Subscribe
Share
Share
Embed
The Diff is a newsletter exploring the technologies, companies, and trends that are making the future high-variance. Posts range from in-depth company profiles, applied financial theory, strategy breakdowns and macroeconomics.
Is "Social" an Enduring Category or Just a Phase?
In 2011, Google tied every employee's bonus to their social efforts and started sticking "+1" buttons onto every other Google product. This effort ended ignominiously, with Google's core social product, Google+, limping along for a few years and then being shut down after Google discovered a security flaw that could expose user data. Google still has products that enable social interaction, like Gmail and whatever the current name of their current most-favored-chat-app is. And YouTube might support more parasocial relationships than any other platform. But the pure bet on "social" didn't work out.
Google, of course, has a reputation for launching splashy products and then quietly killing them (though they seem to have gotten better recently, aside from things like the death of cached pages). But they're not the only ones. Plenty of other companies have started with a peer-to-peer model and moved towards something that looks a lot more like a two-sided network:
Lending Club's original idea was that some people would log in to get better-than-a-certificate-of-deposit returns on fairly short-term loans, while other users could consolidate debts or pay for a big one-time expense. But over time, more and more lending came from credit funds that specialized in crunching numbers on weird debt. Those funds could take some of the models they already used for consumer debt products like credit card ABS and repurpose it to underwrite individual loans at scale.
Airbnb found that side-hustle part-time landlords were less scalable than full-time professionals. There's a wide range there, naturally, from people who downscaled their income and lifestyle by keeping their home in a nice city, buying a smaller and cheaper dwelling in a slightly less trendy location, and using the original home as a source of retirement income and a part-time job.
The original eBay model was that people would clean out their attics and sell their appreciated collectibles to one another. This business still exists, but isn't as important as fixed-price sales from repeat sellers, i.e. small merchants.
And it's also striking to look at two of the biggest growth areas in social media right now. First, short-form video, and second, group chats. These are both on the social media continuum—there are long-tail creators in short video, and some group chats are large enough that they're closer to a miniature Twitter than a one-on-one chat. But they're both at far ends of the "social media" medium. A group chat is close to private communication, and a lot of their growth has come from that same privacy—part of the trouble with seamless shareability in social is that anything you say that only makes sense with context can still reach escape velocity and achieve a mass audience of maximally uncharitable people. And even though video creators tend to start out as independent creators working out of their bedrooms, more and more viewing comes from professionals who operate out of third-party studios or run their own.
This points to an interesting possibility: what if purely social products have a finite life? eBay could be the most illuminating example here: beanie babies, rare coins, no-longer-beloved heirlooms and the like provided the potential energy for a one-time creation of a powerful network. Some parts of the network were one-sided, with inveterate collectors disposing of some products in order to buy something else in the same category. But many more were naturally one-time users: liquidating some assets they didn't really want, or finally getting their hands on something they did.
But what eBay got from that orgy of beanie baby speculation and ruthless rare coin arbitrage was the beginning of a two-sided network: customers shifted to buying more fixed-price products, sellers moved in that direction as well, and now the company is a mature e-commerce platform, one that would be a lot bigger if it had embraced first-party business and logistics the way Amazon did, but still a viable business that generates billions in free cash flow each year.
One of the questions people designing social products have to ask is: how do relationships scale? It's obvious that people can sustain meaningful relationships with close friends, family members, a spouse, etc. But there are clearly limits; someone with thousands of friends usually doesn't have many, or any, close ones. Broadcast as a general concept has existed for a long time (it's arguably not just pre-printing press but pre-literacy, given that some older texts—the Bible, Bhagavad Gita, Odyssey, etc. were oral traditions before being written down, and seem to have had some level of mass distribution without needing to be a mass-produced text. So we've long had distinct categories for flesh-and-blood individuals we interact with regularly, and for quasi-mythic people—the Emperor, the President, Taylor Swift—for whom the relationship is real to us but meaningless to them. Those relationships are closer to media consumption than to personal conversation. They're also much, much more scalable than conversation. It's a necessary consequence of days being 24 hours for everyone that the people with the most interesting lives are the people with the least free time to document them, so the median piece of social media content is, through selection effects alone, pretty boring.
Social media companies have built apps that absorb a billion-plus hours of attention each day because they can highlight the small set of updates that are interesting to people. But there's a whole industry devoted to creating content that's interesting to a large audience: the media business, minus the social. And that is increasingly what social media looks like: some content sponsored by big companies, some content sponsored by smaller ones (who often monetize through bigger ones), and, just to keep things interesting, some life updates from people you know.
This is not just content, though; it's training data. The place where peer-to-peer systems really shine, whether it's early eBay, forum culture, or modern, Twitter, is connecting people to niche interests they didn't realize anyone else would ever share. But over time, the requirement that some other human being share your exact interest gets weaker and weaker, because every interest can be described as some sort of latent space, i.e. a cluster of existing content or statistically-equivalent hypothetical content that satisfies someone's interests. This isn't social, since there isn't an author on the other side of the screen. It's not mass-media, though, because nobody's experiencing the same media. Instead, it's an exit from social.
Social as a model works when people have about as much to offer as they want to receive along a given axis. But no trait is distributed uniformly; there are are outliers in the nice-to-look-at, nice-to-listen-to, nice-to-read, nice-to-get-stock-tips from axes, there's a population that can offer a respectable performance with these traits, and there's a substantial majority with below-mean performance. So, over time, most platforms end up with a more consolidated list of suppliers and a dispersed set of consumers. The shift in progress is that the platform itself is positioned to be a content supplier, too; AI really shines with never-before-seen search queries. And when that happens, the temporary oversupply of some content category that kickstarts social sites won't exist any more, and social will be a strictly inferior model to generative AI. There will still be room for messaging, and for community-based apps and chats, but the communities will be largely preexisting, rather than accreting within a larger platform. There was a time when connecting people online produced value automatically, both socially and for shareholders, but that time was due to particular circumstances that couldn't have held true forever.
Disclosure: Long META