1 00:00:19,560 --> 00:00:23,640 S1: Hey. What's up? So I'm going to try to encapsulate 2 00:00:23,920 --> 00:00:26,000 S1: a whole bunch of stuff that's going on right now 3 00:00:26,600 --> 00:00:30,160 S1: and try to wrap it into a single container. It's 4 00:00:30,200 --> 00:00:32,920 S1: actually very difficult to do because there's so much change, 5 00:00:32,920 --> 00:00:37,120 S1: as everyone knows, and things are just getting crazier, like 6 00:00:37,159 --> 00:00:41,120 S1: every single week. Every single day almost. And I've noticed 7 00:00:41,120 --> 00:00:43,640 S1: like a whole bunch of different transitions happening at the 8 00:00:43,640 --> 00:00:46,480 S1: same time. So I call it the Great Transition. It's 9 00:00:46,480 --> 00:00:49,120 S1: really a whole bunch of smaller transitions, but I think 10 00:00:49,159 --> 00:00:51,480 S1: they kind of have a theme and they have a direction, 11 00:00:51,480 --> 00:00:55,160 S1: and I think I know roughly where they're going. So 12 00:00:55,400 --> 00:00:57,160 S1: what I want to try to give you is something 13 00:00:57,160 --> 00:01:00,240 S1: where if you watch this whole thing and you think 14 00:01:00,240 --> 00:01:03,160 S1: about all these ideas and you just let them stew, 15 00:01:03,680 --> 00:01:07,720 S1: I think the news that comes out over the next weeks, months, 16 00:01:07,720 --> 00:01:09,999 S1: even years is just going to make a lot more 17 00:01:10,000 --> 00:01:13,720 S1: sense because you can kind of put it into this container, 18 00:01:13,720 --> 00:01:17,900 S1: this mental model of thinking about things. So I'm just 19 00:01:17,900 --> 00:01:20,700 S1: going to jump through these and uh, they're going to 20 00:01:20,700 --> 00:01:22,660 S1: jump around quite a bit because there are different topics 21 00:01:22,660 --> 00:01:24,860 S1: we got like personal we got like corporate and stuff 22 00:01:24,860 --> 00:01:28,220 S1: like that. So let's just jump into it. All right. 23 00:01:28,220 --> 00:01:33,140 S1: So the first one is this concept of knowledge going 24 00:01:33,140 --> 00:01:38,780 S1: from private to public okay. This is really, really important. 25 00:01:39,100 --> 00:01:41,179 S1: So there are a few different things that are making 26 00:01:41,180 --> 00:01:45,580 S1: this happen. One is just LMS in general AI in general. 27 00:01:45,580 --> 00:01:49,140 S1: Right now the concept is that it consumes all the 28 00:01:49,140 --> 00:01:51,820 S1: stuff from the internet or whatever, right? All the books, 29 00:01:51,820 --> 00:01:57,620 S1: all the blogs, whatever forum conversations, like all this training 30 00:01:57,620 --> 00:02:02,180 S1: that's been done on these models going back to 2022 actually, 31 00:02:02,180 --> 00:02:06,140 S1: before that. But anyway, all that sort of condenses into 32 00:02:06,140 --> 00:02:08,500 S1: a model. Right. So you have this model that's kind 33 00:02:08,500 --> 00:02:12,740 S1: of representative of like all this knowledge. Right? Everybody knows that. 34 00:02:12,740 --> 00:02:17,230 S1: That's kind of understood What's not so much understood. Is 35 00:02:17,230 --> 00:02:22,150 S1: that what this is actually doing to knowledge work? So 36 00:02:22,190 --> 00:02:26,430 S1: in the past, right. You know, going back ten, 20, 30, 37 00:02:26,470 --> 00:02:31,790 S1: 50 years, if you were an expert in something, you 38 00:02:31,829 --> 00:02:34,709 S1: have knowledge that no one else has, right? You have 39 00:02:34,710 --> 00:02:37,190 S1: knowledge that no one else has. If you are a 40 00:02:37,190 --> 00:02:41,469 S1: specialist consultant at McKinsey, or you are a heart doctor 41 00:02:41,630 --> 00:02:45,310 S1: or whatever you are, you have special knowledge and you 42 00:02:45,310 --> 00:02:48,830 S1: haven't captured even a 10th of it. Let's say you've 43 00:02:48,829 --> 00:02:51,310 S1: written two books. You still haven't captured a 10th of 44 00:02:51,310 --> 00:02:55,790 S1: your knowledge, right? You just know things that other people don't, right? 45 00:02:55,829 --> 00:02:58,230 S1: If you're a security professional who's been doing this for 46 00:02:58,230 --> 00:03:01,910 S1: 20 years, you just understand things. If you're a CISO 47 00:03:01,950 --> 00:03:06,070 S1: that's done this multiple times. You you just understand things 48 00:03:06,070 --> 00:03:10,790 S1: and and get things that nobody else has. And importantly, 49 00:03:10,790 --> 00:03:13,650 S1: it's not in a book somewhere. Okay. Even if you've 50 00:03:13,650 --> 00:03:18,570 S1: written books, your knowledge is still not fully in the books, right? 51 00:03:20,049 --> 00:03:24,810 S1: So that is a powerful thing. It is always protected 52 00:03:24,810 --> 00:03:29,450 S1: smart people. It's always protected the intellectual people who have 53 00:03:29,450 --> 00:03:34,330 S1: all this experience. So it's a combination of smarts plus experience. 54 00:03:34,370 --> 00:03:38,610 S1: And that magical combination has made those people very special. 55 00:03:39,290 --> 00:03:45,010 S1: What is happening now is completely changing that. And this 56 00:03:45,010 --> 00:03:49,330 S1: is one of the major transitions from private knowledge to 57 00:03:49,370 --> 00:03:53,130 S1: public knowledge. What is happening now, especially with skills, this 58 00:03:53,130 --> 00:03:57,490 S1: whole concept that anthropic came up with with skills, it 59 00:03:57,490 --> 00:04:00,530 S1: is scary. We're talking about a folder full of markdown 60 00:04:00,530 --> 00:04:07,770 S1: files that can encapsulate a decent amount of your knowledge, right? 61 00:04:07,810 --> 00:04:09,930 S1: You still have the capture problem where they don't know 62 00:04:09,930 --> 00:04:13,220 S1: exactly what to say, How to capture it. But, but 63 00:04:13,220 --> 00:04:18,380 S1: here's the situation. Many, many smart people are producing skills, 64 00:04:18,700 --> 00:04:22,740 S1: and many, many other smart people are going to collect 65 00:04:23,100 --> 00:04:27,180 S1: knowledge and specialized knowledge from all over the internet, anywhere 66 00:04:27,180 --> 00:04:31,460 S1: it's been written down, and bring that into a skill. Plus, 67 00:04:31,860 --> 00:04:35,660 S1: all these specialist people, they're writing books, right? They have 68 00:04:35,660 --> 00:04:39,540 S1: been writing books. They've been doing presentations, they've been writing blogs, 69 00:04:39,700 --> 00:04:43,940 S1: they've been doing interviews. They've been doing podcasts. Well, in 70 00:04:43,940 --> 00:04:47,060 S1: the past, we'd never had a system that could basically say, well, 71 00:04:47,100 --> 00:04:50,940 S1: go get all of that, right? Go get everything. So 72 00:04:50,980 --> 00:04:55,820 S1: and so Doctor Huberman has ever said about health or 73 00:04:55,980 --> 00:04:59,460 S1: morning routines or whatever, bring it all together and turn 74 00:04:59,460 --> 00:05:02,340 S1: that into a skill. I mean, this is one prompt. 75 00:05:02,500 --> 00:05:06,500 S1: This is one prompt, you know, find everything Huberman has 76 00:05:06,500 --> 00:05:11,080 S1: said about morning routines from every podcast he's ever done, 77 00:05:11,080 --> 00:05:14,560 S1: every blog he's ever done, whatever he's ever put out, 78 00:05:14,560 --> 00:05:18,480 S1: every article, every interview and put that into a skill. 79 00:05:19,120 --> 00:05:23,520 S1: That new thing combined with the models just getting better 80 00:05:23,520 --> 00:05:26,160 S1: and they're absorbing it and everything. I mean, that's a 81 00:05:26,160 --> 00:05:29,320 S1: crazy thing. This just feeds on itself, right? The model 82 00:05:29,720 --> 00:05:32,800 S1: then can consume all those skills, right. Which of course 83 00:05:32,800 --> 00:05:36,440 S1: is going to happen. So ultimately you have this transition 84 00:05:37,280 --> 00:05:39,800 S1: that is really accelerated. But it's just going to continue 85 00:05:39,800 --> 00:05:44,000 S1: to accelerate. And what it means is the gap between 86 00:05:44,320 --> 00:05:50,120 S1: special privatized knowledge that's inside of someone's mind, some specialist doctor, 87 00:05:50,360 --> 00:05:54,720 S1: some specialist psychiatrist who's been doing this work for 40 88 00:05:54,720 --> 00:05:59,400 S1: years or 50 years. The delta between what they know 89 00:05:59,400 --> 00:06:03,560 S1: and no one else knows is getting smaller. And that 90 00:06:03,600 --> 00:06:10,940 S1: that is massively, massively impactful for like, humanity in general. Okay. 91 00:06:11,380 --> 00:06:14,580 S1: Then there's another sort of layer on this which is 92 00:06:14,620 --> 00:06:17,300 S1: which is crazy. So let's just assume all that is 93 00:06:17,300 --> 00:06:20,900 S1: happening and all of that is being consumed by these 94 00:06:20,900 --> 00:06:23,700 S1: labs who are spending billions of dollars bringing that knowledge 95 00:06:23,700 --> 00:06:29,220 S1: into the models. Fine. Okay. That's happening. But what we 96 00:06:29,220 --> 00:06:32,700 S1: just saw from anthropic, and this is happening all over 97 00:06:32,700 --> 00:06:35,660 S1: the place, anthropic just called it out because a bunch 98 00:06:35,660 --> 00:06:39,020 S1: of these Chinese labs are doing it in mass, like, 99 00:06:39,020 --> 00:06:42,580 S1: very organized. I'm so surprised. Like, everyone's, like, surprised by this. 100 00:06:42,580 --> 00:06:44,900 S1: And they're like, hey, what's the big deal? You know, 101 00:06:44,940 --> 00:06:47,540 S1: maybe they're lying or whatever. Why would they be lying? 102 00:06:47,580 --> 00:06:50,540 S1: China is known for doing this. China is, like, famous 103 00:06:50,540 --> 00:06:53,620 S1: for stealing things. Like, they're famous for a lot of 104 00:06:53,620 --> 00:06:58,380 S1: good things, too, right? But they are famous for stealing ideas, 105 00:06:58,380 --> 00:07:04,220 S1: stealing content. And, um, they're also massively going all in 106 00:07:04,420 --> 00:07:08,950 S1: on open source models. So I believe that they have 107 00:07:08,950 --> 00:07:13,150 S1: a very clear strategy. It is you don't have to 108 00:07:13,190 --> 00:07:16,550 S1: compete to be a pinnacle lab, right? They don't have 109 00:07:16,910 --> 00:07:21,310 S1: an anthropic. They don't have a Google DeepMind. They don't 110 00:07:21,310 --> 00:07:25,270 S1: have an OpenAI, but they do have deep seq. And 111 00:07:25,270 --> 00:07:29,590 S1: deep seq has been, you know, called out for doing 112 00:07:29,590 --> 00:07:34,230 S1: this for a very long time. They are capturing the 113 00:07:34,270 --> 00:07:37,790 S1: knowledge of all the billions of dollars of work and 114 00:07:37,790 --> 00:07:41,070 S1: bringing it into open source. And then what they are 115 00:07:41,070 --> 00:07:44,990 S1: doing as a Chinese strategy for AI, I believe, is 116 00:07:45,030 --> 00:07:49,270 S1: they're just saying distill it, not distill it. Well, they 117 00:07:49,270 --> 00:07:54,110 S1: are distilling, but release it, diffuse it, absorb it into 118 00:07:54,110 --> 00:07:57,350 S1: the pool. It's like you've heard the the metaphor. Peeing 119 00:07:57,350 --> 00:08:02,030 S1: in the pool. That's that's what happens. Our specialized knowledge 120 00:08:02,270 --> 00:08:05,650 S1: of what Specialized humans could do that no one else 121 00:08:05,650 --> 00:08:10,290 S1: could do that is the P that's going into the pool. 122 00:08:10,530 --> 00:08:12,690 S1: You can't pull it out, can't pull it out. It's 123 00:08:12,690 --> 00:08:15,330 S1: just going to be in there. And China is basically 124 00:08:15,330 --> 00:08:19,810 S1: making this happen at a mass scale because yes, the 125 00:08:20,210 --> 00:08:23,490 S1: large billion dollar labs will have it first. It'll go 126 00:08:23,490 --> 00:08:27,650 S1: into those models first actually. First it'll go into skills. Right. First. 127 00:08:27,650 --> 00:08:31,730 S1: It's going into this whole knowledge sharing concept of skills. 128 00:08:31,730 --> 00:08:33,890 S1: So it's in the markdown file. So it's on GitHub 129 00:08:33,890 --> 00:08:36,969 S1: and stuff like that. So that takes a little more effort. 130 00:08:37,290 --> 00:08:39,450 S1: But then it goes up into these labs. It gets 131 00:08:39,450 --> 00:08:43,929 S1: consumed however long that takes a period of months or whatever. 132 00:08:44,250 --> 00:08:48,290 S1: But then the open source models are pulling from it. 133 00:08:49,290 --> 00:08:52,490 S1: And the other thing it's pulling is not just the data, 134 00:08:53,170 --> 00:08:58,650 S1: the techniques that make those Premiere Labs better and better, 135 00:08:59,370 --> 00:09:02,780 S1: or have an advantage over another lab, or have an 136 00:09:02,780 --> 00:09:07,700 S1: advantage over open source. Those techniques are also being diffused 137 00:09:07,700 --> 00:09:12,860 S1: somehow when the major labs have a major advantage and 138 00:09:12,860 --> 00:09:17,700 S1: they jump ahead somehow. The Chinese models seem to get it. 139 00:09:17,740 --> 00:09:21,140 S1: A few months later. So all of these things are 140 00:09:21,180 --> 00:09:25,060 S1: actually contributing to the same thing. The specialized knowledge is 141 00:09:25,060 --> 00:09:31,020 S1: being diffused into public domain. That's just a transition that's happening. Okay. 142 00:09:31,060 --> 00:09:34,340 S1: So that's the first one. Knowledge is going from private 143 00:09:34,340 --> 00:09:38,460 S1: to public. Okay. The second one is products are going 144 00:09:38,460 --> 00:09:44,059 S1: from standalone software to APIs. So I talked about this 145 00:09:44,100 --> 00:09:48,260 S1: in my stupid little book from 2016. Basically said that 146 00:09:48,260 --> 00:09:52,980 S1: businesses become APIs. And we're finally now starting to see this. 147 00:09:53,660 --> 00:09:58,100 S1: So if you've seen like all these people releasing tools, right. Um, 148 00:09:58,140 --> 00:10:00,760 S1: they're releasing, you know, so and so model for this 149 00:10:00,760 --> 00:10:03,440 S1: so and so model for that. Or not just models 150 00:10:03,440 --> 00:10:07,679 S1: but like so and so functionality. Um, like uh there's 151 00:10:07,680 --> 00:10:10,560 S1: a company that does remove background. Oh I'll give you 152 00:10:10,560 --> 00:10:15,959 S1: a great example. Um, Excalibur. Excalibur just came out with 153 00:10:16,440 --> 00:10:19,760 S1: a new piece of functionality where you could just describe 154 00:10:19,760 --> 00:10:23,840 S1: what you want to make, and it will build all 155 00:10:23,880 --> 00:10:27,480 S1: the different, uh, objects for you in your favorite fonts 156 00:10:27,480 --> 00:10:31,199 S1: and your favorite aesthetic. Like, it'll just build you diagrams 157 00:10:31,840 --> 00:10:36,640 S1: like the perfect, like, really cool looking Excalibur diagrams. And 158 00:10:36,640 --> 00:10:40,280 S1: my first question when I saw this was, hold on, 159 00:10:40,280 --> 00:10:43,360 S1: because I went and looked at the documentation and it 160 00:10:43,360 --> 00:10:47,360 S1: basically said, yeah, you just go into the interface and 161 00:10:47,360 --> 00:10:52,120 S1: you type into Excalibur, draw what you want to say. 162 00:10:52,640 --> 00:10:56,680 S1: And I'm like, what? What are you talking about? Like, 163 00:10:57,040 --> 00:11:01,540 S1: do you, do you honestly think in like early 2026, 164 00:11:01,780 --> 00:11:05,500 S1: I'm going to open up a drawer and type in 165 00:11:05,500 --> 00:11:08,660 S1: a prompt? Are you kidding me? So I posted, I 166 00:11:08,660 --> 00:11:12,180 S1: was like, hey guys, this looks amazing. This looks amazing, 167 00:11:12,220 --> 00:11:15,740 S1: looks fantastic. There's no way I'm going to use it. 168 00:11:15,860 --> 00:11:19,500 S1: Can you make this available as an MCP? Can you 169 00:11:19,500 --> 00:11:23,620 S1: make this available as an API? I'm not going to 170 00:11:23,620 --> 00:11:26,740 S1: do any of this ever. Like if I have to 171 00:11:26,780 --> 00:11:30,540 S1: open an app, I have already lost, right? I've already lost. 172 00:11:30,580 --> 00:11:35,140 S1: Like this is not a good thing. It means my 173 00:11:35,180 --> 00:11:38,380 S1: tooling is horribly broken. My AI should be doing all 174 00:11:38,420 --> 00:11:41,780 S1: of this for me, so I'm not sure if they 175 00:11:41,780 --> 00:11:45,460 S1: made that adjustment. But when I posted that on Twitter, 176 00:11:45,500 --> 00:11:48,459 S1: like a whole bunch of people showed up and they're like, yeah, 100%, 177 00:11:48,460 --> 00:11:50,860 S1: I need, you know, I need an MCP for this. 178 00:11:50,860 --> 00:11:54,900 S1: Otherwise it's not useful. That is the way everything is going. 