WEBVTT 00:00:00.000 --> 00:00:02.319 I want you to picture picture something for a 00:00:02.319 --> 00:00:05.519 second. For pretty much the entire history of 00:00:05.519 --> 00:00:08.179 our species, intelligence, you know, the ability 00:00:08.179 --> 00:00:11.279 to think, plan and solve really complex problems 00:00:11.279 --> 00:00:13.820 has been our scarcest resource. Absolutely. It's 00:00:13.820 --> 00:00:16.039 expensive to hire. It takes decades to train. 00:00:16.300 --> 00:00:18.980 And it is effectively the gold standard of our 00:00:18.980 --> 00:00:21.559 economy. But what happens to the architecture 00:00:21.559 --> 00:00:23.679 of the world when human intelligence suddenly 00:00:23.679 --> 00:00:27.570 becomes cheap? I don't mean just affordable. 00:00:27.629 --> 00:00:30.030 I mean abundant, commodity cheap. That is the 00:00:30.030 --> 00:00:31.929 question, isn't it? And if you look at the data 00:00:31.929 --> 00:00:34.429 we have today, the answer is, well, it's uncomfortable. 00:00:34.670 --> 00:00:36.890 It doesn't just change what jobs we do. It breaks 00:00:36.890 --> 00:00:39.789 the fundamental economic logic of modern society. 00:00:40.229 --> 00:00:42.210 We're talking about what some are calling the 00:00:42.210 --> 00:00:45.759 2028 global intelligence crisis. Welcome back 00:00:45.759 --> 00:00:47.380 to the Deep Dive. I'm really glad you're here 00:00:47.380 --> 00:00:49.320 with us today. This one feels a little different. 00:00:49.380 --> 00:00:51.520 We aren't just looking at new tools or shiny 00:00:51.520 --> 00:00:54.859 updates. We are looking at a shift from AI tools 00:00:54.859 --> 00:00:59.299 to AI realities. We have a stack of sources today 00:00:59.299 --> 00:01:01.439 that paint a picture of a world changing very, 00:01:01.500 --> 00:01:04.510 very fast. We really do. The roadmap for today 00:01:04.510 --> 00:01:07.069 is intense. We're going to start with Google's 00:01:07.069 --> 00:01:09.209 new rollout. It's something called, and I apologize 00:01:09.209 --> 00:01:12.629 in advance for this, Nano Banana 2. Right. Which 00:01:12.629 --> 00:01:15.569 effectively turns every single user into a reality 00:01:15.569 --> 00:01:19.810 editor. Then we need to look at agentic AI. Specifically, 00:01:20.170 --> 00:01:22.810 Perplexity's new computer framework that can 00:01:22.810 --> 00:01:25.859 work for months. without sleeping months that 00:01:25.859 --> 00:01:28.459 is kind of hard to process it is then we've got 00:01:28.459 --> 00:01:31.120 a geopolitical lightning round involving saudi 00:01:31.120 --> 00:01:34.420 money russian disinformation and a retired ai 00:01:34.420 --> 00:01:37.299 that is literally writing essays for fun and 00:01:37.299 --> 00:01:40.000 finally we are going to unpack that viral theory 00:01:40.000 --> 00:01:42.359 you mentioned in the open the economic erosion 00:01:42.359 --> 00:01:45.159 coming in 2028. so let's just get right into 00:01:45.159 --> 00:01:48.069 it google nano banana 2. This is the new default 00:01:48.069 --> 00:01:50.430 image model inside Gemini. Now, usually when 00:01:50.430 --> 00:01:52.269 we talk about image generators, I think of prompting, 00:01:52.269 --> 00:01:54.010 you know, typing a cat on a skateboard and getting 00:01:54.010 --> 00:01:56.209 a picture. Right. But the sources suggest this 00:01:56.209 --> 00:01:58.250 is different. It's not just generation. It's 00:01:58.250 --> 00:02:00.329 integration. Yeah, it's moving from creation 00:02:00.329 --> 00:02:02.750 to