WEBVTT 00:00:00.000 --> 00:00:01.620 I was looking at my calendar this morning, and 00:00:01.620 --> 00:00:05.540 I had this flashback. It's been, what, three 00:00:05.540 --> 00:00:07.919 years since I spent an entire Sunday copying 00:00:07.919 --> 00:00:11.359 data from PDF reports into a slide deck. You 00:00:11.359 --> 00:00:13.099 know that feeling. Oh, I know it all too well. 00:00:13.220 --> 00:00:16.640 The Sunday scaries powered by copy and paste. 00:00:16.760 --> 00:00:18.739 Exactly. And then I read this stat today that 00:00:18.739 --> 00:00:22.250 just made me stop. that same eight hour job you 00:00:22.250 --> 00:00:23.829 do it now in about three minutes three minutes 00:00:23.829 --> 00:00:26.469 and that's not for like a rough draft full of 00:00:26.469 --> 00:00:29.170 holes we're talking a 22 slide deck yeah table 00:00:29.170 --> 00:00:32.369 of contents conclusion it's all there it's honestly 00:00:32.369 --> 00:00:34.750 it's a little jarring jarring is the word it 00:00:34.750 --> 00:00:36.789 really sinks the stage for today it is tuesday 00:00:36.789 --> 00:00:41.829 january 20 2026 welcome to the deep dive today 00:00:41.829 --> 00:00:43.710 i want to slow things down a bit we're often 00:00:43.710 --> 00:00:45.609 chasing what's new but today we're looking at 00:00:45.609 --> 00:00:48.100 a tool that sort of shifted right under our feet. 00:00:48.259 --> 00:00:50.719 We're unpacking an article by Max Sam called 00:00:50.719 --> 00:00:53.780 I Was Wrong About Notebook LM. It's such a great 00:00:53.780 --> 00:00:57.200 piece. Max just nails this really subtle change. 00:00:57.420 --> 00:01:00.179 He admits he was wrong because he, well, like 00:01:00.179 --> 00:01:02.359 a lot of us, he kind of pigeonholed the tool 00:01:02.359 --> 00:01:05.780 back in 2023, 2024. Right. He frames it as a 00:01:05.780 --> 00:01:07.959 shift from a toy to a factory. Yeah, exactly. 00:01:08.140 --> 00:01:10.280 He basically says, look, in the beginning, Notebook 00:01:10.280 --> 00:01:12.760 LM was just a place to dump your PDFs and ask, 00:01:12.819 --> 00:01:14.620 you know, hey, what's on page seven? A simple 00:01:14.620 --> 00:01:17.680 retrieval tool. Yeah. But his whole point now 00:01:17.680 --> 00:01:21.519 in 2026 is that it's quietly become a production 00:01:21.519 --> 00:01:24.239 studio. It went from a place where you collect 00:01:24.239 --> 00:01:26.680 stuff to a place where you actually produce things. 00:01:27.219 --> 00:01:30.840 Slides, audio, data tables. And that is a huge 00:01:30.840 --> 00:01:33.480 shift in workflow. So here's our roadmap for 00:01:33.480 --> 00:01:35.040 today. We're going to explore that evolution 00:01:35.040 --> 00:01:37.859 from archive to studio. Then we'll tackle slide 00:01:37.859 --> 00:01:41.319 fatigue and, you know, whether these AI slides 00:01:41.319 --> 00:01:43.819 are any good. I'm a little skeptical. Fair enough. 00:01:43.939 --> 00:01:45.739 We'll also look at the new interactive audio. 00:01:45.859 --> 00:01:47.680 And then we have to spend some time on the Gemini 00:01:47.680 --> 00:01:49.739 integration. That feels like the piece that ties 00:01:49.739 --> 00:01:52.079 it all together. That's the missing link. absolutely 00:01:52.079 --> 00:01:54.079 i'm ready okay let's start with the evolution 00:01:54.079 --> 00:01:56.819 max talks about this move from the old notebook 00:01:56.819 --> 00:02:01.579 lm to the new notebook lm of late 2025. for anyone 00:02:01.579 --> 00:02:03.859 who hasn't logged in for a while what's the core 00:02:03.859 --> 00:02:06.299 difference well the old version was pure q a 00:02:06.299 --> 00:02:09.280 you'd upload a doc ask a question get an answer 00:02:09.280 --> 00:02:12.240 it was grounded which was its big strength but 00:02:12.240 --> 00:02:15.460 it was uh passive right the new version is a 00:02:15.460 --> 00:02:19.449 studio it supports up to 50 sources now PDFs, 00:02:19.509 --> 00:02:22.889 URLs, YouTube transcripts, and it treats all 00:02:22.889 --> 00:02:25.830 of them like one collective brain. 