1 00:00:00,080 --> 00:00:02,360 Welcome to The Epicenter, the show which talks about the 2 00:00:02,360 --> 00:00:04,800 technologies, projects, and people driving decentralization 3 00:00:04,880 --> 00:00:07,160 and the blockchain revolution. I'm Sebastian Kucho, and I'm 4 00:00:07,160 --> 00:00:09,880 here today with Ismail, CEO of LaGrange Labs. 5 00:00:10,240 --> 00:00:14,280 LaGrange believes that there is a gap in cryptography within 6 00:00:14,280 --> 00:00:16,400 national security and defence. We realized there was an 7 00:00:16,400 --> 00:00:20,880 opportunity to take things that were developed in crypto that 8 00:00:20,880 --> 00:00:25,040 are frontier hard tech and to purpose that outside of crypto. 9 00:00:25,280 --> 00:00:28,480 The zero knowledge proofs provide this way to glimpse 10 00:00:28,480 --> 00:00:31,600 inside the black box to determine that the output that 11 00:00:31,720 --> 00:00:34,800 you got back has come from the correct model with these set of 12 00:00:34,800 --> 00:00:38,360 correct inputs which allows you to determine why the decision 13 00:00:38,360 --> 00:00:41,040 was made, under what circumstances decision was made 14 00:00:41,320 --> 00:00:44,520 and under what circumstances decision should be adjusted or 15 00:00:44,520 --> 00:00:45,840 things should be changed going forward. 16 00:00:52,280 --> 00:00:55,160 I'm here today with Ismail, CEO of LaGrange Labs. 17 00:00:55,800 --> 00:00:57,080 Hey, how are you doing? Thank. 18 00:00:57,080 --> 00:00:58,520 You so much for having me, Sebastian. 19 00:00:58,720 --> 00:01:02,360 I was on Epicenter recently and there's been so much exciting 20 00:01:02,360 --> 00:01:04,319 progress we've had as a company since then. 21 00:01:04,319 --> 00:01:06,720 I'm so excited to be back and talk about a lot of that today. 22 00:01:07,720 --> 00:01:10,560 This episode is brought to you by Nosis, building the open 23 00:01:10,560 --> 00:01:14,560 Internet one block at a time. Nosis was founded in 2015 and 24 00:01:14,560 --> 00:01:17,320 it's grown from 1 of Ethereum's earliest projects into a 25 00:01:17,320 --> 00:01:20,120 powerful ecosystem for open user owned finance. 26 00:01:20,640 --> 00:01:23,600 Nosis is also the team behind products that had become core to 27 00:01:23,600 --> 00:01:27,120 my business and that are so many others like Safe and Cow Swap. 28 00:01:27,680 --> 00:01:32,080 At the center is Nosis Chain. It's a low fee layer one with 0 29 00:01:32,080 --> 00:01:35,480 downtime in seven years and secured by over 300,000 30 00:01:35,480 --> 00:01:37,720 validators. It's the foundation for real 31 00:01:37,720 --> 00:01:40,800 world financial applications like Nosis Pay and Circles. 32 00:01:41,440 --> 00:01:44,640 All of this is governed by Nosis Dow, a community run 33 00:01:44,640 --> 00:01:48,520 organization where anyone with a GNO token can vote on updates, 34 00:01:48,720 --> 00:01:51,640 fund new projects, and even run a validator from home. 35 00:01:52,200 --> 00:01:55,160 So if you're building a Web 3 or you're just curious about what 36 00:01:55,160 --> 00:01:58,680 financial freedom can look like, start exploring at gnosis dot 37 00:01:58,680 --> 00:02:01,240 IO. OK, Ismael, welcome back. 38 00:02:01,240 --> 00:02:03,760 You've been on just a couple months ago, but so much has been 39 00:02:03,760 --> 00:02:05,960 happening over at the grounds that I want to get you back on 40 00:02:05,960 --> 00:02:08,919 to talk about some of the interesting military and 41 00:02:08,919 --> 00:02:11,720 industrial applications for ZKML that you guys have been working 42 00:02:11,720 --> 00:02:12,880 on. A lot of like exciting 43 00:02:12,880 --> 00:02:14,520 partnerships have been announced. 44 00:02:14,560 --> 00:02:17,680 But I guess like before we before we start like kind of 45 00:02:17,680 --> 00:02:20,760 bouncing off of that conversation from from July 46 00:02:21,240 --> 00:02:24,720 where we talked about deep proof for folks who have who have not 47 00:02:24,720 --> 00:02:27,000 listened to that episode, we should go back and listen to it. 48 00:02:27,040 --> 00:02:32,000 Let's maybe, you know, set the context here for what is LeBron 49 00:02:32,040 --> 00:02:36,680 building and what is deep proof and what are you building in the 50 00:02:36,680 --> 00:02:38,600 space around ZK machine learning? 51 00:02:38,920 --> 00:02:40,800 Yeah. So LaGrange likes to think of 52 00:02:40,800 --> 00:02:44,600 itself as the preeminent company for frontier research in applied 53 00:02:44,600 --> 00:02:48,000 cryptography across both commercial and national security 54 00:02:48,000 --> 00:02:51,360 defense. And what that means is we are 55 00:02:51,360 --> 00:02:55,120 uniquely positioned to build 0 knowledge proofs, things like 56 00:02:55,120 --> 00:02:59,560 fully homomorphic encryption and then consensus in a way that no 57 00:02:59,560 --> 00:03:03,960 other commercial company has the capacity to do really across 58 00:03:03,960 --> 00:03:06,160 anywhere in crypto or or defense right now. 59 00:03:06,640 --> 00:03:09,640 And so the the tech we build, the core of this is something 60 00:03:09,640 --> 00:03:12,720 called deep proof, which is a zero knowledge machine learning 61 00:03:12,720 --> 00:03:15,240 library. Effectively, it takes a model. 62 00:03:15,240 --> 00:03:18,600 Think of an arbitrary machine learning model all the way from 63 00:03:18,600 --> 00:03:23,560 an LLM to a computer vision model used in a drone, or to a 64 00:03:23,720 --> 00:03:27,160 simple MLP used decision. The movement of assets on chain 65 00:03:27,680 --> 00:03:31,280 and we're able to do 2 things. We're able to firstly prove that 66 00:03:31,280 --> 00:03:35,320 that model has executed correctly by effectively 67 00:03:35,320 --> 00:03:38,160 generating A0 knowledge proof of the correctness of the execution 68 00:03:38,160 --> 00:03:40,520 of that model. The same way that AZK roll up 69 00:03:40,520 --> 00:03:43,720 proves that all the transactions executed correctly, ZK machine 70 00:03:43,720 --> 00:03:46,040 learning proves that AI executes correctly. 71 00:03:46,720 --> 00:03:49,080 And secondly, we can do this over a configurable set of 72 00:03:49,080 --> 00:03:51,160 private inputs. The model can be private or the 73 00:03:51,160 --> 00:03:54,200 input data can be private. And this is a very powerful 74 00:03:54,200 --> 00:03:59,640 feature because it in bakes privacy into how AI can be used 75 00:03:59,640 --> 00:04:03,720 across both commercial and then obviously government defense and 76 00:04:03,720 --> 00:04:07,080 national security settings. Privacy of AI is a very, very 77 00:04:07,080 --> 00:04:10,640 big question. And a lot of the incumbent 78 00:04:10,640 --> 00:04:14,600 attempts to add privacy to AI are based on you know, air gap 79 00:04:14,600 --> 00:04:19,240 systems, sequestering data, sequestering cloud environments 80 00:04:19,600 --> 00:04:23,000 but or hardware based security. But they aren't based on 1st 81 00:04:23,000 --> 00:04:24,680 principles, innovation and cryptography. 82 00:04:25,120 --> 00:04:28,240 LaGrange takes the approach that it's going to go to 1st 83 00:04:28,240 --> 00:04:30,720 principles, it's going to rebuild the mathematical 84 00:04:30,720 --> 00:04:33,840 fundamentals of how we can bake privacy into AI, and then we're 85 00:04:33,840 --> 00:04:38,240 going to purpose it as we do across dual use applications. 86 00:04:39,080 --> 00:04:41,760 Let's let's maybe just kind of talk about one of the use cases 87 00:04:41,760 --> 00:04:44,080 that people are most familiar with, which would be like chat 88 00:04:44,080 --> 00:04:48,720 bots, right? Talk to Chachi, BT or Gemini and 89 00:04:50,560 --> 00:04:54,840 how, how the inference is generated in in those cases 90 00:04:54,840 --> 00:05:00,600 versus say some of the, you know, other more private or 91 00:05:00,600 --> 00:05:05,280 privacy preserving chat bots like like Venice, for instance. 92 00:05:05,560 --> 00:05:08,520 And, and contrast that to some like to a first principles 93 00:05:08,520 --> 00:05:11,680 approach, because it sounds like what you're saying that those 94 00:05:11,760 --> 00:05:14,520 systems don't necessarily use a first principles approach. 95 00:05:14,520 --> 00:05:17,640 Is that correct? Yeah, I think Venice uses Tees. 96 00:05:17,640 --> 00:05:20,560 I mean, Venice is like a, they take open source models, they 97 00:05:20,560 --> 00:05:22,560 deploy that on a network and they make it private, which 98 00:05:22,960 --> 00:05:25,680 that's great. I, I think that's fine for some 99 00:05:25,680 --> 00:05:27,920 people. You're, you're not going to be 100 00:05:28,080 --> 00:05:31,840 having a drone swarm that needs to maintain privacy over 101 00:05:31,840 --> 00:05:35,040 computer vision inputs. That's using Venice, right? 102 00:05:35,320 --> 00:05:40,560 You're not going to have an LLM used for, you know, by a field 103 00:05:40,560 --> 00:05:46,440 operator to determine a plan of action for some type of defense 104 00:05:46,760 --> 00:05:50,600 purpose that uses Venice. These are interesting 105 00:05:50,600 --> 00:05:54,640 curiosities that I think are built by crypto companies, but 106 00:05:54,640 --> 00:05:57,280 they tend not to be commercial applications. 107 00:05:58,200 --> 00:06:02,800 I would argue that for something like Open AI or Enthropic or 108 00:06:02,800 --> 00:06:07,800 DeepMind to actually take one of their closed source proprietary 109 00:06:07,800 --> 00:06:11,440 models and add privacy on top of that, it requires a solution 110 00:06:11,640 --> 00:06:15,480 that is low touch, doesn't degrade performance and doesn't 111 00:06:15,480 --> 00:06:19,160 result in the underlying model IP being leaked, right. 112 00:06:19,480 --> 00:06:24,480 If you deploy Open AI's model on Jema, sorry, on Venice, you're 113 00:06:24,480 --> 00:06:27,320 effectively just going to just lose all the IP on that. 114 00:06:27,320 --> 00:06:29,560 The way it's the biases, the architecture all to be public, 115 00:06:29,960 --> 00:06:32,360 right? It's it's a very different 116 00:06:32,360 --> 00:06:34,680 question, right? So if you want to use Frontier 117 00:06:34,680 --> 00:06:38,040 closed source models, you have to be able to bake cryptographic 118 00:06:38,040 --> 00:06:41,440 security in at the level of the model developer. 