WEBVTT 00:00:00.000 --> 00:00:02.560 I mean, you ask an AI a highly complex question. 00:00:03.819 --> 00:00:06.919 You hope for profound clarity, right? Yeah, you 00:00:06.919 --> 00:00:08.619 want to answer that actually connects the dots. 00:00:09.060 --> 00:00:11.140 Exactly. But instead, you just get this massive, 00:00:11.419 --> 00:00:15.400 impenetrable wall of text beat, and then the 00:00:15.400 --> 00:00:17.780 real work begins. Oh, totally. You have to manually 00:00:17.780 --> 00:00:20.320 extract that data line by line. You build the 00:00:20.320 --> 00:00:22.500 presentation yourself. You rebuild the spreadsheet 00:00:22.500 --> 00:00:25.320 yourself. It takes hours of your time. And honestly, 00:00:25.620 --> 00:00:28.079 it's incredibly frustrating. Two secs silence. 00:00:28.920 --> 00:00:30.640 But I want you to imagine a totally different 00:00:30.640 --> 00:00:33.000 scenario today. OK, I'm listening. You ask that 00:00:33.000 --> 00:00:36.520 exact same question, but instead of text, a fully 00:00:36.520 --> 00:00:41.710 editable PowerPoint file just appears. A populated 00:00:41.710 --> 00:00:44.009 Excel spreadsheet materializes right there in 00:00:44.009 --> 00:00:45.810 the chat. I mean, that changes everything about 00:00:45.810 --> 00:00:47.770 how we work. You stop fighting the interface 00:00:47.770 --> 00:00:50.210 entirely, and you start actually collaborating. 00:00:50.649 --> 00:00:52.609 Welcome to the deep dive. I'm very glad you're 00:00:52.609 --> 00:00:55.030 here with us. Our primary mission today is exploring 00:00:55.030 --> 00:00:57.670 that exact shift. We are looking really closely 00:00:57.670 --> 00:01:01.250 at Google's notebook, LM 2 .0, because the source 00:01:01.250 --> 00:01:03.869 material reveals a fascinating transition in 00:01:03.869 --> 00:01:06.090 what this technology is actually doing. Right. 00:01:06.430 --> 00:01:08.790 This tool was previously just a simple document 00:01:08.790 --> 00:01:10.409 reader. It would summarize what you gave it. 00:01:10.590 --> 00:01:13.349 Yeah, but now it operates much closer to a junior 00:01:13.349 --> 00:01:15.730 data analyst. To really trust an AI, to build 00:01:15.730 --> 00:01:17.870 our spreadsheets and parse our data, we first 00:01:17.870 --> 00:01:19.810 have to understand how it suddenly learned to 00:01:19.810 --> 00:01:22.790 do real math. And that requires looking closely 00:01:22.790 --> 00:01:24.569 at the engine powering this whole operation. 00:01:24.750 --> 00:01:27.510 Right, because to understand how it creates these 00:01:27.510 --> 00:01:30.549 tangible files, we need context. We kind of have 00:01:30.549 --> 00:01:33.769 to look at the brain first. Notebook LM has officially 00:01:33.769 --> 00:01:37.069 moved to the Gemini 3 .5 engine, and that brings 00:01:37.069 --> 00:01:40.010 some heavy implications for you. The benchmarks 00:01:40.010 --> 00:01:42.700 in the research are pretty staggering. It's showing 00:01:42.700 --> 00:01:45.959 a 78 .2 % win rate in web research. And when 00:01:45.959 --> 00:01:49.319 it comes to analyzing massive, complex documents, 00:01:49.500 --> 00:01:53.400 it is hitting nearly 70%. But, you know, raw 00:01:53.400 --> 00:01:55.859 numbers only tell part of the story. Yeah. The 00:01:55.859 --> 00:01:58.379 fascinating part is why it is suddenly so much 00:01:58.379 --> 00:02:01.519 more capable. Exactly. Historically, large language 00:02:01.519 --> 00:02:03.939 models are terrible at math. They don't actually 00:02:03.939 --> 00:02:05.939 calculate numbers. No, they