50 sources. 00:02:26.129 --> 00:02:29.349 50. And the core idea is the same. If a fact 00:02:29.349 --> 00:02:32.349 isn't in those 50 sources, the AI will not make 00:02:32.349 --> 00:02:34.129 it up. It's the anti -hallucination machine. 00:02:34.330 --> 00:02:36.610 That grounding is so important, but I want to 00:02:36.610 --> 00:02:39.000 pause on that number, 50. That sounds like a 00:02:39.000 --> 00:02:41.199 lot of information, a lot of noise. It is a lot 00:02:41.199 --> 00:02:43.539 of noise for a human. But this is where I honestly 00:02:43.539 --> 00:02:45.400 had this moment of wonder reading the article. 00:02:45.620 --> 00:02:47.800 Just, you know, imagine trying to hold 50 different 00:02:47.800 --> 00:02:49.479 documents in your head at once. It's impossible. 00:02:49.879 --> 00:02:51.500 Completely. You're trying to find connections 00:02:51.500 --> 00:02:53.719 between a YouTube transcript, a legal PDF, and 00:02:53.719 --> 00:02:56.180 a website. We just can't do it. Our brains aren't 00:02:56.180 --> 00:02:58.900 built to parse that much unstructured data. But 00:02:58.900 --> 00:03:02.960 this tool, it synthesizes it instantly. It connects 00:03:02.960 --> 00:03:04.879 dots we wouldn't even know were on the same page. 00:03:05.419 --> 00:03:08.180 So let me ask you this then. Does this shift 00:03:08.180 --> 00:03:11.439 from just retrieval to production actually change 00:03:11.439 --> 00:03:14.099 the work we do? Or does it just make us better 00:03:14.099 --> 00:03:16.659 hoarders of information? It stops us from hoarding 00:03:16.659 --> 00:03:19.539 info and starts us creating finished work. Okay, 00:03:19.620 --> 00:03:22.580 from hoarding to creating. That leads us right 00:03:22.580 --> 00:03:25.259 into the first big use case Max talks about. 00:03:25.860 --> 00:03:29.659 Slides. The bane of every office worker's existence. 00:03:30.099 --> 00:03:31.740 We mentioned that three -minute number at the 00:03:31.740 --> 00:03:34.879 start. He lays out a scenario. You're doing a 00:03:34.879 --> 00:03:37.360 competitive analysis. Normally that's 50 tabs 00:03:37.360 --> 00:03:40.599 open. You're fighting with text boxes. Oh, the 00:03:40.599 --> 00:03:43.400 worst. In Notebook LM, you just upload the sources 00:03:43.400 --> 00:03:45.840 and click generate slide deck. You wait about 00:03:45.840 --> 00:03:47.479 three minutes. Okay, here's where I have to push 00:03:47.479 --> 00:03:50.060 back a little. I've seen AI generated slides. 00:03:50.360 --> 00:03:52.819 They usually look like a 10th grade science project 00:03:52.819 --> 00:03:56.139 or the text is just off. Is this actually something 00:03:56.139 --> 00:03:58.259 you could use in a boardroom? That's a totally 00:03:58.259 --> 00:04:00.830 fair skepticism. Yeah, and Max is... Honest about 00:04:00.830 --> 00:04:02.550 it, he says, you know, it's not Canva Pro. It's 00:04:02.550 --> 00:04:04.830 not Adobe. You're not going to win a design award 00:04:04.830 --> 00:04:06.830 right out of the box. So what's the value? The 00:04:06.830 --> 00:04:09.310 value isn't the graphic design. It's the structure. 00:04:09.449 --> 00:04:11.389 The structure. Yeah, think about that blank page 00:04:11.389 --> 00:04:13.949 problem. You're staring at an empty PowerPoint. 