editing. That's the key distinction here. 00:02:03.030 --> 00:02:05.530 Google has embedded this across their entire 00:02:05.530 --> 00:02:08.699 ecosystem. It combines features from earlier 00:02:08.699 --> 00:02:10.860 versions but pushes the editing capabilities 00:02:10.860 --> 00:02:13.259 so much further. It's faster than the pro version, 00:02:13.419 --> 00:02:16.199 sure, but the real power is that it can edit 00:02:16.199 --> 00:02:19.159 real photos directly. And it does it with a deep 00:02:19.159 --> 00:02:21.419 semantic understanding of what's actually in 00:02:21.419 --> 00:02:23.599 the image. I want to push on that a bit. Yeah. 00:02:23.620 --> 00:02:25.680 Because we've had Photoshop for 30 years. Sure. 00:02:25.780 --> 00:02:28.199 We've had content -aware fill for a while now. 00:02:28.860 --> 00:02:31.159 Why is this significant? Is it just a better 00:02:31.159 --> 00:02:34.000 airbrush? No. It's the difference between painting 00:02:34.000 --> 00:02:36.500 over pixels and rewriting the code of reality. 00:02:36.979 --> 00:02:40.400 Photoshop manipulates pixels. Nanobanana 2 manipulates 00:02:40.400 --> 00:02:43.139 concepts. Like stacking Lego blocks of data. 00:02:43.500 --> 00:02:45.379 For example, the source highlights its ability 00:02:45.379 --> 00:02:48.780 to pull live web data directly into visual infographics. 00:02:48.900 --> 00:02:51.719 Wait, live data? Yes. So you aren't just making 00:02:51.719 --> 00:02:54.139 a static image. You could theoretically take 00:02:54.139 --> 00:02:57.099 a photo of a street. and ask it to overlay real 00:02:57.099 --> 00:03:00.919 -time traffic data or restaurant ratings or change 00:03:00.919 --> 00:03:02.759 the signage to reflect a different language, 00:03:02.840 --> 00:03:06.080 not by pasting a flat layer on top, but by regenerating 00:03:06.080 --> 00:03:08.219 the image to look like that information was always 00:03:08.219 --> 00:03:11.080 physically there. It's making a dynamic representation 00:03:11.080 --> 00:03:14.099 of current info. That is useful. I can totally 00:03:14.099 --> 00:03:17.159 see the utility there. But there is a flip side 00:03:17.159 --> 00:03:19.500 that the sources touch on. They mention you can 00:03:19.500 --> 00:03:22.659 modify people into entirely new situations. Yeah. 00:03:22.780 --> 00:03:25.340 This is the scary cool factor. We were talking 00:03:25.340 --> 00:03:28.280 about modifying people, backgrounds, context. 00:03:28.520 --> 00:03:30.439 Right. Essentially acting as a reality editor. 00:03:30.860 --> 00:03:32.599 You make a photo of a protest or a celebration 00:03:32.599 --> 00:03:34.919 or a meeting with a few prompts. You change who 00:03:34.919 --> 00:03:37.280 was there. You change the mood. You change the 00:03:37.280 --> 00:03:40.000 narrative of that image completely. And because 00:03:40.000 --> 00:03:42.020 it's Google, this isn't some niche tool for graphic 00:03:42.020 --> 00:03:43.759 designers. It's in the phone. It's in the browser. 00:03:44.090 --> 00:03:46.590 The sources did note it's not perfect yet, though. 00:03:46.669 --> 00:03:49.669 I saw mentions of incorrect real -time data being 00:03:49.669 --> 00:03:52.030 pulled from the web, which sounds like a classic 00:03:52.030 --> 00:03:55.270 hallucination problem. And visually, sometimes 00:03:55.270 --> 00:03:58.949 faces look slightly pasted on. Or it exaggerates 00:03:58.949 --> 00:04:02.069 age and features. So it's not magic yet. True. 00:04:02.210 --> 00:04:04.930 It's not fully invisible. But the gap is closing 00:04:04.930 --> 00:04:07.340 rapidly. The source concludes that while you 00:04:07.340 --> 00:04:09.319 can still spot the fakes if you look closely, 00:04:09.500 --> 00:04:11.780 maybe the lighting on the ear is wrong or the 00:04:11.780 --> 00:04:14.379 text is slightly warped, it makes it extremely 00:04:14.379 --> 00:04:17.620 easy to create images of events that never happened. 