WEBVTT 00:00:00.000 --> 00:00:02.319 Welcome back to the Deep Dive. Today we are looking 00:00:02.319 --> 00:00:04.700 at a website that, and I'm willing to bet on 00:00:04.700 --> 00:00:07.940 this, almost every single person listening has 00:00:07.940 --> 00:00:11.000 on their phone right now. Oh, for sure. You probably 00:00:11.000 --> 00:00:13.140 have a tab open for it on your work computer. 00:00:13.279 --> 00:00:17.539 And if you're like me, you have this. Very specific, 00:00:17.719 --> 00:00:20.420 slightly dreaded relationship with it. I know 00:00:20.420 --> 00:00:21.539 exactly what you're talking about. It's that 00:00:21.539 --> 00:00:23.899 place you go when you need a job or when you 00:00:23.899 --> 00:00:25.420 feel like you should be networking, whatever 00:00:25.420 --> 00:00:27.500 that actually means in practice. Right. It's 00:00:27.500 --> 00:00:30.140 a professional obligation. Exactly. You go there, 00:00:30.320 --> 00:00:32.840 you update your digital resume, you humbly brag 00:00:32.840 --> 00:00:37.039 about a promotion, and you try your best to ignore 00:00:37.039 --> 00:00:40.100 the 15 messages from recruiters who clearly haven't 00:00:40.100 --> 00:00:42.600 read a single word of your profile. You are talking 00:00:42.600 --> 00:00:45.689 about the blue app. LinkedIn. I am talking about 00:00:45.689 --> 00:00:47.649 LinkedIn. And look, I know what everyone is thinking. 00:00:47.789 --> 00:00:49.530 Great, we're going to talk about a job board 00:00:49.530 --> 00:00:51.950 for an hour. How thrilling. Yeah, riveting stuff. 00:00:52.149 --> 00:00:53.649 But here's the thing that really surprised me 00:00:53.649 --> 00:00:55.649 when we started digging into the stack of sources 00:00:55.649 --> 00:00:58.590 for this. We aren't just looking at a place to 00:00:58.590 --> 00:01:02.009 upload a PDF of your CV. No. Not at all. When 00:01:02.009 --> 00:01:04.489 you actually peel back the layers and we're looking 00:01:04.489 --> 00:01:06.849 at Wikipedia archives, financial reports from 00:01:06.849 --> 00:01:10.510 Microsoft, academic sociology papers, it turns 00:01:10.510 --> 00:01:13.530 out we're looking at something much, much bigger. 00:01:13.650 --> 00:01:16.750 It's not a job board. That is, I think, the first 00:01:16.750 --> 00:01:19.069 and most important misconception we need to clear 00:01:19.069 --> 00:01:21.500 up. right away. Right. It is not a job board. 00:01:21.579 --> 00:01:25.000 It's what might be the most comprehensive and 00:01:25.000 --> 00:01:28.680 honestly slightly terrifying digital map of the 00:01:28.680 --> 00:01:31.060 global economy that has ever been built. It's 00:01:31.060 --> 00:01:32.980 the perfect way to frame it. And that right there, 00:01:33.000 --> 00:01:35.739 that is the central tension of this entire deep 00:01:35.739 --> 00:01:38.560 dive. How so? Well. On the surface, to you and 00:01:38.560 --> 00:01:41.280 me, it's a utility. It's a tool. It's a place 00:01:41.280 --> 00:01:43.060 to keep in touch with old colleagues, maybe find 00:01:43.060 --> 00:01:45.700 a new gig. But if you look under the hood, as 00:01:45.700 --> 00:01:48.079 we're going to do today, it is a massive data 00:01:48.079 --> 00:01:51.040 extraction engine. And the scale is just, it's 00:01:51.040 --> 00:01:53.859 hard to process. As of November 2023, we're talking 00:01:53.859 --> 00:01:57.379 about over 1 billion members. 1 billion. Let 00:01:57.379 --> 00:01:59.560 that number sink in. That number is hard to wrap 00:01:59.560 --> 00:02:03.959 your head around. 1 billion human beings. 1 billion 00:02:03.959 --> 00:02:08.389 members across 200 countries. That is... a staggering, 00:02:08.629 --> 00:02:12.150 almost unimaginable amount of human capital data, 00:02:12.270 --> 00:02:15.469 all concentrated in a single database. So the 00:02:15.469 --> 00:02:18.009 question we should be asking isn't how do I optimize 00:02:18.009 --> 00:02:20.629 my profile with the right keywords? Exactly. 00:02:21.159 --> 00:02:24.259 The real question is, what is this machine doing 00:02:24.259 --> 00:02:26.379 with all of that information? Is it a social 00:02:26.379 --> 00:02:29.539 network? Is it a public utility? Or is it what 00:02:29.539 --> 00:02:31.919 they internally call the economic graph? The 00:02:31.919 --> 00:02:34.240 economic graph. It sounds like something out 00:02:34.240 --> 00:02:36.860 of a sci -fi novel. Or a very, very dry economics 00:02:36.860 --> 00:02:39.620 textbook. Right. But as we found in the research, 00:02:39.800 --> 00:02:42.840 that concept is the key. It's the skeleton key 00:02:42.840 --> 00:02:45.060 to understanding everything about this company. 00:02:45.240 --> 00:02:47.849 It really is. So here's our mission for this 00:02:47.849 --> 00:02:50.349 deep dive. We want to understand how this machine 00:02:50.349 --> 00:02:52.750 actually makes money because, spoiler alert, 00:02:52.949 --> 00:02:55.909 it is not just from those little ads for project 00:02:55.909 --> 00:02:58.430 management software you scroll past. Not by a 00:02:58.430 --> 00:03:00.490 long shot. We need to look at the sociology of 00:03:00.490 --> 00:03:02.689 it, how it fundamentally changes the way we hire 00:03:02.689 --> 00:03:04.830 people and the way we get hired. And we have 00:03:04.830 --> 00:03:06.789 to talk about the dark side. Oh, there's definitely 00:03:06.789 --> 00:03:09.610 a dark side. The privacy breaches, the data scraping 00:03:09.610 --> 00:03:14.569 wars, and the geopolitical mess of trying to 00:03:14.569 --> 00:03:18.129 run a global database in a world that is more 00:03:18.129 --> 00:03:20.210 fractured than ever. And we can't ignore the 00:03:20.210 --> 00:03:22.849 security aspect. I mean, think about it. When 00:03:22.849 --> 00:03:25.810 you hold the professional identities of a billion 00:03:25.810 --> 00:03:28.569 people, you become a massive target. A target 00:03:28.569 --> 00:03:31.189 for who? For everyone. A target for hackers, 00:03:31.349 --> 00:03:34.449 for competitors, and even for spies. Spies on 00:03:34.449 --> 00:03:36.090 LinkedIn. Okay, we are definitely getting to 00:03:36.090 --> 00:03:37.409 that. But let's start at the beginning because 00:03:37.409 --> 00:03:39.870 LinkedIn feels like it has been around forever, 00:03:40.110 --> 00:03:41.830 right? Like it's just part of the internet's 00:03:41.830 --> 00:03:43.810 furniture. It does feel that way. But it actually 00:03:43.810 --> 00:03:46.030 has a very specific origin story that explains 00:03:46.030 --> 00:03:48.930 a lot about its corporate DNA. And it all goes 00:03:48.930 --> 00:03:51.090 back to a group of people we've talked about 00:03:51.090 --> 00:03:53.689 before on this show, the PayPal mafia. It is 00:03:53.689 --> 00:03:56.569 fascinating, isn't it? How all roads in Silicon 00:03:56.569 --> 00:04:00.020 Valley seem to lead back to PayPal. You had Elon 00:04:00.020 --> 00:04:03.599 Musk, Peter Thiel, Max Levchin. All the big names? 00:04:03.800 --> 00:04:05.759 All of them. But the key figure for LinkedIn 00:04:05.759 --> 00:04:08.620 is Reid Hoffman. Reid Hoffman. He was the executive 00:04:08.620 --> 00:04:11.099 VP at PayPal, right? That was his role. Correct. 00:04:11.159 --> 00:04:14.259 And before that, he had actually started a dating 00:04:14.259 --> 00:04:17.310 site. called socialnet .com. I did not know that. 00:04:17.550 --> 00:04:19.730 Yeah, it failed, but it taught him a lot about 00:04:19.730 --> 00:04:22.629 matching algorithms and social networks. So in 00:04:22.629 --> 00:04:26.550 December 2002, right after eBay bought PayPal 00:04:26.550 --> 00:04:29.389 and everyone got very, very rich, Hoffman gathered 00:04:29.389 --> 00:04:31.750 a team. And it wasn't just him? No, it was a 00:04:31.750 --> 00:04:34.509 team of people from PayPal and from that old 00:04:34.509 --> 00:04:37.509 social net company. People like Alan Blue, Eric 00:04:37.509 --> 00:04:41.089 Lee, Jean -Luc Fignon. They sat down to build 00:04:41.089 --> 00:04:43.970 something that was intentionally not about social. 00:04:44.439 --> 00:04:47.040 In the fun sense. So not Friendster or Myspace. 00:04:47.319 --> 00:04:49.579 Exactly. It was about social in the utility sense. 00:04:49.779 --> 00:04:51.620 They wanted to map professional relationships. 