How to Blow Up a Timeline

NOTE: I’d been working on this piece on and off for a few weeks while trying to move to NYC and settle into my new apartment, and just as I was about to publish it, Elon rate-limited Twitter and so, sensing a moment of weakness, Meta pulled up its launch date for Threads to yesterday. This piece doesn’t cover Threads directly, nor does it talk about the rate-limiting fiasco. It’s focused on why I think Twitter got so much worse over the past year. I thought about holding off and reworking it entirely to incorporate all that happened this week, but in the end I decided that it was cleaner to publish this one as is. If Twitter hadn’t botched so much over this past year, Threads wouldn’t matter. Still, like past pieces I’ve written on topics related to Twitter, you can apply a lot of the ideas in this piece to analyzing Threads’ prospects. And I’ll push a follow-up piece with my specific reactions to and predictions for Threads soon-ish. Follow me on Substack to get a note when that drops.

“I shall be writing about how cities work in real life, because this is the only way to learn what principles of planning and what practices in rebuilding can promote social and economic vitality in cities, and what practices and principles will deaden these attributes.” — Jane Jacobs, The Death and Life of Great American Cities


Today, I come to bury Twitter, not to appraise him.

Oh, who am I kidding, I’m mostly here to appraise how it blew up.

For years, I thought Twitter would persist like a cockroach because:

  • At its core, it’s a niche experience that alienates most but strongly appeals to a few
  • Those few who love Twitter comprise an influential intellectual and cultural cohort, and at internet scale, even niches can be substantial in size

I’ve written before in Status as a Service or The Network’s the Thing about how Twitter hit upon some narrow product-market fit despite itself. It has never seemed to understand why it worked for some people or what it wanted to be, and how those two were related, if at all. But in a twist of fate that is often more of a factor in finding product-market fit than most like to admit, Twitter's indecisiveness protected it from itself. Social alchemy at some scale can be a mysterious thing. When you’re uncertain which knot is securing your body to the face of a mountain, it’s best not to start undoing any of them willy-nilly. Especially if, as I think was the case for Twitter, the knots were tied by someone else (in this case, the users of Twitter themselves).

But Elon Musk is not one to trust someone else’s knots. He’s made his fortune by disregarding other people’s work and rethinking things from first principles. To his credit, he’s worked miracles in categories most entrepreneurs would never dream of tackling, from electric cars to rockets to satellite internet service. There may be only a handful of people who could’ve pulled off Tesla and SpaceX, and maybe only one who could’ve done both. When the game is man versus nature, he’s an obvious choice. When it comes to man versus human nature, on the other hand…

This past year, for the first time, I could see the end of the road for Twitter. Not in an abstract way; I felt its decline. Don’t misunderstand me; Twitter will persist in a deteriorated state, perhaps indefinitely. However, it's already a pale shadow of what it was at its peak. The cool kids are no longer sitting over in bottle service knocking out banger tweets. Instead, the timeline is filled with more and more strangers the bouncer let in to shill their tweetstorms, many of them Twitter Verified accounts who paid the grand fee of $8 a month for the privilege. In the past year, so many random meetings I have with one-time Twitter junkies begin with a long sigh and then a question that is more lamentation than anything else: “How did Twitter get so bad?”

It’s sad, but it’s also a fascinating case study. The internet is still so young that it’s still momentous to see a social network of some scale and lifespan suddenly lose its vitality. The regime change to Elon and his brain trust and the drastic changes they’ve made constitute a natural experiment we don’t see often. Usually, social networks are killed off by something exogenous, usually another, newer social network. Twitter went out and bought Chekhov’s gun in the first act and use it to shoot itself in the foot in the third act. Zuckerberg can now extend his quip about Twitter being a clown car that fell into a gold mine.

In The Rise and Decline of Nations, Mancur Olson builds on his previous book The Logic of Collective Action: Public Goods and the Theory of Groups to discuss how and why groups form. What are the incentives that guide their behavior?

One of his key insights is what I think of as his theory of group inertia. Groups are hard to form in the first place. Think of how many random Discord communities you were invited into the past few years and how many are still active. “Organization for collective action takes a good deal of time to emerge” observes Olson.

However, inertia works both before and after product-market fit. Once a group has formed, it tends to persist even after the collective good it came together to provide is no longer needed.

The same is true of social networks. As anyone who has tried to start one knows, it’s not easy to jump-start a social graph. But if you manage by some miracle to conjure one from the void, and if you provide that group with a reasonable set of ways for everyone to hang out, network effects can keep the party going long after last call. The group inertia that is your enemy before you’ve coalesced a community is your friend after it’s formed. Anyone who’s ever hosted a party and provided booze knows it’s often hard to get the last stragglers to leave. We are a social species.

No social network epitomizes this more than Twitter. It’s not that Twitter was a group of users that assembled for the explicit goal of producing some collective good. Its rise was too emergent to fit into any such directed narrative. But the early years of inertial drag (for years it was literally inert, inertia and inert sharing the same etymological root) followed by later years of inertial momentum fit the broad arc of Olson’s group theory.

I revisit Olson and Twitter’s history because the specifics of how Twitter found product-market fit are critical to understanding its current dissolution. Social networks are path dependent. This is especially true in the West where social networks are largely ad-subsidized and where they’re almost all built around a singular dominant architecture of an infinite scrolling feed optimized for serving ads on a mobile phone. The path each network took to product-market fit selected for a specific user base. As with any community, but especially ones forced to cluster in close proximity in a singular feed, as is common in the West, the people making up the community go a long way towards determining its tenor and values. Its vibes. The composition of its users then determines how conducive that network is to what types of advertising and at what scale. Finally, closing the circle of life, those ad dynamics then influence the network’s middle age evolution as a service. Money may not begin the conversation, that starts with the users, but money gets the final word.

Of all the social networks that achieved some level of scale in this first era of social media, perhaps no other was tried and abandoned by as many users as Twitter. Except for the extremely online community in which I’m deeply embedded (and that I suspect many of my readers are a part of), most normal, well-adjusted humans churned out of Twitter long ago.One of the trickiest things about projecting off of early growth rates for startups in tech is that even fads can generate massive absolute numbers early on if marketed broadly to a global audience. Without looking at early retention and churn rates, you may extrapolate a much larger terminal user base size than will actually stick around. Think about eBay or Groupon, for example. This same caution needs to be applied to Threads; one of the central questions is whether Twitter reached all the people who enjoy microblogging or whether Meta has some magic formula that will allow it to scale to a much larger population. That’s not ideal from a business perspective, but the upside is that those who made it through that great filter selected hard into Twitter’s unique experience. Most sane people don’t enjoy seeing a bunch of random bursts of text from strangers one after the other, but those that do really really love it. And, despite Twitter’s notoriously slow rate of shipping new features over the years, it eventually offered just enough knobs and dials for its users to wrestle their timelines into a fever dream of cacophonous public discourse that hasn’t been replicated elsewhere. More than any other social network, Twitter was one its users seized control of and crafted into something workable for themselves. To its heaviest and most loyal users, it felt at times like a co-op. Recent events remind us it isn’t.

Out of a petri dish that was lifeless for years emerged a culture of creatives, trolls, humorists, politicians, and other public intellectuals screaming at each other in 140 and later 280-character bursts, with even more users quietly gawking from the sideline. This so-called new town square was a 24/7 nightclub for real-world introverts but textual extroverts. My tribe.

This was as entertaining a spectacle as it was shaky a business. Twitter ads have always been hilariously random, and it’s to the credit of the desirable demographics of many of its users that advertisers continued to stick around to have their brands paraded between sometimes questionable, often horrifyingly offensive tweets. But its poor economics as a business shielded it from direct competition. Even if you could recreate its nerdy gladiatorial vibe, why would you? For years it seemed Twitter might persist in this delicate equilibrium, a Galapagos tortoise sunning on an island all to itself, surrounded by ocean as far as the eye could see.

Back to Olson: “Selective incentives make indefinite survival feasible. Thus those organizations for collective action, at least for large groups, that can emerge often take a long time to emerge, but once established they usually survive until there is a social upheaval or some other form of violence or instability.”

Well, “violence and instability” finally came to Twitter in the form of Elon Musk’s ownership. In almost every way, his stewardship has been the polar opposite of the previous regime’s. Politically, to be sure. But more notably, whereas Twitter was previously known as a company that rarely shipped any substantial changes, new Twitter seemed for months to ship things before having thought them through or even QA’ing them. Random bugs seem to pop up in the app all the time, and changes were pushed out and then reversed within the day. Many a day this past year, Twitter has been the main character of the types of drama it used to serve as the forum to discuss.

In classic Twitter fashion, the irony is that it now seems to be in decline not from doing too little but from doing too much. It turns out the way to overcome Olson’s group inertia is to run in swinging a machete, cutting wires, firing people, unplugging computers, flipping switches, tweaking parameters, anything to upset an ecosystem hanging on by a delicate balance. It was, if nothing else, a fascinating natural experiment in how to nudge a network out of longstanding homeostasis.

Given that Musk ended up having to overpay for Twitter by upwards of 4X, thanks to Delaware Chancery Court, it’s not at all surprising he and his new brain trust might choose to take an active hand in trying to salvage as much of his purchase price as possible.

But this heavy-handed top-down management approach runs counter to how Twitter achieved its stable equilibrium. In this way, Musk’s reign at Twitter resembles one of James Scott’s authoritarian high modernist failures. Twitter may have seemed like an underachieving mess before, but its structure, built up piece by piece by users following, unfollowing, liking, muting, and blocking over years and years in a continuous dialogue with the feed algorithm? That structure had a deceptive but delicate stability. Twitter and its users had assembled a complex but functional community, Jane Jacobs style. Every piece of duct tape and every shim put there by a user had a purpose. It may have been Frankensteinian in its construction, but it was our little monster.

This democratic evolution has long been part of Twitter’s history. Many of Twitter’s primary innovations like hashtags, much of its terminology like the word tweets, seemed to come bottom-up from the community of users and developers. This may have capped its scalability; a lot of its syntax has always seemed obtuse (who can forget how you had to put a period before a username if it opened a tweet so that the network wouldn’t treat it as a reply and hide it in the timeline). But, conversely, the service seemed to mold itself around the users who stuck with its peculiar vernacular. After all, they were often the ones who came up with it.

Olson again:

Stable societies with unchanged boundaries tend to accumulate more collusions and organizations for collective action over time.


What established the boundaries of Twitter? Two things primarily. The topology of its graph, and the timeline algorithm. The two are so entwined you could consider them to be a single item. The algorithm determines how the nodes of that graph interact.

The machine learning algorithms have been crucial to scaling our largest social media feeds. They are among the most enormous social institutions in human history, but we don't often think of them that way. It's often remarked upon that Facebook is larger than any country or government, but it should be remarked upon more? I think it's so shocking and horrifying to so many people that they prefer to block it out of their mind. In a literal sense, Twitter has always just been whose tweets show up in your timeline and in what order.

In the modern world, machine learning algorithms that mediate who interacts with whom and how in social media feeds are, in essence, social institutions. When you change those algorithms you might as well be reconfiguring a city around a user while they sleep. And so, if you were to take control of such a community, with years of information accumulated inside its black box of an algorithm, the one thing you might recommend is not punching a hole in the side of that black box and inserting a grenade.

So of course that seems to have been what the new management team did. By pushing everyone towards paid subscriptions and kneecapping distribution for accounts who don’t pay, by switching a TikTok style algorithm, new Twitter has redrawn the once stable “borders” of Twitter’s communities.

This new pay-to-play scheme may not have altered the lattice of the Twitter graph, but it has changed how the graph is interpreted. There’s little difference. My For You feed shows me less from people I follow, so my effective Twitter graph is diverging further and further from my literal graph. Each of us sits at the center of our Twitter graph like a spider in its web built out of follows and likes, with some empty space made of blocks and mutes. We can sense when the algorithm changes. Something changed. The web feels deadened.

I’ve never cared much about the presence or not of a blue check by a user’s name, but I do notice when tweets from people I follow make up a smaller and smaller percentage of my feed. It’s as if neighbors of years moved out from my block overnight, replaced by strangers who all came knocking on my front door carrying not a casserole but a tweetstorm about how to tune my ChatGPT and MidJourney prompts.

I tried switching to the Following from the For You feed, but it seems the Following feed is strictly reverse chronological. This is a serious regression to the early days of Twitter when you had to check your feed frequently to hope to catch a good tweet from any single person you followed. We tried this before; it was terrible then, it’s terrible now.

This weakening of the follow works in the other direction, too. Many people who follow me tell me they don’t see as many of my tweets as they used to. All my followers are accumulated social capital that seem to have been rendered near worthless by algorithmic deflation.

With every social network, one of the most important questions is how much information the structure of the graph itself contains. Because Twitter allows one-way following, its graph has always skewed towards expressing at least something about the interests of its users. Unlike on Facebook, I didn’t blindly follow people I knew on Twitter. The Twitter graph, more than most, is an interest graph assembled from a bunch of social graphs standing on each other’s shoulders wearing an interest graph costume. Not perfect, but not nothing.

The new Twitter algorithm tossed that out.

If you’re going to devalue the Twitter graph’s core primitive, the act of following someone, you’d better replace it with something great. The name of the new algorithmic feed hints at what they tried: For You. It’s nomenclature borrowed from TikTok, the entertainment sensation of the past few years.

I’ve written tens of thousands of words on TikTok in recent years (my three essays on TikTok are here, here, and here), and I won’t rehash it all here. What prompted my fascination with the app was that it attacked the Western social media incumbents at an oblique angle. In TikTok and the Sorting Hat, I wrote:

The idea of using a social graph to build out an interest-based network has always been a sort of approximation, a hack. You follow some people in an app, and it serves you some subset of the content from those people under the assumption that you’ll find much of what they post of interest to you. It worked in college for Facebook because a bunch of hormonal college students are really interested in each other. It worked in Twitter, eventually, though it took a while. Twitter's unidirectional follow graph allowed people to pick and choose who to follow with more flexibility than Facebook's initial bi-directional friend model, but Twitter didn't provide enough feedback mechanisms early on to help train its users on what to tweet. The early days were filled with a lot of status updates of the variety people cite when criticizing social media: "nobody cares what you ate for lunch."

But what if there was a way to build an interest graph for you without you having to follow anyone? What if you could skip the long and painstaking intermediate step of assembling a social graph and just jump directly to the interest graph? And what if that could be done really quickly and cheaply at scale, across millions of users? And what if the algorithm that pulled this off could also adjust to your evolving tastes in near real-time, without you having to actively tune it?

The problem with approximating an interest graph with a social graph is that social graphs have negative network effects that kick in at scale. Take a social network like Twitter: the one-way follow graph structure is well-suited to interest graph construction, but the problem is that you’re rarely interested in everything from any single person you follow. You may enjoy Gruber’s thoughts on Apple but not his Yankees tweets. Or my tweets on tech but not on film. And so on. You can try to use Twitter Lists, or mute or block certain people or topics, but it’s all a big hassle that few have the energy or will to tackle.


This is more commonly accepted now, but back in 2020 when I wrote this piece, TikTok’s success was still viewed with a lot of skepticism and puzzlement. Since then, we’ve seen Instagram and Twitter both try emulating TikTok’s strategy. Both Instagram and Twitter now serve much less content from people you follow and more posts selected by machine learning algorithms trying to guess your interests.

Instagram has been more successful in part because it has formats like Stories that keep content from one’s follows prominent in the interface. There’s social capital of value embodied in the follow graph, and arguably it’s easier for Instagram to preserve much of that while copying TikTok than it is for TikTok to build a social graph like Instagram.

But that’s a topic for another day. Twitter is the app on trial today. And of all Twitter’s recent missteps, I think this was the most serious unforced error. For a variety of design reasons, Twitter will likely never be as accurate an interest graph as, say, TikTok is an entertainment network.

As I’ve written about before in Seeing Like An Algorithm, Twitter’s interface doesn’t capture sentiment, both positive and negative, as cleanly, as TikTok.

Let’s start with positive sentiment. On this front, Twitter is…fine? It’s not for lack of usage. I’ve used Twitter a ton over more than a decade now, I’ve followed and unfollowed thousands of accounts, liked even more tweets, and posted plenty of tweets and links. I suspect one issue is that many tweets don’t contain enough context to be accurately classified automatically. How would you classify a tweet by Dril?

But perhaps even more damning for Twitter is its inability to see negative sentiment. Allowing users to pay for better tweet distribution leaves the network vulnerable to adverse selection. That’s why the ability to capture negative sentiment, especially passive negative sentiment, is so important to preserving a floor of quality for the Timeline.

Unfortunately, capturing that passive disapproval is something Twitter has never done well. In Seeing Like an Algorithm, I wrote about how critical it was for a service’s design to help machine learning algorithms “see” the necessary feedback from users, both positive and negative. That essay’s title was inspired by Scott’s Seeing Like a State which described how high modernist governments depended on systems of imposed legibility for a particular authoritarian style of governance.

Modern social networks lean heavily on machine learning algorithms to achieve sufficient signal-to-noise in feeds. To manually manage complex adaptive systems at the scale of modern social media networks would be impossible otherwise. One of the critiques of authoritarian technocracies is that they quickly lose touch with the people they rule over. It's no surprise that such governments have also looked at machine learning algorithms paired with the surveillance breadth of the internet as a potential silver bullet to allow them to scale their governance. The two entities that most epitomize each of these both come out of China: Bytedance and the CCP. The latter, in particular, has long been obsessed with cybernetics, despite having followed it down a disastrous policy rabbit hole before.

But these cybernetic systems, in the Norbert Wiener sense, only work well if their algorithms see enough user sentiment and see it accurately. Just as Scott felt high modernism failed again and again because those systems overly simplified complex realities, Twitter’s algorithm operates with serious blind spots. Since every output is an input in a cybernetic system, failure to capture all necessary inputs leads to noise in the timeline.

Twitter doesn’t see a lot of passive negative sentiment; it’s a structural blind spot. In a continuous scrolling interface with multiple tweets on screen at any one time, it’s hard to tell disapproval from apathy or even mild approval because the user will just scroll past a tweet for any number of reasons.

This leads to a For You page that feels like it’s missing my friends and awkwardly misinterpreting my interests. Would you like yet another tweetstorm on AI and how it can change your life? No, well too bad, have another. And another. For someone who claims to be worried about the dangers of AI, Elon’s new platform sure seems to be pushing us to play with it.

In the rush to copy TikTok, many Western social networks have misread how easy it is to apply lessons of a very particular short video experience to social feeds built around other formats. If you’re Instagram Reels and your format and interface are a near carbon copy, then sure, applying the lessons of my three TikTok essays is straightforward. But if you’re Twitter, a continuous scrolling feed of short textual content, you’re dealing with a different beast entirely.

Even TikTok sometimes seems to misunderstand that its strength is its purity of function as an interest/entertainment graph. Its attempts to graft a social graph onto that have struggled because social networking is a different problem space entirely. Pushing me to follow my friends on TikTok muddies what is otherwise a very clear product proposition. Social networking is a complex global maximum to solve for. In contrast, entertaining millions of people with an individual channel personalized to each of them is an agglomeration of millions of local maximums. TikTok’s interface paired with ByteDance’s machine learning algorithms are perfect for solving the latter but much less well-suited towards social networking.

Here’s another way to think about it. The difference between Twitter and an algorithmic entertainment network like TikTok is that you could fairly quickly reconstitute TikTok even without its current graph because its graph is a much less critical input to its algorithm than the user reactions to any random sequence of videos they’re served.

If Twitter had to start over without its graph, on the other hand, it would be dead (which speaks to why Twitter clones like BlueSky which are just Twitter minus the graph and with the same clunky onboarding process seem destined for failure). The new For You feed gives us a partial taste of what that might look like, and it's not pretty.

I ran a report recently on all the accounts I follow on Twitter. I hadn’t realized how many of them had been dormant for months now. Many were people whose tweets used to draw me to the timeline regularly. I hesitate to unfollow them; perhaps they’ll return? But I’m fooling myself. They won’t. Inertia again. A user at rest tends to stay at rest, and a user that flees tends to be gone for good.

Even worse, many accounts I follow look to have continued to tweet regularly over the past year. I just don’t see their tweets anymore. The changes to the Twitter algorithm bulldozed over a decade’s worth of Chesterton fences in a few months.


The other prominent mistake of the Elon era is more commonly cited, and I tend to think it’s overrated, but it certainly didn’t help. It’s the type of mistake only a prominent and polarizing figure running a social network could stumble into: his own participation on the platform he owns. The temptation is understandable. If you overpaid for a social network by tens of billions of dollars, why shouldn’t you be able to use it as you please? Why not boost your own tweets and use it as a personal megaphone? Why buy a McLaren if you take it for a spin and total it? He declared that one of his reasons for purchasing Twitter was to restore it to being a free speech platform, so why not speak his mind?

More than any tech CEO, he’s become a purity test for one’s technological optimism. His acolytes will follow him, perhaps even literally, to Mars, while his critics consider him the epitome of amoral Silicon Valley hubris. That he is discussed in such simplistic, binary terms is ironic; it exemplifies the nature of discourse on Twitter. It’s no surprise that many Twitter alternatives market themselves simply as Twitter minus Elon (though I suspect most people just want, like me, a Twitter with the same graph but minus the new For You algorithm).

But there’s a Heisenberg Uncertainty Principle of social in play here. Every tweet of his alters the fabric of Twitter so drastically that it’s almost impossible for some users to coexist on Twitter alongside him. He singlehandedly brought some users back to Twitter and sent others fleeing for the exits. There are no neutral platforms, as many have noted, but Musk’s gravitational field has warped Twitter’s entire conversational orbit and brand trajectory. Leaving Twitter, or simply refusing to pay for verification, is now treated as an act of resistance. It’s debatable whether that’s fair, but reality doesn’t give a damn.

Some users might have stuck around had Musk used his Twitter account solely for business pronouncements, but that wouldn’t be any fun now would it? He’s always enjoyed trolling his most vocal critics on Twitter, but it hits different when he’s the owner of said platform used by millions of cultural elites the world over.