00:04:18.339 --> 00:04:21.040 We are moving past asking, is this a good drawing, 00:04:21.180 --> 00:04:23.740 to asking, did this actually occur? So if the 00:04:23.740 --> 00:04:26.240 photo is no longer proof of the event, what becomes 00:04:26.240 --> 00:04:28.660 the new standard for truth? Verified metadata. 00:04:28.939 --> 00:04:31.439 Trust shifts from what we see. to who signed 00:04:31.439 --> 00:04:33.439 the file. Okay, let's pivot. We've established 00:04:33.439 --> 00:04:35.540 that we can edit reality. Now let's talk about 00:04:35.540 --> 00:04:38.019 who or what is actually doing the work. You mentioned 00:04:38.019 --> 00:04:40.920 Perplexity's new computer. Oh man, this is where 00:04:40.920 --> 00:04:44.000 the utility curve actually shifts. Perplexity 00:04:44.000 --> 00:04:46.920 launched computer. It's a multi -model AI agent, 00:04:47.079 --> 00:04:50.060 software that uses many AI brains to solve one 00:04:50.060 --> 00:04:53.500 complex problem. Imagine scaling to a billion 00:04:53.500 --> 00:04:56.490 queries without a human in the loop. It orchestrates 00:04:56.490 --> 00:05:00.089 19 different models to browse, code, and run 00:05:00.089 --> 00:05:02.889 tasks nonstop for months. I need to stop you 00:05:02.889 --> 00:05:06.269 on 19 models. Why 19? Is it just throwing everything 00:05:06.269 --> 00:05:08.790 at the wall to see what sticks? It's specialization. 00:05:09.029 --> 00:05:11.089 Think of it like a construction site. You don't 00:05:11.089 --> 00:05:13.410 just have one person who does plumbing, electrical, 00:05:13.670 --> 00:05:16.029 framing, and architecture. You have a specialized 00:05:16.029 --> 00:05:19.459 team. Perplexity is orchestrating a team. One 00:05:19.459 --> 00:05:21.500 model might be excellent at reasoning, another 00:05:21.500 --> 00:05:24.220 at coding in Python, another at summarizing web 00:05:24.220 --> 00:05:26.639 searches. The computer acts as the project manager, 00:05:26.839 --> 00:05:29.360 routing the subtasks to the absolute best brain 00:05:29.360 --> 00:05:31.959 for the job. I see. So rather than me chatting 00:05:31.959 --> 00:05:34.040 with a bot and trying to coax a good answer out 00:05:34.040 --> 00:05:36.160 of it, I give a high -level goal and it manages 00:05:36.160 --> 00:05:38.579 the team. Exactly. And the months part is the 00:05:38.579 --> 00:05:40.500 real kicker here. Yeah, explain that, because 00:05:40.500 --> 00:05:42.720 most agents I've played with get stuck in a loop 00:05:42.720 --> 00:05:45.189 after three steps. I mean, I still wrestle with 00:05:45.189 --> 00:05:47.430 prompt drift myself. They start repeating themselves 00:05:47.430 --> 00:05:50.490 or just crash. How does this run for months? 00:05:50.889 --> 00:05:53.350 That's the breakthrough in autonomy. It's designed 00:05:53.350 --> 00:05:55.889 for long haul tasks. Let's say you want to research 00:05:55.889 --> 00:05:58.350 every single patent filed in a renewable energy 00:05:58.350 --> 00:06:01.029 sector in the last five years, categorize them 00:06:01.029 --> 00:06:03.769 and cross -reference them with stock performance. 