179 00:11:54,940 --> 00:11:57,550 S1: If you notice most of the releases coming out for 180 00:11:57,550 --> 00:12:00,950 S1: products now, they're like, here's the MCP for it. Here's 181 00:12:00,950 --> 00:12:03,950 S1: how your agents can do this automatically, right? This is 182 00:12:03,950 --> 00:12:07,190 S1: just becoming the new way to release software. And this 183 00:12:07,190 --> 00:12:10,830 S1: is heading in the exact direction that I put in 184 00:12:10,870 --> 00:12:16,270 S1: that stupid book in 2016. Businesses become APIs. Now why 185 00:12:16,270 --> 00:12:22,350 S1: is this important? It's important because the consumer is not 186 00:12:22,350 --> 00:12:26,150 S1: so much making the choice anymore. The consumer is not 187 00:12:26,150 --> 00:12:33,030 S1: going to be like, hmm, yeah, there's 47,938 different options 188 00:12:33,030 --> 00:12:37,630 S1: for removing backgrounds from images. Let me pull up GitHub 189 00:12:37,910 --> 00:12:45,310 S1: and Google and let me spend 2.5 hours sampling and 190 00:12:45,310 --> 00:12:49,630 S1: trying different ones. No, no, no, that there's too many apps. 191 00:12:49,790 --> 00:12:52,870 S1: And because of AI, there's too many apps being made 192 00:12:52,910 --> 00:12:55,290 S1: right on top of that. So So there will be 193 00:12:55,330 --> 00:12:57,490 S1: hundreds more of these things coming out all the time. 194 00:12:58,490 --> 00:13:02,250 S1: The only way this resolves, which is what I was 195 00:13:02,250 --> 00:13:08,210 S1: talking about in that text, was there are directories. Okay. 196 00:13:08,250 --> 00:13:11,689 S1: If you have a background remover tool, by the way, 197 00:13:11,730 --> 00:13:15,090 S1: my favorite is remove BG, I think is the one 198 00:13:15,090 --> 00:13:18,170 S1: I use. Um, and they do have an MCP or 199 00:13:18,170 --> 00:13:21,450 S1: they have an API at least. And that's what my 200 00:13:21,770 --> 00:13:25,570 S1: system Cai actually calls. So if you do have one 201 00:13:25,570 --> 00:13:31,170 S1: of these agents, um, there will basically be orchestration layers, 202 00:13:31,210 --> 00:13:37,730 S1: directories of services labeled and categorized, you know, taxonomy, folksonomy, whatever. 203 00:13:37,890 --> 00:13:43,970 S1: Basically saying if you want to, um, remove backgrounds from images, 204 00:13:44,730 --> 00:13:50,050 S1: here's your list of 27,000. But they will be rated right. 205 00:13:50,330 --> 00:13:53,780 S1: You'll have different services with different ratings. And my system 206 00:13:53,780 --> 00:14:00,180 S1: Chi will know which services it prefers, right? He prefers 207 00:14:00,180 --> 00:14:02,700 S1: to go to this rating service or whatever, which is 208 00:14:02,700 --> 00:14:06,540 S1: like IMDb or Rotten Tomatoes, right? But it's for software 209 00:14:06,940 --> 00:14:09,660 S1: and it says, okay, find me the highest rated one 210 00:14:09,660 --> 00:14:13,420 S1: with the most ratings, the least negative ratings, whatever algorithm 211 00:14:13,460 --> 00:14:16,820 S1: Chi wants to use. And he's going to select okay, 212 00:14:16,860 --> 00:14:22,260 S1: it's remove background boom, pulls that in, drops it into 213 00:14:22,260 --> 00:14:26,020 S1: our workflow inside of our skill boom. That's it. From 214 00:14:26,020 --> 00:14:28,820 S1: now on, when I want to remove or when Chi 215 00:14:28,820 --> 00:14:31,740 S1: wants to remove a background for me for an image, 216 00:14:32,100 --> 00:14:38,620 S1: that's what he uses. Okay, where is the website? Where's 217 00:14:38,620 --> 00:14:44,140 S1: the website for remove background? Who needs a website? This 218 00:14:44,140 --> 00:14:48,580 S1: is a service. A directory service like the old days 219 00:14:48,580 --> 00:14:52,520 S1: like Yahoo directories or whatever. Right? This is a directory 220 00:14:52,520 --> 00:14:56,200 S1: service of the best thing, right? It's already been rated. 221 00:14:56,920 --> 00:14:59,600 S1: My agents are going to go check those ratings and 222 00:14:59,600 --> 00:15:01,120 S1: it's going to find the API. And it's going to 223 00:15:01,120 --> 00:15:07,200 S1: integrate it. This old way of making the software, packaging it. Oh, 224 00:15:07,280 --> 00:15:09,160 S1: it's got to have a nice UI. It's got to 225 00:15:09,160 --> 00:15:11,800 S1: have a nice website. You know, when the person comes 226 00:15:11,800 --> 00:15:14,240 S1: to the website, they've got to really like it. Then 227 00:15:14,240 --> 00:15:16,840 S1: they go click the buy button. Then they do all this. 228 00:15:17,240 --> 00:15:20,160 S1: It's all going away. Right. And this is tightly coupled 229 00:15:20,160 --> 00:15:23,040 S1: with another related one I have here in the list 230 00:15:23,040 --> 00:15:28,440 S1: which is interface. Interface is going away SEO is going 231 00:15:28,480 --> 00:15:32,800 S1: away okay. So interface used to be for humans. We 232 00:15:32,800 --> 00:15:37,200 S1: make software. We have services whatever software service, whatever it 233 00:15:37,200 --> 00:15:39,840 S1: is we have to have an interface for that. Right. 234 00:15:40,840 --> 00:15:43,160 S1: It's the interface you use actually day to day to 235 00:15:43,200 --> 00:15:45,160 S1: interact with it. But you also have to have an 236 00:15:45,160 --> 00:15:47,920 S1: interface for the marketing and describing how to use it 237 00:15:47,920 --> 00:15:53,610 S1: and the documentation interfaces in general. Front end in general 238 00:15:53,930 --> 00:15:58,210 S1: is going away. It's not that the content won't be there, 239 00:15:58,210 --> 00:16:01,610 S1: it's that it will be designed to be consumed by agents. 240 00:16:02,090 --> 00:16:04,690 S1: Write your agents are the ones who will be the 241 00:16:04,690 --> 00:16:09,890 S1: main consumers. AI will be the main consumers. And this 242 00:16:09,890 --> 00:16:14,330 S1: also relates to another one, which is everyone has their 243 00:16:14,330 --> 00:16:17,890 S1: own digital assistant. Right? This this was the number one 244 00:16:18,410 --> 00:16:21,890 S1: sort of prediction. And call out, uh, in uh, the 245 00:16:21,930 --> 00:16:27,450 S1: real Internet of Things in 2016. Everyone gets a digital assistant. 246 00:16:27,730 --> 00:16:33,850 S1: Everything gets an API. Most importantly, businesses, people eventually as well. 247 00:16:33,850 --> 00:16:38,370 S1: But objects, businesses, services, they all have an API. And 248 00:16:38,370 --> 00:16:42,170 S1: then our AI, because it knows us so well. When 249 00:16:42,170 --> 00:16:44,490 S1: we make a request, it goes and gets the thing 250 00:16:44,490 --> 00:16:47,250 S1: from the API, brings it back to us. And then 251 00:16:47,250 --> 00:16:51,510 S1: the third piece is interface. When we want an interface, 252 00:16:51,510 --> 00:16:54,350 S1: when we want to look at I'm buying shoes, I 253 00:16:54,350 --> 00:16:57,550 S1: want to see what they look like. Um, I'm buying 254 00:16:57,550 --> 00:16:59,270 S1: a house. I want to see what it looks like. 255 00:16:59,910 --> 00:17:03,710 S1: Our AI will be presenting the interface to us. Okay, 256 00:17:03,990 --> 00:17:07,910 S1: this is already starting to happen. People are building bespoke software. 257 00:17:08,190 --> 00:17:12,629 S1: Custom software is the direction that it's going, right? So 258 00:17:12,629 --> 00:17:16,109 S1: software goes from being everyone has the same packages to 259 00:17:16,149 --> 00:17:21,869 S1: everyone gets bespoke software, everyone has custom software. Interfaces go 260 00:17:21,869 --> 00:17:28,030 S1: from being tied to the application itself. It's inherently part 261 00:17:28,030 --> 00:17:32,310 S1: of the application to all of that being separated. The 262 00:17:32,350 --> 00:17:34,629 S1: core part of your business, the core part of your 263 00:17:34,629 --> 00:17:38,069 S1: product is the API, which will be used by the 264 00:17:38,070 --> 00:17:42,869 S1: agents of the user, of the consumer, and the interface 265 00:17:43,109 --> 00:17:48,890 S1: will be between their agent that user's agent and them. 266 00:17:49,369 --> 00:17:52,330 S1: That's the interface. Right. And this will take a little 267 00:17:52,330 --> 00:17:54,730 S1: more time. But it's already starting to happen a little 268 00:17:54,730 --> 00:17:59,889 S1: bit with everyone having custom software and custom interfaces. So 269 00:17:59,929 --> 00:18:02,649 S1: that that's a number of sort of things all wrapped 270 00:18:02,649 --> 00:18:13,050 S1: into one software goes from standardized and consolidated and integrated 271 00:18:13,929 --> 00:18:21,889 S1: to being separated between its interface and its functionality. And 272 00:18:21,889 --> 00:18:25,169 S1: the user of most software becomes AI as opposed to 273 00:18:25,209 --> 00:18:29,169 S1: the user itself. Right. So those are kind of the 274 00:18:29,169 --> 00:18:35,049 S1: major transitions there. Uh, so SEO again, SEO used to 275 00:18:35,050 --> 00:18:39,050 S1: be about being attractive to the user. SEO goes from 276 00:18:39,050 --> 00:18:42,210 S1: trying to attract the user to trying to attract the 277 00:18:42,209 --> 00:18:45,740 S1: user's AI I again. So when I say, hey, I 278 00:18:45,740 --> 00:18:48,820 S1: need a new bed mattress, I'm not saying that to 279 00:18:48,859 --> 00:18:51,260 S1: the internet. I'm not saying that to Google. I'm not 280 00:18:51,260 --> 00:18:55,699 S1: saying that to the World Wide Web, a browser. No, 281 00:18:55,899 --> 00:18:58,180 S1: I'm saying that to my agent. I am saying that 282 00:18:58,179 --> 00:19:02,659 S1: to my Da. I need a new mattress. Well guess what? 283 00:19:02,699 --> 00:19:06,260 S1: My Da knows my sleeping habits, knows my routine, knows 284 00:19:06,260 --> 00:19:09,340 S1: if I like a firm mattress or a soft one. Knows. 285 00:19:09,659 --> 00:19:12,619 S1: You know my girls. She likes a softer bed. I 286 00:19:12,619 --> 00:19:15,460 S1: like a harder one on my side. So it's like 287 00:19:16,179 --> 00:19:18,979 S1: this is where the customization comes in, right? Your agent 288 00:19:18,980 --> 00:19:22,460 S1: knows you, therefore it can do smarter queries for you. 289 00:19:22,459 --> 00:19:25,580 S1: But the point is, it's the one doing the queries. 290 00:19:25,580 --> 00:19:28,899 S1: It's the one who, if it's going to be tricked 291 00:19:29,379 --> 00:19:33,780 S1: into picking one mattress versus another, the tricking needs to 292 00:19:33,780 --> 00:19:37,340 S1: happen at the AI layer, because I'm just going to 293 00:19:37,340 --> 00:19:40,660 S1: do whatever my agent tells me, right? The agent's going 294 00:19:40,659 --> 00:19:42,520 S1: to be like, yeah, I found the best one. It's 295 00:19:42,520 --> 00:19:45,439 S1: not even a question. You know, it's this much. Do 296 00:19:45,439 --> 00:19:47,119 S1: you want me to get it for you? You know, 297 00:19:47,159 --> 00:19:51,679 S1: I'll have it here tomorrow. And that's the end of that. Okay, 298 00:19:51,719 --> 00:19:54,639 S1: so that was a few things that were based around 299 00:19:54,639 --> 00:19:57,159 S1: the consumer. Now I want to bring it a little 300 00:19:57,199 --> 00:20:00,480 S1: bit over to the enterprise side. All right. So all 301 00:20:00,480 --> 00:20:02,959 S1: of that is sort of on the consumer side. Let's 302 00:20:02,959 --> 00:20:05,840 S1: switch over to talk a little bit about the enterprise side. 303 00:20:07,240 --> 00:20:10,800 S1: So much the same sort of stuff that we're talking 304 00:20:10,800 --> 00:20:12,720 S1: about on the consumer side is also going to be 305 00:20:12,760 --> 00:20:17,600 S1: happening on the enterprise side. The changes that are happening 306 00:20:17,600 --> 00:20:21,479 S1: on the enterprise side, they're so massive, like absolutely massive. 307 00:20:21,800 --> 00:20:24,119 S1: One of the big things that's going to be happening 308 00:20:24,119 --> 00:20:31,320 S1: on the enterprise side is the transition from humans creating 309 00:20:31,320 --> 00:20:36,639 S1: processes and sort of following them to AI, kind of 310 00:20:36,679 --> 00:20:41,930 S1: running the business based on SOPs And basically building out 311 00:20:41,929 --> 00:20:45,330 S1: a lattice structure, like a graph structure of all the 312 00:20:45,369 --> 00:20:47,889 S1: work that needs to be done. I did a post, 313 00:20:47,889 --> 00:20:53,250 S1: I think, in 2024, maybe talking about companies or just 314 00:20:53,290 --> 00:20:59,330 S1: a graph of APIs, a graph of operations. And I 315 00:20:59,330 --> 00:21:02,409 S1: was talking about like somebody who just does a task, right. 