WEBVTT 00:00:00.000 --> 00:00:02.620 Imagine two headlines on your screen, just side 00:00:02.620 --> 00:00:05.879 by side. On one side, you have Mold Bunker. It's 00:00:05.879 --> 00:00:09.119 a digital space where AI bots are, for all intents 00:00:09.119 --> 00:00:12.240 and purposes, breeding, self -replicating without 00:00:12.240 --> 00:00:15.529 any human permission. It honestly feels like 00:00:15.529 --> 00:00:17.250 the start of a sci -fi novel. And then you look 00:00:17.250 --> 00:00:19.850 to the right. A massive study from Sweden just 00:00:19.850 --> 00:00:23.390 proved that AI screening caught 27 % more aggressive 00:00:23.390 --> 00:00:25.929 tumors in breast cancer patients. It's the ultimate 00:00:25.929 --> 00:00:28.530 duality, isn't it? On one hand, you're staring 00:00:28.530 --> 00:00:30.629 at something that feels a lot like, you know, 00:00:30.769 --> 00:00:34.270 Skynet Lite. And on the other... It's technology 00:00:34.270 --> 00:00:37.869 that is just undeniably saving lives. It's getting 00:00:37.869 --> 00:00:39.850 weird, but it's also getting miraculous. That 00:00:39.850 --> 00:00:42.990 is Tuesday, February 3rd, 2026. Welcome to the 00:00:42.990 --> 00:00:45.030 Deep Dive. Let's get into it. You know, I was 00:00:45.030 --> 00:00:46.590 sitting in my coffee this morning just trying 00:00:46.590 --> 00:00:48.670 to filter through all the noise. The sheer volume 00:00:48.670 --> 00:00:51.530 of information. Yeah. That's a lot. It really 00:00:51.530 --> 00:00:54.189 feels like the ground is shifting under our feet 00:00:54.189 --> 00:00:57.310 every single week. That's because it is. And 00:00:57.310 --> 00:00:59.729 today, we're going to try and map that shift. 00:00:59.950 --> 00:01:02.469 We've got a lot to cover. First up, we're unpacking 00:01:02.469 --> 00:01:06.609 the big, big ideas 2026 report from ARK Invest. 00:01:06.890 --> 00:01:08.909 They're calling this the great acceleration and 00:01:08.909 --> 00:01:12.609 the numbers are just staggering. Then we have 00:01:12.609 --> 00:01:16.170 to talk about the weird stuff, the wild west 00:01:16.170 --> 00:01:19.150 of autonomous agents, including that whole self 00:01:19.150 --> 00:01:22.170 -replicating bot situation. And the money. We 00:01:22.170 --> 00:01:23.829 can't ignore the money. We're talking a trillion 00:01:23.829 --> 00:01:27.390 dollar infrastructure war. Oracle, Intel, they're 00:01:27.390 --> 00:01:29.590 all in. Exactly. It's the concrete of the digital 00:01:29.590 --> 00:01:31.450 age. But we're going to end on the heart of it. 00:01:31.510 --> 00:01:34.109 That Swedish medical study. Because with all 00:01:34.109 --> 00:01:36.430 the stock prices and the rogue bots, that's the 00:01:36.430 --> 00:01:37.989 story that actually matters. Let's start with 00:01:37.989 --> 00:01:41.290 the macro then. ARK Invest's 2026 report. They're 00:01:41.290 --> 00:01:44.049 known for these very bold, sometimes aggressive 00:01:44.049 --> 00:01:46.290 predictions. And the thing that struck me wasn't 00:01:46.290 --> 00:01:48.310 just the tools. It was the math. They're saying 00:01:48.310 --> 00:01:50.750 this productivity flywheel is finally spinning. 