119 00:06:41,680 --> 00:06:43,800 And that's what 0 knowledge machine learning does. 120 00:06:44,000 --> 00:06:47,480 It lets you add privacy on top of models and lets you add 121 00:06:47,760 --> 00:06:50,840 verifiability on top of models wherever they're they're 122 00:06:50,840 --> 00:06:53,080 deployed. So just just so we understand 123 00:06:53,320 --> 00:07:00,600 this correctly, a a model that has ZK properties has to be 124 00:07:00,600 --> 00:07:03,320 trained. So from the ground up, it has to 125 00:07:03,320 --> 00:07:05,200 be. Trained as AZK model, you can 126 00:07:05,200 --> 00:07:08,800 take an existing open source model and make AZK. 127 00:07:08,800 --> 00:07:11,400 No, that's not what I'm saying. What what I'm contending is that 128 00:07:11,400 --> 00:07:13,600 you don't want to use an existing open source model 129 00:07:13,880 --> 00:07:16,120 because the performance of existing open source model is 130 00:07:16,120 --> 00:07:18,600 not good. The majority of commercially 131 00:07:18,600 --> 00:07:21,080 interesting models are not existing open source models, 132 00:07:21,600 --> 00:07:23,400 right? Most people when they, they use 133 00:07:23,400 --> 00:07:30,080 chatbot, they're using clawed or they're using, you know, O5 or 134 00:07:30,080 --> 00:07:36,360 they're using, you know, Gemini, they're not using Gemma and GBT 135 00:07:36,360 --> 00:07:39,520 2, right. Those are interesting 136 00:07:39,760 --> 00:07:41,680 applications. If you really want to ask your 137 00:07:41,680 --> 00:07:45,280 LLM something you don't want the world to know, but you, you know 138 00:07:45,320 --> 00:07:48,000 you, you so care about privacy. What you're asking that you, 139 00:07:48,040 --> 00:07:50,760 you, you, you probably shouldn't be asking these questions to 140 00:07:50,760 --> 00:07:56,280 begin with. Generally where I see ZK machine 141 00:07:56,280 --> 00:08:00,480 learning useful is in applications where these are 142 00:08:00,480 --> 00:08:04,680 frontier proprietary models that need to protect both the two 143 00:08:04,680 --> 00:08:08,160 aspects of privacy, both the privacy of the person who wants 144 00:08:08,160 --> 00:08:10,760 to input data sent in and the privacy of the model. 145 00:08:11,360 --> 00:08:13,880 OK, interesting. So what you're saying is that a 146 00:08:13,960 --> 00:08:20,000 in order for AZK model to be any model, any model, right? 147 00:08:20,000 --> 00:08:26,440 Any model? The efficiency of that model and 148 00:08:26,440 --> 00:08:31,280 it's ability to be to be a private model is very much tied 149 00:08:31,280 --> 00:08:33,440 to how it's trained. Or is that? 150 00:08:33,799 --> 00:08:34,919 Are those two things uncorrelated? 151 00:08:35,240 --> 00:08:39,400 Yeah, I, I think for any AI model in the world, right, if 152 00:08:39,400 --> 00:08:41,880 you develop the best AI model to do something, you're not going 153 00:08:41,880 --> 00:08:44,440 to open source it, right, Unless you run a cherry. 154 00:08:44,840 --> 00:08:48,000 But you know, if, if you develop a great generative model for 155 00:08:48,000 --> 00:08:52,360 video, right, VO3 or Nano Banana two, you're not going to open 156 00:08:52,360 --> 00:08:54,440 source that you're not going to send it over to China, you're 157 00:08:54,440 --> 00:08:55,720 not going to give it to your competitor. 158 00:08:55,920 --> 00:08:59,480 You're going to deploy a, a website and you're going to have 159 00:08:59,480 --> 00:09:01,480 an API. And behind that API, your model 160 00:09:01,480 --> 00:09:03,520 is going to sit on a server and people are going to come in, 161 00:09:03,520 --> 00:09:05,160 they're going to ping it and you're going to charge them for 162 00:09:05,160 --> 00:09:06,720 credit. You're not going to give it to 163 00:09:06,720 --> 00:09:09,160 them for free. So you effectively need to 164 00:09:09,160 --> 00:09:11,520 protect privacy of two things. We're going to talk about AI. 165 00:09:11,960 --> 00:09:15,200 You need to protect privacy of IP intellectual property for the 166 00:09:15,200 --> 00:09:18,040 actual model. And you also in many cases want 167 00:09:18,040 --> 00:09:22,200 to protect consumer privacy. So the the data that someone is 168 00:09:22,200 --> 00:09:25,680 sending to that model, a lot of the private AI solutions are 169 00:09:25,680 --> 00:09:29,160 just taking open source models that are publicly available and 170 00:09:29,160 --> 00:09:32,560 deploying those in trust execution environments or in a 171 00:09:32,560 --> 00:09:34,600 cloud somewhere that they call air gapped. 172 00:09:35,160 --> 00:09:40,400 And they are not actually adding privacy on top of commercially 173 00:09:40,400 --> 00:09:42,640 relevant models, right? OK. 174 00:09:42,840 --> 00:09:46,000 They're they're not taking the Frontier models, Nano Banana 2, 175 00:09:46,000 --> 00:09:49,480 VO 3 clawed and they're not adding privacy on those. 176 00:09:49,480 --> 00:09:53,480 They're taking Gemma 3 with a billion parameters that nobody's 177 00:09:53,480 --> 00:09:55,120 going to use for anything, and they're deploying it as a 178 00:09:55,120 --> 00:10:00,680 curiosity. How much of ZK machine learning 179 00:10:00,680 --> 00:10:03,760 and the the ability to do privacy within models is 180 00:10:03,800 --> 00:10:06,800 hardware dependent? None of it, none of it. 181 00:10:07,800 --> 00:10:10,880 It's it's not I mean you can you get faster inference on GPU and 182 00:10:10,880 --> 00:10:12,920 you get faster proofing on GPU, right. 183 00:10:13,680 --> 00:10:16,120 So there there is obviously some questions of hardware that but. 184 00:10:16,120 --> 00:10:19,120 What I meant by hardware dependent is like how much of 185 00:10:19,120 --> 00:10:21,880 that is dependent on the availability of Ttes? 186 00:10:22,080 --> 00:10:23,440 None of it, not at all. No Zeke. 187 00:10:23,440 --> 00:10:25,000 And machine learning is entirely just. 188 00:10:25,000 --> 00:10:27,560 Mathematics got it. OK, which which? 189 00:10:27,680 --> 00:10:31,120 Which is arguably why it's more interesting because it doesn't 190 00:10:31,120 --> 00:10:36,120 require any specific hardware to have provable AI that's also 191 00:10:36,120 --> 00:10:38,760 private. Precisely interesting. 192 00:10:38,760 --> 00:10:43,080 OK. So a lot of the focus around 193 00:10:43,200 --> 00:10:45,320 Lebron's and what you guys were sort of like been focused on in 194 00:10:45,320 --> 00:10:48,320 the last couple months and certainly the like announcements 195 00:10:48,320 --> 00:10:51,640 that front of the company have been around military and 196 00:10:51,640 --> 00:10:54,480 industrial applications. Why is there such an importance 197 00:10:54,480 --> 00:11:00,560 here within these to the very highly sensitive applications 198 00:11:00,560 --> 00:11:04,480 for AI and privacy? What, what's the, what's the 199 00:11:04,480 --> 00:11:05,680 kind of edge that you guys have there? 200 00:11:06,000 --> 00:11:09,360 I love this question. So LaGrange for the majority of 201 00:11:09,360 --> 00:11:12,440 its history has been a commercial focused company. 202 00:11:12,840 --> 00:11:15,920 We've developed privacy solutions and and cryptographic 203 00:11:16,040 --> 00:11:20,160 assurance for AI usage as well as other types of computation 204 00:11:20,160 --> 00:11:22,720 that that were relevant database computation early in our history 205 00:11:23,800 --> 00:11:26,240 for solely commercial applications. 206 00:11:26,960 --> 00:11:29,200 There's a belief that I have though, and this is something 207 00:11:29,200 --> 00:11:32,080 that I've seen talked about by Alex Karp and the Technological 208 00:11:32,080 --> 00:11:33,760 Republic. He's talked about publicly by 209 00:11:33,760 --> 00:11:38,960 Palmer Luckey, a lot of businesses pursuing incremental 210 00:11:39,000 --> 00:11:42,960 and trivial applications of technology that marginally 211 00:11:42,960 --> 00:11:47,360 improve consumer life instead of materially influencing national 212 00:11:47,360 --> 00:11:49,720 security defense, right? This is a generation of 213 00:11:49,720 --> 00:11:53,000 businesses that have been, you know, developing chat apps. 214 00:11:53,520 --> 00:11:57,520 It's a generation of food delivery and, you know, marginal 215 00:11:57,520 --> 00:12:00,520 convenience improvements in in the Bay, right? 216 00:12:00,520 --> 00:12:03,240 It, it, it's the, the generation of engineers who were brought up 217 00:12:03,240 --> 00:12:06,240 to build a slightly better delivery app instead of trying 218 00:12:06,240 --> 00:12:08,760 to close the hypersonic missile gap with China. 219 00:12:09,320 --> 00:12:12,640 It, it, it, it effectively has separated arguably the most 220 00:12:12,960 --> 00:12:16,800 competitive and technologically advanced area of this country 221 00:12:17,000 --> 00:12:19,880 from all areas of national security and defense. 222 00:12:20,280 --> 00:12:24,080 And this has been like what US tech was for the better part of 223 00:12:24,080 --> 00:12:29,000 15 years right before Andrel and Shield and some of the the more 224 00:12:29,000 --> 00:12:32,640 American dynamism and and reindustrialization companies 225 00:12:32,640 --> 00:12:35,120 have started raising very large rounds the last 3-4 years. 226 00:12:35,520 --> 00:12:37,680 There was very little venture financing that deployed into 227 00:12:37,680 --> 00:12:41,200 anything in traditional sectors that was doing aerospace, 228 00:12:41,200 --> 00:12:45,480 defense or national security. And that's not a successful 229 00:12:45,480 --> 00:12:50,160 state of a country, right? What we require as a nation is 230 00:12:50,160 --> 00:12:53,120 for the most innovative people of that country, the people who 231 00:12:53,120 --> 00:12:55,800 build the best technology, who push the state-of-the-art, who 232 00:12:55,800 --> 00:13:00,600 compete to to develop the things that make our nation special. 233 00:13:01,000 --> 00:13:04,360 To purpose that not just to improve the the quality of life 234 00:13:04,760 --> 00:13:07,360 of people in Silicon Valley who want their food delivered on a 235 00:13:07,360 --> 00:13:11,560 drone, but to purpose it to increase freedom, to increase 236 00:13:12,200 --> 00:13:14,720 the hegemony of the United States globally. 237 00:13:15,200 --> 00:13:22,280 And to do that you require dual use applications of frontier 238 00:13:22,280 --> 00:13:25,040 tech. And so when I look at crypto, I 239 00:13:25,040 --> 00:13:29,960 see very much that the state of Silicon Valley 10 years ago, 240 00:13:30,320 --> 00:13:32,080 right? We've had an entire industry 241 00:13:32,080 --> 00:13:36,080 that put a billion dollars into frontier cryptography 242 00:13:36,080 --> 00:13:38,520 development for zero knowledge proofs, fully homomorphic 243 00:13:38,520 --> 00:13:40,640 encryption and consensus well over a billion. 