wavePod

Get Wave AI

Meta - Acquired | Wave AI Podcast Notes

Back to Acquired

Meta

Acquired

Mon Oct 28 2024

Loading summary...

Transcript

Ben Gilbert (0:00)

All right, so I was up late last night. Late for me as a dad is like 11pm But I'm sitting here at my computer in my dark basement pulling notes together. David, what music did I have on? You have one guess.

David Rosenthal (0:14)

Oh. Oh, I know exactly what you had on. You had Trent Reznor Social Network soundtrack.

Ben Gilbert (0:20)

It makes anything you're doing feel twice as important and twice as revolutionary. And it just felt very apt for this episode.

David Rosenthal (0:28)

Oh, man. So I have been listening in the last 24 hours to 50 cent into club, because that came out my freshman year of college. Same year as the facebook.com and like, man, 50 cent, facebook.com can't get any better than that.

Ben Gilbert (0:45)

Perfect match made in heaven. Well, I'll check out our wall to wall from the old days and see if there already posts about that. All right, let's do it.

David Rosenthal (0:54)

Let's do it. Who got the truth? Is it you? Is it you? Is it you? Who got the truth now?

Ben Gilbert (1:03)

Is it you? Is it you? Is it you? Sit me down.

David Rosenthal (1:07)

Say it straight. Another story on the way.

Ben Gilbert (1:10)

Who got the truth? Welcome to the fall 2024 season of acquired, the podcast about great companies and the stories and playbooks behind them. I'm Ben Gilbert. I'm David Rosenthal, and we are your hosts. Today we are studying a company whose products are used by more humans than any other company in history, Meta, of course, formally known as Facebook. So I figured I would contextualize these numbers a little bit. Meta has 4 billion monthly active users.

David Rosenthal (1:42)

And daily active users are over 3 billion.

Ben Gilbert (1:45)

Yeah, nuts. There are only 8 billion humans on Earth. So as I started to brainstorm what the closest competitors could be to serving half of the humans, I thought, surely I can find it in empires or governments from the past where there is some larger percentage.

David Rosenthal (2:03)

Yeah, makes sense.

Ben Gilbert (2:04)

Nope. The Roman Empire at its peak was only 40% of humans, tops. You know, the data's a little bit hard to actually find from that period of time, but the British Empire, which we have a little bit better handle on, at its peak was only 23% of the global human population. So no government, tech company, utility, et cetera, has ever addressed so much of the world.

Ben Gilbert (212:30)

Right. So it's interesting to look at this because Google basically is Facebook's biggest business model. Comp. They show ads on the web and they monetize that really well. This theoretically could have been in their wheelhouse. This is just more real estate to show ads on the web and they've already got all these people with user accounts. You can see why Facebook took this really seriously. We've talked a lot about Facebook's technical prowess. Well, here is an example of where it really kicks in when you need it to to be a key competitive advantage. So they structured themselves in a way that encouraged flexibility of engineers and they really broke from the tide of microservices in this era. They had one monolithic code base that everyone worked out of. And you might say, well that's stupid. Why would they? That's not the way the world was going. But what they basically did was they wanted to encourage portability of teams. If you're an engineer, you get hired into the company, not onto a team. You learn the company's code base, you learn the company's conventions and you can easily sort of move around after that. You do have to deal with the trade off then of you have this big monolithic code base with gigabytes of PHP code for thousands of engineers. What do you do about that? Well, then they just had their cake and eat it too. They would go and have the infrastructure team figure out how to deal with that problem rather than saying, oh, everyone can just work in their little silos. And so that meant that engineers could kind of be quickly reorged. They could have this company wide lockdown to fight Google. And they did all sorts of things. They launched video calls to compete with hangouts, they launched something to compete with Google circles. But either way, they could really quickly reallocate resources and people who sort of knew how to work together to defeat what could have really been an existential threat from their biggest similar company.