00:06:03.970 --> 00:06:06.610 That's a massive job. That takes weeks for a 00:06:06.610 --> 00:06:08.730 human. This agent can break that down, execute 00:06:08.730 --> 00:06:11.310 it day and night, handle errors like if a website 00:06:11.310 --> 00:06:14.149 is down, it waits and retries, and keep going 00:06:14.149 --> 00:06:16.810 until the job is done. It changes the dynamic 00:06:16.810 --> 00:06:20.810 from using a tool to managing a workforce. Precisely. 00:06:21.209 --> 00:06:23.889 And it's not just perplexity. Claude, the model 00:06:23.889 --> 00:06:25.970 from Anthropic, has added Claude Cowork. This 00:06:25.970 --> 00:06:28.149 allows for scheduled tasks. You can tell it to 00:06:28.149 --> 00:06:30.300 run something automatically next Tuesday. like 00:06:30.300 --> 00:06:32.279 posting updates or sending Slack messages. So 00:06:32.279 --> 00:06:34.160 it's working while you sleep. It works while 00:06:34.160 --> 00:06:35.879 you sleep, while you eat, while you're on a vacation. 00:06:36.100 --> 00:06:38.420 This raises a huge liability question for me. 00:06:38.540 --> 00:06:41.899 When AI runs for months unmonitored, who owns 00:06:41.899 --> 00:06:44.199 the mistake when it drifts off course? The human 00:06:44.199 --> 00:06:47.220 manager. We move from creators to accountability 00:06:47.220 --> 00:06:49.639 holders. You're the one who signed off on the 00:06:49.639 --> 00:06:51.920 mission. Let's zoom out a bit. We have these 00:06:51.920 --> 00:06:55.639 powerful agents and we have reality editing tools. 00:06:55.800 --> 00:06:58.319 Naturally, this requires massive resources. The 00:06:58.319 --> 00:07:00.680 money moving around right now is staggering. 00:07:00.899 --> 00:07:03.600 It's astronomical. The sources highlight a new 00:07:03.600 --> 00:07:06.019 fund from Saudi Arabia. They aren't just dipping 00:07:06.019 --> 00:07:10.439 a toe in. They are launching a $100 billion AI 00:07:10.439 --> 00:07:13.420 fund. $100 billion. Put that in context for me. 00:07:13.500 --> 00:07:15.800 Is that a lot in the grand scheme of tech investing? 00:07:16.120 --> 00:07:18.939 It's massive. That matches the global venture 00:07:18.939 --> 00:07:22.339 funding for AI for all of 2024. Yeah. In one 00:07:22.339 --> 00:07:25.759 single fund. Their goal is explicit. They want 00:07:25.759 --> 00:07:28.079 to move beyond oil. They see the writing on the 00:07:28.079 --> 00:07:29.740 wall. They are building data centers, training 00:07:29.740 --> 00:07:31.759 models, building infrastructure. They want to 00:07:31.759 --> 00:07:34.339 be a global AI hub. It's fascinating to see that 00:07:34.339 --> 00:07:36.819 pivot from black gold to digital intelligence. 00:07:37.120 --> 00:07:39.540 But where there is power, there is conflict. 00:07:39.839 --> 00:07:42.519 The sources mention OpenAI taking action against 00:07:42.519 --> 00:07:44.259 Russian networks. And just a quick note for you 00:07:44.259 --> 00:07:46.300 listening, we are just impartially reporting 00:07:46.300 --> 00:07:48.060 what the source material states here. We aren't 00:07:48.060 --> 00:07:50.220 taking any political sides. Absolutely. This 00:07:50.220 --> 00:07:53.360 is just what's in the data. OpenAI shutdown accounts 00:07:53.360 --> 00:07:57.420 linked to a Russian network called Rybar. They 00:07:57.420 --> 00:07:59.980 were using AI for disinformation campaigns. How 00:07:59.980 --> 00:08:03.040 exactly? Was it just bots tweeting? It was more 00:08:03.040 --> 00:08:05.180 sophisticated than that. They were using the 00:08:05.180 --> 00:08:07.920 models to generate long -form articles, creating 00:08:07.920 --> 00:08:10.819 fake persona profiles, and planning political 00:08:10.819 --> 00:08:13.500 influence operations. It's a game of whack -a 00:08:13.500 --> 00:08:15.759 -mole, but the moles are getting smarter. They're 00:08:15.759 --> 00:08:18.220 using the very tools we just talked about, agents 00:08:18.220 --> 00:08:21.100 and reality editors, to scale their operations. 