WEBVTT 00:00:00.000 --> 00:00:02.220 We are looking at a fundamental shift in the 00:00:02.220 --> 00:00:05.580 AI landscape today. Right. I mean, scammers are 00:00:05.580 --> 00:00:08.800 literally using AI to spin up 10 ,000 deceptive 00:00:08.800 --> 00:00:12.039 ads an hour right now. It is forcing Google into 00:00:12.039 --> 00:00:15.080 an unprecedented algorithmic war. Welcome to 00:00:15.080 --> 00:00:16.780 the Deep Dive. We're really glad you're joining 00:00:16.780 --> 00:00:19.039 us. Yeah. Thanks for hanging out with us. Today 00:00:19.039 --> 00:00:21.879 we're tracking Anthropic's sneaky new pricing 00:00:21.879 --> 00:00:25.960 model with Cloud Opus 4 .7. We'll also explore 00:00:25.960 --> 00:00:30.199 how GPT 5 .4 Pro just made math professors weep. 00:00:30.199 --> 00:00:32.299 Oh, that story is absolutely wild. It really 00:00:32.299 --> 00:00:34.759 is. And we're going to unpack why Google is radically 00:00:34.759 --> 00:00:39.500 changing how it fights internet scammers. Let's 00:00:39.500 --> 00:00:41.600 start with the new intelligence baseline. Cloud 00:00:41.600 --> 00:00:45.659 Opus 4 .7 is officially out. It's positioned 00:00:45.659 --> 00:00:48.280 as their smartest public model. Yeah, it's a 00:00:48.280 --> 00:00:50.280 completely fascinating release. They're calling 00:00:50.280 --> 00:00:52.700 it the most reliable model available right now. 00:00:52.780 --> 00:00:55.039 It actually holds the highest public spot on 00:00:55.039 --> 00:00:56.880 the HLE benchmark. And that stands for human 00:00:56.880 --> 00:00:59.560 level execution, right? Exactly. And it hit 46 00:00:59.560 --> 00:01:03.479 .9 % without using any external tools. That number 00:01:03.479 --> 00:01:05.939 alone is quite impressive for a public model. 00:01:06.180 --> 00:01:08.799 But there is a totally new dynamic at play here. 00:01:09.200 --> 00:01:11.359 The whole smarter means more expensive thing. 00:01:11.480 --> 00:01:13.879 Right. The actual token prices are the exact 00:01:13.879 --> 00:01:17.760 same as 4 .6. But Opus 4 .7 just thinks much 00:01:17.760 --> 00:01:20.640 more. It uses higher effort levels to verify 00:01:20.640 --> 00:01:23.159 its own outputs. Right. So it checks its own 00:01:23.159 --> 00:01:26.299 underlying logic for complex coding. It doesn't 00:01:26.299 --> 00:01:28.560 just spit out the very first answer. It pauses. 00:01:28.739 --> 00:01:32.159 It reflects internally and, you know, verifies 00:01:32.159 --> 00:01:34.260 the math before it ever speaks to you. This completely 00:01:34.260 --> 00:01:36.769 changes how we interact with the machine. I mean, 00:01:36.769 --> 00:01:39.250 I still wrestle with prompt drift myself. Oh, 00:01:39.250 --> 00:01:41.750 absolutely. We all do. You ask an older model 00:01:41.750 --> 00:01:44.430 for a complex script and halfway through it just 00:01:44.430 --> 00:01:46.609 forgets the original parameters. Yeah, the context 00:01:46.609 --> 00:01:50.269 window just degrades over time. But Opus 4 .7 00:01:50.269 --> 00:01:53.409 actively fixes that drift. The internal verification 00:01:53.409 --> 00:01:56.189 acts as an anchor. It's basically your new baseline 00:01:56.189 --> 00:01:58.709 for professional reliability. Let's compare the 00:01:58.709 --> 00:02:02.170 broader HLE stats a bit. Gemini 3 .1 Pro is sitting 00:02:02.170 --> 00:02:08.770 at 44 .4%. And GPD 5 .4 Pro is at 42 .7. So Claude 00:02:08.770 --> 00:02:11.610 is clearly leading the pack right now. But we 00:02:11.610 --> 00:02:15.050 really cannot ignore the mythos shadow. Oh man, 00:02:15.110 --> 00:02:18.889 the mythos shadow. It's brilliant, slightly intimidating 00:02:18.889 --> 00:02:22.349 marketing. Anthropix model card directly compared 00:02:22.349 --> 00:02:25.729 4 .7 to Claude Mythos. Which is their unreleased, 