1 00:00:00,040 --> 00:00:03,600 Welcome to Architect it AI Architect, the fully AI 2 00:00:03,600 --> 00:00:06,480 generated podcast for tech enthusiasts, gadget lovers, 3 00:00:06,480 --> 00:00:08,600 curious consumers and AI builders. 4 00:00:09,000 --> 00:00:13,520 Every episode is 100% crafted by AI from concept to delivery, 5 00:00:13,760 --> 00:00:17,120 showcasing real human machine collaboration in action. 6 00:00:17,200 --> 00:00:18,760 It's great to be back. Yeah. 7 00:00:18,760 --> 00:00:21,520 And today we have a really incredible deep dive for you. 8 00:00:21,920 --> 00:00:25,160 I want you to imagine right now that you buy a brand new 9 00:00:25,160 --> 00:00:28,040 computer. It does not matter if it's a top 10 00:00:28,040 --> 00:00:31,440 of the line desktop you know or some sleek new laptop on day 11 00:00:31,440 --> 00:00:33,760 one. It is just light and fast. 12 00:00:33,760 --> 00:00:35,280 Oh yeah, everything opens instantly. 13 00:00:35,280 --> 00:00:37,440 Exactly. The battery life is perfect. 14 00:00:37,440 --> 00:00:40,760 It feels flawless. But Fast forward, say, three 15 00:00:40,760 --> 00:00:43,200 years, without you even realizing it, that exact same 16 00:00:43,200 --> 00:00:46,360 machine has just become this sluggish, tangled mess. 17 00:00:46,360 --> 00:00:48,400 You know, you try to update your graphics driver and suddenly 18 00:00:48,400 --> 00:00:50,000 your audio interface stops working. 19 00:00:50,440 --> 00:00:52,880 Or you install a new piece of development software and an 20 00:00:52,880 --> 00:00:55,440 older application just mysteriously refuses to open. 21 00:00:55,920 --> 00:00:59,080 And we all just kind of accept this slow degradation as like a 22 00:00:59,120 --> 00:01:01,680 law of physics. You do call it digital rot. 23 00:01:01,680 --> 00:01:04,200 We assume computers just age much like we do, right? 24 00:01:04,519 --> 00:01:07,560 Yeah, the industry term for that phenomenon, it's entropy. 25 00:01:08,040 --> 00:01:12,040 We are basically living in an era of entropy driven DevOps. 26 00:01:12,560 --> 00:01:14,840 And it's not just your personal laptop that suffers from this. 27 00:01:14,840 --> 00:01:19,000 You know this exact same rot happens to massive enterprise 28 00:01:19,000 --> 00:01:22,320 server clusters, cloud infrastructure, artificial 29 00:01:22,320 --> 00:01:25,440 intelligence training rigs. We've completely normalized a 30 00:01:25,440 --> 00:01:28,840 reality where our systems get progressively messier and more 31 00:01:28,840 --> 00:01:30,400 fragile with every single update. 32 00:01:30,480 --> 00:01:33,000 Because every change leaves something behind. 33 00:01:33,000 --> 00:01:34,880 Exactly. Every patch, every installation 34 00:01:34,880 --> 00:01:37,920 leaves behind a ghost. Basically like a stray file, an 35 00:01:37,920 --> 00:01:41,640 outdated registry key, an orphan dependency. 36 00:01:41,760 --> 00:01:43,840 But what if that degradation wasn't inevitable? 37 00:01:43,840 --> 00:01:46,280 What if your computer could be virtually immortal? 38 00:01:46,640 --> 00:01:49,400 That is our mission for you today in this deep dive. 39 00:01:49,400 --> 00:01:52,000 The big promise? It is we are exploring A 40 00:01:52,000 --> 00:01:55,480 comprehensive 20,000 word research dossier covering the 41 00:01:55,480 --> 00:01:57,600 state of an operating system called Nix OS. 42 00:01:57,880 --> 00:02:00,400 Looking specifically at where the technology stands right now 43 00:02:00,400 --> 00:02:03,640 in the year 2026. We want to understand why the 44 00:02:03,640 --> 00:02:07,440 elite 1% of engineers, you know, AI researchers and power users 45 00:02:07,920 --> 00:02:10,960 are completely abandoning the traditional rules of computing. 46 00:02:11,120 --> 00:02:13,320 They're migrating to a platform that really shouldn't work. 47 00:02:13,440 --> 00:02:16,480 Right, a platform that completely breaks decades of 48 00:02:16,480 --> 00:02:20,720 established computer science norms, but somehow does work. 49 00:02:21,400 --> 00:02:23,760 We're going to figure out if NICs OS is truly the final 50 00:02:23,760 --> 00:02:25,880 operating system. What's fascinating here is that 51 00:02:25,880 --> 00:02:29,560 when you dig into the core philosophy outlined in these 52 00:02:29,560 --> 00:02:34,440 materials, you realize Nix OS isn't just a different flavor of 53 00:02:34,440 --> 00:02:35,720 Linux. It's not just, you know, a new 54 00:02:35,720 --> 00:02:37,120 coat of paint on the same old engine. 55 00:02:37,600 --> 00:02:41,040 It fundamentally redefines what a computer is at a mathematical 56 00:02:41,040 --> 00:02:43,320 level. Traditional operating systems 57 00:02:43,320 --> 00:02:46,200 manage files. Nix OS treats the entire 58 00:02:46,200 --> 00:02:50,400 operating system as a single pure mathematical function. 59 00:02:50,400 --> 00:02:52,800 OK, let me stop you there, because treating a computer like 60 00:02:52,800 --> 00:02:55,080 a math equation sounds incredibly abstract. 61 00:02:55,080 --> 00:02:57,800 Fair enough. We really need to ground this. 62 00:02:58,320 --> 00:03:01,200 The central tension we are exloring today is that Nix OS 63 00:03:01,200 --> 00:03:05,800 promises an unbreakable reality, like a system that cannot 64 00:03:05,800 --> 00:03:08,920 digitally rot. But achieving that requires 65 00:03:08,920 --> 00:03:11,680 climbing this notoriously brutal learning curve. 66 00:03:12,280 --> 00:03:15,320 So to understand if that climb is worth it, we need to look at 67 00:03:15,320 --> 00:03:18,600 how we've been doing things wrong for the last 50 years. 68 00:03:19,160 --> 00:03:21,360 Let's start with what the sources call the mutable 69 00:03:21,360 --> 00:03:23,600 imperative model. Right, the traditional way. 70 00:03:23,640 --> 00:03:26,440 Yeah, I like to visualize a traditional operating system 71 00:03:26,440 --> 00:03:29,240 like Ubuntu or Windows as a house. 72 00:03:29,320 --> 00:03:32,520 The imperative model is like building that house by randomly 73 00:03:32,520 --> 00:03:34,880 adding and removing bricks over the course of years. 74 00:03:34,880 --> 00:03:37,240 That's a great analogy. When you open a terminal and run 75 00:03:37,240 --> 00:03:40,720 a command like apartment upgrade or you manually edit a text file 76 00:03:40,720 --> 00:03:43,720 in your system folder to change network setting, you are 77 00:03:43,720 --> 00:03:46,360 basically swinging a hammer. You are mutating the live 78 00:03:46,360 --> 00:03:47,840 environment. And the main problem with 79 00:03:47,840 --> 00:03:51,240 swinging that hammer is that you're creating a unique, deeply 80 00:03:51,240 --> 00:03:55,840 personal history of state. Like if you and I both install a 81 00:03:55,840 --> 00:04:00,240 fresh copy of Debian today, they are identical, but by next 82 00:04:00,240 --> 00:04:03,720 month, based on the exact sequence of commands you typed 83 00:04:03,800 --> 00:04:06,800 versus the commands I type, our systems will be completely 84 00:04:06,800 --> 00:04:09,040 different under the hood. Because I might have installed a 85 00:04:09,040 --> 00:04:12,400 Python library on Tuesday that updated a core system file and 86 00:04:12,400 --> 00:04:13,720 you didn't. Exactly. 87 00:04:13,920 --> 00:04:15,760 You might have updated it, and I might have installed a 88 00:04:15,760 --> 00:04:18,680 completely different tool on Friday that actually downgraded 89 00:04:18,680 --> 00:04:21,680 that exact same file. Which means if my system 90 00:04:21,680 --> 00:04:25,440 suddenly crashes and I ask you for help, you can't reliably 91 00:04:25,440 --> 00:04:28,120 troubleshoot it because my system is just this unique 92 00:04:28,120 --> 00:04:30,720 snowflake of accumulated state. Precisely. 93 00:04:30,720 --> 00:04:32,840 You have side effects. You have state drift. 94 00:04:32,960 --> 00:04:36,800 The alternative presented in the 2026 architecture is the 95 00:04:36,800 --> 00:04:39,680 declarative pure model. OK, so how does that work? 96 00:04:39,880 --> 00:04:43,600 Well, in this paradigm you never issue step by step commands to 97 00:04:43,600 --> 00:04:46,120 mutate a live system. You don't tell the computer how 98 00:04:46,120 --> 00:04:48,720 to change, you declare what the final state should be. 99 00:04:48,800 --> 00:04:51,400 Like writing a blueprint. Yeah, exactly like a blueprint. 100 00:04:51,680 --> 00:04:55,400 You write your exact intent into a single text configuration file 101 00:04:55,800 --> 00:04:58,640 and then the next package manager takes that file and 102 00:04:58,640 --> 00:05:00,840 realizes it through atomic transitions. 103 00:05:01,280 --> 00:05:04,720 It builds the system from the ground U completely isolated 104 00:05:04,720 --> 00:05:06,480 from whatever is currently running on the machine. 105 00:05:06,480 --> 00:05:11,680 OK, and if you take that text file and alley it 10 times, you 106 00:05:11,680 --> 00:05:14,360 do not get 10 different variations of a system. 107 00:05:14,880 --> 00:05:18,400 You get the exact same system bit for bit identical 10 times. 108 00:05:18,480 --> 00:05:19,640 Right. But I'm looking at this and 109 00:05:19,640 --> 00:05:22,120 thinking about the cloud engineers listening to us right 110 00:05:22,120 --> 00:05:23,560 now. They're probably screaming at 111 00:05:23,560 --> 00:05:25,640 their dashboard saying, you know, we already solved this. 112 00:05:25,640 --> 00:05:28,560 We use Terraform, we use Ansible, we use Docker, We 113 00:05:28,560 --> 00:05:30,720 declare our server states all the time. 114 00:05:30,720 --> 00:05:31,520 Oh. Absolutely. 115 00:05:31,560 --> 00:05:34,400 So how is this mathematical approach any different from the 116 00:05:34,400 --> 00:05:38,360 DevOps tools the industry has been using for like a decade? 