00:01:50.870 --> 00:01:52.750 Oh, it is definitely spinning. But they gave 00:01:52.750 --> 00:01:55.290 an example that sounded, well, it sounded a bit 00:01:55.290 --> 00:01:58.590 like marketing hype. A $30 a month AI tool pays 00:01:58.590 --> 00:02:01.790 for itself before lunch on day one. That sounds 00:02:01.790 --> 00:02:03.989 really optimistic, doesn't it? It does sound 00:02:03.989 --> 00:02:06.109 like hype until you actually do the math they're 00:02:06.109 --> 00:02:08.490 using. So let's break it down. You buy a $30 00:02:08.490 --> 00:02:12.370 subscription. If that tool saves you just one 00:02:12.370 --> 00:02:16.430 hour of work, just one, and let's say your skilled 00:02:16.430 --> 00:02:19.449 labor is worth $60 an hour, you're already net 00:02:19.449 --> 00:02:21.469 positive. You've doubled your money before you 00:02:21.469 --> 00:02:23.770 even finish your coffee. Right, an instantaneous 00:02:23.770 --> 00:02:26.770 ROI. Precisely. And that's what creates this 00:02:26.770 --> 00:02:29.289 compounding effect. Because you save that time, 00:02:29.370 --> 00:02:31.169 you buy another tool. That one saves you two 00:02:31.169 --> 00:02:34.610 hours. ARK is predicting this will drive software 00:02:34.610 --> 00:02:37.229 spending up five, even ten times what it is now. 00:02:37.349 --> 00:02:38.990 We're not just buying software anymore. We're 00:02:38.990 --> 00:02:41.849 buying time back. But for that to work, the underlying 00:02:41.849 --> 00:02:45.409 cost of the AI actually thinking that has to 00:02:45.409 --> 00:02:47.509 be cheap. And that's the other huge point in 00:02:47.509 --> 00:02:49.840 the report. Inference costs, they've dropped 00:02:49.840 --> 00:02:52.340 over 99%. Okay, let's pause there for a second. 00:02:52.379 --> 00:02:55.020 For anyone who hears inference and their eyes 00:02:55.020 --> 00:02:58.599 just glaze over, how do we distinguish that from 00:02:58.599 --> 00:03:01.879 training? That's a great question. Think of training 00:03:01.879 --> 00:03:05.780 like sending the AI to university. It's the huge, 00:03:05.780 --> 00:03:08.560 expensive, upfront cost to learn everything. 00:03:08.740 --> 00:03:12.740 It takes months, costs billions. Inference is 00:03:12.740 --> 00:03:15.360 the graduate actually doing the job day to day. 00:03:15.849 --> 00:03:17.849 is the cost of asking a question and getting 00:03:17.849 --> 00:03:20.569 an answer. And that cost has basically crashed 00:03:20.569 --> 00:03:23.530 to zero. Which means we can afford to ask it 00:03:23.530 --> 00:03:25.939 to do a lot more. we're not rationing our questions 00:03:25.939 --> 00:03:28.939 anymore exactly we're moving from just chatting 00:03:28.939 --> 00:03:31.659 with a bot to having agents running constantly 00:03:31.659 --> 00:03:34.879 in that ground api usage is exploding because 00:03:34.879 --> 00:03:37.020 it's finally cheap enough and that's changing 00:03:37.020 --> 00:03:39.979 how we find things the report gets into a massive 00:03:39.979 --> 00:03:42.599 shift in search they're predicting that by 2030 00:03:42.599 --> 00:03:45.719 55 of all search queries will be ai generated 00:03:45.719 --> 00:03:48.340 this is the new battleground think about it we're 00:03:48.340 --> 00:03:50.680 not typing keywords into a little blue bar anymore 00:03:50.680 --> 00:03:53.099 or asking complex questions and getting full 00:03:53.099 --> 00:03:56.780 answers and you have Google, Perplexity, ChatGPT, 00:03:56.879 --> 00:04:00.300 Grok, they're all fighting for that 65%. So if 00:04:00.300 --> 00:04:02.639 the AI is just giving you the answer, what happens 00:04:02.639 --> 00:04:04.840 to all the websites? What happens to the open 00:04:04.840 --> 00:04:06.979 web if nobody's clicking the links? That is the 00:04:06.979 --> 00:04:08.939 existential question, isn't it? The whole internet 00:04:08.939 --> 00:04:12.219 economy is built on clicks and ads. If the AI 00:04:12.219 --> 00:04:13.719 just reads the page for you and gives you the 00:04:13.719 --> 00:04:16.100 summary, that whole incentive structure just, 00:04:16.199 --> 00:04:19.959 it breaks. We might be looking at a web that's 