Recorded live at the Fediverse House at SXSW '25, Bluesky CTO Paul Frazee explains how the company, powered by the AT Protocol, is rethinking the internet’s architecture to create a more flexible, user-centric web.
From the outside, Bluesky may seem like a Twitter clone. But anyone who’s close to the technology — and the team — knows that they’re building something much deeper: they’re rethinking the internet’s architecture to create a more flexible, user-centric web.
Bluesky’s CTO Paul Frazee is the perfect person to explain all this, as he’s fantastic at tying technical concepts to their practical application and wider impact. In this interview with Mike McCue, recorded live at the Fediverse House at SXSW 2025, Frazee unpacks Bluesky’s first principles, what makes AT Protocol different from ActivityPub, why identity portability is a radical shift, and how decentralization could lead to more humane social spaces.
Other highlights include:
Mentioned in this episode:
🔎 You can find Paul on Bluesky @pfrazee.com.
✚ You can connect with Mike McCue all across the social web, including on Bluesky @mmccue.bsky.social.
🌊 Catch the wave! Surf the social web and create your own custom feeds at surf.social, a new beta from the people at Flipboard: https://about.surf.social/
This transcript was generated by AI, which may affect its accuracy. As such, we apologize for any errors in the transcript or confusion in the dialogue.
Welcome to Dot Social, the first podcast to explore the world of decentralized social media. Each episode, host Mike McCue talks to a leader in this movement; someone who sees the fediverse’s tremendous potential and understands that this could be the Internet’s next wave.
In this episode, Mike’s talking to Paul Frazee, the CTO at Bluesky. The two chat about Bluesky’s meteoric growth, the principles behind Bluesky’s architecture, their vision for identity on the internet, and much more.
This conversation was recorded inside of the Fediverse House at SXSW so you might hear some background noise and other audio imperfections. There are also a few questions from the audience at the end.
We truly hope you enjoy this conversation.
Paul Frazee:
Gotta put on your wings. Let me see if I can get it over the jacket. This’ll take me about 30 minutes.
Mike McCue:
Paul asked me, you know, do you mind if I put on fly rings before?
Paul Frazee:
But we, one of our users, came up with a good term for our ecosystem, which, this is the ad protocol. They called it the atmosphere, which actually worked really well.
Mike McCue:
That's a great one.
Paul Frazee:
Yeah, that was a really good vibe. I was jealous that I didn't think of that one!
Mike McCue:
And it goes well with the wings as well.
Mike McCue:
So when you think about the sort of the first principles, how you've architected Bluesky, what have been some of the most important first principles for you?
Paul Frazee:
Yeah. So there's a lot of ways to attack this one. I'll start with the in general, the goal of what you're trying to do is establish independent publishing and independent distribution. Right? There's been a push and pull inside of basically all internet technology from the get go where the web really nailed independent publishing, and it still effectively does that. You can still set up a website that you run the servers for and have a lot of control over, but that's only really half the equation. And once we started to get into the more kind of app world, it started to become really obvious that distribution was a huge part of the publishing equation, which is where social networks started to rise, and where you and your career working on RSS from the get go were, like, ahead of the curve is capturing correctly. Like, oh, you know, people are not always going to go to your website, yeah. And so eventually, the social networks really took control over the distribution side, in addition to the other kinds of just socializing aspects of social networks, and what you gotta you gotta push on that you can't leave the distribution to be closed in the way that it is, because that really it's to the extent that minor algorithm changes end up ruining business models, yeah? So you see existing creators pushing back on that in the form of newsletters like sub stacks. That's a really great example of independent distribution. They're just using email, right? That's a for that particular form of blogging. The newsletter model through email is so perfect, yeah, right. But for social networking email doesn't cut it. You say you need something that's a little bit more intelligent and able to persist over what people like about social networks, yeah, while still having a level of independence, and that's what the entire feta verse is focusing on, is making sure that there's an independent method, an open system for the distribution side of things. And so that's, I think that's how to describe that.
Mike McCue:
Yeah. I mean, I do think that's a very, very big deal. You know, earlier today we had Molly White on, and she, if you don't know her, just go to mollywhite.net and you will see everything that she's publishing, all of her writing. She has three blogs, she has YouTube videos that she posts. She's got podcasts that she does, just go to mollywhite.net and how cool is it that I could just say, go to mollywhite.net right? I mean, that's really the way it should be. And she doesn't care whether she writes something that offends Elon, and then Elon she, you know, I asked her an earlier panel, like, hey, how likely is it that Elon's going to ban you at some point in the next year? She's, like, highly likely and but that's not going to ruin her business, right? And that kind of independence for people who are making content is incredibly important. And of course, that's what the web has always been about. But somehow along the way, we sort of lost track of that, like, because we put things on platforms that are, you know, specific to that platform, and then we sort of tie our business to that platform.
