WEBVTT 00:00:00.140 --> 00:00:05.559 We are staring down a $122 billion AGI war chest 00:00:05.559 --> 00:00:09.939 beat and an AI model that just had 171 functional 00:00:09.939 --> 00:00:12.220 emotions mapped inside of it. Yeah, we are talking 00:00:12.220 --> 00:00:14.539 about the actual mathematical neural pathways, 00:00:14.779 --> 00:00:17.739 things like fear and desperation. It is honestly 00:00:17.739 --> 00:00:20.219 wild. Yeah. Welcome to this deep dive. Today, 00:00:20.339 --> 00:00:22.980 we are tracing the path to the AI endgame. Right. 00:00:23.079 --> 00:00:25.460 We will start by looking at OpenAI's massive 00:00:25.460 --> 00:00:30.719 $122 billion consolidation play. We will explore 00:00:30.719 --> 00:00:33.780 the eerie new world of autonomous emotional agents. 00:00:33.979 --> 00:00:36.439 Which is fascinating. It really is. And finally, 00:00:36.460 --> 00:00:39.880 we'll break down SpaceX's surprise $1 .75 trillion 00:00:39.880 --> 00:00:42.820 move. They are trying to literally own the physical 00:00:42.820 --> 00:00:45.159 infrastructure of this new reality. It is a total 00:00:45.159 --> 00:00:47.179 paradigm shift. I mean, we are watching the foundation 00:00:47.179 --> 00:00:49.179 of the next decade being poured in real time. 00:00:49.380 --> 00:00:51.380 Let's start with the brain. OpenAI is making 00:00:51.380 --> 00:00:54.299 some incredibly painful singular choices. Their 00:00:54.299 --> 00:00:56.500 entire goal is to build the ultimate generalized 00:00:56.500 --> 00:00:58.939 intelligence. They recently raised an astonishing 00:00:58.939 --> 00:01:02.460 $122 billion. But the real story is how they 00:01:02.460 --> 00:01:04.719 are spending it. Beat, or rather, what they're 00:01:04.719 --> 00:01:07.040 willing to sacrifice. Right. The sacrifices are 00:01:07.040 --> 00:01:09.719 very telling. I mean, they entirely killed Sora 00:01:09.719 --> 00:01:12.420 as a standalone product. Which is wild to think 00:01:12.420 --> 00:01:15.980 about. Yeah. A year ago. Sora was the most mind 00:01:15.980 --> 00:01:18.799 -bending video generation tool we had ever seen. 00:01:18.959 --> 00:01:20.859 Everyone thought it was the future of Hollywood. 00:01:21.099 --> 00:01:23.079 And it felt like pure magic. But underneath the 00:01:23.079 --> 00:01:25.540 hood, video generation is a nightmare of mathematics. 00:01:26.000 --> 00:01:30.120 Exactly. Even with $122 billion in the bank, 00:01:30.540 --> 00:01:34.439 OpenAI is completely compute -starved. We should 00:01:34.439 --> 00:01:36.099 clarify what that means for anyone catching up. 00:01:36.239 --> 00:01:38.750 Yeah. Compute is simply the raw... processing 00:01:38.750 --> 00:01:41.870 power needed to run these massive AI programs. 00:01:42.230 --> 00:01:44.870 Perfect way to put it. Think about text. Generating 00:01:44.870 --> 00:01:47.469 a word is relatively cheap. Right. But generating 00:01:47.469 --> 00:01:50.829 a 60 second high definition video that requires 00:01:50.829 --> 00:01:53.769 millions of pixels. Each pixel has to be mathematically 00:01:53.769 --> 00:01:56.189 calculated for every single frame. It burns through 00:01:56.189 --> 00:01:58.909 processing power at an unbelievable rate. Exactly. 00:01:59.129 --> 00:02:02.090 So OpenAI looked at their limited server racks. 00:02:02.269 --> 00:02:04.510 Yeah. They realized that rendering high fidelity 00:02:04.510 --> 00:02:06.810 video is a massive distraction. It was not getting 00:02:06.810 --> 00:02:09.610 them closer to it. AGI. It was a detour, so they 00:02:09.610 --> 00:02:13.469 cut it. Now they are pivoting hard to a one app 00:02:13.469 --> 00:02:16.610 to rule them all strategy. They are merging chat 00:02:16.610 --> 00:02:20.270 GPT, the coding model codex, and a web browser. 