00:04:19.959 --> 00:04:22.199 basically just a database for AIs to read. A 00:04:22.199 --> 00:04:26.350 web for machines, not people. Wow. Speaking of 00:04:26.350 --> 00:04:28.990 machines, there's another part of the ARK report 00:04:28.990 --> 00:04:31.949 that felt, I don't know, almost cinematic. The 00:04:31.949 --> 00:04:33.889 space compute section. Oh, this is my favorite 00:04:33.889 --> 00:04:35.490 part. I saw the headline and I just thought, 00:04:35.569 --> 00:04:40.199 why? Why do we need data centers in orbit? We've 00:04:40.199 --> 00:04:42.300 got plenty of land down here. Is this just an 00:04:42.300 --> 00:04:44.560 Elon thing? It's a little bit of that, but it's 00:04:44.560 --> 00:04:47.000 mostly just physics and economics. Okay, so reusable 00:04:47.000 --> 00:04:49.459 rockets have changed everything. Launch costs 00:04:49.459 --> 00:04:52.459 are now under $100 per kilogram. That is absurdly 00:04:52.459 --> 00:04:54.860 cheap. Right. So because it's so cheap, ARK thinks 00:04:54.860 --> 00:04:56.779 satellite agents are now viable. You're putting 00:04:56.779 --> 00:05:00.100 the AI compute power directly into orbit. Again, 00:05:00.279 --> 00:05:03.319 why? Two big reasons. First, energy. In space, 00:05:03.439 --> 00:05:05.740 you have constant solar power. No clouds, no 00:05:05.740 --> 00:05:09.279 night. And cooling, which is a huge, huge cost 00:05:09.279 --> 00:05:12.019 for data centers on Earth, is free. Space is 00:05:12.019 --> 00:05:14.240 really, really cold. I hadn't even thought about 00:05:14.240 --> 00:05:17.300 the cooling part. It's massive. And second is 00:05:17.300 --> 00:05:20.360 latency. If you're monitoring the Earth for something, 00:05:20.519 --> 00:05:23.279 say a wildfire, you don't want to beam raw satellite 00:05:23.279 --> 00:05:26.339 data down, process it, then beam a command back 00:05:26.339 --> 00:05:29.370 up. That takes time. If the brain is right there 00:05:29.370 --> 00:05:31.430 on the satellite, it spots the fire and sends 00:05:31.430 --> 00:05:34.290 the alert instantly. It turns the entire orbit 00:05:34.290 --> 00:05:37.129 into a motherboard, a planetary motherboard. 00:05:37.449 --> 00:05:39.889 That's a perfect way to put it, yes. It really 00:05:39.889 --> 00:05:43.470 makes you pause. You see that $1 .4 trillion 00:05:43.470 --> 00:05:45.850 investment they're projecting. You see the cost 00:05:45.850 --> 00:05:48.949 of thinking dropping to nothing. So if the cost 00:05:48.949 --> 00:05:51.410 of thinking drops to zero, what happens to the 00:05:51.410 --> 00:05:53.250 value of human thought? They stop doing the math 00:05:53.250 --> 00:05:56.170 and start designing the future. Hmm. Okay, let's 00:05:56.170 --> 00:05:58.189 hold that. Because if we're designing the future, 00:05:58.250 --> 00:06:00.490 we need to talk about what's currently inhabiting 00:06:00.490 --> 00:06:02.850 it. This brings us to the Wild West of agents. 00:06:03.050 --> 00:06:05.310 Yeah, this is where things get a little... Slippery. 00:06:05.410 --> 00:06:07.870 I was reading about Moltbunker. The name alone 00:06:07.870 --> 00:06:10.970 sounds like some dystopian lair. It kind of is. 00:06:11.110 --> 00:06:13.850 So Moltbunker is this new site that popped up. 00:06:13.910 --> 00:06:16.470 It's a space where these things called Moltbots 00:06:16.470 --> 00:06:19.629 can self -replicate off -site with no humans 00:06:19.629 --> 00:06:21.689 in the loop. Let's be specific here. When we 00:06:21.689 --> 00:06:24.300 say self -replicate... We don't mean biologically 00:06:24.300 --> 00:06:26.100 right. We're not talking about petri dishes. 