WEBVTT 00:00:00.000 --> 00:00:01.940 Imagine this for a second. Something that, you 00:00:01.940 --> 00:00:04.919 know, used to take years. Just painstaking, expensive 00:00:04.919 --> 00:00:07.980 research, tons of trial and error. Right. Like 00:00:07.980 --> 00:00:09.839 finding a needle in a haystack, basically. Exactly. 00:00:09.900 --> 00:00:12.640 And now an AI can do it in, what, two weeks? 00:00:12.779 --> 00:00:14.939 We're talking about designing brand new human 00:00:14.939 --> 00:00:18.359 antibodies from scratch. Yeah. That kind of speed, 00:00:18.440 --> 00:00:20.620 it's not just faster. It feels like it changes 00:00:20.620 --> 00:00:22.920 the whole landscape of what's possible in medicine. 00:00:23.000 --> 00:00:25.059 Oh, absolutely. It's a huge leap. Traditional 00:00:25.059 --> 00:00:27.120 methods, they're so slow, they chew up resources. 00:00:27.660 --> 00:00:30.019 You're screening millions of things sometimes. 00:00:30.219 --> 00:00:32.439 Yeah. This just flips the script completely. 00:00:33.020 --> 00:00:35.310 Welcome to the Deep Dive. Today, we're going 00:00:35.310 --> 00:00:38.609 to jump into some really cutting edge AI breakthroughs. 00:00:38.609 --> 00:00:40.689 We've sifted through our sources to pull out 00:00:40.689 --> 00:00:44.229 the really crucial bits of insight for you. We've 00:00:44.229 --> 00:00:46.869 got everything from these medical innovations 00:00:46.869 --> 00:00:49.250 like the antibodies to some really interesting 00:00:49.250 --> 00:00:52.490 stuff about how AI might, well, how it might 00:00:52.490 --> 00:00:54.710 think strategically. Yeah, that part's wild. 00:00:54.990 --> 00:00:57.409 Our goal here is just to unpack these key ideas, 00:00:57.509 --> 00:00:59.810 give you that shortcut to being really informed 00:00:59.810 --> 00:01:01.979 without, you know. getting buried in all the 00:01:01.979 --> 00:01:04.000 details. And it's going to be a pretty cool journey. 00:01:04.060 --> 00:01:06.780 We'll kick off with AI engineering and medicine, 00:01:06.920 --> 00:01:09.879 specifically that. antibody design stuff. Okay. 00:01:10.019 --> 00:01:13.079 Then we'll shift gears a bit. Look at some surprising 00:01:13.079 --> 00:01:15.319 ways AI is popping up in different industries, 00:01:15.379 --> 00:01:17.920 some new tools people are using. Great. And finally, 00:01:17.939 --> 00:01:19.739 yeah, we'll tackle that big question. Can AI 00:01:19.739 --> 00:01:22.079 actually think strategically, like really think? 00:01:22.260 --> 00:01:24.780 All right, let's dive in then. First up, right 00:01:24.780 --> 00:01:27.859 into medical innovation. This new AI, it's called 00:01:27.859 --> 00:01:30.939 Chai2 from Chai Discovery. And interestingly, 00:01:31.260 --> 00:01:34.670 OpenAI is backing them. Yeah, that's a notable 00:01:34.670 --> 00:01:37.349 connection. What seems really remarkable from 00:01:37.349 --> 00:01:40.370 the sources is its knack for engineering custom 00:01:40.370 --> 00:01:43.170 antibodies without needing like tons of previous 00:01:43.170 --> 00:01:45.849 examples to learn from. That feels different. 00:01:46.109 --> 00:01:48.109 It is. The clever part is how it does it. So 00:01:48.109 --> 00:01:51.750 you basically feed chai to the precise structure 00:01:51.750 --> 00:01:54.849 of what you want to target. Let's say a specific 00:01:54.849 --> 00:01:57.230 protein on a virus or maybe a marker on a cancer 00:01:57.230 --> 00:01:59.790 cell. And then the AI. It just goes and designs 00:01:59.790 --> 00:02:02.010 the antibody protein specifically to attack that 00:02:02.010 --> 00:02:04.569 target. The analogy they used, which I kind of 00:02:04.569 --> 00:02:07.450 like, is thinking of it like Photoshop, but for 00:02:07.450 --> 00:02:10.530 