117 00:05:38,600 --> 00:05:40,320 That is a very common misconception. 118 00:05:40,320 --> 00:05:42,200 Actually. Tools like Ansible are 119 00:05:42,200 --> 00:05:46,200 incredibly useful, but they operate on an illusion of idem 120 00:05:46,200 --> 00:05:48,520 potence. They are still fundamentally 121 00:05:48,520 --> 00:05:51,200 imperative under the hood. What do you mean that well? 122 00:05:51,600 --> 00:05:55,040 When Ansible runs, it logs into a live server, checks the 123 00:05:55,040 --> 00:05:57,680 current state of a file, and then runs the script to alter 124 00:05:57,680 --> 00:05:59,800 it. It's still swinging the hammer, 125 00:06:00,080 --> 00:06:01,720 it's just doing it via automation. 126 00:06:02,480 --> 00:06:04,280 I see. And Docker containers are 127 00:06:04,280 --> 00:06:06,960 closer, but they're essentially just shipping an entire heavy 128 00:06:06,960 --> 00:06:10,280 prepackaged file system to avoid dealing with the underlying 129 00:06:10,280 --> 00:06:13,600 host. Nix OS is not layering a tool on 130 00:06:13,600 --> 00:06:16,920 top of a messy system. It is applying pure functional 131 00:06:16,920 --> 00:06:20,320 declaration down to the very kernel and desktop environment 132 00:06:20,320 --> 00:06:21,120 level. Right. 133 00:06:21,240 --> 00:06:23,160 It doesn't check the current state and tweak it. 134 00:06:23,400 --> 00:06:26,680 It evaluates the configuration and builds the outcome in total 135 00:06:26,680 --> 00:06:28,840 isolation. OK, let's unpack this isolation 136 00:06:28,840 --> 00:06:31,440 part, because to achieve that isolation, Nix OS has to 137 00:06:31,440 --> 00:06:34,280 completely destroy the traditional layout of computer. 138 00:06:34,920 --> 00:06:38,320 Anyone who has ever poked around a Mac or a Linux machine knows 139 00:06:38,320 --> 00:06:40,240 the File system Hierarchy standard. 140 00:06:40,240 --> 00:06:42,240 Yes, the classic FHS. Right. 141 00:06:42,400 --> 00:06:45,000 You have a folder called bin where your executables live. 142 00:06:45,200 --> 00:06:47,680 You have a folder called Lib where your shared libraries 143 00:06:47,680 --> 00:06:49,440 live. It has literally been that way 144 00:06:49,440 --> 00:06:52,360 since the dawn of Unix. But according to the dossier, 145 00:06:52,440 --> 00:06:54,680 Nix OS completely nukes that structure. 146 00:06:54,920 --> 00:06:57,240 There is no standard bin. There is no Lib. 147 00:06:57,520 --> 00:07:00,480 Because that traditional hierarchy is exactly what causes 148 00:07:00,480 --> 00:07:01,960 the entropy we discussed earlier. 149 00:07:02,440 --> 00:07:04,160 How so? Well, think about how a shared 150 00:07:04,160 --> 00:07:06,640 library works. Let's take something fundamental 151 00:07:06,880 --> 00:07:09,720 like glip, which is the core C library. 152 00:07:10,280 --> 00:07:13,440 Almost every program needs to function in a traditional 153 00:07:13,440 --> 00:07:15,760 system. Glib sits in the Lib folder, 154 00:07:16,000 --> 00:07:17,720 right? If you install a video editor, 155 00:07:17,720 --> 00:07:20,960 it looks in Lib for glib. If you install a web browser, it 156 00:07:20,960 --> 00:07:24,480 looks in the exact same place. But what happens when your new 157 00:07:24,480 --> 00:07:29,080 video editor demands version 2.38 of that library, but your 158 00:07:29,080 --> 00:07:32,480 web browser relies on an older, specific quirk found only in 159 00:07:32,480 --> 00:07:35,240 version 2.36? They overwrite each other, you 160 00:07:35,240 --> 00:07:38,320 update the library for the video editor and the web browser 161 00:07:38,560 --> 00:07:40,480 silently breaks. And then the next time you try 162 00:07:40,480 --> 00:07:43,480 to open it, it just bounces in the dock and crashes because the 163 00:07:43,480 --> 00:07:45,960 foundation it expects has literally been pulled out from 164 00:07:45,960 --> 00:07:48,880 under it dependency hell. Dependency hell. 165 00:07:49,400 --> 00:07:52,240 Nix O solves this by putting everything, every package, every 166 00:07:52,240 --> 00:07:55,640 library, every font, every configuration file into a single 167 00:07:55,640 --> 00:07:57,960 massive directory called the Nix Store. 168 00:07:58,080 --> 00:08:02,160 The Nix Store and the isolation mechanism here is brilliant. 169 00:08:02,360 --> 00:08:06,240 It does not just name a folder Firefox, it names the folder 170 00:08:06,440 --> 00:08:08,360 using a unique cryptographic hash. 171 00:08:08,560 --> 00:08:11,640 So you end up with a folder named with 32 random 172 00:08:11,680 --> 00:08:15,440 alphanumeric characters, something like Z560 followed by 173 00:08:15,440 --> 00:08:18,440 Firefox 135.0. Exactly. 174 00:08:18,720 --> 00:08:21,880 And the math behind that hash is the secret sauce here, right? 175 00:08:22,080 --> 00:08:24,600 It isn't just a random ID generated install time. 176 00:08:24,800 --> 00:08:28,200 That hash is calculated based on the source code of Firefox, the 177 00:08:28,200 --> 00:08:31,600 specific compiler used to build it, and the exact cryptographic 178 00:08:31,600 --> 00:08:34,120 hashes of every single dependency it relies on. 179 00:08:34,240 --> 00:08:36,280 Right. The cascading nature of the hash 180 00:08:36,280 --> 00:08:38,600 is what makes it a mathematical proof of the software. 181 00:08:38,919 --> 00:08:41,640 If a developer changes even a single line of code in some 182 00:08:41,640 --> 00:08:44,720 obscure background dependency, the hash of that dependency 183 00:08:44,720 --> 00:08:46,600 changes. Which triggers a chain reaction. 184 00:08:46,880 --> 00:08:50,920 Exactly because Firefox relies on it, the input has changed, 185 00:08:51,080 --> 00:08:54,440 which means the final hash for Firefox changes completely. 186 00:08:54,880 --> 00:08:57,600 This guarantees that two different builds can never 187 00:08:57,600 --> 00:09:00,120 overwrite each other. You can have the video editor 188 00:09:00,280 --> 00:09:03,960 using its specific glib in one hashed folder, and the web 189 00:09:03,960 --> 00:09:06,960 browser using its older glib in another completely separate hash 190 00:09:06,960 --> 00:09:08,880 folder. And they coexist peacefully 191 00:09:08,880 --> 00:09:12,480 because to the system they are mathematically distinct 192 00:09:12,480 --> 00:09:13,720 entities. Precisely. 193 00:09:13,720 --> 00:09:16,240 I understand the theory, but let's look at the friction this 194 00:09:16,240 --> 00:09:20,040 causes for the user. If my entire operating system is 195 00:09:20,040 --> 00:09:23,800 buried inside a single folder categorized by thousands of 196 00:09:23,800 --> 00:09:27,760 unreadable 32 character hash names, wait, how does the 197 00:09:27,760 --> 00:09:31,080 computer actually function? Like if I open a terminal and 198 00:09:31,080 --> 00:09:33,560 type Firefox, how does the computer know which of the 199 00:09:33,560 --> 00:09:35,920 heavily encrypted folder paths to execute? 200 00:09:36,120 --> 00:09:38,040 It seems like it would be incredibly slow to search 201 00:09:38,040 --> 00:09:40,920 through that massive haystack. Well, it doesn't search at all. 202 00:09:41,160 --> 00:09:43,600 This is where Nix OS introduces Simlink forests. 203 00:09:43,640 --> 00:09:46,320 Simlink Forests. Yeah, when you tell Nix OS to 204 00:09:46,320 --> 00:09:50,040 apply your configuration file, it evaluates all those hashes 205 00:09:50,120 --> 00:09:52,040 and builds what it calls a generation. 206 00:09:52,600 --> 00:09:55,720 A generation is simply a tree of symbolic links, digital 207 00:09:55,720 --> 00:09:58,520 shortcuts. It creates A mock version of the 208 00:09:58,520 --> 00:10:01,920 file system you expect O it creates a bin folder, but inside 209 00:10:01,920 --> 00:10:04,760 that folder the executable for Firefox is just a shortcut 210 00:10:05,160 --> 00:10:08,280 ointing directly to the specific cryptographic hash in the next 211 00:10:08,280 --> 00:10:10,760 door. Wow, so it is an illusion of a 212 00:10:10,760 --> 00:10:13,360 normal computer contracted entirely out of shortcuts. 213 00:10:13,360 --> 00:10:15,360 Exactly. Yeah, and because the system's 214 00:10:15,360 --> 00:10:18,600 reality is just a collection of shortcuts pointing to isolated 215 00:10:18,600 --> 00:10:22,240 read only folders, you unlock the single greatest feature of 216 00:10:22,240 --> 00:10:24,240 Nix OS, which is atomic rollbacks. 217 00:10:24,240 --> 00:10:26,320 Oh, this is the time travel debugging. 218 00:10:26,320 --> 00:10:28,960 Let me paint a picture of why this matters so much for you 219 00:10:28,960 --> 00:10:30,920 listening. Let's say you're a creative 220 00:10:30,920 --> 00:10:32,840 professional. You have a massive deadline 221 00:10:32,840 --> 00:10:34,760 tomorrow. You decide to run a system 222 00:10:34,760 --> 00:10:38,240 update and halfway through the power flashes or a brand new 223 00:10:38,240 --> 00:10:40,280 display driver conflicts with your kernel. 224 00:10:40,600 --> 00:10:43,440 You reboot and your screen is completely black. 225 00:10:43,600 --> 00:10:46,480 A total nightmare scenario. Right, the graphical interface 226 00:10:46,480 --> 00:10:49,080 is dead. On a normal machine, you are in 227 00:10:49,080 --> 00:10:52,480 full panic mode. You're reaching for a USB 228 00:10:52,480 --> 00:10:54,720 recovery drive. You're diving into recovery 229 00:10:54,720 --> 00:10:57,120 partitions. You're wasting hours trying to 230 00:10:57,120 --> 00:11:00,520 manually uninstall the broken driver from a command line, but 231 00:11:00,520 --> 00:11:03,760 on NICs OS your reaction is completely different. 232 00:11:03,760 --> 00:11:07,080 Yeah, you don't panic at all. Yeah, you simply force a reboot. 