WEBVTT 00:00:00.000 --> 00:00:02.919 Imagine writing just a single sentence, a simple 00:00:02.919 --> 00:00:05.480 prompt, and then maybe 50 seconds later you're 00:00:05.480 --> 00:00:07.879 watching a fully functional web app just spring 00:00:07.879 --> 00:00:10.000 to life. And it's not just generating the code, 00:00:10.019 --> 00:00:12.400 it's the whole outcome. I mean, think about a 00:00:12.400 --> 00:00:15.099 personal finance tracker with logs for your income 00:00:15.099 --> 00:00:18.379 and expenses, and it has a working colorful pie 00:00:18.379 --> 00:00:20.260 chart that updates itself. Or something even 00:00:20.260 --> 00:00:22.899 more complex, like a lo -fi beat maker with a 00:00:22.899 --> 00:00:25.899 full retro 80s design, right? With sound and 00:00:25.899 --> 00:00:28.320 a looping play button. Exactly. That kind of 00:00:28.320 --> 00:00:31.000 immediate creation, it used to take hours, maybe 00:00:31.000 --> 00:00:33.579 days of really tedious work. This feels like 00:00:33.579 --> 00:00:36.460 a massive, massive jump. It really does. Let's 00:00:36.460 --> 00:00:39.320 unpack this. Welcome to the deep dive. We know 00:00:39.320 --> 00:00:42.520 the AI news cycle is just relentless. There's 00:00:42.520 --> 00:00:45.119 always some new model, some new feature. But 00:00:45.119 --> 00:00:46.859 what Google has just rolled out with their latest 00:00:46.859 --> 00:00:50.200 model, Gemini 3 .0, this represents a really 00:00:50.200 --> 00:00:52.659 fundamental massive shift in what these things 00:00:52.659 --> 00:00:54.679 can do. This isn't just a small version bump. 00:00:54.939 --> 00:00:57.479 So our mission today is pretty straightforward. 00:00:57.899 --> 00:01:01.079 We need to cut through all the noise and really 00:01:01.079 --> 00:01:04.000 understand the mechanics here. Why is Google 00:01:04.000 --> 00:01:07.439 suddenly leapfrogging the competition? You know, 00:01:07.700 --> 00:01:10.180 winning the major benchmark tests. And what does 00:01:10.180 --> 00:01:12.560 that actually mean for you? We'll start by defining 00:01:12.560 --> 00:01:15.180 the key breakthrough, which researchers are calling 00:01:15.180 --> 00:01:18.079 reasoning. Then we'll dive into the pretty jaw 00:01:18.079 --> 00:01:20.829 -dropping reality of one -shot coding. And finally, 00:01:21.189 --> 00:01:23.549 we'll look at Google's secret advantages, their 00:01:23.549 --> 00:01:25.969 data, and their custom chips, and show you some 00:01:25.969 --> 00:01:28.329 practical tools you can start using today. Let's 00:01:28.329 --> 00:01:30.510 start with that comeback story. Because for a 00:01:30.510 --> 00:01:32.890 long time, it really felt like Google was losing 00:01:32.890 --> 00:01:35.489 the AI race. Oh, they were, relatively speaking. 00:01:35.590 --> 00:01:37.629 I mean, if we go back a couple years, they launched 00:01:37.629 --> 00:01:40.469 BARD. And to be honest, it just missed the mark. 00:01:40.569 --> 00:01:44.299 It was slow. So slow and it made these embarrassing 00:01:44.299 --> 00:01:46.719 errors compared to its rivals It honestly felt 00:01:46.719 --> 00:01:48.920 like Google the company that pioneered so much 00:01:48.920 --> 00:01:51.260 of this research was just falling behind Yeah, 00:01:51.319 --> 00:01:52.799 it felt like they were playing catch -up, which 00:01:52.799 --> 00:01:54.599 was kind of frustrating when you know the resources 00:01:54.599 --> 00:01:57.340 they have but what the sources we looked