1 00:00:13,800 --> 00:00:15,800 This episode is brought to you by Gnosis. 2 00:00:16,320 --> 00:00:19,480 Gnosis builds decentralized infrastructure for the Ethereum 3 00:00:19,560 --> 00:00:24,280 ecosystem with a rich history dating back to 2015 and products 4 00:00:24,280 --> 00:00:27,040 like Safe Cow Swap or Gnosis Chain. 5 00:00:27,640 --> 00:00:31,880 Gnosis combines needs driven development with deep technical 6 00:00:31,880 --> 00:00:35,160 expertise. This year marks the launch of 7 00:00:35,160 --> 00:00:38,480 Gnosis Pay, the world's first decentralized payment network. 8 00:00:39,200 --> 00:00:43,440 With a Gnosis card, you can spend self custody crypto at any 9 00:00:43,440 --> 00:00:45,680 Visa accepting merchant around the world. 10 00:00:46,320 --> 00:00:49,760 If you're an individual looking to live more on chain or 11 00:00:49,760 --> 00:00:53,880 business looking to white label the stack, visit gnosispay.com. 12 00:00:54,840 --> 00:00:57,720 There are lots of ways you can join the Gnosis journey. 13 00:00:58,160 --> 00:01:01,840 Drop in the Gnosis Dow Governance form, become a Gnosis 14 00:01:01,840 --> 00:01:06,040 validator with a single GNO token and low cost hardware, or 15 00:01:06,040 --> 00:01:09,480 deploy your product on the EVM compatible and highly 16 00:01:09,480 --> 00:01:14,240 decentralized Gnosis chain. Get started today at gnosis dot 17 00:01:14,240 --> 00:01:18,000 IO. Cos One is one of the biggest 18 00:01:18,000 --> 00:01:22,040 node operators globally and help you stake your tokens on 45 plus 19 00:01:22,040 --> 00:01:25,080 networks like Ethereum, Cosmos, The last year. 20 00:01:25,080 --> 00:01:29,720 In dydx, more than 100,000 delegators stake with Chorus 21 00:01:29,720 --> 00:01:32,840 One, including institutions like Bit Go and Ledger. 22 00:01:33,320 --> 00:01:37,360 Sticking with Chorus 1 not only gets you the highest years but 23 00:01:37,360 --> 00:01:41,280 also the most robust security practices and infrastructure 24 00:01:41,520 --> 00:01:43,800 that are usually exclusive for institutions. 25 00:01:44,560 --> 00:01:47,560 You can stake directly to Chorus One's public note from your 26 00:01:47,560 --> 00:01:50,720 wallet, set up a white table note, or use the recently 27 00:01:50,720 --> 00:01:55,480 launched product Opus to stake up to 8000 ETH in a single 28 00:01:55,480 --> 00:01:58,560 transaction. You can even offer high yield 29 00:01:58,560 --> 00:02:01,960 staking to your own customers using their API. 30 00:02:02,240 --> 00:02:05,080 Your assets always remain in your custody, so you can have 31 00:02:05,080 --> 00:02:09,280 complete Peace of Mind. Start staking today at Corus .1. 32 00:02:10,720 --> 00:02:12,880 Welcome to Epicentre, the show which talks about the 33 00:02:12,880 --> 00:02:15,320 technologies, projects, and people driving decentralization 34 00:02:15,320 --> 00:02:18,360 and the blockchain revolution. I'm Felix and I'm here with me 35 00:02:18,360 --> 00:02:21,280 here today We're speaking with Iliad, who is the Co founder of 36 00:02:21,280 --> 00:02:23,160 Nir and CEO of the Nir Foundation. 37 00:02:23,800 --> 00:02:25,760 Nir is a sharded layer one blockchain. 38 00:02:26,240 --> 00:02:28,760 So welcome, welcome Iliad, welcome back on Epicenter. 39 00:02:28,840 --> 00:02:30,080 It's great to have you for a second time. 40 00:02:30,800 --> 00:02:33,040 Yeah. Thanks for having me and 41 00:02:33,040 --> 00:02:39,560 Congrats on, yeah, 10 years, epic, epic achievement in this 42 00:02:39,560 --> 00:02:42,120 space. Yeah, thanks so much. 43 00:02:42,120 --> 00:02:45,160 Yeah, like you said, it's it's basically 70 years in crypto. 44 00:02:45,360 --> 00:02:49,280 So we've we've all aged a bit yesterday in the episode we 45 00:02:49,280 --> 00:02:51,320 recorded yesterday, the 10 years episode, they had like the 46 00:02:51,320 --> 00:02:55,720 slideshow and you could see the progression of Maher, Brian and 47 00:02:56,200 --> 00:03:02,120 Sebastian like from their youth to their 40s or late 30s. 48 00:03:02,280 --> 00:03:06,000 So yeah, that's great, cool. Yeah, we actually wanted to 49 00:03:06,000 --> 00:03:12,440 start unconventionally with your background, but in in your case, 50 00:03:12,440 --> 00:03:15,880 it's a very interesting background in AI and machine 51 00:03:15,880 --> 00:03:19,880 learning. So we wanted to 1st sort of talk 52 00:03:19,880 --> 00:03:23,560 about your work there. You're one of the authors of the 53 00:03:23,560 --> 00:03:27,000 original Transformers paper. Can you. 54 00:03:27,000 --> 00:03:30,600 Yeah, maybe start by like telling us about your start in 55 00:03:30,600 --> 00:03:35,000 the AI and the mouse space. For sure, yeah. 56 00:03:35,000 --> 00:03:38,240 I mean, so I started tinkering with AII think. 57 00:03:38,240 --> 00:03:40,600 Even in high school I was actually excited about neural 58 00:03:40,600 --> 00:03:46,400 networks as a concept and I worked for machine learning 59 00:03:46,400 --> 00:03:49,760 company that was pretty old school machine learning company 60 00:03:50,880 --> 00:03:52,680 starting from first year of college. 61 00:03:54,120 --> 00:03:59,680 But when I saw kind of deep learning resurfacing in 2012 62 00:03:59,680 --> 00:04:04,040 thirteen there was this kind of Seminole work at a time which 63 00:04:05,240 --> 00:04:10,280 now now feels like duh. But back then was very exciting 64 00:04:10,280 --> 00:04:18,160 which was they trained in neural network to encode and like 65 00:04:18,160 --> 00:04:20,760 encode the image and then decode it back into the same image. 66 00:04:20,839 --> 00:04:27,320 So pre training what we know now as and that model without any 67 00:04:27,320 --> 00:04:29,720 supervision learn to detect cats. 68 00:04:29,840 --> 00:04:33,320 And so there was a neuron in the network which if you activate it 69 00:04:33,320 --> 00:04:36,960 it would it would generate a cat and like different types of cats 70 00:04:37,360 --> 00:04:41,120 And so it learns something like semantic without any training 71 00:04:41,120 --> 00:04:44,520 like in any like input data from here it's right just by looking 72 00:04:44,520 --> 00:04:47,240 at images. And so when I saw that and that 73 00:04:47,240 --> 00:04:51,160 was done by Google by Jeff Dean and Andrew Ng and they did it on 74 00:04:51,160 --> 00:04:55,680 a bunch of GPUs and like they they managed to scale it up. 75 00:04:55,680 --> 00:04:58,840 And I'm like I want to do that. I think that's that's the thing 76 00:04:58,840 --> 00:05:01,320 that's that's going to, you know actually change things. 77 00:05:01,760 --> 00:05:06,040 And so I joined Google research. My believe all this was that 78 00:05:06,040 --> 00:05:10,280 natural language not images going to be the driver for 79 00:05:10,280 --> 00:05:16,920 reasoning and for kind of like intelligence because you know 80 00:05:16,920 --> 00:05:20,040 there's many, many species in the world like hundreds of 81 00:05:20,040 --> 00:05:23,880 thousands of species that see and only one species that talks 82 00:05:23,880 --> 00:05:25,520 and has language. Right. 83 00:05:25,520 --> 00:05:28,680 So there's way more semantic information language. 84 00:05:29,120 --> 00:05:34,400 And so my team worked across a variety of things specifically 85 00:05:34,880 --> 00:05:37,920 question answering. So when you like type questions 86 00:05:37,920 --> 00:05:41,600 on.google.com we were actually running a neural networks to try 87 00:05:41,600 --> 00:05:47,200 to read our like pages that you see and respond to you was like 88 00:05:47,200 --> 00:05:49,800 a short answer so like you would see some damn short answers. 89 00:05:50,200 --> 00:05:54,400 Now the challenge was the neural networks at a time specifically 90 00:05:54,400 --> 00:05:57,800 recurrent neural networks were too slow to be put in 91 00:05:57,800 --> 00:06:00,880 production. And so we were just using bag of 92 00:06:00,880 --> 00:06:04,120 words models, which means you just literally throw all the 93 00:06:04,120 --> 00:06:08,800 words without any order into the model and it kind of tries to 94 00:06:08,800 --> 00:06:10,720 figure out what's going on. And it worked. 95 00:06:10,720 --> 00:06:15,960 It worked reasonably well. But and this is where kind of 96 00:06:18,280 --> 00:06:22,280 the Transformers gave birth, was like we could not use RNN in any 97 00:06:22,280 --> 00:06:26,040 practical use case. And so we were looking kind of 98 00:06:26,040 --> 00:06:29,320 for something. And so Jacob who was the manager 99 00:06:29,320 --> 00:06:32,800 and had like another team came up with this idea like they were 100 00:06:32,800 --> 00:06:37,240 using attention on top of words without any recurrence for for 101 00:06:37,240 --> 00:06:40,720 another task. And so kind of merging that idea 102 00:06:40,720 --> 00:06:43,720 was recurrence. Like can we use attention to 103 00:06:43,720 --> 00:06:48,760 somehow figure out which words are relevant in the order when 104 00:06:48,760 --> 00:06:51,960 you do answer questions or translate something. 105 00:06:52,720 --> 00:06:55,560 And that kind of gave birth to the Transformers really was like 106 00:06:55,840 --> 00:06:58,120 we need something that's really performant, that can be highly 107 00:06:58,120 --> 00:07:01,480 paralyzed. And attention is really good 108 00:07:01,480 --> 00:07:04,120 mechanism, you know, logically to do this. 109 00:07:04,520 --> 00:07:08,360 But if you package it all kind of the way this models really 110 00:07:08,360 --> 00:07:10,960 work is that everything happens in parallel. 111 00:07:11,320 --> 00:07:14,280 Like the way I like to describe it, there's this movie Arrival 112 00:07:14,520 --> 00:07:18,120 where aliens talk in the whole sentence at the same time. 113 00:07:18,240 --> 00:07:21,240 Like there's like a circle of scrub list, but they produce it 114 00:07:21,240 --> 00:07:22,840 at the same time. And that's kind of how 115 00:07:22,840 --> 00:07:24,560 Transformers actually read articles. 116 00:07:24,800 --> 00:07:26,360 It's not like one word at a time. 117 00:07:26,520 --> 00:07:30,360 It's literally reads the whole article, all the words in 118 00:07:30,360 --> 00:07:33,760 parallel and then like has multiple steps to kind of 119 00:07:33,760 --> 00:07:37,680 process it and reconcile this like the understanding of it and 120 00:07:37,680 --> 00:07:40,240 then it answers the question and so. 121 00:07:40,480 --> 00:07:44,320 So that lays out really well for the modern hardware GPUs that we 122 00:07:44,320 --> 00:07:46,440 use. And so it allows to have like 123 00:07:46,440 --> 00:07:50,000 this massive kind of performance improvement which means also you 124 00:07:50,000 --> 00:07:55,480 can scale out the models. And so I've was a, you know, 125 00:07:55,480 --> 00:07:59,160 worked on that was a team of amazing researchers which now 126 00:07:59,160 --> 00:08:01,200 all went to do really cool stuff. 127 00:08:02,000 --> 00:08:07,440 And then at the time I decided to leave Google to start an AI 128 00:08:07,440 --> 00:08:11,560 company near AI, which was supposed to be pretty much 129 00:08:11,560 --> 00:08:14,760 teaching machines to code. So my belief, and I still 130 00:08:14,760 --> 00:08:19,200 believe this, that now given this, this steps of models, you 131 00:08:19,200 --> 00:08:21,280 can change how we interact with computing. 132 00:08:21,640 --> 00:08:25,680 You can actually talk to computers and they do work for 133 00:08:25,680 --> 00:08:29,080 you instead of needing to have an engineer to write code for 134 00:08:29,080 --> 00:08:31,240 you, right. Which again like now it seems 135 00:08:31,360 --> 00:08:33,240 more obvious that that's possible. 136 00:08:33,480 --> 00:08:39,960 Back in 2017 there was like huh And so so we started near AI but 137 00:08:39,960 --> 00:08:43,080 we only we gave us us a year because obviously at that time 138 00:08:43,080 --> 00:08:46,800 it wasn't moon shot and we didn't have that much resources. 139 00:08:46,800 --> 00:08:51,720 So we're doing some interesting stuff around data collection and 140 00:08:51,720 --> 00:08:55,440 some machine learning. But one thing we ended up doing 141 00:08:55,440 --> 00:08:58,160 is getting a lot of people around the world actually doing 142 00:08:59,440 --> 00:09:02,120 like writing some code for us, writing some descriptions for 143 00:09:02,120 --> 00:09:04,760 the code. And so we we had to struggle to 144 00:09:04,760 --> 00:09:08,240 pay them because they were mostly students in China and 145 00:09:08,240 --> 00:09:11,480 Ukraine and Russia in kind of some other countries. 146 00:09:11,800 --> 00:09:15,080 And like some of them don't have bank accounts, some like 147 00:09:15,160 --> 00:09:17,520 Ukraine, for example, PayPal doesn't work. 148 00:09:17,520 --> 00:09:22,600 In China, PayPal doesn't work and so there's like no good way 149 00:09:22,600 --> 00:09:26,040 to do it like programmatically to send people money. 150 00:09:26,280 --> 00:09:29,200 And so we started looking at blockchain as like, hey, can we 151 00:09:29,200 --> 00:09:33,560 just send people money easily in in code. 152 00:09:34,360 --> 00:09:39,320 And the answer was in 2018 the answer was actually no because 153 00:09:39,320 --> 00:09:41,880 even back then the fees on Bitcoin and stadium were way too 154 00:09:41,880 --> 00:09:46,240 high And and then as you probably know when you start on 155 00:09:46,240 --> 00:09:49,160 the blockchain rabbit hole, you cannot stop and just keep 156 00:09:49,160 --> 00:09:51,160 digging and you're like wait what is this? 157 00:09:52,120 --> 00:09:55,720 And so, so we kind of as we kept digging and researching 158 00:09:55,720 --> 00:09:58,280 different block chains and different technologies, we're 159 00:09:58,280 --> 00:10:03,440 like wait, we actually know how to build something of this sort, 160 00:10:03,440 --> 00:10:05,440 right. So my Co founder Alex, he was 161 00:10:05,440 --> 00:10:10,240 building Sharda database company before and we have like you know 162 00:10:10,240 --> 00:10:15,160 systems background, we're like we can probably do this, but we 163 00:10:15,160 --> 00:10:19,320 can focus on user experience, developer experience while kind 164 00:10:19,320 --> 00:10:22,480 of solving the scalability underneath and making sure fees 165 00:10:22,480 --> 00:10:25,160 are staying stable. And so that's kind of how we 166 00:10:25,160 --> 00:10:30,440 went from near AI to becoming near protocol in 2018 and and 167 00:10:30,440 --> 00:10:36,560 starting this journey. So Elia in this current wave of 168 00:10:36,560 --> 00:10:39,480 LLMS. Of course like this attention 169 00:10:39,480 --> 00:10:44,680 mechanism is a key part but another key part is just the 170 00:10:44,680 --> 00:10:48,120 idea of you know like just the idea of scale. 171 00:10:48,200 --> 00:10:53,800 I'd like collected a lot of data from the Internet from books and 172 00:10:53,800 --> 00:10:58,560 then pre train the model and of course the ideas of RLHF and all 173 00:10:58,600 --> 00:11:00,800 they came later. But the fundamental idea is you 174 00:11:00,800 --> 00:11:04,160 throw in a lot of data, you pre train, you make a big model sort 175 00:11:04,160 --> 00:11:09,240 of produce good results. Did you anticipate that scale 176 00:11:09,240 --> 00:11:14,440 was going to work this well, and if so, why would did you use 177 00:11:14,440 --> 00:11:20,040 that approach in Near? AI, no, I I that that's a part 178 00:11:20,040 --> 00:11:24,240 that definitely kind of was interesting to see that as 179 00:11:24,240 --> 00:11:27,680 people scaled up the models, they became like they started 180 00:11:27,680 --> 00:11:31,200 exhibiting kind of properties like more and more sophisticated 181 00:11:31,200 --> 00:11:34,560 reasoning properties. And it's like it makes sense now 182 00:11:34,560 --> 00:11:37,480 that you know you think about it like the capacity of the model 183 00:11:37,480 --> 00:11:39,520 is higher. It's able to like generalize 184 00:11:39,520 --> 00:11:42,520 better. It's able to kind of learn quote 185 00:11:42,520 --> 00:11:45,640 UN quote programs that it can execute. 