Stage One: Good to Users

But an API can also be informal and improvisational. Hashtags are a kind of API: a developer who scrapes Twitter can use hashtags to figure out what each post is about. APIs can also be added by third parties: a developer who scrapes Twitter and categorizes posts by hashtag can then set up a more formal way for others to pull those out of their own database, without doing all that messy scraping.

At its outset, Twitter was, in fact, an API. The core of Twitter was a database of users’ posts that users themselves had no way to access. In order to post or read tweets, users had to make use of a program that used the Twitter API to access the database, pulling out the relevant entries and presenting them in a human-readable form.

Twitter made one of those API-accessing programs, but it also allowed anyone else to make such programs, and they all accessed the same API. It’s hard to overstate how revolutionary this was: normally companies give themselves privileged access to their own infrastructure. The API they expose to third parties is a weak, thin version of the private tool.

But not Twitter: it gave first-class access to all comers, and developers threw their resources into the project of building all kinds of ways of accessing Twitter. Some of these may be familiar
to you, like TweetDeck, which Twitter eventually acquired and brought in-house. Others were strictly “programmatic” — tools that made it easier for other developers to do cool things with Twitter, like operate bots that answered queries, told jokes, or automated public safety announcements.

Twitter also paid close attention to its users. Users invented “retweeting” by typing “RT” and copying and pasting someone else’s tweet into the composition box. Twitter noticed this and automated the process, creating a one-click system for retweeting anything on the system.

Many of Twitter’s core features were developed in this way, including “quote tweeting”: typing some commentary on a tweet, then pasting in its link and the #QT tag. (Twitter now automates
this process as well.)

Developers call this “paving the desire paths.” That’s a reference to a design principle from physical spaces like parks and campuses. Landscape architects look for places where people have worn the grass thin by cutting across lawns and fields (desire paths) and formalize them by leveling them and paving them or putting down gravel, wood chips, or some other material.

Taken together, these two policies — first-class access to the Twitter API and integration of users’ own innovations — constitute a system of generous value sharing with both business customers (who could create a variety of tools for themselves and for users) and users (whose workarounds were observed and turned into official features).

People loved using Twitter. It was playful. It was fun. It was a party that the whole world was invited to.

Stage Two: Good to Business Customers

But from the start, Twitter also made compromises. When Twitter added advertisements to its service in 2008, it made the decision to open local sales offices in countries around the world, including nations like Turkey, where governments could be relied upon to make censorship and surveillance demands.

This was a way of shifting surplus from users to business customers. Turkish advertisers didn’t need to do business with a Turkish Twitter office, but opening offices in-country made that more convenient for Twitter’s business customers. However, it also put Twitter employees and bank accounts within reach of Turkey’s authoritarian government, which used that reach to compel Twitter to do things that were harmful to its users, like revealing information about the identities of dissidents and removing their speech. (Despite this, Twitter continued to operate at a loss.)

Twitter also engaged in other forms of stage-two enshittification, notably in the staffing of its content moderation division. Content moderation on giant platforms like Twitter is always going to be a difficult proposition, but the more a company spends on moderators, the more moderation it can do. As Twitter’s user base and volume of posts grew, the company did enlarge its moderation team, but not at a rate to match its overall growth. That meant that the ratio of moderators to activity worsened over time.

Of course, all of this was characteristic of Twitter 1.0, the private company that was later taken public through an initial public offering (IPO) and governed by a shareholder-voted board of directors. This arrangement was far from ideal, but compared to what happened next, it was practically Eden.

In 2022, Elon Musk assumed ownership of Twitter. Musk had to borrow $22.4 billion to fund the acquisition. That vast debt exerts enormous pressure on Musk to extract money from Twitter. Remember, shareholders prefer to get paid by the companies they invest in, but their ability to compel the companies they invest in to pay them is limited to voting for company directors who’ll appoint CEOs who’ll agree to give the company’s money to its investors, rather than spending it on product maintenance and development, wages for staff, improved infrastructure, or executive bonuses. But creditors who hold a company’s debt are entitled to regular payments on that debt, and if the company stiffs them, they can ask a court to force the company to cough up, and if the company doesn’t have enough money and can’t borrow or raise it, they can force the company into bankruptcy.