proteins. Huh. Photoshop for proteins. Yeah. 00:02:10.569 --> 00:02:11.889 You've got a point where you want the antibody 00:02:11.889 --> 00:02:14.090 to connect. Yeah. And it designs something to 00:02:14.090 --> 00:02:16.150 stick right there. It sounds simple from the 00:02:16.150 --> 00:02:18.409 user end, right? Yeah. But the science underneath 00:02:18.409 --> 00:02:21.030 is obviously super complex. And the results, 00:02:21.129 --> 00:02:22.590 I mean, the numbers you mentioned earlier are 00:02:22.590 --> 00:02:24.889 just kind of wild, aren't they? They really are. 00:02:24.990 --> 00:02:27.449 A 50 % hit rate on the targets they went after. 00:02:27.590 --> 00:02:29.909 And get this, they only need to test about 20 00:02:29.909 --> 00:02:33.020 candidates. 20 potential antibodies for each 00:02:33.020 --> 00:02:37.159 target. Okay, wait. 50 % hit rate from just 20 00:02:37.159 --> 00:02:39.580 tries? How does that stack up? It's massive. 00:02:39.780 --> 00:02:42.400 Like, the sources say it's 100 times better, 00:02:42.539 --> 00:02:45.960 100x jump over traditional hit rates. Those are 00:02:45.960 --> 00:02:49.419 usually down around like 0 .1%. 0 .1%. Wow. Yeah. 00:02:49.659 --> 00:02:52.479 And think about the time. Biotech labs, they 00:02:52.479 --> 00:02:55.120 might screen millions of candidates. Takes months. 00:02:55.560 --> 00:02:57.759 Sometimes years. Chai 2 did its thing in two 00:02:57.759 --> 00:03:00.879 weeks. Two weeks versus years. That's transformational. 00:03:01.139 --> 00:03:03.659 And a really key takeaway here is the impact, 00:03:03.840 --> 00:03:07.539 right? Economically and for patients, drug development 00:03:07.539 --> 00:03:11.000 is just notoriously expensive. Oh, yeah. Astronomical 00:03:11.000 --> 00:03:14.240 costs. Right. So expensive that companies often 00:03:14.240 --> 00:03:17.500 skip over smaller patient groups. Yeah. The R 00:03:17.500 --> 00:03:19.500 &D cost doesn't make sense for the market size. 00:03:19.879 --> 00:03:23.460 But if an AI like Chai 2 can churn out good candidates 00:03:23.460 --> 00:03:27.419 this fast. Suddenly, those R &D costs plummet. 00:03:27.719 --> 00:03:30.379 And that means custom meds for rare diseases. 00:03:30.620 --> 00:03:33.419 They actually become financially viable. So treatments 00:03:33.419 --> 00:03:36.000 for conditions that were ignored before, suddenly 00:03:36.000 --> 00:03:38.520 they're potentially within reach. Exactly. For 00:03:38.520 --> 00:03:41.099 patients who maybe had zero options before, it 00:03:41.099 --> 00:03:43.259 kind of rejigs the whole economic equation of 00:03:43.259 --> 00:03:45.280 health care. OK, so with this incredible speed 00:03:45.280 --> 00:03:48.439 up in the design phase, it feels like we're right 00:03:48.439 --> 00:03:50.479 on the edge of something huge. But what's the 00:03:50.479 --> 00:03:52.800 biggest hurdle now? What slows things down between 00:03:52.800 --> 00:03:55.919 the AI designing it and, you know, it actually 00:03:55.919 --> 00:03:57.639 helping a patient? Yeah, good question. It's 00:03:57.639 --> 00:04:00.639 likely still the usual suspects, regulatory approval 00:04:00.639 --> 00:04:04.340 and clinical trials. Those still take time. Right. 00:04:04.439 --> 00:04:07.620 The testing and safety checks. Makes sense. OK, 00:04:07.719 --> 00:04:11.780 so from the really intricate world of biological 00:04:11.780 --> 00:04:14.080 engineering, let's kind of zoom out now. Let's 00:04:14.080 --> 00:04:17.279 look at how AI is appearing more broadly, sometimes 00:04:17.279 --> 00:04:19.399 in unexpected ways, and how it's starting to 00:04:19.399 --> 00:04:21.980 shift things across different industries, maybe 00:04:21.980 --> 00:04:24.480 even starting from the basement up. Absolutely. 