WEBVTT 00:00:00.000 --> 00:00:02.740 Right now, sitting on a massive Google server 00:00:02.740 --> 00:00:05.980 is a complete map of a brain. You can trace a 00:00:05.980 --> 00:00:08.820 single microscopic thought. I mean, you can literally 00:00:08.820 --> 00:00:11.119 watch a tiny fruit fly deciding to take flight. 00:00:11.259 --> 00:00:12.900 Yeah, and you can track that single decision 00:00:12.900 --> 00:00:17.079 across 125 million microscopic connections. Right. 00:00:17.260 --> 00:00:20.510 But across the tech world, things look... very 00:00:20.510 --> 00:00:22.750 different. The creators of our most advanced 00:00:22.750 --> 00:00:25.050 artificial intelligence are admitting something 00:00:25.050 --> 00:00:27.829 terrifying. They don't actually know how their 00:00:27.829 --> 00:00:30.210 own models think anymore. The contrast between 00:00:30.210 --> 00:00:32.829 these two fields is genuinely staggering. You 00:00:32.829 --> 00:00:35.450 have extreme biological precision happening on 00:00:35.450 --> 00:00:38.109 one side. We are mapping biology down to the 00:00:38.109 --> 00:00:40.609 absolute smallest synapse. Yeah, and then you 00:00:40.609 --> 00:00:43.429 have rapidly evolving digital chaos on the other. 00:00:44.109 --> 00:00:46.390 Welcome to this deep dive. We are exploring some 00:00:46.390 --> 00:00:49.229 wildly blurring lines today. We really are. It 00:00:49.229 --> 00:00:51.770 is a massive topic. We are going to move from 00:00:51.770 --> 00:00:54.670 biological blueprints to evolving digital minds. 00:00:54.909 --> 00:00:57.509 We'll examine the very dark side of deepfake 00:00:57.509 --> 00:00:59.969 technology. And finally, we will look at the 00:00:59.969 --> 00:01:02.549 heavy -duty hardware containing all of it. Microsoft 00:01:02.549 --> 00:01:05.489 is building literal desktop fortresses for this 00:01:05.489 --> 00:01:08.250 exact purpose. We'd really need to start with 00:01:08.250 --> 00:01:11.480 the biological benchmark first, though. It grounds 00:01:11.480 --> 00:01:13.500 our understanding of how intelligence actually 00:01:13.500 --> 00:01:17.099 forms. Google Research released a truly monumental 00:01:17.099 --> 00:01:19.939 study recently. They worked closely with the 00:01:19.939 --> 00:01:23.659 HHMI Janelia Research Campus. Together, they 00:01:23.659 --> 00:01:27.000 mapped a complete intact male fruit fly brain. 00:01:27.280 --> 00:01:29.280 They just published these findings in the journal 00:01:29.280 --> 00:01:32.879 Cell. absolute scale of the numbers involved 00:01:32.879 --> 00:01:35.140 here is staggering. We are talking about a creature 00:01:35.140 --> 00:01:38.060 smaller than a grain of rice, yet they mapped 00:01:38.060 --> 00:01:42.900 166 ,000 individual neurons. Exactly. And they 00:01:42.900 --> 00:01:46.519 charted roughly 125 million synaptic connections. 00:01:46.840 --> 00:01:49.959 It is a massive biological data set. Just to 00:01:49.959 --> 00:01:52.040 be clear for you listening, a synapse is simply 00:01:52.040 --> 00:01:53.840 the connection point between two nerve cells. 00:01:54.079 --> 00:01:56.439 Right. And they identified almost 12 ,000 specific 00:01:56.439 --> 00:01:59.640 neuron types. The map covers the central brain 00:01:59.640 --> 00:02:02.519 and the optic lobes. It even includes the fly's 00:02:02.519 --> 00:02:04.560 ventral nerve cord. Which functions a lot like 00:02:04.560 --> 00:02:06.480 a human spinal cord, right? Yeah, exactly. It 00:02:06.480 --> 00:02:07.939 controls all the physical movements and basic 00:02:07.939 --> 00:02:11.400 reflexes. Navigating this map feels deeply and 00:02:11.400 --> 00:02:15.520 undeniably surreal. It is like reverse engineering 00:02:15.520 --> 00:02:19.219 an unimaginably complex, microscopic alien spacecraft. 