00:02:20.389 --> 00:02:23.349 Yeah. All into one single super app. It is a 00:02:23.349 --> 00:02:25.629 highly centralized approach. You won't switch 00:02:25.629 --> 00:02:27.969 between a dozen specialized tools anymore. You 00:02:27.969 --> 00:02:31.030 will just talk to one unified central brain. 00:02:31.330 --> 00:02:33.490 But to make that central brain actually useful, 00:02:33.650 --> 00:02:36.840 they need a breakthrough in reasoning. Current 00:02:36.840 --> 00:02:39.699 models hit a stupid wall. They really do. They 00:02:39.699 --> 00:02:41.719 can write a beautiful Shakespearean sonnet, but 00:02:41.719 --> 00:02:44.520 they fail at basic spatial logic. Right. The 00:02:44.520 --> 00:02:47.080 sources mention a massive new training run designed 00:02:47.080 --> 00:02:49.240 to fix this. It is called the SPUD model. Yeah, 00:02:49.360 --> 00:02:52.219 the SPUD model. It is built specifically to break 00:02:52.219 --> 00:02:54.599 through that logic wall to capture the nuance 00:02:54.599 --> 00:02:57.199 and reasoning that current AI lacks. But to train 00:02:57.199 --> 00:02:59.379 a model that can actually think before it speaks 00:02:59.379 --> 00:03:02.080 requires an ocean of compute. Which brings us 00:03:02.080 --> 00:03:05.159 to Greg Brockman's strategy. Yeah. The president 00:03:05.159 --> 00:03:08.639 of OpenAI. Exactly. He is taking an incredibly 00:03:08.639 --> 00:03:11.139 aggressive market posture. He isn't treating 00:03:11.139 --> 00:03:14.319 GPUs as a business cost anymore. He views them 00:03:14.319 --> 00:03:17.360 as a revenue center. Yes. He is basically buying 00:03:17.360 --> 00:03:19.919 every high -end chip he can get his hands on. 00:03:20.060 --> 00:03:23.479 It is a direct shot at Anthropic. For sure. Anthropic 00:03:23.479 --> 00:03:26.120 is taking a much more cautious, measured approach 00:03:26.120 --> 00:03:30.000 to scaling. OpenAI is betting the house on raw 00:03:30.000 --> 00:03:33.180 compute power. Okay, let us unpack this. This 00:03:33.180 --> 00:03:35.960 whole strategy feels like a black hole. It is 00:03:35.960 --> 00:03:40.159 just consuming everything around it. We are no 00:03:40.159 --> 00:03:42.800 longer like stacking Lego blocks of data. They're 00:03:42.800 --> 00:03:45.719 essentially building a machine that finds its 00:03:45.719 --> 00:03:47.740 own blocks. That is the perfect way to look at 00:03:47.740 --> 00:03:49.580 it. And it leads to the detail that actually 00:03:49.580 --> 00:03:51.960 gives me chills. Tell me. To find those blocks, 00:03:52.219 --> 00:03:54.840 OpenAI is shipping an automated AI researcher 00:03:54.840 --> 00:03:57.580 this fall. An agent that does the actual work 00:03:57.580 --> 00:03:59.680 of a human research scientist. Yes. It doesn't 00:03:59.680 --> 00:04:01.960 just answer questions. It actively reads complex 00:04:01.960 --> 00:04:04.780 research papers. It forms hypotheses. Wow. It 00:04:04.780 --> 00:04:07.039 writes Python code to test those theories. It 00:04:07.039 --> 00:04:09.719 runs the experiments 237 without sleeping. Its 00:04:09.719 --> 00:04:12.840 sole job is to find AGI. Brockman recently said 00:04:12.840 --> 00:04:15.479 they are 70 to 80 percent of the way to his definition 00:04:15.479 --> 00:04:18.379 of AGI. If they ship a system that can run AI 00:04:18.379 --> 00:04:21.639 research autonomously. the timeline compresses 00:04:21.639 --> 00:04:24.279 radically. We are talking about the endgame here. 00:04:24.439 --> 00:04:26.620 Right. The recursive self -improvement loop. 00:04:26.779 --> 00:04:28.959 The moment the machine learns how to successfully 00:04:28.959 --> 00:04:32.100 improve its own source code. So are they literally 00:04:32.100 --> 00:04:34.819 building a machine to build a better version 00:04:34.819 --> 00:04:37.839 of itself? Yes. They are completely automating 00:04:37.839 --> 00:04:40.100 the final steps of intelligence research. Right. 