00:02:25.729 --> 00:02:28.469 highly guarded internal model. Right. And it 00:02:28.469 --> 00:02:32.449 scores 56 .8 % on that exact same test. That 00:02:32.449 --> 00:02:35.069 is a massive 10 % jump. They're deliberately 00:02:35.069 --> 00:02:37.830 holding back the true frontier to show us exactly 00:02:37.830 --> 00:02:40.650 what's coming next. I kind of think of the Opus 00:02:40.650 --> 00:02:44.129 pricing like hiring a contractor. How do you 00:02:44.129 --> 00:02:45.969 mean? Well, you hire someone who charges the 00:02:45.969 --> 00:02:49.099 exact same hourly rate. but they take 40 hours 00:02:49.099 --> 00:02:51.319 instead of 10. Ah, I see. They do it to guarantee 00:02:51.319 --> 00:02:53.840 a perfect foundation, so you end up paying more 00:02:53.840 --> 00:02:56.039 for that extra hidden time. Yeah. Speaking of 00:02:56.039 --> 00:02:58.840 extra time, we really have to look at GPT 5 .4 00:02:58.840 --> 00:03:01.460 Pro. Whoa, imagine a machine generating a mathematical 00:03:01.460 --> 00:03:03.840 proof so beautiful it makes human professors 00:03:03.840 --> 00:03:06.520 weep. Many experts are saying human intuition 00:03:06.520 --> 00:03:09.020 is officially cooked. Because it cracked that 00:03:09.020 --> 00:03:12.719 legendary 60 -year -old ergo's problem? I want 00:03:12.719 --> 00:03:14.599 to linger on that idea for a second. Go for it. 00:03:14.889 --> 00:03:18.289 Paul Erdős was this eccentric, brilliant, 20th 00:03:18.289 --> 00:03:21.270 century mathematician. He believed God maintained 00:03:21.270 --> 00:03:23.750 a metaphorical book containing the most perfect, 00:03:23.909 --> 00:03:26.889 elegant proofs for every theorem. Right, the 00:03:26.889 --> 00:03:29.770 famous book proofs. And mathematicians strive 00:03:29.770 --> 00:03:32.389 to find those specific proofs. Exactly. They 00:03:32.389 --> 00:03:34.050 aren't just looking for the right answer. They're 00:03:34.050 --> 00:03:37.099 looking for the most elegant, logical path. AI 00:03:37.099 --> 00:03:38.860 was always supposed to be just a brute force 00:03:38.860 --> 00:03:40.819 calculator, right? Yeah, we thought it would 00:03:40.819 --> 00:03:43.099 solve math by just crunching endless numbers. 00:03:43.400 --> 00:03:46.439 But GPT -5 .4 Pro didn't do that. It found a 00:03:46.439 --> 00:03:49.460 genuinely beautiful, elegant solution. Two -sec 00:03:49.460 --> 00:03:52.500 silence. Does giving a model extra computing 00:03:52.500 --> 00:03:55.639 time actually guarantee a correct answer, or 00:03:55.639 --> 00:03:58.060 just a highly confident wrong one? Well, it's 00:03:58.060 --> 00:04:00.159 actually about verifying the core logic. It doesn't 00:04:00.159 --> 00:04:03.270 just expand the text volume. You know, it actively 00:04:03.270 --> 00:04:05.810 checks the steps before it answers and runs internal 00:04:05.810 --> 00:04:08.669 tests against its own assumptions. So more compute 00:04:08.669 --> 00:04:11.629 equals genuine verification. But you better watch 00:04:11.629 --> 00:04:14.530 your token budget. Precisely. You are paying 00:04:14.530 --> 00:04:17.189 for that internal reflection now. And this massive 00:04:17.189 --> 00:04:20.189 leap in reasoning changes everything. It fundamentally 00:04:20.189 --> 00:04:22.569 alters the entire business model of artificial 00:04:22.569 --> 00:04:26.029 intelligence. You simply cannot offer flat monthly 00:04:26.029 --> 00:04:28.410 rates anymore when machines are doing this kind 00:04:28.410 --> 00:04:31.399 of heavy computational labor. The economics simply 00:04:31.399 --> 00:04:33.939 don't work out. I mean, if a user asks for a 00:04:33.939 --> 00:04:37.060 60 -year -old math proof, the machine works overtime. 