WEBVTT 00:00:00.000 --> 00:00:02.240 You know, while so much of the AI conversation 00:00:02.240 --> 00:00:06.620 seems stuck on which chatbot is the best, Google's 00:00:06.620 --> 00:00:08.900 been doing something else. Something maybe quieter. 00:00:09.259 --> 00:00:12.599 Exactly. Quietly building this whole arsenal 00:00:12.599 --> 00:00:16.280 of really powerful AI tools. They're out there, 00:00:16.300 --> 00:00:19.300 openly available, and honestly, they can give 00:00:19.300 --> 00:00:21.660 you what feels like an almost unfair advantage. 00:00:22.160 --> 00:00:25.140 So today, in this deep dive, we're going to pull 00:00:25.140 --> 00:00:27.379 back the curtain on these tools. They're potent, 00:00:27.500 --> 00:00:29.000 even if they're not always in the headlines. 00:00:29.800 --> 00:00:33.240 Slight pause. Welcome to the Deep Dive. Glad 00:00:33.240 --> 00:00:35.060 to be here. Today, we're digging into some really 00:00:35.060 --> 00:00:38.140 fascinating sources, looking at Google's practical 00:00:38.140 --> 00:00:40.640 AI stuff. We've got insights from conversations 00:00:40.640 --> 00:00:43.380 with a key exec at Google Labs. Right. And the 00:00:43.380 --> 00:00:45.920 mission today. Simple, really. We want to explore 00:00:45.920 --> 00:00:48.759 six specific Google AI tools. You might not be 00:00:48.759 --> 00:00:51.340 using them, but maybe you should be. Game changers, 00:00:51.340 --> 00:00:53.500 potentially. Definitely. We'll look at each one, 00:00:53.600 --> 00:00:55.179 and then we'll talk about what we're calling 00:00:55.179 --> 00:00:58.159 the Voltron strategy. Ah, combining the lions. 00:00:58.940 --> 00:01:01.719 Exactly, how to combine them into this, well, 00:01:01.899 --> 00:01:05.200 almost unstoppable system. Get ready, because 00:01:05.200 --> 00:01:07.640 this might just change how you think about what 00:01:07.640 --> 00:01:10.260 AI can do for you. Let's dive in. Okay, first 00:01:10.260 --> 00:01:14.019 up, Gemini Pro. For a long time, it felt like 00:01:14.019 --> 00:01:16.859 the main question was just, is Gemini as good 00:01:16.859 --> 00:01:19.730 as ChatGPT? Yeah, that comparison dominated. 00:01:19.989 --> 00:01:21.590 But that's kind of missing the bigger picture, 00:01:21.730 --> 00:01:24.250 isn't it? It's like comparing a Starship's, I 00:01:24.250 --> 00:01:27.250 don't know, phasers to its warp drive. Right. 00:01:27.329 --> 00:01:29.329 Different tools for fundamentally different things. 00:01:29.760 --> 00:01:31.780 And with Gemini Pro, especially the paid version, 00:01:32.000 --> 00:01:34.760 the real magic isn't just the chat. It's the 00:01:34.760 --> 00:01:37.680 integration. Exactly. It's deeply natively integrated 00:01:37.680 --> 00:01:40.319 into the whole Google ecosystem. It's less like 00:01:40.319 --> 00:01:42.780 asking a question to a smart stranger. Then more 00:01:42.780 --> 00:01:45.519 like? More like having Iron Man's J -A -R -V 00:01:45.519 --> 00:01:48.120 -I -S. Remember him? The AI co -pilot. Yeah. 00:01:48.219 --> 00:01:50.459 An AI that's designed to plug into your whole 00:01:50.459 --> 00:01:52.920 digital life, your emails, your docs, your calendar, 00:01:52.980 --> 00:01:54.799 and actually work with you. It understands your 00:01:54.799 --> 00:01:57.640 context. Okay, so that deep integration unlocks 00:01:57.640 --> 00:02:00.200 some neat tricks, like scheduled actions? Oh 00:02:00.200 --> 00:02:02.000 yeah, the automated intelligence briefing. This 00:02:02.000 --> 00:02:04.739 is cool. Imagine waking up, checking your email, 00:02:04.920 --> 00:02:07.739 and there's a concise, bulleted summary waiting 00:02:07.739 --> 00:02:11.379 for you. All about your competitors, latest news. 00:02:12.060 --> 00:02:15.659 product launches, funding rounds, maybe big marketing 00:02:15.659 --> 00:02:18.000 pushes. All generated automatically overnight. 00:02:18.219 --> 00:02:21.099 It's like having Q from James Bond working the 00:02:21.099 --> 00:02:23.139 night shift just for you. Think about market 00:02:23.139 --> 00:02:25.879 intelligence firms. They charge thousands a month 00:02:25.879 --> 00:02:27.979 for that kind of service. And here it's just 00:02:27.979 --> 00:02:31.949 a feature built in. It means. You start the day 00:02:31.949 --> 00:02:35.349 already ahead. That's powerful. And it goes deeper. 00:02:35.610 --> 00:02:37.409 It does. Think about the on -demand personal 00:02:37.409 --> 00:02:40.250 tutor capability. This is maybe the closest thing 00:02:40.250 --> 00:02:42.509 we have right now to that matrix -style instant 00:02:42.509 --> 00:02:45.509 learning. Like downloading Kung Fu. Almost. The 00:02:45.509 --> 00:02:47.990 ability to genuinely get up to speed on a whole 00:02:47.990 --> 00:02:50.919 new industry in, say, a week. instead of a year 00:02:50.919 --> 00:02:53.419 that's a massive edge so how does it work you 00:02:53.419 --> 00:02:55.840 can ask gemini pro something like uh teach me 00:02:55.840 --> 00:02:57.759 the basics of quantum computing and make a quiz 00:02:57.759 --> 00:02:59.800 and it doesn't just give you text yeah it builds 00:02:59.800 --> 00:03:02.280 this interactive learning module you get clear 00:03:02.280 --> 00:03:04.740 explanations it embeds relevant youtube videos 00:03:04.740 --> 00:03:07.120 right there for visual learning and then it creates 00:03:07.120 --> 00:03:10.000 an actual interactive quiz with hints if you 00:03:10.000 --> 00:03:11.680 need them that really could slash the