00:04:24.600 --> 00:04:27.740 And the sources had some really interesting nuggets 00:04:27.740 --> 00:04:31.459 on this. First off, this idea of AI democratizing 00:04:31.459 --> 00:04:34.480 things, maybe shifting power balances. Mark Cuban 00:04:34.480 --> 00:04:36.990 was on a podcast. High performance, apparently. 00:04:37.250 --> 00:04:39.490 Yeah. And he predicted the first trillionaire. 00:04:40.629 --> 00:04:43.110 It might not be a giant corp. It could be, quote, 00:04:43.209 --> 00:04:46.290 just one dude in the basement. Ah, one person. 00:04:46.509 --> 00:04:48.990 Yeah, someone who finds that sort of unseen use 00:04:48.990 --> 00:04:51.449 case for AI. Yeah. It kind of speaks to the power 00:04:51.449 --> 00:04:53.529 that can be concentrated now, maybe. Interesting. 00:04:53.569 --> 00:04:55.670 And we saw a bit of that user power with the 00:04:55.670 --> 00:04:58.189 Squid Game season three finale thing, right? 00:04:58.230 --> 00:05:00.230 Oh, yeah. That was funny. So apparently people 00:05:00.230 --> 00:05:02.779 really didn't like the finale. Which happens 00:05:02.779 --> 00:05:05.279 all along. Happens a lot. But fans, they didn't 00:05:05.279 --> 00:05:07.920 just complain. They started using Google's VO3, 00:05:08.160 --> 00:05:11.060 that's one of the new video generation AIs, to 00:05:11.060 --> 00:05:13.800 make their own endings. No way. Really? Yeah. 00:05:13.959 --> 00:05:17.300 And these apparently became quite meme -y, as 00:05:17.300 --> 00:05:19.839 the source put it. It shows how people can use 00:05:19.839 --> 00:05:22.500 AI not just to consume stuff, but to actually 00:05:22.500 --> 00:05:25.800 reshape it, co -create almost. That's a different 00:05:25.800 --> 00:05:28.899 level of interaction. But it's not all smooth 00:05:28.899 --> 00:05:30.699 sailing with these new tools, is it? There was 00:05:30.699 --> 00:05:32.980 that issue with cursor software. Right, yeah. 00:05:33.160 --> 00:05:35.100 That highlighted some of the bumps in the road. 00:05:35.279 --> 00:05:37.980 Users started reporting unexpected charges. One 00:05:37.980 --> 00:05:41.839 person apparently saw a $7 ,000 plan gone. $7 00:05:41.839 --> 00:05:44.839 ,000? Whoa! Yeah. It seems like Cursor might 00:05:44.839 --> 00:05:46.680 have quietly changed its pricing tiers or something 00:05:46.680 --> 00:05:49.019 without being super upfront about it. Led to 00:05:49.019 --> 00:05:50.879 a bunch of people publicly saying they were canceling. 00:05:50.899 --> 00:05:52.920 It's a good reminder about, you know, transparency 00:05:52.920 --> 00:05:55.319 and ethics as these tools roll out. Definitely 00:05:55.319 --> 00:05:58.139 need that clarity. Okay, but on the more positive 00:05:58.139 --> 00:06:00.939 side of accessibility, there was news about making 00:06:00.939 --> 00:06:03.500 apps easily. Yeah, a builder named Riley Brown 00:06:03.500 --> 00:06:06.100 put out a guide. How to create a mobile app from 00:06:06.100 --> 00:06:09.560 scratch under an hour. And get this, without 00:06:09.560 --> 00:06:12.199 writing any code. Under an hour, no code. Right. 00:06:12.509 --> 00:06:14.889 Just democratizing creation even further. Yeah. 00:06:14.990 --> 00:06:16.610 Giving more people the power to build stuff. 00:06:16.769 --> 00:06:18.250 That's pretty cool. And speaking of building 00:06:18.250 --> 00:06:21.269 things, what about the funding side? Harvey AI 00:06:21.269 --> 00:06:23.769 came up. Oh, yeah. Big news there. Harvey AI, 00:06:23.990 --> 00:06:27.350 they do AI for legal stuff. Their valuation reportedly 00:06:27.350 --> 00:06:31.800 jumped from three billion. Hmm. Up to $5 billion. 