00:04:14.250 --> 00:04:17.550 You have no idea where to even start. This tool 00:04:17.550 --> 00:04:19.949 gives you a 22 slide deck that has a logical 00:04:19.949 --> 00:04:23.470 flow. Title, executive summary, market analysis, 00:04:23.629 --> 00:04:26.449 conclusion. It builds the argument for you. So 00:04:26.449 --> 00:04:28.689 it's less about the final polish. And more about 00:04:28.689 --> 00:04:31.250 the skeleton. Okay. Exactly. It changes your 00:04:31.250 --> 00:04:33.089 role. You're not the builder anymore. You're 00:04:33.089 --> 00:04:35.689 the director. You stop moving pixels around and 00:04:35.689 --> 00:04:37.889 you start critiquing the story. You can tell 00:04:37.889 --> 00:04:40.209 it, hey, focus more on the market gaps in slide 00:04:40.209 --> 00:04:43.569 four. And it just, it adjusts. That's a really 00:04:43.569 --> 00:04:45.589 interesting mental shift. You're directing instead 00:04:45.589 --> 00:04:48.870 of doing the labor. Right. And for us non -designers, 00:04:48.870 --> 00:04:50.930 I mean, I can't match colors to save my life. 00:04:51.009 --> 00:04:53.649 Yeah. He brings up the infographic feature. It's 00:04:53.649 --> 00:04:56.470 the same idea. You have this wall of text, maybe 00:04:56.470 --> 00:04:59.029 a dense policy document, and you know nobody 00:04:59.029 --> 00:05:01.649 is going to read it. You click lay plus infographic, 00:05:01.889 --> 00:05:05.149 and it visualizes the key data from what you 00:05:05.149 --> 00:05:07.709 gave it. And again, because it's grounded, it's 00:05:07.709 --> 00:05:09.649 not just inventing numbers to make a pretty chart. 00:05:09.810 --> 00:05:12.689 It's using the actual data from your PDF. That 00:05:12.689 --> 00:05:15.329 reliability is the whole game. So let me ask, 00:05:15.490 --> 00:05:18.709 does this actually replace a human designer? 00:05:19.449 --> 00:05:21.870 Or is it doing something else entirely? It's 00:05:21.870 --> 00:05:24.769 a first draft machine, not a polished final product. 00:05:24.850 --> 00:05:26.790 A first draft machine. I like that phrasing. 00:05:26.829 --> 00:05:29.550 It lowers the barrier to just starting. Let's 00:05:29.550 --> 00:05:31.870 pivot to the second big use case, what Mix calls 00:05:31.870 --> 00:05:34.550 the custom expert. This is probably the most 00:05:34.550 --> 00:05:37.870 2026 feature of the bunch. He sets it against 00:05:37.870 --> 00:05:40.589 generic chatbots. We've all been there. You ask 00:05:40.589 --> 00:05:42.410 a very specific question about your company's 00:05:42.410 --> 00:05:44.949 internal project, and the bot gives you this 00:05:44.949 --> 00:05:47.329 beautifully written, super confident answer that 00:05:47.329 --> 00:05:50.129 is... completely wrong or just gives you the 00:05:50.129 --> 00:05:52.689 most generic advice to manage a project try using 00:05:52.689 --> 00:05:55.009 a calendar right thanks for that it just lacks 00:05:55.009 --> 00:05:58.050 context so notebook LM acts as your project specific 00:05:58.050 --> 00:06:01.670 expert you create a notebook and you only upload 00:06:01.670 --> 00:06:04.970 your relevant files your brand guidelines your 00:06:04.970 --> 00:06:08.529 meeting notes that one messy PDF from HR about 00:06:08.529 --> 00:06:10.490 the new benefits and it just ignores the rest 00:06:10.490 --> 00:06:13.009 of the internet totally It puts on blinders. 00:06:13.009 --> 00:06:14.990 It only knows what you've put in that folder. 