00:04:51.980 --> 00:04:54.399 They launched on May 5, 2003. And I think it's 00:04:54.399 --> 00:04:56.199 really important to contextualize this. This 00:04:56.199 --> 00:04:58.560 is before Facebook. A year before Facebook. Before 00:04:58.560 --> 00:05:01.480 YouTube. This is the very, very early days of 00:05:01.480 --> 00:05:04.589 what we now call Web 2 .0. But unlike the other 00:05:04.589 --> 00:05:06.870 social networks that just exploded overnight, 00:05:07.129 --> 00:05:09.029 LinkedIn was kind of a slow burn, wasn't it? 00:05:09.350 --> 00:05:11.149 Incredibly slow. I mean, in the startup world 00:05:11.149 --> 00:05:13.310 today, if you don't have a million users in a 00:05:13.310 --> 00:05:14.910 month, you're considered dead on arrival. Yeah, 00:05:14.910 --> 00:05:17.389 you're a failure. It took LinkedIn over a year 00:05:17.389 --> 00:05:21.209 until August 2004 just to reach one million users. 00:05:21.930 --> 00:05:24.649 They weren't chasing that viral loop of teenagers 00:05:24.649 --> 00:05:27.490 posting party photos. They were trying to get 00:05:27.490 --> 00:05:30.029 serious, busy professionals to sit down and upload 00:05:30.029 --> 00:05:32.670 their entire resume. That is a much harder ask. 00:05:32.829 --> 00:05:36.170 It's work. It's a huge ask. But because they 00:05:36.170 --> 00:05:38.810 were targeting professionals from day one, the 00:05:38.810 --> 00:05:41.129 economics were completely different. They weren't 00:05:41.129 --> 00:05:43.470 just burning through venture capital cash, hoping 00:05:43.470 --> 00:05:45.410 to figure out a business model later. Which was 00:05:45.410 --> 00:05:47.649 the model for almost everyone else at the time. 00:05:47.730 --> 00:05:50.399 It was. This is a critical distinction. They 00:05:50.399 --> 00:05:52.819 hit their first month of profitability in March 00:05:52.819 --> 00:05:56.500 2006. Wait, 2006? That's only three years after 00:05:56.500 --> 00:05:59.120 they launched. Three years. Think about that. 00:05:59.339 --> 00:06:02.939 Twitter, or X, took over a decade to turn a consistent 00:06:02.939 --> 00:06:04.839 profit. Uber is another one. It took forever. 00:06:05.079 --> 00:06:06.860 LinkedIn was in the black almost immediately. 00:06:07.240 --> 00:06:09.600 They realized very early on that if you have 00:06:09.600 --> 00:06:11.459 high -quality data on high -quality professionals, 00:06:11.839 --> 00:06:15.000 other people, recruiters, salespeople, will pay 00:06:15.000 --> 00:06:17.339 a premium for access to it. So while everyone 00:06:17.339 --> 00:06:19.519 else was trying to sell ads for sneakers and 00:06:19.519 --> 00:06:22.439 soda, LinkedIn was methodically building this 00:06:22.439 --> 00:06:25.660 incredibly valuable database of the global workforce. 00:06:26.000 --> 00:06:28.420 Precisely. Fast forward a few years. They are 00:06:28.420 --> 00:06:31.500 growing. They're opening offices in India, Ireland, 00:06:31.759 --> 00:06:34.639 Australia. It's becoming a global thing. And 00:06:34.639 --> 00:06:38.360 then we get to 2011, the IPO. The IPO moment. 00:06:38.500 --> 00:06:40.720 This seems like the moment the rest of the world, 00:06:40.860 --> 00:06:43.079 especially Wall Street. woke up and realized, 00:06:43.259 --> 00:06:45.819 oh, this isn't just a boring resume site. It 00:06:45.819 --> 00:06:48.600 was a massive wake -up call. They filed in January 00:06:48.600 --> 00:06:52.139 2011 and started trading in May under the ticker 00:06:52.139 --> 00:06:55.819 LNKD. And on the very first day of trading, the 00:06:55.819 --> 00:07:01.160 stock rose 171%. 171 % in a single day. That 00:07:01.160 --> 00:07:04.600 is unheard of. It closed at $94 .25, more than 00:07:04.600 --> 00:07:06.600 double the initial offering price. It was the 00:07:06.600 --> 00:07:09.680 biggest tech IPO since Google. Wow. And the reason 00:07:09.680 --> 00:07:11.620 investors went absolutely crazy for it wasn't 00:07:11.620 --> 00:07:14.220 because the website was fun to use or had amazing 00:07:14.220 --> 00:07:17.139 engagement metrics. No, it's pretty dry. It was 00:07:17.139 --> 00:07:19.779 because they saw the moat. They saw that LinkedIn 00:07:19.779 --> 00:07:22.259 had convinced millions and millions of professionals 00:07:22.259 --> 00:07:24.920 to voluntarily maintain and update their own 00:07:24.920 --> 00:07:27.040 data for free. They outsourced their data entry 00:07:27.040 --> 00:07:30.199 to the entire world. Essentially, yes. And by 00:07:30.199 --> 00:07:32.459 2011, they were actually earning more from advertising 00:07:32.459 --> 00:07:35.060 revenue than Twitter did, which shocks most people. 00:07:35.220 --> 00:07:37.300 That is shocking. I always think of Twitter as 00:07:37.300 --> 00:07:40.769 the ad machine from that era. But I guess high 00:07:40.769 --> 00:07:44.410 value targets, you know, CEOs, VPs, decision 00:07:44.410 --> 00:07:47.509 makers are worth a lot more to advertisers than 00:07:47.509 --> 00:07:50.069 random teenagers. An order of magnitude more. 00:07:50.189 --> 00:07:52.430 But that independent era didn't last forever. 00:07:52.970 --> 00:07:54.870 We have to talk about the biggest shift in the 00:07:54.870 --> 00:07:57.490 company's entire history. The moment they got 00:07:57.490 --> 00:07:59.930 swallowed by a giant. The Microsoft acquisition. 00:07:59.990 --> 00:08:02.009 I remember when this happened in 2016 and the 00:08:02.009 --> 00:08:04.629 price tag was just, it was eye -watering. June 00:08:04.629 --> 00:08:07.689 2016, Microsoft announces they're acquiring LinkedIn 00:08:07.689 --> 00:08:11.449 for $196 per share. The total deal was valued 00:08:11.449 --> 00:08:16.490 at $26 .2 billion. $26 .2 billion in cash. All 00:08:16.490 --> 00:08:18.670 cash. Even for a company the size of Microsoft, 00:08:18.870 --> 00:08:22.310 that is a massive check to write. Why? What was 00:08:22.310 --> 00:08:25.470 the strategic logic there? At the time, it was 00:08:25.470 --> 00:08:28.290 Microsoft's largest acquisition ever. It's only 00:08:28.290 --> 00:08:30.290 been surpassed recently by the Activision Blizzard 00:08:30.290 --> 00:08:33.409 deal. Right. To understand the why, you have 00:08:33.409 --> 00:08:35.929 to understand Satya Nadella's strategy for Microsoft. 00:08:36.750 --> 00:08:39.330 Microsoft already owned the professional cloud. 00:08:39.549 --> 00:08:44.029 They have Office, Word, Excel, Outlook. That's 00:08:44.029 --> 00:08:46.269 the tool set of work. It's what you use to do 00:08:46.269 --> 00:08:49.250 your job. Exactly. But they didn't own the network 00:08:49.250 --> 00:08:52.230 of work. They had the documents, but they didn't 00:08:52.230 --> 00:08:54.250 have the people. Oh, okay. That makes sense. 00:08:54.470 --> 00:08:56.090 Imagine the synergy they were pitching to their 00:08:56.090 --> 00:08:58.590 board. You're in Outlook. You're writing an email 00:08:58.590 --> 00:09:01.690 to a potential client. Microsoft wanted a world 00:09:01.690 --> 00:09:03.470 where you didn't just see an email address. You 00:09:03.470 --> 00:09:05.529 saw their entire LinkedIn profile right there 00:09:05.529 --> 00:09:07.110 in the sidebar. You'd see where they went to 00:09:07.110 --> 00:09:09.009 school, who you know in common, their recent 00:09:09.009 --> 00:09:11.750 posts. All of it. It was about integrating that 00:09:11.750 --> 00:09:14.529 social graph directly into the productivity suite. 00:09:14.669 --> 00:09:17.490 It connects the human context to the otherwise 00:09:17.490 --> 00:09:21.429 dry document or email. That was the vision. And 00:09:21.429 --> 00:09:24.350 surprisingly, for a deal this big, Microsoft 00:09:24.350 --> 00:09:27.549 took a very hands -off approach. That's not usually 00:09:27.549 --> 00:09:30.370 how these things go. No. Usually when a big company 00:09:30.370 --> 00:09:32.889 buys a smaller one, they gut it, they integrate 00:09:32.889 --> 00:09:34.789 it, and they completely destroy the culture. 00:09:34.870 --> 00:09:36.990 They absorb it. Right. Microsoft did the opposite. 00:09:37.070 --> 00:09:39.850 They let LinkedIn run as a distinct, independent 00:09:39.850 --> 00:09:43.570 entity. Jeff Wehner, the CEO at the time, stayed 00:09:43.570 --> 00:09:46.610 on and reported directly to Satya Nadella. They 00:09:46.610 --> 00:09:48.429 kept their brand, their culture, their offices. 