Is Bluesky really decentralized? It's complicated
A couple of weeks ago, I wrote at The Torment Nexus about whether Bluesky could become the new Twitter, and whether that would be a good thing or not. Since then, the network has just continued to ramp up its growth — it now has more than 23 million members, up from 15 million when I wrote that first piece — and so I wanted to go a little bit deeper and look under the hood at how Bluesky actually works, and how that compares not just to something like Twitter or Threads but also to other social networks such as Mastodon that are often referred to as "federated" or "decentralized." Before I do, I should note that I am not a programmer or social networking expert, and so it's entirely possible that I may describe some of this inaccurately or just plain get things wrong and for that I apologize in advance. But I think the differences in how they are perceived versus how they actually work are important.
A network like Twitter or Threads is relatively easy to understand. There's a company that owns everything (including the actual user accounts, as Elon Musk is arguing in a brief related to The Onion's acquisition of InfoWars) and it controls who gets to post, what they get to say, where the messages go, and so on. If Meta or Musk want to make the network either unusable or actively hostile, or nuke your account and everything you've ever said and all the contacts you've made, there is absolutely nothing you can do about it. This fits the definition of centralized pretty well. Yes, you can export your tweets etc., but it is difficult (but not impossible) to import them into some other network, and even if you do you lose any related content and connections.
Note: In case you are a first-time reader, or you forgot that you signed up for this newsletter, this is The Torment Nexus. You can find out more about me and this newsletter — and why I chose to call it that — in this post. This newsletter survives solely on your contributions, so please sign up for a paying subscription or visit my Patreon, which you can find here.
This structure makes it a relative no-brainer to use and understand, and I think that helps explain why Threads has more than 275 million accounts (according to Threads honcho Adam Mosseri, in November alone the network added the same number of accounts as Bluesky had in total). But having an account is one thing, and actually using it is another — you'll notice that I said Threads has 275 million accounts rather than users, and that's because, despite its size, the activity level on Threads seems to be significantly lower than on Bluesky. According to estimates from Similarweb, daily use on Bluesky hit 3.5 million recently, while on Threads it was just over 4 million, despite the fact that Threads has an order of magnitude more users than Bluesky does.
Decentralization is part of the appeal
It's worth noting that Bluesky's growth has come with some of the problems that scale brings: according to one estimate, the number of fake accounts imitating celebrities has been growing rapidly, to the point where almost a quarter of the accounts that represented a person with a name had at least one doppelgänger. There has also been an influx of unsavoury content, and so the company told Casey Newton, who writes the Platformer newsletter, that it plans to expand its moderation team from the current level of 25 to more than 100 (Bluesky has said in that past that it also plans to rely on users to do some of the moderating, although so far it has nothing like Twitter's crowd-powered Community Notes program, formerly known as Birdwatch). Chinese state media is also reportedly looking at Bluesky now that everyone is leaving Twitter.
Bluesky's rapid growth seems to have caught the attention of Meta, to the point where Threads is testing a number of features that it has borrowed from Bluesky, including the ability to make the "following" feed — the one that is just a stream of posts from people you follow in reverse chronological order — the main feed in Threads. This is a pretty big departure from the algorithm-first approach that Meta has taken in the past with Facebook, Instagram, and Threads. It is also adding custom feeds based on keywords, another feature people seem to like about Bluesky, and Mosseri said that Threads has tweaked the algorithm so that it will show more content from people that a user follows, and less of the algorithm-driven clickbait that some dislike. Whether this creates more of a Bluesky-like environment remains to be seen.
The relative ghost-town aspect of Threads as compared to Bluesky could be a result of the algorithm-driven nature of the Meta app. Or it could be a result of people setting up Threads accounts because Twitter is a garbage fire and they want to have an exit strategy if it finally implodes completely, but they don't actually use it. And there are probably lots of bots, as there are on Twitter. On Bluesky, however, I get the sense that (at least for now) a significant number of users are there because they want something different from Threads or Twitter. Many of the users I've encountered are attracted by the fact that Bluesky is not just run by what seem like decent people, but is also run — and in fact constructed — in a way that is inherently decentralized, so that it's more difficult or even impossible for someone like Musk to degrade it, and easier for someone to export their account if that does happen.
A protocol not a platform
This ability to make what Bluesky calls a "credible exit" is a key factor in how the service was built. As some people know already, the app was created by Jack Dorsey as a skunkworks project within Twitter, based on a seminal paper my friend Mike Masnick of Techdirt wrote in 2019, called Protocols Not Platforms (Dorsey is not involved any more, but Masnick is — he joined the Bluesky board in August). In a nutshell, the idea behind the paper was that it would be beneficial not just for users of social networks but for society as a whole if the structure of the social arena moved closer to the way the internet works, with open protocols like SMTP (the simple mail transfer protocol) for email, and HTTP (hypertext transfer protocol) for the web. As Masnick wrote:
Moving to a world where protocols and not proprietary platforms dominate would solve many issues currently facing the internet today. Rather than relying on a few giant platforms to police speech online, there could be widespread competition, in which anyone could design their own interfaces, filters, and additional services, allowing whichever ones work best to succeed, without having to resort to outright censorship for certain voices. It would allow end users to determine their own tolerances for different types of speech but make it much easier for most people to avoid the most problematic speech, without silencing anyone entirely or having the platforms themselves make the decisions about who is allowed to speak. In short, it would push the power and decision making out to the ends of the network, rather than keeping it centralized among a small group of very powerful companies.
