1 00:00:00,000 --> 00:00:12,800 Trust is a function. It is more than just a feeling. In a world of infinite noise, 2 00:00:12,800 --> 00:00:20,440 the only signal that matters is the one you verify. We are here to map the graph, 3 00:00:21,180 --> 00:00:26,000 order the protocol, and build the immune system for the decentralized web. 4 00:00:26,000 --> 00:00:29,760 Welcome to the signal processing layer. 5 00:00:30,860 --> 00:00:33,720 This is Say What? 6 00:00:41,100 --> 00:00:44,460 Welcome to the second episode of Say What? 7 00:00:45,060 --> 00:00:51,880 Today we're debugging the intersection of AI and the web of trust with the architect Jesus. 8 00:00:51,880 --> 00:01:01,120 We're tracing his build history through the protocol, from projects like DVM-CP and ContextVM to his latest one, Relator. 9 00:01:01,780 --> 00:01:10,200 This is a technical deep dive into how machine intelligence can be engineered to serve human reputation rather than capture it. 10 00:01:10,760 --> 00:01:13,900 A quick protocol reminder before we execute. 11 00:01:14,620 --> 00:01:17,260 There are no corporate sponsors in this bunker. 12 00:01:17,900 --> 00:01:20,940 This frequency is strictly peer-to-peer. 13 00:01:20,940 --> 00:01:26,880 If you extract actionable signal from this conversation, consider returning the value. 14 00:01:27,720 --> 00:01:33,900 Boosts, zaps, and streaming sats are the only proof of work that keep this node running. 15 00:01:34,520 --> 00:01:42,140 You can also commit to the signal by locking in a subscription to PlepChain Radio on Fountain. 16 00:01:42,800 --> 00:01:44,780 Let's get into the code. 17 00:01:45,260 --> 00:01:47,520 Jesus, welcome to Say What. 18 00:01:47,520 --> 00:01:57,620 hey thank you for having me here yes thank you for being here uh jesus i've uh i came across 19 00:01:57,620 --> 00:02:04,200 i've obviously seen your name on nostril i've been on nostril for three years and it's come up in on 20 00:02:04,200 --> 00:02:11,320 the timeline uh but i think i started paying attention around a year or so ago or it might 21 00:02:11,320 --> 00:02:17,740 have been even less than a year when I saw something you posted about DVMCP. 22 00:02:19,120 --> 00:02:23,280 And then after that, that sort of morphed into ContextVM. 23 00:02:23,480 --> 00:02:26,520 And now here we are with your latest project, Relator. 24 00:02:26,760 --> 00:02:31,760 But before we get into all of that, maybe let's just talk about your build history on 25 00:02:31,760 --> 00:02:32,020 Nostro. 26 00:02:32,020 --> 00:02:35,340 When did you join and what was your first project? 27 00:02:36,620 --> 00:02:37,100 Nice. 28 00:02:37,300 --> 00:02:37,540 Yes. 29 00:02:37,540 --> 00:02:40,660 so I joined Nosser for a first time 30 00:02:40,660 --> 00:02:42,780 three years ago as well 31 00:02:42,780 --> 00:02:44,480 something around that 32 00:02:44,480 --> 00:02:46,720 the first time that I heard something 33 00:02:46,720 --> 00:02:47,760 about Nosser was 34 00:02:47,760 --> 00:02:49,480 from a podcast 35 00:02:49,480 --> 00:02:52,020 of a Spanish podcaster 36 00:02:52,020 --> 00:02:54,240 he's a good friend 37 00:02:54,240 --> 00:02:56,240 and he did a 38 00:02:56,240 --> 00:02:58,320 shout out Luna, he's a good friend too 39 00:02:58,320 --> 00:02:58,600 yeah 40 00:02:58,600 --> 00:03:01,740 so he did a podcast with 41 00:03:01,740 --> 00:03:02,980 Kameri 42 00:03:02,980 --> 00:03:05,540 three years ago 43 00:03:05,540 --> 00:03:06,540 something like that 44 00:03:06,540 --> 00:03:15,740 and that was the first time that I heard about Nocer and yeah, my head exploded at that time 45 00:03:15,740 --> 00:03:23,580 and I started to think about that more and more and at that time I was working for a company of a 46 00:03:23,580 --> 00:03:32,380 friend where we were doing open source software and consultancy. Then I discovered Nocer and I 47 00:03:32,380 --> 00:03:38,880 I started to put more time and effort into basically learning everything around it. 48 00:03:40,140 --> 00:03:45,340 You know, everything related to how keepers work. 49 00:03:46,840 --> 00:03:52,460 And, you know, identifying the relation between Nocer, Bitcoin. 50 00:03:53,280 --> 00:04:02,140 And, you know, I had this aha moment after working in my first project in Nocer. 51 00:04:02,140 --> 00:04:07,180 that was basically a little key storage CLI 52 00:04:07,180 --> 00:04:09,980 that I call Nozzle Starter. 53 00:04:11,620 --> 00:04:14,320 I created that because at that time 54 00:04:14,320 --> 00:04:16,100 in Nozzle we didn't have too many clients 55 00:04:16,100 --> 00:04:19,460 and the one that I used the most at that time 56 00:04:19,460 --> 00:04:22,220 was one called Nozzle Console 57 00:04:22,220 --> 00:04:25,240 which was a CLI client 58 00:04:25,240 --> 00:04:31,880 created by Vichel, I think that is the name. 59 00:04:32,140 --> 00:04:39,140 It's a developer that is not currently in Nosr, I don't see him very actively active lately. 60 00:04:39,140 --> 00:04:50,140 But I started with that, I started with a key storage solution because in order to log in to Nosr console 61 00:04:50,140 --> 00:04:56,140 you had to basically copy and paste the NSEC in the terminal. 62 00:04:56,140 --> 00:05:12,260 And you know, bash and the terminal have a history, so it felt a bit insecure to let that private key float in plain text on the history of my terminal. 63 00:05:12,260 --> 00:05:14,280 so yeah 64 00:05:14,280 --> 00:05:16,240 so then I created this 65 00:05:16,240 --> 00:05:18,320 little case storage basically 66 00:05:18,320 --> 00:05:20,260 to load NOSER 67 00:05:20,260 --> 00:05:21,800 console without 68 00:05:21,800 --> 00:05:26,380 letting a trace of 69 00:05:26,380 --> 00:05:28,420 the private key in the 70 00:05:28,420 --> 00:05:29,340 history of the terminal 71 00:05:29,340 --> 00:05:32,440 yeah that was it 72 00:05:32,440 --> 00:05:34,420 that was my first exploration of NOSER 73 00:05:34,420 --> 00:05:37,800 by the way 74 00:05:37,800 --> 00:05:39,120 can I just 75 00:05:39,120 --> 00:05:42,220 the key sort of floating around in plain 76 00:05:42,220 --> 00:05:44,300 text. This is not as 77 00:05:44,300 --> 00:05:45,640 uncommon even today. 78 00:05:46,660 --> 00:05:48,280 There are some setups that 79 00:05:48,280 --> 00:05:49,360 I've, you know, 80 00:05:49,840 --> 00:05:52,140 so this stream now 81 00:05:52,140 --> 00:05:53,920 is being live streamed, our conversation 82 00:05:53,920 --> 00:05:56,180 on zap.stream, and 83 00:05:56,180 --> 00:05:58,100 it's on my personal 84 00:05:58,100 --> 00:05:58,580 server. 85 00:06:00,400 --> 00:06:02,020 So I run my own instance of 86 00:06:02,020 --> 00:06:04,000 zap.stream, and in the initial 87 00:06:04,000 --> 00:06:04,540 setup, 88 00:06:05,760 --> 00:06:07,460 it required me to 89 00:06:07,460 --> 00:06:09,220 paste my raw NSEC 90 00:06:09,220 --> 00:06:11,180 into the config file 91 00:06:11,180 --> 00:06:17,400 obviously that's not the case anymore but it's it just goes to show that even today 92 00:06:17,400 --> 00:06:25,300 your nsec floating around uh in in plain text is uh is a fairly is not that uncommon 93 00:06:25,300 --> 00:06:35,280 no definitely not um i think that that is uh one of the most kind of tricky points of 94 00:06:35,280 --> 00:06:37,380 nocer of the user experience 95 00:06:37,380 --> 00:06:39,540 because, you know, it's a new 96 00:06:39,540 --> 00:06:41,360 paradigm. Basically, you have keys 97 00:06:41,360 --> 00:06:43,160 and the user is responsible 98 00:06:43,160 --> 00:06:45,280 for managing those keys. 99 00:06:46,020 --> 00:06:47,620 And right now, we have very good solutions 100 00:06:47,620 --> 00:06:49,500 like things 101 00:06:49,500 --> 00:06:51,420 like Amber and Remote Sign-In 102 00:06:51,420 --> 00:06:52,860 that I think that is 103 00:06:52,860 --> 00:06:55,460 the good, the best 104 00:06:55,460 --> 00:06:57,000 UX for 105 00:06:57,000 --> 00:06:59,340 key management. But 106 00:06:59,340 --> 00:07:01,500 it's true. Like today, we're still 107 00:07:01,500 --> 00:07:03,580 having private keys 108 00:07:03,580 --> 00:07:05,280 in plain text floating around. 109 00:07:06,940 --> 00:07:07,200 Yeah. 110 00:07:07,540 --> 00:07:09,600 It's a task to educate the user 111 00:07:09,600 --> 00:07:10,720 and everyone to 112 00:07:10,720 --> 00:07:12,440 use the tools. 113 00:07:15,120 --> 00:07:15,320 Yeah. 114 00:07:15,820 --> 00:07:17,340 So carry on, Jesus. 115 00:07:17,900 --> 00:07:19,740 So once you did that key storage 116 00:07:19,740 --> 00:07:21,520 for the command line tool that you were 117 00:07:21,520 --> 00:07:23,880 working on, what was... 118 00:07:23,880 --> 00:07:24,000 Yeah. 119 00:07:25,100 --> 00:07:26,600 Yeah, I started with that. 120 00:07:27,460 --> 00:07:29,800 After that, I was 121 00:07:29,800 --> 00:07:31,260 creating Nostry. 122 00:07:31,980 --> 00:07:33,340 Nostry with 123 00:07:33,340 --> 00:07:34,720 basically was 124 00:07:34,720 --> 00:07:37,280 an alternative to Linktree 125 00:07:37,280 --> 00:07:39,260 which I assume 126 00:07:39,260 --> 00:07:40,300 you already know 127 00:07:40,300 --> 00:07:43,360 this service to publish a list 128 00:07:43,360 --> 00:07:45,320 of links for your 129 00:07:45,320 --> 00:07:46,520 socials and 130 00:07:46,520 --> 00:07:49,360 websites and anything 131 00:07:49,360 --> 00:07:50,860 is very 132 00:07:50,860 --> 00:07:53,080 used a lot in 133 00:07:53,080 --> 00:07:55,220 the music industry by 134 00:07:55,220 --> 00:07:56,900 groups and so on 135 00:07:56,900 --> 00:07:59,240 basically to share links for their work 136 00:07:59,240 --> 00:08:00,740 and 137 00:08:00,740 --> 00:08:02,540 I created this nostril 138 00:08:02,540 --> 00:08:06,320 was my first social client in Nosser. 139 00:08:08,100 --> 00:08:10,340 Yeah, that was well received. 140 00:08:12,180 --> 00:08:13,980 People are still using it today. 141 00:08:15,060 --> 00:08:18,780 I'm a bit ashamed that I haven't 142 00:08:18,780 --> 00:08:22,480 did any maintenance on it for some time, 143 00:08:22,760 --> 00:08:24,020 quite a long time. 144 00:08:25,180 --> 00:08:28,500 But I really need to revive it 145 00:08:28,500 --> 00:08:31,580 because there is a lot of value there 146 00:08:31,580 --> 00:08:36,260 and people are still using it, even if I'm not maintaining it too much, 147 00:08:36,880 --> 00:08:42,120 which is a good sign because everything is still working and nothing broke. 148 00:08:43,080 --> 00:08:46,980 But definitely it deserves a maintenance. 149 00:08:48,580 --> 00:08:54,080 Can I, you know, something very useful, and I don't know if Nostry, I've seen Nostry, 150 00:08:54,600 --> 00:09:01,020 I haven't used it extensively, but what would be super useful is if someone creates a Nostry, 151 00:09:01,580 --> 00:09:05,160 and then they're at a meetup or a conference or something, 152 00:09:05,200 --> 00:09:07,520 and they want to exchange contact info, 153 00:09:07,740 --> 00:09:09,720 and they pull up a QR code, 154 00:09:10,540 --> 00:09:12,740 and that takes them to their nostril. 155 00:09:12,740 --> 00:09:15,320 And then there should be an export feature 156 00:09:15,320 --> 00:09:22,020 that's able to export all of that into a contact link on their phone. 157 00:09:22,420 --> 00:09:24,420 And I don't know how standardized that is. 158 00:09:24,600 --> 00:09:27,440 I mean, this is not an area I'm particularly familiar with, 159 00:09:27,940 --> 00:09:30,160 the contact data standard, 160 00:09:30,160 --> 00:09:35,820 but whether it's portable across iOS, Android, Graphene, and so on. 161 00:09:36,340 --> 00:09:38,160 But that would be kind of a cool thing. 162 00:09:38,260 --> 00:09:41,460 You have your Nostri, it's a QR code that takes you there, 163 00:09:41,520 --> 00:09:44,340 and then there's a button that allows you to export it directly into contacts. 164 00:09:45,680 --> 00:09:47,320 Yeah, yeah, indeed that would be nice. 165 00:09:48,140 --> 00:09:53,040 I don't know how it would work the support function in the different platforms, 166 00:09:53,280 --> 00:09:56,420 but definitely it will work. 167 00:09:57,480 --> 00:10:00,080 But yeah, Nostri was something good. 168 00:10:00,160 --> 00:10:02,540 still something very valuable, I think 169 00:10:02,540 --> 00:10:03,620 it's a good tool 170 00:10:03,620 --> 00:10:05,760 and I have a bunch of ideas 171 00:10:05,760 --> 00:10:08,420 because, you know 172 00:10:08,420 --> 00:10:10,680 ahead from Nostry 173 00:10:10,680 --> 00:10:12,600 I have been exploring the 174 00:10:12,600 --> 00:10:13,800 contact 175 00:10:13,800 --> 00:10:16,380 apps of Nostry 176 00:10:16,380 --> 00:10:18,020 I, during 177 00:10:18,020 --> 00:10:20,240 an edition of Cyberding and Generating, I was 178 00:10:20,240 --> 00:10:22,620 working on an unreleased 179 00:10:22,620 --> 00:10:24,020 app that 180 00:10:24,020 --> 00:10:25,880 probably Gigi 181 00:10:25,880 --> 00:10:27,620 will 182 00:10:27,620 --> 00:10:29,920 shame me for 183 00:10:29,920 --> 00:10:32,080 for that because I never released it. 184 00:10:32,080 --> 00:10:34,100 I'm still there in the vault. 185 00:10:35,000 --> 00:10:38,500 But I was creating an app called Colorillo. 186 00:10:39,900 --> 00:10:41,780 Basically, creating... 187 00:10:41,780 --> 00:10:42,880 What was the name of it? 188 00:10:43,740 --> 00:10:44,160 Sorry. 189 00:10:44,380 --> 00:10:44,860 Colorillo. 190 00:10:45,280 --> 00:10:48,440 Colorillo, like a color in Spanish. 191 00:10:48,580 --> 00:10:48,700 Colorillo, okay. 192 00:10:49,140 --> 00:10:49,340 Yeah. 193 00:10:50,080 --> 00:10:57,160 But the idea behind that app was basically to create a contact list application for you. 194 00:10:57,160 --> 00:11:06,140 So you have all the details of your contacts, you can navigate them by a system of colors. 195 00:11:06,440 --> 00:11:11,900 That's basically why the name, because at the time I was exploring the, you know, 196 00:11:11,980 --> 00:11:19,780 extracting an X color string from the first six characters of the X public key. 197 00:11:19,780 --> 00:11:22,760 and basically with that 198 00:11:22,760 --> 00:11:24,660 representing all the contacts 199 00:11:24,660 --> 00:11:27,260 in a palette of colors 200 00:11:27,260 --> 00:11:30,120 basically to 201 00:11:30,120 --> 00:11:33,040 have a way to navigate the contacts 202 00:11:33,040 --> 00:11:35,140 without needing to search 203 00:11:35,140 --> 00:11:36,960 I was exploring basically the relation 204 00:11:36,960 --> 00:11:38,880 between the colors and the contacts 205 00:11:38,880 --> 00:11:42,760 in some way to understand 206 00:11:42,760 --> 00:11:43,600 if that 207 00:11:43,600 --> 00:11:46,700 was a good system to remember 208 00:11:46,700 --> 00:11:48,860 what contacts 209 00:11:48,860 --> 00:11:52,420 belong to what color and easily or quickly navigate them. 210 00:11:53,260 --> 00:11:57,760 And also that app, the goal or the vision behind that app 211 00:11:57,760 --> 00:12:00,640 was to help you clean the contact list 212 00:12:00,640 --> 00:12:04,480 because, you know, I'm sure that, 213 00:12:04,900 --> 00:12:06,360 I don't know how is your contact list, 214 00:12:06,360 --> 00:12:14,080 but mine is full of dead people, dead endpaps, 215 00:12:14,340 --> 00:12:18,380 because, you know, my public is very old, 216 00:12:18,860 --> 00:12:21,180 It's three years old. 217 00:12:22,980 --> 00:12:30,140 There is a lot of people that I have been following that they are not in NOSR anymore. 