00:09:48.509 --> 00:09:51.110 It was a very smart move. Speaking of Jeff Weiner, 00:09:51.210 --> 00:09:53.509 he seems like a really pivotal character here. 00:09:53.629 --> 00:09:55.990 He ran the company for 11 years before stepping 00:09:55.990 --> 00:09:59.049 down in 2020. He did. He became executive chairman 00:09:59.049 --> 00:10:02.049 and Ryan Roslansky stepped up as CEO. And in 00:10:02.049 --> 00:10:04.110 the research, it's Weiner who really pushed this 00:10:04.110 --> 00:10:07.710 concept of the economic graph. We dropped that 00:10:07.710 --> 00:10:09.370 term in the intro, but we need to really dig 00:10:09.370 --> 00:10:11.809 into it. What is the economic graph, practically 00:10:11.809 --> 00:10:14.330 speaking? So in 2012, long before the Microsoft 00:10:14.330 --> 00:10:17.899 deal. Jeff Weiner articulated his vision. He 00:10:17.899 --> 00:10:19.960 said he didn't want to just build a social network. 00:10:20.179 --> 00:10:22.919 He wanted to create a digital map of the entire 00:10:22.919 --> 00:10:26.039 global economy. A map of the economy? What does 00:10:26.039 --> 00:10:29.279 that even mean? He defined the nodes of this 00:10:29.279 --> 00:10:31.799 map very specifically. Nodes. OK, we're getting 00:10:31.799 --> 00:10:33.879 into graph theory here. It's simpler than it 00:10:33.879 --> 00:10:37.700 sounds. A node is just a connection point. Weiner 00:10:37.700 --> 00:10:39.960 said they needed to digitally represent and map 00:10:39.960 --> 00:10:43.039 every single job in the world. Every job. Every 00:10:43.039 --> 00:10:45.539 skill required for those jobs, every company, 00:10:45.639 --> 00:10:48.480 every professional, and every single educational 00:10:48.480 --> 00:10:51.820 institution. Okay, so it's not just John Smith 00:10:51.820 --> 00:10:54.539 works at Google. No, it's much deeper. It's John 00:10:54.539 --> 00:10:58.240 Smith, who is Node A, has Python skills, Node 00:10:58.240 --> 00:11:01.899 G, works at Google Node C, which is located in 00:11:01.899 --> 00:11:04.720 Mountain View, Node D, and is currently hiring 00:11:04.720 --> 00:11:07.960 for a product manager, Node E. which requires 00:11:07.960 --> 00:11:10.460 Python skills node B again. And all those nodes 00:11:10.460 --> 00:11:12.940 are interconnected. Exactly. And when you map 00:11:12.940 --> 00:11:14.840 all of those relationships at a scale of a billion 00:11:14.840 --> 00:11:17.059 people, you can start to do things that were 00:11:17.059 --> 00:11:19.039 previously impossible, science fiction even. 00:11:19.120 --> 00:11:21.559 Like what? The stated goal was to make the labor 00:11:21.559 --> 00:11:23.779 market more efficient by making it transparent, 00:11:24.080 --> 00:11:27.860 to remove friction. Imagine if a government or 00:11:27.860 --> 00:11:30.899 a company could see in real time that there is 00:11:30.899 --> 00:11:34.190 a shortage of 5 ,000 nurses in London. but a 00:11:34.190 --> 00:11:37.070 surplus of 5 ,000 nurses in Manchester. If you 00:11:37.070 --> 00:11:40.070 have that data, you can theoretically fix that 00:11:40.070 --> 00:11:42.389 inefficiency. You can target training programs 00:11:42.389 --> 00:11:44.830 or relocation incentives. That's the ambition. 00:11:44.970 --> 00:11:47.210 It's like they're trying to play SimCity with 00:11:47.210 --> 00:11:49.389 the actual global labor market. But to do that, 00:11:49.389 --> 00:11:51.330 you need incredible technology. You can't just 00:11:51.330 --> 00:11:53.289 run a standard database search on that level 00:11:53.289 --> 00:11:55.870 of complexity and scale. No, you can't. Which 00:11:55.870 --> 00:11:59.070 is why in 2014, they unveiled a new custom -built 00:11:59.070 --> 00:12:02.169 search architecture they called Killeen. Killeen. 00:12:02.190 --> 00:12:04.450 What makes that special? Well, think about a 00:12:04.450 --> 00:12:06.210 normal Google search. You type in a word, it 00:12:06.210 --> 00:12:08.629 finds documents with that word. That's a relatively 00:12:08.629 --> 00:12:10.990 easy problem to solve at scale. All right. But 00:12:10.990 --> 00:12:12.889 a graph search is fundamentally different. You're 00:12:12.889 --> 00:12:14.750 asking much more complex questions. You're asking 00:12:14.750 --> 00:12:16.490 things like, find me all of my second degree 00:12:16.490 --> 00:12:18.509 connections who live in the Bay Area, have worked 00:12:18.509 --> 00:12:21.250 at Apple in the past, no C++ art, and went to 00:12:21.250 --> 00:12:23.250 Stanford. That's a lot of different variables. 00:12:23.409 --> 00:12:27.529 It requires traversing this massive web of relationships 00:12:27.529 --> 00:12:31.480 almost instantly. Galeen is the custom built 00:12:31.480 --> 00:12:34.759 engine that parses that billion node web to find 00:12:34.759 --> 00:12:37.080 the needle in the haystack in milliseconds. So 00:12:37.080 --> 00:12:38.980 when I go to the LinkedIn search bar and start 00:12:38.980 --> 00:12:41.840 adding filters for people, company, and location, 00:12:42.240 --> 00:12:45.220 that's Galeen in the background sprinting through 00:12:45.220 --> 00:12:47.840 a billion connections to give me a list. That's 00:12:47.840 --> 00:12:51.340 Galeen. And that technology is what enables the 00:12:51.340 --> 00:12:54.779 entire business model. But, and here's where 00:12:54.779 --> 00:12:57.000 we need to get critical, building the map is 00:12:57.000 --> 00:12:59.740 one thing. How human beings actually behave on 00:12:59.740 --> 00:13:02.360 that map is something else entirely. Right, the 00:13:02.360 --> 00:13:04.200 human element. We have to look at the sociology 00:13:04.200 --> 00:13:06.259 of LinkedIn. This was the part of the research 00:13:06.259 --> 00:13:08.740 that I found most fascinating. Because we tend 00:13:08.740 --> 00:13:11.419 to think of getting a job as a meritocracy. You 00:13:11.419 --> 00:13:13.379 know, I have the right skills, therefore I get 00:13:13.379 --> 00:13:16.100 the job. But the academic research suggests it's 00:13:16.100 --> 00:13:18.799 a lot messier than that. It's incredibly messy. 00:13:18.960 --> 00:13:21.220 The source material brings up the weak ties theory. 00:13:21.399 --> 00:13:24.009 This is classic, isn't it? It's one of the most 00:13:24.009 --> 00:13:26.610 famous concepts in sociology, developed by Mark 00:13:26.610 --> 00:13:30.269 Granovetter way back in the 1970s. And LinkedIn 00:13:30.269 --> 00:13:33.190 is basically the industrialization of this theory. 00:13:33.350 --> 00:13:37.389 So what's the core idea? The idea is this. Your 00:13:37.389 --> 00:13:39.990 strong ties, your best friends, your family, 00:13:40.090 --> 00:13:42.409 your closest co -workers, they all inhabit the 00:13:42.409 --> 00:13:45.429 same world you do. They know the same people. 00:13:45.570 --> 00:13:47.970 They read the same news. They hear about the 00:13:47.970 --> 00:13:50.509 same job openings. So if I'm looking for a new 00:13:50.509 --> 00:13:53.059 opportunity. Asking my best friend is actually 00:13:53.059 --> 00:13:54.679 kind of inefficient because he probably only 00:13:54.679 --> 00:13:57.440 knows what I already know. Exactly. The information 00:13:57.440 --> 00:14:00.279 is redundant. But your weak ties, that's the 00:14:00.279 --> 00:14:02.460 key. The guy you met at a conference three years 00:14:02.460 --> 00:14:04.659 ago, the woman you worked with on one short project 00:14:04.659 --> 00:14:06.799 at a previous job. People on the periphery of 00:14:06.799 --> 00:14:08.639 your network. Right. They move in different circles. 00:14:08.799 --> 00:14:11.159 They have access to new information, information 00:14:11.159 --> 00:14:13.870 that you do not have. And research has proven 00:14:13.870 --> 00:14:17.690 this out on LinkedIn specifically. Yes. A study 00:14:17.690 --> 00:14:20.929 published in Research Policy in 2017 analyzed 00:14:20.929 --> 00:14:24.070 PhD holders on the platform. It confirmed that 00:14:24.070 --> 00:14:26.850 those who had larger, more diverse networks of 00:14:26.850 --> 00:14:30.169 weak ties were significantly more likely to transition 00:14:30.169 --> 00:14:33.350 successfully from academia into industry jobs. 00:14:33.960 --> 00:14:36.500 So the platform's real value isn't actually keeping 00:14:36.500 --> 00:14:39.259 in touch with your friends. It's hoarding acquaintances. 