WEBVTT 00:00:00.000 --> 00:00:02.439 Okay, so let's try to unpack this. There's this 00:00:02.439 --> 00:00:06.860 strange contradiction happening in AI right now. 00:00:06.980 --> 00:00:09.560 A really strange one. On one hand, you've got 00:00:09.560 --> 00:00:14.359 the biggest rivals. We're talking open AI, anthropic, 00:00:14.380 --> 00:00:17.579 block all making peace to standardize their future. 00:00:17.699 --> 00:00:20.449 Right. building the serious infrastructure. And 00:00:20.449 --> 00:00:22.989 then at the same time, you have Sam Altman, their 00:00:22.989 --> 00:00:26.089 leader on late night TV. And he's admitting he 00:00:26.089 --> 00:00:29.309 uses his, you know, genius level AI to figure 00:00:29.309 --> 00:00:31.829 out what his baby's poop means. It's just it's 00:00:31.829 --> 00:00:33.869 high stakes standardization meeting this deeply 00:00:33.869 --> 00:00:37.049 mundane, very human need. Exactly. Welcome back 00:00:37.049 --> 00:00:40.030 to the deep dive. Our mission, as always, is 00:00:40.030 --> 00:00:42.049 to take that whole complex stack of information 00:00:42.049 --> 00:00:44.770 and find the clear currents, you know, signals 00:00:44.770 --> 00:00:46.829 guiding the industry. And today those currents 00:00:46.829 --> 00:00:48.810 are pulling in two totally opposite directions. 00:00:49.710 --> 00:00:51.789 Consolidation on one side. And humanization on 00:00:51.789 --> 00:00:53.729 the other. We're looking at a field that's maturing 00:00:53.729 --> 00:00:57.109 incredibly fast, but maybe not always so gracefully. 00:00:57.289 --> 00:00:59.810 So our roadmap. We're going to start with that 00:00:59.810 --> 00:01:01.909 Ben's complex move to standardize what they're 00:01:01.909 --> 00:01:04.510 calling agentic AI. These are the systems designed 00:01:04.510 --> 00:01:06.730 to connect and actually act on their own. Right. 00:01:07.079 --> 00:01:10.000 Then we'll pivot to how you can navigate all 00:01:10.000 --> 00:01:12.260 this change. We'll talk about the real professional 00:01:12.260 --> 00:01:14.900 value of getting quick validation, like with 00:01:14.900 --> 00:01:17.140 the new Google Gemini certificate. And finally, 00:01:17.219 --> 00:01:19.420 we're going to get into the really clever PR 00:01:19.420 --> 00:01:23.200 move behind giving AGI a more human face with 00:01:23.200 --> 00:01:26.879 Sam Altman's parenting stories as like the key 00:01:26.879 --> 00:01:30.420 exhibit. OK, before we jump in, let's just nail 00:01:30.420 --> 00:01:33.000 down that piece of jargon. When we say agentic 00:01:33.000 --> 00:01:36.400 AI, we're just describing. Models that use protocols 00:01:36.400 --> 00:01:38.799 to connect with and use other software and data 00:01:38.799 --> 00:01:41.180 tools. They're the AIs that do things, not just 00:01:41.180 --> 00:01:44.280 talk. Exactly. They execute tasks. So segment 00:01:44.280 --> 00:01:47.439 one, the standardization wars. This is a big 00:01:47.439 --> 00:01:49.519 one. It's all centered on this new agendic AI 00:01:49.519 --> 00:01:52.519 foundation, the AAIF. And it's under the Linux 00:01:52.519 --> 00:01:55.609 foundation. And the members, OpenAI, Anthropic, 00:01:55.829 --> 00:01:58.370 Block. I mean, that's a surprising trio given 00:01:58.370 --> 00:02:00.469 how much they compete. It really is. To me, this 00:02:00.469 --> 00:02:02.609 is just them admitting reality. They know they 00:02:02.609 --> 00:02:05.310 can't scale the business of autonomous AI if 00:02:05.310 --> 00:02:07.450 everyone's building on, you know, a different 00:02:07.450 --> 00:02:10.349 shaky foundation. Fragmentation is a liability 00:02:10.349 --> 00:02:12.650 now. A huge one. And they didn't just join. They 00:02:12.650 --> 00:02:15.090 immediately donated these foundational tools 00:02:15.090 --> 00:02:17.689 to get it all started. So what are they contributing? 