244 00:13:40,640 --> 00:13:45,520 If you include consensus, probably a billion for ZK couple 245 00:13:45,520 --> 00:13:50,040 100 million for FHE and funding. And maybe you know, 5-6 billion 246 00:13:50,040 --> 00:13:52,960 for things that are raised with novel consensus, maybe 3 or 4 247 00:13:52,960 --> 00:13:55,840 billion. But all of that was purpose for 248 00:13:55,840 --> 00:13:58,880 one thing in crypto. How can I launch a more 249 00:13:58,880 --> 00:14:02,880 efficient token? And so we've put effectively the 250 00:14:02,920 --> 00:14:07,360 the America's best in cryptography, consensus 251 00:14:07,480 --> 00:14:11,840 mechanism design, distributed systems, all of this not the 252 00:14:11,840 --> 00:14:14,800 building things that make America safer, make America more 253 00:14:14,800 --> 00:14:18,280 secure, allow us to dominate internationally, but we've 254 00:14:18,280 --> 00:14:22,640 instead purposed at the things that just flip a quick buck for 255 00:14:22,640 --> 00:14:27,280 investors wanting a token. When we looked at this, we 256 00:14:27,280 --> 00:14:30,400 realized there was an opportunity to take things that 257 00:14:30,400 --> 00:14:34,760 were developed in crypto that are frontier hard tech and to 258 00:14:34,760 --> 00:14:38,760 purpose that outside of crypto, not just to commercial, but for 259 00:14:38,760 --> 00:14:42,160 dual use for defense, national security, military and 260 00:14:42,160 --> 00:14:46,520 government, and and that's why LaGrange started doing this. 261 00:14:47,320 --> 00:14:50,240 We have 3 professors in applied cryptography in the US, Our 262 00:14:50,240 --> 00:14:52,960 research teams led by Babis Papa Monttu, who chairs the 263 00:14:52,960 --> 00:14:56,360 cryptography department at Yale. We have 8 PhDs in applied 264 00:14:56,360 --> 00:15:01,040 cryptography on our team. We have multiple patents we file 265 00:15:01,040 --> 00:15:02,960 in the US on novel proof systems. 266 00:15:03,200 --> 00:15:05,840 We have 5 or 6 papers we've authored on the company's 267 00:15:05,840 --> 00:15:08,960 history. And yet like many other 268 00:15:08,960 --> 00:15:14,200 companies, all we were purposing is for was launching tokens. 269 00:15:15,280 --> 00:15:18,320 Now I think there's a tremendous commercial opportunity to 270 00:15:18,320 --> 00:15:21,800 instead purpose that not just us but the whole industry, the 271 00:15:21,800 --> 00:15:23,160 things that are a lot more important. 272 00:15:25,320 --> 00:15:27,920 Yeah. I mean, I think it resonates 273 00:15:27,920 --> 00:15:31,680 with me this this idea that a lot of our a lot of the 274 00:15:31,680 --> 00:15:34,200 innovation in tech in the last 20 years has been around 275 00:15:34,200 --> 00:15:38,320 increasing convenience. Maybe this is also sort of a 276 00:15:38,320 --> 00:15:43,640 reflection of just like how cheap capital has been in the 277 00:15:43,640 --> 00:15:51,000 last, you know, 20-30 years and also like a relatively peaceful 278 00:15:51,000 --> 00:15:54,440 state of the world in the last 20 years. 279 00:15:54,440 --> 00:15:55,520 I mean, certainly since the Cold War. 280 00:15:55,520 --> 00:15:57,800 Debatable. Yeah, right. 281 00:15:57,800 --> 00:16:01,480 I mean debatable, but I mean like people's everyday sort of 282 00:16:01,480 --> 00:16:06,120 like considerations were not were no longer whether or not, 283 00:16:06,200 --> 00:16:08,520 you know, a nuclear bomb would be dropped on their heads where 284 00:16:08,520 --> 00:16:10,960 which argued it was the case, you know, for like a generation 285 00:16:10,960 --> 00:16:17,160 before us now that that has that that the world dynamic is 286 00:16:17,160 --> 00:16:20,880 definitely different now. And I think that the that is 287 00:16:20,880 --> 00:16:26,280 reflected in the fact that there's more capital for, for 288 00:16:26,280 --> 00:16:28,360 military and industrial applications. 289 00:16:29,480 --> 00:16:32,520 But the, the return profile on those, I think are much 290 00:16:32,520 --> 00:16:37,600 different than the, you know, the delivery app type of things 291 00:16:37,600 --> 00:16:39,360 for investors. How do you think that's going to 292 00:16:39,360 --> 00:16:42,840 change? I mean, already I think we're 293 00:16:42,840 --> 00:16:45,760 seeing that that change in, in, in capital funding in crypto. 294 00:16:47,600 --> 00:16:50,800 How do you think that's going to change the entire sort of VC 295 00:16:50,800 --> 00:16:56,440 landscape where you know, the, the time to return is probably a 296 00:16:56,440 --> 00:17:00,680 lot longer in a defense company that's building, you know, 297 00:17:00,680 --> 00:17:04,920 drumstorm technology over, you know, delivery app that's going 298 00:17:04,920 --> 00:17:08,480 to be sold at private equity in like a couple years or maybe 299 00:17:08,720 --> 00:17:10,680 sold off on like the next funding round? 300 00:17:11,200 --> 00:17:12,680 Yeah, I I would actually disagree. 301 00:17:12,680 --> 00:17:18,480 I, I think that the reforms in procurement have allowed defense 302 00:17:18,480 --> 00:17:22,359 and national security focus companies to see substantial 303 00:17:22,359 --> 00:17:27,480 growth at rates that are comparable to even the delivery 304 00:17:27,480 --> 00:17:29,760 apps, right? You look at public markets, 305 00:17:29,760 --> 00:17:32,640 Palantir has been one of the fastest growing companies on the 306 00:17:32,640 --> 00:17:35,520 equity markets. I mean, Palantir is a comp of, 307 00:17:35,680 --> 00:17:38,240 you know, what you want your public company to do and you 308 00:17:38,240 --> 00:17:41,560 compare Palantir's performance over the last, you know, let's 309 00:17:41,560 --> 00:17:46,640 say 48 months compared to Uber, DoorDash and other more 310 00:17:46,640 --> 00:17:48,080 convenience oriented companies there. 311 00:17:48,080 --> 00:17:50,280 There's there's no question that you would you would you would 312 00:17:50,280 --> 00:17:52,920 want your money parked in Palantir. 313 00:17:54,160 --> 00:17:55,760 It's it's a hard tech problem, right? 314 00:17:55,760 --> 00:17:59,880 You look at another dual use company, SpaceX, right? 315 00:18:00,160 --> 00:18:03,640 SpaceX is IPO and it's going to be one of the largest IP OS of 316 00:18:03,640 --> 00:18:06,200 all time, right? I'm here people saying 1 1/2 317 00:18:06,200 --> 00:18:08,680 trillion has to be the valuation, right? 318 00:18:08,840 --> 00:18:12,760 I mean, these are, these are incredibly hard to fathom 319 00:18:12,760 --> 00:18:15,680 outcuts, right? I mean, Uber is what like a 50 320 00:18:15,680 --> 00:18:20,800 billion market cap business? And you know, Palantir, I think 321 00:18:20,800 --> 00:18:25,280 is over a trillion, SpaceX could be over a trillion and the rolls 322 00:18:25,280 --> 00:18:28,360 going to are going to come out probably in, within 24 months. 323 00:18:28,760 --> 00:18:31,240 And it's a once again going to be a fantastically large 324 00:18:31,240 --> 00:18:32,440 business. Their growth has just been 325 00:18:32,440 --> 00:18:35,960 tremendous. I I think the, the fact that the 326 00:18:35,960 --> 00:18:40,760 prime defense contractors, the, the Lockheed, the Raytheon, the 327 00:18:40,760 --> 00:18:44,240 Northrop Brummins, the General Dynamics, the General Atomics, 328 00:18:44,560 --> 00:18:48,800 they have much more competition now and is a much larger 329 00:18:48,800 --> 00:18:50,680 willingness within the Department of War and other 330 00:18:50,680 --> 00:18:54,960 agencies to purchase technology and to, to, to solicit bids for 331 00:18:54,960 --> 00:18:58,760 technology from non traditional defense contractors. 332 00:18:59,280 --> 00:19:02,520 And that is going to mean significantly more innovation, 333 00:19:02,840 --> 00:19:05,920 significantly more alignment between US private sector 334 00:19:05,920 --> 00:19:10,600 innovation and U.S. military and US defense capabilities. 335 00:19:11,880 --> 00:19:14,200 I think that just starts with, with the willingness of any 336 00:19:14,200 --> 00:19:16,920 administration to to, to purchase defense relevant 337 00:19:16,920 --> 00:19:21,400 technology from from non prime contractors. 338 00:19:21,400 --> 00:19:23,840 We've seen that, that, that that is a trend. 339 00:19:24,120 --> 00:19:28,680 So I do think the return profile there is going to be very, very 340 00:19:28,680 --> 00:19:30,160 strong. And I think that's why we've 341 00:19:30,160 --> 00:19:31,840 seen so much investment in this, right. 342 00:19:32,240 --> 00:19:34,760 I think last week it was, I can't remember the name 343 00:19:34,760 --> 00:19:37,040 hypersonic missile company raised that was closing the gap, 344 00:19:37,040 --> 00:19:41,600 you know, 3400 million and you don't see rounds like that for 345 00:19:41,600 --> 00:19:44,360 things outside of American dynamics that are just, you 346 00:19:44,360 --> 00:19:48,280 know, foundation models in Web 2 venture right now. 347 00:19:50,400 --> 00:19:53,680 As as as you were talking, I was looking at the market cap for 348 00:19:53,680 --> 00:19:57,280 Lockheed Martin and Raytheon. I had no idea what these were, 349 00:19:57,280 --> 00:20:03,240 but yeah, Raytheon is like 240 billion, Lockheed's 100 billion. 350 00:20:04,960 --> 00:20:06,280 Kind of interesting though. That's. 351 00:20:08,120 --> 00:20:12,640 That software companies in the space like Palantir are, you 352 00:20:12,640 --> 00:20:15,600 know, chasing valuations that's like over a trillion. 353 00:20:16,240 --> 00:20:17,840 Or I think 4:50. I just looked it up. 354 00:20:18,280 --> 00:20:21,600 I was wrong. Oh, OK, 450 billion for yeah, 355 00:20:21,600 --> 00:20:23,640 450 billion, yeah for Palantir, Yeah. 356 00:20:24,720 --> 00:20:30,040 While the, you know, the guys building jets are, are far below 357 00:20:30,040 --> 00:20:33,400 that. Like, I guess like in terms of 358 00:20:33,400 --> 00:20:37,320 the, the, you know, the geopolitical aspect here. 359 00:20:37,320 --> 00:20:40,400 You, you wrote somewhere, I think it was like a corn market 360 00:20:40,400 --> 00:20:44,520 cap article. I've got the quote here that was 361 00:20:44,520 --> 00:20:46,200 interesting. So superiority and software has 362 00:20:46,200 --> 00:20:49,280 become the most efficient way to sustain US leadership worldwide. 