00:14:39.419 --> 00:14:42.120 It is about maintaining a low -maintenance digital 00:14:42.120 --> 00:14:45.600 Rolodex of people you barely know, because statistically, 00:14:45.799 --> 00:14:48.539 one of them is going to be your bridge to your 00:14:48.539 --> 00:14:51.210 next opportunity. But there's a darker side to 00:14:51.210 --> 00:14:53.570 this digital map. It's not a neutral map, is 00:14:53.570 --> 00:14:56.070 it? The research from sociologist Ofer Sharon 00:14:56.070 --> 00:14:58.289 talks about something called the filtration effect. 00:14:58.690 --> 00:15:01.409 This made me really rethink how we present ourselves 00:15:01.409 --> 00:15:04.250 online. Ofer Sharon's work is absolutely essential 00:15:04.250 --> 00:15:06.830 here. He didn't just look at the data. He interviewed 00:15:06.830 --> 00:15:09.210 unemployed workers and recruiters to understand 00:15:09.210 --> 00:15:11.690 how the hiring process actually works in the 00:15:11.690 --> 00:15:13.909 LinkedIn era. And what did he find? He found 00:15:13.909 --> 00:15:15.990 that platforms like this create all these new 00:15:15.990 --> 00:15:17.929 structural barriers that have almost nothing 00:15:17.929 --> 00:15:20.559 to do with your actual... ability to do the job. 00:15:20.679 --> 00:15:23.379 What kind of barriers? Visuals, for one. The 00:15:23.379 --> 00:15:26.419 profile picture. In the old days of paper resumes, 00:15:26.759 --> 00:15:29.159 at least in the U .S., you never included a photo. 00:15:29.340 --> 00:15:32.139 It was considered a way to avoid bias based on 00:15:32.139 --> 00:15:37.019 race, age, gender. Sure. On LinkedIn, a photo 00:15:37.019 --> 00:15:39.259 is basically mandatory if you want to be taken 00:15:39.259 --> 00:15:42.720 seriously. Sharon found that people get filtered 00:15:42.720 --> 00:15:46.080 out consciously or subconsciously, based on the 00:15:46.080 --> 00:15:49.039 perceived professionalism of their headshot. 00:15:49.200 --> 00:15:51.000 And also based on the personal narrative, the 00:15:51.000 --> 00:15:53.539 about section. Yes, how you construct your personal 00:15:53.539 --> 00:15:56.440 brand. So if you are an amazing engineer, but 00:15:56.440 --> 00:15:58.399 you're bad at taking selfies, and you're not 00:15:58.399 --> 00:16:00.440 great at writing marketing copy about yourself, 00:16:00.740 --> 00:16:03.100 you lose. You become invisible. You're invisible. 00:16:03.769 --> 00:16:05.950 The algorithm and the recruiters filter you out 00:16:05.950 --> 00:16:08.129 before they even look at your GitHub or your 00:16:08.129 --> 00:16:11.070 actual credentials. It exerts this massive pressure 00:16:11.070 --> 00:16:13.429 on every worker to become their own personal 00:16:13.429 --> 00:16:15.509 brand manager. You have to perform your professional 00:16:15.509 --> 00:16:18.009 identity. Constantly. It's a performance of employability. 00:16:18.129 --> 00:16:20.070 That explains all those posts we see, right? 00:16:20.190 --> 00:16:22.929 The I'm humbled and honored to announce. It's 00:16:22.929 --> 00:16:25.539 a performance because we all know. We're being 00:16:25.539 --> 00:16:28.159 watched and judged by the algorithm and by a 00:16:28.159 --> 00:16:31.179 hundred potential future employers. And another 00:16:31.179 --> 00:16:33.759 study from 2019 showed that this creates something 00:16:33.759 --> 00:16:36.960 called a bifurcated market. Bifurcated, meaning 00:16:36.960 --> 00:16:40.120 split in two. Yes. How does that work? The researchers 00:16:40.120 --> 00:16:42.799 found a stark class divide in how the platform 00:16:42.799 --> 00:16:46.220 is used. For high -skilled, high -demand workers, 00:16:46.559 --> 00:16:50.340 think AI engineers, senior executives, recruiters 00:16:50.340 --> 00:16:53.649 use LinkedIn to actively poach them. These are 00:16:53.649 --> 00:16:56.029 the passive candidates. Exactly. These people 00:16:56.029 --> 00:16:58.009 aren't even looking for jobs. The jobs come to 00:16:58.009 --> 00:17:00.549 them through in -mail. For them, the platform 00:17:00.549 --> 00:17:03.009 is a red carpet. Must be nice. But for everyone 00:17:03.009 --> 00:17:06.089 else, lower skilled workers, generalists, people 00:17:06.089 --> 00:17:08.789 in the administrative roles, LinkedIn works completely 00:17:08.789 --> 00:17:10.960 differently. Recruiters don't hunt for them. 00:17:11.039 --> 00:17:13.000 They just post an ad and let thousands of people 00:17:13.000 --> 00:17:15.380 apply and fight over it. So for that group, it's 00:17:15.380 --> 00:17:18.380 not a red carpet. It's a gladiator arena. That 00:17:18.380 --> 00:17:20.759 is a depressing but very accurate way to put 00:17:20.759 --> 00:17:22.599 it. And there was one more piece of research 00:17:22.599 --> 00:17:25.019 that totally contradicted the advice everyone 00:17:25.019 --> 00:17:27.740 gives you. We always hear, get a referral. A 00:17:27.740 --> 00:17:30.680 referral is the golden ticket. But the data says, 00:17:30.920 --> 00:17:34.259 maybe not. This was a fascinating study from 00:17:34.259 --> 00:17:36.720 the Foster School of Business. They found that 00:17:36.720 --> 00:17:39.720 job seekers were actually less likely to get 00:17:39.720 --> 00:17:42.200 a referral from a connection if they were in 00:17:42.200 --> 00:17:45.019 the same field. Wait, that seems completely backwards. 00:17:45.440 --> 00:17:47.859 If I'm a graphic designer, wouldn't I want to 00:17:47.859 --> 00:17:49.980 help another graphic designer get a job at my 00:17:49.980 --> 00:17:53.079 company? You would think so. But the researchers 00:17:53.079 --> 00:17:55.799 called it competition self -protection. Competition 00:17:55.799 --> 00:17:59.410 self -protection. If I refer you... and you do 00:17:59.410 --> 00:18:02.269 exactly what I do, subconsciously, I might be 00:18:02.269 --> 00:18:04.170 worried you'll replace me or you might compete 00:18:04.170 --> 00:18:06.089 with me for the best projects or for the next 00:18:06.089 --> 00:18:08.829 promotion. So people tend to hoard opportunities 00:18:08.829 --> 00:18:11.190 when it comes to their direct peers. So much 00:18:11.190 --> 00:18:13.849 for community. It's really, I'll help you as 00:18:13.849 --> 00:18:15.670 long as you're not a direct threat to my paycheck. 00:18:16.009 --> 00:18:17.950 You and nature doesn't magically change just 00:18:17.950 --> 00:18:19.869 because you put it on a website with a blue logo. 00:18:20.299 --> 00:18:23.539 Okay, so we've established the sociology, the 00:18:23.539 --> 00:18:25.859 human behavior on the map. Now let's talk about 00:18:25.859 --> 00:18:28.259 the actual product features and how they monetize 00:18:28.259 --> 00:18:30.500 this. Because I think most people don't realize 00:18:30.500 --> 00:18:34.039 that we, the users, are not the customers. That 00:18:34.039 --> 00:18:37.039 is rule number one of the internet. If you aren't 00:18:37.039 --> 00:18:39.140 paying for the product, you are the product. 00:18:39.380 --> 00:18:42.079 So who is the customer? LinkedIn has a few revenue 00:18:42.079 --> 00:18:45.359 streams, but the biggest one historically has 00:18:45.359 --> 00:18:48.579 always been talent solutions. Which is just a 00:18:48.579 --> 00:18:51.359 nice corporate phrase for selling our data to 00:18:51.359 --> 00:18:53.980 recruiters. That's exactly what it is. If you're 00:18:53.980 --> 00:18:56.559 a corporate recruiter, you pay thousands, sometimes 00:18:56.559 --> 00:18:59.460 tens of thousands of dollars a year for a special, 00:18:59.599 --> 00:19:02.160 super -powered version of LinkedIn called Recruiter. 00:19:02.259 --> 00:19:04.720 And what does that get you? It unlocks the full 00:19:04.720 --> 00:19:07.180 power of the economic graph. It allows you to 00:19:07.180 --> 00:19:09.240 see everyone's full profile, search with all 00:19:09.240 --> 00:19:10.839 those advanced filters we talked about thanks 00:19:10.839 --> 00:19:13.319 to Galeen, and crucially, message people who 00:19:13.319 --> 00:19:16.009 aren't in your network using InMail. That is 00:19:16.009 --> 00:19:18.390 the cash cow. They are selling the map back to 00:19:18.390 --> 00:19:19.990 the companies that are on the map. That's the 00:19:19.990 --> 00:19:22.589 business model. Then you have premium subscriptions 00:19:22.589 --> 00:19:25.289 for job seekers and sales professionals. And 00:19:25.289 --> 00:19:27.509 of course, advertising like sponsored updates. 00:19:27.769 --> 00:19:29.690 And it's a big business. The numbers are huge. 00:19:29.930 --> 00:19:34.230 Oh, yeah. In 2022, revenue hit $13 .8 billion. 