00:02:17.930 --> 00:02:20.550 What are these core technical pieces that are 00:02:20.550 --> 00:02:23.250 going to form the new standard? OK, so Anthropic 00:02:23.250 --> 00:02:25.569 contributed something called the Model Context 00:02:25.569 --> 00:02:30.129 Protocol or MCP. Right. This is so crucial. It's 00:02:30.129 --> 00:02:32.569 basically the universal language standard. It 00:02:32.569 --> 00:02:35.909 lets AI models negotiate context and access all 00:02:35.909 --> 00:02:38.669 these external tools and data sources. It handles 00:02:38.669 --> 00:02:41.250 the boring stuff like state management and authentication 00:02:41.250 --> 00:02:45.210 in a shared way. So it's it's not just a translator, 00:02:45.310 --> 00:02:47.370 is it? It's more like a shared set of rules for 00:02:47.370 --> 00:02:49.370 trust and access between different models and 00:02:49.370 --> 00:02:51.860 other software. Precisely. It just gets rid of 00:02:51.860 --> 00:02:53.599 so much developer perfection. Then you've got 00:02:53.599 --> 00:02:56.280 Block, Jack Dorsey's company. They open source 00:02:56.280 --> 00:02:58.780 something called Goose. And you can think of 00:02:58.780 --> 00:03:01.659 this as the essential starter pack for any developer 00:03:01.659 --> 00:03:04.400 building agents. It just cuts out months of basic 00:03:04.400 --> 00:03:06.740 setup work. And OpenAI's contribution, agents 00:03:06.740 --> 00:03:10.680 .ind. It's the simplest, but maybe the most strategically 00:03:10.680 --> 00:03:14.080 clever part of this. Oh, it's very smart. It's 00:03:14.080 --> 00:03:16.199 just a plain text file that lives inside your 00:03:16.199 --> 00:03:19.659 code repository. And its only job is to clearly 00:03:19.659 --> 00:03:23.259 define how any AI coding tool is allowed to behave 00:03:23.259 --> 00:03:26.460 inside that specific project. It's a way to manage 00:03:26.460 --> 00:03:29.719 trust by, like, sandboxing the model's actions. 00:03:30.139 --> 00:03:33.469 The thing is, the significance here... goes way 00:03:33.469 --> 00:03:35.729 beyond just these three companies because the 00:03:35.729 --> 00:03:37.330 Linux Foundation is driving it. And they have 00:03:37.330 --> 00:03:40.349 a history here. A huge history. They turn messy, 00:03:40.550 --> 00:03:44.069 fragmented ideas into core global infrastructure. 00:03:44.490 --> 00:03:46.689 Well, think about Kubernetes. It started as this 00:03:46.689 --> 00:03:49.930 super complex, messy thing for container orchestration. 00:03:50.169 --> 00:03:52.550 The Linux Foundation gave it that neutral home, 00:03:52.710 --> 00:03:54.849 and now Kubernetes basically runs most of the 00:03:54.849 --> 00:03:58.509 world's modern cloud, this AAF thing. It feels 00:03:58.509 --> 00:04:02.409 like the Kubernetes moment for agentic AI. That 00:04:02.409 --> 00:04:05.009 makes the stakes very clear. But let's push on 00:04:05.009 --> 00:04:06.550 the competitive side of this for a second. Okay. 00:04:06.629 --> 00:04:08.930 It's great for developers, great for interoperability. 00:04:09.110 --> 00:04:12.030 But isn't this also a strategic move for these 00:04:12.030 --> 00:04:14.490 three big players to kind of control the infrastructure? 