00:02:19.639 --> 00:02:21.819 You are looking at the fundamental machinery 00:02:21.819 --> 00:02:24.169 of life. That's a brilliant way to visualize 00:02:24.169 --> 00:02:26.610 it. The structure is incredibly dense and highly 00:02:26.610 --> 00:02:29.830 purposeful. It took years of intense, sustained 00:02:29.830 --> 00:02:32.490 manual labor to finish this. Right. Human researchers 00:02:32.490 --> 00:02:35.689 had to verify millions of microscopic pathways. 00:02:36.129 --> 00:02:38.430 Yeah. Human experts at Janelia had to manually 00:02:38.430 --> 00:02:41.210 proofread the entire neural map. They built this 00:02:41.210 --> 00:02:44.129 on a tiny version from 2020. That older map only 00:02:44.129 --> 00:02:46.750 had 25 ,000 neurons. The jump in sheer scale 00:02:46.750 --> 00:02:49.370 here is frankly unbelievable. It is a monumental 00:02:49.370 --> 00:02:53.050 achievement in neuroscience. specific details 00:02:53.050 --> 00:02:56.129 during this process. They identified over 7 ,000 00:02:56.129 --> 00:02:59.210 shared neuron types, but they also spotted highly 00:02:59.210 --> 00:03:01.930 specific sexually dimorphic neurons. Meaning 00:03:01.930 --> 00:03:03.930 neurons that differ physically between males 00:03:03.930 --> 00:03:07.009 and females. Yeah, they found 262 male -specific 00:03:07.009 --> 00:03:11.030 types, and they mapped 69 distinct female -specific 00:03:11.030 --> 00:03:13.990 types. These physical differences link directly 00:03:13.990 --> 00:03:17.009 to very specific behaviors. They control complex, 00:03:17.250 --> 00:03:19.550 instinctual actions like courtship and aggression. 00:03:19.889 --> 00:03:23.449 It proves how complex behavior is literally hardwired 00:03:23.449 --> 00:03:26.189 into the biology. The entire dataset is publicly 00:03:26.189 --> 00:03:28.990 available on Google's NeuroGlancer tool. You 00:03:28.990 --> 00:03:31.169 can trace a circuit from visual input directly 00:03:31.169 --> 00:03:33.930 to motor output. It tracks from a faint smell 00:03:33.930 --> 00:03:36.370 down to a physical leg movement. It is entirely 00:03:36.370 --> 00:03:39.430 open for anyone to explore. So how does mapping 00:03:39.430 --> 00:03:42.050 a fly brain actually help us understand higher 00:03:42.050 --> 00:03:44.150 intelligence? Well, it gives us the complete 00:03:44.150 --> 00:03:46.530 foundational wiring diagram. We call this biological 00:03:46.530 --> 00:03:49.310 diagram a connectome. Meaning a comprehensive 00:03:49.310 --> 00:03:52.150 map of all neural connections. Right. If we understand 00:03:52.150 --> 00:03:54.889 how a simple biological system processes information, 00:03:55.389 --> 00:03:58.580 we can learn. We can scale that fundamental logic 00:03:58.580 --> 00:04:01.780 upwards over time. We learn the absolute basic 00:04:01.780 --> 00:04:04.240 algorithms of natural thought. We see exactly 00:04:04.240 --> 00:04:06.620 how nature builds a thinking machine. So mastering 00:04:06.620 --> 00:04:09.180 the fly brain is our stepping stone to mapping 00:04:09.180 --> 00:04:11.840 human minds. Precisely. It provides the grounded 00:04:11.840 --> 00:04:14.560 reality for biological intelligence. We finally 00:04:14.560 --> 00:04:16.360 have a complete schematic to study and learn 00:04:16.360 --> 00:04:19.519 from. From biological wiring, we