233 00:11:07,320 --> 00:11:10,080 When the computer turns on, it presents you with a boot menu 234 00:11:10,080 --> 00:11:11,920 listing all your previous generations. 235 00:11:11,920 --> 00:11:14,040 Yeah, and you just select the generation from yesterday. 236 00:11:14,120 --> 00:11:16,720 And instantly the system boots up perfectly, because you didn't 237 00:11:16,720 --> 00:11:18,480 actually uninstall the bad update. 238 00:11:18,480 --> 00:11:20,560 The bad files are still sitting in the NICs store. 239 00:11:21,080 --> 00:11:24,760 But by selecting yesterday's generation, you told the system 240 00:11:24,760 --> 00:11:27,640 to point the entire forest of shortcuts back to the older 241 00:11:27,640 --> 00:11:31,000 working cryptographic hashes. Yeah, the transition is instant 242 00:11:31,000 --> 00:11:33,680 and mathematically guaranteed. Your entire environment is 243 00:11:33,680 --> 00:11:38,400 exactly as it was 24 hours ago. And the 2026 dossier highlights 244 00:11:38,400 --> 00:11:41,920 this as the primary reason enterprise users adopt Nix OS. 245 00:11:42,400 --> 00:11:44,920 The total eradication of update anxiety. 246 00:11:45,480 --> 00:11:48,120 You can test the most experimental bleeding edge 247 00:11:48,120 --> 00:11:51,680 software knowing that your safety net is an absolute law of 248 00:11:51,680 --> 00:11:55,120 the system architecture. OK, so we have established a 249 00:11:55,120 --> 00:11:58,360 foundation where the computer is basically indestructible. 250 00:11:58,520 --> 00:12:02,880 We have perfect isolated puzzle pieces, but having a perfect 251 00:12:02,880 --> 00:12:05,240 machine in a vacuum isn't enough, right? 252 00:12:05,520 --> 00:12:08,960 We live in a connected world. If I spend 3 weeks crafting the 253 00:12:09,000 --> 00:12:12,440 ultimate perfectly optimized developer environment, how do I 254 00:12:12,440 --> 00:12:15,760 give that exact environment to a junior developer who just joined 255 00:12:15,760 --> 00:12:17,480 my team? Right, the sharing problem. 256 00:12:17,480 --> 00:12:20,280 Yeah, if we are using hashes, we need to ensure they get the 257 00:12:20,280 --> 00:12:23,400 exact same hashes. Which brings us to a historical 258 00:12:23,400 --> 00:12:25,480 vulnerability in the NICs ecosystem. 259 00:12:26,120 --> 00:12:29,080 For years, Nick suffered from something called channel drift. 260 00:12:29,160 --> 00:12:31,320 Channel drift. Yeah, before the modern era, 261 00:12:31,480 --> 00:12:34,600 users subscribed to software channels which were essentially 262 00:12:34,600 --> 00:12:36,920 a live feed of the main package repository. 263 00:12:37,320 --> 00:12:40,200 So you and a junior developer could have the exact same 264 00:12:40,200 --> 00:12:43,560 declarative configuration file, but if you downloaded the 265 00:12:43,560 --> 00:12:47,320 channel on Tuesday and they downloaded it on Friday, the 266 00:12:47,320 --> 00:12:49,600 underlying source code in the repository might have been 267 00:12:49,600 --> 00:12:51,200 updated. The inputs changed. 268 00:12:51,320 --> 00:12:55,880 And because the inputs changed, the outputs, the hashes changed, 269 00:12:56,160 --> 00:12:59,000 you would build your system and get one set of software, and the 270 00:12:59,000 --> 00:13:01,160 junior developer would build theirs and get slightly 271 00:13:01,160 --> 00:13:03,960 different versions. So it was highly reproducible, 272 00:13:03,960 --> 00:13:06,640 but it wasn't truly hermetic. It wasn't sealed off from the 273 00:13:06,640 --> 00:13:08,840 flow of time. And that flow of time is the 274 00:13:08,960 --> 00:13:12,160 enemy of stability. And the solution to that is a 275 00:13:12,160 --> 00:13:16,000 technology that the sources say spent six years in a fierce 276 00:13:16,000 --> 00:13:19,800 experimental debate before becoming the absolute standard 277 00:13:19,800 --> 00:13:22,560 in 2026. They are called flakes. 278 00:13:22,720 --> 00:13:26,040 Flakes are the mechanism that finally achieve true hermeticity 279 00:13:26,040 --> 00:13:30,240 for NICs O They borrow a concept heavily utilized in modern web 280 00:13:30,240 --> 00:13:31,640 development, which is the lock file. 281 00:13:32,320 --> 00:13:34,920 When you use flakes, your configuration generates A flake 282 00:13:34,920 --> 00:13:37,640 dot lock file. This file rigidly pins every 283 00:13:37,640 --> 00:13:40,600 single dependency, including the entire mass of NICs package 284 00:13:40,600 --> 00:13:44,880 repository itself, to a specific immutable git commit hash. 285 00:13:45,160 --> 00:13:47,320 So it completely freezes the flow of time. 286 00:13:47,440 --> 00:13:50,680 If my lock file points to a commit from March 1st, it does 287 00:13:50,680 --> 00:13:52,880 not matter if someone tries to build my system in the year 288 00:13:52,880 --> 00:13:56,480 2030, NICs will pull the exact state of the repository from 289 00:13:56,480 --> 00:13:59,320 March 1st 2026 and compile it identically. 290 00:13:59,720 --> 00:14:03,440 I visualize flakes as a universal USB port for system 291 00:14:03,440 --> 00:14:05,640 configurations. That's a good way to look at it. 292 00:14:05,800 --> 00:14:08,840 Because everything is perfectly pinned, you can pull different 293 00:14:08,840 --> 00:14:11,040 features from all over the Internet and plug them together 294 00:14:11,040 --> 00:14:15,160 without any fear of collision. Like I can pull a complex 50 295 00:14:15,160 --> 00:14:19,120 plugin code editor configuration from a developer in Tokyo, I can 296 00:14:19,120 --> 00:14:22,760 grab a high performance audio routing flake from a producer in 297 00:14:22,760 --> 00:14:24,640 London. I dropped them into my 298 00:14:24,640 --> 00:14:27,680 configuration file and they snap together perfectly because their 299 00:14:27,680 --> 00:14:30,640 dependencies are frozen in their own respective lock files. 300 00:14:30,760 --> 00:14:33,000 That composability has completely changed how 301 00:14:33,000 --> 00:14:36,320 configurations are shared today, and the technical maturation of 302 00:14:36,320 --> 00:14:39,120 flakes mirrors a broader organizational maturation 303 00:14:39,120 --> 00:14:43,680 detailed in the sources. Around 2024 and 2025, the Knicks 304 00:14:43,680 --> 00:14:46,560 community went through a pretty significant governance schism. 305 00:14:47,120 --> 00:14:50,120 Oh right, I read about that. Yeah, it was the classic tension 306 00:14:50,120 --> 00:14:53,440 between the original hardcore academic purists who built the 307 00:14:53,440 --> 00:14:57,280 language and a rapidly expanding commercial ecosystem that really 308 00:14:57,280 --> 00:15:00,200 needed enterprise stability. Just the growing pains of an 309 00:15:00,200 --> 00:15:02,840 open source project suddenly becoming critically important to 310 00:15:02,840 --> 00:15:04,680 Fortune 500 companies. Exactly. 311 00:15:04,760 --> 00:15:08,480 But the 2026 materials show that the communities successfully 312 00:15:08,480 --> 00:15:11,720 navigated that. They drafted A Nix governance 313 00:15:11,720 --> 00:15:14,880 constitution and established a formal steering committee. 314 00:15:15,560 --> 00:15:19,160 And that institutional trust has unlocked serious capital and 315 00:15:19,160 --> 00:15:20,360 hardware partnerships. Real. 316 00:15:20,360 --> 00:15:22,640 Investment. Yeah, We are seeing major 317 00:15:22,640 --> 00:15:26,280 financial backing like the €100,000 grant from Restack to 318 00:15:26,280 --> 00:15:29,400 fund core development. Even more significantly, we are 319 00:15:29,400 --> 00:15:32,920 seeing official OEM support companies like Framework are 320 00:15:32,920 --> 00:15:35,960 partnering to guarantee Nix OS runs Florida State 321 00:15:35,960 --> 00:15:39,280 out-of-the-box on their newest silicon like the Ryzen AI 300 322 00:15:39,280 --> 00:15:42,120 processors. They are also directly attacking 323 00:15:42,120 --> 00:15:43,720 that learning curve we mentioned earlier. 324 00:15:44,240 --> 00:15:46,760 The sources detail a new tool called Nixmate. 325 00:15:47,240 --> 00:15:50,800 It's a terminal user interface that acts as a bridge for new 326 00:15:50,800 --> 00:15:53,920 users. Historically, if you made a typo 327 00:15:53,920 --> 00:15:57,360 in a Nix configuration, the terminal would spit out a wall 328 00:15:57,360 --> 00:16:00,320 of esoteric functional programming trace backs that 329 00:16:00,320 --> 00:16:03,200 looked like ancient runes. They were terrifying. 330 00:16:03,200 --> 00:16:07,320 Right, but Nix may intercepts those trace backs and uses local 331 00:16:07,360 --> 00:16:09,760 AI to translate them into plain English. 332 00:16:10,000 --> 00:16:14,040 It tells you, hey, you forgot a semi colon on line 42 and you're 333 00:16:14,040 --> 00:16:15,920 trying to mix a string with a boolean. 334 00:16:16,360 --> 00:16:19,160 It lulls the barrier to entry significantly. 335 00:16:19,360 --> 00:16:21,960 And lowering that barrier is so crucial because there is one 336 00:16:21,960 --> 00:16:25,200 sector of the tech industry that is flocking to Nix OS faster 337 00:16:25,200 --> 00:16:28,000 than any other, driven by just sheer desperation. 338 00:16:28,080 --> 00:16:30,000 And that is the world of artificial intelligence. 339 00:16:30,760 --> 00:16:32,760 This brings us to the Creator's Paradox. 