218 00:12:31,140 --> 00:12:33,000 I have thousands of contacts. 219 00:12:34,940 --> 00:12:44,700 That's ensure that I'm putting a lot of pressure in social clients because every time I enter in a social client, they have to pull all the information from that. 220 00:12:45,360 --> 00:12:48,840 And there are thousands of users of MPUBs. 221 00:12:48,860 --> 00:12:56,220 my contact list so the idea behind colorillo was create a tool to help you navigate that contact 222 00:12:56,220 --> 00:13:05,340 list and clean clean clean it up from all contacts from your own criteria like you use 223 00:13:06,140 --> 00:13:08,380 the idea that is that you was able to 224 00:13:10,700 --> 00:13:18,140 select or create your criteria to determine what are dead endpaps and whatnot and filter them out 225 00:13:18,860 --> 00:13:19,600 easily. 226 00:13:22,300 --> 00:13:23,040 So yeah, 227 00:13:23,140 --> 00:13:24,860 I think that I'm going to 228 00:13:24,860 --> 00:13:26,920 mix both. I'm going to mix 229 00:13:26,920 --> 00:13:28,860 Nostry and the idea behind 230 00:13:28,860 --> 00:13:31,120 Colorillo and extend 231 00:13:31,120 --> 00:13:32,700 it through Web of Trust. 232 00:13:32,800 --> 00:13:34,800 That is something that I'm working on. 233 00:13:35,780 --> 00:13:36,680 So that was, 234 00:13:36,820 --> 00:13:38,940 I was actually going to say, when we get to the Web of 235 00:13:38,940 --> 00:13:41,000 Trust part, this to 236 00:13:41,000 --> 00:13:43,120 me sounds like a pretty good use case, actually. 237 00:13:43,540 --> 00:13:45,140 But also, do you 238 00:13:45,140 --> 00:13:46,820 know the Daniel on Nostry? 239 00:13:47,820 --> 00:13:48,300 Yeah. 240 00:13:48,860 --> 00:13:59,610 He I think created an app a couple of months back called Something Zombies right And the idea is very similar Yeah yeah yeah I really like the idea 241 00:14:00,610 --> 00:14:04,530 I think that's Zombies versus, I don't know. 242 00:14:04,610 --> 00:14:06,090 I don't remember the name exactly. 243 00:14:06,270 --> 00:14:07,110 I don't want to be wrong. 244 00:14:07,750 --> 00:14:10,450 But yeah, I know what app you're talking about. 245 00:14:10,550 --> 00:14:11,310 It's very fun. 246 00:14:11,830 --> 00:14:12,650 And I really like it. 247 00:14:12,890 --> 00:14:14,930 So yeah, it's a very similar concept. 248 00:14:15,550 --> 00:14:17,890 I don't know exactly what is the criteria behind. 249 00:14:18,310 --> 00:14:21,450 the Daniels app, but it's similar. 250 00:14:21,590 --> 00:14:24,630 It's basically helping you to clean the contact list. 251 00:14:27,930 --> 00:14:33,610 And that's an important point for a lot of non-technical users 252 00:14:33,610 --> 00:14:38,250 to be aware of, which is every time your contact, 253 00:14:38,790 --> 00:14:40,470 and you mentioned this, right? 254 00:14:41,890 --> 00:14:44,470 Well, first of all, it's nice to have a clean contact list. 255 00:14:44,550 --> 00:14:47,530 Why do you want to have dead zombie endpubs in there, 256 00:14:47,530 --> 00:14:49,850 like people who abandoned the protocol and have disappeared. 257 00:14:50,850 --> 00:14:55,110 But the other thing is you put a load on a relay every single time. 258 00:14:55,330 --> 00:14:56,530 That thing loads, right? 259 00:14:56,890 --> 00:14:57,150 Yeah. 260 00:14:57,950 --> 00:14:58,290 Oh, yeah. 261 00:14:59,130 --> 00:15:02,170 And it's also a kind of a fingerprint, you know. 262 00:15:03,930 --> 00:15:08,870 Your social graph is somehow a fingerprint of yours, 263 00:15:09,570 --> 00:15:10,390 a social fingerprint. 264 00:15:10,390 --> 00:15:17,390 That can be some privacy concerns there as well. 265 00:15:17,530 --> 00:15:26,850 But yeah, the main case for me to clean your contact list is to keep it healthy, because 266 00:15:26,850 --> 00:15:33,810 the contact list is in Noster now more than else with the, you know, web of trust and 267 00:15:33,810 --> 00:15:38,210 rankings and trust desertions and so on. 268 00:15:38,210 --> 00:15:43,050 That contact list is being used for different stuff. 269 00:15:43,050 --> 00:15:50,550 it clean is basically helped Nasser to be more healthy to have better 270 00:15:50,550 --> 00:15:58,170 calculations or things because that data is used as a source of 271 00:15:58,170 --> 00:16:04,290 truth for these computations so it really makes sense to have it clean so 272 00:16:04,290 --> 00:16:10,170 having tools to clean it to clean that is something important in my 273 00:16:10,170 --> 00:16:10,570 point of view. 274 00:16:14,050 --> 00:16:16,010 So, Coloreo is 275 00:16:16,010 --> 00:16:18,170 waiting for web of trust 276 00:16:18,170 --> 00:16:18,690 integration. 277 00:16:19,790 --> 00:16:22,190 So, we'll get to that when we talk about Relator 278 00:16:22,190 --> 00:16:23,410 and the work you're doing there. 279 00:16:23,710 --> 00:16:25,210 So, what happened after Coloreo? 280 00:16:27,110 --> 00:16:27,850 I mean, 281 00:16:28,290 --> 00:16:29,850 a lot of things happened 282 00:16:29,850 --> 00:16:32,170 after and before 283 00:16:32,170 --> 00:16:33,250 Coloreo. But 284 00:16:33,250 --> 00:16:36,130 at that time, I was already participating 285 00:16:36,130 --> 00:16:37,530 in sovereign engineering. 286 00:16:37,530 --> 00:16:47,670 and I was part of the first, the participants of the first cohort, was lucky enough to be part of 287 00:16:47,670 --> 00:16:55,170 that and it was magical to be honest, kind of turning point to meet all that talented people 288 00:16:55,170 --> 00:17:01,190 and share a space with them and share discussions and ponchas of course. 289 00:17:01,190 --> 00:17:05,710 so yeah I have been working 290 00:17:05,710 --> 00:17:07,650 a lot in Nosser 291 00:17:07,650 --> 00:17:08,610 in different projects 292 00:17:08,610 --> 00:17:11,210 all of them are public and open source 293 00:17:11,210 --> 00:17:13,810 I have been 294 00:17:13,810 --> 00:17:15,570 helping and 295 00:17:15,570 --> 00:17:17,410 contributing to different projects 296 00:17:17,410 --> 00:17:19,230 in the Nosser ecosystem 297 00:17:19,230 --> 00:17:21,590 I was working 298 00:17:21,590 --> 00:17:21,950 for 299 00:17:21,950 --> 00:17:24,130 I was working with 300 00:17:24,130 --> 00:17:27,590 guys from Preveo Markets at some point 301 00:17:27,590 --> 00:17:29,750 also 302 00:17:29,750 --> 00:17:31,510 with other people. 303 00:17:32,610 --> 00:17:33,570 Now I'm 304 00:17:33,570 --> 00:17:36,050 contributing to Marmot ES 305 00:17:36,050 --> 00:17:37,630 as well. There is 306 00:17:37,630 --> 00:17:39,550 Marmot protocol, 307 00:17:40,070 --> 00:17:41,870 MLS and white noise. 308 00:17:43,110 --> 00:17:43,830 We are doing 309 00:17:43,830 --> 00:17:46,050 this TypeScript 310 00:17:46,050 --> 00:17:48,090 implementation of Marmot 311 00:17:48,090 --> 00:17:50,030 protocol, which is 312 00:17:50,030 --> 00:17:50,490 very cool. 313 00:17:51,290 --> 00:17:53,950 Soon we'll flourish. 314 00:17:56,190 --> 00:17:56,790 Excellent. 315 00:17:56,990 --> 00:17:58,910 I think a lot of us 316 00:17:58,910 --> 00:18:04,070 are waiting with bated breath for Marmot to manifest itself 317 00:18:04,070 --> 00:18:06,870 so that it can finally stop using Signal and Telegram 318 00:18:06,870 --> 00:18:09,650 and move to something Nostro native. 319 00:18:10,570 --> 00:18:11,970 Yeah, it's coming along pretty nicely. 320 00:18:13,510 --> 00:18:19,730 We already have working demos for basically a chat application, 321 00:18:20,290 --> 00:18:21,670 everything running in the browser, 322 00:18:22,470 --> 00:18:26,490 implemented with our Marmot TS implementation. 323 00:18:26,490 --> 00:18:30,010 it's pretty nice to see 324 00:18:30,010 --> 00:18:30,750 to be honest 325 00:18:30,750 --> 00:18:33,510 I think that that will unlock 326 00:18:33,510 --> 00:18:35,210 a lot of things 327 00:18:35,210 --> 00:18:36,890 So let me 328 00:18:36,890 --> 00:18:39,670 I should probably get Jeff G at some point 329 00:18:39,670 --> 00:18:41,110 to talk to him about this 330 00:18:41,110 --> 00:18:42,410 but from your understanding 331 00:18:42,410 --> 00:18:45,950 obviously White Noise is 332 00:18:45,950 --> 00:18:48,070 a reference implementation of the Mahmoud 333 00:18:48,070 --> 00:18:49,570 protocol but there will be others 334 00:18:49,570 --> 00:18:52,090 like the TypeScript one that you're working on 335 00:18:52,090 --> 00:18:54,050 or maybe you're working on libraries 336 00:18:54,050 --> 00:18:55,630 someone's going to take your TypeScript libraries 337 00:18:55,630 --> 00:19:00,590 and actually build another reference app on the Marmot protocol, right? 338 00:19:02,310 --> 00:19:08,510 Will messages across these implementations of Marmot be interoperable 339 00:19:08,510 --> 00:19:09,890 or will they be siloed? 340 00:19:11,390 --> 00:19:14,930 Yeah, yeah, they will be interoperable totally. 341 00:19:15,730 --> 00:19:16,950 I mean, they should be, definitely. 342 00:19:19,090 --> 00:19:21,530 MLS is a quite complex protocol. 343 00:19:22,730 --> 00:19:24,970 It's very cool, but quite complex. 344 00:19:25,630 --> 00:19:33,990 and this complexity means that there are subtle details that you need to take care about 345 00:19:33,990 --> 00:19:36,710 to basically be interpretable 346 00:19:36,710 --> 00:19:46,010 and those little nuances are the things that are difficult to manage 347 00:19:46,010 --> 00:19:53,570 or to get fully because the RFC of MLS is a very long document 348 00:19:53,570 --> 00:19:57,270 and very difficult to transverse. 349 00:19:59,010 --> 00:20:01,030 But it's a common protocol. 350 00:20:03,110 --> 00:20:05,130 If we get it right, 351 00:20:05,270 --> 00:20:07,150 and I think that we are doing it right, 352 00:20:07,890 --> 00:20:10,350 definitely it should be interoperable. 353 00:20:11,070 --> 00:20:12,250 It will be interoperable, yes. 354 00:20:13,050 --> 00:20:15,410 And I understand it in theory, right, Jesus? 355 00:20:15,410 --> 00:20:17,170 But the reason I'm asking is, 356 00:20:18,270 --> 00:20:20,690 DMing is so broken on NOSTA right now. 357 00:20:20,750 --> 00:20:22,390 And I'm not just talking about NIPO 4. 358 00:20:22,390 --> 00:20:28,530 right is there's there's some clients that implement gift wrap their nip 17 the multi 359 00:20:28,530 --> 00:20:35,150 malmi uh bless him is working on on double ratchet for iris yeah and now there's going to be these 360 00:20:35,150 --> 00:20:41,590 marmot clients so like so someone obviously so for folks i'm sure everyone knows as well just 361 00:20:41,590 --> 00:20:46,570 it's worth saying if you're using a nostra client that only supports nip or four and someone sends 362 00:20:46,570 --> 00:20:51,910 you a nip 17 dm you're not going to see it right and the other way around i mean typically 363 00:20:51,910 --> 00:20:56,050 yes, a lot of the NIP 17 clients have just said, you know what, I'm not even going to deal with 364 00:20:56,050 --> 00:21:03,550 NIP 4. So if you, there's no communication. So if some, for example, someone's iOS user and you're 365 00:21:03,550 --> 00:21:11,830 using Nostor, I believe Nostor has moved almost fully to NIP 17. I don't know if they've still, 366 00:21:12,130 --> 00:21:16,730 Fabian has still kept the support for NIP 4. But let's just say they've moved fully to NIP 17. 367 00:21:16,730 --> 00:21:22,030 If I send them a NIPA 4 DM from Primal or from Darmus, 368 00:21:22,150 --> 00:21:23,690 they're not going to see it, right? 369 00:21:23,830 --> 00:21:25,590 So it's just so broken. 370 00:21:26,010 --> 00:21:27,790 There's no interoperability in DMs. 371 00:21:28,010 --> 00:21:30,410 And there are all of these different implementations coming out. 372 00:21:30,650 --> 00:21:33,150 So that's kind of where my question is coming from, 373 00:21:33,230 --> 00:21:39,090 given how complex Marmot is, or the MLS is, rather, right? 374 00:21:40,130 --> 00:21:44,690 So I think if someone can solve that problem properly and cleanly, 375 00:21:44,690 --> 00:21:45,570 it'll be amazing. 376 00:21:45,570 --> 00:21:54,130 yeah yeah definitely um you know i have a very controversial opinion about dm and on nocer 377 00:21:54,130 --> 00:22:01,670 because definitely nipo 4 is something very bad for your privacy but it's the the communication 378 00:22:01,670 --> 00:22:08,190 channel of last resort because everyone is going to be able to see it and even if even if you are 379 00:22:08,190 --> 00:22:16,530 and someone could carry your message to the receiver. 380 00:22:18,190 --> 00:22:22,910 Even that, you have very bad privacy, but it works reliably. 381 00:22:23,870 --> 00:22:28,570 It's something that we should not use for regular communication, 382 00:22:29,030 --> 00:22:36,650 but I think that is a very interesting perspective to have about Nipo4 383 00:22:36,650 --> 00:22:38,790 that is the last resort communication. 384 00:22:39,590 --> 00:22:40,830 It's very simple. 385 00:22:41,650 --> 00:22:42,490 It works. 386 00:22:42,990 --> 00:22:45,490 It leaks metadata, for sure. 387 00:22:46,250 --> 00:22:47,430 But it works. 388 00:22:47,990 --> 00:22:51,010 But ahead of that controversial take, 389 00:22:52,510 --> 00:22:54,350 what we are doing in Marmot at the end 390 00:22:54,350 --> 00:22:56,630 is creating high-level libraries. 391 00:22:57,410 --> 00:22:59,610 So developers doesn't need to worry about 392 00:22:59,610 --> 00:23:03,410 the low-level details of the protocol of MLS. 393 00:23:03,610 --> 00:23:04,490 That is very complex. 394 00:23:06,650 --> 00:23:16,410 They can use built on top of the high level libraries that they care about, or the lower level details. 395 00:23:17,410 --> 00:23:28,130 I hope that someday we have a consistent environment for the DMs of Nosr, 396 00:23:28,130 --> 00:23:31,890 because I think that there is a very good use case. 397 00:23:31,890 --> 00:23:44,050 I don't know if the answer is to put things on relays, because that intrinsically means that the data in some way is public. 398 00:23:45,430 --> 00:23:55,730 So maybe the solution is to, I don't know, use ephemeral events, use relays as a transport, but not relying on them for storage. 