learning 00:03:11.680 --> 00:03:14.620 curve okay and then there's the big one true 00:03:14.620 --> 00:03:19.800 personal context this feels like the end game 00:03:19.800 --> 00:03:22.319 for personal AI. It's certainly heading that 00:03:22.319 --> 00:03:25.259 way. This is where if you give permission, Gemini 00:03:25.259 --> 00:03:27.639 starts integrating with your personal Google 00:03:27.639 --> 00:03:30.400 stuff, your photos, your search history, YouTube 00:03:30.400 --> 00:03:32.400 watches, even Gmail. And when it has all that 00:03:32.400 --> 00:03:35.379 context, what happens? Well, it starts to feel 00:03:35.379 --> 00:03:38.060 less like just an assistant and more like, remember 00:03:38.060 --> 00:03:40.979 the Oracle from the Matrix or maybe a Benet Gesserit 00:03:40.979 --> 00:03:43.840 from Dune? Meaning it gives eerily relevant advice. 00:03:44.020 --> 00:03:46.120 Exactly. Yeah. Its insights can be strikingly 00:03:46.120 --> 00:03:48.389 prescient because they're based on you. So a 00:03:48.389 --> 00:03:50.629 generic AI might plan a family trip to France. 00:03:50.849 --> 00:03:54.409 Fine. But Gemini Pro. Knowing your context. It 00:03:54.409 --> 00:03:56.250 sees from Google Photos you guys love hiking. 00:03:56.430 --> 00:03:58.530 It notices from YouTube you're into cooking shows. 00:03:59.009 --> 00:04:01.830 So it designs a totally custom trip. Like skipping 00:04:01.830 --> 00:04:05.289 Paris. Maybe. It might suggest a food -themed 00:04:05.289 --> 00:04:07.370 road trip through Provence and the French Alps. 00:04:07.650 --> 00:04:10.870 A guided hike in the Gorge du Verdun. A cooking 00:04:10.870 --> 00:04:14.430 class in Lyon. Maybe even foraging near Anatelassie. 00:04:14.889 --> 00:04:17.230 It's tailored. That level of personalization 00:04:17.230 --> 00:04:19.470 is. Yeah. Yeah. That's beyond what we usually 00:04:19.470 --> 00:04:21.870 see. And you can ramp this up further. Definitely. 00:04:22.189 --> 00:04:24.430 Combine those scheduled actions with this personal 00:04:24.430 --> 00:04:26.589 context. Now you've got what we call the proactive 00:04:26.589 --> 00:04:29.050 opportunity finder. Okay. Like an automated bat 00:04:29.050 --> 00:04:31.870 computer. Kind of. Imagine you're an investor. 00:04:32.490 --> 00:04:35.110 You set a scheduled action, scan tech news and 00:04:35.110 --> 00:04:37.470 financial reports daily for companies matching 00:04:37.470 --> 00:04:40.329 my investment thesis in, say, renewable energy 00:04:40.329 --> 00:04:42.629 startups in the Pacific Northwest. And it would. 00:04:42.769 --> 00:04:44.910 It identifies new companies or funding rounds 00:04:44.910 --> 00:04:47.629 matching your criteria, maybe even cross -references 00:04:47.629 --> 00:04:50.149 with Google Maps for location analysis, and then 00:04:50.149 --> 00:04:52.290 emails you a summary of the top three opportunities 00:04:52.290 --> 00:04:55.170 each morning. Wow. Okay, so probing question 00:04:55.170 --> 00:04:58.660 time. What makes Gemini Pro's integration so 00:04:58.660 --> 00:05:01.439 impactful beyond just being a chatbot? Its deep 00:05:01.439 --> 00:05:04.319 ecosystem access provides personalized, proactive 00:05:04.319 --> 00:05:06.680 insights. Got it. Personalized and proactive. 00:05:07.040 --> 00:05:10.600 Okay, let's shift gears. AI video. Right. AI 00:05:10.600 --> 00:05:13.480 video generation that, frankly, doesn't look 00:05:13.480 --> 00:05:16.579 like AI anymore. VO3. We've seen AI video before, 00:05:16.699 --> 00:05:19.660 and it often felt a bit off. Young Candy Valley, 00:05:19.899 --> 00:05:21.920 yeah. For a couple of years, it was like a cool 00:05:21.920 --> 00:05:24.399 tech demo, but you could always tell it was machine 00:05:24.399 --> 00:05:27.079 -made. But that's changing. That era is basically 00:05:27.079 --> 00:05:31.379 over. Google's latest, VO3, it's a quantum leap. 00:05:31.620 --> 00:05:34.819 Think of the jump from, like, 90s CGI in the 00:05:34.819 --> 00:05:38.339 lawnmower to the photorealistic stuff you see 00:05:38.339 --> 00:05:42.319 in a modern Marvel movie. VO3 crosses that believability 00:05:42.319 --> 00:05:45.339 threshold. The output isn't just interesting, 00:05:45.480 --> 00:05:47.980 it's genuinely usable for professional stuff. 00:05:48.240 --> 00:05:50.199 How good are we talking? We're talking 8 -second 00:05:50.199 --> 00:05:53.779 clips, 720p resolution, but with stunning cinematic 00:05:53.779 --> 00:05:56.860 quality. Seriously, it rivals good stock footage. 00:05:57.040 --> 00:05:59.899 And it understands things like lighting, physics. 00:06:00.000 --> 00:06:02.139 Yeah, it has a real grasp of lighting, composition, 00:06:02.459 --> 00:06:04.980 even basic physics. A bouncing ball looks right, 00:06:05.060 --> 00:06:07.180 a flowing river, a car speeding up the motion 00:06:07.180 --> 00:06:09.100 feels natural. Impressive. What about sound? 00:06:09.420 --> 00:06:11.839 It can even generate synchronized audio that 00:06:11.839 --> 00:06:14.399 actually matches the action on screen, another 00:06:14.399 --> 00:06:16.639 layer of realism. Okay, so if the quality is 00:06:16.639 --> 00:06:19.879 there, what's the implication? Huge business 00:06:19.879 --> 00:06:22.790 opportunity, honestly. It feels like the end 00:06:22.790 --> 00:06:25.449 of the gatekeepers for video. Meaning? For decades, 00:06:25.689 --> 00:06:27.990 high quality video needed big budgets, fancy 00:06:27.990 --> 00:06:31.350 gear, specialized crews. VO3 disrupts that. It 00:06:31.350 --> 00:06:34.009 completely democratizes video creation. So for 00:06:34.009 --> 00:06:37.410 a small business. Or a solo entrepreneur. It's 00:06:37.410 --> 00:06:40.189 massive. You can suddenly create an endless stream 00:06:40.189 --> 00:06:43.790 of unique, eye -catching video ads, social posts, 00:06:44.110 --> 00:06:46.730 stuff that used to cost thousands. You can make 00:06:46.730 --> 00:06:48.970 pro -looking product demos without hiring anyone, 00:06:49.170 --> 00:06:51.110 maybe even without the physical product handy. 