00:06:15.189 --> 00:06:17.589 So if you ask it, what is our refund policy for 00:06:17.589 --> 00:06:21.269 legacy clients? It doesn't just Google refund 00:06:21.269 --> 00:06:23.689 policies. It looks at the handbook you uploaded 00:06:23.689 --> 00:06:26.339 and gives you the actual policy. You know, I 00:06:26.339 --> 00:06:28.180 have to admit something here. I still wrestle 00:06:28.180 --> 00:06:30.279 with prompt drift all the time. I'll be working 00:06:30.279 --> 00:06:32.860 with a generic AI, trying to get it to stick 00:06:32.860 --> 00:06:35.779 to a certain tone or follow some rules. And five 00:06:35.779 --> 00:06:38.480 prompts in, it just forgets. It goes right back 00:06:38.480 --> 00:06:41.439 to being a generic helper. It's exhausting trying 00:06:41.439 --> 00:06:43.839 to keep it on track. It's so frustrating. And 00:06:43.839 --> 00:06:45.899 that's the beauty of grounding. It can't drift. 00:06:46.000 --> 00:06:48.579 It has nowhere else to go. Max points to a perfect 00:06:48.579 --> 00:06:51.199 application for this. training packages walk 00:06:51.199 --> 00:06:52.860 us through that one okay so imagine you're an 00:06:52.860 --> 00:06:55.600 hr manager you upload the company handbook and 00:06:55.600 --> 00:06:58.199 maybe a few process videos you can ask notebook 00:06:58.199 --> 00:07:01.300 lm to generate a study guide for new hires based 00:07:01.300 --> 00:07:03.939 only on that content okay then ask it to make 00:07:03.939 --> 00:07:06.899 a quiz then an faq but in the company's voice 00:07:06.899 --> 00:07:09.699 you're taking one source of truth your documents 00:07:09.699 --> 00:07:11.819 and you're just refracting it into all these 00:07:11.819 --> 00:07:14.639 different useful formats so every new hire gets 00:07:14.639 --> 00:07:17.019 the exact same information from the exact same 00:07:17.019 --> 00:07:20.410 source 100 no more Game of telephone where the 00:07:20.410 --> 00:07:22.449 policy changes a little bit every time someone 00:07:22.449 --> 00:07:25.250 explains it. Okay. So what is the singular advantage 00:07:25.250 --> 00:07:28.410 of restricting the AI to only your uploaded documents? 00:07:28.730 --> 00:07:31.269 It eliminates hallucinations and guarantees answers 00:07:31.269 --> 00:07:34.089 align with company policy. Precision over creativity. 00:07:34.370 --> 00:07:37.629 But speaking of a different kind of engagement, 00:07:37.709 --> 00:07:40.889 I guess, let's talk about audio. Oh, the feature 00:07:40.889 --> 00:07:42.569 that kind of broke the internet for a minute 00:07:42.569 --> 00:07:45.170 there. The audio. overview i remember when that 00:07:45.170 --> 00:07:48.389 came out it was so uncanny two ai hosts sounding 00:07:48.389 --> 00:07:50.930 a bit like us just bantering about a chemistry 00:07:50.930 --> 00:07:54.009 textbook it was a cool novelty at first but in 00:07:54.009 --> 00:07:56.930 2026 max's point is that it's evolved into a 00:07:56.930 --> 00:08:00.449 serious learning tool the big upgrade is interactivity 00:08:00.449 --> 00:08:03.050 interactivity because before it was basically 00:08:03.050 --> 00:08:05.569 just a radio show you couldn't talk back to exactly 00:08:05.569 --> 00:08:08.370 you just listened to a 15 -minute generated podcast 00:08:08.370 --> 00:08:10.850 and hope it covered what you cared about now 00:08:10.850 --> 00:08:13.459 you can actually stop them OK, give me a real 00:08:13.459 --> 00:08:16.259 world scenario. How does that work? So picture 00:08:16.259 --> 00:08:19.220 this. You're on a walk listening to an audio 00:08:19.220 --> 00:08:21.860 overview of some dense legal contract you need 00:08:21.860 --> 00:08:24.779 to understand for a meeting. The AI host says 00:08:24.779 --> 00:08:28.100 something like section 4 .2 has a strict liability 00:08:28.100 --> 00:08:31.279 clause. You can literally tap your headphones 00:08:31.279 --> 00:08:35.110 and say, whoa. Pause. Explain strict liability 00:08:35.110 --> 00:08:38.029 like I'm five. And the AI just responds. The 00:08:38.029 --> 00:08:39.690 other AI host will jump in and say something 00:08:39.690 --> 00:08:42.370 like, good question. Basically, it means you're 00:08:42.370 --> 00:08:44.629 responsible even if it wasn't your fault. Yeah. 00:08:44.710 --> 00:08:46.669 And then it will cite the specific paragraph 00:08:46.669 --> 00:08:49.070 in the contract where it got that information. 