at reveal 00:01:57.340 --> 00:01:59.519 is that they didn't stop they just went quiet 00:01:59.519 --> 00:02:03.680 and they leveraged that massive financial power 00:02:03.680 --> 00:02:06.780 and that just unparalleled pile of data they're 00:02:06.780 --> 00:02:08.759 sitting on. And the result is this current version 00:02:08.759 --> 00:02:11.819 of Gemini. This is the comeback. It is now consistently 00:02:11.819 --> 00:02:15.199 beating every other top -tier model on all the 00:02:15.199 --> 00:02:17.379 major intelligence tests that researchers use. 00:02:17.520 --> 00:02:19.939 I mean, the shift is just undeniable. But, you 00:02:19.939 --> 00:02:22.860 know, beating academic tests doesn't always translate 00:02:22.860 --> 00:02:25.620 to useful intelligence. What really defines this 00:02:25.620 --> 00:02:29.020 new model beyond just raw speed? The key breakthrough, 00:02:29.099 --> 00:02:31.719 like you said, is genuine reasoning. So older 00:02:31.719 --> 00:02:34.219 models were basically brilliant prediction machines. 00:02:34.819 --> 00:02:37.199 They just guessed the next most probable word 00:02:37.199 --> 00:02:39.840 in a sentence based on statistics. They were 00:02:39.840 --> 00:02:42.680 excellent at imitating, but honestly, pretty 00:02:42.680 --> 00:02:44.879 poor at thinking. So what does this new ability 00:02:44.879 --> 00:02:48.449 to reason actually look like in practice? It 00:02:48.449 --> 00:02:51.210 means the AI takes a kind of internal pause. 00:02:51.409 --> 00:02:54.069 It takes time to think before generating an answer. 00:02:54.610 --> 00:02:57.229 It'll run internal checks, process the information 00:02:57.229 --> 00:02:59.849 more deeply, and create a logical plan before 00:02:59.849 --> 00:03:01.629 it executes. So it's like moving from a need 00:03:01.629 --> 00:03:04.629 your guess to a deep reflective pause, where 00:03:04.629 --> 00:03:07.210 you actually verify your own steps. Exactly. 00:03:07.550 --> 00:03:10.849 And that ability fundamentally changes how reliable 00:03:10.849 --> 00:03:13.740 the complex outputs are. With an older model, 00:03:13.759 --> 00:03:16.080 if it made a mistake in step one of a 10 -step 00:03:16.080 --> 00:03:19.340 process, that error would just cascade down and 00:03:19.340 --> 00:03:21.639 ruin the entire result. Right. Because Gemini 00:03:21.639 --> 00:03:24.560 3 .0 reasons, it can often spot and correct those 00:03:24.560 --> 00:03:26.639 internal errors before they become a bigger problem. 00:03:26.699 --> 00:03:28.419 And what's fascinating is that this reasoning 00:03:28.419 --> 00:03:31.180 power is deeply tied to its multimodal capabilities, 00:03:31.419 --> 00:03:33.419 right? It's precisely. It's inherently multimodal. 00:03:33.439 --> 00:03:36.000 That means it doesn't just process text. It perfectly 00:03:36.000 --> 00:03:39.520 understands images, videos, audio, physical concepts. 00:03:39.659 --> 00:03:42.569 So give us an example. How does training on specifically 00:03:42.569 --> 00:03:45.210 help it reason. Okay, think about the broken 00:03:45.210 --> 00:03:47.669 bicycle example. You show it a short video of 00:03:47.669 --> 00:03:50.129 a bike, and you point out that the chain is slipping 00:03:50.129 --> 00:03:53.189 off the sprocket. The AI isn't just reading a 00:03:53.189 --> 00:03:55.289 description, it's literally watching the physics 00:03:55.289 --> 00:03:57.930 of failure. It understands how the world moves. 