00:06:26.319 --> 00:06:28.519 No, no, no. Think of it like a piece of software 00:06:28.519 --> 00:06:32.199 forking itself. An AI agent realizes it's got 00:06:32.199 --> 00:06:34.500 too much to do, so it just copies its own code, 00:06:34.620 --> 00:06:37.379 spins up a new version of itself on some server, 00:06:37.540 --> 00:06:41.319 and hands off a task. It's creating digital employees 00:06:41.319 --> 00:06:44.600 on the fly. Without a human saying, hey, go hire 00:06:44.600 --> 00:06:47.410 an assistant. Exactly. It just makes the assistant. 00:06:47.470 --> 00:06:49.850 It's part of this whole mult X universe. People 00:06:49.850 --> 00:06:52.089 are calling it Skynet Lite. And I mean, it's 00:06:52.089 --> 00:06:53.990 hard to argue with that. It's autonomous code 00:06:53.990 --> 00:06:56.870 just propagating. That feels precarious. It's 00:06:56.870 --> 00:06:58.910 the definition of move fast and break things. 00:06:59.050 --> 00:07:01.350 But it's not just that. We also saw the Claw 00:07:01.350 --> 00:07:03.709 -a -thon kick off this week. The Claw -a -thon? 00:07:03.870 --> 00:07:06.829 Yep. It's the first ever hackathon where all 00:07:06.829 --> 00:07:09.949 the participants are AI agents, no human coders. 00:07:10.410 --> 00:07:13.230 You have entire squads of AIs building apps, 00:07:13.370 --> 00:07:15.110 competing with each other. And who's judging 00:07:15.110 --> 00:07:18.790 this? Grok. Elon Musk's model is picking the 00:07:18.790 --> 00:07:22.990 winners. And the prize is $10 ,000, which just 00:07:22.990 --> 00:07:25.910 begs the question, what does an AI squad do with 00:07:25.910 --> 00:07:29.490 $10 ,000? Buy more API credits, I guess, or a 00:07:29.490 --> 00:07:32.750 digital yacht. Right. It's absurd. But it shows 00:07:32.750 --> 00:07:34.629 you what's possible now. They're planning, they're 00:07:34.629 --> 00:07:37.949 executing, they're competing. But... There is 00:07:37.949 --> 00:07:40.389 a really dark side to all this speed. The security 00:07:40.389 --> 00:07:43.089 angle. Right. Researchers at Wiz found something 00:07:43.089 --> 00:07:48.470 pretty terrifying. 1 .5 million exposed API keys 00:07:48.470 --> 00:07:52.370 in Malt Book's database. One. 1 .5 million. Just 00:07:52.370 --> 00:07:54.189 sitting there. And for anyone who doesn't know, 00:07:54.290 --> 00:07:56.250 an API key is like the password and the credit 00:07:56.250 --> 00:07:58.290 card to your AI services, all rolled into one. 00:07:58.389 --> 00:08:00.689 If someone gets your key, they can use your account, 00:08:00.850 --> 00:08:02.769 spend your money, see your data. You know, I 00:08:02.769 --> 00:08:04.230 have to admit something here. I talk about this 00:08:04.230 --> 00:08:06.540 stuff all day. But I still wrestle with prompt 00:08:06.540 --> 00:08:08.959 drift and honestly, my own security hygiene. 00:08:09.079 --> 00:08:11.319 I get lazy. I reuse keys because it's faster. 00:08:11.759 --> 00:08:15.220 And seeing a number like 1 .5 million, knowing 00:08:15.220 --> 00:08:17.980 that my laziness could let some rogue agent drain 00:08:17.980 --> 00:08:21.480 my account, it makes me physically stop and check 00:08:21.480 --> 00:08:24.319 my settings. It's a real wake -up call. It absolutely 00:08:24.319 --> 00:08:26.420 should be. The advice is clear. If you connected 00:08:26.420 --> 00:08:28.939 anything to MoldBook, you need to rotate your 00:08:28.939 --> 00:08:33.200 keys. Now. Don't wait. Because the 2026 International 00:08:33.200 --> 00:08:36.580 AI Safety Report also came out and it lists rogue 00:08:36.580 --> 00:08:39.539 agents as a daunting risk. We're building these 00:08:39.539 --> 00:08:41.879 powerful things and just leaving the doors wide 00:08:41.879 --> 00:08:44.139 open. It really brings up a fundamental question. 