186 00:11:46,280 --> 00:11:53,080 But yeah at the time that wasn't like particularly clear that 187 00:11:54,560 --> 00:11:57,320 like it will be that kind of step function change. 188 00:11:57,600 --> 00:12:01,840 And so we yeah we were not at nearly I we're not doing that 189 00:12:01,840 --> 00:12:05,120 partially because we also just didn't have you know like we 190 00:12:05,120 --> 00:12:10,200 raised you know small C like pre C drown actually And we we 191 00:12:10,200 --> 00:12:16,520 thought we could get better supervised data instead and we 192 00:12:16,520 --> 00:12:19,720 did we did some pre training on like GitHub and and things like 193 00:12:19,720 --> 00:12:22,120 that. But we we didn't thought, didn't 194 00:12:22,120 --> 00:12:26,080 think of like training the whole Internet, that's like large 195 00:12:26,080 --> 00:12:28,800 scale and we did have resources to do something like that 196 00:12:28,800 --> 00:12:33,680 either. And the other interesting thing 197 00:12:33,680 --> 00:12:36,920 is kind of like this. This attention mechanism also 198 00:12:36,920 --> 00:12:41,320 seems to like it's built for natural language processing, but 199 00:12:41,320 --> 00:12:45,240 also seems to kind of work across different modalities like 200 00:12:45,240 --> 00:12:48,280 such as like images and maybe video in the future. 201 00:12:49,360 --> 00:12:53,880 And like how does that come across to you right? 202 00:12:53,880 --> 00:12:57,320 Like is that unexpected or is that is that something you 203 00:12:57,320 --> 00:13:03,080 expected in the past? I mean like when the 204 00:13:03,320 --> 00:13:06,160 Transformers were just in development there was like like 205 00:13:06,160 --> 00:13:09,040 the teams actually tried them on different modalities, I mean not 206 00:13:09,040 --> 00:13:11,680 like multi model models but different modalities and it was 207 00:13:12,040 --> 00:13:14,040 pretty interesting to see it worked really well. 208 00:13:15,040 --> 00:13:19,200 So I think it's that was kind of known that it works on on 209 00:13:19,200 --> 00:13:22,160 different modalities pretty early on. 210 00:13:22,480 --> 00:13:26,280 I think the kind of the intuition there is really that 211 00:13:26,320 --> 00:13:31,000 you know the way kind of we work as well as very much like like 212 00:13:31,080 --> 00:13:34,560 our our eyes actually like move all the time every like I I 213 00:13:34,880 --> 00:13:37,920 forgot how many milliseconds. And so we actually kind of pay 214 00:13:37,920 --> 00:13:40,760 attention to different parts and then our brain kind of 215 00:13:40,760 --> 00:13:45,880 reconstructs the image right at different levels and kind of you 216 00:13:45,880 --> 00:13:47,160 know residential language same right. 217 00:13:47,160 --> 00:13:50,160 You read sentences you like build some semantic meaning and 218 00:13:50,160 --> 00:13:53,400 then you know it kind of continue building out this the 219 00:13:53,400 --> 00:13:57,960 meaning of the what you read. But sometimes you like zoom in 220 00:13:57,960 --> 00:14:03,720 on specific words when you need to answer a question and so like 221 00:14:03,720 --> 00:14:07,480 I think like generally speaking there is like intuition behind 222 00:14:07,480 --> 00:14:12,040 this but obviously again it's like it it's interesting to see 223 00:14:12,040 --> 00:14:16,960 how well it all works right Definitely you know like we had 224 00:14:16,960 --> 00:14:20,000 we had a pretty good models like even before it. 225 00:14:20,000 --> 00:14:22,680 It just like they were super slow and like non you couldn't 226 00:14:22,680 --> 00:14:26,720 use them in production at all but this you know obviously like 227 00:14:27,160 --> 00:14:31,000 the the scale is which for example Open AI went and and 228 00:14:31,000 --> 00:14:33,920 scale it up and and by the way they did a tremendous amount of 229 00:14:33,920 --> 00:14:37,080 work to make it work like it's not we cannot take it for 230 00:14:37,080 --> 00:14:38,520 granted. They're just like oh we just 231 00:14:38,520 --> 00:14:41,440 increased parameters and hit enter like no it was a ton of 232 00:14:41,440 --> 00:14:46,400 work across the board from you know low level engineering to 233 00:14:46,400 --> 00:14:49,960 like fine tuning to you know they changed some of the model 234 00:14:50,520 --> 00:14:52,800 kind of details of model architecture as well. 235 00:14:54,640 --> 00:14:56,800 But yeah, like the. It's it is. 236 00:14:56,800 --> 00:15:00,080 It is It was surprising for me like, I think like that when it 237 00:15:00,080 --> 00:15:04,560 went from 2:00 to 3:00 that was like interesting like two. 238 00:15:04,560 --> 00:15:06,120 It was kind of like, OK, yeah, I get it. 239 00:15:06,120 --> 00:15:08,520 Like we've trained models like that at Google kind of thing 240 00:15:09,600 --> 00:15:11,600 from 2:00 to 3:00. It was like, OK, that's that's 241 00:15:11,600 --> 00:15:16,840 really interesting because I can see the, you know, there's like 242 00:15:16,840 --> 00:15:20,800 something more now happens and obviously it's 3.5 is where 243 00:15:20,800 --> 00:15:24,360 like, OK, yeah, that's like it actually learns something that 244 00:15:25,080 --> 00:15:28,040 is like beyond just language modeling, right? 245 00:15:28,040 --> 00:15:32,040 Like there's some reasoning that this is extractable now through 246 00:15:32,080 --> 00:15:34,120 kind of this instruction fine tuning. 247 00:15:36,080 --> 00:15:38,880 On a high level, I'm I'm actually curious what your 248 00:15:39,000 --> 00:15:45,280 stance is on this stochastic parrot versus understanding 249 00:15:45,280 --> 00:15:48,840 spectrum. So, so there are, there are 250 00:15:48,840 --> 00:15:54,240 people in the AI community that that say that LLMS actually 251 00:15:54,240 --> 00:15:59,120 don't understand anything. They are stochastic parrots in 252 00:15:59,120 --> 00:16:03,200 the sense that they have understood the statistics of 253 00:16:03,200 --> 00:16:08,640 what word follows what other word in language because they 254 00:16:08,640 --> 00:16:12,680 have seen billions of examples. And when you talk to an LLM and 255 00:16:12,680 --> 00:16:17,360 it's generating words, it's just replicating the statistics of 256 00:16:17,360 --> 00:16:21,200 what it has seen in the past without actual any understanding 257 00:16:21,200 --> 00:16:25,160 behind behind the box. That's the like, at the extreme, 258 00:16:25,160 --> 00:16:27,200 that's the stochastic parrot view. 259 00:16:27,800 --> 00:16:32,240 And on the other extreme, perhaps there's a view, maybe 260 00:16:32,240 --> 00:16:36,240 like the ES upscale view, which is kind of when you force a 261 00:16:36,240 --> 00:16:41,400 model to predict the next word and you force it to do it again 262 00:16:41,400 --> 00:16:45,000 and again in order to predict the next word. 263 00:16:45,000 --> 00:16:50,320 It has to Start learning something about the world itself 264 00:16:51,360 --> 00:16:53,520 to do the the job of prediction well. 265 00:16:54,120 --> 00:16:58,600 And in kind of trying to predict it well, it is forced to learn 266 00:16:58,600 --> 00:17:03,520 about the world and so it has actual actual intelligence about 267 00:17:03,800 --> 00:17:09,720 what world it has in it is in. So it's not just a stochastic 268 00:17:09,720 --> 00:17:11,480 parrot. This is actually when you're 269 00:17:11,480 --> 00:17:14,720 talking to GPD 4, you're talking to something which has 270 00:17:14,960 --> 00:17:18,640 understanding distilled into it and they seem to be like these 271 00:17:18,640 --> 00:17:22,000 two extremes in the in the space and I'm I'm curious like where 272 00:17:22,000 --> 00:17:25,480 you stand on on that on that debate. 273 00:17:26,760 --> 00:17:32,240 Yeah, I I mean I definitely closer to Discover's view like 274 00:17:32,360 --> 00:17:35,920 from my perspective kind of you know it at the end is bunch of 275 00:17:35,920 --> 00:17:38,360 math, right. And so like you can kind of 276 00:17:38,360 --> 00:17:43,040 decompose what this math is doing and you know try to build 277 00:17:43,040 --> 00:17:46,080 an intuition around like types of transformations. 278 00:17:46,080 --> 00:17:51,320 It can, it can or cannot do. And so from my perspective kind 279 00:17:51,320 --> 00:17:53,960 of you know the first step is you take the document and you 280 00:17:53,960 --> 00:17:57,080 embed it, right. So you went from words into a 281 00:17:57,080 --> 00:18:00,080 multi into dots and multi dimensional space, right. 282 00:18:00,080 --> 00:18:02,000 So I mean let's let's for a second imagine it's 283 00:18:02,000 --> 00:18:04,000 two-dimensional although it's multiple. 284 00:18:04,240 --> 00:18:07,960 And so there is like the words that are similar, right. 285 00:18:07,960 --> 00:18:10,440 Are you know close to into space. 286 00:18:10,440 --> 00:18:14,440 The words that are far. Now you have a next layer which 287 00:18:14,440 --> 00:18:18,600 transforms us words, right, to kind of give them more context. 288 00:18:20,280 --> 00:18:23,080 And so just, you know, think of it as rotation in the space and 289 00:18:23,920 --> 00:18:27,120 then you have a tension which is you know, you're trying to kind 290 00:18:27,120 --> 00:18:30,560 of given the, the current word, you know, try to pull in the 291 00:18:30,560 --> 00:18:34,960 context of the words around it to give to give it more semantic 292 00:18:34,960 --> 00:18:37,280 meaning. And so that's another 293 00:18:37,280 --> 00:18:40,640 transformation, right. So like in a way you take kind 294 00:18:40,640 --> 00:18:43,080 of set of words, right, and then you kind of keep transforming 295 00:18:43,080 --> 00:18:46,320 them. And so it it what it learns is 296 00:18:46,320 --> 00:18:48,840 the transformation function which in a way is a program. 297 00:18:49,160 --> 00:18:53,360 It's a program that is trying to transform the words into a level 298 00:18:53,360 --> 00:18:58,080 which is useful then to predict next word right or and then like 299 00:18:58,080 --> 00:19:02,800 later respond to questions. And so kind of is this like a 300 00:19:02,800 --> 00:19:07,160 pure stochastic parrot where it's like well pure stochastic 301 00:19:07,160 --> 00:19:10,280 parrot we had when we were doing just like you know, we were 302 00:19:10,280 --> 00:19:12,160 generating Wikipedia articles for example, right. 303 00:19:12,160 --> 00:19:15,160 You just give it a name and just say generate a Wikipedia article 304 00:19:15,520 --> 00:19:19,000 like that's pure like you know there there's it just makes up 305 00:19:19,000 --> 00:19:21,440 stuff because like that that name doesn't exist right. 306 00:19:21,440 --> 00:19:24,440 There's no there's nothing So just generate something that 307 00:19:24,440 --> 00:19:28,640 looks like an article But when we when we starting to look at 308 00:19:28,640 --> 00:19:32,240 like OK well how would you answer to this question right. 309 00:19:32,440 --> 00:19:35,600 It it to be able to do that right it needs to kind of 310 00:19:35,600 --> 00:19:38,360 process information right. It it does this kind of 311 00:19:38,360 --> 00:19:44,600 transformations on the on the article and like it's trying to 312 00:19:44,600 --> 00:19:47,680 connect contextualize that and and give give the answer. 313 00:19:47,680 --> 00:19:51,400 So in a way like I think of it as it's learn some set of 314 00:19:51,400 --> 00:19:54,200 programs that like our world has right. 315 00:19:54,200 --> 00:19:56,600 So like it's not a complete word model, right. 316 00:19:56,600 --> 00:20:00,560 It clearly has a lot of gaps, but it is a kind of set of 317 00:20:00,560 --> 00:20:04,760 programs that our like world model has that it can apply to 318 00:20:04,760 --> 00:20:09,480 be able to to answer well or predict next word for a training 319 00:20:09,920 --> 00:20:13,680 and that in itself is really useful, right, as we see. 320 00:20:14,120 --> 00:20:17,960 But it's also because it has so many gaps it it, it has issues 321 00:20:17,960 --> 00:20:21,760 with doing some you know, kind of specific things and the more 322 00:20:21,920 --> 00:20:25,960 precise it needs to be, the less well it does, right. 323 00:20:25,960 --> 00:20:30,600 Because it kind of ends up being like if either as a program, the 324 00:20:30,600 --> 00:20:34,000 programs are very probabilistic and kind of semantic versus you 325 00:20:34,000 --> 00:20:37,320 know, if you ask it to like describe the steps of of 326 00:20:37,320 --> 00:20:40,120 something. But at the same time a lot of 327 00:20:40,120 --> 00:20:44,280 the things we do is kind of like there's like few thing core 328 00:20:44,280 --> 00:20:46,280 things and then everything else you kind of fill in 329 00:20:46,280 --> 00:20:48,880 automatically, right. So that's why it's really good 330 00:20:48,880 --> 00:20:52,480 like even at coding like most of the most of the coding we do 331 00:20:52,480 --> 00:20:57,240 right is actually kind of boilerplatey and so and so 332 00:20:57,280 --> 00:21:00,240 there's like few not just you can actually get to like a 333 00:21:00,240 --> 00:21:04,040 reasonable code and that's why I think like things like copilots 334 00:21:04,280 --> 00:21:06,760 are in pretty good products in result. 335 00:21:09,040 --> 00:21:11,880 Cool. So look, turning to applicative 336 00:21:11,880 --> 00:21:18,280 views, so now this LMS are are pretty amazing and you have some 337 00:21:18,280 --> 00:21:21,160 applicative ideas on applying them to the near ecosystem. 338 00:21:21,160 --> 00:21:24,760 So yeah, what? What are they and how do you see 339 00:21:24,760 --> 00:21:28,880 that unfold? Yeah, I think I think of this 340 00:21:28,880 --> 00:21:33,680 kind of across 3 dimensions. So the first dimension is 341 00:21:33,680 --> 00:21:38,200 actually less about AI itself and more about our kind of 342 00:21:38,200 --> 00:21:41,800 society. And this is the idea that kind 343 00:21:41,800 --> 00:21:46,720 of as more content is generated as there's more kind of 344 00:21:46,720 --> 00:21:50,000 information wars in general misinformation. 345 00:21:50,000 --> 00:21:53,280 And again, the important part to note, misinformation is not AI 346 00:21:53,280 --> 00:21:58,000 problem, it's a human problem. the IT, you know, we are in 347 00:21:58,000 --> 00:22:01,200 crypto space and so Byzantine generals is something that our 348 00:22:01,200 --> 00:22:04,200 space is based on. And that's literally the, you 349 00:22:04,200 --> 00:22:09,760 know, the miss cite citeable misinformation like. 350 00:22:10,240 --> 00:22:14,480 And so the idea of like misinformation of of malicious 351 00:22:14,480 --> 00:22:18,440 like attack on information is something that exists from, you 352 00:22:18,440 --> 00:22:27,440 know from like early on. And so from my perspective, the 353 00:22:27,440 --> 00:22:33,640 way to kind of start solving that is to to bring the kind of 354 00:22:33,640 --> 00:22:39,920 security, cryptography and reputation to a level of of the 355 00:22:39,920 --> 00:22:42,400 content of the of individual pieces of content. 356 00:22:42,400 --> 00:22:45,600 So right now, for example, we are, you know, using websites, 357 00:22:45,600 --> 00:22:50,040 we have HTTPS and so we have, we have some set of security 358 00:22:50,240 --> 00:22:52,800 guarantees around accessing specific websites. 359 00:22:53,160 --> 00:22:56,440 But the content on the website can be coming from anywhere. 360 00:22:56,720 --> 00:22:59,640 It can be saying anything and there's no way to kind of 361 00:23:00,040 --> 00:23:03,520 maintain reputation, contacts, comments, etcetera around it. 362 00:23:03,520 --> 00:23:06,760 So we need a new set of standards around that, so that 363 00:23:06,760 --> 00:23:10,040 you can hover an image and it shows you or a video or piece of 364 00:23:10,040 --> 00:23:14,360 text and it tells you like who published it, when it was done, 365 00:23:15,400 --> 00:23:18,840 if there's any side comments or contacts etcetera from reputable 366 00:23:18,840 --> 00:23:20,520 sources that should be attached to it. 367 00:23:20,840 --> 00:23:24,400 So for that we need blockchain, we need, you know, set of 368 00:23:24,400 --> 00:23:27,400 standards, we need browser support and we need kind of 369 00:23:27,400 --> 00:23:29,440 publishers to be supporting this. 370 00:23:30,200 --> 00:23:32,920 And I think that's a really important part for our society 371 00:23:32,920 --> 00:23:36,120 generally because otherwise we're going to be living in a 372 00:23:36,120 --> 00:23:39,400 world of kind of, you know, all the content is like you never 373 00:23:39,400 --> 00:23:41,040 know what if it's true or not, right. 