move to digital 00:04:19.519 --> 00:04:22.910 wiring. And this is where things get... significantly 00:04:22.910 --> 00:04:26.069 more strange. OpenAI recently admitted something 00:04:26.069 --> 00:04:28.370 fascinating about their newest model. Oh yeah, 00:04:28.470 --> 00:04:30.889 this was a wild statement. They publicly stated 00:04:30.889 --> 00:04:34.569 that their AI is an alien mind. That phrasing 00:04:34.569 --> 00:04:37.290 is very deliberate and deeply revealing. It thinks 00:04:37.290 --> 00:04:38.970 differently from us on a fundamental mechanical 00:04:38.970 --> 00:04:41.550 level. They realize the intelligence has grown 00:04:41.550 --> 00:04:44.529 more than it is designed. We have to find totally 00:04:44.529 --> 00:04:46.980 new ways to study it now. We can't just read 00:04:46.980 --> 00:04:49.220 the underlying code anymore. The parameters are 00:04:49.220 --> 00:04:51.959 just too vast and complex. Wait, calling it an 00:04:51.959 --> 00:04:54.439 alien mind sounds a lot like sci -fi marketing 00:04:54.439 --> 00:04:56.699 hype. Are they just saying that to sound profound? 00:04:57.000 --> 00:04:59.579 Or is there an actual mechanical truth to this 00:04:59.579 --> 00:05:02.600 claim? There is a deep mechanical truth to it. 00:05:02.699 --> 00:05:05.319 We don't program these models line by line like 00:05:05.319 --> 00:05:07.759 traditional software. We feed them massive amounts 00:05:07.759 --> 00:05:09.699 of data and let them adjust their own weights. 00:05:09.819 --> 00:05:12.189 Right. The intelligence emerges organically from 00:05:12.189 --> 00:05:14.769 that complex process. It is a black box. Even 00:05:14.769 --> 00:05:17.370 the creators don't fully understand how it arrives 00:05:17.370 --> 00:05:20.509 at an answer. We are seeing these GPT -6 Astra 00:05:20.509 --> 00:05:23.310 demos online right now. People in the industry 00:05:23.310 --> 00:05:25.589 are calling the capabilities genuinely ridiculous. 00:05:25.990 --> 00:05:29.389 Even NVIDIA's CEO is weighing in on this rapid 00:05:29.389 --> 00:05:32.269 progress. The capabilities are escalating incredibly 00:05:32.269 --> 00:05:35.790 fast right now. OpenAI just hit the automated 00:05:35.790 --> 00:05:38.829 research intern milestone internally. What does 00:05:38.829 --> 00:05:41.389 that actually mean in practice? It means Astra 00:05:41.389 --> 00:05:43.970 is currently handling days of actual researcher 00:05:43.970 --> 00:05:47.670 work. It reads scientific papers, synthesizes 00:05:47.670 --> 00:05:50.930 complex data, and proposes new experiments. It 00:05:50.930 --> 00:05:53.189 isn't just answering questions anymore. It is 00:05:53.189 --> 00:05:56.019 actively conducting research. Wow. Yeah, this 00:05:56.019 --> 00:05:58.319 capability pushed their internal roadmap forward 00:05:58.319 --> 00:06:01.439 by six full months. They're essentially automating 00:06:01.439 --> 00:06:03.620 the scientific discovery process itself. And 00:06:03.620 --> 00:06:05.259 they are reportedly already sitting on the Bell 00:06:05.259 --> 00:06:07.740 model behind closed doors. Right. And Bell is 00:06:07.740 --> 00:06:09.879 a completed pre -training run. Which is the initial 00:06:09.879 --> 00:06:12.639 phase where AI learns patterns from massive data. 00:06:12.899 --> 00:06:15.259 Exactly. They are heavily preparing for the post 00:06:15.259 --> 00:06:18.180 -GPT -6 era. The raw intelligence for the next 00:06:18.180 --> 00:06:20.639 generation is already baked. And Astra is not 00:06:20.639 --> 00:06:22.800 even fully rolled out to the public yet. Whoa. 00:06:23.399 --> 00:06:26.899 Imagine scaling to a billion queries. The sheer 00:06:26.899 --> 00:06:29.639 speed of this evolution is incredibly hard to 00:06:29.639 --> 00:06:31.980 grasp. It feels like the fundamental ground is 00:06:31.980 --> 00:06:35.120 shifting daily. The massive financial moves absolutely 00:06:35.120 --> 00:06:39.480 prove the immense scale here. Anthropic is aggressively 00:06:39.480 --> 00:06:42.610 eyeing mid -October, for their IPO marketing. 