00:04:14.789 --> 00:04:17.459 Well, that's always the risk, right? potentially 00:04:17.459 --> 00:04:20.180 lock out smaller competitors who weren't at the 00:04:20.180 --> 00:04:21.839 table when the rules were written. That's the 00:04:21.839 --> 00:04:23.839 inherent risk of standardizing early. You're 00:04:23.839 --> 00:04:25.699 right. You build the roads before everyone else 00:04:25.699 --> 00:04:28.480 has a car. But the immediate driver here really 00:04:28.480 --> 00:04:31.019 is business growth. We know that the AI coding 00:04:31.019 --> 00:04:34.079 market just exploded this year. We saw the figures. 00:04:34.259 --> 00:04:37.639 Yeah. AI coding spend went from, what, $550 million 00:04:37.639 --> 00:04:42.189 to $4 billion in one year. That is just explosive 00:04:42.189 --> 00:04:44.930 growth. And when you have that much money flying 00:04:44.930 --> 00:04:48.050 around, fragmentation, you know, where a tool 00:04:48.050 --> 00:04:50.709 only works with cloud or only in VS Code or only 00:04:50.709 --> 00:04:52.810 in the cloud, that becomes a massive barrier. 00:04:53.029 --> 00:04:55.170 You need standardization because that $4 billion 00:04:55.170 --> 00:04:57.649 is only going to grow if the tools are reliable 00:04:57.649 --> 00:04:59.689 and they work across all the different software 00:04:59.689 --> 00:05:02.769 stacks. So if we strip away the money in the 00:05:02.769 --> 00:05:06.250 competition for a moment, what's the one major 00:05:06.250 --> 00:05:09.839 underlying... technical pain point that this 00:05:09.839 --> 00:05:12.319 whole open source effort is trying to fix. It 00:05:12.319 --> 00:05:14.560 fixes the challenge of different AI tools and 00:05:14.560 --> 00:05:17.279 models needing a reliable shared language to 00:05:17.279 --> 00:05:20.079 connect securely across any software stack. That's 00:05:20.079 --> 00:05:23.800 the core of it. Okay, so from the complex job 00:05:23.800 --> 00:05:27.279 of standardizing how we build AI. We pivot. We 00:05:27.279 --> 00:05:29.980 pivot to standardizing how we learn about it 00:05:29.980 --> 00:05:32.180 with this new credential from Google. I love 00:05:32.180 --> 00:05:34.459 this story because it's so actionable for anyone 00:05:34.459 --> 00:05:38.000 listening. Google AI Education just launched 00:05:38.000 --> 00:05:40.660 the Gemini Educator Certificate. And the key 00:05:40.660 --> 00:05:42.860 detail. The key detail is that it's completely 00:05:42.860 --> 00:05:47.180 free until December 31st, 2025. Wow. After that, 00:05:47.279 --> 00:05:50.920 it's $25. That's a really generous window, a 00:05:50.920 --> 00:05:52.980 low -friction way for people to validate their 00:05:52.980 --> 00:05:55.339 skills. So what's the actual value proposition 00:05:55.339 --> 00:05:58.009 here for someone learning? I think it's got three 00:05:58.009 --> 00:06:00.310 tiers of benefit. First, just the practical utility. 00:06:00.470 --> 00:06:02.569 You actually gain mastery of the Gemini platform 00:06:02.569 --> 00:06:04.310 and that's becoming a really essential tool. 00:06:04.490 --> 00:06:07.110 Second, the profile boost. It's a Google credential 00:06:07.110 --> 00:06:09.410 that just carries some weight on a resume, you 00:06:09.410 --> 00:06:11.410 know. And maybe the most important part, in a 00:06:11.410 --> 00:06:13.769 really crowded job market, it signals something 00:06:13.769 --> 00:06:18.180 critical to employers. Yes. This is it. It signals 00:06:18.180 --> 00:06:21.360 proactive learning. If you have a pool of candidates 00:06:21.360 --> 00:06:24.000 who are all equally skilled, the one who took 00:06:24.000 --> 00:06:26.459 the initiative to learn and validate their knowledge 00:06:26.459 --> 00:06:29.100 on these new tools, they just gave themselves 00:06:29.100 --> 00:06:32.379 an edge. It shows curiosity. You know, I'll