399 00:23:55,730 --> 00:23:57,950 and in some way 400 00:23:57,950 --> 00:23:59,830 MLS or Marmot 401 00:23:59,830 --> 00:24:01,830 it's like that 402 00:24:01,830 --> 00:24:03,670 because a lot of the thing 403 00:24:03,670 --> 00:24:06,170 that allows users to use 404 00:24:06,170 --> 00:24:08,650 MLS or Marmot 405 00:24:08,650 --> 00:24:10,510 protocol happens client 406 00:24:10,510 --> 00:24:10,890 side 407 00:24:10,890 --> 00:24:13,470 what 408 00:24:13,470 --> 00:24:16,350 is transmitted by relays are 409 00:24:16,350 --> 00:24:18,830 very cryptic 410 00:24:18,830 --> 00:24:20,530 blocks of encrypted 411 00:24:20,530 --> 00:24:21,010 data 412 00:24:21,010 --> 00:24:24,750 which doesn't mean nothing if the client 413 00:24:24,750 --> 00:24:29,530 if the user client doesn't have the correct data. 414 00:24:30,070 --> 00:24:32,510 So there is a lot going on client-side. 415 00:24:33,630 --> 00:24:37,450 What happens on the relays is very cryptic. 416 00:24:39,670 --> 00:24:41,150 So yeah, I don't know. 417 00:24:42,210 --> 00:24:43,490 Definitely it's a good solution. 418 00:24:45,290 --> 00:24:47,930 But yeah, let's see how it plays out. 419 00:24:50,330 --> 00:24:50,970 Yeah. 420 00:24:50,970 --> 00:24:56,870 All right, so let's get to, let's fast forward. 421 00:24:57,870 --> 00:25:00,210 You were part of the first Sorbonne Engineering Cohort. 422 00:25:00,450 --> 00:25:01,150 So when was that? 423 00:25:01,250 --> 00:25:03,110 Like late 2023, early 2024? 424 00:25:03,550 --> 00:25:07,110 I think that was the one that you had that huge presentation 425 00:25:07,110 --> 00:25:08,590 at Bitcoin Atlantis, right? 426 00:25:09,010 --> 00:25:11,590 Like it was supposed to be half an hour, one hour presentation 427 00:25:11,590 --> 00:25:13,670 and it ended up being like four hours or something. 428 00:25:14,630 --> 00:25:15,750 Yeah, yeah, yeah. 429 00:25:15,750 --> 00:25:16,390 That was really fun. 430 00:25:17,230 --> 00:25:17,910 That was really fun. 431 00:25:18,710 --> 00:25:20,610 That room was, I tried to get in. 432 00:25:20,970 --> 00:25:24,050 I couldn't get in because there was like people spilling out of that room. 433 00:25:25,270 --> 00:25:25,790 Nice. 434 00:25:26,410 --> 00:25:27,510 Yeah, yeah, yeah. 435 00:25:27,550 --> 00:25:28,570 That was really good. 436 00:25:29,910 --> 00:25:35,090 I mean, the experience in sovereign engineering, the first one and the rest of them, 437 00:25:35,210 --> 00:25:37,610 I have been participating on almost all of them. 438 00:25:37,610 --> 00:25:47,050 But the first one was kind of special because of the Bitcoin Atlantis happening just straight after the cohort. 439 00:25:47,050 --> 00:25:49,090 so it was 440 00:25:49,090 --> 00:25:50,610 really magical 441 00:25:50,610 --> 00:25:53,570 to basically end the cohort 442 00:25:53,570 --> 00:25:55,670 with a bunch of 443 00:25:55,670 --> 00:25:57,550 new ideas, with a bunch of 444 00:25:57,550 --> 00:25:58,250 new friends 445 00:25:58,250 --> 00:26:01,730 and be able 446 00:26:01,730 --> 00:26:03,010 to present all of that 447 00:26:03,010 --> 00:26:05,390 in front of people that 448 00:26:05,390 --> 00:26:07,530 cares about the thing that you are 449 00:26:07,530 --> 00:26:07,850 building 450 00:26:07,850 --> 00:26:11,470 and yeah that was really nice 451 00:26:11,470 --> 00:26:13,570 I will have good 452 00:26:13,570 --> 00:26:14,450 experiences about 453 00:26:14,450 --> 00:26:15,890 Sovereign and Geren 454 00:26:15,890 --> 00:26:18,290 and recommend everyone to 455 00:26:18,290 --> 00:26:19,830 I mean every engineer 456 00:26:19,830 --> 00:26:22,430 interested in us to participate there 457 00:26:22,430 --> 00:26:24,230 because it's a good experience 458 00:26:24,230 --> 00:26:26,870 And we're only seeing the benefits 459 00:26:26,870 --> 00:26:28,510 of some of the work that went 460 00:26:28,510 --> 00:26:30,550 there now, when I say we 461 00:26:30,550 --> 00:26:32,450 I mean as just regular users 462 00:26:32,450 --> 00:26:34,790 the people, right, like stuff like Blossom 463 00:26:34,790 --> 00:26:37,110 came out of, it might have come out of the first one 464 00:26:37,110 --> 00:26:38,830 right Jesus, or was it the second 465 00:26:38,830 --> 00:26:39,150 Yeah 466 00:26:39,150 --> 00:26:41,950 Yeah, yeah, first one 467 00:26:41,950 --> 00:26:44,390 Yeah, so Blossom came out of that 468 00:26:44,390 --> 00:26:48,050 other things which I'm forgetting right now. 469 00:26:48,250 --> 00:26:51,430 And we're beginning to experience the benefits now 470 00:26:51,430 --> 00:26:54,730 of all of those things a year and a half later. 471 00:26:55,210 --> 00:26:56,950 So that's really cool stuff. 472 00:26:57,510 --> 00:27:00,390 For sure, people should check out Sovereign Engineering. 473 00:27:00,390 --> 00:27:05,410 Shout out Andre Loja as well from Free Madeira, Bitcoin Atlantis. 474 00:27:05,770 --> 00:27:10,930 He's one of the unspoken heroes of that, of Sovereign Engineering. 475 00:27:12,090 --> 00:27:13,210 Yeah, yeah, yeah. 476 00:27:13,210 --> 00:27:23,130 I mean everything happens in his co-work. He's a very nice guy. Yeah, I mean he's really nice. 477 00:27:27,850 --> 00:27:39,370 Free Madeira is a very cool group of people and they do meetups. There is also the people 478 00:27:39,370 --> 00:27:50,620 from the bitcoin work there in madeira which is really nice there is a good community there Yeah for sure 479 00:27:53,020 --> 00:27:56,120 Let's fast forward a little bit by a few months. 480 00:27:56,320 --> 00:27:58,100 Bitcoin Atlantis was March of 2024. 481 00:27:58,460 --> 00:28:02,200 Let's get to September, October of 2024, 482 00:28:02,540 --> 00:28:05,080 when really, to me, 483 00:28:05,420 --> 00:28:09,920 a lot of the AI acceleration became apparent to people. 484 00:28:09,920 --> 00:28:13,740 Before that, people were using ChatGPT for silly things. 485 00:28:14,060 --> 00:28:16,480 But then the reasoning models came out. 486 00:28:16,600 --> 00:28:17,980 Then MCP servers came out. 487 00:28:18,000 --> 00:28:21,020 All of that was happening in late 2024, right? 488 00:28:21,480 --> 00:28:27,140 So I think early 2025, you launched DVMCP, correct? 489 00:28:28,340 --> 00:28:28,680 Yeah. 490 00:28:29,060 --> 00:28:34,720 I think that's a week or two before servering Engineering 3. 491 00:28:34,720 --> 00:28:38,160 because I remember to 492 00:28:38,160 --> 00:28:40,320 to see 493 00:28:40,320 --> 00:28:42,560 the specification of MCP 494 00:28:42,560 --> 00:28:44,700 because 495 00:28:44,700 --> 00:28:46,460 that came out as you said 496 00:28:46,460 --> 00:28:48,080 late 2024 497 00:28:48,080 --> 00:28:50,740 the first specification of MCP 498 00:28:50,740 --> 00:28:52,840 and then I 499 00:28:52,840 --> 00:28:53,940 got it 500 00:28:53,940 --> 00:28:55,800 used earlier 501 00:28:55,800 --> 00:28:57,140 2025 502 00:28:57,140 --> 00:29:00,580 and started with the specification 503 00:29:00,580 --> 00:29:02,100 and that was used 504 00:29:02,100 --> 00:29:03,840 some weeks before 505 00:29:03,840 --> 00:29:05,660 Sovereign Engineering 3 start. 506 00:29:07,500 --> 00:29:10,100 I was reading the spec, you know, 507 00:29:10,220 --> 00:29:14,960 and I was realizing, okay, this is JSON RPC. 508 00:29:16,040 --> 00:29:17,760 All the communication is JSON RPC. 509 00:29:18,120 --> 00:29:19,180 Everything is JSON. 510 00:29:19,780 --> 00:29:21,060 JSON schemas. 511 00:29:21,700 --> 00:29:22,280 Wait a second. 512 00:29:22,500 --> 00:29:25,460 I know a protocol that is JSON-based 513 00:29:25,460 --> 00:29:27,140 and is transmitted by the race. 514 00:29:28,100 --> 00:29:32,800 And, you know, then I have that spark in my head 515 00:29:32,800 --> 00:29:36,340 and I said, okay, this is an interesting way to, 516 00:29:37,760 --> 00:29:40,420 it's something interesting to bridge, 517 00:29:40,700 --> 00:29:41,960 basically, NCP and Nasser. 518 00:29:43,360 --> 00:29:44,540 And yeah, I started to work. 519 00:29:45,060 --> 00:29:46,860 First iteration was called DBNCP. 520 00:29:49,340 --> 00:29:53,120 Yeah, then I moved to Context VM, 521 00:29:53,560 --> 00:29:55,240 the rebranding for different reasons. 522 00:29:55,460 --> 00:29:56,600 We can talk about that. 523 00:29:57,540 --> 00:29:59,820 But yeah, that was more or less a timeline. 524 00:29:59,820 --> 00:30:07,400 And maybe let's talk about ContextVM, and then we'll get to Relator right after that, Jesus. 525 00:30:08,420 --> 00:30:14,960 For someone who's listening, and maybe they've seen some of your posts, but they haven't actually sat down and used ContextVM, 526 00:30:16,360 --> 00:30:25,060 what exactly is it? How can, well, first of all, is it a tool for regular users, or do you need to be a technical person to use it? 527 00:30:25,060 --> 00:30:36,660 I mean, context-vm itself is a protocol, basically put together in OCR and MCP. 528 00:30:36,660 --> 00:30:49,580 A bit of history about this and why I moved from dbmcp to context-vm is that at the beginning 529 00:30:49,580 --> 00:30:52,620 when MCP was something unexplored, 530 00:30:53,660 --> 00:30:56,440 I wanted to learn about the protocol, 531 00:30:56,440 --> 00:30:59,240 you know, about the insights of the protocol, 532 00:30:59,420 --> 00:31:01,860 about the weeds, getting to the weeds of it. 533 00:31:02,840 --> 00:31:07,580 And I developed DB MCP as a kind of following 534 00:31:07,580 --> 00:31:10,160 a kitchen sink approach, you know, 535 00:31:10,280 --> 00:31:12,880 like re-implementing big parts of everything 536 00:31:12,880 --> 00:31:15,680 in order to learn. 537 00:31:15,680 --> 00:31:18,880 then 538 00:31:18,880 --> 00:31:21,580 after working with 539 00:31:21,580 --> 00:31:22,520 MCP 540 00:31:22,520 --> 00:31:25,840 for quite a while in DBNCP 541 00:31:25,840 --> 00:31:27,560 I realized that 542 00:31:27,560 --> 00:31:29,520 you know 543 00:31:29,520 --> 00:31:31,680 the best place for no service 544 00:31:31,680 --> 00:31:33,940 is to sit at the transport 545 00:31:33,940 --> 00:31:35,840 level of MCP 546 00:31:35,840 --> 00:31:37,560 and just worry 547 00:31:37,560 --> 00:31:39,940 about carrying messages 548 00:31:39,940 --> 00:31:42,000 through the channel 549 00:31:42,000 --> 00:31:43,100 to the transport layer 550 00:31:43,100 --> 00:31:45,200 which in this case is relays 551 00:31:45,200 --> 00:31:50,580 and it used basically a public key cryptography 552 00:31:50,580 --> 00:31:52,920 of Nocer for authentication, authorization. 553 00:31:56,240 --> 00:31:59,580 And basically, when I realized that, 554 00:32:00,220 --> 00:32:04,500 when I realized what is the best design 555 00:32:04,500 --> 00:32:06,940 or the better design for this, 556 00:32:07,300 --> 00:32:09,380 I moved to ContextVM. 557 00:32:09,540 --> 00:32:10,840 I rebranded to ContextVM. 558 00:32:11,480 --> 00:32:12,720 That was one of the reasons. 559 00:32:12,720 --> 00:32:23,540 other of the reasons that dbmcp started following a more dbm-ish approach or you know where dbms 560 00:32:23,540 --> 00:32:32,460 have this job kinds for for sending requests a different job kind for for receiving the request 561 00:32:32,460 --> 00:32:35,000 and another kind for feedback. 562 00:32:36,100 --> 00:32:40,580 So you have three kinds for a normal operation 563 00:32:40,580 --> 00:32:47,640 and you also have segregated kinds for jobs. 564 00:32:48,580 --> 00:32:51,320 So for example, if you want algorithmic feeds, 565 00:32:52,000 --> 00:32:55,240 you have a kind, but if you want a translation, 566 00:32:55,780 --> 00:32:59,460 you have a different kind for the job that you are requesting. 567 00:32:59,460 --> 00:33:03,760 So when I was developing the BNCP, I followed that pattern. 568 00:33:04,740 --> 00:33:13,460 And then moving forward to Context BN, when I realized the best place for NOSER in the MCP ecosystem, 569 00:33:14,400 --> 00:33:23,080 which is sitting at the transport level, I moved also to a more minimalistic approach, 570 00:33:23,080 --> 00:33:27,240 where I moved to ephemeral events, 571 00:33:27,920 --> 00:33:30,860 which are events that are not supposed to be stored in relays. 572 00:33:31,720 --> 00:33:35,780 And also I started to use one kind, 573 00:33:36,140 --> 00:33:39,180 one single kind for all the communications, 574 00:33:39,380 --> 00:33:43,220 for requests, responses, everything in the middle. 575 00:33:43,960 --> 00:33:47,560 So basically I was treating that kind as a kind of socket, 576 00:33:47,560 --> 00:33:52,460 as a kind of pipe with relays. 577 00:33:53,080 --> 00:34:02,380 And all of that seemed enough reasons to move and to rebrand the project. 578 00:34:03,380 --> 00:34:11,640 And also, at that time, there was a project called DB-MCP as well in the MCP world 579 00:34:11,640 --> 00:34:14,820 that stands for Damn Vulnerable MCP. 580 00:34:14,820 --> 00:34:21,860 which was a kind of hack-the-box MCP 581 00:34:21,860 --> 00:34:26,780 basically to discover and to explore the security model of MCP. 582 00:34:27,660 --> 00:34:30,480 I didn't want it to be related with that. 583 00:34:32,380 --> 00:34:36,300 And different things, so I decided to move forward 584 00:34:36,300 --> 00:34:39,600 and to rebrand it as Context VM. 585 00:34:40,420 --> 00:34:42,000 And I think that was a good decision, to be honest. 586 00:34:44,820 --> 00:34:56,440 So if I wanted to use Context VM today, and let's just say I'm a semi-technical user, what type of use case could I use it for and what would my steps be? 587 00:34:58,920 --> 00:35:08,100 I mean, the cool part of this is that you can do anything you want using this server client architecture. 588 00:35:08,100 --> 00:35:13,300 because one thing that I haven't mentioned yet is that, you know, everyone, 589 00:35:13,780 --> 00:35:17,620 and this is one of the questions that I received the most, one of the most common doops, 590 00:35:19,000 --> 00:35:23,460 is that, okay, MCP is something related with IA. 591 00:35:26,300 --> 00:35:31,700 So I don't want this to play with LLMs or nothing like that. 592 00:35:31,700 --> 00:35:40,560 and I mean the thing to understand here is that MCP is a protocol for communicating 593 00:35:40,560 --> 00:35:45,160 and it has a client-server topology 594 00:35:45,160 --> 00:35:53,200 and it uses JSON-RPC as the underlying schema or shapes of the messages 595 00:35:53,200 --> 00:35:56,700 so it's really a protocol to connect things 596 00:35:56,700 --> 00:36:04,540 you know uh it used that it was created specifically for for lms uh to connect with 597 00:36:04,540 --> 00:36:11,500 external tools but what it really is is a protocol to connect tools you know or external things 598 00:36:12,140 --> 00:36:21,260 between between client and server um this is what this is why you can why it got so so hyped 599 00:36:21,260 --> 00:36:28,700 because basically everyone was able to connect their Gmail account 600 00:36:28,700 --> 00:36:34,100 to their LNMs or their calendars or their databases or... 601 00:36:34,100 --> 00:36:35,960 File system, which is a big one. 602 00:36:35,960 --> 00:36:40,620 Yeah, their file system, exactly, or domotics, you know, through NCP. 603 00:36:41,060 --> 00:36:46,980 And that really pumped the protocol 604 00:36:46,980 --> 00:36:51,480 because it was the first standard way to provide these tools to LLMs. 605 00:36:51,480 --> 00:36:56,740 Before that, we had a tool calling or function calling 606 00:36:56,740 --> 00:36:59,660 that is something similar to NCP, 607 00:37:00,000 --> 00:37:04,760 but in a more lowest-level approach. 