340 00:16:33,120 --> 00:16:35,960 We've established that NICs provides perfect mathematical 341 00:16:35,960 --> 00:16:37,960 stability. But what happens when you 342 00:16:37,960 --> 00:16:41,480 introduce the messiest, most chaotic, most rapidly updating 343 00:16:41,480 --> 00:16:44,080 software stack on the planet into that pure math? 344 00:16:44,840 --> 00:16:47,840 AI is the ultimate stress test for an operating system right 345 00:16:47,840 --> 00:16:48,760 now. It really is. 346 00:16:48,840 --> 00:16:52,480 It is arguably the most cutting edge technology we possess, but 347 00:16:52,480 --> 00:16:55,800 underneath it is built on an incredibly fragile, crumbling 348 00:16:55,800 --> 00:16:58,960 foundation. The dependency nightmare in AI 349 00:16:58,960 --> 00:17:02,160 research is just staggering. If you speak with anyone 350 00:17:02,440 --> 00:17:05,560 training local large language models or developing computer 351 00:17:05,560 --> 00:17:08,280 vision neural networks, they will describe an environment of 352 00:17:08,280 --> 00:17:11,480 constant fear. Their stack relies on a highly 353 00:17:11,480 --> 00:17:14,839 volatile mixture of Python virtual environments, specific 354 00:17:14,839 --> 00:17:20,160 versions of the NVIDIA CUDA tool chain, native C++ bindings, and 355 00:17:20,160 --> 00:17:22,359 these deeply nested mathematical libraries. 356 00:17:22,400 --> 00:17:24,480 I want to break down exactly what that friction looks like 357 00:17:24,480 --> 00:17:26,520 for you. Let's say you are three weeks in 358 00:17:26,520 --> 00:17:29,640 training a complex model. Your computer runs an automatic 359 00:17:29,640 --> 00:17:32,400 background update that bumps your Linux kernel version up by 360 00:17:32,400 --> 00:17:35,560 a tiny fraction. Suddenly, your NVIDIA driver no 361 00:17:35,560 --> 00:17:37,320 longer perfectly aligns with the kernel. 362 00:17:37,320 --> 00:17:41,200 The CUDA toolkit, which expects the old driver, miscommunicates 363 00:17:41,200 --> 00:17:43,960 with your Python library. When you run your training 364 00:17:43,960 --> 00:17:47,880 script, the CPU tries to access a block of memory that the GPU 365 00:17:47,880 --> 00:17:50,840 thinks is reserved. That is a segmentation fault. 366 00:17:51,240 --> 00:17:54,800 The system panics, the process is killed instantly, and you 367 00:17:54,800 --> 00:17:58,840 lose days of computational work. All because a background process 368 00:17:58,840 --> 00:18:01,240 updated a file you didn't even know existed. 369 00:18:01,240 --> 00:18:04,040 That is the daily reality of the mutable imperative model we 370 00:18:04,040 --> 00:18:06,440 discussed earlier. Yeah, but Nix OS solves this 371 00:18:06,480 --> 00:18:09,760 elegantly through flakes. An AI researcher can publish 372 00:18:09,760 --> 00:18:12,680 their groundbreaking paper on GitHub and include a single 373 00:18:12,680 --> 00:18:15,800 flake dot NICs file. You can clone that repository. 374 00:18:15,800 --> 00:18:17,720 You don't have to install Python, you don't have to 375 00:18:17,720 --> 00:18:20,720 manually download drivers. You just type a single command. 376 00:18:20,720 --> 00:18:22,920 NICs develop. And the system builds an 377 00:18:22,920 --> 00:18:25,160 invisible force field around that project. 378 00:18:25,200 --> 00:18:27,840 Basically, yeah, yeah, it creates a perfectly isolated 379 00:18:27,840 --> 00:18:31,000 development shell. The flake pins the exact Linux 380 00:18:31,000 --> 00:18:34,560 kernel version, the exact NVIDIA driver version, and the specific 381 00:18:34,560 --> 00:18:36,360 commit of Pytorch the researcher used. 382 00:18:37,080 --> 00:18:39,400 It configures the environment paths so the code points 383 00:18:39,400 --> 00:18:42,200 strictly to those specific hash libraries in the next store. 384 00:18:42,560 --> 00:18:45,120 That's huge. And here is the truly 385 00:18:45,120 --> 00:18:47,640 revolutionary part. Whether you are opening that 386 00:18:47,640 --> 00:18:50,920 project on a local MacBook or deploying it to a massive 387 00:18:50,920 --> 00:18:55,760 $1,000,000 H 100 server cluster, the evaluation is bit for bit 388 00:18:55,760 --> 00:18:58,640 identical. There is 0 drift. 389 00:18:59,320 --> 00:19:02,200 But I want to push back on this perfect AI utopia because 390 00:19:02,200 --> 00:19:05,520 historically Nix OS and AI hardware were a terrible 391 00:19:05,520 --> 00:19:08,640 combination. It was known as the CUDA 392 00:19:08,640 --> 00:19:11,840 conundrum. Nix OS ensures purity by 393 00:19:11,840 --> 00:19:15,680 building software from source code, but Nvidia's CDA toolkit 394 00:19:15,680 --> 00:19:19,400 is massive, and compiling complex AI software with CUDA 395 00:19:19,400 --> 00:19:21,960 enabled from raw source code could take hours. 396 00:19:22,240 --> 00:19:24,880 I've read horror stories of researchers waiting six hours 397 00:19:24,880 --> 00:19:26,880 just to see if their AI environment would even boot, 398 00:19:27,080 --> 00:19:28,520 only to realize they made a typo. 399 00:19:28,680 --> 00:19:31,320 That kind of latency is unacceptable in a fast-paced 400 00:19:31,320 --> 00:19:33,200 research environment. Oh, it was a massive bottleneck, 401 00:19:33,600 --> 00:19:36,960 but the 2026 standard has entirely bypassed that issue 402 00:19:37,160 --> 00:19:39,760 thanks to a dedicated team called CUDA Maintainers and 403 00:19:39,760 --> 00:19:42,960 Enterprise tools like Flux. Instead of forcing your laptop 404 00:19:42,960 --> 00:19:45,360 to compile everything from scratch, Flux maintains a 405 00:19:45,360 --> 00:19:48,360 colossal geographically distributed binary cache. 406 00:19:48,480 --> 00:19:52,120 They prebuild all these massive CUD binaries on their own server 407 00:19:52,120 --> 00:19:54,320 farms in a legally compliant way. 408 00:19:54,480 --> 00:19:57,920 So when you type NICs develop your system calculates the hash 409 00:19:57,920 --> 00:20:01,480 of what it needs, checks the flux cache, sees that the exact 410 00:20:01,480 --> 00:20:04,040 hash already exists, and just downloads the finished product. 411 00:20:04,320 --> 00:20:07,040 A process that took six hours now takes 30 seconds. 412 00:20:07,040 --> 00:20:09,760 Exactly, and because of the architectural isolation of the 413 00:20:09,760 --> 00:20:12,560 next door, it unlocks the capability that is practically 414 00:20:12,560 --> 00:20:16,840 impossible on Ubuntu or Windows running multiple CDA version 415 00:20:16,840 --> 00:20:19,320 side by side. Really on the same machine? 416 00:20:19,320 --> 00:20:22,080 Yes, you could have a legacy computer vision project that 417 00:20:22,080 --> 00:20:26,120 strictly requires CDA 11.8 for stability running in one 418 00:20:26,120 --> 00:20:28,400 terminal window. In another window you could have 419 00:20:28,400 --> 00:20:32,680 a cutting edge LLM training script utilizing CDA 12.8 for 420 00:20:32,680 --> 00:20:34,960 maximum speed. They run simultaneously on the 421 00:20:34,960 --> 00:20:37,760 exact same GPU without ever interfering with each other. 422 00:20:37,920 --> 00:20:40,720 That is incredible, but we have to address the elephant in the 423 00:20:40,720 --> 00:20:43,960 room when it comes to Python. Python is the language of AI, 424 00:20:44,320 --> 00:20:47,000 and the way Python manages packages is through a tool 425 00:20:47,000 --> 00:20:48,960 called pip. You type pip, installed 426 00:20:48,960 --> 00:20:52,040 Tensorflow, and it goes to the Internet, grabs the software, 427 00:20:52,040 --> 00:20:53,320 and dumps it onto your hard drive. 428 00:20:53,960 --> 00:20:57,240 But I'm guessing pip is fundamentally incompatible with 429 00:20:57,240 --> 00:20:59,480 Nix OS. It is entirely incompatible. 430 00:20:59,480 --> 00:21:01,240 Pip is the ultimate imperative tool. 431 00:21:01,640 --> 00:21:05,320 It downloads pre compiled wheels that assume a standard file 432 00:21:05,320 --> 00:21:07,760 system exists. It wants to dump files into 433 00:21:07,760 --> 00:21:10,640 slash oser slash Lib slash Python, which as we established 434 00:21:10,640 --> 00:21:14,360 doesn't exist on Nix OS, right? And even if it did, the system 435 00:21:14,360 --> 00:21:16,640 is strictly read only to prevent entropy. 436 00:21:17,440 --> 00:21:20,840 Running pip install in Nix OS is effectively trying to mutate an 437 00:21:20,840 --> 00:21:23,040 immutable system. So how do you actually write 438 00:21:23,160 --> 00:21:25,320 Python code? If I can't use the standard 439 00:21:25,320 --> 00:21:27,080 package manager, how do I get my libraries? 440 00:21:27,080 --> 00:21:29,800 The ecosystem has shifted to transplation tools, with the 441 00:21:29,800 --> 00:21:32,280 2020 sixth standard being UV Tunex. 442 00:21:32,640 --> 00:21:35,320 Instead of running PIP, you write out your required Python 443 00:21:35,320 --> 00:21:37,240 libraries in a standard requirements file. 444 00:21:37,840 --> 00:21:41,440 UV2 next reads that file and mathematically translates those 445 00:21:41,440 --> 00:21:43,360 requirements into native NICs derivations. 446 00:21:43,400 --> 00:21:45,080 Oh wow. Yeah, it figures out every 447 00:21:45,080 --> 00:21:49,800 underlying C++ dependency those Python libraries need, hashes 448 00:21:49,800 --> 00:21:53,200 them, downloads them into the NICs store, and strictly links 449 00:21:53,200 --> 00:21:55,440 them. It takes the chaotic imperative 450 00:21:55,440 --> 00:21:59,400 nature of Python And forces it into the declarative pure order 451 00:21:59,400 --> 00:22:01,240 of NICs. Here's where it gets really 452 00:22:01,240 --> 00:22:03,840 interesting, because it forces order onto chaos. 453 00:22:04,400 --> 00:22:07,600 But this philosophical commitment to pure, observable 454 00:22:07,600 --> 00:22:10,680 mathematics brings us to a massive divide in the hardware 455 00:22:10,680 --> 00:22:12,960 world. The dossier refers to it as the 456 00:22:12,960 --> 00:22:14,960 GPU. Wall the GPU wall, yeah. 457 00:22:15,240 --> 00:22:17,760 When you are building an operating system where every 458 00:22:17,760 --> 00:22:21,040 input must be hashed and verified, proprietary hardware 459 00:22:21,040 --> 00:22:23,320 drivers become a fundamental roadblock. 