admit 00:06:32.379 --> 00:06:35.360 I still wrestle with prompt drift myself sometimes, 00:06:35.519 --> 00:06:37.720 finding the best way to talk to these things 00:06:37.720 --> 00:06:40.160 as they evolve. We all do. So I genuinely appreciate 00:06:40.160 --> 00:06:44.240 these low stakes ways to learn a new platform 00:06:44.240 --> 00:06:47.259 without. you know, the pressure of a huge tuition 00:06:47.259 --> 00:06:49.560 bill. The only barrier to entry here is time 00:06:49.560 --> 00:06:52.040 and curiosity. That's a great democratizing move. 00:06:52.259 --> 00:06:54.939 So for anyone interested, what are the key logistical 00:06:54.939 --> 00:06:56.920 things they need to know about taking it? Okay, 00:06:56.980 --> 00:06:59.600 watch the clock. You get 120 minutes for the 00:06:59.600 --> 00:07:01.660 exam, but people who've done well say it can 00:07:01.660 --> 00:07:04.389 be done in about 30. But the big warning. The 00:07:04.389 --> 00:07:07.310 retake policy. The retake policy. If you fail 00:07:07.310 --> 00:07:09.430 that first time, you have to wait eight days 00:07:09.430 --> 00:07:11.750 before you can try again. Given that restriction 00:07:11.750 --> 00:07:13.750 and the professional benefit we're talking about, 00:07:13.889 --> 00:07:16.430 should people be prioritizing a really high score? 00:07:16.649 --> 00:07:19.110 Or is it more about just showing you're curious 00:07:19.110 --> 00:07:22.470 and proactive by getting it done? Focus on signaling 00:07:22.470 --> 00:07:24.649 curiosity and showing you're actively engaging 00:07:24.649 --> 00:07:29.079 with new and evolving AI tools. We're back. We've 00:07:29.079 --> 00:07:31.779 covered the technical standardization that requires 00:07:31.779 --> 00:07:34.379 engineering collaboration. Now we're shifting 00:07:34.379 --> 00:07:37.620 to this powerful current of AI humanization that 00:07:37.620 --> 00:07:40.259 requires public relatability. And this is where 00:07:40.259 --> 00:07:43.240 Sam Altman, the guy building AGI, makes his debut 00:07:43.240 --> 00:07:45.560 on Late Night with Jimmy Fallon. And the topic 00:07:45.560 --> 00:07:48.459 wasn't, you know, the next LLM benchmark or the 00:07:48.459 --> 00:07:50.319 future of superintelligence. No, it was diapers 00:07:50.319 --> 00:07:54.800 and pizza. He framed ChatGPT as his parenting 00:07:54.800 --> 00:07:57.560 sidekick for his newborn. So the man leading 00:07:57.560 --> 00:07:59.579 one of the highest stakes tech projects in human 00:07:59.579 --> 00:08:02.459 history is using it for these deeply mundane, 00:08:02.680 --> 00:08:05.000 stressful parenting moments. The stories he told 00:08:05.000 --> 00:08:07.279 were perfect PR. He talked about asking why his 00:08:07.279 --> 00:08:09.439 baby was laughing while throwing pizza on the 00:08:09.439 --> 00:08:11.620 floor. I mean, that is just a genuinely human, 00:08:11.720 --> 00:08:14.120 chaotic moment. Or the other one, the universal 00:08:14.120 --> 00:08:17.040 but kind of gross question. Asking his genius 00:08:17.040 --> 00:08:20.019 -level AI about the color of baby poop. Right. 00:08:20.199 --> 00:08:23.300 He even admitted he feels kind of bad asking 00:08:23.300 --> 00:08:25.980 this super advanced system such dumb questions. 00:08:26.180 --> 00:08:28.879 And that little admission of vulnerability, that's 00:08:28.879 --> 00:08:30.959 the whole strategy. That's the core of it. The 00:08:30.959 --> 00:08:34.419 punchline is... It works. It's deeply relatable. 00:08:34.659 --> 00:08:37.120 It's incredibly effective, especially when you 00:08:37.120 --> 00:08:39.620 remember this is a company that has spent years 00:08:39.620 --> 00:08:43.059 in these serious, sometimes fearful public debates 00:08:43.059 --> 00:08:46.700 about existential risk and model races with Google. 