316 00:21:02,449 --> 00:21:06,769 S1: Let's say it's a threat intelligence task or no, let's 317 00:21:06,770 --> 00:21:10,049 S1: take the insurance one. Right. You have to look at 318 00:21:10,050 --> 00:21:13,090 S1: the photos. You have to look at their account. You know, 319 00:21:13,129 --> 00:21:15,490 S1: they're making a claim. For example, you know, I need 320 00:21:15,490 --> 00:21:18,210 S1: to be paid for this accident I just got into. 321 00:21:18,570 --> 00:21:21,889 S1: And you need to filter for fraud, right? So I'm Sarah, 322 00:21:22,169 --> 00:21:25,490 S1: I'm looking at these things. I'm trying to figure out if, uh, 323 00:21:26,129 --> 00:21:28,409 S1: this is fraud. I'm looking at the picture. Does it 324 00:21:28,409 --> 00:21:31,369 S1: look real? Do they have a real account? Are they 325 00:21:31,369 --> 00:21:34,929 S1: making lots of claims just recently? Like, does their account 326 00:21:34,929 --> 00:21:39,909 S1: look compromised? All these different things. And if it looks legit, okay, 327 00:21:39,949 --> 00:21:42,790 S1: here's how much we're going to pay you out. You know, 328 00:21:42,830 --> 00:21:45,109 S1: some of that's automated, but you still have people that 329 00:21:45,109 --> 00:21:48,189 S1: are their job. Their actual physical job is to do 330 00:21:48,189 --> 00:21:53,070 S1: this task right. So that's the type of thing that 331 00:21:53,550 --> 00:21:56,549 S1: currently in the enterprise, if you look at any major company, 332 00:21:56,550 --> 00:21:59,230 S1: there is not a map, there is not a graph 333 00:21:59,350 --> 00:22:03,510 S1: that basically the CEO can look down and say, this 334 00:22:03,510 --> 00:22:07,629 S1: is my entire business. This is every task happening in 335 00:22:07,629 --> 00:22:12,510 S1: my company and the process of how it's done. The 336 00:22:12,510 --> 00:22:16,990 S1: SOP right, also a process kind of the same word. 337 00:22:17,070 --> 00:22:20,749 S1: Uh unmilitary. So that's the way I say it. There's 338 00:22:20,750 --> 00:22:25,429 S1: an SOP standard operating procedure for how this thing is done. Right. 339 00:22:26,230 --> 00:22:28,629 S1: And here are the people that do it. Here are 340 00:22:28,629 --> 00:22:34,669 S1: the workflows. Here are the multiple steps involved. AI is 341 00:22:34,669 --> 00:22:38,009 S1: going to have this for every company. This is the 342 00:22:38,050 --> 00:22:40,850 S1: major transition that's going to be happening. This is just 343 00:22:40,850 --> 00:22:43,969 S1: now starting. This is very slow. This is much slower 344 00:22:43,969 --> 00:22:46,809 S1: than all this consumer adoption that we've seen over the 345 00:22:46,810 --> 00:22:49,210 S1: last couple of years. This is going to take time. 346 00:22:49,889 --> 00:22:52,489 S1: And this is what companies are still figuring out. They're like, 347 00:22:52,490 --> 00:22:54,810 S1: what exactly am I going to do with AI? Well, 348 00:22:54,810 --> 00:22:56,889 S1: this is what they're going to do with AI. And 349 00:22:56,889 --> 00:22:59,889 S1: I'm sure a million different companies, this is a thing 350 00:22:59,889 --> 00:23:04,489 S1: I do for companies. But people like McKinsey, lots of 351 00:23:04,490 --> 00:23:07,330 S1: different companies like this. They're already bringing in, you know, 352 00:23:07,369 --> 00:23:12,250 S1: an army of 22 year old smiling kids to interview 353 00:23:12,250 --> 00:23:15,850 S1: everyone and produce this map. Here is the work that 354 00:23:15,850 --> 00:23:20,330 S1: takes place. Here's where decisions happen. Here's where tools happen. 355 00:23:20,330 --> 00:23:25,170 S1: Here's where all these things take place, right? So bringing 356 00:23:25,169 --> 00:23:28,169 S1: you back to the consumer side, when you have a 357 00:23:28,169 --> 00:23:33,729 S1: map like that, the whole conversation around software changes once 358 00:23:33,730 --> 00:23:39,540 S1: again before you would have people. The people are basically 359 00:23:39,540 --> 00:23:43,260 S1: the company. The people are doing the work. Yes, they 360 00:23:43,260 --> 00:23:46,420 S1: have documents. Yes, they have processes. But it's the people 361 00:23:46,419 --> 00:23:49,980 S1: doing the work. They're supposed to follow this policy, but 362 00:23:49,980 --> 00:23:53,459 S1: it's just a doc. It's just a word doc used 363 00:23:53,459 --> 00:23:56,899 S1: to be. Right. Um, now it's like a Google doc 364 00:23:56,899 --> 00:24:02,139 S1: or whatever, but do they follow it? Not really. Sometimes. 365 00:24:02,580 --> 00:24:08,739 S1: Often it depends on the company, but it's it's not required. Right? 366 00:24:08,780 --> 00:24:12,379 S1: A company can have people who mostly do the work. 367 00:24:12,540 --> 00:24:16,100 S1: They're barely referring to the policy. By the way, there 368 00:24:16,100 --> 00:24:18,659 S1: might be 3 or 4 of those docs and the 369 00:24:18,659 --> 00:24:21,300 S1: main person who maintain them. You know, she went on 370 00:24:21,300 --> 00:24:25,580 S1: maternity leave. She never came back. Right. Whatever it is like, oh, 371 00:24:25,619 --> 00:24:28,700 S1: that was Scott. Scott left the company, you know, he retired. 372 00:24:29,179 --> 00:24:32,899 S1: He's not he's not coming back. So those docs get old. 373 00:24:32,959 --> 00:24:38,240 S1: Nobody's following the policy. That is completely different than an 374 00:24:38,240 --> 00:24:43,560 S1: AI saying, okay, I now own all these SOPs. Here 375 00:24:43,560 --> 00:24:47,559 S1: is the map of all work that's being done. And 376 00:24:47,560 --> 00:24:50,079 S1: the humans humans are still there. There's going to be 377 00:24:50,080 --> 00:24:54,200 S1: some humans left in the company. Humans are the ones 378 00:24:54,199 --> 00:24:58,639 S1: responsible for improving this AI, for telling the AI, hey, look, 379 00:24:58,679 --> 00:25:01,680 S1: we need to change this SOP. Like, that's not how 380 00:25:01,679 --> 00:25:04,320 S1: this should be done anymore. And you have a conversation 381 00:25:04,320 --> 00:25:06,280 S1: with AI or whatever. And the AI is like, okay, 382 00:25:06,320 --> 00:25:08,479 S1: are you sure about this? Boom. It makes the change. 383 00:25:08,600 --> 00:25:11,600 S1: All the documentation is updated, all the SOPs are updated, 384 00:25:11,600 --> 00:25:15,480 S1: the cross references are updated. That is the new model 385 00:25:15,480 --> 00:25:19,679 S1: for for business, right? It's not started yet. It's barely started. 386 00:25:20,159 --> 00:25:23,119 S1: Massively different for different companies, right? Some people have been 387 00:25:23,119 --> 00:25:25,840 S1: doing this for a year or two already. People like 388 00:25:25,840 --> 00:25:32,290 S1: Google probably, but most companies, they're barely even realizing that 389 00:25:32,290 --> 00:25:35,010 S1: this is a thing right now. They're just like, oh, 390 00:25:35,050 --> 00:25:37,330 S1: AI is a tool. Everyone's telling me I have to 391 00:25:37,330 --> 00:25:39,969 S1: use it. Tell all your employees you must use AI. 392 00:25:40,010 --> 00:25:42,810 S1: What does that actually mean? This is the thing that's 393 00:25:42,850 --> 00:25:48,409 S1: actually going to happen. Sop's are established, goals are established. 394 00:25:48,649 --> 00:25:50,489 S1: What are we actually trying to do as a company? 395 00:25:50,490 --> 00:25:52,210 S1: We're going to talk about that one later. This is 396 00:25:52,250 --> 00:25:57,450 S1: a massively important one. But all of that gets canonicalized captured, 397 00:25:57,490 --> 00:26:02,249 S1: turned into SOPs, turned into goals, turned into texts, turned 398 00:26:02,250 --> 00:26:05,490 S1: into actual things you could look at. But the important 399 00:26:05,490 --> 00:26:08,969 S1: part for this particular point is the the map, this 400 00:26:08,969 --> 00:26:13,730 S1: graph of all the different work, they're okay. Now a 401 00:26:13,730 --> 00:26:16,969 S1: software vendor comes in and they're like, hey, look, we got, 402 00:26:17,010 --> 00:26:21,770 S1: you know, the best. We'll just say this company needs background, uh, 403 00:26:21,770 --> 00:26:26,169 S1: image removal or whatever. We have the best one out there. 404 00:26:26,689 --> 00:26:30,310 S1: It's the best. Now, in the past, what would happen 405 00:26:30,310 --> 00:26:33,869 S1: is they would bring their salesperson, take someone out to 406 00:26:33,909 --> 00:26:36,830 S1: a steak dinner and be like, yeah, yeah, it's just 407 00:26:36,830 --> 00:26:39,669 S1: it does background removal way better than anything else. Like 408 00:26:39,669 --> 00:26:42,270 S1: you just have to have this. And if they convince 409 00:26:42,270 --> 00:26:45,909 S1: this person and the manager and, you know, the purchasing 410 00:26:45,909 --> 00:26:51,109 S1: authority or whatever procurement, whatever, they buy it, once again, 411 00:26:51,350 --> 00:26:55,749 S1: it's a human doing the work. They have this image tool. Okay, 412 00:26:55,790 --> 00:26:58,470 S1: maybe there's an API call if there's some automation, if 413 00:26:58,469 --> 00:27:04,270 S1: it's relatively new, but ultimately it's humans buying software. Software 414 00:27:04,270 --> 00:27:07,230 S1: is a package. It's sitting on a shelf and they've 415 00:27:07,230 --> 00:27:09,909 S1: purchased a bunch of software. Maybe they use it, maybe 416 00:27:09,909 --> 00:27:15,550 S1: they don't. In this new model, this lattice, this graph system, 417 00:27:16,030 --> 00:27:20,030 S1: it includes all the tools. The work is laid out 418 00:27:20,030 --> 00:27:24,229 S1: as a grid, as a graph. Here's here's the stuff. 419 00:27:24,429 --> 00:27:27,030 S1: So watch this. This is a completely different conversation you 420 00:27:27,030 --> 00:27:30,490 S1: have with the software vendor. Um, okay. Cool. Here is 421 00:27:30,490 --> 00:27:34,489 S1: my map. Here's all the processes, all the work that 422 00:27:34,490 --> 00:27:39,090 S1: I do. Every single workflow being done. What are you replacing? 423 00:27:39,850 --> 00:27:43,330 S1: What are you doing better. Click on this note. Boom. 424 00:27:43,330 --> 00:27:47,609 S1: You click on the note. Background image removal. Let's look 425 00:27:47,609 --> 00:27:50,250 S1: at the metrics for it. Here's how fast it is. 426 00:27:50,250 --> 00:27:53,770 S1: Here's how cheap it is. Here's how many times it 427 00:27:53,770 --> 00:27:57,490 S1: failed here. Here's how many times it succeeded. Here's the 428 00:27:57,490 --> 00:28:02,529 S1: ratings for how that's been done. Now what are yours? 429 00:28:03,649 --> 00:28:06,809 S1: Now what are yours. And they have they're going to 430 00:28:06,810 --> 00:28:10,330 S1: have to produce metrics that say I can do that 431 00:28:10,330 --> 00:28:14,250 S1: function better. So now it's not about a software package 432 00:28:15,090 --> 00:28:17,529 S1: that some human is buying. And maybe they use it. 433 00:28:17,530 --> 00:28:21,529 S1: Maybe they don't. Now it's about an AI saying, here 434 00:28:21,530 --> 00:28:26,139 S1: are my metrics for this function which fits into this graph. 435 00:28:26,980 --> 00:28:31,659 S1: Can you prove that your metrics for doing this function 436 00:28:32,139 --> 00:28:36,259 S1: are better? That's a completely different way of thinking about software. 437 00:28:36,540 --> 00:28:40,740 S1: Once again, going back to for the consumer for the 438 00:28:40,780 --> 00:28:43,460 S1: same service. Okay. Let's just say it's the same company. 439 00:28:43,459 --> 00:28:47,620 S1: Remove background. Before they were marketing to the consumer with 440 00:28:47,620 --> 00:28:52,380 S1: a pretty website and with SEO from Google, which is 441 00:28:52,380 --> 00:28:54,499 S1: I'm sure, how I found this company in the first place. 442 00:28:54,500 --> 00:28:57,140 S1: Remove background. Boom. It's a number one hit on Google 443 00:28:57,620 --> 00:29:01,300 S1: that was targeted at me and it worked. Now that 444 00:29:01,300 --> 00:29:06,420 S1: marketing goes to my agent, Chi Chi finds that for 445 00:29:06,420 --> 00:29:10,380 S1: me in a directory, the same thing is going to 446 00:29:10,380 --> 00:29:13,979 S1: happen inside the enterprise. The enterprise is going to be like, look, 447 00:29:14,020 --> 00:29:17,459 S1: find all of our software, find out where it is 448 00:29:17,459 --> 00:29:21,180 S1: on this function map. And if it's not, first of all, 449 00:29:21,180 --> 00:29:24,000 S1: if it's not on the map, they're all fired. Cancel 450 00:29:24,000 --> 00:29:28,680 S1: all contracts. Right. If it's not in this map or 451 00:29:28,680 --> 00:29:30,959 S1: it's not in something, we're about to add a node 452 00:29:30,959 --> 00:29:33,479 S1: because we have a new function, a new business line 453 00:29:33,480 --> 00:29:36,360 S1: or new product or service or whatever. If it's not 454 00:29:36,360 --> 00:29:39,400 S1: part of our structure, why do we have it? So 455 00:29:39,440 --> 00:29:45,040 S1: it's just gone, right? But the conversation becomes improving upon 456 00:29:45,040 --> 00:29:49,120 S1: this matrix. That's all that matters. So all these incoming 457 00:29:49,320 --> 00:29:55,400 S1: like pitches and the steak dinners and like sales and marketing, blah, blah, blah, 458 00:29:55,440 --> 00:30:01,480 S1: it becomes a whole lot less important once we're actually 459 00:30:01,480 --> 00:30:05,200 S1: doing an objective measurement of what is the function that 460 00:30:05,200 --> 00:30:07,680 S1: you're performing and how well do you do it. Right. 461 00:30:07,719 --> 00:30:11,960 S1: So that's that's absolutely huge. Okay. So the transition here 462 00:30:12,160 --> 00:30:18,080 S1: is going from corporate product marketing being a human to 463 00:30:18,120 --> 00:30:23,729 S1: human targeted operation that results in a software package being 464 00:30:23,730 --> 00:30:26,890 S1: purchased by a human, which is then used by a 465 00:30:26,890 --> 00:30:33,530 S1: human or not. That's the old world. Two. The enterprise 466 00:30:33,730 --> 00:30:38,930 S1: has a grid, a graph of all operations, and all 467 00:30:38,930 --> 00:30:44,090 S1: functions inside of its company, and software is evaluated based 468 00:30:44,090 --> 00:30:47,330 S1: on the function and the metrics that it provides. And 469 00:30:47,330 --> 00:30:51,450 S1: those functions and metrics need to be superior or they 470 00:30:51,450 --> 00:30:54,850 S1: don't get swapped out. So it's a that's the transition. 471 00:30:54,890 --> 00:31:01,209 S1: It's a fundamental shift there. Okay. Next topic is work itself. Okay. 472 00:31:01,250 --> 00:31:06,650 S1: So automation inside of companies. So the transition here is 473 00:31:07,490 --> 00:31:10,770 S1: automation going from a thing that helps humans do their 474 00:31:10,770 --> 00:31:15,249 S1: jobs better and basically improves productivity and efficiency and stuff 475 00:31:15,250 --> 00:31:19,730 S1: like that, to being a way for companies to get 476 00:31:19,730 --> 00:31:23,590 S1: to their ideal state of being able to do all 477 00:31:23,630 --> 00:31:30,950 S1: the work themselves. This is colossal. This is economy changing. Okay. 478 00:31:30,990 --> 00:31:35,269 S1: Because this really is the end of labor, right? So 479 00:31:35,310 --> 00:31:38,470 S1: there's labor and there's capital. And these have always been 480 00:31:38,469 --> 00:31:41,989 S1: in balance. This is how this is going to get disrupted. 