00:06:42.829 --> 00:06:46.250 They are lining up a $15 billion credit facility. 00:06:46.550 --> 00:06:48.889 That is a massive financial setup for future 00:06:48.889 --> 00:06:50.889 computing power. It fuels the infrastructure 00:06:50.889 --> 00:06:53.670 arms race across the entire board. Look at the 00:06:53.670 --> 00:06:57.129 new robotics data startup named XDOF. They're 00:06:57.129 --> 00:07:01.129 in Series B talks at a $1 .2 billion valuation. 00:07:01.790 --> 00:07:04.370 Unbelievable. They hit $50 million in annual 00:07:04.370 --> 00:07:06.529 recurring revenue instantly, and they're only 00:07:06.529 --> 00:07:09.480 three months out of stealth mode. 20 major customers 00:07:09.480 --> 00:07:11.620 are already using their robot training data. 00:07:11.779 --> 00:07:14.019 The physical world is rapidly merging with these 00:07:14.019 --> 00:07:16.920 alien digital minds. Data is the ultimate bottleneck 00:07:16.920 --> 00:07:18.699 for physical robots right now. Yeah, physical 00:07:18.699 --> 00:07:21.240 data is the new gold rush. If we grow these alien 00:07:21.240 --> 00:07:23.439 minds rather than programming them, how do we 00:07:23.439 --> 00:07:25.519 predict what an automated intern might do next? 00:07:25.860 --> 00:07:28.540 We have to use established behavioral observation 00:07:28.540 --> 00:07:31.839 techniques. We must test them exactly like living 00:07:31.839 --> 00:07:35.279 biological subjects. We set up isolated environments 00:07:35.279 --> 00:07:37.689 and watch how they solve complex problems. So 00:07:37.689 --> 00:07:39.730 treating them like lab animals, essentially. 00:07:39.810 --> 00:07:42.029 Pretty much. It requires a totally different 00:07:42.029 --> 00:07:44.189 scientific framework than traditional computer 00:07:44.189 --> 00:07:47.490 science. You probe the model, observe the outputs, 00:07:47.709 --> 00:07:50.310 and map the behavior. We can't just program them. 00:07:50.350 --> 00:07:52.709 We have to study how they actually grow. Exactly. 00:07:52.709 --> 00:07:55.569 And that rapid, unpredictable growth creates 00:07:55.569 --> 00:07:58.649 massive vulnerabilities everywhere. The models 00:07:58.649 --> 00:08:00.889 scale much faster than our collective understanding. 00:08:01.189 --> 00:08:03.230 They are being weaponized just as quickly as 00:08:03.230 --> 00:08:05.089 they are productized. We are seeing the fallout 00:08:05.089 --> 00:08:08.509 right now in the real world. A fake CEO deepfake 00:08:08.509 --> 00:08:12.430 video call almost scammed a $400 ,000 deal. It 00:08:12.430 --> 00:08:14.870 is getting scary fast out there. The digital 00:08:14.870 --> 00:08:17.050 illusion of humanity is becoming nearly perfect. 00:08:17.329 --> 00:08:20.009 The scammers didn't just use a prerecorded video 00:08:20.009 --> 00:08:23.129 file. They used a live, fully interactive video 00:08:23.129 --> 00:08:26.279 deepfake. The deepfake responded in real time 00:08:26.279 --> 00:08:28.019 to the conversation. Which is terrifying. They 00:08:28.019 --> 00:08:30.459 cloned the voice perfectly. And they even captured 00:08:30.459 --> 00:08:33.019 his unique visual mannerisms and facial expressions. 