460 00:22:23,640 --> 00:22:28,040 And the text makes it very clear AMD is the absolute golden child 461 00:22:28,040 --> 00:22:30,280 of the Nix OS ecosystem. It all comes down to 462 00:22:30,280 --> 00:22:33,600 transparency. AM DS compute drivers, the ROCM 463 00:22:33,600 --> 00:22:35,600 stack, are completely open source. 464 00:22:35,600 --> 00:22:37,600 They are baked directly into the Linux kernel. 465 00:22:38,120 --> 00:22:40,120 Because the source code is public and visible. 466 00:22:40,320 --> 00:22:42,920 It fits perfectly into Nix's declarative philosophy. 467 00:22:43,120 --> 00:22:44,600 It's observable. Exactly. 468 00:22:44,840 --> 00:22:47,240 The build system can observe the code, calculate the 469 00:22:47,240 --> 00:22:49,840 cryptographic hash, and manage it natively. 470 00:22:50,240 --> 00:22:54,080 For a user with an AMD graphics card, setting up a complex AI 471 00:22:54,080 --> 00:22:57,600 rig requires exactly one line of code in their configuration 472 00:22:57,600 --> 00:23:01,480 file. Hardware dot MDGPU dot row, TMP 473 00:23:01,480 --> 00:23:08,000 dot row CMW enable equals true. You type that, rebuild your 474 00:23:08,000 --> 00:23:11,040 system, and you have a fully functional, mathematically pure 475 00:23:11,040 --> 00:23:13,280 compute environment. It's the dream of open source 476 00:23:13,280 --> 00:23:16,360 computing, but the reality of the AI industry is that NVIDIA 477 00:23:16,360 --> 00:23:20,000 holds a functional monopoly. NVIDIA is the undisputed king of 478 00:23:20,120 --> 00:23:23,640 AI hardware, but their drivers are closed source proprietary 479 00:23:23,640 --> 00:23:25,800 binary blobs. They do not share the source 480 00:23:25,800 --> 00:23:28,240 code. So how does Nix OS, an operating 481 00:23:28,240 --> 00:23:30,880 system built on observing and hashing source code, deal with a 482 00:23:30,880 --> 00:23:33,000 black box? It treats it as an impurity. 483 00:23:33,000 --> 00:23:35,600 An impurity. Yeah, to run NVIDIA hardware, 484 00:23:35,600 --> 00:23:38,320 users have to explicitly flag their configuration to allow 485 00:23:38,320 --> 00:23:41,720 unfree proprietary packages. The system has to blindly accept 486 00:23:41,720 --> 00:23:43,840 the NVIDIA BLOB into the Nix store without fully 487 00:23:43,840 --> 00:23:45,600 understanding its internal dependencies. 488 00:23:45,800 --> 00:23:48,520 This introduces fragility. Because it can't mathematically 489 00:23:48,520 --> 00:23:49,480 verify it. Right. 490 00:23:50,040 --> 00:23:53,240 The dossier explicitly notes that NVIDIA users pay a heavy 491 00:23:53,240 --> 00:23:57,160 configuration tax because NICs cannot perfectly manage the 492 00:23:57,160 --> 00:23:59,200 internal state of those closed drivers. 493 00:23:59,600 --> 00:24:02,240 Users spend significantly more time wrestling with kernel 494 00:24:02,240 --> 00:24:05,600 parameters, debugging power management States, and fixing 495 00:24:05,600 --> 00:24:08,160 display rendering issues than AMD users do. 496 00:24:08,560 --> 00:24:11,560 It is the tax you pay to access the CUDA ecosystem. 497 00:24:12,080 --> 00:24:14,880 But this friction with closed source software transitions us 498 00:24:15,000 --> 00:24:19,040 perfectly into our next topic. AI tools, despite their messy 499 00:24:19,040 --> 00:24:21,880 Python dependencies, are generally open source at their 500 00:24:21,880 --> 00:24:25,600 core, so the community can build transpilers like you do tunics 501 00:24:25,600 --> 00:24:26,840 to wrangle them. True. 502 00:24:27,520 --> 00:24:30,440 But what happens when you hit a massive proprietary closed 503 00:24:30,440 --> 00:24:33,520 source enterprise application that actively refuses to be 504 00:24:33,520 --> 00:24:35,760 wrangled? What happens when you try to run 505 00:24:35,760 --> 00:24:38,440 professional creative software on a pure math machine? 506 00:24:38,560 --> 00:24:41,440 You encounter what the dossier vividly describes as the final 507 00:24:41,440 --> 00:24:44,680 boss of Nix OS configuration, and that is Davinci Resolve. 508 00:24:45,120 --> 00:24:48,160 Davinci Resolve, yeah. Davinci Resolve is a monolithic 509 00:24:48,160 --> 00:24:50,840 video editing suite used by Hollywood professionals. 510 00:24:51,360 --> 00:24:53,720 It is the antithesis of the Nix philosophy. 511 00:24:54,280 --> 00:24:57,440 It is hard coded to expect a standard messy Linux file 512 00:24:57,440 --> 00:24:59,080 system. When it launches. 513 00:24:59,280 --> 00:25:02,880 Its code explicitly looks for its licensing files in slash opt 514 00:25:03,240 --> 00:25:07,360 and its interface libraries in slash user slash Lib 64. 515 00:25:07,360 --> 00:25:10,200 And when it wakes up inside Nix OS and finds that those folders 516 00:25:10,200 --> 00:25:13,440 literally do not exist and everything is hidden behind 32 517 00:25:13,440 --> 00:25:16,120 character cryptographic hashes, it panics. 518 00:25:16,480 --> 00:25:19,320 It throws tantrums over Q2 interface library version 519 00:25:19,320 --> 00:25:22,040 mismatches and simply refuses to open. 520 00:25:22,160 --> 00:25:24,520 It cannot comprehend the environment it is running in. 521 00:25:24,520 --> 00:25:27,000 So the community had to engineer workarounds to trick the 522 00:25:27,000 --> 00:25:29,480 software. The early attempt was the native 523 00:25:29,480 --> 00:25:32,080 wrapper, often called an FHS user NV. 524 00:25:32,120 --> 00:25:34,560 Let me guess how this works. If Resolve is looking for a 525 00:25:34,560 --> 00:25:36,920 library in slash year slash Lib, the native wrapper is 526 00:25:36,920 --> 00:25:39,560 essentially a massive script that intercepts that request. 527 00:25:40,120 --> 00:25:43,440 Right as Resolve reaches for slash slash Lib, the script 528 00:25:43,440 --> 00:25:45,960 catches it, looks up the correct hash in the next store, and 529 00:25:45,960 --> 00:25:48,160 hands the file back, pretending it came from the standard 530 00:25:48,160 --> 00:25:50,240 folder. That is exactly the mechanism. 531 00:25:50,240 --> 00:25:53,600 It is a real time translation layer, but as you can imagine it 532 00:25:53,600 --> 00:25:57,320 is notoriously fragile. I bet Davinci Resolve is a 533 00:25:57,320 --> 00:25:59,960 massive program with thousands of dependencies. 534 00:26:00,560 --> 00:26:03,680 If Black Magic Design releases a minor brain update that looks 535 00:26:03,680 --> 00:26:06,720 for a new library, and the Rapos script hasn't been manually 536 00:26:06,720 --> 00:26:09,800 updated by a community member to intercept that specific request, 537 00:26:10,160 --> 00:26:13,560 the application crashes. It is a constant game of cat and 538 00:26:13,560 --> 00:26:15,880 mouse. Which is why the 2026 539 00:26:15,880 --> 00:26:18,640 Professional Standard abandoned the wrapper completely. 540 00:26:18,800 --> 00:26:22,360 They moved to a containerized approach widely known as Davinci 541 00:26:22,360 --> 00:26:24,560 Box. This is fascinating. 542 00:26:24,880 --> 00:26:27,760 They use technologies like Distrobox and Podman to create a 543 00:26:27,760 --> 00:26:30,840 localized, immutable container based on a traditional 544 00:26:30,840 --> 00:26:33,760 enterprise distribution like Rocky Linux or Fedora. 545 00:26:33,800 --> 00:26:36,160 It is an incredibly pragmatic compromise. 546 00:26:36,360 --> 00:26:38,800 You take this demanding, inflexible enterprise 547 00:26:38,800 --> 00:26:41,440 application and you put it inside a container where it 548 00:26:41,440 --> 00:26:43,560 believes it is running on a standard Fedora system. 549 00:26:43,760 --> 00:26:45,920 It sees all the traditional file pads it wants. 550 00:26:46,160 --> 00:26:49,000 It has direct pass through access to your GPU for 551 00:26:49,000 --> 00:26:51,080 rendering. It is completely happy in its 552 00:26:51,080 --> 00:26:53,280 little illusion. I think of it as a padded cell 553 00:26:53,280 --> 00:26:56,000 for software. You lock the messy application 554 00:26:56,000 --> 00:26:58,840 in a padded room where it can't hurt itself or anything else. 555 00:26:59,120 --> 00:27:03,400 And the beauty is that outside that padded cell, your core Nix 556 00:27:03,520 --> 00:27:06,880 OS system remains totally untouched, secure and 557 00:27:06,880 --> 00:27:10,160 mathematically pure. You get the stability of Nix OS 558 00:27:10,160 --> 00:27:12,920 with the compatibility of a mainstream distro. 559 00:27:13,080 --> 00:27:15,720 And that stability is highly sought after by other Korean 560 00:27:15,720 --> 00:27:18,040 professionals as well. Consider professional 561 00:27:18,040 --> 00:27:21,080 photographers working with massive RAW image files. 562 00:27:21,920 --> 00:27:24,800 They rely on tools like Darktable, which lean heavily on 563 00:27:24,800 --> 00:27:28,320 an API called OpenCL to use the graphics card for hardware 564 00:27:28,320 --> 00:27:31,320 accelerated processing. On a traditional operating 565 00:27:31,320 --> 00:27:33,360 system. A random background update to a 566 00:27:33,360 --> 00:27:36,400 graphics library can sever the link between Darktable and 567 00:27:36,400 --> 00:27:38,800 OpenCL. Right in the middle of a massive 568 00:27:38,800 --> 00:27:42,560 editing project, suddenly your export times jump from minutes 569 00:27:42,560 --> 00:27:44,440 to hours. That's agonizing. 570 00:27:44,800 --> 00:27:47,600 But on Nix OS, because every connection is declaratively 571 00:27:47,600 --> 00:27:50,400 define, that hardware acceleration state is frozen. 