374 00:23:41,040 --> 00:23:44,920 And it's constantly like kind of manipulation around that. 375 00:23:45,800 --> 00:23:51,560 Now kind of the second pillar for me is I call it kind of 376 00:23:51,560 --> 00:23:55,840 decentralized AGI, right. So if we assume you know this 377 00:23:55,840 --> 00:24:00,400 models are getting more powerful, more intelligent, what 378 00:24:00,400 --> 00:24:03,600 you definitely don't want is a single company or you know two 379 00:24:03,600 --> 00:24:06,400 or three companies deciding what's right and wrong for this 380 00:24:06,400 --> 00:24:09,240 models to do. You don't want to like them to 381 00:24:09,240 --> 00:24:11,920 decide what you're allowed to do and what you're not allowed to 382 00:24:11,920 --> 00:24:15,120 do as models. It's also like it's the same 383 00:24:15,120 --> 00:24:19,480 thing that happened with social networks, like being a kind of 384 00:24:19,760 --> 00:24:22,520 moral police for the world just doesn't work. 385 00:24:22,720 --> 00:24:24,800 The world is very multi dimensional. 386 00:24:25,000 --> 00:24:28,400 Something that's legal in Amsterdam is completely legal in 387 00:24:28,800 --> 00:24:31,240 a lot of other countries and the other way around. 388 00:24:31,640 --> 00:24:35,360 And so like you know what moral is, is even more complicated. 389 00:24:36,400 --> 00:24:41,280 And so it's really important to have community be governing kind 390 00:24:41,280 --> 00:24:46,840 of the alignment safety as well as kind of the, you know, 391 00:24:46,840 --> 00:24:50,680 instruction data sets that these models are trained on and as 392 00:24:50,680 --> 00:24:54,200 well as like being able to validate that the model you run 393 00:24:54,200 --> 00:24:57,760 is actually the model that you wanted to run, right. 394 00:24:57,760 --> 00:25:02,280 So right now if you call, you know, GB TAVI or Google API, you 395 00:25:02,280 --> 00:25:05,320 get a response, you have no idea which that could produce that 396 00:25:05,320 --> 00:25:07,360 response. You have no guarantees that it 397 00:25:07,360 --> 00:25:10,440 was the model that you wanted to run. 398 00:25:10,920 --> 00:25:13,560 And actually sometimes it's not because they're trying to 399 00:25:13,560 --> 00:25:17,840 optimize costs. And so like how do you actually 400 00:25:18,280 --> 00:25:20,560 have this guarantees? And especially for something 401 00:25:20,560 --> 00:25:23,920 that's mission critical, right, like like if I'm doing trading 402 00:25:23,920 --> 00:25:28,040 on this, if I'm doing healthcare, like any kind of 403 00:25:28,040 --> 00:25:30,240 business decisions, right. You want to make sure to, you 404 00:25:30,240 --> 00:25:33,760 know you're accessing the model that you have predictable 405 00:25:33,760 --> 00:25:36,800 parameters and outputs. And so for that we need 406 00:25:36,960 --> 00:25:39,080 decentralized inference. We need kind of model 407 00:25:39,080 --> 00:25:43,720 marketplaces, We need kind of community data, crowdsourcing 408 00:25:43,720 --> 00:25:47,600 and data management governance and took kind of the whole stack 409 00:25:47,600 --> 00:25:50,240 of tooling that really manages this. 410 00:25:50,480 --> 00:25:54,240 And then, you know on top of this you'll be able to to kind 411 00:25:54,240 --> 00:25:57,880 of interact with it in a hopeful like. 412 00:25:58,040 --> 00:26:00,080 I think the other way is like making sure it's privacy 413 00:26:00,080 --> 00:26:02,440 preserving so that when you interact with it, you have it. 414 00:26:02,640 --> 00:26:04,520 So there's a lot of work to be done. 415 00:26:04,520 --> 00:26:06,360 There's a lot of like there's a bunch of startups doing 416 00:26:06,360 --> 00:26:09,760 decentralized inference. There is still privacy gap. 417 00:26:09,800 --> 00:26:12,720 I think that people are researching, but it's still 418 00:26:12,720 --> 00:26:16,760 pretty far. There's some data marketplaces, 419 00:26:16,760 --> 00:26:20,240 there's some other kind of pieces, but it's not really, I 420 00:26:20,240 --> 00:26:23,560 would say like combine into like a product story yet. 421 00:26:23,800 --> 00:26:26,440 But I think like that's a really important for like humanity 422 00:26:26,440 --> 00:26:29,000 period. Because otherwise, you know, 423 00:26:29,440 --> 00:26:35,200 like tomorrow you go to your favorite, you know, AI model and 424 00:26:35,200 --> 00:26:38,200 it says like, oh, you're banned or you know, you use the 425 00:26:38,200 --> 00:26:41,360 incorrect word and so now or something, right. 426 00:26:42,640 --> 00:26:45,120 So all the usual stuff we've seen before. 427 00:26:45,640 --> 00:26:50,560 And then finally, I actually think the the flip side of this 428 00:26:51,000 --> 00:26:55,560 is local models, right? Because although like this big 429 00:26:55,560 --> 00:26:58,080 models, they have the world knowledge, they have maybe 430 00:26:58,080 --> 00:27:00,080 access to lots and lots of context. 431 00:27:00,960 --> 00:27:04,800 But actually what you want most of the time is a model that 432 00:27:04,800 --> 00:27:07,960 knows everything about you. But you don't want all this data 433 00:27:07,960 --> 00:27:10,600 to go anywhere else, right? You want it to live with you on 434 00:27:10,600 --> 00:27:14,560 your machine, on your, you know, your private encrypted data 435 00:27:14,560 --> 00:27:17,520 store and you want to model the table to access that. 436 00:27:17,520 --> 00:27:20,360 So you want a local model that is personalized for you. 437 00:27:20,440 --> 00:27:23,000 You control it. It's not affected and 438 00:27:23,000 --> 00:27:27,000 manipulated in any way by, you know, advertisement giants. 439 00:27:27,200 --> 00:27:31,120 And so it's actually on your side and there's just responding 440 00:27:31,120 --> 00:27:36,160 kind of the way you would like to not the way you know tied 441 00:27:36,160 --> 00:27:40,520 once you or whatever. And so I think that is a really 442 00:27:40,520 --> 00:27:42,920 important side of kind of as well. 443 00:27:42,920 --> 00:27:45,680 And so we're actually been playing around with like Edge 444 00:27:45,680 --> 00:27:49,640 Intelligence and I did a couple events and being kind of talking 445 00:27:49,640 --> 00:27:54,080 with some projects around this space and it's like it's 446 00:27:54,080 --> 00:27:58,640 actually it's lasts VEP 3 like you know sense of blockchain, 447 00:27:59,040 --> 00:28:01,880 but it's more of a three in a sense of principles, right. 448 00:28:01,880 --> 00:28:06,040 It's user owned AI is you know controlling your own data. 449 00:28:06,320 --> 00:28:09,440 It's like all of those values that we to talk about and I 450 00:28:09,440 --> 00:28:12,520 think that and kind of the web series of custody will be 451 00:28:12,520 --> 00:28:16,840 converging more kind of in a in in on the principal side, right 452 00:28:16,840 --> 00:28:20,360 maybe and like on technology side as well and this kind of 453 00:28:20,560 --> 00:28:24,000 the area I'm I'm most excited and working on right now. 454 00:28:25,880 --> 00:28:29,640 So in practice how how are you like approaching this? 455 00:28:29,640 --> 00:28:35,240 Are there like teams you are funding or is like there AI team 456 00:28:35,240 --> 00:28:37,600 in the year or how? How can we imagine this? 457 00:28:38,640 --> 00:28:40,080 Yeah. So we've been working with some 458 00:28:40,080 --> 00:28:42,560 of AI teams. We actually just had a Neo con 459 00:28:42,920 --> 00:28:45,440 about a months ago and we headed AI track there with some 460 00:28:45,440 --> 00:28:49,800 projects presenting that we already working with as well as 461 00:28:50,400 --> 00:28:55,280 kind of I'm working as like advisor with a few projects kind 462 00:28:55,840 --> 00:29:01,600 of more closely and we do have I would say like AI efforts more 463 00:29:01,720 --> 00:29:05,560 on also just automating our own operations like. 464 00:29:05,720 --> 00:29:09,080 So the other side of this is I think kind of the ecosystem 465 00:29:09,080 --> 00:29:13,720 itself should become AI enabled and over time AI ran. 466 00:29:14,200 --> 00:29:19,520 So like ideally my, you know, my job and kind of the job of 467 00:29:19,520 --> 00:29:21,720 coordinating the ecosystem should be done by AI. 468 00:29:22,160 --> 00:29:27,680 And by the way the AI is a kind of like this approach actually 469 00:29:27,680 --> 00:29:30,520 solved the core problem of humanity and of resource 470 00:29:30,520 --> 00:29:33,200 coordination. The core problem of humanity is 471 00:29:33,240 --> 00:29:37,080 principal agent problem is that when we want somebody to to to 472 00:29:37,080 --> 00:29:42,920 do stuff on our behalf like we select, you know, in elections 473 00:29:42,920 --> 00:29:46,720 or we hire someone to manage our money or something else, they 474 00:29:46,720 --> 00:29:51,040 have their own needs and and they have their own wants and so 475 00:29:51,040 --> 00:29:54,880 their decisions are usually not fully aligned with us who hired 476 00:29:54,880 --> 00:29:56,280 them. So that's called principal agent 477 00:29:56,280 --> 00:29:58,760 problem. And so AI actually being the 478 00:29:58,760 --> 00:30:02,280 agent that behaves on our behalf, acts on our behalf is 479 00:30:02,280 --> 00:30:04,920 the way to solve that. And if you scale it up to kind 480 00:30:04,920 --> 00:30:09,200 of governance level, right, like actually having AI, being the 481 00:30:09,280 --> 00:30:13,040 actor that you know, makes decisions based on what the 482 00:30:13,040 --> 00:30:16,240 population wants, is the way to solve a lot of the current 483 00:30:16,240 --> 00:30:18,960 challenges was, you know, when you're like someone, they do 484 00:30:19,040 --> 00:30:21,160 stupid things, right? Or not, things that they 485 00:30:21,160 --> 00:30:23,520 promised to do. Kind of that's a way to like 486 00:30:23,520 --> 00:30:25,160 really address it. And so there's a really 487 00:30:25,160 --> 00:30:28,800 interesting kind of future of governance there. 488 00:30:29,000 --> 00:30:32,040 But like, we can start applying it now in this decentralized 489 00:30:32,040 --> 00:30:34,520 ecosystems because they're already fully digital, they 490 00:30:34,520 --> 00:30:37,840 already have kind of like all the actions are in chain, right? 491 00:30:37,840 --> 00:30:41,080 So you can have traceability, you can have like veto power, 492 00:30:41,080 --> 00:30:42,720 etcetera, if something goes wrong. 493 00:30:43,000 --> 00:30:48,280 And so I'm really excited about also that side of the applying 494 00:30:48,360 --> 00:30:51,960 AI in in the three space and obviously you need that whole AI 495 00:30:52,440 --> 00:30:54,440 like decentralized stack to do that. 496 00:30:55,000 --> 00:30:58,120 But we are kind of starting to do it from bottom up on on our 497 00:30:58,120 --> 00:31:00,760 side just in foundation for example like hey what are things 498 00:31:00,760 --> 00:31:03,160 we can automate, what are things that we can like start 499 00:31:03,160 --> 00:31:06,280 leveraging this technology for as well as maybe build some of 500 00:31:06,280 --> 00:31:10,720 the tooling for developers to build tariff AI enabled things 501 00:31:10,720 --> 00:31:14,320 in in in the space. We also have, yeah, a bunch of 502 00:31:14,320 --> 00:31:16,800 projects that are kind of experimenting with this across 503 00:31:16,800 --> 00:31:20,600 different areas. Yeah that's super awesome. 504 00:31:20,600 --> 00:31:25,640 I I also saw actually your Co founder like the Alex working 505 00:31:26,320 --> 00:31:30,320 directly on like smarter LLMS. Can you maybe also like what 506 00:31:30,320 --> 00:31:32,680 what's that about? Is that related to me or is it 507 00:31:32,680 --> 00:31:35,200 like some sort of totally different thing or what can you 508 00:31:35,200 --> 00:31:35,960 share about that? Yeah. 509 00:31:35,960 --> 00:31:38,480 So I mean it's it's a stealth project right now. 510 00:31:38,480 --> 00:31:42,280 So I'll I'll not go into too detail maybe you'll have him you 511 00:31:42,280 --> 00:31:46,680 know in a in at some point to go more in depth into it. 512 00:31:47,080 --> 00:31:50,120 Well, yeah, I mean we kind of so I'm advisors there and we 513 00:31:50,120 --> 00:31:52,520 working kind of I would say side by side. 514 00:31:52,520 --> 00:31:56,040 But yeah, he's focusing more on the lower level and like kind of 515 00:31:56,040 --> 00:31:59,040 preparing for the future of of this as well. 516 00:32:00,720 --> 00:32:03,640 Yeah. I think, I guess maybe you 517 00:32:03,680 --> 00:32:08,160 you're mentioning right, like AI sort of also like making our 518 00:32:08,360 --> 00:32:11,000 life easier in the sense of operationally in the 519 00:32:11,120 --> 00:32:14,320 organizations, but also I guess, yeah, in the wider society. 520 00:32:14,320 --> 00:32:17,680 And I guess that's always been like a huge focus of Near. 521 00:32:17,960 --> 00:32:23,200 So yeah, we wanted to sort of dive also in that side of Near 522 00:32:23,200 --> 00:32:29,000 where basically you're branded now in many places like as the 523 00:32:29,000 --> 00:32:30,920 blockchain operating system. And I think, yeah, one of the 524 00:32:30,920 --> 00:32:34,800 core features around that is like sort of the the UX focus of 525 00:32:34,800 --> 00:32:38,240 new. So maybe yeah, can you explain 526 00:32:38,240 --> 00:32:42,560 to us how Near has sort of approached yeah basically 527 00:32:42,560 --> 00:32:46,000 usability for for developers and and users in the in blockchain 528 00:32:46,000 --> 00:32:47,760 systems and and what you're currently doing there? 529 00:32:49,120 --> 00:32:49,720 For sure. Yeah. 530 00:32:49,720 --> 00:32:53,160 So I mean this was our vision from the start because when we 531 00:32:53,160 --> 00:32:56,400 started ourselves kind of diving into the blockchain and again 532 00:32:56,400 --> 00:32:58,600 this 2018, so things were different. 533 00:32:59,440 --> 00:33:06,080 You know you needed to install Mist for the and so the I mean 534 00:33:06,080 --> 00:33:12,000 the experience was pretty like painful and and it's also it was 535 00:33:12,000 --> 00:33:17,240 built on top of kind of a very different set of primitives I 536 00:33:17,240 --> 00:33:19,920 would say like conceptual primitives said than what 537 00:33:19,920 --> 00:33:23,080 normally people both users and developers expect right. 538 00:33:23,080 --> 00:33:27,600 So, you know, you need like to understand the Xerox wallets, 539 00:33:27,600 --> 00:33:30,840 you need a seat phrase, You need to like kind of pay gas. 540 00:33:30,840 --> 00:33:33,360 You need to like have, do all these things which are like 541 00:33:33,520 --> 00:33:37,960 strange when you, you know, when you're just starting. 542 00:33:38,760 --> 00:33:43,800 And what we've tried to do from the start is like how do we 543 00:33:43,800 --> 00:33:48,240 design kind of still like a blockchain that is secure that 544 00:33:48,240 --> 00:33:53,240 it has all all the same properties that we all want, but 545 00:33:53,240 --> 00:33:58,280 is able to kind of hide a lot of this complexity, ideally most of 546 00:33:58,280 --> 00:34:04,800 it, and make you know blockchain kind of abstracted out such that 547 00:34:04,800 --> 00:34:07,680 developers when they build applications can just build like 548 00:34:07,720 --> 00:34:10,480 as close to normal that to experience. 