just predict the 00:02:05.939 --> 00:02:08.219 next most likely word in a sequence based on 00:02:08.219 --> 00:02:10.520 their training data. Right. It is basically highly 00:02:10.520 --> 00:02:12.800 educated guessing. Which is why you could never 00:02:12.800 --> 00:02:16.120 truly cross them with your accounting or your 00:02:16.120 --> 00:02:18.560 precise data sets. But the architecture here 00:02:18.560 --> 00:02:21.120 has fundamentally changed. Every single notebook 00:02:21.120 --> 00:02:24.340 now gets its own sandbox. Right. Which is a safe, 00:02:24.819 --> 00:02:26.919 isolated space for running code. Right. And that 00:02:26.919 --> 00:02:30.000 is the actual game changer here. Before, the 00:02:30.000 --> 00:02:33.439 AI just estimated math from text. Now... It writes 00:02:33.439 --> 00:02:36.300 and runs its own code. It relies on over 100 00:02:36.300 --> 00:02:39.680 curated software skills to compute absolute answers. 00:02:40.159 --> 00:02:42.639 Whoa. I mean, imagine it just spinning up its 00:02:42.639 --> 00:02:45.500 own secure cloud computer to calculate your invoices. 00:02:45.520 --> 00:02:48.599 That is wild. You upload three years of complex 00:02:48.599 --> 00:02:50.979 billing records. Instead of summarizing what 00:02:50.979 --> 00:02:53.500 might be in them, it takes physical action. It 00:02:53.500 --> 00:02:56.020 writes a unique script. It runs it in that isolated 00:02:56.020 --> 00:02:58.379 sandbox. And it hands back a mathematically perfect 00:02:58.379 --> 00:03:01.370 finished report. Beat. Okay, let's untack this. 00:03:01.710 --> 00:03:03.409 It stops being a guessing machine and becomes 00:03:03.409 --> 00:03:06.069 a computation engine. But wait, if it is writing 00:03:06.069 --> 00:03:08.129 code to achieve this, am I expected to know how 00:03:08.129 --> 00:03:10.770 to prompt in Python or manually trigger these 00:03:10.770 --> 00:03:13.289 specific skills? Not at all. That is the beauty 00:03:13.289 --> 00:03:16.289 of the system design. The AI automatically evaluates 00:03:16.289 --> 00:03:18.870 your prompt, realizes it needs hard math instead 00:03:18.870 --> 00:03:21.650 of text, and silently writes and executes the 00:03:21.650 --> 00:03:24.050 necessary scripts in the background. So it silently 00:03:24.050 --> 00:03:26.469 picks the right tools without you lifting a finger. 00:03:26.750 --> 00:03:29.349 Precisely. It completely abstracts the coding 00:03:29.349 --> 00:03:32.250 layer away from you. But an engine that powerful 00:03:32.250 --> 00:03:35.310 is useless without the right fuel. So how do 00:03:35.310 --> 00:03:38.289 we actually feed this new system? The fundamental 00:03:38.289 --> 00:03:40.789 rule of data analysis has not changed at all. 00:03:41.009 --> 00:03:43.620 The quality of your sources still absolutely 00:03:43.620 --> 00:03:45.580 matters. Yeah, the old adage applies perfectly 00:03:45.580 --> 00:03:49.020 here. Barbage in, garbage out. Notebook LM simply 00:03:49.020 --> 00:03:51.389 amplifies the foundation you handed. Let's make 00:03:51.389 --> 00:03:53.530 this concrete for you. Imagine you're building 00:03:53.530 --> 00:03:56.009 a new course on freelance copywriting. OK, you 00:03:56.009 --> 00:03:59.150 need a very solid, multifaceted knowledge base 00:03:59.150 --> 00:04:02.110 for that. Exactly. So you upload competitor course 00:04:02.110 --> 00:04:05.210 pages and their pricing tiers. You drop in lengthy 00:04:05.210 --> 00:04:07.490 Reddit threads where real freelancers are asking 00:04:07.490 --> 00:04:09.430 questions. You add your own scattered notes from 00:04:09.430 --> 00:04:11.710 past client work. You include broad industry 00:04:11.710 --> 00:04:14.810 reports on freelance income trends. It actively 00:04:14.810 --> 00:04:17.149 connects all of those disparate pieces together. 