WEBVTT 00:00:00.000 --> 00:00:02.540 Picture this. It's 3 a .m. You're staring at 00:00:02.540 --> 00:00:04.519 a screen. Your eyes are just blurring You've 00:00:04.519 --> 00:00:07.940 got a deadline at 8 a .m. The slide deck is a 00:00:07.940 --> 00:00:12.380 total mess. The research feels thin and You're 00:00:12.380 --> 00:00:15.599 just exhausted. We've all been there. Oh, that 00:00:15.599 --> 00:00:18.179 is a universal feeling. Yeah, but then contrast 00:00:18.179 --> 00:00:21.079 that with a Reality from one of our case studies 00:00:21.079 --> 00:00:23.899 today where that same workload the one that used 00:00:23.899 --> 00:00:27.440 to take 10 You know grueling hours is finished 00:00:27.440 --> 00:00:30.260 in less than two You're just done. Equality is 00:00:30.260 --> 00:00:32.280 higher. And you've gotten eight hours of your 00:00:32.280 --> 00:00:35.600 life back to sleep or see your family. And that's 00:00:35.600 --> 00:00:37.820 the promise we're digging into today. Welcome 00:00:37.820 --> 00:00:40.570 to the Deep Dive. We're covering something that 00:00:40.570 --> 00:00:42.990 I think feels a little different. Usually we're 00:00:42.990 --> 00:00:45.530 dissecting history or these complex market trends, 00:00:45.630 --> 00:00:48.270 but today we're looking at what our source material 00:00:48.270 --> 00:00:51.570 calls a survival skill. We've got this whole 00:00:51.570 --> 00:00:53.750 stack of research on mastering AI productivity 00:00:53.750 --> 00:00:55.509 tools. And I have to be honest with you, when 00:00:55.509 --> 00:00:57.649 I first saw this topic, I was a little skeptical. 00:00:58.070 --> 00:01:00.250 I've been that person who, well, I download the 00:01:00.250 --> 00:01:02.490 apps, I try them for five minutes, get frustrated. 00:01:02.570 --> 00:01:05.069 And then delete them. And delete them. because 00:01:05.069 --> 00:01:07.569 the learning curve just feels so steep. You're 00:01:07.569 --> 00:01:09.370 definitely not alone in that. Okay. It's the 00:01:09.370 --> 00:01:11.989 shiny object center, right? But looking at the 00:01:11.989 --> 00:01:14.010 data, we are way past the point where these are 00:01:14.010 --> 00:01:16.890 just, you know, fun toys. Right. The source material 00:01:16.890 --> 00:01:19.430 frames this as a really fundamental shift in 00:01:19.430 --> 00:01:22.390 how work gets done. Not in a doom and gloom, 00:01:22.549 --> 00:01:24.769 robots are coming for your job way. Thank goodness. 00:01:24.909 --> 00:01:28.090 But in a don't you want your life back way. These 00:01:28.090 --> 00:01:30.870 aren't just software. They're pitched as partners. 00:01:31.129 --> 00:01:33.329 That distinction is interesting. Partners, not 00:01:33.329 --> 00:01:36.609 tools. Because a tool, it just sits there until 00:01:36.609 --> 00:01:39.209 you pick it up. A partner contributes. Exactly. 00:01:39.390 --> 00:01:41.950 We need to stop thinking of AI as a calculator 00:01:41.950 --> 00:01:46.349 and start thinking of it as a collaborator. So 00:01:46.349 --> 00:01:48.370 today is about breaking down the toolkit for 00:01:48.370 --> 00:01:51.109 what's being called the 2026 standard. We're 00:01:51.109 --> 00:01:53.409 going to look at four distinct pillars. Research, 00:01:53.989 --> 00:01:57.450 voice, video, and presentation. And then this 00:01:57.450 --> 00:01:58.909 is the important part. We're going to stitch 00:01:58.909 --> 00:02:01.439 them all together into a single workflow. OK, 00:02:01.480 --> 00:02:03.060 let's unpack this, because the biggest hurdle 00:02:03.060 --> 00:02:05.019 for me, and I think for a lot of people, is just 00:02:05.019 --> 00:02:08.219 the sheer noise. There are too many tools. So 00:02:08.219 --> 00:02:10.680 let's start with the foundation research. The 00:02:10.680 --> 00:02:13.319 foundation of everything. The old way was, what, 00:02:13.520 --> 00:02:17.379 Googling for a week, opening 50 tabs, and just 00:02:17.379 --> 00:02:21.020 praying the sources were real. But our sources 00:02:21.020 --> 00:02:24.400 highlight this huge problem with early AI tools. 