00:19:35.009 --> 00:19:38.130 That was up from $10 .3 billion just the year 00:19:38.130 --> 00:19:42.329 before in 2021. The growth under Microsoft has 00:19:42.329 --> 00:19:45.460 been astronomical. And what about the features 00:19:45.460 --> 00:19:49.759 for us regular mortals? I feel like the whole 00:19:49.759 --> 00:19:51.859 definition of a connection has changed over the 00:19:51.859 --> 00:19:54.940 years. It has shifted dramatically. In the beginning, 00:19:55.079 --> 00:19:56.900 LinkedIn was what they called a gated network. 00:19:57.180 --> 00:19:59.839 Gated access. The philosophy was that you should 00:19:59.839 --> 00:20:01.680 only connect with people you actually know and 00:20:01.680 --> 00:20:04.589 trust. It was about replicating your real world 00:20:04.589 --> 00:20:07.250 trusted professional network online. I remember 00:20:07.250 --> 00:20:09.029 that. You used to have to know someone's email 00:20:09.029 --> 00:20:10.829 address to even send them an invitation. You 00:20:10.829 --> 00:20:12.750 couldn't just click a connect button. Exactly. 00:20:12.950 --> 00:20:15.710 But somewhere around 2017, they clearly realized 00:20:15.710 --> 00:20:18.769 that trusted networks don't grow very fast and 00:20:18.769 --> 00:20:21.990 social media is a game of scale. So they threw 00:20:21.990 --> 00:20:24.250 open the gates. And now you can have up to 30 00:20:24.250 --> 00:20:27.829 ,000 connections. 30 ,000, which completely changes 00:20:27.829 --> 00:20:30.349 the dynamic. It has moved from being a private 00:20:30.349 --> 00:20:33.329 Rolodex to a public broadcasting platform. That's 00:20:33.329 --> 00:20:36.150 why you see so many influencers and content creators 00:20:36.150 --> 00:20:39.990 on there now. And the endorsements feature. I 00:20:39.990 --> 00:20:41.789 have to laugh about this one. My profile says 00:20:41.789 --> 00:20:43.549 I'm endorsed for skills I haven't used since 00:20:43.549 --> 00:20:45.130 college from people I haven't spoken to in a 00:20:45.130 --> 00:20:47.150 decade. The endorsement feature is a perfect 00:20:47.150 --> 00:20:50.230 case study in gamification gone wrong. It was 00:20:50.230 --> 00:20:52.150 supposed to be a lightweight way to verify skills. 00:20:52.569 --> 00:20:54.849 But it just became a game. It became a game of... 00:20:55.130 --> 00:20:57.509 I'll endorse you for strategic planning if you 00:20:57.509 --> 00:21:00.289 endorse me for public speaking. It's reciprocal 00:21:00.289 --> 00:21:02.569 backscratching. It's meaningless. Pretty much. 00:21:02.809 --> 00:21:06.369 In 2016, LinkedIn tried to fix it by highlighting 00:21:06.369 --> 00:21:09.109 endorsements from mutual connections or colleagues 00:21:09.109 --> 00:21:11.410 from the same company, trying to add some social 00:21:11.410 --> 00:21:14.289 proof, some weight to it. But honestly, most 00:21:14.289 --> 00:21:15.990 recruiters I've spoken to say they completely 00:21:15.990 --> 00:21:18.619 ignore them. They're just vanity metrics. They 00:21:18.619 --> 00:21:20.980 also made a big push to become a media company 00:21:20.980 --> 00:21:24.220 with the Influencers Program. Yes, bringing in 00:21:24.220 --> 00:21:27.019 people like Bill Gates, Richard Branson, Arianna 00:21:27.019 --> 00:21:29.339 Huffington. This was a really smart play to increase 00:21:29.339 --> 00:21:31.900 time on site. Right, to keep you on the app longer. 00:21:32.180 --> 00:21:34.119 If you just go to LinkedIn to update your resume, 00:21:34.339 --> 00:21:37.039 you might visit once every six months. But if 00:21:37.039 --> 00:21:39.200 you go there to read Bill Gates' latest thoughts 00:21:39.200 --> 00:21:41.799 on climate change, you might visit every day. 00:21:42.279 --> 00:21:44.519 It turned the platform from a static profile 00:21:44.519 --> 00:21:48.140 page into a dynamic content feed. And then there's 00:21:48.140 --> 00:21:50.440 the whole education piece, LinkedIn Learning. 00:21:50.559 --> 00:21:52.599 Which was born from the acquisition of lynda 00:21:52.599 --> 00:21:57.680 .com back in 2015 for $1 .5 billion. A huge purchase 00:21:57.680 --> 00:21:59.960 at the time. It was, and it completes the loop 00:21:59.960 --> 00:22:02.579 for them. You look for a job on LinkedIn, you 00:22:02.579 --> 00:22:04.839 see you're missing a required skill, you pay 00:22:04.839 --> 00:22:06.859 LinkedIn to learn that skill through LinkedIn 00:22:06.859 --> 00:22:09.180 Learning, and then you apply for the job on LinkedIn. 00:22:09.200 --> 00:22:11.460 They want to own the entire life cycle of your 00:22:11.460 --> 00:22:14.799 career. From education to application to networking. 00:22:15.099 --> 00:22:17.859 The whole thing. And recently, in October 2025, 00:22:18.420 --> 00:22:20.759 they launched the Career Hub, which seems to 00:22:20.759 --> 00:22:23.200 be doubling down on this all -in -one ecosystem 00:22:23.200 --> 00:22:26.079 approach. But as they gather more and more of 00:22:26.079 --> 00:22:29.039 our data to power these features, they run into 00:22:29.039 --> 00:22:33.339 massive legal and ethical walls. And this brings 00:22:33.339 --> 00:22:35.140 us to the part of the story that makes me the 00:22:35.140 --> 00:22:37.940 most nervous. The privacy part. Yeah, let's get 00:22:37.940 --> 00:22:40.640 into the dark side. Because for a company that 00:22:40.640 --> 00:22:43.220 relies so heavily on the idea of professional 00:22:43.220 --> 00:22:45.960 trust, they have done some pretty shady things 00:22:45.960 --> 00:22:48.960 to grow. We have to talk about the Perkins lawsuit. 00:22:49.319 --> 00:22:52.579 This is a classic example of growth hacking crossing 00:22:52.579 --> 00:22:55.680 the line into illegality. In the early days, 00:22:55.859 --> 00:22:59.500 LinkedIn had that feature. To find friends, you'd 00:22:59.500 --> 00:23:01.440 grant them access to your email address book, 00:23:01.559 --> 00:23:03.799 and they would see which of your contacts were 00:23:03.799 --> 00:23:06.019 already on the platform. Which sounds standard. 00:23:06.059 --> 00:23:08.039 A lot of apps do that. The first part was standard. 00:23:08.259 --> 00:23:10.960 But what users didn't realize, because it was 00:23:10.960 --> 00:23:13.299 buried deep in the terms of service, was that 00:23:13.299 --> 00:23:15.759 LinkedIn would then take the email addresses 00:23:15.759 --> 00:23:18.500 of everyone in your contacts who was not on LinkedIn. 00:23:18.680 --> 00:23:21.660 Oh, no. And send them an invite email. And if 00:23:21.660 --> 00:23:23.500 they didn't join after the first email, they 00:23:23.500 --> 00:23:25.279 would send another one. And then a third one. 00:23:25.400 --> 00:23:27.200 And the worst part was that the emails looked 00:23:27.200 --> 00:23:29.940 like they came directly from me personally. Yes. 00:23:30.299 --> 00:23:32.660 The subject line was literally, host speaker 00:23:32.660 --> 00:23:36.339 has invited you to join LinkedIn. It felt like 00:23:36.339 --> 00:23:38.079 you were personally nagging your friends and 00:23:38.079 --> 00:23:41.319 colleagues. People were furious. They felt humiliated. 00:23:41.400 --> 00:23:43.599 I can imagine. So a class action lawsuit was 00:23:43.599 --> 00:23:46.240 filed. Perkins versus LinkedIn core. That's right. 00:23:46.359 --> 00:23:49.160 And the court argument was, I agreed to connect 00:23:49.160 --> 00:23:52.299 with my friends. I did not agree to spam my entire 00:23:52.299 --> 00:23:55.519 address book, including my grandma. And how did 00:23:55.519 --> 00:23:57.960 the court rule? The court agreed with the users. 00:23:58.220 --> 00:24:00.680 They ruled that while a user might have consented 00:24:00.680 --> 00:24:03.039 to that first invitation email, they absolutely 00:24:03.039 --> 00:24:05.420 did not consent to the follow -up reminder emails. 00:24:05.900 --> 00:24:09.960 LinkedIn settled in 2015 for $13 million. Which 00:24:09.960 --> 00:24:12.119 for them is a rounding error. A slap on the wrist 00:24:12.119 --> 00:24:14.579 financially. But it was a huge reputational hit, 00:24:14.680 --> 00:24:16.940 and it forced them to stop that specific aggressive 00:24:16.940 --> 00:24:19.279 behavior. But that was about how they misused 00:24:19.279 --> 00:24:21.980 internal data. There's also this huge war over 00:24:21.980 --> 00:24:24.039 who owns the data once it's made public, the 00:24:24.039 --> 00:24:26.819 HiQ Labs versus LinkedIn case. This one feels 00:24:26.819 --> 00:24:28.799 like it sets a massive precedent for the entire 00:24:28.799 --> 00:24:31.200 Internet. It really does. This is all about the 00:24:31.200 --> 00:24:33.599 practice of data scraping. So HiQ Labs was a 00:24:33.599 --> 00:24:35.740 data analytics startup. And what did they do? 00:24:35.940 --> 00:24:38.660 They built a business model around scraping public 00:24:38.660 --> 00:24:41.819 LinkedIn profiles to analyze workforce trends. 