00:04:17.589 --> 00:04:20.889 It treats all of it. as one connected mind. Yeah, 00:04:20.990 --> 00:04:22.889 rather than separate files sitting in different 00:04:22.889 --> 00:04:26.470 folders. But, you know, human research is rarely 00:04:26.470 --> 00:04:28.990 ever perfect. There are almost always missing 00:04:28.990 --> 00:04:31.230 pieces in our logic. Right, and this is where 00:04:31.230 --> 00:04:33.069 the self -directed research feature comes in. 00:04:33.290 --> 00:04:36.269 You can simply ask the chat what is missing from 00:04:36.269 --> 00:04:39.250 your own brain trust. It reviews your uploaded 00:04:39.250 --> 00:04:42.750 sources alongside your conversation. For our 00:04:42.750 --> 00:04:45.689 copywriting course example, it easily spots two 00:04:45.689 --> 00:04:49.449 clear structural gaps. First, it identifies completely 00:04:49.449 --> 00:04:51.970 missing financial data. Right. You lack customer 00:04:51.970 --> 00:04:54.329 acquisition cost numbers and you lack average 00:04:54.329 --> 00:04:57.379 conversion rates for sales pages. It finds the 00:04:57.379 --> 00:04:59.420 financial weaknesses in your business plan. It 00:04:59.420 --> 00:05:01.959 really does. Then it finds a blind spot in your 00:05:01.959 --> 00:05:04.879 focus. Your research perfectly explains how students 00:05:04.879 --> 00:05:07.500 will acquire freelance clients. But it entirely 00:05:07.500 --> 00:05:10.480 fails to explain how you, the creator, will acquire 00:05:10.480 --> 00:05:12.699 those students in the first place. I genuinely 00:05:12.699 --> 00:05:15.180 love that level of critical pushback from a tool. 00:05:15.540 --> 00:05:17.779 It's not just agreeing with me. No, it's finding 00:05:17.779 --> 00:05:19.759 the holes in your thinking. And you can tell 00:05:19.759 --> 00:05:21.920 it to actively chase down one of those gaps. 00:05:22.519 --> 00:05:24.560 Let's say you pick customer acquisition costs. 00:05:24.800 --> 00:05:28.379 It searches the live web for those exact missing 00:05:28.379 --> 00:05:31.100 metrics. It comes back with a tailored report 00:05:31.100 --> 00:05:33.939 and suggested external sources to fill the hole. 00:05:34.160 --> 00:05:37.759 Two -sec silence. I have to admit something vulnerable 00:05:37.759 --> 00:05:41.000 here. I still wrestle with the fear of AI hallucinating 00:05:41.000 --> 00:05:44.279 data, so approving sources manually is huge for 00:05:44.279 --> 00:05:47.000 me. If it's searching the open web to fill these 00:05:47.000 --> 00:05:49.579 gaps, how do I know it's not quietly slipping 00:05:49.579 --> 00:05:52.180 unverified junk in my course outline? Because 00:05:52.180 --> 00:05:54.300 of the hard gatekeeping built into the interface, 00:05:54.519 --> 00:05:57.600 it will suggest those web sources, yes, but it 00:05:57.600 --> 00:05:59.759 physically cannot incorporate that data into 00:05:59.759 --> 00:06:01.680 your Notebook's core knowledge base until you 00:06:01.680 --> 00:06:04.600 explicitly click approve. You remain the strict 00:06:04.600 --> 00:06:06.839 gatekeeper for every single piece of data. You 00:06:06.839 --> 00:06:09.379 do. Every answer stays completely grounded in 00:06:09.379 --> 00:06:11.800 material you have personally reviewed. This deep 00:06:11.800 --> 00:06:13.639 dive is