WEBVTT 00:00:00.000 --> 00:00:05.580 In 2026, we have a $2 .4 trillion AI startup 00:00:05.580 --> 00:00:08.640 ecosystem. And they're building lasers that literally 00:00:08.640 --> 00:00:12.140 zap weeds on farms. But at the exact same time... 00:00:12.140 --> 00:00:14.300 At the exact same time, people are running AI 00:00:14.300 --> 00:00:16.640 detectors to argue over a single punctuation 00:00:16.640 --> 00:00:19.679 mark in a Nike tweet? Yeah, the whiplash between 00:00:19.679 --> 00:00:21.820 those two realities is just, it's pretty wild. 00:00:21.899 --> 00:00:24.079 We are automating physical reality, yet we're 00:00:24.079 --> 00:00:27.140 still hyper fixating on like... Human fingerprints 00:00:27.140 --> 00:00:30.980 in text. Welcome to a new deep dive. We have 00:00:30.980 --> 00:00:33.500 a massive stack of updates for you today. So 00:00:33.500 --> 00:00:36.000 let's lay out the roadmap. First, we are looking 00:00:36.000 --> 00:00:39.119 at a fundamental power shift. Huge shift. Anthropic 00:00:39.119 --> 00:00:41.920 just dethroned OpenAI on the CNBC Disruptor 50 00:00:41.920 --> 00:00:43.920 list. We're going to trace where that staggering 00:00:43.920 --> 00:00:47.280 $2 .4 trillion is actually flowing. Then we will 00:00:47.280 --> 00:00:49.320 look at how Google is radically changing access 00:00:49.320 --> 00:00:52.500 with their IO 2026 subscriptions. Yeah, and then 00:00:52.500 --> 00:00:54.640 we'll unpack the cutthroat talent wars that are 00:00:54.640 --> 00:00:56.780 driving the bleeding edge of these models. And 00:00:56.780 --> 00:00:58.939 finally, we will explore the hyperlocal tools 00:00:58.939 --> 00:01:01.240 hitting your desktop right now. The ground is 00:01:01.240 --> 00:01:03.500 definitely shifting under our feet. I mean, the 00:01:03.500 --> 00:01:05.760 sheer speed of this evolution is really what 00:01:05.760 --> 00:01:08.659 we are trying to decode today. Let's start with 00:01:08.659 --> 00:01:12.799 that macro shift. The CNBC Disruptor 50 list 00:01:12.799 --> 00:01:16.659 for 2026 just dropped. And the headline is impossible 00:01:16.659 --> 00:01:20.060 to ignore. Anthropic has officially taken the 00:01:20.060 --> 00:01:22.680 number one spot. Yeah, they push OpenAI down 00:01:22.680 --> 00:01:25.340 to number two, which is, you know, a seismic 00:01:25.340 --> 00:01:28.260 shift in Silicon Valley. Right. For years, OpenAI 00:01:28.260 --> 00:01:30.459 was just the undisputed heavyweight champion. 00:01:30.579 --> 00:01:33.129 They completely defined the category. And now 00:01:33.129 --> 00:01:35.930 the challenger wears the crown. And the money 00:01:35.930 --> 00:01:38.469 following these companies, well, it's almost 00:01:38.469 --> 00:01:41.109 hard to comprehend. It really is. Total funding 00:01:41.109 --> 00:01:44.569 for the top 50 jumped from $127 billion last 00:01:44.569 --> 00:01:48.670 year all the way to $337 billion this year. Wow. 00:01:49.159 --> 00:01:52.439 That drives a $2 .4 trillion implied value. Just 00:01:52.439 --> 00:01:54.980 think about that capital concentration. It is 00:01:54.980 --> 00:01:57.340 completely unprecedented. And geographically, 00:01:57.400 --> 00:01:59.079 it's highly localized, too. Yeah, California 00:01:59.079 --> 00:02:01.280 claims nearly half the list, right? Exactly. 00:02:01.519 --> 00:02:04.900 There are 23 California companies in total, and 00:02:04.900 --> 00:02:07.560 14 of those are sitting right in San Francisco 00:02:07.560 --> 00:02:11.479 alone. It's a modern gold rush. But the tools 00:02:11.479 --> 00:02:13.280 have completely changed. I mean, if you look 00:02:13.280 --> 00:02:14.979 at the new categories forming, they're wild. 00:02:15.560 --> 00:02:18.080 We have prediction markets now. Yeah, like Polymarket 00:02:18.080 --> 00:02:20.979 and Kalshi. They're tracking real world events 00:02:20.979 --> 00:02:24.120 and just aggregating human belief. But then we 00:02:24.120 --> 00:02:26.349 also have something called vibe coding. Companies 00:02:26.349 --> 00:02:29.449 like Cursor and Lovable are really leading this 00:02:29.449 --> 00:02:32.449 charge. It's so cool. Vibe coding basically means 00:02:32.449 --> 00:02:36.050 writing software using only plain everyday language. 