00:08:49.470 --> 00:08:51.710 That changes everything. You're not just a passive 00:08:51.710 --> 00:08:54.090 listener. You're actively interrogating the source 00:08:54.090 --> 00:08:56.289 material. That's it. It turns your passive time, 00:08:56.389 --> 00:08:58.870 your commute, your walk into active study time. 00:08:58.970 --> 00:09:00.750 You can go down these little rabbit holes. You 00:09:00.750 --> 00:09:02.370 can even challenge it. Are you sure? That's what 00:09:02.370 --> 00:09:04.710 it says. And it will double check. It's like 00:09:04.710 --> 00:09:06.830 we're finally closing the loop between reading 00:09:06.830 --> 00:09:09.389 something and then discussing it. So is this 00:09:09.389 --> 00:09:11.190 just a neat tech demo or does it really change 00:09:11.190 --> 00:09:15.269 how we consume dense information? It makes academic 00:09:15.269 --> 00:09:17.690 and legal texts approachable for normal people. 00:09:17.830 --> 00:09:20.710 It democratizes density. I like that. Now let's 00:09:20.710 --> 00:09:22.649 get to the most, maybe the most complex part 00:09:22.649 --> 00:09:24.610 of this whole system, the Gemini integration. 00:09:25.110 --> 00:09:28.029 The power combo. Max's article treats this like 00:09:28.029 --> 00:09:30.960 it's the real missing link. But for a while, 00:09:31.039 --> 00:09:34.059 these felt like two separate tools. You had Notebook 00:09:34.059 --> 00:09:37.980 LM, the safe librarian, and Gemini, the creative 00:09:37.980 --> 00:09:41.100 artist. Why mash them together? Because most 00:09:41.100 --> 00:09:43.919 of the time, you need both. Yeah. But you need 00:09:43.919 --> 00:09:46.120 them in the right order. Let's take the workflow 00:09:46.120 --> 00:09:49.879 he describes for writing, say, a crisis PR email. 00:09:50.330 --> 00:09:52.350 Something went wrong and you have to talk to 00:09:52.350 --> 00:09:54.330 customers. High stakes. You can't get the facts 00:09:54.330 --> 00:09:56.090 wrong there. Not at all. If you just go to a 00:09:56.090 --> 00:09:57.889 creative AI and say, write an apology email, 00:09:58.090 --> 00:10:00.529 it might invent a refund policy you don't have. 00:10:00.730 --> 00:10:03.129 It could make promises you can't keep. It's dangerous. 00:10:03.429 --> 00:10:06.330 So you start in Notebook LM. You start there. 00:10:06.750 --> 00:10:08.669 You upload the incident report, the customer 00:10:08.669 --> 00:10:11.870 service logs, your legal guidelines. You use 00:10:11.870 --> 00:10:14.529 it to synthesize the actual facts. You get a 00:10:14.529 --> 00:10:17.210 bulleted list of exactly what happened and exactly 00:10:17.210 --> 00:10:20.169 what you were allowed to say or offer. The grounding 00:10:20.169 --> 00:10:23.350 phase. Right. Then, and this is the magic button, 00:10:23.389 --> 00:10:26.370 you click open in Gemini. It takes that safe 00:10:26.370 --> 00:10:29.090 fact -checked context and pushes it straight 00:10:29.090 --> 00:10:30.970 into the creative engine. And then you just tell 00:10:30.970 --> 00:10:33.610 Gemini, write an empathetic email based on these 00:10:33.610 --> 00:10:37.169 specific facts. Yes. And because Gemini is the 00:10:37.169 --> 00:10:40.529 artist, it handles the tone, the empathy, the 00:10:40.529 --> 00:10:43.990 professional language. But, and this is key, 00:10:44.190 --> 00:10:46.889 it is constrained by the facts that Notebook 00:10:46.889 --> 00:10:50.250 LM provided. It can't invent a new refund offer 00:10:50.250 --> 00:10:52.490 because the notebook never gave it one. So it 00:10:52.490 --> 00:10:54.990 filters out the potential for hallucination before 00:10:54.990 --> 00:10:57.529 the creative work even starts. Precisely. You 00:10:57.529 --> 00:10:59.389 get the safety of the librarian with the flair 00:10:59.389 --> 00:11:01.730 of the artist. It's a workflow that solves both 00:11:01.730 --> 00:11:04.149 the blank page problem and the lying chatbot 00:11:04.149 --> 00:11:07.179 problem. in one go. That distinction is so important. 