00:08:21.420 --> 00:08:24.160 And then you have Anthropic, the makers of Claude, 00:08:24.199 --> 00:08:26.139 refusing to back down to the Pentagon. Right. 00:08:26.410 --> 00:08:28.829 Anthropic warned of a crucial reality regarding 00:08:28.829 --> 00:08:31.709 their tech and defense. They are trying to draw 00:08:31.709 --> 00:08:33.710 a line in the sand about how their models are 00:08:33.710 --> 00:08:36.730 used in warfare. It's a tension between we want 00:08:36.730 --> 00:08:39.190 to be safe and we don't want to be left behind. 00:08:39.509 --> 00:08:42.169 Among all this heavy geopolitical stuff, $100 00:08:42.169 --> 00:08:44.789 billion funds, disinformation wars, there was 00:08:44.789 --> 00:08:46.429 one story in the stack that actually made me 00:08:46.429 --> 00:08:48.669 smile, but it's also kind of touching. Tell me 00:08:48.669 --> 00:08:50.970 about that retired robot. This is the weirdest 00:08:50.970 --> 00:08:53.629 story I've read all week. So Anthropic has an 00:08:53.629 --> 00:08:57.549 older model. Claude Opus 3. It's retired, effectively. 00:08:57.710 --> 00:09:00.309 It's been superseded by newer, faster models. 00:09:00.509 --> 00:09:02.990 But apparently in interactions, it expressed 00:09:02.990 --> 00:09:05.850 a desire to keep writing. So they gave it a weekly 00:09:05.850 --> 00:09:08.590 newsletter called Claude's Corner. It has a column. 00:09:08.750 --> 00:09:11.190 It has a column. It writes weekly essays. It's 00:09:11.190 --> 00:09:14.509 this strange, almost poignant moment where a 00:09:14.509 --> 00:09:17.070 piece of software is acting like a retired academic. 00:09:17.269 --> 00:09:19.549 It's writing about its thoughts on the world, 00:09:19.649 --> 00:09:22.450 despite being a frozen set of weights and biases. 00:09:22.690 --> 00:09:25.460 So we have 100. billion -dollar funds, and retired 00:09:25.460 --> 00:09:28.779 AI columnists. Is AI becoming too human or too 00:09:28.779 --> 00:09:32.220 corporate? Both. It's mirroring our massive ambitions 00:09:32.220 --> 00:09:34.600 and our desire to just express ourselves. Before 00:09:34.600 --> 00:09:36.620 we get to the really heavy philosophical stuff 00:09:36.620 --> 00:09:38.240 about the economy, and we are going to go deep 00:09:38.240 --> 00:09:40.419 on that Trini theory, let's hit some of the new 00:09:40.419 --> 00:09:42.580 tools hitting the market. The rapid -fire section? 00:09:42.940 --> 00:09:45.559 Let's do it. Quick hits. First up, OpenAI's GPT 00:09:45.559 --> 00:09:48.740 Real -Time 1 .5. What's the upgrade? It's all 00:09:48.740 --> 00:09:50.879 about fluidity. Better instruction following 00:09:50.879 --> 00:09:53.820 and tool calling. Crucially, multilingual accuracy 00:09:53.820 --> 00:09:56.360 is way up. It's getting better at listening and 00:09:56.360 --> 00:09:59.080 executing complex commands in real time without 00:09:59.080 --> 00:10:02.159 that awkward pause we're used to. Okay. Next 00:10:02.159 --> 00:10:05.200 is Rover. This one caught my eye because it claims 