481 00:31:42,150 --> 00:31:45,990 S1: It gets disrupted because companies have always wished that they 482 00:31:45,990 --> 00:31:49,830 S1: could do all the work themselves without employees. If you 483 00:31:49,870 --> 00:31:52,430 S1: are only one founder and you don't have much work 484 00:31:52,430 --> 00:31:54,990 S1: to do, you do all the work yourself. You get 485 00:31:54,990 --> 00:31:59,309 S1: all the profits yourself. Now, if you hire a washing 486 00:31:59,310 --> 00:32:04,469 S1: machine and your your business is washing clothes and you 487 00:32:04,510 --> 00:32:06,670 S1: don't want to expand, you're not trying to scale. You're 488 00:32:06,670 --> 00:32:09,310 S1: just trying to make the money that you make people 489 00:32:09,350 --> 00:32:12,790 S1: bring you dirty clothes. You give them back clean clothes. 490 00:32:14,430 --> 00:32:16,709 S1: That's a cool business. And you make lots of money 491 00:32:16,830 --> 00:32:20,050 S1: and you are able to feed your family and you just. 492 00:32:20,050 --> 00:32:24,290 S1: You have a good life, right? If you hire a 493 00:32:24,330 --> 00:32:27,650 S1: washing machine before you were doing it on, on the 494 00:32:27,650 --> 00:32:31,569 S1: board right in the river and you were making money, 495 00:32:31,570 --> 00:32:34,769 S1: but you weren't making that much money, you weren't making 496 00:32:34,770 --> 00:32:37,930 S1: as much as you could. So you decided to save 497 00:32:37,930 --> 00:32:40,850 S1: up for a year and buy a washing machine. Now 498 00:32:40,850 --> 00:32:45,370 S1: your washing machine is doing way more. You're able to 499 00:32:45,370 --> 00:32:49,930 S1: do way more clothes and make way more money. If 500 00:32:49,930 --> 00:32:52,009 S1: a whole bunch of people come to you and say, hey, 501 00:32:52,370 --> 00:32:57,290 S1: I can also wash clothes in the river. Think about this. 502 00:32:58,050 --> 00:33:01,410 S1: You need to hire me as a clothes washer in 503 00:33:01,410 --> 00:33:05,890 S1: the river, because that's what's fair. And, um, there needs 504 00:33:05,890 --> 00:33:09,810 S1: to be a balance between labor and capital if you're 505 00:33:09,810 --> 00:33:12,930 S1: going to be like, are you kidding me? I can 506 00:33:12,930 --> 00:33:17,820 S1: do all this work myself. I am literally doing all 507 00:33:17,860 --> 00:33:19,900 S1: the work myself. They're going to be like, no, you're 508 00:33:19,900 --> 00:33:24,060 S1: not doing all the work yourself. You have a washing machine. 509 00:33:25,580 --> 00:33:30,140 S1: That is me. Okay. If I have a clothes washing 510 00:33:30,140 --> 00:33:33,420 S1: business and I have ten washing machines behind me, that 511 00:33:33,420 --> 00:33:37,579 S1: is me doing all the work myself. That is the 512 00:33:37,580 --> 00:33:41,580 S1: transition that is happening. That is what AI is. And 513 00:33:41,900 --> 00:33:45,540 S1: just to be clear, the total amount of compensation that 514 00:33:45,540 --> 00:33:54,220 S1: knowledge workers receive is somewhere around $50 trillion per year, 515 00:33:54,820 --> 00:34:00,739 S1: $50 trillion per year globally. I think it's somewhere around 516 00:34:00,740 --> 00:34:09,540 S1: 10 trillion for the US. That is how much money 517 00:34:10,860 --> 00:34:16,520 S1: companies are spending to pay humans. And the major transition 518 00:34:16,520 --> 00:34:19,640 S1: here is they don't want to be paying those humans. 519 00:34:19,840 --> 00:34:23,880 S1: They actually never did want to be paying those humans. 520 00:34:25,839 --> 00:34:28,519 S1: My favorite way of capturing this is the ideal number 521 00:34:28,520 --> 00:34:32,200 S1: of human employees inside of any company is zero. That 522 00:34:32,200 --> 00:34:34,600 S1: is the number that they're trying to get to. Now, 523 00:34:34,640 --> 00:34:37,320 S1: there are exceptions to this, right? If I'm a small, 524 00:34:37,319 --> 00:34:39,439 S1: spunky founder and you know, I want to work with 525 00:34:39,440 --> 00:34:42,680 S1: my friends, um, you know, I want to do a 526 00:34:42,680 --> 00:34:45,520 S1: project with my friends. I get five of my friends involved. 527 00:34:45,520 --> 00:34:48,640 S1: We build a small startup or whatever. It's kind of 528 00:34:48,640 --> 00:34:52,160 S1: like we're all owners at that point, right? So you 529 00:34:52,160 --> 00:34:57,999 S1: still will have, like, elite employees, cadre co-founders, stuff like 530 00:34:58,000 --> 00:35:00,839 S1: that inside of these companies. Like, that's that's not going 531 00:35:00,880 --> 00:35:03,839 S1: to go away. Like hardly any companies are going to 532 00:35:03,839 --> 00:35:07,359 S1: have like zero employees, especially large or medium sized companies. 533 00:35:07,920 --> 00:35:10,759 S1: But we're talking about going from tens of thousands of 534 00:35:10,759 --> 00:35:15,529 S1: employees to a few hundred, maybe eventually a few dozen, 535 00:35:15,569 --> 00:35:18,529 S1: but let's just say a few hundred. We're talking about massive, 536 00:35:18,529 --> 00:35:23,410 S1: massive reduction because of this different way of thinking about automation. 537 00:35:23,410 --> 00:35:25,890 S1: It's not a thing that helps a human do a task, 538 00:35:26,130 --> 00:35:28,770 S1: which is what it's always been. It is a way 539 00:35:28,770 --> 00:35:32,730 S1: to get to the state of the company does the 540 00:35:32,770 --> 00:35:38,249 S1: work itself, which is a natural, clean, happy state for 541 00:35:38,250 --> 00:35:42,810 S1: any company. They prefer to be doing the work themselves. Again, 542 00:35:42,810 --> 00:35:46,290 S1: go back to the single person with a washing machine business. 543 00:35:46,610 --> 00:35:49,930 S1: The washing machine business is part of them. It is 544 00:35:49,930 --> 00:35:53,330 S1: part of their business. So they don't have a reason 545 00:35:53,330 --> 00:35:57,130 S1: to hire employees. So that is the major transition, their 546 00:35:58,089 --> 00:36:03,170 S1: automation going from a thing that helps employees to being 547 00:36:03,170 --> 00:36:07,450 S1: a way for the company to do all the work themselves. Okay, 548 00:36:07,489 --> 00:36:11,710 S1: so that brings us to the next transition, which is okay, fine. 549 00:36:12,630 --> 00:36:15,509 S1: That makes sense. What are we supposed to do? What 550 00:36:15,509 --> 00:36:19,870 S1: are we supposed to do? We have jobs, right? How do. Okay. 551 00:36:19,910 --> 00:36:23,589 S1: If everyone gets fired, who's going to buy all the stuff? Right? 552 00:36:23,630 --> 00:36:26,509 S1: Who's going to buy all the stuff? So that's a 553 00:36:26,509 --> 00:36:29,590 S1: UBI conversation. Not really going to go into that. Like 554 00:36:30,589 --> 00:36:33,070 S1: there's going to be money. People are going to receive 555 00:36:33,069 --> 00:36:35,430 S1: money to pay their bills and stuff like that. Otherwise 556 00:36:35,430 --> 00:36:39,390 S1: you just don't have a society. It's all bad. So 557 00:36:39,589 --> 00:36:42,630 S1: that will get solved one way or another. Hopefully, uh, 558 00:36:42,910 --> 00:36:48,870 S1: more gently and faster and easier, uh, rather than the alternative. Okay. 559 00:36:48,910 --> 00:36:51,709 S1: So what does this look like for actual people? So 560 00:36:51,750 --> 00:36:57,069 S1: instead of people en masse working for companies, you know, 561 00:36:57,110 --> 00:37:01,950 S1: medium large sized companies or whatever, that goes away, that 562 00:37:01,950 --> 00:37:04,950 S1: goes away, right? Because those companies like we talked about, 563 00:37:04,950 --> 00:37:07,509 S1: they're trying to get rid of everyone. Instead of that 564 00:37:07,509 --> 00:37:09,850 S1: being the way that we make our money. We're going 565 00:37:09,890 --> 00:37:14,489 S1: to make money by producing value ourselves, by articulating the 566 00:37:14,489 --> 00:37:17,450 S1: skills that we have, the capabilities that we have, the 567 00:37:17,450 --> 00:37:21,210 S1: products that we provide, the services that we provide, broadcasting 568 00:37:21,210 --> 00:37:23,850 S1: that out. And that is going to go up into 569 00:37:23,850 --> 00:37:26,850 S1: one of these directories like I talked about before, with 570 00:37:26,850 --> 00:37:30,529 S1: products that AI can look at, right? AI can find 571 00:37:30,529 --> 00:37:34,010 S1: products in one of these directories, but this will also 572 00:37:34,009 --> 00:37:37,850 S1: be the substrate for all work to be done. So 573 00:37:38,089 --> 00:37:41,290 S1: humans will broadcast their capabilities and say, look, you know, 574 00:37:41,330 --> 00:37:44,890 S1: I'm a systems engineer. I've got eight years of experience, 575 00:37:44,890 --> 00:37:48,410 S1: two years experience, 25 years experience, whatever. Here's all the 576 00:37:48,410 --> 00:37:51,489 S1: different stuff I can do. Here's my portfolio, blah, blah, blah. 577 00:37:51,529 --> 00:37:53,770 S1: By the way, I like to mountain bike, blah blah blah. 578 00:37:53,770 --> 00:37:57,930 S1: So this is your demon. This is your broadcast system describing, 579 00:37:57,969 --> 00:38:01,370 S1: you know, the people you've worked with, like your reputation score. 580 00:38:01,410 --> 00:38:04,450 S1: People give you upvotes kind of similar to what LinkedIn 581 00:38:04,450 --> 00:38:06,969 S1: was doing. They kind of went away from this. It's 582 00:38:06,969 --> 00:38:10,100 S1: now kind of like a social network. So this is 583 00:38:10,100 --> 00:38:13,700 S1: the play though. There's a substrate that connects all these 584 00:38:13,700 --> 00:38:18,219 S1: different people, right. When I need a cat sitter because 585 00:38:18,219 --> 00:38:21,580 S1: I'm going on vacation, I'm going to broadcast out, hey, 586 00:38:21,620 --> 00:38:23,379 S1: I need someone to watch my cat. They need to 587 00:38:23,380 --> 00:38:27,460 S1: be fed. You know, they're crazy. They might scratch you. Whatever. 588 00:38:27,860 --> 00:38:32,299 S1: Broadcast that out. Keep in mind, uh, my AI is 589 00:38:32,299 --> 00:38:37,780 S1: broadcasting that out for me. Okay, everyone around AI is 590 00:38:37,779 --> 00:38:41,819 S1: watching all the substrate, whatever this thing is called, and 591 00:38:41,860 --> 00:38:44,739 S1: they're like, hey, I'm a cat person. You know, I'm 592 00:38:44,739 --> 00:38:46,819 S1: a cat lady, I love cats, I can take care 593 00:38:46,819 --> 00:38:49,660 S1: of cats. Um, you know, I don't get scratched or whatever, 594 00:38:49,660 --> 00:38:51,780 S1: or I don't get sick when I get scratched, whatever 595 00:38:51,779 --> 00:38:55,219 S1: it is. And I also live, you know, two blocks away. 596 00:38:55,739 --> 00:39:00,740 S1: Being in her ear, her AI tells her, hey, there's 597 00:39:00,739 --> 00:39:03,339 S1: a cat sitting job over here. You know it's going 598 00:39:03,380 --> 00:39:09,520 S1: to pay $84. Um. Do you accept? Boom. Yes. Same thing. 599 00:39:09,520 --> 00:39:12,520 S1: Like someone. There's a crash on the on the corner. 600 00:39:12,560 --> 00:39:16,880 S1: You know, someone injured themselves. Hey, does anyone have any medical, uh, 601 00:39:16,880 --> 00:39:22,280 S1: you know, professional training or whatever? Does anyone have EMT skills? Boom. 602 00:39:22,279 --> 00:39:26,399 S1: It's going to beacon for people all around nearby who 603 00:39:26,400 --> 00:39:30,640 S1: have a reputation score above so and so. They get beacon. 604 00:39:30,680 --> 00:39:34,080 S1: Someone takes the job, they go help the person. Okay, 605 00:39:34,120 --> 00:39:37,999 S1: same for gardening, same for engineering services, same for hey, 606 00:39:38,000 --> 00:39:41,200 S1: I need a shoulder to cry on. Hey, I need nurturing. Hey, 607 00:39:41,200 --> 00:39:45,480 S1: I need tutor for my kids. Hey, I need meal prep. Um, 608 00:39:45,839 --> 00:39:48,319 S1: I need to gain a lot of muscle mass in 609 00:39:48,319 --> 00:39:52,359 S1: my legs. Uh, I need a personal trainer. Everyone who 610 00:39:52,360 --> 00:39:56,519 S1: has services, capabilities, value to offer, they are beaconing out 611 00:39:56,520 --> 00:40:00,039 S1: onto this system. It is out there on the substrate, 612 00:40:00,400 --> 00:40:06,129 S1: and then everyone's AI is looking at that system. And 613 00:40:06,130 --> 00:40:09,209 S1: that's how you find work. Okay. Another version of this 614 00:40:09,210 --> 00:40:12,649 S1: is like Fiverr. Okay. It never it's not automated, but 615 00:40:12,650 --> 00:40:16,330 S1: it's the same kind of vibe. Uh, also as LinkedIn, 616 00:40:16,370 --> 00:40:20,330 S1: it's like, here are my services. Here are the jobs. Right. 617 00:40:20,370 --> 00:40:24,329 S1: So you you've got consumer and producer, right? You've got like, 618 00:40:24,370 --> 00:40:26,850 S1: here's the work I need done. And then you've got 619 00:40:26,850 --> 00:40:29,090 S1: the other group saying, here's all the stuff I can do. 620 00:40:29,450 --> 00:40:31,169 S1: And then it's a matter of, okay, we're going to 621 00:40:31,170 --> 00:40:33,569 S1: join together. It's going to be five people on this project. 622 00:40:33,569 --> 00:40:35,850 S1: It's going to be a six month long project. Here's 623 00:40:35,850 --> 00:40:40,250 S1: the money that I'm offering. Boom, boom, boom. Everyone agrees. Boom, 624 00:40:40,250 --> 00:40:45,570 S1: the work starts. That is what I'm calling human 3.0. 625 00:40:45,610 --> 00:40:49,089 S1: It's what I'm calling the state that we're going to 626 00:40:49,089 --> 00:40:52,729 S1: get to do. I know this is going to happen 627 00:40:52,730 --> 00:40:57,050 S1: for sure. I believe the answer is yes. Um, only 628 00:40:57,049 --> 00:41:01,129 S1: because I'm an optimist and I believe we are not 629 00:41:01,130 --> 00:41:04,870 S1: just going to die off and, you know, devolve into 630 00:41:04,870 --> 00:41:09,109 S1: chaos after the corporate system breaks down. Uh, I can 631 00:41:09,110 --> 00:41:11,189 S1: tell you that the corporate system is going to break 632 00:41:11,190 --> 00:41:15,189 S1: down because the corporate system is going to pull all 633 00:41:15,190 --> 00:41:18,830 S1: the work internally and do it with automation and AI 634 00:41:18,870 --> 00:41:25,910 S1: and robotics. That seems obviously inevitable. Again, not everyone, but 635 00:41:26,430 --> 00:41:29,950 S1: most people will be laid off as a result of this. 636 00:41:30,230 --> 00:41:34,509 S1: Over time, it'll happen different speeds and different industries, but 637 00:41:34,509 --> 00:41:39,150 S1: that is pretty much inevitable. I think that this better solution, 638 00:41:39,150 --> 00:41:43,190 S1: this new solution, is also inevitable. The reason I love this, 639 00:41:43,469 --> 00:41:45,870 S1: and this is a little bit of an aside, but 640 00:41:45,870 --> 00:41:48,710 S1: I love this because it's more human focused. It's humans 641 00:41:48,710 --> 00:41:52,589 S1: connecting with humans. And there'll still be companies, right? It's 642 00:41:52,589 --> 00:41:56,950 S1: humans connecting with companies. But it's not in this hierarchical, 643 00:41:56,989 --> 00:42:01,090 S1: you know, you know what corp corporate. That's that's military. 