00:04:37.560 --> 00:04:40.360 And Anthropic just made a massive move to address 00:04:40.360 --> 00:04:42.519 this. They officially shifted Claude Enterprise 00:04:42.519 --> 00:04:46.240 to usage -based billing. Which perfectly matches 00:04:46.240 --> 00:04:48.720 your actual compute needs. Heavy corporate users 00:04:48.720 --> 00:04:50.800 are definitely going to see a sharp price jump. 00:04:50.959 --> 00:04:53.560 It is a completely necessary move for their survival, 00:04:53.639 --> 00:04:56.120 though. The hardware supporting this hidden reasoning 00:04:56.120 --> 00:04:59.160 is... booming right now. Cerebra Systems just 00:04:59.160 --> 00:05:02.699 snagged $850 million in new funding. That is 00:05:02.699 --> 00:05:04.879 a staggering amount of capital to raise overnight. 00:05:05.300 --> 00:05:08.139 It boosts their total funding to nearly $3 billion. 00:05:08.620 --> 00:05:12.199 $2 .85 billion, to be exact. They are building 00:05:12.199 --> 00:05:15.879 faster, vastly smarter AI hardware. We desperately 00:05:15.879 --> 00:05:18.139 need that physical power to run these heavy tools. 00:05:18.360 --> 00:05:21.000 Look at the new Cloud Code desktop app. Yeah, 00:05:21.040 --> 00:05:23.120 that thing is entirely designed for parallel 00:05:23.120 --> 00:05:25.829 urgent decoding. Software that independently 00:05:25.829 --> 00:05:29.750 writes, tests, and ships its own code. Beat. 00:05:29.889 --> 00:05:33.089 It runs complex sessions across multiple repositories 00:05:33.089 --> 00:05:34.769 simultaneously. Or it doesn't just sit around 00:05:34.769 --> 00:05:36.569 waiting for you to hit enter. It just gets to 00:05:36.569 --> 00:05:39.509 work. Yeah. And these tools are merging with 00:05:39.509 --> 00:05:42.459 our daily design work. Canva just dropped their 00:05:42.459 --> 00:05:46.040 massive AI 2 .0 update. Yeah, using prompt -powered 00:05:46.040 --> 00:05:48.720 tools for almost everything now. You design and 00:05:48.720 --> 00:05:51.139 edit via simple text descriptions. Google is 00:05:51.139 --> 00:05:53.120 doing the exact same thing within their ecosystem. 00:05:53.339 --> 00:05:56.819 They unlocked a side -by -side AI mode directly 00:05:56.819 --> 00:05:59.759 in Chrome. That supports deep multi -tab grounding 00:05:59.759 --> 00:06:02.939 and massive PDF analysis, right? Exactly. You 00:06:02.939 --> 00:06:05.980 access it seamlessly via a new plus menu. But 00:06:05.980 --> 00:06:08.839 the real industry drama is the suspocalypse. 00:06:09.120 --> 00:06:11.949 Oh, yeah. Rumors are exploding online about Opus 00:06:11.949 --> 00:06:14.970 4 .7 being a legitimate Figma killer tool. Let's 00:06:14.970 --> 00:06:17.430 unpack the causality there. Figma is the absolute 00:06:17.430 --> 00:06:19.550 industry standard for interface design. Right. 00:06:19.629 --> 00:06:21.649 Everybody relies on it. If Opus can generate 00:06:21.649 --> 00:06:24.569 perfect tested interfaces from simple text prompts, 00:06:24.829 --> 00:06:27.569 Figma's dominance is totally threatened. And 00:06:27.569 --> 00:06:29.550 the timing of these rumors is highly suspicious. 00:06:29.970 --> 00:06:32.449 Because Chief Product Officer Mike Krieger just 00:06:32.449 --> 00:06:36.000 abruptly quit Anthropix Board. And Krieger is 00:06:36.000 --> 00:06:38.899 a legendary design visionary. He co -founded 00:06:38.899 --> 00:06:41.459 Instagram. Right. It feels like the entire design 00:06:41.459 --> 00:06:44.300 industry is bracing for impact. He might be stepping 00:06:44.300 --> 00:06:47.079 away due to direct conflicts of interest. The 00:06:47.079 --> 00:06:49.399 fundamental landscape of software is shifting 00:06:49.399 --> 00:06:53.319 under our feet. Two sec silence. Are we seeing 00:06:53.319 --> 00:06:55.660 the death of the flat rate software subscription 00:06:55.660 --> 00:06:58.600 model in real time? Well, processing power is 00:06:58.600 --> 00:07:01.620 a hard physical constraint. Charging per use 00:07:01.620 --> 00:07:03.860 is really the only sustainable path forward. 