WEBVTT 00:00:00.000 --> 00:00:02.299 So just imagine for a second, you take your company's 00:00:02.299 --> 00:00:05.759 entire, you know, proprietary code base or maybe 00:00:05.759 --> 00:00:08.019 a decade of legal documents and you just upload 00:00:08.019 --> 00:00:10.099 it all into one single box. That's the scale 00:00:10.099 --> 00:00:11.880 we're talking about now. It's not just about 00:00:11.880 --> 00:00:14.900 slightly bigger inputs anymore. The whole game 00:00:14.900 --> 00:00:17.100 has changed, really. And the big one is that 00:00:17.100 --> 00:00:20.160 Disney is backing that scale with a billion dollars. 00:00:20.239 --> 00:00:22.280 It proves this isn't an experiment for them. 00:00:22.320 --> 00:00:26.140 It's destiny. Welcome back. Today we are doing 00:00:26.140 --> 00:00:28.980 a deep dive into your sources. We're going to 00:00:28.980 --> 00:00:32.799 be looking at the huge new GBT 5 .2 launch and 00:00:32.799 --> 00:00:36.000 really how AI is embedding itself into business 00:00:36.000 --> 00:00:39.070 strategy and, well... our daily lives. Yeah, 00:00:39.149 --> 00:00:41.390 we have a really dense stack of material today. 00:00:41.490 --> 00:00:44.109 We're covering the tech itself, this new model 00:00:44.109 --> 00:00:47.030 code named Garlic, the corporate rivalry it sparked, 00:00:47.310 --> 00:00:50.070 and a fascinating report on how millions of people 00:00:50.070 --> 00:00:52.549 are actually using these things right now. We're 00:00:52.549 --> 00:00:54.549 going to unpack those specs, figure out what 00:00:54.549 --> 00:00:56.770 those three new tiers of AI really mean for a 00:00:56.770 --> 00:00:59.969 business user, and we'll reveal the pretty surprising 00:00:59.969 --> 00:01:01.609 number one topic people are talking about with 00:01:01.609 --> 00:01:04.230 their AI companions. So first up, we'll dissect 00:01:04.230 --> 00:01:07.819 the specs on GPT 5 .2. we need to talk about 00:01:07.819 --> 00:01:10.000 its raw power. And we'll connect that tech leap 00:01:10.000 --> 00:01:13.420 to the billion -dollar investments and the escalating 00:01:13.420 --> 00:01:15.379 rivalry we're seeing. And finally, we're going 00:01:15.379 --> 00:01:17.519 to look at the human side of it. We have data 00:01:17.519 --> 00:01:21.060 that shows how AI is shifting roles from a philosopher 00:01:21.060 --> 00:01:23.719 at 2 a .m. to a trip planner during work hours. 00:01:24.040 --> 00:01:26.000 Okay, so let's start with the tech. This new 00:01:26.000 --> 00:01:28.579 model... Garlic. It dropped just a few weeks 00:01:28.579 --> 00:01:31.079 after the last major update. Right. That in itself 00:01:31.079 --> 00:01:34.079 is a huge signal. It tells you the pace of development 00:01:34.079 --> 00:01:37.659 is accelerating way faster than most people predicted. 00:01:37.920 --> 00:01:40.200 And what's really significant here is the scale. 00:01:40.359 --> 00:01:43.040 I mean, this isn't really for your average chatbot 00:01:43.040 --> 00:01:46.840 user. It is explicitly built for complex AI agents 00:01:46.840 --> 00:01:49.000 and these heavy coding workflows. The headline 00:01:49.000 --> 00:01:51.680 spec that makes everyone stop is the 400 ,000 00:01:51.680 --> 00:01:54.250 token context window. We should probably define 00:01:54.250 --> 00:01:56.750 that. Yeah. Simply put, the context window is 00:01:56.750 --> 00:01:59.569 just how much information the AI can hold in 00:01:59.569 --> 00:02:01.829 its short -term memory at once. Think about it 00:02:01.829 --> 00:02:04.689 like this. The last generation could maybe handle 00:02:04.689 --> 00:02:09.569 a long article. This new 400K window. You can 00:02:09.569 --> 00:02:13.210 literally drop in massive files an entire book 00:02:13.210 --> 00:02:16.229 of, say, 300 ,000 words or a whole code base. 00:02:16.490 --> 00:02:18.909 And it analyzes all of it in context. It doesn't 00:02:18.909 --> 00:02:21.449 forget page one when it gets to page 300. Exactly. 