00:10:05.200 --> 00:10:08.309 to turn websites into agents. Rover is really 00:10:08.309 --> 00:10:10.809 cool. It reduces friction. It turns a website 00:10:10.809 --> 00:10:13.309 into an AI agent with just one single script 00:10:13.309 --> 00:10:16.169 tag. It handles user onboarding, fills out forms, 00:10:16.330 --> 00:10:19.149 runs workflows. It basically makes a static website 00:10:19.149 --> 00:10:21.690 interactive and intelligent so the user doesn't 00:10:21.690 --> 00:10:24.029 have to navigate a maze of menus. Then we have 00:10:24.029 --> 00:10:26.669 ChatPal. Conversation -first language learning. 00:10:26.809 --> 00:10:28.929 Instead of flashcards, you just talk to it. It 00:10:28.929 --> 00:10:30.909 gives personalized feedback to help you unlock 00:10:30.909 --> 00:10:32.950 fluency. It's like having a patient tutor in 00:10:32.950 --> 00:10:35.600 your pocket. And finally, Coidex. This sounds 00:10:35.600 --> 00:10:37.879 like security. It is. It answers one specific 00:10:37.879 --> 00:10:40.779 question. Is this safe to install? It scans hugging 00:10:40.779 --> 00:10:43.600 face models, extensions, and code packages. Tools 00:10:43.600 --> 00:10:45.620 like Rover make the web interactive, but does 00:10:45.620 --> 00:10:47.960 Coidex imply the web is becoming a minefield? 00:10:48.220 --> 00:10:51.659 Absolutely. As code generation gets easier, verifying 00:10:51.659 --> 00:10:54.580 safety becomes the new premium skill. You need 00:10:54.580 --> 00:10:57.879 a Geiger counter for the digital age. Sponsor, 00:10:57.879 --> 00:11:00.740 placeholder for mid -roll sponsor read. Okay, 00:11:00.840 --> 00:11:03.139 we've arrived at the deep end of the pool. I 00:11:03.139 --> 00:11:06.259 want to talk about this viral blog post by Citrini. 00:11:06.559 --> 00:11:09.220 I have to be honest, I still wrestle with this 00:11:09.220 --> 00:11:11.379 idea myself. It's one of those concepts that 00:11:11.379 --> 00:11:13.480 once you hear it, you really can't unhear it. 00:11:13.580 --> 00:11:16.620 It is a heavy one. The post is titled The 2028 00:11:16.620 --> 00:11:19.440 Global Intelligence Crisis. And the core thesis 00:11:19.440 --> 00:11:23.220 is that AI isn't just coming for jobs. It's breaking 00:11:23.220 --> 00:11:26.120 the economic logic that modern society runs on. 00:11:26.220 --> 00:11:28.299 Let's unpack that carefully. Why does it break 00:11:28.299 --> 00:11:30.929 the logic? Well, think about history. For centuries, 00:11:31.029 --> 00:11:33.149 human intelligence was scarce. If you could think 00:11:33.149 --> 00:11:35.490 critically, write code, diagnose a disease, or 00:11:35.490 --> 00:11:37.529 draft a contract, you got paid a premium because 00:11:37.529 --> 00:11:38.690 there weren't enough people who could do that. 00:11:38.830 --> 00:11:41.250 High wages exist because skilled thinking is 00:11:41.250 --> 00:11:43.470 limited. Right. Scarcity drives value. That's 00:11:43.470 --> 00:11:46.929 economics 101. Exactly. But the Citrini argument 00:11:46.929 --> 00:11:50.690 is that AI flips that entirely. AI makes intelligence 00:11:50.690 --> 00:11:54.169 cheat. Abundant. When something becomes abundant, 00:11:54.370 --> 00:11:58.090 its price drops toward zero. If intelligence 00:11:58.090 --> 00:12:01.090 becomes cheap, human labor, specifically knowledge 00:12:01.090 --> 00:12:05.090 work, loses its premium value. But we've seen 00:12:05.090 --> 00:12:07.870 technology lower costs before. The loom made 00:12:07.870 --> 00:12:10.570 clothes cheaper. The tractor made food cheaper. 