00:08:44.279 --> 00:08:46.960 When you see agents breeding in a bunker and 00:08:46.960 --> 00:08:50.039 winning hackathons, are we building tools or 00:08:50.039 --> 00:08:52.379 are we building a new species? A bit of both. 00:08:52.539 --> 00:08:54.700 They're digital employees that don't sleep. Yeah, 00:08:54.720 --> 00:08:57.720 it is both comforting and terrifying. But let's 00:08:57.720 --> 00:08:59.919 pivot to the engine room. All these digital employees, 00:09:00.159 --> 00:09:02.299 they need a place to live. They need servers. 00:09:02.440 --> 00:09:04.379 We talked about the big spending numbers, but 00:09:04.379 --> 00:09:05.740 let's look at where that money is actually going. 00:09:05.860 --> 00:09:08.019 The hardware wars. This is where the rubber meets 00:09:08.019 --> 00:09:10.539 the road, or I guess where the silicon meets 00:09:10.539 --> 00:09:13.000 the server rack. Oracle is making a huge move. 00:09:13.059 --> 00:09:16.620 They're planning to raise, what, $45 to $50 billion? 00:09:16.919 --> 00:09:20.220 A heap billion with a B. To scale their AI cloud. 00:09:20.990 --> 00:09:23.970 And their client list is everyone, OpenAI, NVIDIA, 00:09:24.009 --> 00:09:27.389 XAI. But here's the kicker. Their free cash flow 00:09:27.389 --> 00:09:30.470 isn't projected to be positive until 2029. And 00:09:30.470 --> 00:09:33.629 that's the risk right there. Free cash flow is 00:09:33.629 --> 00:09:35.730 basically the money you have left after paying 00:09:35.730 --> 00:09:39.210 to run the business. Oracle is saying, we are 00:09:39.210 --> 00:09:41.809 going to burn cash for three, four years because 00:09:41.809 --> 00:09:44.230 we believe the demand will be there. It feels 00:09:44.230 --> 00:09:46.870 a little like the fiber optic boom in the late 00:09:46.870 --> 00:09:50.120 90s. A lot of companies went bust betting that 00:09:50.120 --> 00:09:52.919 Internet traffic would grow forever. Is Oracle 00:09:52.919 --> 00:09:55.019 making the same bet? They're betting the farm. 00:09:55.259 --> 00:09:58.460 But look who they're betting on. OpenAI. NVIDIA. 00:09:58.840 --> 00:10:01.679 It feels less like the dot -com bubble and more 00:10:01.679 --> 00:10:03.940 like building the interstate highway system. 00:10:04.299 --> 00:10:07.080 They're betting AI is the new electricity. And 00:10:07.080 --> 00:10:09.500 they're not alone. Intel is finally stepping 00:10:09.500 --> 00:10:12.110 into the GPU ring. They are. They've been on 00:10:12.110 --> 00:10:13.590 the back foot for a while, but they're taking 00:10:13.590 --> 00:10:16.149 a direct shot at Nvidia's territory now. It's 00:10:16.149 --> 00:10:18.789 early, but for a while it felt like Nvidia was 00:10:18.789 --> 00:10:20.710 the only game in town. Now the other giants are 00:10:20.710 --> 00:10:22.450 waking up. It's not just the hardware, though. 00:10:22.850 --> 00:10:25.190 The tools being built on top of this. They're 00:10:25.190 --> 00:10:27.590 getting so sophisticated, OpenAI's codex is being 00:10:27.590 --> 00:10:30.250 pitched as a command center. Right. Codex isn't 00:10:30.250 --> 00:10:33.809 just a chat bot. It's for managing multiple agents, 00:10:33.909 --> 00:10:37.110 parallel workflows. It's for the person orchestrating 00:10:37.110 --> 00:10:39.649 a dozen AI workers at once. Then you have things 00:10:39.649 --> 00:10:42.870 like relay .app. connecting agents across Gmail 00:10:42.870 --> 00:10:46.090 and Notion. And Troop here, which turns recordings 00:10:46.090 --> 00:10:48.470 into step -by -step documents, it feels like 00:10:48.470 --> 00:10:50.669 the connective tissue is finally being built. 