completely supported by our listener 00:06:13.639 --> 00:06:15.879 community. Your backing keeps this conversation 00:06:15.879 --> 00:06:19.639 independent, ad -free, and deeply curious. We 00:06:19.639 --> 00:06:21.379 appreciate you being part of this journey with 00:06:21.379 --> 00:06:24.240 us. And we are back. Let's keep exploring this 00:06:24.240 --> 00:06:27.560 architecture. We now have a perfectly curated 00:06:27.560 --> 00:06:30.439 and vetted brain of sources. Right. But how do 00:06:30.439 --> 00:06:32.779 we actually extract the work from it? Right. 00:06:36.350 --> 00:06:38.610 downloadable files. What's fascinating here is 00:06:38.610 --> 00:06:40.769 the physical reality of the files it generates. 00:06:41.069 --> 00:06:43.050 This all happens within the studio panel on the 00:06:43.050 --> 00:06:45.149 right side of your screen. Let's look at the 00:06:45.149 --> 00:06:48.290 first major example. Say you want an Excel spreadsheet. 00:06:48.470 --> 00:06:51.389 Okay. You need a 12 -month revenue projection 00:06:51.389 --> 00:06:53.810 broken down by specific marketing campaigns. 00:06:54.250 --> 00:06:56.129 You just select the approved sources holding 00:06:56.129 --> 00:06:58.209 your figures and you give the instructions in 00:06:58.209 --> 00:07:00.750 the chat. It pulls those scattered numbers into 00:07:00.750 --> 00:07:03.560 one unified Excel sheet. And importantly, the 00:07:03.560 --> 00:07:06.399 formulas and the numbers are real data. They 00:07:06.399 --> 00:07:08.220 automatically update when you edit them later 00:07:08.220 --> 00:07:10.740 in Excel. Yeah, it performs the tedious structural 00:07:10.740 --> 00:07:13.160 formatting work for you. Saves hours of cell 00:07:13.160 --> 00:07:15.519 linking. And then there's the PowerPoint generation, 00:07:15.579 --> 00:07:18.160 which is equally impressive. You need a 10 slide 00:07:18.160 --> 00:07:20.680 pitch deck for investors. You tell it to focus 00:07:20.680 --> 00:07:23.079 heavily on the business model. You instruct it 00:07:23.079 --> 00:07:24.899 to skip the company history section entirely 00:07:24.899 --> 00:07:27.699 because investors don't care. A real formatted 00:07:27.699 --> 00:07:29.759 file shows up a few minutes later. But there 00:07:29.759 --> 00:07:31.839 is a significant cache we must discuss here. 00:07:31.980 --> 00:07:35.199 It is a known limitation that many users eventually 00:07:35.199 --> 00:07:38.959 run into. There is. It involves how the AI handles 00:07:38.959 --> 00:07:42.420 the PowerPoint text boxes. They often generate 00:07:42.420 --> 00:07:45.459 as flat images packed inside a PowerPoint frame 00:07:45.459 --> 00:07:48.240 rather than editable text fields. It's like stacking 00:07:48.240 --> 00:07:50.779 Lego blocks of data. You can't just mold the 00:07:50.779 --> 00:07:52.519 plastic, you have to swap the blog entirely. 00:07:53.019 --> 00:07:54.959 Because large language models struggle with spatial 00:07:54.959 --> 00:07:57.500 bounding boxes, they often just render the text 00:07:57.500 --> 00:08:00.120 as a picture. Exactly. You cannot simply click 00:08:00.120 --> 00:08:02.480 and type directly on the slide to fix a typo. 00:08:02.839 --> 00:08:05.000 You must use the revise button located in the 00:08:05.000 --> 00:08:07.680 chat interface. You ask the chat to change the 00:08:07.680 --> 00:08:09.899 specific slide. And it regenerates the whole 00:08:09.899 --> 00:08:12.300 block. The revise button becomes your primary 00:08:12.300 --> 00:08:14.860 tool for edits. It rewrites the slide instead 