00:02:36.189 --> 00:02:38.210 Which completely democratizes creation. Exactly. 00:02:38.210 --> 00:02:40.509 You do not need to know syntax anymore. You just 00:02:40.509 --> 00:02:42.629 need to know what you want to build and the AI 00:02:42.629 --> 00:02:44.569 handles the rest. But then on the other end, 00:02:44.590 --> 00:02:47.620 we have everyday automation. AI is hitting the 00:02:47.620 --> 00:02:50.479 physical world hard. Carbon robotics is using 00:02:50.479 --> 00:02:53.379 those lasers we mentioned to clear weeds on farms. 00:02:53.539 --> 00:02:55.680 And Harvey is upgrading the massive infrastructure 00:02:55.680 --> 00:02:58.419 of the entire legal industry. It's a profound 00:02:58.419 --> 00:03:01.460 collision, honestly. We are moving way past just 00:03:01.460 --> 00:03:04.139 chatbots on screens. We are entering physical 00:03:04.139 --> 00:03:06.580 reality. Okay, let's unpack this for a second. 00:03:06.960 --> 00:03:10.219 Why is something as abstract as vibe coding suddenly 00:03:10.219 --> 00:03:12.460 sharing the stage with heavy physical automation? 00:03:12.860 --> 00:03:15.699 Like... Why is capital splitting between software 00:03:15.699 --> 00:03:18.219 that writes itself and weed -zapping lasers on 00:03:18.219 --> 00:03:20.560 a farm? If we connect this to the bigger picture, 00:03:20.680 --> 00:03:22.639 it's because they are deeply connected. How so? 00:03:23.099 --> 00:03:25.919 Well, as software generation becomes effortless, 00:03:26.060 --> 00:03:29.360 capital naturally flows to harder physical world 00:03:29.360 --> 00:03:33.020 problems. We have largely solved text. So now 00:03:33.020 --> 00:03:35.719 we solve reality. So software creates itself, 00:03:35.860 --> 00:03:38.639 freeing AI to tackle physical reality. Exactly. 00:03:38.639 --> 00:03:40.900 And if capital is flooding into physical automation 00:03:40.900 --> 00:03:43.560 because software is solving itself, that leaves 00:03:43.560 --> 00:03:45.860 a massive question. Right. How are the legacy 00:03:45.860 --> 00:03:48.699 tech giants actually packaging this new reality 00:03:48.699 --> 00:03:51.500 for the average consumer? Exactly. Which brings 00:03:51.500 --> 00:03:55.080 us right to Google's massive IO 2026 announcements. 00:03:55.300 --> 00:03:57.539 We are seeing a major subscription shake up here. 00:03:57.780 --> 00:04:00.360 Google is fundamentally changing how they monetize 00:04:00.360 --> 00:04:03.340 these tools for you. They introduced a new $100 00:04:03.340 --> 00:04:06.560 AI Ultra plan. This is built specifically for 00:04:06.560 --> 00:04:09.759 creators and like tech leads. The specs on this 00:04:09.759 --> 00:04:12.740 are aggressive. I mean, you get five times higher 00:04:12.740 --> 00:04:15.439 usage limits than the standard pro plan. Yep. 00:04:15.800 --> 00:04:19.000 And it integrates Gemini 3 .5 flash. Plus, you 00:04:19.000 --> 00:04:22.319 get a massive 20 terabytes of storage. 20 terabytes. 00:04:22.699 --> 00:04:26.040 And you get priority access to Google anti -gravity 00:04:26.040 --> 00:04:28.699 plus full YouTube premium. But they also dropped 00:04:28.699 --> 00:04:31.259 the price on the absolute top tier Ultra plan. 00:04:31.339 --> 00:04:34.300 It went from $250 down to $200. Right. And it 00:04:34.300 --> 00:04:36.959 keeps the 20x usage limits. And it has Project 00:04:36.959 --> 00:04:39.480 Genie. And even the standard paid pro subscribers 00:04:39.480 --> 00:04:42.259 get a perk. They now get YouTube premium light. 