00:06:51.350 --> 00:06:53.970 Whoa, imagine generating infinite video variations 00:06:53.970 --> 00:06:56.350 for A -B testing that used to be only for companies 00:06:56.350 --> 00:06:58.730 with multi -million dollar ad campaigns. Exactly. 00:06:58.769 --> 00:07:00.829 That's the A -B testing on steroids workflow. 00:07:01.189 --> 00:07:03.449 That's a game changer. Tell me more. Instead 00:07:03.449 --> 00:07:05.360 of blowing your budget on one experiment, You 00:07:05.360 --> 00:07:08.600 generate maybe 20 versions in an afternoon. Red 00:07:08.600 --> 00:07:11.819 car versus blue car. Sunrise versus sunset. Male 00:07:11.819 --> 00:07:13.800 actor versus female. And then you test them all. 00:07:14.019 --> 00:07:16.439 Yep. Test them all with minimal ad spend and 00:07:16.439 --> 00:07:18.339 just let the market data tell you what works 00:07:18.339 --> 00:07:21.920 best. Okay. So question. How does VO3 fundamentally 00:07:21.920 --> 00:07:24.060 change video marketing for small businesses? 00:07:24.500 --> 00:07:27.920 It democratizes pro video, enabling infinite 00:07:27.920 --> 00:07:30.480 affordable testing. Democratizes and enables 00:07:30.480 --> 00:07:34.230 testing. Nice. Okay, so if Veo is the camera... 00:07:34.230 --> 00:07:36.529 Then Flow is the studio. Flow. Tell me about 00:07:36.529 --> 00:07:39.730 Flow. If Gemini and Veo are like this incredibly 00:07:39.730 --> 00:07:43.129 advanced camera system, Flow is Google's cloud 00:07:43.129 --> 00:07:46.949 -based AI filmmaking studio. Think script development, 00:07:47.230 --> 00:07:49.870 character creation, scene generation, editing, 00:07:50.089 --> 00:07:53.449 all integrated. So it takes AI video generation 00:07:53.449 --> 00:07:56.069 and turns it into a proper creative suite. Exactly. 00:07:56.069 --> 00:07:58.509 It's Google's dedicated platform for that. And 00:07:58.509 --> 00:08:00.509 it has features specifically for storytellers. 00:08:00.550 --> 00:08:02.709 Like what? What's in the director's toolkit? 00:08:02.949 --> 00:08:05.769 Okay, first, it tackles a big AI video problem. 00:08:06.029 --> 00:08:08.490 Consistent characters. Flow can actually maintain 00:08:08.490 --> 00:08:10.370 the same character, let's call him Detective 00:08:10.370 --> 00:08:12.689 Harding, across multiple different scenes. So 00:08:12.689 --> 00:08:14.310 Detective Harding can go from the crime scene 00:08:14.310 --> 00:08:16.430 to the car chase, and it's still recognizably 00:08:16.430 --> 00:08:18.870 him. Precisely. Builds narrative coherence. Then 00:08:18.870 --> 00:08:21.629 there's scene extension. This is cool. You can 00:08:21.629 --> 00:08:23.430 treat your generated scene like a storyboard 00:08:23.430 --> 00:08:26.149 and just add stuff. Like with a prompt. Yeah. 00:08:26.540 --> 00:08:29.860 Like, make a huge menacing vulture come flying 00:08:29.860 --> 00:08:33.200 in from the sky. It layers it in. Like, plussing 00:08:33.200 --> 00:08:36.679 an idea at Pixar. Okay. And multi -video generation. 00:08:36.799 --> 00:08:40.240 It can generate, say, four different takes on 00:08:40.240 --> 00:08:43.259 the same scene simultaneously. Great for exploring 00:08:43.259 --> 00:08:45.279 options or breaking creative blocks. All right. 00:08:45.299 --> 00:08:47.539 Interesting tools. But what's the secret weapon? 00:08:47.860 --> 00:08:51.559 Ah, Flow TV. This is brilliant. What is it? A 00:08:51.559 --> 00:08:54.200 gallery. It's a public gallery showcasing the 00:08:54.200 --> 00:08:57.399 best stuff people are making with flow. But it's 00:08:57.399 --> 00:09:00.259 way more than just inspiration. How so? It's 00:09:00.259 --> 00:09:02.779 like an open source, reverse -engineerable Netflix. 00:09:03.299 --> 00:09:05.500 For almost every single video you see there, 00:09:05.559 --> 00:09:07.620 you can click and see the exact prompt the creator 00:09:07.620 --> 00:09:10.769 used. No way. Seriously? Seriously. It's like 00:09:10.769 --> 00:09:12.990 looking over the shoulder of thousands of creative 00:09:12.990 --> 00:09:15.210 people, learning their tricks, remixing their 00:09:15.210 --> 00:09:16.990 ideas. It's an incredible learning resource. 00:09:17.289 --> 00:09:19.309 That is a game changer. Okay, so how do you use 00:09:19.309 --> 00:09:21.429 this at a pro level? Right, the pro level upgrade, 00:09:21.590 --> 00:09:24.490 the brand story generator. This uses Flow's ability 00:09:24.490 --> 00:09:27.210 to handle longer narratives. For businesses? 00:09:27.230 --> 00:09:29.889 Yeah, creating a really high value asset. Step 00:09:29.889 --> 00:09:33.990 one, feed a powerful AI. like Gemini, your company's 00:09:33.990 --> 00:09:36.889 About Us page, maybe some customer testimonials, 00:09:36.970 --> 00:09:39.909 prompted to write a compelling story arc, maybe 00:09:39.909 --> 00:09:42.370 three minutes, five scenes about the company's 00:09:42.370 --> 00:09:43.830 founding or mission. Okay, you've got a script. 