00:02:25.450 --> 00:02:27.650 hallucinations yeah can we define that really 00:02:27.650 --> 00:02:30.189 quick for everyone sure so in a large language 00:02:30.189 --> 00:02:33.110 model a hallucination is when the AI generates 00:02:33.110 --> 00:02:36.250 something that sounds totally confident totally 00:02:36.250 --> 00:02:39.430 plausible but it's factually wrong it just made 00:02:39.430 --> 00:02:42.259 it up why does it do that It happens because 00:02:42.259 --> 00:02:44.939 the model is just predicting the next likely 00:02:44.939 --> 00:02:47.879 word. It's not actually querying a database of 00:02:47.879 --> 00:02:50.039 truth. So if you ask for a citation, it might 00:02:50.039 --> 00:02:52.639 just invent a paper title that sounds academic. 00:02:53.099 --> 00:02:55.319 But the paper itself never existed. Which is 00:02:55.319 --> 00:02:57.419 a complete nightmare if you're doing actual work. 00:02:57.520 --> 00:02:59.219 It's a career -ender in some fields. I mean, 00:02:59.319 --> 00:03:01.139 if you're doing academic research or a deep marker 00:03:01.139 --> 00:03:03.620 report, accuracy is the only metric that matters. 00:03:03.759 --> 00:03:05.860 And that's where our first tool comes in. Elicit. 00:03:06.080 --> 00:03:08.509 Elicit. So how is this different from just asking 00:03:08.509 --> 00:03:11.090 chat GPT a question? OK, the fundamental difference 00:03:11.090 --> 00:03:13.849 is its architecture. The list doesn't just scrape 00:03:13.849 --> 00:03:16.409 the open web. It's basically a search engine 00:03:16.409 --> 00:03:19.270 over these massive libraries of scientific papers. 00:03:19.430 --> 00:03:22.389 So you type in a question. It finds related papers. 00:03:23.229 --> 00:03:25.990 But then, and this is the key, it summarizes 00:03:25.990 --> 00:03:28.629 the abstracts of those specific papers. So it's 00:03:28.629 --> 00:03:31.909 built for honesty. It's built for rigor. It constrains 00:03:31.909 --> 00:03:34.289 the AI's creativity to the text of the source 00:03:34.289 --> 00:03:37.330 documents. If it lists a citation, you can click 00:03:37.330 --> 00:03:39.810 it and read the actual PDF. OK, but let's be 00:03:39.810 --> 00:03:42.669 real. Most of us aren't writing PhD theses every 00:03:42.669 --> 00:03:44.110 day. What if I just want to know what's happening 00:03:44.110 --> 00:03:46.349 in the news? Elicit sounds like overkill for 00:03:46.349 --> 00:03:49.330 that. It is. And that is exactly where perplexity 00:03:49.330 --> 00:03:51.830 comes in. You can think of perplexity as the 00:03:51.830 --> 00:03:53.909 bridge between a standard Google search and a 00:03:53.909 --> 00:03:56.449 chat bot. I've been seeing this name pop up everywhere 00:03:56.449 --> 00:03:59.669 in text circles. For good reason. Its superpower 00:03:59.669 --> 00:04:01.860 is the fact check. You ask it a question, it 00:04:01.860 --> 00:04:04.039 browses the live internet news sites, Reddit, 00:04:04.159 --> 00:04:06.759 whatever, and it compiles an answer. But every 00:04:06.759 --> 00:04:08.560 sentence has a little footnote number. Oh, that's 00:04:08.560 --> 00:04:10.659 smart. You click the number, you see the original 00:04:10.659 --> 00:04:13.199 website. It lets you verify the AI immediately. 