1 00:00:00,880 --> 00:00:02,899 Join us as we gather around the hedge, 2 00:00:02,960 --> 00:00:05,940 where we dig into technology, business, and culture 3 00:00:06,000 --> 00:00:08,339 with the finest minds in computer networking. 4 00:00:20,814 --> 00:00:21,954 Well, hello, Yvonne. 5 00:00:22,335 --> 00:00:23,934 It's good to see that you're back with 6 00:00:23,934 --> 00:00:27,074 us from your visiting from your other famous 7 00:00:27,535 --> 00:00:28,035 and 8 00:00:28,440 --> 00:00:29,900 highly successful podcast. 9 00:00:30,280 --> 00:00:32,520 Ah, well, that's a bit of a I 10 00:00:32,520 --> 00:00:34,120 don't know. I don't know what I'll say 11 00:00:34,120 --> 00:00:35,079 to that. It's, 12 00:00:36,840 --> 00:00:38,859 I'm glad to be here. How's that? 13 00:00:39,640 --> 00:00:42,679 Yeah. I'm I'm famous. It sure feels like 14 00:00:42,679 --> 00:00:44,905 it. Yeah. No. I am happy to be 15 00:00:44,905 --> 00:00:46,664 here with with you guys. For us, it 16 00:00:46,664 --> 00:00:49,384 is a Friday afternoon, and it couldn't come 17 00:00:49,384 --> 00:00:52,664 any sooner. Hello, Tom and Audrey. So speaking 18 00:00:52,664 --> 00:00:54,905 of famous, I have to say this morning, 19 00:00:55,304 --> 00:00:57,864 at the start of another recording, another guest 20 00:00:57,864 --> 00:00:59,939 was like, is that the plant? 21 00:01:03,520 --> 00:01:06,079 So That I actually missed because of work. 22 00:01:06,079 --> 00:01:08,319 But anyway, sorry about that. But yeah. If 23 00:01:08,319 --> 00:01:08,979 you notice, 24 00:01:09,359 --> 00:01:10,180 my plant 25 00:01:10,719 --> 00:01:11,780 is no more. 26 00:01:12,644 --> 00:01:13,144 No. 27 00:01:13,525 --> 00:01:15,864 I know. It's terrible. So 28 00:01:16,244 --> 00:01:18,805 we had a really big freeze back a 29 00:01:18,805 --> 00:01:19,864 couple months ago, 30 00:01:20,244 --> 00:01:20,744 and 31 00:01:21,125 --> 00:01:23,704 my heater in my shed stopped working. 32 00:01:24,325 --> 00:01:25,625 So I came out, 33 00:01:26,349 --> 00:01:28,129 and then it had been all weekend, 34 00:01:28,590 --> 00:01:31,009 and my plant was frozen solid. 35 00:01:31,549 --> 00:01:34,450 It was, like, 17 degrees out here, so 36 00:01:34,989 --> 00:01:37,649 it did not survive. So it's springtime. 37 00:01:39,255 --> 00:01:41,015 I need to go find another plant, but 38 00:01:41,015 --> 00:01:42,234 it did not make it. 39 00:01:42,935 --> 00:01:44,875 So I'm sorry, everybody. The Dieffenbachia 40 00:01:45,174 --> 00:01:46,935 is no more. I was just gonna say, 41 00:01:46,935 --> 00:01:49,015 don't worry, Audrey. You're safe in here. You 42 00:01:49,015 --> 00:01:51,195 don't need to worry about that. Yeah. Cover 43 00:01:52,055 --> 00:01:53,435 its ears her ears. 44 00:01:56,760 --> 00:01:59,079 So believe it or not, this is when 45 00:01:59,079 --> 00:01:59,819 this publishes, 46 00:02:00,120 --> 00:02:02,520 in theory, the three hundredth episode of the 47 00:02:02,520 --> 00:02:03,020 hedge. 48 00:02:05,000 --> 00:02:05,980 Isn't that crazy? 49 00:02:06,704 --> 00:02:08,484 I feel I feel old. 50 00:02:08,944 --> 00:02:11,185 Okay. I am old, so never mind. Hey. 51 00:02:11,185 --> 00:02:12,944 I hit a bunch of big milestones this 52 00:02:12,944 --> 00:02:15,584 year myself. So, yeah, we don't need to 53 00:02:15,584 --> 00:02:16,644 talk about that anymore. 54 00:02:21,584 --> 00:02:23,439 Yeah. I was at the gym yesterday, and 55 00:02:23,439 --> 00:02:25,439 I was doing something. And somebody said I 56 00:02:25,439 --> 00:02:26,719 was on the bench press, and I was 57 00:02:26,719 --> 00:02:28,560 doing something. I was doing my reps, and 58 00:02:28,560 --> 00:02:31,439 somebody walked over and said, alright. Man, you're 59 00:02:31,439 --> 00:02:33,680 kinda old, like, for me in that kinda 60 00:02:33,680 --> 00:02:35,615 way. And I was like, I'm just gonna 61 00:02:35,615 --> 00:02:37,694 smack you upside the head. Maybe you meant 62 00:02:37,694 --> 00:02:39,375 that as a compliment, but we need to 63 00:02:39,375 --> 00:02:41,314 talk about how this works. Yeah. 64 00:02:43,615 --> 00:02:45,615 You get around good for an old guy, 65 00:02:45,615 --> 00:02:46,115 Russ. 66 00:02:51,759 --> 00:02:52,259 Wow. 67 00:02:52,639 --> 00:02:53,139 Okay. 68 00:02:53,519 --> 00:02:56,000 So we're gonna we're gonna start today talking 69 00:02:56,000 --> 00:02:58,079 about prompt injection in Bell Labs. And I 70 00:02:58,079 --> 00:02:59,139 think this is interesting. 71 00:03:00,324 --> 00:03:02,164 Well and then promptly destroyed it for a 72 00:03:02,164 --> 00:03:04,164 profit. This is an article on LinkedIn, by 73 00:03:04,164 --> 00:03:05,764 the way, if you wanna find it. I'll 74 00:03:05,764 --> 00:03:07,284 try to put a link to it in 75 00:03:07,284 --> 00:03:08,584 the show notes if I remember. 76 00:03:08,884 --> 00:03:09,625 But anyway, 77 00:03:10,324 --> 00:03:12,884 it's interesting to me well, let's talk about 78 00:03:12,884 --> 00:03:14,344 the the what they've done 79 00:03:14,669 --> 00:03:16,370 and what they did, and, 80 00:03:17,150 --> 00:03:18,669 then we can talk about, like, why I 81 00:03:18,669 --> 00:03:20,509 think this is this is or why we 82 00:03:20,509 --> 00:03:21,889 think this is really interesting 83 00:03:22,590 --> 00:03:25,069 in the in the realm of technology or 84 00:03:25,069 --> 00:03:25,569 networking. 85 00:03:26,189 --> 00:03:28,030 Lost my tab, guys. So 86 00:03:29,235 --> 00:03:31,634 I know. It's terrible. I'm having to use 87 00:03:31,634 --> 00:03:33,495 search to find my tab. Okay. 88 00:03:33,875 --> 00:03:36,694 I've got it now. Yeah. It's terrible. Gosh. 89 00:03:36,834 --> 00:03:38,354 Yeah. The whole the whole point of the 90 00:03:38,354 --> 00:03:39,735 article is that, 91 00:03:40,275 --> 00:03:42,860 it says Bell Labs self prompt injection 92 00:03:43,560 --> 00:03:45,580 back in 1976. 93 00:03:46,040 --> 00:03:47,180 And the whole 94 00:03:47,480 --> 00:03:48,860 theory of the article 95 00:03:49,319 --> 00:03:49,819 is 96 00:03:50,280 --> 00:03:51,900 they invented technologies 97 00:03:52,280 --> 00:03:54,919 back then. SS seven is a specific one 98 00:03:54,919 --> 00:03:57,875 that we're talking about, where you had one 99 00:03:57,935 --> 00:04:00,354 control channel and one data channel. 100 00:04:00,655 --> 00:04:02,354 So all of the instructions 101 00:04:03,455 --> 00:04:06,175 for the protocol and the service happened on 102 00:04:06,175 --> 00:04:08,495 one channel. It ran its own protocol that 103 00:04:08,495 --> 00:04:10,034 was completely architecturally 104 00:04:11,060 --> 00:04:13,080 independent from the data channel. 105 00:04:13,460 --> 00:04:15,300 And the and the the two did not 106 00:04:15,300 --> 00:04:17,699 twain. Right? So you had your control channel 107 00:04:17,699 --> 00:04:20,120 and your data channel, and those were separate, 108 00:04:20,740 --> 00:04:21,240 which 109 00:04:22,100 --> 00:04:23,319 pretty much prevents 110 00:04:24,020 --> 00:04:25,160 the the possibility 111 00:04:26,875 --> 00:04:27,454 of injecting, 112 00:04:29,675 --> 00:04:30,495 bad instructions 113 00:04:31,754 --> 00:04:33,754 in the data plane, right, or in in 114 00:04:33,754 --> 00:04:36,794 the in in the information exchange because there's 115 00:04:36,794 --> 00:04:39,214 a separate channel specifically for 116 00:04:39,699 --> 00:04:40,199 control. 117 00:04:41,540 --> 00:04:42,439 And, of course, 118 00:04:44,420 --> 00:04:47,160 the business said what Well Go ahead. 119 00:04:47,540 --> 00:04:50,180 The business said no. Yeah. So but but 120 00:04:50,180 --> 00:04:51,699 to to even go to go back in 121 00:04:51,699 --> 00:04:53,904 history and understand why this happened. When SS 122 00:04:53,904 --> 00:04:55,524 seven was developed originally, 123 00:04:56,225 --> 00:04:56,725 the 124 00:04:57,345 --> 00:04:57,845 original 125 00:04:58,865 --> 00:04:59,365 digital 126 00:05:00,225 --> 00:05:02,704 signaling protocols, they weren't they weren't really digital 127 00:05:02,704 --> 00:05:04,805 in those days, but let's let's say digital, 128 00:05:05,745 --> 00:05:07,399 just to be kind. 129 00:05:08,259 --> 00:05:10,660 You could call you pick up a phone 130 00:05:10,660 --> 00:05:13,480 and and and dial certain tones 131 00:05:13,860 --> 00:05:15,560 or inject certain tones 132 00:05:15,939 --> 00:05:17,939 and make a payphone do a long distance 133 00:05:17,939 --> 00:05:19,319 phone call, for instance, 134 00:05:19,699 --> 00:05:22,199 without putting a quarter and without being charged. 