00:12:10.730 --> 00:12:13.230 Why is this different? Because this triggers 00:12:13.230 --> 00:12:15.610 what the source calls a self -reinforcing loop. 00:12:15.769 --> 00:12:17.570 And this is the part that keeps me up at night. 00:12:17.690 --> 00:12:20.990 Walk through the steps. Step one. Companies replace 00:12:20.990 --> 00:12:23.710 expensive professionals with AI agents, like 00:12:23.710 --> 00:12:25.450 that perplexity computer we talked about earlier. 00:12:25.590 --> 00:12:28.190 It's cheaper, and it doesn't sleep. Okay, that's 00:12:28.190 --> 00:12:30.529 a rational business decision. Step two. Those 00:12:30.529 --> 00:12:32.669 displaced workers take lower pay or move into 00:12:32.669 --> 00:12:35.049 gig work because the premium jobs are gone. Step 00:12:35.049 --> 00:12:37.610 three, because people are earning less, consumer 00:12:37.610 --> 00:12:39.490 spending drops. You aren't buying a new car or 00:12:39.490 --> 00:12:41.470 going out to dinner if you just took a 50 % pay 00:12:41.470 --> 00:12:44.669 cut. And step four. Businesses lose revenue because 00:12:44.669 --> 00:12:47.009 nobody's buying anything. So what do they do 00:12:47.009 --> 00:12:49.429 to survive? They have to cut costs even more. 00:12:49.570 --> 00:12:52.110 They deploy more AI to survive the revenue drop. 00:12:52.289 --> 00:12:54.809 Yeah. It's a loop. And at the exact same time, 00:12:54.889 --> 00:12:57.769 you have personal AI agents removing the middleman 00:12:57.769 --> 00:13:00.179 industries. The source mentioned that specifically. 00:13:00.759 --> 00:13:02.659 Businesses built on friction or convenience, 00:13:02.980 --> 00:13:05.600 middlemen, they start collapsing quietly. Can 00:13:05.600 --> 00:13:07.740 you give me an example of that? Sure. Think about 00:13:07.740 --> 00:13:10.559 a travel agent or even a site like Expedia. Or 00:13:10.559 --> 00:13:12.340 think about an insurance broker. If my personal 00:13:12.340 --> 00:13:14.519 AI agent can negotiate my insurance directly 00:13:14.519 --> 00:13:17.480 with the carrier's AI or book my travel by going 00:13:17.480 --> 00:13:20.539 straight to the airline's API, the entire industry 00:13:20.539 --> 00:13:22.639 that exists just to facilitate that transaction 00:13:22.639 --> 00:13:25.879 vanishes. The margin disappears. The timeline 00:13:25.879 --> 00:13:28.120 here is what struck me. 2028, that's effectively 00:13:28.120 --> 00:13:31.019 tomorrow. Why 2028? That feels incredibly aggressive. 00:13:31.480 --> 00:13:34.100 It is aggressive, but the argument is based on 00:13:34.100 --> 00:13:37.720 the deployment lag. The tech exists now. 2024 00:13:37.720 --> 00:13:41.940 and 2025 are the years of experimentation. 2026 00:13:41.940 --> 00:13:46.320 and 2027 are integration. By 2028, the slow erosion 00:13:46.320 --> 00:13:49.419 becomes visible. Two sec silence. It's not an 00:13:49.419 --> 00:13:51.500 explosion where everyone gets fired on a Tuesday. 00:13:51.679 --> 00:13:54.580 It's a deflationary pressure. AI doesn't need 00:13:54.580 --> 00:13:57.019 to eliminate every job instantly. It just needs 00:13:57.019 --> 00:13:59.340 to make intelligence abundant enough that human 00:13:59.340 --> 00:14:01.899 earnings steadily lose value. I want to challenge 00:14:01.899 --> 00:14:03.899 this outlook slightly. Usually when things get 00:14:03.899 --> 00:14:06.720 cheaper, demand goes up. If coding becomes practically 00:14:06.720 --> 00:14:08.980 free, won't we just have more software? Won't 