Ben Gilbert (254:47)

Yes. So listeners, of course we don't actually ever know what anybody's intent is or what they're thinking when they decide to buy a company or something like that. This is just sort of, David and I guessing at strategy from the outside. We do have, thanks to a court case, an actual email from mark Zuckerberg on February 28th. Oh yeah, this is so good 2012. To their then CFO talking about @ the time, they're sort of discussing the Instagram acquisition, but laying out the idea behind an acquisition strategy. The basic plan would be to buy these companies and leave their products running while over time incorporating the social dynamics they've invented into our core products. One thing that may make neutralizing a potential competitor more reasonable here is that there are network effects around social products and a finite number of different social mechanics to invent. Once someone wins at a specific mechanic, it's difficult for others to supplant them without doing something different. It's possible someone beats Instagram by building something that is better to the point where they get network migration. But this is harder as long as Instagram keeps running as a product pause, which is pretty interesting, that is the argument of why to keep Instagram separate and running as its own product is because Instagram's already discovered this fascinating new mechanic around publishing one image at a time with these beautiful filters. If anybody else tries to come after them, they're already ahead. So actually the Best thing to do is own Instagram and let it keeps doing its thing anyway. Resuming integrating their products with ours to improve the service is also a factor. But in reality we already know these companies social mechanics and we will integrate them over the next 12 to 24 months anyway. The integration plan involves building their mechanics into our products rather than directly integrating their products, if that makes sense. By a combination of these two things. Neutralizing a potential competitor, integrating their products with ours to improve the service. One way of looking at this is that what we're really buying is time. Even if some new competitor springs up, buying Instagram, Path, Foursquare, etc. Now will give us a year or more to integrate their dynamics before anyone can get close to their scale again. Within that time, if we incorporate the social mechanics they are using, those new products won't get much traction because we will already have their mechanics deployed at scale. It is goddamn brilliant. David. Here's my commentary based on all this. It is so smart to basically say, well if we buy them we basically get two strategies that we get to execute at the same time. One, we just leave them alone and let it keep succeeding. It's actually not a current potential threat. We don't know if it will turn into a potential threat. Hell, Instagram wasn't making any money yet. WhatsApp didn't have a feed. It was just a way people connected. It's not like just because they had a network, they were going to turn it into a feed. So there's this idea that it's not a competitor. Now by buying it, we basically get this option on if it becomes a competitor, if we completely leave them alone and let them decide what to do. But then there's also. Look, we will integrate those mechanics into our core product, Facebook, the blue app that already has well built out network effects. And by owning this thing that could become a competitor. There isn't white space in the middle. Anyone just like us is also going to take time and they're not going to do it as good as the original and they don't have our network. Therefore, whether the winner turns out to be the original product or us incorporating the mechanic into the blue app, we've won either way. And probably what's going to happen is both.

David Rosenthal (266:05)

Okay, so that I think is dimension one, that TikTok is completely orthogonal to Facebook. Dimension two is actually organizationally. So TikTok, as we chronicled on our episode about it a couple years ago, came out of Bytedance. Well, it was an acquisition of Musical Ly by Bytedance and then Bytedance turned it into TikTok. Bytedance had Toutiao before having TikTok. Bytedance is an AI company. Bytedance is not a product company. And the core product that ByteDance makes is its AI, along with all sorts of other stuff. TikTok is like the product vector, one of many. And the Biggest by which they deploy their actual product, which is AI. This is very, very different from Facebook, which has always been a product company. And so like organizationally, the way ByteDance is set up is there are these centers of excellence within the company where the algorithms, the AI that is controlled by a centralized team within the company, at least that's my perception of how they're different organizationally from how Facebook used to be. Facebook and Facebook leadership was obviously watching this very, very closely. And if you look at how Meta is organized now, it looks a hell of a lot more like that. So Meta today is organized. There are centers of excellence across the company company. Same way that I was just talking about ByteDance. There's fair, that's the fundamental AI research team. There's infrastructure shared across the whole company, deployed to all the products and apps. There's revenue and monetization shared across the whole company, deployed to all the products and apps. There's growth shared across the whole company, deployed across the various products and apps. There's integrity and safety, which is a big, big, big advantage that Facebook has. They've reorganized how the whole company looks. There's still all the individual products in the family of apps. Those all have their own heads and they all roll up to Chris Cox who has had a product for the whole company and AI too actually reports to him. But all the infrastructure, all the revenue, all the growth, all the integrity and safety are decoupled from those actual products now and live in these centers of excellence across the company.

