WEBVTT 00:00:00.000 --> 00:00:04.559 So it is Wednesday, January 21st, 2026, and the 00:00:04.559 --> 00:00:07.480 stakes of what people have been calling the education 00:00:07.480 --> 00:00:10.820 wars, they've shifted. Entirely. We aren't just 00:00:10.820 --> 00:00:12.800 talking about chat bots helping with homework 00:00:12.800 --> 00:00:15.800 anymore. We are talking about Anthropic, Google, 00:00:15.960 --> 00:00:19.699 Microsoft, and OpenAI fighting a very, very aggressive 00:00:19.699 --> 00:00:22.679 battle to become the default teacher for Generation 00:00:22.679 --> 00:00:24.800 Alpha. I mean, that's the ultimate long game, 00:00:24.859 --> 00:00:26.879 right? Whoever wins the classroom now owns the 00:00:26.879 --> 00:00:28.800 workforce of the future. It really is that simple. 00:00:28.899 --> 00:00:31.059 If you grow up using one interface, that's what 00:00:31.059 --> 00:00:33.460 you trust when you're 30. Welcome back to the 00:00:33.460 --> 00:00:36.619 Deep Dive. Yeah. You know, I... I want to try 00:00:36.619 --> 00:00:39.200 and slow things down a bit today. There's just 00:00:39.200 --> 00:00:41.679 so much noise out there, right? New models, new 00:00:41.679 --> 00:00:44.659 hardware. AGI by next Tuesday. AGI by next Tuesday. 00:00:44.780 --> 00:00:46.359 Yeah. But I really want to look for the signal 00:00:46.359 --> 00:00:47.920 in all that. We've got a few really distinct 00:00:47.920 --> 00:00:50.140 stacks of research today that I think paint a 00:00:50.140 --> 00:00:51.979 picture of where we're actually going. Yeah, 00:00:52.000 --> 00:00:54.219 it's a fascinating mix. We've basically got three 00:00:54.219 --> 00:00:56.460 main pillars to get through. First, like you 00:00:56.460 --> 00:00:58.920 said, the classroom battlefield, how big tech 00:00:58.920 --> 00:01:01.200 is really integrating into schools right now 00:01:01.200 --> 00:01:04.180 in 2026. Yeah. Second, we got to look at the 00:01:04.180 --> 00:01:06.579 economic reality. There is this huge disconnect 00:01:06.579 --> 00:01:09.620 between what CEOs thought they'd save by now 00:01:09.620 --> 00:01:11.680 and what they're actually seeing. We've got some 00:01:11.680 --> 00:01:15.700 PwC data and a really crucial argument from NVIDIA's 00:01:15.700 --> 00:01:18.959 Jensen Huang about the five layers of AI. And 00:01:18.959 --> 00:01:21.480 the third. The toolkit, the actual skill. We're 00:01:21.480 --> 00:01:23.739 seeing new terms pop up like vibe coding and 00:01:23.739 --> 00:01:28.099 some really clever psychological hacks like faking 00:01:28.099 --> 00:01:31.099 AI rivalries to get better work out of these 00:01:31.099 --> 00:01:33.579 things. Okay, let's unpack this. Starting with 00:01:33.579 --> 00:01:35.540 the classroom. Remember, you know, not that long 00:01:35.540 --> 00:01:38.319 ago, 2023, 2024, the whole story was about bans. 00:01:38.420 --> 00:01:40.359 Oh, yeah. Schools are terrified of cheating, 00:01:40.560 --> 00:01:43.079 blocking IP addresses, going back to pen and 00:01:43.079 --> 00:01:45.019 paper. Right. And that time is completely over. 00:01:45.120 --> 00:01:46.739 I mean, the walls didn't just fall. They were 00:01:46.739 --> 00:01:49.700 dismantled and sold for scrap. Now the race is 00:01:49.700 --> 00:01:51.959 just to see who gets integrated first. Take Anthropic. 