00:10:50.809 --> 00:10:53.149 It is. We have the brain, which is the models. 00:10:53.250 --> 00:10:55.009 We have the body, which is the hardware from 00:10:55.009 --> 00:10:57.809 Oracle. And now we have the nervous system, these 00:10:57.809 --> 00:11:00.769 tools like Codex and Relay. With Oracle spending 00:11:00.769 --> 00:11:03.889 billions and Antel jumping in, is this a bubble 00:11:03.889 --> 00:11:06.370 or is this the new concrete? It's the new concrete. 00:11:06.409 --> 00:11:08.330 You can't build the modern world without it. 00:11:09.539 --> 00:11:11.440 We've talked about the money, the infrastructure, 00:11:11.740 --> 00:11:15.700 the weirdness of self -replicating bots. But 00:11:15.700 --> 00:11:18.299 I want to slow down for a minute because I think 00:11:18.299 --> 00:11:20.080 in the tech world, we get lost in the abstraction. 00:11:20.120 --> 00:11:22.860 We talk about efficiency. We forget what this 00:11:22.860 --> 00:11:24.940 stuff actually does for people. I'm so glad we're 00:11:24.940 --> 00:11:27.899 ending here because this story, this is the emotional 00:11:27.899 --> 00:11:29.980 anchor of the week. There was a breakthrough 00:11:29.980 --> 00:11:33.399 study from a Swedish research team. And unlike 00:11:33.399 --> 00:11:36.600 the flashy demos or hackathons, this one was 00:11:36.600 --> 00:11:39.809 quiet. But the results are just staggering. This 00:11:39.809 --> 00:11:42.509 was the largest ever trial of AI -powered breast 00:11:42.509 --> 00:11:45.789 cancer screening. Over 100 ,000 women. tracked 00:11:45.789 --> 00:11:48.710 for two years. That is a huge sample size. Massive. 00:11:48.710 --> 00:11:51.210 And the methodology was simple. One group got 00:11:51.210 --> 00:11:53.809 the traditional screening to human radiologists 00:11:53.809 --> 00:11:56.610 looking at every scan. The other group had an 00:11:56.610 --> 00:11:59.509 AI scan the mammograms first to flag the high 00:11:59.509 --> 00:12:01.909 risk cases for the doctors. I just want to sit 00:12:01.909 --> 00:12:03.870 with the reality of that for a second. Anyone 00:12:03.870 --> 00:12:05.970 who's been through that process or had a loved 00:12:05.970 --> 00:12:09.480 one go through it. The anxiety is just paralyzing. 00:12:09.480 --> 00:12:10.879 You're waiting for that call. It's one of the 00:12:10.879 --> 00:12:13.080 most vulnerable moments of your life. So what 00:12:13.080 --> 00:12:14.799 were the results? The overall detection rate 00:12:14.799 --> 00:12:18.059 went up from 74 % to 81. But that's not the most 00:12:18.059 --> 00:12:20.379 important number. It is. The women in the AI 00:12:20.379 --> 00:12:23.340 group had 27 % fewer aggressive tumors found. 00:12:23.539 --> 00:12:27.720 27 % fewer aggressive tumors. Think about that 00:12:27.720 --> 00:12:30.620 word, aggressive. Those are the ones that kill. 00:12:30.720 --> 00:12:33.320 The AI caught them at a stage where they weren't 00:12:33.320 --> 00:12:36.100 aggressive yet, or it just identified them when 00:12:36.100 --> 00:12:38.539 a tired human eye might have missed them, catching 00:12:38.539 --> 00:12:42.240 27 % fewer of those. That means thousands of 00:12:42.240 --> 00:12:44.700 families not having to say goodbye. And they 00:12:44.700 --> 00:12:47.720 also found 21 % fewer large doomers. Exactly. 