By taking on tens of billions of dollars in debt, Musk was setting the company up for a world of hurt. Twitter (or, for some reason, X) will need to come up with large sums of money every year to service its debts, or its creditors can kill the company.  (Of course, if they do that, they will wipe out any chance of getting paid back. Forcing the company into bankruptcy would likely mean a fire sale of Twitter to someone else, with a share of the proceeds going to the creditors. By contrast, if they let Musk stiff
them, they at least have the hope of getting paid in the future if he turns the company around — or if he makes it structurally important to a future federal government, as Musk was actively trying to do with the second Trump administration as of early 2025.)

There’s not much to be gained by trying to read the minds of tech CEOs to determine which of their espoused views are sincerely held and what they’re merely saying to win points with some group of users, customers, lawmakers, investors, or peers.

And honestly, it doesn’t matter whether Musk’s strident pronouncements about gender politics, “wokeness,” and other subjects are his true feelings, transient blurts, or acts of calculated image-craſting.

Stage Three: A Giant Pile of Shit

Rather than playing Twitter Kremlinology, let’s instead look at how Musk’s handling of Twitter post-acquisition is an example of how an enshittification speedrun can boomerang on the enshittifier — and how those bad choices can nevertheless inflict serious harms on users.

Musk’s tenure at Twitter’s helm is best understood as a rapid, indiscriminate, and clumsy series of value transfers from end users to Twitter (which is to say Musk, his investors, and his creditors).

At the outset, Musk shed the vast majority of Twitter’s content moderation team. These workers were charged with maintaining an environment that was both hospitable to users and “brand safe.”

This is an absurdly difficult balancing act. Life is not brand safe, and many of the least brand-safe parts of our lives matter the most to us. When platforms seek to ensure that their advertisers’ materials only run alongside upliſting user content, they have to suppress or block users’ posts about their sex lives, their political fears, or the disasters they’re in the midst of.

While users don’t necessarily want the platforms to block their own frustrated or angry posts, they also don’t want to be abused, dogpiled, or doxed. They don’t want their feeds filled with unsolicited gore, sexually explicit content, or extremist hatred.

So moderators have to figure out how to trade off the interests of advertisers against the sensitivities of users. On top of that, moderators are charged with weeding out bad ads — frauds and scams, paid disinformation, and ads for illegal products — as well as weeding out bad nonadvertising content like spam, scams, and unpaid (but coordinated) disinformation.

No platform does this well. Pre-Musk, Twitter was somewhere in the middle of the pack, having made its share of high-profile gaffes and missteps but also routinely blocking millions of posts of the sort that it intended to block, consistent with its own decisions about the interests of users and advertisers.

Getting rid of the moderators made Twitter instantly, permanently, significantly worse for both users and advertisers. Users were — and are — inundated with ads for scams, counterfeits, and hoaxes. Advertisers find their messages attached to posts featuring gore, Holocaust denial, and pornography.

Everybody loses. Including Twitter. This is where things get weird. For all that Facebook has strip-mined its service, sucking out the value that its business customers and end users created, it did so gradually. Facebook remained (very) profitable, even through its missteps. But under Musk, Twitter speedran the enshittification curve, and it removed value so quickly that it sparked a mass exodus of advertisers and a collapse in revenue that more than offset any savings from firing the workers who maintained quality under the old guard.

Again, I’m not going to try to peer into Musk’s state of mind here. He frequently claimed that his actions were in service to his appreciation of free speech, and at one point publicly swore at a roomful of executives of companies that advertised with Twitter who’d gathered for a New York Times summit, telling them that he didn’t want their money if it was contingent on a well-moderated Twitter. In the profane onstage rant, Musk personally singled out Disney CEO Bob Iger, telling him “Fuck you” and
“Go fuck yourself.”