Earlier this year, it appeared that Musk had comped Twitter Verified blue checkmarks to prominent public figures like Stephen King, some of whom had repeatedly criticized him. This led to the absurd and prolonged spectacle of dozens of famous people asserting over and over that they had absolutely not paid the meager sum of $8 a month for the scarlet, err, baby blue checkmark that now adorned their profiles, not to be confused with the blue checkmark that formerly appeared in the same spot that they hadn’t paid for. This made the blue checkmark a sort of Veblen good; more people seemed to want one when you couldn’t buy one, when it was literally priceless.The price is an odd one. $8 a month is not expensive enough to be a wealth signal, but it’s enough to feel like an insult to users who feel like they subsidized the popularity of Twitter over the years with their pro bono wit. I believe it was Groucho Marx who once said something to the effect of not wanting to belong to any club that would accept him as a member for the tidy sum of $8 a month.

This culminated in one weekend when Musk engaged in a protracted back and forth with Twitter celebrity shitposter Dril, pinning a Twitter Blue badge on his profile over and over only to have Dril remove it by changing his profile description. This went on for hours, and some of us followed along, like kids on the playground watching a schoolboy chase a girl holding a frog. This was bad juju and everyone knew it.


I’ll miss old Twitter. Even now, in its diminished state, there isn’t any real substitute for the experience of Twitter at its peak. Compared to its larger peers in the social media space, Twitter always reminded me of Philip Seymour Hoffman’s late-night speech as Lester Bangs in Almost Famous, delivered over the phone to the Cameron Crowe stand-in William Miller, warning him about having gotten seduced by Stillwater, the band Miller was profiling for The Rolling Stone:

Oh man, you made friends with them. See, friendship is the booze they feed you. They want you to get drunk on feeling like you belong. Because they make you feel cool, and hey, I met you. You are not cool. We are uncool. Women will always be a problem for guys like us, most of the great art in the world is about that very problem. Good-looking people they got no spine, their art never lasts. They get the girls but we’re smarter. Great art is about guilt and longing. Love disguised as sex and sex disguised as love. Let’s face it, you got a big head start. I’m always home, I’m uncool.

The only true currency in this bankrupt world is what you share with someone else when you’re uncool. My advice to you: I know you think these guys are your friends. You want to be a true friend to them? Be honest and unmerciful.

In the world of Almost Famous, Instagram would be the social network for the Stillwaters, the Russell Hammonds, the Penny Lanes. Beautiful people, cool people. Twitter was for the uncool, the geeks, the wonks, the wits, the misfits. Twitter was honest and unmerciful, sometimes cruelly so, but at its best it felt like a true friend.

It was striking how many of Elon’s early tweets about Twitter’s issues seemed to pin Twitter’s underperformance on engineering problems. Response times, things of that nature. But Twitter’s appeal was never a pure feat of engineering, nor were its problems solely the fault of engineering malpractice. They were human in nature. Twitter isn’t, as many have noted, rocket science, making it a particularly tricky domain for a CEO of, among other things, a rocket company. Ironically, Norbert Wiener, often credited as the father of cybernetics, a field which has lots of relevance to analyzing social networks, worked on anti-aircraft weapons during World War II. So if you really want to nitpick, your vast conspiracy board might somehow connect running a social network to rocket science. You can test unmanned rockets, and if they blow up on take-off or re-entry, you’ve learned something, no harm done. But running the same test on a social media service is like testing rockets with your users as passengers. Crash a rocket and those users aren’t going to be around for the next test flight.

It’s not clear there will ever be a Twitter replacement. If there is one, it won’t be the same. It may look the same, but it will be something else. The internet is different now, and the conditions that allowed Twitter to emerge in the first place no longer exist. The Twitter diaspora has scattered to all sorts of subscale clones or alternatives, with no signs of agreeing on where to settle. As noted social analyst Taylor Swift said, “We are never ever getting back together.”

For this reason, Twitter won’t ever fully vanish unless management pulls the plug. None of the contenders to replace Twitter has come close to replicating its vibe of professional and amateur intellectuals and jesters engaged in verbal jousting in a public global tavern, even as most have lifted its interface almost verbatim. Social networks aren’t just the interface, or the algorithm, they’re also about the people in them. When I wrote “The Network’s the Thing” I meant it; the graph is inextricable from the identity of a social media service. Change the inputs of such a system and you change the system itself.

Thus Twitter will drift along, some portion of its remaining users hanging out of misguided hope, others bending the knee to the whims of the new algorithm.

But peak Twitter? That’s an artifact of history now. That golden era of Twitter will always be this collective hallucination we look back on with increasing nostalgia, like alumni of some cult. With the benefit of time, we’ll appreciate how unique it was while forgetting its most toxic dynamics. Twitter was the closest we’ve come to bottling oral culture in written form.

Media theorist Harold Innis distinguished between time-biased and space-biased media:

The concepts of time and space reflect the significance of media to civilization. Media that emphasize time are those durable in character such as parchment, clay and stone. The heavy materials are suited to the development of architecture and sculpture. Media that emphasize space are apt to be less durable and light in character such as papyrus and paper. The latter are suited to wide areas in administration and trade. The conquest of Egypt by Rome gave access to supplies of papyrus, which became the basis of a large administrative empire. Materials that emphasize time favour decentralization and hierarchical types of institutions, while those that emphasize space favour centralization and systems of government less hierarchical in character.

Twitter always intrigued me because it has elements of both. It always felt like it compressed space—the timeline felt like a single lunch room hosting a series of conversations we were all participating in or eavesdropping on—and time—every tweet seemed to be uttered to us in the moment, and so much of it was about things occurring in the world at that moment (one of the challenges of machine learning applied to news and Tweets both is how much of it has such a short half-life versus the more evergreen nature of TikToks, YouTube videos, movies, and music. A lot of Twitter was textual, but the character limit and the ease of replying lent much of it an oral texture. It felt like a live, singular conversation.

When reviewing a draft of this piece, my friend Tianyu wrote the following comment, which I’ll just cite verbatim, it’s so good:

Twitter feels like a perfect example of what James W. Carey calls the "ritual view of communication" (see Communication as Culture). Its virality doesn't come from transmission alone, but rather the quasi-religiosity of it; scrolling Twitter while sitting on the toilet is like attending a mass every Sunday morning. Like religions, Twitter formulates participatory rituals that come with a public culture of commonality and communitarianism. These rituals are then taken for granted—they become how people on the internet consume information and interact with one another by default.

Religious rituals rise and fall. Today all major religions have, at some point, become a global mimesis through missionary work, political power, and imperial expansions. Musk's regime is basically saying, 'oh well, Christianity isn't expanding fast enough. What we need to do is to rewrite the Bible and abolish the clergy. That'll do the work.'

Carey often notes that communication shares the same roots as words like common, community, and communion. Combine the ritualistic nature of Twitter with its sense of compressing space and time and you understand why its experience was such a convincing illusion of a single global conversation. I suspect Carey would argue that the simulacrum of such a conversation effectively created and maintained a community.

Even the vocabulary used to describe Twitter reinforced its ritualistic nature. Who would be today’s main character, we’d ask, as if that day’s Twitter drama was a single narrative we were all reading. We’d go to see the list of Trending Topics for the day as if looking to see who was being tarred and feathered in the Twitter town square that day. There was always a mob to join if you wanted to cast a stone, or a meme template of the day to borrow.

Friends would forward me tweets, and at some point I stopped replying “Oh yeah I saw that one already” because we had all seen all of them already. Twitter was small, but more importantly, it felt small. Users often write about how Twitter felt worse once they exceeded some number of followers, and while there are obvious structural reasons why mass distribution can be unpleasant, one underrated drawback of a mass following was the loss of that sense of speaking to a group of people you mostly knew, if not personally, then through their tweets.

In a way, Twitter’s core problem is so different than that of something like TikTok, which, as I noted earlier, is a challenge of creating a local maximum for each user. Twitter at its best felt, like Tianyu described it to me, a global optimum. In reality, it’s never so binary. Even in a world of deep personalization, we want shared entertainment and grand myths, and vice versa. TikTok has its globally popular trends and Twitter its micro-communities. But a TikTok-like algorithm was always going to be particularly susceptible to ruining the cozy, communal feel of a scaled niche like Twitter.

I’ve met more friends in the internet era through Twitter than any other social media app. Some of my closest friends today first entered my life by sliding into my DM’s, and it saddens me to see the place emptying out.

All of this past year, as a slow but steady flow of Twitter’s more interesting users has made their way to the exits, unwilling to fight to be heard anymore, or just stopped tweeting, I’ve still opened the app daily out of habit, and to research for pieces like this. But the vibes are all off. I haven’t churned yet, but at the very least, I’ve asked the bartender to close out my tab.

If Twitter’s journey epitomizes the sentimental truism that the real treasure was the friends we made along the way, then the story of its demise will begin the moment we could no longer find those friends on that darkened timeline.


ACKNOWLEDGMENTS: Thanks to my friends Li and Tianyu for reading drafts of this piece at various stages and offering such rapid feedback. Considering the length of my pieces, that's no small thing. Their encouragement and useful notes and questions helped me refine and clarify my thinking. Also, if it wasn’t for Twitter, I probably wouldn’t know either of them today.

Inspiration for the title of this post comes from this which is based on this.

As my own Twitter usage fades, I plan to ramp back up writing on my website. If you're interested in keeping up, follow my Substack which I plan to spin back up to keep folks updated on my latest writing and where I’ll drop, among other things, a follow-up to this piece with my thoughts on Threads.

My Favorite Movies of 2021

A second year of the pandemic passed in which I didn’t attend any film festivals in person. I miss it. My viewing output of is lower than usual but still much much higher than that of the median filmgoer.

Film is one category of media in which human recommendations still feel superior to algorithmic ones. It is notable that none of my favorite Netflix movies this year came via their recommendations. Some I might have never heard of had some critic or friend not written about them.

Film remains a difficult category for machine learning to crack. Most people only watch movies once. In a category like music, people listen to their favorite tracks repeatedly. Films are very long while music tracks only last a few minutes. As a result, the frequency of feedback is much higher for music than film.

Viewers generally provide a single point of feedback on a film, if they even choose to sample it: they either finish the movie or they don’t. In music, you not only gather many more data points per hour because of the short duration of each track, but you gather feedback within each piece. People hit skip, or rewind, or repeat. People add songs to playlists or ask their streaming service to generate radio stations off of that track.

As I’ve written before about TikTok, one of its most critical design choices was to full-screen videos, allowing it to gather really accurate signal from the viewer on each video. TikTok videos are even shorter than music tracks, but they often contain snippets of music tracks. In many ways a TikTok is about as short a piece of media as could be designed that can be said to still tell a narrative (though maybe a dating app profile photo is even more concise).

The ways that music tracks resemble each other feel easier to see with math. This makes it easier to generate a playlist of similar tracks even before gathering listener feedback. Machine learning algorithms have learned to write music that often sounds like specific composer and musicians. I’ve yet to see an algorithm that can just spit out a Wes Anderson-like movie.

It’s no surprise to me that Netflix seems largely to have given up on much of the work that came out of the Netflix Prize and instead focuses on using the massive funnel of its above-the-fold home screen real estate to push its latest original production. I didn’t like Red Notice, but I can understand what types of metrics would lead Netflix to just splash it across every subscriber’s eyeballs.

Film is also a category in which we still haven’t fully understood the variation in people’s aesthetic preferences. Even people I consider to share many of my movie tastes will disagree with me vociferously on particular movies. I doubt anyone will agree with all my movie choices below.

Rather than a bug, this variance in taste is to be treasured. I’m not interested in terse recommendations like “this film is good” or “this film was terrible.” Given the individuality of aesthetic preferences, there’s little signal in a binary expression of one person’s preferences.

Instead, give me a review which can articulate why someone enjoyed a film or not. Some of my favorite reviews are pans of movies I loved, or vice versa. It’s a rare gift for someone to be able to express just how a film works on them given the subconscious and emotional nature of the medium. Moving images are pre-verbal. Something is almost always lost in translation to text. It’s even rarer for someone to be able to tie that to film craft given how visually illiterate our educational systems have left us.

This doesn’t mean I rely exclusively on professional film critics. More and more, I’ve come to rely on the film buffs of Letterboxd to guide my film choices. Unlike Rotten Tomatoes or MetacriticThe way those two sites compress the quality of film into a single numeric score has always been reductive. That's by design, but my aesthetic response to a film can't be mapped that way. Some of my favorite restaurants and books don't rate highly on Yelp or Amazon. Similarly, often it's the movie that's divisive that I find most compelling. Sometimes what you want is a work that attracts you with equal force as it repels others., you can curate your own panel of people to follow and filter film reviews by their tastes. Since many of the members are not professional critics, they don’t feel a need to conform to some standard review template. Many reviews are just humorous quips. Many are just a line or two. But taken as a group, they simulate that feeling of standing outside on the sidewalk after a festival screening, debating the movie with other film buffs.

Pauline Kael made famous a particular type of deeply subjective film criticism. Along with Susan Sontag, she treated as legitimate her very personal aesthetic response to art. The logical successor to that is not any single film critic today but the pluralistic critical response of the public via the Internet. Sometimes it can be toxic and suffocating as in the angrier strains of franchise fandom. Other times, it can feel like a warm fellowship, trying to tease out why some films work for some of us and not others, the nature of the medium's alchemy.

That’s a community I’ve missed these past two years. The pandemic accelerated many trends, and the decline in theater-going is one of them. Studios adapted by pushing even more movies day-and-date. I’ll always prefer to see a movie in theaters, but more than that I just appreciate being able to see movies. Bemoan the death of the mid-budget adult drama all you want, but complaining is not a strategy. I’ve worked too long in the technology industry to know how this plays out. The world changes, and you either change with it or get left behind. These forces sweeping Hollywood are exogenous to its world and will sweep it along regardless of what it does.

For example, the traditional release model for prestige films has always been festival to limited release in NY and LA and then much later to wider release. The pandemic brought some films to VOD more quickly, even day-and-date at times, but in 2021 most prestige indies are still next to impossible to watch unless you live in NYC or LA.

It’s long past the time when this model should be updated. I often hear buzz out of festivals for movies like The Worst Person in the World or Licorice Pizza but then realize I won’t be able to see them until months later, sometimes not until the following year. That type of delayed anticipation is fine for a blockbuster like Batman, but for indie films it is questionable at best. Sometimes I don’t realize that a movie released in theaters until it has already come and gone. That used to never happen in the era before the Big Bang of Content.

In a previous era, this staged build-up of anticipation worked for indie films. Now, it actively hurts them. When the public is bombarded with what is effectively an infinite number of contenders for their attention, movies need publicity and availability to crest together.

Furthermore, the idea of a movie moving through a period of unavailability because of a gap in release windows is just absurd in an age of abundance. Once a movie has left theaters, it should always be available somewhere for people who want to seek it out. Windowing worked great in a content scarce world where people would wait patiently for some piece of media to hit the market, but nowadays, it just means an audience whose attention will get diverted elsewhere. I’m still amazed by how many movies I can’t find streaming anywhere even though they’ve left their theatrical run. This hurts a specific type of movie more than others, and it’s not the Spider-Man: No Way Home’s of the world.

In his new book The Nineties, Chuck Klosterman chronicles how the rise of a the video rental store like Blockbuster spawned a new and specific type of cinephile. Never before had so many movies been available to watch on near demand, and people from Kevin Smith to Quentin Tarantino had their film tastes broadened by exposure to movies from around the world, in all sorts of genres. The combination of the VCR and video stores enabled an explosion of cinephilia. I was one of those freshly minted film buffs, birthed in dimly lit aisles housing one box cover after another of films I'd never heard of.It began for me in grade school when my father would rent films from the library, then Hollywood and Blockbuster video, and would reach full bloom when I moved to Seattle to work at Amazon and discovered Scarecrow Video. It was there that I'd rent Criterion edition Laserdiscs of movies and a Laserdisc player to play them on. I have such fond memories of putting down deposits of a few hundred dollars in case I somehow absconded with the the Criterion Edition Laserdisc of John Woo's The Killer or something like that.

We now have, via the internet, the ability to make every film available on demand at all times. We've already seen what Netflix licensing and streaming content from all over the world has done for people's exposure to international film and television. Studios need to ensure that it's as easy as possible to fall into a lifelong romance with the medium. This is an aesthetic abundance strategy for an industry which spent its entire history built around scarcity-based business models. It's not that I don't love the occasional screening of a rare 70mm print of some film. It's that withholding things in an age of abundance is more likely to make the public forget it entirely than to seek it out.

My last memory from this past year is the escalation in what’s commonly referred to as the Discourse (capital D because it’s a very specific, modern form I’m referencing). It’s not just the world of film that’s been prey to this as it’s an output of the structure of Western social media.

Any film lover on social media will be familiar with some forms of it. The most prominent was the Scorsese versus MCU debate. Then it was the debate over The Oscars, and this past week it's arguments over whether Steven Spielberg should direct a remake of BullittEven if you aren't a fan of Spielberg's sentimentality, he is an S-Tier mover of the camera. The way some people worship the linguistic stylings of certain writers, I know few film buffs who don't stand in awe of how Spielberg chooses to cover a scene. We need an 80 hour documentary that consists solely of Spielberg and his DP's discussing how and why they choose to move the camera a specific way in every scene of every movie he ever directed. Purity tests are especially useful when roaming a threat-filled landscape, to separate friend from foe. It just so happens that Western social media is just such a post-apocalyptic desert of tribal warfare.

The Scorsese-MCU debate is an ideal purity test because the MCU movies are the most watched films in the world now. Almost anyone has at least heard of if not seen at least one MCU movie. That means even a casual filmgoer can be tossed in the water like a witch to see if they float. Spoiler alert: everyone floats.

As with a mistaken mass bcc: email incident, the only way to make the Discourse stop is to ignore it. But given enough participants, it can't be helped. Someone always presses reply all to request to be removed from the distribution, which leads to people asking to be removed, which leads to other people telling them to stop replying all.

This same snowball effect propels arguments like the Scorsese-MCU debate. Like Neil McCauley headed to freedom near the end of Heat, with Eady in the passenger seat next to him, we should just drive on. But it’s irresistible to weigh in, and so we yank the steering wheel to the right, cut through three lanes of traffic to the exit, just so we can hunt down Waingro to let him know that Scorsese possesses more talent in his pinkie than every MCU director put together, or that Spider-Man: No Way Home deserves the Best Picture Oscar, or whatever.

This type of exhausting Discourse is a headless, distributed phenomenon. It’s a monster we animate, and it only lives because we keep feeding it our own anger. Even complaining about the Discourse is part of the Discourse. It changes nothing except to punish and exhaust the participants.

In Simulacra and Simulation, Baudrillard writes of "the map that precedes the territory—precession of simulacra—that engenders the territory...it is the territory whose shreds slowly rot across the extent of the map." Today the Discourse begins as an illusory shadow online and then assumes corporeal form. This is the reflexive loop between the internet and the world at large: we put the ghost into the machine, then we pull it out of the machine with a look of surprise.

I beg of you, don’t feed the Discourse. We’re all better than that. I confront enough tribal debate in every other aspect of my online life, I just want to preserve movies as a peaceful corner of civilized dialogue. The worst type of prisoner’s dilemma is one which the two prisoners construct themselves, where they defect against each other when there are no prison guards or police to enforce any judgment. We're playing ourselves.

Movies I Enjoyed This Year

In no particular order...

The Power of the Dog

I love Westerns, one of the most storied of Western film genres, and this year added a new entry to the syllabus.

The Power of the Dog’s violence is of the psychological variety. If your ideal Western consists of six-shooters at high noon, just know that much of the conflict in this movie is waged via banjos, pianos, the occasional venomous quip, and leather weaving. I mean, one of this movie’s main characters is present only via his old saddle hanging in a barn.

Some people won't consider this much of a Western at all. But the power of this genre is its ability to speak to so much of the human condition. The West has always represented the frontier in the American imagination, a place where one goes to try to escape structure, the place of maximal freedom, but it also represents a site where society can be built anew. That tug and pull is core to the genre.

Campion explores this tension in a new way. Many of the characters in this film, from Benedict Cumberbatch and Jesse Plemons' Phil and George Burbank to Kirsten Dunst's Rose Gordon to Kodi Smit-McPhee's Peter Gordon, are in search of the both the freedom and the community promised by the West. But each runs into invisible structures imposed by society and culture, and each tries to cope in their own way.

Benedict Cumberbatch’s acting style often feels overly theatrical. In many movies it’s a distraction. Here, it’s perfect. His Phil Burbank’s cruelty is itself a conscious pose, for reasons we learn by movie’s endThe least believable thing in the movie is that Benedict Cumberbatch and Jesse Plemons could be biological brothers. Someone check a photo of the Pony Express delivery guy.. Cumberbatch menaces every frame of this movie; he is Chekhov’s gun, or so we’re meant to believe.

Campion is a master of visual iconography that lends her films a psychoanalytic portent. Who can forget Holly Hunter underwater, tied to the anchor of her titular piano? In The Power of the Dog, a character stumbles upon a tunnel in the woods near town. To enter it and traverse to the other side is to crawl back through a birth canal into a mother’s womb, to a place of psychic security and unconditional love. I won’t ruin who built the tunnel or where it leads, but the movie is full of imagery that burrows into your subconscious. Even the title is cryptic, forebodingIt comes from Psalm 22:20. “Deliver my soul from the sword; my precious life from the power of the dog.”. In the Bible it references Jesus on the cross, his is the precious life. In the case of Campion's film, there is more than one person who could be the precious life, and more than one person or force who could be the power of the dog. To say more would be a spoiler; the fun is in working it out for yourself by movie's end..

The end of the movie is a bit of a shock, but walk the movie back in your head and the clues were there all along.

The Lost Daughter

I don’t know that Netflix has to continue to fund arthouse films in an effort to win a Best Picture Oscar, but I understand the impulse. Despite the precipitous decline in the ratings of the Oscars, almost every one I know would lose their minds just to attend the ceremony. Hollywood’s ability to manufacture its own cultural prestige will live long past the decline of the mid-budget adult drama.