00:11:07.360 --> 00:11:09.299 It's not about one tool doing everything. It's 00:11:09.299 --> 00:11:11.620 about a pipeline. A production line. So why is 00:11:11.620 --> 00:11:14.419 the integration with Gemini necessary if Notebook 00:11:14.419 --> 00:11:17.980 LM is so good on its own? Notebook LM analyzes 00:11:17.980 --> 00:11:21.360 facts. Gemini adds the creative flair and expansion. 00:11:21.679 --> 00:11:24.320 The analyst and the copywriter. Okay, we're going 00:11:24.320 --> 00:11:26.740 to take a very brief pause here. When we come 00:11:26.740 --> 00:11:29.580 back, we'll dig into the hidden truth. Why, if 00:11:29.580 --> 00:11:32.720 this tool is so powerful, Max claims almost nobody 00:11:32.720 --> 00:11:37.429 is using it yet. Stay with us. And we are back. 00:11:37.789 --> 00:11:39.769 We've been talking about this transformation 00:11:39.769 --> 00:11:43.110 of Notebook LM into a full on content factory. 00:11:43.269 --> 00:11:45.529 But there's this section in Max's article that's 00:11:45.529 --> 00:11:47.889 a bit of a puzzle. He calls it the hidden truth. 00:11:48.230 --> 00:11:51.370 He asks, if this is so great, why isn't it everywhere? 00:11:51.690 --> 00:11:53.789 Why isn't every knowledge worker using this all 00:11:53.789 --> 00:11:55.409 day? It's the billion dollar question, right? 00:11:55.470 --> 00:11:57.289 He points to three main barriers. The first one 00:11:57.289 --> 00:11:59.870 is just classic Google. They are terrible at 00:11:59.870 --> 00:12:01.330 marketing their own stuff. That's not exactly 00:12:01.330 --> 00:12:03.779 breaking news. No. They launched these incredible 00:12:03.779 --> 00:12:05.879 features like the interactive audio, and it just 00:12:05.879 --> 00:12:08.299 sort of appears in a menu one day. There's no 00:12:08.299 --> 00:12:11.000 big announcement, no tour. If you weren't actively 00:12:11.000 --> 00:12:13.179 looking for a generate slide deck button, you 00:12:13.179 --> 00:12:15.259 would never know it's there. A discovery problem. 00:12:15.500 --> 00:12:18.600 Okay, what's the second reason? Friction. And 00:12:18.600 --> 00:12:21.480 this one is really interesting. Notebook LM requires 00:12:21.480 --> 00:12:23.980 some upfront work. You can't just open it and 00:12:23.980 --> 00:12:26.220 type, tell me about the Roman Empire. You have 00:12:26.220 --> 00:12:28.600 to find your sources. You have to upload them. 00:12:28.620 --> 00:12:30.600 You have to give the notebook a name. In a world 00:12:30.600 --> 00:12:33.360 of instant gratification where ChatGPT gives 00:12:33.360 --> 00:12:35.860 you an answer in two seconds, that five minutes 00:12:35.860 --> 00:12:39.580 of setup can feel like... Forever. It does. But 00:12:39.580 --> 00:12:41.740 Max makes the argument that this friction is 00:12:41.740 --> 00:12:45.000 actually a feature, not a bug. How so? Because 00:12:45.000 --> 00:12:47.899 the prep phase is a filter. By forcing you to 00:12:47.899 --> 00:12:51.080 choose your sources to say, use this PDF, not 00:12:51.080 --> 00:12:53.379 that website, you're doing the quality control 00:12:53.379 --> 00:12:55.399 yourself. You're making sure the output isn't 00:12:55.399 --> 00:12:57.899 built on garbage. The friction is the price you 00:12:57.899 --> 00:13:00.419 pay for accuracy. That's a great reframing. The 00:13:00.419 --> 00:13:03.259 friction is a safety rail. And the third reason. 