00:04:42.649 --> 00:04:45.129 bundled right in. The pricing is a clear land 00:04:45.129 --> 00:04:48.930 grab, in my opinion. But the capabilities are 00:04:48.930 --> 00:04:51.209 what actually matter here. They announced Gemini 00:04:51.209 --> 00:04:54.550 Omni. Oh, this is a major upgrade for video creation. 00:04:54.810 --> 00:04:57.209 You can flawlessly blend real -world footage 00:04:57.209 --> 00:04:59.829 with AI in Google Flow. And it keeps perfect 00:04:59.829 --> 00:05:01.829 character consistency across all the scenes. 00:05:02.069 --> 00:05:03.649 Which solves one of the biggest headaches in 00:05:03.649 --> 00:05:06.209 AI video. Consistency was always the breaking 00:05:06.209 --> 00:05:08.750 point. Before this, characters would morph slightly 00:05:08.750 --> 00:05:11.649 in every single frame. Right. And Gemini Omni 00:05:11.649 --> 00:05:14.689 solved that character consistency problem by 00:05:14.689 --> 00:05:17.089 fundamentally changing its spatial memory. Yeah, 00:05:17.129 --> 00:05:18.829 it doesn't just generate frame by frame anymore. 00:05:18.930 --> 00:05:21.149 It anchors the 3D geometry of the character. 00:05:22.300 --> 00:05:25.699 What's fascinating here is the shift from just 00:05:25.699 --> 00:05:29.769 generation to persistent memory. But Gemini Spark 00:05:29.769 --> 00:05:32.910 is the real game changer for daily life. Totally. 00:05:32.990 --> 00:05:35.250 It is available for the ultra tiers, U .S. only 00:05:35.250 --> 00:05:38.189 right now, though. It's a 24 -7 autonomous agent. 00:05:38.370 --> 00:05:41.110 It connects the dots across all your Google products 00:05:41.110 --> 00:05:44.089 to handle complex tasks entirely in the background. 00:05:44.290 --> 00:05:47.569 They also announced the new AI inbox and daily 00:05:47.569 --> 00:05:50.750 brief. So Gmail will now highlight your priorities 00:05:50.750 --> 00:05:52.629 for you. It will draft replies for you, too. 00:05:52.689 --> 00:05:54.930 And it gives you a clean morning digest of your 00:05:54.930 --> 00:05:57.250 calendar and your past chats. It's doing a lot 00:05:57.250 --> 00:06:00.550 of heavy lifting. I have to admit, I still wrestle 00:06:00.550 --> 00:06:02.910 with trusting an AI to fully manage my inbox 00:06:02.910 --> 00:06:05.350 without sounding like a robot. Yeah, that makes 00:06:05.350 --> 00:06:07.649 sense. The nuance of human communication is just 00:06:07.649 --> 00:06:10.350 so hard to fake. But Google is pushing really 00:06:10.350 --> 00:06:12.990 hard to ease that friction. They are. They're 00:06:12.990 --> 00:06:15.110 also changing how they handle limits. Usage caps 00:06:15.110 --> 00:06:17.490 now refresh every five hours. Right. And if you 00:06:17.490 --> 00:06:19.970 do hit your cap, they do not just lock you out 00:06:19.970 --> 00:06:23.490 anymore. Google seamlessly shifts you to a smaller, 00:06:23.610 --> 00:06:25.970 ultra -fast model. So your work never stops. 00:06:26.480 --> 00:06:29.360 But let me challenge that for a second. Do these 00:06:29.360 --> 00:06:31.939 rolling five -hour limits and fallbacks finally 00:06:31.939 --> 00:06:35.040 solve the dreaded cap anxiety for power users? 00:06:35.339 --> 00:06:38.639 Or is it just a masked paywall that ruins your 00:06:38.639 --> 00:06:41.000 output quality? I think it actually solves it. 00:06:41.040 --> 00:06:44.000 Because seamless fallback to smaller models keeps 00:06:44.000 --> 00:06:46.160 workflows alive rather than hitting a sudden 00:06:46.160 --> 00:06:48.639 paywall. That's true. And the smaller models 00:06:48.639 --> 00:06:50.519 are smart enough for basic routing now anyway. 