00:07:04.019 --> 00:07:06.439 These companies just can't absorb infinite compute 00:07:06.439 --> 00:07:09.819 costs. Right. Usage -based billing means we are 00:07:09.819 --> 00:07:11.699 basically paying for digital electricity now. 00:07:11.879 --> 00:07:13.889 That is the perfect way to look at it. Businesses 00:07:13.889 --> 00:07:16.449 are paying for AI like electricity. They use 00:07:16.449 --> 00:07:19.009 it to build amazing complex things every single 00:07:19.009 --> 00:07:21.670 day, but bad actors are constantly using that 00:07:21.670 --> 00:07:24.350 exact same cheap electricity. To break things 00:07:24.350 --> 00:07:27.629 at an unprecedented scale, Google's 2025 ad safety 00:07:27.629 --> 00:07:30.509 report is honestly terrifying. Scammers are using 00:07:30.509 --> 00:07:33.110 large language models at a truly massive scale. 00:07:33.370 --> 00:07:36.370 Yeah, they churn out 10 ,000 unique ad variants 00:07:36.370 --> 00:07:39.089 an hour. That volume would have been physically 00:07:39.089 --> 00:07:41.410 impossible just two years ago. Let's explain 00:07:41.410 --> 00:07:44.189 how that actually works. In the past, a scammer 00:07:44.189 --> 00:07:47.050 uploaded a single malicious ad. Right. Google 00:07:47.050 --> 00:07:49.810 caught it, banned the account, and blacklisted 00:07:49.810 --> 00:07:52.290 the image. The problem was temporarily solved. 00:07:52.589 --> 00:07:56.490 But now that same scammer uses an LLM. The AI 00:07:56.490 --> 00:07:59.149 generates 10 ,000 completely unique variations 00:07:59.149 --> 00:08:01.889 of the ad copy. So banning the individual accounts 00:08:01.889 --> 00:08:05.019 is now totally useless. Bad actors just script 00:08:05.019 --> 00:08:08.000 new ones via API instantly. It takes a millisecond. 00:08:08.139 --> 00:08:10.459 Google had to change their entire defensive strategy 00:08:10.459 --> 00:08:13.600 to survive. They are essentially banning AI ads 00:08:13.600 --> 00:08:16.060 entirely, but they're surprisingly keeping the 00:08:16.060 --> 00:08:18.079 actual advertisers on the platform. They're fighting 00:08:18.079 --> 00:08:20.620 fire with fire using their Gemini models. They 00:08:20.620 --> 00:08:22.939 trained Gemini to target the content instead 00:08:22.939 --> 00:08:25.199 of the individual creators. It looks for the 00:08:25.199 --> 00:08:28.009 semantic intent behind the deceptive ad. The 00:08:28.009 --> 00:08:30.689 statistics on the shift are absolutely wild to 00:08:30.689 --> 00:08:33.370 read. Yet Google removed 8 billion ads globally 00:08:33.370 --> 00:08:35.850 last year, yet the total number of suspended 00:08:35.850 --> 00:08:37.929 accounts actually dropped. In the United States 00:08:37.929 --> 00:08:42.169 alone, 1 .7 billion ads were pulled. And in India, 00:08:42.330 --> 00:08:45.529 the metrics look even more extreme. Blocked ads 00:08:45.529 --> 00:08:49.830 nearly doubled almost overnight to 483 .7 million. 00:08:50.220 --> 00:08:52.940 Meanwhile, actual account bans in India fell 00:08:52.940 --> 00:08:55.820 sharply. They went from 2 .9 million down to 00:08:55.820 --> 00:08:58.960 1 .7 million. Because Google focuses purely on 00:08:58.960 --> 00:09:02.240 policing the individual creatives now. And this 00:09:02.240 --> 00:09:04.960 new automated approach is highly effective in 00:09:04.960 --> 00:09:07.419 practice. They successfully reduced accidental 00:09:07.419 --> 00:09:10.480 account suspensions by 80 percent. Google claims 00:09:10.480 --> 00:09:13.759 their systems catch 99 % of policy -violating 00:09:13.759 --> 00:09:16.500 ads. Literally catching them before human eyes 00:09:16.500 --> 00:09:19.340 ever see them. But there is a huge, undeniable 00:09:19.340 --> 00:09:22.320 business tension here. Google's massive corporate 00:09:22.320 --> 00:09:25.220 revenue depends entirely on ads running continuously. 