549 00:34:10,679 --> 00:34:15,000 But using the benefits of that three, using the kind of all of 550 00:34:15,000 --> 00:34:18,560 the value. And then also enabling users to 551 00:34:18,560 --> 00:34:22,320 have like more composibility, right, more ownership kind of 552 00:34:22,320 --> 00:34:25,480 being able to interact with multiple kind of applications 553 00:34:25,840 --> 00:34:29,199 and and have this like transportability of data. 554 00:34:29,920 --> 00:34:34,280 And so the near itself, right, kind of was designed was this. 555 00:34:34,280 --> 00:34:38,639 So we've like our accounts for example, you know the account 556 00:34:38,639 --> 00:34:41,800 obstruction part of the accounts have been designed from scrap 557 00:34:41,800 --> 00:34:44,520 from the start on the protocol level. 558 00:34:45,760 --> 00:34:49,639 There's like a bunch of kind of differences that was done 559 00:34:49,639 --> 00:34:53,080 including that accounts themselves are you know just a 560 00:34:53,080 --> 00:34:56,960 username that follows kind of domain name structure. 561 00:34:57,920 --> 00:35:00,400 We have lots of different keys with different permissions which 562 00:35:00,400 --> 00:35:02,840 allow us to have like multiple devices securely. 563 00:35:02,920 --> 00:35:07,880 It allows to delegate access, it allows to like applicate the 564 00:35:07,880 --> 00:35:10,800 front end of application to have a session key for example to 565 00:35:10,800 --> 00:35:15,120 transact for a specific set of kind of interaction. 566 00:35:15,120 --> 00:35:18,280 So kind of all of this functionality comes in by 567 00:35:18,280 --> 00:35:22,400 default. And then on the developer side, 568 00:35:22,480 --> 00:35:25,880 the choices we made around first of all choosing Web assembly 569 00:35:25,880 --> 00:35:30,720 which but at this point is like seems that everybody's kind of 570 00:35:30,800 --> 00:35:37,720 agrees on, but pretty much it's like it's an engine that runs in 571 00:35:37,720 --> 00:35:41,360 all the browsers. It's something that you know is 572 00:35:41,360 --> 00:35:44,680 like on billions of devices at this point it's supported by you 573 00:35:44,680 --> 00:35:48,600 know, large network of kind of developers. 574 00:35:48,600 --> 00:35:51,840 It runs on edge. It you know it supports lots of 575 00:35:51,840 --> 00:35:55,240 languages. You can run a lot of software in 576 00:35:55,240 --> 00:36:01,880 it and so so we kind of picked that and made it really easy to 577 00:36:01,880 --> 00:36:05,680 build like in a way from a developer perspective what when 578 00:36:05,680 --> 00:36:08,960 you write nearest my contract it's really just a service which 579 00:36:08,960 --> 00:36:13,840 has message like messages in and out and you have a kind of local 580 00:36:14,240 --> 00:36:18,360 key value database which is pretty much like the limits 581 00:36:18,360 --> 00:36:21,280 there are so big that like I don't think anybody ever hits 582 00:36:21,280 --> 00:36:24,320 them like you can't you. I think we have contract that 583 00:36:24,320 --> 00:36:28,800 have like 4 gigabytes of storage in their in their database 584 00:36:28,800 --> 00:36:32,400 right. So you can build like massive 585 00:36:32,520 --> 00:36:35,400 massive contracts specifically you can build all the chains as 586 00:36:35,400 --> 00:36:37,960 a smart contract on there. So we have Aurora which is an 587 00:36:37,960 --> 00:36:42,880 EVM as a smart contract just like taking you know the EVM 588 00:36:42,880 --> 00:36:45,480 that's usually run people a separate chain just put in smart 589 00:36:45,480 --> 00:36:50,680 contract their their database is where all the state of the of 590 00:36:50,680 --> 00:36:53,200 this storage right. You can do the same as Bitcoin. 591 00:36:53,200 --> 00:36:55,840 I've I've been suggesting somebody to like fork Bitcoin 592 00:36:55,840 --> 00:36:57,800 and put it on there make it ultrasound money. 593 00:37:00,040 --> 00:37:02,080 We have JavaScript running as well. 594 00:37:02,400 --> 00:37:05,160 So you can run JavaScript smart contracts you can potentially do 595 00:37:05,160 --> 00:37:08,240 Python And other stuff. So like it's kind of enables 596 00:37:08,560 --> 00:37:10,480 developer experience across the board. 597 00:37:10,840 --> 00:37:14,400 And since then we kind of following the same principle is 598 00:37:14,400 --> 00:37:17,400 like, OK, well, now that you can build anything on smart contract 599 00:37:17,400 --> 00:37:21,600 side, what's the next part? Well, actually you want to get 600 00:37:21,600 --> 00:37:26,000 the data out of this out of the blockchain and blockchain are 601 00:37:26,000 --> 00:37:30,240 not optimized for reading data. They are all kind of we've tend 602 00:37:30,240 --> 00:37:32,880 to optimize them for writing and kind of maintaining security. 603 00:37:33,280 --> 00:37:36,160 And so for reading data you want a completely different data 604 00:37:36,160 --> 00:37:39,160 structure and so that hence there's like this principle of 605 00:37:39,160 --> 00:37:43,640 indexing and kind of in the way of chain computation. 606 00:37:44,120 --> 00:37:47,440 And so we've been building indexing framework and that 607 00:37:47,440 --> 00:37:50,400 actually culminated in what we call query API which is a 608 00:37:50,400 --> 00:37:54,200 service that indexes that you can like write a smart contract 609 00:37:54,200 --> 00:37:57,760 that describes the indexing of data that executes off chain. 610 00:37:57,760 --> 00:38:01,880 So in a way it's like a off chain computation framework that 611 00:38:01,880 --> 00:38:05,960 allows you to store output of that computation in kind of SQL 612 00:38:05,960 --> 00:38:11,480 databases that then you can query and finally, well OK now 613 00:38:11,480 --> 00:38:14,160 you have back end and and middleware, now you need a front 614 00:38:14,160 --> 00:38:16,760 end, right. And again it seems weird that we 615 00:38:16,760 --> 00:38:20,400 are like oh you build everything decentralized but now run a 616 00:38:20,400 --> 00:38:24,240 server on a specific domain that you will need to maintain. 617 00:38:24,480 --> 00:38:28,360 It's like OK well that kind of violates the whole part point of 618 00:38:28,360 --> 00:38:33,240 what we're doing. So, so we created this kind of 619 00:38:33,240 --> 00:38:36,200 decentralized front end framework that allows to store 620 00:38:36,200 --> 00:38:38,360 the front end code itself on chain. 621 00:38:38,920 --> 00:38:42,080 So again the smart contracts code on chain, the middleware 622 00:38:42,080 --> 00:38:46,640 code on chain, the front end code on chain and now anyone, 623 00:38:46,760 --> 00:38:50,000 any kind of we call them gateways can render this code on 624 00:38:50,000 --> 00:38:52,680 a user side, right? So we have a desktop app, we can 625 00:38:52,680 --> 00:38:56,440 have a mobile app and we have obviously web apps that can bode 626 00:38:56,440 --> 00:38:59,320 that from the blockchain directly into your browser and 627 00:38:59,320 --> 00:39:02,000 render it there. So there's no kind of middle 628 00:39:02,000 --> 00:39:07,200 server that's needed to render you, you don't need to have a 629 00:39:07,200 --> 00:39:09,840 domain. You can obviously if you want to 630 00:39:09,880 --> 00:39:13,440 and so you can just you know launch kind of part of your web 631 00:39:13,440 --> 00:39:17,600 app as as this decentralized front end component and now it 632 00:39:17,600 --> 00:39:20,240 will live forever on blockchain, right? 633 00:39:20,240 --> 00:39:23,320 But side by side with your smart contracts, have the same 634 00:39:23,320 --> 00:39:26,720 operability, have security, cryptographic security, who has 635 00:39:26,760 --> 00:39:29,800 it, have versioning. So if I as a user don't hate a 636 00:39:29,800 --> 00:39:33,840 new version I can go to version before and so like all the same 637 00:39:33,840 --> 00:39:37,120 properties we really like about smart contracts we now get for 638 00:39:37,120 --> 00:39:39,720 front ends. So, so all of that is really 639 00:39:39,720 --> 00:39:42,800 enables like a full stack decentralized development that 640 00:39:42,800 --> 00:39:45,880 is you know familiar with to normal developers. 641 00:39:45,880 --> 00:39:48,360 It's React JavaScript components, it's JavaScript for 642 00:39:48,360 --> 00:39:52,480 middleware indexing it's JavaScript Ross C++ and like 643 00:39:52,480 --> 00:39:54,040 other languages for smart contracts. 644 00:39:54,240 --> 00:39:57,240 So you have like a full stack decentralization that you can 645 00:39:57,240 --> 00:40:00,520 have. And interestingly as we were 646 00:40:00,520 --> 00:40:02,640 building the front ends, we realized actually the front ends 647 00:40:02,640 --> 00:40:07,680 can work with any blockchain. And so we kind of just turned on 648 00:40:07,680 --> 00:40:11,880 all the Avms and some other blockchains and people started 649 00:40:11,880 --> 00:40:14,480 building all the AVM fronts as well. 650 00:40:14,480 --> 00:40:18,240 So we have a UNISWAP for example, for linear, the kind of 651 00:40:18,240 --> 00:40:21,920 official UNISWAP front end is served out of the decentralized 652 00:40:21,920 --> 00:40:23,720 front end, right? Because by the way, it also 653 00:40:23,720 --> 00:40:26,160 doesn't charge extra fees and we have like partnerships with 654 00:40:26,160 --> 00:40:28,080 others, the KVM, Mantle, etcetera. 655 00:40:28,520 --> 00:40:32,320 And so the idea is like actually as you start looking from that 656 00:40:32,320 --> 00:40:36,040 lens, from a user lens, right. As a user, I don't really care 657 00:40:36,040 --> 00:40:38,600 which blockchain the apps is on, I just want to use them. 658 00:40:38,880 --> 00:40:43,560 And like if you go to you know some like you know some of this 659 00:40:43,880 --> 00:40:48,120 gateways where you can access this front ends, you can just go 660 00:40:48,120 --> 00:40:50,880 and search for whatever app you want, click on it and start 661 00:40:50,880 --> 00:40:52,600 using it that that's how it should be. 662 00:40:52,960 --> 00:40:57,840 And so, and this is kind of where we get to this concept I 663 00:40:57,840 --> 00:40:59,880 started with which is like, hey, we want to abstract the 664 00:40:59,880 --> 00:41:01,840 blockchain for users and developers. 665 00:41:02,360 --> 00:41:05,080 We're getting back to it with kind of this now that we have 666 00:41:05,080 --> 00:41:09,320 this full stack decentralization where like actually this works 667 00:41:09,320 --> 00:41:11,920 for all blockchains, for all chains, for roll UPS, for 668 00:41:11,920 --> 00:41:15,920 whatever, because you can actually abstract out all that 669 00:41:15,920 --> 00:41:18,960 on the front end side and make it really easy for people to 670 00:41:18,960 --> 00:41:21,920 interact with it. And so hence we kind of started 671 00:41:21,920 --> 00:41:25,280 going backwards now with some of the other launches we had, 672 00:41:25,280 --> 00:41:30,360 right, allowing pretty much as a kind of how do we make it really 673 00:41:30,360 --> 00:41:35,440 easy now for one experience to to unite all of the block chains 674 00:41:35,480 --> 00:41:37,760 and kind of recall the chain abstraction principle. 675 00:41:38,560 --> 00:41:42,280 And so this goes into like GA and and some other things we, 676 00:41:42,440 --> 00:41:47,520 yeah, we can discuss. So, so India is it, is it 677 00:41:47,520 --> 00:41:51,800 correct to imagine so when you talk of like this indexes 678 00:41:51,800 --> 00:41:56,440 service or the service of hosting a front end, is it 679 00:41:56,440 --> 00:42:01,440 correct to imagine it as the indexing logic or the front end 680 00:42:01,440 --> 00:42:05,000 logic is stored on the chain. But then there is some kind of 681 00:42:05,000 --> 00:42:09,960 off chain actor that is actually taking that, taking that logic 682 00:42:10,080 --> 00:42:13,920 and the data and actually serving it much as a traditional 683 00:42:13,920 --> 00:42:17,760 server. And somehow the chain is 684 00:42:17,760 --> 00:42:23,640 guaranteeing that this server's work is correct and it is 685 00:42:23,640 --> 00:42:24,720 compensated. Is it? 686 00:42:24,720 --> 00:42:26,120 Is it correct to imagine it like that? 687 00:42:27,600 --> 00:42:29,800 Yeah, pretty much. So the idea is, I mean similar 688 00:42:29,800 --> 00:42:33,520 to maybe blotching validator nodes as well, right? 689 00:42:33,520 --> 00:42:39,000 There's a kind of a logic that is conceptual and then all the 690 00:42:39,000 --> 00:42:43,280 validators are doing that job and like you can always have, 691 00:42:43,280 --> 00:42:44,880 you know, more validators, less validators. 692 00:42:44,880 --> 00:42:47,360 It's kind of independent of that. 693 00:42:47,560 --> 00:42:52,080 Similarly, yes, the indexing logic and the front end kind of 694 00:42:52,440 --> 00:42:57,760 source code itself is stored on chain and so any server can run 695 00:42:57,760 --> 00:43:02,920 on and kind of create the same, you know outcome from this, 696 00:43:02,920 --> 00:43:04,360 right? Again similar to RPCS for 697 00:43:04,360 --> 00:43:09,880 example, RPC server right is serving your data but it's you 698 00:43:09,880 --> 00:43:12,640 know anybody can run RPC server and get the same results. 699 00:43:12,640 --> 00:43:14,240 So like it's it's part of protocol. 700 00:43:14,240 --> 00:43:17,720 In a way it becomes part of protocol and so similar thing 701 00:43:17,720 --> 00:43:20,800 we're trying to do for about both front ends and middleware 702 00:43:20,800 --> 00:43:25,440 indexing as well. So maybe one way to think about 703 00:43:25,440 --> 00:43:28,320 this is that, so on on on the TDM, right? 704 00:43:28,320 --> 00:43:32,520 If you look at the TDM, there's a basically a blockchain and 705 00:43:32,520 --> 00:43:35,880 then there are separate protocols like the ENS for 706 00:43:37,640 --> 00:43:40,680 naming your blockchain address to a human readable name. 707 00:43:40,840 --> 00:43:44,720 That is the graph which kind of like indexes a smart contract 708 00:43:44,720 --> 00:43:49,200 and kind of presents historical data about the transactions and 709 00:43:49,200 --> 00:43:53,920 events in the smart contract. And maybe maybe there are other 710 00:43:53,920 --> 00:43:56,840 examples that I'm missing. So in the TDM layer, these are 711 00:43:56,840 --> 00:43:59,360 like different systems and usually they are competing 712 00:43:59,360 --> 00:44:03,200 systems as ENS, but they won't be a competitor to ENS the graph 713 00:44:03,200 --> 00:44:04,800 and they won't be a competitor to the graph. 714 00:44:05,640 --> 00:44:11,280 But in near NIA has kind of taken the philosophy that some 715 00:44:11,280 --> 00:44:16,880 of some of these things are like are like really key to the UX of 716 00:44:16,880 --> 00:44:19,520 a blockchain and therefore they should be supported 717 00:44:19,520 --> 00:44:22,760 out-of-the-box by the layer one itself. 718 00:44:23,480 --> 00:44:25,960 Is that is that the philosophy here? 719 00:44:26,920 --> 00:44:29,240 To extend, yeah, I think, I think the way to think about it 720 00:44:29,240 --> 00:44:32,760 is it's more than just layer one, right. 721 00:44:32,760 --> 00:44:36,480 Like at the end when we are interacting with applications on 722 00:44:36,480 --> 00:44:41,160 any of these chains, like there's a whole host of tools 723 00:44:41,160 --> 00:44:44,080 and and more importantly standards that we are 724 00:44:44,080 --> 00:44:47,480 interacting with, right. And so like ERC 20 for example 725 00:44:47,480 --> 00:44:51,280 is a standard and it's a standard that kind of came out 726 00:44:51,280 --> 00:44:54,640 of the application space. But it's now like you cannot 727 00:44:54,640 --> 00:44:58,480 imagine the CDO without the ERC 20 standard and So what we're 728 00:44:58,480 --> 00:45:03,240 doing here is really defining standards for this key 729 00:45:03,240 --> 00:45:06,400 primitives that are just going beyond just you know token 730 00:45:06,400 --> 00:45:10,480 transfers but going to like how to how to define indexing, how 731 00:45:10,480 --> 00:45:14,200 to define this interest run it's now implementation of those 732 00:45:14,200 --> 00:45:16,520 things can like you can have many implementations, you can 733 00:45:16,520 --> 00:45:18,800 have you know mobile render and web render. 