00:09:44.070 --> 00:09:46.889 Step two, generate each of those five scenes 00:09:46.889 --> 00:09:49.509 in Flow. Make sure you use a consistent animated 00:09:49.509 --> 00:09:51.590 brand mascot throughout. All right, consistent 00:09:51.590 --> 00:09:55.009 character. Step three, use Flow's tools to stitch 00:09:55.009 --> 00:09:57.789 those scenes together. Add a professional voiceover, 00:09:57.870 --> 00:10:01.009 maybe AI -generated. maybe human, add royalty 00:10:01.009 --> 00:10:03.610 -free music score. And the result? A high -quality 00:10:03.610 --> 00:10:06.370 animated brand story video done in maybe a day, 00:10:06.389 --> 00:10:08.610 not a month, for a tiny fraction of what that 00:10:08.610 --> 00:10:11.769 usually costs. Wow. Okay, probing question. What's 00:10:11.769 --> 00:10:13.950 the biggest creative breakthrough Flow offers 00:10:13.950 --> 00:10:16.070 storytellers? Consistent characters and direct 00:10:16.070 --> 00:10:18.909 access to successful creative prompts. Consistency 00:10:18.909 --> 00:10:21.370 in learning. Makes sense. Let's move on. Notebook 00:10:21.370 --> 00:10:24.450 LM. Ah, Notebook LM. The research revolution 00:10:24.450 --> 00:10:27.389 of this one. This might be Google's most underrated 00:10:27.389 --> 00:10:30.149 tool. Seriously powerful potential here. Underrated. 00:10:30.289 --> 00:10:33.129 How so? Well, most AI tools feel like a smart 00:10:33.129 --> 00:10:35.870 person you talk to, right? Yeah, like a chatbot. 00:10:36.110 --> 00:10:38.330 Notebook LM feels more like a smart room you 00:10:38.330 --> 00:10:40.950 work in. It's a platform designed to help you 00:10:40.950 --> 00:10:43.169 think better with your own information. Oh, okay. 00:10:43.409 --> 00:10:45.470 Intriguing. What makes it so special? Lay it 00:10:45.470 --> 00:10:47.269 out for me. The core idea is straightforward. 00:10:47.789 --> 00:10:51.039 You create a notebook. for a project. Then you 00:10:51.039 --> 00:10:53.580 upload your source materials, could be PDFs, 00:10:53.620 --> 00:10:57.659 web links you've saved, project docs, even transcripts 00:10:57.659 --> 00:10:59.720 from YouTube videos. So all your research in 00:10:59.720 --> 00:11:02.679 one place. Exactly. And then Notebook LM basically 00:11:02.679 --> 00:11:05.200 becomes an expert on your stuff. It's like having 00:11:05.200 --> 00:11:07.419 research librarians who have memorized your entire 00:11:07.419 --> 00:11:09.980 personal library and are ready to discuss it. 00:11:10.120 --> 00:11:12.159 Okay, that sounds useful. What are the killer 00:11:12.159 --> 00:11:16.139 features? I'd highlight five. First, Cited answers. 00:11:16.279 --> 00:11:19.279 This is huge for trust. The AI trust crisis, 00:11:19.460 --> 00:11:22.840 right? Hallucinations. Exactly. Notebook LM combats 00:11:22.840 --> 00:11:25.399 that. Every answer it gives based on your sources 00:11:25.399 --> 00:11:27.879 includes a direct citation. Click it, and it 00:11:27.879 --> 00:11:29.740 takes you to the exact passage in the original 00:11:29.740 --> 00:11:31.500 document. It's like getting a receipt for the 00:11:31.500 --> 00:11:33.879 information. Okay, that's number one. What else? 00:11:34.220 --> 00:11:37.639 Second, automatic source discovery. If the documents 00:11:37.639 --> 00:11:39.519 you uploaded can't fully answer your question, 00:11:39.870 --> 00:11:42.570 It doesn't just stop. It proactively suggests 00:11:42.570 --> 00:11:45.250 searching the web for new, relevant articles 00:11:45.250 --> 00:11:47.889 it thinks might help. Smart. Number three. Mind 00:11:47.889 --> 00:11:50.509 mapping. It analyzes all your uploaded sources 00:11:50.509 --> 00:11:53.309 and automatically generates a mind map showing 00:11:53.309 --> 00:11:56.570 connections, themes, links between concepts that 00:11:56.570 --> 00:11:58.750 you might have missed. Helps you literally connect 00:11:58.750 --> 00:12:02.590 the dots. Nice visualization. Four. Personalized 00:12:02.590 --> 00:12:06.009 podcasts. This is a productivity game changer. 00:12:06.149 --> 00:12:08.769 Wait, it makes podcasts? Yeah. You can ask it 00:12:08.769 --> 00:12:10.929 to create an audio podcast -style discussion, 00:12:11.149 --> 00:12:13.950 summarizing key findings from your sources. Imagine 00:12:13.950 --> 00:12:17.669 turning a dense 100 -page report into a 20 -minute 00:12:17.669 --> 00:12:19.769 audio briefing you can listen to on your commute. 00:12:19.950 --> 00:12:22.750 Whoa, okay. That's very cool. And number five. 00:12:23.110 --> 00:12:25.090 Collaborative knowledge building. You can share 00:12:25.090 --> 00:12:27.769 notebooks, so a whole team can dump their research, 00:12:27.990 --> 00:12:30.789 notes, and documents into one shared space. It 00:12:30.789 --> 00:12:32.950 becomes like a collective war room for understanding 00:12:32.950 --> 00:12:35.629 a complex topic together. Okay, that's a solid 00:12:35.629 --> 00:12:37.529 set of features. How about a pro tip? Try the 00:12:37.529 --> 00:12:39.950 Socratic Debate Club upgrade. This is for really 00:12:39.950 --> 00:12:42.169 high -level thinking. How does that work? Get 00:12:42.169 --> 00:12:46.029 your team, upload two conflicting reports, maybe 00:12:46.029 --> 00:12:48.289 different strategy proposals or market analyses, 00:12:48.549 --> 00:12:52.059 into a shared notebook. Then... prompt the AI 00:12:52.059 --> 00:12:54.860 to act as a neutral moderator. Ask it to argue. 