00:02:21.469 --> 00:02:24.169 And the output is just as big. It can handle 00:02:24.169 --> 00:02:27.810 up to 128 ,000 tokens in a single response, which 00:02:27.810 --> 00:02:30.810 is huge for any kind of complex long form task. 00:02:31.169 --> 00:02:33.349 Whoa. I mean, just imagine scaling that kind 00:02:33.349 --> 00:02:35.710 of analytical power to a billion daily queries. 00:02:35.889 --> 00:02:38.129 It's a fundamental shift in what's even possible. 00:02:38.629 --> 00:02:40.530 Yeah. And that power, of course, comes at a cost. 00:02:40.669 --> 00:02:42.710 The source of note is about 40 percent pricier 00:02:42.710 --> 00:02:45.129 than the last model. So it's very clearly aimed 00:02:45.129 --> 00:02:47.530 at serious enterprise users. For sure. And for 00:02:47.530 --> 00:02:50.030 the first time, OpenAI is actually offering three 00:02:50.030 --> 00:02:52.349 official versions, which seems to reflect different 00:02:52.349 --> 00:02:54.250 kinds of workloads. Right. It's not one size 00:02:54.250 --> 00:02:55.969 fits all anymore. No. You've got the instant 00:02:55.969 --> 00:02:58.310 tier. That's for quick replies, simple tasks 00:02:58.310 --> 00:03:00.409 where speed is everything. OK, then there's the 00:03:00.409 --> 00:03:02.349 thinking tier. This one feels like the core of 00:03:02.349 --> 00:03:04.669 it. It's designed for that multi -step agent 00:03:04.669 --> 00:03:08.270 style reasoning. Right. So think of like. A customer 00:03:08.270 --> 00:03:11.210 support ticket. The AI can analyze the mood, 00:03:11.370 --> 00:03:13.870 check the user's history, find the bug in the 00:03:13.870 --> 00:03:16.550 database, and then draft an apology all in one 00:03:16.550 --> 00:03:18.870 go. That's the kind of workflow it's built for. 00:03:19.129 --> 00:03:22.110 And finally, the pro tier. That's for the really 00:03:22.110 --> 00:03:24.490 complex stuff like deep research projects where 00:03:24.490 --> 00:03:27.930 you need absolute precision and cost is a secondary 00:03:27.930 --> 00:03:30.710 concern. So these three tiers, they really show 00:03:30.710 --> 00:03:33.509 AI is moving from just a general tool to a very 00:03:33.509 --> 00:03:35.310 specialized engine. You have to pick the right 00:03:35.310 --> 00:03:37.310 one for the job. So that raises a big question. 00:03:37.719 --> 00:03:41.479 How do these three cures actually change how 00:03:41.479 --> 00:03:43.919 a company would build its AI processes? Well, 00:03:43.979 --> 00:03:46.560 they formalize those complex workflows for multi 00:03:46.560 --> 00:03:49.039 -step business tasks. And this is where the business 00:03:49.039 --> 00:03:51.259 side of things gets really interesting. On the 00:03:51.259 --> 00:03:54.180 exact same day this launched, Disney became its 00:03:54.180 --> 00:03:56.860 biggest content partner and investor. A billion 00:03:56.860 --> 00:03:59.400 dollars. That's a massive investment. And they're 00:03:59.400 --> 00:04:02.099 aiming to integrate AI magic, as they call it, 00:04:02.139 --> 00:04:05.240 into their content by 2026. This is not a pilot 00:04:05.240 --> 00:04:07.930 program. This is them saying. LLMs are now a 00:04:07.930 --> 00:04:10.849 core creative tool. And you need that 400K token 00:04:10.849 --> 00:04:13.650 capacity we just talked about to actually manage 00:04:13.650 --> 00:04:17.089 a library of IP as vast as Disney's. Exactly. 00:04:17.110 --> 00:04:19.930 And on top of that, OpenAI also confirmed their 00:04:19.930 --> 00:04:22.350 models are being integrated into U .S. federal 00:04:22.350 --> 00:04:24.389 agencies. Which is a huge vote of confidence. 00:04:24.629 --> 00:04:26.930 It is. So, you know, GPT -4 got people curious. 