As I’m writing this, it’s been about 3 months since Elon Musk took Twitter private for $44 billion and started shaping it to his liking. In that short amount of time, I don’t think anyone, outside of a handful of Elon sycophants, would say things have gone well: it’s done several uneven and poorly-planned layoffs; lost a bunch of users upon the takeover and even more after burning bridges with its developer community; lost advertisers; have seen its parts of its services slowly-but-surely degrade; and stopped paying their office rent. Whatever short-term cuts have been made to try to contain costs have been more than offset by the destruction in revenue and enterprise value—some of the financiers of the privatization have already substantially marked down their investments.
I’ve been a Twitter user for over a decade, though not in heavy usage or follower count. Like many others, I hedged my short-form social media activities by signing up for Mastodon, and have been trying to use both to organically evolve my usage of the platforms. In just a month, I’ve seen Mastodon pick up steam; whereas posting activity was initially 90:10 in favor of Twitter, it’s now something like 60:401. On Twitter’s side of the world, my chronological feed has been taken over by automated media outlet promos2, celebrities, and politics.
Software engineers are familiar with the reference in the title. For well over a decade, it’s been en vogue for software architectures to transition from a singular service—a monolith—into a constellation of purposefully smaller and limited systems—microservices. It’s a direct reflection of a core computer science principle: the separation of concerns, whereby each service has clear responsibilities and boundaries and can be built, run, deployed, and maintained as its entity3.
Those who embark on this task know it is a long and arduous journey. µservices have to be planned out in advance: each one is spun up to replicate a limited, targeted subset of functionality, which upon reaching stability allows for that part of the monolith to be cleaved out of the codebase and carefully deprecated. Repeat this process a bunch of times, and you end up with a collection of such µservices…along with a thinner monolith powering everything that couldn’t be separated. It ends up being quite difficult to shrink a monolith down to 0, and not worth it for most engineering organizations.
Twitter’s slow and steady fragmentation is reminiscent of the same processes and is perhaps fated for the same end state. Twitter is a singular app owned by a singular company, but its utility comes from capturing its users’ interests, where users have self-organized into groups. Semi-officially, there’s: NBA Twitter; Black Twitter; Tech Twitter; Politics Twitter; Car Twitter—and many other groups defined through common interests. As with most things Twitter, most of these groups arose organically, well before there was any official platform support, making do with basic features like hashtags and user follows to establish communities.
The groups are now fragmenting away onto other social media platforms. Techies have migrated over to Mastodon en masse, while gamers have gathered on a newer app called Hive. Discord has always been an alternative as well as private group chats, and there’s Substack and Tumblr and Reddit which have already cultivated communities and would easily absorb Twitter refugees. There’s a rough analog to µservices, in the sense that each app is a bit more focused on its niche, without trying to be everything to everyone.
There’s also less drama on these other platforms. To be fair, there’s a newcomer effect that keeps users initially civil, and the new owner of the Twitterverse relishes in picking fights and ginning up controversy ex nihilo. But one of the spectacles central to Twitter—dunking on someone for a bad take—has yet to infect migrant destinations. Twitter is still the destination for Internet drama, and that may be why topics like politics and celebrities are some of the more prominent remainders of the Twitter monolith: they thrive on drama and emotion, which is particularly acute when expressed in 280 characters.
The other factor to consider, then, is whether the drama can sustain the service as the user count continues to dwindle over time, as many predicted when the takeover originally took place. Elon seems to be following the same playbook as conservative social networks—loosening moderation in the name of free speech—but ending up attracting trolls and the worst of humanity. In a similar vein, as Twitter is alienating itself from its formerly diverse user base, there are just fewer fights to be had, not to mention the now smaller audience to bear witness. Anecdotally, the parody account @bestofdyingtwit started strong in the Elon-Twitter era but has run out of steam as there’s just less activity to poke fun at—and fewer people to appreciate the schadenfreude.
Though I’ve been more active in curating connections with the latter social network.↩
The service recently announced they’re charging for their API, so that’ll likely go away too.↩
Critics have questioned whether the added complexity and footprint is worth these benefits in the first place.↩