00:06:50.800 --> 00:06:53.019 A softer landing instead of a hard stop on your 00:06:53.019 --> 00:06:55.279 workflow. Precisely. You get to keep your momentum 00:06:55.279 --> 00:06:58.259 going. So while Google focuses on consumer packaging, 00:06:58.500 --> 00:07:01.139 the underlying engines powering those tools are 00:07:01.139 --> 00:07:04.620 fighting a brutal, invisible war. Oh, the talent 00:07:04.620 --> 00:07:07.079 and compute arms race behind the scenes is incredibly 00:07:07.079 --> 00:07:09.600 cutthroat. Let's look at the bleeding edge. The 00:07:09.600 --> 00:07:12.959 talent shift is a massive story right now. Andres 00:07:12.959 --> 00:07:15.579 Karpathy, who co -founded OpenAI, just joined 00:07:15.579 --> 00:07:18.220 Anthropic. Yeah, Karpathy joining Anthropic is 00:07:18.220 --> 00:07:20.560 making major headlines everywhere. Here's where 00:07:20.560 --> 00:07:24.769 it gets really interesting, though. $2 .4 trillion 00:07:24.769 --> 00:07:29.149 ecosystem. Does one researcher changing jerseys 00:07:29.149 --> 00:07:32.529 actually shift the balance of power? Or is this 00:07:32.529 --> 00:07:34.930 just Silicon Valley theater? It is definitely 00:07:34.930 --> 00:07:38.110 not just theater. Bringing his specific pre -training 00:07:38.110 --> 00:07:40.769 expertise to the company, already at number one, 00:07:40.910 --> 00:07:43.610 essentially cements their lead. Right. He is 00:07:43.610 --> 00:07:46.329 joining their pre -training team. That is a crucial 00:07:46.329 --> 00:07:48.790 detail. Yeah, because pre -training is the foundational 00:07:48.790 --> 00:07:51.970 phase. It's the ingestion of raw data. That phase 00:07:51.970 --> 00:07:55.189 builds the core worldview of the model. It really 00:07:55.189 --> 00:07:57.730 decides who wins the entire frontier race. Right, 00:07:57.810 --> 00:08:00.189 and getting the guy who literally wrote the playbook 00:08:00.189 --> 00:08:02.470 on early model training is a massive strategic 00:08:02.470 --> 00:08:05.089 advantage. The brain behind the models just changed 00:08:05.089 --> 00:08:07.470 jerseys to the winning team. Exactly, and the 00:08:07.470 --> 00:08:09.550 results of this talent war are already showing 00:08:09.550 --> 00:08:12.250 up in raw speed. A company called Descartes just 00:08:12.250 --> 00:08:15.149 raised $300 million. Yeah, they are backed by 00:08:15.149 --> 00:08:17.990 Amazon, NVIDIA, and Sequoia. They are building 00:08:17.990 --> 00:08:20.550 faster systems with their DOS 2 .0 platform. 00:08:21.079 --> 00:08:23.360 The numbers on this are just staggering. They 00:08:23.360 --> 00:08:26.379 are reportedly delivering over 1 ,600 tokens 00:08:26.379 --> 00:08:29.360 per second. And just to clarify, tokens per second 00:08:29.360 --> 00:08:32.299 measures how fast an AI reads and writes text 00:08:32.299 --> 00:08:35.960 words. Whoa. I mean, imagine 1 ,600 tokens per 00:08:35.960 --> 00:08:39.860 second. That is real -time instantaneous thought 00:08:39.860 --> 00:08:42.740 generation. It's incredible. It outpaces human 00:08:42.740 --> 00:08:45.019 reading speed by orders of magnitude. It means 00:08:45.019 --> 00:08:47.399 the bottleneck is no longer the machine's thinking 00:08:47.399 --> 00:08:50.279 time. And the speed is translating to other modalities, 00:08:50.340 --> 00:08:52.779 too. Microsoft just dropped a local AI model 00:08:52.779 --> 00:08:55.779 that generates detailed 3D objects from a single 00:08:55.779 --> 00:08:58.179 photo. Yeah, and it takes just three seconds. 00:08:58.320 --> 00:09:00.740 Three seconds for a 3D asset. That used to take 00:09:00.740 --> 00:09:03.059 a human artist days of modeling and texturing. 