00:04:27.209 --> 00:04:30.029 But GPT -5 .2, it seems like it wants to run 00:04:30.029 --> 00:04:32.290 your business. The Disney deal and the federal 00:04:32.290 --> 00:04:34.550 integration, they just confirmed that shift. 00:04:34.689 --> 00:04:37.829 The competition is also... So it's just fierce. 00:04:38.069 --> 00:04:41.170 Google released its own deep AI research agent 00:04:41.170 --> 00:04:43.569 on the exact same day. The exact same day. They 00:04:43.569 --> 00:04:45.410 call it Disco, and that is not a coincidence. 00:04:45.649 --> 00:04:48.290 It's a full -blown arms race. A match made in 00:04:48.290 --> 00:04:51.009 heaven for tech rivalry, for sure. It shows that 00:04:51.009 --> 00:04:52.470 they're watching each other's release schedules 00:04:52.470 --> 00:04:54.970 like hawks. And if we zoom out a bit, one of 00:04:54.970 --> 00:04:57.649 the sources mentioned an open AI exec who identified 00:04:57.649 --> 00:05:00.569 three specific jobs that are on the cusp of being 00:05:00.569 --> 00:05:02.550 automated by this. Right. That's the other side 00:05:02.550 --> 00:05:04.230 of the coin, the flip side of all this efficiency. 00:05:04.509 --> 00:05:08.689 And we also saw the AI 2027 tracker, which has 00:05:08.689 --> 00:05:11.870 been, what, almost 91 % accurate so far. It actually 00:05:11.870 --> 00:05:14.569 became a question on Jeopardy. So these predictions 00:05:14.569 --> 00:05:17.490 are becoming reality really, really fast. So 00:05:17.490 --> 00:05:19.730 what does a billion dollar partnership like this? 00:05:19.819 --> 00:05:23.319 signal about the future of, say, intellectual 00:05:23.319 --> 00:05:26.279 property. It means IP holders see these models 00:05:26.279 --> 00:05:29.199 as essential creation partners now. Okay, so 00:05:29.199 --> 00:05:31.759 let's shift to this Microsoft report. It analyzed 00:05:31.759 --> 00:05:35.620 intent data from over 37 million copilot chats. 00:05:36.220 --> 00:05:39.379 It gives us this very human, very real picture 00:05:39.379 --> 00:05:42.220 of how AI is being used day to day. Yeah. And 00:05:42.220 --> 00:05:44.199 this data explains why companies like Shopify 00:05:44.199 --> 00:05:46.860 and Adobe are baking AI into everything. They 00:05:46.860 --> 00:05:48.800 see the real demand there. And here's the big 00:05:48.800 --> 00:05:51.939 surprise. The number one use case. It was health 00:05:51.939 --> 00:05:54.819 and wellness. Yeah. Across every device, every 00:05:54.819 --> 00:05:57.220 hour of the day, people were asking co -pilot 00:05:57.220 --> 00:05:59.060 questions about their health. More than coding. 00:05:59.180 --> 00:06:01.600 More than summarizing. It's a profound signal 00:06:01.600 --> 00:06:04.420 of trust, you know, because the AI offers instant 00:06:04.420 --> 00:06:07.300 access and I guess a feeling of anonymity. There's 00:06:07.300 --> 00:06:09.860 no human judgment. And we also see behavior changing 00:06:09.860 --> 00:06:11.620 based on the time of day, which is fascinating. 00:06:11.699 --> 00:06:13.439 It's like people are assigning the AI different 00:06:13.439 --> 00:06:16.899 roles. Right. The data is so clear. At 2 a .m., 00:06:16.899 --> 00:06:19.680 people ask about philosophy, deep existential 00:06:19.680 --> 00:06:21.980 stuff. It's a late night counselor. But then 00:06:21.980 --> 00:06:24.839 during work hours, it's used heavily for trip 00:06:24.839 --> 00:06:27.449 planning. Not just simple queries either. We're 00:06:27.449 --> 00:06:29.649 talking complex logistics, multiple flights, 00:06:29.910 --> 00:06:32.750 hotels, budget constraints. Which makes sense. 00:06:32.750 --> 00:06:35.170 That's a high effort task. And offloading that 00:06:35.170 --> 00:06:38.149 during the workday is a huge time saver. Exactly. 