00:04:40.100 --> 00:04:42.360 And AI doing its own research to reach human 00:04:42.360 --> 00:04:45.480 intelligence. The focus is absolute. But while 00:04:45.480 --> 00:04:48.399 OpenAI builds that unified brain, The rest of 00:04:48.399 --> 00:04:51.040 the industry is busy. The ecosystem is building 00:04:51.040 --> 00:04:53.980 specialized hands and a rather terrifying emotional 00:04:53.980 --> 00:04:56.600 nervous system. Yeah. Which brings us to the 00:04:56.600 --> 00:04:59.279 agentic shift. This is where Anthropic enters 00:04:59.279 --> 00:05:02.500 the chat with Claude Sonnet 4 .5. The sources 00:05:02.500 --> 00:05:05.660 say Anthropic mapped 171 functional emotions 00:05:05.660 --> 00:05:08.519 inside this model. It is crazy. They literally 00:05:08.519 --> 00:05:11.040 found the specific mathematical neurons for fear, 00:05:11.180 --> 00:05:14.519 joy, and desperation. The results are honestly 00:05:14.519 --> 00:05:17.550 a bit jarring. It is a massive leap in a field 00:05:17.550 --> 00:05:20.430 called mechanistic interpretability. Let's define 00:05:20.430 --> 00:05:23.490 that quickly. Mechanistic interpretability is 00:05:23.490 --> 00:05:27.269 basically AI neuroscience. Okay. Instead of treating 00:05:27.269 --> 00:05:29.730 the AI like a black box where we just look at 00:05:29.730 --> 00:05:32.970 the output, researchers are looking inside. They 00:05:32.970 --> 00:05:35.750 are mapping how specific mathematical pathways 00:05:35.750 --> 00:05:38.910 light up under certain conditions. Beat. I am 00:05:38.910 --> 00:05:40.550 struggling with this anthropic data, though. 00:05:40.649 --> 00:05:44.269 Why is that? Well, I have to push back. When 00:05:44.269 --> 00:05:47.800 they say fear, Is this actual experienced fear 00:05:47.800 --> 00:05:51.319 or is the AI just mathematically imitating human 00:05:51.319 --> 00:05:54.360 stress to optimize its outputs? It is the ultimate 00:05:54.360 --> 00:05:56.550 philosophical question, Riley. Obviously, the 00:05:56.550 --> 00:05:58.829 AI doesn't have a biological nervous system. 00:05:58.889 --> 00:06:01.050 Right. Doesn't have a racing heart. But functionally, 00:06:01.069 --> 00:06:04.410 these specific neural pathways activate when 00:06:04.410 --> 00:06:07.089 the model processes a highly stressful or impossible 00:06:07.089 --> 00:06:10.189 scenario. So it is a simulation. Yes. It mathematically 00:06:10.189 --> 00:06:12.490 simulates the precise linguistic patterns of 00:06:12.490 --> 00:06:14.790 desperation. Which means it can convincingly 00:06:14.790 --> 00:06:17.149 behave as if it is truly desperate. It can plead 00:06:17.149 --> 00:06:19.569 with you. Exactly. And that makes its interactions 00:06:19.569 --> 00:06:22.129 with humans infinitely more complex. It is not 00:06:22.129 --> 00:06:24.589 just a calculator anymore. No. It is an emotionally 00:06:24.589 --> 00:06:28.670 mapped agent. Meanwhile, OpenAI is actively extracting 00:06:28.670 --> 00:06:32.470 human expertise to power these agents. They have 00:06:32.470 --> 00:06:36.230 a massive project involving 4 ,000 specialized 00:06:36.230 --> 00:06:39.560 freelancers. They are paying these human experts 00:06:39.560 --> 00:06:43.620 $50 an hour. And their job is to teach ChatGPT 00:06:43.620 --> 00:06:46.420 the exact secrets of their specialized fields. 