00:13:03.399 --> 00:13:05.639 People mistake it for a search bar. They think 00:13:05.639 --> 00:13:07.860 it's just another way to ask, you know, who won 00:13:07.860 --> 00:13:10.480 the World Series in 1998. But it's not a search 00:13:10.480 --> 00:13:12.279 engine. It's a production studio. If you treat 00:13:12.279 --> 00:13:14.279 it like Google search, you'll be let down. If 00:13:14.279 --> 00:13:16.220 you treat it like a research assistant, you'll 00:13:16.220 --> 00:13:18.259 be blown away. This all brings us to the core 00:13:18.259 --> 00:13:20.860 idea Max leaves us with. He talks about synthesis. 00:13:21.679 --> 00:13:24.500 And I want to reflect on that for a moment. In 00:13:24.500 --> 00:13:27.519 2026, our problem isn't a shortage of information. 00:13:27.679 --> 00:13:31.129 We are drowning in it. Right. The problem is 00:13:31.129 --> 00:13:33.289 not creating new text. We have more than enough 00:13:33.289 --> 00:13:36.289 text. The problem is compression. Compression. 00:13:36.529 --> 00:13:38.750 Yeah. Taking the insane amount of data we already 00:13:38.750 --> 00:13:41.610 have, all the PDFs, the videos, the Slack threads, 00:13:41.769 --> 00:13:44.250 and compressing it into something useful, something 00:13:44.250 --> 00:13:46.990 actionable. That's the big idea here. Notebook 00:13:46.990 --> 00:13:49.190 LM isn't about adding to the noise. It's about 00:13:49.190 --> 00:13:51.980 making sense of the noise you already own. That 00:13:51.980 --> 00:13:55.039 is a really profound distinction. It moves the 00:13:55.039 --> 00:13:57.740 value away from generation, just making more 00:13:57.740 --> 00:14:00.220 words and toward understanding. Exactly. And 00:14:00.220 --> 00:14:01.500 that's why it's a tool that's going to stick 00:14:01.500 --> 00:14:03.480 around while a lot of the other flashy stuff 00:14:03.480 --> 00:14:06.240 fades away. So as we close out this deep dive, 00:14:06.379 --> 00:14:08.220 I want to leave you with a choice that Max poses. 00:14:08.480 --> 00:14:11.379 Do you need speed or do you need depth? If you 00:14:11.379 --> 00:14:13.419 need a quick answer to a trivia question, use 00:14:13.419 --> 00:14:15.879 ChatGPT. It's faster. But if you're building 00:14:15.879 --> 00:14:18.740 a training course or analyzing a competitor or 00:14:18.740 --> 00:14:21.690 trying to understand a complex document, You 00:14:21.690 --> 00:14:24.429 need depth. You need that grounding. So my challenge 00:14:24.429 --> 00:14:27.610 to you, the listener, is to try the content factory 00:14:27.610 --> 00:14:30.769 approach just one time this week. Don't just 00:14:30.769 --> 00:14:33.990 ask a question. Go upload a brand handbook or 00:14:33.990 --> 00:14:37.070 a project archive and then ask for a blog post 00:14:37.070 --> 00:14:40.009 outline or a slide deck just to feel the difference 00:14:40.009 --> 00:14:42.509 between grounded and the open web. Just give 00:14:42.509 --> 00:14:44.450 it a shot. I mean, even if you don't end up using 00:14:44.450 --> 00:14:47.009 the slides, just seeing the structure it creates 00:14:47.009 --> 00:14:49.399 is worth the three minutes. It's been a fascinating 00:14:49.399 --> 00:14:51.419 look at how our tools are starting to shape our 00:14:51.419 --> 00:14:54.159 thinking. From simple archives to interactive 00:14:54.159 --> 00:14:56.340 studios, I'm really eager to see what you all 00:14:56.340 --> 00:14:58.379 build with it. See you in the studio. Until next 00:14:58.379 --> 00:14:59.720 time, keep diving deep.