Elon Musk didn't ruin Twitter
Who doesn't like a Stoke Newington drinks party?
The past two weeks has seen the largest exodus from Twitter since Ello back in 2014, with many people moving to Bluesky. I haven’t actively been using the ex-bird-app since December 2022, but I went back only to see one person who likes spending eight dollars a month on virtual pixels arguing that people are leaving because they could not enact or sustain cancel culture on the platform. I closed the app.
No, they are not. They are leaving because Elon Musk exacerbated Twitter’s already existing problems that were caused when the platform realised back in, roughly, 2016, that bottomless VC money would, at some point dry up. This was peak-ZIRP, remember, and it was tough to see how old Twitter could monetise its users. With a virtually unlimited API, heavy users would not use the default app, so Twitter had little control over the user experience, and, crucially, how it could be monetised - i.e. deliver ads. The demise of app.net - a story for another day - taught them that putting it behind a subscription would likely not work, and remember, this was 2016, subscriptions were just beginning to become ubiquitous, so why scare off your users with a pay-wall?
So, Twitter began to limit API calls and the number of users a 3rd party client could have. Tweetbot, Tweetdeck, Echofon, and many other clients were suddenly curtailed. Features disappeared, and new users were limited from using the app. Some of these ended up being acquired by Twitter. It was not the case that the default Twitter app was bad - but it was worse than all the others out there. The walls around the third-party apps continued to grow, so new users would just inevitably gravitate to using the default app.
The default app at this point still had a chronological timeline. You could simply open the app, continue where you left off, and ‘catch up’ on what people were discussing. This was the optimal use case - every tweet would get, roughly, the same amount of exposure in a timeline, and it was the right way to use the platform and follow discussions. This is why the ‘digital town square’ idea was appealing. Admittedly, there was the ‘discover’ panel but this was pre-for-you-pages, and realistically, if you cared about topics that you wanted to hear the latest and greatest topics about, you could just click on trending topics, or search by hashtags and see the most popular tweets. But people would be engaged with the platform as long as there was still thins to catch up on.
The chronological timeline also made sense for those who live-tweeted events, like sports games, Apple keynotes, Eurovision, Game of Thrones - you get the idea. You could keep half an eye on what was going on, and you could immediately keep your finger on the pulse of what people thought. Slate finds this convincing too.
In 2016, Twitter tried to make the algorithmic timeline default before facing backlash and restoring the chronological one.
“Twitter helps you see what’s happening by showing the best Tweets for you based on your interactions", they wrote after restoring it. But this built on a fundamental misunderstanding of how the heavy users approached Twitter. Sure, it worked as a news aggregator, but the timeline was a way to see what other people thought about ongoing events. Hot takes would be part of the platform, and sure, they would garner attention, but they would generate discourse rather than just empty re-tweets and hundreds of likes (favourites).
Some people began gaming the system, and nobody was better than this than @realDonaldTrump, who gained more and more attention during the 2016 primaries by posting incendiary content. More and more of Twitter was just about him - his latest tweets would get retweets and quotes with commentaries, corrections, but it was always one-way and reactionary rather than conversational, which characterised the pre-2016 Twitter.
This was when the platform started becoming worse and worse - instead of using it for conversations, people sought more and more attention. My argument is not that Trump ruined Twitter, but catalysed an already ongoing process that had begun with the increasing importance of engagement. This feeds into the biological pleasure of being pinged a like, a retweet, or seeing response triggering dopamine release. The reason why the past two week’s exodus is unprecedented is because the biggest posters are simply addicted to the pleasure of knowing that hundreds of thousands, if not millions of people, are seeing their content, and re-building the same following elsewhere is a tough ask.
Then, after the Musk acquisition, Twitter made the algorithmic timeline default, which changed the incentive structure around posting. Instead of being part of the conversation - users seeking popularity would try to dominate it by posting, often, very basic and flawed analysis, exacerbating hot take culture further. Nuance fell further by the wayside and those offering insight or corrections could often find themselves simply unseen.