00:04:13.599 --> 00:04:16.079 It's great for just scanning news, checking viewpoints 00:04:16.079 --> 00:04:18.379 without opening those 50 tabs we talked about. 00:04:19.060 --> 00:04:21.740 So Illicit is the library, Perplexity is the 00:04:21.740 --> 00:04:24.120 newsstand. What about the big players? I mean, 00:04:24.259 --> 00:04:28.319 Google is pushing Gemini so hard. Gemini is interesting. 00:04:28.420 --> 00:04:31.720 Their deep research feature is incredibly fast 00:04:31.720 --> 00:04:34.899 because it has this huge context window. It can 00:04:34.899 --> 00:04:36.819 hold a lot of information in its memory at once. 00:04:37.680 --> 00:04:40.680 But, and this is a big warning from our source 00:04:40.680 --> 00:04:42.779 material. Here's where the nuance comes in. You 00:04:42.779 --> 00:04:45.360 have to be careful. The sources warn that Gemini, 00:04:45.439 --> 00:04:48.160 because it's a more general creative model, still 00:04:48.160 --> 00:04:50.720 carries a higher risk of hallucination than illicit. 00:04:51.240 --> 00:04:53.040 It might just invent a source if the question 00:04:53.040 --> 00:04:55.660 is too niche. So what's the advice? Use Gemini 00:04:55.660 --> 00:04:58.079 for broad strokes, getting an outline, brainstorming 00:04:58.079 --> 00:04:59.680 angles, that kind of thing. But don't rely on 00:04:59.680 --> 00:05:01.540 it for your final facts without checking. That 00:05:01.540 --> 00:05:03.759 makes perfect sense. Fast engine for the rough 00:05:03.759 --> 00:05:06.839 draft, precise engine for the facts. So let me 00:05:06.839 --> 00:05:08.860 just boil this down. If accuracy is the only 00:05:08.860 --> 00:05:11.079 metric that matters, if I absolutely cannot afford 00:05:11.079 --> 00:05:14.319 to be wrong, Where do I go? Elicit for deep science, 00:05:14.860 --> 00:05:17.839 perplexity for checking news facts. Short and 00:05:17.839 --> 00:05:20.660 sweet. OK. So we've done a research. Now we need 00:05:20.660 --> 00:05:23.100 to communicate it. And the sources talk about 00:05:23.100 --> 00:05:26.600 this huge shift from text to voice. Yeah, this 00:05:26.600 --> 00:05:28.800 is one of the fastest changing areas. We are 00:05:28.800 --> 00:05:30.899 moving away from that era where you needed a 00:05:30.899 --> 00:05:34.060 soundproof booth, an expensive microphone, and 00:05:34.060 --> 00:05:38.490 some distinct radio voice to get pro audio. looking 00:05:38.490 --> 00:05:40.670 into hiring a voice actor for a project once 00:05:40.670 --> 00:05:44.250 the cost was just astronomical and the turnaround 00:05:44.250 --> 00:05:46.889 was weeks now we're talking about cloning a voice 00:05:46.889 --> 00:05:49.269 we are and this brings up the whole uncanny valley 00:05:49.269 --> 00:05:51.370 problem right that creepy feeling when something 00:05:51.370 --> 00:05:54.449 sounds almost human but a little robotic mm -hmm 00:05:54.449 --> 00:05:56.670 the goal is to cross that valley and the gold 00:05:56.670 --> 00:05:58.509 standard right now according to our material 00:05:58.509 --> 00:06:01.449 is 11 labs I've heard this one creates voices 00:06:01.449 --> 00:06:04.290 that actually sound human. It's not just about 00:06:04.290 --> 00:06:06.350 sounding human, it's about performance. That's 00:06:06.350 --> 00:06:09.149 the key. With 11 Labs, you can upload