00:03:57.990 --> 00:04:00.789 I see. And from that, it can then generate step 00:04:00.789 --> 00:04:02.750 -by -step instructions on how to fix it, because 00:04:02.750 --> 00:04:05.590 it's drawing on this deep understanding of mechanical 00:04:05.590 --> 00:04:09.129 action, not just keywords. So if the AI takes 00:04:09.129 --> 00:04:13.360 time to reason, How does that fundamentally change 00:04:13.360 --> 00:04:16.920 the reliability of its complex outcomes? It processes 00:04:16.920 --> 00:04:20.660 complex inputs by running internal checks, minimizing 00:04:20.660 --> 00:04:23.740 the potential for cascading logical errors. Okay, 00:04:23.939 --> 00:04:25.680 let's talk about building things. This next part 00:04:25.680 --> 00:04:27.060 is where it gets really interesting because it 00:04:27.060 --> 00:04:29.519 just changes the efficiency equation completely. 00:04:29.790 --> 00:04:32.629 It accelerates it by an order of magnitude, truly. 00:04:33.029 --> 00:04:35.430 When we first started writing code with AI tools, 00:04:35.490 --> 00:04:37.810 even the best ones, it was this frustrating back 00:04:37.810 --> 00:04:39.730 and forth. Oh, yeah. You'd ask it to build something. 00:04:39.769 --> 00:04:41.470 It would generate code. You'd find four bugs. 00:04:41.509 --> 00:04:43.810 You'd ask it to fix them. It would break something 00:04:43.810 --> 00:04:46.550 else. I mean, it took hours. Right? And I still 00:04:46.550 --> 00:04:48.850 wrestle with prompt drift myself. You know, you 00:04:48.850 --> 00:04:50.670 try to keep a conversation going for too long, 00:04:50.930 --> 00:04:53.629 and halfway through, the AI just forgets the 00:04:53.629 --> 00:04:55.470 original constraints you gave it. I think that's 00:04:55.470 --> 00:04:58.920 a... real, vulnerable admission for anyone who 00:04:58.920 --> 00:05:01.800 uses these tools a lot. That whole painful cycle 00:05:01.800 --> 00:05:05.060 is what we're now calling the old way. Now we're 00:05:05.060 --> 00:05:08.040 seeing one -shot coding. You give one detailed, 00:05:08.240 --> 00:05:10.639 well -structured instruction, just a single prompt, 00:05:11.019 --> 00:05:14.699 and the entire job is done right in one go, in 00:05:14.699 --> 00:05:16.779 less than a minute. Which means the skill shifts 00:05:16.779 --> 00:05:20.100 entirely. It's no longer about correcting the 00:05:20.100 --> 00:05:22.899 AI's mistakes. It's about crafting that perfect, 00:05:23.160 --> 00:05:25.420 clear instruction the first time. Let's look 00:05:25.420 --> 00:05:27.459 at that personal finance dashboard example because 00:05:27.459 --> 00:05:29.660 it really shows the depth here. The prompt was 00:05:29.660 --> 00:05:31.740 really demanding. It had to be a functional dashboard 00:05:31.740 --> 00:05:35.160 all in a single HTML file. It had to have dark 00:05:35.160 --> 00:05:37.800 mode, income and expense tracking, and automatically 00:05:37.800 --> 00:05:40.720 updating colorful pie chart, and crucially, data 00:05:40.720 --> 00:05:43.980 persistence. That is a huge load of complexity. 00:05:44.360 --> 00:05:46.379 Can you define data persistence for someone who 00:05:46.379 --> 00:05:49.160 might not know? Sure. Persistence just means 00:05:49.160 --> 00:05:51.740 that the data you input you know, your expense 00:05:51.740 --> 00:05:54.000 tracking, it doesn't vanish if you refresh the 00:05:54.000 --> 00:05:57.779 browser page. For a simple HTML file, the AI 00:05:57.779 --> 00:06:00.540 has to correctly implement the browser's storage 00:06:00.540 --> 00:06:03.480 logic, like local storage, to save that info. 00:06:04.240 --> 00:06:06.319 That's a technically tricky detail that older 00:06:06.319 --> 00:06:08.480 models always struggled with in one command. 