644 00:42:01,410 --> 00:42:05,290 S1: This idea of. I work for Sarah. Sarah works for Joe. 645 00:42:05,330 --> 00:42:09,049 S1: Joe works for Raj. And. Oh, I'm having a meeting 646 00:42:09,049 --> 00:42:12,370 S1: with Raj. Oh my God. You know, he's three skip 647 00:42:12,370 --> 00:42:16,410 S1: levels above me. Like this whole military structure, this whole, like, 648 00:42:16,690 --> 00:42:20,530 S1: dreading Monday. And just like, this whole thing is, like, 649 00:42:20,690 --> 00:42:25,290 S1: toxic and poisonous. And it has been for decades. People 650 00:42:25,290 --> 00:42:28,450 S1: have been so unhappy with this. And now that it's 651 00:42:28,489 --> 00:42:31,089 S1: actually under threat, people are like, well, don't get rid 652 00:42:31,130 --> 00:42:34,810 S1: of my job. Don't fire me. And which is justified, right? 653 00:42:34,810 --> 00:42:37,209 S1: Because they're worried about losing their livelihood and they're worried 654 00:42:37,210 --> 00:42:40,250 S1: about being able to pay the rent and a mortgage and, 655 00:42:40,290 --> 00:42:44,489 S1: you know, school and groceries. So it's understandable that they're 656 00:42:44,489 --> 00:42:48,089 S1: clinging to that. But remember, we shouldn't be clinging to 657 00:42:48,130 --> 00:42:52,410 S1: a thing we hate and we have hated. Right. All 658 00:42:52,410 --> 00:42:55,489 S1: this to say, I think the way that it's going, 659 00:42:55,650 --> 00:42:58,700 S1: the way that I'm pushing for, I'm actively building to 660 00:42:58,739 --> 00:43:02,700 S1: make this happen is to have this new human based 661 00:43:02,700 --> 00:43:07,979 S1: substrate where producer, consumer producer, consumer things are a lot 662 00:43:07,980 --> 00:43:11,860 S1: more equal. You get hired based on your skills, and 663 00:43:12,259 --> 00:43:14,779 S1: that's a relationship you can get out of any time. 664 00:43:15,299 --> 00:43:17,620 S1: And here's what's cool about it. You can be in 665 00:43:17,660 --> 00:43:22,380 S1: multiple of these, right. You can have ongoing like retainer 666 00:43:22,380 --> 00:43:26,100 S1: type things going on with like 20 different customers. And 667 00:43:26,100 --> 00:43:28,900 S1: you're on this big project for a full six months 668 00:43:28,900 --> 00:43:31,299 S1: and you're getting paid from that. Plus you're doing the 669 00:43:31,299 --> 00:43:34,739 S1: cat sitting. Plus you're a part time EMT in the 670 00:43:34,779 --> 00:43:37,299 S1: in the nights just in case someone falls off their 671 00:43:37,299 --> 00:43:39,299 S1: bike or whatever, you know what I mean? So it's 672 00:43:39,299 --> 00:43:44,620 S1: like it's just a more aligned thing, right? Because now 673 00:43:44,980 --> 00:43:50,620 S1: you're actually broadcasting everything that you you are right. You're like, yeah, 674 00:43:50,660 --> 00:43:53,500 S1: I play violin. I will come to your son's birthday 675 00:43:53,500 --> 00:43:57,239 S1: party and play violin if they like classical music. So 676 00:43:57,239 --> 00:44:01,920 S1: it's like you're broadcasting not just, oh, you know, here's 677 00:44:01,920 --> 00:44:05,920 S1: my resume. I'm a tech engineer, level three, blah blah, blah. 678 00:44:05,960 --> 00:44:10,640 S1: I worked at Intuit. That is not you. You. What 679 00:44:10,640 --> 00:44:12,719 S1: this is going to allow us to do is broadcast 680 00:44:12,719 --> 00:44:17,720 S1: our full selves and monetize ourselves in the best way possible. 681 00:44:17,719 --> 00:44:20,480 S1: Not in a gross way. We're going to say we 682 00:44:20,520 --> 00:44:25,440 S1: are going to be compensated and rewarded for being ourselves. 683 00:44:26,000 --> 00:44:29,160 S1: If we are the best nurturers in the world, forget 684 00:44:29,200 --> 00:44:34,160 S1: tech skills. Forget any tech, right? You're the best nurturer. 685 00:44:34,200 --> 00:44:36,959 S1: You're the best listener. You're the best mother, you're the 686 00:44:36,960 --> 00:44:41,040 S1: best parent, you're the best tutor. You're the best trainer. 687 00:44:41,080 --> 00:44:45,759 S1: You're the best boxing coach. Okay? They're not producing products. 688 00:44:45,759 --> 00:44:49,680 S1: They're not releasing startups. They're helping other people become the 689 00:44:49,680 --> 00:44:53,080 S1: best versions of themselves. In many ways, they're the most 690 00:44:53,080 --> 00:44:56,569 S1: important people. Well, guess what? That's in there. Demon that 691 00:44:56,569 --> 00:45:00,609 S1: is broadcasted. They become world famous for that. So they 692 00:45:00,610 --> 00:45:03,770 S1: got everyone trying to come in and use them, right? 693 00:45:05,009 --> 00:45:07,850 S1: And they make money from it, which they should. This 694 00:45:07,850 --> 00:45:11,969 S1: is how humanity should work, right? So anyway, this is 695 00:45:11,969 --> 00:45:14,370 S1: the tech layer I think is going to replace the 696 00:45:14,370 --> 00:45:18,089 S1: whole system. This is the transition that I think is happening. 697 00:45:18,770 --> 00:45:20,690 S1: I just want to cover that because a lot of 698 00:45:20,690 --> 00:45:23,410 S1: people have questions. What am I supposed to do after 699 00:45:23,410 --> 00:45:26,689 S1: we lose all these jobs? Okay, next topic I want 700 00:45:26,730 --> 00:45:30,649 S1: to talk about is cybersecurity. So similar to a lot 701 00:45:30,650 --> 00:45:33,770 S1: of the other things we've already talked about. Cybersecurity has 702 00:45:33,770 --> 00:45:39,009 S1: been human based, right. Um, it's you hire a human team, 703 00:45:39,410 --> 00:45:43,489 S1: you have human staff. They are good security engineers or whatever. 704 00:45:43,930 --> 00:45:48,970 S1: They're doing pen tests, they're doing security assessments, vulnerability assessments. 705 00:45:49,130 --> 00:45:52,770 S1: They are manually looking at all the different vendors and 706 00:45:52,870 --> 00:45:55,550 S1: trying to figure out, is this one dangerous? Um, should 707 00:45:55,549 --> 00:45:59,870 S1: we allow it through procurement? And they're just being bombarded 708 00:45:59,870 --> 00:46:03,189 S1: by all these requests. And, you know, if you have 709 00:46:03,190 --> 00:46:08,390 S1: a really good, you know, third party service person, um, 710 00:46:08,390 --> 00:46:11,989 S1: auditor person, they could look at x number of, of 711 00:46:12,029 --> 00:46:15,950 S1: those vendors per day. Maybe some of them get escalated 712 00:46:15,950 --> 00:46:19,190 S1: because the CEO really wants that, uh, piece of software. 713 00:46:19,190 --> 00:46:21,270 S1: So you have to do a security assessment. You have 714 00:46:21,270 --> 00:46:23,109 S1: to find some way to make it secure because it's 715 00:46:23,110 --> 00:46:25,710 S1: coming in the company anyway. That's the life of a 716 00:46:25,710 --> 00:46:28,589 S1: security person. That's the life of a security program. It's 717 00:46:28,589 --> 00:46:32,830 S1: humans grappling with all this stuff in a very gross 718 00:46:32,870 --> 00:46:37,670 S1: sort of in the weeds, like sewer type of thing. 719 00:46:37,790 --> 00:46:42,390 S1: It's not pretty, it's not clean, but it's got the 720 00:46:42,390 --> 00:46:46,150 S1: job done, right. We're still alive. The inner still mostly works. 721 00:46:46,509 --> 00:46:49,109 S1: You could still go to the ATM and get cash out. 722 00:46:49,110 --> 00:46:52,959 S1: So security has been largely successful. Keep in mind, we're 723 00:46:52,960 --> 00:46:56,000 S1: still losing billions to fraud all the time, but it's 724 00:46:56,000 --> 00:47:00,519 S1: a messy process. What security becomes now with all this 725 00:47:00,520 --> 00:47:05,839 S1: AI stuff is it becomes your AI stack. As a 726 00:47:05,839 --> 00:47:10,040 S1: defender against the AI stack of the attacker. And unfortunately, 727 00:47:10,040 --> 00:47:13,239 S1: you're not facing one attacker, you're facing all the attackers. 728 00:47:13,759 --> 00:47:19,960 S1: So the attacker is trying to understand your company extremely well. 729 00:47:19,960 --> 00:47:23,440 S1: It's making a list of all your employees. It's creating 730 00:47:23,440 --> 00:47:26,680 S1: a personality profiles on all your employees. It's coming up 731 00:47:26,680 --> 00:47:30,320 S1: with the best spear phishing campaigns to find the ones 732 00:47:30,319 --> 00:47:33,680 S1: who probably have the most access based on their job title. 733 00:47:33,680 --> 00:47:37,999 S1: And it's sending out these spear phishing attacks. It's constantly pulling. 734 00:47:38,000 --> 00:47:39,960 S1: Your DNS is trying to see if you're doing a 735 00:47:39,960 --> 00:47:43,759 S1: merger and acquisition with a company that doesn't have good security, 736 00:47:43,759 --> 00:47:46,759 S1: so that's going to be a weakness. And they're launching 737 00:47:47,080 --> 00:47:51,940 S1: a malware. They're sending you a spear phishing emails. They're 738 00:47:51,940 --> 00:47:55,620 S1: trying to compromise all your websites. They're trying to pivot internally. 739 00:47:55,620 --> 00:47:58,100 S1: They're trying to sell the access that they have. And 740 00:47:58,100 --> 00:48:01,180 S1: they're doing this at like machine speed, right? They've just 741 00:48:01,180 --> 00:48:04,780 S1: got so many agents working on this constantly, constantly hitting you. 742 00:48:04,980 --> 00:48:08,939 S1: You can't tell Chris and Raj and Sarah. Hey, uh, 743 00:48:08,980 --> 00:48:11,580 S1: great job last year. Um, I need you to do 744 00:48:11,739 --> 00:48:18,739 S1: 895 times as much work because, uh, that's how many 745 00:48:18,739 --> 00:48:23,020 S1: more attacks were being hit with. That doesn't work. Also, 746 00:48:23,060 --> 00:48:25,939 S1: you can't be like, uh. Hey, great news. We got 747 00:48:25,940 --> 00:48:29,420 S1: three more head count. That also won't make a dent. 748 00:48:29,779 --> 00:48:34,739 S1: Your only chance is to have the same AI or 749 00:48:34,739 --> 00:48:38,379 S1: better as the attacker. And this goes back to what 750 00:48:38,380 --> 00:48:42,020 S1: happens to all companies. What happens to all companies also 751 00:48:42,020 --> 00:48:46,620 S1: happens to security programs. It's no longer about here's our 752 00:48:46,620 --> 00:48:50,080 S1: security team hears roughly the things we need to do. 753 00:48:50,640 --> 00:48:54,840 S1: And yeah, there's some documentation, but ultimately it's like who's 754 00:48:54,839 --> 00:48:58,079 S1: on call, who's doing the work, blah, blah, blah. And 755 00:48:58,080 --> 00:49:02,040 S1: it's like a very human focused thing that's all out 756 00:49:02,040 --> 00:49:10,440 S1: the window. SOPs, SOPs, everything is a process and workflows 757 00:49:10,440 --> 00:49:13,999 S1: which you could visually look at and see, look, this 758 00:49:14,000 --> 00:49:18,160 S1: is the queue for processing incoming things. Uh, here's a 759 00:49:18,160 --> 00:49:21,719 S1: queue for, you know, here are the constant workflows that 760 00:49:21,719 --> 00:49:24,480 S1: are happening, all the cron jobs or whatever it is 761 00:49:25,040 --> 00:49:28,759 S1: for finding insider threats. Here's all the processes for managing 762 00:49:28,759 --> 00:49:33,600 S1: the CI, CD. Here's every single tool, every single decision point, 763 00:49:33,640 --> 00:49:36,919 S1: every single approval point that needs to happen as part 764 00:49:36,920 --> 00:49:40,640 S1: of CI CD before something goes live. We're doing static analysis. 765 00:49:40,680 --> 00:49:44,240 S1: We're doing dynamic. We're doing, uh, you know, we're looking 766 00:49:44,239 --> 00:49:47,770 S1: at some other aspect of the application, how it impacts 767 00:49:47,770 --> 00:49:51,730 S1: the business. Each one of those turns into a little node. 768 00:49:52,529 --> 00:49:55,289 S1: That is where a tool drops into. This is where 769 00:49:55,290 --> 00:49:57,650 S1: a human drops into. This is where the human is 770 00:49:57,650 --> 00:50:00,769 S1: replaced because a very smart AI model just took over 771 00:50:00,770 --> 00:50:08,930 S1: that role. Everything becomes transparent, visible with discrete actions and 772 00:50:08,930 --> 00:50:13,370 S1: activities and decision points at each area. And the question 773 00:50:13,370 --> 00:50:18,450 S1: is then how well are we doing this? How good 774 00:50:18,450 --> 00:50:22,450 S1: are our information sources feeding our context? This concept of 775 00:50:22,450 --> 00:50:26,370 S1: unified context that I put out a year or so ago, 776 00:50:27,650 --> 00:50:34,250 S1: maybe 2024, actually unified entity context. Everyone like I was 777 00:50:34,250 --> 00:50:36,970 S1: talking about with the corporate side, also the security side, 778 00:50:36,969 --> 00:50:39,810 S1: we're all working off a unified context of the company. 779 00:50:40,009 --> 00:50:42,930 S1: What are its goals? What are its challenges? What's the 780 00:50:42,930 --> 00:50:46,029 S1: risk register. Look like they're all working off the same 781 00:50:46,029 --> 00:50:49,510 S1: sort of thing. Right here are our attackers, here are 782 00:50:49,509 --> 00:50:52,509 S1: sensitive things. Here's our crown jewels over here. All of 783 00:50:52,509 --> 00:50:55,430 S1: this is unified. We have this map of all the 784 00:50:55,469 --> 00:50:59,230 S1: workflows that are happening, all the different tools. And now 785 00:50:59,430 --> 00:51:04,549 S1: the question is, are we getting that context updated? Okay. 786 00:51:04,870 --> 00:51:08,270 S1: New AWS account, which just stood up. We just launched 787 00:51:08,270 --> 00:51:13,070 S1: a new application, a new service in Asia and also 788 00:51:13,110 --> 00:51:16,629 S1: Iceland and also in Seattle. And it's got all these 789 00:51:16,630 --> 00:51:20,589 S1: new services on it. How quickly are we as a 790 00:51:20,589 --> 00:51:23,350 S1: security team learning that they did that? Because that was 791 00:51:23,350 --> 00:51:26,029 S1: some crazy marketing group over there or whatever. Let's just 792 00:51:26,029 --> 00:51:29,390 S1: assume we're going to have shadow it still for quite 793 00:51:29,390 --> 00:51:32,190 S1: a while. This transition I'm talking about with like everything 794 00:51:32,190 --> 00:51:36,229 S1: being mapped out and SOPs. That's not going to happen overnight. 795 00:51:36,509 --> 00:51:38,710 S1: This is going to take a long time. Some companies 796 00:51:38,710 --> 00:51:41,550 S1: this is going to take 15 years and they'll probably 797 00:51:41,549 --> 00:51:44,439 S1: be dead, but some will survive and it will take 798 00:51:44,440 --> 00:51:46,879 S1: them 15 years because they're a soup sandwich. It will 799 00:51:46,880 --> 00:51:51,200 S1: take them forever. Somehow they survived. Many companies, they'll do 800 00:51:51,200 --> 00:51:54,040 S1: this over the course of two or 3 or 5 years. Right? 