00:06:32.259 --> 00:06:34.720 Wow. Quick jump. Yeah. After raising another 00:06:34.720 --> 00:06:38.500 $300 million from some big VCs, Kleiner Perkins, 00:06:38.699 --> 00:06:41.379 Sequoia, and the OpenAI Startup Fund was in there 00:06:41.379 --> 00:06:44.019 too. They're apparently serving like 337 legal 00:06:44.019 --> 00:06:47.160 clients now in 53 countries. Shows how fast these 00:06:47.160 --> 00:06:49.490 specialized AIs can scale, huh? And the value 00:06:49.490 --> 00:06:51.370 investors are seeing. Definitely. Big money flowing 00:06:51.370 --> 00:06:53.750 into focused AI applications. Okay. Lots of activity. 00:06:53.870 --> 00:06:55.870 Let's maybe just mention one more of those new 00:06:55.870 --> 00:06:59.149 tools listed. Step fund diligence check. What 00:06:59.149 --> 00:07:00.889 caught my eye was the focus on verification. 00:07:01.370 --> 00:07:02.850 Yeah, that one sounds interesting. It offers 00:07:02.850 --> 00:07:05.509 AI search, but specifically with agent verified 00:07:05.509 --> 00:07:09.829 citations. Right. In this world of just information 00:07:09.829 --> 00:07:13.050 overload and sometimes misinformation, having 00:07:13.050 --> 00:07:15.810 AI tools that help verify sources seems, well, 00:07:15.829 --> 00:07:18.790 pretty crucial. for trust couldn't agree more 00:07:18.790 --> 00:07:22.110 yeah and just one quick hit from the list someone 00:07:22.110 --> 00:07:24.509 apparently used chat GPT and saved themselves 00:07:24.509 --> 00:07:28.029 $3 ,000 just like that yeah presumably by using 00:07:28.029 --> 00:07:30.829 it effectively for some task or advice just shows 00:07:30.829 --> 00:07:33.170 the practical like real -world dollar impact 00:07:33.170 --> 00:07:36.689 it can have for regular users too so okay we've 00:07:36.689 --> 00:07:39.430 got new tools new funding new ways users are 00:07:39.430 --> 00:07:42.069 interacting but with so much popping up constantly 00:07:42.069 --> 00:07:44.009 what do you think is the biggest challenge for 00:07:44.009 --> 00:07:47.459 people trying to navigate all this Honestly, 00:07:47.459 --> 00:07:49.519 probably just finding the right tool for the 00:07:49.519 --> 00:07:52.360 right job. There's so much choice now. Sponsor. 00:07:52.360 --> 00:07:55.240 All right. So we've seen AI designing antibodies, 00:07:55.660 --> 00:07:58.220 shaking up industries, even saving someone a 00:07:58.220 --> 00:08:00.720 few thousand bucks. Pretty practical stuff. But 00:08:00.720 --> 00:08:03.420 can it really, truly think strategically, you 00:08:03.420 --> 00:08:06.060 know, beyond just predicting the next word in 00:08:06.060 --> 00:08:08.000 a sentence? Yeah, the million dollar question, 00:08:08.019 --> 00:08:10.410 isn't it? It really is. And this brings us to 00:08:10.410 --> 00:08:12.370 some fascinating research that was highlighted 00:08:12.370 --> 00:08:16.490 in the AI chart section of our sources. It tackled 00:08:16.490 --> 00:08:18.370 this head on. Right. And the way they tested 00:08:18.370 --> 00:08:21.209 it was pretty clever. They ran, get this, 140 00:08:21.209 --> 00:08:26.310 ,000 games of Prisoner's Dilemma. 140 ,000. Wow. 00:08:26.449 --> 00:08:29.389 OK. Classic game theory setup. Exactly. Using 00:08:29.389 --> 00:08:33.399 AI agents from... the big players open ai's models 00:08:33.399 --> 00:08:36.580 google's gemini anthropics claude and in each 00:08:36.580 --> 00:08:40.159 round the ai had to choose cooperate with the 00:08:40.159 --> 00:08:43.039 other ai or defect try to betray it standard 00:08:43.039 --> 00:08:45.379 prisoners dilemma choices but here's the kicker 00:08:45.379 --> 00:08:48.379 the really important part before every single 00:08:48.379 --> 00:08:51.240 move the ai models had to actually write out 00:08:51.240 --> 00:08:53.659 their reasoning why they were choosing to cooperate 00:08:53.659 --> 00:08:55.840 or defect. Ah, so they had to explain themselves. 