00:08:47.049 --> 00:08:49.610 And now the CEO is talking about nursery stress. 00:08:49.690 --> 00:08:53.789 This is absolutely a clever, soft PR move. The 00:08:53.789 --> 00:08:56.190 goal is to shift the public's focus away from 00:08:56.190 --> 00:08:58.830 that scary theoretical future and bring it right 00:08:58.830 --> 00:09:01.710 back to immediate, friendly, practical use. They're 00:09:01.710 --> 00:09:04.610 working to make ChatGPT feel like a non -intimidating 00:09:04.610 --> 00:09:06.970 tool again. And the choice of Fallon's show for 00:09:06.970 --> 00:09:09.970 distribution is just genius. It completely bypasses 00:09:09.970 --> 00:09:11.809 the tech press and goes straight to millions 00:09:11.809 --> 00:09:14.029 of everyday people. The tech community doesn't 00:09:14.029 --> 00:09:16.610 need to be convinced. No, but millions of late 00:09:16.610 --> 00:09:19.490 night viewers just met this technology in the 00:09:19.490 --> 00:09:22.529 lowest stakes way possible. It normalizes it 00:09:22.529 --> 00:09:25.009 through humor. It suggests, hey, the CEO uses 00:09:25.009 --> 00:09:26.889 it to check on his kid. It can't be that scary. 00:09:27.129 --> 00:09:31.179 So does this PR shift suggest that. Public fear, 00:09:31.259 --> 00:09:34.080 or maybe just intimidation about AI, was a bigger 00:09:34.080 --> 00:09:36.080 challenge for the company than the actual technological 00:09:36.080 --> 00:09:39.080 capability. Yes, it suggests establishing broad 00:09:39.080 --> 00:09:42.460 public relatability is now a primary goal, maybe 00:09:42.460 --> 00:09:45.340 even equal to scaling the tech itself. Okay, 00:09:45.440 --> 00:09:48.240 let's wrap our deep dive with a rapid fire segment. 00:09:48.360 --> 00:09:51.320 Some interesting industry insights, tools, and 00:09:51.320 --> 00:09:54.440 predictions. Quick hits. Let's start with a really 00:09:54.440 --> 00:09:57.240 powerful new tool for anyone doing serious research. 00:09:57.320 --> 00:09:59.879 It's a Chrome extension called the Slopivator. 00:10:00.240 --> 00:10:02.399 The name alone just perfectly captures the need. 00:10:02.539 --> 00:10:04.539 It really does. It filters out AI -generated 00:10:04.539 --> 00:10:07.139 content and only shows you search results that 00:10:07.139 --> 00:10:09.440 were published before ChatGPT launched. So it's 00:10:09.440 --> 00:10:12.159 a dedicated tool for finding pure research before 00:10:12.159 --> 00:10:14.740 the floodgates of synthetic text open. Which 00:10:14.740 --> 00:10:17.259 says a lot about data integrity now, right? It's 00:10:17.259 --> 00:10:19.480 fascinating. We need a tool to deliberately look 00:10:19.480 --> 00:10:22.059 backwards just to trust what we're reading. Okay, 00:10:22.100 --> 00:10:25.110 what's next? On the topic of getting better outputs, 00:10:25.470 --> 00:10:28.190 an OpenAI co -founder gave some great advice. 00:10:28.590 --> 00:10:31.870 A simple instruction. Stop using the phrase, 00:10:32.090 --> 00:10:35.070 what do you think? Exactly. Just eliminate that 00:10:35.070 --> 00:10:37.490 request for an AI's opinion. And you will immediately 00:10:37.490 --> 00:10:41.129 get higher quality, less biased, more objective 00:10:41.129 --> 00:10:43.830 outputs. That's just actionable advice for anyone 00:10:43.830 --> 00:10:46.309 using these tools professionally. And in hardware, 00:10:46.549 --> 00:10:48.190 a lot of money is flowing toward efficiency. 