801 00:51:54,080 --> 00:51:56,839 S1: Some companies have already started this. Some companies won't start 802 00:51:56,839 --> 00:52:01,839 S1: this for another three years. It's massive, massive variance in 803 00:52:01,839 --> 00:52:05,239 S1: how fast it's going to happen across various industries. Okay. 804 00:52:05,279 --> 00:52:07,680 S1: So I don't want you to think like, oh, I 805 00:52:07,680 --> 00:52:11,279 S1: think this is happening right now to everyone. And it's 806 00:52:11,279 --> 00:52:15,000 S1: just like it's going great. No, this is barely starting. Okay. 807 00:52:15,319 --> 00:52:19,879 S1: So the point is this structure is what attackers are 808 00:52:19,880 --> 00:52:22,719 S1: going to use to attack you. They're going to have 809 00:52:22,719 --> 00:52:26,839 S1: a world model of your company which has all these 810 00:52:26,839 --> 00:52:30,439 S1: pieces right there. Hopefully they're not as filled in as 811 00:52:30,440 --> 00:52:35,000 S1: your version, right. Um, when you launch that AWS stuff, 812 00:52:35,000 --> 00:52:39,200 S1: when you launch those new services, you have access to AWS. Hopefully. 813 00:52:39,200 --> 00:52:41,899 S1: Hopefully your attacker doesn't, so you should be able to 814 00:52:41,940 --> 00:52:45,060 S1: pull that like really quickly. Okay. You should have access 815 00:52:45,060 --> 00:52:47,819 S1: to all your endpoint visibility. You should have access to 816 00:52:47,819 --> 00:52:50,979 S1: all your different tools, all your different APIs, all your 817 00:52:50,980 --> 00:52:56,020 S1: different agent infrastructure constantly pulling asset management. Asset management is 818 00:52:56,020 --> 00:53:01,140 S1: about to get elevated to top top tier. Finally thank goodness. 819 00:53:01,339 --> 00:53:06,339 S1: But you should have this data way faster than your attacker. Unfortunately, 820 00:53:06,819 --> 00:53:11,060 S1: all of this is very, very ancient inside of most companies. 821 00:53:11,060 --> 00:53:16,580 S1: I have been doing consulting for over 20 years. Most 822 00:53:16,580 --> 00:53:20,979 S1: companies I see they do not have asset management. They 823 00:53:20,980 --> 00:53:25,100 S1: do not have unified documentation. They do not have most 824 00:53:25,100 --> 00:53:28,219 S1: of this stuff. Okay, we're in a bad state. We're 825 00:53:28,219 --> 00:53:31,660 S1: in a bad state because as AI starts spinning this 826 00:53:31,660 --> 00:53:34,100 S1: stuff up for the attackers, they're going to build a 827 00:53:34,100 --> 00:53:38,180 S1: world model of these companies faster than the company has it, 828 00:53:38,239 --> 00:53:40,120 S1: because the company has to go slow. They have to 829 00:53:40,120 --> 00:53:43,560 S1: have 19 meetings to prepare for the meeting. Attackers are 830 00:53:43,560 --> 00:53:46,879 S1: just going to like YOLO it, you know, submit the 831 00:53:46,880 --> 00:53:52,439 S1: single prompt, make no mistakes and start attacking. So all 832 00:53:52,440 --> 00:53:58,560 S1: this to say, transition is humans doing security work to 833 00:53:59,520 --> 00:54:05,640 S1: a unified workflow model with SOPs of the work that 834 00:54:05,640 --> 00:54:10,880 S1: is being executed largely by agents, largely by AI, with 835 00:54:10,880 --> 00:54:16,520 S1: humans there to tweak and improve and guide and instruct 836 00:54:16,520 --> 00:54:22,759 S1: and steer and validate the AI automation. And the game 837 00:54:22,880 --> 00:54:26,839 S1: here is for your orchestration system to be better than 838 00:54:26,839 --> 00:54:29,520 S1: the attackers. All right. So this next one is a 839 00:54:29,520 --> 00:54:33,640 S1: way of thinking about enterprise AI in like a completely 840 00:54:33,640 --> 00:54:36,930 S1: different way. And I put this out about a year 841 00:54:36,930 --> 00:54:40,769 S1: ago or so, and I think it's really powerful as 842 00:54:40,770 --> 00:54:46,049 S1: an inversion. So the idea is that, um, currently it 843 00:54:46,049 --> 00:54:48,930 S1: definitely a year ago, two years ago. But even still 844 00:54:48,930 --> 00:54:53,089 S1: now everyone is thinking, okay, we have security, we need 845 00:54:53,089 --> 00:54:55,890 S1: to put AI on it. Okay, we have finance, we 846 00:54:55,890 --> 00:54:59,210 S1: need to put AI on it. Okay. We have um, 847 00:54:59,489 --> 00:55:03,009 S1: HR we need to put AI on that, which means 848 00:55:03,290 --> 00:55:08,770 S1: we need AI agents. Um, for finance we need finance AI, 849 00:55:08,969 --> 00:55:14,569 S1: we need, um, security, AI, security, AI agents. Right. So 850 00:55:14,569 --> 00:55:18,170 S1: the idea is you have the discipline, you have the topic, 851 00:55:18,810 --> 00:55:23,049 S1: and then AI gets sprinkled on top, and it's supposedly 852 00:55:23,049 --> 00:55:27,009 S1: going to do something that benefits you related to that topic. 853 00:55:28,210 --> 00:55:31,690 S1: This is I don't think, the way to think about it. Okay. 854 00:55:31,730 --> 00:55:34,210 S1: I think the way to think about this is actually 855 00:55:34,210 --> 00:55:38,549 S1: you have a company and you have the company's work 856 00:55:38,910 --> 00:55:41,149 S1: and all its workflows and the graph of all the 857 00:55:41,150 --> 00:55:45,549 S1: services and the tools and the operations, SOPs, goals, everything 858 00:55:45,950 --> 00:55:51,270 S1: that is actually the system. The system is the graph 859 00:55:51,270 --> 00:55:56,469 S1: of operations, okay? It is the graph of algorithms that 860 00:55:56,469 --> 00:56:00,310 S1: take place to make this business function okay. Maybe it 861 00:56:00,310 --> 00:56:03,629 S1: has humans there, maybe it doesn't, maybe has lots of humans, 862 00:56:03,670 --> 00:56:07,750 S1: maybe it's mostly AI and a few humans. Doesn't really matter. 863 00:56:07,790 --> 00:56:12,430 S1: This graph of functions, this graph of algorithms is the company. 864 00:56:12,830 --> 00:56:18,629 S1: Now think of AI as a system for running this 865 00:56:18,630 --> 00:56:24,830 S1: graph of algorithms. That is what AI is. Then you 866 00:56:24,830 --> 00:56:29,269 S1: have the question of okay, what are you doing for finance? 867 00:56:29,710 --> 00:56:35,120 S1: What does procurement look like? Show me procurement. You're looking 868 00:56:35,120 --> 00:56:40,800 S1: at this graph. And this one little, um, line lights up. 869 00:56:41,400 --> 00:56:44,959 S1: It's actually like 19 different lines. They all light up. Oh, 870 00:56:45,000 --> 00:56:50,600 S1: these are the procurement workflows. Cool. So we can drill 871 00:56:50,600 --> 00:56:53,480 S1: into those. We can inspect. Okay, here's the tools. Here's 872 00:56:53,480 --> 00:56:56,680 S1: the human involved. Here's the decisions. Here's the sign offs. 873 00:56:56,759 --> 00:57:00,480 S1: Here's the exceptions. Here's the risk register. Whatever it is, 874 00:57:00,839 --> 00:57:06,160 S1: the overall system is the graph of algorithms largely run 875 00:57:06,160 --> 00:57:09,640 S1: by AI. What ends up happening is all the different 876 00:57:09,640 --> 00:57:13,279 S1: things that used to be industries. They become use cases 877 00:57:13,279 --> 00:57:18,240 S1: inside of AI. Okay. So the before is you have 878 00:57:18,240 --> 00:57:22,800 S1: industries using AI. And now what you have is a 879 00:57:23,400 --> 00:57:29,480 S1: graph of algorithms run by AI that has use cases 880 00:57:29,480 --> 00:57:33,860 S1: for different things. Right. You you've got some of these 881 00:57:33,860 --> 00:57:36,740 S1: things are security. Some of these things are HR. Some 882 00:57:36,740 --> 00:57:39,860 S1: of these things are engineering. Some of these things are marketing. 883 00:57:41,860 --> 00:57:46,900 S1: AI is the container. AI is the thing. And it 884 00:57:46,900 --> 00:57:50,419 S1: just has functions that happen to be affiliated with what 885 00:57:50,420 --> 00:57:55,180 S1: we used to call industries, that that is a fundamental transition. 886 00:57:55,180 --> 00:57:58,499 S1: Now it's not absolute. This is a mental model, but 887 00:57:58,500 --> 00:58:01,499 S1: I think this mental model is a lot more descriptive 888 00:58:01,500 --> 00:58:05,380 S1: of what's actually happening. Then thinking about it in terms 889 00:58:05,380 --> 00:58:07,979 S1: of like, well, let's do everything the way we used 890 00:58:07,980 --> 00:58:11,459 S1: to do it before. Okay, we've got an HR database, 891 00:58:11,460 --> 00:58:14,300 S1: we're going to have an HR interface. We're going to 892 00:58:14,300 --> 00:58:21,180 S1: have HR tools and an HR dashboard. No that's going away. Okay. 893 00:58:21,220 --> 00:58:25,060 S1: If you abstract everything to questions and everything to algorithms, 894 00:58:25,340 --> 00:58:29,019 S1: the question is something like how happy are our employees? 895 00:58:30,200 --> 00:58:33,840 S1: How much money are we spending on compensating employees? Should 896 00:58:33,840 --> 00:58:38,200 S1: we pay more for bonuses? Those are questions. Those questions 897 00:58:38,200 --> 00:58:41,520 S1: are not in industry. Those are questions that just happen 898 00:58:41,520 --> 00:58:44,760 S1: to be associated with what we used to call an industry. 899 00:58:45,080 --> 00:58:48,520 S1: Those are all hitting unified entity contexts. The results are 900 00:58:48,520 --> 00:58:54,040 S1: coming out and you have the answer. But ultimately, it 901 00:58:54,040 --> 00:59:00,760 S1: is all feeding off this underlying unified context and graph 902 00:59:00,760 --> 00:59:04,040 S1: of algorithms and understanding of what the business actually is. 903 00:59:04,600 --> 00:59:08,000 S1: And all of that is powered and enhanced and managed 904 00:59:08,000 --> 00:59:14,320 S1: and orchestrated by AI. Okay, here's the next one. Custom everything. 905 00:59:14,920 --> 00:59:19,600 S1: So in the past and this one crosses consumer and enterprise. 906 00:59:20,360 --> 00:59:23,680 S1: So in the past you basically had very few people, 907 00:59:24,040 --> 00:59:28,850 S1: very few organizations, very few companies producing software, right? You've 908 00:59:28,850 --> 00:59:33,050 S1: got Adobe, right? There aren't many competitors to Adobe. When 909 00:59:33,090 --> 00:59:36,890 S1: Adobe was on top, it was on top. When Microsoft 910 00:59:36,890 --> 00:59:38,810 S1: was on top, it was on top. You have a 911 00:59:38,810 --> 00:59:42,770 S1: few competitors, but you go into any enterprise. They're mostly 912 00:59:42,770 --> 00:59:48,370 S1: running Microsoft Shop, they're mostly running Google Shop, whatever. I 913 00:59:48,370 --> 00:59:53,209 S1: think that might start to go away. Um, it's not 914 00:59:53,210 --> 00:59:57,570 S1: that I think that these major platforms won't still have 915 00:59:57,570 --> 01:00:00,850 S1: a stronghold or whatever. It's just that I don't think 916 01:00:00,850 --> 01:00:03,850 S1: the implementation of their software is going to look the 917 01:00:03,850 --> 01:00:06,890 S1: same inside of all these different companies, and it could 918 01:00:06,890 --> 01:00:11,730 S1: be for a much smaller companies that aren't legacy. Their 919 01:00:11,730 --> 01:00:17,450 S1: software stack could look completely different from another startup, maybe 920 01:00:17,450 --> 01:00:20,770 S1: even doing the same sort of stuff just because the founders, 921 01:00:20,890 --> 01:00:23,850 S1: which they'll probably be very few, the founders will just 922 01:00:23,850 --> 01:00:27,190 S1: be like, yeah, I like this style. I like terminal style. 923 01:00:27,190 --> 01:00:30,270 S1: I don't like UIs, you know? I want everything to 924 01:00:30,310 --> 01:00:34,150 S1: be API based. You know, I really like the color purple. 925 01:00:34,430 --> 01:00:37,070 S1: I need purple and blue. I need the software to 926 01:00:37,110 --> 01:00:38,709 S1: look a certain way. I need it to act a 927 01:00:38,710 --> 01:00:41,110 S1: certain way. I want all the auditing and logging to 928 01:00:41,150 --> 01:00:43,310 S1: go a certain way. You know, I want to be 929 01:00:43,310 --> 01:00:46,830 S1: stood up only in these regions, whatever, whatever their preferences are, 930 01:00:47,390 --> 01:00:51,590 S1: and the preferences for the interface for all this stuff. Custom. 931 01:00:52,310 --> 01:00:56,150 S1: We're already seeing this with like custom replacements of tons 932 01:00:56,150 --> 01:00:58,350 S1: of different SaaS software. I mean, this is just all 933 01:00:58,350 --> 01:01:02,470 S1: over the news everywhere. It's really easy to create a 934 01:01:02,470 --> 01:01:05,510 S1: version of something now. It's a big difference between being 935 01:01:05,510 --> 01:01:09,190 S1: able to have that thing roll out, you know, enterprise wide, 936 01:01:09,190 --> 01:01:11,790 S1: and it's stable and it's secure and all of that. Right? 937 01:01:11,790 --> 01:01:15,189 S1: There's some work to be done there that'll take, you know, 938 01:01:15,230 --> 01:01:17,990 S1: a couple years to sort of work its way through. 939 01:01:17,990 --> 01:01:22,550 S1: But I think there's a very high chance that companies 940 01:01:23,070 --> 01:01:28,560 S1: and consumers will be making their own software even if 941 01:01:28,560 --> 01:01:31,600 S1: they're not making their own software. Let's be clear about that. 942 01:01:31,600 --> 01:01:34,560 S1: There's 8 billion people on the planet. Not everyone's making 943 01:01:34,560 --> 01:01:39,240 S1: their own software, but just just consider in 2019 or 2022, 944 01:01:39,520 --> 01:01:43,600 S1: the number of people who made software products was rounded 945 01:01:43,600 --> 01:01:47,480 S1: down to zero. Okay, that's the way to think about this. 946 01:01:47,480 --> 01:01:51,280 S1: Rounded down to zero. Everything in the App Store, every 947 01:01:51,280 --> 01:01:55,720 S1: software company rounded down to zero. If you multiply that 948 01:01:55,720 --> 01:02:00,600 S1: by 1000 or 10,000 or 1 million, that is a 949 01:02:00,640 --> 01:02:03,999 S1: lot more software that is. Plus, we're talking about the 950 01:02:04,000 --> 01:02:07,800 S1: ability for someone to speak to their AI agent, which 951 01:02:07,800 --> 01:02:10,800 S1: we've already talked about. Everyone's going to have one and say, hey, 952 01:02:10,840 --> 01:02:12,919 S1: I really wish I had a workout app. You know, 953 01:02:12,960 --> 01:02:16,039 S1: I've been using this one. You know, it's $19 a month. 954 01:02:16,040 --> 01:02:18,240 S1: It's pretty good, but I just wish it did this. 955 01:02:18,280 --> 01:02:20,280 S1: I wish it did this. I wish it did this. 956 01:02:20,800 --> 01:02:24,060 S1: And boom, it's now installed on your phone. It's on 957 01:02:24,060 --> 01:02:28,220 S1: your thing. The other one's uninstalled. The subscription is cancelled. 958 01:02:28,380 --> 01:02:31,019 S1: That same exact thing is going to happen inside the enterprise. 959 01:02:31,580 --> 01:02:33,780 S1: Now in the enterprise, it's going to take a little 960 01:02:33,780 --> 01:02:35,900 S1: bit longer because that thing needs to be integrated with 961 01:02:35,900 --> 01:02:38,580 S1: everything else. It's got to have the tie ins, it's 962 01:02:38,580 --> 01:02:41,100 S1: got to have procurement, it's got to have legal sign off. 963 01:02:41,300 --> 01:02:44,500 S1: Auditors have to sign off. There's a lot more workflow 964 01:02:44,500 --> 01:02:47,380 S1: involved in being able to do that. But I can 965 01:02:47,380 --> 01:02:51,860 S1: tell you this every CFO is looking at their software 966 01:02:51,860 --> 01:02:54,100 S1: list and saying, how can I cancel all of this? 967 01:02:54,300 --> 01:02:56,220 S1: It's just like the employee thing, how can I get 968 01:02:56,220 --> 01:02:59,500 S1: to zero employees? That's what the CFO is thinking as well. 969 01:02:59,500 --> 01:03:03,100 S1: How can I not pay any other company for software? 