00:12:55.179 --> 00:12:58.039 Sort of. Ask things like, you're a neutral moderator. 00:12:58.659 --> 00:13:01.659 Based on these two docs, what's the strongest 00:13:01.659 --> 00:13:04.600 argument from doc A that refutes the main point 00:13:04.600 --> 00:13:08.370 of doc B? Now do the reverse. Ah, so it identifies 00:13:08.370 --> 00:13:11.129 the core points of conflict. Exactly. Then follow 00:13:11.129 --> 00:13:13.970 up. Okay, now synthesize a third hybrid perspective 00:13:13.970 --> 00:13:16.509 that incorporates the most valid points from 00:13:16.509 --> 00:13:19.009 both sides. It becomes an engine for critical 00:13:19.009 --> 00:13:21.529 thinking, pressure testing ideas. That sounds 00:13:21.529 --> 00:13:23.690 really useful for strategy sessions. You know, 00:13:23.710 --> 00:13:25.870 even with these amazing tools, I still wrestle 00:13:25.870 --> 00:13:27.710 sometimes with getting my initial prompts just 00:13:27.710 --> 00:13:29.970 right, making sure they're precise enough to 00:13:29.970 --> 00:13:32.230 unlock what the AI can really do. It's a skill 00:13:32.230 --> 00:13:34.850 in itself. Absolutely. Prompting is key. So quick 00:13:34.850 --> 00:13:36.950 question on Notebook LM. How does it transform 00:13:36.950 --> 00:13:38.769 the way we interact with our own information? 00:13:39.210 --> 00:13:42.009 It makes your personal data searchable, analyzable, 00:13:42.129 --> 00:13:44.850 and synthesizable instantly. Searchable, analyzable, 00:13:44.929 --> 00:13:47.509 synthesizable. Got it. Okay, next up, Project 00:13:47.509 --> 00:13:51.509 Mariner. Your army of AI interns. This one sounds 00:13:51.509 --> 00:13:54.289 different, more experimental. It is Google's 00:13:54.289 --> 00:13:56.350 most experimental public lab tool right now, 00:13:56.450 --> 00:13:59.659 yeah. But the potential is huge. maybe the most 00:13:59.659 --> 00:14:02.419 transformative for actual day -to -day business 00:14:02.419 --> 00:14:05.580 tasks. So what is Project Mariner? It's a system 00:14:05.580 --> 00:14:08.279 that lets you deploy autonomous AI agents. These 00:14:08.279 --> 00:14:10.759 agents can navigate the web and complete complex 00:14:10.759 --> 00:14:14.220 multi -step tasks for you. Like actual tasks 00:14:14.220 --> 00:14:17.139 on the internet? Yeah. Think of it like being 00:14:17.139 --> 00:14:19.899 able to instantly spin up an army of hyper -fast, 00:14:20.000 --> 00:14:22.320 super -efficient virtual interns whenever you 00:14:22.320 --> 00:14:24.500 need them. Okay, autonomous agents sounds a little 00:14:24.500 --> 00:14:27.960 scary. Like Skynet. No, and that's the key difference. 00:14:28.159 --> 00:14:30.740 What makes Mariner feel practical and, frankly, 00:14:30.820 --> 00:14:33.139 usable is its human -in -the -loop approach. 00:14:33.519 --> 00:14:36.200 Meaning it's not totally autonomous. Exactly. 00:14:36.360 --> 00:14:38.620 It's not fire -and -forget. It's designed for 00:14:38.620 --> 00:14:41.360 collaboration, with human oversight built right 00:14:41.360 --> 00:14:43.659 in. Think of it like a really good self -driving 00:14:43.659 --> 00:14:46.080 car. Great on the highway, but it knows when 00:14:46.080 --> 00:14:48.419 to flag the human driver if things get confusing, 00:14:48.559 --> 00:14:50.919 like unexpected construction. So if the agent 00:14:50.919 --> 00:14:52.740 gets stuck... It doesn't just fail silently. 00:14:53.289 --> 00:14:56.929 If it hits, say, a KPTCHA it can't solve or it 00:14:56.929 --> 00:14:59.509 needs logging credentials for a site, it pauses, 00:14:59.629 --> 00:15:03.210 raises a little flag, and asks you, the human, 00:15:03.370 --> 00:15:06.649 for help or guidance. Ah, that safety net makes 00:15:06.649 --> 00:15:09.409 it actually practical for real tasks? Precisely. 00:15:09.409 --> 00:15:11.690 It builds in reliability. Okay, give me a real 00:15:11.690 --> 00:15:14.450 -world example. How would someone use this? Let's 00:15:14.450 --> 00:15:16.710 say you want an automated lead researcher. You 00:15:16.710 --> 00:15:18.909 give a mariner agent a mission, something like, 00:15:18.950 --> 00:15:22.009 go to climate base. Find two or three recent 00:15:22.009 --> 00:15:24.049 job postings for software engineers in the San 00:15:24.049 --> 00:15:26.769 Francisco Bay Area. Summarize the key requirements 00:15:26.769 --> 00:15:29.110 and how to apply for each. And you can watch 00:15:29.110 --> 00:15:31.490 it work? Yep. There's often a real -time feed. 00:15:31.590 --> 00:15:33.950 You see it open a browser window, maybe access 00:15:33.950 --> 00:15:35.950 your Google Drive if you told it where your resume 00:15:35.950 --> 00:15:38.850 template is. Navigate to ClimateBase, perform 00:15:38.850 --> 00:15:41.210 the search, scrape the relevant info from the 00:15:41.210 --> 00:15:43.190 job listings, and then compile it all into a 00:15:43.190 --> 00:15:45.789 neat, structured list or document for you. So 00:15:45.789 --> 00:15:48.789 it delegates those kinds of high -value but tedious 00:15:48.789 --> 00:15:51.740 research tasks? Perfectly. Now, the real power 00:15:51.740 --> 00:15:54.279 move, the swarm intelligence workflow, comes 00:15:54.279 --> 00:15:55.799 when you start chaining these tools together. 00:15:56.120 --> 00:15:58.399 Combining Mariner with other things. Exactly. 