00:24:42.079 --> 00:24:44.079 So they could tell their clients things like, 00:24:44.140 --> 00:24:46.259 engineers at this company are likely to quit 00:24:46.259 --> 00:24:50.109 soon, based on changes to their profiles. And 00:24:50.109 --> 00:24:52.430 to be clear, they weren't hacking into LinkedIn's 00:24:52.430 --> 00:24:55.009 servers. No, not at all. They were just writing 00:24:55.009 --> 00:24:57.349 automated scripts to visit the public facing 00:24:57.349 --> 00:25:00.069 pages and copy the information that anyone could 00:25:00.069 --> 00:25:02.589 see. They were doing what a human could do, but 00:25:02.589 --> 00:25:04.349 at the speed of light. LinkedIn didn't like that. 00:25:04.529 --> 00:25:07.289 LinkedIn hated it. They sent a cease and desist 00:25:07.289 --> 00:25:09.769 letter citing the Computer Fraud and Abuse Act, 00:25:09.869 --> 00:25:13.230 the CFAA, which is an anti -hacking law. They 00:25:13.230 --> 00:25:15.650 basically said, this is our data. It's on our 00:25:15.650 --> 00:25:18.390 servers. You are breaking into our house. And 00:25:18.390 --> 00:25:20.539 the court said. The court sided with Haikyuu. 00:25:20.819 --> 00:25:23.480 They granted a preliminary injunction that forced 00:25:23.480 --> 00:25:25.160 LinkedIn to allow them to continue scraping. 00:25:26.059 --> 00:25:28.740 The reasoning was profound for the future of 00:25:28.740 --> 00:25:30.960 the web. What was the reasoning? The court argued. 00:25:31.549 --> 00:25:33.869 That if a user chooses to make their profile 00:25:33.869 --> 00:25:36.450 public visible to the entire world without a 00:25:36.450 --> 00:25:38.950 password, then the company, in this case LinkedIn, 00:25:39.250 --> 00:25:41.950 cannot use anti -hacking laws to stop someone 00:25:41.950 --> 00:25:44.569 else from looking at it. That is huge. It basically 00:25:44.569 --> 00:25:46.190 says if you put something in the public square, 00:25:46.269 --> 00:25:47.970 you can't be mad when someone takes a picture 00:25:47.970 --> 00:25:49.950 of it. That's the perfect analogy. Thank you. 00:25:50.269 --> 00:25:52.769 But this creates a security nightmare. If scraping 00:25:52.769 --> 00:25:55.369 public data is legal, how do you distinguish 00:25:55.369 --> 00:25:58.369 between a startup like HiQ and a bad actor from 00:25:58.369 --> 00:26:01.329 Russia harvesting data for identity theft or 00:26:01.329 --> 00:26:03.569 phishing campaigns? And we know the bad actors 00:26:03.569 --> 00:26:05.529 are there. We absolutely have to talk about the 00:26:05.529 --> 00:26:08.940 2012 hack. The 2012 hack was the moment. LinkedIn, 00:26:08.940 --> 00:26:11.299 I think, lost its innocence in the public eye. 00:26:11.599 --> 00:26:15.759 Hackers stole 6 .4 million passwords. But the 00:26:15.759 --> 00:26:18.079 real scandal wasn't just the theft. It was the 00:26:18.079 --> 00:26:20.000 sheer incompetence of their security practices. 00:26:20.279 --> 00:26:23.240 They were storing passwords as unsalted SHA -1 00:26:23.240 --> 00:26:25.900 hashes. They were. Okay, we need to pause and 00:26:25.900 --> 00:26:28.319 explain that. For the listener who isn't a cryptographer, 00:26:28.500 --> 00:26:31.740 what is unsalted SHA -1 and why is it so bad? 00:26:32.000 --> 00:26:34.259 Okay, think of hashing a password like putting 00:26:34.259 --> 00:26:36.839 through a shredder. You can't tape it back together 00:26:36.839 --> 00:26:39.299 to read the original password, but if you have 00:26:39.299 --> 00:26:41.859 the shredded remains, you can check if they match 00:26:41.859 --> 00:26:44.980 another shredded document. SHA -1 was the type 00:26:44.980 --> 00:26:47.940 of shredder they were using, and by 2012, it 00:26:47.940 --> 00:26:50.940 was already known to be a weak, outdated, and 00:26:50.940 --> 00:26:53.160 insecure type of shredder. And what about the 00:26:53.160 --> 00:26:56.680 salt? Salting is like throwing a handful of unique, 00:26:56.759 --> 00:27:00.019 random glitter into the shredder along with each 00:27:00.019 --> 00:27:02.539 piece of paper. It makes the resulting shredded 00:27:02.539 --> 00:27:05.279 mess unique to that specific piece of paper. 00:27:05.460 --> 00:27:07.579 So even if two people have the same password, 00:27:07.799 --> 00:27:09.859 the shredded result will be different. Exactly. 00:27:09.859 --> 00:27:12.059 If you don't salt the passwords, then the password 00:27:12.059 --> 00:27:15.700 123 always shreds into the exact same string 00:27:15.700 --> 00:27:18.460 of characters. Hackers have these massive lists 00:27:18.460 --> 00:27:21.119 of pre -shredded common passwords. They're called 00:27:21.119 --> 00:27:23.819 rainbow tables. Because LinkedIn didn't salt 00:27:23.819 --> 00:27:26.160 them, the hackers could just look up the hashes 00:27:26.160 --> 00:27:28.480 in a table and crack millions of them almost 00:27:28.480 --> 00:27:31.470 instantly. That is a security 101 failure. It's 00:27:31.470 --> 00:27:35.190 basic stuff. It was grossly negligent. And the 00:27:35.190 --> 00:27:37.950 fallout from that one breach lasted for years. 00:27:38.509 --> 00:27:43.630 In 2016, a new data dump appeared online. 117 00:27:43.630 --> 00:27:46.390 million login credentials, email and password 00:27:46.390 --> 00:27:48.890 combinations, likely from that original 2012 00:27:48.890 --> 00:27:51.720 breach, appeared for sale on the dark web. For 00:27:51.720 --> 00:27:57.079 how much? For about $2 ,200. $2 ,200 for 117 00:27:57.079 --> 00:28:00.119 million accounts. That is terrifyingly cheap. 00:28:00.299 --> 00:28:02.579 It shows how commoditized our personal data has 00:28:02.579 --> 00:28:05.359 become. And the security issues didn't stop there. 00:28:05.500 --> 00:28:08.640 In 2021, you might remember reports surfacing 00:28:08.640 --> 00:28:11.339 of 500 to 700 million user records being leaked. 00:28:11.440 --> 00:28:13.059 I do remember that. It was all over the news. 00:28:13.380 --> 00:28:15.359 Now, LinkedIn's defense to this was very interesting. 00:28:15.460 --> 00:28:18.000 They said, we weren't hacked. This was just aggregated 00:28:18.000 --> 00:28:20.150 scraped data. So they were falling back on the 00:28:20.150 --> 00:28:22.089 high Q ruling saying that someone just scraped 00:28:22.089 --> 00:28:24.069 what was already public and put it all in one 00:28:24.069 --> 00:28:26.690 big convenient database for criminals. That was 00:28:26.690 --> 00:28:29.109 their argument. But to you, the user, does it 00:28:29.109 --> 00:28:31.450 really matter if your phone number, your email, 00:28:31.549 --> 00:28:33.589 your entire work history and your location are 00:28:33.589 --> 00:28:36.250 for sale on a hacker forum? Do you care if the 00:28:36.250 --> 00:28:38.970 data was technically obtained via a hack or just 00:28:38.970 --> 00:28:41.809 a scrape? The end result is the same. My privacy 00:28:41.809 --> 00:28:44.490 is compromised. The result is identical. And 00:28:44.490 --> 00:28:46.549 there was one other feature that really creeped 00:28:46.549 --> 00:28:47.990 me out when I was reading through the source 00:28:47.990 --> 00:28:51.519 material. The intro feature from 2013. Oh, this 00:28:51.519 --> 00:28:54.480 was wild. This is peak Silicon Valley hubris. 00:28:54.680 --> 00:28:57.200 What did it do? So LinkedIn wanted to integrate 00:28:57.200 --> 00:28:59.980 more deeply with the iPhone mail app. They wanted 00:28:59.980 --> 00:29:02.440 to show you the LinkedIn profiles of people who 00:29:02.440 --> 00:29:05.000 were emailing you right inside the app. Okay, 00:29:05.039 --> 00:29:07.700 that sounds potentially useful. How do they do 00:29:07.700 --> 00:29:10.579 it? To do that, they rerouted all of your incoming 00:29:10.579 --> 00:29:13.480 and outgoing emails through LinkedIn's own servers 00:29:13.480 --> 00:29:16.039 first so they could inject the necessary HTML 00:29:16.039 --> 00:29:18.440 code. Wait, they were intercepting your email? 