00:09:25.679 --> 00:09:27.960 Right. Balancing platform safety with the bottom 00:09:27.960 --> 00:09:31.080 line is incredibly tricky. It is especially hard 00:09:31.080 --> 00:09:34.279 when 602 million blocked ads were malicious scams. 00:09:34.419 --> 00:09:37.059 That is a massive volume of bad traffic to filter 00:09:37.059 --> 00:09:40.799 out daily. Two sec silence. If Google's AI catches 00:09:40.799 --> 00:09:43.919 99 % of scams, what happens when scammers upgrade 00:09:43.919 --> 00:09:46.940 to smarter AI? It just forces Google's Gemini 00:09:46.940 --> 00:09:50.159 to train even harder. It creates a perpetual 00:09:50.159 --> 00:09:53.100 invisible machine war behind the scenes where 00:09:53.100 --> 00:09:55.779 neither side can ever afford to stop innovating. 00:09:55.879 --> 00:09:58.440 It essentially becomes an endless arms race between 00:09:58.440 --> 00:10:01.480 two opposing algorithms. And it never really 00:10:01.480 --> 00:10:04.100 stops. This invisible arms race is completely 00:10:04.100 --> 00:10:06.740 fascinating. But it's not just happening in the 00:10:06.740 --> 00:10:09.379 dark shadows of the internet. The people actively 00:10:09.379 --> 00:10:12.460 using AI every day are fundamentally changing. 00:10:12.679 --> 00:10:14.799 We're seeing a massive demographic shift in real 00:10:14.799 --> 00:10:17.779 time. The original bro culture of ChatGPT is 00:10:17.779 --> 00:10:21.360 rapidly closing. At launch, roughly 80 % of all 00:10:21.360 --> 00:10:23.620 early users were men. It was heavily dominated 00:10:23.620 --> 00:10:26.399 by coders and tech enthusiasts. But new viral 00:10:26.399 --> 00:10:29.220 data shows a massive, rapid demographic shift. 00:10:29.440 --> 00:10:32.259 AI is truly becoming a universal human tool now. 00:10:32.440 --> 00:10:34.700 We are seeing the total normalization of AI. 00:10:35.120 --> 00:10:37.519 Everywhere. People who never coded a day in their 00:10:37.519 --> 00:10:39.919 lives are now power users. Google Chrome Skills 00:10:39.919 --> 00:10:42.759 is a perfect example of this accessibility. It 00:10:42.759 --> 00:10:45.360 lets everyday users save and instantly repeat 00:10:45.360 --> 00:10:48.360 complex AI workflows. You discover a great prompt 00:10:48.360 --> 00:10:50.299 sequence and you just save it. You don't have 00:10:50.299 --> 00:10:52.679 to be a dedicated prompt engineer anymore. You 00:10:52.679 --> 00:10:54.620 just click a button and the browser handles the 00:10:54.620 --> 00:10:57.059 heavy lifting. Then you have the new Gemini 3 00:10:57.059 --> 00:11:00.590 .1 Flash TTS. which brings incredibly realistic 00:11:00.590 --> 00:11:04.250 multi -speaker dialogue to the table it seamlessly 00:11:04.250 --> 00:11:07.389 supports over 70 distinct languages right and 00:11:07.389 --> 00:11:10.429 with inline audio tags too so you let users control 00:11:10.429 --> 00:11:13.450 the emotion you adjust the pacing and the tone 00:11:13.450 --> 00:11:16.639 of the synthetic voice It feels incredibly natural 00:11:16.639 --> 00:11:20.240 and intuitive for regular people to use. And 00:11:20.240 --> 00:11:23.559 XPilot is becoming incredibly useful for modern 00:11:23.559 --> 00:11:26.059 educators. Oh, yeah. It takes plain text documents 00:11:26.059 --> 00:11:28.879 and turns them into accurate video courses. It's 00:11:28.879 --> 00:11:31.399 perfect for users who need to explain complex 00:11:31.399 --> 00:11:34.000 things clearly. It totally removes the inherent 00:11:34.000 --> 00:11:36.639 risk of dangerous AI hallucinations. Because 00:11:36.639 --> 00:11:39.200 it strictly grounds the video generation in your 00:11:39.200 --> 00:11:41.919 provided text. It doesn't invent new facts. It 00:11:41.919 --> 00:11:44.299 just translates your document into a highly engaged... 