a small 00:06:09.149 --> 00:06:11.649 sample of a voice, even your own, and it creates 00:06:11.649 --> 00:06:14.050 a clone that you can really manipulate. Speed, 00:06:14.329 --> 00:06:16.970 pitch, stability. When you say stability, what 00:06:16.970 --> 00:06:19.589 does that mean in audio? Stability basically 00:06:19.589 --> 00:06:23.250 controls how much emotion the AI adds. So low 00:06:23.250 --> 00:06:26.000 stability means the AI takes risks. It might 00:06:26.000 --> 00:06:28.740 shout or whisper or even crack its voice to simulate 00:06:28.740 --> 00:06:32.040 drama. High stability, it stays consistent like 00:06:32.040 --> 00:06:34.180 a news reader. So if I'm telling a story and 00:06:34.180 --> 00:06:37.939 I need the narrator to sound devastated or ecstatic, 00:06:38.240 --> 00:06:40.439 Eleven Labs can actually do that. It captures 00:06:40.439 --> 00:06:42.680 that nuance. It's the tool for storytelling, 00:06:42.680 --> 00:06:45.220 but it's not the only player. The sources also 00:06:45.220 --> 00:06:47.759 talk about Minimax. Minimax, that sounds like 00:06:47.759 --> 00:06:50.060 a villain in a kids movie. It's the speed specialist. 00:06:50.620 --> 00:06:52.939 If Eleven Labs is the method actor, Minimax is 00:06:52.939 --> 00:06:56.040 the 24 -hour news anchor. It just processes data 00:06:56.040 --> 00:06:58.860 incredibly fast and the sound is very clean, 00:06:59.160 --> 00:07:01.779 very articulate. But does it have the soul? Not 00:07:01.779 --> 00:07:04.050 really. That's the limitation they cite. It lacks 00:07:04.050 --> 00:07:06.470 that deep emotional range, that breathiness of 00:07:06.470 --> 00:07:08.850 11 Labs. But if you have a massive amount of 00:07:08.850 --> 00:07:11.110 text to process, say converting a long report 00:07:11.110 --> 00:07:13.410 into an audio summary for your commute. Then 00:07:13.410 --> 00:07:15.670 Minimax is the efficient choice. Exactly. Cost 00:07:15.670 --> 00:07:18.370 effective and fast. OK. So is there a clear dividing 00:07:18.370 --> 00:07:20.930 line for when to use which one? Yes. Emotion 00:07:20.930 --> 00:07:23.689 and story go to 11 Labs. Speed and volume go 00:07:23.689 --> 00:07:27.050 to Minimax. Got it. OK. Let's pivot to what the 00:07:27.050 --> 00:07:31.279 source calls the trend of 2026. Video. The document 00:07:31.279 --> 00:07:35.579 says, if 2025 was images and text, 2026 is the 00:07:35.579 --> 00:07:38.240 year of AI video. This is where we see the biggest 00:07:38.240 --> 00:07:41.339 wow factor, but also the most confusion. We're 00:07:41.339 --> 00:07:44.339 moving so fast from those weird glitchy AI clips 00:07:44.339 --> 00:07:46.500 where people had seven fingers. And their faces 00:07:46.500 --> 00:07:49.399 would just melt. Yeah. to send them at a quality 00:07:49.399 --> 00:07:51.800 that takes minutes to generate. I have to admit, 00:07:51.839 --> 00:07:54.740 the first time I saw high -end AI video recently, 00:07:54.899 --> 00:07:57.100 I was genuinely shocked. It didn't look like 00:07:57.100 --> 00:07:59.500 a cartoon. It looked like B -roll footage from 00:07:59.500 --> 00:08:02.240 a real documentary. And that gap is closing so 00:08:02.240 --> 00:08:04.680 fast. The source highlights three main tools, 00:08:04.740 --> 00:08:06.980 and they each have a different superpower. First, 00:08:07.040 --> 00:08:09.379 there's cling AI. Cling? What's its claim to 00:08:09.379 --> 00:08:11.699 fame? Physics and motion control. One of the 00:08:11.699 --> 00:08:13.879 biggest issues with early AI video was something 00:08:13.879 --> 00:08:16.639 called temporal consistency. That's a fancy term. 00:08:16.759 --> 00:08:19.040 It just means... Does the object stay the same 00:08:19.040 --> 00:08:21.779 object over time? In older models, a character 00:08:21.779 --> 00:08:24.259 would walk and their leg might vanish, or their 00:08:24.259 --> 00:08:26.540 face would morph into someone else by frame 50. 