Ben Gilbert (285:41)

Yes. So as you can imagine, I'm preparing for this episode and I'm like, okay, I just heard Mark Zuckerberg say on another podcast, by the way, great interview with Alex Heath at the Verge after Meta Connect about this, he dropped this line. People thought that all this data had been taken and it had been used in this campaign. And it turned out the data wasn't and the data wasn't even accessible to the developer. So David, I'm listening to that and I stop because I'm thinking, well, that's not what anybody thinks because there was a ton of reporting on the fact that they did so what actually happened here? So here are the chain of events. Back in the 2000 and tens, Facebook built the platform that we talked about that was incredibly permissive. In the early days, when you authed your profile against an app, you gave it all of your profile data. And in those earliest days it could actually see your whole friend list too. Facebook eventually made their API much more closed off, but it took a couple of years to do that. Now there was a terms of service that required app developers to delete any data they had after a period of time and only use it for very specific purposes. But of course, Facebook couldn't actually guarantee that people were complying with the terms of service. It was just, you know, you were a breaching contract. So at some point a developer made an app that was a quiz. Users could opt into taking that quiz. The quiz asked you questions and by combining the answers to those questions with the profile data you authenticated from Facebook, that app developer then tried to label you with a pretty basic psychographic profile. And then ultimately that derivative data, the labels of personality characteristics that were derived from Facebook data and your quiz answers is what Cambridge Analytica had, as best I could tell. So I read a good amount of the report that the UK government put out about this. The conclusion was that the quiz was taken by 320,000 people. Since those people had access to friends. The total surface area was the public profile data of 87 million Facebook users. They also found that Facebook had requested that Cambridge Analytica delete all the data, both the actual Facebook data and anything derived from the Facebook Data back in 2015, before the election even started. So it was yes, true an application got access to Facebook data on a lot of users. It is also true that it violated terms of service to use that data for other purposes and they did not comply with a delete request or they may not have complied. It's kind of difficult to reconstruct. On top of all this, it supposes that Cambridge Analytica's method of taking these quiz answers and translating them into something that could impact voting worked like it was effective. And that's pretty thin, right?

Ben Gilbert (368:16)

And then as we drift to a close here, the engineering culture and being a technology company at their core has been essential. Early on, they really did manage to hire only A and then stay A after that forever and ever and ever. It was just this badge of honor if you were an engineer at Facebook for product design, too. I mean, they just had such a great talent density. The set of things that they did on the technical side were over and over Again, a way to have their cake and eat it too. If you can move faster, you can learn more through your multiple iterations. And so speed of development comes from having great tools. One very great shining example of this is something called hip hop for php. I know this is very esoteric, but in the late 2000s they had this crazy idea that what we should do, instead of switching to C or C or Java, we want our engineers to keep writing php. So they wrote a compiler to C so they didn't have to take the performance hit from running php, which was an interpreted language. But it also solved this scalability problem because then they didn't need to go hire all these systems level programmers. They could hire web developers who wanted to move with that pace and flexibility while also having the infrastructure to run these massive systems and scale really efficiently. And then that solved the scaling and performance problem. But again, in 2014 they realized, oh crap, we're big enough and we now have enough sensitive data that we really should switch to a statically typed language like Java or C. But again, they didn't want to force their engineers to learn that they didn't want to. People who code in those languages have a different culture than existed at Facebook too. They couldn't really recruit people, so they invented a new language called Hack that was very similar to PHP but had static type over and over again. I mean, Tao was another example of this NoSQL database. They just keep finding ways where they invent new technology to solve a problem that probably only exists for them. And then they create this whole boutique system that allows A having world class talent, B to have a ton of them, C everyone gets to move fast, but then D it's all unbelievably performant and efficient and they just don't have to make trade offs. It is wild.