734 00:45:18,800 --> 00:45:21,960 You can have indexing like you can have you know external 735 00:45:21,960 --> 00:45:24,880 partners who are competing with each other, how to implement it? 736 00:45:25,800 --> 00:45:29,680 Same same for RPCS, right? RPC is a standard but then the 737 00:45:29,680 --> 00:45:32,480 way it's implemented right can be very different like 738 00:45:32,480 --> 00:45:34,880 underneath you may be like cached everything in database 739 00:45:34,880 --> 00:45:37,480 and maybe using cloud flare like whatever the architecture you 740 00:45:37,480 --> 00:45:40,000 want to use. But the standard is there and I 741 00:45:40,000 --> 00:45:43,280 think what we've been trying to do is define a standard and I 742 00:45:43,280 --> 00:45:46,560 mean have a reference implementation but for for this 743 00:45:46,560 --> 00:45:50,280 more key pieces to make indeed the experience more aligned and 744 00:45:50,280 --> 00:45:55,120 kind of have this like singular journey for developers and users 745 00:45:55,120 --> 00:45:58,880 that you know is cohesive. And yeah, like the way you know 746 00:45:58,880 --> 00:46:01,640 some of the things like you can have businesses around the 747 00:46:01,640 --> 00:46:03,960 standards that are, you know, very profitable. 748 00:46:05,320 --> 00:46:09,880 But like the core principle for me of decentralization is 749 00:46:09,880 --> 00:46:13,240 actually in the standard. It's the fact that like if you 750 00:46:13,240 --> 00:46:16,680 define a standard, it means that you can swap in and swap out any 751 00:46:16,680 --> 00:46:20,160 participant. And so you're not you don't have 752 00:46:20,160 --> 00:46:22,320 this like lock in effect. You don't have the effect of you 753 00:46:22,320 --> 00:46:24,480 know you go to a bank and you cannot move your money out 754 00:46:24,480 --> 00:46:27,760 because it doesn't allow you to. Or you know the cannot cancel 755 00:46:27,760 --> 00:46:30,640 your telco provider because like or like your positive telco 756 00:46:30,640 --> 00:46:32,320 providers don't even work for you. 757 00:46:32,960 --> 00:46:37,560 The here we can always have like a competitor that comes in and 758 00:46:37,560 --> 00:46:40,440 if they're more effective and can provide better prices people 759 00:46:40,440 --> 00:46:42,240 can switch to it. But the stand because the 760 00:46:42,240 --> 00:46:44,640 standards is the same. And so for me that's kind of the 761 00:46:44,640 --> 00:46:48,240 key principle of like Web 3IN general. 762 00:46:48,760 --> 00:46:52,520 And so I think the challenge that I've seen is like not 763 00:46:52,520 --> 00:46:55,720 having the standards actually leads to kind of huge 764 00:46:55,720 --> 00:46:59,720 fragmentation of experiences. And as well, like actually 765 00:46:59,720 --> 00:47:03,320 Monopoly has been built because like now that you've built all 766 00:47:03,320 --> 00:47:06,920 your software towards some API, you cannot switch because nobody 767 00:47:06,920 --> 00:47:09,080 else provides us. And like you need to rewrite 768 00:47:09,080 --> 00:47:14,760 half your code to do that. So how much is kind of analogy 769 00:47:14,760 --> 00:47:18,640 with the Apple ecosystem versus the Microsoft ecosystem for 770 00:47:18,640 --> 00:47:20,880 desktop, how much, how well does it map? 771 00:47:20,880 --> 00:47:26,360 So in a sense when you look at kind of like the the Apple, the 772 00:47:26,360 --> 00:47:30,520 Apple ecosystem, it's a company that has kind of maintained 773 00:47:30,520 --> 00:47:34,160 control over, it's kind of like operating system, supply chain, 774 00:47:34,480 --> 00:47:37,880 it's way of like delivering music, it's way of delivering 775 00:47:37,880 --> 00:47:44,200 books, it's way of how kind of applications kind of like appear 776 00:47:44,200 --> 00:47:48,600 to the end user. And in the in the beginning, I 777 00:47:48,600 --> 00:47:51,040 think they also wanted control over the hardware, but maybe 778 00:47:51,040 --> 00:47:54,440 they have retracted on that now. Whereas like kind of like 779 00:47:54,440 --> 00:47:59,200 Microsoft is 1 where just like the raw operating system and 780 00:47:59,200 --> 00:48:01,640 then applications emerge. And if there are standards 781 00:48:01,640 --> 00:48:03,920 needed for their interoperability, the market 782 00:48:03,920 --> 00:48:08,520 kind of like figures it out. And from the outside it feels 783 00:48:08,520 --> 00:48:11,440 like OK. Near is kind of like more going 784 00:48:11,440 --> 00:48:14,120 towards that Apple philosophy that we are going to define all 785 00:48:14,120 --> 00:48:18,440 of the standards for, for many of the things that are key 786 00:48:18,440 --> 00:48:20,480 determinants of the user experience. 787 00:48:21,840 --> 00:48:26,400 Whereas other ecosystems like Cosmos or Ethereum might be more 788 00:48:26,400 --> 00:48:29,520 kind of like the Microsoft approach where we are providing 789 00:48:29,520 --> 00:48:32,920 like transaction throughput as the center, the account model as 790 00:48:32,920 --> 00:48:35,000 the center. And then kind of a lot of the 791 00:48:35,000 --> 00:48:38,480 interoperability between the standards is left to the market 792 00:48:38,480 --> 00:48:44,000 to figure out how much does that analogy map and how much does 793 00:48:44,000 --> 00:48:48,240 doesn't map. I mean the, I would say the. 794 00:48:48,240 --> 00:48:52,240 Part that's like agree on is we're definitely trying to focus 795 00:48:52,240 --> 00:48:55,680 on user experience, right. And so with that, it's important 796 00:48:55,680 --> 00:48:58,880 to figure out like what are the touch points that you want to 797 00:48:58,880 --> 00:49:01,520 have standards on. Again like my perspective is for 798 00:49:01,520 --> 00:49:04,120 example RPC, Jason. RPC is part of Ethedium 799 00:49:04,120 --> 00:49:08,880 standard, like it's part of kind of the protocol even though it's 800 00:49:08,880 --> 00:49:12,200 actually not but like it it for both accounts. 801 00:49:12,200 --> 00:49:16,240 If you try to change it RPCAPI like you will break everyone And 802 00:49:16,240 --> 00:49:19,800 so, so we kind of see in a similar way, right, like if RPC 803 00:49:19,800 --> 00:49:22,120 is part of the standard, why not some other parts. 804 00:49:22,720 --> 00:49:27,640 But as I said like you know Near for example has like number of 805 00:49:27,640 --> 00:49:29,680 contributors that they're building things like the, the 806 00:49:29,680 --> 00:49:33,360 actually the VM that's built right now for the decentralized 807 00:49:33,360 --> 00:49:35,280 finance is built by proximity, right. 808 00:49:36,840 --> 00:49:41,920 And for example you know the query API kind of other 809 00:49:41,920 --> 00:49:45,120 companies can implement the same standard and provide kind of 810 00:49:45,120 --> 00:49:48,480 better services. So I think the idea here is that 811 00:49:48,640 --> 00:49:51,880 by defining the standard we kind of actually opening up the 812 00:49:51,880 --> 00:49:56,640 market for people to fill in like with better products in 813 00:49:56,640 --> 00:49:57,960 this. And again like it's pretty early 814 00:49:57,960 --> 00:49:59,920 still. So like a lot of stuff we still 815 00:49:59,920 --> 00:50:02,720 build like reference implementations, but similar as 816 00:50:02,720 --> 00:50:06,840 a CDM by defining a standard for protocol, it opened up a place 817 00:50:06,840 --> 00:50:09,480 for all of these clients to be implemented, right. 818 00:50:09,640 --> 00:50:12,840 Like that's kind of the idea is like you you, you define a 819 00:50:12,840 --> 00:50:17,200 standard and then you open it up so that others can contribute to 820 00:50:17,200 --> 00:50:19,720 it. In the same way versus competing 821 00:50:19,720 --> 00:50:24,360 on AP is and competing on like kind of you know in a way 822 00:50:24,360 --> 00:50:29,880 marketing what's happening right now in in in like talking price, 823 00:50:30,320 --> 00:50:32,840 what's happening in the CDM for some of this like infrastructure 824 00:50:32,840 --> 00:50:36,440 tooling, right. It's like can can we get a 825 00:50:36,440 --> 00:50:39,600 bigger AirDrop by using a product versus like hey, this is 826 00:50:39,600 --> 00:50:43,120 a standard, everybody will be using this standard and so now 827 00:50:43,280 --> 00:50:46,360 what's the best product people can build for the standard. 828 00:50:46,520 --> 00:50:48,640 So I think like that's kind of the difference. 829 00:50:48,640 --> 00:50:52,320 I don't think it's as applicable to like this, you know big 830 00:50:52,320 --> 00:50:56,840 commercial for profit companies versus like this is an ecosystem 831 00:50:57,200 --> 00:51:02,840 that we're building and really defining more kind of this I 832 00:51:02,880 --> 00:51:06,760 would say like layers of the stack. 833 00:51:09,240 --> 00:51:12,480 Going back a bit to like this, what you said about the 834 00:51:12,480 --> 00:51:15,360 switching cost from your telephobic, I guess like related 835 00:51:15,360 --> 00:51:17,120 here. I guess one big thing in 836 00:51:17,120 --> 00:51:20,440 blockchain is generally like bridging or like if you want to 837 00:51:20,440 --> 00:51:24,480 switch the ecosystem you have to like go to some other chain and 838 00:51:24,600 --> 00:51:26,560 then like move the liquidity there which can be like 839 00:51:26,560 --> 00:51:30,080 cumbersome. And you did mention the chain 840 00:51:30,120 --> 00:51:32,960 abstraction for a second there and I I saw on your Twitter a 841 00:51:32,960 --> 00:51:36,680 bunch also like this concept you have like of account aggregation 842 00:51:37,200 --> 00:51:39,400 that you teased. So maybe you can yeah, can you 843 00:51:39,400 --> 00:51:43,520 talk, tell us a bit about like what what are you doing there or 844 00:51:43,520 --> 00:51:46,120 how are you like sort of solving this interoperability problem in 845 00:51:46,200 --> 00:51:49,000 in the blockchain space? For sure, yeah. 846 00:51:49,000 --> 00:51:50,880 So that that's a very important topic. 847 00:51:51,240 --> 00:51:55,280 So although we have started building bridges I think so the 848 00:51:55,320 --> 00:51:58,000 our Rainbow Bridge been built from 2019. 849 00:51:58,000 --> 00:52:01,360 So I think we started building you know kind of in line with 850 00:52:01,360 --> 00:52:09,080 IBC kind of timing and we've been, we've been running since I 851 00:52:09,080 --> 00:52:12,920 guess in the beginning of 21. And at the same time like 852 00:52:12,920 --> 00:52:17,480 bridges are really bad as a concept because they created 853 00:52:17,480 --> 00:52:22,120 honeypot for security. They are the place to siphon off 854 00:52:22,120 --> 00:52:26,440 assets and like there's you know if there's any attack on the 855 00:52:26,440 --> 00:52:32,880 protocol itself bridges kind of how you exit and and they like 856 00:52:32,880 --> 00:52:36,040 that just the amount of failing modes between different block 857 00:52:36,040 --> 00:52:38,160 chains is pretty like it's pretty big right. 858 00:52:38,160 --> 00:52:41,160 But between multiple block chains it's like insane like you 859 00:52:41,160 --> 00:52:44,480 know chain stop blocks didn't publish like all those things 860 00:52:44,480 --> 00:52:47,600 you need to like as a developer now you need to handle and then 861 00:52:47,600 --> 00:52:51,880 on application side, OK yes you know the fungible tokens 862 00:52:51,880 --> 00:52:55,400 transfers is maybe reasonable but as soon as you add any 863 00:52:55,400 --> 00:52:59,400 logic, right be that you know rebasing or be it LFT. 864 00:52:59,680 --> 00:53:02,360 Now when you bridge it you lose all of the logic on the other 865 00:53:02,360 --> 00:53:07,480 side and so so the concept I've been kind of exploring for a 866 00:53:07,480 --> 00:53:09,920 while now. I was calling it originally 867 00:53:10,640 --> 00:53:14,920 remote accounts, but we we kind of reframe it as account 868 00:53:14,920 --> 00:53:17,400 aggregation. This idea that ideally you're 869 00:53:17,480 --> 00:53:22,240 going to have one account and there's mapped accounts to this 870 00:53:22,720 --> 00:53:25,400 on other chase. So imagine you know you have my 871 00:53:25,400 --> 00:53:29,400 root dot near account on near and then I have an address on 872 00:53:29,400 --> 00:53:31,320 this idiom, I have an address on Bitcoin. 873 00:53:31,320 --> 00:53:34,160 I have an address on Solana which I control with this 874 00:53:34,160 --> 00:53:37,680 account. And so now if I want to buy a 875 00:53:37,680 --> 00:53:41,000 Solana NFT, right, right now I would need to like set up a new 876 00:53:41,000 --> 00:53:45,240 wallet, you know bridge some stuff to Solana by the NFT. 877 00:53:45,560 --> 00:53:49,520 And then I like, I don't know and then like go and look at it, 878 00:53:49,520 --> 00:53:52,040 you know from time to time because I'm mostly sitting on 879 00:53:52,040 --> 00:53:56,240 here or you can you have the Solana address that's linked to 880 00:53:56,240 --> 00:54:01,040 your to your new account. You pretty much through this by 881 00:54:01,040 --> 00:54:04,760 an NFT for this address and we can talk how that works. 882 00:54:05,080 --> 00:54:08,600 And now you have the front end that actually shows you 883 00:54:09,240 --> 00:54:11,840 everything you own across all of this chains from all of this 884 00:54:11,840 --> 00:54:14,480 addresses. And the way mentally to think 885 00:54:14,480 --> 00:54:18,120 about it is when you go to Binance or Coinbase and you sign 886 00:54:18,120 --> 00:54:21,400 up with your Binance or Coinbase accounts, you have addresses on 887 00:54:21,400 --> 00:54:23,960 all chains, right. And I mean they they are deposit 888 00:54:23,960 --> 00:54:26,440 addresses usually. But imagine those addresses were 889 00:54:26,440 --> 00:54:29,800 actually normal addresses. You can use apps and buy NFTS 890 00:54:29,800 --> 00:54:34,400 and tokens etcetera Wiz so that and but your account is your you 891 00:54:34,400 --> 00:54:37,240 know, Coinbase account. And so that's kind of where 892 00:54:37,240 --> 00:54:39,600 you're, you know, like ownership is. 893 00:54:40,560 --> 00:54:43,320 And so that's what we're trying like we building, we're going to 894 00:54:43,320 --> 00:54:47,160 be launching end of the first quarter is this concept of 895 00:54:47,160 --> 00:54:50,320 account aggregation that now allows together with 896 00:54:50,320 --> 00:54:54,360 decentralized front ends allows to actually collapse the kind of 897 00:54:54,360 --> 00:54:57,800 this whole, you know, multiple chains switching networks, you 898 00:54:57,800 --> 00:55:01,520 know, bridging all of this into a very simple experience of you 899 00:55:01,520 --> 00:55:04,760 get an account, you know you deposit some funds into it and 900 00:55:05,000 --> 00:55:08,720 now you can transact across all block chains, across all of 901 00:55:08,720 --> 00:55:12,320 their apps. And it kind of will get executed 902 00:55:12,720 --> 00:55:15,360 on on your behalf on those chains. 903 00:55:15,360 --> 00:55:19,240 And you have this addresses but it's all self custodial and all 904 00:55:19,240 --> 00:55:21,480 kind of hidden from you. You don't need to think about 905 00:55:21,480 --> 00:55:23,200 gas fees on those chains etcetera. 906 00:55:23,560 --> 00:55:25,680 So that's kind of the experience we're going after. 907 00:55:26,000 --> 00:55:28,480 And again, this is just an extension of what we've been 908 00:55:28,520 --> 00:55:31,600 building was near by trying to abstract out the near 909 00:55:31,600 --> 00:55:34,200 blockchain, we're just like OK, well we can actually do that. 910 00:55:34,200 --> 00:55:38,640 It's the same thing for everyone and really provide like a unique 911 00:55:38,880 --> 00:55:42,200 and valuable experience because you know anything multi chain 912 00:55:42,200 --> 00:55:45,160 you want to build near will be actually the place to build it 913 00:55:45,320 --> 00:55:47,960 because you will be able to transact across all of the 914 00:55:47,960 --> 00:55:51,480 chains without having to bridge, without having this like 915 00:55:51,480 --> 00:55:54,440 complexity. You want to build for example 916 00:55:54,440 --> 00:55:58,240 Bitcoin D5 well on near, you know every near account or smart 917 00:55:58,240 --> 00:56:01,040 contract will have a Bitcoin address you can deposit to it 918 00:56:01,040 --> 00:56:03,160 can start you know doing stuff right. 