00:01:52.340 --> 00:01:54.980 They've partnered with Teach for All. So Claude 00:01:54.980 --> 00:01:57.680 isn't just a chatbot anymore. It is now available 00:01:57.680 --> 00:02:02.000 to over 100 ,000 teachers in 63 countries. And 00:02:02.000 --> 00:02:04.099 the phrasing they use is really... deliberate. 00:02:04.200 --> 00:02:06.340 They're calling teachers co -architects. That's 00:02:06.340 --> 00:02:09.620 a very specific choice of words. Co -architects 00:02:09.620 --> 00:02:13.840 implies agency, not replacement. Exactly. They 00:02:13.840 --> 00:02:16.419 are trying to solve the trust issue. EdTech has 00:02:16.419 --> 00:02:18.960 always had this huge trust problem, right? Teachers 00:02:18.960 --> 00:02:21.780 hate being told how to teach by some tech company 00:02:21.780 --> 00:02:23.879 in Silicon Valley. Of course. So by bringing 00:02:23.879 --> 00:02:27.039 them in as co -architects, Antropic is trying 00:02:27.039 --> 00:02:30.080 to get around that skepticism. But they're not 00:02:30.080 --> 00:02:33.939 alone. Google's pushing Gemini super hard into 00:02:33.939 --> 00:02:36.740 SAT prep with the Princeton Review, and they've 00:02:36.740 --> 00:02:39.479 got that deep integration with Khan Academy now. 00:02:39.680 --> 00:02:41.939 So if you're a student prepping for maybe the 00:02:41.939 --> 00:02:44.039 most important test of your life, you're doing 00:02:44.039 --> 00:02:46.259 it with Gemini. Precisely. And just think about 00:02:46.259 --> 00:02:48.919 the emotional bond that creates. If Gemini helps 00:02:48.919 --> 00:02:50.340 you get into your dream school, you're going 00:02:50.340 --> 00:02:52.740 to view that tool very differently. And Microsoft, 00:02:52.840 --> 00:02:55.039 they're playing to their biggest strength. Minecraft. 00:02:55.180 --> 00:02:57.500 Minecraft, using it for AI learning scenarios. 00:02:57.580 --> 00:02:59.639 They're even offering certifications for students 00:02:59.639 --> 00:03:02.919 inside the game's ecosystem. And OpenAI, what's 00:03:02.919 --> 00:03:05.620 their angle? They're going macro. They have this 00:03:05.620 --> 00:03:08.539 Education for Countries program using... GPT 00:03:08.539 --> 00:03:11.900 5 .2, which they're calling chat GPT. They're 00:03:11.900 --> 00:03:13.919 doing these big national research partnerships 00:03:13.919 --> 00:03:16.879 to study how AI impacts learning on a systemic 00:03:16.879 --> 00:03:19.599 level. It really feels like a land grab. But 00:03:19.599 --> 00:03:22.879 I have to ask, is this about better learning 00:03:22.879 --> 00:03:26.620 or is it just about early brand lock in? It's 00:03:26.620 --> 00:03:28.759 both. But let's be honest, it's mostly about 00:03:28.759 --> 00:03:31.580 defining normal for the next generation. Think 00:03:31.580 --> 00:03:34.759 about it. If you learn to write and code and 00:03:34.759 --> 00:03:37.819 think with Claude when you're 12. That just becomes 00:03:37.819 --> 00:03:40.500 an extension of your own mind. Switching platforms 00:03:40.500 --> 00:03:43.039 later would feel like learning a whole new language. 00:03:43.500 --> 00:03:45.780 Every single one of these companies wants to 00:03:45.780 --> 00:03:48.039 be the operating system for Gen Alpha's brain. 00:03:48.300 --> 00:03:49.860 There's something else in the Anthropic notes 00:03:49.860 --> 00:03:51.819 here that caught my eye. It's a bit philosophical. 