I hope this isn't too loud for folks in the room. The speaker here is busted, so we can't. Okay, so we're 0 for two. Okay. Okay, I will try to be loud but while speaking into a mic, this feels very counterintuitive. Hi folks, uh it is nice to see every all your wonderful faces. I should start off that this is um the same presentation that I gave yesterday. So if anyone was here yesterday, I'm very flattered you should probably leave because you will just be here. Or hey, stay, I I love the sound of my own voice, maybe you'll love hear more of it too.
Um this is part of the at science conference, and so I toss in this little cheeky joke towards at social science because I am a social scientist. I'm a PhD student at the University of Pennsylvania where I can study communications uh specifically the political economy of online spaces. Um, we think about uh some of the benefits uh from the app protocol, there are kind of two and a half real main benefits um for social science researchers. Um the first one is data access. So there is this kind of long history of I I should say long, long within the history of internet research, um, concern over the access to public social media data.
Um, and over the last five to ten years, a lot of that has been constricted nowadays. If you're interested in understanding the kind of handful of dominant platforms, what you end up having to do is either in the case of Meta, use a really clunky system that like is really limited in terms of the data analysis you're able to do and the amount of uh data you're able to sample, uh, you have to uh ask them for permission in order to conduct any analysis. So if you want to do critical analysis of TikTok, you have to get TikTok's permission, which feels very counterintuitive, um, or they just charge you a bunch of money, like in the case of uh what was formerly known as Twitter, they charge you an absurd amount of money that most researchers cannot afford uh if you want to do things the right way.
So the first thing that's really nice about Bluesky is that okay, let's put a caveat of permission spaces, but currently all the data is public, which means as a social scientist, your ability to look at user behavior is really uh helped. Also, uh Sophie Greenwood was talking about this yesterday, but if you want to do experiments with different algorithms, you also have an ability in the generation of custom feeds to do some of that testing. What I'm interested in though is the ability to ask structural questions. And so there's a long history of uh media scholars looking at how does the ownership and concentration of media systems affect the quality of journalism, uh, the quality of conversation and reporting, um, but then also like what are the kind of broader implications for democratic theory and democracy.
And it's a lot of folks have thought about this generally in the context of the internet. Uh that is, I mean, arguably one of the reasons that Mike Masnik wrote that piece in the first place. But up until recently, we haven't had a lot of alternative uh sources to look at. We can go and look at the history of the 90s and the early 2000s to get a kind of indication of this, but for the most part, it's been really hard to ask these structural questions of what is a deconcentrated media system actually look like. And so it was in thinking of this.
I was looking at a paper uh called with a great name, stuck in the middleware with you, uh, where some social scientists looked at the current status of feeds, uh starter packs, moderation lists, and labels, and basically made the argument that most of these were run as gift economies. If anyone was here in the conversation earlier, um the the the PhD student from the University of Washington was making a very similar point, which is currently feeds are mostly volunteer work, right? Folks are putting these together because they care about community, they're paying the hosting costs. Now if you got a larger feed, um I was talking uh with Emily earlier who runs um uh the astronomy feed.
You know, there are some of these that are more sustainable, and sometimes donations can be a real uh viable pathway. But in that kind of long tail of feeds that are created, oftentimes it's you're volunteering or you just happen to care about something specific. Um also supply often outstrips demand, and it is that these do tend to be rather centralized groups where a handful of feeds get used by a lot large majority of folks. There's a long tail. So naturally this brought me to the question of my presentation today, which is how to centralize this Bluesky.
I mean, we talk about it in the context of being this really decentralized platform, um, but it's not always clear that that is the case. Um I want to say that I got completely nerve-sniped in this topic. I'm gonna be referencing um I mentioned his name yesterday, but Kuba uh has put together who is not sadly not here, has put together a really good dashboard that today I saw both Rudy and Paul Fraser, uh Paul Frazy uh reference in their own, so please go check that out. And then eight days before this conference, uh Lawrence came and uh wrote a blog post about exactly this concern, so I got completely completely nerd sniped.
But um let's talk about it. So there are kind of two ways you can measure decentralization. The first is you go to the relay and you say, hey, can you give me all your list of users? And then you or you give me the list of all the PDSs that you were querying through, and you go to those PDSs and you say, hey, can you give me your list of users? And you count that all up and you say, hey, this is how many people are in all these different PDSs. You combine that, you get your number.
The challenge is that, especially for smaller data servers, not there the they aren't often completing every part of the API, and so you're not gonna get that information fully, or they will have there will be a lot of uh PDSs where that information is just not as clear. So the other thing you can do if you just assume that most people are using uh DID uh PLC DIDs uh is you can just go through the list of exports and you can say, hey, uh what does this tell me about the current state of the network? Because then you will see every time a new identity was created and uh with what PDS it is associated.
So 99% of 99.7% of users have their information stored on Blue Sky's network. I should say this information is a little bit out of date. I would guess it's probably closer to like below 99 point something percent. But the point is that's a large number. Now I know what everyone in this room is saying is thinking. Wait, did he really give this talk yesterday? Why is he repeating himself? But the thing you're also thinking is, well, this might be technically on paper true, but maybe this doesn't reflect the actual behavior. Maybe the folks who are posting more are the ones who are uh like who are willing to go out of their way to self-host their data.