00:06:08.720 --> 00:06:11.250 And what was the result? Took about 50 seconds. 00:06:11.550 --> 00:06:13.449 The math was right, the design was polished, 00:06:13.509 --> 00:06:15.689 and the data actually saved perfectly between 00:06:15.689 --> 00:06:18.389 refreshes. A task that might take a junior developer 00:06:18.389 --> 00:06:21.529 half a day was done in under a minute. That is 00:06:21.529 --> 00:06:24.589 genuinely astonishing. But if it's that fast, 00:06:24.990 --> 00:06:28.490 doesn't that speed also create enormous risks? 00:06:28.930 --> 00:06:31.069 What about security or just the quality of the 00:06:31.069 --> 00:06:34.399 code? That's a very valid question. The observation 00:06:34.399 --> 00:06:36.180 right now is that because the reasoning model 00:06:36.180 --> 00:06:38.300 is more robust, the code it produces actually 00:06:38.300 --> 00:06:40.060 tends to be logically cleaner than the older 00:06:40.060 --> 00:06:42.800 models. And just to test the limits of its logic 00:06:42.800 --> 00:06:45.120 and its aesthetic sense, they asked it to build 00:06:45.120 --> 00:06:47.639 a lo -fi beat maker. OK, tell me what that prompt 00:06:47.639 --> 00:06:50.519 demanded. It needed sound interaction, a 16 -button 00:06:50.519 --> 00:06:53.300 grid, a looping play button, and a very specific 00:06:53.300 --> 00:06:56.459 aesthetic. A retro 80s synthesizer look with 00:06:56.459 --> 00:06:59.459 neon colors, so it had to understand complex 00:06:59.459 --> 00:07:02.279 musical timing and style constraints at the same 00:07:02.279 --> 00:07:05.149 time. Whoa. Imagine scaling that complexity. 00:07:05.430 --> 00:07:09.069 It finished in maybe 90 seconds, and it produced 00:07:09.069 --> 00:07:11.930 a working drum machine with correct timing, sound, 00:07:12.089 --> 00:07:15.019 and the requested 80s vibe. That's the moment 00:07:15.019 --> 00:07:17.699 of wonder, right? It gets deep logic like music 00:07:17.699 --> 00:07:19.800 loops and precise styling in the same breath. 00:07:20.040 --> 00:07:22.040 Does this mean the AI is only good at building 00:07:22.040 --> 00:07:24.920 simple, self -contained apps? No. It demonstrates 00:07:24.920 --> 00:07:27.660 understanding of deep logic like music loops 00:07:27.660 --> 00:07:29.819 and precise styling, paving the way for larger 00:07:29.819 --> 00:07:32.399 projects. So let's shift to the engine under 00:07:32.399 --> 00:07:35.240 the hood. How did Google get so far ahead so 00:07:35.240 --> 00:07:37.459 fast? It really boils down to two huge advantages. 00:07:37.980 --> 00:07:40.779 data and chips. Right. And the data is the raw 00:07:40.779 --> 00:07:42.779 fuel for learning. We all know Google basically 00:07:42.779 --> 00:07:45.079 has the library of the world. They absolutely 00:07:45.079 --> 00:07:47.220 own it. Google Search, Scholar, Google Books, 00:07:47.279 --> 00:07:49.579 that gives them an unmatched depth of text data. 00:07:49.879 --> 00:07:51.620 But the sources we looked at really highlight 00:07:51.620 --> 00:07:54.019 YouTube as the secret weapon for this new generation 00:07:54.019 --> 00:07:57.819 of multimodal AI. How does YouTube, a video platform, 00:07:58.240 --> 00:08:01.129 help an AI reason better? Well, YouTube