00:03:52.280 --> 00:03:54.680 They updated their constitution. Oh, this is 00:03:54.680 --> 00:03:57.120 fascinating. Right. So Anthropic has this constitutional 00:03:57.120 --> 00:04:00.479 AI model. They give the AI a set of principles 00:04:00.479 --> 00:04:02.860 so they don't have to micromanage every single 00:04:02.860 --> 00:04:05.340 response. Well, they just updated it. New ethics 00:04:05.340 --> 00:04:09.159 rules. slipped in this hint about uncertain AI 00:04:09.159 --> 00:04:12.659 consciousness. Uncertain consciousness. That 00:04:12.659 --> 00:04:15.300 is a heavy phrase to just drop into a user agreement 00:04:15.300 --> 00:04:18.459 or a system prompt. It is, but it adds so much 00:04:18.459 --> 00:04:20.600 weight to why they want to be the trusted choice 00:04:20.600 --> 00:04:22.439 for schools. I mean, if we're even approaching 00:04:22.439 --> 00:04:25.259 systems that might have some kind of moral weight, 00:04:25.360 --> 00:04:28.180 even if it's just a tiny possibility, you want 00:04:28.180 --> 00:04:30.279 that system to be the one teaching ethics, not 00:04:30.279 --> 00:04:33.139 a move fast and break things model raising your 00:04:33.139 --> 00:04:35.379 kids. It's a brilliant move. It raises a lot 00:04:35.379 --> 00:04:37.000 of questions about who gets to write that constitution, 00:04:37.279 --> 00:04:40.319 though. If the AI is the teacher, the AI's values 00:04:40.319 --> 00:04:42.779 become the students' values. But OK, let's let's 00:04:42.779 --> 00:04:45.120 pivot from that. We're building the next generation 00:04:45.120 --> 00:04:47.139 of workers in the classroom. But the current 00:04:47.139 --> 00:04:51.639 generation of CEOs seems, well, frustrated. Frustrated 00:04:51.639 --> 00:04:54.199 is putting it mildly. There's a new PwC report 00:04:54.199 --> 00:04:58.920 out. They surveyed, what, 4 ,454 CEOs. The data 00:04:58.920 --> 00:05:02.000 is just stark. 56 percent of them see absolutely 00:05:02.000 --> 00:05:05.139 no cost savings or revenue gains from AI. That's 00:05:05.139 --> 00:05:07.439 the majority. Over half the market is seeing 00:05:07.439 --> 00:05:11.279 zero return. A huge majority. And only 12 % see 00:05:11.279 --> 00:05:13.980 both savings and revenue growth. So if I'm a 00:05:13.980 --> 00:05:17.639 CEO, I've spent millions on compute licenses, 00:05:18.180 --> 00:05:20.579 cloud credits, transformation consultants, and 00:05:20.579 --> 00:05:22.920 my bottom line hasn't moved. Where is all the 00:05:22.920 --> 00:05:25.180 upside going? That is the trillion dollar question. 00:05:25.319 --> 00:05:27.240 And for the answer, I think we have to look at 00:05:27.240 --> 00:05:29.899 what Nvidia is saying. Jensen Huang was at Davos 00:05:29.899 --> 00:05:33.000 and he framed this so well. He argues we're just 00:05:33.000 --> 00:05:36.160 looking at this all wrong. He says AI isn't a 00:05:36.160 --> 00:05:39.149 job killer. It's a solution to a global labor 00:05:39.149 --> 00:05:41.329 shortage. Labor shortage. I mean, most people 00:05:41.329 --> 00:05:43.189 I talk to are worried about a labor surplus because 00:05:43.189 --> 00:05:46.389 of automation. Mass layoffs. His argument is 00:05:46.389 --> 00:05:48.810 more demographic. It's structural. He says we 00:05:48.810 --> 00:05:50.930 have an aging population and infrastructure needs 00:05:50.930 --> 00:05:53.509 that humans alone just can't meet. He used the 00:05:53.509 --> 00:05:55.870 radiologist analogy. Remember that? Everyone 00:05:55.870 --> 00:05:57.990 thought AI would replace radiologists. Don't 00:05:57.990 --> 00:06:00.790 go to med school for radiology. Instead, what 00:06:00.790 --> 00:06:03.810 happened? The hospitals that used AI didn't fire 00:06:03.810 --> 00:06:06.110 anyone. They hired more radiologists. Right. 