00:09:03.500 --> 00:09:05.440 and it is fully open source and commercially 00:09:05.440 --> 00:09:07.919 usable. Cursor is pushing the boundaries on long 00:09:07.919 --> 00:09:10.980 -term tasks too. They launched Composer 2 .5. 00:09:11.039 --> 00:09:13.360 It's their strongest coding model yet. This is 00:09:13.360 --> 00:09:16.000 huge for agentic tasks. If you've ever tried 00:09:16.000 --> 00:09:19.440 to build an app and watch the AI completely forget 00:09:19.440 --> 00:09:22.559 your architecture by the third prompt, you know 00:09:22.559 --> 00:09:25.139 exactly how frustrating prompt drift is. Oh, 00:09:25.279 --> 00:09:27.759 it's the worst. But Composer 2 .5 is designed 00:09:27.759 --> 00:09:31.080 specifically to fix that memory loss. They also 00:09:31.080 --> 00:09:33.940 teased a massive new model. Cursor is partnering 00:09:33.940 --> 00:09:37.279 with Elon's SpaceX AI. Yeah, they are reportedly 00:09:37.279 --> 00:09:40.470 using 10x more compute for that. the sheer scale 00:09:40.470 --> 00:09:43.590 of compute being deployed right now is well it's 00:09:43.590 --> 00:09:46.210 hard to wrap your head around entire power grids 00:09:46.210 --> 00:09:48.009 are being dedicated to training these systems 00:09:48.009 --> 00:09:50.809 and yet With billions of dollars in compute power, 00:09:51.070 --> 00:09:53.889 we still get tangled up in the most human things. 00:09:54.110 --> 00:09:56.570 Like, look at the recent Nike drama. Oh, man, 00:09:56.669 --> 00:10:00.330 it is fascinating. A single em dash in Nike's 00:10:00.330 --> 00:10:03.129 latest social media post restarted a massive 00:10:03.129 --> 00:10:05.629 debate online. Yeah, people are literally running 00:10:05.629 --> 00:10:08.549 AI writing detectors on a shoe ad. Some people 00:10:08.549 --> 00:10:11.250 are calling it blatant chat GPT output. Others 00:10:11.250 --> 00:10:13.970 argue it's just completely normal human copywriting. 00:10:14.330 --> 00:10:16.529 It's wild that we are arguing over a punctuation 00:10:16.529 --> 00:10:18.970 mark. It just shows we are hypersensitive to 00:10:18.970 --> 00:10:21.110 the boundary between human and machine text right 00:10:21.110 --> 00:10:22.870 now. We are desperately looking for fingerprints. 00:10:23.090 --> 00:10:25.149 We want to know who is really speaking to us. 00:10:25.269 --> 00:10:27.570 We see this massive compute power fighting at 00:10:27.570 --> 00:10:29.909 the top. We see billions of dollars in funding. 00:10:30.289 --> 00:10:32.710 But the most practical shift for you listening 00:10:32.710 --> 00:10:35.610 right now is how that massive power is trickling 00:10:35.610 --> 00:10:38.500 down. Yeah, it is finally hitting hyper specific 00:10:38.500 --> 00:10:41.460 local tools right on your own machine. This is 00:10:41.460 --> 00:10:43.820 where it gets incredibly practical. Anthropic 00:10:43.820 --> 00:10:46.659 just introduced new features in cloud managed 00:10:46.659 --> 00:10:49.820 agents. They now offer self -hosted sandboxes 00:10:49.820 --> 00:10:53.200 and MCP tunnels. Let's break that down. MCP tunnels 00:10:53.200 --> 00:10:56.539 are basically secure paths connecting AI directly 00:10:56.539 --> 00:10:59.120 to private company data. Right. Running cloud 00:10:59.120 --> 00:11:02.759 managed agents locally via MCP tunnels is essentially 00:11:02.759 --> 00:11:05.919 like. bringing a master chef into your own log 00:11:05.919 --> 00:11:08.059 kitchen. They get to use all your secret family 00:11:08.059 --> 00:11:10.480 recipes, but because of the tunnel, they cannot 00:11:10.480 --> 00:11:12.679 leave the house and share that recipe with Anthropic. 00:11:12.860 --> 00:11:14.879 That's a great way to put it. It completely solves 00:11:14.879 --> 00:11:17.259 the enterprise privacy problem. You get the reasoning 00:11:17.259 --> 00:11:19.639 power of a frontier model without handing over 00:11:19.639 --> 00:11:21.860 the keys to your internal databases. We're also 00:11:21.860 --> 00:11:24.480 seeing incredible workflow hacks happening locally. 