608 00:37:07,200 --> 00:37:10,480 But yeah, basically, NCP is a protocol to connect things, 609 00:37:11,340 --> 00:37:15,540 to connect clients and servers to perform this communication. 610 00:37:16,980 --> 00:37:36,160 It's used that it's designed for LLens, which is something interesting because it comes with different fields in the schemas of tools and the server, which are thought basically to put natural language there. 611 00:37:36,160 --> 00:37:42,960 So you can put titles for the tools, descriptions. 612 00:37:44,100 --> 00:37:48,860 You have a field that is called instructions for the server. 613 00:37:49,660 --> 00:37:59,920 So it's really a mix of natural language and JSON schema for defining the shapes and the types and so on. 614 00:37:59,920 --> 00:38:04,660 which is very interesting because it creates a self-describing server 615 00:38:04,660 --> 00:38:12,040 or a self-describing service that doesn't require a standard source of documentation 616 00:38:12,040 --> 00:38:16,320 to understand how it works or what each tool does 617 00:38:16,320 --> 00:38:21,400 because it already comes with a natural language description. 618 00:38:21,840 --> 00:38:22,640 So it's very interesting. 619 00:38:22,640 --> 00:38:30,040 um yeah so has this let me ask you this uh just a specific question about this 620 00:38:30,040 --> 00:38:36,020 was one and you know when mcps came came out it was over a year ago right late 2024 621 00:38:36,020 --> 00:38:45,640 yeah the cool thing was you could then use your uh cursor or uh or even visual studio right if you 622 00:38:45,640 --> 00:38:51,240 visual studio code if you if you use some a plugin like cline you add your mcp servers then you can 623 00:38:51,240 --> 00:38:54,460 point it to a folder or a directory on your computer. 624 00:38:54,620 --> 00:38:56,200 It can look at an Excel file. 625 00:38:56,280 --> 00:38:57,100 It can do stuff for you. 626 00:38:57,120 --> 00:39:01,100 You write it in natural language, and then all of these cool things are happening, right? 627 00:39:01,120 --> 00:39:03,820 You can aggregate receipts and all of that stuff. 628 00:39:04,580 --> 00:39:09,020 One thing I haven't found, and maybe this is just I haven't searched properly, 629 00:39:09,020 --> 00:39:13,620 or I just have had bad luck in finding it, is when you're building an app, 630 00:39:14,420 --> 00:39:17,460 Vibe Coding an app, let's just say, and you deploy it somewhere, 631 00:39:17,460 --> 00:39:19,620 is there an MCP server 632 00:39:19,620 --> 00:39:21,260 or can I use context VM 633 00:39:21,260 --> 00:39:24,640 that allows me to plug directly 634 00:39:24,640 --> 00:39:25,680 into the browser console 635 00:39:25,680 --> 00:39:29,580 and look at the errors that come there? 636 00:39:29,680 --> 00:39:32,720 Because that's kind of a big deal, right? 637 00:39:33,300 --> 00:39:35,160 There's a lot of these vibe-coded apps 638 00:39:35,160 --> 00:39:36,200 when you deploy them 639 00:39:36,200 --> 00:39:37,680 like you've got to go through 640 00:39:37,680 --> 00:39:40,020 the way I do it is I go to the console 641 00:39:40,020 --> 00:39:41,080 I copy the error 642 00:39:41,080 --> 00:39:42,840 I put it back in the LLM 643 00:39:42,840 --> 00:39:44,000 and the LLM has to interpret 644 00:39:44,000 --> 00:39:45,820 and it's a complete nuisance. 645 00:39:45,820 --> 00:39:48,620 is there a way of automating that? 646 00:39:49,660 --> 00:39:52,280 Yeah, there are different MCP servers 647 00:39:52,280 --> 00:39:54,800 that give you that 648 00:39:54,800 --> 00:39:58,240 I think that the first option 649 00:39:58,240 --> 00:40:01,900 that we had for that 650 00:40:01,900 --> 00:40:05,940 was the Playwright MCP 651 00:40:05,940 --> 00:40:10,780 Playwright is a kind of framework 652 00:40:10,780 --> 00:40:12,760 to do browser automation 653 00:40:12,760 --> 00:40:16,000 through scripting. 654 00:40:17,120 --> 00:40:19,360 Basically, you can use that to 655 00:40:19,360 --> 00:40:21,600 automate your browser 656 00:40:21,600 --> 00:40:24,840 and, you know, get all the details 657 00:40:24,840 --> 00:40:29,000 from executing or from loading 658 00:40:29,000 --> 00:40:32,400 a page or an address. 659 00:40:33,340 --> 00:40:36,400 And that's already something that you can use for that 660 00:40:36,400 --> 00:40:39,940 to basically debug web applications. 661 00:40:40,760 --> 00:40:42,200 So yeah, there are solutions already. 662 00:40:42,760 --> 00:40:44,760 You don't need a context VM for that. 663 00:40:45,760 --> 00:40:52,760 The use case for context VM is basically to have a remote service or things that are running elsewhere 664 00:40:53,760 --> 00:41:00,760 and to be able to connect like you are connecting to a local server. 665 00:41:01,760 --> 00:41:09,760 Because even if it's remote, you can connect to it without any hassle or any difficult configuration. 666 00:41:09,760 --> 00:41:17,000 because everything works over Nocer and use relays as a message buses. 667 00:41:17,000 --> 00:41:25,120 So you don't need to worry about IP addresses, DNS, firewalls or nothing like that. 668 00:41:25,120 --> 00:41:33,500 You could basically spin up a context VM server from a coffee shop or something and it will 669 00:41:33,500 --> 00:41:51,090 be reachable instantly through Nocer This is quite powerful and I think that there have been a lot of different and crazy experiments around that Pete Wien that is a good friend and I also a big fan of ContextVM he did an experiment 670 00:41:51,090 --> 00:41:58,530 on basically running a ContextVM server over the WiFi of a plane. 671 00:41:58,530 --> 00:42:02,530 He was able to receive a request during the flight. 672 00:42:02,530 --> 00:42:05,030 It's pretty crazy to think about. 673 00:42:06,150 --> 00:42:08,190 Mile-high club for ContextVM. 674 00:42:10,010 --> 00:42:10,810 What's that? 675 00:42:12,090 --> 00:42:16,790 Now, the mile-high in the air 676 00:42:16,790 --> 00:42:20,550 and you're receiving ContextVM requests. 677 00:42:21,110 --> 00:42:22,030 But yeah, that's pretty cool. 678 00:42:23,130 --> 00:42:23,690 Yeah. 679 00:42:24,590 --> 00:42:29,410 I mean, you can run it in a server in your phone as well. 680 00:42:29,690 --> 00:42:31,750 You can run it in a Raspberry Pi. 681 00:42:32,530 --> 00:42:40,770 The possibilities that come with basically putting a service online without any hassle 682 00:42:40,770 --> 00:42:43,650 I think that are quite interesting 683 00:42:43,650 --> 00:42:48,770 There are different people working on projects apart of me 684 00:42:48,770 --> 00:42:53,410 that are really creating some really cool stuff 685 00:42:53,410 --> 00:42:58,690 But yeah, I don't know if I explain this properly 686 00:42:58,690 --> 00:43:03,890 please just ask if you think that I can explain it better. 687 00:43:05,110 --> 00:43:07,290 No, that's great. That's super helpful. 688 00:43:08,670 --> 00:43:10,290 So, Jesus, let's move to Relator. 689 00:43:10,950 --> 00:43:12,030 And I think that's the one, 690 00:43:13,770 --> 00:43:17,670 that's the Web of Trust solution that you rolled out. 691 00:43:18,270 --> 00:43:20,470 That was, what, a month or two ago? 692 00:43:20,950 --> 00:43:22,650 Right? Relator is pretty new, isn't it? 693 00:43:23,730 --> 00:43:24,090 Yeah. 694 00:43:24,750 --> 00:43:25,150 Yeah. 695 00:43:25,150 --> 00:43:27,530 Yeah, I think something like that. 696 00:43:27,530 --> 00:43:32,030 I don't remember the exact dates, but yeah, a month or two. 697 00:43:34,290 --> 00:43:41,890 Yeah, Relator was the result of different conversations that we had during Sovereign Engineering 5, 698 00:43:43,050 --> 00:43:50,250 especially around conversations with Gigi and some of them with Hazard as well. 699 00:43:50,250 --> 00:43:53,590 you know, it came from 700 00:43:53,590 --> 00:43:55,450 a natural frustration 701 00:43:55,450 --> 00:43:56,070 because 702 00:43:56,070 --> 00:43:59,350 in Nosser, currently 703 00:43:59,350 --> 00:44:01,510 we don't have an open source 704 00:44:01,510 --> 00:44:03,230 solution to do 705 00:44:03,230 --> 00:44:05,250 what Vertex 706 00:44:05,250 --> 00:44:06,990 started to do, basically. 707 00:44:07,750 --> 00:44:09,330 And the frustration was to 708 00:44:09,330 --> 00:44:10,930 basically have something open source 709 00:44:10,930 --> 00:44:13,330 that you can run and that you can understand 710 00:44:13,330 --> 00:44:14,130 and reason about 711 00:44:14,130 --> 00:44:16,910 that everyone can run 712 00:44:16,910 --> 00:44:18,430 or anyone can run. 713 00:44:18,430 --> 00:44:21,030 and it offers 714 00:44:21,030 --> 00:44:23,610 same kind of service 715 00:44:23,610 --> 00:44:24,650 that Vertex 716 00:44:24,650 --> 00:44:27,590 which is a pretty good service 717 00:44:27,590 --> 00:44:29,290 shout out to Pip 718 00:44:29,290 --> 00:44:32,530 but yeah 719 00:44:32,530 --> 00:44:35,650 it's something close 720 00:44:35,650 --> 00:44:37,770 and the idea 721 00:44:37,770 --> 00:44:39,710 was basically put together 722 00:44:39,710 --> 00:44:41,050 something open source 723 00:44:41,050 --> 00:44:43,230 fully self-hostable 724 00:44:43,230 --> 00:44:46,810 with the idea of 725 00:44:46,810 --> 00:44:52,470 that it could run anywhere in the computer of a pleb. 726 00:44:52,470 --> 00:44:58,610 You don't need a very expensive machine to run this, 727 00:45:00,130 --> 00:45:03,750 but you can run it in a Raspberry Pi even if you want. 728 00:45:04,550 --> 00:45:09,630 I mean, the public instance of Relator that we are running 729 00:45:09,630 --> 00:45:13,530 which receives a lot of requests 730 00:45:13,530 --> 00:45:17,250 because I integrated with Waterlight. 731 00:45:17,690 --> 00:45:21,810 There's another project that came from this exploration. 732 00:45:22,770 --> 00:45:27,590 But basically, my instance of Relator is receiving a lot of requests 733 00:45:27,590 --> 00:45:32,110 and it's running in a very cheap $2 VPS. 734 00:45:33,630 --> 00:45:36,630 It's still performing very well because, you know, 735 00:45:36,630 --> 00:45:43,170 it has response times of 50 milliseconds or 40 milliseconds for search. 736 00:45:43,530 --> 00:45:46,410 it's behaving pretty well, 737 00:45:47,030 --> 00:45:49,450 it's quite performant and efficient. 738 00:45:50,270 --> 00:45:53,750 So it can be run anywhere. 739 00:45:55,050 --> 00:45:58,510 That was the idea behind Redator, create this. 740 00:46:00,190 --> 00:46:02,710 So actually, let's talk about that, 741 00:46:02,950 --> 00:46:04,770 the compute requirement, right? 742 00:46:04,770 --> 00:46:08,290 So you mentioned Vertex, which uses personalized page rank, 743 00:46:09,070 --> 00:46:13,010 which is like a take on Google's original search algorithm, 744 00:46:13,010 --> 00:46:20,870 right page rank the other web of trust service provider is uh is brainstorm which which is 745 00:46:20,870 --> 00:46:27,770 nosfabrica is developing that one uh it looks at follows meets today anyway it's extensible but 746 00:46:27,770 --> 00:46:33,210 today just looks at follows meets and reports you can obviously add other signals like zaps and so 747 00:46:33,210 --> 00:46:38,010 on if you want to looks at those it comes up with scores completely personalized and then publishes 748 00:46:38,010 --> 00:46:43,590 them using trusted assertions and then there's relator what is the algorithm so well actually 749 00:46:43,590 --> 00:46:50,070 before you answer that both in the case of personalized page rank and in the case of 750 00:46:50,070 --> 00:46:58,030 brainstorm which is which produces grape rank you know it's computationally incredibly intensive 751 00:46:58,030 --> 00:47:03,410 because you have to go through this entire graph database right and crunch through all of these 752 00:47:03,410 --> 00:47:06,090 numbers and it's 753 00:47:06,090 --> 00:47:08,030 tricky to run those computations 754 00:47:08,030 --> 00:47:10,050 on a Raspberry Pi for instance 755 00:47:10,050 --> 00:47:11,830 so what is the algorithm 756 00:47:11,830 --> 00:47:14,070 in the case of 757 00:47:14,070 --> 00:47:16,090 Relator and how are you able to get it to 758 00:47:16,090 --> 00:47:16,450 run 759 00:47:16,450 --> 00:47:19,990 in a low compute 760 00:47:19,990 --> 00:47:20,430 environment 761 00:47:20,430 --> 00:47:23,170 yes so 762 00:47:23,170 --> 00:47:24,270 basically 763 00:47:24,270 --> 00:47:27,010 yeah I think that 764 00:47:27,010 --> 00:47:29,830 the algorithm behind Relator is 765 00:47:29,830 --> 00:47:31,510 very close in the spirit 766 00:47:31,510 --> 00:47:33,950 to grape rank 767 00:47:33,950 --> 00:47:35,990 because both 768 00:47:35,990 --> 00:47:37,890 basically use different 769 00:47:37,890 --> 00:47:40,110 signals to compose 770 00:47:40,110 --> 00:47:41,670 the thrust or the score 771 00:47:41,670 --> 00:47:45,110 but the 772 00:47:45,110 --> 00:47:47,810 algorithm is basically 773 00:47:47,810 --> 00:47:50,010 a weight formula 774 00:47:50,010 --> 00:47:51,750 of different components 775 00:47:51,750 --> 00:47:54,090 that at the end are 776 00:47:54,090 --> 00:47:56,330 crystallized 777 00:47:56,330 --> 00:47:58,170 and zoomed in a 778 00:47:58,170 --> 00:47:59,750 number between 779 00:47:59,750 --> 00:48:01,250 1 and 0 780 00:48:01,250 --> 00:48:06,850 You know, everything is a floating point, a floating point number, a float. 781 00:48:08,910 --> 00:48:16,950 And basically, it takes different signals or different validators, validation signals, 782 00:48:17,850 --> 00:48:26,250 and the perspective of a source pubkey, which basically is used as a kind of rule of thumb 783 00:48:26,250 --> 00:48:30,610 to have a component in this formula 784 00:48:30,610 --> 00:48:34,210 that weighs 50% 785 00:48:34,210 --> 00:48:40,150 in the case of the instance of Relator that we are running. 786 00:48:40,710 --> 00:48:42,030 Because the cool part of this is that 787 00:48:42,030 --> 00:48:45,830 by running Relator, anyone can tune that, 788 00:48:46,030 --> 00:48:46,930 can tune the algorithm, 789 00:48:47,190 --> 00:48:51,410 and can tune the weights that compose the algorithm. 790 00:48:52,350 --> 00:48:55,950 So basically, you have this component 791 00:48:55,950 --> 00:49:04,750 that is the primary component, which is the social graph distance between the source public key 792 00:49:05,390 --> 00:49:10,350 and the target public key that you are attempting to perform the calculation. 