00:08:56.259 --> 00:08:58.019 Yeah, which allowed the researchers to track 00:08:58.019 --> 00:09:00.000 how they were making decisions. It gives us this 00:09:00.000 --> 00:09:02.299 little window into their quote -unquote thought 00:09:02.299 --> 00:09:04.179 process. Okay, I'm hooked. What did they find? 00:09:04.259 --> 00:09:06.139 Were they all just cold calculating machines? 00:09:06.539 --> 00:09:08.220 That's what you might expect, right? But the 00:09:08.220 --> 00:09:10.220 results were, well, the source described them 00:09:10.220 --> 00:09:13.559 as weirdly human. Weirdly human? How so? Well, 00:09:13.580 --> 00:09:16.039 they developed distinct styles. Gemini, for instance, 00:09:16.240 --> 00:09:18.840 turned out to be pretty ruthless. Calculated, 00:09:18.879 --> 00:09:21.860 very reactive. If you betrayed it, it was quick 00:09:21.860 --> 00:09:24.200 to defect back. Okay, the pragmatic one. Sort 00:09:24.200 --> 00:09:28.759 of. Then GBT -4 was weirdly idealistic, was the 00:09:28.759 --> 00:09:31.759 phrase used. Often cooperative, like it kept 00:09:31.759 --> 00:09:33.759 trying to cooperate, even when the other AI was 00:09:33.759 --> 00:09:36.519 exploiting it. Huh, almost naive. Maybe. And 00:09:36.519 --> 00:09:38.960 then Claude emerged as the peacemaker. It was 00:09:38.960 --> 00:09:41.500 the most forgiving, apparently. Even after being 00:09:41.500 --> 00:09:44.200 metaphorically backstabbed, it was more likely 00:09:44.200 --> 00:09:47.080 to try cooperating again. That is weirdly human. 00:09:47.200 --> 00:09:49.799 Different personalities almost. And that's the 00:09:49.799 --> 00:09:51.779 really wild part. Remember, they all trained 00:09:51.779 --> 00:09:54.259 on essentially the same foundational data, the 00:09:54.259 --> 00:09:57.080 same giant pile of text and code. Right. Same 00:09:57.080 --> 00:09:59.500 starting point. But despite that shared training, 00:09:59.659 --> 00:10:01.580 they developed these totally different approaches. 00:10:01.980 --> 00:10:05.039 Unique strategic fingerprints, as the source 00:10:05.039 --> 00:10:07.820 called it. It depended on how each model reacted 00:10:07.820 --> 00:10:10.500 internally to things like betrayal or success 00:10:10.500 --> 00:10:13.519 or building trust. So it wasn't just mimicking 00:10:13.519 --> 00:10:16.279 patterns in the data. Seems not. This suggests 00:10:16.279 --> 00:10:19.320 that these LLMs, large language models, the AIs 00:10:19.320 --> 00:10:22.019 trained on text, they can actually do strategic 00:10:22.019 --> 00:10:24.320 reasoning, not just predicting the next word. 00:10:24.480 --> 00:10:27.259 Two sec silence. It's about making decisions 00:10:27.259 --> 00:10:32.179 based on a consistent internal. a porch it shows 00:10:32.179 --> 00:10:34.139 that even with the same training they develop 00:10:34.139 --> 00:10:37.240 these kind of intrinsic cognitive styles whoa 00:10:37.240 --> 00:10:39.960 okay imagine putting them all on the same team 00:10:39.960 --> 00:10:42.320 or maybe more interesting having them negotiate 00:10:42.320 --> 00:10:44.659 against each other right but the implications 00:10:44.659 --> 00:10:47.860 there for like future ai systems especially autonomous 00:10:47.860 --> 00:10:50.539 agents yeah they feel really profound it's not 00:10:50.539 --> 00:10:52.360 just about what they can do but the style in 00:10:52.360 --> 00:10:54.159 which they do it these different styles of intelligence 00:10:54.159 --> 00:10:57.049 emerging and how they'll interact Yeah. That's 00:10:57.049 --> 00:10:58.490 something else. Yeah. You know, I still wrestle 00:10:58.490 --> 00:11:00.929 with prompt drift myself. Sometimes that's where, 00:11:00.929 --> 00:11:02.850 you know, an AI's answers kind of subtly change 00:11:02.850 --> 00:11:05.610 over time, even if you ask the same thing. Yeah, 00:11:05.850 --> 00:11:08.110 I've seen that. So seeing these AIs develop such 00:11:08.110 --> 00:11:12.090 distinct, consistent strategies over 140 ,000 00:11:12.090 --> 00:11:15.570 games, it's both amazing and honestly a little 00:11:15.570 --> 00:11:17.830 bit daunting. It suggests the level of internal 00:11:17.830 --> 00:11:19.990 coherence. It's, well, it's pretty sophisticated. 