00:08:27.240 --> 00:08:28.939 Kling has solved a lot of that. It keeps the 00:08:28.939 --> 00:08:31.240 character's shape stable. So if I need a shot 00:08:31.240 --> 00:08:34.120 of a person walking down a hallway, and I need 00:08:34.120 --> 00:08:36.240 them to actually look like the same person at 00:08:36.240 --> 00:08:39.659 the end of it, Kling is the go -to. Precisely. 00:08:39.840 --> 00:08:42.679 It can handle complex actions. Walking, turning, 00:08:42.940 --> 00:08:46.049 moving back and forth without breaking the physics 00:08:46.049 --> 00:08:48.090 of the image. Then we have Runway. I feel like 00:08:48.090 --> 00:08:50.110 Runway has been around for a while, at least 00:08:50.110 --> 00:08:52.950 in AI years. Runway is positioned as the tool 00:08:52.950 --> 00:08:56.169 for the artist. It's less about just generating 00:08:56.169 --> 00:08:58.980 a clip and more about visual thinking. It creates 00:08:58.980 --> 00:09:01.399 these scenes with a lot of depth, a very expensive 00:09:01.399 --> 00:09:04.620 cinematic look. Plus, it has editing tools built 00:09:04.620 --> 00:09:06.879 right into the browser. So you can fix and cut 00:09:06.879 --> 00:09:09.100 the video right there instead of exporting it 00:09:09.100 --> 00:09:11.399 to Premiere or Final Cut? Correct. It's a creative 00:09:11.399 --> 00:09:13.620 suite. But then if you just want something done 00:09:13.620 --> 00:09:16.159 exactly as you asked for, without all the artistic 00:09:16.159 --> 00:09:19.500 flair, there's Veo from Google. The obedient 00:09:19.500 --> 00:09:22.250 one? The obedient one. Vio is described as following 00:09:22.250 --> 00:09:24.690 prompts exactly. It's a great budget option, 00:09:24.970 --> 00:09:27.049 perfect for high -volume social media drafts, 00:09:27.190 --> 00:09:29.250 where you just need the video to exist and match 00:09:29.250 --> 00:09:31.870 the text. OK, so let's bring it back to the practical 00:09:31.870 --> 00:09:34.330 test. If I need a character to walk across a 00:09:34.330 --> 00:09:37.750 room without glitching out, who wins? Kling AI 00:09:37.750 --> 00:09:40.789 is the current king of motion control and stability. 00:09:40.950 --> 00:09:43.850 OK. We're halfway through the toolkit. We have 00:09:43.850 --> 00:09:46.750 the research, the voice, the video. Let's take 00:09:46.750 --> 00:09:48.769 a very brief moment before we get into the final 00:09:48.769 --> 00:09:51.529 piece of this puzzle presentation. Mid -roll 00:09:51.529 --> 00:09:57.149 sponsor, Reed Placeholder. And we are back. So 00:09:57.149 --> 00:09:59.070 we have all these raw materials. We have our 00:09:59.070 --> 00:10:01.190 facts, our audio, our visuals. But eventually 00:10:01.190 --> 00:10:03.529 you have to present this stuff to a boss, a client, 00:10:03.690 --> 00:10:07.309 a team. And that usually means. the dreaded slide 00:10:07.309 --> 00:10:09.669 deck. The nightmare of manual formatting. Oh, 00:10:09.769 --> 00:10:12.370 it is the worst. Aligning text boxes, choosing 00:10:12.370 --> 00:10:14.769 fonts, trying to find images that don't look 00:10:14.769 --> 00:10:17.190 like those cheesy stock photos. The source says 00:10:17.190 --> 00:10:20.169 we can hand off 80 % of this work. At least 80%. 