793 00:49:11,390 --> 00:49:19,550 And the distance between the source and the target is then translated to a float, 794 00:49:20,510 --> 00:49:27,950 basically using a decay function. So for example, if the source pub key and the target pub key 795 00:49:29,390 --> 00:49:40,750 they are one of distance, the output is one, okay, the result of that component. Then if the 796 00:49:40,750 --> 00:49:43,110 if the distance is two, 797 00:49:44,270 --> 00:49:47,710 the decay function basically map that 798 00:49:47,710 --> 00:49:50,130 to a float number as well. 799 00:49:51,190 --> 00:49:53,490 In the case of the data... 800 00:49:53,490 --> 00:49:57,090 Sorry, just so I understand the concept of distance, 801 00:49:57,410 --> 00:49:58,430 you mean hops, right? 802 00:49:58,510 --> 00:50:00,010 So if you're following someone, 803 00:50:00,150 --> 00:50:00,830 distance is one, 804 00:50:01,310 --> 00:50:03,610 a follow of a follow is a distance of two, correct? 805 00:50:04,130 --> 00:50:04,270 Okay. 806 00:50:04,670 --> 00:50:05,250 Yeah, exactly. 807 00:50:05,250 --> 00:50:07,890 Basically, distance one 808 00:50:07,890 --> 00:50:10,970 is what the source path key follows. 809 00:50:12,270 --> 00:50:20,190 And distance two is what the follows of the source follows. 810 00:50:21,130 --> 00:50:22,610 So, yeah, hops, basically. 811 00:50:23,550 --> 00:50:23,870 Yeah. 812 00:50:24,870 --> 00:50:28,310 Yeah, so basically, as I was saying, 813 00:50:28,910 --> 00:50:34,490 we have a decay function that basically maps the distance 814 00:50:34,490 --> 00:50:36,410 to a floating point, to a float. 815 00:50:36,410 --> 00:50:38,630 so if you are at distance 1 816 00:50:38,630 --> 00:50:41,230 you have 1, if you are at distance 2 817 00:50:41,230 --> 00:50:42,310 you have 818 00:50:42,310 --> 00:50:43,870 0.85 819 00:50:43,870 --> 00:50:46,550 or something around that 820 00:50:46,550 --> 00:50:48,590 if you are at distance 3 821 00:50:48,590 --> 00:50:50,810 you have 0.7 822 00:50:50,810 --> 00:50:52,250 or something like that 823 00:50:52,250 --> 00:50:53,710 and then 824 00:50:53,710 --> 00:50:56,470 Is this some Fibonacci decay 825 00:50:56,470 --> 00:50:57,210 or something? 826 00:50:57,690 --> 00:51:00,010 1 to 0.0 827 00:51:00,010 --> 00:51:02,550 Not exactly 828 00:51:02,550 --> 00:51:04,510 I mean I tried to 829 00:51:04,510 --> 00:51:14,590 model the curve of the decay function to basically decay harder when the distance is more than 830 00:51:14,590 --> 00:51:25,250 three. So ahead from three, the result of that component is bigger, start to be more 831 00:51:25,250 --> 00:51:26,650 pronunciated 832 00:51:26,650 --> 00:51:28,690 and the 833 00:51:28,690 --> 00:51:31,270 6 that is out of the graph 834 00:51:31,270 --> 00:51:33,230 and this is based on different 835 00:51:33,230 --> 00:51:34,990 studies and different papers 836 00:51:34,990 --> 00:51:37,550 So you've put 837 00:51:37,550 --> 00:51:39,570 together this decay function 838 00:51:39,570 --> 00:51:41,030 and it sounds like some kind of 839 00:51:41,030 --> 00:51:43,790 I don't think it sounds like some kind of stepwise 840 00:51:43,790 --> 00:51:44,870 function right like 841 00:51:44,870 --> 00:51:47,130 if you're saying after 3 it's more 842 00:51:47,130 --> 00:51:48,950 than it's before 3 843 00:51:48,950 --> 00:51:51,290 so it's not some kind of 844 00:51:51,290 --> 00:51:52,770 existing 845 00:51:52,770 --> 00:51:57,110 curve, right? You've looked at different studies and you've sort of pieced it together. 846 00:51:58,210 --> 00:52:01,490 No, I try to basically collapse 847 00:52:01,490 --> 00:52:05,430 there, different things about social dynamics and things 848 00:52:05,430 --> 00:52:09,170 that we can extract from the graph and map that to 849 00:52:09,170 --> 00:52:12,210 a number, to a float from 0 to 1. 850 00:52:14,350 --> 00:52:15,470 Basically because 851 00:52:15,470 --> 00:52:21,610 if you are starting to be, if you have a target pub key that is 852 00:52:21,610 --> 00:52:24,310 up to four hops 853 00:52:24,310 --> 00:52:26,010 that means that 854 00:52:26,010 --> 00:52:28,630 a lot of people 855 00:52:28,630 --> 00:52:30,810 in your graph 856 00:52:30,810 --> 00:52:31,770 in your social graph 857 00:52:31,770 --> 00:52:33,810 is not following it 858 00:52:33,810 --> 00:52:35,770 because for example 859 00:52:35,770 --> 00:52:39,390 if the source pub key 860 00:52:39,390 --> 00:52:40,750 and the target pub key 861 00:52:40,750 --> 00:52:42,270 are at one 862 00:52:42,270 --> 00:52:43,590 distance one 863 00:52:43,590 --> 00:52:46,470 the source pub key 864 00:52:46,470 --> 00:52:48,170 can control that distance 865 00:52:48,170 --> 00:52:50,450 so if the source pub key 866 00:52:50,450 --> 00:52:54,370 unfollows you, you are going to be at distance 2. 867 00:52:55,150 --> 00:52:58,450 Okay? But if you are at distance 2, 868 00:52:58,950 --> 00:53:02,590 then it's much harder to go 869 00:53:02,590 --> 00:53:06,370 from 2 to 3. Because to go from 870 00:53:06,370 --> 00:53:10,490 2 to 3, all the follows of the 871 00:53:10,490 --> 00:53:14,590 source path key should unfollow you at the same time. 872 00:53:15,570 --> 00:53:18,230 And to go from 3 to 4, 873 00:53:18,230 --> 00:53:20,410 the same but exponentially 874 00:53:20,410 --> 00:53:23,030 in the hub 875 00:53:23,030 --> 00:53:26,570 then from 4 to 5 876 00:53:26,570 --> 00:53:28,390 more exponential and 6 877 00:53:28,390 --> 00:53:30,610 is out of the graph 878 00:53:30,610 --> 00:53:32,110 is all of Noster 879 00:53:32,110 --> 00:53:33,970 yeah exactly 880 00:53:33,970 --> 00:53:35,490 yeah because 881 00:53:35,490 --> 00:53:38,370 this is basically 882 00:53:38,370 --> 00:53:40,470 the rule of thumb 883 00:53:40,470 --> 00:53:42,690 in related 884 00:53:42,690 --> 00:53:44,770 because 885 00:53:44,770 --> 00:53:45,990 basically this 886 00:53:45,990 --> 00:53:47,790 weighted formula 887 00:53:47,790 --> 00:53:57,250 take these components as a 50% of the final scoring or the final number and 888 00:53:57,250 --> 00:54:07,350 the other 50% are these trust signals that users or operators of Relator can 889 00:54:07,350 --> 00:54:16,850 modify. Basically those signals are like if the user... in the case of the 890 00:54:16,850 --> 00:54:23,410 main relator or the public relator instance that we are running. The components of the 891 00:54:23,410 --> 00:54:28,450 signals that we are taking there to do the computation is basically if the user have a 892 00:54:28,450 --> 00:54:39,730 published relay list, if the user have a valid NIPO5, if the user have reciprocity with the 893 00:54:39,730 --> 00:54:42,130 the source pub key. 894 00:54:44,410 --> 00:54:49,210 Yeah, if it has a lightning address set up in its profile. 895 00:54:51,730 --> 00:54:53,090 And yeah, that's it. 896 00:54:53,770 --> 00:54:59,970 So those signals are also mapped to a float number. 897 00:55:01,010 --> 00:55:06,250 And together with the components, the other 50% components are zoomed up 898 00:55:06,250 --> 00:55:08,970 and the result is the final score. 899 00:55:09,730 --> 00:55:12,970 So that makes complete sense, right? 900 00:55:12,970 --> 00:55:17,770 In some sense, when you map those, whatever signals you take, right? 901 00:55:17,830 --> 00:55:21,050 Like you said, the existence of a lightning address, NIPO5, 902 00:55:21,210 --> 00:55:23,250 and you could do anything, how many times they zapped. 903 00:55:23,650 --> 00:55:25,730 You can pick anything as a signal, for example. 904 00:55:25,970 --> 00:55:29,030 Yeah, you can count GNs, you know, if you want. 905 00:55:29,170 --> 00:55:31,850 That could be a good trust signal. 906 00:55:32,790 --> 00:55:33,210 Sure. 907 00:55:33,240 --> 00:55:41,940 So those have weights of their own, and for today you assign those weights, right, to these signals? 908 00:55:43,100 --> 00:55:52,160 Okay, and then those weights get multiplied with the distance weights, and it collapses into one score per relationship, right? 909 00:55:52,160 --> 00:55:55,060 So I want to make sure I understand. 910 00:55:55,820 --> 00:55:58,880 So if I'm the one who's requesting the scores, 911 00:55:59,380 --> 00:56:02,260 I have, let's say, 300 people at distance one, 912 00:56:02,760 --> 00:56:04,660 maybe 20,000 people at distance two, 913 00:56:05,000 --> 00:56:07,880 like whatever, 300,000 people at distance three. 914 00:56:08,940 --> 00:56:14,280 If I request a score, I need to get all 300,000 scores, right? 915 00:56:15,120 --> 00:56:17,660 And those scores get calculated 916 00:56:17,660 --> 00:56:21,940 by doing this weighted average across distances 917 00:56:21,940 --> 00:56:23,640 and signal weights as well. 918 00:56:24,020 --> 00:56:25,200 Yes, exactly. 919 00:56:25,500 --> 00:56:27,420 So this is what I mean. 920 00:56:27,460 --> 00:56:29,700 This is like a very complicated computation. 921 00:56:30,160 --> 00:56:31,320 How are you able to get it? 922 00:56:33,180 --> 00:56:38,000 Because what is this magic that you're able to get it to do on a Raspberry Pi? 923 00:56:39,340 --> 00:56:44,640 Yeah, I mean, I think it's not that complex at the end. 924 00:56:44,640 --> 00:57:05,840 If you are able to organize and to have a clear life cycle of the whole process and model the whole process in a way that is conservative with the resources that it consumes, the computation that it needs to develop in order to have all the computation done. 925 00:57:05,840 --> 00:57:08,980 you can get it to work 926 00:57:08,980 --> 00:57:11,300 well and perform 927 00:57:11,300 --> 00:57:13,200 well in resource 928 00:57:13,200 --> 00:57:13,800 constrained 929 00:57:13,800 --> 00:57:16,460 environments 930 00:57:16,460 --> 00:57:19,100 because 931 00:57:19,100 --> 00:57:20,080 basically 932 00:57:20,080 --> 00:57:25,260 Relator is not doing 933 00:57:25,260 --> 00:57:27,700 real-time computations all the time 934 00:57:27,700 --> 00:57:29,200 like for every request 935 00:57:29,200 --> 00:57:31,880 you have a TTL 936 00:57:31,880 --> 00:57:33,280 time to leave 937 00:57:33,280 --> 00:57:34,500 for the 938 00:57:34,500 --> 00:57:38,940 score and the result of the validations. 939 00:57:40,120 --> 00:57:45,220 That TTL determines the freshness of the output. 940 00:57:46,020 --> 00:57:49,360 So from time to time there is a 941 00:57:49,360 --> 00:57:52,880 schedule job that 942 00:57:52,880 --> 00:57:58,600 basically refresh the stale 943 00:57:58,600 --> 00:58:02,740 computations. And it does that 944 00:58:02,740 --> 00:58:10,740 in a conservative way so it really doesn't put a lot of pressure on the machine. 945 00:58:10,740 --> 00:58:19,140 The other part is the stack that I choose for this which is specifically designed for 946 00:58:19,140 --> 00:58:29,740 work on resource-constrained environments which is basically DACDB which is an embedded 947 00:58:29,740 --> 00:58:38,740 database for all app in process for specifically design for analytics that it performs really well 948 00:58:38,740 --> 00:58:46,540 on the edge you know and is embedded embedded means that you don't need to run a separate 949 00:58:46,540 --> 00:58:54,060 service a site so everything is contained in the same program the database is a file 950 00:58:54,060 --> 00:58:56,640 you can copy that file 951 00:58:56,640 --> 00:58:57,920 you can download that file 952 00:58:57,920 --> 00:58:59,280 you can transmit that file 953 00:58:59,280 --> 00:59:01,200 it's a file 954 00:59:01,200 --> 00:59:04,440 it's not like a Postgres database 955 00:59:04,440 --> 00:59:09,260 or a Neo4j database 956 00:59:09,260 --> 00:59:11,640 or a Redis 957 00:59:11,640 --> 00:59:12,740 or nothing like that 958 00:59:12,740 --> 00:59:13,780 it's embedded 959 00:59:13,780 --> 00:59:16,000 indeed 960 00:59:16,000 --> 00:59:18,640 DACDB is a very interesting database 961 00:59:18,640 --> 00:59:19,900 for this kind of things 962 00:59:19,900 --> 00:59:22,720 because they take seriously 963 00:59:22,720 --> 00:59:25,300 these things about 964 00:59:25,300 --> 00:59:25,840 because 965 00:59:25,840 --> 00:59:28,300 DuckDB is a single node 966 00:59:28,300 --> 00:59:30,700 engine which means that 967 00:59:30,700 --> 00:59:32,700 it's supposed not to be shared 968 00:59:32,700 --> 00:59:34,460 it's just to be run 969 00:59:34,460 --> 00:59:36,820 in the same thing, single node 970 00:59:36,820 --> 00:59:38,660 and they run very crazy 971 00:59:38,660 --> 00:59:40,340 benchmarks on phones 972 00:59:40,340 --> 00:59:42,580 with huge data sets and 973 00:59:42,580 --> 00:59:44,760 sorting that stuff and doing 974 00:59:44,760 --> 00:59:45,600 different operations 975 00:59:45,600 --> 00:59:48,760 they are 976 00:59:48,760 --> 00:59:50,700 very cool, they are really cool, they are fully 977 00:59:50,700 --> 00:59:51,500 open source as well 978 00:59:51,500 --> 01:00:10,060 So I'm very careful with the stack that issues for Relator and Context VM because one of the things that open source developers tend to forget is that the license of the libraries that you use determines the final license that your software can have. 979 01:00:11,500 --> 01:00:20,360 All the software that I'm using, all the libraries and everything, I'm very careful with that because I really want the stack to be fully open source. 980 01:00:20,360 --> 01:00:24,520 but yeah that is the magic behind 981 01:00:24,520 --> 01:00:26,200 Relator and why 982 01:00:26,200 --> 01:00:28,860 it works well 983 01:00:28,860 --> 01:00:30,420 in this kind of environment 984 01:00:30,420 --> 01:00:35,080 magic is .db 985 01:00:35,080 --> 01:00:36,800 it sounds like and a few other things 986 01:00:36,800 --> 01:00:37,400 so 987 01:00:37,400 --> 01:00:40,100 I mean I can 988 01:00:40,100 --> 01:00:41,760 if you are interested I think that 989 01:00:41,760 --> 01:00:44,860 I can just give a previous 990 01:00:44,860 --> 01:00:45,680 historic 991 01:00:45,680 --> 01:00:48,820 timeline of how 992 01:00:48,820 --> 01:00:51,040 Relator was developed because 993 01:00:51,040 --> 01:00:52,540 at the beginning I didn't start 994 01:00:52,540 --> 01:00:54,480 with TagTV. 995 01:00:54,780 --> 01:00:56,460 I started with SQLite 996 01:00:56,460 --> 01:00:58,340 and 997 01:00:58,340 --> 01:01:00,540 Martin Malmi's 998 01:01:00,540 --> 01:01:03,080 Nocell Social Graph library 999 01:01:03,080 --> 01:01:04,480 which 1000 01:01:04,480 --> 01:01:06,600 in Spirit is the same that 1001 01:01:06,600 --> 01:01:08,920 a library that I created for 1002 01:01:08,920 --> 01:01:11,200 Relator that is called Nocell Social 1003 01:01:11,200 --> 01:01:12,800 DAC which is 1004 01:01:12,800 --> 01:01:15,060 the same that Martin Malmi's 1005 01:01:15,060 --> 01:01:16,460 do but 1006 01:01:16,460 --> 01:01:18,680 instead of loading 1007 01:01:18,680 --> 01:01:19,800 everything on memory 1008 01:01:19,800 --> 01:01:22,420 just use 1009 01:01:22,420 --> 01:01:24,320 DACDB as the engine for 1010 01:01:24,320 --> 01:01:25,680 the graph computations 1011 01:01:25,680 --> 01:01:27,500 you know 1012 01:01:27,500 --> 01:01:29,800 everything in Related started 1013 01:01:29,800 --> 01:01:32,640 well with the Martis 1014 01:01:32,640 --> 01:01:33,860 libraries 1015 01:01:33,860 --> 01:01:36,560 and SQLite database 1016 01:01:36,560 --> 01:01:38,080 but 1017 01:01:38,080 --> 01:01:39,860 it started to show 1018 01:01:39,860 --> 01:01:40,940 the cracks 1019 01:01:40,940 --> 01:01:44,140 very quickly because 1020 01:01:44,140 --> 01:01:46,180 you know search 1021 01:01:46,180 --> 01:01:49,680 Search is a heavy topic. 