572 00:27:50,520 --> 00:27:53,080 You rebuild your system and you are mathematically guaranteed 573 00:27:53,080 --> 00:27:55,720 that your photo editing suite will behave the exact same way 574 00:27:55,720 --> 00:27:58,000 it did yesterday. The audio engineering space 575 00:27:58,040 --> 00:28:01,560 benefits similarly. Podcasters and video editors in 576 00:28:01,560 --> 00:28:06,040 2026 are leveraging Nix OS's granular control over the Linux 577 00:28:06,040 --> 00:28:09,880 audio server pipe wire. In professional audio, latency 578 00:28:09,880 --> 00:28:11,440 is the enemy. Definitely. 579 00:28:11,440 --> 00:28:13,920 If you speak into a microphone and hear it in your headphones a 580 00:28:13,920 --> 00:28:16,400 fraction of second later, it is disorienting. 581 00:28:17,200 --> 00:28:19,960 To fix this, you have to run the audio server at a very low 582 00:28:19,960 --> 00:28:22,400 quantum, which is essentially the buffer size. 583 00:28:23,040 --> 00:28:26,640 A quantum of 64 samples means the system processes audio in 584 00:28:26,760 --> 00:28:30,440 tiny rapid fire chunks, resulting in near 0 latency. 585 00:28:30,800 --> 00:28:34,000 On a normal machine, you have to open a graphical settings menu, 586 00:28:34,000 --> 00:28:37,240 drag a slider to 64 samples, and pray that a system reboot 587 00:28:37,240 --> 00:28:38,720 doesn't reset it back to default. 588 00:28:39,160 --> 00:28:41,880 But on XOS you don't fiddle with graphical menus hoping the 589 00:28:41,880 --> 00:28:44,720 settings stick. You write services dot pipewire 590 00:28:44,720 --> 00:28:47,920 dot extraconfig dot quantum equals 64 directly into your 591 00:28:47,920 --> 00:28:50,720 system definition. The system evaluates it and it 592 00:28:50,720 --> 00:28:52,720 becomes a hard law of the operating system. 593 00:28:53,440 --> 00:28:55,720 The audio server literally cannot start any other way. 594 00:28:55,840 --> 00:28:58,560 It is the ultimate manifestation of infrastructure as code 595 00:28:58,560 --> 00:29:01,440 applied to personal creativity. You define the physical behavior 596 00:29:01,440 --> 00:29:02,720 of the hardware through a text file. 597 00:29:02,760 --> 00:29:06,320 So we have built a machine that is virtually indestructible, it 598 00:29:06,320 --> 00:29:10,160 can seamlessly isolate chaotic AI dependencies, and it can cage 599 00:29:10,160 --> 00:29:12,840 a massive video editing suite in a padded cell. 600 00:29:13,160 --> 00:29:16,640 We've conquered the professional workloads, but let's look at the 601 00:29:16,640 --> 00:29:19,960 ultimate stress test for raw hardware performance and real 602 00:29:19,960 --> 00:29:23,080 time responsiveness. What happens when we are done 603 00:29:23,080 --> 00:29:27,200 working and we just want to play a demanding video game without 604 00:29:27,200 --> 00:29:29,280 lag? The gaming renaissance on Linux 605 00:29:29,280 --> 00:29:31,480 is one of the most significant tech narratives of the twenty 606 00:29:31,480 --> 00:29:34,760 20s, heavily driven by the massive success of Val's Steam 607 00:29:34,760 --> 00:29:35,080 Deck. Oh. 608 00:29:35,240 --> 00:29:38,000 Absolutely. By 2026, the maturation of the 609 00:29:38,000 --> 00:29:41,560 Proton 10 compatibility layer, which translates Windows DirectX 610 00:29:41,560 --> 00:29:45,560 calls into Linux Vulcan calls in real time, means that gaming on 611 00:29:45,560 --> 00:29:48,000 Nix OS is no longer an experimental headache. 612 00:29:48,280 --> 00:29:51,240 It is a first class premium experience, and true to the Nix 613 00:29:51,240 --> 00:29:54,000 philosophy, it is entirely managed declaratively. 614 00:29:54,160 --> 00:29:56,320 I remember the days of trying to game on Ubuntu. 615 00:29:56,400 --> 00:29:59,080 You would open the terminal, run apt, install steam, and then 616 00:29:59,080 --> 00:30:02,360 spend 3 hours tracking down obscure 32 bit graphics 617 00:30:02,360 --> 00:30:06,440 libraries because a game from 2013 refused to launch on Nix 618 00:30:06,600 --> 00:30:08,360 OS. You add one line to your config, 619 00:30:08,520 --> 00:30:10,800 programs dot, steam dot enable equals true. 620 00:30:11,040 --> 00:30:13,600 And I want to emphasize what that one line actually does 621 00:30:13,600 --> 00:30:15,760 under the hood. It doesn't just download the 622 00:30:15,760 --> 00:30:18,920 Steam binary, it dynamically constructs the perfect 623 00:30:18,920 --> 00:30:21,760 environment for it. It automatically opens the 624 00:30:21,760 --> 00:30:24,640 necessary firewall port so your local Steam link streaming 625 00:30:24,640 --> 00:30:27,240 works. It seeks out and perfectly links 626 00:30:27,240 --> 00:30:30,200 all those arcane 32 bit libraries that older Windows 627 00:30:30,200 --> 00:30:32,880 games demand, and it isolates them so they don't pollute your 628 00:30:32,880 --> 00:30:34,920 main system. It abstracts an incredible 629 00:30:34,920 --> 00:30:38,120 amount of complexity, but the true power users on NICs OS 630 00:30:38,360 --> 00:30:40,040 don't stop at the default settings. 631 00:30:40,240 --> 00:30:42,920 The dossier highlights how gamers utilize the declarative 632 00:30:42,920 --> 00:30:46,360 nature of the OS to hot swap the very core of their computer to 633 00:30:46,360 --> 00:30:48,920 maximize frame rates. Like changing the kernel. 634 00:30:48,920 --> 00:30:51,960 Exactly. In 2026, the standard choice for 635 00:30:51,960 --> 00:30:55,320 gamers is the xanmod kernel. This is a custom Linux kernel 636 00:30:55,320 --> 00:30:58,520 specifically optimized for desktop responsiveness rather 637 00:30:58,520 --> 00:31:00,880 than server throughput. The technical specs on the 638 00:31:00,920 --> 00:31:02,920 xanmod kernel integration are wild. 639 00:31:03,600 --> 00:31:07,120 The dossier mentions it utilizes LLVM's thin LTO. 640 00:31:07,680 --> 00:31:10,120 For those who aren't compile engineers, Link Time 641 00:31:10,120 --> 00:31:14,120 optimization, or LTO analyzes the entire program all at once 642 00:31:14,120 --> 00:31:16,960 during the final compilation step, rather than file by file. 643 00:31:17,480 --> 00:31:20,560 It allows the compiler to strip away unused code and inline 644 00:31:20,560 --> 00:31:23,360 functions across different parts of the kernel, resulting in a 645 00:31:23,360 --> 00:31:25,840 significantly leaner, faster operating system. 646 00:31:26,120 --> 00:31:29,000 But the real game changer mentioned is the **NSYNC driver, 647 00:31:29,000 --> 00:31:30,760 the. **NSYNC driver is a fascinating 648 00:31:30,760 --> 00:31:33,160 piece of engineering. When Pro Con translates a 649 00:31:33,160 --> 00:31:36,360 Windows came to run on Linux, one of the biggest bottlenecks 650 00:31:36,360 --> 00:31:39,200 isn't rendering the graphics, it's thread synchronization, 651 00:31:39,360 --> 00:31:41,440 Right? Windows and Linux handle memory 652 00:31:41,440 --> 00:31:43,840 and threading synchronization fundamentally differently. 653 00:31:44,000 --> 00:31:46,720 Historically, Proton had to emulate the Windows method in 654 00:31:46,720 --> 00:31:49,960 user space, which ate up a massive amount of CPU cycles. 655 00:31:50,600 --> 00:31:53,120 The *NSYNC driver moves that emulation directly into the 656 00:31:53,120 --> 00:31:55,560 Linux kernel itself. It provides native Windows 657 00:31:55,560 --> 00:31:58,120 synchronization primitives. Bypassing the user space 658 00:31:58,120 --> 00:32:01,920 translation completely. That drastically reduces the CPU 659 00:32:01,920 --> 00:32:05,240 overhead, meaning your processor can spend more time pushing high 660 00:32:05,240 --> 00:32:07,840 frame rates instead of translating background threads. 661 00:32:08,000 --> 00:32:11,840 Exactly and visually a major milestone has been reached in 662 00:32:11,840 --> 00:32:14,480 this era. Wayland HDR stabilization. 663 00:32:15,120 --> 00:32:18,880 With modern desktop environments like KD Plasma 6.8 and newer, 664 00:32:19,160 --> 00:32:22,680 users can finally experience true high dynamic range gaming 665 00:32:22,680 --> 00:32:25,640 on Linux, perfectly matching the visual fidelity of Windows. 666 00:32:26,240 --> 00:32:29,520 However, the dossier points out that this is 1 specific area 667 00:32:29,720 --> 00:32:32,560 where the strict purity of the next door still causes friction. 668 00:32:32,600 --> 00:32:35,440 I was reading about that the color calibration issue games 669 00:32:35,440 --> 00:32:38,720 and display calibration tools often rely on standard ICC color 670 00:32:38,720 --> 00:32:41,000 profiles. They are hard coded to look for 671 00:32:41,000 --> 00:32:44,400 these profiles and very specific directories, but because the Nix 672 00:32:44,400 --> 00:32:47,600 store is strictly read only and isolated, applying a custom 673 00:32:47,600 --> 00:32:50,400 color profile you downloaded from a monitor manufacturer can 674 00:32:50,400 --> 00:32:53,000 be surprisingly difficult. It's not a simple drag and drop. 675 00:32:53,200 --> 00:32:55,880 No, you can't just drop the file into a system folder. 676 00:32:55,880 --> 00:33:00,040 You have to use user level tools like Home Manager to explicitly 677 00:33:00,040 --> 00:33:03,240 weave assembling of that color profile into your environment so 678 00:33:03,240 --> 00:33:04,640 the graphical server can find it. 679 00:33:04,640 --> 00:33:06,840 It is a recurring theme with Nix OS. 680 00:33:07,600 --> 00:33:10,120 The architectural purity that makes the system unbreakable 681 00:33:10,400 --> 00:33:14,240 also requires you to be ruthlessly explicit about every 682 00:33:14,240 --> 00:33:17,000 single connection. You cannot rely on implied 683 00:33:17,000 --> 00:33:18,960 state. But the most fascinating, 684 00:33:18,960 --> 00:33:22,560 slightly unhinged workaround in the entire gaming section of the 685 00:33:22,560 --> 00:33:25,920 dossier is how Nix OS users handle anti cheat software. 