Ben Gilbert (370:44)

Okay then my second technical one in addition to this is they are their own customer. We've talked about this. They're a AWS scale technology company, but they don't take outside customers, so they only have to build for their own internal use cases. Now this sounds great, but it actually does have these big trade offs. You can't dramatically change what your infrastructure is used for. It is purpose built, but it does let you be incredibly efficient and have high performance if you have good communication between the customers or the app team or the backend service and the designers of that data center. This is completely the opposite of Amazon. Amazon uses interface so teams don't have to talk to each other at Meta. They require incredibly tight Communication. It's a very different organizational philosophy where they're like, no, no, no. Not only do we not have external customers, we want this insane type coupling between our infrastructure in our internal customers. Anyway, my last one. This is a company that grows intentionally. It would be easy to look at this company and say, wow, what a viral product. What a universally applicable product. That is not the case. It is unnatural to have connected 4 billion humans. This is a freak of nature. This is not just something that people adopted. And so while I think it is totally fair to say wow, it just like blew up at Harvard. Facebook is the story of 50 different growth tactics in different eras, all carefully constructed and iterated upon wait lists at colleges, strategically launching at every single college, picking each one for a reason. Translation internationalization, zero rating with carriers to bring new people online, figuring out when they needed to acquire versus build carefully split testing every change, aligning the whole company on specific networks, on specific metrics. I mean, building relationships with governments in all these different countries and at the very least complying with local laws on where should we have certain speech laws versus not. I mean, it is completely unnatural for them to have done what they've done. And it's all been very, very intentional to connect the world.

Ben Gilbert (378:11)

Oh, yes, yes. My second quick one is the Dwarkesh podcast. I love the Dwarkesh podcast. I also love Dwarkesh, and I think that if you like this show, you'll love listening to the interviews he does. Most recent one, or maybe just a couple ago, is with Daniel Juergen, who is the author of the Prize, which is a book. David, you and I almost read a whole bunch of it for Standard Oil. And then we realized Standard Oil is sort of over at least the chapter of Standard Oil that we were covering within the first two chapters of his book. And so it's basically everything from the end of our Standard Oil episode forward on the geopolitics of oil that ended up shaping and informing our world today. And Dwarkesh is just an amazing interviewer and conversationalist. All right, well, with that, listeners, a huge thank you to our partners JP Morgan Payments, Crusoe, Statsig and Huntress. You can click the link in the show notes to learn more. We talked to a ton of people for research on this one, and while we can't mention everyone in part because the list would just be too long, in other part made some folks ask not to be thanked, we do have some specific ones that we want to give a shout out to. So Alex Schultz, the cmo, head of Growth analytics and Internationalization. Great to talk with Boz. Andrew Bosworth, the cto Stephen Levy, who wrote the book the Inside Story, was generous with his time. Jim Breyer, who led Excel's Series A investment in Facebook. Jan Lecun met his chief AI scientist, Alex Heath at the Verge for spending his time with me. To a friend of the show, Arvind Navarathnam from Worldly Partners who wrote an excellent research report kind of chronicling everything. Well, it's almost like a written version of this podcast. It's like a 100 page PDF that was awesome to consume, kind of to help me remember all the big beats of the story and that his research report is linked in the show notes to Arielle Zuckerberg, Mark's sister Sheryl Sandberg, obviously longtime COO Mike Schretfer, the former CTO and now senior fellow to Pete Hunt, early engineer who transferred from Facebook to Instagram Post acquisition Naomi Gleit, who we talked about, founding member of the Growth team and the longest employee, longest.