The Lost Daughter is an example of a book adaptation that honors the tone of the source material while recognizing the unavoidable differences in film as a medium. The book is told in the first person, but absent a voice-over, movies have to externalize that type of subjectivity. Maggie Gyllenhaal, in her directorial debut, succeeds in doing so through shot choices and the sheer acting chops of Olivia Colman. Much of the text of the movie consists of long, wordless, tight closeups of Colman’s face.

A latent dread haunts Ferrante’s novels. The Lost Daughter honors that. All the mothers out there who’ve been trapped at home with young children going on some two years now will look upon Colman and think, I understand. Damn it momma, I understand.

The Worst Person in the World

They have trouble making decisions. They would rather hike in the Himalayas than climb a corporate ladder. They have few heroes, no anthems, no style to call their own. They crave entertainment, but their attention span is as short as one zap of a TV dial. They hate yuppies, hippies and druggies. They postpone marriage because they dread divorce. They sneer at Range Rovers, Rolexes and red suspenders. What they hold dear are family life, local activism, national parks, penny loafers and mountain bikes. They possess only a hazy sense of their own identity but a monumental preoccupation with all the problems the preceding generation will leave for them to fix.

This appeared in Time Magazine in July, 1990Quentin Tarantino is clearly Gen X by this definition. His Once Upon a Time in Hollywood is one of the most anti-hippie movies I've ever seen. At movie's end, Brad Pitt and Leonardo Dicaprio take comically violent revenge on the hippies for ruining Hollywood. It takes the place of Dirty Harry, my previous benchmark for most anti-hippie film.

. It was pointed at twenty-somethings at that time, or, as we know them today, Generation XI'm a member of one of the later cohorts of Generation X. I recently read Chuck Klosterman's The Nineties and he makes the point that Generation X is the least annoying of those yet living because we are the smallest in population, exceeded in size by both the Boomers and the Millenials that sandwich us. Since this is a positive rather than descriptive statement, I declare it indisputable. (BANGS GAVEL).

A common description of The Worst Person in the World, the final chapter of Joachim Trier’s Oslo Trilogy, is that it’s about millennial late-twenties, early-thirties indecision. What should I do with my life? Who should I settle down with? When should I have kids, if at all?

But as the Time excerpt shows, most of the same critiques of millennials were directed at the previous generation. This type of pre-midlife-crisis of indecision is more and more common in any post-modern Western society. Think of it as a type of post-industrial paradox of choice. Free of religious, societal, institutional, and cultural guide rails as to how to lead our lives, we find ourselves, like this movie’s protagonist Julie (Renata Reinsve), wandering a maze of options at the age of 30 in a haze of existential bewilderment.

The decline of power structures can be a blessing and a curse. On the one hand, many of them were coercive. In another era, Julie’s career options would have been curtailed even more by sexual discrimination or societal notions of what a woman could be. Back then, even those who earned a taste of freedom had to wait until after their kids had left the nest. Today, for many, the mid-life crisis has been pulled up by two decades. Absent tradition and authority, told we can be anything we want to be, we are trapped by our freedom. The neoliberal marketplace tells us to follow our own desires while assailing us with imagery of what we should covet.

The internet has turned this dynamic collective. Social media cocoons us in the never-ending hall of mirrors of other people’s lives. It has never been easier to visualize the opportunity costs of our own choices, so much so that we gave it an explicit name: FOMO. In the moment, we feel a momentary fear of missing out, but over time, we’re even more haunted by a persistent fear of having missed out for good. It turns out that the promise of unfettered pleasure and choice of the postmodern age was a mirage for many.

Marriage, a stable job, children, all the things Julie foregoes as she explores her freedom are structures that organize the span of one’s life. They are anchor points in one’s timeline. Without them, one’s life can flow any which way. That is both blessing and a curse, as you can feel unmoored, destabilized. The fact that the movie is structured into twelve chapters and an epilogue, with specific titles, is ironic. That the movie is able to impose an artificial structure to what is otherwise a life of spontaneity is only because it is a work of art, created from an explicit artist’s mind. It is less certain whether Julie herself can find a coherent arc in herself.

The description of Liquid Love by Zygmunt Bauman reads:

This book is about the central figure of our contemporary, ‘liquid modern’ times – the man or woman with no bonds, and particularly with none of the fixed or durable bonds that would allow the effort of self-definition and self-assertion to come to a rest. Having no permanent bonds, the denizen of our liquid modern society must tie whatever bonds they can to engage with others, using their own wits, skill and dedication. But none of these bonds are guaranteed to last. Moreover, they must be tied loosely so that they can be untied again, quickly and as effortlessly as possible, when circumstances change – as they surely will in our liquid modern society, over and over again.

Within, Bauman writes:

The principal hero of this book is human relationship. This book’s central characters are men and women, our contemporaries, despairing at being abandoned to their own wits and feeling easily disposable, yearning for the security of togetherness and for a helping hand to count on in a moment of trouble, and so desperate to ‘relate’; yet wary of the state of ‘being related’ and particularly of being related ‘for good’, not to mention forever – since they fear that such a state may bring burdens and cause strains they neither feel able nor are willing to bear, and so may severely limit the freedom they need – yes, your guess is right – to relate…

He references a laboratory study which epitomizes this tension:

In their famous experiment, Miller and Dollard saw their laboratory rats ascending the peak of excitement and agitation when ‘the adiance equalled the abiance’ – that is, when the threat of electric shock and the promise of tasty food were finely balanced…

No wonder that ‘relationships’ are one of the main engines of the present-day ‘counselling boom’. The complexity is too dense, too stubborn and too difficult to unpack or unravel for individuals to do the job unassisted. The agitation of Miller and Dollard’s rats all too often collapsed into a paralysis of action. An inability to choose between attraction and repulsion, between hopes and fears, rebounded as an incapacity to act.

Julie seems to be one of Bauman’s liquid loversSpeaking of our modern liquid times, what could epitomize that more than dating apps? With one swipe, another option appears at our fingertips on our phone screen. Apps like Tinder and Hinge not only represent the postmodern allure of infinite choice but also the triumph of neoliberalism. It's not a coincidence that we use the term dating economy when referring to modern courtship. The market is our solution to everything. It's the same with gaming, where we refer to virtual economies. Baudrillard would surely be both delighted and horrified that so many modern games center around repetitive work. We complain about our actual jobs but embrace virtual work in grind games like Farmville.

Any decision that forecloses future options both attracts and repulses in equal measure. Julie dives in and then flees for another life again and again. In his review of the film in The American Conservative, Matthew Schmitz notes:

Julie knows the risks of intimacy. Love causes suffering. It brings with it the shadow of death, and not just because we injure others and are injured by them. Love requires us to die to self, a foretaste of the death all experience.

Schmitz points at millennial precarity as a subject of the film:

The middle-class life that was the classic setting of the mid-life crisis has become less attainable for millennials, a fact reflected in Julie’s transition from the financially independent Aksel to the hourly worker Eivind. Soon the majority of my fellow millennials will have turned 35, the age Julie is approaching at the end of the film. The oldest millennials are already in their forties. Social scientists have painstakingly described our low rates of marriage, childbearing, and homeownership. Trier gets at something that is harder to capture: the ambivalent experience of people who came of age in these years.

It seemed that we could do what we wanted, except form lasting relationships; go where we liked, unless it was home. For no other generation have the possibilities been so limitless and the reality so limited. The Supreme Court proclaimed that anyone could marry, even as marriage became unattainable for the poor. AirBnB opened up houses across the world, even as houses became something that fewer could afford.

I’m less certain that’s the primary preoccupation of The Worst Person in the World given Norway’s renowned social safety net. Instead, Julie’s story embodies one of the popular critiques of 60’s and 70’s postmodernism which urged a rejection of elite authorities in favor of following our desires. What was promised was a liberation and authenticity.

In Anti-Oedipus: Capitalism and Schizophrenia Gilles Deleuze and Felix Guattari wrote of desire:

It is explosive; there is no desiring-machine capable of being assembled without demolishing entire social sectors. Despite what some revolutionaries think about this, desire is revolutionary in its essence – desire, not left-wing holidays! – and no society can tolerate a position of real desire without its structures of exploitation, servitude, and hierarchy being compromised.

Earlier, they wrote:

Courage consists, however, in agreeing to flee rather than live tranquilly and hypocritically in false refuges. Values, morals, homelands, religions, and these private certitudes that our vanity and our complacency bestow generously on us, have many deceptive sojourns as the world arranges for those who think they are standing straight and at ease, among stable things.

Follow your desires instead of the herd, don’t be a sheep, be your authentic self. A few decades later, similar slogans permeate corporate culture slogans and self-help paeans.

Anti-Oedipal theories promised to throw off our shackles. What we ended up with is more ambivalent. Freud wrote:

...we must begin to love in order not to fall ill, and we are bound to fall ill if, in consequence of frustration, we are unable to love.

Two images are shared most often from The Worst Person in the World. One is the opening shot, of Julie in profile, wearing a black dress, standing alone on a balcony, a cityscape behind her. She holds a cigarette in one hand and her cell phone in the other, and she seems bored. After a few beats, she swipes open her cell phone and starts tapping away.

If the film were in black and white she could be one of Antonioni’s post-modern heroines, wandering vast cities alone, disinterested but free of burdens, smothered by a vague sense of alienation.

This is no coincidence. Julie is a spiritaul descendant of Antonioni’s figures of anomie from his Trilogy of DecadenceThe three films in this trilogy were L'Avventura, La Notte, and L'Eclisse, shot one a year from 1960-2. Quite a three year run. For my money, though, his greatest postmodern classic is The Passenger. I'm waiting for someone to direct an update called The Influencer, a masterpice capturing the fluid identity construction of the 2010's and 2020's., updated for a more precarious and distraction-filled age. At the briefest sense of boredom, today’s Westerner turns to her cell phone for relief.

The other image, the one on the movie poster, is of Julie running with a smile on her face. In that magical realist scene, she is running from one life to another through a world frozen around her. It’s a way of capturing that sense of breaking off from the world when in the early throes of love.

But her smile is also that of the joy of leaving a life behind. Julie is reveling in that sense of freedom, the power of being able to hit the shuffle button on life and skip to a new track. At that moment, mid-film, the opportunity costs of her freedom, and the specter of mortality, have yet to bubble up. They will.

The epilogue is a bit tidy and blunt. It’s the only chapter that feels forced. Julie is confronted with a coincidental and convenient Sliding Doors-style vision of what her life could have been if she settled down and had kids. By that point, Reinsve has long since let made it clear she’s conscious of the trade-offs in her life, if not at peace with them.

A lot of people I know found the movie’s title off-putting. It sounds like a Buzzfeed article. While there were more understated alternatives, it captures an important sense of societal judgment that envelops women who embrace their freedom to its fullest and choose more unconventional life paths. At least some of Julie’s regret arises from the general fog of impatience of those around her, from her boyfriend to her friends to her parents.

It’s not nearly as glum or didactic as it sounds. Its vibe is a sweet melancholy, and occasionally, like one particular meet-cute, it sparkles.

Give Reinsve a lot of credit for that. Director Joachim Trier said he wrote the movie with her in mind, and it shows. Her face is incapable of emotional dishonesty. She’s the friend you can’t help rooting for even as she stacks one uncertain life choice atop the next. She deserved a Best Actress nomination. Alas. Neon holding the movie back from wide release until this weekend didn’t help. At least she’ll always have her Cannes Best Actress win. She shouldn’t have to wait long for her next role.

I love all three of Trier’s Oslo Trilogy, the two previous entries being Reprise and Oslo, August 31. Very few directors of his age can channel the crippling weight of twenty-something identity crises with such empathy. But this entry is particularly apt for this moment. Film has, to date, tended to focus on the traditional notion of hedonic marriage in genres like the romantic comedy. We need more movies, like this one, that contemplate the actual lives many young people are choosing in post-industrial societies.

Drive My Car
Wheel of Fortune and Fantasy

The 2020 winner of the pandemic’s honorary “Shakespeare wrote King Lear during a plague” productivity award was Taylor Swift for Folklore and Evermore. 2021’s winner is Ryusuke Hamagachi for directing two critically-acclaimed movies.

I once thought that Haruki Murakami’s novels and short stories were ill-suited to film, but after seeing Drive My Car and Burning, I’ve done a one-eighty. More of his work should be adapted.

Drive My Car (coming to HBO Max in a few days on Mar 2) feels like a dialogue between the sometimes whimsical urban alienation of Murakami and the disillusionment of Chekhov’s Uncle Vanya. Like the Murakami short story on which it’s based, this movie is the duck gliding placidly across the surface of a pond while subtext churns furiously beneath the surface.

At 3 hours long, it will be too slow for many audiences. For those struggling with the now two years of pandemic life, however, it maps one cathartic path out of stasis and tragedy.

I’ve always loved Sonya’s speech from Uncle Vanya, but I never thought I’d see a new rendition as moving as the one in the film, performed via Korean Sign Language. For me, it was the most rapturous moment in cinema all year.

And when our final hour comes, we shall meet it humbly, and there beyond the grave, we shall say that we have known suffering and tears, that our life was bitter. And God will pity us.

Wheel of Fortune and Fantasy, a triptych of short films, is not an adaptation of Murakami short stories, but it feels as if it could be. The middle short of the three, “Door Wide Open,” is more interesting on cancel culture than the usual squabbles online. If you’re tired of the all-too predictable Joe Rogan-Spotify arguments, watch this as a palate cleanser.

The closing short “Once Again” concerns a coincidence at a high school reunion. I won’t ruin the plot, but it is wise to how much easier it is to help others with their problems than it is to solve our own. It’s a great argument for therapy.

Dune

If you’ve never read the book, I can understand why this Part One might feel slow. Having read the book multiple times, the first time as a high school freshman during my formative years as a science fiction reader, I carried the anticipation and context of the book’s back story to every scene. Whereas the novel has multiple long appendices and even a glossary, for me the entire novel was the appendix to the movie.

Leonardo Dicaprio pointing meme from Once Upon a Time in Hollywood

Fans of Dune the novel any time any iconic character or scene is referenced in the movie. If you hadn’t read the novel, you might have found the movie lacking in action. I don’t blame you, but I was the annoying Leo pointing meme throughout, and I apologize for nothing.

I’ve long thought Dune should be adapted as a miniseries instead of a film. There’s just so much ground to cover, especially in world building. Much of what bring me back to the book again and again is the journey its hero Paul traverses, to synthesize the divergent teachings of his Father and Mother and the two hemispheres of his brain, to achieve hyper consciousness and through it a form of transcendent mastery of his own mind and emotions. In decisive moments, when the stakes couldn’t be higher, Paul enters a flow state that connects him to the world. Though it’s referred to as a sci-fi novel, Dune’s beating heart is mystical, spiritual.

At the preview screening I attended, I had no idea the movie only covered the first 60% or so of the book. Near 3 hours in, with my bladder about to explode, I was never so relieved to see a To Be Continued appear on screen. News of an HBO prequel series built around the women of Dune is good news, though Dune as IP really drops off quickly in appeal after that. I never made it past the third book in the series as a kid, and it’s not clear it’s even worth adapting the second book.

What Denis Villeneuve channels best from the novel is a sense of pervasive political intrigue built up over centuries of jockeying between noble houses. When House Atreides is granted control of Arrakis, Duke Leto just assumes it’s a plot against him. This is how deep the rot goes. This is when you live life at the efficient frontier of the prisoner’s dilemma, defecting over and over, as the game theory predicts, because you know your opponent already has.

Of all the major narrative feature films I’ve seen, Dune features more IMAX footage than any I’ve seen. It’s a different film in IMAX in so many ways. Director of Photography Greg Fraser tried something new to me. He shot digitally, processed it, filmed it out to film stock, then scanned it back to digital to do the final color grade. It’s a sort of variant of scanning analog film grain and then overlaying it on digital images so they aren’t quite so clean. I only saw Dune in IMAX once, and to my eye the results were striking. To date, I continue to prefer the output from shooting digitally on location to shooting film against green screen.Some inematography buffs found the single light source setup of lots of Dune to be a flaw. Crafts people love to recognize higher degrees of difficulty like certain shots in West Side Story, for example. I was less bothered. The heavy shadows work in the traditional film noir way to visualize the political threats from every direction. And you're in the desert, where often there is just one light source, the sun, and it is relentless.

The French Dispatch

Richard Brody of The New Yorker named this the best movie of the year, which, as it’s a movie inspired by The New Yorker, feels like a mild conflict of interest. But damn if Wes Anderson didn’t make a movie that captures the feeling of The New Yorker’s house style, its meandering, understated rhetorical authority.

Anderson’s signature visual tropes, the perpendicular camera angles and symmetrical framing, the muted line readings, are both a signature of his individual style and a way of producing a sort of neutrality. The same could be said of The New Yorker’s plain house style.

The fidelity of this aesthetic homage was so pleasing to me as a longtime New Yorker reader that it functioned as a sort of ASMR. This is what a New Yorker article looks and sounds like.

My sister fell asleep watching the movie. The New Yorker doesn't use exclamation points. These things are correlated.

Wrath of Man

Jason Statham’s still, focused intensity is the oak tree that all the other twitchy, male violence wraps itself around in this slow-burn thriller. He never seems nonplussed; this is because he is badder than the other people around him and he knows it. This gives him the zen-like calm of a monk; his gleaming bald head is appropriate.

Also, for once, a Guy Ritchie film without some oddball speaking in an indecipherable accent. Instead, just a thrilling meditation on the corrosive nature of greed. Every scene constricts the suspense one notch tighter. If you have a subwoofer it will get a workout, like someone sounding the horns of hell.

The spatial geometry of the climactic set piece could be cleaner, but otherwise this is a tonal territory I’d love to see Ritchie revisit. Scott Eastwood was meant to play a dirtbag.

Bergman Island
The Souvenir: Part II

You don’t need to be an Ingmar Bergman fan to enjoy Bergman Island (though it’s recommended on its own terms). More useful might be knowing that director Mia Hansen-Love was once a partner to director Olivier Assayas, for whom she acted before she became a director, and that this movie is based loosely on and haunted by the dissolution of their relationship.

On the other hand, I would recommend you watch The Souvenir before watching The Souvenir Part II.

Both are films about filmmaking, but more than that, about how artists make sense of their lives through their work. It’s often said that creatives draw inspiration from their lives, but the creative process isn’t just a form of transcription. Often, the act of creation is how the artist makes sense of life.

ABBA has my favorite musical cue of the year in Bergman Island, and I can never get enough of Vicky Krieps and Mia Wasikowska. Bergman Island understands this paradox of love, that we can be haunted by the one who got away and why they never loved us while also being puzzled by how we ever loved the person we ended up with.

The Souvenir: Part II had me both laughing and convulsing in horror at its dead accurate depiction of the insufferable drama on film school sets. But while those scenes seemed lifted from my days on film school sets, they also reminded me of so many heated tech company meetings. A director struggling to articulate their artistic vision to her cast and crew is like a CEO or VP of Product who can’t articulate product vision to engineering and design.

In my favorite moment in the movie, the protagonist Julie runs into an older director named Patrick for whom she has been crewing on a studio project. He is pretentious, a tyrant, and because of that the studio has cut him out of post-production. Chastened, and in a self-reflective mood, he offers her some much needed perspective.

“Did you resist the urge to be obvious?” he asks about her just completed student thesis. All during her tumultuous shoot, her cast and crew pestered her to clarify what her movie was about. What Patrick recognizes, and what she has come to peace with, is how to preserve an individuality of expression in what is a collaborative creative process.

The Novice

Whiplash but if the J.K. Simmons and Miles Teller characters were one person. A tactile film about that particular type of obsession in which we hurtle ourselves against the limits of our bodies. But also, perhaps more than that, about how obsessive ambition is viewed as treachery in a zero-sum environment.

Any type-A high-achiever will recognize some of themselves in Alex. She’s a freshman who walks on to her college crew team and sets her sights on making the rare first-year leap to varsity. Through much of life, you can compete on all sorts of achievement ladders to surpass those around you, but true transcendence and grace comes when your ambitions are those you’d pursue when no one is watching. Except you.

Isabelle Fuhrman stars in The Novice

In The Novice, Isabelle Fuhrman confronts us with the question of what you call it if you Tiger Mom yourself

A Hero

Not my favorite Farhadi, but as with many of his movies, an X-ray one how financial and social capital interact within Iranian society and institutions. His movies have a Chekhovian soul.

The lesson here is as timeless as it is difficult for us to accept. One’s reputation is contextual, relative. It is defined, in large part, by others. Our character is absolute. Our integrity is all we can grasp in full.

TV Shows I Enjoyed This Year

Okay, I lied, this post isn’t just about 2021 movies I enjoyed. What qualifies as TV instead of film? It matters less than it once did. Here are a handful of episodic works I enjoyed.

Can’t Get You Out of My Head

(All six parts of this series are on YouTube. Not sure if they are there legally, but they haven’t been pulled, so ¯\_(ツ)_/¯. Here’s Part 1, for example.)

Adam Curtis dances in that shared territory between stark raving mad conspiracy theory and sweeping grand history narrative. One thing that separates him from other charismatic intellectuals seeking to connect the dots in history is his access to copious archival video footage and music and his willingness to wade through it. You think writing an essay is hard, try creating one in film.

A lot of this six episode history feels like a film yarn-and-headshot conspiracy wall, but damn if that signature Adam Curtis montage style isn’t a real vibe. At times, when Curtis’ signature voiceover drops out and all we see is grainy footage from various eras of history spliced together one after the other while “Song for Zula” by Phosphorescent plays in the background, what lingers in the memory is not airtight logic but the kind of associative implication that seems especially profound when in a pot-induced haze. Curtis’s coherence is an aesthetic one.

In our increasingly multimedia saturated discourse online, it’s not surprising to see memes come to dominate. But underrated is a style of argumentation built on vibes. TikTok is just the latest platform that enables this type of hyper-emotive rhetoric at scale. In every era, but especially this one, underestimate the emotional high ground at your peril.