00:08:14.860 --> 00:08:17.899 of you typing on it. If I am staring at a newly 00:08:17.899 --> 00:08:20.480 generated deck right before a meeting, how do 00:08:20.480 --> 00:08:22.899 I quickly figure out if I am looking at editable 00:08:22.899 --> 00:08:25.399 text or just a baked -in image block? It is a 00:08:25.399 --> 00:08:28.240 very simple manual test. You download the file 00:08:28.240 --> 00:08:31.040 and just click any line of text. If your cursor 00:08:31.040 --> 00:08:33.200 does not let you type directly into the paragraph, 00:08:33.399 --> 00:08:36.659 it is an image block. So a quick click... instantly 00:08:36.659 --> 00:08:39.340 reveals if it is an image block. Exactly right. 00:08:39.860 --> 00:08:42.419 That need for structural revision is a good segue 00:08:42.419 --> 00:08:45.379 into how it handles truly dense contradictory 00:08:45.379 --> 00:08:48.740 information. Because sometimes even a perfectly 00:08:48.740 --> 00:08:51.740 formatted spreadsheet is still too dense. Right. 00:08:51.740 --> 00:08:53.799 We need to turn that density into a cohesive 00:08:53.799 --> 00:08:56.440 story we can understand at a single glance. Let's 00:08:56.440 --> 00:08:59.259 look at a very specific data comparison example 00:08:59.259 --> 00:09:02.580 from the research. You upload two distinct files 00:09:02.580 --> 00:09:05.220 into your notebook. First, your monthly Facebook 00:09:05.220 --> 00:09:08.360 ad spend broken down by campaign. Second, your 00:09:08.360 --> 00:09:10.679 daily course enrollments exported from your hosting 00:09:10.679 --> 00:09:13.240 platform. Two completely different formats of 00:09:13.240 --> 00:09:15.220 data. Right. And you ask it to compare those 00:09:15.220 --> 00:09:18.039 two specific data sets. The resulting analysis 00:09:18.039 --> 00:09:21.350 is incredibly sharp. It calculates the cost per 00:09:21.350 --> 00:09:23.690 enrollment for each month by synthesizing the 00:09:23.690 --> 00:09:26.509 two files. It discovers the cost drops from roughly 00:09:26.509 --> 00:09:31.529 $79 .59 in January, all the way down to $31 .91 00:09:31.529 --> 00:09:35.289 by June. The advertising is clearly getting much 00:09:35.289 --> 00:09:40.009 more efficient over time. But here is the truly 00:09:40.009 --> 00:09:42.529 critical moment in this interaction. It's the 00:09:42.529 --> 00:09:44.570 moment that proves how the architecture has changed. 00:09:45.049 --> 00:09:47.590 You ask the AI which specific campaign drove 00:09:47.590 --> 00:09:50.580 the most signups. And the system admits outright 00:09:50.580 --> 00:09:53.759 that it cannot answer that question. The uploaded 00:09:53.759 --> 00:09:56.639 data only shows total daily signups. It does 00:09:56.639 --> 00:09:59.240 not explicitly link those individual signups 00:09:59.240 --> 00:10:03.019 to specific ad campaigns. Beat, the AI does not 00:10:03.019 --> 00:10:05.240 try to guess. Yeah, that algorithmic honesty 00:10:05.240 --> 00:10:07.820 is absolutely vital. Historically, AI models 00:10:07.820 --> 00:10:09.879 are people pleasers. They are designed to give 00:10:09.879 --> 00:10:11.580 you an answer, even if they have to hallucinate 00:10:11.580 --> 00:10:13.620 to do it. But this system is programmed to hit 00:10:13.620 --> 00:10:16.350 a wall when the uploaded data stops. doesn't 00:10:16.350 --> 00:10:18.409 invent a bridge to cross it. Exactly. Which is 00:10:18.409 --> 00:10:20.509 why I'm so used to these models trying to impress 00:10:20.509 --> 00:10:22.690 me with an answer no matter what. Why is the 00:10:22.690 --> 00:10:25.070 AI refusing to answer the campaign question actually 00:10:25.070 --> 00:10:28.289 a massive feature, not a bug? Because its primary 00:10:28.289 --> 00:10:31.129 directive is grounded analysis, not conversation. 