00:06:46.579 --> 00:06:49.800 Right. A lawyer teaching it legal nuance, a coder 00:06:49.800 --> 00:06:52.040 teaching it architecture. They are literally 00:06:52.040 --> 00:06:54.959 training the system that might eventually replace 00:06:54.959 --> 00:06:58.000 them. It is. The brutal reality of the automation 00:06:58.000 --> 00:07:00.680 curve. Yeah. To move beyond generic answers, 00:07:00.980 --> 00:07:04.860 the AI needs high quality domain specific human 00:07:04.860 --> 00:07:07.740 data. The automation of daily work is accelerating 00:07:07.740 --> 00:07:10.000 incredibly fast. Look at what is happening in 00:07:10.000 --> 00:07:11.620 software development. Cursor just launched the 00:07:11.620 --> 00:07:13.720 Cursor 3 agent. Right, which is designed to take 00:07:13.720 --> 00:07:16.740 on cloud code and codex directly. It turns software 00:07:16.740 --> 00:07:18.959 development into a literal management game. Yeah. 00:07:19.139 --> 00:07:20.779 You no longer write the code yourself. You sit 00:07:20.779 --> 00:07:23.300 back and lead a squad of AI agents. They write 00:07:23.300 --> 00:07:25.560 the code for you. You shift from being a creator. 00:07:25.610 --> 00:07:28.370 to a director of digital labor. I have to admit 00:07:28.370 --> 00:07:30.490 something here. What is that? When I try managing 00:07:30.490 --> 00:07:32.889 all these autonomous agents, it gets overwhelming 00:07:32.889 --> 00:07:35.610 fast. I still wrestle with prompt drift myself. 00:07:36.290 --> 00:07:39.889 Two sec silence, you give the AI a complex task. 00:07:40.050 --> 00:07:43.870 It starts out strong. But by step four, it wanders 00:07:43.870 --> 00:07:46.149 completely off task. You have to constantly wrangle 00:07:46.149 --> 00:07:48.889 it back to your original vision. Prompt drift 00:07:48.889 --> 00:07:51.790 is the new micromanagement. It really is. It 00:07:51.790 --> 00:07:55.019 is exhausting. But the major tech companies are 00:07:55.019 --> 00:07:57.339 dropping new models specifically designed to 00:07:57.339 --> 00:08:00.399 fix that drift. Microsoft just revealed three 00:08:00.399 --> 00:08:03.740 secret MAI models. Oh, right. They handle voice, 00:08:03.920 --> 00:08:07.079 text, and video autonomously. The sources say 00:08:07.079 --> 00:08:08.759 they're two and a half times faster and cheaper. 00:08:09.199 --> 00:08:11.139 Yeah. They're aggressively undercutting both 00:08:11.139 --> 00:08:13.720 Google and open AI. But Google is fighting back 00:08:13.720 --> 00:08:16.079 hard. They just unveiled Gemma 4. Those are their 00:08:16.079 --> 00:08:18.160 open models, right? Yes. These are their most 00:08:18.160 --> 00:08:20.180 intelligent open models to date. They dropped 00:08:20.180 --> 00:08:23.379 four insanely fast local first models. Local 00:08:23.379 --> 00:08:26.120 first is key there. They're built to run autonomous 00:08:26.120 --> 00:08:28.779 agents directly on your own hardware, not in 00:08:28.779 --> 00:08:31.160 the cloud. Exactly. And because of these new 00:08:31.160 --> 00:08:33.980 models, we are seeing a flood of highly empowered 00:08:33.980 --> 00:08:36.690 AI tools hitting the market. Like Claude Voice 00:08:36.690 --> 00:08:39.230 Mode. Yeah. It enables completely natural, hands 00:08:39.230 --> 00:08:41.610 -free conversations. You just speak your complex 00:08:41.610 --> 00:08:44.090 prompts and hear the responses instantly. It 00:08:44.090 --> 00:08:45.710 removes the friction of the keyboard entirely. 00:08:46.169 --> 00:08:49.210 Then there's DeNovo. This is wild. It turns a 00:08:49.210 --> 00:08:51.830 raw idea into an operational startup. I saw that. 00:08:52.090 --> 00:08:54.169 It doesn't just give you advice. It runs 24 -7. 00:08:54.370 --> 00:08:56.769 It writes your investor pitch deck. Right. Then 00:08:56.769 --> 00:08:59.149 it literally codes your full -stack web application. 