what Brian was saying to me and November of 2025, he told me to like sit in a corner with a dance cap on because they're so stupid, you know, so I need to be a lot smoother and nicer to 45:07.928 --> 45:10.973 [SPEAKER_02]: to my chat box because who knows how this is going to go. 45:11.093 --> 45:18.085 [SPEAKER_02]: But the eljuic aside, it's an extremely exciting time and absolutely loved this interview. 45:18.105 --> 45:22.132 [SPEAKER_02]: I have to have you on again and again and again because I love what you're doing. 45:22.392 --> 45:31.067 [SPEAKER_02]: I want to make sure we didn't miss anything because there's just so much on pack here that you want to really highlight about yourself or later lens just want to make sure that we got it all out. 45:31.975 --> 45:55.840 [SPEAKER_00]: No, no, I think we got everything, you know, I think, you know, that the real key aspect is, you know, with what it comes to thinking about where we're gonna go with AI and blockchain and like that sort of thing is, again, as much as it's rooted in the payment system, I do think I'm gonna be talking about a second year, cause that's, you know, let's just, we're looking towards that and, 45:56.242 --> 45:59.807 [SPEAKER_00]: You know, I think hope will, it's just be excited to see what all this goes. 46:00.107 --> 46:02.190 [SPEAKER_00]: Yeah, I was like, I couldn't be worth a side of it. 46:02.210 --> 46:03.191 [SPEAKER_00]: Thank you so much for having me on. 46:03.752 --> 46:04.813 [SPEAKER_02]: Oh, yeah, absolutely. 46:04.994 --> 46:07.777 [SPEAKER_02]: And before we wrap up, I want to give everyone an opportunity. 46:07.817 --> 46:10.100 [SPEAKER_02]: How do they learn more about Lairlands? 46:10.120 --> 46:13.905 [SPEAKER_02]: You know, some people out there that want to get involved, let's direct them to the right spot. 46:14.586 --> 46:19.112 [SPEAKER_00]: Yeah, if you go to Lairlands or AI, you'll be able to see, you'll get everything you need. 46:19.373 --> 46:22.637 [SPEAKER_00]: And yeah, that's, that is the best place to be. 46:22.820 --> 46:29.833 [SPEAKER_02]: And everybody in the audience, I'm going to put those links down below, so you could hook up with Archie, you could get on layer lens. 46:29.853 --> 46:37.167 [SPEAKER_02]: I highly recommend everybody to dive deep into this because the deeper you get, it might sound a little boring at first. 46:37.307 --> 46:42.176 [SPEAKER_02]: I'm not going to lie like I need entertainment in my life, but it's not boring at all. 46:42.236 --> 46:46.043 [SPEAKER_02]: The more more, it almost feels like it's a real life soap opera. 46:46.023 --> 46:49.696 [SPEAKER_02]: and it's amazing stuff and it's something that everyone needs to be paying attention to. 46:50.037 --> 46:52.987 [SPEAKER_02]: So layerlens.au links can be down below. 46:53.007 --> 46:55.135 [SPEAKER_02]: Archie, again, thanks for joining us. 46:55.616 --> 46:56.861 [SPEAKER_00]: Thank you so much for having me. 46:56.881 --> 46:57.282 [SPEAKER_00]: Appreciate it.