135 00:05:23,355 --> 00:05:25,915 And it was because the signaling was on 136 00:05:25,915 --> 00:05:26,495 the same 137 00:05:27,595 --> 00:05:28,095 channel 138 00:05:28,475 --> 00:05:30,154 as the voice data. You pick up the 139 00:05:30,154 --> 00:05:31,835 voice data. You pick up the voice voice 140 00:05:31,835 --> 00:05:33,995 call, you pick up the handset, for those 141 00:05:33,995 --> 00:05:35,535 of you who remember handsets, 142 00:05:36,040 --> 00:05:36,860 speaking of 143 00:05:38,120 --> 00:05:38,620 old, 144 00:05:39,879 --> 00:05:42,459 you had a direct line to the switch, 145 00:05:43,319 --> 00:05:45,240 to the telco switch, and you were talking 146 00:05:45,240 --> 00:05:46,139 to the controller. 147 00:05:47,240 --> 00:05:49,339 And so yes. So SS seven was developed 148 00:05:49,800 --> 00:05:50,540 in part 149 00:05:51,105 --> 00:05:52,884 against these tone kitties 150 00:05:53,665 --> 00:05:55,264 or, you know, what we would call script 151 00:05:55,264 --> 00:05:57,345 kitties today. Did you know that if you 152 00:05:57,345 --> 00:05:58,805 pick up the phone and you go 153 00:05:59,264 --> 00:06:01,605 at a certain frequency, it makes it bypass 154 00:06:02,064 --> 00:06:04,305 the coin drop? You know, that was the 155 00:06:04,305 --> 00:06:06,004 kind of thing that was passed around. 156 00:06:06,479 --> 00:06:08,160 And so, yeah, they did they did this 157 00:06:08,160 --> 00:06:08,899 by splitting 158 00:06:09,839 --> 00:06:11,220 the control plane 159 00:06:11,680 --> 00:06:13,939 from the data plane, and that still exists 160 00:06:14,639 --> 00:06:16,259 in telco systems today. 161 00:06:17,039 --> 00:06:18,899 They don't pay attention to it. 162 00:06:20,014 --> 00:06:22,175 And if and if you're not familiar with 163 00:06:22,175 --> 00:06:24,574 the story of Captain Crunch and the whistle 164 00:06:24,574 --> 00:06:26,915 that came in the Captain Crunch box and 165 00:06:26,975 --> 00:06:27,634 the lore, 166 00:06:28,495 --> 00:06:30,814 behind that, it's it's worth looking up and 167 00:06:30,814 --> 00:06:32,654 reading about because that's where a lot of 168 00:06:32,654 --> 00:06:33,154 the, 169 00:06:33,790 --> 00:06:35,089 you know, hacker 170 00:06:35,470 --> 00:06:35,970 mentality 171 00:06:36,430 --> 00:06:38,189 came from. It was one of the really 172 00:06:38,189 --> 00:06:39,410 first well known, 173 00:06:40,189 --> 00:06:42,990 hacks that got well publicized, and, and they 174 00:06:42,990 --> 00:06:45,629 exploited this exact phenomenon that Russ has been 175 00:06:45,629 --> 00:06:46,129 describing. 176 00:06:47,985 --> 00:06:51,125 Yeah. Yeah. The the captain crunch whistle, and 177 00:06:51,264 --> 00:06:53,665 people would take their modems, and there were 178 00:06:53,665 --> 00:06:55,524 little programs you could get on 179 00:06:56,064 --> 00:06:58,404 BBC, bulletin board systems, BBSs, 180 00:06:59,105 --> 00:07:01,685 that would emit the correct tones. Right. 181 00:07:02,530 --> 00:07:04,250 So that you could dial into the b 182 00:07:04,370 --> 00:07:06,370 BBS system for free. Yeah. The thing I 183 00:07:06,370 --> 00:07:08,050 I think is interesting about this bit of 184 00:07:08,050 --> 00:07:09,970 history is we often think about, 185 00:07:10,610 --> 00:07:12,689 rule 11 that there's not really new ideas. 186 00:07:12,689 --> 00:07:14,530 They're just kind of recycled ideas and new 187 00:07:14,530 --> 00:07:15,030 packages. 188 00:07:15,464 --> 00:07:17,464 But I it extends to lots of different 189 00:07:17,464 --> 00:07:19,324 things, not just protocol design. 190 00:07:19,865 --> 00:07:21,865 It it extends in this case to how 191 00:07:21,865 --> 00:07:23,865 to how to attack the vulnerabilities of a 192 00:07:23,865 --> 00:07:24,365 system. 193 00:07:24,665 --> 00:07:26,904 Like, it is it is you know, they're 194 00:07:26,985 --> 00:07:29,490 the recipes for attacking a system also don't 195 00:07:29,490 --> 00:07:31,329 change that much. There are new classes of 196 00:07:31,329 --> 00:07:33,889 attacks, but, like, the the first principles of 197 00:07:33,889 --> 00:07:34,789 attacking systems, 198 00:07:35,970 --> 00:07:38,289 stayed basically the same. I think we're learning 199 00:07:38,289 --> 00:07:40,129 new things. I'm not saying that there's only 200 00:07:40,129 --> 00:07:41,669 old knowledge, but, certainly, 201 00:07:42,064 --> 00:07:43,824 if you were trying to attack something, you'd 202 00:07:43,824 --> 00:07:45,664 start from the playbook of what has worked 203 00:07:45,664 --> 00:07:48,144 in the past. And you you a part 204 00:07:48,144 --> 00:07:50,144 of the reason you would succeed is because 205 00:07:50,144 --> 00:07:51,824 we don't learn the lessons of the past. 206 00:07:51,824 --> 00:07:53,284 We're not very good at that. 207 00:07:56,689 --> 00:07:59,009 Yeah. Exactly. Or we don't know how to 208 00:07:59,009 --> 00:08:00,629 apply them. Or we think 209 00:08:01,650 --> 00:08:03,189 but this is so different. 210 00:08:03,650 --> 00:08:06,069 The problem we're solving is so different. 211 00:08:06,850 --> 00:08:08,914 And, you know, and we we've come up 212 00:08:08,914 --> 00:08:10,535 with a really unique solution. 213 00:08:11,474 --> 00:08:12,535 No. Really, honestly, 214 00:08:13,394 --> 00:08:16,115 you haven't. I promise you haven't come up 215 00:08:16,115 --> 00:08:17,495 with a really unique solution. 216 00:08:18,274 --> 00:08:21,040 Right? It's like it's like reading Shakespeare and 217 00:08:21,040 --> 00:08:23,060 you hear people modern people go, well, 218 00:08:23,839 --> 00:08:25,759 for the last thousand years, people have read 219 00:08:25,759 --> 00:08:27,520 Shakespeare this way, but I found a new 220 00:08:27,520 --> 00:08:29,759 way to read Shakespeare that actually explains what 221 00:08:29,759 --> 00:08:30,420 it means. 222 00:08:31,759 --> 00:08:32,420 I promise 223 00:08:32,720 --> 00:08:33,460 you haven't. 224 00:08:37,415 --> 00:08:40,295 So yeah. So continue, Yvonne. Sorry. You know, 225 00:08:40,295 --> 00:08:42,054 they they broke it. So it took years. 226 00:08:42,215 --> 00:08:44,154 Gosh. Where did I leave off? So, 227 00:08:45,894 --> 00:08:48,949 yeah, so this we we we, ultimately, what 228 00:08:48,949 --> 00:08:49,449 happened 229 00:08:50,069 --> 00:08:53,429 is the business decided, like, hey. We've got 230 00:08:53,429 --> 00:08:55,589 extra bandwidth out here. We've got this whole 231 00:08:55,589 --> 00:08:56,089 channel 232 00:08:56,470 --> 00:08:58,809 that we're using just for signaling. 233 00:08:59,750 --> 00:09:01,144 We can shove 234 00:09:01,605 --> 00:09:05,045 more revenue generating data down that channel without 235 00:09:05,045 --> 00:09:06,425 spending any more money. 236 00:09:06,805 --> 00:09:09,225 And so that's what the business did. 237 00:09:10,485 --> 00:09:12,585 So and what that did 238 00:09:13,059 --> 00:09:15,960 is it broke this separation in architecture 239 00:09:16,580 --> 00:09:18,899 where you had control on one channel and 240 00:09:18,899 --> 00:09:21,320 you had your data on another. And 241 00:09:21,860 --> 00:09:24,100 here we are again, back to the same 242 00:09:24,100 --> 00:09:26,920 place where you can inject 243 00:09:28,115 --> 00:09:29,174 control language 244 00:09:29,875 --> 00:09:33,174 into a control channel or into a shared 245 00:09:33,235 --> 00:09:34,695 data and control channel 246 00:09:35,315 --> 00:09:35,554 and, 247 00:09:36,434 --> 00:09:39,014 and and break things or do bad things. 248 00:09:39,955 --> 00:09:40,455 Yeah. 249 00:09:41,009 --> 00:09:43,810 Yeah. I mean, originally, when they did this, 250 00:09:43,810 --> 00:09:44,550 I bet 251 00:09:45,170 --> 00:09:47,330 there was a conversation. He doesn't talk about 252 00:09:47,330 --> 00:09:48,230 it in this article. 253 00:09:49,170 --> 00:09:49,670 But 254 00:09:50,850 --> 00:09:53,009 I bet there was a conversation that went 255 00:09:53,009 --> 00:09:55,090 around a whiteboard or around a conference room 256 00:09:55,090 --> 00:09:55,590 table 257 00:09:55,975 --> 00:09:57,355 that said something like, 258 00:09:57,735 --> 00:10:00,134 but doesn't this allow people to go back 259 00:10:00,134 --> 00:10:00,875 to doing 260 00:10:01,414 --> 00:10:02,475 the the, 261 00:10:03,414 --> 00:10:05,434 the whistle in the phone thing? 