00:10:31.649 --> 00:10:35.049 It is designed to build profound trust. You know 00:10:35.049 --> 00:10:37.289 it will not invent convenient answers when your 00:10:37.289 --> 00:10:39.909 data falls short. honesty about its limits proves 00:10:39.909 --> 00:10:42.710 it will never invent fake data. Exactly. And 00:10:42.710 --> 00:10:44.970 you can visualize this data density easily, too. 00:10:45.049 --> 00:10:46.750 You just ask it to turn this comparison into 00:10:46.750 --> 00:10:48.889 a visual chart. The chart appears within a minute, 00:10:48.970 --> 00:10:50.909 ready to download. And it shows the ad spend 00:10:50.909 --> 00:10:55.539 climbing from $3 ,900 to $9 ,000. but the enrollments 00:10:55.539 --> 00:11:00.940 rocket from 49 to 282. That is a massive 475 00:11:00.940 --> 00:11:03.879 % increase. One glance proves the fundamental 00:11:03.879 --> 00:11:06.379 efficiency of the business. You see both lines 00:11:06.379 --> 00:11:08.519 rise, but the enrollments grow significantly 00:11:08.519 --> 00:11:11.149 faster. Yeah. You can run other incredibly valuable 00:11:11.149 --> 00:11:13.610 comparisons too. Weekly website traffic versus 00:11:13.610 --> 00:11:15.909 newsletter signups. Customer support ticket volume 00:11:15.909 --> 00:11:18.509 against average response times. Or department 00:11:18.509 --> 00:11:20.950 budgets against actual quarterly spending. It 00:11:20.950 --> 00:11:23.629 turns a dense written analytical answer into 00:11:23.629 --> 00:11:26.409 actionable clarity. Trust and clarity are great 00:11:26.409 --> 00:11:29.289 in theory. But we need to ground this entirely 00:11:29.289 --> 00:11:32.509 in practical reality before we wrap up. Who can 00:11:32.509 --> 00:11:35.330 actually use this system right now? And what 00:11:35.330 --> 00:11:37.649 is the best mental framework to approach it with? 00:11:37.870 --> 00:11:40.370 There is a paywall we need to acknowledge up 00:11:40.370 --> 00:11:42.750 front. This specific update is not available 00:11:42.750 --> 00:11:45.690 to everyone just yet. It requires Google AI Ultra. 00:11:46.230 --> 00:11:49.289 Or you need to be a workspace business customer 00:11:49.289 --> 00:11:52.429 with AI Ultra or Expanded Access. Broad access 00:11:52.429 --> 00:11:54.669 is planned, but it's currently limited. Let's 00:11:54.669 --> 00:11:56.769 quickly recap the ideal workflow you can copy 00:11:56.769 --> 00:11:59.330 today if you have access. Say you run a small 00:11:59.330 --> 00:12:01.759 online store. you want to decide whether to expand 00:12:01.759 --> 00:12:03.899 your product line. You start a notebook with 00:12:03.899 --> 00:12:06.679 your existing documents, supplier catalogs, historical 00:12:06.679 --> 00:12:09.899 sales reports, customer reviews, competitor pricing 00:12:09.899 --> 00:12:12.200 matrices. Then you actively ask the tool what 00:12:12.200 --> 00:12:14.970 strategic gaps exist. You carefully review and 00:12:14.970 --> 00:12:16.970 approve any new web sources before adding them 00:12:16.970 --> 00:12:19.309 to the brain. Right. Then you generate a slide 00:12:19.309 --> 00:12:22.269 deck for a supplier pitch. You build a data comparison 00:12:22.269 --> 00:12:24.490 table comparing your past sales against ad spend. 