Paul Frazee:
Yeah, it's understandable that it happened to, like, the it's the work that's been going on in all the different federated technologies, like it's not exactly easy, and so those companies were tapping into the fastest way possible to build the kind of experiences that people wanted to have, and now just the technology standards are just catching up to it. So in many ways, this is about the commodification of the patterns that we establish work that people want to have, which is, you know, happening also at the same time that there's dissatisfaction with structural aspects of it. So great, you know, let's go ahead and make the move.
Mike McCue:
Yeah, yeah. And, and, you know, by the way, how many any founders, any entrepreneurs out there. All right, nice, cool, excellent, and we're gonna do Q and A here in, you know, about halfway through the discussion and but I if you are someone that wants to build an app that relies on community, you know, it used to be that you had to build the entire community staff yourself, and then hope that people would come to that community and build a profile figure out who to follow, right? Yeah, but that's not how. That's not no longer the case.
Paul Frazee:
Yeah, exactly. That's it. Like, one way I like to say this. I don't know if this resonates with anybody when I say it, but it's like teaching the internet how to do social because what the internet does is it's a giant, collectively shared and own thing. Yeah. So whenever you start to tap into these fediverse technologies, you're tapping straight into that existing network, you know. And getting the user base initially is such a challenge, getting an initial enough activity going on things like that. So even if they're not your users, you're, you know, and we ran into that cold start problem ourselves, because we actually did a, you know, it's a separate technology with the ad protocol. So for a while, people are just like, yeah, I showed up and I looked for my favorite people. They weren't there, so I bounced right. So that should just stop happening and stop being a factor entirely, and how these things go. So now you're lowering the threshold costs for new products to come into play, and that means that there's just way more opportunity for people to build something better. And if it's not us, it's the next person, right with each of these projects. In some ways, we kind of want to make sure that we stick to well established patterns with what we're doing with our work, because we know that works and but even more adventurous people can come afterwards, and that's great. That's absolutely what happened. But it's not just the people working on application experience, so that's important. But like, also, again, I'm going to harp on this, like, independent distribution thing you've seen things like, YouTube creators are very frequently trying to solve their independent distribution. There's a project called Nebula, kind of like, also what musicians try to do with title, like, look, this is the distribution. We need to have some control over this thing. If the cost of building your own sort of focused on your art in your craft platform that gives you control of your distribution can happen, and the cost can get low enough that you're able to integrate right back into the whole network, and it starts to become way more viable for that to happen, which is so important for people that are trying to have online business as their you know, their actual business, and rely on that revenue. So lowering the barriers to people building applications, I think, is just a very important next step for where the internet needs to go.
Mike McCue:
Yeah, and you're one of the blog posts. I saw your team post several months ago. I thought was fantastic. Is a call for developers, and it outlined, you should check this out if you're interested in building new apps in the social space, either apps or backend services. And I loved the all the different ideas that you guys outlined for what people could be. Yeah, there's a huge kind of green field of opportunity for new kinds of services, like you said, more adventurous services. Yeah, it was amazing.
Paul Frazee:
Yeah. I mean, the again, this is shared across all of the federal technologies that like, they're designed with the ability to go off of the micro blogging form, the micro blogging kind of like the obvious entry app into it all this infrastructure can be repurposed in so many different ways. A lot of applications share the same shape. It's just a different focus on different kinds of, you know, media, yeah. But then, you know, okay, now get into forums. Now get into code sharing and collaboration tools there. There's a lot of different things you can do. I personally, my the one that I'm really waiting for is, this is maybe a bit of a flight of fancy, but I think micro podcasting app should exist. You tie that to a sort of Clubhouse style. You know, you get into a room and you talk, and then you snapshot it and onto a feed, waiting for somebody to do that…
Mike McCue:
…on AT Proto?
Paul Frazee:
Whichever one works for, you know? But yeah, like, it's enough of a hair-brained thing that I'm hoping that, you know, yeah, that's the one I'm hoping somebody will do at some point.
Mike McCue:
Yeah, there's, there's an app called Air talk. Have you seen that one?
Paul Frazee:
Air talk, or air chat?
Mike McCue:
Air chat.
Paul Frazee:
That one was interesting. It was close, yeah, probably what I wanted, yeah, not quite, yeah.
Mike McCue:
And also, again, it's you have to, like, create another account and get people to follow you. And, like, you know, anyone would do that now.