00:15:58.600 --> 00:16:01.860 Imagine this. Step one, deploy a small swarm, 00:16:02.080 --> 00:16:05.120 maybe five Mariner interns, to scrape the top 00:16:05.120 --> 00:16:08.259 50 customer reviews for a competitor's new product 00:16:08.259 --> 00:16:11.259 from three different e -commerce sites. That's 00:16:11.259 --> 00:16:14.120 your raw data collection. Okay. Lots of unstructured 00:16:14.120 --> 00:16:17.460 text. Step two, that raw data automatically gets 00:16:17.460 --> 00:16:21.429 sent over to Notebook LM, its job. Analyze all 00:16:21.429 --> 00:16:24.470 those reviews and spit out a concise, cited summary 00:16:24.470 --> 00:16:26.649 of the top five most common customer complaints. 00:16:26.889 --> 00:16:30.129 Ah, market intelligence. Step three. That summary 00:16:30.129 --> 00:16:32.190 of complaints automatically gets sent to VO3 00:16:32.190 --> 00:16:35.409 with a prompt like. Create a 15 -second video 00:16:35.409 --> 00:16:37.750 ad that directly addresses these five complaints 00:16:37.750 --> 00:16:40.389 and positions our product as the solution. Whoa. 00:16:40.570 --> 00:16:43.210 So it goes from scraping reviews to a finished 00:16:43.210 --> 00:16:45.230 video ad. Pretty much automatically built an 00:16:45.230 --> 00:16:46.950 end -to -end pipeline. Right. Market intelligence 00:16:46.950 --> 00:16:49.289 feeding directly into content creation. Minimal 00:16:49.289 --> 00:16:51.590 human intervention needed. That's a system. Okay, 00:16:51.649 --> 00:16:54.210 probing question. What makes Project Mariner's 00:16:54.210 --> 00:16:56.450 human -in -the -loop so crucial for real -world 00:16:56.450 --> 00:16:59.350 tasks? It ensures reliability and safety by asking 00:16:59.350 --> 00:17:02.090 for human help when needed. Reliability and safety 00:17:02.090 --> 00:17:06.230 through collaboration. Okay. Next tool. Stitch. 00:17:06.589 --> 00:17:10.490 Stitch. Your personal AI -powered UI UX designer. 00:17:10.769 --> 00:17:12.650 This tackles another common pain point, right? 00:17:12.769 --> 00:17:16.609 Design. Oh, a huge one. Stitch addresses that 00:17:16.609 --> 00:17:19.009 frustrating bottleneck in building anything new. 00:17:19.250 --> 00:17:22.490 The design phase. How many great ideas stall 00:17:22.490 --> 00:17:24.809 because they're waiting for a designer? Or worse, 00:17:24.990 --> 00:17:27.089 the engineer tries to design it themselves. Right. 00:17:27.130 --> 00:17:28.890 You get something functionally brilliant, but, 00:17:28.910 --> 00:17:32.319 uh... visually challenging. Stitch is Google's 00:17:32.319 --> 00:17:34.119 attempt to solve that. The goal is basically 00:17:34.119 --> 00:17:37.920 to kill the app. forever bold goal how does it 00:17:37.920 --> 00:17:40.740 work it promises from idea to figma in 60 seconds 00:17:40.740 --> 00:17:43.059 pretty much you describe the app or interface 00:17:43.059 --> 00:17:45.599 you want in plain english just type it out and 00:17:45.599 --> 00:17:47.920 stitch acts like your instant ui ux designer 00:17:47.920 --> 00:17:50.599 ui ux being user interface and user experience 00:17:50.599 --> 00:17:53.559 how it looks and feels to use exactly and it 00:17:53.559 --> 00:17:55.119 doesn't just give you one screen it generates 00:17:55.119 --> 00:17:57.160 multiple professional looking screens that are 00:17:57.160 --> 00:17:59.180 connected logically like a prototype think of 00:17:59.180 --> 00:18:01.500 it like getting a high -end prefabricated house 00:18:01.500 --> 00:18:03.990 kit You're not starting from a blank plot of 00:18:03.990 --> 00:18:06.410 land. You get professionally designed modules, 00:18:06.670 --> 00:18:09.890 maybe 80 % of the structure, done in 5 % of the 00:18:09.890 --> 00:18:12.450 usual time. You just customize from there. And 00:18:12.450 --> 00:18:15.190 Figma is that popular design tool lots of teams 00:18:15.190 --> 00:18:17.329 use. Right. A collaborative web -based design 00:18:17.329 --> 00:18:20.349 tool. So you might prompt Stitch. Make an app 00:18:20.349 --> 00:18:22.990 for finding local hiking trails. Main screen 00:18:22.990 --> 00:18:25.490 needs a map view, a list of nearby trails with 00:18:25.490 --> 00:18:27.750 difficulty ratings, and a way to save favorite 00:18:27.750 --> 00:18:30.769 hikes. And it generates. A whole series of connected, 00:18:30.849 --> 00:18:32.930 professional -looking mock -ups based on best 00:18:32.930 --> 00:18:35.410 practices. But here are the two really interesting 00:18:35.410 --> 00:18:38.930 bits. Okay. One, it exports directly to fully 00:18:38.930 --> 00:18:41.839 editable, multi -layered Figma files. Not just 00:18:41.839 --> 00:18:44.460 flat images. Your designer can jump right in 00:18:44.460 --> 00:18:46.400 and tweak everything. That's useful. And two? 00:18:46.519 --> 00:18:49.920 It also generates real front -end code. The actual 00:18:49.920 --> 00:18:51.980 code developers use for the part of the app the 00:18:51.980 --> 00:18:54.380 user sees and interacts with. Wait, usable code? 00:18:54.579 --> 00:18:56.779 Usable front -end code that a developer can take 00:18:56.779 --> 00:18:59.119 and immediately start connecting to a back -end 00:18:59.119 --> 00:19:01.240 database, the part that stores all the data. 00:19:01.529 --> 00:19:04.230 So engineers must love this, surprisingly. Yeah, 00:19:04.329 --> 00:19:05.829 it's interesting. You'd think designers would 00:19:05.829 --> 00:19:08.109 be the main audience, but engineers have really 00:19:08.109 --> 00:19:10.910 latched onto it. Why? Isn't it replacing designers? 00:19:11.450 --> 00:19:13.630 It's not really positioned that way. It's more 00:19:13.630 --> 00:19:16.230 of a productivity and collaboration tool. It 00:19:16.230 --> 00:19:19.109 bridges that gap between engineering and design. 