I don’t doubt that if you asked Curtis to write an essay on these same ideas, with copious footnotes, his arguments would feel more convincing in some ways and diminished in others. The medium is the message, as they say.

Imagine what types of video essayists we’d unlock if we made it easier to access and use archival footage. When Musical.ly which then became TikTok licensed music tracks from the labels for its users to deploy in their videos, they subsidized millions of creatives with one of the most powerful elements of film, commercial music for the soundtrack. In the same way that the video store birthed a new form of cinephilia, unlocking or shared film and television corpus for easier sampling would unlock a new level of visual discourse.

Get Back

The year’s best series about the upside of in-person work. A movie like The Lighthouse hinted at the same but by depicting the negative; it gave us one hellish vision of the effects of the prolonged isolation of remote work.

Get Back is also a testament to the power of editing since much of the same event was assembled into a movie with a much different valence decades earlier. And to think, they plan to remove editing from this year's live Oscars broadcast.

I’ve long yearned for more slow cinema about craftsmanship. Instead of a puff piece of an hour-and-a-half documentary with dozens of talking heads praising some master of their craft, just show me 20 hours of unedited footage of them actually working. This documentary, to me, is some proof that this genre would act as am ambient boost to societal productivity.

I’m not a Beatles-phile by any stretch, so much of the narrative drama is lost to me. But even minus that context, the frissons and frictions of their creative process mesmerized me.Ian Leslie's "The Banality of Genius" is a great long read from someone much more well-versed in the Beatles history and mythology. When people ask writers “how did you write this?” it can feel as if you’re being asked to describe a color to someone who can’t see. But Get Back may be as close an answer to “how did you make this album” as anything we’ve seen yet.

Succession

In this age of streaming on demand, there is a nostalgic comfort in Sunday night prestige television that some critical mass of urban elites (I plead guilty) keep as appointment viewing. Succession was one of the only candidates in 2021. Thank goodness it was an operatic banger.

This season attracted some grumbling about the show’s circularity. Third seasons can be that way. But in many ways, this is the show’s theme, that the hell of the wealthy really just is an endless death match for Daddy’s love, or better yet, the keys to his kingdom. In this respect, the rich are like us; they too crave status.

What they don’t struggle with are material needs. From episode to episode this season they hop helicopters and private jets from one exotic locale to the next. When more and more TV is shot against green screen, and while I was stuck at home waiting out the pandemic, Succession's world-hopping felt like a treat. In many ways, the distinguishing feature of the elites of society is the amount of time they spend in limousines, helicopters, private jets, and yachts traveling from one meeting to the next. How bodies move through space will become an even more scarce status signal in this post-pandemic age. Already Zoom is beginning to feel like the low budget metaversal compromise for the masses.

As to that fantastic Jeremy Strong profile in The New Yorker, it’s the rare celebrity profile that enhanced my enjoyment of the show. That Strong is hardcore method on set, to the likely annoyance of his fellow cast-mates, is some bizarro parallel to the way Kendall drives the rest of the Roy clan insane. When I picture Strong hearing the news that Al Pacino has absconded with the chalice for the made-up award they used to entice Pacino to Yale, what I picture is Kendall Roy’s hangdog face.

The series also feels like a critique of postmodern irony. Logan is old-school, crass, but virile, direct, the canonical lion. He’s the decisive man of action who constantly cuts deliberation short. His children, in contrast, especially Roman, crack quips and snide remarks, reveling in each other’s hypocrisy and faults. But when push comes to shove, none of them seem to have any strong beliefs. In key business strategy sessions, they constantly waffle and hedge. A lifetime of Logan withholding his love has left them with a sort of PTSD. They’re the hectoring foxes, nipping at Logan’s lion until he swats them away.

Logan senses his children’s impotence and deploys it against them. Kendall becomes some social justice activist against Waystar RoyCo not because he believes in the cause but as a way of acting out. But both of them know the sword of Damocles hovering over Kendall’s head: it’s Daddy who bailed him out of his personal Chappaquiddick.

Shiv acts like a girlboss except when in the presence of her father, who alternately flatters and debases her. She hangs on to the emotional yo-yo for dear life. Her only means of avoiding spiraling in shame is to take her frustrations out on her husband Tom. In his spineless bureaucrat’s nature she is confronted with her own weakness and it disgusts her. By demeaning him she finds some relative high ground from which to avoid wading through her own humiliation.

Roman is the purest postmodern ironist. His soul seems corroded beyond repair. A lifetime of paternal abuse has left him unable to speak to his siblings except in the rhetoric of contempt. This also manifests in his odd sexual proclivities, especially in his Oedipal, S&M relationship to Gerri. She is the nurturing parent he never had as a child, but what he wants from her is a variant of what his father has always given him: humiliation. He could have a surrogate mother, but he wants a dominatrix. Roman is the living embodiment of the “men will literally X instead of going to therapy” meme.

If Kendall is oddly sympathetic, it’s because he’s the only one of the Roy clan who occasionally buckles under the weight of self-awareness. At times, he sees himself for who he really is, and it crushes him. Near season’s end, he was in such a spiral of despair that viewers spent a week debating whether he’d killed himself.

Everyone finds some emotional vindication in the series. By season’s end, it’s never been more evident that mommy and daddy don’t love their kids. It’s the Ok Boomer vibe on an operatic scale for this generation of kids who feel betrayed by their parents. The Roys are all wealthy, but technically the Roy children are also part of this first U.S. generation that is less well off than their parents. For the Boomers, the Roy children seem like the purest distillation of the entitled millennial archetype. For those of lesser means, it’s reassurance that the rich may have finer linens but burn in a hell of their own making.

It’s as acidic a show as I can remember, devoid of love. Few shows capture the feeling of Western culture at this moment better. It reminds me of Twitter.

And You Will Know Us by the Company We Keep

It feels as if we're at the tail end of the first era of social media in the West. Looking back at the companies that have survived, certain application architectural choices are ubiquitous. By now, we're all familiar with the infinite vertical scrolling feed of content units, the likes, the follows, the comments, the profile photos and usernames, all those signature design tropes of this Palaeozoic era of social.

But just as there are reasons why these design patterns won out, we shouldn't let survivor bias blind us to their inherent tradeoffs. The next wave of social startups should learn from the weaknesses of some of these choices of our current social incumbentsIt's easy to point out where our incumbent social networks went wrong. Of course, to be where they are today, they had to do a hell of a lot right, too. A lot of mistakes are understandable in hindsight given that online networks of this scale hadn't been built in history before. Still, it's easier to learn from where they went wrong if we're to head towards greener pastures.. It's never smart to tackle powerful incumbents head on anyway. The converged surface area in the design of all these apps suggest oblique vectors of attack.

While many of these flaws have already been pointed out and discussed in various places, one critical design mistake keeps rearing its head in many of the social media Testflights sent my way. I've mentioned it in various passing conversations online before. I refer to this as the problem of graph design:

When designing an app that shapes its user experience off of a social graph, how do you ensure the user ends up with the optimal graph to get the most value out of your product/service?


The fundamental attribution error has always been one of my cautionary mental models. The social media version of this is over-attributing how people behave on a social app to their innate nature and under-attributing it to the social context the app places them in. Perhaps the single most important contextual influence in social media is one's social graph. Who they follow and who follows them.

Just as some sharks that stop moving dieSome sharks rely on ram ventilation must swim in order to push water over their gills to breathe. But many shark species do not. Maybe we should refer to social apps that rely on a graph to work as "graph ventilated.", most Western social media apps must build a graph or die. This is because most of the most well-known Western social apps chose to interlace two things: the social graph and the content feed. That is, the most social media apps serve up an infinite vertical scrolling feed populated by content posted by the accounts the user follows. In my essay series on TikTok (in order, they are TikTok and the Sorting Hat, Seeing Like an Algorithm, and American Idle), I refer to this as approximating an interest graph using a social graph.

You can see this time-tested design, for example, in Facebook, Twitter, and Instagram. It is particularly suited to mobile phones, which dominate internet usage today, and which offer a vertical viewport when held in portrait orientation, as they most often are.

We'll return, in a second, to whether this choice makes sense. For now, just note that this architecture behooves these apps to prioritize scaling of the social graph. It's imperative to get users to follow people from the jump. Otherwise, by definition, their feeds will be empty.

This is the classic social media chicken-and-egg cold start problem. Every Silicon Valley PM has likely heard the stories about how Twitter and Facebook's critical keystone metrics were similar: get a user to follow some minimum number of accounts. Achieve that and those users turn into WAUs, or even better, DAUs. Users failing to follow enough accounts were the most likely to churn. Many legendary growth teams built their entire reputations inducing tens or hundreds of millions users to follow as many other users as possible.

But, again, this obligation derives entirely from the choice to build the feed directly off of the social graph. In TikTok and the Sorting Hat, I wrote:

But what if there was a way to build an interest graph for you without you having to follow anyone? What if you could skip the long and painstaking intermediate step of assembling a social graph and just jump directly to the interest graph? And what if that could be done really quickly and cheaply at scale, across millions of users? And what if the algorithm that pulled this off could also adjust to your evolving tastes in near real-time, without you having to actively tune it?


The problem with approximating an interest graph with a social graph is that social graphs have negative network effects that kick in at scale. Take a social network like Twitter: the one-way follow graph structure is well-suited to interest graph construction, but the problem is that you’re rarely interested in everything from any single person you follow. You may enjoy Gruber’s thoughts on Apple but not his Yankees tweets. Or my tweets on tech but not on film. And so on. You can try to use Twitter Lists, or mute or block certain people or topics, but it’s all a big hassle that few have the energy or will to tackle.


Almost all feeds end up vying with each other in the zero sum attention landscape, and as such, they all end up getting pulled into competing on the same axis of interest or entertainment. Head of Instagram Adam Mosseri recently announced a series of priorities for the app in the coming year, one of them being an increased focus on video. “People are looking to Instagram to be entertained, there’s stiff competition and there’s more to do,” Mosseri said. “We have to embrace that, and that means change.”

In my post Status as a Service, I noted that social networks tend to compete on three axes: social capital, entertainment, and utility. Focusing just on entertainment, the problem with building a content feed off of a person's social graph is that, to be blunt, we don't always find the people we know to be that entertaining. I love my friends and family. That doesn't mean I want to see them dancing the nae nae. Or vice versaEDITOR'S NOTE: It's not just people who know him. No one wants to see Eugene dance the nae nae.. Who we follow has a disproportionate effect on the relevance and quality of what we see on much of Western social media because the apps were designed that way.

At the same time, who follows us may be just as consequential. We tend to neglect that in our discussions of social experiences, perhaps because it's a decision over which users have even less control than who they choose to follow. Yet it shouldn't come as a surprise that what we are willing to post on social media depends a lot on who we believe might see it. Our followers are our implied audience.

To take the most famous example, the root of Facebook's churn issues began when their graph burgeoned to encompass everyone in one's life. As noted above, just because we are friends with someone doesn't mean we want to see everything they post about in our News Feed. In the other direction, having many more people from all spheres of our lives follow us created a massive context collapse. It wasn't just that everyone and their mother had joined Facebook, it was specifically that everyone's mother had joined Facebook.There's some generalizable form of Groucho's Marx quip about not refusing to join any club that would have him as a member. Namely, that most people don't want to belong to a club where they're the highest status member. Because, by definition, the median status of a member of the club is lowering their own. That's not to say it can't be a stable configuration. Networks based more around utility, like WeChat, aren't driven as much by status dynamics. Not surprisingly, they are less focused on a singular feed.

It's difficult, when you're starting out on a social network, to imagine that having more followers could be a bad thing. Yet many Twitter users complain after they surpass 20K, then 50K, then 100K followers or more. Suddenly, a lot of your hot takes attract equally hot pushback. Suddenly, it isn't so fun yeeting your ideas out into the ether. I know. Boo hoo on the smallest violin. But regardless of whether you think this is a first world problem, it's indicative of how phase shifts in the experience of social media are difficult to detect until long after they've occurred.

To put it even stronger, graph design problems are particularly dangerous to social companies because they fall into that class of mistakes that are difficult to reverse. Jeff Bezos wrote, in his 1997 Amazon letters to shareholders, about two types of decisions.

Some decisions are consequential and irreversible or nearly irreversible – one-way doors – and these decisions must be made methodically, carefully, slowly, with great deliberation and consultation. If you walk through and don’t like what you see on the other side, you can’t get back to where you were before. We can call these Type 1 decisions. But most decisions aren’t like that – they are changeable, reversible – they’re two-way doors. If you’ve made a suboptimal Type 2 decision, you don’t have to live with the consequences for that long. You can reopen the door and go back through. Type 2 decisions can and should be made quickly by high judgment individuals or small groups.


Graph design problems are one-way mistakes in large part because users make them so. Most social media users don't unfollow people after following them. Much of this comes down to social conformity. It's awkward and uncomfortable to do so, especially if you'll run into them. Anytime I unfollow someone I might run into, I imagine them cornering me like Larry David at the water cooler, eyebrows raised, with that signature tone of voice he mastered on Curb Your Enthusiasm, an equal mix of indignation at being slighted and glee at having caught you in an act of hypocrisy. "So, Eugene, I notice you unfollowed me. Pret-tay, pret-tay interesting."

If people tend to add to their social graphs more than they prune them, the social graph you help your users design should be treated as a one-way decision. And as Bezos noted, one-way decisions should be treated with care.Once Twitter started posting tweets to my timeline simply because people I followed had liked them, even if they were tweets from people I didn't follow myself, I started getting very confused. If you're angry I don't follow you, it may be that I think I already follow you.

Many social apps, because of how they're configured, undergo phase shifts as the graph scales. The user experience at the start, when you have few friends and followers, changes as those figures rise. At first, it's more lively with more people. Now the party's getting started. But beyond some scale, negative network effects creep in. And if you don't change how you handle it, before you know it, you find yourself pronouncing that you're taking a break from social media for your mental health.

Not only do users not notice it happening, like the proverbial slow boiling frog, the people operating the apps may be oblivious to the phase shifts until it's too late. Social graphs are path dependent.

A classic example, though I don't know if this still persists, is how Pinterest skewed heavily towards female users at launch, losing lots of potential male users in the process. This was a function of building their feed off of each user's social graph. Men would see a flood of pins from the females in their network as women were some of the strongest earlier adopters of pinning. This created a reflexive loop in which Pinterest was perceived as a female-centric social app, which chased off some male users, thus becoming self-fulfilling stereotype. An alternate content selection heuristic for the feed could have corrected for this skew.

But again, this is a problem unique to Western social media design. In conflating the social graph and the interest graph, we've introduced a content matching problem that needn't exist. I don't get upset that my friends don't follow me on TikTok or Reddit or what I think of as purer interest and/or entertainment networks. It's very clear in those products that each person should follow their own interests.

The way China has built out its social infrastructure is, in at least this respect, more logical. WeChat owns the dominant social graph, and it acts as an underlying social infrastructure to the rest of the Chinese internetThough not always a reliable one. If you're a WeChat competitor in any category, they may block links to your apps, as they've done with Douyin and Taobao in the past. This is always the danger of a private company owning the dominant social graph, and where regulators need to step in.. Rather than duplicate the social graph of everyone, which WeChat owns, other apps can focus on what they do best, which might or might not require an alternate graph.

Western social apps also rely much more heavily on advertising revenue. The lifeblood of their income statement is traffic to the feed. This means feed relevance is paramount. Anywhere one's social graph drifts from one's interests, boring content invades the feed. The signal to noise ratio shifts the wrong direction. Instead of pruning and tuning their social graphs to fix their feeds, most users do the next easiest thing: they churn.

As a product manager or designer on social app, you might object. The user chooses who to follow, and other users choose to follow them. It's out of your control. But this ignores all the ways in which apps put their hands on the scale to nudge each user towards a specific type of graph.

Take initial user sign-up flows. Every week, I seem to encounter this modal dialog on a new social app Testflight:

ios-contact-access-permission.png

What I look for is where this request appears and how the app frames it. Most times, users are asked to grant access to their contact book and to follow any matching users (or even worse, to spam their contact list with invites) before they have any idea of what the app is even about. In pushing people to duplicate their contact book, these apps are explicitly choosing to build off of people's real-world social graphs.

It's not surprising that social apps prioritize this permission as a critical one in the sign-up flow. The iOS contact book is now the only "open-source" social graph that a new app can work from to jumpstart their own. In The Network's the Thing, I argued that the network itself provides the lion's share of the value for a social network, that arguments about what types of content to allow in feeds, how those were formatted, were of much less importance. For a brief window, massive social graphs like Facebook or Twitter allowed third-party apps to tap into those graphs, even to duplicate them wholesale.

Instagram famously got a nice head start on building out their own social graph by siphoning off of Twitter's. It didn't take long for those companies to realize that they were arming their future competition. They clamped down on graph access hard. You can still offer Facebook or Twitter auth as an option for your app, but if you want a social graph of your own, the mobile contact book is the easiest to tap into nowadays.

Another way apps really influence the shape of their social graph is with suggested follow lists. These often appear in the first-time user walkthrough, interspersed in the feed, and sometimes alongside the feed.

Early Twitter users fortunate enough to be on the first versions of Twitter's suggested follow list today have hundreds of thousands or even millions of followers because they were paraded in front of every new user.

It was a massive social capital subsidy, but I find a lot of selections on that list puzzling. A few years ago, a friend set up a Twitter account for the first time and showed me the list of accounts Twitter suggested to them during sign up. It included Donald Trump. Which, regardless of your political leanings, is a dubious choice. Let's just shove every new user in the direction of politics Twitter (I'd be skeptical of a suggestion of Biden, too), one of the worst Twitters there is. Cool, cool.

For some people, like those who frequent fight clubs on weekends, politics Twitter might be the perfect dopamine fix, but when a user is signing up for the first time and Twitter knows nothing about them, that's a bizarre gamble to take.

For years, people marveled at Facebook's Suggested Friends widget. Wow, how did they know that I knew that person, yes, of course I'll friend them. And yet, as noted earlier, that may have been a graph design mistake given the way the News Feed was being constructed.

In the other direction, it's also important to help a users acquire the right types of followers. Cults are held together by a bi-directional influence. Cult leaders use their charisma to grow a following, then those followers shape the cult leader in return. It's a symbiotic feedback loop, not always a healthy one.

Besides being one-way mistakes, graph design errors are also pernicious because the tend to manifest only after an app has achieved some level of product-market fit. By that point, not only is it difficult to undo the social graph that has crystallized, to do so would violate the expectations of the users who've embraced the app as it is. It's a double bind, you're damned if you do and damned if you don't. Apps that achieve some level of product-market fit, even if it's a local maximum, require real courage to revert.

This doesn't stop social apps from trying to fix the problem. Reduced traffic to the feed is existential for many social apps. Instead of fixing the root problem of the graph design, however, most apps opt instead to patch the problem. The most popular method is to switch to an algorithmic, rather than chronological, feed. The algorithm is tasked with filtering the content from the accounts you've chosen to follow. It tries to restore signal over noise. To determine what to keep and what to toss, feed algorithms look at a variety of signals, but at a basic level they are all trying to guess what will engage you.

Still, this is a band-aid on an upstream error. Look at Facebook oscillating every few years between news content and more personal content from people you know. Until they acknowledge that the root problem lies in sourcing stories for News Feed from their monolithic social graph, they'll never truly solve their churn. And yet, to walk away from this fundamental architecture of their News Feed would be the boldest decision they've made in their long historyIronically, shifting to the News Feed itself was perhaps their previous boldest decision.. Not just because almost all their revenue comes from News Feed as it works now, but also since assembling a monolithic graph might be their strongest architectural defense against government antitrust action.

Twitter, unlike Facebook with its predominant two-way friending, is built on a graph assembled from one-way follows. In theory, this should reduce its exposure to graph design problems. However, it suffers from the same flaw that any interest graph has when built on a social graph. You may be interested in some of a person's interests but not their others. Twitter favors pure play Twitter accounts that focus on one niche. But most people don't opt to operate multiple Twitter accounts to cleanly separate the topics they like to tweet about.

One of my favorite heuristics for spotting flaws in a system is to look at those trying to break it. Advanced social media users have long tried to hack their away around graph design problems. Users who create finsta's or alt Twitter accounts are doing so, in part, to create alternative graphs more suited to particular purposes. One can imagine alternative social architectures that wouldn't require users to create multiple accounts to implement these tactics. But in this world where each social media account can only be associated with one identity, users are locked into a single graph per account.

One clever way an app might help solve the graph design problem is by removing the burden of unfollowing accounts that no longer interest users. Just as our social graphs change throughout our lives, so could our online social graphs. Our set of friends in kindergarten tend not to be the same friends we have in grade school, high school, college, and beyond.

A higher fidelity social product would automatically nip and tuck our social graphs over time as they observed our interaction patterns. Imagine Twitter or Instagram just silently unfollowing accounts you haven't engaged with in a while, accounts that have gone dormant, and so on. Twitter and Facebook offer methods like muting to reduce what we see from people without unfriending or unfollowing, but it's a lot of work, and frankly I feel like a coward using any of those.

Messaging apps, by virtue of focusing on direct communication between two people or among groups, naturally achieve this by pushing the threads with the latest messages to the top of their application windows. People who fall out of our lives just fall off the bottom of the screen. LIFO has always been a reasonably effective general purpose relevance heuristic.

Another possible solution to the graph design problem is to decouple a users content feed from their social graph. In my three pieces on TikTok, I wrote about how that app's architecture is fundamentally different from that of most Western social media. TikTok doesn't need you to follow any accounts to construct a relevant feed for you. Instead, it does two things.

First, it tries to understand what interests you by observing how you react to everything it shows you. It tries to learn your taste, and it does a damn good job of it. TikTok is an interest graph built as an interest graph.