363 00:20:50,400 --> 00:20:53,080 I mean, I think that's true. I mean, to a large extent, a lot 364 00:20:53,080 --> 00:20:56,480 of that software superiority has gone into things that are maybe 365 00:20:56,840 --> 00:20:59,800 more convenience driven and, and you know, now that the world is 366 00:20:59,800 --> 00:21:02,720 changing or like people are, there's a reaction to that. 367 00:21:02,720 --> 00:21:06,280 And that is going now into more military and industrial 368 00:21:06,280 --> 00:21:11,280 applications, you know, to the extent that that has been 369 00:21:11,280 --> 00:21:18,400 possible because of the availability of chips due to, 370 00:21:18,400 --> 00:21:21,400 you know, somewhat from the commercial relationships with 371 00:21:21,400 --> 00:21:24,320 China and you know, by extension Taiwan. 372 00:21:24,640 --> 00:21:27,640 The ability for the US to be superior in software has been 373 00:21:27,640 --> 00:21:29,720 dependent on its on access to chips. 374 00:21:30,000 --> 00:21:33,480 And that access is now, you know, possibly compromised in 375 00:21:33,480 --> 00:21:36,920 2027. Yeah, I, I, I, I would say 376 00:21:36,920 --> 00:21:39,760 broadly that the, the, the ability to be superior in 377 00:21:39,760 --> 00:21:42,800 software does come out of chips, but it also comes down to energy 378 00:21:42,800 --> 00:21:47,800 independence, which which China also has risks if they, if they 379 00:21:47,800 --> 00:21:50,280 move on Taiwan, right. The majority of Chinese oil 380 00:21:50,280 --> 00:21:53,520 moves through the Malacca Strait, which you know, 381 00:21:53,720 --> 00:21:58,000 effectively if, if, if Chinese energy was was significantly 382 00:21:58,000 --> 00:22:03,120 impaired, having large amounts of, of chip access or would not 383 00:22:03,120 --> 00:22:07,000 necessarily improve their, their global position, software or AI 384 00:22:07,000 --> 00:22:09,400 development. But yes, it would substantially 385 00:22:09,400 --> 00:22:13,480 harm the United States position in AI development if, if, if 386 00:22:13,480 --> 00:22:17,160 chip access from Taiwan was significantly reduced or cut off 387 00:22:17,800 --> 00:22:19,720 and chip manufacturing in Taiwan was cut off. 388 00:22:19,920 --> 00:22:21,960 And I think there's there's a recognition of this and there's 389 00:22:21,960 --> 00:22:25,280 efforts that, that both private companies and then obviously the 390 00:22:25,280 --> 00:22:28,880 US government is taking to improve domestic chip 391 00:22:28,880 --> 00:22:31,240 manufacturing and build foundation for US electronics 392 00:22:31,240 --> 00:22:33,600 chip manufacturing. But it's very unlikely that this 393 00:22:33,600 --> 00:22:37,120 will catch up to the capacity that that that that Taiwan can 394 00:22:37,120 --> 00:22:41,680 produce within a reasonable period of time. 395 00:22:42,840 --> 00:22:45,200 There's there is, there is the, I mean, there is like the time 396 00:22:45,200 --> 00:22:48,480 it takes to reach capacity, which I think, you know, we're 397 00:22:48,480 --> 00:22:50,880 probably talking about, you know, in, in the decades, 398 00:22:50,880 --> 00:22:53,120 decades. But there's also access to the 399 00:22:53,120 --> 00:22:57,440 rare minerals that make those trips possible to produce. 400 00:22:57,440 --> 00:23:03,720 And to I think mostly are Chinese is, is the West kind of 401 00:23:04,120 --> 00:23:11,600 by that by by that fact doomed long term unless there are there 402 00:23:11,600 --> 00:23:17,120 are, you know, sustainable business relationship with 403 00:23:17,120 --> 00:23:19,520 China, commercial relationship with China that allow them to 404 00:23:19,520 --> 00:23:24,040 purchase these these rare metals and and also manufacture chips 405 00:23:24,040 --> 00:23:25,720 at a scale that allows us all to grow. 406 00:23:26,240 --> 00:23:28,880 No, I, I would, I would go the other way. 407 00:23:29,200 --> 00:23:35,560 And I would say that if China is not able to affect a military 408 00:23:35,560 --> 00:23:39,400 action against Taiwan by 20-30 and they've, they've, they've 409 00:23:39,400 --> 00:23:42,760 publicly said 2027 is the, the, the point that they're going to 410 00:23:42,760 --> 00:23:45,880 do this. And most intelligence analysts 411 00:23:45,880 --> 00:23:51,560 expect 2027. China has a substantial risk of 412 00:23:51,640 --> 00:23:55,760 never being able to and therein significantly reducing its 413 00:23:56,000 --> 00:23:59,360 prominence internationally. The Chinese naval's effectively 414 00:23:59,360 --> 00:24:00,920 bottled up in the South China Sea. 415 00:24:01,280 --> 00:24:06,200 They're blocked off by the ring of islands consisting of Japan, 416 00:24:06,200 --> 00:24:11,400 Taiwan, and the Philippines. All of their oil moves to 417 00:24:11,400 --> 00:24:13,560 Malacca Straits. Their population is aging. 418 00:24:13,840 --> 00:24:16,760 Within 15 years, the average population in China is going to 419 00:24:16,760 --> 00:24:19,160 be the same as the average population in Japan, which is 420 00:24:19,160 --> 00:24:23,560 not very good. Chinese equities have not been 421 00:24:23,560 --> 00:24:25,640 doing well. The problem with the SEO system, 422 00:24:26,280 --> 00:24:28,640 with state owned enterprises that that that China has is that 423 00:24:28,640 --> 00:24:32,720 they are they are not as competitive at implementing and 424 00:24:32,720 --> 00:24:35,160 developing new technology as US counterparts. 425 00:24:35,400 --> 00:24:37,680 They don't have an open and free market economy. 426 00:24:37,680 --> 00:24:41,480 There is brain drain from China. There's a lot of intellectual 427 00:24:41,480 --> 00:24:44,160 talent in China that would rather study and and work in the 428 00:24:44,160 --> 00:24:48,120 United States in a free country than they would within the 429 00:24:48,160 --> 00:24:50,280 hierarchy of a state owned enterprise in China. 430 00:24:50,280 --> 00:24:53,120 I actually don't think that China's position is necessarily 431 00:24:53,120 --> 00:24:56,200 very strong. We also have seen Chinese 432 00:24:56,200 --> 00:25:00,680 military equipment fail. Pakistan was mostly using 433 00:25:00,680 --> 00:25:02,880 Chinese military equipment during the recent conflict with 434 00:25:02,880 --> 00:25:04,520 India and most of the equipment didn't work. 435 00:25:06,400 --> 00:25:10,160 With the recent strikes by Israel on Iran, most of the anti 436 00:25:10,160 --> 00:25:13,720 aircraft systems that were being used by Iran were Chinese and 437 00:25:13,720 --> 00:25:17,760 most of them didn't work. So there's levels of grift and 438 00:25:17,760 --> 00:25:21,480 corruption and inefficiencies in the Chinese system that are 439 00:25:21,480 --> 00:25:26,200 intrinsic to how the Chinese government structures control 440 00:25:26,240 --> 00:25:29,480 and enterprise. And I think a lot of that boils 441 00:25:29,480 --> 00:25:34,960 down to historical issues, issues over freedom and issues 442 00:25:34,960 --> 00:25:38,360 over free markets. And so I, I think that the 443 00:25:38,360 --> 00:25:42,240 reindustrialization narrative that the US private sector has 444 00:25:42,240 --> 00:25:43,920 been financing very aggressively. 445 00:25:44,120 --> 00:25:47,760 So building US domestic manufacturing capacity, building 446 00:25:47,840 --> 00:25:51,600 AI manufacturing, building the capacity to produce the mined 447 00:25:51,600 --> 00:25:55,400 refined rare minerals in the United States, building chip 448 00:25:55,400 --> 00:25:58,480 production in the US, building drone production in the US, 449 00:25:58,680 --> 00:26:01,120 building decentralized manufacturing so that they can 450 00:26:01,120 --> 00:26:04,960 tested time. There's a very hard to knockout 451 00:26:05,280 --> 00:26:11,240 manufacturing base plus energy independence, small modular 452 00:26:11,240 --> 00:26:15,320 reactors. US is very far ahead, I would 453 00:26:15,320 --> 00:26:20,160 argue, in the economic structures that will allow it 454 00:26:20,160 --> 00:26:23,240 over significant periods of time to innovate and outperform 455 00:26:23,280 --> 00:26:26,680 China. One area that that I think is, 456 00:26:26,720 --> 00:26:30,840 is, is, is kind of interesting is this is well, I mean, I guess 457 00:26:30,840 --> 00:26:33,960 a military development that has really accelerated since the war 458 00:26:33,960 --> 00:26:37,320 in Ukraine. Is, is, is drone technology and, 459 00:26:37,320 --> 00:26:41,320 and, and particularly they the use of a low cost drones to 460 00:26:41,320 --> 00:26:47,400 achieve like all, all of the, sort of like all the, all the 461 00:26:47,400 --> 00:26:51,160 positive advancements we've seen in, in, in, in Ukraine there. 462 00:26:51,360 --> 00:26:55,400 And, and it seems like, I mean, from where I'm sitting, the US 463 00:26:55,400 --> 00:26:59,480 approach has been to kind of like throw lots of money at this 464 00:26:59,480 --> 00:27:02,920 problem and with, with like no boots on the ground, right? 465 00:27:02,920 --> 00:27:07,960 Whereas like Ukraine has been actively pursuing this problem 466 00:27:08,080 --> 00:27:12,120 and, and demonstrating that they're able to like, utilize 467 00:27:12,280 --> 00:27:18,360 cheap commercial drones for all sorts of military applications. 468 00:27:19,120 --> 00:27:22,240 Yeah. What's the like, to what extent 469 00:27:22,240 --> 00:27:27,760 is the bottoms up approach that we're seeing in Ukraine a more 470 00:27:27,760 --> 00:27:32,600 effective approach to, to generating outcomes over like 471 00:27:32,600 --> 00:27:35,280 spending billions of dollars on like new drone technology that 472 00:27:35,280 --> 00:27:38,440 may or may not be, I mean, maybe used in conflict, But like by 473 00:27:38,440 --> 00:27:40,440 the time that technology gets used in conflict, there's been 474 00:27:40,440 --> 00:27:42,320 like billions of dollars spent on that technology. 475 00:27:42,560 --> 00:27:44,480 We'll we'll only know if it's effective, right? 476 00:27:44,480 --> 00:27:46,280 Like when it's been used. See what I'm saying? 