00:10:48.899 --> 00:10:52.840 Unconventional AI just raised $475 million. And 00:10:52.840 --> 00:10:56.019 that huge funding round is specifically for creating 00:10:56.019 --> 00:10:58.759 brain -inspired chips. Chips designed to be more 00:10:58.759 --> 00:11:01.320 energy efficient, more eco -friendly. Which proves 00:11:01.320 --> 00:11:03.360 efficiency is now a multi -million dollar concern, 00:11:03.480 --> 00:11:05.759 not just a philosophical one. Right. But here's 00:11:05.759 --> 00:11:07.879 the prediction that really stopped me. Time's 00:11:07.879 --> 00:11:11.440 2025 person of the year, the betting favorite. 00:11:11.720 --> 00:11:15.419 AI itself, at 40%. The head of actual people 00:11:15.419 --> 00:11:18.710 like Jensen Huang and Sam Altman. Whoa. Just 00:11:18.710 --> 00:11:21.629 stop for a second. Imagine scaling the very concept 00:11:21.629 --> 00:11:24.710 of personhood to an algorithm. That reflects 00:11:24.710 --> 00:11:27.610 how pervasive this tech has become in our collective 00:11:27.610 --> 00:11:29.950 imagination. It's a genuine moment of wonder 00:11:29.950 --> 00:11:32.309 at the speed of this change. So with all these 00:11:32.309 --> 00:11:36.450 quick hits from filtering data to venture deals 00:11:36.450 --> 00:11:39.029 for the average professional, what's the single 00:11:39.029 --> 00:11:42.559 most practical insight here? The actionable strategy 00:11:42.559 --> 00:11:44.919 to refine your prompts by eliminating vague, 00:11:45.139 --> 00:11:48.379 opinion -seeking questions. Precise input guarantees 00:11:48.379 --> 00:11:51.120 better output. We covered a huge amount of ground 00:11:51.120 --> 00:11:53.840 today. We really did. From rival companies agreeing 00:11:53.840 --> 00:11:57.320 on foundational rules to a CEO talking about 00:11:57.320 --> 00:11:59.539 his personal life. And to recap the big idea. 00:12:00.029 --> 00:12:02.649 The world of AI is simultaneously consolidating 00:12:02.649 --> 00:12:05.330 its infrastructure. That AAIF alliance proves 00:12:05.330 --> 00:12:07.690 it. And it's dissolving its intimidating public 00:12:07.690 --> 00:12:10.830 image through this very calculated PR. The core 00:12:10.830 --> 00:12:13.149 takeaway here is that the tools for both building 00:12:13.149 --> 00:12:16.070 AI and for using AI are becoming so much more 00:12:16.070 --> 00:12:18.470 accessible. But that accessibility only helps 00:12:18.470 --> 00:12:20.750 you if you stay curious and proactive. Which 00:12:20.750 --> 00:12:23.029 brings us right back to that free Gemini certificate. 00:12:23.450 --> 00:12:25.110 Yeah, if you're curious, go check it out. It's 00:12:25.110 --> 00:12:27.470 a really low friction way to show you're engaged 00:12:27.470 --> 00:12:30.080 before that deadline is up. And before we sign 00:12:30.080 --> 00:12:32.679 off, we want to leave you with one final provocative 00:12:32.679 --> 00:12:35.559 thought from this week's news. A Google VP of 00:12:35.559 --> 00:12:37.899 ads recently denied the rumor about integrating 00:12:37.899 --> 00:12:40.779 ads deep inside their personalized AI models. 00:12:41.360 --> 00:12:44.500 But the key phrase they used was, no current 00:12:44.500 --> 00:12:47.080 plans. We encourage you to think about what that 00:12:47.080 --> 00:12:50.299 very deliberate, nuanced phrasing implies about 00:12:50.299 --> 00:12:53.039 the inevitable future of advertising. What happens 00:12:53.039 --> 00:12:55.820 when our trusted, personalized study buddy also 00:12:55.820 --> 00:12:57.940 becomes an incredibly well -informed salesman? 00:12:58.190 --> 00:13:00.389 Keep digging into your own source material. We'll 00:13:00.389 --> 00:13:01.129 catch you on the next one.