The identities of both the fediverse and the atmosphere have been strongly influenced by Elon Musk’s acquisition of Twitter. Bluesky’s growth trajectory, Mastodon’s cultural identity, and the entire discourse and self-understanding of making social platforms resistant to billionaire purchase are all downstream of that transfer of ownership of Twitter. Three years after Twitter changed hands, we now have a second case study, with TikTok’s transition from the Chinese Bytedance to a majority American ownership last week. The culture of ActivityPub and ATProto are shaped by what it means to watch a dominant Big Tech platform become owned by a fascist oligarch, and the question now is whether TikTok will leave a similar mark. Early signals of the impact of the transfer of ownership is how users claim that TikTok is now blocking mentions of Epstein and ICE, or Emmy-winning Palestinian journalist Bisan Owda reporting a permanent ban from her 1.4 million follower account.
The standard story about why X is hard to leave centers on network effects, the idea that the network is valuable because everyone else is there. This isn’t exactly wrong, but misses a nuance in what makes a platform sticky. People don’t make individual choices based on the user count of the entire network, they make choices based on their perception of the network effects. People perceive which accounts matter to them, and are making their platform choice based on that. Crucially, this perception is shaped by the platform’s architecture.
Twitter and X’s chronological timeline (and even to an extent their algorithmic feed, more so for Twitter than for X) make the network legible. You can see who of your follows is actually there, and who is posting. This also makes it visible when prominent accounts go quiet or announce their departure. This legibility is what made coordinated migration possible: you could see others leaving, which gave you permission to leave.
Beyond this legibility, Twitter had also accumulated a particular political significance, it had created the common knowledge that it was the place where public discourse happened, where news broke, and where politicians and journalists gathered. This common knowledge was never fully accurate, but accurate enough to make Twitter feel like the agreed-upon stage for public political life.
Together, this created a unique set of circumstances for the case of staying or leaving Twitter: leaving means abandoning the agreed-upon place for public political life. Moreso, it was a visible act that others could see you do. This move of people towards Mastodon and Bluesky was as much about joining a new platform as it was about making a collective political starement.
TikTok’s structure differs from X in two meaningful ways. The first is that its TikTok’s architecture makes the network much more invisible to its own users. The For You Page algorithm is not strongly influenced by the accounts you follow, instead it responds instantly to the videos you most recently watched and liked. This makes it very hard for people to build a mental model of what their network on TikTok actually is. TikTok’s algorithm selects from a pool so vast that individual departures are virtually undetectable by people. You cannot see who has left, because it is very hard to build a stable sense of who was there in the first place. The ban of Bisan Owda got the attention of Al Jazeera, but not every ban will get that attention. Most followers might never notice her absence, with the algorithm simply serving them a never ending supply of other videos.
This creates a coordination problem for migrating to other platform. While Twitter and X were not great for coordinating collective actions to join other platforms like Mastodon and Bluesky, it was in fact possible, as both platforms did experience such spikes in people joining. This pattern has died down in the last year for people going from X to either Bluesky or Mastodon, suggesting that collective actions to join other platforms are both possible but not guaranteed to happen and can be disrupted.
The second way that a migration away from TikTok might look different than X, is that where Twitter and X are the centers of political discourse, TikTok is the center of culture. When journalists and politicians left Twitter, they were making a statement about the public political sphere. When creators leave TikTok, they are making a career decision about where their audiences are. These actions have different motivations, with different symbolism, and will likely not create the type of collective identity that has developed on Bluesky and Mastodon as a result of people leaving Twitter and X. The fediverse and Bluesky both carry the imprint of being “the place people went when they refused to stay on Musk’s platform.” This does not mean that TikTok users won’t move to differrent platforms. It means that this move, if it happens, likely won’t follow the same “leaving” narrative that has shaped open social’s current identity.
There’s a lesson here from what actually happened with the moves away from Twitter. In raw numbers, Meta won, as Threads has now overtaken X in daily active users, and has around 140M daily active users on mobiel, compared to Bluesky’s 3 million total daily active users. Bluesky, although much smaller than both other platforms, does hold outsized political impact already, especially for politicians on the left side of the political spectrum. While Threads did win the ‘number-go-up’ game, it’s political and cultural relevance is surprisingly low in Western countries.
Meta is well-positioned to repeat this numerical victory with TikTok, and Instagram Reels already competes directly for the same attention and the same creators. If TikTok users move away due to privacy concerns or content moderation frustrations, Reels is the lowest-friction alternative option, together with YouTube Shorts. But this tells us little about whether open social will benefit, because open social platforms select heavily for early adopters and the type of people who want to built the future of social platforms.
There’s a another structural difference worth noting, beyond the nature of the migration itself, namely that video is harder than text when it comes to running independent social networking platforms. Text-based social media is relatively cheap, and accessible for self-hosting. You can run a Mastodon or GoToSocial instance on a cheap VPS, especially if the server is only for a few people, same with self-hosting a ATProto PDS. Video is in an entirely different category, with storage costs, bandwidth costs, transcoding costs, CDN costs that are both much larger than text, and scale superlinearly with usage. Then there is moderation, and where the moderation of text is already diffecult and expensive, video moderation either requires massive (and expensive) compute for automated systems, as well large-scale human labor for manual review.
Skylight, the ATProto-based TikTok alternative, crossed 380,000 users this week with around 95,000 monthly active users. The pattern of signups is arguably just as interesting as the numbers itself, with CEO Tori White says that compared to previous signup waves, this one is more sustained, with continuing elevated signups of around 4k new people per day, whereas previous waves had a more spike-and-crash pattern. What makes this possible for Skylight is ATProto’s infrastructure model. Skylight does not need to set up their own complete infrastructure stack, instead it uses Bluesky’s relay, CDN and AppView.
This model, where an app on ATProto can start out as a client, grow its own user base, and gradually built out infrastructure later is a genuine novel pattern that we’ve not really seen before. It does create dependencies however, where Skylight’s existence is contingent on Bluesky’s continued funding and moderation.
ActivityPub has it’s own TikTok-style video platform with Loops, which has a different path to viability. The fediverse model assumes instances run by independent operators, which are largely hobbyist volunteers. This works well enough for text-based platforms, where costs are manageable. For video at scale, the question of ‘who is going to pay for this’ becomes unavoidable however. Loops cannot bootstrap on shared infrastructure the way Skylight can, and each instance must bear its own costs from the start. This means for Loops to become a meaningful TikTok competitor, someone needs to commit serious money to operate a video platform at scale.
Twitter’s ownership transfer has become an integral part of the story for both the fediverse and the atmosphere. It established the concept that growth for these networks happens as a result of people ‘leaving’ Big Tech platforms, and that decentralised open protocols are a recognisable alternative. TikTok’s ownership transfer is unlikely to produce the same dynamic. The structural differences, namely the lower visibility of what your ‘network is’, the lower presence of political signals and the much higher cost of video infrastructure all suggest that the dynamics will be different. This does not mean that platform migrations won’t happen, but does mean that it will be driven by different forces, and produce different kinds of communities as well. Skylight’s sustained growth indicates that people are aware of the issue and looking for alternatives. TikTok will likely not produce another Twitter effect for the open social networks for structural reasons, and video on open social networks might have to grow without one.