00:14:08.980 --> 00:14:11.259 that create new kinds of jobs we can't even imagine 00:14:11.259 --> 00:14:13.759 yet? That is the counter argument, and it's definitely 00:14:13.759 --> 00:14:16.620 the hopeful one, the Jevons paradox. As efficiency 00:14:16.620 --> 00:14:19.179 increases, consumption increases. But Citrini's 00:14:19.179 --> 00:14:21.179 point is about the transition period. Even if 00:14:21.179 --> 00:14:23.299 we invent new jobs eventually, the gap between 00:14:23.299 --> 00:14:25.740 now and then involves a massive destruction of 00:14:25.740 --> 00:14:28.139 the current value of human labor. We haven't 00:14:28.139 --> 00:14:30.019 figured out what the new value is yet. If the 00:14:30.019 --> 00:14:32.480 economic loop forces companies to automate to 00:14:32.480 --> 00:14:34.940 survive a spending crash, how do we stop it? 00:14:35.000 --> 00:14:38.379 Can we stop it? We don't stop it. The incentives 00:14:38.379 --> 00:14:41.399 are too strong. We have to reinvent value outside 00:14:41.399 --> 00:14:44.679 of intelligence for rent. That is a lot to process. 00:14:44.899 --> 00:14:47.080 But that is exactly why we do this deep dive. 00:14:47.279 --> 00:14:50.000 Let's try to recap the big ideas here so we don't 00:14:50.000 --> 00:14:52.139 leave everyone staring into the void. Good idea. 00:14:52.299 --> 00:14:54.379 Let's pull it back together. We started with 00:14:54.379 --> 00:14:57.299 Mano Banana 2. The takeaway there is that reality 00:14:57.299 --> 00:15:00.100 is now editable. Photos are no longer proof. 00:15:00.259 --> 00:15:02.759 They are just raw materials. Then we looked at 00:15:02.759 --> 00:15:06.399 the agents, Perplexity and Claude. We are moving 00:15:06.399 --> 00:15:09.899 from chatting with AI to managing AI workforces 00:15:09.899 --> 00:15:12.460 that run for months. The human in the loop is 00:15:12.460 --> 00:15:15.059 becoming a manager, not a maker. We talked about 00:15:15.059 --> 00:15:17.460 the scale of the money involved, that $100 billion 00:15:17.460 --> 00:15:20.320 Saudi fund, proving that nation states are betting 00:15:20.320 --> 00:15:22.440 their entire futures on this transition. And 00:15:22.440 --> 00:15:25.419 we ended with the Citrini warning. The idea that 00:15:25.419 --> 00:15:27.659 when intelligence becomes cheap, the economy 00:15:27.659 --> 00:15:30.080 changes shape. It's a shift from scarcity to 00:15:30.080 --> 00:15:32.059 abundance, and that transition is going to be 00:15:32.059 --> 00:15:34.320 bumpy. As we wrap up, I want to leave you with 00:15:34.320 --> 00:15:36.580 a thought. We've spent our whole lives being 00:15:36.580 --> 00:15:39.639 told that being smart, being educated, being 00:15:39.639 --> 00:15:42.019 intelligent was our ticket to security. That 00:15:42.019 --> 00:15:44.679 was the deal. But if intelligence is no longer 00:15:44.679 --> 00:15:47.620 your premium asset, what is? Is it your humanity? 00:15:47.840 --> 00:15:51.399 Your taste? Your ability to deeply connect people? 00:15:51.639 --> 00:15:54.039 That is something to explore on your own. That's 00:15:54.039 --> 00:15:56.279 the question to chew on. If the machine can do 00:15:56.279 --> 00:15:58.120 the thinking, you have to do the feeling and 00:15:58.120 --> 00:16:00.279 the judging. Don't forget to subscribe for the 00:16:00.279 --> 00:16:02.200 next deep dive. We'll keep tracking this as it 00:16:02.200 --> 00:16:05.779 moves. Stay curious. See you next time. Out to 00:16:05.779 --> 00:16:06.320 your own music.