Secondly, TikTok runs every candidate video through a two-stage screening process. First, it runs videos through one of the most terrifying, vicious quality filters known to man: a panel of a few hundred largely Gen Z users.Okay, yes, that's not quite right. Anyone can be on this test audience for a video. It just happens, however, that TikTok's user base skews younger, so most of the people on that panel will be Gen Z. Also, it's a known fact that a pack of Gen Z users muttering "OK Boomer" is the most terrifying pack hunter in the animal kingdom after hyenas and murder hornets. If those test viewers don't show any interest, the video is yeeted into the dustbin of TikTok, never to be seen again except if someone seeks it out directly on someone's profile.

Secondly, it then uses its algorithm to decide whether that video would interest each user based on their taste profile. Even if you don't follow the creator of a video, if TikTok's algorithm thinks you'll enjoy it, you'll see it in your For You Page.

Recently, Instagram announced it would start showing its users posts from accounts they don't follow. In many ways, this is as close to a concession as we'll see from Instagram to the superiority of TikTok's architecture for pure entertainment.

Some apps use some sort of topic or content picker. Tell us what music or film genres you like. What news topics interest you. Then they try to use machine learning and signals from their entire user base to serve you a relevant feed.

The effectiveness of this approach varies widely. Why does a playlist generated off a single song on Spotify work so well and yet its podcast recommendations feel generic? Why, after spending years and millions of dollars on research, including the fabled Netflix prize, do Netflix's recommendations still feel generic, and why doesn't it really matter? Why are book recommendations on Amazon solid while article recommendations on news sites feel random? It would take an entire separate piece just to dig into why some content recommendations work so much better than others, so complex is the topic.

In this piece focused on graph design, what matters is that things like content pickers explicitly veer away from the social graph. Twitter allowing you to follow topics in addition to accounts can be seen as one attempt to move a half step towards being a pure interest graph.

It's not that apps can't be more fun when social, or that people don't share some overlapping interests with people they know. We all care both our interests and the people in our lives. When they overlap, even better. It's just that after more than a decade of living with our current social apps, we have ample case studies illustrating the downsides of assuming they are perfectly correlated.

A secondary consideration is what type of interaction an application is building towards in the long run. Is is about one-to-one interactions or broadcasting to large audiences? What percentage of your users do you want creating as opposed to just consuming? Is your app best served by a graph of people who know each other in real life or by a graph that connects strangers who share common interests? Or some mix of both? Is your app for people from the same company or organization? Will the interactions cut across cultures and national borders, or is it best if various geographies are segregated into their own graphs?

The next generation of social product teams can and should be more proactive about thinking through what type of social graph will offer the best user experience in the long run.

I'm not certain, but it doesn't feel, based on the histories I've heard, that many social networks built their graphs with a particular design in mind. This makes graph design an exercise with more open questions than answers. In some ways, Facebook being built for just Harvard students in the beginning may have imposed some helpful graph design constraints by chance.

Unlike some types of design, graph design doesn't lend itself easily to prototyping. Social networks are at least in part complex adaptive systems, making it difficult to prototype what types of interactions will occur if and when the graph achieves scale.

But whereas traditional complex adaptive systems are so complex that predictions are futile, social networks are different in two ways. One is that human nature is consistent. The second is that we have numerous super scaled social networks to study. They're massive real world test cases for what happens when you make certain choices in graph design.

They also exist in multiple markets around the world. This makes it possible to study distinct path dependencies, especially when comparing across cultures and market conditions as unique as China versus the U.S. Despite all the variations in context, issues like trolling seem universal, suggesting that some potent underlying mechanisms are at work.

Once you tug on the threads surrounding graph design, you can burrow deep down many rabbit holes. If the people connected are going to be complete strangers, how will you establish sufficient trust (e.g. through a reputation system)? If the trunk of the app is a content feed, does that feed have to draw exclusively from stories posted by accounts followed by the user? Does it have to pluck candidates from those accounts at all? Is a feed even the right architecture for healthy interactions among your users?

Whose job is it to consider the problem of graph design? And when? To take one example, growth team strategies should be informed by your graph design. Growth shouldn't be treated as a rogue team whose only job is to extend the graph in every possible direction. They need to know what both good and harmful graph growth looks like so they can craft strategies more aligned with the long term vision.

Recently, TikTok started pushing me to connect more with people I know IRL. I've gotten prompts asking me to follow people I may know, and now when I share videos with people, I often get a notification telling me they've watched the video I've shared. Often these notifications are the only way I know they even have a TikTok account and what their username is.

To date, I've enjoyed TikTok without really following any people I know IRL. Perhaps TikTok is trying to make sharing of its videos endogenous to the app itself. But by this point in my piece, it should be obvious that I consider any changes to the graph of any social product to be moves that should be treated with greater caution. Most people I know don't make any TikToks (I know, I know, this is how you can tell I'm old), so following them won't impact my FYP much. For a younger cohort, where users make TikToks at a much higher rate, following each other may make more sense.

On the other hand, any app with a default public graph structure plays into the innate human impulse to judge. Wait, this person I know follows which accounts on TikTok?! Tsk tsk.

The answer to whether TikTok should push its users to replicate their real world social graphs isn't cut and dried. I bring it up only to illustrate that graph design is a discipline that requires deeper consideration. It could use, as its name implies, some design.


The term "follow" is fitting. Who we follow can become a self-fulfilling prophecy. First you build your graph, then your graph builds you. Plenty of research shows that humans tend to oscillate at the same frequency as the people they spend the most time with. Silicon Valley sage Naval Ravikant popularized the 5 Chimp Theory from zoology, which says you can deduce the mood and behavior of any single chimp by observing by which five chimps they hang out with the most.

The social media version of this is that we can predict how any user will behave on an app by the people they follow, the people who follow them, and the "space" they're forced to interact with those people in, be it a Facebook News Feed or Twitter Timeline or other architecture. We all know people who are the worst versions of themselves on social media. The fundamental attribution error predicts we'll think they're terrible by nature when they may just be responding to their environment and incentives.

Humans aren't chimps, we tend to juggle membership in dozens of different social groups at a time. Reed's Law predicts that the utility of networks scales exponentially because not only can each person in a network connect with every other node, but the number of possible subgroups is 2^N-N-1 where N is the number of people in that network.

But whether a social app allows such subgroups to form easily is a design problem. Monolithic feeds tend to force people into larger subgroups than is optimal for healthy interaction. While every user sees a different Twitter Timeline or Facebook News Feed, the illusion is still of a large public commons. Because anyone might see something you post, you should operate as if everyone will.

Messaging apps, in contrast, tend to allow users themselves to form the subgroups most relevant to them. Facebook Groups is a more flexible architecture than News Feed. Humans contain multitudes, and social apps should flex to their various communication privacy needs.

It's no surprise that many tech companies install Slack and then suddenly find themselves, shortly thereafter, dealing with employee uprisings. When you rewire the communications topology of any group, you alter the dynamic among the members. Slack's public channels act as public squares within companies, exposing more employees to each other's thoughts. This can lead to an employee finding others who share what they thought were minority opinions, like reservations about specific company policies. We're only now seeing how many companies operated in relative peace in the past in large part because of the privacy inherent in e-mail as a communications technology.

In many ways, graph design was always bound to be more important in Western social media now, in the year 2021, than in the early days of social media. In the early days of the internet, the public social graph was sparse to non-existent. For the most part, our graphs were limited to the email addresses we knew and the occasional username of someone in our favorite news groups. It's hard to explain to a generation that grew up with the internet what a secret thrill every new connection online was in the early days of the internet. How hard it was to track down someone online if all you knew was their name.

Today, we have more than enough ways to connect to just about anybody in the world. Adding someone to my address book feels almost unnecessary when I can likely reach any person with a smartphone and internet access any of a dozen ways.

In a world where finding someone online is a commodityOne sign that it is a commodity is that messaging apps, while massive, are for the most part lousy businesses that generate little in revenue. That's the financial profile you'd expect of a commodity business., the niftier trick is connecting to the right people in the right context. I have over a dozen messaging apps installed on my phone, they all look roughly the same. While I've discussed graph design largely defensively here—how to avoid mistakes in graph design—the positive view is to use graph design offensively. How do you craft a unique graph whose very structure encodes valuable, and more importantly, unique intelligence?

LinkedIn may be the social app Silicon Valley product people like to grouse about the most, but while many of the complaints are valid, its sizable market cap is testament to the value of its graph. It turns out if you map out the professional graph, not just today but also across long temporal and organizational dimensions, recruiters will pay a lot of money to traverse it.

For all the debate over whether our current social networks are good for society, I prefer to focus on the potential we've yet to realize. We have the miracle of Wikipedia, yes, but aren't there more types of mass scale collaboration to be enabled?

Every other week or so, I am introduced to someone amazing, or an account I've never heard of before that blows me away. That social networks themselves aren't facilitating these introductions leaves me less sad than hopeful. In a decade, today's social graphs will look like blunt instruments, so primitive were their configurations.

We'll also look back over that decade, see how many more amazing people we finally met at the right time and the right context, and realize that indeed, the real treasure was the friends we made along the way.

American Idle

I promised one final piece on TikTok, focused primarily on the network effects of creativity. And this is that, in part. But it discusses a bunch of other topics, some only tangentially related to TikTok.

All the points I wanted to cover seem hyperlinked in a sprawling loose tangle. This could easily have been several standalone posts. I've been stuck on how to structure it.

Some people find my posts too long. I’m sympathetic to the modern plague of shortened attention spans, but I also don’t want lazy readers. At the same time, this piece felt like it was missing a through line that would help pull a reader through.

And then I had a minor epiphany, or perhaps it was a moment of delusion. Either way, it provided an organizing conceit: I decided to write this piece in the style of the TikTok FYP feed. That is, a series of short bits, laid out vertically in a long scrolling feed.

This piece is long, but if you get bored in any one section, you can just scroll on the next one; they're separated by horizontal rules for easy visual scanning. You can also read them out of order. There are lots of cross-references, though, so if you skip some of the segments, others may not make complete sense. However, it’s ultimately not a big deal.

If I had more time, I might have built this essay as a series of full-screen cards that you could swipe from one to the next. Or perhaps tap from one to the next, like Robin Sloan’s tap essay (I wish there a way to export this piece into a form like that, if someone built that already let me know). And if I were even more ambitious, I would've used some Anki-like spaced repetition algorithm to randomize the order in which the following text chunks are presented to you, shuffling it each time a reader jumped in.The most meta way for me to ship this essay would have been as a series of TikTok videos. It would have been the Snowfall of TikTok essays. That would have also taken a year of my life (which, being locked inside because of a pandemic might be the time to attempt something like that?). Also, I am camera shy.

But as it is, this is what you get.


By network effects of creativity, I mean that every additional user on TikTok makes every other user more creative.

This exists in a weak form on every social network and on the internet at large. The connected age means we are exposed to so much from so many more people than at any point in human history. That can't help but compound creativity.

Various memes and trends pass around on networks like Instagram and Twitter. But there, you still have to create your own version of a meme from scratch, even if, on Twitter, it's as simple as copying and pasting.

But TikTok has a strong form of this type of network effect. They explicitly lower the barrier to the literal remixing of everyone else's content. In their app, they have a wealth of features that make it dead simple to grab any element from another TikTok and incorporate it into a new TikTok.


The barrier to entry in editing video is really high as anyone who has used a non-linear editor like Premiere or compositing software like After Effects can attest. TikTok abstracted a lot of formerly complex video editing processes into effects and filters that even an amateur can use.

Instagram launched one-click photo filters (after Hipstamatic, of course, though Hipstamatic lacked the feed which is like the spine of modern social apps), and later Instagram added additional features for editing Stories, and even some separate apps like Boomerang that were later re-incorporated back into Instagram as features.

Snapchat has a gazillion video filters, too, though many are what I think of as simple facial cosmetic FX.

YouTube has launched almost no creator tools of note ever. WTF.

TikTok launches seemingly a new video effect or filter every week. I regularly log in and see creators using some filter I've never heard of, and some of them are just flat out bonkers. What creators can accomplish with some of these filters I can't even fathom how I'd replicate in something like the Adobe Creative Suite.

Kili So Silly (@kili.so.silly) has created a short video on TikTok with music original sound. | #stitch with @xxelacxx #TimeWarpScan #fyp #foryou

JeremyLynch (@jeremylynch) has created a short video on TikTok with music Despicable Me (From "Despicable Me"). | This freaks me out watching it back 😅 #timewarp #timewarpchallenge

TikTok’s Warp Scan filter is a bizarre concept for a filter in and of itself, but the myriad of ways TikTok users put it to use just shows what happens when you throw random tools to the masses and allow for emergent creativity. It only takes a handful of innovators to unleash a meme tsunami.


A longstanding economics debate is why we haven't seen the effects of the internet in our productivity figures. I won't rehash every side of every argument there.

But I know this: to take someone else's video and insert a reaction video of my own playing alongside it on the same screen is not easy in a traditional NLE. I'm not saying it's the moon landing, but it's not trivial.

On TikTok, you can just press the Duet button and start talking into your phone, and soon you have a side-by-side of the original video and your reaction video (you can choose from any number of their preset layouts for reactions). That's an explicit productivity boost; I can measure it in time saved for the same output.

You won't see that show up in GDP per capita figures, but it's real.


Remember 2 Girls, 1 Cup? If you've seen it, how could you not?

What interested me was less the video, which just horrified me, but the reaction videos of people watching it. Because 2 Girls, 1 Cup was a short video, I think it was a minute or two long, you could simply watch the face of someone watching the video and sync every reaction to every horrific beat of the video now forever haunting your memory, even though the original video wasn't visible on screen. The fun of the 2 Girls 1 Cup reaction video, but reaction videos in general, is that shared context.

Until TikTok came along, there wasn't an easy way to do reaction videos to other videos and have them make sense unless the original video had so much distribution that it was common knowledge. Or you could put the reaction video alongside or on top of or beneath the original video, but that required skill in using a non-linear video editor to lay those out and synchronize their timelines.

With TikTok's Duet feature, you can instantly record a side-by-side reaction video to anyone else's video. Duet is the quote tweet of TikTok. Or you don't have to do a reaction video at all. The Duet feature is designed simply to allow you to record a video that will play back alongside another video. It can be used for reaction videos, sure, but also to just provide a running commentary on other videos, and there are entire accounts built around both concepts.

But again, the Duet feature is built at such a low level that you can treat the feature as a primitive to replicate any number of other editing tools.


One such tool is to use the Duet feature as a dynamic matte. Since you know where your video will be placed in relation to the original poster's video, you can build a video mosaic.


Another is to use the Duet feature to, well, literally record duets.

But if you allow Duets to stack, well, eventually, one Wellerman can bring the whole chorus to your yard.

Someone truly ambitious could adjust the playback speed of various levels of Inception from the film Inception and stack them and synchronize them in TikTok using the Duet feature. If I had more time I'd do this myself, but the time has come for some time-rich kid out there to take this on.


Knowing that others can Duet your video means you can post any number of videos as prompts.

For example, you can read one side of the dialogue in a two-hander.

sara (@saraecheagaray) has created a short video on TikTok with music 人生のメリーゴーランド (Jazzical Lounge ver.) [『ハウルの動く城』より]. | #duet with @thechrisbarnett NOW I can say my accent sucks #fyp #foryou #acting


Knowing that TikTok has a Stitch feature, you can also post a question in a video and expect that some number of people will use Stitch an answer to your question and distribute that as a new video.

A popular prompt is "Tell me you're X without telling me you're X" or any number of its variants like "Show me you're X without showing me you're X."

Stitch wasn't necessarily designed to be used in this way, but as a primitive it's well-suited to any number of uses, including making TikTok a sort of video Quora.


Video prompts can come from not only other TikTok videos but commenters.

Some TikTok videos are made in response to requests posted in the comments. The comment is excerpted and published as an on screen text overlay at the beginning of the response video.

This is another of the nested feedback loops within the global feedback loop that is the FYP talent show. Once one example of this went viral, then the entire community adopted this as one of the norms of the community.


The Daily Show with Jon Stewart was a show almost entirely built around the reaction video. Stewart would play some clip of a politician being a hypocrite, or some Fox News anchor spouting their usual performative indignation, and then the camera would cut back to Stewart, his face frozen in some emoji mask of shock: eyes wide, mouth agape.

Social networks, and entertainment networks like TikTok, have completed the work of democratizing reactions. Yes, there's no reason you need to react to everything. But it's human nature. This is the social contract of the social media era. If you dare to shout your opinion or publish your work to the masses, the masses can choose to shout back.

Gossip litigates and fleshes out the boundaries of acceptable behavior within groups. Whereas gossip used to be contained, social networks now give it global distribution. This is one reason of many we've seen in-group and out-group boundaries drawn in bolder weight in this era. For every wide-eyed look of horror by Jon Stewart, you had the furrowed brow of disbelief that is Tucker Carlson's signature look, like someone in his elevator car passed gas.

Now extend that to clapbacks on the internet and you have a world in which back-channel gossip, a useful release valve and distribution channel for information about our peers, has become an open dialogue. The grapevine became the public feed, and every day, kangaroo court is in session.


TikTok's Duet feature belongs in the social media hall of fame of primitives alongside features like Follow and the Like button.

What feature better epitomizes the remix, react culture of the internet? Paul Ford once wrote that "Why Wasn't I Consulted?" is the fundamental question of the web. By then, social networks were well on their way to taking over from the web, and in the process, installing the plumbing by which the masses could finally directly opine to the masses, who could, in return, directly consult back. The "reply guy" is the consultant class of the internet, and mansplaining is its verb.

Yes, there are quote tweets and replies, but the TikTok Duet is the video analog, so simple and elegant in its design that you wonder why YouTube didn't launch it ten years ago, and then you remember that YouTube hasn't launched any creator tools of note since...ever.


What the Duet feature does, as described by how it would be done in a traditional non-linear editing program like Adobe Premiere, is the following:

  1. Copies the original file
  2. Inserts a new video track and a new audio track on top of the originals
  3. Allows you to lay down a new video on those new tracks
  4. Performs a whole series of steps to arrange the videos side-by-side on screen

TikTok abstracts a bunch of steps into a single function.


Yes, yes, some of these features in TikTok came from Musical.ly. But that's just a meta form of the theme of this piece! TikTok sampled from Musical.ly and improved upon it. They remixed a remix app.

But also, isn't this how innovation happens? We stand on the shoulder of giants and all that? Good artists copy, great artists steal?

TikTok enables, for video and audio, the type of combinatorial evolution that Brian Arthur describes as the underlying mechanism of the tech industry's innovation.


How many truly original ideas are there in Silicon Valley? Very few. Most have been tried umpteenth times in the past. Much of finding product-market fit in tech is context and timing. And people always underestimate the market side of product-market fit. When something fails, people tend to blame the product, but we should blame the market more often. The pull of the market is usually as important, if not more so, than the push from a product.

One day, the conditions are finally right, and an idea that has failed ten times before suddenly breaks out. Sometimes it's a tweak in execution, maybe it's an advance in complementary or enabling technology, sometimes it's a cultural shift.

Most of the best ideas in tech first appeared in science fiction books in the 1960s, and many of those are still waiting for their time to come. This is why rejecting companies that are trying something that's been tried before is so dangerous. It's lazy pattern-matching.

I do like Jeff Bezos' principle on when he decides to finally give up on an idea: "When the last smart person in the room gives up on the idea." But it also implies that you should bring some ideas back when a new smart person, or maybe a naive overconfident one, enters the room and champions the idea.


Given we know innovation compounds as more ideas from more people collide, it's stunning how many tech firms, even ones that ostensibly tout the value of openness, have launched services that do a better job of letting their users exchange ideas than any internal tool does for their own employees’ ideas.

How many employees join a firm and then spend a week in orientation learning where to get lunch, how to file expense reports, mundane trivialities like that. How many sessions are led by random trainers who don't even work at the company?

If you think of a company as an organism, and new employees as new brain cells, it's staggering how many join the company and begin from an absolute cold start. It's as if the company has chronic amnesia. What has the company learned from its past, what is its culture? When employees take months or even years to get up to speed at a company, companies should be embarrassed. Instead, it's treated as normal.

The free flow of ideas outside a company shouldn't, or in apps like TikTok, shouldn't exceed the rate at which knowledge flows inside a company, but I see it happen time and again.


The toughest job for any creative is the cold start. The blinking cursor on the blank page in a new document. Granted, writing a tweet, or even shooting an Instagram photo, isn't like composing the great American novel. But we tend to underrate the extent to which new users often churn without having ever posted anything to a social network because we only focus on those who do.

Now imagine trying to make a TikTok from scratch if you're older than, say, 19. The creative bar is high, you don't know how to dance, you're not up on the latest memes or popular music. Even if you're a teen, it's not easy to come up with a 0 to 1 TikTok.

But the beauty of TikTok's FYP algorithm and the Discover page is that you don't have to create a TikTok from scratch. The vast majority of TikToks are riffs on memes and trends that other users originate. It's no shame to be a 1 to n TikToker. Many on the platform achieve their first viral hit riffing on an existing meme.

Charli didn't invent the Renegade dance, Jalaiah Harmon did, but Charli made it famous. A lot of Charli and Addison's most popular TikToks are their interpretation of dances other people choreographed to songs other people composed.The ongoing debate on cultural appropriation seems to have no end in sight, but at least on TikTok there is a chance, with time stamps and some of the literal links the app creates between videos, to trace the origin of memes more easily.


Richard Dawkins introduced the term meme in his classic The Selfish Gene, defining it as a unit of information that spreads via imitation. He noted that memes evolve via natural selection just as in evolution. This memetic evolution happens via the same mechanisms as biological evolution, via variation, mutation, competition, and inheritance.

The internet writ large has always been fertile ground for the accelerated breeding of memes (cue toothless old prospector: "Back in my day sonny boy we had to spread memes via email chain letters"). But the TikTok app is perhaps the most evolved meme ecosystem to date.


Assisted evolution occurs when humans intervene to accelerate the pace of natural evolution.

TikTok is a form of assisted evolution in which humans and machine learning algorithms accelerate memetic evolution. The FYP algorithm is TikTok's version of selection pressure, but it's aided by the feedback of test audiences for new TikToks.