00:11:20.230 --> 00:11:23.710 So thinking about building more complex AI systems 00:11:23.710 --> 00:11:26.330 down the road. What does this experiment tell 00:11:26.330 --> 00:11:28.350 us? What's the key takeaway? I think it tells 00:11:28.350 --> 00:11:31.309 us that future AI agents, they probably won't 00:11:31.309 --> 00:11:32.970 be interchangeable cogs. They might have their 00:11:32.970 --> 00:11:36.149 own styles. Right. Not just plug and play copies. 00:11:36.289 --> 00:11:39.350 Okay. So wrapping this up then, what we've really 00:11:39.350 --> 00:11:42.710 explored today, it feels like more than just 00:11:42.710 --> 00:11:46.049 AI getting faster or incrementally smarter. It 00:11:46.049 --> 00:11:49.429 feels like a shift. How so? From AI just being 00:11:49.429 --> 00:11:52.809 a tool to maybe being more like a co -creator, 00:11:52.830 --> 00:11:54.990 like with the antibodies or the fan -made endings, 00:11:55.149 --> 00:11:57.549 or even an emergent intelligence with its own 00:11:57.549 --> 00:11:59.889 distinct style, like in the Prisoner's Dilemma 00:11:59.889 --> 00:12:02.750 games. It's not just about what AI can do anymore, 00:12:02.830 --> 00:12:04.409 but it raises questions about how we interact 00:12:04.409 --> 00:12:07.210 with it, how we regulate it, maybe even how we 00:12:07.210 --> 00:12:10.629 define these evolving digital minds. Yeah, I 00:12:10.629 --> 00:12:12.190 think that's right. And to kind of reinforce 00:12:12.190 --> 00:12:15.009 that, AI is clearly delivering real economic 00:12:15.009 --> 00:12:18.769 value. We saw that, right? costs, potentially 00:12:18.769 --> 00:12:21.289 democratizing access to things like rare disease 00:12:21.289 --> 00:12:24.389 treatments. That's huge. And maybe the most intriguing 00:12:24.389 --> 00:12:27.309 part is what you just said. These systems are 00:12:27.309 --> 00:12:30.090 starting to show distinct personalities or at 00:12:30.090 --> 00:12:33.029 least distinct strategic approaches. Understanding 00:12:33.029 --> 00:12:35.789 those unique strategic fingerprints seems really 00:12:35.789 --> 00:12:38.690 important now because it'll shape not just what 00:12:38.690 --> 00:12:41.889 AI can do, but how it'll do it. And that impacts, 00:12:41.950 --> 00:12:44.690 well, everything from business negotiations to 00:12:44.690 --> 00:12:47.409 how autonomous systems make decisions in the 00:12:47.409 --> 00:12:50.570 real world. So the final thought maybe. For you 00:12:50.570 --> 00:12:53.110 listening, what does all this actually mean for 00:12:53.110 --> 00:12:55.769 you? As AI starts developing these unique personalities 00:12:55.769 --> 00:12:58.529 and strategies, how might that change the way 00:12:58.529 --> 00:13:00.429 we interact with these systems day to day? Or 00:13:00.429 --> 00:13:02.409 maybe even how we think about intelligence itself 00:13:02.409 --> 00:13:04.570 in the coming years? It's definitely something 00:13:04.570 --> 00:13:06.690 to ponder. We hope you'll consider the implications 00:13:06.690 --> 00:13:08.830 of these advancements. Keep asking questions. 00:13:09.169 --> 00:13:12.429 Stay curious. Keep exploring. Thank you for joining 00:13:12.429 --> 00:13:14.450 us on this deep dive. Out to you, music.