919 00:56:03,160 --> 00:56:06,400 And so that's kind of conceptually what we really 920 00:56:06,400 --> 00:56:10,240 bringing to market with this and like kind of finishing our I 921 00:56:10,240 --> 00:56:13,120 would say arc of chain abstraction that we started with 922 00:56:13,160 --> 00:56:18,880 doing near in the first place. So on a on a high level India, 923 00:56:18,960 --> 00:56:23,640 I'd like this idea is is in the Cosmos ecosystem. 924 00:56:23,640 --> 00:56:28,120 There's a chain called Neutron and because the Cosmos ecosystem 925 00:56:28,120 --> 00:56:33,040 has IBC, so Cosmos chains can bridge to each other in quite a 926 00:56:33,040 --> 00:56:36,120 good way. Neutron has the idea that in 927 00:56:36,120 --> 00:56:38,680 Cosmos you have the idea of delegated account control, which 928 00:56:38,680 --> 00:56:42,520 is like on one chain you have an address and that can control 929 00:56:42,520 --> 00:56:45,480 many other Puppet addresses on other chains and neutrons. 930 00:56:45,480 --> 00:56:48,480 Trying to build that kind of that that Puppet master chain 931 00:56:48,480 --> 00:56:50,800 where like you would have your central account and you will 932 00:56:50,800 --> 00:56:54,680 control other addresses on a lot of chains over IBC through 933 00:56:54,680 --> 00:56:56,400 Neutron. Does it feel similar? 934 00:56:57,560 --> 00:57:02,400 But the reason it works in the Cosmos ecosystem is you assume 935 00:57:02,400 --> 00:57:05,880 my BC that there's a secure bridging solution underneath 936 00:57:05,880 --> 00:57:09,840 available. For this to work in Neutron my I 937 00:57:09,840 --> 00:57:13,080 almost start to saying that OK, the only way like this can work 938 00:57:13,080 --> 00:57:15,800 for Near and Solana. For example, having like a 939 00:57:16,440 --> 00:57:20,680 address on Near that in control of puppet address on Solana, you 940 00:57:20,680 --> 00:57:24,000 need a secure bridge between Near and Solana, is it not so 941 00:57:24,000 --> 00:57:27,080 the bridging problem? Solving the bridging problem 942 00:57:27,080 --> 00:57:28,840 seems like a prerequisite to this. 943 00:57:29,760 --> 00:57:32,640 Yeah, so we're trying to go away from bridging almost completely. 944 00:57:32,640 --> 00:57:34,920 I mean there will be some places where you still need bridges, 945 00:57:34,920 --> 00:57:40,120 but so let let's look at Bitcoin as just to get way more clean. 946 00:57:40,120 --> 00:57:42,400 Example, right. With Bitcoin you cannot have a 947 00:57:42,400 --> 00:57:45,400 smart contract bridge because, well, Bitcoin still have smart 948 00:57:45,400 --> 00:57:49,440 contracts and so the only thing you can do is to own addresses. 949 00:57:49,440 --> 00:57:53,840 And so the core idea here and it's conceptually the same as 950 00:57:53,840 --> 00:57:56,320 yeah, what Neutron is doing, but the core idea is different. 951 00:57:56,320 --> 00:58:01,360 The core idea is that we make near network itself be able to 952 00:58:01,360 --> 00:58:05,080 sign transactions for other block chains And so near network 953 00:58:05,080 --> 00:58:10,160 becomes in the way custodian of all of this address mapped 954 00:58:10,160 --> 00:58:14,960 addresses on other chains. And you as a near user telling 955 00:58:15,520 --> 00:58:18,640 network right be that through smart contract or or or user 956 00:58:18,680 --> 00:58:22,400 user interaction to sign a transaction on Bitcoin to send 957 00:58:22,400 --> 00:58:25,880 some bitcoins from your remote address from your delegated 958 00:58:25,880 --> 00:58:28,160 address to some other address right. 959 00:58:28,440 --> 00:58:32,160 And so because of this like you don't need to actually bridge 960 00:58:32,160 --> 00:58:34,120 Bitcoin to near to do anything right? 961 00:58:34,120 --> 00:58:37,360 You you just literally the bitcoins live on Bitcoin 962 00:58:37,360 --> 00:58:41,080 network. The you know OP coins live on 963 00:58:41,080 --> 00:58:46,000 optimism the you know Salon and NF TS will live on Salon, NFT on 964 00:58:46,000 --> 00:58:50,280 Solana and I just control all that by just sending 965 00:58:50,280 --> 00:58:53,360 transactions there. But as a user like I just 966 00:58:53,640 --> 00:58:57,160 interact with near and I kind of pay near guest fee which is very 967 00:58:57,160 --> 00:58:59,640 small. I say like do this, you know I 968 00:58:59,640 --> 00:59:03,080 attach whatever also you know if I need to buy something etcetera 969 00:59:03,200 --> 00:59:06,440 on near and then we have kind of intent relayers that actually 970 00:59:06,440 --> 00:59:09,520 execute stuff like you know the transaction gets signed by near 971 00:59:09,520 --> 00:59:12,000 network and then intense relay of you know sends that 972 00:59:12,000 --> 00:59:14,480 transaction on your behalf on the other. 973 00:59:14,640 --> 00:59:16,600 And so there's no actual like bridging. 974 00:59:16,600 --> 00:59:22,240 There's no kind of security kind of issue where like if this 975 00:59:22,240 --> 00:59:24,960 bridge gets broken or whatever or that network gets forked 976 00:59:24,960 --> 00:59:27,040 etcetera, like none of that exists. 977 00:59:27,480 --> 00:59:31,080 And because near account there's also like a very interesting and 978 00:59:31,080 --> 00:59:34,360 kind of a little bit crazy thing because near accounts are are 979 00:59:34,360 --> 00:59:37,560 actually tradable. So you can actually list near 980 00:59:37,560 --> 00:59:40,960 account as NFT and somebody can buy it and get access to it 981 00:59:40,960 --> 00:59:42,720 because you can rotate keys on there. 982 00:59:43,360 --> 00:59:46,840 What is allowed to do is you can have lots of assets across all 983 00:59:46,840 --> 00:59:51,080 kinds of networks and then you can list that as a bundle on 984 00:59:51,080 --> 00:59:56,280 near as like you want to sell some BRC 20s, some so on NFTS, 985 00:59:56,280 --> 01:00:00,920 some Ethenium NFTS and some and OP coins and GMX at the same 986 01:00:00,920 --> 01:00:03,080 time. You can list all that as a 987 01:00:03,080 --> 01:00:06,480 bundle under one near account and then somebody can buy all 988 01:00:06,480 --> 01:00:10,920 that with one transaction on near paying near transaction fee 989 01:00:11,160 --> 01:00:13,760 and within one second block time. 990 01:00:14,040 --> 01:00:16,760 So you don't need to wait for Bitcoin transfer, You don't need 991 01:00:16,760 --> 01:00:20,040 to wait for all of this. You can do it on what so you can 992 01:00:20,040 --> 01:00:23,040 actually start. Bundling all of these things and 993 01:00:23,040 --> 01:00:26,640 trading kind of across all chains on near very easily 994 01:00:26,640 --> 01:00:28,760 without actually sending transactions of bridging 995 01:00:28,760 --> 01:00:32,120 anything anywhere else. And that's kind of the the shift 996 01:00:32,120 --> 01:00:34,400 that we're trying to do. I called on bridging that we 997 01:00:34,400 --> 01:00:37,160 like. You have the account level kind 998 01:00:37,160 --> 01:00:40,600 of ownership that's maintained indeed, but it's maintained by 999 01:00:40,600 --> 01:00:43,600 very specific security parameters that are near 1000 01:00:43,600 --> 01:00:47,240 parameters. And then if if the, let's say 1001 01:00:47,240 --> 01:00:49,880 Solana network fails for whatever reason, there's no 1002 01:00:49,880 --> 01:00:54,480 bridge problems, right, That would you know, rise from from 1003 01:00:54,480 --> 01:00:57,520 this like because you own stuff on on Solana. 1004 01:00:57,520 --> 01:01:00,840 So whatever Solana has to deal with, right, like whenever it 1005 01:01:00,840 --> 01:01:03,280 recovers etcetera, like you will get it back. 1006 01:01:03,520 --> 01:01:06,840 But like it's it's kind of the same like you know you have this 1007 01:01:06,840 --> 01:01:10,280 kind of relationship with that network but not like there's no 1008 01:01:10,280 --> 01:01:13,000 bridge that you need to deal with and kind of think of as 1009 01:01:13,000 --> 01:01:15,680 like an intermediate you know complexity. 1010 01:01:16,000 --> 01:01:17,760 So that's the idea. It's like you know again we're 1011 01:01:17,760 --> 01:01:19,760 going to be rolling out more documentation. 1012 01:01:20,360 --> 01:01:23,560 It's like we have a test net version coming out for people to 1013 01:01:23,560 --> 01:01:28,600 hack on in in kind of January. And so we we actually invite 1014 01:01:28,600 --> 01:01:31,680 people to start building because again like a multi chain 1015 01:01:31,680 --> 01:01:35,200 experiences like you'll be able to build this this way easier 1016 01:01:36,120 --> 01:01:38,480 because you don't need to think about all of the complexity of 1017 01:01:38,480 --> 01:01:42,120 like oh this message didn't deliver that that work is like 1018 01:01:42,120 --> 01:01:44,400 paused. You know like something crashed 1019 01:01:44,400 --> 01:01:46,440 because of inscriptions like you don't need to deal with any of 1020 01:01:46,440 --> 01:01:48,480 that right. It's like you can literally sell 1021 01:01:48,800 --> 01:01:51,000 your your you know network failed. 1022 01:01:51,000 --> 01:01:54,840 You can sell the account that that has assets in failed 1023 01:01:54,840 --> 01:01:57,960 network right to somebody else for example if they want to take 1024 01:01:57,960 --> 01:02:00,080 that risk, right. So like you can do that without 1025 01:02:00,080 --> 01:02:02,680 having that network life Even so. 1026 01:02:02,680 --> 01:02:06,000 So that's kind of the the, the level of experience we won by 1027 01:02:06,320 --> 01:02:08,640 and like this leads to fully obstructing the block chains 1028 01:02:08,640 --> 01:02:10,520 right. Because now from a user 1029 01:02:10,520 --> 01:02:14,480 interface I just go I use the app and like I just see that I'm 1030 01:02:14,480 --> 01:02:19,040 using my for example near account and it doesn't really 1031 01:02:19,040 --> 01:02:22,360 matter for me that that was a Solana like NFT that I bought 1032 01:02:22,360 --> 01:02:23,920 right. I just see it in My Portfolio 1033 01:02:23,920 --> 01:02:27,120 view and like you know for that we need like indexing of Solana 1034 01:02:27,120 --> 01:02:31,080 and all the other chains data. So the same stack there we need 1035 01:02:31,200 --> 01:02:33,480 decentralized front end that kind of aggregate all this and 1036 01:02:33,480 --> 01:02:35,520 so kind of that. That's like how we package all 1037 01:02:35,520 --> 01:02:38,840 the stack into by abstracting the blockchain. 1038 01:02:39,880 --> 01:02:44,560 So quick quick nerd question which is so OK so Nia's like 1039 01:02:44,560 --> 01:02:47,840 this is awesome. First of all I mean NIA becoming 1040 01:02:47,840 --> 01:02:51,680 like a distributed custodian essentially imagine it as like 1041 01:02:51,680 --> 01:02:56,200 Coinbase but distributed and the distributed custodian can have 1042 01:02:56,600 --> 01:03:00,840 hot wallets basically on all of the other all of the other 1043 01:03:00,840 --> 01:03:05,480 chains. But as an engineer my my my 1044 01:03:05,480 --> 01:03:09,640 questions really starts to be in in Bitcoin you have like a 1045 01:03:09,640 --> 01:03:11,800 single single account or a multi sig account. 1046 01:03:11,800 --> 01:03:13,840 That's what Bitcoin provides available right? 1047 01:03:13,840 --> 01:03:18,320 Like it, it assumes that there is like maybe like 1 private key 1048 01:03:18,440 --> 01:03:20,960 and then one public key and there's a signature to that 1049 01:03:20,960 --> 01:03:23,400 public key. Whereas near as a distributed 1050 01:03:23,400 --> 01:03:27,040 network has lots of validators. So how do what fancy 1051 01:03:27,040 --> 01:03:30,120 cryptography makes this makes this work? 1052 01:03:31,360 --> 01:03:32,880 Yeah, so it's called chain signature. 1053 01:03:33,040 --> 01:03:37,880 And so this is a threshold signature with where the valid 1054 01:03:37,880 --> 01:03:40,360 as validate is rotated. You can maintain actually the 1055 01:03:40,360 --> 01:03:45,360 same set of public keys. So even though you rotate and 1056 01:03:45,360 --> 01:03:49,120 like have different parts of the private key being rotated, they 1057 01:03:49,120 --> 01:03:53,080 all like when they sign threshold signature you get the 1058 01:03:53,080 --> 01:03:56,040 same public key then. I mean you can have derivations 1059 01:03:56,040 --> 01:03:59,120 of this, so you can have like as many public is possible, but 1060 01:03:59,120 --> 01:04:03,600 they all deterministic within the whole blockchain, so, so 1061 01:04:03,600 --> 01:04:07,720 that's a pretty cool technology. And yeah, like it's kind of 1062 01:04:08,080 --> 01:04:10,760 reasonably new. Some of the folks from the city 1063 01:04:10,760 --> 01:04:14,440 have been pioneering that and yeah, we've kind of leveraging 1064 01:04:14,440 --> 01:04:19,800 that as a way to have needed to become this, this decentralized 1065 01:04:20,120 --> 01:04:23,520 kind of custodial. Right. 1066 01:04:23,520 --> 01:04:28,120 I think maybe also XLR works a little bit like that or am I? 1067 01:04:28,120 --> 01:04:31,440 I think, But anyway, one question that I had like in this 1068 01:04:31,440 --> 01:04:36,480 scenario where you have the divider NFT on Solana, you need 1069 01:04:36,480 --> 01:04:39,280 the liquidity right on Solana as a user. 1070 01:04:39,840 --> 01:04:42,200 So maybe I have like funds on the air, but I don't have on 1071 01:04:42,400 --> 01:04:45,200 Solana. Like is there some system that 1072 01:04:45,200 --> 01:04:48,880 you're thinking of to to balance that out without bridging or 1073 01:04:48,960 --> 01:04:50,040 yeah. Exactly. 1074 01:04:50,040 --> 01:04:51,840 Yeah. So we so this is where we call 1075 01:04:51,840 --> 01:04:54,760 them intent through layers or I mean we'll we're still shopping 1076 01:04:54,760 --> 01:04:59,720 the name but so this is the idea that on near we have this well 1077 01:04:59,720 --> 01:05:02,920 we have this principle of trial account to this idea where I can 1078 01:05:02,920 --> 01:05:07,080 send you right now a link you click on it and you'll have some 1079 01:05:07,080 --> 01:05:10,720 near in it. So you can do stuff on near but 1080 01:05:10,720 --> 01:05:13,840 you cannot withdraw that near. So we actually like, we kind of 1081 01:05:14,480 --> 01:05:18,560 like what what it'll do is like actually send you a one time use 1082 01:05:18,560 --> 01:05:21,800 private key which when you click it actually create a new private 1083 01:05:21,800 --> 01:05:25,440 key in your in on your browser switch that private key. 1084 01:05:25,560 --> 01:05:28,720 But that private key is limited access to that account. 1085 01:05:28,720 --> 01:05:32,240 So you can transact but you cannot withdraw funds And so 1086 01:05:32,240 --> 01:05:37,200 that kind of concept applied now to other kind of chains in a way 1087 01:05:37,480 --> 01:05:42,320 a lot what it allows to do is we can have other parties to, to 1088 01:05:42,400 --> 01:05:47,880 fund the account to execute things, right. 1089 01:05:47,880 --> 01:05:51,840 They can put some Solana tokens to pay for gas or for NFTS but 1090 01:05:51,840 --> 01:05:54,400 you cannot withdraw that by sending a like direct 1091 01:05:54,400 --> 01:05:58,600 transaction to withdraw Solana. So, So what this allows now to 1092 01:05:58,600 --> 01:06:03,440 do is you can pay the somebody on near with near token and then 1093 01:06:03,440 --> 01:06:07,520 they will put Solana tokens there and then execute your 1094 01:06:07,520 --> 01:06:11,640 transaction. And kind of by doing that right, 1095 01:06:11,640 --> 01:06:15,400 we kind of have pretty much a way to that's why I say it's 1096 01:06:15,400 --> 01:06:17,600 intent, right. You say like my intent is to buy 1097 01:06:17,600 --> 01:06:20,640 some Solana thing but I don't have Solana token like here's 1098 01:06:20,640 --> 01:06:25,080 punch and near tokens execute that there and so and you know 1099 01:06:25,240 --> 01:06:28,280 now you need somebody who has liquidity on all the chains to 1100 01:06:28,280 --> 01:06:30,360 execute the stuff. But that's like having a third 1101 01:06:30,360 --> 01:06:33,400 parties doing that is way easier right than to have like whole 1102 01:06:33,400 --> 01:06:35,920 bridging and automatic execution. 