477 00:27:47,400 --> 00:27:49,920 Yeah, I, I, I would say it's slightly different than that, 478 00:27:49,920 --> 00:27:51,280 right. So a lot of the drones being 479 00:27:51,280 --> 00:27:54,520 used in Ukraine right now are repurposed civilian drones, 480 00:27:54,520 --> 00:27:57,080 right? So DJI drones like the DJI Mavic 481 00:27:57,600 --> 00:28:00,800 and a lot of these are FPV drones that are being, you know, 482 00:28:00,800 --> 00:28:04,200 flown by an operator like the racing drones and FPV view that 483 00:28:04,200 --> 00:28:08,720 they strap the RPG on top of and in contested environments strap 484 00:28:08,720 --> 00:28:12,240 a fiber optic cable on. And so these are like very 485 00:28:12,240 --> 00:28:15,600 unsophisticated, but very effective improvised munitions, 486 00:28:16,080 --> 00:28:22,320 right, which are, I think in many ways a great piece of 487 00:28:22,320 --> 00:28:25,640 evidence on how warfare is evolving and how battlefield 488 00:28:25,640 --> 00:28:27,880 strategy is evolving. And I would argue that U.S. 489 00:28:27,880 --> 00:28:31,200 drone doctrine has historically prioritized very large 490 00:28:31,200 --> 00:28:33,960 platforms, right? The US Predator drones and the 491 00:28:33,960 --> 00:28:38,840 Reaper drones have been highly effective at pinpoint strikes on 492 00:28:38,840 --> 00:28:41,400 targets in the Middle East on the War on Terror. 493 00:28:43,120 --> 00:28:45,160 You know, the, the, the, the General Atomics, Reaper 494 00:28:45,160 --> 00:28:49,000 platforms and Predator platforms. 495 00:28:49,320 --> 00:28:54,440 Now we're seeing both Air Force and Navy prioritize new CCA 496 00:28:54,440 --> 00:28:57,320 initiative, the Collaborative combat aircrafts, which are 497 00:28:57,320 --> 00:29:02,840 effectively unmanned UAVs that pair with existing US fires. 498 00:29:03,240 --> 00:29:08,080 So you'll have an F35 and you'll have 5 CCA drones, which are 499 00:29:08,160 --> 00:29:11,480 very large combat aircrafts flying alongside of it, which is 500 00:29:11,480 --> 00:29:16,080 going to hopefully increase the capacity of, of, of US aerial 501 00:29:16,080 --> 00:29:21,160 programs to defend and to affect missions in highly adversarial 502 00:29:21,160 --> 00:29:24,560 environments, both in terms of combat with other aircrafts and 503 00:29:24,560 --> 00:29:26,240 then combat on, on ground strikes. 504 00:29:27,080 --> 00:29:30,080 But so it is true that a lot of the, the US platforms are, are 505 00:29:30,080 --> 00:29:33,240 much larger than the platforms that are being used in Ukraine 506 00:29:33,240 --> 00:29:35,000 right now, right? The the, the majority of the, 507 00:29:35,040 --> 00:29:40,200 the programs, the the shield, AI expat and V bat, the Andrel 508 00:29:41,840 --> 00:29:47,720 Fury, the Andrel Omen, these are like group three, group 5 UAVs. 509 00:29:47,720 --> 00:29:50,040 You know, the, the US has groupings of what they'd 510 00:29:50,040 --> 00:29:51,840 consider UAVs. Most of the ones used in Ukraine 511 00:29:51,840 --> 00:29:54,320 right now are the Group 1 and the group 2 UAV. 512 00:29:54,320 --> 00:29:56,360 So very small, managed by a single operator. 513 00:29:57,000 --> 00:30:01,360 Now it is true that the US has less of a of a small man, less 514 00:30:01,360 --> 00:30:04,240 of a manufacturing capacity in small drones than China does and 515 00:30:04,240 --> 00:30:07,080 also has less competitive swarm technology than China does. 516 00:30:07,440 --> 00:30:09,640 Which is actually one of the problems that LeBron just is 517 00:30:09,640 --> 00:30:13,040 currently working on, which is how can you have swarms that 518 00:30:13,040 --> 00:30:15,520 operate in highly contested environments when base links are 519 00:30:15,520 --> 00:30:17,400 knocked out? How do they effectively form 520 00:30:17,400 --> 00:30:21,360 consensus and reorganize themselves to be able to affect 521 00:30:21,360 --> 00:30:24,320 emission outcome even if half the drones get shot down, lit on 522 00:30:24,320 --> 00:30:27,320 fire, jammed, etcetera. But I would say that US 523 00:30:27,320 --> 00:30:30,760 battlefield doctrine is changing and there's going to be a 524 00:30:30,880 --> 00:30:32,960 prioritization of both very large platform and small 525 00:30:32,960 --> 00:30:34,680 platforms. Hag Seth has talked publicly a 526 00:30:34,680 --> 00:30:36,720 lot about that. I think there's needs to be 527 00:30:36,720 --> 00:30:38,320 catching up that the US has to do here. 528 00:30:38,920 --> 00:30:41,280 But I I wouldn't say the US isn't learning from from the 529 00:30:41,280 --> 00:30:42,720 Ukraine more. I think it's probably one of 530 00:30:42,720 --> 00:30:45,760 the, the areas that that the US is studying most aggressively. 531 00:30:46,320 --> 00:30:49,600 It's also worth mentioning that that probably the one of the 532 00:30:49,600 --> 00:30:53,160 most effective and under mentioned pieces of, of 533 00:30:53,160 --> 00:30:57,280 battlefield technology in the war in Ukraine is electronic 534 00:30:57,280 --> 00:31:00,800 warfare, jamming things that knock out drones. 535 00:31:01,000 --> 00:31:04,320 And the US has substantial work that it's doing in this end, 536 00:31:04,520 --> 00:31:05,840 right? So how can you effectively knock 537 00:31:05,840 --> 00:31:10,360 down large amounts large swarms of of of unmanned vehicles as 538 00:31:10,360 --> 00:31:13,080 they try to effect some action on the battlefield? 539 00:31:14,520 --> 00:31:18,920 So let's talk about verifiable AI for defense and some of the 540 00:31:19,000 --> 00:31:21,680 work you've been doing with with Andrel. 541 00:31:22,280 --> 00:31:26,520 You, you announced recently that that the Grange was building 542 00:31:26,520 --> 00:31:30,000 within their Lattice SDK, which I'm not super familiar with. 543 00:31:30,000 --> 00:31:34,600 Maybe we've been talking about what what that SDK is and and 544 00:31:35,240 --> 00:31:38,640 also how LaGrange is is integrating with an SDK and and 545 00:31:38,640 --> 00:31:41,960 and and how how relevant is DKML in that context? 546 00:31:42,360 --> 00:31:45,160 Yeah. So to take a step back, one of 547 00:31:45,160 --> 00:31:50,120 Lagrange's core theses is that there is a gap in cryptography 548 00:31:50,440 --> 00:31:52,120 within national security defense. 549 00:31:52,840 --> 00:31:56,400 Effectively the the, the modern paradigms for implementing 550 00:31:56,400 --> 00:31:59,040 cryptographic security and defense systems are based on air 551 00:31:59,040 --> 00:32:02,240 gap, air gap in the system. So just private clouds, private 552 00:32:02,240 --> 00:32:06,480 networks, private links and the assumption that these these 553 00:32:06,480 --> 00:32:08,800 communication links can't be broken since they're encrypted, 554 00:32:08,920 --> 00:32:12,360 which broadly is true. Where we believe this type of 555 00:32:12,360 --> 00:32:16,280 framing falls down is in highly decentralized combat 556 00:32:16,280 --> 00:32:20,440 environments, which is what we see in modern drone warfare. 557 00:32:20,840 --> 00:32:24,400 When you have hundreds, if not thousands of different drones 558 00:32:24,560 --> 00:32:27,800 across potentially different manufacturers that need the 559 00:32:27,800 --> 00:32:34,400 capacity to form consensus or form agreements on the state of 560 00:32:34,400 --> 00:32:36,800 a battlefield. And then be able to effect 561 00:32:36,800 --> 00:32:41,760 action subject to the onboard AI or whatever command and control 562 00:32:41,760 --> 00:32:44,360 system provided them with, with a mission objective. 563 00:32:45,200 --> 00:32:48,800 And so when we look at that problem and it's very analogous 564 00:32:48,800 --> 00:32:52,480 in our mind to the problems we see in crypto, which is you have 565 00:32:54,120 --> 00:32:59,400 10,100 thousand nodes globally that need to reach consensus in 566 00:32:59,400 --> 00:33:03,640 the state of a shared Ledger very, very quickly based on a 567 00:33:04,200 --> 00:33:06,960 set of transactions that could be submitted at any single node 568 00:33:06,960 --> 00:33:10,280 at any point in time. It's very similar in our view to 569 00:33:10,280 --> 00:33:13,240 to the question of how do drones reach coordination in a tested 570 00:33:13,240 --> 00:33:16,240 environments. And so a lot of the technology 571 00:33:16,240 --> 00:33:19,960 that LaGrange is building orients itself around adding 572 00:33:19,960 --> 00:33:24,800 enhanced cryptographic security to the use of autonomous weapons 573 00:33:24,800 --> 00:33:27,760 and drones in highly adversarial environments. 574 00:33:28,440 --> 00:33:32,560 And so that's things as simple as proving the computer vision 575 00:33:32,560 --> 00:33:38,320 models that run on top of drones and edge devices, or proving the 576 00:33:38,320 --> 00:33:41,440 correctness of command and control systems and AI that runs 577 00:33:41,440 --> 00:33:46,360 on command and control systems. So Anderl's Lattice SDK is an 578 00:33:46,360 --> 00:33:52,640 open SDK that is developed by Anderl to coordinate across 579 00:33:52,640 --> 00:33:55,440 assets and drones that are built by Anderl as well as by other 580 00:33:55,440 --> 00:34:01,640 companies, wherein drones can interface, coordinate and allow 581 00:34:01,640 --> 00:34:06,120 an operator to control those drones and complete downstream 582 00:34:06,440 --> 00:34:10,320 mission objectives. This is everything from their 583 00:34:10,320 --> 00:34:14,880 Seabed Sentry and Uvs all the way through Omen and Fury and a 584 00:34:14,880 --> 00:34:16,760 lot of their their larger airborne platforms. 585 00:34:18,000 --> 00:34:22,520 And So what LaGrange built was a proof of concept demonstrating 586 00:34:22,520 --> 00:34:26,120 how command and control system that made relevant battlefield 587 00:34:26,120 --> 00:34:29,639 decisions with AI could be made accountable and provably 588 00:34:29,639 --> 00:34:31,600 correct. We've open sources, we've 589 00:34:31,600 --> 00:34:34,560 released it publicly, and we think this is hopefully the 590 00:34:34,560 --> 00:34:38,679 start of enhancing how cryptographic security is going 591 00:34:38,679 --> 00:34:44,760 to be used in defense relevant programs that involve autonomous 592 00:34:44,760 --> 00:34:48,040 systems. You you guys have written about 593 00:34:50,199 --> 00:34:54,600 accountable autonomy and and why that's important. 594 00:34:54,600 --> 00:34:57,480 So what is that? What does that mean and and and 595 00:34:57,480 --> 00:35:03,960 why is it relevant to to the context of AI and and defense? 596 00:35:04,600 --> 00:35:06,200 Yeah. There's a few places where 597 00:35:06,200 --> 00:35:08,360 accountable autonomy really matters, right? 598 00:35:08,640 --> 00:35:12,040 If your baseline assumption is that in a contested environment 599 00:35:12,040 --> 00:35:15,680 you only have crash fault tolerant requirements, so 600 00:35:15,680 --> 00:35:19,280 there's no Byzantine actors, then you you assume that the 601 00:35:19,280 --> 00:35:21,960 drones on board AI can't be compromised. 