00:10:20.169 --> 00:10:21.909 There are three tools here, and they all tackle 00:10:21.909 --> 00:10:24.289 this from different angles. First up, gamma AI. 00:10:24.870 --> 00:10:27.490 Gamma. I've heard this described as notion meets 00:10:27.490 --> 00:10:30.009 PowerPoint. That's a good analogy. Gamma is the 00:10:30.009 --> 00:10:32.090 wow factor tool because it completely changes 00:10:32.090 --> 00:10:34.409 the paradigm. You don't drag and drop boxes anymore. 00:10:34.429 --> 00:10:36.289 You just talk to it. What do you mean? You give 00:10:36.289 --> 00:10:38.610 it a topic where you paste in your rough notes 00:10:38.610 --> 00:10:41.710 and it builds the deck for you. It handles the 00:10:41.710 --> 00:10:44.610 layout, the text, the images automatically. It 00:10:44.610 --> 00:10:46.970 does it. It does it. It's designed for pure speed. 00:10:47.629 --> 00:10:49.490 If you need a presentation in 10 minutes for 00:10:49.490 --> 00:10:52.210 a meeting, Gamma is your answer. It looks modern. 00:10:52.230 --> 00:10:54.629 It's polished. The only downside is sometimes 00:10:54.629 --> 00:10:57.649 you have to swap out a few images it picks. But 00:10:57.649 --> 00:10:59.789 all the heavy lifting is done. That sounds like 00:10:59.789 --> 00:11:02.289 a dream for a quick pitch. But what if I have 00:11:02.289 --> 00:11:05.690 that 50 -page PDF we talked about? Gamma might 00:11:05.690 --> 00:11:07.889 gloss over the details. Then you want Notebook 00:11:07.889 --> 00:11:11.110 LM. This one is from Google. It's less of a designer 00:11:11.110 --> 00:11:13.169 and more of a disciplined student. Disciplined 00:11:13.169 --> 00:11:15.809 how? It used this technique called RAG, Retrieval 00:11:15.809 --> 00:11:18.629 Augmented Generation. Basically, you upload your 00:11:18.629 --> 00:11:21.509 specific sources, your PDFs, your docs, and it 00:11:21.509 --> 00:11:24.210 reads them. So when it creates a summary or slides, 00:11:24.289 --> 00:11:26.789 it is grounded entirely in that data. It doesn't 00:11:26.789 --> 00:11:29.250 hallucinate outside info. Oh, interesting. It's 00:11:29.250 --> 00:11:31.169 perfect for students who are technical presentations, 00:11:31.250 --> 00:11:33.809 where accuracy is way more important than aesthetics. 00:11:34.330 --> 00:11:37.210 So gamma for the sales pitch, notebook LM for 00:11:37.210 --> 00:11:39.809 the quarterly technical report. I assume Canva 00:11:39.809 --> 00:11:42.450 is still in the mix? Canva is for the control 00:11:42.450 --> 00:11:46.049 freak, or let's be nicer, the designer. OK. They've 00:11:46.049 --> 00:11:49.210 added AI tools, Magic Studio, that suggest layouts, 00:11:49.409 --> 00:11:51.980 but you control every single pixel. If you need 00:11:51.980 --> 00:11:54.940 a very specific brand look, or you want to design 00:11:54.940 --> 00:11:57.659 a unique infographic, Canva is still the best. 00:11:58.159 --> 00:12:00.600 It's just slower than Gamma. Okay, so here's 00:12:00.600 --> 00:12:04.299 the scenario. I have a dense 50 -page PDF, and 00:12:04.299 --> 00:12:06.740 I need a summary deck for a board meeting. Which 00:12:06.740 --> 00:12:09.759 tool? Notebook LM. It reads the file and stays 00:12:09.759 --> 00:12:12.519 disciplined to the source data. Okay. We have 00:12:12.519 --> 00:12:14.559 covered the individual tools, but here's where 00:12:14.559 --> 00:12:17.220 it gets really interesting for me. The source 00:12:17.220 --> 00:12:19.320 material talks about the integrated workflow. 