00:12:24.950 --> 00:12:27.409 Finally, you turn that comparison into a visual 00:12:27.409 --> 00:12:30.529 chart for your team. But you must remember the 00:12:30.529 --> 00:12:34.149 absolute golden rule here. You must treat everything 00:12:34.149 --> 00:12:36.990 this system produces as a powerful first draft. 00:12:37.149 --> 00:12:40.009 You cannot treat this as a perfectly finished, 00:12:40.389 --> 00:12:43.129 infallible product. You always verify the Excel 00:12:43.129 --> 00:12:46.389 math against your raw sources. You always test 00:12:46.389 --> 00:12:48.750 those PowerPoint editing capabilities before 00:12:48.750 --> 00:12:51.269 walking into a presentation. Treating it as a 00:12:51.269 --> 00:12:53.610 strong first draft keeps the entire workflow 00:12:53.610 --> 00:12:57.389 trustworthy. It saves immense time, but it doesn't 00:12:57.389 --> 00:13:00.110 replace careful human verification. With all 00:13:00.110 --> 00:13:02.669 this capability, from coding in sandboxes to 00:13:02.669 --> 00:13:05.269 building pitch decks, does a tool like this effectively 00:13:05.269 --> 00:13:07.750 replace a human data analyst on a small team? 00:13:07.970 --> 00:13:10.570 Not at all. It just closes the massive gap between 00:13:10.570 --> 00:13:12.830 having a business problem and holding a workable 00:13:12.830 --> 00:13:15.690 draft solution. The human still completely owns 00:13:15.690 --> 00:13:18.129 the final strategic decisions. It does the grunt 00:13:18.129 --> 00:13:20.529 work, but you make the final strategic decisions. 00:13:20.629 --> 00:13:22.710 Perfectly said. The critical thinking remains 00:13:22.710 --> 00:13:25.370 entirely yours. Let's take a measured calm moment 00:13:25.370 --> 00:13:27.950 here at the end. Let's reflect on the overarching 00:13:27.950 --> 00:13:30.490 theme of this deep dive. We have witnessed a 00:13:30.490 --> 00:13:33.009 profound architectural shift in consumer technology 00:13:33.009 --> 00:13:35.669 today. We really have. We have officially moved 00:13:35.669 --> 00:13:39.110 from AI as a simple conversational partner, a 00:13:39.110 --> 00:13:42.929 glorified chatbot, to AI as an active synthesizer 00:13:42.929 --> 00:13:45.669 and real file creator. It uses isolated secure 00:13:45.669 --> 00:13:48.850 computation to do real verifiable math. It spots 00:13:48.850 --> 00:13:51.610 the logical gaps in our own human research. And 00:13:51.610 --> 00:13:54.009 it outputs highly tangible tools like bitch decks 00:13:54.009 --> 00:13:57.990 and complex charts. to sex silence. But all of 00:13:57.990 --> 00:14:00.429 this extraordinary capability comes with one 00:14:00.429 --> 00:14:03.370 fundamental condition. We must give it a solid 00:14:03.370 --> 00:14:06.220 factual foundation to stand on. I want to leave 00:14:06.220 --> 00:14:08.059 you with a final thought to ponder as you go 00:14:08.059 --> 00:14:10.799 about your day. If an AI can now perfectly connect 00:14:10.799 --> 00:14:12.960 the dots between your scattered notes and turn 00:14:12.960 --> 00:14:15.200 them into a beautifully polished investor pitch 00:14:15.200 --> 00:14:18.259 in seconds, it raises a profound question. Yeah. 00:14:18.500 --> 00:14:20.460 Will the true future skill be less about knowing 00:14:20.460 --> 00:14:23.320 how to format a presentation and entirely about 00:14:23.320 --> 00:14:25.799 how deeply you understand the original raw ideas 00:14:25.799 --> 00:14:28.039 you feed into the machine? Keep questioning your 00:14:28.039 --> 00:14:30.539 assumptions. Keep exploring these new frontiers. 00:14:31.059 --> 00:14:32.779 Thank you for joining us on this deep dive.