262 00:10:06,134 --> 00:10:09,100 And somebody probably answered and said, well, that's 263 00:10:09,100 --> 00:10:11,019 not a part of the attack surface. This 264 00:10:11,019 --> 00:10:12,960 SMS service is text only. 265 00:10:13,340 --> 00:10:15,259 And the only way you'll get into this 266 00:10:15,259 --> 00:10:17,420 SMS channel or be able to send SMS 267 00:10:17,420 --> 00:10:17,920 data 268 00:10:18,379 --> 00:10:20,480 is by is by being on the phone 269 00:10:20,940 --> 00:10:23,084 and typing this stuff in. Like, there's no 270 00:10:23,084 --> 00:10:23,584 interface 271 00:10:24,044 --> 00:10:24,945 into that. 272 00:10:25,485 --> 00:10:26,784 And, you know, 273 00:10:27,725 --> 00:10:29,264 it's it's crazy 274 00:10:29,804 --> 00:10:30,865 because anytime 275 00:10:31,245 --> 00:10:31,745 anytime 276 00:10:32,764 --> 00:10:35,024 that I have access to the physical endpoint 277 00:10:35,245 --> 00:10:36,065 of the circuit, 278 00:10:36,379 --> 00:10:38,539 I can inject anything I want to into 279 00:10:38,539 --> 00:10:39,200 the circuit. 280 00:10:39,820 --> 00:10:41,580 We we tend we tend to have this, 281 00:10:41,580 --> 00:10:42,080 like, 282 00:10:42,459 --> 00:10:44,879 oh, no. No. Nobody can get to that. 283 00:10:45,179 --> 00:10:47,120 Well and and in my youth, 284 00:10:47,659 --> 00:10:50,139 I've made this mistake as well. I remember 285 00:10:50,139 --> 00:10:50,639 back, 286 00:10:50,985 --> 00:10:54,264 it used to be that email systems were 287 00:10:54,264 --> 00:10:55,164 purely text. 288 00:10:55,705 --> 00:10:58,264 Right? Like, you you you read your email 289 00:10:58,264 --> 00:10:59,565 in a text reader. 290 00:11:00,024 --> 00:11:03,544 There was no ability to execute anything from 291 00:11:03,544 --> 00:11:04,845 an email, and I remember 292 00:11:05,470 --> 00:11:06,529 boldly declaring 293 00:11:08,110 --> 00:11:11,389 that viruses could not be transmitted via email 294 00:11:11,389 --> 00:11:13,570 because it was purely a text system. 295 00:11:14,190 --> 00:11:16,850 Well, what happens is we get richer, 296 00:11:18,029 --> 00:11:18,929 email readers, 297 00:11:19,535 --> 00:11:22,575 richer text readers that are more integrated with 298 00:11:22,575 --> 00:11:24,835 the client. And then all of a sudden, 299 00:11:25,295 --> 00:11:26,675 you can start embedding 300 00:11:27,774 --> 00:11:30,575 content that be can be clicked or launched 301 00:11:30,575 --> 00:11:33,215 or executed from inside of an email. And 302 00:11:33,215 --> 00:11:35,850 so it was really just a matter of 303 00:11:35,850 --> 00:11:38,169 time. And, of course, I was the, 304 00:11:39,129 --> 00:11:39,629 smart, 305 00:11:40,169 --> 00:11:42,330 you know, computer science college student who was 306 00:11:42,330 --> 00:11:44,250 like, oh, no. Email doesn't work that way. 307 00:11:44,250 --> 00:11:46,204 You can never exploit it for by and, 308 00:11:46,204 --> 00:11:48,757 you know, sending a virus in it because 309 00:11:48,757 --> 00:11:51,630 there's no execute there's nowhere to do the 310 00:11:51,630 --> 00:11:54,183 execution of whatever the thing is. Well, sure 311 00:11:54,183 --> 00:11:56,737 enough, just one more small evolution ahead, and 312 00:11:56,737 --> 00:11:59,290 all of a sudden, you have a whole 313 00:11:59,290 --> 00:12:00,829 new attack service. 314 00:12:01,209 --> 00:12:03,289 And then that's not even accounting for things 315 00:12:03,289 --> 00:12:03,789 like, 316 00:12:04,169 --> 00:12:07,129 you know, phishing attacks or social engineering attacks 317 00:12:07,129 --> 00:12:08,809 or any of those things that can happen 318 00:12:08,809 --> 00:12:11,549 over email. But it's it's really tempting 319 00:12:12,329 --> 00:12:14,409 to fall for this kind of thinking. And 320 00:12:14,409 --> 00:12:14,990 I think 321 00:12:15,654 --> 00:12:17,495 most of us have probably done it in 322 00:12:17,495 --> 00:12:18,554 one way or another 323 00:12:19,014 --> 00:12:19,514 as, 324 00:12:20,134 --> 00:12:22,634 what I'll say, early in career technologists. 325 00:12:26,054 --> 00:12:27,870 Oh, we do we do it as late, 326 00:12:27,870 --> 00:12:29,649 Yeah. We do it as late in career. 327 00:12:29,870 --> 00:12:32,589 Hubris as we get more advanced may actually 328 00:12:32,589 --> 00:12:36,350 make us more susceptible, not less, because we 329 00:12:36,350 --> 00:12:38,370 think we know how it all works. Yeah. 330 00:12:38,589 --> 00:12:39,089 Right. 331 00:12:39,710 --> 00:12:41,389 Yeah. Oh, yeah. Yeah. Yeah. Yeah. I mean, 332 00:12:41,389 --> 00:12:44,074 I remember I remember people deploying 333 00:12:44,615 --> 00:12:46,634 individual boxes out to 334 00:12:47,014 --> 00:12:48,235 rental car agencies 335 00:12:48,694 --> 00:12:50,375 and saying, well, it has a password on 336 00:12:50,375 --> 00:12:52,774 it, so who cares? You can't break into 337 00:12:52,774 --> 00:12:54,314 it with a password on it. 338 00:12:54,774 --> 00:12:56,694 Like, no. If I have physical access to 339 00:12:56,694 --> 00:12:58,299 the box, I'm in. 340 00:12:58,679 --> 00:13:00,759 Right? There are very few things you can 341 00:13:00,759 --> 00:13:02,839 do. I mean, there are things you can 342 00:13:02,839 --> 00:13:03,339 do. 343 00:13:04,199 --> 00:13:05,799 You know, when I was at a company 344 00:13:05,799 --> 00:13:07,799 once, we designed a system where you couldn't 345 00:13:07,799 --> 00:13:10,220 break in, but it was because there were 346 00:13:10,835 --> 00:13:11,555 there were, 347 00:13:12,035 --> 00:13:13,735 motion sensors on the box. 348 00:13:14,514 --> 00:13:16,595 And if you move the box, it erases 349 00:13:16,595 --> 00:13:17,254 the data. 350 00:13:17,875 --> 00:13:19,555 We'll break in all you want to now. 351 00:13:19,555 --> 00:13:21,315 We don't really care because there's nothing on 352 00:13:21,315 --> 00:13:23,875 there anymore. It has no circuit access anymore 353 00:13:23,875 --> 00:13:25,495 because the data's gone. 354 00:13:25,929 --> 00:13:27,850 And so, I mean, there are things you 355 00:13:27,850 --> 00:13:30,350 can do, but they're radical. Like, they're crazy, 356 00:13:31,769 --> 00:13:33,389 things that that are difficult. 357 00:13:33,850 --> 00:13:35,610 But then this brings up the whole concept 358 00:13:35,610 --> 00:13:36,590 of attack surfaces 359 00:13:37,210 --> 00:13:39,309 and thinking through attack surfaces. 360 00:13:39,985 --> 00:13:42,304 And then the idea of how you manage 361 00:13:42,304 --> 00:13:44,404 most attack surfaces is via segmentation. 362 00:13:45,665 --> 00:13:47,904 That's that's what we do. It's interesting because 363 00:13:47,904 --> 00:13:49,504 you as you read the article, he talks 364 00:13:49,504 --> 00:13:51,605 about this is actually is an architectural problem. 365 00:13:52,679 --> 00:13:55,980 And specifically in in in relation to 366 00:13:56,679 --> 00:13:59,240 prompt injection attempts. One example of this is 367 00:13:59,240 --> 00:14:00,700 the GitHub data heist thing, 368 00:14:01,720 --> 00:14:03,720 where you can you can you can put 369 00:14:03,720 --> 00:14:05,975 stuff in a prompt that will that will 370 00:14:05,975 --> 00:14:08,394 sort through GitHub that'll that'll find things, 371 00:14:08,855 --> 00:14:09,355 and 372 00:14:09,894 --> 00:14:12,375 you can you can inject prompts. It's just 373 00:14:12,375 --> 00:14:14,774 like it smells just like a SQL injection 374 00:14:14,774 --> 00:14:16,054 attack to me. You know, in back in 375 00:14:16,054 --> 00:14:17,720 the old days, you'd put in a a 376 00:14:17,720 --> 00:14:20,120 drop whatever semi colon in the input field, 377 00:14:20,120 --> 00:14:21,480 and they hadn't scrubbed it. And so you 378 00:14:21,480 --> 00:14:23,080 could do stuff to the database. Like, it 379 00:14:23,080 --> 00:14:24,920 smells exactly like that to me. But the 380 00:14:24,920 --> 00:14:26,759 thing that I thought was really interesting was 381 00:14:26,840 --> 00:14:28,759 so there's a there's a whole series that, 382 00:14:29,080 --> 00:14:31,480 Docker has put up called the, MCP horror 383 00:14:31,480 --> 00:14:34,285 stories. It's pretty interesting. And they describe their 384 00:14:34,285 --> 00:14:37,085 solution to this, and their solution to this 385 00:14:37,085 --> 00:14:39,565 is put a gateway in. Put a gateway 386 00:14:39,565 --> 00:14:41,325 in. It's just like a WAF is what 387 00:14:41,325 --> 00:14:43,165 it reads like to me. We'll just inspect 388 00:14:43,165 --> 00:14:45,165 all the control stuff that's going through this 389 00:14:45,165 --> 00:14:48,039 consolidated control data channel, and every single control 390 00:14:48,039 --> 00:14:49,559 message will just make sure that it's that 391 00:14:49,559 --> 00:14:51,720 it's okay, which okay. And that's a pretty 392 00:14:51,720 --> 00:14:54,679 normal reaction, but this is not a problem. 