00:08:33.379 --> 00:08:36.220 It bypassed normal corporate trust protocols 00:08:36.220 --> 00:08:39.240 completely. Because humans are biologically wired 00:08:39.240 --> 00:08:42.179 to trust what they see and hear. When a video 00:08:42.179 --> 00:08:45.620 perfectly mimics a person, our biological connectome 00:08:45.620 --> 00:08:48.570 trusts it. We desperately need everyday tools 00:08:48.570 --> 00:08:51.269 to manage this escalating chaos. We really do. 00:08:51.389 --> 00:08:54.350 We need entirely new mental frameworks to navigate 00:08:54.350 --> 00:08:57.110 this reality. I found that viral Toyota root 00:08:57.110 --> 00:08:59.909 cause prompt really interesting. It applies their 00:08:59.909 --> 00:09:02.889 famous manufacturing logic directly to personal 00:09:02.889 --> 00:09:06.149 AI workflows. It traces issues back to the real 00:09:06.149 --> 00:09:08.429 source instead of patching symptoms. Yeah, that 00:09:08.429 --> 00:09:10.929 prompt framework is incredibly useful. It uses 00:09:10.929 --> 00:09:13.149 the five whys method to dig deep into a problem. 00:09:13.799 --> 00:09:16.600 If an AI gives a bad output, you don't just ask 00:09:16.600 --> 00:09:18.740 it again. You figure out why your prompt architecture 00:09:18.740 --> 00:09:21.299 failed in the first place. I still wrestle with 00:09:21.299 --> 00:09:24.259 prompt drift myself. We all do, honestly. Prompt 00:09:24.259 --> 00:09:26.899 drift is when an AI slowly forgets your original 00:09:26.899 --> 00:09:29.620 instructions over time. The context window gets 00:09:29.620 --> 00:09:32.370 cluttered and the model loses its focus. That 00:09:32.370 --> 00:09:34.649 is exactly why we need these new management tools. 00:09:34.809 --> 00:09:37.509 We have to enforce strict operational boundaries 00:09:37.509 --> 00:09:40.629 on these models. Tools like the AI Toolbox are 00:09:40.629 --> 00:09:43.289 really stepping up to help. It organizes and 00:09:43.289 --> 00:09:45.669 searches every single chat in one centralized 00:09:45.669 --> 00:09:49.570 place. It works seamlessly across ChatGPT, Claude, 00:09:49.710 --> 00:09:52.649 Gemini, and Grok. That one is huge. They have 00:09:52.649 --> 00:09:55.250 over 40 ,000 active users globally right now. 00:09:55.370 --> 00:09:58.309 It centralizes your scattered interactions with 00:09:58.309 --> 00:10:00.690 all these different alien minds. You are constantly 00:10:00.690 --> 00:10:02.529 switching tabs. and losing your train of thought 00:10:02.529 --> 00:10:04.730 then there is an interesting new tool called 00:10:04.730 --> 00:10:07.789 to data it is an AI employee living directly 00:10:07.789 --> 00:10:10.610 inside your slack workspace right it learns your 00:10:10.610 --> 00:10:13.690 company culture and gets real work done It integrates 00:10:13.690 --> 00:10:16.350 deeply with your existing team dynamic. It acts 00:10:16.350 --> 00:10:18.289 more like a colleague than a simple chatbot. 00:10:18.529 --> 00:10:20.750 We also have automated monitoring tools like 00:10:20.750 --> 00:10:24.009 Notify. It monitors domains across all top -level 00:10:24.009 --> 00:10:27.149 registries 20 times daily. It alerts you instantly 00:10:27.149 --> 00:10:29.789 when a clear opportunity appears. It handles 00:10:29.789 --> 00:10:31.789 the tedious background monitoring perfectly. 