This is, at the very least, evidence that the subject of free expression rouses strong views in Musk. Nevertheless, his record on free speech and moderation has been poor. He’s booted journalists who criticize him, caved to censorship requests from oppressive governments with poor human rights records, and pursued a vendetta against an account that published public records about the movement of Musk’s private jet. In the first days of the second Trump administration, he suspended users who identified government contractors whom Musk hired on behalf of the so-called Department of Government Efficiency (DOGE) and gave unsupervised access to sensitive government payment systems.

One of Musk’s first official acts was to sell “verification” — that is, blue tick marks — to all comers. These blue ticks were initially created by (old) Twitter as a way of helping users spot fake accounts. Twitter users who had a high public profile and were worried about being impersonated on the platform could apply for a blue tick, which was awarded aſter a dedicated Twitter team took some steps to verify the account’s authenticity.

This was a far-from-perfect process. Twitter’s standards for notability were opaque and unevenly applied. Behind the scenes, Twitter’s verification process was a sloppy mess.

In 2011, I myself was impersonated on Twitter by a user who created a fake account to attack others as me and to post false information about me. I wrote to the company and was told that it would take action against the impersonator only if I applied for verification, and that doing so entailed faxing the company a
copy of my driver’s license, because “email isn’t secure.” When I told the people at Twitter that I couldn’t send them a fax because my time machine was broken, they had a ready answer: yes, many users don’t have fax access, but here’s a free email-to-fax service run by some parties unknown. Just email your driver’s license scan to this address, and they’ll fax it to us.

I was living in Europe at the time, so I wrote to a friend at Twitter and pointed out that this was radioactively illegal under the European Union’s General Data Protection Regulation. A few weeks later, Twitter changed its policy and let me send in a scan of my driver’s license by email. Shortly thereaſter, I got my blue tick, and the impersonation account was deleted.

But under Musk, this verification process was swept away entirely. Users could receive a blue tick by paying a monthly fee, which rose steeply for businesses such as publishers. Aſter a grace period, all the “legacy” blue ticks were deleted.

Again, this was a measure that clawed away value from business customers and end users. Business customers had to pay a monthly fee to fight impersonation. End users found themselves deceived by pranksters and fraudsters who bought blue ticks to lend verisimilitude to their deception.

Musk continued to turn the screws. Users who didn’t pay for blue ticks had their posts suppressed on the platform — meaning that their posts were less likely to be seen by other users who followed them, or to be recommended to strangers — while users who paid were prioritized and had their posts shoved into other users’ feeds across the platform. This was a godsend for fraudsters, griefers, and other bad actors, who could put shock images or fraudulent solicitations at the top of millions of strangers’ feeds.

But for legitimate business customers, having to pay a fee to reach the people who’d asked to hear from them was a double insult. On the one hand, Musk had been vocally contemptuous of the press, which muddied the waters as to whether the fees he demanded were a form of ritual humiliation or merely a way to ransom publishers’ subscribers to them. On the other hand, the proliferation of fraud and trolling among blue ticks meant that publishers who bought a blue tick actually made themselves look less reliable. (Musk eventually “solved” this by adding a feature that let users disguise the fact that they were paying him, which speaks volumes about the message that his new blue ticks sent.)

Users suffered as well: the quantum of content in their feed that they had asked to see, or that the algorithm predicted they would enjoy, dwindled to a bare minimum. This left a void that could be filled with ads and artificially boosted posts from blue ticks (which, again, were likely to be ads, frauds, gore, or porn).

Musk’s leadership since has continued in this vein. While the public attention focuses on the flashy changes, such as renaming the company the letter X, the most consequential changes have to do with making things worse for users, while keeping users locked in to the platform. (Aſter all, if users are still locked in to the platform, they’ll keep at least some advertisers stuck to it as well.)

For example, in 2022, Musk suspended some high-profile Twitter users and stated that this reflected a new policy prohibiting users from listing their accounts on rival platforms (like Mastodon, Bluesky, and Threads) in their bios, usernames, or  posts.