The platform flooded with porn, attention bait and only polarised further. A friend who had already described it as ‘bad’ a year ago told me that it got even worse since the recent US election. There’s a new working paper out that shows a “structural engagement shift” on the 13th of July, 2024, when Musk endorsed the republican candidate. After the endorsement, Musk’s posts begun to see a ‘significant additional boost’ in the frequency of their presentation. After this date, "Republican-leaning accounts [started] exhibiting a post-change increase [in view counts] relative to Democrat-leaning accounts”.
This finding suggests a possible recommendation bias favouring Republican content in terms of visibility, potentially via recommendation mechanisms such as the "For You" feed. Conversely, retweet and favourite counts did not display the same group-specific boost, indicating a more balanced distribution of engagement across political alignments.
Bluesky, since its inception, has felt more like 2013-2016 era Twitter than anything else - including a strong left-leaning poster base. Which is fine. As one skeeter puts it:
More fundamentally, the timeline is still chronological. I still see the engagement numbers a post generates but the platform does not decide for me that they are somehow more important than others. There is even a Quiet Posters feed which I love reading because many of the accounts I follow do not skeet often, making their contributions even more valuable.
Back in September, there was a really good article on the Financial Times that argued that “nobody cares about being in an echo chamber on Bluesky” which caused a lot of discourse on the platform. But I don’t use social media to constantly be challenged and I don’t always want to think about politics. If I do to think critically, I want to choose to engage with that kind of content, rather than it being shovelled in my face.
I’m not sure I have ever felt more like I’m at a Stoke Newington drinks party than when I’m browsing Bluesky (including when tucking into Perelló olives and truffle-flavoured Torres crisps in actual N16).
I loved this quote because that is exactly the kind of environment I am after. I want witty, middle-to-upper class discourse with people I know I will likely agree with on certain basic values. I don’t want to see justifications for restricting women’s health, or videos of war, or the latest thing this podcaster said that I should care about for whatever reason. I have yet to fully formalise my view on why I go on textual social media that has very few of my friends.
But I want to close the app feeling like I came across interesting, thought-provoking - but not anger-provoking content, just like I’d leave a N16 drinks party after a good number of interesting conversations. I have a friend who hosts the best parties because of this - he knows who he is inviting, and he can safely assume that his friends will get along well with each other.
I want this experience on social media rather than doom-scrolling because an algorithm knows what kind of content to serve me to keep me engaged and enraged. Bluesky is also better for credibility, as Jason Barnard, CEO of Kalicube told Sky News. Barnard spent nine years collecting three billion data points to understand what Google decides is factual information.
"Bluesky is 20 times smaller in terms of the number of people on the platform," he says. "If you search for people [on Google], you will find Bluesky 10 times less often than you will find X.
"But," he says, "it's 10 times more important to Google today for factual information."
So, for now, it remains a promising platform with a skyrocketing number of users. But the monetisation question is coming, and, with all of us having seen what Twitter grew into, many will be happy to pay a small monthly fee to keep Bluesky functioning as it is today.
And hey, if you are joining, welcome to the perennial Stokey drinks party. Leave your shacket by the front door and please don’t vape indoors.
On a serious note - if you’re not finding enough people to follow, look up Starter Packs!
Things I’ve enjoyed recently
My intention was to begin writing these posts weekly, but that was before my personal laptop’s display was shattered by a falling Apple charger in August. Fortunately, I have a new laptop on the way, so I am aiming to write more often.
With that, links.
A working paper from NBER that shows a causal relationship between looks, the number of friends, and gaming. (Does anyone want my old gaming PC?)
China is selling record highs of EVs.
A collection of insults from Marx and Engels.
An article from The Economist arguing that energy transition will be cheaper than we think.
Austrian Airlines’s X-odus.
A thread on the worst examples of data visualisation.
Treat yourself to Christmas coffee from DAK. It’s really good.
Some good takes on the election results:
Finally, Žižek.