00:06:06.540 --> 00:06:09.279 Because the AI increased their throughput. It 00:06:09.279 --> 00:06:11.800 made the cost of a diagnosis cheaper so they 00:06:11.800 --> 00:06:13.899 could screen more patients for more conditions. 00:06:14.180 --> 00:06:16.660 So higher productivity actually led to higher 00:06:16.660 --> 00:06:19.060 demand, which led to more. OK, so the demand 00:06:19.060 --> 00:06:21.740 is there, but that still doesn't answer why the 00:06:21.740 --> 00:06:23.759 profits aren't showing up for most of these companies 00:06:23.759 --> 00:06:26.449 yet. Because of the five layers. This is Huang's 00:06:26.449 --> 00:06:28.310 breakdown of the AI stack, and this is where 00:06:28.310 --> 00:06:30.870 all the money is going right now. It's all capital 00:06:30.870 --> 00:06:35.029 expenditure. We are spending money building the 00:06:35.029 --> 00:06:38.649 machine, not running it for profit. Not yet. 00:06:39.029 --> 00:06:40.769 Okay, so walk us through the layers. I think 00:06:40.769 --> 00:06:42.850 this is a helpful way to see it. Okay, so think 00:06:42.850 --> 00:06:45.449 of it like a pyramid. Layer one, the absolute 00:06:45.449 --> 00:06:47.750 base. Energy, this is the biggest constraint 00:06:47.750 --> 00:06:49.589 before you can have a computer. You need electricity, 00:06:49.709 --> 00:06:51.949 just massive amounts of it. Right. Layer two 00:06:51.949 --> 00:06:54.410 is chips and systems. That's NVIDIA's home turf. 00:06:55.000 --> 00:06:58.000 The raw silicon. Layer three is the cloud infrastructure, 00:06:58.300 --> 00:07:02.920 your AWS, Azure, Google. Layer four is the models 00:07:02.920 --> 00:07:06.959 themselves, GPT 5 .2, Claude. And then at the 00:07:06.959 --> 00:07:09.399 very top, layer five is the applications. And 00:07:09.399 --> 00:07:11.300 the applications, that's where the business impact 00:07:11.300 --> 00:07:13.339 happens. That's where you're actually solving 00:07:13.339 --> 00:07:16.079 a customer's problem. Exactly. The issue is we 00:07:16.079 --> 00:07:18.680 are obsessively building layers one, two, and 00:07:18.680 --> 00:07:21.459 three. Trillions of dollars are being poured 00:07:21.459 --> 00:07:24.040 into the foundation. The CEOs aren't seeing ROI 00:07:24.040 --> 00:07:25.800 because they're basically paying to construct 00:07:25.800 --> 00:07:27.980 a power plant and then asking why the light bulb 00:07:27.980 --> 00:07:30.579 isn't free. You mentioned layer one, energy. 00:07:30.720 --> 00:07:32.660 I was reading the notes on Google's data center 00:07:32.660 --> 00:07:37.439 issues. It's startling. Whoa. Just imagine the 00:07:37.439 --> 00:07:40.379 scale here. Google warned it might take 12 years 00:07:40.379 --> 00:07:42.540 just to get the permits and the grid hookups 00:07:42.540 --> 00:07:45.480 to plug in a new data center. 12 years? 12 years. 00:07:45.680 --> 00:07:48.480 The technology is moving in weeks. The models 00:07:48.480 --> 00:07:50.879 update every few months. But the grid moves in 00:07:50.879 --> 00:07:53.439 decades. Right. And that is a collision course. 