Memes can start from almost anything on TikTok. It can be the lyrics of a song, or just the vibe of a track, or both. A user can post a question or a challenge. In a single session on TikTok, you'll find videos of all types, most being riffs on existing memes (the variation).

Regardless of the provenance, any video, once loaded into TikTok, is subject to the assisted evolutionary forces in the app. Software tools like the Duet or Stitch feature and all of TikTok's other video editing tools assist in mutation and inheritance, and each remix of a source video becomes a source video for others to remix, generating further variation. Meanwhile, the competition on the FYP feed is fierce, and the survivors of that extreme selection pressure are memes of uncommon fitness.


In this assisted evolutionary ecosystem that is TikTok, and with an...umm...assist from the pandemic that kept hundreds of millions of people locked inside scrolling their phones, we've seen a marked contraction in the half-life of memes.

Memes used to dominate TikTok for what felt like weeks, and now it seems the memetic zeitgeist on TikTok shifts every few days, if not nightly. If I don't check back on TikTok every day, I find myself scrambling to catch up to the meta when I finally do open the app.


Of course, people grab TikToks and share them on YouTube or Twitter or as Reels on Instagram, but those apps receive flattened video files and can’t break them into component parts to be remixed the way you can on TikTok. Those other services are fine endpoints for distribution, but the creativity happens on TikTok. Don't get me started on apps like Triller (which feels like a Ponzi scheme).

People will litigate Instagram copying Snapchat's Stories feature until the end of time, but the fact is that format wasn't ever going to be some defensible moat. Ephemerality is a clever new dimension on which to vary social media, but it's easily copiable.

This is why TikTok's network effects of creativity matter. To clone TikTok, you can't just copy any single feature. It's all of that, and not just the features, but how users deploy them and how the resultant videos interact with each other on the FYP feed. It's replicating all the feedback loops that are built into TikTok's ecosystem, all of which are interconnected. Maybe you can copy some of the atoms, but the magic lives at the molecular level.

TikTok has a a series of flywheels that interconnect, and there isn't any single feature you can copy to recreate the ecosystem. Meanwhile, Reels has to try to compete while being one of like a half dozen things jammed into the Instagram app.


Markets in the internet and technology age are conducive to winner-take-all effects thanks to preferential attachment. This means that if you are first to stumble upon some flywheelMany like my friend Kevin use the term loops. I use flywheel merely to indicate I'm referring to positive feedback loops since loops can also be negative feedback. Also, I had to make that damn Amazon flywheel diagram for way too many presentations back in the day, it's mounted on a wall in my brain. in your business, the returns are even greater and accumulate more quickly than they would've in any other era in history.

Building a flywheel, though, often requires connecting a series of features at once. When I advise various companies, big and small, I often run into objections to my recommendations because of the popularity of agile or other incremental development philosophies. We end up at loggerheads on the V of MVP (minimum viable product), V having always been contextually determined.

If a flywheel requires three or four or even more things to connect in your app, it takes more work to ship all of them at once, and that feels like a riskier expenditure of your team's time. But, I'd counter: 1) often, testing a flywheel by definition means you have to build multiple features that work together 2) the returns of achieving a flywheel are often so high as to be worth the risk and 3) if you don't achieve any flywheels you are, as investor updates are so fond of saying, default dead.


Instagram famously has never had its version of resharing (e.g. retweeting). This reduced the velocity of photos and later videos on the service, a sort of brake on spam and misinformation and other possible such downsides.

But after using TikTok, it does feel odd to go through Instagram and not be able to grab anyone's photo to remix. Imagine you could grab someone's photo and apply your own filters, or grab just one element of the photo and use it in your photo.

Once we all live in the metaverse, this type of infinite replication and remixability will be something we take for granted, but even now, we're starting to see an early version of it on TikTok. This type of native remixability feels like it will be table stakes in future creative networks.


Fanfic is one text version of sampling and remixing. It doesn't require much more than your imagination.

It's always been really expensive, in both time and legal costs, to sample and remix film and television. TikTok has, with its short video format and tools, made remixing of premium video easier and safer. In Harry Potter TikTok, and its sub-genus Draco Malfoy TikTok, creators pull from the repository of the Harry Potter film universe as if it were on GitHub and merge themselves into branching storylines in which, well, creators become students at Hogwarts and catch the romantic interests of one Draco Malfoy.


The Discover page acts as the Fed in the central economy of memes on TikTok, while the FYP algorithm is the interest rate on meme distribution.

The Discover Page features hashtags. By the very act of featuring a hashtag, they signal to creators that if they create using that hashtag, they will get the distribution boost of that hashtag being featured on the Discover Page. Which raises the age-old conundrum, which came first, the Discover Page hashtag placement, or the hashtag's trending? The answer is yes. It's circular, an ouroboros of virality.

TikTok also posts the number of collective views on videos with that hashtag, helping creators gauge the potential distribution value of climbing aboard that trend.

TikTok is a mix of a centrally planned economy and a free market, much like many multiplayer video games where the game publisher manages the price and availability of various assets like weapons and armor while the players put them to use in the virtual economy.

The Discover Page is also where TikTok will feature corporate challenges. Yes, it's a paid placement, but the creative output is collective and distributed.


Because the most popular memes get super-distribution via the FYP algorithm, you can assume common knowledge of the meme among your viewers and just cut to the punchline. You don't need a bunch of what would be the video equivalent of exposition upfront. This keeps the majority of videos on TikTok compact, critical to the high cadence of the FYP feed. TikTok feels fast. Almost manic.

It also gives viewers that hit of in-group dopamine when they already know the references in your video.


If you don't understand a TikTok video and its references, you can trace the provenance from within the app in any number of ways. You can follow the hashtags in the caption or tap the sound icon and see all the other videos which have been made in that meme branch. Often that's enough to derive the context.

Or you can just read the comments. You'll find you usually aren't alone, someone will almost always have posted a comment like "In here before the smart people arrive" and then below that will be comments that explain the video to everyone else.


The internet, and the assumption of the internet, allowed for more complex and long linear narratives in television, shows like Lost and Game of Thrones. The assumption of Know Your Meme, or just knowledgeable commenters in the TikTok comments, allows for less expository and more compact, obscure TikToks. TikTok comments are a form of distributed annotation.


This technique of offloading the setup for a joke to the internet allows TikTok's, or even Tweets or Instagram posts to take on a form of what I call compressed narrative.

The old format of a joke, with a setup—A man walks into the doctor's office wearing only underwear made of Saran wrap—and then the punchline—and the doctor said "I can clearly see your nuts."—is dead. The internet killed the "joke."

Instead, the internet is mostly punchline, with the barest of setup, if any. It's on you to know the context. Go Google it.

And if you still don't get it, you weren't meant to.


An example is the "I ain't ever seen two pretty best friends" meme that went around on TikTok for a hot minute and has since just become a base trope of the TikTok creative universe. Videos started taking more and more circuitous routes to end with that punchline, throwing all sorts of sleight of hands before dropping it, out of nowhere, like an M. Night Shyamalan film twist.

If you hadn't heard of the meme or didn't know the reference, these videos would be complete mysteries. Even now, if you don't know what I'm talking about, this section will make no sense. Nor will comments like "We found the two pretty best friends" on various videos.


One of the better pieces I read last year was this on the death of political humor in the age of Trump. My favorite turn of phrase from the piece is that "Irony in politics, meanwhile, has reversed its polarity." David Foster Wallace predicted the death of irony, of cynicism, after an initial boom when the internet was coming of age. The lament of the humorist is that figures like Trump are beyond the reach of irony because they are already satires of themselves.I often lament when I refer to as fortune cookie Twitter, and to combat this, I think Twitter should set up a GPT-3 bot that constantly trains on each account, and the moment most of your followers can no longer distinguish between the GPT-3 spoof of your account and your actual account, you should be forced to vacate your account and allow the GPT-3 bot to replace you. You will have literally become a parody of yourself. Also, if for some reason I ever hacked my way into a famous person's account, my goal would not to be to request BTC or post something offensive. Instead, my goal would be to post a tweet that so resembles their voice that no one, not even the person who owned that account, could tell. They'd just think, wow, that's strange, I don't remember posting that, but it is something I'd post, so ¯_(ツ)_/¯

To me, humor has always depended on creating a gap and then helping your audience to hurdle it. In a traditional joke, the gap is the space between the setup and the punchline. When the audience's mind comprehends the joke, they soar across that gap, and the exhilaration is released as laughter. You don't want to carry them across, you want to do just enough to let them take that last leap themselves.

A comedian like Chris Rock will take something from real life and just point out the hidden social truth beneath it, and your mind gets that dopamine hit of acknowledging a social fiction that you'd otherwise observe without question. Like Moses, comedians part the sea of taboo and let you stroll through, laughing all the way at being able to get away with it.

Pre-cancellation Louis C.K. also lived in this space, exposing something of your nature that you were embarrassed to acknowledge. Either he'd absolve you of your shame by absorbing it all himself in a performance of self-loathing, or he'd just forgive you that fault by making it seem universal. Comedians let you look at yourself from outside yourself, creating a gap between you and your own nature.

Trump killed humor by closing that gap entirely, becoming such a parody of himself that shows like Veep seemed less dark satire than some form of fatuous cosplay even though they came first.

But humor is not so easily killed. You just need new ways to restore the comedic gap. Much of TikTok humor is oblique in form, making references that flatter you if you understand them and puzzle you if you don't. But for the latter, you then must set off on a journey to traverse that gap. And when you've completed that journey, you get the delayed satisfaction of getting the joke but also the pleasure of now being in the in-group.

But more sophisticated creators can also play with that expectation, setting off on what seems like a familiar meme, then subverting audience expectations.


Douyin, the Chinese version of TikTok, from the same parent company Bytedance, provides an interesting contrast in the styles of humor between China and America.

A lot of comedic videos in China use a laugh track sound effect. I can't remember the last time I heard a laugh track in a TikTok.

I want to draw some conclusion here, but I don't feel confident enough. Someone more familiar with the cultural differences in Chinese and American humor might clarify this for me.


Netflix brings international programs to the U.S. TikTok brings some Chinese programming to the States also.

TikToker @funcolle makes a sort of hyper-compressed episodic detective series that is filmed in China and spoken in Mandarin, but it works on U.S. TikTok thanks to onscreen subtitles. The sound she uses is, by now, as memorable to me as the theme song to any number of popular TV series like Game of Thrones.

If you can't solve these really short single TikTok video mysteries, you can turn to the comments section to get help from all the other viewers who've pored over the videos in detail and raced to post the solution.

One measure of a platform's power is the number of things people make with it that you had never been made before. Every week, I find videos on TikTok that I can't imagine having been made on any other app.

Funcolle (@funcolle) has created a short video on TikTok with music original sound. | Anything wrong with this room? Come on, my detectives!#foryou


On TikTok, the comments have become creative terrain in their own right. Somewhere along the line, riffing on someone else's TikTok no longer required you to make a TikTok. Instead, you can just go into the comments and tack on a punchline to the punchline of the video and rack up hundreds of thousands of likes. Writing the most clever comment on a TikTok video has become its own art form.

I can't remember the last time I watched a good TikTok video without then opening up the comments to see what the peanut gallery came up with. Sometimes I read the comments before even finishing the video. TikTok's method of ranking comments almost always surfaces the best and most relevant comments to the top. However you feel about a video, it's uncanny how often one of the top five comments encapsulates it perfectly.

It's difficult in a video to feel the presence of other viewers in a tangible, meaningful way. The Twitch comment bar gives you a visible if somewhat bewildering waterfall of text as evidence of their presence, and the hearts on something like an Instagram Live or the bullet comments on Bilibili videos do the same.

TikTok comments, though, feel ike they extend the canvas of the video. Just as talent shows like The Voice require both contestants and voices to work, more and more it feels as if the TikTok experience is about watching the performers and then listening to the judges (all of us viewers) render their opinions via the comments. There isn't one Simon Cowell on TikTok, but in any comments section of any TikTok video, someone will play that role.

Never read the comments. Unless you're on TikTok, in which case, always read the comments.


Reading the comments on TikTok serves a communal function. It's like hearing the laughter of the crowd at a comedy show.

One of the existential challenges of life is truly connecting with other people's thoughts. Who can ever know that series of emotions and thoughts and dreams we call our consciousness? True human connection seems always out of grasp.

The pandemic exacerbates that sense of isolation. When most of our interactions are with flat faces on video screens, it feels either like we're living in a simulation or some solipsistic nightmare.

Before I check the comments on a TikTok I've just watched, I almost always have a strong reaction to that video. That's why opening the comments and finding that one of the first few comments perfectly encapsulates your reaction, then seeing it already has tens or hundreds of thousands of likes, is so comforting. This confirmation of a shared response creates, asynchronously, a passing score on a form of the Voight-Kampff test. It's a checksum on your humanity.

Many comments have begun using the inclusive second person singular, literally speaking for the rest of the viewers. These comments often begin with "POV:" as in "POV: You're lying bed at 2am scrolling TikTok." It's presumptive, and yet the best TikToks evoke such a consistent multiple-choice checklist of responses that it's rare the times I can think of an original comment that isn't already posted above the fold.


The sense of collective response in TikTok comments and the publicly visible view and like counts have been around long enough that users now assume enough others have encountered enough of the same memes despite everyone's FYP algorithm being tailored to their individual tastes. Many a comment on a viral TikTok will read like "Oh we're back here again."

Though I have said that TikTok isn't a social network—I don't know most people on the app, I don't have to follow anyone to have a good experience—the algorithm does create, through its efficient sorting, a sense of traveling through subcultural neighborhoods as you scroll down one TikTok at a time.

Users have adopted spatial or geographic language to describe this sense of shared viewing spaces. Various subcultures are described by appending -tok or TikTok behind a descriptor. Someone commenting on a particularly high-quality video might say "I've finally gotten Premium TikTok." People share weird niches they're on by saying things like "I'm deep into carpet cleaning tok" or "I don't know how but I've found music theory tok." Sometimes it's just one word, like "Sportstok or Liberaltok." Tok has almost come to be a suffix meaning "neighborhood" or "community," almost like Disney uses -land to describe themed areas in its parks like Frontierland or Tomorrowland.

Of course, we're all just in our FYP feeds, which just scrolls up endlessly, so it isn't an actual space. But we trust the visible view counts as evidence FYP is doing its job getting many of us with the same tastes in front of the same videos, and so this evidence of common knowledge creates a liminal third place that exists [waves hands at the air in front of me] out there.

I’ve tended to think of social networks as being built by people assembling a graph of people bottoms up, but perhaps I’ve been too narrow-minded. TikTok might not qualify by that definition, but it feels social, with FYP as village matchmaker.


There's been a lot written on Warner Media's decision to move some films from theatrical only windows to having a concurrent release on HBO Max. A lot of conclusions were drawn about theatrical's future based on Wonder Woman 84’s Christmas premiere in theaters and on HBO Max day-and-date. A lot of it is the usual knee jerk extrapolation that the internet is famous for, despite confounding circumstances like a pandemic, and despite Wonder Woman 84 being a single data point.

But one thing I'm confident of is that something is lost in not having the audible feedback of a hundred or more humans around you when you watch something, especially from genres that are built to elicit frequent emotional feedback, like comedies and horror films. At some point, perhaps we'll crack the nut on social viewing and how to make it more, umm, social, but for now, pre-VR metaverse, it's a shoddy facsimile of a crowd.

Look, I've streamed my share of concerts during this past year, and I don't miss standing for an hour between sets in a crowded club or bar, nursing a $9 beer in a plastic cup, waiting for my band of choice to get on stage.

And yet, I miss standing in that bar, my shoes sticking to the beer-soaked floor, trying to look at ease in my own skin while gawking at other humans.


In a year where we've been trapped inside for nearly a year now, there's something about the chaotic collectivist media art form that is TikTok that felt most joyful and genuine.

Thumbing through the FYP feed one portrait-oriented rectangle at a time felt like swiping from one bedroom window to the next on a tall skyscraper, peering into one user's bedroom after another (literally, as the bedroom is the most common space in which teens do their creative work). It's like a Chris Ware comic strip, with its architectural design, navigated one window pane at a time.

Because it's full screen, it can feel like my phone screen is literally a rectangular porthole. As if one user after another is hijacking my rear-facing camera and turning it into their rear or front-facing camera.

There's something about media like TikTok or ChatRoulette or Omegle, where so much of what you see is a creator directly addressing the camera, breaking the fourth wall from the start, that is immediately intimate.


One thing I wish TikTok would do is make it easier to trace multi-part videos from creators. Nothing drives me crazier than videos that end with "Stay tuned for Part 2" or "Like for Part 2" and then you spend like ten minutes browsing their profile trying to find the second part.

I understand that it's a sort of view count hack on the part of creators, but some videos do need to be broken up across installments. TikTok needs to add some sort of concept of a pointer or link to make it easier to jump directly to the next installment in a series. Perhaps it could be done via a playlist feature.

For now, the best way to trace linked videos is to visually scan the thumbnails on a person's profile and search for onscreen text reading "Part #" or just click on every video with the same visual grammar, the user in the same outfit in the same room with the same lighting.

(Since I wrote the note above, the app has added a way to highlight, on a creator's profile page, the video you just watched, and since videos are sorted reverse chronologically by creation date on the profile, often part II can be found next to the video you just watched, which is handy).


In another example of the community coming up with creative solutions, commenters on the first in a multi-part video series where the next part has yet to be published will now leave a comment saying that they'll promise to tag people who like their comment once the next installment is posted. In other words, users are serving as Mechanical Turk notification bots.


Another feature I wish TikTok would add is the ability to sort by descending popularity on any grid of videos, like on sound or profile pages. Please.


TikTok's needs to improve its search ranking algorithm. Trying to find popular TikTok's I remembered seeing back in the day was much harder than it should have been using TikTok's native search. A couple that I wanted to use I just couldn't locate, and even Google and YouTube didn't turn them up (a thing you realize after trying to do it more than once is how hard it is to create a comprehensible search query for certain TikTok's).


Network effects are powerful, but there are so many distinct types. It's important to understand exactly what type of network effect you have because they all scale and operate differently.

For example, Dunbar's number is just one form of limit on a very specific type of network effect. But there are dozens and dozens of network effects, all with their distinct quirks. Someone could make a lot of money just making a reference book of the taxonomy of network effect varietals in the world.

TikTok is an extreme experiment in not only making creative network effects endogenous to its app but to the medium of video. Like some video Minecraft, almost everything in the app is a replicable chunk of bits that you can combine into a larger configuration of bits, and the resulting creation becomes, itself, a chunk that anyone can take and splice or mutate or combine however they want.

This is anathema to old media, most of whom have spent hundreds of millions of dollars to lock up their content behind copyright law, DRM, and any number of other mechanisms meant to slow the rate of reproduction and iteration of their work. It has the effect of slowing the evolutionary feedback loops on all of that work.

TikTok's "OODA loop" is collective and distributed, and it spins thousands of times faster than that of big media.


When I first joined the Amazon Web Services team in 2003, it was still a small Jeff Bezos-sponsored project. There were only some 15 people or so on the team at the time under the leadership of now Amazon CEO Andy Jassy.

A book Jeff had us read, one which he said should serve as an inspiration for how we'd design AWS, was Creation: Life and How to Make It by Steve Grand. It's a book about programming artificial life, but the core principle that Jeff wanted us to take from it was the idea that complex things like life forms are built from very simple building blocks or primitives. It's the same thesis as that in Stephen Wolfram's A New Kind of Science.

The key implication for AWS from the book was about how to design the first AWS primitives. Jeff urged us to include only what was necessary and nothing more. If you were designing a storage service, like S3, you'd need functions like get, write, delete, but you wouldn't want to layer in things that weren't part of storage, like security. That should be a separate primitive.

The reason to design your primitives with the utmost elegance is to maximize combinatorial optionality.


This is one of the most elegant things about TikTok's design. It includes a ton of primitives, and they are almost all ones you can combine or link.

More than that, every element in a TikTok is a building block you can replicate and use in your own TikTok. The most important of these is the soundtrack or sound of your TikTok.


Be careful of taking this idea of building primitives too far. In many ways, choosing what level of abstraction to stake your ground on is one of the most important questions any company must answer.

The answer is contextual. Abstract at too high a level and someone can come in beneath you, with something like AWS. In some ways this is a form of disruption.

Build at too low a level, however, and often someone will abstract at a level above you and siphon all the value of that market above your product. Many of TikTok’s filters are abstractions of a lot of things, almost like Lightroom Presets. As many of us learned early in this pandemic, maybe paying a few bucks for a loaf of bread is preferable to having to spend hours of our free time mastering baking.


When I think about modern remix culture and apps like TikTok, I often think back to Mixel, an app designer Khoi Vinh launched years ago. It was an iPad collage app.

In his blog post introducing Mixel, Vinh wrote:

Because of the componentized nature of collage, we can add new social dimensions that aren’t currently possible in any other network, art-based or not. Mixel keeps track of every piece of every collage, regardless of who uses it or how it’s been cropped. That means, in a sense, that the image pieces within Mixel have a social life of their own. Anyone can borrow or re-use any other piece; you’re free to peruse all the collages (we call them “mixels”) and pick up literally any piece and use it in your own mixel. If you don’t like the crop, the full, unedited original is stored on the server, so you can open it back up in an instant and cut out just the parts you like. Mixel can even show you everywhere else a particular image has been used, so you can follow it throughout the network to see how other people have cropped it and combined it with other elements.

The thread view turns collaging into a visual conversation, where anyone can remix anyone else’s work.


Though Mixel is no longer around, what he describes presages modern meme culture and TikTok.

Inherent to digital culture is the remix.