602 00:35:22,320 --> 00:35:26,960 So if the groans onboard AI can be compromised and in theory 603 00:35:26,960 --> 00:35:30,840 someone can jam the thing, they can, they can spoof 604 00:35:30,840 --> 00:35:33,360 communication with it, take it over, update the software. 605 00:35:33,560 --> 00:35:36,800 Then it's very, very important that you can prove the 606 00:35:36,800 --> 00:35:39,560 correctness of the model running there such that the output that 607 00:35:39,560 --> 00:35:42,880 you're getting back, if you know another UAV is communicating 608 00:35:42,880 --> 00:35:45,960 and, and, and, and, and interacting with, with the, the 609 00:35:45,960 --> 00:35:49,040 first asset is provably correct, right? 610 00:35:49,240 --> 00:35:51,040 It's the same thing as crypto, right? 611 00:35:51,040 --> 00:35:54,720 You have, how do you verify that an AI model was run correctly on 612 00:35:54,720 --> 00:35:56,720 the blockchain? We can't have everybody rerun 613 00:35:56,720 --> 00:35:58,280 it. So you have to have the person 614 00:35:58,280 --> 00:36:00,520 who ran it prove it. That's very true. 615 00:36:00,520 --> 00:36:01,840 If you're assuming Byzantine assumptions. 616 00:36:01,840 --> 00:36:05,440 Now it's debatable if you should assume Byzantine assumptions in 617 00:36:05,480 --> 00:36:09,080 a war zone or not. We're not going to get into 618 00:36:09,080 --> 00:36:10,480 that. There's some people who think 619 00:36:10,480 --> 00:36:11,560 you should, some people think you shouldn't. 620 00:36:13,760 --> 00:36:15,560 Someone once said to me, we can trust the hardware. 621 00:36:15,560 --> 00:36:18,840 The hardware blows up, so there you go. 622 00:36:19,560 --> 00:36:24,600 But now, if you have a command and control system and you want 623 00:36:24,600 --> 00:36:28,520 to ensure that the outputs that are being consumed by the end 624 00:36:28,520 --> 00:36:31,600 device or the edge device are correct, that's another place 625 00:36:31,600 --> 00:36:33,800 where you need 0 knowledge machine learning. 626 00:36:34,240 --> 00:36:38,120 Effectively, how can you ensure that one system that's 627 00:36:38,120 --> 00:36:41,160 controlling a bunch of downstream assets or one system 628 00:36:41,160 --> 00:36:43,760 that an operator is using and one AI model of an opera is 629 00:36:43,760 --> 00:36:46,480 using to control a bunch of downstream assets is in fact 630 00:36:46,520 --> 00:36:49,800 only making for examle kill chain relevant decisions as a 631 00:36:49,800 --> 00:36:53,920 result of the correct AI models being used in the command and 632 00:36:53,920 --> 00:36:56,760 control system? That's one of the use cases that 633 00:36:56,880 --> 00:36:58,440 we think is the most relevant, right? 634 00:36:58,640 --> 00:37:01,880 So assume you have an LLM that's job is to coordinate submission 635 00:37:01,880 --> 00:37:04,240 objective. It's supposed to take in a bunch 636 00:37:04,240 --> 00:37:08,160 of sensor input from a bunch of UU VS and then coordinate some 637 00:37:08,160 --> 00:37:12,200 substantive UU VS to go in, you know, effect some some 638 00:37:12,200 --> 00:37:16,680 battlefield action. In this case, you would require 639 00:37:16,880 --> 00:37:20,560 that the Uuvs that are affecting that battlefield action would 640 00:37:20,560 --> 00:37:22,800 only affect that battlefield action if there is 641 00:37:22,800 --> 00:37:26,160 accountability that the correct model using the correct inputs 642 00:37:26,320 --> 00:37:29,480 has in fact recommended action. Otherwise, the kill chain 643 00:37:29,480 --> 00:37:32,400 decision shouldn't be made unless the correct chain of 644 00:37:32,400 --> 00:37:36,240 command, the correct chain of of both data custody and inference 645 00:37:36,240 --> 00:37:39,200 custody has been followed. This is when we talk about 646 00:37:39,200 --> 00:37:42,440 accountable autonomy, that autonomous systems are not just 647 00:37:42,440 --> 00:37:45,880 operating as these black boxes making, you know, random 648 00:37:45,880 --> 00:37:49,720 battlefield decisions, but that they're made entirely with an 649 00:37:49,720 --> 00:37:53,240 operator's decision or entirely by the decision of a model 650 00:37:53,400 --> 00:37:56,640 wherein there's hard constraints put in place by an operator. 651 00:37:57,920 --> 00:37:59,400 Oh, sorry, what's a kill chain decision? 652 00:38:00,200 --> 00:38:03,280 A decision to so to affect, let's say, a kill, a kill 653 00:38:03,280 --> 00:38:06,560 action, to to, to fire a missile, to to to to, to use 654 00:38:06,560 --> 00:38:08,960 some armament for a battlefield purpose. 655 00:38:09,600 --> 00:38:13,240 OK, and in and in this context? I guess the assumptions that the 656 00:38:13,440 --> 00:38:17,480 AI is providing that kill chain decision and then yeah, an 657 00:38:17,480 --> 00:38:20,320 operator is deciding to execute that decision or not. 658 00:38:20,600 --> 00:38:21,880 That's often how it works, right? 659 00:38:21,880 --> 00:38:24,720 So it's a structure of multi stage process describing how 660 00:38:25,040 --> 00:38:30,080 adversaries can conduct attacks. So you know under what 661 00:38:30,080 --> 00:38:33,560 circumstance, under what assumptions, under what sensor 662 00:38:33,560 --> 00:38:40,960 feed inputs should a autonomous system engage in an attack, 663 00:38:41,640 --> 00:38:43,600 right? Should an operative in a loop 664 00:38:43,760 --> 00:38:46,960 should just entirely be AI? If it's entirely AI, this is a 665 00:38:46,960 --> 00:38:50,680 place where we highly believe that verifiability is paramount. 666 00:38:51,640 --> 00:38:55,640 Right. I I guess like so I'm not super 667 00:38:55,640 --> 00:38:59,240 filming with this, but I I suspect there are some sort of 668 00:38:59,240 --> 00:39:04,320 Geneva Convention laws that relate to the use of AI in 669 00:39:04,320 --> 00:39:08,400 military applications. I'll have to check. 670 00:39:09,480 --> 00:39:11,480 I don't believe. Yeah. 671 00:39:12,160 --> 00:39:16,640 I don't like let's let's let's let's assume that you know, 672 00:39:16,640 --> 00:39:20,440 there are or there may be like in the future, some Geneva 673 00:39:20,440 --> 00:39:24,680 Convention rules around like the use of AI in war. 674 00:39:25,480 --> 00:39:34,120 To what extent is provability and, and to what I sense is 675 00:39:34,120 --> 00:39:38,280 provability and accountable autonomy important then in 676 00:39:38,600 --> 00:39:42,920 ensuring that decisions made by AI are in fact lawful? 677 00:39:43,480 --> 00:39:48,880 And I guess this this could also extend maybe to, to domestic use 678 00:39:48,880 --> 00:39:52,480 of AI in, in policing, for example, it's not only in the 679 00:39:52,480 --> 00:39:55,960 military context, but also inside like national like sort 680 00:39:55,960 --> 00:39:59,200 of policing. Yeah, this is a great question. 681 00:39:59,440 --> 00:40:05,400 So I would say that the the the use of autonomous weapons. 682 00:40:05,760 --> 00:40:10,080 Is very, very contentious right now, especially autonomous 683 00:40:10,080 --> 00:40:11,600 weapons that don't have an operator to loop. 684 00:40:12,200 --> 00:40:15,040 So effectively a say a drone that can make a kill train 685 00:40:15,040 --> 00:40:18,880 decision without an operator saying, hey, yes, I approve this 686 00:40:18,920 --> 00:40:22,400 to occur. The USI believe doesn't have 687 00:40:22,400 --> 00:40:24,920 autonomous weapons that can make kill train decisions about 688 00:40:24,920 --> 00:40:28,000 operators being in the loop that that as part of U.S. military 689 00:40:28,000 --> 00:40:30,000 doctrine, there always must be an operator in the loop. 690 00:40:30,760 --> 00:40:32,720 That's my current understanding. I don't know if that's changed, 691 00:40:34,160 --> 00:40:37,640 but you know, to be honest, and Andrew has talked about this 692 00:40:37,640 --> 00:40:40,720 shield has talked about this, the use of autonomous weapons is 693 00:40:40,720 --> 00:40:42,920 not a new thing. It just feels new, right? 694 00:40:43,280 --> 00:40:46,840 You know, when, when Caesar was, was fighting versus Gedricks, he 695 00:40:47,160 --> 00:40:52,640 dug a spike pit surrounding the the, the Gaelic army, right? 696 00:40:52,640 --> 00:40:54,240 That that is nothing more than autonomous weapon. 697 00:40:54,240 --> 00:40:56,160 You walk and you fall into the spikes, right? 698 00:40:56,160 --> 00:40:57,880 They don't have to be there right there. 699 00:40:57,880 --> 00:41:00,040 You can go anywhere you want. The middle of the night, there's 700 00:41:00,200 --> 00:41:02,040 a spike that you can step on, right? 701 00:41:02,040 --> 00:41:06,080 World War One, World War 2, the use of mines, all of these are 702 00:41:06,080 --> 00:41:08,360 autonomous weapons. You get too close to a mine, you 703 00:41:08,360 --> 00:41:12,880 could be a school bus, you could be a family of, of, of refugees 704 00:41:13,200 --> 00:41:16,000 and horrific, there's horrific implications. 705 00:41:16,400 --> 00:41:20,240 So the, the idea that you know, an armament now is has a 706 00:41:20,240 --> 00:41:24,520 computer vision model that says, hey, if this is a Russian tank, 707 00:41:24,760 --> 00:41:29,400 yes, explode. If this is a school bus, don't 708 00:41:29,880 --> 00:41:33,080 that that's a net positive that's substantially better the 709 00:41:33,080 --> 00:41:36,600 state-of-the-art of a of a mind that doesn't have any type of AI 710 00:41:36,600 --> 00:41:39,280 usage right. But just the the question over, 711 00:41:39,280 --> 00:41:44,040 hey, we're going to we're going to to outright ban the use of AI 712 00:41:44,040 --> 00:41:48,800 and weapons is is is actually much worse than the alternative. 713 00:41:49,640 --> 00:41:54,240 I'd much rather that a missile has on board guidance to to 714 00:41:54,320 --> 00:41:57,800 reroute and avoid hitting the wrong target, right. 715 00:41:58,080 --> 00:42:02,160 If it's if it's firing at a target, then the, the, the comms 716 00:42:02,160 --> 00:42:04,840 link is jammed. I'd much rather correctly rearm 717 00:42:04,840 --> 00:42:07,160 and hit the correct target instead of, you know, hitting a 718 00:42:07,160 --> 00:42:09,080 school. All of this stuff I think is 719 00:42:09,080 --> 00:42:11,440 super important. The more precise weapons can be 720 00:42:11,440 --> 00:42:14,840 and AI is a is incredible way to increase precision of weapons. 