provides 00:08:01.129 --> 00:08:04.209 billions of videos. So Gemini 3 .0 has essentially 00:08:04.209 --> 00:08:06.550 watched the entire world in motion. This helps 00:08:06.550 --> 00:08:08.829 it understand how objects behave, how the world 00:08:08.829 --> 00:08:11.509 interacts, not just how we write about it. So 00:08:11.509 --> 00:08:14.310 if you train an AI on text, you learn that fire 00:08:14.310 --> 00:08:16.810 is hot. But if you train it on YouTube video, 00:08:17.209 --> 00:08:19.829 you learn that fire spreads, objects move when 00:08:19.829 --> 00:08:22.589 you push them, friction... causes wear and tear. 00:08:23.069 --> 00:08:25.970 Exactly. It understands physics, geometry, real 00:08:25.970 --> 00:08:28.529 world cause and effect. Competitors have to pay 00:08:28.529 --> 00:08:31.089 these insane fees or scrape the web to get similar 00:08:31.089 --> 00:08:33.669 training data, while Google already owns the 00:08:33.669 --> 00:08:36.730 largest library of human action in motion. That 00:08:36.730 --> 00:08:39.049 data moat is just impenetrable. And the other 00:08:39.049 --> 00:08:40.850 side of that coin is the custom hardware they 00:08:40.850 --> 00:08:43.210 run it all on. Most other companies are relying 00:08:43.210 --> 00:08:45.629 on NVIDIA GPUs, which are, you know, fantastic 00:08:45.629 --> 00:08:48.210 chips. Think of NVIDIA GPUs as a really high 00:08:48.210 --> 00:08:50.629 -end off -the -rack suit. It's great, but it's 00:08:50.629 --> 00:08:53.350 made for general -purpose computing. Google made 00:08:53.350 --> 00:08:55.490 a strategic bet years ago and designed its own 00:08:55.490 --> 00:08:57.789 chips from the ground up. They're called TPUs, 00:08:57.970 --> 00:09:00.610 or Tensor Processing Units. So their TPUs are 00:09:00.610 --> 00:09:03.070 the custom -tailored suit. A perfect analogy. 00:09:03.710 --> 00:09:06.750 These chips are built specifically and only to 00:09:06.750 --> 00:09:09.070 handle the unique massive calculations their 00:09:09.070 --> 00:09:12.059 AI models need. That specialization gives them 00:09:12.059 --> 00:09:14.820 a profound competitive edge. And what's the practical 00:09:14.820 --> 00:09:17.679 benefit of that? It lets Google run these colossal 00:09:17.679 --> 00:09:20.320 models faster and, this is the important part, 00:09:20.700 --> 00:09:23.259 cheaper than anyone else who is relying on external 00:09:23.259 --> 00:09:26.460 hardware. That combination of speed and low cost 00:09:26.460 --> 00:09:29.299 lets them iterate, test, and improve their models 00:09:29.299 --> 00:09:31.759 way faster than any competitor could afford to. 00:09:31.899 --> 00:09:34.059 So how does Google's control over both hardware 00:09:34.059 --> 00:09:36.860 and data widen their competitive advantage? They 00:09:36.860 --> 00:09:39.299 control their own destiny. Competitors rely on 00:09:39.299 --> 00:09:41.779 outside chips and resources, which slows them 00:09:41.779 --> 00:09:44.240 down and increases their costs. For the learner 00:09:44.240 --> 00:09:46.299 or the student or the professional who just needs 00:09:46.299 --> 00:09:48.659 to synthesize information faster, Google has 00:09:48.659 --> 00:09:51.559 now baked this immense power into a new set of 00:09:51.559 --> 00:09:53.960 practical tools. Yeah, let's focus on deep research, 00:09:54.019 --> 00:09:56.519 which is a massive leap beyond simple search. 