393 00:14:54,679 --> 00:14:56,679 Like, you can you can somewhat solve it 394 00:14:56,679 --> 00:14:58,279 with this idea of a gateway with the 395 00:14:58,279 --> 00:14:59,975 middle box. But what the author of the 396 00:14:59,975 --> 00:15:01,975 LinkedIn article is pointing out is there's actually 397 00:15:01,975 --> 00:15:03,355 an architectural solution. 398 00:15:03,815 --> 00:15:05,735 And and what you were saying, Russ, like, 399 00:15:05,735 --> 00:15:07,095 yeah. We'll we'll we'll build it so you 400 00:15:07,095 --> 00:15:09,014 can't break into it physically. That's a that's 401 00:15:09,014 --> 00:15:11,894 a architectural solution. That's a, you know, how 402 00:15:11,894 --> 00:15:13,274 the thing works. Yeah. 403 00:15:13,700 --> 00:15:14,200 Quick. 404 00:15:14,740 --> 00:15:16,820 Yeah. You can Well, we've also heard about 405 00:15:17,059 --> 00:15:18,740 and I don't know that that we've reached 406 00:15:18,740 --> 00:15:20,500 the attack level of this kind of thing 407 00:15:20,500 --> 00:15:22,120 yet, but we've certainly heard 408 00:15:22,740 --> 00:15:24,919 about AI systems misbehaving 409 00:15:26,014 --> 00:15:28,754 because the context window gets so 410 00:15:29,134 --> 00:15:31,794 large or or gets so congested 411 00:15:32,335 --> 00:15:35,394 that the initial instructions that provided the constraints 412 00:15:35,455 --> 00:15:36,355 for the system 413 00:15:36,815 --> 00:15:39,794 fall out of the context window. Right? So, 414 00:15:40,059 --> 00:15:42,940 hey. Don't don't or always ask before you 415 00:15:42,940 --> 00:15:43,759 make changes. 416 00:15:44,139 --> 00:15:46,480 Well, all of a sudden, you've you've you've 417 00:15:46,779 --> 00:15:48,620 put so much data into the system that 418 00:15:48,620 --> 00:15:51,019 you lose that instruction out of the context 419 00:15:51,019 --> 00:15:53,120 window. And then all of a sudden, 420 00:15:53,534 --> 00:15:54,034 your, 421 00:15:54,815 --> 00:15:56,894 AI system starts behaving in ways that you 422 00:15:56,894 --> 00:15:59,934 don't expect. Well, that sounds a lot to 423 00:15:59,934 --> 00:16:00,434 me 424 00:16:00,975 --> 00:16:02,355 like a buffer overrun. 425 00:16:03,294 --> 00:16:06,355 Right? Like Yeah. It's it's it's the same 426 00:16:06,414 --> 00:16:07,475 kind of thing, 427 00:16:08,480 --> 00:16:09,299 where you're 428 00:16:09,839 --> 00:16:10,339 expanding 429 00:16:11,200 --> 00:16:14,879 the understood boundaries of the system and causing 430 00:16:14,879 --> 00:16:16,339 it to behave erratically. 431 00:16:17,039 --> 00:16:19,700 That's what a buffer overflow attack is. 432 00:16:20,559 --> 00:16:22,100 Yeah. It's the same concept. 433 00:16:23,865 --> 00:16:25,544 And so yeah. So and and think about 434 00:16:25,544 --> 00:16:26,605 something like HTML. 435 00:16:27,065 --> 00:16:28,745 Right? So we have the same sorts of 436 00:16:28,745 --> 00:16:31,225 problems like like Tom was saying with, SQL 437 00:16:31,225 --> 00:16:34,024 injection. SQL is the same thing. We use 438 00:16:34,024 --> 00:16:36,284 a markup language that combines 439 00:16:36,985 --> 00:16:37,485 the 440 00:16:38,450 --> 00:16:38,950 instructions 441 00:16:39,970 --> 00:16:41,649 with the met with the data that the 442 00:16:41,649 --> 00:16:43,410 instruction is supposed to work on. It's very 443 00:16:43,410 --> 00:16:46,290 object oriented. And we think, oh, object oriented 444 00:16:46,290 --> 00:16:48,870 is a much better model for programming 445 00:16:49,330 --> 00:16:51,910 because you contain the data with the code 446 00:16:52,254 --> 00:16:54,414 and they're one unit. Right? So they always 447 00:16:54,414 --> 00:16:56,574 know what each other are doing. But the 448 00:16:56,574 --> 00:16:58,735 problem with that the the trade off there 449 00:16:58,735 --> 00:17:00,654 is is exactly what this article is talking 450 00:17:00,654 --> 00:17:01,394 about. Right? 451 00:17:01,855 --> 00:17:04,595 You inject the data. Like, if we made 452 00:17:05,309 --> 00:17:08,210 HTML, if we broke HTML into two pieces 453 00:17:09,149 --> 00:17:11,649 and you said, I'm gonna have a 454 00:17:11,950 --> 00:17:13,649 I'm gonna have a control channel 455 00:17:14,109 --> 00:17:16,130 that says render this 456 00:17:16,750 --> 00:17:19,230 and it says token $1.04 5 or token 457 00:17:19,230 --> 00:17:22,315 $9.08 6 or whatever, and then you carry 458 00:17:22,315 --> 00:17:24,634 the actual data for that token in a 459 00:17:24,634 --> 00:17:25,534 separate channel, 460 00:17:26,714 --> 00:17:28,954 and the tokens were well formatted and well 461 00:17:28,954 --> 00:17:29,454 understood, 462 00:17:29,835 --> 00:17:31,934 you'd get rid of 90% of the SQL 463 00:17:31,994 --> 00:17:34,809 side channel, all these other weirdo attacks 464 00:17:35,190 --> 00:17:37,990 that we get in CSS and SQL and 465 00:17:37,990 --> 00:17:40,170 just all this stuff that we run into 466 00:17:40,869 --> 00:17:41,930 on the Internet. 467 00:17:42,789 --> 00:17:45,590 It's the same principle. Right? It's just separation 468 00:17:45,590 --> 00:17:48,410 of concerns. It's using segmentation to your advantage. 469 00:17:49,295 --> 00:17:51,855 And so yeah. I mean and by the 470 00:17:51,855 --> 00:17:53,535 way, all this holds true in networks too. 471 00:17:53,535 --> 00:17:56,275 If your network has been breached by somebody, 472 00:17:57,775 --> 00:17:59,694 the number one way you can stop that 473 00:17:59,694 --> 00:18:01,950 breach from being a real problem is by 474 00:18:01,950 --> 00:18:04,029 segmenting so that people can't get done with 475 00:18:04,029 --> 00:18:05,169 things. And Tom 476 00:18:05,470 --> 00:18:05,869 was, 477 00:18:06,429 --> 00:18:09,710 using the wife analogy. Right? And so, yeah, 478 00:18:09,710 --> 00:18:11,549 you have the you have an option to 479 00:18:11,549 --> 00:18:14,269 fix these kinds of vulnerabilities, weaknesses in a 480 00:18:14,269 --> 00:18:15,250 system architecturally, 481 00:18:16,204 --> 00:18:18,144 or you can fix them some other way. 482 00:18:18,444 --> 00:18:21,244 But when you fix them whatever that other 483 00:18:21,244 --> 00:18:22,065 way is, 484 00:18:22,605 --> 00:18:23,105 typically, 485 00:18:23,805 --> 00:18:27,164 you're injecting more state into the system, which 486 00:18:27,164 --> 00:18:28,785 is gonna make it more fragile, 487 00:18:29,569 --> 00:18:32,150 or you're injecting more complexity 488 00:18:34,049 --> 00:18:35,029 or computational 489 00:18:35,490 --> 00:18:36,549 power required, 490 00:18:37,089 --> 00:18:39,430 which is gonna make the system less performant, 491 00:18:40,609 --> 00:18:41,109 and 492 00:18:41,410 --> 00:18:41,910 really 493 00:18:42,289 --> 00:18:42,789 is 494 00:18:43,994 --> 00:18:46,015 gonna work most of the time, 495 00:18:46,394 --> 00:18:49,295 but probably not a 100% of the time. 496 00:18:49,515 --> 00:18:51,914 Right? And we've seen this over and over 497 00:18:51,914 --> 00:18:54,095 and over with security technologies, 498 00:18:54,394 --> 00:18:55,214 whether it's 499 00:18:55,515 --> 00:18:56,894 stateful packet inspection, 500 00:18:57,434 --> 00:18:59,214 whether it's proxy servers, 501 00:18:59,730 --> 00:19:00,470 whether it's, 502 00:19:02,049 --> 00:19:03,809 you know, I'm I'm thinking about it. So 503 00:19:03,809 --> 00:19:05,650 then then then we go to what? 504 00:19:06,369 --> 00:19:09,890 Layer seven firewalls, right, that inspect all the 505 00:19:09,890 --> 00:19:12,284 traffic. And then we, you know, and then 506 00:19:12,284 --> 00:19:15,804 we've got, you know, SQL injection protection. Right? 507 00:19:15,804 --> 00:19:17,664 All the different layers 508 00:19:18,204 --> 00:19:18,944 of protection 509 00:19:19,565 --> 00:19:20,065 that 510 00:19:20,764 --> 00:19:21,585 are imperfect 511 00:19:22,444 --> 00:19:23,904 and are ultimately 512 00:19:24,284 --> 00:19:24,784 workarounds 513 00:19:25,085 --> 00:19:26,605 to an architectural solution, 514 00:19:28,039 --> 00:19:30,299 that could be more elegant, more performant. 