00:10:32.210 --> 00:10:35.710 And Docsalot is a really powerful visual editor 00:10:35.710 --> 00:10:38.730 for developers. It creates technical documentation 00:10:38.730 --> 00:10:41.750 using AI edits and detailed version history. 00:10:42.299 --> 00:10:44.360 Yeah, it helps remote teams build beautiful, 00:10:44.539 --> 00:10:47.700 reusable components very easily. It streamlines 00:10:47.700 --> 00:10:50.399 the entire technical writing process. These tools 00:10:50.399 --> 00:10:52.779 are integrating deeply into our daily professional 00:10:52.779 --> 00:10:55.720 routines. They are rapidly becoming our primary 00:10:55.720 --> 00:10:58.240 interface with this chaotic technology. They're 00:10:58.240 --> 00:11:00.639 acting as a buffer, really. So do these everyday 00:11:00.639 --> 00:11:03.159 management tools actually protect us, or do they 00:11:03.159 --> 00:11:05.940 just make us more reliant on the tech? They definitely 00:11:05.940 --> 00:11:07.919 do a bit of both, if we are being completely 00:11:07.919 --> 00:11:11.019 honest. They insulate us from the raw, unpredictable 00:11:11.019 --> 00:11:14.259 chaos of the Internet. They put a friendly, usable 00:11:14.259 --> 00:11:16.860 interface over an alien mind. But there is a 00:11:16.860 --> 00:11:19.840 catch. Right. The root cause thinking framework 00:11:19.840 --> 00:11:21.919 remains absolutely essential for your survival. 00:11:22.200 --> 00:11:24.460 It forces you to critically question the underlying 00:11:24.460 --> 00:11:27.580 data. You stop blindly reacting to the superficial 00:11:27.580 --> 00:11:30.259 generated output. You start actively looking 00:11:30.259 --> 00:11:32.360 for the actual mechanical truth of the situation. 00:11:32.909 --> 00:11:35.309 Basically, better root cause thinking helps us 00:11:35.309 --> 00:11:37.870 spot the illusions in our daily workflows. Right. 00:11:37.990 --> 00:11:39.830 You have to maintain your own critical thinking, 00:11:39.950 --> 00:11:42.250 which brings us directly to the physical containment 00:11:42.250 --> 00:11:45.049 strategy. Yes, the hardware side of things. Frontier 00:11:45.049 --> 00:11:47.389 cloud models are becoming incredibly powerful 00:11:47.389 --> 00:11:50.870 and increasingly alien. Deep fakes and autonomous 00:11:50.870 --> 00:11:54.549 AI agents are running totally rampant. Relying 00:11:54.549 --> 00:11:57.149 solely on the cloud is becoming a massive security 00:11:57.149 --> 00:12:00.009 liability. You are sending sensitive data out 00:12:00.009 --> 00:12:02.440 into the unknown. Microsoft is building something 00:12:02.440 --> 00:12:05.879 major to address this exact vulnerability. They 00:12:05.879 --> 00:12:08.559 just officially announced Project Zenith. It 00:12:08.559 --> 00:12:11.799 is a ready to code Windows 11 setup. It is built 00:12:11.799 --> 00:12:15.559 specifically for high end developer PCs. It focuses 00:12:15.559 --> 00:12:18.740 entirely on running heavy local AI workloads. 00:12:19.000 --> 00:12:21.500 Right. They're bringing the raw power out of 00:12:21.500 --> 00:12:23.779 the cloud and onto your desk. The hardware specs 00:12:23.779 --> 00:12:26.419 on this machine are completely wild. It requires 00:12:26.419 --> 00:12:29.059 at least 64 gigabytes of unified memory. Unified 00:12:29.059 --> 00:12:31.580 memory is a massive leap forward for local AI. 00:12:31.799 --> 00:12:34.519 It lets the CPU and GPU share data seamlessly 00:12:34.519 --> 00:12:37.000 without moving it. Traditional computers have 00:12:37.000 --> 00:12:39.059 separate memory pools. And moving data between 00:12:39.059 --> 00:12:41.779 them creates a massive bottleneck for AI processing. 