00:09:56.879 --> 00:09:59.179 This tool isn't for a quick answer. It's a long 00:09:59.179 --> 00:10:01.799 analysis tool. It takes 10 to 15 minutes and 00:10:01.799 --> 00:10:04.539 it reads 50, 100, sometimes 200 different websites 00:10:04.539 --> 00:10:07.220 and synthesizes all that complex data for you. 00:10:07.320 --> 00:10:10.720 And the use case is so important here. You wouldn't 00:10:10.720 --> 00:10:13.539 ask it what time the sun rises. You'd ask it 00:10:13.539 --> 00:10:15.879 something complex like, find the best places 00:10:15.879 --> 00:10:18.320 to live in Southeast Asia for a family of four. 00:10:18.980 --> 00:10:21.940 And you want it to synthesize details from government 00:10:21.940 --> 00:10:24.899 visa sites, school pricing, weather reports, 00:10:25.179 --> 00:10:27.539 all of that. Right. It saves you from opening 00:10:27.539 --> 00:10:30.600 50 browser tabs, fighting with conflicting info, 00:10:30.639 --> 00:10:32.539 and trying to pull all those threads together. 00:10:33.100 --> 00:10:35.700 Deep research just does the synthesis for you, 00:10:35.840 --> 00:10:38.220 and it organizes it logically. It's knowledge 00:10:38.220 --> 00:10:41.679 acquisition at speed. And search itself is changing 00:10:41.679 --> 00:10:44.360 with this visual search integration. It feels 00:10:44.360 --> 00:10:47.019 like we're moving from a search engine that finds 00:10:47.019 --> 00:10:49.759 links to one that delivers answers. That's a 00:10:49.759 --> 00:10:51.639 great way to put it. If you ask Google, how does 00:10:51.639 --> 00:10:54.480 a car engine work? You might now see a little 00:10:54.480 --> 00:10:56.940 dynamic animation of the pistons moving right 00:10:56.940 --> 00:10:59.100 there on the results page instead of just a Wikipedia 00:10:59.100 --> 00:11:02.019 link. That instant visual comprehension is incredibly 00:11:02.019 --> 00:11:04.419 powerful for lightning. Okay, so if we want to 00:11:04.419 --> 00:11:06.899 access this power right now, what's the entry 00:11:06.899 --> 00:11:09.820 point? The easiest way is the web app at Gemini 00:11:09.820 --> 00:11:12.820 .Google .com. But, and this is important, you 00:11:12.820 --> 00:11:15.340 need to select the correct mode to get the reasoning 00:11:15.340 --> 00:11:17.559 power we've been talking about. You have to select 00:11:17.559 --> 00:11:20.019 the full power version, which is labeled Thinking 00:11:20.019 --> 00:11:22.620 Mode or Pro. And what do we lose if we pick the 00:11:22.620 --> 00:11:24.340 other option? Well, there's a faster, cheaper 00:11:24.340 --> 00:11:27.259 mode called Flash. And while Flash is speedy 00:11:27.259 --> 00:11:29.919 and it's great for simple queries, it lacks that 00:11:29.919 --> 00:11:32.100 deep reasoning ability and the error checking. 00:11:32.639 --> 00:11:35.779 If you want that complex one -shot coding power, 00:11:36.299 --> 00:11:39.350 you need Pro. And for students? or anyone dealing 00:11:39.350 --> 00:11:41.850 with dense documents, what about Notebook LM? 00:11:42.289 --> 00:11:45.250 This is an incredibly powerful tool for academia, 00:11:45.730 --> 00:11:48.210 for heavy reading. You upload your lecture notes, 00:11:48.389 --> 00:11:51.149 your textbooks, your complex PDFs, up to 50 documents. 00:11:51.490 --> 00:11:53.669 And the crucial part is that the AI will only 00:11:53.669 --> 00:11:55.830 answer questions based on those documents. That 00:11:55.830 --> 00:11:58.690 is the key distinction. It ensures accuracy by 00:11:58.690 --> 00:12:01.350 preventing what we call hallucination. Exactly. 