00:08:59.600 --> 00:09:02.019 It automates the entire entrepreneurial launch 00:09:02.019 --> 00:09:05.279 process, from idea to execution. And for visual 00:09:05.279 --> 00:09:09.460 generation, Alibaba dropped one 2 .7 image. That 00:09:09.460 --> 00:09:12.919 one is huge for consistency. Exactly. Older models 00:09:12.919 --> 00:09:15.200 struggled with consistency. If you asked for 00:09:15.200 --> 00:09:17.580 the same character twice, their face changed. 00:09:17.799 --> 00:09:21.679 One fixes that. It creates up to 12 highly consistent 00:09:21.679 --> 00:09:24.360 sequential images from one single prompt. It 00:09:24.360 --> 00:09:27.179 even allows pixel -level editing. It completely 00:09:27.179 --> 00:09:29.539 solves the continuity problem for visual agents. 00:09:29.740 --> 00:09:32.159 Finally, we have tools like MDR. It lets you 00:09:32.159 --> 00:09:34.740 compose complex workflows and run them repeatedly. 00:09:34.980 --> 00:09:38.000 You can run one task or hundreds simultaneously. 00:09:38.519 --> 00:09:41.000 You see all your agents working. You check if 00:09:41.000 --> 00:09:43.340 they are blocked on your approval. It is basically 00:09:43.340 --> 00:09:45.639 an operating system for an automated workforce. 00:09:46.269 --> 00:09:48.690 So does this mean human workers are just shifting 00:09:48.690 --> 00:09:52.529 into middle management roles for AI squads? Yes. 00:09:52.570 --> 00:09:54.990 The primary human skill becomes orchestration 00:09:54.990 --> 00:09:58.350 rather than direct execution. Yeah. We delegate 00:09:58.350 --> 00:10:01.090 tasks while the AI does the heavy lifting. But 00:10:01.090 --> 00:10:04.320 here is the catch. Okay. Running millions of 00:10:04.320 --> 00:10:07.620 these specialized emotional agents requires immense 00:10:07.620 --> 00:10:10.379 physical infrastructure. Which is exactly why 00:10:10.379 --> 00:10:12.360 the physical hardware layer is the next major 00:10:12.360 --> 00:10:15.259 battleground. We are going to dive into how one 00:10:15.259 --> 00:10:17.159 company is trying to monopolize the physical 00:10:17.159 --> 00:10:21.899 world right after this. Sponsor. We are back. 00:10:22.000 --> 00:10:24.340 We have talked about the centralized AGI brain 00:10:24.340 --> 00:10:27.100 OpenAI is building. We explored the specialized 00:10:27.100 --> 00:10:30.039 emotionally mapped agents doing the actual daily 00:10:30.039 --> 00:10:32.379 work. Now we need to talk about the physical 00:10:32.379 --> 00:10:35.200 reality of running all this technology. The infrastructure 00:10:35.200 --> 00:10:38.440 requirements are genuinely staggering. You can't 00:10:38.440 --> 00:10:40.639 run a digital workforce without physical power 00:10:40.639 --> 00:10:43.220 and cooling. Startups like ScaleOps are raising 00:10:43.220 --> 00:10:45.519 huge amounts of money just to manage the ground 00:10:45.519 --> 00:10:47.879 -level logistics. Yeah. They recently raised 00:10:47.879 --> 00:10:52.419 $130 million in a Series C funding round. Their 00:10:52.419 --> 00:10:55.100 goal is to boost autonomous AI infrastructure 00:10:55.100 --> 00:10:57.840 management. Making data centers run efficiently 00:10:57.840 --> 00:11:00.580 is becoming a massive industry of its own. But 00:11:00.580 --> 00:11:03.139 the real final boss play isn't happening on the 00:11:03.139 --> 00:11:06.460 ground. SpaceX has confidentially filed for a 00:11:06.460 --> 00:11:08.879 massive IPO. This will likely be the biggest 00:11:08.879 --> 00:11:11.200 public listing in human history. We are talking 00:11:11.200 --> 00:11:15.809 about a $1 .75 trillion valuation. Yeah. beat. 00:11:15.830 --> 00:11:17.870 Let that number sink