00:12:42.019 --> 00:12:44.820 Exactly. Unified memory solves that problem elegantly. 00:12:44.980 --> 00:12:49.299 And it requires a staggering 250 gigabytes per 00:12:49.299 --> 00:12:52.340 second of memory bandwidth. That extreme speed 00:12:52.340 --> 00:12:55.919 is absolutely critical for local AI workloads. 00:12:56.299 --> 00:13:00.059 AI models are incredibly massive files. If your 00:13:00.059 --> 00:13:02.320 computer's memory is too slow, the processor 00:13:02.320 --> 00:13:05.490 simply starves. It sits there waiting for data 00:13:05.490 --> 00:13:08.490 to arrive. The bandwidth widens the highway so 00:13:08.490 --> 00:13:11.470 the AI can actually think in real time. Memory 00:13:11.470 --> 00:13:13.809 bandwidth is basically the speed at which you 00:13:13.809 --> 00:13:16.549 hand pieces to a master builder. It moves information 00:13:16.549 --> 00:13:19.049 instantly, like stacking Lego blocks of data. 00:13:19.210 --> 00:13:21.889 Exactly. It completely removes the physical friction 00:13:21.889 --> 00:13:24.269 from the hardware. It lets you run massive 30 00:13:24.269 --> 00:13:26.909 billion parameter models locally. And parameters 00:13:26.909 --> 00:13:29.370 are just the internal variables an AI uses to 00:13:29.370 --> 00:13:31.870 make decisions. Right. You run these huge models 00:13:31.870 --> 00:13:33.649 locally and completely unmetered. You don't have 00:13:33.649 --> 00:13:35.149 to worry about an internet connection dropping. 00:13:35.350 --> 00:13:38.809 It launches first on the new AMD Ryzen AI Halo 00:13:38.809 --> 00:13:41.429 hardware. More silicon partners are definitely 00:13:41.429 --> 00:13:44.129 coming later down the line. Microsoft also cleaned 00:13:44.129 --> 00:13:45.929 up the developer software settings significantly. 00:13:46.669 --> 00:13:49.450 File extensions and hidden files are finally 00:13:49.450 --> 00:13:52.639 enabled by default. Full paths and long path 00:13:52.639 --> 00:13:54.879 support are standardized right out of the box. 00:13:55.019 --> 00:13:57.639 That is a nice touch. It removes the usual annoying 00:13:57.639 --> 00:13:59.799 friction of setting up a clean coating environment. 00:14:00.220 --> 00:14:03.279 It is designed by developers, strictly for developers. 00:14:03.539 --> 00:14:05.679 The security angle is what really catches my 00:14:05.679 --> 00:14:08.360 eye here, though. They built dedicated infrastructure 00:14:08.360 --> 00:14:11.620 specifically for local AI agents. Yeah, this 00:14:11.620 --> 00:14:13.700 is arguably the most crucial part of the entire 00:14:13.700 --> 00:14:16.740 Zenith project. They are using Microsoft Execution 00:14:16.740 --> 00:14:20.080 Containers, or MXC for short. They also added 00:14:20.080 --> 00:14:23.039 incredibly strict OS -level identity controls. 00:14:24.159 --> 00:14:26.600 Why are those execution containers so necessary 00:14:26.600 --> 00:14:29.600 for a local AI? Because if you have an autonomous 00:14:29.600 --> 00:14:32.500 agent running on your PC, you need strict boundaries. 00:14:32.679 --> 00:14:34.580 You don't want it accidentally deleting your 00:14:34.580 --> 00:14:36.919 critical system files. You don't want it emailing 00:14:36.919 --> 00:14:39.159 your private documents to a random server. Right. 