721 00:42:15,720 --> 00:42:17,880 The the, the, the less collateral damage and 722 00:42:17,880 --> 00:42:19,520 casualties. You'll, you'll, you'll see. 723 00:42:20,680 --> 00:42:24,280 And so that that's something I do firmly believe, but I, I, to 724 00:42:24,280 --> 00:42:27,880 your point, I do believe that there has to be conventions and 725 00:42:27,880 --> 00:42:32,040 organization around what or how these decisions can be made such 726 00:42:32,040 --> 00:42:34,520 that they're lawful. How is auditability conducted? 727 00:42:34,800 --> 00:42:37,400 How can we quickly audit to make sure that all of the use of 728 00:42:37,400 --> 00:42:41,080 autonomous weapons are following a predefined several rules, 729 00:42:41,720 --> 00:42:43,680 right? Is the AI operating correctly 730 00:42:43,920 --> 00:42:48,480 right? If if a weapon missed, is it the 731 00:42:48,480 --> 00:42:50,120 fault of an operator? Is it the fault of the onboard 732 00:42:50,200 --> 00:42:53,840 AI? Is it the fault of an adversary? 733 00:42:53,840 --> 00:42:56,200 Can taking over the system? Is the fault of the system being 734 00:42:56,200 --> 00:42:58,040 jammed or having a kinetic interference? 735 00:42:58,880 --> 00:43:00,360 Now all of this stuff is super important. 736 00:43:00,640 --> 00:43:04,680 And when you add AI, you open up a bunch of questions over what 737 00:43:04,680 --> 00:43:07,960 verifiability and auditability of AI means. 738 00:43:08,520 --> 00:43:10,960 And that's where zero knowledge machine learning is super, super 739 00:43:10,960 --> 00:43:13,680 important. And that's been one of our very 740 00:43:13,760 --> 00:43:16,560 large pitches that we've made to to almost everyone we can talk 741 00:43:16,560 --> 00:43:20,440 to in defense that zero knowledge proofs provide this 742 00:43:20,440 --> 00:43:24,520 this way to glimpse inside the black box to determine that the 743 00:43:24,520 --> 00:43:30,440 output that you got back that indicated that an action should 744 00:43:30,440 --> 00:43:33,520 be taken has come from the correct model with these set of 745 00:43:33,520 --> 00:43:37,120 correct inputs, which allows you to determine why the decision 746 00:43:37,120 --> 00:43:39,840 was made, under what circumstances decision was made 747 00:43:40,080 --> 00:43:43,680 and under what circumstances decision should be adjusted, the 748 00:43:43,680 --> 00:43:46,080 model should be adjusted or things should be changed going 749 00:43:46,080 --> 00:43:48,120 forward. Yeah, yeah. 750 00:43:49,080 --> 00:43:52,600 I was thinking about this like, as with a lot of things with AI, 751 00:43:52,600 --> 00:43:56,480 I think they, it quickly becomes a race to the bottom where, you 752 00:43:56,480 --> 00:44:01,080 know, if we have increasingly more, more and more AI in 753 00:44:01,160 --> 00:44:05,160 military decision making, then like at the bottom, it just 754 00:44:05,160 --> 00:44:08,120 becomes about taking out the other person's AI, right? 755 00:44:09,040 --> 00:44:12,080 If, if we're narrowing every military decision to like, is 756 00:44:12,080 --> 00:44:15,840 this, is this a military target? And at the same time you have 757 00:44:15,840 --> 00:44:18,920 those like increasing use of AI, then at the end, you're just 758 00:44:18,920 --> 00:44:21,040 trying to take out the other person's AI or the other 759 00:44:21,040 --> 00:44:23,440 country's AI And, and you're no longer trying to take out any 760 00:44:23,440 --> 00:44:27,320 one of their like military or, you know, infrastructure. 761 00:44:27,320 --> 00:44:29,080 It's all about just taking out the AI. 762 00:44:29,720 --> 00:44:32,160 But I, I would say taking out AI has a lot to do with 763 00:44:32,160 --> 00:44:34,520 infrastructure, right? I mean, I think that's the whole 764 00:44:34,520 --> 00:44:36,600 room industrialization. It's the energy grid. 765 00:44:36,600 --> 00:44:39,400 It's the the where you train the AI, it's where you run the 766 00:44:39,400 --> 00:44:41,760 inference and data centers. It's it's it's where you I met 767 00:44:41,760 --> 00:44:45,040 I. Met Milac area infrastructure, 768 00:44:45,040 --> 00:44:47,600 but Oh yeah, I got you of course, like yeah, they're you 769 00:44:47,600 --> 00:44:50,200 know, electricity infrastructure, IT 770 00:44:50,240 --> 00:44:52,200 infrastructure, but like yeah, at the end of the day, it comes 771 00:44:52,200 --> 00:44:56,400 down to that where then I guess like, you know, having having 772 00:44:56,400 --> 00:44:59,680 decentralized AI that's able to function autonomously and in a 773 00:44:59,680 --> 00:45:03,200 more of a mesh system becomes more interesting. 774 00:45:03,200 --> 00:45:06,560 It does. I want to talk about the token 775 00:45:06,560 --> 00:45:09,960 here because, you know I know you guys still have some 776 00:45:10,200 --> 00:45:13,440 applications and kind of products that are geared to more 777 00:45:13,520 --> 00:45:16,320 towards the more crypto application crypto use cases. 778 00:45:16,720 --> 00:45:20,920 But you know, when you're talking to Andrew or or Lockheed 779 00:45:21,240 --> 00:45:25,600 about implementing LaGrange technology and the token has a 780 00:45:25,600 --> 00:45:31,160 lot to do and a lot to sort of do in the, in the decentralized 781 00:45:31,160 --> 00:45:37,680 nature of the of the of the ZK verification aspect of the 782 00:45:37,680 --> 00:45:39,480 product. How's that conversation going? 783 00:45:39,560 --> 00:45:42,200 And I mean does the token have anything to do with the military 784 00:45:42,200 --> 00:45:47,360 applications or is it strictly for for the crypto related use 785 00:45:47,360 --> 00:45:48,360 cases? Yeah. 786 00:45:48,360 --> 00:45:51,080 So the token is staked into Lagrange's Prover network with 787 00:45:51,120 --> 00:45:53,640 which is a decentralized network of operators that can generate 788 00:45:53,640 --> 00:45:57,320 proofs within an auction mechanism that that has been 789 00:45:57,320 --> 00:45:59,720 designed and developed by our team where work is assigned into 790 00:45:59,720 --> 00:46:01,360 the network. The network generates proofs and 791 00:46:01,360 --> 00:46:04,200 sends those proofs back to whoever requested them. 792 00:46:04,600 --> 00:46:07,240 The decision of whether or not you want to deploy Deep Proof or 793 00:46:07,240 --> 00:46:09,880 any of Lagrange's technology and decentralized or centralized 794 00:46:09,880 --> 00:46:13,720 capacity just depends on whether or not the person in question 795 00:46:14,120 --> 00:46:16,560 would prefer one of the other. I would say for all of our 796 00:46:16,560 --> 00:46:19,760 crypto use cases, which is a lot, we work with probably every 797 00:46:19,760 --> 00:46:21,440 AI X crypto company in the space right now. 798 00:46:21,880 --> 00:46:25,760 We've announced stuff with Sentient, Gaia, Mira, 0G, Open 799 00:46:25,760 --> 00:46:29,800 Ledger, really all of these crypto X AI companies. 800 00:46:30,120 --> 00:46:32,600 It all is going to be running through and runs through our 801 00:46:32,600 --> 00:46:35,400 proven network and where our proven network powers all of the 802 00:46:35,400 --> 00:46:37,040 ZK machine learning that we do crypto. 803 00:46:37,480 --> 00:46:40,520 Now the defense applications, it'll just depend on the on the 804 00:46:40,520 --> 00:46:42,760 decisions that our defense counterparties want to make. 805 00:46:43,840 --> 00:46:45,120 I can't promise one way or the other. 806 00:46:45,240 --> 00:46:47,200 We have decentralized deployments and our software 807 00:46:47,200 --> 00:46:48,920 also works on centralized places as well. 808 00:46:49,240 --> 00:46:52,640 LaGrange Foundation is the foundation of issues of token 809 00:46:52,960 --> 00:46:56,120 and LaGrange Labs is a development entity for the the 810 00:46:56,120 --> 00:46:58,640 LaGrange Prover network as well as deep proven a bunch of other 811 00:46:58,640 --> 00:47:02,560 technology and cryptography and we are very excited to see 812 00:47:02,560 --> 00:47:05,560 growth of our technology across dual use purposes, right. 813 00:47:05,880 --> 00:47:09,520 Think of Palantir, there's very different deployments of 814 00:47:09,520 --> 00:47:13,040 Palantir for public sector and for private sector. 815 00:47:14,240 --> 00:47:17,960 You know, Gotham for example, is a platform that Palantir 816 00:47:17,960 --> 00:47:19,640 develops entirely for police forces. 817 00:47:20,320 --> 00:47:23,760 Something like Foundry and AIP are really strongly commercial, 818 00:47:23,760 --> 00:47:25,680 but also have government application as well. 819 00:47:26,240 --> 00:47:29,800 And I think that LaGrange is committed to being a dual use 820 00:47:29,800 --> 00:47:32,840 company, which means that we will continue to expand our 821 00:47:32,840 --> 00:47:35,800 adoption in crypto, we'll continue to expand our adoption 822 00:47:35,800 --> 00:47:38,440 in enterprise and we will continue to expand our adoption 823 00:47:38,440 --> 00:47:41,240 in government defense. What are your thoughts on Europe 824 00:47:41,240 --> 00:47:44,120 in the sort of military geopolitical landscape, 825 00:47:44,120 --> 00:47:48,240 specifically when it comes to AI driven military technology? 826 00:47:48,480 --> 00:47:49,440 Yeah. I mean, look, I think there's 827 00:47:49,440 --> 00:47:52,120 some, there's some great military and defense contractors 828 00:47:52,160 --> 00:47:58,880 out of out of the EU, out of Europe, Rafael Talis in France, 829 00:47:59,560 --> 00:48:03,520 BA Systems in the UKI would say that these are more legacy style 830 00:48:03,520 --> 00:48:07,560 primes and more so than they are, you know, up and coming 831 00:48:08,280 --> 00:48:12,560 highly competitive manufacturers of, of, of, of new drone 832 00:48:12,560 --> 00:48:15,600 platforms, new AI platforms. But you know, we know the teams, 833 00:48:15,600 --> 00:48:18,560 we've met the teams at, at, at these companies and we have, you 834 00:48:18,560 --> 00:48:19,840 know, think they're fantastic, frankly. 835 00:48:21,240 --> 00:48:26,320 But I, I would say that, you know, where I think a lot of the 836 00:48:26,320 --> 00:48:28,680 innovation is happening in Europe in, in, in modern 837 00:48:28,680 --> 00:48:31,400 military doctrine or on drones, obviously in, in Ukraine's, we 838 00:48:31,400 --> 00:48:36,400 touched on before, right, Ukraine uses AI and drones. 