1022 01:01:50,380 --> 01:01:53,640 You know, it's a very difficult task indeed 1023 01:01:53,640 --> 01:01:58,420 to basically have a huge data set of metadata 1024 01:01:58,420 --> 01:02:05,680 and just perform a search that supposedly is going to return to you 1025 01:02:05,680 --> 01:02:11,620 the most relevant results from that huge set of data. 1026 01:02:11,620 --> 01:02:18,280 and when I was the first iterations of Relator when it used the the Martis 1027 01:02:18,280 --> 01:02:27,160 Marmies library and SQLite I had a very huge bottleneck there because basically 1028 01:02:27,160 --> 01:02:34,960 in order to perform a search I had to first go to SQLite and do a full-text 1029 01:02:34,960 --> 01:02:40,480 search in the database in the in the table of the metadata then get an 1030 01:02:40,480 --> 01:02:43,180 an arbitrary number of candidates, 1031 01:02:44,080 --> 01:02:49,360 then those candidates resolve the score for all of them, 1032 01:02:50,420 --> 01:02:53,820 sort the results, and then slice it 1033 01:02:53,820 --> 01:02:56,940 to basically respect the limit of the filter. 1034 01:02:58,320 --> 01:03:02,060 That was painful because, you know, 1035 01:03:02,060 --> 01:03:06,580 you never was sure about the number of candidates 1036 01:03:06,580 --> 01:03:08,180 to take from the database. 1037 01:03:08,180 --> 01:03:16,860 because maybe you take a hundred but you know the the most relevant result is 1038 01:03:16,860 --> 01:03:27,320 101 you know is at 101 so it's quite arbitrary and that was creating a bad experience and a huge 1039 01:03:27,320 --> 01:03:35,660 performance bottleneck because every time you have to compute trust or the scores scores for 1040 01:03:35,660 --> 01:03:39,580 a random number or an arbitrary number of candidates 1041 01:03:39,580 --> 01:03:43,940 hoping to have the best or the most relevant results 1042 01:03:43,940 --> 01:03:47,560 in that candidate, which is arbitrary. 1043 01:03:48,100 --> 01:03:51,060 So then I moved to Nocell Social DAC. 1044 01:03:51,340 --> 01:03:55,560 That is the same concept that the Nocell Social Graph library 1045 01:03:55,560 --> 01:04:00,700 from Martin Malmys, but everything is in DACDB. 1046 01:04:00,700 --> 01:04:03,880 so I moved from SQLite 1047 01:04:03,880 --> 01:04:05,420 plus Martins Malme 1048 01:04:05,420 --> 01:04:06,860 library to 1049 01:04:06,860 --> 01:04:08,600 use DAGDB 1050 01:04:08,600 --> 01:04:10,240 and everything there 1051 01:04:10,240 --> 01:04:12,760 so it allowed me to 1052 01:04:12,760 --> 01:04:15,060 do joins 1053 01:04:15,060 --> 01:04:17,100 and perform everything 1054 01:04:17,100 --> 01:04:19,180 in SQL 1055 01:04:19,180 --> 01:04:21,520 to perform searches 1056 01:04:21,520 --> 01:04:23,200 and that really 1057 01:04:23,200 --> 01:04:25,460 improved the experience by a lot 1058 01:04:25,460 --> 01:04:26,700 and the performance because 1059 01:04:26,700 --> 01:04:28,480 now instead of 1060 01:04:28,480 --> 01:04:30,800 do the candidate things, 1061 01:04:31,580 --> 01:04:35,080 I could basically get the most relevant results 1062 01:04:35,080 --> 01:04:38,100 from the database straight out of the database 1063 01:04:38,100 --> 01:04:41,860 without needing to do all the score computation 1064 01:04:41,860 --> 01:04:43,600 after getting the candidates. 1065 01:04:45,180 --> 01:04:48,640 So yeah, all of that are the engineering decisions 1066 01:04:48,640 --> 01:04:55,640 and all of them was in order to accomplish 1067 01:04:55,640 --> 01:05:01,280 accomplish the vision behind it later of being able to run everywhere on any device. 1068 01:05:03,300 --> 01:05:03,780 That's great. 1069 01:05:04,260 --> 01:05:08,680 So we have, as of today, there's three Web of Trust service providers 1070 01:05:08,680 --> 01:05:17,060 that essentially you can, as a user, can request the scores, I guess, 1071 01:05:17,980 --> 01:05:21,740 for your Web of Trust, for your entire network. 1072 01:05:21,940 --> 01:05:24,020 Now, I think Vertex works a little differently. 1073 01:05:24,020 --> 01:05:26,720 it's an API call and I don't think 1074 01:05:26,720 --> 01:05:28,200 they put a number on a person. 1075 01:05:29,480 --> 01:05:30,300 I should talk to, 1076 01:05:30,540 --> 01:05:32,660 should have Pip on the show just so that he does 1077 01:05:32,660 --> 01:05:34,600 a deep dive on Vertex. 1078 01:05:34,920 --> 01:05:35,540 But in your case, 1079 01:05:36,660 --> 01:05:38,620 do you support 1080 01:05:38,620 --> 01:05:40,620 an API call or have you gone straight to 1081 01:05:40,620 --> 01:05:41,540 trusted assertions? 1082 01:05:43,220 --> 01:05:44,600 No, no, no. I mean, the first, 1083 01:05:44,600 --> 01:05:46,380 the main API of 1084 01:05:46,380 --> 01:05:47,640 Relator is ContextVM. 1085 01:05:49,060 --> 01:05:50,640 Everything, that was 1086 01:05:50,640 --> 01:05:51,020 the first 1087 01:05:51,020 --> 01:05:56,020 the first API, now I integrated with Trusted Assertions, 1088 01:05:59,060 --> 01:06:04,060 which basically publish events on relays above the ranks. 1089 01:06:04,960 --> 01:06:23,362 So yeah these two interfaces are complementary because you know Context VM gives you a request pattern similar to the VMs or RPC systems and that is necessary for things like a search 1090 01:06:24,022 --> 01:06:32,882 I cannot rely on a produce-consumer approach, which is basically what is NOS or the trusted 1091 01:06:32,882 --> 01:06:42,622 assertions approach to perform search because you need to basically publish all the events 1092 01:06:42,622 --> 01:06:48,702 with all the possible searches that someone could do, which is impossible and doesn't 1093 01:06:48,702 --> 01:06:49,182 make sense. 1094 01:06:49,182 --> 01:06:55,162 So for that, the request response pattern is necessary. 1095 01:06:55,162 --> 01:06:57,762 but the other 1096 01:06:57,762 --> 01:06:59,942 the produce consumer pattern 1097 01:06:59,942 --> 01:07:02,162 that fits more 1098 01:07:02,162 --> 01:07:03,102 naturally in Nosser 1099 01:07:03,102 --> 01:07:05,962 the concept behind 1100 01:07:05,962 --> 01:07:07,102 the trusted assertions 1101 01:07:07,102 --> 01:07:09,622 is complementary because 1102 01:07:09,622 --> 01:07:11,682 it allows 1103 01:07:11,682 --> 01:07:13,822 a much easier integration 1104 01:07:13,822 --> 01:07:15,202 in any client because 1105 01:07:15,202 --> 01:07:17,162 clients can use 1106 01:07:17,162 --> 01:07:19,322 create filters 1107 01:07:19,322 --> 01:07:21,662 as they create for 1108 01:07:21,662 --> 01:07:23,202 render a fit 1109 01:07:23,202 --> 01:07:25,782 but instead of getting the feed 1110 01:07:25,782 --> 01:07:28,502 they can just get the trusted assertions 1111 01:07:28,502 --> 01:07:29,402 events 1112 01:07:29,402 --> 01:07:31,762 but yeah 1113 01:07:31,762 --> 01:07:33,842 both interfaces are present in Reader 1114 01:07:33,842 --> 01:07:35,562 you can request ranks 1115 01:07:35,562 --> 01:07:38,342 or you can fetch ranks from relays 1116 01:07:38,342 --> 01:07:40,562 yeah 1117 01:07:40,562 --> 01:07:43,542 it contains both APIs 1118 01:07:43,542 --> 01:07:44,842 and you're right 1119 01:07:44,842 --> 01:07:46,762 it's different use cases 1120 01:07:46,762 --> 01:07:48,822 different complementary use cases 1121 01:07:48,822 --> 01:07:51,882 let's talk about the spam mitigation one 1122 01:07:51,882 --> 01:07:56,042 which I think trusted assertions are probably better suited for. 1123 01:07:56,502 --> 01:07:58,362 And I'll lay it out here, right? 1124 01:07:58,682 --> 01:08:04,822 So let's say I'm a user and I want to calculate the web of trust scores for my network 1125 01:08:04,822 --> 01:08:13,122 because I'm seeing a lot of junk on my, you know, just bots shilling, you know, crypto stuff. 1126 01:08:13,322 --> 01:08:15,742 They're shilling some pills, this and that, right? 1127 01:08:15,822 --> 01:08:17,582 So I'm like, I don't want to see any of this stuff. 1128 01:08:18,082 --> 01:08:19,342 Why are they getting into my comments? 1129 01:08:19,802 --> 01:08:21,002 So I go to Relator. 1130 01:08:21,002 --> 01:08:24,002 I choose Relator as my service provider 1131 01:08:24,002 --> 01:08:27,382 and I get my scores calculated 1132 01:08:27,382 --> 01:08:30,202 and let's just say I'm an Amethyst user. 1133 01:08:30,822 --> 01:08:32,682 Relator blasts out the trusted assertions 1134 01:08:32,682 --> 01:08:35,942 and then I go to Amethyst 1135 01:08:35,942 --> 01:08:37,942 and I choose Relator as my service provider. 1136 01:08:38,802 --> 01:08:40,562 That's how the flow works, right? 1137 01:08:41,642 --> 01:08:42,302 More or less. 1138 01:08:42,902 --> 01:08:45,082 The way that Amethyst is implemented 1139 01:08:45,082 --> 01:08:48,402 the user doesn't need to configure nothing on Amethyst. 1140 01:08:48,402 --> 01:08:51,122 the user can go to 1141 01:08:51,122 --> 01:08:52,982 brainstorm for example 1142 01:08:52,982 --> 01:08:53,382 that is 1143 01:08:53,382 --> 01:08:56,422 one or the other 1144 01:08:56,422 --> 01:08:59,022 service of Web of Trust 1145 01:08:59,022 --> 01:08:59,882 that we talk about 1146 01:08:59,882 --> 01:09:02,222 or relator 1147 01:09:02,222 --> 01:09:03,502 XYZ 1148 01:09:03,502 --> 01:09:05,322 and 1149 01:09:05,322 --> 01:09:08,682 indeed I'm not sure if in brainstorm 1150 01:09:08,682 --> 01:09:09,202 you can 1151 01:09:09,202 --> 01:09:11,142 I think that you can 1152 01:09:11,142 --> 01:09:13,222 but basically edit 1153 01:09:13,222 --> 01:09:16,482 your list of 1154 01:09:16,482 --> 01:09:17,542 trusted providers 1155 01:09:17,542 --> 01:09:26,402 and you can do that in Relator XC for sure. You can just go there and select or add to your list 1156 01:09:26,402 --> 01:09:36,862 of trusted providers the pub key of the Relator server and Amethyst will just pick that up and 1157 01:09:36,862 --> 01:09:46,382 render things, render numbers on your screen and use that for the Web of Trust stuff. So yeah, 1158 01:09:46,382 --> 01:09:50,862 you don't really need to modify or to edit nothing in Ametiz. 1159 01:09:51,022 --> 01:09:52,802 Everything happens elsewhere. 1160 01:09:53,862 --> 01:09:55,902 I don't know if Vitor will implement something 1161 01:09:55,902 --> 01:09:59,182 to allow, modify the list of trusted providers. 1162 01:09:59,542 --> 01:10:02,142 But yeah, in any case, the flow is that. 1163 01:10:02,262 --> 01:10:07,582 The flow is that the user find a provider. 1164 01:10:08,582 --> 01:10:14,682 The provider publish these events, these trusted assertions. 1165 01:10:14,682 --> 01:10:17,082 and then the user have to 1166 01:10:17,082 --> 01:10:19,062 publish another event that 1167 01:10:19,062 --> 01:10:20,822 is a list of the 1168 01:10:20,822 --> 01:10:22,122 trust sets providers 1169 01:10:22,122 --> 01:10:24,742 and yeah that's all 1170 01:10:24,742 --> 01:10:26,622 and then 1171 01:10:26,622 --> 01:10:28,822 so then I go back to Amethyst and 1172 01:10:28,822 --> 01:10:30,942 I don't know if VTOR 1173 01:10:30,942 --> 01:10:32,822 has implemented the filtering yet 1174 01:10:32,822 --> 01:10:34,402 right now it's just I can see 1175 01:10:34,402 --> 01:10:35,922 when I see a comment 1176 01:10:35,922 --> 01:10:38,902 I'll still see the comment but I'll see a number 1177 01:10:38,902 --> 01:10:40,882 on top of that of whoever that 1178 01:10:40,882 --> 01:10:42,702 pub key is that npub is 1179 01:10:42,702 --> 01:10:43,022 right 1180 01:10:43,022 --> 01:10:47,102 but presumably he can add an option in settings 1181 01:10:47,102 --> 01:10:47,982 where I can go and say, 1182 01:10:48,262 --> 01:10:50,582 don't show me comments for anyone whose score is below 20 1183 01:10:50,582 --> 01:10:53,282 because they're probably bots or something like that. 1184 01:10:54,542 --> 01:10:54,802 Yes. 1185 01:10:56,022 --> 01:10:56,642 Yeah, exactly. 1186 01:10:58,022 --> 01:10:59,022 Let's talk about... 1187 01:10:59,702 --> 01:11:00,382 Sorry, go ahead. 1188 01:11:00,522 --> 01:11:01,882 Finish your thought, Jesus. 1189 01:11:02,322 --> 01:11:02,942 Sorry, no. 1190 01:11:02,942 --> 01:11:08,822 I don't know if you were going to jump from the spam prevention topic, 1191 01:11:08,822 --> 01:11:19,862 but I can share another thing that I'm working on that is related to the water lie. 1192 01:11:19,862 --> 01:11:34,782 And water lie is an experiment of a relay that is moderated by the social dynamics of these scores and these web of trust things. 1193 01:11:34,782 --> 01:11:50,002 Because basically, I picked the example of Pip from the relay framework that he had for creating relays. 1194 01:11:51,002 --> 01:11:52,562 And he had an example there. 1195 01:11:53,342 --> 01:11:59,542 And that is basically a bucket algorithm for spam prevention and rate limiting. 1196 01:11:59,542 --> 01:12:04,862 then I took that example 1197 01:12:04,862 --> 01:12:11,422 I refactored it to work with later 1198 01:12:11,422 --> 01:12:15,282 and then what I got is a relay 1199 01:12:15,282 --> 01:12:23,282 that doesn't declare or set very hard rules 1200 01:12:23,962 --> 01:12:27,682 or specific rules for who can publish or what not 1201 01:12:27,682 --> 01:12:35,702 and it is moderated by the social dynamics that are expressed as numbers or ranks 1202 01:12:35,702 --> 01:12:38,742 that comes from relator in this case. 1203 01:12:39,102 --> 01:12:44,942 So the bucket algorithm is basically imagine that you have a bucket 1204 01:12:44,942 --> 01:12:52,082 and the capacity of the bucket is the capacity that you have to publish in this case in the relay. 1205 01:12:53,042 --> 01:12:55,542 You have a refill rate. 1206 01:12:55,542 --> 01:13:02,022 okay and that refill rate if you for example spend all the capacity of your 1207 01:13:02,022 --> 01:13:12,262 bucket it will be refilled at a constant rate so what this relay is doing the 1208 01:13:12,262 --> 01:13:20,342 water relay that I'm running a public instance at water.relator.hyc 1209 01:13:20,342 --> 01:13:37,682 What it's doing is that when you send an event, the relay is contacting with the relator, getting the score or the rank for that pubkey and determining the size of the bucket. 