00:29:18.740 --> 00:29:20.279 They were putting themselves in the middle of 00:29:20.279 --> 00:29:22.779 your private email conversations. Basically, 00:29:22.859 --> 00:29:24.839 yes. That's it. Security firms went absolutely 00:29:24.839 --> 00:29:27.980 nuclear. Uh -huh. Bishop Fox, a well -respected 00:29:27.980 --> 00:29:31.119 security consultancy, described it as being architecturally 00:29:31.119 --> 00:29:33.500 similar to a man -in -the -middle attack. That's 00:29:33.500 --> 00:29:36.599 a hacking term. It's a classic hacking term for 00:29:36.599 --> 00:29:39.259 when a third party secretly relays and possibly 00:29:39.259 --> 00:29:42.259 alters communication between two parties who 00:29:42.259 --> 00:29:45.170 believe they're communicating directly. LinkedIn 00:29:45.170 --> 00:29:47.609 was performing a man in the middle attack on 00:29:47.609 --> 00:29:50.650 its own users and calling it a feature. Unbelievable. 00:29:50.890 --> 00:29:52.809 They eventually killed the feature after the 00:29:52.809 --> 00:29:55.150 backlash, but it just showed this mindset of 00:29:55.150 --> 00:29:57.829 we will break the fundamental security model 00:29:57.829 --> 00:29:59.990 of email just to show you a profile picture. 00:30:00.210 --> 00:30:02.589 It's that constant tension, isn't it? Convenience 00:30:02.589 --> 00:30:06.130 versus privacy and privacy almost always loses. 00:30:06.289 --> 00:30:09.039 Almost always. Now, I want to zoom out to the 00:30:09.039 --> 00:30:11.240 biggest picture possible. We've talked about 00:30:11.240 --> 00:30:13.779 users. We've talked about the company. But LinkedIn 00:30:13.779 --> 00:30:16.660 is a global entity, which means it has to deal 00:30:16.660 --> 00:30:19.619 with global politics. And this is where the lofty 00:30:19.619 --> 00:30:21.859 mission of connecting the world's professionals 00:30:21.859 --> 00:30:25.019 crashes into the hard reality of authoritarian 00:30:25.019 --> 00:30:28.980 regimes. Specifically. China. The China conundrum. 00:30:29.059 --> 00:30:31.279 This is a fascinating case study in corporate 00:30:31.279 --> 00:30:33.519 compromise. For a long time, LinkedIn was the 00:30:33.519 --> 00:30:36.440 only major American social network allowed to 00:30:36.440 --> 00:30:38.279 operate in China. That's right. Facebook was 00:30:38.279 --> 00:30:40.859 banned. Twitter was banned. Google was banned. 00:30:41.180 --> 00:30:44.099 But LinkedIn survived. How? Did they have some 00:30:44.099 --> 00:30:45.720 kind of special deal with the government? They 00:30:45.720 --> 00:30:48.559 made a deal. They compromised. When they launched 00:30:48.559 --> 00:30:51.180 their localized Chinese version called Nying 00:30:51.180 --> 00:30:54.599 back in 2014, the CEO at the time, Jeff Weiner, 00:30:54.740 --> 00:30:57.619 was very transparent about it. He wrote a blog 00:30:57.619 --> 00:31:00.940 post saying effectively, we will have to censor 00:31:00.940 --> 00:31:03.259 some content to comply with local Chinese law. 00:31:03.440 --> 00:31:06.099 But we believe providing some access is better 00:31:06.099 --> 00:31:08.799 than providing no access at all. That is a heavy 00:31:08.799 --> 00:31:11.359 utilitarian calculation to make. We will agree 00:31:11.359 --> 00:31:13.759 to silence some people in order to help millions 00:31:13.759 --> 00:31:15.859 of others get jobs and connect with the world. 00:31:15.980 --> 00:31:18.579 It's a massive ethical tightrope. And for a while, 00:31:18.599 --> 00:31:20.859 it seemed to work. But the demands of the Chinese 00:31:20.859 --> 00:31:23.000 government are never static. They always want 00:31:23.000 --> 00:31:26.079 more control. When what happened? In 2021, the 00:31:26.079 --> 00:31:29.140 crackdown intensified. LinkedIn started blocking 00:31:29.140 --> 00:31:31.599 the profiles of U .S. journalists, people like 00:31:31.599 --> 00:31:34.579 Bethany Allen Abrahamian and Melissa Chan, blocking 00:31:34.579 --> 00:31:37.299 them from being viewed inside China because their 00:31:37.299 --> 00:31:40.140 profiles mentioned sensitive topics like their 00:31:40.140 --> 00:31:42.279 coverage of human rights abuses. So an American 00:31:42.279 --> 00:31:44.500 company was actively censoring American citizens 00:31:44.500 --> 00:31:47.079 on an American -owned platform to appease the 00:31:47.079 --> 00:31:49.000 government. in Beijing. That was the breaking 00:31:49.000 --> 00:31:52.160 point. The bad press was overwhelming. The operational 00:31:52.160 --> 00:31:55.140 compliance was becoming impossible. And so late 00:31:55.140 --> 00:31:58.680 in 2021, Microsoft finally pulled the plug. They 00:31:58.680 --> 00:32:00.440 shut it down. They shut down the social networking 00:32:00.440 --> 00:32:03.559 service in China. They replaced it with a bare 00:32:03.559 --> 00:32:06.980 -bones app called InJobs, which had no social 00:32:06.980 --> 00:32:09.799 feed, no ability to share posts or articles. 00:32:09.880 --> 00:32:12.380 It was just a sterile job board. And even that 00:32:12.380 --> 00:32:16.000 didn't last, did it? No. In 2023, they discontinued 00:32:16.000 --> 00:32:18.579 the local job apps entirely. It was the end of 00:32:18.579 --> 00:32:20.819 the experiment. And it proved, I think, that 00:32:20.819 --> 00:32:23.119 you cannot run a Western -style social network, 00:32:23.319 --> 00:32:25.920 even a professional one, inside a regime that 00:32:25.920 --> 00:32:28.799 demands total information control. The two models 00:32:28.799 --> 00:32:31.200 are fundamentally incompatible. And Russia was 00:32:31.200 --> 00:32:32.819 even stricter, right? They didn't even get to 00:32:32.819 --> 00:32:34.819 the censorship phase. They just got banned outright. 00:32:35.059 --> 00:32:37.700 Russia kicked them out in 2016. No. Russia passed 00:32:37.700 --> 00:32:40.539 a data sovereignty law that requires any company 00:32:40.539 --> 00:32:42.940 holding the personal data of Russian citizens 00:32:42.940 --> 00:32:45.839 to store that data on servers physically located 00:32:45.839 --> 00:32:48.039 within Russia. And LinkedIn refused. LinkedIn 00:32:48.039 --> 00:32:50.440 refused to move the data. So Putin pulled the 00:32:50.440 --> 00:32:53.240 plug. LinkedIn is blocked in Russia to this day. 00:32:53.359 --> 00:32:55.039 It was removed from the App Store and Google 00:32:55.039 --> 00:32:58.359 Play there. If you're a professional in Moscow 00:32:58.359 --> 00:33:00.900 and you want to update your resume, You have 00:33:00.900 --> 00:33:03.680 to use a VPN. It really shows that the economic 00:33:03.680 --> 00:33:06.779 graph has hard borders. It is not truly global. 00:33:06.980 --> 00:33:09.119 But where there are borders and where there's 00:33:09.119 --> 00:33:12.000 valuable data, there are spies. And this part 00:33:12.000 --> 00:33:13.839 of the outline reads like it's straight out of 00:33:13.839 --> 00:33:17.119 a spy novel. LinkedIn is apparently a playground 00:33:17.119 --> 00:33:19.680 for espionage. It makes perfect sense if you 00:33:19.680 --> 00:33:21.779 think about it. If you're an intelligence officer 00:33:21.779 --> 00:33:24.740 and your mission is to recruit a source inside, 00:33:24.880 --> 00:33:27.970 say, the U .S. Department of Defense, or inside 00:33:27.970 --> 00:33:30.970 a major European telecom company, where do you 00:33:30.970 --> 00:33:33.049 go to find your targets? You go to the global 00:33:33.049 --> 00:33:35.109 directory where everyone conveniently lists their 00:33:35.109 --> 00:33:37.670 job title, their security clearance level, and 00:33:37.670 --> 00:33:39.869 the specific sensitive projects they've worked 00:33:39.869 --> 00:33:42.970 on. It's a menu for spies, a targeting database. 00:33:43.269 --> 00:33:46.130 And it's not just the bad guys, so to speak. 00:33:46.230 --> 00:33:48.390 There was a famous case revealed by the Snowden 00:33:48.390 --> 00:33:50.930 documents called Operation Socialist. And who 00:33:50.930 --> 00:33:53.829 was behind that? This was GCHQ British Intelligence. 00:33:54.539 --> 00:33:57.759 They set up fake LinkedIn pages to target employees 00:33:57.759 --> 00:34:00.880 of Belgacom, a major Belgian telecom company. 