00:11:25.120 --> 00:11:27.299 Developers are combining Claude code with Gemma 00:11:27.299 --> 00:11:30.389 4. That is a brilliant hybrid approach. Anthropic 00:11:30.389 --> 00:11:32.710 actually shared best practices for using clod 00:11:32.710 --> 00:11:36.169 code in large code bases. By routing the smaller 00:11:36.169 --> 00:11:38.610 repetitive coding tasks to a local model like 00:11:38.610 --> 00:11:41.950 Gemma 4, you just stop wasting expensive cloud 00:11:41.950 --> 00:11:44.070 credits. Exactly. You only ping clod for the 00:11:44.070 --> 00:11:47.460 heavy lifting. There is also a new laptop LLM 00:11:47.460 --> 00:11:49.360 ranking tool that just launched. Oh, this is 00:11:49.360 --> 00:11:52.120 a lifesaver. If you have ever spent hours downloading 00:11:52.120 --> 00:11:54.779 a local model only to realize your specific GPU 00:11:54.779 --> 00:11:57.159 cannot even run it, this changes everything. 00:11:58.009 --> 00:12:00.509 It analyzes your exact hardware and ranks the 00:12:00.509 --> 00:12:02.750 best models your laptop can truly handle. It 00:12:02.750 --> 00:12:05.309 is all about local empowerment. And we are seeing 00:12:05.309 --> 00:12:07.769 everyday tools get agentic features too, like 00:12:07.769 --> 00:12:10.509 look at Amazon's Alexa Plus. Alexa Plus is acting 00:12:10.509 --> 00:12:13.169 as a fully autonomous AI podcast creator now. 00:12:13.309 --> 00:12:15.950 You just pick a topic. Two AI hosts generate 00:12:15.950 --> 00:12:18.269 a custom, human -sounding audio conversation 00:12:18.269 --> 00:12:21.649 on demand. It is very similar to Notebook LM's 00:12:21.649 --> 00:12:24.129 viral audio feature, but integrated directly 00:12:24.129 --> 00:12:26.779 into your smart speaker. We also have PolyReach. 00:12:26.799 --> 00:12:29.259 It gives your AI agent a real phone number and 00:12:29.259 --> 00:12:31.559 a synthetic voice. It can literally answer your 00:12:31.559 --> 00:12:34.399 incoming phone calls 24 -7. And it speaks over 00:12:34.399 --> 00:12:36.879 50 languages. Think about the implications of 00:12:36.879 --> 00:12:39.220 that for small businesses. You have a tireless 00:12:39.220 --> 00:12:41.639 receptionist that never sleeps and speaks every 00:12:41.639 --> 00:12:43.700 language your customers do. We are seeing it 00:12:43.700 --> 00:12:46.950 in software development, too. Drizzt is reshaping 00:12:46.950 --> 00:12:50.250 how we test apps. It's an AI mobile test platform 00:12:50.250 --> 00:12:52.870 built around intent -based testing. Yeah, it 00:12:52.870 --> 00:12:55.389 adapts to dynamic user interfaces automatically. 00:12:55.889 --> 00:12:58.710 Plus, tools like MampleChat are collapsing the 00:12:58.710 --> 00:13:00.710 software stack. They are combining messaging, 00:13:01.029 --> 00:13:04.629 AI models, and autonomous agents all into one 00:13:04.629 --> 00:13:07.450 shared workspace. This raises an important question 00:13:07.450 --> 00:13:10.029 for the future of the industry, though. We have 00:13:10.029 --> 00:13:12.289 all this local power now on our own machines. 00:13:12.610 --> 00:13:16.610 Does the rise of highly capable local LLMs eventually 00:13:16.610 --> 00:13:19.289 replace these massive cloud subscriptions entirely? 00:13:19.809 --> 00:13:22.269 I think they will likely work in tandem. Cloud 00:13:22.269 --> 00:13:24.389 for the heavy lifting, local for privacy and 00:13:24.389 --> 00:13:26.970 speed. You will always need the massive server 00:13:26.970 --> 00:13:29.070 farms for the hardest problems, but everyday 00:13:29.070 --> 00:13:32.070 tasks move to the edge. Cloud for the heavy lifting, 00:13:32.289 --> 00:13:35.070 local for privacy and speed? Exactly. It's a 00:13:35.070 --> 00:13:37.389 hybrid future. The ecosystem is finding its balance. 