This step coincided with measures that blocked many business customers from using Twitter’s API and made API usage far more expensive for those who were still able to use it. Remember, Twitter started out as an API-first company, designed to be built upon, tinkered with, improved, and extended by a constellation of complementary companies, tinkerers, and users.

Together, this ban on publishing off-Twitter “forwarding addresses,” coupled with severe API restrictions, killed off the burgeoning ecosystem of automated tools to help users migrate off Twitter. These tools (such as Twitodon and Fedifinder) requested your Twitter login, then used the API to pull down a list of all the accounts you followed and added them to the people you followed on Mastodon.

If all of that sounds confusingly technical, let me describe the process from your point of view. Say you get tired of Twitter, so you set up an account on Mastodon, a Twitter-like service that is built on open-source soſtware and has lots of different servers you can sign up with, run by individuals, companies, co-ops, and nonprofits.

So now you have a Mastodon account and a Twitter account. You log in to a migration tool and tell it how to find your old Twitter account and your new Mastodon account, and click “Go.” A few minutes later, your Mastodon account has been updated to follow everyone you know on Twitter who has also set up a Mastodon account.

When Musk targeted this practice, he made it clear that he viewed his path to profitability as depending on making it hard for users to leave.

In the years since, Musk has introduced a series of “antifeatures” that give users and/or business customers plenty of reasons to quit. For example, in 2023, Musk abruptly altered the way that link previews worked; tweets that contained web links would automatically feature an image but no header or text snippet from the linked page.

Headers and snippets helped users make better estimates of whether they wanted to follow a link before clicking it. But they were even more important for publishers, a key constituency of Twitter’s business customers, whom Musk hoped he could coerce into paying high monthly “validation” fees. By stripping headers out of link previews, Musk drastically reduced the likelihood that users would click on the links, resulting in a near-total collapse of readership coming from Twitter posts. Publishers that had spent years cultivating a large Twitter subscriber base in order to drive traffic to their sites took a huge monetary loss from this decision.

Around the same time, Twitter also rolled out extremely long posts, with extensive formatting options. Musk then told publishers who were worried about collapsing Twitter traffic that they should reproduce the contents of their websites in very long tweets. Of course, this would mean eschewing the subscription and advertising revenue publishers got on their own sites. (Musk offered them a share of the revenue from the content on the site, contingent on their paying for validation and calculated using a complex and opaque formula.)

Again, this is the same tactic that Facebook tried, but Facebook did it by degrees . . . and did it first. When Musk tried it, publishers were already familiar with this snare, and they declined to stick their feet in it.

Eventually Musk had to back off: later in 2023, he announced that he would add blue ticks to “prominent” accounts that had enough followers. These were typically accounts that had been “verified” under the old Twitter management. A very small number of these verified users had chosen to pay to keep the status under Musk, and, aſter months of moribund sales, Musk suddenly and nonconsensually gave these users a tick mark that had come to mean “I am someone who tolerates — or even likes — Musk’s unhinged management of Twitter and/or his personal views on race, immigration, gender, and workers’ rights. I am voluntarily giving money to the Hitler-salute guy, every month.”

This was yet another impetuous, irrational policy U-turn for Musk. His initial gambit had failed: he didn’t get hundreds of thousands of internationally notable people to pay for blue ticks, thus creating a penumbra of desirability around his flagship product that would lure in millions of everyday people to splurge on blue ticks of their own.

Without the cohort of notable and verified pre-Musk blue-tick users, the blue ticks became associated with spam, porn, trolling, and fraud, which made them especially unattractive for everyday users and the influencers who Musk hoped would bring in business.

Musk’s restoration of (unpaid) blue ticks to the original cohort now takes on a desperate air — like he’s hoping that he can dilute the cesspool that most people associate with blue ticks until they become a credible product. At the time of this writing, this gambit is failing, and I am reasonably certain that it will continue to be failing by the time you read this, notwithstanding Musk’s increasingly alarming early 2025 role as shadow president of the United States.