1210 01:13:37,682 --> 01:13:40,042 so if you are a reputable 1211 01:13:40,042 --> 01:13:41,182 pubkey 1212 01:13:41,182 --> 01:13:44,382 with a high score 1213 01:13:44,382 --> 01:13:46,402 you're going to get a big 1214 01:13:46,402 --> 01:13:47,982 bucket, let's say 1215 01:13:47,982 --> 01:13:50,542 thousands of posts a day 1216 01:13:50,542 --> 01:13:52,582 and 1217 01:13:52,582 --> 01:13:54,342 if you are a non-reputable 1218 01:13:54,342 --> 01:13:56,482 or a pubkey 1219 01:13:56,482 --> 01:13:58,342 with a very low rank 1220 01:13:58,342 --> 01:14:00,542 you will have 1221 01:14:00,542 --> 01:14:02,102 less capacity to 1222 01:14:02,102 --> 01:14:04,142 chat 1223 01:14:04,142 --> 01:14:05,202 or to publish 1224 01:14:05,202 --> 01:14:07,622 but, for part 1225 01:14:07,622 --> 01:14:16,602 and if you have a zero rank you're still allowed to publish just that you are restricted to kind 1226 01:14:16,602 --> 01:14:23,422 ones and kind ones without links in this case because I noticed that this is one of the the 1227 01:14:23,422 --> 01:14:30,322 shape of the spans of the spam in Nocer is kind ones with with loads of links and things or 1228 01:14:30,322 --> 01:14:39,802 photographs or images or media. Basically, pubkeys with rank zero are still able to publish 1229 01:14:39,802 --> 01:14:48,762 just to allow new people to solve the bootstrapping problem, where a new pubkey that doesn't 1230 01:14:48,762 --> 01:14:54,442 have a sociograph or someone that doesn't have any of these components that no one knows 1231 01:14:54,442 --> 01:14:59,302 can just publish a GM, for example, or any kind of node 1232 01:14:59,302 --> 01:15:01,682 and start to create connections. 1233 01:15:02,982 --> 01:15:07,702 You know, as that unknown pubkey start to create connections 1234 01:15:07,702 --> 01:15:11,782 and start to follow people and people start to follow that pubkey, 1235 01:15:12,602 --> 01:15:14,822 the rank will evolve 1236 01:15:14,822 --> 01:15:19,282 and the size of the bucket will evolve as well. 1237 01:15:19,282 --> 01:15:22,242 So the publish capacity will evolve with that. 1238 01:15:22,242 --> 01:15:25,162 that is the concept 1239 01:15:25,162 --> 01:15:27,022 the experiment behind the 1240 01:15:27,022 --> 01:15:29,142 waterline which is 1241 01:15:29,142 --> 01:15:30,142 this 1242 01:15:30,142 --> 01:15:33,182 moderation by social 1243 01:15:33,182 --> 01:15:34,202 dynamics basically 1244 01:15:34,202 --> 01:15:37,042 and this logic 1245 01:15:37,042 --> 01:15:38,862 lives on the relay 1246 01:15:38,862 --> 01:15:41,062 for, ok 1247 01:15:41,062 --> 01:15:41,962 that's interesting 1248 01:15:41,962 --> 01:15:45,122 yeah it's a very decoupled 1249 01:15:45,122 --> 01:15:46,942 architecture because you have the relay 1250 01:15:46,942 --> 01:15:48,382 in one side 1251 01:15:48,382 --> 01:15:50,942 which is just cares 1252 01:15:50,942 --> 01:15:52,642 about Relay stuff. 1253 01:15:53,482 --> 01:15:54,982 And in the other side, you have Relator 1254 01:15:54,982 --> 01:15:56,882 that is running as a separate service 1255 01:15:56,882 --> 01:15:58,402 through ContextVM. 1256 01:15:59,342 --> 01:16:01,242 And the Relay can basically, 1257 01:16:01,582 --> 01:16:03,462 when it have to request 1258 01:16:03,462 --> 01:16:05,562 a new rank computation or something, 1259 01:16:06,282 --> 01:16:07,522 it use ContextVM, 1260 01:16:07,882 --> 01:16:08,762 gets the rank, 1261 01:16:09,062 --> 01:16:12,062 and use that rank for this bucket algorithm. 1262 01:16:13,222 --> 01:16:15,882 So they are very decoupled things. 1263 01:16:16,702 --> 01:16:17,702 And for example, 1264 01:16:17,702 --> 01:16:18,902 if you want to run 1265 01:16:18,902 --> 01:16:22,262 a relator instance 1266 01:16:22,262 --> 01:16:24,942 and you run a waterline instance, 1267 01:16:25,042 --> 01:16:26,102 you can connect both 1268 01:16:26,102 --> 01:16:29,822 or do any kind of combination 1269 01:16:29,822 --> 01:16:31,022 in that sense. 1270 01:16:32,702 --> 01:16:32,902 Yeah. 1271 01:16:34,362 --> 01:16:36,742 Yeah, that's super interesting, Jesus. 1272 01:16:37,082 --> 01:16:37,662 So let's actually, 1273 01:16:37,762 --> 01:16:38,642 so that ties into 1274 01:16:38,642 --> 01:16:40,762 the question I was about to ask you 1275 01:16:40,762 --> 01:16:42,582 before you talked about this, 1276 01:16:42,582 --> 01:16:43,202 which is, 1277 01:16:45,162 --> 01:16:45,742 you know, this is, 1278 01:16:45,942 --> 01:16:46,942 people can say, 1279 01:16:47,122 --> 01:17:00,024 now this might not be true But the perception of something like this is you doing some kind of rationing right This bucket size based on score Similarly on Amethyst if you put a number on someone face it not very nice right 1280 01:17:00,044 --> 01:17:01,584 It's like this is a social credit score. 1281 01:17:02,064 --> 01:17:11,024 And I think, you know, a lot of us who are working in this world, we know that it's not, it's all personalized, right? 1282 01:17:11,024 --> 01:17:17,924 So all of these scores are calculated from a point of view of the individual who's requesting the scores. 1283 01:17:18,164 --> 01:17:18,944 It's not being assigned. 1284 01:17:19,064 --> 01:17:20,744 Amethyst is not giving you a score. 1285 01:17:20,884 --> 01:17:23,244 Relator is not giving you a score. 1286 01:17:23,384 --> 01:17:29,404 Relator is just providing the mathematical framework for a user to calculate a score from their perspective. 1287 01:17:30,564 --> 01:17:31,344 Yeah, and it's transparent. 1288 01:17:31,744 --> 01:17:32,744 The algorithm is transparent. 1289 01:17:32,984 --> 01:17:38,624 And the components that compose that rank, final rank, is transparent. 1290 01:17:40,064 --> 01:17:40,544 Right. 1291 01:17:40,544 --> 01:17:43,864 Right. But what, you know, we can explain that. 1292 01:17:43,944 --> 01:17:49,384 But like to an average user, non-technical user, they come in and let's say they're new to Nostra. 1293 01:17:49,564 --> 01:17:53,064 And they're like, well, your bucket is really small because no one knows you. 1294 01:17:53,484 --> 01:17:55,524 And I'm like, my God, I'm being rationed. 1295 01:17:55,824 --> 01:17:58,804 Then they go to Amethyst. Sorry about that. That's my dog. 1296 01:17:59,444 --> 01:18:06,124 Then they go to Amethyst and they see a number on their face or on people's faces. 1297 01:18:06,124 --> 01:18:09,584 And like, what is this? It's like social credit score. I'm being rationed here. 1298 01:18:09,584 --> 01:18:18,444 How would you sort of talk to a non-technical new user and just maybe, you know, help them through the sphere? 1299 01:18:19,844 --> 01:18:26,584 Yeah. So I think that my rationale here is, you know, trust is a scaling solution. 1300 01:18:28,224 --> 01:18:37,704 Thanks to trust, we can go further scale system, especially decentralized systems, right? 1301 01:18:37,704 --> 01:18:48,344 Because if you have a centralized platform that manages that for you and is totally opaque, 1302 01:18:50,824 --> 01:19:00,264 yeah, I mean, that is a very dark pattern, right? 1303 01:19:00,264 --> 01:19:06,084 but we have to recognize that trust is a scaling solution 1304 01:19:06,084 --> 01:19:10,244 to basically perform actions between individuals 1305 01:19:10,244 --> 01:19:13,164 ahead from its own trust 1306 01:19:13,164 --> 01:19:15,684 or its own vision of the world 1307 01:19:15,684 --> 01:19:18,584 it allows me to collaborate with you for example 1308 01:19:18,584 --> 01:19:20,344 I don't know you personally 1309 01:19:20,344 --> 01:19:25,564 but I know people that are my friends that trust you 1310 01:19:25,564 --> 01:19:29,604 so that trust that my friend has 1311 01:19:29,604 --> 01:19:32,844 for you allows me to trust you. 1312 01:19:32,844 --> 01:19:36,724 So it's just my trust has scale 1313 01:19:36,724 --> 01:19:39,844 and we can collaborate without friction 1314 01:19:40,484 --> 01:19:43,064 or knowing that we want to collaborate 1315 01:19:43,064 --> 01:19:45,764 with the person that we want to collaborate 1316 01:19:45,764 --> 01:19:48,584 and thanks to this trust. 1317 01:19:49,264 --> 01:19:51,104 This is especially true, I think, 1318 01:19:51,524 --> 01:19:55,004 for, you know, protocol like NOSER 1319 01:19:55,004 --> 01:20:00,224 where everything is scattered, 1320 01:20:01,764 --> 01:20:04,564 it's centralized, there is no central authorities. 1321 01:20:06,404 --> 01:20:07,864 If we don't have this, 1322 01:20:08,424 --> 01:20:10,604 or if we don't have a way to express this, 1323 01:20:11,024 --> 01:20:12,464 to express this trust, 1324 01:20:12,684 --> 01:20:14,924 we can't escape relations between people 1325 01:20:14,924 --> 01:20:18,344 and everything feels death in that sense. 1326 01:20:18,484 --> 01:20:20,984 Because you can't solve the problem. 1327 01:20:21,124 --> 01:20:23,024 It's a chicken and egg problem, right? 1328 01:20:23,024 --> 01:20:27,104 You first have to interact, people have to see you. 1329 01:20:29,904 --> 01:20:36,024 Thanks to trust, we can create staff that are more advanced 1330 01:20:36,024 --> 01:20:41,644 and can surpass the limits of the decentralized environments like this. 1331 01:20:42,344 --> 01:20:42,624 Like Nasser. 1332 01:20:44,604 --> 01:20:46,344 Especially, it's true because... 1333 01:20:46,344 --> 01:20:48,244 No, no, go ahead. Go ahead, Jesus. 1334 01:20:48,244 --> 01:20:50,084 no it's just that 1335 01:20:50,084 --> 01:20:52,404 you know 1336 01:20:52,404 --> 01:20:54,424 a rank a number can be scary 1337 01:20:54,424 --> 01:20:56,304 and can feel like a social credit 1338 01:20:56,304 --> 01:20:58,464 but I think that the important 1339 01:20:58,464 --> 01:21:00,444 part here is to understand 1340 01:21:00,444 --> 01:21:01,324 that 1341 01:21:01,324 --> 01:21:04,264 if that rank is produced 1342 01:21:04,264 --> 01:21:06,424 by the perspective 1343 01:21:06,424 --> 01:21:08,184 of someone that is 1344 01:21:08,184 --> 01:21:10,444 an individual as well 1345 01:21:10,444 --> 01:21:12,144 as you and it has its own 1346 01:21:12,144 --> 01:21:14,224 criteria to determine what 1347 01:21:14,224 --> 01:21:16,144 it wants to trust and what not 1348 01:21:16,144 --> 01:21:18,504 and all of that is 1349 01:21:18,504 --> 01:21:20,904 transparent in the sense that 1350 01:21:20,904 --> 01:21:22,944 you understand what is behind that 1351 01:21:22,944 --> 01:21:25,004 number or what components 1352 01:21:25,004 --> 01:21:26,744 have been taken 1353 01:21:26,744 --> 01:21:28,684 into account to basically 1354 01:21:28,684 --> 01:21:30,744 get that output 1355 01:21:30,744 --> 01:21:32,944 that 1356 01:21:32,944 --> 01:21:34,704 from my point of view that is 1357 01:21:34,704 --> 01:21:36,964 much less scary and not 1358 01:21:36,964 --> 01:21:38,444 scary at all because it's 1359 01:21:38,444 --> 01:21:39,524 basically 1360 01:21:39,524 --> 01:21:42,244 something natural and organic 1361 01:21:42,244 --> 01:21:43,384 in human relations 1362 01:21:43,384 --> 01:21:48,444 like my perspective is something relative to me 1363 01:21:48,444 --> 01:21:52,844 if I choose how to model that 1364 01:21:52,844 --> 01:21:54,864 it's my decision 1365 01:21:54,864 --> 01:21:56,524 and you can see that 1366 01:21:56,524 --> 01:21:58,244 you can agree with that 1367 01:21:58,244 --> 01:22:00,184 or not 1368 01:22:00,184 --> 01:22:01,524 there is no problem 1369 01:22:01,524 --> 01:22:02,804 you can run your own 1370 01:22:02,804 --> 01:22:05,364 there is nothing imposed top down 1371 01:22:05,364 --> 01:22:08,084 and everything is transparent 1372 01:22:08,084 --> 01:22:11,004 and I think that is the important thing 1373 01:22:11,004 --> 01:22:13,804 to educate someone who might be new to this, 1374 01:22:13,864 --> 01:22:17,864 who might have seen how centralized systems impose these cores 1375 01:22:17,864 --> 01:22:20,264 or manifest these cores. 1376 01:22:21,044 --> 01:22:24,004 Completely agree there on the education part. 1377 01:22:24,984 --> 01:22:27,704 And also agree with your earlier point, 1378 01:22:28,604 --> 01:22:32,524 which is I think that the user experience on NOSTA 1379 01:22:32,524 --> 01:22:36,784 has been suffering from spam and just poor discovery 1380 01:22:36,784 --> 01:22:38,024 and poor curation. 1381 01:22:38,024 --> 01:22:45,264 because these trust-based solutions, personalized trust-based solutions, have not existed. 1382 01:22:45,964 --> 01:22:51,044 Which leads me to, I guess, a follow-on question, which is, 1383 01:22:51,304 --> 01:22:53,304 you've been here three years, I've been here three years. 1384 01:22:53,444 --> 01:22:56,044 You know, a lot of us have been here for that long. 1385 01:22:56,604 --> 01:23:00,084 Why has it taken this long for this realization to come that, 1386 01:23:00,084 --> 01:23:07,184 look, we need this layer of trust to actually make NOSTA meaningful and usable for most people? 1387 01:23:08,024 --> 01:23:15,304 I think that is the uniqueness of the challenge, you know. 1388 01:23:17,904 --> 01:23:22,364 Before Nosher, the most similar thing we had is PGP, right? 1389 01:23:23,264 --> 01:23:27,384 And PGP in spirit is very similar to Nosher. 1390 01:23:28,244 --> 01:23:34,504 And they were the ones that came up with this concept of Web of Trust. 1391 01:23:34,504 --> 01:23:57,404 And Web of Trust for them was basically key parties where people in real life, in mid-space, show up with the fingerprint of their public keys, basically to allow their peers to establish a relation between the key and the person that is behind that. 1392 01:23:58,824 --> 01:24:04,424 That was called Web of Trust and came from that epoch, from that time. 1393 01:24:04,504 --> 01:24:22,804 That was a unique challenge and I think that we are reviving that in Nozer with even more uniqueness because we surpassed some of the limitations that PGP had. 1394 01:24:22,804 --> 01:24:32,704 because basically to use PGP you have to use a medium to transmit an encrypted message. 1395 01:24:33,604 --> 01:24:41,844 I think that the most popular medium was email, basically to encrypt the emails between people. 1396 01:24:41,844 --> 01:24:46,604 but Noceris is a 1397 01:24:46,604 --> 01:24:47,804 decentralized 1398 01:24:47,804 --> 01:24:51,364 protocol 1399 01:24:51,364 --> 01:24:52,764 where people 1400 01:24:52,764 --> 01:24:55,804 run on mpups that are just numbers 1401 01:24:55,804 --> 01:24:57,544 that anyone can generate for free 1402 01:24:57,544 --> 01:24:59,204 can just take 1403 01:24:59,204 --> 01:25:02,044 and sign messages 1404 01:25:02,044 --> 01:25:04,244 and publish them in relays 1405 01:25:04,244 --> 01:25:05,724 and 1406 01:25:05,724 --> 01:25:08,004 I think that the trust 1407 01:25:08,004 --> 01:25:09,604 problem is 1408 01:25:09,604 --> 01:25:16,024 such unique or have present unique challenges for an environment like Nosser. 1409 01:25:16,824 --> 01:25:26,164 And I think that there is a lot there because the why behind this taking so long. 1410 01:25:26,644 --> 01:25:33,904 Because we need to realize things before and also we need data basically to have to compute things. 1411 01:25:34,444 --> 01:25:37,284 Because it's also a chicken and egg problem. 1412 01:25:37,284 --> 01:25:42,744 but now we have all this history behind 1413 01:25:42,744 --> 01:25:47,904 and people can find different ways to represent trust 1414 01:25:47,904 --> 01:25:50,944 and the data wasn't present before. 1415 01:25:51,364 --> 01:25:54,704 So the uniqueness of the challenge 1416 01:25:54,704 --> 01:25:56,924 and the amount of data present, 1417 01:25:57,544 --> 01:26:01,104 I think that is the reason why this took so long. 1418 01:26:02,364 --> 01:26:04,784 Yeah, I think you hit the nail on the head, Jesus. 