Paul Frazee:
Yes, so the Yeah, it should just be, you do your sign in it. Ideally, it's going to be, like, a single sign on kind of thing. So, like, you're used to, we're just pressing a couple buttons. You're not going to enter anything, yeah, and then, and then it's just everybody that you already engaged with is there? It's going to be, you're still going to, like, there still are they hanging out in that? So the further away from micro blogging that they go it may, if it was like, audio things, you may have to convince them to just, like, open up this app and start to engage there, right? So, like, a kind of locality of like, I'll go big with it. When you have a connected network, compatibility becomes the walls. So it's about like, are you engaging on the same kind of subset of the experience? Yeah. So there's still that aspect of it, but boy, it's a lower barrier that used to be.
Mike McCue:
Yeah, yeah. And what for developers who were thinking about building on protocols directly, yeah, which you don't necessarily have to do, there are APIs. There's other other kinds of extraction layers, I'm sure coming. But yeah, what? How do you what advice or coaching do you have for folks in terms of ActivityPub versus AT Proto? How do you think about those, the differences between the technology, yeah, and when to use one or the other, even though they both have the same purpose and the same vision?
Paul Frazee:
Yeah, yeah, they are very similar in their goals the the boy. So I would say that ActivityPub definitely has a really great strength on this kind of notion of communities that are represented by the service themselves, and then the ability for them to bridge between each other through the perimeter. And so there's something actually very intuitive about that design, that you set up a service somewhere and then just speak to the other services. So I certainly think there's a lot of intuition there, and in some ways, I would be inclined to expect that it might might be an easier sort of secondary addition into a more traditional approach to a service. I think you have to buy in a little bit more into the AT Protocol to leverage it. The AT Protocol tends to think more in terms of aggregating high scales of user data. Because what it's trying to do is make the entire experience feel like one shared network, and the way that it does that is by actually functioning really similarly to the web, where every application is almost like like a search engine, and it crawls across all of the activity of people that is being hosted on their own, sort of not quite websites, but they're almost like websites. So the the application experience is just crawling across all the activity, gathering it together in big indexes and building an application experience out of that, which means that the applications are sort of detached, actually from your account, which is also in some ways a benefit, but it plays well into the high scales. Doesn't play as well yet, into like selective sharing of information. So if you're talking about right now, we it's on our roadmap to get like selective sharing. You're talking right now with thAT Protocol, you're going to want to do public posting.
Paul Frazee:
And yeah, I think that's how I would pose it.
Mike McCue:
And how do you think about the bridge and bridging strategies?
Paul Frazee:
There's incredible stuff happening there. What is Ryan's thing called?
Mike McCue:
It's called Bridgyfed, exactly. And there's a nonprofit called A New Social, yeah.
Paul Frazee:
Did he form that?
Mike McCue:
He did yes with Anuj, yes. And really, guys are doing great work.
Paul Frazee:
Yeah, so there are people out there building bridges directly between the two different protocols that, to a shocking degree, actually abstracts over the difference. When he first showed up talking about that project, I was good luck. You know. I think it could work, maybe, but I wasn't so sure that I can, like, definitely do that. And he's surprised the heck out of me with that. So as a consequence of that, I think the way that you walked into it, at least in the Bluesky AT Protocol world, you just follow the account, and then it like, okay, great, we're gonna follow you, and it creates sort of like, another account over on the other protocol. So if I'm a Bluesky user, it'll create an ActivityPub count, or an ActivityPub account. It'll create a Bluesky account that just sort of mirrors over the activity and a lot of the like. It works pretty well. Like, I think I've seen likes transfer over. Am I right about that?
Mike McCue:
Likes and replies and replies, yeah.
Paul Frazee:
It's really pretty good.
Mike McCue:
Yeah, it's pretty good.
Paul Frazee:
So cool. So it's definitely doable, yeah.
Mike McCue:
The only little wrinkle in the bridging thing is that you do end up having these two accounts, and you're like, wait a minute, which account am I looking at here, right? So I mean, though I do think there are things that can be done in the UI for Bluesky and other apps to kind of like, only show one of the accounts and, you know, show the relevant one and things like that.
Paul Frazee:
This sounds like a call that we need to have. Yeah, what's being set up right now?
Mike McCue:
Exactly, exactly, we definitely do. And it's a little bit also of a sort of lead into identity, yeah? Because that's a, you know, if I, like, for example, I have a Mastodon account, yeah, I have a Flipboard account. I have a Bluesky account, yeah, Pixelfed account. If you search for me on the social web, you're gonna see like, eight versions of me, right? Sure, sure, and then, but there's no, there's no way to just follow all of my accounts, right? Even though they're, you know, they are, they can be bridged. And my Pixelfed account, by the way, is bridged to Bluesky, yeah. And, by the way, for those of you who want to do that, it's really easy. You just have Pixelfed follow the Bridgy Fed account, and it's automatically then bridged to Bluesky, right? So, and, and it looks fantastic, good on Bluesky. It's fantastic. So I linked to it in my Bluesky profile, right, right? But what are your thoughts around like, where do you see this going further? This is a little bit, you know, kind of a hacky scenario now, where how important and central is identity for you guys?