839 00:48:37,320 --> 00:48:41,600 One of the very useful things you can use for AI and drones 840 00:48:41,600 --> 00:48:44,720 is, is allowing those drones to operate in jammed environments, 841 00:48:45,160 --> 00:48:47,440 right? So, you know, assume you have a 842 00:48:47,440 --> 00:48:49,760 bunch of electronic warfare jammers that are trying to knock 843 00:48:49,760 --> 00:48:52,560 out and operate like, right. The, the, the one way you can do 844 00:48:52,560 --> 00:48:55,120 that in a very naive way is you just attach fiber optic cable to 845 00:48:55,120 --> 00:48:56,400 this. This is what a lot of the FTV 846 00:48:56,400 --> 00:48:59,160 drones are being done. And you know, now, now Russia 847 00:48:59,160 --> 00:49:01,280 has launched these like big mothership drones that have like 848 00:49:01,280 --> 00:49:05,440 6 kilometers of, of, of fiber optic cable attached on smaller 849 00:49:05,440 --> 00:49:07,200 drones. You know, I personally don't 850 00:49:07,200 --> 00:49:09,640 think that that's a very future proof strategy. 851 00:49:09,840 --> 00:49:13,120 There's, you know, kinetic issues when you have a bunch of 852 00:49:13,600 --> 00:49:16,920 fiber optics attached. But AI provides another 853 00:49:16,920 --> 00:49:19,320 alternative, which is you can still jam the operator link. 854 00:49:19,840 --> 00:49:22,160 You can have a bunch of electronic warfare jamming it. 855 00:49:22,440 --> 00:49:26,160 But if the drone can complete its objective in the last mile 856 00:49:26,160 --> 00:49:30,680 without needing an operator link because it has onboard AI, then 857 00:49:30,680 --> 00:49:35,000 it can then it can do this. This is if you remember the the 858 00:49:35,000 --> 00:49:38,680 Ukraine operation when they attacked the Russian jets, it 859 00:49:38,680 --> 00:49:42,960 took out the Russian jets. All of those drones, I believe 860 00:49:42,960 --> 00:49:46,800 were using on board AI since the Russian air base was obviously 861 00:49:46,800 --> 00:49:48,880 jamming. And so these models were trained 862 00:49:48,880 --> 00:49:50,840 to recognize weak points in the Russian jets. 863 00:49:50,840 --> 00:49:53,680 They got close enough and then the jet was able to kind of 864 00:49:53,880 --> 00:49:59,800 loiter as required, then be able to, to affect that nation on top 865 00:49:59,800 --> 00:50:01,440 of the Russian jet at the weak spots. 866 00:50:01,760 --> 00:50:03,840 And that wouldn't have been possible without onboard AI. 867 00:50:04,720 --> 00:50:08,240 So I, I think that there is substantial military innovation 868 00:50:08,240 --> 00:50:10,720 in, in Europe right now. And I think a lot of it's driven 869 00:50:10,720 --> 00:50:13,360 by the war in Ukraine. But I'd say like the the world 870 00:50:13,360 --> 00:50:17,760 leader right now, I'd argue in drone innovation is the US and 871 00:50:17,760 --> 00:50:22,320 China. With the advent of of AI in 872 00:50:22,720 --> 00:50:26,120 military and military technologies, it, it feels like 873 00:50:26,240 --> 00:50:29,720 a huge turning point to most of the military technologies that 874 00:50:29,720 --> 00:50:31,680 we've had up until now that have been mostly kinetic. 875 00:50:31,800 --> 00:50:34,120 To what extent do you think there's this sort of like 876 00:50:34,120 --> 00:50:38,640 getaway phenomenon or a runaway phenomenon where when one sort 877 00:50:38,640 --> 00:50:43,080 of country or, or military alliance really takes off, the 878 00:50:43,080 --> 00:50:46,960 other ones can no longer catch up just because their their 879 00:50:46,960 --> 00:50:50,440 superiority that that AI superiority just makes them so 880 00:50:50,440 --> 00:50:53,040 much more powerful. Is is there a new sort of 881 00:50:53,040 --> 00:50:56,720 dynamic at play here that really changes the trajectory of 882 00:50:56,720 --> 00:51:00,640 military advancement in a way that we haven't seen previously? 883 00:51:01,200 --> 00:51:02,080 I mean, I think. Military. 884 00:51:02,080 --> 00:51:05,000 Developments I, I think warfare continually evolves, right? 885 00:51:06,240 --> 00:51:09,200 And I, I think the, the, the strategies that are employed by 886 00:51:09,200 --> 00:51:14,040 nations to counter the strengths of other nations militaries just 887 00:51:14,040 --> 00:51:16,960 continually change, right? I mean, the entire US power 888 00:51:16,960 --> 00:51:21,400 projection platform often predicates on the US Air Force 889 00:51:21,920 --> 00:51:24,920 or US aircraft carriers, wherein, you know, US carrier 890 00:51:24,920 --> 00:51:29,680 group can effectively go very deep blue waters and, and 891 00:51:29,920 --> 00:51:33,080 project the US authority and US power anywhere in the world. 892 00:51:33,240 --> 00:51:35,480 And So what has China done to try to combat that? 893 00:51:35,720 --> 00:51:38,920 Well, they've developed, you know, cutting edge hypersonic 894 00:51:38,920 --> 00:51:41,760 missile technology and in the conflict with China. 895 00:51:41,760 --> 00:51:43,440 And then Pete Hags has talked about this. 896 00:51:43,440 --> 00:51:47,600 So this is not me making it up. It would be very, very hard for 897 00:51:47,600 --> 00:51:52,800 for for US carrier groups to be able to combat just a large 898 00:51:52,800 --> 00:51:57,920 barrage of hypersonic missiles that were fired by China at US 899 00:51:57,920 --> 00:52:00,840 aircraft carriers. And so if the 16 US aircraft 900 00:52:00,840 --> 00:52:03,440 carriers could be knocked out in the 1st 20 minutes of a conflict 901 00:52:03,440 --> 00:52:06,040 with China, it would substantially harm EU s s 902 00:52:06,040 --> 00:52:08,400 position, right? And so there's obviously a lot 903 00:52:08,400 --> 00:52:10,720 of work been done in the US to improve the US domestic 904 00:52:11,280 --> 00:52:14,320 hypersonic missile manufacturing capacity as well, right? 905 00:52:14,600 --> 00:52:17,840 You look at, for example, the role of tanks in the war in 906 00:52:17,840 --> 00:52:20,480 Ukraine, I would argue is analogous to this, right? 907 00:52:22,440 --> 00:52:27,840 A lot of these have been effectively left to become 908 00:52:27,840 --> 00:52:32,040 artillery more so than they are now kind of these forward moving 909 00:52:32,440 --> 00:52:38,800 heavily armored platforms since FPV drones that cost $1000 in 910 00:52:38,800 --> 00:52:42,160 RPG attached can disable and take out these tanks at will. 911 00:52:42,400 --> 00:52:45,080 So a lot of these tanks have been now both on the Ukrainian 912 00:52:45,080 --> 00:52:48,640 side and the Russian side set back under effectively like 913 00:52:48,640 --> 00:52:51,600 mosquito netting to stop drones from getting through where they 914 00:52:51,600 --> 00:52:55,280 can lob long range effectively artillery shells. 915 00:52:55,640 --> 00:52:58,800 But they aren't kind of these forward advancing units. 916 00:52:59,640 --> 00:53:01,120 And it's an entirely different paradigm. 917 00:53:01,120 --> 00:53:02,760 And I, I think this stuff continually changes. 918 00:53:02,760 --> 00:53:07,440 So I, I, you look at like what I would argue makes the US so 919 00:53:07,440 --> 00:53:10,200 special and what result in the US being the strongest country 920 00:53:10,200 --> 00:53:13,840 in the world and for, for hundreds of, you know, a couple 921 00:53:13,840 --> 00:53:17,920 100 years is effectively the, the, the market. 922 00:53:18,440 --> 00:53:24,800 It's the fact the US can commercially compete by having a 923 00:53:24,800 --> 00:53:27,640 tight alignment between both public sector and private 924 00:53:27,640 --> 00:53:29,640 sector. This is the same story of the US 925 00:53:29,640 --> 00:53:32,320 military manufacturing base for US manufacturing base for World 926 00:53:32,320 --> 00:53:34,840 War 2. It's the same story of, of DARPA 927 00:53:34,840 --> 00:53:39,320 funding research across, across, you know, Massachusetts and San 928 00:53:39,840 --> 00:53:42,000 Francisco areas during, during the Cold War. 929 00:53:42,000 --> 00:53:45,160 And it's what we're seeing now with, with the manufacturing of, 930 00:53:45,280 --> 00:53:51,120 of, of weapons and autonomous weapons in these kind of 931 00:53:51,120 --> 00:53:53,520 reindustrialization in American dynamism companies. 932 00:53:54,320 --> 00:53:56,960 When you align these things and you have one of the largest, 933 00:53:56,960 --> 00:53:58,960 most productive and most competitive free market in the 934 00:53:58,960 --> 00:54:02,840 world that is putting resources behind it, you, you're able to 935 00:54:02,840 --> 00:54:05,040 produce stuff that other countries cannot, right? 936 00:54:05,040 --> 00:54:07,560 This is, This is why the, the, you know, I tend to be very 937 00:54:07,560 --> 00:54:10,600 short on the, the whole state of enterprise durability long term. 938 00:54:10,920 --> 00:54:13,560 I don't think you can centrally plan an economy long term. 939 00:54:13,880 --> 00:54:15,480 I agree and we'll have to leave it there. 940 00:54:15,480 --> 00:54:19,120 Thank you so much, Ismail, for coming back on and sharing your 941 00:54:19,120 --> 00:54:23,600 thoughts on on this like really interesting topic, which I think 942 00:54:23,600 --> 00:54:25,680 is only going to get more interesting as time goes on. 943 00:54:26,480 --> 00:54:28,960 Unfortunately, I mean, fortunately or unfortunately, 944 00:54:29,680 --> 00:54:33,760 you know, war, war continues to be something that happens across 945 00:54:33,760 --> 00:54:36,760 the world. And I don't think that's going 946 00:54:36,760 --> 00:54:40,880 to go away anytime soon. And if if the Western world can 947 00:54:40,880 --> 00:54:45,720 continue to project its values through through these 948 00:54:45,720 --> 00:54:50,080 technologies, then I guess that's a better state of affairs 949 00:54:50,080 --> 00:54:54,200 than than the alternative. Yeah, I, I don't think anyone in 950 00:54:54,200 --> 00:54:56,360 crypto wants to live in a world where the West doesn't win. 951 00:54:57,040 --> 00:55:01,000 It just is not amenable to the quality of life to standard of 952 00:55:01,000 --> 00:55:04,320 living and, and the the, the freedoms and civil liberties 953 00:55:04,320 --> 00:55:05,720 that that we've grown accustomed to. 954 00:55:06,240 --> 00:55:08,960 And I think there's a moment where there's an opportunity for 955 00:55:08,960 --> 00:55:12,160 companies in crypto like LaGrange and hopefully others to 956 00:55:12,360 --> 00:55:16,880 assist with ensuring that all of us and as well as all of our 957 00:55:16,880 --> 00:55:19,360 descendants can benefit from the same liberties that we have. 958 00:55:21,240 --> 00:55:21,520 Thank you.