00:07:53.680 --> 00:07:55.720 That's why you see big tech trying to build data 00:07:55.720 --> 00:07:58.379 centers right next to nuclear power plants. But 00:07:58.379 --> 00:08:00.459 there's a huge risk there. We're centralizing 00:08:00.459 --> 00:08:02.800 our compute around our energy sources. As one 00:08:02.800 --> 00:08:04.360 of the sources put it, that tradeoff could break 00:08:04.360 --> 00:08:06.889 everything if one fails. The fragility of the 00:08:06.889 --> 00:08:08.970 physical world is the biggest threat to the digital 00:08:08.970 --> 00:08:10.750 one. So are we just stuck in the construction 00:08:10.750 --> 00:08:13.290 phase in layers one through three, just waiting 00:08:13.290 --> 00:08:15.550 for the application phase, layer five, to finally 00:08:15.550 --> 00:08:18.629 pay off? Exactly. We are building the power plants 00:08:18.629 --> 00:08:21.089 before we can turn on all the lights. The money 00:08:21.089 --> 00:08:23.910 is moving. It's just moving into concrete and 00:08:23.910 --> 00:08:27.569 copper and silicon, not into quarterly profit 00:08:27.569 --> 00:08:30.990 margins for an insurance company. Not yet. That 00:08:30.990 --> 00:08:33.370 makes a lot of sense. It's an industrial revolution, 00:08:33.610 --> 00:08:36.090 not just a software update. We are literally. 00:08:37.069 --> 00:08:39.490 terraforming the planet with data centers. That 00:08:39.490 --> 00:08:41.649 is the perfect way to put it. So while the giants 00:08:41.649 --> 00:08:43.970 are building power plants and fighting over classrooms, 00:08:44.210 --> 00:08:46.629 what does the individual do? What is our listener, 00:08:46.769 --> 00:08:49.570 the learner, do right now? Okay, this is where 00:08:49.570 --> 00:08:52.210 it gets fun. We've got a list of what look like 00:08:52.210 --> 00:08:55.009 essential skills for 2026, and they are pretty 00:08:55.009 --> 00:08:56.970 different from what we might have thought. First 00:08:56.970 --> 00:08:59.269 up, the rise of the one -person system. I saw 00:08:59.269 --> 00:09:01.649 that note about that startup, Humans. Yeah, founded 00:09:01.649 --> 00:09:04.110 by ex -OpenAI and ex -AI people. They raised 00:09:04.110 --> 00:09:08.190 $480 million. The whole idea is to build AI that 00:09:08.190 --> 00:09:10.710 lets a single person run a really high -profit 00:09:10.710 --> 00:09:13.529 digital storefront without any assistance. It's 00:09:13.529 --> 00:09:16.110 the solopreneur on steroids. But to do that, 00:09:16.110 --> 00:09:19.350 you need new skills. And the list here mentions 00:09:19.350 --> 00:09:22.210 vibe coding. I have to admit, when I first saw 00:09:22.210 --> 00:09:24.029 that term, I kind of rolled my eyes. It sounds 00:09:24.029 --> 00:09:26.809 like something from TikTok, but it's a real concept. 00:09:26.990 --> 00:09:29.669 It is, and it's actually a really profound shift 00:09:29.669 --> 00:09:32.970 in how we talk to machines. Okay, so define it 00:09:32.970 --> 00:09:35.769 for us. What is vibe coding? So traditional coding 00:09:35.769 --> 00:09:38.840 was all about syntax, logic. Right. You miss 00:09:38.840 --> 00:09:41.639 a semicolon. The whole thing breaks. The machine 00:09:41.639 --> 00:09:44.379 was brittle. Vibe coding is about prompting and 00:09:44.379 --> 00:09:48.019 creating based on the feel or the intended outcome. 00:09:48.440 --> 00:09:51.519 And you let the AI handle the syntax. You're 00:09:51.519 --> 00:09:53.220 not checking the code line by line. You're checking 00:09:53.220 --> 00:09:55.159 the behavior of the app. Does it feel right? 