1419 01:26:05,504 --> 01:26:09,624 I think it took this long for people to actually realize how to solve it. 1420 01:26:09,624 --> 01:26:16,204 And even if they had realized this two years ago, right, how best to approach this, the data wasn't there. 1421 01:26:16,784 --> 01:26:22,984 So I think it's a nice confluence of these two things that are happening, which is people have given it some thought over the last three years. 1422 01:26:23,244 --> 01:26:29,564 And I think people, as you can see, like with Relator and Brainstorm, it's kind of converging on a similar type thing. 1423 01:26:29,564 --> 01:26:35,124 personalized scores using an extensible set of signals, fully transparent. 1424 01:26:35,364 --> 01:26:39,024 You apply a weight based on distance, based on the type of signal. 1425 01:26:39,544 --> 01:26:41,144 Again, the weights are transparent. 1426 01:26:41,564 --> 01:26:45,524 You can, and I'm sure with Relator, you can change the weights if you wanted to, right, 1427 01:26:45,644 --> 01:26:46,824 on each of these signals. 1428 01:26:47,384 --> 01:26:50,084 I think that, so, you know, you're seeing that happening with Relator. 1429 01:26:50,184 --> 01:26:51,304 You're seeing that happening with Brainstorm. 1430 01:26:51,304 --> 01:26:56,764 I'm sure the next one, that next provider comes out will, you know, have a, you know, 1431 01:26:56,764 --> 01:27:01,724 that approach to transparency and portability, 1432 01:27:01,904 --> 01:27:03,764 composability, and so on, extensibility. 1433 01:27:05,524 --> 01:27:07,824 So I think that realization, 1434 01:27:08,944 --> 01:27:11,504 it's almost a collective realization that's happening now. 1435 01:27:11,964 --> 01:27:15,364 It's taken three years, and then we have the data to act on it. 1436 01:27:16,584 --> 01:27:17,684 Yeah, yeah, exactly. 1437 01:27:18,684 --> 01:27:19,064 Exactly. 1438 01:27:19,064 --> 01:27:22,864 I really believe that was one of the main reasons. 1439 01:27:22,864 --> 01:27:44,806 Also because you know exactly for the same reason of the uniqueness of the challenge Nostar itself is something new it just has four years something very young very young protocol and it really presents challenges in its topology and how it works and so on 1440 01:27:44,806 --> 01:28:00,946 So I think that it's a combination of all of that, but principally the trust, how to express it, we are understanding that now, and also the data, the amount of data that was there. 1441 01:28:00,946 --> 01:28:11,506 talking about something that really excites me that is coming to Relator is basically a new 1442 01:28:11,506 --> 01:28:21,446 system that I'm creating for plugins or for these two sorry so basically to configure what are the 1443 01:28:21,446 --> 01:28:29,706 signals that you take to for the for the computations because currently Relator allows you to 1444 01:28:29,706 --> 01:28:35,886 to modify what are the signals that you take for the computation, 1445 01:28:36,066 --> 01:28:37,786 the signal that we talked about before, 1446 01:28:38,366 --> 01:28:43,966 that currently by default are if the user has a relay list, 1447 01:28:43,966 --> 01:28:48,806 if the user has a lining address set in their profiles, 1448 01:28:49,606 --> 01:28:52,546 if the ones that we talked about before. 1449 01:28:53,206 --> 01:28:57,986 And developers could modify and add new validations, 1450 01:28:57,986 --> 01:29:04,606 but in the current version of Relator that would mean to modify the source code of Relator 1451 01:29:04,606 --> 01:29:12,046 which nowadays is fairly easy with LLMs but still presenting friction 1452 01:29:12,046 --> 01:29:18,206 I'm very excited for a plugin system that I'm developing for Relator 1453 01:29:18,206 --> 01:29:24,806 which is based on an expression language that is called ELO or ELO 1454 01:29:24,806 --> 01:29:38,326 and it basically allows me to design a system where these signals can be expressed in very concise expressions 1455 01:29:38,326 --> 01:29:42,446 that can be bundled as NOSER events. 1456 01:29:42,446 --> 01:29:51,726 So basically the idea is that anyone can create these programs for signals, 1457 01:29:51,726 --> 01:29:54,726 publish them on Nozir 1458 01:29:54,726 --> 01:29:58,526 and people can use them 1459 01:29:58,526 --> 01:30:00,286 in their relator instances 1460 01:30:00,286 --> 01:30:04,626 so if I'm a relator operator 1461 01:30:04,626 --> 01:30:06,566 starting to decide 1462 01:30:06,566 --> 01:30:08,786 what trust means for me 1463 01:30:08,786 --> 01:30:11,066 or what the rank 1464 01:30:11,066 --> 01:30:13,586 that a relator is outputting 1465 01:30:13,586 --> 01:30:15,166 what that means 1466 01:30:15,166 --> 01:30:18,106 I will be able to go to Nozir 1467 01:30:18,106 --> 01:30:21,826 in a marketplace manner 1468 01:30:21,826 --> 01:30:26,206 and see what validators or what things are there, 1469 01:30:26,586 --> 01:30:27,886 what people have been created, 1470 01:30:29,506 --> 01:30:31,986 just with the ease of a click 1471 01:30:31,986 --> 01:30:37,226 or configure the algorithm 1472 01:30:37,226 --> 01:30:40,726 and set the weights for all those signals. 1473 01:30:40,726 --> 01:30:44,646 So this effectively creates a marketplace of these signals 1474 01:30:44,646 --> 01:30:53,586 and it allows people to create, to curate, to share, to use. 1475 01:30:54,666 --> 01:31:02,086 This excites me a lot because I think that this is the materialization of the vision behind Relator 1476 01:31:02,086 --> 01:31:04,866 and with a good experience for the user 1477 01:31:04,866 --> 01:31:15,346 because you will be able to determine, tune, and understand why the rank 1478 01:31:15,346 --> 01:31:17,946 or what is the rank that is coming out from Relator. 1479 01:31:17,946 --> 01:31:24,566 So for example, if I'm publishing a signal or an expression 1480 01:31:24,566 --> 01:31:34,406 that measures the amount of GMs or the amount of subs that a user received, 1481 01:31:34,866 --> 01:31:38,806 I can just pick those, configure them in Relator, 1482 01:31:39,506 --> 01:31:41,446 because they are no-serie events, 1483 01:31:41,626 --> 01:31:43,926 and Relator can just fetch them from Relator, 1484 01:31:44,926 --> 01:31:47,926 and use them for the computations. 1485 01:31:49,386 --> 01:31:50,826 I think that's really cool. 1486 01:31:52,566 --> 01:31:54,146 Is this feature out yet? 1487 01:31:55,266 --> 01:31:55,526 No. 1488 01:31:56,446 --> 01:32:01,446 No, I was trying to put it together for this podcast, 1489 01:32:01,446 --> 01:32:06,866 but I was thrilled by different things, by work. 1490 01:32:07,506 --> 01:32:13,546 Also, yesterday we released a new CLI for a set of skills for ContextVM, 1491 01:32:13,666 --> 01:32:15,846 for developing things on ContextVM. 1492 01:32:17,886 --> 01:32:23,866 This new feature for Relator needs a bit more of polishing. 1493 01:32:24,706 --> 01:32:26,686 It's basically to provide... It's working already. 1494 01:32:27,086 --> 01:32:28,546 It's totally feasible. 1495 01:32:28,546 --> 01:32:33,226 I have tests and indeed makes Relator even more efficient and performance. 1496 01:32:35,166 --> 01:32:37,286 But it's not released yet. 1497 01:32:37,966 --> 01:32:44,646 I need to polish the user experience in the sense of publishing 1498 01:32:44,646 --> 01:32:48,886 and setting up these signals from Nocer. 1499 01:32:49,506 --> 01:32:51,046 But it's coming along nicely. 1500 01:32:52,486 --> 01:32:55,646 Nice. Yeah, certainly something to look forward to. 1501 01:32:55,646 --> 01:32:57,146 Jesus 1502 01:32:57,146 --> 01:32:58,946 this has been great 1503 01:32:58,946 --> 01:33:00,926 great conversation 1504 01:33:00,926 --> 01:33:02,386 thank you for educating us 1505 01:33:02,386 --> 01:33:04,946 the audience about Relator 1506 01:33:04,946 --> 01:33:07,046 and all the work that you've done 1507 01:33:07,046 --> 01:33:08,226 on NOSTA 1508 01:33:08,226 --> 01:33:10,886 with Context VM, DVM-CP 1509 01:33:10,886 --> 01:33:12,106 maybe one day we'll see 1510 01:33:12,106 --> 01:33:12,966 Coloria 1511 01:33:12,966 --> 01:33:16,286 I think that 1512 01:33:16,286 --> 01:33:18,226 it's going to be 1513 01:33:18,226 --> 01:33:20,286 everything part of NOSTA 1514 01:33:20,286 --> 01:33:22,026 and when I 1515 01:33:22,026 --> 01:33:23,626 revive NOSTA 1516 01:33:23,626 --> 01:33:26,066 it's going to come with these features. 1517 01:33:27,886 --> 01:33:28,886 So, yeah. 1518 01:33:29,706 --> 01:33:33,526 Relator scores, kill the zombies with all the relator scores, 1519 01:33:33,646 --> 01:33:34,346 and that's going to be great. 1520 01:33:35,806 --> 01:33:37,726 Yeah, exactly. 1521 01:33:37,926 --> 01:33:40,766 I mean, Nosser is very fun, definitely. 1522 01:33:41,586 --> 01:33:48,466 If I can say something from this time working on Nosser, 1523 01:33:49,386 --> 01:33:52,186 it is very fun to develop on Nosser. 1524 01:33:52,186 --> 01:33:54,646 it's also quite frustrating 1525 01:33:54,646 --> 01:33:55,246 sometimes 1526 01:33:55,246 --> 01:33:58,246 because the challenges 1527 01:33:58,246 --> 01:33:59,306 that it presents 1528 01:33:59,306 --> 01:34:02,526 how standards 1529 01:34:02,526 --> 01:34:04,266 are standardized 1530 01:34:04,266 --> 01:34:05,826 talking about MIPS 1531 01:34:05,826 --> 01:34:08,026 and so on 1532 01:34:08,026 --> 01:34:09,906 so sometimes it's very frustrating 1533 01:34:09,906 --> 01:34:11,566 very frustrating 1534 01:34:11,566 --> 01:34:13,446 but very fun as well 1535 01:34:13,446 --> 01:34:15,126 so it's a balance 1536 01:34:15,126 --> 01:34:16,906 no for sure 1537 01:34:16,906 --> 01:34:19,886 it's the emergent and messy 1538 01:34:19,886 --> 01:34:21,866 consensus of being part of 1539 01:34:21,866 --> 01:34:23,706 a decentralized 1540 01:34:23,706 --> 01:34:25,146 group. 1541 01:34:26,406 --> 01:34:28,406 Look, I think people have the right attitude 1542 01:34:28,406 --> 01:34:30,486 here, because there's two ways of looking 1543 01:34:30,486 --> 01:34:32,186 at it. You can have a decentralized 1544 01:34:32,186 --> 01:34:34,446 mob, and it's complete 1545 01:34:34,446 --> 01:34:36,506 chaos, or you can 1546 01:34:36,506 --> 01:34:38,486 have a decentralized mob, and it's only a little 1547 01:34:38,486 --> 01:34:39,906 bit of chaos, but mostly fun, 1548 01:34:40,446 --> 01:34:42,186 which thankfully, at least 1549 01:34:42,186 --> 01:34:43,866 until now, that's what Nostro has been. 1550 01:34:45,186 --> 01:34:45,386 Yeah. 1551 01:34:46,646 --> 01:34:48,466 My only critic 1552 01:34:48,466 --> 01:34:50,426 for Nostro is 1553 01:34:50,426 --> 01:34:58,546 that the governance or at least the standards are not easy to navigate. 1554 01:34:58,846 --> 01:35:01,346 If you are someone new, like a new developer, 1555 01:35:01,746 --> 01:35:05,046 trying to figure out how to develop something on Nocer, 1556 01:35:05,686 --> 01:35:08,186 it's going to have a hard time trying to navigate Nibs 1557 01:35:08,186 --> 01:35:10,146 and to understand how to do things. 1558 01:35:10,846 --> 01:35:15,226 I have expertise now after three years working on that, 1559 01:35:15,366 --> 01:35:20,026 working on this, and I understand how things work and so on. 1560 01:35:20,026 --> 01:35:23,866 But yeah, someone new wanting to develop something in Nozir, 1561 01:35:24,286 --> 01:35:26,406 it's going to you see a lot of mess. 1562 01:35:27,406 --> 01:35:29,806 And that presents a lot of friction. 1563 01:35:31,546 --> 01:35:36,906 Yeah, also I wrote an article some time ago, 1564 01:35:36,906 --> 01:35:39,446 I think during Sovereign Engineering 5, 1565 01:35:40,526 --> 01:35:42,786 that is called Enemies of Nozir. 1566 01:35:43,726 --> 01:35:48,466 It's a reflection about open source protocols. 1567 01:35:50,026 --> 01:35:51,846 and how they get captured, 1568 01:35:52,986 --> 01:35:57,546 and how the same history can repeat on Nostra. 1569 01:35:59,446 --> 01:36:03,026 So I don't know if you read that, but... 1570 01:36:03,026 --> 01:36:05,846 I remember seeing it. 1571 01:36:05,886 --> 01:36:06,946 It was a while back, right? 1572 01:36:07,846 --> 01:36:09,706 Yeah, some months ago, yeah. 1573 01:36:10,646 --> 01:36:11,986 I'll tell you what, Jesus, 1574 01:36:12,246 --> 01:36:14,866 I will put all of those links in the show notes, 1575 01:36:14,866 --> 01:36:16,546 and I'll certainly read it as well, 1576 01:36:16,626 --> 01:36:17,506 The Enemies of Nostra. 1577 01:36:17,926 --> 01:36:19,986 I'll link that, I'll link ContextVM, 1578 01:36:20,026 --> 01:36:24,866 I'll link Relator, Nostri, anything else you want me to link. 1579 01:36:24,986 --> 01:36:26,946 I'll put all of that in the show notes 1580 01:36:26,946 --> 01:36:30,566 so that our audience can check it out as well. 1581 01:36:31,866 --> 01:36:33,366 No, that sounds good. 1582 01:36:34,266 --> 01:36:34,626 Awesome. 1583 01:36:35,546 --> 01:36:38,086 All right, Jesus, again, thank you. 1584 01:36:38,586 --> 01:36:40,766 Great conversation, very educational. 1585 01:36:40,906 --> 01:36:44,166 Thank you for everything you do to make Nostri a better place, 1586 01:36:44,906 --> 01:36:46,966 make it more fun despite the challenges. 1587 01:36:46,966 --> 01:36:48,426 Thank you for being patient. 1588 01:36:48,426 --> 01:36:53,046 and not running away and going crazy. 1589 01:36:53,406 --> 01:36:54,326 It's worth it. 1590 01:36:54,566 --> 01:36:55,246 It's worth it. 1591 01:36:55,706 --> 01:36:58,666 It's worth it because we are leaning towards 1592 01:36:58,666 --> 01:37:04,086 a very scary thing. 1593 01:37:04,706 --> 01:37:06,086 But yeah, thank you so much. 1594 01:37:06,206 --> 01:37:07,906 I appreciate it so much, your words. 1595 01:37:09,186 --> 01:37:11,106 Yeah, let's continue building. 1596 01:37:11,866 --> 01:37:13,526 Let's continue building indeed. 1597 01:37:13,526 --> 01:37:17,266 Thank you to our live stream audience for listening. 1598 01:37:17,266 --> 01:37:18,566 Thank you for the zaps. 1599 01:37:18,806 --> 01:37:20,566 If you are listening to this recording, 1600 01:37:20,786 --> 01:37:23,986 I hope you're doing so on a modern Podcasting 2.0 app 1601 01:37:23,986 --> 01:37:25,126 such as Fountain. 1602 01:37:25,846 --> 01:37:27,966 If you found value in this conversation, 1603 01:37:28,186 --> 01:37:29,906 if it gave you a new mental model, 1604 01:37:30,126 --> 01:37:31,226 a moment of clarity, 1605 01:37:31,826 --> 01:37:34,226 or just a respite from the noise, 1606 01:37:34,586 --> 01:37:36,926 then please consider returning the value. 1607 01:37:37,466 --> 01:37:40,406 You can stream sats as you listen on Fountain 1608 01:37:40,406 --> 01:37:43,466 or another Podcasting 2.0 app, 1609 01:37:43,886 --> 01:37:46,626 or you could become a monthly subscriber 1610 01:37:46,626 --> 01:37:49,066 to keep the signal strong. 1611 01:37:50,146 --> 01:37:51,386 Thank you, folks. 1612 01:37:51,926 --> 01:37:52,806 See you next time.