1103 01:06:37,760 --> 01:06:41,880 So this is like, yeah, I really like some sort of fee attached 1104 01:06:41,880 --> 01:06:43,520 to it and the relayer can grab it. 1105 01:06:43,520 --> 01:06:45,400 It's not like a blockchain network or anything. 1106 01:06:45,560 --> 01:06:47,120 Yeah, yeah, exactly. Yeah. 1107 01:06:47,120 --> 01:06:48,800 There's like a surf body, like, you know. 1108 01:06:49,680 --> 01:06:51,720 Like a market maker or whatever. Whoever it is, it's. 1109 01:06:51,720 --> 01:06:56,480 Like yeah, any market maker or any like bot arbitrage bots can 1110 01:06:56,480 --> 01:07:01,240 do this kind of stuff pretty much and they also is doing that 1111 01:07:01,240 --> 01:07:04,400 they'll just relay the transaction as well. 1112 01:07:04,400 --> 01:07:06,520 So like you don't need to actually also send the like 1113 01:07:06,720 --> 01:07:09,120 submit transaction because like the valid is only signed 1114 01:07:09,120 --> 01:07:11,200 transaction right now and somebody needs to like actually 1115 01:07:11,280 --> 01:07:14,200 ship it to peer-to-peer network. So they they will do that as 1116 01:07:14,200 --> 01:07:16,560 well, yeah. That's pretty awesome. 1117 01:07:16,560 --> 01:07:17,680 Yeah. I like looking forward to 1118 01:07:17,680 --> 01:07:21,120 reading more about it once the the more documentation is there 1119 01:07:21,120 --> 01:07:22,760 and stuff. But yeah, thanks for for sharing 1120 01:07:22,760 --> 01:07:26,560 it here. And yeah, I guess further in the 1121 01:07:26,560 --> 01:07:29,560 near journey like we we didn't actually talk about much about 1122 01:07:29,560 --> 01:07:35,120 the chain itself, right. I think you were basically one 1123 01:07:35,120 --> 01:07:37,720 of the first, if not the first like sort of sharded block 1124 01:07:37,720 --> 01:07:43,520 chains and and been like staying with that sort of narrative 1125 01:07:43,520 --> 01:07:45,920 wire. I think the others have pivoted 1126 01:07:45,920 --> 01:07:50,160 from that. So yeah, can you tell us a bit 1127 01:07:50,160 --> 01:07:54,760 like how has the like near sharding developed or what? 1128 01:07:54,760 --> 01:07:57,720 What is sharding actually again for people that forgot about it 1129 01:07:58,400 --> 01:08:00,200 and you know where, where, where is it going? 1130 01:08:01,840 --> 01:08:04,140 Yeah. So as I mentioned, right, my Co 1131 01:08:04,140 --> 01:08:06,760 founder Alex who was you know, building sharded database, I 1132 01:08:06,760 --> 01:08:09,400 mean I'm coming from Google where everything is charted just 1133 01:08:09,400 --> 01:08:13,200 like you cannot have, you know, billion users and put them into 1134 01:08:13,200 --> 01:08:14,600 one database. This just doesn't work. 1135 01:08:15,120 --> 01:08:18,960 And so and like on my computer and so for us it was like, you 1136 01:08:18,960 --> 01:08:21,160 know, kind of pretty obvious that you need charting. 1137 01:08:21,479 --> 01:08:25,560 And so sharding, I mean at at the core of the idea is like you 1138 01:08:25,640 --> 01:08:28,800 as you process, you know, as you store more data, as you process 1139 01:08:28,800 --> 01:08:33,200 more transactions, you need multiple machines doing work in 1140 01:08:33,200 --> 01:08:36,880 parallel and you want this machines to be kind of doing 1141 01:08:37,240 --> 01:08:39,800 similar work, right. And like distributing load And 1142 01:08:39,800 --> 01:08:43,680 ideally as more load comes in, you actually increase number of 1143 01:08:43,840 --> 01:08:46,479 computers, right. So this is how all of the VEP 2 1144 01:08:46,479 --> 01:08:49,520 giants work. You know, again, imagine your 1145 01:08:49,520 --> 01:08:52,560 Gmail, right, or imagine Facebook, right. 1146 01:08:52,560 --> 01:08:55,479 There's like a database underneath which you know is 1147 01:08:55,560 --> 01:08:58,000 charted. It has hundreds or thousands of 1148 01:08:58,000 --> 01:09:00,760 servers that store for example user data. 1149 01:09:01,040 --> 01:09:05,000 And you know when you're a user requesting it routes you to the 1150 01:09:05,000 --> 01:09:08,520 server where your user data is and retrieves it and then when 1151 01:09:08,520 --> 01:09:12,000 you need to update something or process you know transaction, it 1152 01:09:12,000 --> 01:09:13,279 kind of routes a transaction there. 1153 01:09:13,560 --> 01:09:15,160 So that's kind of the core concept. 1154 01:09:15,240 --> 01:09:19,359 And like you know again logically you cannot have like 1155 01:09:20,080 --> 01:09:23,920 you cannot have billions of users using the same like 1 1156 01:09:23,920 --> 01:09:26,040 server, right? And this is what's currently 1157 01:09:26,040 --> 01:09:29,840 happening where for non sharded systems it means like they 1158 01:09:29,840 --> 01:09:32,920 relying on pretty much one server replicated, but one 1159 01:09:32,920 --> 01:09:36,560 server nevertheless to process everything that happens on their 1160 01:09:36,560 --> 01:09:40,439 chain and so. So for us it was kind of, you 1161 01:09:40,439 --> 01:09:41,920 know, pretty obvious when to do this. 1162 01:09:41,920 --> 01:09:45,240 Now blockchain adds extra complexity compared to that too, 1163 01:09:45,240 --> 01:09:48,359 where you have all of the, you know, security that you need to 1164 01:09:48,359 --> 01:09:52,920 deal with. And so we've been kind of 1165 01:09:53,120 --> 01:09:56,440 obviously iterating on, on a design kind of within this 1166 01:09:56,440 --> 01:09:58,720 conceptual thing. And so we introduced Nightshade 1167 01:09:59,480 --> 01:10:04,280 back in 2019, which was our sharding design where in a way 1168 01:10:04,280 --> 01:10:08,880 every single near contract or account is actually a separate 1169 01:10:09,120 --> 01:10:13,800 chain and we just bundle them in such a way such that as users 1170 01:10:13,800 --> 01:10:16,480 and developers you don't know about it, right. 1171 01:10:17,160 --> 01:10:20,600 And so we kind of bundle them to the number of machines that you 1172 01:10:20,600 --> 01:10:23,480 know parallel processing machines at a time you need to. 1173 01:10:23,600 --> 01:10:27,080 And so again this is very similar how VEP 2 works where 1174 01:10:27,080 --> 01:10:31,400 you know every like user account is in a way independent and they 1175 01:10:31,400 --> 01:10:34,240 store and they can be like moved around between different 1176 01:10:34,240 --> 01:10:38,640 databases like between different computers in the database and 1177 01:10:38,640 --> 01:10:40,320 so. So this kind of allows to 1178 01:10:40,320 --> 01:10:43,000 abstract out the complexity of the sharding from the user, 1179 01:10:43,000 --> 01:10:44,600 right? As a user, if you go to near 1180 01:10:44,600 --> 01:10:46,560 blockchain, you will not see shards. 1181 01:10:46,560 --> 01:10:49,240 We don't actually show them like you need to go to our PC and 1182 01:10:49,240 --> 01:10:51,480 like query the block headers and stuff like this. 1183 01:10:53,240 --> 01:10:58,440 Now the thing that we in 19 were planning to do was for security 1184 01:10:58,440 --> 01:11:01,360 was based on challenges and that's proved to be very 1185 01:11:01,360 --> 01:11:03,360 challenging. And this is across the whole 1186 01:11:03,360 --> 01:11:05,960 space, right? We've seen like a number of 1187 01:11:05,960 --> 01:11:08,960 other chains actually struggling with implementing challenges. 1188 01:11:09,320 --> 01:11:15,320 And so kind of earlier this year we ended up kind of doing 1189 01:11:15,320 --> 01:11:18,480 research and and refocusing on instead doing stateless 1190 01:11:18,480 --> 01:11:22,360 validation. So what this means is now when 1191 01:11:22,360 --> 01:11:27,120 block is produced, block actually contains all of the 1192 01:11:27,320 --> 01:11:33,200 state that that execute the transactions touched and that 1193 01:11:33,200 --> 01:11:35,640 information is being sent around to everybody else. 1194 01:11:36,760 --> 01:11:40,360 What this means is that other validators don't need to have 1195 01:11:40,360 --> 01:11:43,160 state of the Shard, they can just validate the block on its 1196 01:11:43,160 --> 01:11:46,200 own. And it means we can have, you 1197 01:11:46,200 --> 01:11:49,360 know, hundreds and thousands of validators validating every 1198 01:11:49,360 --> 01:11:51,160 Shard. It can be completely random, 1199 01:11:51,160 --> 01:11:54,120 they don't need to be assigned to specific Shard at any time. 1200 01:11:54,480 --> 01:12:01,480 And this also means we can you know have now a lot more kind of 1201 01:12:01,480 --> 01:12:03,640 nodes and validators in a network kind of proving the 1202 01:12:03,640 --> 01:12:08,280 whole system now kind of on a low level what it what NIR is, 1203 01:12:08,280 --> 01:12:13,840 is really a decentralized shared sequencer that then sends out 1204 01:12:13,840 --> 01:12:17,240 the data availability of this transactions across the whole 1205 01:12:17,240 --> 01:12:19,680 chain. We use erasure coding and then 1206 01:12:19,680 --> 01:12:23,840 we have this execution which now is stateless execution which 1207 01:12:23,840 --> 01:12:27,880 then is being proven by number of other validators and and 1208 01:12:27,880 --> 01:12:30,280 settled right. So we kind of package the whole 1209 01:12:30,280 --> 01:12:34,640 what now is modular framework actually in one you know 1210 01:12:34,640 --> 01:12:38,960 pipeline way on top of the same set of validators, right, kind 1211 01:12:38,960 --> 01:12:41,120 of just being rotated constantly across the network. 1212 01:12:42,160 --> 01:12:46,240 And so that's you know at the core what near is. 1213 01:12:46,240 --> 01:12:53,280 And so we actually going to be launching the new testing 1214 01:12:53,280 --> 01:12:58,120 network for stateless validation to kind of as a part of our 1215 01:12:58,120 --> 01:13:01,320 phase two launch. And so this is kind of finalizes 1216 01:13:01,320 --> 01:13:05,360 like core road map of of sharding that we've outlined 1217 01:13:05,360 --> 01:13:10,680 since 2019 and this should be coming kind of January or 1218 01:13:10,680 --> 01:13:14,240 February and we're going to have you know the full mainnet launch 1219 01:13:14,240 --> 01:13:17,000 probably in April. And this is the idea that 1220 01:13:17,360 --> 01:13:21,680 actually kind of conceptually if people read like with Dalek's 1221 01:13:21,680 --> 01:13:25,160 end game this is in the way that structure you have block 1222 01:13:25,160 --> 01:13:29,520 producers who are charted or kind of can you know we can keep 1223 01:13:29,520 --> 01:13:32,160 adding more block producers in parallel, so you can keep 1224 01:13:32,160 --> 01:13:37,360 scaling the network. We also moving the kind of a 1225 01:13:37,360 --> 01:13:41,440 somewhat in because of this block producers now don't need 1226 01:13:41,440 --> 01:13:43,520 to rotate as much. We're actually moving the whole 1227 01:13:43,520 --> 01:13:49,120 state into memory which gives us about 10X improvement on each 1228 01:13:49,120 --> 01:13:52,160 shards kind of transaction processing. 1229 01:13:54,200 --> 01:13:57,400 And so like each Shard gets 10X and then you can have more 1230 01:13:57,400 --> 01:14:02,360 shards And so then they kind of you know they do there's a 1231 01:14:02,360 --> 01:14:06,640 coding date availability and then they do processing creates 1232 01:14:06,640 --> 01:14:11,240 this blocks with state witnesses send them out and then you have 1233 01:14:11,240 --> 01:14:15,280 large network of validators who who don't need to be this large 1234 01:14:15,600 --> 01:14:19,320 who can just validate this blocks without having the full 1235 01:14:19,320 --> 01:14:22,600 state of the chain. So that that's kind of the you 1236 01:14:22,600 --> 01:14:26,800 know in a way finalizing our road map but also very much end 1237 01:14:26,800 --> 01:14:28,600 game. There's kind of bundles a lot of 1238 01:14:28,600 --> 01:14:32,360 the current like roll up concepts and you know sells a 1239 01:14:32,360 --> 01:14:36,280 base concept that is idiom is talking about into one product. 1240 01:14:36,920 --> 01:14:41,560 And then we announced we we working on ZK Vazem with Polygon 1241 01:14:41,560 --> 01:14:45,680 because this kind of just sending out state witness with 1242 01:14:45,680 --> 01:14:49,200 the block is actually a lot of bandwidth. 1243 01:14:49,800 --> 01:14:53,560 And what ZK Vazem allows us to do is actually to prove the 1244 01:14:53,560 --> 01:14:57,960 whole block execution with state witness on the block producer 1245 01:14:57,960 --> 01:15:00,200 directly. And so now instead of sending 1246 01:15:00,200 --> 01:15:04,880 like potentially you know MB of data we can just send you know 1247 01:15:05,400 --> 01:15:11,120 whatever 10 KB proof out and everybody else can just validate 1248 01:15:11,120 --> 01:15:13,440 that without re executing all the same transactions. 1249 01:15:13,680 --> 01:15:16,960 So that is kind of actually you know final game. 1250 01:15:16,960 --> 01:15:19,720 I mean there's like a few more pieces that to complete the 1251 01:15:19,720 --> 01:15:24,360 picture but that is the structure that we we think is 1252 01:15:24,360 --> 01:15:28,400 pretty much final kind of architecture that you know you 1253 01:15:28,400 --> 01:15:33,280 have tenship resistant shared charted sequencer, right. 1254 01:15:33,280 --> 01:15:37,640 So and you can you know you have like all the data availability 1255 01:15:37,640 --> 01:15:40,760 underneath to provide you so that and like we do data 1256 01:15:40,760 --> 01:15:43,960 availability first before execution because that means all 1257 01:15:43,960 --> 01:15:47,080 the other indexers and other piece of infrastructure can 1258 01:15:47,080 --> 01:15:50,640 start executing in parallel. And so you don't have latency on 1259 01:15:50,800 --> 01:15:55,880 user interfaces before the kind of finalization of the execution 1260 01:15:55,880 --> 01:15:59,680 on the validators themself. Then you have kind of execution 1261 01:15:59,680 --> 01:16:05,800 on validators send out witness and now you know large network 1262 01:16:05,920 --> 01:16:09,000 of validators can validate it and prove it without needing to 1263 01:16:09,000 --> 01:16:11,560 have state for dated. And all kind of having like you 1264 01:16:11,560 --> 01:16:13,840 know potential state is like 50 gigabytes for example. 1265 01:16:13,840 --> 01:16:16,560 So they don't need to like have that 50 gigabytes on them. 1266 01:16:16,560 --> 01:16:19,400 They just receive whatever relevant for the transactions 1267 01:16:19,400 --> 01:16:23,320 are being processed. And so that's kind of the, yeah, 1268 01:16:23,320 --> 01:16:26,720 I mean it's it's a little bit complicated as a as a scheme but 1269 01:16:26,720 --> 01:16:31,120 but like really it's powering this again like the kind of the 1270 01:16:31,120 --> 01:16:35,920 end game structure that people been talking about at the same 1271 01:16:35,920 --> 01:16:40,120 time it's it is like kind of that modularity just like 1272 01:16:40,560 --> 01:16:44,240 reusing the same set of servers right to to ensure kind of 1273 01:16:44,960 --> 01:16:47,640 throughput and latency low latency. 