00:19:19.630 --> 00:19:23.170 Wow. It handles the tedious foundational stuff 00:19:23.170 --> 00:19:26.450 layout, spacing, basic components, making sure 00:19:26.450 --> 00:19:29.250 things look clean. This frees up the engineer 00:19:29.250 --> 00:19:32.250 to focus on the logic, the hard stuff under the 00:19:32.250 --> 00:19:34.930 hood. And the designer. By the time a human designer 00:19:34.930 --> 00:19:37.410 gets involved, they're not starting from zero 00:19:37.410 --> 00:19:40.609 or worse from an engineer's messy first draft. 00:19:41.130 --> 00:19:43.990 They start with a beautiful, functional baseline. 00:19:44.539 --> 00:19:46.779 They can focus on the higher level user experience, 00:19:47.000 --> 00:19:49.160 the branding, the polish. Okay, that makes sense. 00:19:49.339 --> 00:19:51.980 So the pro level upgrade here. The full stack 00:19:51.980 --> 00:19:54.440 prototype workflow. This is where you really 00:19:54.440 --> 00:19:56.500 see the power of chaining these tools. Putting 00:19:56.500 --> 00:19:59.400 Stitch together with other things. Yep. Imagine 00:19:59.400 --> 00:20:01.740 a startup team trying to build a functional prototype 00:20:01.740 --> 00:20:04.460 over a weekend hackathon. Team member one gives 00:20:04.460 --> 00:20:06.579 Stitch a detailed prompt for their app idea. 00:20:06.980 --> 00:20:10.119 Minutes later, professional Figma design and 00:20:10.119 --> 00:20:13.140 the front end code. Nice. While maybe a designer 00:20:13.140 --> 00:20:15.900 quickly tweaks the visuals in Figma, a developer 00:20:15.900 --> 00:20:18.259 takes that front -end code and starts connecting 00:20:18.259 --> 00:20:21.339 it to a backend -as -a -service platform, something 00:20:21.339 --> 00:20:24.579 like Supabase or Firebase. Easy database setup. 00:20:24.779 --> 00:20:26.480 Okay, front end and back end coming together. 00:20:26.799 --> 00:20:29.480 And simultaneously, maybe team member three uses 00:20:29.480 --> 00:20:32.599 ChatGPT or Gemini Pro to generate all the text 00:20:32.599 --> 00:20:35.240 for the app, the marketing copy, the onboarding 00:20:35.240 --> 00:20:38.200 messages, button labels, everything. So in one 00:20:38.200 --> 00:20:40.859 afternoon. A small team can go from just an idea 00:20:40.859 --> 00:20:44.160 to a visually appealing, functional front end 00:20:44.160 --> 00:20:46.960 prototype connected to a real database with all 00:20:46.960 --> 00:20:48.920 the copy written. That used to take weeks, maybe 00:20:48.920 --> 00:20:51.180 months. That dramatically speeds things up. Okay, 00:20:51.200 --> 00:20:53.430 a question. How does Stitch Excel... the entire 00:20:53.430 --> 00:20:55.650 product development lifecycle, not just the design 00:20:55.650 --> 00:20:57.750 part. It provides professional design and usable 00:20:57.750 --> 00:21:00.210 code, bridging design and development. Design 00:21:00.210 --> 00:21:02.089 and code bridge. Got it. Okay, we've looked at 00:21:02.089 --> 00:21:04.730 six powerful tools individually. Gemini, Veo, 00:21:04.910 --> 00:21:08.890 Flow, Notebook LM, Mariner, Stitch. Quite the 00:21:08.890 --> 00:21:12.200 lineup. Indeed. But the real power move, the 00:21:12.200 --> 00:21:15.400 thing that separates maybe the casual users from 00:21:15.400 --> 00:21:17.799 the people building serious things. Is combining 00:21:17.799 --> 00:21:20.559 them. Is stopping thinking of them as separate 00:21:20.559 --> 00:21:22.519 instruments and seeing them as components of 00:21:22.519 --> 00:21:25.220 one single integrated machine. This is what we're 00:21:25.220 --> 00:21:28.299 calling the Voltron strategy. I like it. Assembling 00:21:28.299 --> 00:21:31.299 the lions. Exactly. Combining these specialized 00:21:31.299 --> 00:21:34.539 AI lions, each good at its own thing, to form 00:21:34.539 --> 00:21:38.240 one giant. business building super robot. Creating 00:21:38.240 --> 00:21:40.779 a self -perpetuating flywheel, right? Where the 00:21:40.779 --> 00:21:43.220 output of one tool feeds directly into the input 00:21:43.220 --> 00:21:45.500 of the next. An automated system. Precisely. 00:21:45.500 --> 00:21:47.759 Let's walk through a concrete example. Let's 00:21:47.759 --> 00:21:50.839 call it the idea to launch flywheel. Imagine 00:21:50.839 --> 00:21:53.099 a solo entrepreneur trying to launch a niche 00:21:53.099 --> 00:21:55.640 product maybe over a single weekend. Okay, starting 00:21:55.640 --> 00:21:57.359 from scratch, where do they begin? Step one, 00:21:57.480 --> 00:22:00.269 the scout. Project Mariner. Give it a mission. 