515 00:19:32,200 --> 00:19:34,140 Yeah. Yeah. Yep. 516 00:19:34,679 --> 00:19:36,700 And performance is a big issue too. 517 00:19:37,000 --> 00:19:39,400 Like, you know, a lot of DDoS protection 518 00:19:39,400 --> 00:19:41,579 things, they do things that are definitely 519 00:19:42,039 --> 00:19:42,779 they reduce 520 00:19:43,615 --> 00:19:44,115 performance 521 00:19:44,894 --> 00:19:45,394 overall 522 00:19:45,775 --> 00:19:47,954 in order to protect you from DDoS. 523 00:19:48,654 --> 00:19:51,375 And that's, you know, alright. There's the trade 524 00:19:51,375 --> 00:19:52,654 off. Right? I don't I don't want the 525 00:19:52,654 --> 00:19:54,034 DDoS attack, but then 526 00:19:54,734 --> 00:19:56,994 do I always want the performance hit? 527 00:19:57,430 --> 00:19:59,349 And we you know, we don't really it's 528 00:19:59,349 --> 00:20:01,930 another issue that we don't really think about. 529 00:20:02,549 --> 00:20:04,549 Yeah. So I think this is I think 530 00:20:04,549 --> 00:20:06,250 this is incredibly interesting, 531 00:20:07,109 --> 00:20:08,330 just from a design 532 00:20:08,869 --> 00:20:09,369 paradigm, 533 00:20:09,670 --> 00:20:10,170 understanding 534 00:20:10,950 --> 00:20:12,490 how to design things, 535 00:20:13,884 --> 00:20:14,384 perspective 536 00:20:14,924 --> 00:20:15,424 on, 537 00:20:16,204 --> 00:20:17,345 you know, not just 538 00:20:17,644 --> 00:20:18,144 protocols, 539 00:20:19,164 --> 00:20:19,904 not just, 540 00:20:22,045 --> 00:20:23,105 the networks, protocols, 541 00:20:23,404 --> 00:20:23,904 applications, 542 00:20:24,845 --> 00:20:26,765 security, they're all of the above. Even your 543 00:20:26,765 --> 00:20:27,664 physical security. 544 00:20:28,220 --> 00:20:30,880 Even your physical security falls to this. Right? 545 00:20:31,580 --> 00:20:32,700 You you have to you have to make 546 00:20:32,700 --> 00:20:33,599 sure that your 547 00:20:33,980 --> 00:20:36,000 control plane is not tied up 548 00:20:36,380 --> 00:20:37,359 with your 549 00:20:37,820 --> 00:20:40,299 data channel as much as you possibly can. 550 00:20:40,299 --> 00:20:41,980 There are some places where you can't avoid 551 00:20:41,980 --> 00:20:43,759 it at the point of access. 552 00:20:44,345 --> 00:20:45,945 But once you get behind the point of 553 00:20:45,945 --> 00:20:46,445 access, 554 00:20:46,984 --> 00:20:49,164 like, you know, these should be different things. 555 00:20:50,105 --> 00:20:51,945 So, yeah, I think it's I think it's 556 00:20:51,945 --> 00:20:52,924 an important thing. 557 00:20:54,424 --> 00:20:55,164 So okay. 558 00:20:55,705 --> 00:20:56,605 Second topic 559 00:20:57,970 --> 00:20:58,789 is unfortunate. 560 00:21:00,529 --> 00:21:02,069 It's sad and unfortunate. 561 00:21:03,809 --> 00:21:06,390 So I saw this article in Network World. 562 00:21:07,329 --> 00:21:07,829 I 563 00:21:08,130 --> 00:21:10,369 didn't get I I guess I could suppose 564 00:21:10,369 --> 00:21:13,025 it's by rational effect. It's not by rational 565 00:21:13,025 --> 00:21:16,144 effects but the analysis is done by rational 566 00:21:16,144 --> 00:21:16,644 effects. 567 00:21:17,184 --> 00:21:19,205 And so I'm not entirely certain 568 00:21:19,664 --> 00:21:21,744 the timelines of all these and everything like 569 00:21:21,744 --> 00:21:24,630 that. But the article headline is is that 570 00:21:24,869 --> 00:21:27,910 there are tech playoffs surpassed 45,000 571 00:21:27,910 --> 00:21:29,609 in early twenty twenty six. 572 00:21:32,630 --> 00:21:33,769 And the immediate 573 00:21:34,150 --> 00:21:36,490 reasoning that that is used here 574 00:21:36,869 --> 00:21:37,609 is that 575 00:21:38,904 --> 00:21:42,505 firms across the industry restructure operations and shift 576 00:21:42,505 --> 00:21:43,644 resources towards 577 00:21:44,345 --> 00:21:45,964 AI and automation investments, 578 00:21:48,505 --> 00:21:50,345 which I think is, like I think a 579 00:21:50,345 --> 00:21:52,039 lot of people are saying that. I think 580 00:21:52,039 --> 00:21:53,799 it's about what what is provable at this 581 00:21:53,799 --> 00:21:55,339 point. Yeah. I think a lot yeah. 582 00:21:56,519 --> 00:21:57,019 Yeah. 583 00:21:57,480 --> 00:21:59,419 Well, essentially, they say investments, 584 00:21:59,799 --> 00:22:00,299 not 585 00:22:00,919 --> 00:22:01,900 they're not saying 586 00:22:02,759 --> 00:22:05,480 we're replacing the people with AI specifically, although 587 00:22:05,480 --> 00:22:07,244 that has been a thing in a lot 588 00:22:07,244 --> 00:22:07,904 of layoffs. 589 00:22:08,845 --> 00:22:11,085 But that we don't have money for people 590 00:22:11,085 --> 00:22:12,945 because we're investing in AI. 591 00:22:13,484 --> 00:22:14,865 Because what? We 592 00:22:15,644 --> 00:22:16,625 expect AI 593 00:22:16,924 --> 00:22:17,424 to, 594 00:22:18,605 --> 00:22:19,265 you know, 595 00:22:19,880 --> 00:22:21,880 take over? Like, what what are we thinking 596 00:22:21,880 --> 00:22:24,299 here? Well, I think depending on the organization, 597 00:22:24,440 --> 00:22:26,539 some of it is just flat out CapEx. 598 00:22:26,919 --> 00:22:27,819 Right? Like, 599 00:22:28,200 --> 00:22:30,200 we're letting people go because we need more 600 00:22:30,200 --> 00:22:30,700 GPUs. 601 00:22:31,319 --> 00:22:33,319 You know? I I I think some of 602 00:22:33,319 --> 00:22:34,619 it is going to CapEx. 603 00:22:36,404 --> 00:22:38,505 I I think some of it is opportunistic. 604 00:22:39,525 --> 00:22:41,865 And what I mean by that is, 605 00:22:42,484 --> 00:22:45,525 hey. We we've overhired, or we're not sure 606 00:22:45,525 --> 00:22:47,365 that we're getting the value out of all 607 00:22:47,365 --> 00:22:49,204 the people that we have on payroll, and 608 00:22:49,204 --> 00:22:52,390 this gives us a an opportunity to to 609 00:22:52,390 --> 00:22:55,029 trim things down. I think we are we 610 00:22:55,029 --> 00:22:56,549 are still in the middle of an era 611 00:22:56,549 --> 00:22:57,930 where that's happening. 612 00:22:58,869 --> 00:23:00,630 And and then I think some of it 613 00:23:00,630 --> 00:23:01,690 too is 614 00:23:02,470 --> 00:23:03,930 there's this promise, 615 00:23:04,994 --> 00:23:06,674 some of which I think will be realized 616 00:23:06,674 --> 00:23:08,375 that AI is gonna optimize 617 00:23:08,835 --> 00:23:10,134 some knowledge work. 618 00:23:10,595 --> 00:23:11,494 And I think, 619 00:23:12,195 --> 00:23:15,174 you know, there's this there's this leadership philosophy 620 00:23:15,315 --> 00:23:17,890 that, like, peep people people use the time 621 00:23:17,890 --> 00:23:20,210 that they have. People use the resources that 622 00:23:20,210 --> 00:23:22,049 they have. It's like, you know, time is 623 00:23:22,049 --> 00:23:24,390 kinda like closet space. You're gonna absorb 624 00:23:25,089 --> 00:23:27,190 whatever you have with what you're given. 625 00:23:27,650 --> 00:23:30,130 And so I think organizations are thinking, well, 626 00:23:30,130 --> 00:23:32,884 if we if we cut headcount, that's gonna 627 00:23:32,884 --> 00:23:35,065 force the people that we have today 628 00:23:35,684 --> 00:23:37,924 to be more efficient. And I think I 629 00:23:37,924 --> 00:23:40,424 I think there's there's some truth to that, 630 00:23:41,125 --> 00:23:42,744 but I don't know that 631 00:23:44,500 --> 00:23:46,980 some of our organizations really even understand what's 632 00:23:46,980 --> 00:23:49,859 going on enough to make those decisions in 633 00:23:49,859 --> 00:23:51,480 the best best way. 634 00:23:51,940 --> 00:23:53,539 But I think all of those things are 635 00:23:53,539 --> 00:23:54,919 part of what's happening. 636 00:23:55,940 --> 00:23:58,414 I I I find it interesting if if 637 00:23:58,414 --> 00:24:01,055 somebody is saying we're gonna force our people 638 00:24:01,055 --> 00:24:01,715 to use, 639 00:24:02,414 --> 00:24:03,875 AI by giving them more 640 00:24:04,174 --> 00:24:04,674 work. 641 00:24:05,295 --> 00:24:06,974 You know, but most people see when they 642 00:24:06,974 --> 00:24:08,974 see a layoff is an input to the 643 00:24:08,974 --> 00:24:11,535 amygdala in their brain that says, danger. You're 644 00:24:11,535 --> 00:24:13,529 probably next. I don't know of 645 00:24:14,150 --> 00:24:15,289 anybody who makes fundamental 646 00:24:15,670 --> 00:24:17,289 paradigm shifts in how they work 647 00:24:17,829 --> 00:24:18,650 under threat. 