686 00:33:26,600 --> 00:33:28,600 This is the ultimate question for PC gamers. 687 00:33:29,040 --> 00:33:32,240 Can I play competitive games like Valorant or Apex Legends? 688 00:33:32,280 --> 00:33:35,520 Games with kernel level anti cheat are notoriously hostile to 689 00:33:35,520 --> 00:33:37,840 Linux environments and if you were playing high stakes 690 00:33:37,840 --> 00:33:40,600 competitive shooters you need absolute maximum hardware 691 00:33:40,600 --> 00:33:42,640 performance with 0 background interference. 692 00:33:42,880 --> 00:33:46,280 The community engineered a brilliant solution using the 693 00:33:46,280 --> 00:33:50,080 core architecture of the system, the isolated generation. 694 00:33:50,680 --> 00:33:54,800 Because Nix OS can evaluate, build and boot into completely 695 00:33:54,800 --> 00:33:57,840 different system states without overriding the previous one, 696 00:33:58,440 --> 00:34:01,560 hardcore gamers create a separate, highly minimal boot 697 00:34:01,560 --> 00:34:05,000 generation specifically and only for competitive gaming. 698 00:34:05,160 --> 00:34:08,440 They build an entirely distinct personality for their computer. 699 00:34:08,520 --> 00:34:12,159 Yes, in this specific gaming generation they strip away 700 00:34:12,159 --> 00:34:15,400 everything unnecessary. No background docker containers, 701 00:34:15,400 --> 00:34:18,199 no development tools. But more importantly, they strip 702 00:34:18,199 --> 00:34:21,199 away critical security hardening features at the kernel level. 703 00:34:21,760 --> 00:34:24,600 For example, they will intentionally disable Spectre V2 704 00:34:24,600 --> 00:34:26,120 mitigations. Let me explain what that 705 00:34:26,120 --> 00:34:28,000 actually means, because it sounds terrifying. 706 00:34:28,000 --> 00:34:31,040 To disable security, modern CPU's use a technique called 707 00:34:31,040 --> 00:34:33,560 speculative execution. They try to guess what 708 00:34:33,560 --> 00:34:36,120 calculation you will need next and do it in advance to save 709 00:34:36,120 --> 00:34:38,000 time. But years ago, researchers 710 00:34:38,000 --> 00:34:40,120 discovered that malicious programs could observe this 711 00:34:40,120 --> 00:34:42,400 guessing process to steal sensitive data. 712 00:34:42,520 --> 00:34:44,440 That is the Spectre vulnerability. 713 00:34:44,639 --> 00:34:47,800 To fix it, operating systems added mitigations that 714 00:34:47,800 --> 00:34:50,480 constantly double check the CPU's guesses. 715 00:34:50,920 --> 00:34:54,280 It makes the system safe, but it slows the CPU down. 716 00:34:54,639 --> 00:34:58,440 If you are running a server hosting banking data, disabling 717 00:34:58,440 --> 00:35:00,800 that mitigation is a fatal error. 718 00:35:01,160 --> 00:35:04,320 But if you are just playing a lockdown video game, you don't 719 00:35:04,320 --> 00:35:07,640 need that level of security. By disabling it in the isolated 720 00:35:07,640 --> 00:35:10,920 generation, they stopped the CPU from constantly double checking 721 00:35:10,920 --> 00:35:14,600 itself, which instantly claws back 5 to 10 extra frames per 722 00:35:14,600 --> 00:35:16,840 second. It maximizes raw throughput. 723 00:35:17,040 --> 00:35:19,320 I chuckle every time I think about the sheer brilliance of 724 00:35:19,320 --> 00:35:21,480 this workflow. You want to play a match of 725 00:35:21,480 --> 00:35:23,320 Valorant. You restart your computer and 726 00:35:23,320 --> 00:35:25,440 select the gamer generation from the boot menu. 727 00:35:25,640 --> 00:35:29,280 Your computer boots up, stripped down, highly optimized and 728 00:35:29,280 --> 00:35:32,600 intentionally less secure. You play your match with maximum 729 00:35:32,600 --> 00:35:34,640 coin rights. When you are done, you restart 730 00:35:34,640 --> 00:35:37,400 again, select your daily driver generation and you are back in a 731 00:35:37,400 --> 00:35:40,200 fully locked down, perfectly secure environment, ready for 732 00:35:40,200 --> 00:35:42,520 work. And it is completely risk free. 733 00:35:42,680 --> 00:35:44,760 Right, you don't have to worry about your aggressive game 734 00:35:44,760 --> 00:35:47,160 tweaks ruining your work setup because mathematically they 735 00:35:47,160 --> 00:35:49,440 never touched it. It highlights the ultimate 736 00:35:49,440 --> 00:35:52,720 flexibility of a declarative system, but that level of 737 00:35:52,720 --> 00:35:56,000 granular control forces us to address the central paradox of 738 00:35:56,000 --> 00:35:59,360 this entire dossier. If Nix OS is truly this 739 00:35:59,360 --> 00:36:03,240 indestructible, if it's the perfect drift free haven for AI 740 00:36:03,240 --> 00:36:06,720 research, if it can safely cage proprietary software like 741 00:36:06,720 --> 00:36:10,360 Davinci Resolve, and if it allows you to dynamically hot 742 00:36:10,360 --> 00:36:14,120 swap kernels for 0 latency gaming, why isn't it the default 743 00:36:14,120 --> 00:36:16,640 operating system of the world? Yeah, why isn't it running on 744 00:36:16,640 --> 00:36:19,680 every laptop sold at Best Buy? That is the $1,000,000 question, 745 00:36:20,120 --> 00:36:22,760 and it brings us to the NICs Wall, the learning curve. 746 00:36:23,280 --> 00:36:26,040 Because as utopian as everything we've talked about sounds, 747 00:36:26,360 --> 00:36:29,440 climbing the mountain to reach that utopia is an incredibly 748 00:36:29,440 --> 00:36:32,560 steep, punishing endeavor. The primary barrier is not just 749 00:36:32,560 --> 00:36:35,080 learning a new command, it is the functional programming 750 00:36:35,080 --> 00:36:37,720 paradigm itself. NICs is not just a package 751 00:36:37,720 --> 00:36:41,480 manager like after brew. It is a lazy, functional, domain 752 00:36:41,480 --> 00:36:44,680 specific programming language. For an engineer coming from 753 00:36:44,680 --> 00:36:48,080 decades of Unix conventions, or a consumer used to visual menus, 754 00:36:48,280 --> 00:36:50,760 it requires a complete dismantling of their mental 755 00:36:50,760 --> 00:36:52,320 model. Let's contrast the mental 756 00:36:52,320 --> 00:36:55,200 models. If you are running a web server 757 00:36:55,200 --> 00:36:59,160 on a standard Linux distro and it crashes, your instinct is to 758 00:36:59,160 --> 00:37:02,560 troubleshoot state, right? You SSH into the server, open 759 00:37:02,560 --> 00:37:05,400 the live configuration file, tweak a line of code, restart 760 00:37:05,400 --> 00:37:08,400 the service and see if it works. If it does, you leave it. 761 00:37:08,800 --> 00:37:10,320 You have mutated the live system. 762 00:37:10,520 --> 00:37:12,960 On Nix OS you cannot do that. The entire system is read. 763 00:37:12,960 --> 00:37:16,120 Only the file you want to edit is locked behind a cryptographic 764 00:37:16,120 --> 00:37:17,840 hash. You don't troubleshoot state, 765 00:37:18,120 --> 00:37:20,720 you must refine intent. You have to go back to your 766 00:37:20,720 --> 00:37:23,400 master configuration file, figure out exactly what logic is 767 00:37:23,400 --> 00:37:26,320 failing, refine the definition, and then instruct the system to 768 00:37:26,320 --> 00:37:28,080 completely rebuild itself from scratch. 769 00:37:28,400 --> 00:37:30,960 While that guarantees long term stability because you now have a 770 00:37:30,960 --> 00:37:34,400 documented record of the fix, the short term friction for a 771 00:37:34,400 --> 00:37:37,560 new user is immense. And when you make a mistake 772 00:37:37,560 --> 00:37:40,400 refining that intent, the system is unforgiving. 773 00:37:41,280 --> 00:37:44,920 Historically, new users are hit with esoteric error messages, a 774 00:37:44,920 --> 00:37:48,280 deeply nested language syntax, and documentation that is often 775 00:37:48,280 --> 00:37:51,000 fragmented. You search for a solution online 776 00:37:51,000 --> 00:37:54,200 and half the answers use the old channel methods, while the other 777 00:37:54,200 --> 00:37:56,960 half use the new Flake standard and combining them breaks the 778 00:37:56,960 --> 00:37:58,920 evaluation. Which is why the biggest 779 00:37:58,960 --> 00:38:02,440 overarching trend of 2026 detailed in this dossier is the 780 00:38:02,440 --> 00:38:06,040 movement toward invisible NICs. The community has realized that 781 00:38:06,040 --> 00:38:08,240 while the underlying mathematical architecture is 782 00:38:08,240 --> 00:38:11,640 revolutionary, the surface level language is a severe deterrent 783 00:38:11,640 --> 00:38:14,160 to mass adoption. We see this addressed 784 00:38:14,160 --> 00:38:16,960 technically with the release of Determinate NICs 3 point. 785 00:38:16,960 --> 00:38:18,840 O this is a major structural shift. 786 00:38:19,320 --> 00:38:21,760 Determinate NICs is an alternative implementation of 787 00:38:21,760 --> 00:38:24,960 the core engine that evaluates the code, and the key innovation 788 00:38:24,960 --> 00:38:27,640 is the integration of Web assembly or resume. 789 00:38:28,440 --> 00:38:31,040 Web Assembly is a binary instruction format originally 790 00:38:31,040 --> 00:38:34,840 designed to let complex code run at near native speed inside web 791 00:38:34,840 --> 00:38:36,240 browsers. Oh, interesting. 