00:22:00.690 --> 00:22:03.250 Scrape the top 20 posts from the last month in 00:22:03.250 --> 00:22:06.190 the, say, Roltralite subreddit. Identify the 00:22:06.190 --> 00:22:08.849 top five most discussed new or emerging product 00:22:08.849 --> 00:22:11.750 types. Market research. Okay. Mariner brings 00:22:11.750 --> 00:22:15.829 back raw data discussion topics. Step two. The 00:22:15.829 --> 00:22:19.170 strategist. Notebook LM. Feeded that raw data 00:22:19.170 --> 00:22:22.309 prompt. Analyze this data. Synthesize the key 00:22:22.309 --> 00:22:24.690 findings. Identify the single biggest underserved 00:22:24.690 --> 00:22:27.369 product opportunity here and find three external 00:22:27.369 --> 00:22:29.630 articles supporting this market trend. Notebook 00:22:29.630 --> 00:22:31.849 LM clarifies the opportunity and gives supporting 00:22:31.849 --> 00:22:34.970 evidence. Step three, the ideator, Gemini Pro. 00:22:35.470 --> 00:22:39.500 Take Notebook LM's final summary. Command. Based 00:22:39.500 --> 00:22:42.000 on this research, brainstorm five potential brand 00:22:42.000 --> 00:22:44.559 names and write a core value proposition for 00:22:44.559 --> 00:22:47.000 a new ultralight backpacking gadget that solves 00:22:47.000 --> 00:22:49.420 this specific problem. Naming and positioning 00:22:49.420 --> 00:22:53.400 done. Step four. The designer. Stitch. Take the 00:22:53.400 --> 00:22:56.180 chosen brand name and value prop. Prompt. Create 00:22:56.180 --> 00:22:58.519 a three -page website design for a new e -commerce 00:22:58.519 --> 00:23:00.799 brand called Brand Name selling this ultralight 00:23:00.799 --> 00:23:03.480 gadget. Minutes later. Pro Figma design and front 00:23:03.480 --> 00:23:05.579 -end code for the landing page. Website visuals 00:23:05.579 --> 00:23:08.170 and code. Ready to go. Step five. The marketing 00:23:08.170 --> 00:23:11.150 studio, VO3 and Flow. Feed them the product concept, 00:23:11.329 --> 00:23:13.690 the branding from Stitch. Task. Generate five 00:23:13.690 --> 00:23:16.109 short, compelling video ads for social media 00:23:16.109 --> 00:23:18.250 and maybe a slightly longer animated explainer 00:23:18.250 --> 00:23:20.410 video using Flow. Marketing assets generated. 00:23:20.710 --> 00:23:25.210 Finally, step six. The manager. Gemini Pro scheduled 00:23:25.210 --> 00:23:28.650 action. Set up a daily scheduled action in Gemini 00:23:28.650 --> 00:23:31.839 Pro. Monitor traffic and conversion rates on 00:23:31.839 --> 00:23:34.000 the new landing page and the performance of the 00:23:34.000 --> 00:23:36.519 video ad campaigns. Deliver a summary report 00:23:36.519 --> 00:23:39.299 each morning. Automated monitoring and reporting. 00:23:39.519 --> 00:23:41.579 Read the loop. Yeah, it's a continuous automated 00:23:41.579 --> 00:23:44.599 cycle. Market research leads to ideation, leads 00:23:44.599 --> 00:23:47.140 to design, leads to marketing, leads to performance 00:23:47.140 --> 00:23:49.700 monitoring, which then informs the next cycle. 00:23:49.880 --> 00:23:52.259 It lets you generate and test new business ideas 00:23:52.259 --> 00:23:54.500 at a speed and scale that was just unthinkable 00:23:54.500 --> 00:23:57.450 before. Truly. It's an entire system. So the 00:23:57.450 --> 00:24:00.029 probing question for this strategy, what's the 00:24:00.029 --> 00:24:02.910 biggest shift in mindset needed to leverage this 00:24:02.910 --> 00:24:06.029 Voltron approach? Viewing AI tools as interconnected 00:24:06.029 --> 00:24:08.890 parts of an automated end -to -end system. Exactly. 00:24:08.890 --> 00:24:11.069 Seeing the whole system, not just the individual 00:24:11.069 --> 00:24:14.190 tools. So wrapping up, what's the big takeaway 00:24:14.190 --> 00:24:16.309 here? I think the biggest idea is that the future 00:24:16.309 --> 00:24:18.410 of AI, or at least a huge part of it, is moving 00:24:18.410 --> 00:24:22.009 beyond just the chat box. Chat's a powerful interface, 00:24:22.190 --> 00:24:24.150 sure. But maybe not the only one, or always the 00:24:24.150 --> 00:24:27.180 best one. Right. Google's demonstrating that 00:24:27.180 --> 00:24:30.400 for complex tasks, specialized purpose -built 00:24:30.400 --> 00:24:33.019 interfaces, often with visual controls, clear 00:24:33.019 --> 00:24:35.759 workflows, transparent processes like citations, 00:24:35.880 --> 00:24:38.880 can create a far superior user experience. While 00:24:38.880 --> 00:24:41.180 everyone's debating which chatbot writes better 00:24:41.180 --> 00:24:43.799 poems. Google is quietly building the actual 00:24:43.799 --> 00:24:47.079 infrastructure for the future of work. This suite 00:24:47.079 --> 00:24:49.900 of tools we just discussed. Think about the cost 00:24:49.900 --> 00:24:52.839 just two years ago. Software licenses, specialized 00:24:52.839 --> 00:24:56.660 agency fees, salaries. It would have been hundreds 00:24:56.660 --> 00:24:59.619 of thousands of dollars. And now? Now it's accessible 00:24:59.619 --> 00:25:02.339 for, what, roughly the price of a Netflix subscription. 00:25:02.660 --> 00:25:05.180 It's a staggering democratization of capability. 00:25:05.700 --> 00:25:08.059 And the implication. The businesses, the entrepreneurs, 00:25:08.319 --> 00:25:10.900 the creators who take the time to master these 00:25:10.900 --> 00:25:13.579 interconnected tools now, they're going to gain 00:25:13.579 --> 00:25:15.640 an almost insurmountable advantage in the coming 00:25:15.640 --> 00:25:17.740 years. So the call to action is pretty clear 00:25:17.740 --> 00:25:20.519 then. I think so. The best way to really grasp 00:25:20.519 --> 00:25:22.900 this opportunity isn't just listening to us talk 00:25:22.900 --> 00:25:25.900 about it or reading reviews. It's to start experimenting 00:25:25.900 --> 00:25:29.059 yourself. Get hands on. Head over to labs .google. 00:25:29.599 --> 00:25:31.880 Dive in. See what you can build. See how these 00:25:31.880 --> 00:25:34.160 pieces fit together for you. Because, well, here's 00:25:34.160 --> 00:25:36.339 the thing. Your competition isn't waiting. Your 00:25:36.339 --> 00:25:39.279 competition certainly isn't waiting. That's our 00:25:39.279 --> 00:25:41.720 deep dive for today. Thanks for joining us. Thanks, 00:25:41.779 --> 00:25:43.359 everyone. Outiro music.