00:34:01.039 --> 00:34:02.940 Wait, the British were hacking the Belgians, 00:34:02.960 --> 00:34:05.019 aren't they allies? In the world of signals intelligence, 00:34:05.539 --> 00:34:09.219 everyone is a potential target. GCHQ used these 00:34:09.219 --> 00:34:12.539 fake profiles to lure Belgacom engineers to a 00:34:12.539 --> 00:34:14.960 malware -infected page so they could infiltrate 00:34:14.960 --> 00:34:17.579 their network. They used LinkedIn as the honeypot. 00:34:17.880 --> 00:34:21.039 And I assume the Chinese and Russians and Iranians 00:34:21.039 --> 00:34:23.260 are doing the same thing? At an industrial scale. 00:34:23.739 --> 00:34:25.500 The French intelligence services have issued 00:34:25.500 --> 00:34:27.579 public warnings that Chinese spies use LinkedIn 00:34:27.579 --> 00:34:30.179 to contact thousands of French officials. There 00:34:30.179 --> 00:34:32.639 was an Iranian hacking group, Threat Group 2889, 00:34:33.019 --> 00:34:36.159 that created these incredibly detailed fake personas. 00:34:36.559 --> 00:34:38.340 What did they look like? Fully fleshed out profiles, 00:34:38.659 --> 00:34:41.039 professional headshots, endorsements from real 00:34:41.039 --> 00:34:43.400 people, fake work histories at real companies. 00:34:43.619 --> 00:34:46.079 They look like legitimate recruiters. They would 00:34:46.079 --> 00:34:48.380 connect with you, build trust over a few weeks, 00:34:48.460 --> 00:34:50.340 and then send you a job description document 00:34:50.340 --> 00:34:52.500 that was actually a malware payload designed 00:34:52.500 --> 00:34:55.099 to steal your... It really makes you think twice 00:34:55.099 --> 00:34:57.219 about accepting that random connection request 00:34:57.219 --> 00:34:59.940 from the recruiter with the vague profile picture. 00:35:00.219 --> 00:35:03.199 You absolutely should be skeptical. Professional 00:35:03.199 --> 00:35:06.639 networking as a concept relies on trust on the 00:35:06.639 --> 00:35:08.860 heuristic that a friend of a friend is probably 00:35:08.860 --> 00:35:12.260 safe. Espionage operations are designed to exploit 00:35:12.260 --> 00:35:15.079 that exact heuristic. Before we wrap up, we need 00:35:15.079 --> 00:35:17.119 to look at the present day, or at least the very 00:35:17.119 --> 00:35:20.159 near future. The source material mentions some 00:35:20.159 --> 00:35:22.820 pretty significant changes in 2025 that seem 00:35:22.820 --> 00:35:24.739 to reflect the shifting political landscape, 00:35:24.980 --> 00:35:27.420 particularly here in the U .S. Yes, these big 00:35:27.420 --> 00:35:29.900 tech platforms are often barometers of the broader 00:35:29.900 --> 00:35:33.059 political climate. They're not neutral. In January 00:35:33.059 --> 00:35:35.099 2025, during the second Trump administration, 00:35:35.820 --> 00:35:38.380 LinkedIn quietly deleted its entire diversity, 00:35:38.500 --> 00:35:41.579 equity, and inclusion, or DEI, webpage. Just 00:35:41.579 --> 00:35:44.820 scrubbed it from the site. Gone. And then in 00:35:44.820 --> 00:35:47.500 July 2025, they removed specific protections 00:35:47.500 --> 00:35:49.599 against misgendering or deadnaming transgender 00:35:49.599 --> 00:35:52.199 users from their professional community policies. 00:35:52.420 --> 00:35:54.780 That's a huge rollback. It's a significant reversal. 00:35:55.119 --> 00:35:58.219 For years, corporate America, especially big 00:35:58.219 --> 00:36:00.739 tech, was moving in a very specific direction 00:36:00.739 --> 00:36:04.260 on these social issues. Now, these sources suggest 00:36:04.260 --> 00:36:06.739 they are pivoting rapidly to align with a new 00:36:06.739 --> 00:36:09.599 political reality. It shows that even these neutral 00:36:09.599 --> 00:36:12.719 professional platforms are deeply political in 00:36:12.719 --> 00:36:14.219 how they choose to govern their communities. 00:36:14.440 --> 00:36:16.940 And finally, we have to talk about the AI elephant 00:36:16.940 --> 00:36:19.079 in the room. We are all feeding the machine. 00:36:19.340 --> 00:36:22.460 We are all the training data. In 2024, there 00:36:22.460 --> 00:36:25.119 was a controversy in the UK because it was revealed 00:36:25.119 --> 00:36:28.940 that LinkedIn had quietly opted in all users 00:36:28.940 --> 00:36:31.659 by default to have their public profile data. 00:36:32.030 --> 00:36:34.590 used for AI training. So my profile, my posts, 00:36:34.670 --> 00:36:36.769 my work history, my skills, it's all being fed 00:36:36.769 --> 00:36:39.030 into their large language models. Unless you 00:36:39.030 --> 00:36:41.449 explicitly go deep into your settings and find 00:36:41.449 --> 00:36:44.269 the button to turn it off. Yes. The UK regulators 00:36:44.269 --> 00:36:46.289 pushed back and forced them to pause it there. 00:36:46.409 --> 00:36:49.070 But for most of the world, we are the fuel for 00:36:49.070 --> 00:36:51.650 their AI ambitions. The economic graph is becoming 00:36:51.650 --> 00:36:54.130 the brain of an AI. So we've covered the history, 00:36:54.250 --> 00:36:56.769 the sociology, the money, the lawsuits, the spies 00:36:56.769 --> 00:36:59.449 and the politics. It is a massive, complex and 00:36:59.449 --> 00:37:02.900 powerful beast. It is so much more than a digital 00:37:02.900 --> 00:37:05.800 resume. So let's try to synthesize this for the 00:37:05.800 --> 00:37:08.400 listener. When they open that blue app tomorrow 00:37:08.400 --> 00:37:10.659 morning to scroll through the feed, how should 00:37:10.659 --> 00:37:12.639 they see it differently? I think you have to 00:37:12.639 --> 00:37:16.260 see it as a trade, a very explicit trade. You 00:37:16.260 --> 00:37:20.059 are trading your data, your entire professional 00:37:20.059 --> 00:37:23.699 life history, updated in real time for access 00:37:23.699 --> 00:37:26.130 to the network. and the opportunities it provides. 00:37:26.469 --> 00:37:29.210 And that trade can have real value. It can. It 00:37:29.210 --> 00:37:30.949 can get you a job. It can help you build your 00:37:30.949 --> 00:37:34.230 career. But you cannot be naive about the cost 00:37:34.230 --> 00:37:37.750 of that trade. You are a node in a graph that 00:37:37.750 --> 00:37:40.650 is being sold to recruiters, scraped by startups, 00:37:40.869 --> 00:37:42.949 and monitored by governments. Oh, be intentional 00:37:42.949 --> 00:37:45.670 about it. Be very intentional. Curate what information 00:37:45.670 --> 00:37:47.889 you choose to make public. Understand that your 00:37:47.889 --> 00:37:50.309 weak ties are powerful, but the platform itself 00:37:50.309 --> 00:37:52.989 is not your friend. It is a marketplace, and 00:37:52.989 --> 00:37:55.039 you are the inventory. I want to leave us with 00:37:55.039 --> 00:37:57.079 a final provocative thought based on where this 00:37:57.079 --> 00:38:00.019 is all going. We mentioned the new career hub 00:38:00.019 --> 00:38:03.320 and the AI agents that LinkedIn is rolling out 00:38:03.320 --> 00:38:06.400 in 2024 and 2025. If LinkedIn has all the data, 00:38:06.500 --> 00:38:08.699 they know who I am, they know my skills, they 00:38:08.699 --> 00:38:10.599 know what jobs are open, and they know who the 00:38:10.599 --> 00:38:12.659 recruiters are. I see exactly where you're going 00:38:12.659 --> 00:38:14.920 with this. Are we moving toward a future where 00:38:14.920 --> 00:38:17.820 I don't even apply for jobs anymore? Where the 00:38:17.820 --> 00:38:20.440 LinkedIn AI just... talks to the recruiters, 00:38:20.440 --> 00:38:22.420 they negotiate in the background and they just 00:38:22.420 --> 00:38:24.559 tell me where to show up for work on Monday. 00:38:24.699 --> 00:38:28.059 That is the logical conclusion of the economic 00:38:28.059 --> 00:38:31.639 graph, a completely frictionless labor allocation 00:38:31.639 --> 00:38:35.300 system. Yeah. But if we ever get there, if the 00:38:35.300 --> 00:38:37.300 machine does the networking and the applying 00:38:37.300 --> 00:38:40.659 for us, does the platform actually build community 00:38:40.659 --> 00:38:44.820 or have we just built a massive, silent, automated 00:38:44.820 --> 00:38:47.320 vending machine for human capital? A vending 00:38:47.320 --> 00:38:49.500 machine for human capital. That is a haunting 00:38:49.500 --> 00:38:52.039 image and definitely something to mull over the 00:38:52.039 --> 00:38:53.880 next time you get an email asking you to endorse 00:38:53.880 --> 00:38:57.619 someone for strategic planning. Thanks for listening 00:38:57.619 --> 00:38:59.480 to The Deep Dive. We'll see you next time. Goodbye.