But there is one respect in which Twitter is thriving: it complements Musk’s financial influence by giving him a huge megaphone. In combination, Musk’s money and his platform have allowed him to establish himself as a kingmaker, “first bro” to Trump, and key political ally for far-right figures around the world, whose ideology universally involves wrecking things, safe in the knowledge that the people who care about these things can’t escape. I could go on for pages more about the various enshittificatory gambits that Musk has assayed on Twitter, but that would be beside the point.

I don’t recite Musk’s fumbles to make the point that he is stupid and incompetent (though, for the record, I think he is both very stupid and very incompetent). Rather, this is all to build up to the following: people still use Twitter.

Hundreds of millions of people are wading through the enshittification, which rises every day, and continuing to use the service. All kinds of people continue to use it: the marginalized groups that have endured racist, gendered, homophobic, and transphobic hate campaigns are still there. So are the journalists who Musk denigrates with every opportunity, and whose work he has gone to great lengths to devalue. So too are the performers who espouse progressive values antithetical to those that Musk promotes on the platform. Even people on Mastodon or Bluesky are usually maintaining their Twitter accounts. Same for the millions of users who were bootstrapped onto Meta’s Twitter clone, Threads. People who deplore Musk’s politics, his reckless and unlawful seizure of entire US agencies, his fire hose of lies, and his support for neo-Nazi parties all over the world are still using Twitter.

As of this writing, I’m still on Twitter. Why are we still there? Switching costs. Collective action problems. Think of the parallels to the offline world: Why do marginalized groups stay in regions where they are openly despised and subject to harassment and discrimination? Because if you have to live with continuous harassment and discrimination, you absolutely rely on your community for your sanity.

The only thing worse than being a member of an oppressed minority is being an isolated member of an oppressed minority. A community — or, for a creative worker, an audience — is a vital lifeline, but it’s also an anchor. It’s impossible to overstate how hard it is to coordinate an exodus of people, even people who love and rely on one another, even when things are terrible.

I call this “the Fiddler on the Roof problem.” In the musical, we meet a group of Ukrainian Jews living in a shtetl called Anatevka. Anatevka isn’t a very nice place to live. It’s poor, it’s primitive, and, of course, it’s subject to the czar’s Cossacks, who ride through the frame every fiſteen minutes or so and kick six kinds of shit out of the Anatevkans.

So why do the Anatevkans stay put? We learn the answer to that in the melancholy final scene. The czar has finally pulled the plug on Anatevka by ordering a purge of all the Jews. As the villagers make ready for their departures, they have one final farewell:

LAZAR: Tevye! Tevye, I’m on my way.

TEVYE: Where are you going?

LAZAR: Chicago, in America.

TEVYE: Chicago, America? We are going to New York, America.

LAZAR: We’ ll be neighbors. My wife, Fruma Sarah, may she rest in peace, has a brother there.

TEVYE: That’s nice.

LAZAR: I hate him, but a relative is a relative.

It’s quite a tearjerker! We have spent the past three hours coming to understand just how much these people rely on one another to get through their brutal, impoverished daily lives. Now we must reckon with the fact that they are going on to a new stage of their lives — one that will be every bit as brutal and impoverished, with the difference that they won’t have one another to help.

That’s why people are still on Twitter. It’s not that they like the service — it’s that they like one another. And leaving one another is especially hard in moments when things are especially terrible — say, when Elon Musk and Donald Trump are dismantling whole swaths of the US government in a blatantly undemocratic way. Those moments of existential terror are exactly when you need your community the most.

Enshittification — deliberately worsening a service — is only possible when people value that service to begin with. Enshittification is a game of seeking an equilibrium between how much people like the thing that locks them to the service (oſten that’s other people) and how much they hate the management of that service.

Twitter is a cautionary tale. It tells us that the “market forces” that we’d expect to kill off services that turn into piles of shit have been neutralized. We are living in an age of zombie platforms: platforms that shamble on long aſter they should have been double-tapped and stuffed in a shallow grave.

The force that animates those zombies is desperation: not the desperation of the platforms’ owners; rather, the desperation of the platforms’ users and business customers, who can’t live without one another and don’t know how to leave without losing one another.