00:12:19.639 --> 00:12:21.879 It's not just about using these in isolation. 00:12:22.000 --> 00:12:24.139 It's about connecting them. Right. If you use 00:12:24.139 --> 00:12:25.940 them separately, you're just saving minutes here 00:12:25.940 --> 00:12:27.919 and there. But if you connect them, you change 00:12:27.919 --> 00:12:30.179 the entire production method. You stop being 00:12:30.179 --> 00:12:32.480 a writer or an editor, and you start being a 00:12:32.480 --> 00:12:35.000 producer. The source outlines a 30 -minute A 00:12:35.000 --> 00:12:37.820 to Z process. Let's walk through this step by 00:12:37.820 --> 00:12:39.779 step. Let's say we're making a short explainer 00:12:39.779 --> 00:12:41.460 video about, I don't know, the future of coffee. 00:12:41.879 --> 00:12:45.539 OK, great example. Step one, research. You don't 00:12:45.539 --> 00:12:48.009 Google. You use perplexity. You ask it for the 00:12:48.009 --> 00:12:50.129 latest trends in coffee production, climate impact, 00:12:50.610 --> 00:12:52.409 market data. You get the facts and the sources 00:12:52.409 --> 00:12:54.870 in minutes. Okay, data acquired. I know the story. 00:12:55.029 --> 00:12:58.950 Step two, script. You feed that specific data 00:12:58.950 --> 00:13:02.149 into Gemini. You ask it to write a 60 -second 00:13:02.149 --> 00:13:04.490 video script based only on the facts you just 00:13:04.490 --> 00:13:06.730 found. You can tweak it, give it some personality. 00:13:06.889 --> 00:13:10.220 Script is done. Step three, voice. You copy that 00:13:10.220 --> 00:13:12.639 script into 11 labs. You pick a narrator voice, 00:13:12.779 --> 00:13:14.639 maybe a gritty documentary style, maybe a bright 00:13:14.639 --> 00:13:16.899 commercial one, hit generate. Now you have a 00:13:16.899 --> 00:13:19.159 professional voiceover. No studio needed. Step 00:13:19.159 --> 00:13:22.740 four, video. This is the magic. You take the 00:13:22.740 --> 00:13:24.500 script scenes, you know, farmer walking in a 00:13:24.500 --> 00:13:26.759 field, coffee beans roasting, and you describe 00:13:26.759 --> 00:13:29.580 them to cling AI. You generate the visuals to 00:13:29.580 --> 00:13:32.200 match your story. And finally. Step five, packaging. 00:13:32.840 --> 00:13:34.679 You take your key data points and the script. 00:13:34.830 --> 00:13:37.610 And you feed that outline into gamma AI to create 00:13:37.610 --> 00:13:39.590 a presentation back to pitch the whole idea. 00:13:39.809 --> 00:13:41.990 And the claim is this whole process can take 00:13:41.990 --> 00:13:45.669 30 minutes. From an idea to a rough presentable 00:13:45.669 --> 00:13:48.990 product. Yeah. Yes. The friction is just gone. 00:13:49.289 --> 00:13:51.169 The information flows from one tool right to 00:13:51.169 --> 00:13:54.210 the next. So the key is the handoff between tools. 00:13:54.470 --> 00:13:57.750 Exactly. Data flows from research to voice to 00:13:57.750 --> 00:14:01.129 video in one single stream. You're not context 00:14:01.129 --> 00:14:04.179 switching all the time. That is wild. It really 00:14:04.179 --> 00:14:06.720 reframes the whole narrative. It sounds less 00:14:06.720 --> 00:14:09.879 like AI will take our jobs and more like AI will 00:14:09.879 --> 00:14:11.679 do the boring parts so we can actually think. 00:14:12.000 --> 00:14:14.879 That is the core philosophy here. And the source 00:14:14.879 --> 00:14:16.700 warns against trying to use everything at once. 00:14:16.940 --> 00:14:19.299 That just leads to confusion. But if you master 00:14:19.299 --> 00:14:21.639 the workflow, you're not working more. You're 00:14:21.639 --> 00:14:24.259 working smarter. Which brings us to the big recap. 