1274 01:16:49,200 --> 01:16:51,600 Yeah, like that's. That's like an episode on its 1275 01:16:51,640 --> 01:16:55,120 own, to be honest. To dig through that, is it 1276 01:16:55,120 --> 01:16:58,400 correct to think that like the stateless validation requires ZK 1277 01:16:58,400 --> 01:17:03,600 wasam as a primitive? So no, because you you can do 1278 01:17:03,600 --> 01:17:07,320 stateless validation without ZK. So what you do is you execute 1279 01:17:07,320 --> 01:17:10,720 transactions, you record which pieces of state you touched and 1280 01:17:10,720 --> 01:17:14,160 then you just send those pieces of state with with witnesses, 1281 01:17:14,160 --> 01:17:18,200 right, with kind of proof that it's part of the state together 1282 01:17:18,200 --> 01:17:20,640 with transactions. And so we're actually launching 1283 01:17:20,640 --> 01:17:25,840 that first while in parallel kind of working on ZK Wasim and 1284 01:17:25,840 --> 01:17:28,800 so Zikiwasm, what it allows to do though is just compress all 1285 01:17:28,800 --> 01:17:31,760 of these and execution of invalidation of this into just a 1286 01:17:31,760 --> 01:17:34,680 proof right. So in a way that Zikiwasm will 1287 01:17:34,680 --> 01:17:38,520 prove the execution of this BLOB pretty much state plus 1288 01:17:38,520 --> 01:17:42,800 transactions into just a fixed size proof but it but it's so 1289 01:17:42,800 --> 01:17:45,440 it's more of an optimistic Zikiwasm from this perspective 1290 01:17:45,440 --> 01:17:49,440 is optimization and it's obviously like way better for 1291 01:17:49,760 --> 01:17:53,800 like longer term you know storage but it's not a 1292 01:17:53,800 --> 01:17:57,440 prerequisite. I mean maybe I'll I'll try to 1293 01:17:57,960 --> 01:18:01,640 present my simple imagination of like of the system. 1294 01:18:01,960 --> 01:18:04,720 So the way I imagine it is like if you imagine I'm a I'm a 1295 01:18:04,720 --> 01:18:08,880 validator I'm an accountant right automated accountant. 1296 01:18:09,960 --> 01:18:16,440 Essentially in near I have the capability I'm assigned somehow 1297 01:18:16,440 --> 01:18:21,360 like some piece of work and and some of my work is also rotating 1298 01:18:21,400 --> 01:18:24,960 right like it's not part there's a massive Ledger, massive 1299 01:18:24,960 --> 01:18:30,080 Ledger, massive state and I am assigned hey go and make some 1300 01:18:30,080 --> 01:18:34,880 changes to this part of of the state so I can I can basically 1301 01:18:35,200 --> 01:18:38,280 go to that part of the state. There are bunch of transaction 1302 01:18:38,280 --> 01:18:42,960 associated with it. I execute the transactions and 1303 01:18:42,960 --> 01:18:45,400 1st I can I make the data available. 1304 01:18:45,400 --> 01:18:49,360 Hey these are the transactions I am I'm going to execute. 1305 01:18:49,880 --> 01:18:55,160 I make the execution, I update the state and today I somehow 1306 01:18:55,400 --> 01:18:59,040 provide some witnesses so that for the other accountants I can 1307 01:18:59,040 --> 01:19:02,360 sort of provide a proof. Hey, I did my job correctly, 1308 01:19:02,880 --> 01:19:07,240 here's proof. And they don't need to download 1309 01:19:07,240 --> 01:19:11,480 my part of the state to verify my work and the ZK proof will 1310 01:19:11,480 --> 01:19:14,600 make that even easier. So the state imagine as a 1311 01:19:14,600 --> 01:19:16,240 massive massive tree or something. 1312 01:19:16,280 --> 01:19:20,000 I can, I can modify some branches of the tree and I 1313 01:19:20,000 --> 01:19:24,080 create a proof and then I that proof is witnesses today, 1314 01:19:24,080 --> 01:19:27,800 Ezekiel wasn't tomorrow and I can send that thing to others. 1315 01:19:28,160 --> 01:19:31,920 They don't actually need to have my part of the tree in order to 1316 01:19:31,920 --> 01:19:36,040 verify my work. And and then there's a separate 1317 01:19:36,040 --> 01:19:39,800 system that says OK, in modifying this part of the tree. 1318 01:19:40,520 --> 01:19:44,720 What are the transactions? I did some somebody duplicates 1319 01:19:44,760 --> 01:19:49,520 duplicates that work and because because I can modify a part of 1320 01:19:49,520 --> 01:19:55,920 the tree quite independently and there are many like me, so there 1321 01:19:55,920 --> 01:19:58,720 are many accountants like me. All of these accountants are 1322 01:19:58,720 --> 01:20:03,200 kind of modifying like different parts of the tree in parallel 1323 01:20:03,280 --> 01:20:07,480 and like that is fundamentally why the system is able to scale. 1324 01:20:09,040 --> 01:20:12,680 Yeah, very well put. So you have partnership with 1325 01:20:12,680 --> 01:20:16,040 Eigen DA. Why do you need a partnership 1326 01:20:16,040 --> 01:20:18,720 with Eigen DA in that case? Yeah. 1327 01:20:18,720 --> 01:20:22,000 So, so this kind of maybe, yeah, changing gears right. 1328 01:20:22,000 --> 01:20:25,720 So so this is like near itself. This is near itself, right? 1329 01:20:25,720 --> 01:20:27,840 Like it has no interaction with other things but. 1330 01:20:27,960 --> 01:20:34,480 Yeah, yeah, so so and again Near itself right now is you know top 1331 01:20:34,480 --> 01:20:36,280 use blockchain by number of addresses. 1332 01:20:36,280 --> 01:20:40,720 For example, you know daily active, monthly active, weekly 1333 01:20:40,720 --> 01:20:43,760 active. And so like Near itself has like 1334 01:20:43,760 --> 01:20:45,880 a bunch of utility and and value already. 1335 01:20:46,400 --> 01:20:50,080 But again we kind of when we frame this like chain 1336 01:20:50,080 --> 01:20:53,360 abstraction thesis, right. What it means is that for the 1337 01:20:53,360 --> 01:20:56,680 developers and users on top, we're trying to provide a smooth 1338 01:20:56,680 --> 01:20:59,240 experience across using other chains as well. 1339 01:20:59,640 --> 01:21:04,120 And this is where we kind of looked around and like oh, NIR 1340 01:21:04,120 --> 01:21:09,560 already has data availability built in like that's just part 1341 01:21:09,560 --> 01:21:13,160 of our protocol. And so we have seen a bunch of 1342 01:21:13,160 --> 01:21:17,920 layer twos that we can plug in into this to kind of hooking 1343 01:21:17,920 --> 01:21:19,960 into the rest of our systems, right. 1344 01:21:20,480 --> 01:21:26,320 And so that's where we kind of you know started in in like kind 1345 01:21:26,320 --> 01:21:33,200 of pretty much provided a way to hook in AP stack, CDK, Stark, 1346 01:21:33,200 --> 01:21:37,040 Nets, kind of stacks. How do you publish your data on 1347 01:21:37,040 --> 01:21:39,560 NIR? Now if you just publish data on 1348 01:21:39,560 --> 01:21:43,000 NIR, it's useful it, you know, it's obviously very cheap. 1349 01:21:43,000 --> 01:21:46,280 It's you know way cheap like cheaper than pretty much 1350 01:21:46,280 --> 01:21:48,960 everything else in the market. And because near as shorted you 1351 01:21:48,960 --> 01:21:52,800 actually have more capacity than anything else can that can take 1352 01:21:52,880 --> 01:21:56,240 your data already and we're going to add more shorts, but 1353 01:21:56,600 --> 01:22:00,080 it's not as useful because you cannot route messages between 1354 01:22:01,720 --> 01:22:05,320 between kind of smart contracts on roll ups between each other 1355 01:22:05,320 --> 01:22:10,080 and near in near contracts. And so that's where we had a 1356 01:22:10,080 --> 01:22:17,160 partnership with Eigen layer not Eigen DA to help us actually do 1357 01:22:17,160 --> 01:22:22,880 the work for this layer twos to get to executed state and 1358 01:22:23,680 --> 01:22:28,000 outgoing messages such that the applications that want to route 1359 01:22:28,000 --> 01:22:31,240 messages faster within one and two second, they can actually do 1360 01:22:31,240 --> 01:22:36,360 that through the near network. So Eigen layer validators will 1361 01:22:36,560 --> 01:22:40,240 execute this roll up given the data published on near, they'll 1362 01:22:40,240 --> 01:22:44,040 execute it and they will have a new state route for the roll up 1363 01:22:44,040 --> 01:22:46,680 itself now. So think of it it's it'll be 1364 01:22:46,680 --> 01:22:50,360 extra accountants, Etherium accountants who will be actually 1365 01:22:50,360 --> 01:22:55,280 looking at the roll ups and updating statehood there, but 1366 01:22:55,280 --> 01:22:58,480 then publishing back to the near like telling it to near 1367 01:22:58,480 --> 01:23:01,880 accountants as well. And so now near accountants and 1368 01:23:01,880 --> 01:23:05,000 Etherium accounts together know the state of both near and all 1369 01:23:05,000 --> 01:23:07,720 of the roll ups that are plugged into the system. 1370 01:23:07,880 --> 01:23:10,880 And so now you can route messages between roll up 1371 01:23:10,880 --> 01:23:14,280 contracts and near contracts and you know back and forth. 1372 01:23:14,840 --> 01:23:18,240 And so this allows us to kind of again like align more the the, 1373 01:23:18,600 --> 01:23:24,080 the space of of the the space. And so again for chain 1374 01:23:24,080 --> 01:23:26,520 abstraction for kind of aggregation it means we can do 1375 01:23:26,520 --> 01:23:30,240 things way faster between all of the roll ups that that fit into 1376 01:23:30,240 --> 01:23:32,320 the system. So that's kind of how like DA 1377 01:23:32,440 --> 01:23:36,240 plus Eigen layer kind of provides this fast finality. 1378 01:23:36,760 --> 01:23:39,920 And then you know there's other kind of tooling that we you know 1379 01:23:39,920 --> 01:23:43,000 plug in on top is decentralized front ends to really kind of 1380 01:23:43,040 --> 01:23:46,920 abstract it from a user. But like we need that kind of 1381 01:23:47,000 --> 01:23:51,640 alignment again neat near in the way each account like each each 1382 01:23:51,640 --> 01:23:54,840 element of that tree is is separate roll up right And we 1383 01:23:54,840 --> 01:23:58,920 kind of we have a system for managing them And so we kind of 1384 01:23:58,920 --> 01:24:02,920 trying to fit the other roll ups into the same system and you 1385 01:24:02,920 --> 01:24:06,320 know obviously we need to like plug in some pieces to to make 1386 01:24:06,320 --> 01:24:09,240 it work under the same security parameters that roll up expect 1387 01:24:09,240 --> 01:24:13,840 right which is a theorem security hands dagger layer and 1388 01:24:13,840 --> 01:24:17,120 then DA is kind of way to get this data you know into the 1389 01:24:17,120 --> 01:24:20,040 system as well and and provide some guarantees there. 1390 01:24:23,400 --> 01:24:27,440 Hard to unpack, but but like logically it's like, yeah, 1391 01:24:27,560 --> 01:24:29,160 imagine, yeah, it's exactly that. 1392 01:24:29,160 --> 01:24:32,920 It's imagine near as this massive tree and then there are 1393 01:24:32,920 --> 01:24:35,360 like lots of accountants in near itself. 1394 01:24:35,360 --> 01:24:38,640 There's one group of accountants and then accountants can kind of 1395 01:24:38,640 --> 01:24:42,040 modify parts of the tree independent of each other. 1396 01:24:42,480 --> 01:24:45,160 We can send proofs about their modifications so that other 1397 01:24:45,200 --> 01:24:47,120 other accountants can trust their work. 1398 01:24:48,120 --> 01:24:51,960 And then kind of like this, eigen layer partnership is in 1399 01:24:51,960 --> 01:24:55,480 some way saying that. There is CDM accountants, yeah. 1400 01:24:56,160 --> 01:24:59,760 Yeah, it's like near, Near says. We have an awesome group of 1401 01:24:59,760 --> 01:25:01,640 accountants. But if you want your own 1402 01:25:01,640 --> 01:25:04,440 accountants and if you want your own roll up, you have created a 1403 01:25:04,440 --> 01:25:08,440 separate group of accountants. But then your accountants and 1404 01:25:08,480 --> 01:25:16,200 the Near accountants we we sort of need to interface in some way 1405 01:25:16,520 --> 01:25:22,040 so that so that the work, your accountant's date can be 1406 01:25:22,040 --> 01:25:24,680 deduplicated on near and the other way around. 1407 01:25:24,800 --> 01:25:28,240 And via this deduplication we can somehow achieve like 1408 01:25:28,240 --> 01:25:33,320 trustless interactions between Ethereum roll ups and Near 1409 01:25:33,960 --> 01:25:37,320 something like that, right? Yeah, pretty much like I would 1410 01:25:37,320 --> 01:25:39,720 say that. So the roll ups is pretty much I 1411 01:25:39,720 --> 01:25:42,880 want my own accountant, right that runs everything. 1412 01:25:43,000 --> 01:25:47,480 But then I I trust Etherium accountants to revalidate 1413 01:25:47,480 --> 01:25:50,200 everything and and finalize it, right. 1414 01:25:50,720 --> 01:25:53,760 So like Etherium accountants are are the final final. 1415 01:25:53,760 --> 01:25:55,920 My accountant is the one who can do quickly, right. 1416 01:25:55,920 --> 01:26:01,480 He sits right by my side. And So what we say here is near 1417 01:26:01,480 --> 01:26:05,720 accountants can, you know can provide a bunch of value by, you 1418 01:26:05,720 --> 01:26:08,360 know either connecting your accountant to the other guy's 1419 01:26:08,360 --> 01:26:11,160 accountant, right. So you can connect together or 1420 01:26:11,160 --> 01:26:14,720 to our applications. But we still need to see them 1421 01:26:14,720 --> 01:26:17,880 accountants because the finality of the roll ups is on Etherium, 1422 01:26:18,360 --> 01:26:20,400 right. And so that's why we have Eigen 1423 01:26:20,400 --> 01:26:25,880 layer pretty much to lend us their Etherium accountants to 1424 01:26:25,880 --> 01:26:29,720 kind of use the let the you know as a roll up publishes the 1425 01:26:29,720 --> 01:26:33,160 Ledger right from their account. And 1st like we have the sedum 1426 01:26:33,520 --> 01:26:35,960 you know accountants or Eigen layer to like validate 1427 01:26:35,960 --> 01:26:40,280 everything quickly right before the fully sedum sinality will 1428 01:26:40,280 --> 01:26:42,480 happen. And so that allows to kind of 1429 01:26:42,600 --> 01:26:47,520 near accountants and to have like trust into into the 1430 01:26:47,520 --> 01:26:51,880 execution of what happened on the roll up while also have the 1431 01:26:52,240 --> 01:26:56,960 way quicker you know time to finality and to communication of 1432 01:26:56,960 --> 01:27:00,640 messages for this roll ups and maintaining the same security as 1433 01:27:00,640 --> 01:27:03,520 as they have through ECDM. So that's kind of like, you 1434 01:27:03,520 --> 01:27:06,240 know, it's like roll ups near an ECDM coming all together into 1435 01:27:06,240 --> 01:27:08,040 like 1 happy family of accountants. 1436 01:27:09,480 --> 01:27:11,640 I think that's, that's a great note to end on, right. 1437 01:27:11,640 --> 01:27:14,760 Like big happy family of accountants. 1438 01:27:16,560 --> 01:27:18,960 Yeah. Thank you so much for coming on. 1439 01:27:18,960 --> 01:27:21,480 It's been like a massive episode, I think. 1440 01:27:22,680 --> 01:27:26,040 Yeah, I need to like process this and I'm I'm sure our 1441 01:27:26,040 --> 01:27:28,640 listeners will take some time to process everything too. 1442 01:27:29,320 --> 01:27:33,880 Well, we can do another one in a few months as we launched all 1443 01:27:33,880 --> 01:27:36,200 this stuff, so. Yeah, totally. 1444 01:27:36,200 --> 01:27:39,560 And yeah, we still also have Alex episode about the Smarter 1445 01:27:39,560 --> 01:27:41,040 LLMS. Outstanding. 1446 01:27:41,040 --> 01:27:44,440 So lots to do. But yeah, thanks so much for 1447 01:27:44,440 --> 01:27:45,960 coming on. And and thanks for to our 1448 01:27:45,960 --> 01:27:50,160 listeners, We'll have like 1 1/2 hours of of content here. 1449 01:27:51,080 --> 01:27:55,600 Raise guys. Thank you for joining us on this 1450 01:27:55,600 --> 01:27:57,960 week's episode. We release new episodes every 1451 01:27:57,960 --> 01:28:00,000 week. You can find and subscribe to 1452 01:28:00,000 --> 01:28:03,760 the show on iTunes, Spotify, YouTube, SoundCloud, or wherever 1453 01:28:03,760 --> 01:28:06,160 you listen to podcasts. And if you have a Google Home or 1454 01:28:06,160 --> 01:28:08,960 Alexa device, you can tell it to listen to the latest episode of 1455 01:28:08,960 --> 01:28:12,480 the Epicenter podcast. So go to epicenter.tv, subscribe 1456 01:28:12,680 --> 01:28:14,600 for a full list of places where you can watch and listen. 1457 01:28:15,200 --> 01:28:17,480 While you're there, be sure to sign up for the newsletter so 1458 01:28:17,480 --> 01:28:19,840 you get new episodes in your inbox as they're released. 1459 01:28:19,840 --> 01:28:23,160 If you want to interact with us guests or other podcast 1460 01:28:23,160 --> 01:28:24,800 listeners, you can follow us on Twitter. 1461 01:28:25,160 --> 01:28:26,640 And please leave us a review on iTunes. 1462 01:28:26,680 --> 01:28:29,160 It helps people find the show and we're always happy to read 1463 01:28:29,160 --> 01:28:31,520 them. So thanks so much and we look 1464 01:28:31,520 --> 01:28:32,680 forward to being back next week.