648 00:24:19,269 --> 00:24:22,230 And, like, so the so I I totally 649 00:24:22,230 --> 00:24:23,910 believe that what you're like, some people are 650 00:24:23,910 --> 00:24:24,970 using this rationalization, 651 00:24:25,430 --> 00:24:26,970 Yvonne. Like, it seems like a very 652 00:24:27,585 --> 00:24:29,025 rational thing for someone to say, oh, yeah. 653 00:24:29,025 --> 00:24:30,484 We'll do this, and then this will happen. 654 00:24:30,544 --> 00:24:32,065 But I don't know. I have a hard 655 00:24:32,065 --> 00:24:34,065 time seeing that actually laying out like they 656 00:24:34,065 --> 00:24:36,244 see like, they're predicting in reality 657 00:24:36,704 --> 00:24:38,884 just because of how people look how people 658 00:24:39,105 --> 00:24:40,785 look at things. And not right now is 659 00:24:40,785 --> 00:24:42,144 not a great time to lose your job. 660 00:24:42,144 --> 00:24:44,269 And, you know, I I don't think people 661 00:24:44,269 --> 00:24:46,109 think straight when you might get you might 662 00:24:46,109 --> 00:24:48,190 get laid off next week. I think that's 663 00:24:48,190 --> 00:24:50,269 absolutely true. I actually went on a rant 664 00:24:50,269 --> 00:24:52,930 privately with a friend today about my frustration 665 00:24:53,230 --> 00:24:54,450 with engineers 666 00:24:55,150 --> 00:24:57,410 who make decisions that impact people 667 00:24:57,795 --> 00:24:58,295 and 668 00:24:58,755 --> 00:25:01,315 don't think through the human elements of it. 669 00:25:01,315 --> 00:25:02,134 Right? Like, 670 00:25:02,434 --> 00:25:02,934 people 671 00:25:03,555 --> 00:25:05,715 first of all, like, reads some Kahneman and 672 00:25:05,715 --> 00:25:06,215 Diverski. 673 00:25:06,674 --> 00:25:07,735 People aren't rational. 674 00:25:08,195 --> 00:25:11,075 They don't make economic decisions rationally. They don't 675 00:25:11,075 --> 00:25:13,390 necessarily make career decision rationally. 676 00:25:14,009 --> 00:25:17,609 And we we rationalize things as if people 677 00:25:17,609 --> 00:25:19,390 are going to decide to behave rationally. 678 00:25:19,929 --> 00:25:22,109 They do not, first of all. 679 00:25:22,490 --> 00:25:23,230 And then, 680 00:25:23,929 --> 00:25:25,230 also, there are, 681 00:25:26,345 --> 00:25:29,244 second and third order order effects of of 682 00:25:29,545 --> 00:25:32,184 these kinds of changes that may not be 683 00:25:32,184 --> 00:25:36,045 seen immediately, but ultimately bear themselves out in 684 00:25:36,105 --> 00:25:37,565 compliance and reliability 685 00:25:38,585 --> 00:25:39,085 and, 686 00:25:40,539 --> 00:25:43,019 you know, the the ability to execute in 687 00:25:43,019 --> 00:25:43,839 certain ways. 688 00:25:44,539 --> 00:25:46,539 Yeah. So it's, it's just a really interesting 689 00:25:46,539 --> 00:25:47,359 time. Yeah. 690 00:25:49,740 --> 00:25:51,599 So if you're saying to somebody, 691 00:25:52,619 --> 00:25:53,359 for instance, 692 00:25:54,954 --> 00:25:57,115 you can do four times as much using 693 00:25:57,115 --> 00:25:57,615 AI. 694 00:25:59,034 --> 00:26:01,355 In the inverse, if I were to think 695 00:26:01,355 --> 00:26:03,115 about that, and there's a way to think 696 00:26:03,115 --> 00:26:05,755 about that, which is that my time is 697 00:26:05,755 --> 00:26:07,294 worth four times less. 698 00:26:08,950 --> 00:26:10,169 And and so 699 00:26:10,549 --> 00:26:12,710 there's a bit of a there's a bit 700 00:26:12,710 --> 00:26:14,650 of a struggle there on how, 701 00:26:15,109 --> 00:26:17,429 as a human, you should see that and 702 00:26:17,429 --> 00:26:18,569 understand that 703 00:26:19,589 --> 00:26:20,970 and not be defensive 704 00:26:21,269 --> 00:26:22,089 about it. 705 00:26:22,804 --> 00:26:25,044 And and I'm not sure exactly how you 706 00:26:25,044 --> 00:26:25,785 do that 707 00:26:26,164 --> 00:26:26,984 as a person. 708 00:26:28,484 --> 00:26:30,724 And then, again, the person, as Yvonne is 709 00:26:30,724 --> 00:26:32,345 saying, is making the decision, 710 00:26:32,724 --> 00:26:35,285 and they're not thinking about second and third 711 00:26:35,285 --> 00:26:37,544 order effects. There's been this thing about, 712 00:26:39,019 --> 00:26:40,779 Sam Altman said the other day on some 713 00:26:40,779 --> 00:26:43,259 clip, he said, you know, we see knowledge 714 00:26:43,259 --> 00:26:45,099 as a service that you can buy like 715 00:26:45,099 --> 00:26:46,079 water or electricity 716 00:26:46,859 --> 00:26:47,359 effectively 717 00:26:47,819 --> 00:26:49,980 or or intelligence. He didn't say knowledge. He 718 00:26:49,980 --> 00:26:50,720 said intelligence. 719 00:26:51,579 --> 00:26:53,200 And I thought, wow. 720 00:26:53,659 --> 00:26:54,319 You know? 721 00:26:54,775 --> 00:26:57,355 I'm not sure that intelligence is is amenable 722 00:26:57,414 --> 00:26:58,315 to that kind 723 00:26:58,934 --> 00:26:59,674 of commoditization. 724 00:27:01,015 --> 00:27:03,515 And and then other people will say, well, 725 00:27:03,575 --> 00:27:05,494 once we get AIs in place, I can 726 00:27:05,494 --> 00:27:08,134 cut my workforce in half and sell just 727 00:27:08,134 --> 00:27:10,210 as much. And then then you have to 728 00:27:10,210 --> 00:27:10,710 ask, 729 00:27:11,250 --> 00:27:12,710 who are you selling to? 730 00:27:14,049 --> 00:27:16,549 If everybody cuts their workforce in half 731 00:27:17,329 --> 00:27:19,009 Mhmm. Then who has the money to buy 732 00:27:19,009 --> 00:27:20,230 your products? Right. 733 00:27:21,250 --> 00:27:21,750 Like, 734 00:27:22,115 --> 00:27:23,255 how does that work? 735 00:27:23,634 --> 00:27:26,035 So, like, there's all these weird second order 736 00:27:26,035 --> 00:27:26,934 knock on effects 737 00:27:27,875 --> 00:27:28,375 that 738 00:27:28,994 --> 00:27:31,475 are un I'm sure they're unintended consequences, but 739 00:27:31,475 --> 00:27:32,535 nonetheless, like, 740 00:27:32,914 --> 00:27:35,019 there are things that are happening. I think 741 00:27:35,019 --> 00:27:37,420 there's another knock on effect, which is what 742 00:27:37,420 --> 00:27:39,980 happens when so right now, your shareholders would 743 00:27:39,980 --> 00:27:42,240 be very pleased with you cutting cost massively 744 00:27:42,299 --> 00:27:44,859 because they're not in your business. They're in 745 00:27:44,859 --> 00:27:46,460 the business of turning their money into more 746 00:27:46,460 --> 00:27:48,400 money. So the shareholders like to see, 747 00:27:48,835 --> 00:27:50,515 cost go down because they can at any 748 00:27:50,515 --> 00:27:51,894 time get out 749 00:27:52,195 --> 00:27:53,795 and get what get you know, give make 750 00:27:53,795 --> 00:27:56,115 whatever gains they were gonna make, whereas employees 751 00:27:56,115 --> 00:27:58,275 can't just get out instantly on a dime 752 00:27:58,275 --> 00:28:01,049 like investors can, like shareholders can. And so 753 00:28:01,169 --> 00:28:04,049 but what happens when you when you gut 754 00:28:04,049 --> 00:28:06,369 innovation? I saw another I saw another article, 755 00:28:06,369 --> 00:28:07,889 maybe we could talk about it another time, 756 00:28:07,889 --> 00:28:08,389 about, 757 00:28:08,849 --> 00:28:09,829 research and development, 758 00:28:10,369 --> 00:28:12,690 someone who worked at HP who sees what's 759 00:28:12,690 --> 00:28:13,669 happening at Intel. 760 00:28:14,265 --> 00:28:16,105 You know, what happens if you if you've 761 00:28:16,105 --> 00:28:18,025 got research and development. But in this case, 762 00:28:18,025 --> 00:28:19,404 you're going to get innovation 763 00:28:19,865 --> 00:28:21,305 if you tell people if you if you 764 00:28:21,305 --> 00:28:22,744 do all these things to the workforce, you're 765 00:28:22,744 --> 00:28:23,644 gonna get innovation. 766 00:28:23,945 --> 00:28:26,184 And that works for the short term. You 767 00:28:26,184 --> 00:28:28,105 probably will cut cost. Your investors will be 768 00:28:28,105 --> 00:28:28,605 happy. 769 00:28:29,099 --> 00:28:31,119 Everything will be fine. But once everything's commoditized 770 00:28:31,339 --> 00:28:33,339 down to the same large language model doing 771 00:28:33,339 --> 00:28:34,799 the same thing for all the companies, 772 00:28:35,980 --> 00:28:37,740 then you've lost the ability to climb out 773 00:28:37,740 --> 00:28:39,259 of that hole. And then what happens then? 774 00:28:39,259 --> 00:28:41,099 Now this happens in ten years or twenty 775 00:28:41,099 --> 00:28:43,375 years. It doesn't happen this quarter. But I 776 00:28:43,375 --> 00:28:45,634 think that's a I don't know. Long term 777 00:28:45,934 --> 00:28:47,295 I don't I don't I wonder how many 778 00:28:47,295 --> 00:28:49,295 people are thinking long term about what these 779 00:28:49,295 --> 00:28:51,075 strategies are leading us towards. 