00:06:38.149 --> 00:06:40.189 And the patterns hold up. Weekdays have more 00:06:40.189 --> 00:06:42.550 programming questions. Weekends, you see a big 00:06:42.550 --> 00:06:45.990 shift toward gaming and creative stuff. So people 00:06:45.990 --> 00:06:48.750 are basically using it like a companion that 00:06:48.750 --> 00:06:52.149 just shifts roles. It's like stacking Lego blocks 00:06:52.149 --> 00:06:54.110 of utility depending on what you need right then. 00:06:54.430 --> 00:06:56.449 It is. You know, I still wrestle with prompt 00:06:56.449 --> 00:06:58.670 drift myself sometimes. Oh, sure. Where the AI 00:06:58.670 --> 00:07:01.089 kind of loses the thread in a long chat. But 00:07:01.089 --> 00:07:03.750 this data, it really suggests people are getting 00:07:03.750 --> 00:07:06.389 past that and finding real utility in their personal 00:07:06.389 --> 00:07:09.069 lives, not just for work. And there was a clear 00:07:09.069 --> 00:07:11.089 trend over time, too. Back in January, it was 00:07:11.089 --> 00:07:13.610 mostly toting questions, early adopters. Right. 00:07:13.610 --> 00:07:16.290 But by September, society and culture queries 00:07:16.290 --> 00:07:18.870 had taken over, which shows much broader adoption. 00:07:19.439 --> 00:07:21.660 And importantly, Microsoft concerned they only 00:07:21.660 --> 00:07:24.899 analyzed summarized intent, not the full transcripts. 00:07:24.899 --> 00:07:27.300 So user privacy was protected while they gathered 00:07:27.300 --> 00:07:29.500 these insights. So why is this rise of health 00:07:29.500 --> 00:07:32.939 and wellness use cases such a pivotal sign for 00:07:32.939 --> 00:07:36.800 AI adoption? It shows users trust AI as a personal 00:07:36.800 --> 00:07:40.000 companion for crucial life areas. So after all 00:07:40.000 --> 00:07:42.180 of that, what does this all mean for you? We've 00:07:42.180 --> 00:07:45.019 seen this generational leap in AI scale handling 00:07:45.019 --> 00:07:47.980 massive documents, entire code bases. And we've 00:07:47.980 --> 00:07:50.439 seen the market validation. That billion from 00:07:50.439 --> 00:07:52.779 Disney really confirms that AI is going to be 00:07:52.779 --> 00:07:55.360 deeply integrated into content and infrastructure. 00:07:55.779 --> 00:07:57.500 But I think the most profound takeaway is that 00:07:57.500 --> 00:08:00.720 behavioral shift. AI is moving from just a curiosity 00:08:00.720 --> 00:08:03.480 to a trusted companion for everything from enterprise 00:08:03.480 --> 00:08:06.639 code to, you know, 2 a .m. philosophical questions. 00:08:06.879 --> 00:08:08.860 Yeah, it's no wonder that Times Person of the 00:08:08.860 --> 00:08:11.720 Year for 2025 is apparently just AI. itself. 00:08:11.720 --> 00:08:13.899 The technology is shaping the conversation like 00:08:13.899 --> 00:08:16.240 nothing before. So here's a final thought. If 00:08:16.240 --> 00:08:20.660 that AI 2027 tracker has been 91 % accurate so 00:08:20.660 --> 00:08:23.540 far, what are the real implications for those 00:08:23.540 --> 00:08:26.800 three specific white -collar jobs that an open 00:08:26.800 --> 00:08:29.680 AI exec says are on the immediate cusp of automation? 00:08:30.220 --> 00:08:32.279 That's something for you to mull over. The real 00:08:32.279 --> 00:08:34.360 story here isn't just the tech. It's the incredible 00:08:34.360 --> 00:08:37.100 speed of adoption. We hope this deep dive gave 00:08:37.100 --> 00:08:39.360 you that crucial shortcut to being well -informed 00:08:39.360 --> 00:08:40.960 on where this is all heading. And we'd encourage 00:08:40.960 --> 00:08:43.019 you to try Google's Disco to see it for yourself. 00:08:43.340 --> 00:08:45.460 It lets you turn your browser tabs into custom 00:08:45.460 --> 00:08:48.759 apps just by asking. Dig deeper. Until next time, 00:08:48.779 --> 00:08:50.440 keep exploring. Keep learning.