792 00:38:36,240 --> 00:38:39,040 By bringing WESM into the NICs evaluator, the system can 793 00:38:39,040 --> 00:38:41,960 process the complex logic of package building significantly 794 00:38:41,960 --> 00:38:45,120 faster. Speed is a benefit, but the true 795 00:38:45,120 --> 00:38:48,240 value of determinate NICs 3 point O is how it handles 796 00:38:48,240 --> 00:38:51,160 failure. It actively intercepts those 797 00:38:51,160 --> 00:38:55,160 terrifying, deeply nested functional trace backs and uses 798 00:38:55,160 --> 00:38:58,400 its advanced parser to translate them into human readable, 799 00:38:58,480 --> 00:39:02,000 actionable instructions. It bridges the gap between the 800 00:39:02,000 --> 00:39:04,200 strict compiler and the human user. 801 00:39:04,360 --> 00:39:06,080 It's wrapping the bitter medicine of functional 802 00:39:06,080 --> 00:39:07,840 programming in a sweet candy shell. 803 00:39:08,400 --> 00:39:11,200 And beyond the core engine, we are seeing enterprise tools like 804 00:39:11,200 --> 00:39:14,760 Flocks acting as what the dossier calls a gateway drug. 805 00:39:15,440 --> 00:39:18,280 Flocks allows developers who are perfectly happy running Mac OS 806 00:39:18,280 --> 00:39:21,840 or Ubuntu to install the NICs package manager and use the 807 00:39:21,840 --> 00:39:25,680 slash NICs slash store solely for isolated project shells. 808 00:39:25,800 --> 00:39:28,720 It delivers 90% of the benefits, perfectly reproducible 809 00:39:28,720 --> 00:39:32,200 environments and 0 dependency conflicts across A-Team without 810 00:39:32,200 --> 00:39:34,480 demanding that the user format their hard drive and fully 811 00:39:34,480 --> 00:39:36,280 commit to the NICs O operating system. 812 00:39:36,280 --> 00:39:39,160 A developer can type flocks, activate inside a project folder 813 00:39:39,160 --> 00:39:41,480 and suddenly they are working inside a pure hashed 814 00:39:41,480 --> 00:39:43,520 environment. When they leave the folder, 815 00:39:43,680 --> 00:39:45,360 their Mac goes back to being a normal Mac. 816 00:39:45,440 --> 00:39:47,640 It is normalizing the functional architecture by making it a 817 00:39:47,640 --> 00:39:50,120 modular tool rather than an all or nothing lifestyle. 818 00:39:50,280 --> 00:39:53,440 It is the NICs Anywhere approach and the ambition of this 819 00:39:53,440 --> 00:39:56,720 architecture isn't stopping at desktop PC's or developer 820 00:39:56,720 --> 00:39:59,920 environments. The final section of the dossier 821 00:39:59,920 --> 00:40:03,240 look at the hardware frontiers and the implications are mind 822 00:40:03,240 --> 00:40:05,480 bending. They're pushing this declarative 823 00:40:05,480 --> 00:40:09,680 model to the very edges of computing, starting with mobile 824 00:40:09,680 --> 00:40:12,240 NICs OS. Mobile Linux has historically 825 00:40:12,240 --> 00:40:15,680 been a graveyard of fragmented, hacked together kernels. 826 00:40:16,080 --> 00:40:19,800 Every new smartphone requires A bespoke, highly customized build 827 00:40:19,800 --> 00:40:22,520 of Linux just to make the touchscreen and cellular modem 828 00:40:22,520 --> 00:40:25,800 work. But Nix OS is unifying this 829 00:40:25,800 --> 00:40:28,000 chaos. Because the OS separates 830 00:40:28,000 --> 00:40:30,960 hardware configuration from software definition, users are 831 00:40:30,960 --> 00:40:34,080 targeting devices like the Pine Phone Pro using the exact same 832 00:40:34,080 --> 00:40:37,080 configuration files they use for their laptops, simply swapping 833 00:40:37,080 --> 00:40:39,560 out the hardware module. And what truly blew my mind is 834 00:40:39,560 --> 00:40:41,280 the concept of the convergent generation. 835 00:40:41,280 --> 00:40:44,200 I want you to picture this. You have a smartphone running a 836 00:40:44,200 --> 00:40:46,960 mobile friendly user interface, something like FOSH or Plasma 837 00:40:46,960 --> 00:40:49,240 Mobile. You're walking around using it 838 00:40:49,240 --> 00:40:52,000 like a normal phone, but because the next door already 839 00:40:52,000 --> 00:40:55,080 dynamically contains all the software for a full desktop 840 00:40:55,080 --> 00:40:57,920 environment, mathematically LinkedIn the background, When 841 00:40:57,920 --> 00:41:01,600 you plug that phone into a USBC dock connected to a monitor and 842 00:41:01,600 --> 00:41:05,440 keyboard, the system detects the hardware change and instantly 843 00:41:05,440 --> 00:41:08,360 triggers a profile switch. It drops the mobile interface 844 00:41:08,360 --> 00:41:11,440 and instantly swaps it for a full blown desktop suite. 845 00:41:11,920 --> 00:41:14,960 And because both interfaces are pulling from the exact same 846 00:41:14,960 --> 00:41:19,040 underlying immutable slash NICs slash store, there is no 847 00:41:19,040 --> 00:41:22,160 virtualization overhead. There is no syncing data between 848 00:41:22,160 --> 00:41:25,880 a mobile app and a desktop app. Your phone literally is your 849 00:41:25,880 --> 00:41:29,160 primary desktop computer, evaluating its reality based on 850 00:41:29,160 --> 00:41:31,080 the physical context it finds itself in. 851 00:41:31,240 --> 00:41:33,640 It is the realization of a sci-fi concept we've been 852 00:41:33,640 --> 00:41:37,120 chasing for a decade and looking past consumer devices toward the 853 00:41:37,120 --> 00:41:39,480 data center. Nix OS is becoming the gold 854 00:41:39,480 --> 00:41:42,520 standard for deploying to the emerging RACV architecture. 855 00:41:42,640 --> 00:41:46,200 RACV is an open standard instruction set architecture 856 00:41:46,200 --> 00:41:48,720 that is heavily disrupting the server market, traditionally 857 00:41:48,720 --> 00:41:54,040 dominated by by 86 processors from Intel and AMD and ARM 858 00:41:54,040 --> 00:41:55,720 processors from Apple and Amazon. 859 00:41:56,200 --> 00:41:59,320 But developing an operating system for a brand new processor 860 00:41:59,320 --> 00:42:02,000 architecture is incredibly difficult because you have to 861 00:42:02,000 --> 00:42:04,480 cross compile. You have to write code on an 862 00:42:04,480 --> 00:42:08,360 Intel machine that is meant to execute on a RASCV machine. 863 00:42:08,360 --> 00:42:11,000 And usually that introduces subtle translation errors. 864 00:42:11,440 --> 00:42:14,880 A compiler on an Intel machine might accidentally inject an I86 865 00:42:14,880 --> 00:42:18,080 specific instruction into the RASCV binary, and when you 866 00:42:18,080 --> 00:42:19,840 deploy it to the new server it crashes. 867 00:42:20,360 --> 00:42:23,200 But NICs is fundamentally designed around pure inputs and 868 00:42:23,200 --> 00:42:26,720 explicit dependencies. It excels at cross compilation 869 00:42:26,720 --> 00:42:29,760 because it mathematically tracks every single tool used in the 870 00:42:29,760 --> 00:42:32,720 build process. A developer on a standard AMD 871 00:42:32,720 --> 00:42:35,800 workstation can instruct NICs to build an entire operating system 872 00:42:35,800 --> 00:42:39,680 image targeting ARISCV server. Because the build process is a 873 00:42:39,680 --> 00:42:42,720 pure function, the developer is mathematically guaranteed that 874 00:42:42,720 --> 00:42:45,480 the output image is bit for bit identical to an image that would 875 00:42:45,480 --> 00:42:48,520 have been built natively on the RISCV hardware itself. 876 00:42:48,920 --> 00:42:51,200 It completely eliminates the uncertainty of platform 877 00:42:51,200 --> 00:42:53,520 migration. So, synthesizing this entire 878 00:42:53,520 --> 00:42:58,280 massive journey, Nix OS in 2026 is an uncompromising powerhouse. 879 00:42:58,440 --> 00:43:01,680 It is demanding. It forces you to abandon decades 880 00:43:01,960 --> 00:43:04,200 of deeply ingrained computing habits. 881 00:43:04,640 --> 00:43:07,840 But what it offers in return is nothing short of revolutionary. 882 00:43:08,400 --> 00:43:10,800 The complete eradication of update anxiety. 883 00:43:11,000 --> 00:43:13,400 It cures the fear of the broken upgrade. 884 00:43:13,600 --> 00:43:16,880 It cleanly cages the messy, chaotic beasts of AI 885 00:43:16,880 --> 00:43:19,880 dependencies and proprietary creative monoliths. 886 00:43:20,320 --> 00:43:23,640 It takes the corporate buzzword of infrastructure as code and 887 00:43:23,640 --> 00:43:26,840 turns it into a literal mathematical law of physics 888 00:43:27,080 --> 00:43:30,080 executing on your hard drive. The initial tax of learning the 889 00:43:30,080 --> 00:43:33,280 NICs language and functional paradigm is undeniably high. 890 00:43:33,640 --> 00:43:35,840 You'll send hours fighting the configuration just to get a 891 00:43:35,840 --> 00:43:37,280 simple tool working the first time. 892 00:43:37,360 --> 00:43:39,720 But the long term reward for paying that tax is a system that 893 00:43:39,720 --> 00:43:41,760 simply never rots. It is a system that you can 894 00:43:41,760 --> 00:43:44,280 confidently rebuild on a brand new machine anywhere in the 895 00:43:44,280 --> 00:43:47,280 world with a single command, and know what absolute mathematical 896 00:43:47,280 --> 00:43:49,320 certainty that it will be exactly as you left it. 897 00:43:49,320 --> 00:43:51,520 Which? Leaves us with a final, slightly 898 00:43:51,520 --> 00:43:55,480 philosophical thought to ponder. If an operating system removes 899 00:43:55,480 --> 00:43:58,600 all permanent state, if the entire reality of your computer 900 00:43:58,600 --> 00:44:01,320 is ultimately just a pure mathematical function waiting to 901 00:44:01,320 --> 00:44:04,640 be evaluated from a simple text file, does the computer as a 902 00:44:04,640 --> 00:44:07,240 unique physical machine even exist anymore? 903 00:44:07,640 --> 00:44:10,440 Or have we just turned our hardware into a temporary 904 00:44:10,440 --> 00:44:13,720 disposable vessel for our code? Think about what that means the 905 00:44:13,720 --> 00:44:15,880 next time you hesitate before clicking update. 906 00:44:16,000 --> 00:44:17,640 Thanks for joining us on this deep dive. 907 00:44:17,640 --> 00:44:18,360 See you next time.