00:14:24.340 --> 00:14:26.159 We've covered a lot of ground here. We've talked 00:14:26.159 --> 00:14:28.580 about illicit for deep science, cling for stable 00:14:28.580 --> 00:14:31.940 video, gamma for instant slides. But the overarching 00:14:31.940 --> 00:14:34.740 message, it seems to be about agency. It's about 00:14:34.740 --> 00:14:37.419 removing barriers. I mean, think about it. Previously, 00:14:37.460 --> 00:14:39.159 if you wanted to make a film, you needed a camera 00:14:39.159 --> 00:14:41.019 crew. Right. If you wanted a professional slide 00:14:41.019 --> 00:14:43.440 deck, you needed a graphic designer. Deep research, 00:14:43.440 --> 00:14:45.299 you needed a library card in weeks of your time. 00:14:45.320 --> 00:14:49.059 Now, those barriers are just gone. The only barrier 00:14:49.059 --> 00:14:51.000 left is the willingness to actually learn the 00:14:51.000 --> 00:14:54.769 tool. Precisely. The takeaway isn't that AI replaces 00:14:54.769 --> 00:14:58.029 humans. It's that humans who use AI replace those 00:14:58.029 --> 00:15:00.789 who don't. It's a force multiplier. And the source 00:15:00.789 --> 00:15:03.309 gives some specific advice on how to start, right? 00:15:03.350 --> 00:15:05.370 Because this list we just went through can feel 00:15:05.370 --> 00:15:08.850 pretty overwhelming. The advice is simple. Don't 00:15:08.850 --> 00:15:11.769 try to build the 30 -minute workflow today. Start 00:15:11.769 --> 00:15:15.269 slowly. Pick one pain point. Do you hate making 00:15:15.269 --> 00:15:18.330 slides? Download gamma. Do you struggle finding 00:15:18.330 --> 00:15:22.559 good sources? Try perplexity. Master one tool, 00:15:22.799 --> 00:15:25.039 then expand. I love that. Don't boil the ocean, 00:15:25.100 --> 00:15:26.759 just pick one thing. And remember that these 00:15:26.759 --> 00:15:29.539 are partners. Treat them like assistants. You're 00:15:29.539 --> 00:15:31.860 still the director. You have to guide them, check 00:15:31.860 --> 00:15:33.779 their work, and provide that creative spark. 00:15:34.159 --> 00:15:35.539 So here's something I want you to think about 00:15:35.539 --> 00:15:37.820 as you walk away from this deep dive. We always 00:15:37.820 --> 00:15:40.440 talk about AI in terms of efficiency saving time, 00:15:40.820 --> 00:15:43.919 but what if we thought about it in terms of capability? 00:15:44.480 --> 00:15:46.299 What's the project you've had on a shelf for 00:15:46.299 --> 00:15:48.139 five years because you didn't have the budget 00:15:48.139 --> 00:15:51.190 or the team or the skills to do it? That's the 00:15:51.190 --> 00:15:53.490 provocative question, isn't it? With this stack 00:15:53.490 --> 00:15:57.389 of tools, cling, 11 labs, gamma, that project 00:15:57.389 --> 00:15:59.649 isn't impossible anymore. It's just a matter 00:15:59.649 --> 00:16:02.049 of sitting down and actually doing it. The cost 00:16:02.049 --> 00:16:05.110 of failure has dropped to zero. The cost of experimentation 00:16:05.110 --> 00:16:08.149 is just your time. So that's our challenge to 00:16:08.149 --> 00:16:11.309 you. Pick one tool we mentioned, just one, and 00:16:11.309 --> 00:16:13.629 try it on your next project. See if you can reclaim 00:16:13.629 --> 00:16:15.850 those eight hours. And let us know how it goes. 00:16:16.309 --> 00:16:18.110 Thanks for listening to the Deep Dive. We'll 00:16:18.110 --> 00:16:18.769 see you next time.