And so if you look at that as well, you grab a sample of the fire host stream and you say, okay, let's look at uh where folks are liking uh where folks are liking, who they're following um, and who's posting, maybe this will get you different numbers. Unfortunately, that's really not the case. Uh followers, about 99.9% of follows are going to focus on blue skies, uh, same with likes, and then posts are a little bit higher in that only 98.9% of posts are coming from folks on Blue Sky's servers. Um this has a few issues behind it.
Um and I want to talk about this in the language that we so often do in Bluesky, which is what if 2022 Twitter happened again? What if an investor came in or somebody felt the need suddenly to make a bunch of money and they want to take over this network, um, how credible is our exit as users? How sustainable is this network to that threat. Um the first concern, and this is one that Paul identified in a blog post back in February, um, is that still it isn't there's not a lot of adoption. That's what the measuring decentralization actually gets you, and part of that goes with these overarching questions of funding, but also part of that is just a challenge for users, right?
Users are not always clear about the value they get from moving their data abroad. I think all of us understand the value of hosting your data yourself or going to something like Eurosky or BlackSky or North Sky and having somebody else host your data, but that's not always clear to users who might not be willing to adopt it. Um and if you think, if you don't believe me in that, uh I would encourage you to do what I did a week ago, which is I was visiting my grandparents, go to a retirement community, try to explain the concept of Bluesky to them.
I do not think it will take you long to realize why adoption might be challenging. The second is that there is just practically a limit to what a protocol can get you. Protocols always exist within these kind of broader socio-technical systems, and so there's going to be a limit to the extent to which a protocol actually indemnifies you against these concerns. Um the third thing, this is to get just be a brief legal nerd for a second, um, Bluesky is incorporated in uh Delaware, and under Delaware law, Bluesky is what's called a public benefit corporation.
Now this means that if you do some if you can have a mission and if you do something that's against that mission, your shareholders can sue you for violating your mission. That is great in theory. There are two problems with this. The first one is that you no longer you can change your mission at any point if you have over 50% of the shareholder voting, including getting rid of your mission entirely. And so if 50 point 50.1% of Bluesky shareholders said, hey, actually we don't want to be a public benefit corporation anymore, that could be done tomorrow, and there's nothing any of us could do to stop it.
The second concern, and that didn't used to be the change, that's a 2019 law change. The second concern is that we don't know what percent of like we know the list of owners vaguely, we can see the list of investors, and they have posted on they have posted before about who owns it, but we don't know how that breaks down percentage-wise. Right? We don't know what percent of Bluesky J owns or Mike. Very smart people, I've met all of them today, very nice people, but we don't know how that actually breaks down. And so it's unclear if that 50.1% number rep would have to would represent a coalition of most of the owners, or if that could be achieved by a couple of investors working together.
Unfortunately, I'm also not trying to put a turn in the punch pole here, so I'm gonna end with a little bit of optimism. Um I've seen a lot of stuff over the last couple days that have really inspired me that have made me feel a lot better about this concern, though not entirely fixed. The first is the idea, and I know Daniel's gonna talk more about this in this uh talk tomorrow, uh, of independent PLC uh governance. And so that does give you some indemnification where if the PLC directories were uh wholly owned by another organization in this kind of catastrophic platform takeover situation, maybe you could get them, you could get um those your identities would be indemnified against that concern, right?
So if we have a different organization that controls all these identities, in this worst case scenario, we at least have a third party who we can go to and say, hey, Blue Sky's trying to fuck us all over, let's let's fix that issue. Um the second concern with that is even if you get your identity migrated over, you still have this concern about data ownership. And that's where projects like Hubble are really valuable if we get this public copy of all the data on the platform. That way, if you move, you would have a third party in uh in the uh independent PLC organization that would give you your identity back, and then using the data backups, somebody could put together a tool to recover your data.
Now, this gets a little bit more complicated because of permission spaces, but if anyone here is technical and wants to think about how you could under the new permission data permission space structure, you could help people set up their own backups. Please talk to me. I think that's a great idea. Third is the rise of community PDSs like Eurosky, BlackSky, North Sky, all those wonderful folks, which present are really promising. The more we can encourage adoption of these third-party things, the better. And last but not least, because I realize I'm getting close to time, um, is the broader community.
It has been a really inspiring few days talking to everyone here. I've had a lot of wonderful conversations. After I gave this talk yesterday, I was surrounded by the mob of three different Bluesky engineers who threatened to break my knees. I'm kidding, they did not. Um, but they're everybody is is equally as concerned about this, and folks are working really hard. Um I would just say that there's a limit to what uh we can do technically, but what we can do technically is still really important. Because we have to do is we have to decrease the lift on users to adopt these alternatives, but we also have to increase the we all reduce the burden and increase their incentive to do so.
Um and I will leave you with a quote from Lawrence who totally nerd-sniped me. Thank you.