780 00:28:53,615 --> 00:28:55,490 Yeah. I was also part of a a 781 00:28:55,490 --> 00:28:57,890 conversation with a colleague the other day, and 782 00:28:57,890 --> 00:28:59,430 we were talking about 783 00:29:00,450 --> 00:29:01,750 gains in 784 00:29:02,049 --> 00:29:02,549 in, 785 00:29:02,930 --> 00:29:04,630 specifically, developer productivity 786 00:29:05,090 --> 00:29:07,430 and what that means, but, also, 787 00:29:09,035 --> 00:29:09,694 the entire 788 00:29:09,994 --> 00:29:12,335 system that exists around the developer. 789 00:29:12,795 --> 00:29:14,734 Right? And so if you think about 790 00:29:15,275 --> 00:29:19,295 agile programming methodology and what existed before that, 791 00:29:19,355 --> 00:29:21,615 all of that really came out of 792 00:29:21,994 --> 00:29:24,329 what we call, like, the Toyota Way, books 793 00:29:24,329 --> 00:29:27,289 like the goal by Goldratt. Right? And what 794 00:29:27,289 --> 00:29:29,470 we what we learned from 795 00:29:29,769 --> 00:29:30,750 that manufacturing 796 00:29:31,690 --> 00:29:32,190 world 797 00:29:32,890 --> 00:29:33,710 is that 798 00:29:34,890 --> 00:29:38,589 part of pushing code through a system is 799 00:29:38,875 --> 00:29:40,414 doing things like reducing 800 00:29:40,875 --> 00:29:41,934 work in progress, 801 00:29:42,714 --> 00:29:44,255 managing your backlog, 802 00:29:44,954 --> 00:29:46,255 making sure that 803 00:29:46,794 --> 00:29:49,615 every stage in the development process 804 00:29:50,714 --> 00:29:52,174 is working together, 805 00:29:52,609 --> 00:29:54,369 and you're not you've not got these big 806 00:29:54,369 --> 00:29:57,509 huge backlogs of work that are piling up. 807 00:29:58,049 --> 00:30:00,390 Well, what AI is doing 808 00:30:00,690 --> 00:30:03,490 is it's moving the choke point. In other 809 00:30:03,490 --> 00:30:05,590 words, maybe writing the code 810 00:30:06,130 --> 00:30:07,269 is no longer 811 00:30:07,730 --> 00:30:08,230 the 812 00:30:08,875 --> 00:30:11,134 slowest part of the process, 813 00:30:11,674 --> 00:30:14,095 but there are other parts of that process, 814 00:30:14,154 --> 00:30:16,474 whether they're QA and testing, whether it's the 815 00:30:16,474 --> 00:30:18,875 speed at which users can consume the changes, 816 00:30:18,875 --> 00:30:20,654 whether it's, you know, 817 00:30:21,329 --> 00:30:24,289 how you're meeting your compliance requirements and standards. 818 00:30:24,289 --> 00:30:26,230 There's a lot of work that goes around 819 00:30:26,929 --> 00:30:28,789 the work of generating code. 820 00:30:29,410 --> 00:30:30,549 And so, ultimately, 821 00:30:32,130 --> 00:30:33,750 if all we're doing is 822 00:30:34,505 --> 00:30:36,684 shoving more code through the system, 823 00:30:36,984 --> 00:30:39,304 we're just creating new places for there to 824 00:30:39,304 --> 00:30:40,125 be backlogs. 825 00:30:40,505 --> 00:30:43,404 And so it's not gonna improve the overall 826 00:30:43,625 --> 00:30:45,244 speed of the system 827 00:30:45,704 --> 00:30:47,384 the way we think it is. And in 828 00:30:47,384 --> 00:30:48,444 so many ways, 829 00:30:49,299 --> 00:30:51,220 a lot of what's going on with AI 830 00:30:51,220 --> 00:30:53,799 right now, we're just creating new local optimization 831 00:30:54,180 --> 00:30:54,680 problems. 832 00:30:55,140 --> 00:30:57,779 Right? Like, because we're optimizing in a different 833 00:30:57,779 --> 00:30:58,279 place, 834 00:30:58,820 --> 00:31:01,380 but that doesn't mean that the whole system 835 00:31:01,380 --> 00:31:03,904 is speeding up. And I don't think anybody 836 00:31:03,904 --> 00:31:05,744 can know what that looks like yet because 837 00:31:05,744 --> 00:31:06,884 it's so new. 838 00:31:07,345 --> 00:31:09,585 And so we're making assumptions that the whole 839 00:31:09,585 --> 00:31:11,845 system is gonna improve in productivity. 840 00:31:12,545 --> 00:31:15,105 I don't believe that that is gonna be 841 00:31:15,105 --> 00:31:17,359 the case. I think there will be improvements 842 00:31:17,359 --> 00:31:19,200 in productivity, but I think we got a 843 00:31:19,200 --> 00:31:20,339 long way off 844 00:31:20,640 --> 00:31:22,880 before we fix the whole system. And so 845 00:31:22,880 --> 00:31:24,799 I think we're making some assumptions that are 846 00:31:24,799 --> 00:31:25,779 gonna end up 847 00:31:26,240 --> 00:31:26,740 causing 848 00:31:27,519 --> 00:31:29,460 challenges in the future because 849 00:31:29,759 --> 00:31:31,034 we're we're gonna end up with huge backsell 850 00:31:31,274 --> 00:31:33,294 backlogs of work just in different places. 851 00:31:33,674 --> 00:31:35,994 Well, it's and, yeah, it's even I think 852 00:31:35,994 --> 00:31:37,755 it's actually even worse than that in a 853 00:31:37,755 --> 00:31:38,654 sense that 854 00:31:39,115 --> 00:31:41,294 we used to avoid tech debt 855 00:31:41,835 --> 00:31:42,815 like the plague. 856 00:31:43,339 --> 00:31:45,980 Like, tech debt was horrible. We would spend 857 00:31:45,980 --> 00:31:46,720 many, many, 858 00:31:47,820 --> 00:31:51,440 many, many hours of developer time refactoring code, 859 00:31:51,900 --> 00:31:54,480 not because it was broken per se, 860 00:31:55,285 --> 00:31:57,224 but because nobody understood it 861 00:31:57,525 --> 00:31:59,765 because it was tech debt. Now we almost 862 00:31:59,765 --> 00:32:01,704 seem to celebrate tech debt. 863 00:32:02,005 --> 00:32:04,184 Oh, you don't know how it works? Yay. 864 00:32:04,964 --> 00:32:05,464 Like, 865 00:32:06,005 --> 00:32:06,505 really? 866 00:32:07,570 --> 00:32:10,289 And I'm gonna go, well, the the the 867 00:32:10,289 --> 00:32:13,170 humans can can code review the code that 868 00:32:13,170 --> 00:32:16,130 they don't understand because some AI system is 869 00:32:16,130 --> 00:32:16,950 writing it. 870 00:32:17,730 --> 00:32:19,410 Yeah. I'm not I'm not really sure how 871 00:32:19,410 --> 00:32:20,950 this is gonna work. I 872 00:32:21,424 --> 00:32:23,904 worry a lot about that sooner. Thought all 873 00:32:23,904 --> 00:32:25,904 the way through this analogy. And and But 874 00:32:25,904 --> 00:32:27,684 it's just like a different kind of inflation. 875 00:32:28,065 --> 00:32:29,845 Right? Like, the perception 876 00:32:30,144 --> 00:32:32,785 of the value of that the the perception 877 00:32:32,785 --> 00:32:34,565 of the cost of that debt is lower. 878 00:32:36,720 --> 00:32:39,140 But is it really? Right? Or is just 879 00:32:39,440 --> 00:32:39,940 everything 880 00:32:41,200 --> 00:32:43,920 more expensive? Right? Like, there's is is is 881 00:32:43,920 --> 00:32:46,400 there just more money in the system? Right? 882 00:32:46,400 --> 00:32:48,820 And so it seems like it's cheaper, 883 00:32:49,534 --> 00:32:50,994 but there's just more in the system. 884 00:32:51,615 --> 00:32:53,615 But we don't really know whether it has 885 00:32:53,615 --> 00:32:56,255 the same value it used to. I actually 886 00:32:56,255 --> 00:32:57,474 think that analogy 887 00:32:58,015 --> 00:32:59,694 might hold together better than I thought it 888 00:32:59,694 --> 00:33:01,315 did when I first said it out loud. 889 00:33:01,839 --> 00:33:03,839 Yeah. No. No. It does. Because because it's 890 00:33:03,839 --> 00:33:05,299 the same thing with photography 891 00:33:05,839 --> 00:33:08,899 and motion pictures and music and, 892 00:33:09,200 --> 00:33:10,419 you know, even, 893 00:33:10,960 --> 00:33:12,819 dare I say it, even truth itself 894 00:33:13,119 --> 00:33:16,179 to some degree. Right? Even truth itself. 895 00:33:16,525 --> 00:33:17,825 When you get to the point 896 00:33:18,125 --> 00:33:18,625 where 897 00:33:19,005 --> 00:33:21,505 you wanna cover up one incident 898 00:33:22,045 --> 00:33:22,944 by publishing 899 00:33:24,045 --> 00:33:24,785 10,000 900 00:33:24,924 --> 00:33:27,265 similar incidents, none of which were true, 901 00:33:28,285 --> 00:33:30,545 and that the one the ten thousandth becomes 902 00:33:30,684 --> 00:33:31,184 implausible 903 00:33:31,930 --> 00:33:34,009 because there were 999,000 904 00:33:34,009 --> 00:33:36,250 whatever it is, nine nine thousand nine hundred 905 00:33:36,250 --> 00:33:38,109 ninety nine that were not true, 906 00:33:39,609 --> 00:33:41,230 there's an entire different 907 00:33:42,330 --> 00:33:42,830 perception 908 00:33:43,450 --> 00:33:44,190 or texture 909 00:33:44,994 --> 00:33:47,015 to the to the way we see things 910 00:33:47,315 --> 00:33:48,375 once that happens. 