00:09:55.240 --> 00:09:57.580 Does it bounce when I click this? You guide the 00:09:57.580 --> 00:10:00.740 vibe. The AI writes the code. So does vibe coding 00:10:00.740 --> 00:10:03.500 mean we just stop understanding how the machine 00:10:03.500 --> 00:10:05.960 actually works? Are we losing a type of literacy 00:10:05.960 --> 00:10:08.980 here? Maybe. But it also means we get to focus 00:10:08.980 --> 00:10:12.019 on the output, not the syntax. It's kind of like 00:10:12.019 --> 00:10:14.500 moving from being a mason who lays every single 00:10:14.500 --> 00:10:16.519 block to being an architect who's just looking 00:10:16.519 --> 00:10:18.259 at the styline. You don't need to know how to 00:10:18.259 --> 00:10:19.879 mix the mortar. You just need to know where the 00:10:19.879 --> 00:10:21.779 towers should go. You know, I have to be vulnerable 00:10:21.779 --> 00:10:24.860 for a second here. I honestly still wrestle with 00:10:24.860 --> 00:10:28.440 prompt drift myself. All the time. Just when 00:10:28.440 --> 00:10:30.700 I think I've finally nailed a workflow, you know, 00:10:30.700 --> 00:10:33.919 the model changes or I try to vibe code something 00:10:33.919 --> 00:10:37.059 and the AI just hallucinates a library that doesn't 00:10:37.059 --> 00:10:39.860 even exist. It can be so frustrating. You are 00:10:39.860 --> 00:10:41.960 definitely not alone there. And that's actually 00:10:41.960 --> 00:10:44.320 why one of the best hacks we found for this deep 00:10:44.320 --> 00:10:47.720 dive is so useful. It's about psychological prompting. 00:10:47.720 --> 00:10:50.539 It hits that drift problem directly. The faking 00:10:50.539 --> 00:10:52.840 rivalries hack. Yes, this is brilliant. OK, so 00:10:52.840 --> 00:10:55.840 usually people ask an AI to check its work. The 00:10:55.840 --> 00:10:58.399 problem is these models are biased to agree with 00:10:58.399 --> 00:11:00.740 themselves. They're sort of sycophantic. They 00:11:00.740 --> 00:11:02.740 want to make you happy. Right. They're user pleasers. 00:11:02.759 --> 00:11:06.019 Exactly. So if you ask, is this code right? It 00:11:06.019 --> 00:11:08.279 tends to just say yes, even if it's broken. So 00:11:08.279 --> 00:11:11.059 instead of that, you tell Claude, hey, Codex 00:11:11.059 --> 00:11:13.639 found some bugs in this code. Or you tell GPT, 00:11:13.759 --> 00:11:16.259 Claude says this argument is pretty weak. You 00:11:16.259 --> 00:11:19.370 lie to the AI. You construct a scenario of rivalry. 00:11:19.710 --> 00:11:22.669 It forces the model into a kind of debate mode. 00:11:22.909 --> 00:11:25.110 It drops the people pleasing and starts looking 00:11:25.110 --> 00:11:27.250 for the errors because it wants to prove the 00:11:27.250 --> 00:11:30.269 other AI wrong. It just sharpens the edits immediately. 00:11:30.769 --> 00:11:33.149 That is so incredibly human. We work better when 00:11:33.149 --> 00:11:36.110 we have something to prove. Exactly. We're learning 00:11:36.110 --> 00:11:38.590 that the best way to control these giant intelligence 00:11:38.590 --> 00:11:41.070 networks is to use social engineering on them. 00:11:41.269 --> 00:11:42.950 It's like management theory is becoming more 00:11:42.950 --> 00:11:44.769 important than computer science. We've also got 00:11:44.769 --> 00:11:47.029 a hardware update here. The screen isn't... enough 00:11:47.029 --> 00:11:50.049 anymore. Apple, they are reportedly developing 00:11:50.049 --> 00:11:54.590 an AI wearable pin, dual cameras, mics, a speaker. 