In Mark Ronson's TED Talk on How Sampling Transformed Music, he says:

That's what the past 30 years of music has been. That's the major thread. See, 30 years ago, you had the first digital samplers, and they changed everything overnight. All of a sudden, artists could sample from anything and everything that came before them, from a snare drum from the Funky Meters, to a Ron Carter bassline, the theme to "The Price Is Right." Albums like De La Soul's "3 Feet High and Rising" and the Beastie Boys' "Paul's Boutique" looted from decades of recorded music to create these sonic, layered masterpieces that were basically the Sgt. Peppers of their day. And they weren't sampling these records because they were too lazy to write their own music. They weren't sampling these records to cash in on the familiarity of the original stuff. To be honest, it was all about sampling really obscure things, except for a few obvious exceptions like Vanilla Ice and "doo doo doo da da doo doo" that we know about. But the thing is, they were sampling those records because they heard something in that music that spoke to them that they instantly wanted to inject themselves into the narrative of that music. They heard it, they wanted to be a part of it, and all of a sudden they found themselves in possession of the technology to do so, not much unlike the way the Delta blues struck a chord with the Stones and the Beatles and Clapton, and they felt the need to co-opt that music for the tools of their day. You know, in music we take something that we love and we build on it.


One of the most revolutionary aspects of TikTok is how effortless it makes it to sample or interpolate any other TikTok video.


Anyone who's used a non-linear editor like Adobe Premiere, Final Cut Pro, or Avid Media Composer knows the standard multi-pane interface. And any editor knows that editing begins with importing all the media, the shots from dailies, the temp music, and so on, into your media bin. From there, you drag elements onto the timeline to compose the edit.

Much of the pain of creating memes is gathering all the components, like images, from the web. In the modern networked age, though, the media bin should really just be the entirety of the internet. Anything you want should just be a short search away. We're starting to get closer, though the library of material is still sparse, and many pieces, especially video, still require chasing down.

Someday, any sort of remix will just be a GPT-3 like interface away from composing. You'll just be able to write "This is Fine cartoon but the dog's face is Donald Trump" and it will just spit it out for you. If you're building this, please let me know, I'll write you a seed check.


The Verge interviewed a TikTok beatmaker named Ricky Desktop.

What makes a great TikTok beat?

You need concrete, sonic elements that dancers can visually engage with on a person-by-person basis. I know that sounds super scientific, but that is how I think about it. If you’re trying to make a viral beat, it’s got to correspond with the viral dance.

In order to lock in on that, you need elements of the music to hit. So for example, I have this beat called “The Dice Beat.” I added a flute sound, which in my head was like, “Okay, people will pretend to play the flute.” And then there’s the dice sound, where they’ll roll the dice. It was super calculated. I would create the music with the dance in mind.

I developed this little pattern. I pioneered the “triple woah” thing where in all the beats there’s three kicks — bum-bum-bum. So typically, when the bass drop hits, the dancers will do the woah (ed: an accentuated arm and elbow movement popular in TikTok dances) to emphasize the bass drop. Usually, the beat will keep going after that. But what I did, I would add three more bass hits, super calculated, so that dancers could do the woah three times or do three concrete dance accents.


The woah inspired Ricky Desktop to develop a score for the triple woah which then actually inspired dancers to choreograph and perform an actual triple woah.

Can you program human movement with music? It turns out you can. You use an API called TikTok. That's delightful.


TikTok beatmaker Ricky Desktop pictured, in his head, dancers performing some movement. Then he wrote a piece of music that included a musical cue intended to elicit that exact movement.

Then, later, some dancers on TikTok performed the movement he had pictured, exactly at the moment he had inserted the musical prompt. It's not just that he choreographed the human body via music, but how he did it. Ricky Desktop is a marionettist manipulating human bodies not via strings but music.


Ricky Desktop:

So I would post my beat and say, “Anyone trying to help me make this beat go viral?” Or I would say, “Who’s gonna create a dance to this new banger?” I’m giving an action item to whoever’s watching. And that’s important because it gives the person watching something to do.


The message is in the medium. That is, Ricky Desktop issues these to-dos inside of the video he uses to release his various sounds.


Ricky Desktop:

What makes a great TikTok beat?

You need concrete, sonic elements that dancers can visually engage with on a person-by-person basis. I know that sounds super scientific, but that is how I think about it. If you’re trying to make a viral beat, it’s got to correspond with the viral dance.

In order to lock in on that, you need elements of the music to hit. So for example, I have this beat called “The Dice Beat.” I added a flute sound, which in my head was like, “Okay, people will pretend to play the flute.” And then there’s the dice sound, where they’ll roll the dice. It was super calculated. I would create the music with the dance in mind.


In filmmaking, when you want a score for your film, you bring the latest cut of your film to a composer's studio, and they start riffing based on what they see on screen, incorporating some of the themes you're trying to evoke in that scene.

What Ricky Desktop talks about above is a different process in which he scores to visuals that only exist in his imagination, generic dance tropes like "pretend to play the flute".

This is a form of "inverted scoring." Or, if you prefer to go from the other direction, what TikTok dancers do with sounds is "visualizing."

The program WinAmp used to do software visualizations of music. TikTok is like Mechanical Turk for visualizing music.


If you've watched any amount of TikTok, you've doubtless seen someone answering questions by dancing and pointing to floating text overlays.

Now, they could easily just speak the questions and answer them verbally. There's no reason to have to dance to music while answering the questions.

To which I say, no one knows what it means, but it's provocative, it gets the people going!

Kylie Jenner (@kyliejenner) has created a short video on TikTok with music original sound. | i'm still a supermodel on the inside | INSTAGRAM MODEL | SUPERMODEL | DADS FAV MOMS FAV | ...

This is one of many TikTok survey or poll formats, all devised by the users. On one hand, there are simpler ways to share this information. On the other hand, this is much more entertaining than a Twitter poll.


On the other hand, maybe all this choreographed dancing is something more of us should be doing to make our messages land. A teacher went viral on TikTok this year for filming herself trying to teach her class remotely over Zoom. Seeing her precise and broad gestures paired with her sharply articulated speech, you couldn't help but feel empathy for what a burden we've placed on our teachers, trying to make remote classes engaging over Zoom.

But perhaps we just lose some of our childlike exuberance and joy expressiveness as we age? Perhaps if we were more animated in our delivery, more people would remember what we said.

One of the most common weaknesses among managers and leaders is the illusion of transparency, though it is a problem for most people. It is the tendency to overestimate how much people know what you're thinking. It can ruin marriages or relationships, and it leads to a healthy market for therapy.

Young children have the a strong form of this illusion which is why in early childhood they are so frustrated when you don't understand why they're upset (and parents are likewise just as exasperated that their children can't verbalize why they're freaking out). Until later in life, children think you should know exactly what they're feeling, and it takes a bit of coaxing to tease out their inner emotional state. Ironically, despite their illusion of transparency, kids tend to be much more emotionally transparent and thus expressive.

It's when they finally realize that no one can see into their heads that they learn to lie. It's then that you wish they still had the illusion of transparency. When they become teenagers, the battle over transparency into their lives becomes literal: parents yell at their teenagers to keep their bedroom doors open, and those same doors slam shut after heated arguments. Their bedrooms become, like their thoughts, spaces they wish to protect from prying eyes.

This is all a roundabout way to say that a CEO communicating a company's top goal for the coming year in a TikTok dance, pointing to on-screen captions, isn't the worst idea in the world? Maybe this is the new Amazon 6-page memo.


Study any high-level memory competitor and they'll all say the same thing. Humans' visual memories are far superior to their memories for abstractions. It's one of the core lessons from the great book Moonwalking with Einstein. It's the reason people who have to try to memorize a thousand digits of pi or the order of a deck of cards turn numbers and letters into images which they place spatially in memory palaces.

In its heyday, which coincided with my childhood, MTV was dominated by music videos, and each of those was essentially a visualization of a musical track. To this day, I can't hear a song like A-Ha's "Take on Me" without picturing its music video. I haven't seen it in decades, but its cartoon sketches come to life are forever how I "see" the song. Likewise, I can't hear Michael Jackson's Thriller without conjuring its epic music video of nearly 14 minutes.

It doesn't even have to be a music video. A song incorporated in a film can permanently bond with the moving images on the screen. For example, I can't hear three tracks, one each by Huey Lewis, Genesis, and Whitney Houston, respectively, without picturing Christian Bale and quoting Patrick Bateman, and then being filled with a sense of self-loathing for having been indicted as someone who turns to the appreciation of cultural artifacts as a substitute for personality. If I mention Celine Dion's song "My Heart Will Go On," what do you see in your mind's eye?

TikTok is the modern MTV because (1) it increases consumption of music tracks that go viral on its platform as sounds and (2) any number of songs will forever summon the accompanying meme and visual choreography from my memory.


When Charli and other TikTokers formed the Hype House in Los Angeles, they were experimenting with IRL creative network effects. They created what was efffectively a commune to produce the D'Amelio TikTok Universe with Charli at the center as, I don't know, Tony Stark or something.

They started guest-starring in each other's TikTok's, some of them started dating and hooking up, and soon, to follow the entire extended narrative, you had to follow each other's accounts. Studios have tried to push out fictional versions of such networked series, but Charli et al just created it bottoms up, with TikTok as the distributor.

The Kardashian-Jenner clan are the clear predecessors who ran this type of crossover mindshare grab, but they're family. This new generation of influencers often aren't related, their common bond is just that they're young and famous in the age of social media and so they already all live together in a virtual universe held together by the gravity of popularity.


In Status as a Service, I wrote about how social networks require some proof of work to gain status.

A lot of TikTok's have the caption "I spent way too long on this" as a sort of plea for likes, but that wouldn't land if the proof of work wasn't visible on the screen. It is, and even non-creators can see it. Some TikToks seem like they took days to produce.


Have you tried using the in-app TikTok video editor? In some ways, it's loaded with really first-rate filters and effects, but in many ways, its user interface is inscrutable. I went to editing school and have used a variety of non-linear editors (NLEs) like FCP, Premiere, and Avid to edit video in a previous life, and I still tear my hair out trying to use TikTok's native editor.

The easiest videos to make are just ones where you film yourself live and apply a filter, but if you want to bring in pre-recorded video and mix them with other graphical elements, like text boxes, it is very painful to assemble them properly. My kingdom for a persistent timeline with a scrubber in the TikTok editor.

In one sense, it's staggering to ponder how many more videos TikTok would have if its video editor were more usable. On the other hand, every video that does make it onto the app feels like a miracle. The proof of work is in the pain.


If you're a movie star like Will Smith and you get a VFX studio to produce some whiz-bang TikTok for you, it will feel off, like driving a Ferrari down the street in Omaha. Authenticity or at least the sheen of one's own sweat equity is part of the TikTok aesthetic, and the canonical backdrop for any TikTok video is always some teenager's somewhat messy bedroom, just as it was in the heyday of the YouTube vlog.

On Instagram, you can get away with proof of wealth, but the TikTok aesthetic is proof of creative labor. The verdict is a bit more mixed on proof of hotness, though. I still think Instagram is a more welcoming home for pure thirst trap content than TikTok, where, if you want to honeytrap the simps, you're going to have to dance for it.


Something about a feed that can hit you with such a variety of styles and moods in such quick succession makes TikTok feel like the most modern of media channels. One second you're watching a dog communicate with their owner using a language mat, the next second some high school girl is roasting one of her classmates, the next you see a teen making an earnest confessional deprecating their own looks (only to have thousands of commenters offering affirmation), and then you might see a boat chase that you later realize is some drug cartel member filming a TikTok as police boats give chase (even Narcos be chasing them likes). At times it feels as if the FYP feed is a pastiche generator.

It is equal parts ironic and earnest, having long since surpassed its label as the cringey social network.

Whereas Instagram is performative, TikTok is performative and self-aware. It’s not that any single creator is self-aware, but that the Greek chorus in the comments will descend on anyone with the slightest bit of hubris like a pack of harpies.

In this rectangular proscenium that is TikTok, the fickle god of Zeus is played, of course, by the FYP algorithm. Everyone offers up their sacrifices of time and labor in the hopes of being graced by its favor, but its whims remain just capricious enough to keep everyone grinding.


If your FYP feed is dialed in to your tastes, you start to pre-react to videos purely based on the like count visible on the right hand side of the screen.

If a video has a high like count, even if it starts slowly I'll tend to give it the benefit of the doubt and stick around to the end, simply because this statistic has proved, in my experience, reliable evidence of a worthwhile payoff. The larger the figure, the more I anticipate a strong punchline or close. I'm like Tom Cruise in Minority Report, already having seen the precog verdict printed on that ball.

Conversely, when I'm the test audience for a little-seen video (a dead giveaway is it has almost no likes yet), I tend to be merciless in skipping ahead if it doesn't hold my attention after a few seconds.

This creates a ruthless rich-get-richer dynamic, but that's by design. Bytedance as a company has built its products around pitiless algorithms enforcing a high Gini coefficient economy of entertainment. It's a marketplace in which the supply side—the TikTok videos from creators—can be shown to an unlimited number of viewers. Much of the content is evergreen, so there is almost no end to the leverage TikTok can get off any single good video.

Imagine if YouTube's key metric was to show every good video in its entire catalog to every viewer that would enjoy it. If you view the TikTok mission that way, even if no one submitted another new video for the next year, its FYP algorithm would still have an almost infinite supply of short videos to show to hundreds of millions of users for that entire dry spell.


Because sounds become the genesis of particular memes, when you start watching a TikTok video and hear a familiar sound, you anticipate the moment of that sound when the punchline will happen. It's Pavlovian.

The kismet shoe transition, for example, causes you to anticipate the pleasure of that exact moment when the performer will go from looking plain to looking EXPENSIVE. There are only so many plots in Hollywood, but we go see genre films precisely for the story beats we know are coming.

On TikTok, sounds and memes are almost inseparable. The sound is the meme is the sound.

TikTok sounds are often the most pleasing snippets from pop songs, and listening to one catchy loop after another is like listening to a pop radio channel that doesn't play entire songs, only plays bass drops and choruses. The time between anticipation and payoff is so short that scrolling the feed can feel like pressing the button on some sonic IV drip over and over. Just inject it into my ears.


In Infinite Jest, David Foster Wallace describes a film called Infinite Jest which is so entertaining people lose all will to do anything except watch it until they die. He had often written about the addictiveness of television and may have been extrapolating to the future, projecting the entertainment value of entertainment increasing until it surpassed some threshold where you'd lose all will to do anything except consume. In that way, he predicted binge watching.

But the earliest form of entertainment that conjured the addictive properties of his fictional film (referred to in the novel as "the Entertainment") was video games. I read stories about players who died after playing games for so long without eating and, recalling some game binge sessions from my youth, could imagine myself trapped in a similar dark loop.

TikTok is the second form of entertainment that brings DFW's fictional entertainment to mind. In hindsight, it seems obvious that a personalized feed of video, tailored to your tastes, would be the addictive end state of entertainment. And, considering the rise of social media and the smartphone, it would make sense that the videos might all be short, like pellets of rain, sliding comfortably into every spare pocket of time in our day, of which we have so many.

One of my favorite paragraphs of recent years was one describing the miracle that are Cheetos:

To get a better feel for their work, I called on Steven Witherly, a food scientist who wrote a fascinating guide for industry insiders titled, “Why Humans Like Junk Food.” I brought him two shopping bags filled with a variety of chips to taste. He zeroed right in on the Cheetos. “This,” Witherly said, “is one of the most marvelously constructed foods on the planet, in terms of pure pleasure.” He ticked off a dozen attributes of the Cheetos that make the brain say more. But the one he focused on most was the puff’s uncanny ability to melt in the mouth. “It’s called vanishing caloric density,” Witherly said. “If something melts down quickly, your brain thinks that there’s no calories in it . . . you can just keep eating it forever.”

TikTok is entertainment Cheetos. Each video requires so little cognitive exertion and reaches its climax so quickly that it feels like we could keep watching forever, each punch line scored to the most satisfying bass drop or stanza from every pop song. TikTok delivers dopamine hits with a metronomic rhythm, and as soon as we swipe up the previous one melts in our memory.


It's always been the case, but especially in this networked age, that every piece of entertainment is its own social network. The network effects of a story arise from shared consumption. The more people watch Star Wars, the more people I can talk to about particular scenes or compare costumes with at a convention. The more people that watch Game of Thrones, the more my Game of Thrones memes will land.

TikTok is personalized, yet through its algorithm it creates shared stories of real scale. Some of these shared stories occur on the creative side in duets and trims that connect creators to each other literally and metaphorically. The FYP algorithm also aggregates large communities of viewers for the hottest TikTok videos. It's not uncommon now for me to send a TikTok to a friend who's already seen it, or vice versa. Not always, but enough that the audience now assumes enough common knowledge to foster that sense of shared experience.

Despite having what must be a gazillion videos in its catalog, watch TikTok enough and you'll be able to refer to something like Sea Shanty TikTok and feel reasonably confident other TikTok addicts get the reference. In contrast, people regularly send me YouTube videos with like millions of views that I've never even heard of.

It is algorithms that may be tearing us apart. But maybe it's also algorithms that reassemble us, albeit in smaller unit sizes. 330M Americans feel like too large an optimal governance size if we're going to let social media algorithms just run amok, but I find some comfort sometimes when I find some TikTok that feels so catered to my tastes that it must be a micro-niche and then see it has millions of likes.


The term binge-watching typically refers to watching multiple episodes of a series in one sitting, but perhaps the act of watching dozens of TikTok videos in a row is the purest form of this type of entertainment gluttony.

Other types of social media like Instagram and Twitter are also series of really compact units of media. When I scroll Twitter or Instagram, I often feel like an elephant, standing there placidly, as various people toss individual packing peanuts at my forehead (let’s call these people the peanut gallery?).

TikTok videos are, for the most part, a bit longer. Their compressed narratives are still, nevertheless, complete, with some full story arc to traverse. In its rhythm, binge-watching TikTok reminds me of watching a standup comedy set, but instead of watching one comedian, I’m watching a whole series of them, each on stage just long enough to tell one joke. And if they bore me, I can press a button and, like a Looney Tunes cartoon, a cane whisks them off the stage and a new comedian pops out from the floor to take their place and start right into their joke.

Someone told me that if you watch TikTok for over an hour it posts a warning asking you to consider taking a break. I'm not sure if that's the case, but I'm glad I've never encountered it yet.


TikTok can only match you with videos it has, and for some people, there may not be enough relevant content in the TikTok catalog to sustain a feed. But that pool of videos has grown by an astonishing amount in a short amount of time.

I'm an easy mark for the sort of wry, sometimes savage humor of TikTok, especially when it skews almost post-modern in its awareness of its own form. It's both a community that constantly tries to legislate its own social norms of decency—any video of someone making fun of how they look using a supposed beauty filter will be flooded with comments like "You're a queen", the comments section being sort of a rolling floor vote on what the acceptable response is—and also a bloodbath of Gen Z violence. The kids will be alright, but that's in part because they're savage. Every generation learns it has to fend for itself.

During a pandemic when most of social media feels even more nakedly performative than usual, as we sit inside day after day for month after month, my occasional sessions on TikTok have been one of the only pastimes to reliably make me laugh, and it's not particularly close.

Twitter has reached a crest of fortune cookie thinkboi bait when it's not subsumed in petty high school lunchroom culture war fistfights. Seemingly every day, a playground brawl breaks out and we all form a circle to gawk, but at the back of our minds is always the threat that we'll be the next to be sucker-punched and forced to throw down. Outrage porn is exhausting and also not that fun?

When viewed from the eye of a global pandemic, Instagram feels like a horrifying Truman Show of idyllic capitalist showboating. Life must go on, influencers gotta influence, but I'm also not weeping any tears when people get chastised for renting private islands and posting photos of themselves partying during a pandemic.

Andrew Niccol, the screenwriter of The Truman Show, once said, "When you know there is a camera, there is no reality.” The most absurd but popular tag on visual social media is #nofilter, a hashtag that aspires to a pretense of truth when there is almost nothing on an app like Instagram that isn’t production-designed within an inch of its life.

TikTok, by virtue of its high bar to even produce a video that anyone will see (FYP algo is like "That's a no for me dawg" on almost every video), is upfront about what it is: a global talent show to entertain the masses. In a pandemic where much of the U.S. lives in eternal lockdown, TikTok is the 24/7 channel where the American Idle entertain each other from their bedrooms. I laughed, and then I laughed some more.

Seeing Like an Algorithm

In my previous post on TikTok I discussed why its For You Page algorithm is the connective tissue that makes TikTok work. It is the bus on its motherboard that connects and closes all its feedback loops.

But in the breathless rush to understand why companies might want to acquire TikTok, should ByteDance be forced to divest itself of the popular short video app, the hype around its algorithm has taken on a bit of exoticization that often characterizes Western analysis of the Chinese tech scene these days.I kept holding off on publishing this piece because every day seemed to bring some new development in the possible ban of TikTok in the U.S. And instead of writing any introduction that would become instantly outdated, I'll just leave this sidenote here to say that as of publishing this entry, it seems Oracle will take over the TikTok cloud computing deal while also joining Wal-Mart and some VCs in assuming some ownership stake in TikTok Global. But it won't surprise me one bit if we find out even more bizarre details over the next week. This is the type of deal that I would have thought could only happen in Succession, but even in that satire it would seem hyperbolic. The 2020 Writer's Room is undefeated.

In this post, I want to discuss exactly how the design of TikTok helps its algorithm work as well as it does. Last time I discussed why the FYP algorithm is at the heart of TikTok’s flywheel, but if the algorithm wasn’t effective then the whole feedback loop would collapse. Understanding how the algorithm achieves its accuracy matters even if you’re not interested in TikTok or the short video space because more and more, companies in all industries will be running up against a competitor whose advantage centers around a machine learning algorithm.

What I want to discuss is how TikTok’s design helps its algorithm “see.”


Seeing Like a State by James C. Scott is one of those books that turns you into one of those Silicon Valley types that use (abuse?) the term legibility. I first heard about it after reading Venkatesh Rao’s summary of its main themes, and that piece remains a good tldr primer on the book if you don’t plan to read the text (Scott Alexander's review of the book is also good though is long enough that it could almost justify its own tldr). However, I recommend that you do.