00:13:37.799 --> 00:13:40.080 We have covered a staggering amount of ground 00:13:40.080 --> 00:13:42.740 today. Let's tie all these threads together. 00:13:43.000 --> 00:13:45.000 Yeah, if you step back and look at the macro 00:13:45.000 --> 00:13:47.779 picture, the clear through line is that AI has 00:13:47.779 --> 00:13:50.659 definitively shifted. It is no longer a novelty 00:13:50.659 --> 00:13:53.159 chatbot where you type a prompt and get a funny 00:13:53.159 --> 00:13:57.080 poem. It has become a continuous background operating 00:13:57.080 --> 00:14:00.639 system. We see that massive shift in the $2 .4 00:14:00.639 --> 00:14:03.419 trillion market valuations of the Disruptor 50. 00:14:03.639 --> 00:14:06.059 Capital is betting heavily on physical automation 00:14:06.059 --> 00:14:09.259 and vibe coding. we see it in google's packaging 00:14:09.259 --> 00:14:13.379 too the 24 7 gemini spark agent is designed to 00:14:13.379 --> 00:14:16.059 manage our digital lives seamlessly it operates 00:14:16.059 --> 00:14:18.799 constantly in the background it summarizes drafts 00:14:18.799 --> 00:14:20.919 and connects the docs before you even ask and 00:14:20.919 --> 00:14:23.120 we see it in the private local environments running 00:14:23.120 --> 00:14:25.960 right on our own laptops tools like claude managed 00:14:25.960 --> 00:14:29.039 agents and local combinations like gemma 4 give 00:14:29.039 --> 00:14:31.899 us secure persistent intelligence right in our 00:14:31.899 --> 00:14:34.220 own kitchens it is the normalization of agentic 00:14:34.220 --> 00:14:36.830 behavior The friction of interacting with computers 00:14:36.830 --> 00:14:39.289 is just disappearing entirely. So what does this 00:14:39.289 --> 00:14:41.769 all mean for you? It means we are moving from 00:14:41.769 --> 00:14:44.470 prompting machines to actively collaborating 00:14:44.470 --> 00:14:47.889 with them. The underlying infrastructure is finally 00:14:47.889 --> 00:14:50.889 mature enough to support actual autonomy. I want 00:14:50.889 --> 00:14:53.169 to leave you with a final thought today. We're 00:14:53.169 --> 00:14:55.850 building incredibly powerful autonomous systems. 00:14:56.409 --> 00:14:59.409 Alexa Plus can now generate fully custom human 00:14:59.409 --> 00:15:02.490 sounding podcasts on demand. Yeah. Agents like 00:15:02.490 --> 00:15:04.970 Gemini Spark and Polyreach can seamlessly manage 00:15:04.970 --> 00:15:08.470 your digital life, triage your emails and literally 00:15:08.470 --> 00:15:12.269 answer your phone calls. 2047 in 50 languages. 00:15:12.509 --> 00:15:14.330 It's wild to think about. So if the machines 00:15:14.330 --> 00:15:16.470 are perfectly capable of just talking to each 00:15:16.470 --> 00:15:18.490 other, researching for each other and summarizing 00:15:18.490 --> 00:15:20.549 for each other, what happens to our own internal 00:15:20.549 --> 00:15:23.220 voice? What happens? to our fundamental human 00:15:23.220 --> 00:15:25.639 desire to learn when the friction of ignorance 00:15:25.639 --> 00:15:28.039 is completely removed. It is the ultimate question 00:15:28.039 --> 00:15:31.539 of the 2026 landscape. We have successfully outsourced 00:15:31.539 --> 00:15:33.980 the friction of daily work, but we cannot afford 00:15:33.980 --> 00:15:36.720 to outsource our own curiosity. Keep asking questions. 00:15:36.899 --> 00:15:39.000 Keep exploring these boundaries on your own. 00:15:39.120 --> 00:15:41.899 The tools are undeniably powerful, but your human 00:15:41.899 --> 00:15:43.779 perspective is what actually gives them purpose. 00:15:43.960 --> 00:15:45.620 Thanks for listening to this deep dive.