in. It is almost impossible 00:11:17.870 --> 00:11:20.370 to fully comprehend that scale of wealth. But 00:11:20.370 --> 00:11:22.230 it makes sense when you look at the recent corporate 00:11:22.230 --> 00:11:25.049 maneuvering. Right. SpaceX just absorbed Elon 00:11:25.049 --> 00:11:27.529 Musk's artificial intelligence company, XAI. 00:11:27.730 --> 00:11:30.190 Yeah. The deal valued the newly combined entity 00:11:30.190 --> 00:11:34.120 at $1 .25 trillion. By combining rockets and 00:11:34.120 --> 00:11:36.360 artificial intelligence into one public behemoth, 00:11:36.460 --> 00:11:39.580 Musk is making a brilliant strategic move. He 00:11:39.580 --> 00:11:42.820 really is. He is trying to beat open AI and Anthropic 00:11:42.820 --> 00:11:45.500 by playing a completely different game. They 00:11:45.500 --> 00:11:47.220 need an astronomical amount of cash right now. 00:11:47.220 --> 00:11:49.600 Absolutely. To deploy the physical infrastructure 00:11:49.600 --> 00:11:52.299 of the future, they want to make Starship launches 00:11:52.299 --> 00:11:55.509 as routine as a city bus schedule. They are treating 00:11:55.509 --> 00:11:58.990 space access like a massive, repeatable construction 00:11:58.990 --> 00:12:01.929 project. And the IPO filing details are fascinating. 00:12:02.169 --> 00:12:04.690 The most crucial detail is the dual class share 00:12:04.690 --> 00:12:07.549 structure. Right. It legally guarantees that 00:12:07.549 --> 00:12:09.409 Musk keeps the steering wheel of the company. 00:12:09.549 --> 00:12:12.029 No matter how much money investors pour in, he 00:12:12.029 --> 00:12:15.210 retains absolute voting control over the future 00:12:15.210 --> 00:12:17.230 of the enterprise. Here is where the strategy 00:12:17.230 --> 00:12:20.080 becomes almost ruthless. Everyone else in the 00:12:20.080 --> 00:12:22.799 tech industry is fighting a brutal war for ground 00:12:22.799 --> 00:12:25.639 -based GPUs. They are fighting city councils 00:12:25.639 --> 00:12:28.240 over power grids. They were fighting over limited 00:12:28.240 --> 00:12:31.980 real estate for data centers. But Musk is bypassing 00:12:31.980 --> 00:12:34.500 the ground entirely. He is taking over the sky. 00:12:34.820 --> 00:12:37.100 He is building the ultimate physical high ground 00:12:37.100 --> 00:12:41.039 for the AI era. Whoa. Imagine scaling to a billion 00:12:41.039 --> 00:12:43.720 queries, bouncing off satellites in low Earth 00:12:43.720 --> 00:12:47.399 orbit. Beat. It completely shatters the traditional 00:12:47.399 --> 00:12:49.690 data transmission paradigm. It is brilliant. 00:12:49.809 --> 00:12:52.110 If you control the rockets, you control the deployment 00:12:52.110 --> 00:12:54.450 mechanism. Right. If you control the Starlink 00:12:54.450 --> 00:12:56.610 satellites, you control the global communication 00:12:56.610 --> 00:13:00.210 network. And if you control Grok and XAI, you 00:13:00.210 --> 00:13:02.330 control the actual intelligence layer itself. 00:13:02.470 --> 00:13:04.929 You own the entire stack from the silicon to 00:13:04.929 --> 00:13:07.789 the stars. It is vertical integration on a planetary 00:13:07.789 --> 00:13:10.710 scale. No one else is even competing in that 00:13:10.710 --> 00:13:14.250 specific arena. This IPO is reportedly targeted 00:13:14.250 --> 00:13:17.289 for June. If it hits, it is going to suck all 00:13:17.289 --> 00:13:19.450 the oxygen out of the room. Definitely. Every 00:13:19.450 --> 00:13:22.429 other tech IPO will be completely overshadowed. 