The subtitle of Scott’s book is “How Certain Schemes to Improve the Human Condition Have Failed.” In particular, Scott dissects a failure state that recurs across a number of domains, in which a governing body like the nation-state turns to what Scott terms high modernism in an effort to increase legibility of whatever it is they are trying to exert control over, whether for the purposes of taxation or conscription or any number of goals. In doing so, they impose a false sense of order on a reality more complex than they can imagine.It would really be fascinating to hear from Scott on the case of modern China, under CCP rule, with modern technology for surveillance, and whether he thinks they will prove or violate his thesis in the fullness of time.

It’s a book that raises one’s awareness of all sorts of examples of unintended consequences in day-to-day life. We all could use a healthier does of humility when we are too flush with great man hubris. The world is richer and more complicated than we give it credit for.

As an example, much of what Scott discusses has relevance to some of the hubris of our modern social networking giants. These dominant apps are designed to increase legibility of their user bases for, among other things, driving engagement, preventing churn, and ultimately, serving targeted advertisements. That, in turn, has led their parent companies into a thicket of problems which they’re grappling with constantly now.

But that is a topic for another post, another day. Whereas Scott focuses in on how the nation-state uses simplifying abstractions to “see” its citizens at a synoptic level, I want to discuss how TikTok’s application design allows its algorithm to “see” all the detail it needs to perform its matchmaking job efficiently and accurately. If Seeing Like a State is about a common failure state, this post is about a new model for getting the most leverage from machine learning algorithms in the design of applications and services.I’m aware of the irony that the controversy around TikTok was the potential of user data being accessed by the CCP, or being “seen by that state.” Or that one of the sticking points of this new Cold War is the Chinese Firewall, which selects what the citizens of China “see.” And which most U.S. tech companies sit outside of, looking in.


In recent years, one of the realizations in machine learning, at least to an outsider to the subject like myself, is just how much progress was possible just by increasing the volume of training data by several orders of magnitude. That is, even if the algorithms themselves aren’t that different than they were a few years ago, just by training them on a much larger datasets, AI researchers have achieved breakthroughs like GPT-3 (which temporarily gave tech Twitter a tantric orgasm).

When people say that TikTok’s algorithms are key to its success, many picture some magical block of code as being the secret sauce of the company. The contemporary postmodernist Russian writer Viktor Pelevin has said that the protagonist of all modern cinema is a briefcase full of money. From the briefcase of radioactive material (I think that’s what it was?) in Kiss Me Deadly to the briefcase of similarly glowing who knows what (Marcellus Wallace’s soul?) in Pulp Fiction, from the Genesis equation in The Formula to the secret financial process in David Mamet’s The Spanish Prisoner, we’ve long been obsessed in cinema with the magical McGuffin. In recent weeks, discussion of TikTok’s algorithm has elevated it into something similar, akin to one of those mystical archaeological artifacts in one of the Indiana Jones films, like the Ark of the Covenant, the Holy Grail, or the lingam Shivling.

But most experts in the field doubt that TikTok has made some hitherto unknown advance in machine learning recommendations algorithms. In fact, most of them would say that TikTok is likely building off of the same standard approaches to the problem that others are.

But recall that the effectiveness of a machine learning algorithm isn’t a function of the algorithm alone but of the algorithm after trained on some dataset. GPT-3 may not be novel, but trained on an enormous volume of data, and with a massive number of parameters, its output is often astonishing.

Likewise, the TikTok FYP algorithm, trained on its dataset, is remarkably accurate and efficient at matching videos with those who will find them entertaining (and, just as importantly, at suppressing the distribution of videos to those who won’t find them entertaining).

For some domains, like text, good training data is readily available in large volumes. For example, to train an AI model like GPT-3, you can turn to the vast corpus of text already available on the internet, in books, and so on. If you want to train a visual AI, you can turn to the vast supply of photos online and in various databases. The training is still expensive, but at least copious training data is readily at hand.

But for TikTok (or Douyin, its Chinese clone), who needed an algorithm that would excel at recommending short videos to viewers, no such massive publicly available training dataset existed. Where could you find short videos of memes, kids dancing and lip synching, pets looking adorable, influencers pushing brands, soldiers running through obstacle courses, kids impersonating brands, and on and on? Even if you had such videos, where could you find comparable data on how the general population felt about such videos? Outside of Musical.ly’s dataset, which consisted mostly of teen girls in the U.S. lip synching to each other, such data didn’t exist.

In a unique sort of chicken and egg problem, the very types of video that TikTok’s algorithm needed to train on weren’t easy to create without the app’s camera tools and filters, licensed music clips, etc.

This, then, is the magic of the design of TikTok: it is a closed loop of feedback which inspires and enables the creation and viewing of videos on which its algorithm can be trained.

For its algorithm to become as effective as it has, TikTok became its own source of training data.


To understand how TikTok’s created such a potent flywheel of learning, we need to delve into its design.

The dominant school of thought when it comes to UI design in tech, at least that I’ve grown up with the past two decades, has centered around removing friction for users in accomplishing whatever it is they’re trying to do while delighting them in the process. The goal has been design that is elegant, in every sense of the word: intuitive, ingenious, even stylish.

Perhaps no company has more embodied this school of design than Apple. At its best, Apple makes hardware and software that is pleasingly elegant—“it just works”—but also sexy in a way that makes its users feel tasteful. Apple’s infamous controlling style—no replaceable batteries for its phones and laptops, the current debate over its App Store rules—put the company squarely in the camp of what Scott in Seeing Like a State refers to as high modernism. Is there any reason to show a video of how the new MacBook Pro body is crafted from one solid block of aluminum (besides the fact that Jony Ive cooing “a-loo-MIN-eee-um” is ASMR to Apple fans) when unveiling it at an Apple keynote? How about because it’s sexy AF to see industrial lasers carving that unibody out of a solid chunk of aluminum? And later, when you’re cranking out an email at a coffee shop on said laptop, some residual memory of that video in your unconscious will give you just the slightest hit of dopamine?

There’s a reason this user-centric design model has been so dominant for so long, especially in consumer tech. First, it works. Apple’s market cap was, at last check, over 2 trillion dollars. Remember when fake Sean Parker said a billion dollars was cool? That was just a decade ago and a billion dollars is no longer S-Tier. The wealth meta moves fast. Furthermore, we live in the era of massive network effects, where tech giants who apply Ben Thompson’s aggregation theory and acquire a massive base of users can exert unbelievable leverage on the markets they participate in. One of the best ways to do that is to design products and services that do what users want better than your competitors.

This school of design has been so dominant for so long that I’ve almost managed to forget some of the brutal software design that used to the norm in a bygone era.Not to be confused with brutalist design, which can be quite beautiful in its own respect, like its architectural cousins.

But what if the key to serving your users best depends in large part upon training a machine learning algorithm? What if that ML algorithm needs a massive training dataset? In an age when machine learning is in its ascendancy, this is increasingly a critical design objective.

More and more, when considering how to design an app, you have to consider how best to help an algorithm “see.” To serve your users best, first serve the algorithm.

TikTok fascinates me because it is an example of a modern app whose design, whether by accident or, uhh, design, is optimized to feed its algorithm as much useful signal as possible. It is an exemplar of what I call algorithm-friendly design.I thought about calling it algorithm-centric design but felt it went too far. Ultimately, a design that helps an algorithm see is still doing so in service of providing the user with the best possible experience. This might still be considered just a variant of user-centric design, but for those teams working on products with a heavy machine learning algorithm component, it may be useful to acknowledge explicitly. After all, when a product manager, designer, and engineer meet to design an app, the algorithm isn't in attendance. Yet its training needs must be represented.

James Scott speaks of “seeing like a state,” of massive shifts in fields like urban design that made quantities like plots of land and their respective owners “legible” to tax collectors. TikTok’s design makes its videos, users, and user preferences legible to its For You Page algorithm. The app design fulfills one of its primary responsibilities: “seeing like an algorithm.”

Let’s take a closer look. TikTok opens into the For You Page and goes right into a video. This is what it looks like.

This is, as of right now, the most popular TikTok ever. By the time I publish this post, its 34.1M likes will likely be outdated. You can read the story of how this TikTok even came to be and it will still feel like a cultural conundrum wrapped in a riddle stuffed in a paradox, and you love to see it. I showed this to my niece, we looped it a few dozen times, then we started chanting “M to the B, M to the B” and laughing our asses off and it was one of the only times in this pandemic I’ve truly felt anything other than despair.

The entire screen is filled with one video. Just one. It is displayed fullscreen, in vertical orientation. This is not a scrolling feed. It’s paginated, effectively. The video autoplays almost immediately (and the next few videos are loaded in the background so that they, too, can play quickly when it’s their turn on stage).

This design puts the user to an immediate question: how do you feel about this short video and this short video alone?

Everything you do from the moment the video begins playing is signal as to your sentiment towards that video. Do you swipe up to the next video before it has even finished playing? An implicit (though borderline explicit) signal of disinterest.

Did you watch it more than once, letting it loop a few times? Seems that something about it appealed to you. Did you share the video through the built-in share pane? Another strong indicator of positive sentiment. If you tap the bottom right spinning LP icon and watch more videos with that same soundtrack, that is additional signal as to your tastes. Often the music cue is synonymous with a meme, and now TikTok has another axis on which to recommend videos for you. Did you tap into the video creator’s profile page? Did you watch other videos of theirs, and did you then follow them? In addition to enjoying the video, perhaps you appreciate them in particular.

But let’s step back even earlier, before you’re even watching the video, and understand how the TikTok algorithm “sees” the video itself. Before the video is even sent down to your phone by the FYP algorithm, some human on TikTok’s operations team has already watched the video and added lots of relevant tags or labels.

Is the video about dancing? Lip synching? Video games? A kitten? A chipmunk? Is it comedic? Is the subject a male or female? What age, roughly? Is it a group video? Where is it set? What filters or visual effects are used? If there’s food involved, what kind? And so on. All of these labels become features that the algorithm can now see.

Vision AI also does a pass on the video, and to the extent it can, contributes what it sees. Some of TikTok’s camera filters are designed to track human faces or hands or gestures so vision AI is often invoked even earlier, at the point of creation.

The algorithm can also see what TikTok already knows about you. What types of videos have you enjoyed in the past? What demographic or psychographic information is known about you? Where are you watching the video? What type of device do you have? And so on. Beyond that, what other users are similar to you?

Let's jump back to the moment you watch that video on your phone in TikTok. The FYP algorithm can now close all the feedback loops. It takes every one of the actions you take on the video and can guess how you, with all your tastes, feels about this video, with all its attributes.

None of these individual steps sounds like rocket science, especially to anyone who works on any algorithmic social feed today.In my previous piece I noted that TikTok doesn’t really have a strong social graph. One of the reasons the app is as effective as it is is that it doesn’t try to pretend to be what it isn’t. That is, people already have a gazillion other social graphs and ways to share with people they know. Rather than force people to do so within the TikTok app, they make it dead simple to download videos or share them through those external channels. What TikTok keeps, however, is the signal that you chose to share that video. That data feeds their algorithm and their algorithm alone. Since the videos are watermarked, they also get a nice hit of free publicity from the share. In fact, TikTok has published a blog post describing essentially how their FYP algorithm works, and I doubt anyone in tech will find the description anything but obvious.

But contrast what TikTok's FYP algorithm sees with what a comparable recommendation algorithm sees on most other social networking feeds.

The default UI of our largest social networks today is the infinite vertically scrolling feed (I could have easily used a screenshot of Facebook above, for example). Instead of serving you one story at a time, these apps display multiple items on screen at once. As you scroll up and past many stories, the algorithm can’t “see” which story your eyes rest on. Even if it could, if the user doesn’t press any of the feedback buttons like the Like button, is their sentiment towards that story positive or negative? The signal of user sentiment isn’t clean.

If you subscribe to the idea that UI's should remove friction, the infinite scrolling feed is ideal. It offers a sense of uninhibited control of the pace of consumption. The simulated physics that result from flicking a feed with your thumb and seeing it scroll up like the drum of the Big Wheel from the Price is Right Showcase Showdown with the exact rotational velocity implied by the speed of your initial gesture, seeing that software wheel gradually slow down exactly as it would if encountering constant physical friction, it’s one of the most delightful user interactions of the touchscreen era. You can scroll past a half dozen tweets or Facebook feed items in no time. Wheeeeeeee!

A paginated design, in which you could only see one story at a time, where each flick of the finger would only advance the feed one item at a time, would be a literal and metaphoric drag.

On the other hand, maybe you wouldn’t mind reading one tweet at a time if they were better targeted, and maybe they would be better targeted if Twitter knew more about which types of tweets really interest you. And maybe Twitter would know more about what really interested you if you had to give explicit and implicit positive or negative signals on every tweet.

Even on a story a user does engage with, judging sentiment is a challenge. Most apps only have positive feedback mechanisms, most typically some form of a like button. Since apps like Facebook, Instagram, and Twitter are built around social graphs, it’s obvious why they might opt not to offer dislike buttons.

But, as Stephen King wrote in On Writing, "If you expect to succeed as a writer, rudeness should be the second-to-least of your concerns. The least of all should be polite society and what it expects. If you intend to write as truthfully as you can, your days as a member of polite society are numbered, anyway."

By relying on a long scrolling feed with mostly explicit positive feedback mechanisms, social networks like Facebook, Twitter, and Instagram have made a tradeoff in favor of lower friction scanning for users at the expense of a more accurate read on negative signal.You see another variant of this tradeoff at longstanding companies with the same founding CEO. That person tends to surround themselves with a C-Suite that follows their lead, works well with them. The danger of being surrounded by yes-men is not having anyone to challenge the blindspots in your thinking. It's always worth asking who the people are who are powerful enough to actually change the minds of people like Bezos, Cook, Zuckerberg, Musk. Often the answer is no one, so their blindspots become the blindspots of the company.

Networks that are built around interest graphs, like Reddit, do tend to incorporate down voting mechanisms because their prime directive to keep users from churning is to serve them the most interesting content. That means weeding out uninteresting content as much as it does surfacing appealing content.

TikTok doesn’t have an explicit downvote button, but by serving you just one video at a time, they can infer your lack of interest in any single video based on whether you churn out of that video quickly A quick swipe up before a video has completed is like swiping left on Tinder. The best TikTokers have an intuitive sense of the narrative pace that is appropriate for that platform. How long can you drag out the punching or payoff without losing the viewer, how do you have a set up that keeps the user involved. Using a music cue that has already been co-opted into a meme helps because the bass drop or musical payoff foreshadows when the punchline of the video will drop; a viewer knows how much longer before they reach the payoff. Also viewers may stick around just for the pleasure of hearing that musical resolution.
and by which positive actions you don’t take.

If you click into a text post by someone on Facebook but don’t comment or like the post, how can Facebook judge your sentiment toward that post? Maybe you thought about disagreeing violently in comments, but the person is a coworker or friend of a friend and you decided the better of it. That negative sentiment is difficult to capture; the algorithm can’t “see” your feelings.Most social networks have explicit reporting features for reporting offensive and/or abusive content, but those features are buried and most users don’t resort to them. By the time someone does use a feature like that, you’ve usually already made a grave mistake far upstream and it’s too late to salvage most of the damage that’s been done.

It’s the content that’s boring or causes mild displeasure that is the slow killer. In my previous post, I noted that content derived from a social graph can drift away from a user’s true interests because of the mismatch between your own interests and those of people you know. The switch from a chronological to algorithmic feed is often the default defensive move against such drift.

But if the algorithm isn’t "seeing" signals of a user’s growing disinterest, if only positive engagement is visible, some amount of divergence is unavoidable. You might see that a user is slowly losing interest, not liking as many items, not opening your app as often, but precisely which stories are driving them away may be unclear. By the time they're starting to exhibit those signs of churn, it's often too late to reverse the bleeding.

Algorithm-friendly design need not be user-hostile. It simply takes a different approach as to how to best serve the user’s interests. Pagination may insert some level of friction to the user, but in doing so, it may provide the algorithm with cleaner signal that safeguards the quality of the feed in the long run.

Minimizing friction is merely one means to a great user experience. The goal of any design is not to minimize friction, it’s to help the user achieve some end. Reducing friction is often consistent with that end, but not always. You might say that the quote tweet reduces the friction of manually copying someone else’s tweet, but reducing friction to organizing a mob to pile on someone might not be a core mechanic you want to encourage if your goal is civil public discourse. Some forms of friction are good.

You'll hear many power Twitter users counseling others to make use of muting and blocking early and often.Some users even make liberal use of soft blocking to surreptitiously remove followers.
Users proudly tweet screenshots of words they've muted as a sign of their displeasure with some popular topic of discussion (or their intellectual superiority to said topic). Non sports fans tweet about "sportsball," others tweet "I'll bite, what's X?" where X is something everyone is discussing. Some people have gone so far as to unfollow everyone and start their following from scratch again.At some point, and likely because it A/B tested well, Twitter started showing users tweets that people they followed had liked, even from people that user didn't follow themselves. This does occasionally show me tweets of interest, but what it also does is increase, on an absolute basis, the number of tweets I have no interest in and have to scroll past. I'm a broken record on this: no two people have the exact same interests. The launch of this feature has me really considering unfollowing everyone and starting from scratch, but I also worry about hurting people's feelings, because I'm a softie. If Twitter were structured differently this wouldn't be an issue.

I sometimes think about adopting some or all of these strategies myself, but for Twitter, the necessity of these is itself a failure of the service. If the algorithm were smarter about what interested you, it should take care of muting topics or blocking people on your behalf, without you having to do that work yourself. As I wrote last time, that you have to follow people at all on Twitter to get interesting content is, one could argue, a design flaw for what could be a powerful interest graph.

Not only does TikTok capture very clean signals of sentiment from its users, it also gathers a tremendous volume of them per session. Videos on TikTok are so short that even in a brief session, TikTok can gather a lot of feedback on your tastes.

The process is relatively painless, too. At worst, a few videos might bore you, but swiping them away makes for relatively painless work, and since the algorithm listens closely to your feedback, you may even enjoy dismissing videos knowing that the app will register your displeasure and act on it.Short video happens to be a category quite suited to this type of machine learning-driven recommendation. By no means would I imply that it would work for every type of category. Music works well. It is short in duration so the sampling cost is low, and the repeat consumption value is high. Musical similarities tend to be mathematically detectable. My Spotify Radio recommendations are solid. On the other hand, algorithmic movie recommendations have never really felt magical to me. Movies are very long, the sampling cost is very high. The corpus is small, and only something like 500 or so movies come out each year, of which most people only see a handful. This entire subject is worth a separate post.

By the way, TikTok isn’t the only app with an interface that is optimized for the task of matching, with an interface that shows you one entity at a time so as to be more clear on how you feel. Before TikTok, we had a whole category in which the one-item-at-a-time audition-style UI was dominant.

There’s a reason swipe right and swipe left have become shorthand slang for signaling approval and disapproval, generally. Tinder came up with what feels like a design primitive on a touchscreen UI for binary voting.

In this software era, true competitive advantages, or moats, are increasingly illusory. Most software features or UI designs can be copied easily by an incumbent or competitor overnight. All you will have done is test the impact of the design for them.On one of my trips to China, I was at a dinner with a large group of Chinese entrepreneurs, and I mentioned the hubbub over Instagram copying Stories from Snapchat. One of the chief product officers of one of China’s top companies laughed and remarked, “In China, if your competitor doesn’t copy one of your successful features inside of two weeks, they must be incompetent.” In many ways, the Chinese tech scene is the true Darwinian marketplace of ideas that Silicon Valley thinks of itself as. This bodes poorly for the relative output of Silicon Valley because the rate of idea spread and mutation occurs more quickly in China. Silicon Valley is often said to have taken over as the geographic center of technology innovation from Boston’s Route 128 in part because Silicon Valley’s more open labor markets allowed ideas to move freely among companies. China has taken that playbook and pushed it even further. Surviving the competitive landscape of the Chinese tech scene is like trying to climb out of that pit in The Dark Knight Rises. Terrifying.

But if you can create a flywheel, like TikTok’s, it becomes much harder for a competitor like Reels or Triller to catch up. Triller may pay some influencers from TikTok to come over and make videos there, Reels might try to draft off of existing Instagram traffic, but what makes TikTok work is the entire positive feedback loop connecting creators, videos, and viewers via the FYP algorithm.

In tech, an industry that epitomizes Brian Arthur’s Increasing Returns and Path Dependence in the Economy perhaps more than any other, the first competitor to achieve product-market fit can run away from the pack. If more and more markets feel like they are winner-take-all, or winners-take-all, that is because in an increasingly interconnected world, they are.

Bytedance is often described as the algorithm company, and TikTok has been described over the past few weeks as powered by just such algorithmic black magic. Many have gone so far as to say that TikTok wouldn’t be worth purchasing if the algorithm weren’t included.

That’s a mistake, in my opinion. Yes, retraining the FYP recommendations algorithm might take so long that some users would churn. I don’t mean to trivialize that task. But the actual magic is how every element of TikTok's design and processes connect with each other to create a dataset with which the algorithm trains itself into peak performance. No single step in that loop is beyond the capabilities of any of the many U.S. suitors. All that’s needed is an understanding of how the flywheel works and a commitment to keep every element and process in it functioning.

All around me, I encounter products or services that seem to have hit a ceiling in the quality of their algorithmic recommendations: Yelp, OpenTable, Amazon, Google, Netflix, and on and on. Don't get me wrong, some of them are at rest in a good place. But I can't help but feel there is another leap to be made in some of these, and that perhaps more algorithm-friendly design might be one of the possible solutions.

To recap, in part one of my series on TikTok, I discussed how the algorithm acts as an matching mechanism that makes TikTok such a scalable entertainment network. In comparison, social networks have to approximate an interest graph using a social graph, with all the problems that come with that. In this second piece on TikTok, I’ve focused on how its design helps its machine learning FYP algorithm “see” what it needs to see to do its job so effectively. An algorithm-friendly design ethos may become a model of how other companies in other verticals might achieve an edge in the age of machine learning.

But there’s one final reason I find TikTok such a fascinating and anomalous case study. It has to do less with software and algorithms and more with something that the cultural determinist in me will never tire of studying: the network effects of creativity. That will be the subject of my third and final part of this series on TikTok.