00:12:01.710 --> 00:12:03.990 It stops the AI from just making things up or 00:12:03.990 --> 00:12:06.389 pulling from the wider web, which is so important 00:12:06.389 --> 00:12:08.629 in fields where precision is everything. You 00:12:08.629 --> 00:12:11.210 can ask it to make me a study guide for the final 00:12:11.210 --> 00:12:14.210 based only on Chapter 5 or explain the hardest 00:12:14.210 --> 00:12:16.690 concept in this paper like I'm five years old 00:12:16.690 --> 00:12:19.389 and it draws exclusively from your material. 00:12:19.850 --> 00:12:22.230 For students, how crucial is it that Notebook 00:12:22.230 --> 00:12:26.299 LM only uses uploaded documents? It ensures accuracy, 00:12:26.639 --> 00:12:29.419 preventing the AI from fabricating or making 00:12:29.419 --> 00:12:32.419 things up that could jeopardize study results. 00:12:32.580 --> 00:12:34.960 So to quickly synthesize the biggest takeaway 00:12:34.960 --> 00:12:37.419 here, the combination of custom hardware, the 00:12:37.419 --> 00:12:40.460 TPUs, and that unparalleled data from YouTube 00:12:40.460 --> 00:12:43.100 means Google has fundamentally shifted the AI 00:12:43.100 --> 00:12:46.149 landscape. Their new models are focused on genuine 00:12:46.149 --> 00:12:48.769 reasoning, not just prediction. And the implication 00:12:48.769 --> 00:12:51.370 of that is just monumental. The barrier to entry 00:12:51.370 --> 00:12:53.690 for building complex functional software is, 00:12:53.690 --> 00:12:56.009 well, it's nearly gone. The power to create these 00:12:56.009 --> 00:12:58.190 systems is now available to anyone who can just 00:12:58.190 --> 00:13:00.149 clearly describe what they want it to do. Let's 00:13:00.149 --> 00:13:01.850 give the listeners some practical advice here. 00:13:02.529 --> 00:13:05.590 Use the multimodal power. Don't just type out 00:13:05.590 --> 00:13:08.149 your instructions. Use your phone, upload a quick 00:13:08.149 --> 00:13:10.429 photo sketch of a website design you want, and 00:13:10.429 --> 00:13:12.629 just say, build a website that looks like this 00:13:12.629 --> 00:13:15.269 sketch using these colors. That's a massive shortcut. 00:13:15.549 --> 00:13:18.029 And don't forget the power for boring tasks. 00:13:18.269 --> 00:13:21.929 Use that reasoning to organize or rename 50 messy 00:13:21.929 --> 00:13:24.409 files on your computer based on what's inside 00:13:24.409 --> 00:13:27.250 them, not just their file names. It saves a ton 00:13:27.250 --> 00:13:30.110 of admin time. The final provocative thought 00:13:30.110 --> 00:13:32.750 here is this. The future of work isn't about 00:13:32.750 --> 00:13:35.830 mastering complex syntax. It's not about laying 00:13:35.830 --> 00:13:38.370 every single brick anymore. It's about becoming 00:13:38.370 --> 00:13:41.750 the architect who directs the AI to build the 00:13:41.750 --> 00:13:44.409 walls. That's the skill shift. We really encourage 00:13:44.409 --> 00:13:46.850 you go right now and try the free version. Ask 00:13:46.850 --> 00:13:49.190 it to explain a complex hobby you have, write 00:13:49.190 --> 00:13:51.769 a dense meal plan, or help with a tricky piece 00:13:51.769 --> 00:13:54.190 of homework. The best way to truly understand 00:13:54.190 --> 00:13:56.370 this acceleration is to just use it yourself. 00:13:56.549 --> 00:13:58.590 The future is here, waiting for you to tie a 00:13:58.590 --> 00:14:00.490 pillow. We'll see you on the next deep dive.