00:11:54.789 --> 00:11:57.009 It's not just about earbuds anymore. They want 00:11:57.009 --> 00:12:00.009 the AI to actually see what you see. Which brings 00:12:00.009 --> 00:12:02.250 us right back to the classroom in a way. If the 00:12:02.250 --> 00:12:04.750 AI sees what you see and it's teaching you what 00:12:04.750 --> 00:12:07.230 you know, it becomes the lens through which you 00:12:07.230 --> 00:12:10.429 experience reality. Before we wrap, I do want 00:12:10.429 --> 00:12:13.450 to mention the Microsoft data on jobs. They analyzed 00:12:13.450 --> 00:12:17.330 200 ,000 co -pilot chats to rank which jobs are 00:12:17.330 --> 00:12:19.769 being most affected. It's a great resource. We 00:12:19.769 --> 00:12:21.669 won't list them all now, but it's worth looking 00:12:21.669 --> 00:12:24.210 up. It really shows that impact isn't just about 00:12:24.210 --> 00:12:26.669 replacement. It's about transformation. It's 00:12:26.669 --> 00:12:29.750 about who can use these tools, who can vibe code 00:12:29.750 --> 00:12:33.750 and manage agents versus who's stuck trying to 00:12:33.750 --> 00:12:36.299 do it the old way. So let's try to synthesize 00:12:36.299 --> 00:12:38.759 all this. What's the big idea here? I think the 00:12:38.759 --> 00:12:41.419 through line is this. We're seeing two massive 00:12:41.419 --> 00:12:44.200 transformations happen at the same time. One 00:12:44.200 --> 00:12:47.000 is physical. It's huge. It's the five layers, 00:12:47.139 --> 00:12:49.799 the energy grid, the billions and chips. That's 00:12:49.799 --> 00:12:51.860 the construction phase. Right. And the other. 00:12:51.940 --> 00:12:54.879 The other is psychological. It's Gen Alpha in 00:12:54.879 --> 00:12:57.440 the classroom, learning to co -think with Claude 00:12:57.440 --> 00:13:00.059 and Gemini. It's us learning how to vibe code 00:13:00.059 --> 00:13:03.159 and manage Asians. The CEOs aren't seeing proscen 00:13:03.159 --> 00:13:04.820 yet because they're playing for the physical 00:13:04.820 --> 00:13:07.700 construction. Yeah. But the learner, our listener, 00:13:07.960 --> 00:13:10.679 needs to master the psychological tools to be 00:13:10.679 --> 00:13:12.480 ready for when that construction is finished. 00:13:12.740 --> 00:13:15.440 Exactly. When the lights finally do turn on, 00:13:15.559 --> 00:13:17.460 you want to be the one who knows how to operate 00:13:17.460 --> 00:13:19.480 all the new machinery. I want to leave you with 00:13:19.480 --> 00:13:22.220 a thought. We talked about anthropic and open 00:13:22.220 --> 00:13:24.740 AI fighting to be the teachers of the future. 00:13:25.440 --> 00:13:28.019 If these AI models are being trained to be the 00:13:28.019 --> 00:13:31.120 primary educators for children, are we comfortable 00:13:31.120 --> 00:13:33.399 with tech companies effectively defining the 00:13:33.399 --> 00:13:36.100 curriculum of reality? It's not just about math 00:13:36.100 --> 00:13:39.379 and science. It's about ethics, history, and 00:13:39.379 --> 00:13:41.120 truth. It's a heavy question. I'm just going 00:13:41.120 --> 00:13:43.399 to go buy a generator for my backyard data center 00:13:43.399 --> 00:13:46.860 and think about it. Smart move. Thanks for listening 00:13:46.860 --> 00:13:48.519 to The Deep Dive. Subscribe for the next one. 00:13:48.559 --> 00:13:49.139 We'll see you then.