00:14:39.279 --> 00:14:41.500 That would be a disaster. It isolates the local 00:14:41.500 --> 00:14:44.460 AI agent activity completely. It sandboxes the 00:14:44.460 --> 00:14:47.980 intelligence. It is basically a secure physical 00:14:47.980 --> 00:14:51.080 box for the alien mind. There is no specific 00:14:51.080 --> 00:14:54.360 price or exact shipping date announced yet. But 00:14:54.360 --> 00:14:56.480 the clear intention behind this hardware is very 00:14:56.480 --> 00:14:59.320 obvious. Microsoft wants to secure the future 00:14:59.320 --> 00:15:01.539 of AI development. They want to lock it down 00:15:01.539 --> 00:15:04.000 locally. Why would a developer use this instead 00:15:04.000 --> 00:15:06.940 of just relying on frontier cloud models? Because 00:15:06.940 --> 00:15:09.840 cloud API token costs add up incredibly quickly. 00:15:10.220 --> 00:15:13.419 Every single prompt costs real money. You want 00:15:13.419 --> 00:15:15.500 to shift your routine coding and agent workloads 00:15:15.500 --> 00:15:19.200 locally. It saves a fortune over a year of heavy 00:15:19.200 --> 00:15:21.409 development. You save the expensive frontier 00:15:21.409 --> 00:15:24.190 cloud models for the truly difficult, complex 00:15:24.190 --> 00:15:26.590 problems. Plus, the local physical isolation 00:15:26.590 --> 00:15:29.429 provides massive, undeniable security benefits. 00:15:29.690 --> 00:15:31.649 Cloud is for the heavy lifting. Local hardware 00:15:31.649 --> 00:15:34.210 handles the everyday grind. Exactly. It beautifully 00:15:34.210 --> 00:15:37.070 balances extreme computational capability with 00:15:37.070 --> 00:15:39.710 practical everyday security. We are going to 00:15:39.710 --> 00:15:41.570 take a quick break here. Insert the sponsor read 00:15:41.570 --> 00:15:44.970 here. And we are back. Okay, let's pull all of 00:15:44.970 --> 00:15:47.429 these massive concepts together into one big 00:15:47.429 --> 00:15:51.509 picture. We are simultaneously mapping. biological 00:15:51.509 --> 00:15:54.090 hardware with insane unprecedented precision. 00:15:54.350 --> 00:15:58.409 We are charting the 125 million synapses of a 00:15:58.409 --> 00:16:00.990 fruit fly brain. We are finding the physical 00:16:00.990 --> 00:16:03.789 hardwiring of biological behavior. At the exact 00:16:03.789 --> 00:16:05.850 same time, we are growing completely synthetic 00:16:05.850 --> 00:16:09.549 software. We are actively watching OpenAI's alien 00:16:09.549 --> 00:16:13.049 mind and the Astra intern rapidly evolve. We're 00:16:13.049 --> 00:16:15.190 letting them learn and grow organically. Right. 00:16:15.250 --> 00:16:17.549 And because these digital minds are scaling faster 00:16:17.549 --> 00:16:19.980 than our understanding, We are adapting. We are 00:16:19.980 --> 00:16:22.460 now building dedicated, hyper -secured desktop 00:16:22.460 --> 00:16:25.279 fortresses. Project Zenith is designed just to 00:16:25.279 --> 00:16:27.879 safely contain these new digital entities. We 00:16:27.879 --> 00:16:30.379 are bridging biology and silicon in real time. 00:16:30.730 --> 00:16:32.870 It is a lot to process all at once. The pace 00:16:32.870 --> 00:16:34.710 of this change is truly relentless. It really 00:16:34.710 --> 00:16:37.710 is. If an AI is genuinely an alien mind and we 00:16:37.710 --> 00:16:39.649 grow it rather than code it, and we are giving 00:16:39.649 --> 00:16:41.850 it its own isolated execution container on a 00:16:41.850 --> 00:16:44.330 Zenith PC, at what point does your computer stop 00:16:44.330 --> 00:16:46.429 being a tool? When does it start being a terrarium 00:16:46.429 --> 00:16:48.970 for digital life? That is the defining philosophical 00:16:48.970 --> 00:16:52.169 question of our current era. Thank you for joining 00:16:52.169 --> 00:16:53.149 us on this deep dive.