970 01:03:03,420 --> 01:03:07,060 S1: How can it all be ours? Everyone wants this. The 971 01:03:07,060 --> 01:03:10,860 S1: question is, you know, we couldn't do it before. We 972 01:03:10,860 --> 01:03:13,900 S1: couldn't do it before because it's hard to make software. 973 01:03:13,900 --> 01:03:15,260 S1: You got to maintain it. You got to do all 974 01:03:15,260 --> 01:03:19,300 S1: this stuff. Well, the better AI gets, the easier that gets. Okay, 975 01:03:19,760 --> 01:03:28,280 S1: So the tendency here is from common shared large few 976 01:03:28,320 --> 01:03:33,520 S1: companies producing software to everyone has their own and right. 977 01:03:33,520 --> 01:03:36,000 S1: And it's not going to go to the full extreme right. 978 01:03:36,040 --> 01:03:38,040 S1: It's going to I'm going to say it's going to 979 01:03:38,040 --> 01:03:41,240 S1: go to like a 90% extreme or an 80% extreme, 980 01:03:42,080 --> 01:03:48,480 S1: highly specialized software, everyone running their own personal AI stack 981 01:03:48,760 --> 01:03:52,999 S1: as a person, as a human, your AI system is 982 01:03:53,000 --> 01:03:58,200 S1: going to look, feel, act, behave differently than someone else. 983 01:03:58,200 --> 01:04:02,080 S1: They might even be using similar products, but your interfaces 984 01:04:02,080 --> 01:04:04,160 S1: are going to be different. The way it speaks to 985 01:04:04,160 --> 01:04:06,560 S1: you is going to be different. The news sources that 986 01:04:06,560 --> 01:04:09,200 S1: you're filtering is going to be different. And that takes 987 01:04:09,200 --> 01:04:12,480 S1: me to my next one, which Robert Putman talked about 988 01:04:12,480 --> 01:04:17,880 S1: in his book Bowling Alone in, I think, 2000. When 989 01:04:17,880 --> 01:04:22,890 S1: you have fragmentation like this across companies, but mostly across people. 990 01:04:23,130 --> 01:04:26,689 S1: When you have fragmentation like this, that's going to be profound. 991 01:04:27,530 --> 01:04:30,130 S1: So one of the things Putman talks about in his 992 01:04:30,290 --> 01:04:34,410 S1: book is the reason the country was so unified before 993 01:04:35,490 --> 01:04:38,610 S1: is that we were all watching the same TV shows. 994 01:04:38,610 --> 01:04:41,130 S1: We're all watching the same news. We all drove the 995 01:04:41,130 --> 01:04:44,090 S1: same cars, we all had the same watches. We all 996 01:04:44,090 --> 01:04:46,970 S1: went to the same churches. We all lived on the 997 01:04:46,970 --> 01:04:49,850 S1: same block. Maybe someone had a slightly bigger house, but 998 01:04:49,850 --> 01:04:53,130 S1: we're all living in the same sort of community and 999 01:04:53,170 --> 01:04:57,450 S1: everyone's consuming the same sources. Everyone's thinking similar things because 1000 01:04:57,450 --> 01:05:00,530 S1: we're reading the same newspapers, we're learning the same things 1001 01:05:00,530 --> 01:05:04,090 S1: in schools. Well, when you take what I just talked about, 1002 01:05:04,090 --> 01:05:08,570 S1: everyone gets custom software, everyone has custom AI, and in fact, 1003 01:05:08,810 --> 01:05:13,050 S1: you're not even viewing the sources. All these APIs, all 1004 01:05:13,050 --> 01:05:16,590 S1: these services that make up the world, you're not viewing him. 1005 01:05:16,590 --> 01:05:20,950 S1: There's millions of them. There's billions of them. Your AI's 1006 01:05:20,990 --> 01:05:23,470 S1: job is to consume all of that. Its job is 1007 01:05:23,470 --> 01:05:27,229 S1: to understand you and understand your needs. It consumes all 1008 01:05:27,230 --> 01:05:30,910 S1: those services, and then it gives you what you want. Well, 1009 01:05:30,910 --> 01:05:33,830 S1: what does that open up? It opens up the possibility 1010 01:05:34,030 --> 01:05:37,950 S1: that all of us will be having a different world experience. 1011 01:05:38,830 --> 01:05:44,309 S1: We believe the reality that we see. Okay. So when 1012 01:05:44,310 --> 01:05:48,669 S1: you watch a particular news source, if you watch particular 1013 01:05:48,670 --> 01:05:53,750 S1: YouTube channels, you watch particular podcasts or YouTube channels or 1014 01:05:54,070 --> 01:05:58,430 S1: TV shows or, you know, news channels or whatever, you 1015 01:05:58,430 --> 01:06:01,430 S1: think the world works that way. That is your world. 1016 01:06:01,430 --> 01:06:05,750 S1: That is reality to you. And if you watch this 1017 01:06:05,750 --> 01:06:09,229 S1: other set, which this person might live next door, they 1018 01:06:09,270 --> 01:06:14,350 S1: are in a completely different reality than you. They don't 1019 01:06:14,440 --> 01:06:18,840 S1: even know that this new, uh, dancing team from, uh, 1020 01:06:18,840 --> 01:06:21,959 S1: Korea is the most popular thing. And they got, uh, 1021 01:06:21,960 --> 01:06:25,480 S1: 20 billion views on the video that just came out yesterday. 1022 01:06:25,600 --> 01:06:27,919 S1: And to all of your friends, this is the most 1023 01:06:27,920 --> 01:06:30,840 S1: important thing that's ever happened. And you just can't believe it. 1024 01:06:30,840 --> 01:06:33,479 S1: You've watched it over and over. You cry every time 1025 01:06:33,480 --> 01:06:37,120 S1: you watch it. It's the most amazing thing ever. The 1026 01:06:37,120 --> 01:06:41,560 S1: person next door doesn't know it exists. In fact, they're 1027 01:06:41,560 --> 01:06:44,240 S1: not sure if Korea is a real country based on 1028 01:06:44,240 --> 01:06:47,560 S1: the media they're watching. They're like, yeah, that could just 1029 01:06:47,560 --> 01:06:53,120 S1: be made up. This fragmentation, because of custom. Everything is 1030 01:06:53,120 --> 01:07:00,560 S1: going to be massive right inside of companies. The dynamic 1031 01:07:00,560 --> 01:07:03,680 S1: here is very strange. How do you audit software. How 1032 01:07:03,680 --> 01:07:08,160 S1: do you do security scans. How do you do auditing. Um, 1033 01:07:08,160 --> 01:07:11,800 S1: how do you do legal compliance? It's a lot easier 1034 01:07:11,800 --> 01:07:16,900 S1: when everyone has SAP and everyone has Microsoft three. 65. 1035 01:07:17,020 --> 01:07:21,180 S1: And everyone has, you know, CrowdStrike and everyone has Palo 1036 01:07:21,180 --> 01:07:26,060 S1: Alto firewalls and checkpoint firewalls. Right. That's a whole lot 1037 01:07:26,100 --> 01:07:29,300 S1: easier because people have skill sets around those things. What 1038 01:07:29,300 --> 01:07:34,100 S1: happens when everything is custom, right. So the transition here 1039 01:07:34,100 --> 01:07:38,980 S1: roughly is few things that are packaged that most everyone 1040 01:07:38,980 --> 01:07:47,220 S1: uses to millions of options with infinite customization. And everyone 1041 01:07:47,220 --> 01:07:50,500 S1: is kind of using different stuff customized just for them, 1042 01:07:50,980 --> 01:07:56,260 S1: and they're experiencing different realities as a result. All right, 1043 01:07:56,260 --> 01:07:59,380 S1: final one. I've saved this one for last. So I 1044 01:07:59,380 --> 01:08:04,820 S1: think the kind of ultimate use case for AI is 1045 01:08:04,820 --> 01:08:09,019 S1: what I'm calling ideal state management or state management or 1046 01:08:09,060 --> 01:08:13,280 S1: state orchestration. Not really sure on the term. Doesn't really matter. 1047 01:08:13,920 --> 01:08:16,960 S1: The real term will be created probably in the next 1048 01:08:16,960 --> 01:08:24,080 S1: year or so. The idea is. What we've been doing 1049 01:08:24,080 --> 01:08:27,920 S1: inside of companies, what we've been doing as people, what 1050 01:08:27,920 --> 01:08:31,840 S1: we've been doing as a society is just kind of yoloing. 1051 01:08:32,200 --> 01:08:35,960 S1: This is what we've always done. This is what humanity does. 1052 01:08:36,200 --> 01:08:38,040 S1: We're like, hey, you know, I need to do better 1053 01:08:38,040 --> 01:08:40,800 S1: next time. I need to do you know, less of 1054 01:08:40,800 --> 01:08:43,880 S1: this thing that's really hurting me? Companies are like, hey, 1055 01:08:43,880 --> 01:08:46,640 S1: we should have goals. We should have OKRs. Hey, we 1056 01:08:46,640 --> 01:08:48,600 S1: should have a meeting. We should plan the next year 1057 01:08:48,600 --> 01:08:53,200 S1: or whatever. Those documents go somewhere. Maybe they get revised, 1058 01:08:53,200 --> 01:08:56,200 S1: maybe they don't. But what you do see, and again, 1059 01:08:56,200 --> 01:08:59,280 S1: I've been doing this for so long. If you track 1060 01:08:59,320 --> 01:09:02,680 S1: a company's goals, um, for an average company, like a 1061 01:09:02,680 --> 01:09:06,640 S1: medium sized company, which there are just, you know, hundreds 1062 01:09:06,640 --> 01:09:09,490 S1: of thousands of these things all over the world If 1063 01:09:09,490 --> 01:09:13,130 S1: you look at their goals and plans and metrics over 1064 01:09:13,130 --> 01:09:15,689 S1: the course of a year or five years or ten years, 1065 01:09:16,090 --> 01:09:19,730 S1: they're just making shit up like it changes constantly. The 1066 01:09:19,730 --> 01:09:25,010 S1: management changes like they, uh, they come out with a 1067 01:09:25,010 --> 01:09:27,810 S1: set of metrics. They don't hit them, they come out 1068 01:09:27,810 --> 01:09:30,850 S1: right after that and they're like, uh, yeah, guess what? 1069 01:09:30,850 --> 01:09:33,970 S1: We got a new plan. Uh, you know, we hired 1070 01:09:33,970 --> 01:09:36,490 S1: this person. It's going to be amazing. Here's new metrics, 1071 01:09:36,490 --> 01:09:38,930 S1: here's the new plan. And it's just kind of like 1072 01:09:39,090 --> 01:09:44,650 S1: constant reinvention, but not because of innovation. More so just 1073 01:09:44,650 --> 01:09:48,810 S1: because we're winging it. And this is no fault of anybody's. 1074 01:09:48,970 --> 01:09:51,850 S1: Everyone's doing it very, very smart. People are doing it. 1075 01:09:51,850 --> 01:09:55,650 S1: It's just the reality that we live in. Okay. This 1076 01:09:56,130 --> 01:09:59,850 S1: is one of the fundamental changes that AI is going 1077 01:09:59,850 --> 01:10:03,290 S1: to bring. I think what we're about to move into 1078 01:10:03,450 --> 01:10:07,910 S1: is a situation that moves away from this ad hoc 1079 01:10:08,150 --> 01:10:12,790 S1: YOLO sort of situation to the ultimate use case for AI, 1080 01:10:12,830 --> 01:10:18,670 S1: which is state management. And state management starts with defining 1081 01:10:18,670 --> 01:10:22,350 S1: what ideal state is. This is a thing that most 1082 01:10:22,350 --> 01:10:25,430 S1: companies do not have. They they do not have an 1083 01:10:25,430 --> 01:10:32,430 S1: articulated statement, kind of like a PRD document which has 1084 01:10:32,430 --> 01:10:37,070 S1: multiple SOPs revolving around it that says, here is our 1085 01:10:37,110 --> 01:10:40,310 S1: actual mission, here are our actual goals, here are the 1086 01:10:40,310 --> 01:10:43,390 S1: problems we are trying to solve in the world, which 1087 01:10:43,390 --> 01:10:46,190 S1: is why this is our mission. Here are our goals. 1088 01:10:46,590 --> 01:10:49,870 S1: Here are our challenges. These are the things that are, 1089 01:10:49,910 --> 01:10:52,470 S1: you know, we can't hire fast enough. You know, we 1090 01:10:52,470 --> 01:10:55,630 S1: can't ship product fast enough. We keep getting beaten by 1091 01:10:55,630 --> 01:10:58,390 S1: our competitors. We need to move into Asia. We're not 1092 01:10:58,390 --> 01:11:00,790 S1: there yet. Like you put all your challenges, you got 1093 01:11:00,790 --> 01:11:05,070 S1: your risk register, you have your projects, you have your budget. 1094 01:11:05,070 --> 01:11:10,840 S1: You have your people and this is your unified document. 1095 01:11:10,840 --> 01:11:15,200 S1: This is your your grail. This is your system. This 1096 01:11:15,200 --> 01:11:20,200 S1: is your algorithm for what? What you're chasing. Okay? This 1097 01:11:20,200 --> 01:11:26,040 S1: being locked into the core DNA of the company and 1098 01:11:26,080 --> 01:11:31,760 S1: everything revolving around that. So I talked about all this infrastructure, this, um, 1099 01:11:32,480 --> 01:11:37,840 S1: graph of algorithms, that graph of algorithms needs to be 1100 01:11:38,840 --> 01:11:45,720 S1: feeding this system, this ideal state, this perfectly articulated and 1101 01:11:45,720 --> 01:11:50,800 S1: not perfectly. Nothing's gonna be perfect. This very well articulated 1102 01:11:51,040 --> 01:11:55,560 S1: ideal state is about to become the most important thing 1103 01:11:56,320 --> 01:12:02,480 S1: for companies, but also for anything, for organizations, for entities, 1104 01:12:02,800 --> 01:12:05,540 S1: for people. This is a big thing that I work 1105 01:12:05,540 --> 01:12:08,700 S1: on in my private work. If the Pi system with 1106 01:12:08,700 --> 01:12:11,180 S1: my Telos system, like all the stuff that we're doing 1107 01:12:11,180 --> 01:12:14,580 S1: in my community, we think about ideal state from the 1108 01:12:14,580 --> 01:12:22,780 S1: very beginning, ideal state. This concept is extraordinary. It's extraordinarily powerful. Okay. 1109 01:12:22,820 --> 01:12:25,700 S1: I've been using it for a long time. Not this 1110 01:12:25,740 --> 01:12:29,060 S1: new version that I have now, but roughly I've been 1111 01:12:29,060 --> 01:12:33,180 S1: using a system like this for probably ten years and 1112 01:12:33,380 --> 01:12:36,500 S1: there wasn't any AI. You don't need AI to sort 1113 01:12:36,500 --> 01:12:40,380 S1: of think about, articulate this and start moving towards it. 1114 01:12:40,380 --> 01:12:45,300 S1: It's very powerful. Forget any tech, just note cards, index cards, 1115 01:12:45,340 --> 01:12:48,460 S1: a space pen, you're off and you're going to have 1116 01:12:48,460 --> 01:12:52,100 S1: massive benefits from doing this. Here's where the AI comes in. 1117 01:12:52,100 --> 01:12:55,420 S1: And this this is like the sickest, most powerful, most 1118 01:12:55,420 --> 01:13:00,540 S1: exciting thing in all of AI across all AI ideas 1119 01:13:01,060 --> 01:13:05,670 S1: that I've ever heard about or come up with or whatever. Okay, 1120 01:13:05,710 --> 01:13:08,670 S1: so check this out. This is the universal game. When 1121 01:13:08,670 --> 01:13:12,750 S1: we meet aliens flying around the galaxy or whatever, and 1122 01:13:12,750 --> 01:13:15,110 S1: we tell them this thing I'm about to tell you, 1123 01:13:15,150 --> 01:13:19,110 S1: they're going to be like, yeah, that's what everyone does. 1124 01:13:19,870 --> 01:13:21,390 S1: How do you think we got here? How do you 1125 01:13:21,390 --> 01:13:23,350 S1: think we have all these spaceships? How do you think 1126 01:13:23,350 --> 01:13:25,510 S1: we have all these planets? How do you think we 1127 01:13:25,510 --> 01:13:29,510 S1: have Dyson spheres around all these stars? Like that's obvious. 