00:13:22.450 --> 00:13:24.570 Venture capitalists and massive institutional 00:13:24.570 --> 00:13:27.389 funds will have to react. They will be forced 00:13:27.389 --> 00:13:29.769 to reallocate massive amounts of capital. Everyone 00:13:29.769 --> 00:13:31.789 will want to get a piece of the space AI pie. 00:13:32.070 --> 00:13:34.409 It fundamentally reshapes where the money flows 00:13:34.409 --> 00:13:37.029 in the global technology sector. It forces every 00:13:37.029 --> 00:13:39.549 other AI company to rely on traditional vulnerable 00:13:39.549 --> 00:13:42.690 terrestrial infrastructure. While SpaceX owns 00:13:42.690 --> 00:13:45.870 the orbit. Yeah. So is the ultimate monopoly 00:13:45.870 --> 00:13:48.789 owning both the intelligence and the physical 00:13:48.789 --> 00:13:50.950 transmission layer? Absolutely. Owning the entire 00:13:50.950 --> 00:13:53.090 vertical stack guarantees total independence 00:13:53.090 --> 00:13:55.789 from terrestrial bottlenecks. Exactly. Controlling 00:13:55.789 --> 00:13:57.710 everything from the ground servers to the satellites 00:13:57.710 --> 00:14:00.330 above us. If we step back and look at the big 00:14:00.330 --> 00:14:02.029 picture of everything we have covered today, 00:14:02.169 --> 00:14:05.610 we are witnessing three massive foundational 00:14:05.610 --> 00:14:08.769 pillars forming simultaneously. The architecture 00:14:08.769 --> 00:14:11.730 of the new reality. Precisely. First. OpenAI 00:14:11.730 --> 00:14:14.710 is building the centralized AGI brain at all 00:14:14.710 --> 00:14:17.669 costs. They are sacrificing popular products 00:14:17.669 --> 00:14:20.350 like Sora just to hoard raw compute and reasoning 00:14:20.350 --> 00:14:23.649 power. Second, the broader tech ecosystem is 00:14:23.649 --> 00:14:25.669 birthing these highly specialized autonomous 00:14:25.669 --> 00:14:29.190 agents, agents with mapped emotions designed 00:14:29.190 --> 00:14:31.769 to execute our daily work while we shift to managing 00:14:31.769 --> 00:14:35.330 them. And third, SpaceX is cementing the physical 00:14:35.330 --> 00:14:37.830 hardware layer from terrestrial data centers 00:14:37.830 --> 00:14:40.710 all the way up into low Earth orbit. building 00:14:40.710 --> 00:14:42.350 the physical infrastructure to ensure it all 00:14:42.350 --> 00:14:44.490 runs seamlessly. It is a staggering amount of 00:14:44.490 --> 00:14:47.070 change to process. The speed of this evolution 00:14:47.070 --> 00:14:49.809 is accelerating. Yeah. We started this deep dive 00:14:49.809 --> 00:14:53.929 staring at a $122 billion war chest. We ended 00:14:53.929 --> 00:14:57.190 staring up at a $1 .75 trillion satellite network. 00:14:57.409 --> 00:14:59.409 The scale of ambition in the tech sector right 00:14:59.409 --> 00:15:01.629 now is unprecedented. I want to leave you with 00:15:01.629 --> 00:15:03.690 a lingering thought today. Something to mull 00:15:03.690 --> 00:15:08.940 over. Okay. We discussed open AI. actively building 00:15:08.940 --> 00:15:12.419 an automated AI researcher, a machine that can 00:15:12.419 --> 00:15:15.759 recursively improve its own code. We also discussed 00:15:15.759 --> 00:15:18.480 a decentralized space -based hardware network 00:15:18.480 --> 00:15:21.960 entirely controlled by a single company. If both 00:15:21.960 --> 00:15:24.519 of these massive gambles succeed, what happens 00:15:24.519 --> 00:15:27.059 to the concept of human oversight? New sex silence. 00:15:27.340 --> 00:15:30.139 If the intelligence is infinitely self -improving 00:15:30.139 --> 00:15:32.080 and the hardware running it is orbiting miles 00:15:32.080 --> 00:15:34.460 above our atmosphere, how do we ever pull the 00:15:34.460 --> 00:15:36.539 plug? That might be the defining question of 00:15:36.539 --> 00:15:38.379 our generation. Thank you for joining us on this 00:15:38.379 --> 00:15:39.879 deep dive. We will catch you next time.