911 00:33:48,914 --> 00:33:51,474 And that's that that worries me a lot. 912 00:33:51,474 --> 00:33:53,075 Maybe that's a different show, though, is to 913 00:33:53,075 --> 00:33:53,815 talk about 914 00:33:54,115 --> 00:33:55,654 AI and perception and 915 00:33:56,434 --> 00:33:58,375 and What is reality, Russ? 916 00:33:59,639 --> 00:34:01,639 Stuff like that. Just unpack that for us 917 00:34:01,639 --> 00:34:02,940 and we'll call it a day. 918 00:34:06,279 --> 00:34:08,199 Well, are you a realist or an not 919 00:34:08,519 --> 00:34:09,659 or an anti realist? 920 00:34:10,359 --> 00:34:11,424 Do you do do you are you a 921 00:34:11,424 --> 00:34:13,744 nominalist or a practical nominalist? You knew what 922 00:34:13,744 --> 00:34:15,505 you were getting into, Yvonne, when you asked 923 00:34:15,505 --> 00:34:17,125 that question. Yeah. Yeah. 924 00:34:17,664 --> 00:34:18,804 I know. I 925 00:34:19,344 --> 00:34:20,964 I was just trying to get a laugh. 926 00:34:26,789 --> 00:34:28,510 So alright. Cool. So I think we've hit 927 00:34:28,510 --> 00:34:30,750 both of those pretty well. So anything else 928 00:34:30,750 --> 00:34:33,630 to anybody wanna, like, say anything about those 929 00:34:33,630 --> 00:34:36,030 two before we call show 300 930 00:34:36,030 --> 00:34:36,530 done? 931 00:34:37,070 --> 00:34:38,050 Nice round 932 00:34:38,349 --> 00:34:40,734 number. I I will end by saying I 933 00:34:40,734 --> 00:34:42,594 don't think the end of civilization 934 00:34:43,054 --> 00:34:44,195 as we know it 935 00:34:44,574 --> 00:34:46,655 is as near as we believe. I think 936 00:34:46,655 --> 00:34:49,534 it's probably gonna be different. Like, you know, 937 00:34:49,534 --> 00:34:51,934 just like the just like the Internet changed 938 00:34:51,934 --> 00:34:55,450 civilization and just like mobile changed civilization. I 939 00:34:55,450 --> 00:34:57,630 think AI is going to make some changes, 940 00:34:58,090 --> 00:34:59,289 but I think at the end of the 941 00:34:59,289 --> 00:34:59,789 day, 942 00:35:00,489 --> 00:35:02,170 we're still gonna be people, and there's still 943 00:35:02,170 --> 00:35:04,170 gonna be interesting work to do. And it 944 00:35:04,170 --> 00:35:04,670 may, 945 00:35:05,849 --> 00:35:07,769 shove around in the different places and look 946 00:35:07,769 --> 00:35:10,065 a little different in the future, but I 947 00:35:10,065 --> 00:35:10,625 think the, 948 00:35:11,585 --> 00:35:14,065 the the predictions of the end of society 949 00:35:14,065 --> 00:35:15,445 as we know it are 950 00:35:15,984 --> 00:35:16,644 a bit 951 00:35:17,184 --> 00:35:17,684 overblown. 952 00:35:18,144 --> 00:35:19,445 That's what I will say. 953 00:35:20,304 --> 00:35:21,844 Yeah. We'll see. 954 00:35:22,430 --> 00:35:25,250 I also feel pretty optimistic about things. 955 00:35:25,869 --> 00:35:26,530 I think, 956 00:35:27,230 --> 00:35:29,070 you know, this for for many people, they 957 00:35:29,070 --> 00:35:31,070 didn't live through most of us didn't live 958 00:35:31,070 --> 00:35:33,869 through the transition from the agricultural age to 959 00:35:33,869 --> 00:35:36,105 the industrial age. And so we don't know 960 00:35:36,244 --> 00:35:37,844 what it was like we didn't witness the 961 00:35:37,844 --> 00:35:40,325 human race surviving that. And so, 962 00:35:40,965 --> 00:35:41,625 I think 963 00:35:42,164 --> 00:35:44,005 I think we're there's just a lot of 964 00:35:44,005 --> 00:35:45,965 opportunity. I think there's and in fact and 965 00:35:45,965 --> 00:35:47,684 I think there's a lot of opportunity if 966 00:35:47,684 --> 00:35:50,180 you think differently than other people do. If 967 00:35:50,180 --> 00:35:52,579 your background and your training is different than 968 00:35:52,579 --> 00:35:53,559 the mass of people, 969 00:35:54,260 --> 00:35:55,780 or if you if you're just if you're 970 00:35:55,780 --> 00:35:57,380 just not on the same path everybody else 971 00:35:57,380 --> 00:35:58,500 is on, right now is a great time 972 00:35:58,500 --> 00:35:59,940 to be you. I think we have to 973 00:35:59,940 --> 00:36:01,700 think about problems. Now we're gonna be forced 974 00:36:01,700 --> 00:36:03,559 to think about problems from different perspectives. 975 00:36:04,034 --> 00:36:06,434 And because the traditional ones have all been 976 00:36:06,434 --> 00:36:06,934 commoditized 977 00:36:07,234 --> 00:36:08,674 and people who are a little different, I 978 00:36:08,674 --> 00:36:10,054 think, are gonna have an edge. 979 00:36:11,234 --> 00:36:13,655 Hooray for optimism. Yeah. Could be. 980 00:36:16,430 --> 00:36:18,829 Alright. Well, Yvonne, where can people follow you 981 00:36:18,829 --> 00:36:20,190 if they want to or get in touch 982 00:36:20,190 --> 00:36:22,349 with people? I'm on LinkedIn. You can find 983 00:36:22,349 --> 00:36:25,250 me at Yvonne Sharp, and that other podcast 984 00:36:25,309 --> 00:36:27,390 you heard Russ refer to is called the 985 00:36:27,390 --> 00:36:29,970 Cloud Gambit over on the Packet Pusher's network. 986 00:36:30,335 --> 00:36:32,275 I hang out. I look I look 987 00:36:32,574 --> 00:36:35,614 on x Twitter x quite a bit, at 988 00:36:35,614 --> 00:36:36,434 Sharp Network. 989 00:36:37,054 --> 00:36:38,494 I don't post as much as I have 990 00:36:38,494 --> 00:36:40,255 in the past, but you can certainly find 991 00:36:40,255 --> 00:36:41,554 me there. So, yeah, 992 00:36:42,014 --> 00:36:43,900 I'm I'm around it online. Yeah. You can 993 00:36:43,900 --> 00:36:46,719 follow Yvonne on her on her famous podcast. 994 00:36:46,940 --> 00:36:48,300 Yeah. I was gonna say, I thought we 995 00:36:48,300 --> 00:36:50,139 should call it the podcast that shall not 996 00:36:50,139 --> 00:36:50,800 be named. 997 00:36:57,074 --> 00:36:59,074 I have nothing to say. Just kidding. The 998 00:36:59,074 --> 00:36:59,574 podcast 999 00:37:03,795 --> 00:37:05,554 we should tell Ethan to change the name 1000 00:37:05,554 --> 00:37:07,554 of Packet Pushers to the podcast form There 1001 00:37:07,554 --> 00:37:09,494 you go. On his Packet Pushers. Nah. 1002 00:37:10,195 --> 00:37:10,695 Nah. 1003 00:37:11,000 --> 00:37:12,760 There's hey. It's a big tent. There's room 1004 00:37:12,760 --> 00:37:14,219 for all of us. Come on. 1005 00:37:17,880 --> 00:37:19,719 Alright, Tom. Where can people get in touch 1006 00:37:19,719 --> 00:37:21,019 with you or follow you? 1007 00:37:21,880 --> 00:37:22,380 LinkedIn. 1008 00:37:25,175 --> 00:37:28,054 I think Audrey holds the timer. She does. 1009 00:37:28,054 --> 00:37:29,655 That's that's what I think it is. I 1010 00:37:29,655 --> 00:37:32,235 think I I think Audrey has the timer. 1011 00:37:32,534 --> 00:37:34,135 Alright. I'm Russ White. You can find me 1012 00:37:34,135 --> 00:37:35,655 on LinkedIn. You can find me at routing 1013 00:37:35,655 --> 00:37:37,974 geek on x occasionally, not all the time. 1014 00:37:37,974 --> 00:37:39,574 I do post there from time to time, 1015 00:37:39,574 --> 00:37:42,230 but not all the time. And 1016 00:37:42,690 --> 00:37:44,610 you can always find my email address or 1017 00:37:44,610 --> 00:37:46,369 whatever if you wanna get in touch. I'm 1018 00:37:46,369 --> 00:37:48,530 super, super hard to find. Trust me. I'm 1019 00:37:48,530 --> 00:37:51,410 I'm very hard to find. So, anyway, we 1020 00:37:51,410 --> 00:37:53,250 know we live in an attention driven economy. 1021 00:37:53,250 --> 00:37:54,769 We thank you very much for listening all 1022 00:37:54,769 --> 00:37:56,175 the way to the bitter end of this 1023 00:37:56,175 --> 00:37:59,934 episode of the hedge. Hedge 300. There's 300. 1024 00:37:59,934 --> 00:38:01,695 If you have not listened to all 300 1025 00:38:01,695 --> 00:38:02,355 of these, 1026 00:38:03,215 --> 00:38:04,914 this is what you're doing this weekend. 1027 00:38:05,375 --> 00:38:05,875 And 1028 00:38:08,780 --> 00:38:10,699 so thanks for listening, and we will catch 1029 00:38:10,699 --> 00:38:11,599 you next time.