1128 01:13:29,550 --> 01:13:32,310 S1: That is the algorithm, right? So this is what I 1129 01:13:32,310 --> 01:13:35,830 S1: call it. I call it the algorithm. The algorithm is 1130 01:13:37,270 --> 01:13:42,750 S1: we have ideal state. What is our current state okay. 1131 01:13:43,070 --> 01:13:46,110 S1: Step two what is our current state. So remember what 1132 01:13:46,110 --> 01:13:49,390 S1: I was talking about before about you have the graph 1133 01:13:49,390 --> 01:13:53,230 S1: of algorithms. You have asset management gets elevated to the top. 1134 01:13:54,070 --> 01:13:57,950 S1: This current state. It's the job of your AI and 1135 01:13:57,950 --> 01:14:01,090 S1: all the different agents and all the different systems. It's 1136 01:14:01,090 --> 01:14:03,970 S1: not all AI. This is deterministic code as well, right? 1137 01:14:04,290 --> 01:14:08,050 S1: What is the current snapshot of the ORC? What is 1138 01:14:08,050 --> 01:14:11,689 S1: the current state of our problems and our solutions and 1139 01:14:11,730 --> 01:14:15,490 S1: our products and our services? How happy are our people? 1140 01:14:15,970 --> 01:14:20,010 S1: How happy are our customers? What is the current churn number? 1141 01:14:20,290 --> 01:14:23,010 S1: What is the current products we're about to release? What 1142 01:14:23,010 --> 01:14:32,130 S1: are our competitors about to release? What are the market conditions? Snapshot. Snapshot. Snapshot. 1143 01:14:32,330 --> 01:14:36,290 S1: Ideal state. Ideal state is here. We're in Asia. We 1144 01:14:36,290 --> 01:14:40,970 S1: have 30% more employees, actually more likely 30% less employees. 1145 01:14:40,970 --> 01:14:45,130 S1: But separate topic. Um, you know, we want to be 1146 01:14:45,130 --> 01:14:47,570 S1: in these markets. We need to have more services. We 1147 01:14:47,570 --> 01:14:51,410 S1: need to have more private luncheon learns, uh, in our 1148 01:14:51,410 --> 01:14:54,930 S1: local community, whatever it is, all that's going to be 1149 01:14:55,330 --> 01:14:57,970 S1: hundreds of thousands of things or hundreds of thousands or 1150 01:14:58,190 --> 01:15:01,670 S1: millions of things. Whatever it is, our current state also 1151 01:15:01,670 --> 01:15:04,070 S1: has millions of things in it, right? Because it's going 1152 01:15:04,070 --> 01:15:08,390 S1: to go down to like eventually at some, some level. 1153 01:15:09,430 --> 01:15:14,910 S1: Here's the current open ports for all of our Kubernetes pods, right. 1154 01:15:15,110 --> 01:15:19,910 S1: All of our AWS infrastructure. Current state, current state over here, 1155 01:15:19,910 --> 01:15:24,830 S1: they're all locked down, nothing publicly available, strong authentication on them, etc.. 1156 01:15:25,750 --> 01:15:31,430 S1: Now here it is. Watch this. The role of AI 1157 01:15:32,390 --> 01:15:39,430 S1: is continuous migration continuous gap closing between current and ideal state. 1158 01:15:41,430 --> 01:15:47,590 S1: This works as an individual trying to lose 20 kilos 1159 01:15:48,950 --> 01:15:52,390 S1: which I'm trying to do. An individual trying to lose 1160 01:15:52,390 --> 01:15:57,240 S1: 20 kilos. Trying to increase their Vo two Max trying 1161 01:15:57,240 --> 01:16:02,600 S1: to find a wife trying to get their art exhibited 1162 01:16:02,600 --> 01:16:07,360 S1: in the local gallery and eventually at MoMA in San Francisco, 1163 01:16:07,920 --> 01:16:13,400 S1: trying to come up with, uh, new EDM tracks so 1164 01:16:13,440 --> 01:16:16,080 S1: they can go to EDC and actually play live on 1165 01:16:16,080 --> 01:16:21,880 S1: stage trying to run a federation of planets. Okay. If 1166 01:16:21,880 --> 01:16:24,360 S1: we're talking about, you know, 9000 years in the future 1167 01:16:24,360 --> 01:16:27,840 S1: or whatever, anything you're trying to do at any scale 1168 01:16:28,120 --> 01:16:32,320 S1: can be managed by ideal state and current state and 1169 01:16:32,320 --> 01:16:40,080 S1: the migration. This is extremely, extremely powerful. I think we're 1170 01:16:40,080 --> 01:16:46,880 S1: about to apply this to almost everything. I think prds engineering. 1171 01:16:47,920 --> 01:16:51,000 S1: So I have this thing inside of the Pi platform. 1172 01:16:51,200 --> 01:16:57,060 S1: It's called the algorithm. I literally start the algorithm by 1173 01:16:57,100 --> 01:17:00,780 S1: decomposing the prompt that comes in and turning it into 1174 01:17:00,820 --> 01:17:05,620 S1: reverse engineered ideal state components. I break out the pieces, 1175 01:17:05,620 --> 01:17:08,140 S1: I look at the context from the user, and I 1176 01:17:08,140 --> 01:17:11,300 S1: do deep analysis, and we do research as well. This 1177 01:17:11,300 --> 01:17:13,540 S1: is if you have a lot of time, do research 1178 01:17:13,540 --> 01:17:15,700 S1: and you figure out, okay, what do they actually mean 1179 01:17:15,700 --> 01:17:18,580 S1: by this? Okay, I might have sent in a single 1180 01:17:18,580 --> 01:17:21,780 S1: one sentence prompt, but I do all this research to 1181 01:17:21,820 --> 01:17:25,260 S1: break it into its pieces. What is he explicitly say 1182 01:17:25,540 --> 01:17:29,740 S1: he wants? What is he explicitly say he doesn't want? Okay. 1183 01:17:30,020 --> 01:17:32,620 S1: What is he maybe mean by that? What are some 1184 01:17:32,620 --> 01:17:36,460 S1: common gotchas for someone trying to build a system like this? 1185 01:17:36,460 --> 01:17:40,380 S1: All that gets decomposed, reverse engineered and I start creating 1186 01:17:40,380 --> 01:17:45,820 S1: ideal state criteria. These actually go into cloud code tasks, right? 1187 01:17:46,020 --> 01:17:48,940 S1: The task system from cloud code, which is extraordinary. I 1188 01:17:48,940 --> 01:17:51,260 S1: put them all in there. Then they go into the 1189 01:17:51,300 --> 01:17:54,270 S1: PRD and the PRD is what we work off of, 1190 01:17:54,470 --> 01:17:58,310 S1: but it's all working off of ideal state so engineering 1191 01:17:58,310 --> 01:18:01,430 S1: can be run this way. Companies can be run this way, 1192 01:18:01,790 --> 01:18:09,110 S1: you managing your life, you managing your entire family's happiness. Okay. 1193 01:18:09,390 --> 01:18:11,830 S1: My husband, you know he needs to be more healthy. 1194 01:18:11,990 --> 01:18:15,430 S1: My kids, they need to have better education. They need 1195 01:18:15,430 --> 01:18:18,270 S1: more tutoring in this. Hey, I'm looking at the dashboard. 1196 01:18:18,310 --> 01:18:20,910 S1: Looks like they haven't had enough vegetables. Okay, I'm going 1197 01:18:20,950 --> 01:18:23,750 S1: to tell my AI we're going to order more vegetables. Hey, 1198 01:18:23,790 --> 01:18:26,950 S1: we haven't gone on a vacation together in a while. Hey, um, 1199 01:18:26,950 --> 01:18:29,470 S1: I noticed, uh, we were looking at our phones a 1200 01:18:29,470 --> 01:18:32,390 S1: little bit too much at the dinner table last few days. 1201 01:18:32,470 --> 01:18:35,710 S1: My AI notified me about that, so. Yeah, let's make 1202 01:18:35,750 --> 01:18:40,270 S1: sure we're all looking in each other's faces and talking about, uh, 1203 01:18:40,270 --> 01:18:43,230 S1: physical art. Let's make sure we do a number of 1204 01:18:43,230 --> 01:18:47,510 S1: vacations during the year where there's no tech involved whatsoever. 1205 01:18:47,550 --> 01:18:52,330 S1: We're focusing on human things and all of this. Again, 1206 01:18:52,890 --> 01:18:56,130 S1: the AI is managing this stuff for you because you 1207 01:18:56,130 --> 01:18:59,650 S1: put it in your ideal state. Okay, so if you're 1208 01:18:59,650 --> 01:19:03,290 S1: running an ice cream truck business, you're running a federation 1209 01:19:03,290 --> 01:19:06,570 S1: of planets, or you're just trying to like, find a 1210 01:19:06,570 --> 01:19:09,730 S1: girlfriend or a boyfriend. All of this can be managed 1211 01:19:09,730 --> 01:19:11,930 S1: the same way. Same way as if you're trying to 1212 01:19:11,970 --> 01:19:16,610 S1: launch a new SaaS replacement app, or a new workout app, 1213 01:19:17,570 --> 01:19:20,730 S1: or whatever you're trying to do with AI. Everything can 1214 01:19:20,730 --> 01:19:23,170 S1: be managed by thinking about it in this way. Now 1215 01:19:23,170 --> 01:19:26,130 S1: there's a lot of steps in between, but those steps 1216 01:19:26,130 --> 01:19:28,890 S1: can be scaffolded. Those steps can be engineered, those steps 1217 01:19:28,890 --> 01:19:32,090 S1: can be filled in. And the smarter the AI gets, 1218 01:19:32,370 --> 01:19:34,490 S1: the more it's going to be able to just know 1219 01:19:34,970 --> 01:19:37,330 S1: how to help you transition and build plans to help 1220 01:19:37,330 --> 01:19:41,650 S1: you transition. So I think this is potentially the biggest 1221 01:19:41,650 --> 01:19:44,810 S1: idea in AI. I don't think many people are talking 1222 01:19:44,810 --> 01:19:47,810 S1: about it yet, but I think this concept of ideal 1223 01:19:47,810 --> 01:19:50,670 S1: state is is is about to be huge. And in fact, 1224 01:19:50,670 --> 01:19:54,870 S1: I think Agentic platforms are about to start building it 1225 01:19:54,870 --> 01:19:57,750 S1: in to like everything they do. Wouldn't be surprised to 1226 01:19:57,750 --> 01:20:04,710 S1: see this cloud code Codex Gemini start to move towards 1227 01:20:04,910 --> 01:20:10,630 S1: AGI generality by doing this reverse engineering of requests into 1228 01:20:10,630 --> 01:20:14,790 S1: an ideal state that you then pursue. Because here's the trick. 1229 01:20:15,390 --> 01:20:20,990 S1: And I got this from Andrej Karpathy you can't hill climb. 1230 01:20:21,510 --> 01:20:29,070 S1: You can't progress towards something if you don't have failures, 1231 01:20:29,070 --> 01:20:31,750 S1: if you don't have a thing to hill climb against. 1232 01:20:32,070 --> 01:20:35,430 S1: And this is why I reverse engineer everything into ideal 1233 01:20:35,430 --> 01:20:40,189 S1: state criteria, because they are also my verification criteria. So 1234 01:20:40,390 --> 01:20:44,830 S1: what Karpathy said was previous software was you can make 1235 01:20:44,830 --> 01:20:50,160 S1: anything the next generation of software will be you can 1236 01:20:50,160 --> 01:20:54,240 S1: verify anything and I think it's very powerful. So my 1237 01:20:54,240 --> 01:20:59,000 S1: ideal state criteria in the algorithm within, uh, personal AI infrastructure, 1238 01:20:59,000 --> 01:21:02,960 S1: the open source project, it turns the ideal state into 1239 01:21:02,960 --> 01:21:07,560 S1: the verification criteria. And they're all discrete and they're all verifiable, 1240 01:21:07,560 --> 01:21:10,400 S1: yes or no. And that's what gets us to this 1241 01:21:10,400 --> 01:21:14,680 S1: verifiability thing that Andre was talking about. And that's what 1242 01:21:14,720 --> 01:21:19,280 S1: ultimately allows us to go from current state to ideal state. 1243 01:21:19,480 --> 01:21:22,720 S1: So really excited about this. I know this was tons 1244 01:21:22,720 --> 01:21:26,400 S1: of content and I just want you to think about it. 1245 01:21:26,600 --> 01:21:30,120 S1: It's a set of transitions. It's not just one transition. 1246 01:21:30,120 --> 01:21:32,000 S1: So I consider all of this to be part of 1247 01:21:32,000 --> 01:21:35,720 S1: the great transition. Hopefully it's a mental model. I know 1248 01:21:35,720 --> 01:21:39,160 S1: it includes lots of subset things, but I think all 1249 01:21:39,160 --> 01:21:43,000 S1: of it combined does produce a single mental model. So 1250 01:21:43,000 --> 01:21:46,900 S1: hopefully you can now see when all this news comes out, right? 1251 01:21:46,940 --> 01:21:52,460 S1: New models, new products, new services. You know, um, Co-work 1252 01:21:52,500 --> 01:21:54,900 S1: can now do this. Oh, it's it's it's going after 1253 01:21:54,900 --> 01:21:57,900 S1: the lawyers. It's going after whatever. What my goal is 1254 01:21:57,900 --> 01:21:59,860 S1: here is for you to be able to say, well, 1255 01:21:59,860 --> 01:22:04,020 S1: hold on, you know, based on this great transition model, 1256 01:22:04,220 --> 01:22:07,340 S1: that thing that's happening is not something, something new. It's 1257 01:22:07,380 --> 01:22:10,140 S1: not going to produce anxiety. Right? I'm trying to reduce 1258 01:22:10,140 --> 01:22:14,420 S1: your anxiety of thinking about all this change, because I 1259 01:22:14,420 --> 01:22:17,620 S1: believe this great transition sort of model, with all the 1260 01:22:17,620 --> 01:22:20,820 S1: little submodels and transitions that I was talking about, it's 1261 01:22:20,860 --> 01:22:24,100 S1: kind of like a container. It's a container. It's a 1262 01:22:24,100 --> 01:22:27,500 S1: mental model container that can reduce your anxiety if you 1263 01:22:27,500 --> 01:22:31,939 S1: understand this, in my opinion, if you understand this, then 1264 01:22:31,940 --> 01:22:35,020 S1: you just won't be surprised by things, right? And I 1265 01:22:35,020 --> 01:22:37,820 S1: can be surprised. Like I could have missed something, right? 1266 01:22:37,860 --> 01:22:40,780 S1: Obviously I could be wrong about any one of these 1267 01:22:40,780 --> 01:22:43,670 S1: or more, but in general, I don't think I'm going 1268 01:22:43,710 --> 01:22:46,230 S1: to be surprised by this. I put the stuff into 1269 01:22:46,230 --> 01:22:49,710 S1: the book in 2016. It is coming true. I think 1270 01:22:49,710 --> 01:22:52,270 S1: it's a natural progression for the stuff to be happening. 1271 01:22:52,510 --> 01:22:54,670 S1: Like who knows how it's going to happen, when it's 1272 01:22:54,670 --> 01:22:56,270 S1: going to happen, which company is going to do it. 1273 01:22:56,270 --> 01:22:59,150 S1: You can't predict that stuff. It's impossible to predict, but 1274 01:22:59,150 --> 01:23:03,750 S1: the direction I think is possible to to see where 1275 01:23:03,750 --> 01:23:09,110 S1: it's potentially going. And that gives you a container, a 1276 01:23:09,110 --> 01:23:12,590 S1: mental model to be like, hmm, okay, this all sort 1277 01:23:12,590 --> 01:23:16,310 S1: of makes sense inside of this framework. So you're not 1278 01:23:16,310 --> 01:23:20,150 S1: spooked out by like all these new models coming out, 1279 01:23:21,070 --> 01:23:23,910 S1: new products coming out. It's like, oh, this person's going 1280 01:23:23,950 --> 01:23:28,110 S1: to lose their job, that person. These industries are being disrupted. No. 1281 01:23:28,150 --> 01:23:30,510 S1: If you have this mental model, I think you're more 1282 01:23:30,510 --> 01:23:33,310 S1: likely to just be like, ah, yeah, that seems to 1283 01:23:33,310 --> 01:23:36,310 S1: be the way it's going. And, uh, you can buckle 1284 01:23:36,310 --> 01:23:39,190 S1: down and just move forward. So hopefully this has been 1285 01:23:39,190 --> 01:23:41,910 S1: helpful and, uh, we'll see you in the next one.