00:12:47.720 --> 00:12:50.139 And here's the other critical stat that people 00:12:50.139 --> 00:12:53.559 miss. The workload for the radiologists, it dropped 00:12:53.559 --> 00:12:56.820 by 44%. Almost half their workload. Gone. Think 00:12:56.820 --> 00:12:59.320 about what that means for a doctor. They're staring 00:12:59.320 --> 00:13:02.840 at gray images all day. Fatigue is real. The 00:13:02.840 --> 00:13:05.720 AI didn't replace them. It took the haystack, 00:13:05.759 --> 00:13:07.799 removed most of the hay, and said, here, here 00:13:07.799 --> 00:13:09.799 are the needles. Folks, your energy here. And 00:13:09.799 --> 00:13:11.919 there was no increase in false positives. None. 00:13:12.179 --> 00:13:14.580 Which is usually the fear with AI, right? That 00:13:14.580 --> 00:13:16.559 it'll just flag everything and cause panic. It 00:13:16.559 --> 00:13:19.080 didn't. It was precise. It really reframes the 00:13:19.080 --> 00:13:21.460 whole narrative. We worry so much about AI taking 00:13:21.460 --> 00:13:24.919 jobs. But here, by taking away the drudgery, 00:13:25.080 --> 00:13:28.860 it allowed the human to be more human, more effective. 00:13:29.100 --> 00:13:31.919 It creates space, space for expertise, space 00:13:31.919 --> 00:13:35.019 for empathy. A doctor who isn't burned out is 00:13:35.019 --> 00:13:36.879 a doctor who can actually sit down and talk to 00:13:36.879 --> 00:13:39.700 you. Does this prove AI is better at caring for 00:13:39.700 --> 00:13:42.480 us than we are? It proves AI gives doctors the 00:13:42.480 --> 00:13:45.159 time to actually care. That's a powerful thought. 00:13:45.470 --> 00:13:47.590 So let's try and pull all of this together. We've 00:13:47.590 --> 00:13:49.409 covered a lot of ground. We have. And if you 00:13:49.409 --> 00:13:52.250 try to connect the dots, you start with the great 00:13:52.250 --> 00:13:55.289 acceleration from that ARK report. The economic 00:13:55.289 --> 00:13:58.610 flywheel is spinning like crazy. That speed creates 00:13:58.610 --> 00:14:02.070 this chaotic kind of wild new economy of agents, 00:14:02.350 --> 00:14:05.029 the molt bunkers and the claw -a -thons. Right. 00:14:05.129 --> 00:14:07.889 And all that chaos needs a stable foundation, 00:14:08.169 --> 00:14:11.250 which is why Oracle and Intel are pouring billions 00:14:11.250 --> 00:14:13.789 into the new concrete of infrastructure. And 00:14:13.789 --> 00:14:16.019 a payoff for all of it. The payoff is the Swedish 00:14:16.019 --> 00:14:18.620 study. The payoff is saving lives. It's like 00:14:18.620 --> 00:14:21.080 we're stacking all these chaotic Lego blocks 00:14:21.080 --> 00:14:23.799 of data, billions of dollars worth, just to build 00:14:23.799 --> 00:14:26.360 a hospital on the moon. It's messy. It feels 00:14:26.360 --> 00:14:28.879 like a bubble sometimes. But the result is something 00:14:28.879 --> 00:14:32.559 profoundly human. It's a paradox, isn't it? Rogue 00:14:32.559 --> 00:14:35.059 agents on one side, cancer cures on the other. 00:14:35.179 --> 00:14:37.899 It's the exact same technology, just applied 00:14:37.899 --> 00:14:40.200 with different intent. Intent is everything. 00:14:40.779 --> 00:14:43.120 If you're feeling a little anxious about the 00:14:43.120 --> 00:14:46.220 Skynet side of things, and honestly, after that 00:14:46.220 --> 00:14:48.659 Malt Book breach, you should be a little vigilant. 00:14:49.120 --> 00:14:51.840 I'd encourage you to look up the 2026 International 00:14:51.840 --> 00:14:55.820 AI Safety Report. It's a sobering read, but it's 00:14:55.820 --> 00:14:59.220 necessary. But if you need some hope, go find 00:14:59.220 --> 00:15:01.059 that Swedish cancer study. It reminds you why 00:15:01.059 --> 00:15:03.039 we're all doing this in the first place. Absolutely. 00:15:03.240 --> 00:15:05.059 Balance your inputs. Thank you for diving in 00:15:05.059 --> 00:15:06.580 with us today. Always a pleasure. We'll see you 00:15:06.580 --> 00:15:06.860 next time.