Without going into too much detail, Bluesky is built on something called the AT protocol, which is short for Authenticated Transfer. It is an open standard, which means that anyone can build something with it, or build something that integrates with Bluesky — including custom feeds, custom moderation features, etc. In addition to the openness and customizability, Bluesky is built in such a way that it decouples some of the things that are central to a social network from each other, so that each one can (theoretically at least) be run independently of the others. So the thing that hosts or stores the posts or messages is separate from the thing that distributes them, and both of them are separate from the thing that moderates them. The AT protocol specs say that it was constructed so that "speech and reach are separate layers."
Cosplaying at decentralization?
Bluesky co-founder and CEO Jay Graber (whose first name is Lantian, which means "blue sky" in Mandarin) told Wired writer Steven Levy that the whole point of building the network was not to recreate Twitter but to reshape social media on the principle of openness and user control. To verify their identity, for example, users don't have to pay the social network's mentally unstable overlord for a blue checkmark icon, they can simply connect a domain name that they own. And one of the overarching principles, Graber said, is that the best way to keep users is to allow them to leave and take their network with them. As Levy put it, "if they’re fed up with Bluesky, they can pick up and go elsewhere, taking their contacts and network activity with them."
This all sounds wonderful. But is it true? In other words, if I wanted to use the AT protocol to build my own Bluesky, or to run my own server hosting my own posts and replies, could I do that? And could I leave and take not just my account, but also all of my posts and connections? This (according to some) is part of what “decentralization" implies. But network programmers say Bluesky doesn't really achieve that, at least not yet. One says that the service is "cosplaying at decentralization." Others, like Christine Lemmer-Webber — one of the programmers who helped create the ActivityPub protocol, the federated system behind Mastodon — are more diplomatic, but the point remains that true decentralization is something Bluesky wants to offer, but it realistically either isn't possible, or is prohibitively difficult and/or expensive.
For example, if you wanted to set up your own relay or duplicate the network itself to compete with Bluesky, you would need a server capable of storing and modifying terabytes of information in real time, which would cost an estimated $55,000 a year or more. You could run your own PDS (personal data store), which is the term for the server that hosts your messages and replies, and then connect it to the Bluesky distribution system. But if Bluesky goes away, will that have any meaning? (For more context on this, see the part below about the centralization of message-handling data). And of course many of these numbers are increasing as more people join Bluesky.
Note: I've updated this paragraph quite a bit from the original, after a number of people explained that it confused the cost of running a PDS with your own messages vs. mirroring the entire network.
Centralized in fundamental ways
As Lemmer-Webber notes (and a Bluesky engineer confirms) direct messages on Bluesky, which the service added in May, are a completely separate system that is controlled by Bluesky, and therefore couldn't be exported or handled by anyone else, at least not right now. In addition, open source programmer Michal Wozniak (the one who said Bluesky is cosplaying decentralization) noted that the reference points that refer to messages, which are called DIDs — short for "decentralized identifiers" — aren't actually decentralized at all. They are controlled by something Bluesky originally called "DID Placeholder," a centralized way of handling them until the company built something that was actually decentralized, but recently it said that it was sticking with the placeholder (renamed the "public ledger of credentials"). As Lemmer-Webber puts it:
Part of the concern I have with Bluesky presently is thus that people are gaining the impression that it's a decentralized system in ways that it is not. There are multiple ways this could end up being a problem for the decentralized world; one irritating way is that people might believe there's an "easy decentralized way to do things" that Bluesky has discovered which isn't actually that at all, and another is that Bluesky could collapse at some point and that people might walk away with the impression of "oh well, we tried decentralization and that didn't work... remember Bluesky?"
The most likely future, Lemmer-Webber says, is that there will always have to be a large corporation at the heart of Bluesky or the AT protocol, and the network will have to rely on that corporation to control things like identity, illegal content and spam. This may be a good enough for most users (many of whom likely don't know or care about decentralization or protocols, etc.) but based on the economics, it's likely to be a centralized system that relies on trusting a central authority. So not that different from Threads or Twitter in that respect. Decentralized in theory, but centralized in practice (I should point out that a "fediverse" service like Mastodon isn't really decentralized in some important ways either, including the fact that the admin of the server you sign up with can shut it down and take all your messages with it).
Lemmer-Webber notes that even if a user wishes to switch away from Bluesky, it "probably has effective permanent control over that user's identity destiny, removing the reassurance that one need not trust Bluesky as a corporation in the long term." I'm not arguing (and I don't think Lemmer-Webber is arguing), that Bluesky is not a great improvement over Twitter, or that the people running it aren't well-intentioned. But it's not accurate to say that it is decentralized in some important ways, and I think that matters. That's not to say it will be inherently bad, just that we should go into it with our eyes open.
Got any thoughts or comments? Feel free to either leave them here, or post them on Substack or on my website, or you can also reach me on Twitter, Threads, BlueSky or Mastodon. And thanks for being a reader.