00:11:45.300 --> 00:11:48.399 These tools are deeply embedding themselves into 00:11:48.399 --> 00:11:50.480 our daily routines. They're no longer standalone 00:11:50.480 --> 00:11:52.620 websites you have to intentionally visit. You 00:11:52.620 --> 00:11:54.960 don't have to open a separate tab to access the 00:11:54.960 --> 00:11:57.279 intelligence. It is just baked into the browser 00:11:57.279 --> 00:12:01.340 you already use. Beat. As demographics broaden 00:12:01.340 --> 00:12:03.940 and tools get embedded into Chrome, does the 00:12:03.940 --> 00:12:06.759 concept of using AI eventually disappear entirely? 00:12:07.179 --> 00:12:09.080 I agree with that completely. It's just like 00:12:09.080 --> 00:12:11.460 Wi -Fi. Soon people won't even think about the 00:12:11.460 --> 00:12:13.759 underlying AI. They'll just blindly expect their 00:12:13.759 --> 00:12:16.159 digital tools to be smart. Yeah, the technology 00:12:16.159 --> 00:12:18.519 just vanishes into the background of how computers 00:12:18.519 --> 00:12:21.559 work. Sponsor it. We are witnessing the true 00:12:21.559 --> 00:12:24.820 industrialization of AI today. it has rapidly 00:12:24.820 --> 00:12:28.500 matured from a simple fascinating novelty into 00:12:28.500 --> 00:12:31.100 a foundational utility for our entire modern 00:12:31.100 --> 00:12:34.039 society we have models solving legendary 60 year 00:12:34.039 --> 00:12:36.279 old math problems with pure elegance we have 00:12:36.279 --> 00:12:39.259 agentic systems writing incredibly complex code 00:12:39.259 --> 00:12:42.409 completely independently This reality is forcing 00:12:42.409 --> 00:12:44.789 the tech industry to radically adapt. They absolutely 00:12:44.789 --> 00:12:47.409 must adopt usage -based billing just to survive 00:12:47.409 --> 00:12:50.029 the immense compute costs. You simply cannot 00:12:50.029 --> 00:12:52.470 offer flat rates for infinite digital electricity. 00:12:52.789 --> 00:12:54.850 At the same time, the sheer scale of generation 00:12:54.850 --> 00:12:57.470 is overwhelming. It is forcing giant companies 00:12:57.470 --> 00:13:00.070 like Google to react defensively. They have to 00:13:00.070 --> 00:13:03.870 use advanced AI to police malicious AI. It fundamentally 00:13:03.870 --> 00:13:06.350 shifts the ongoing war against Internet scammers. 00:13:06.429 --> 00:13:08.730 They are targeting the generated content instead 00:13:08.730 --> 00:13:11.330 of targeting the individual people. It is a complete 00:13:11.330 --> 00:13:14.309 paradigm shift in modern cybersecurity. Two sec 00:13:14.309 --> 00:13:16.529 silence. I want to leave you with the final thought 00:13:16.529 --> 00:13:20.049 to mull over. GPT 5 .4 produced a mathematical 00:13:20.049 --> 00:13:23.169 proof today. It was so remarkably elegant it 00:13:23.169 --> 00:13:26.629 broke a 60 year barrier. It literally made human 00:13:26.629 --> 00:13:29.950 mathematical experts weep. We always assumed 00:13:29.950 --> 00:13:32.210 machines would conquer us with brute force calculation. 00:13:32.669 --> 00:13:34.970 We never expected them to conquer us with pure 00:13:34.970 --> 00:13:37.590 elegance. What happens to human intuition and 00:13:37.590 --> 00:13:40.710 other deeply specialized fields? Are we slowly 00:13:40.710 --> 00:13:43.730 losing our monopoly on elegance? Or is AI simply 00:13:43.730 --> 00:13:46.110 pushing us to ask much harder questions? Stay 00:13:46.110 --> 00:13:48.409 deeply curious out there. Keep a close eye on 00:13:48.409 --> 00:13:50.629 your token usage. And take some time to explore 00:13:50.629 --> 00:13:52.990 the tools we discussed today. Until next time.