Paul Frazee:
Well, very much. So, yeah. So from the perspective of the AT Protocol, like I mentioned when we were doing the side by side just a second ago, scale was a big part was to make sure that identity is portable. That was a really big thing that we were working on right from the get go. Yeah. So we solved that through these persistent identifiers, these, this kind of like a giant unreadable number that everybody's got that is has the goal of being one thing that you can reuse across multiple different applications, right? And it gets assigned associated with the domain name. So I'm petefrazee.com and that just actually maps internally to this, like long Yeah. And so the hope with that is that if you're migrating around between services or changing your handle at any point, which happens, yes, that it still persists that identity, because, again, lowering friction is such a big part of what we're trying to do. And so if you start out with us, and then at some point, you're like, I can't anymore with these folks that if you wanted to move somewhere else, you could do that and have really zero pain, right? All of your previous activity, right? And all your follows and followers and everything that is a part of your social account, the hosting should just not be, not be something you're hard bound to because, again, easy choice ensures that you're focusing on what does matter, which is what these companies are doing, and are they representing your values effectively? And what you need out of it, the protesting with your feet just needs to be baked in right from the good right, while your identity remains persistent throughout right. And it really has to do with like, trying to set a baseline of like an individual, like rights given to people, right that you should be able to exist as a person outside of a corporate innovate, that you are a, maybe not even a high, pretty low priority customer for, yeah, right. Because otherwise, like, how much more do we want to, like, hand off any kind of individual rights to these, to these companies, yeah, and speaking as a company, it's like, fine, because our job is, should be to provide a service to people not to own their online identity, right? I think that is a more balanced and kind of just long term, feasible relationship that should exist there.
Mike McCue:
Yeah, 100% Yeah, yeah. You know the, I think the fact that you guys adopted that, did you know W3C standard, definitely worth looking at this because, you know, it's sort of like, it's analogous to contact, your contact book, your address book and phone numbers. So, you know, I could, you know, maybe, you know, Paul decides to change his name and I change his name in the my address book, but I still has the same phone number. I can still reach you. And you might change carriers. You might decide you're on horizon or AT&T, I can still reach you. I don't have to think about that, right? And that got that unique identifier number that's super powerful.
Paul Frazee:
It didn't used to be that way, right? Like, there was a period where phone numbers were carrier specific. So whenever you change your carrier, you boy starting over, and you're getting those random texts, yeah? Like, hey, it's my new phone number. What did that change?
Mike McCue:
Yeah, it was a very important change. I think it was about 20 years ago, yeah, something in that zone. Think I remember a number portability, yeah, no. So they had to, like, do a bit. It was a government, it was a regulatory, yeah, that happened.
Paul Frazee:
I actually should look up sometime, how they solved it, but, like, yeah, they eventually introduced the ability for the number portability that it carries over with your carriers, pretty good for keeping a competitive market. Yeah, yeah, exactly. It's that's friction that should not be a factor in a choice.
Mike McCue:
Yeah. I personally think that your choice around doing identity, the way you guys have is going to pay off to be one of the most important, best decisions you guys have made. I think it's incredibly thoughtful and bulletproof for the future, and something I hope, I hope we can, you know, adopt also in ActivityPub world and in other places. So there's an opportunity, I think, to rally around that holistically, one of the one of the, one of the other things I always hear Bluesky people talk about is the concept of composability. Sure, yeah. And for people who don't know exactly what that means, can you just elaborate a little bit on what is that? How do you get why do you talk about that so much?
Paul Frazee:
All right, so now I'm thinking about the hands raised earlier, and I have to kind of think about how to communicate this for the varied audience. All right, so a big part of what we have focused on with this is making it so that composability. In this case, it's about collaboration amongst different services. The I'll start with an example the way that feeds work in the Bluesky application is that anytime. So we have this idea of feeds, where you can go to a page inside the app that calls itself a feed, and it can be about sort of anything that the creator wants it to be about. It could be a cat pick speed. It could be fungus pictures. I think we have one of those. It could be the kid literature community. And it can also be sort of algorithms, right? Like one that is pretty good is Quiet Posters. I love that one that's that one's a hit, and that one is just like the people that you follow that don't post as much, this feed is just their post, and they do post because so easy to miss them in the you got some post, lots of don't and the way that all of them work across the board is that they are running on somebody else's servers. So somebody, at some point, sets up a server somewhere that is creating these feeds, and the Bluesky application reaches out to them to find out what should show up whenever our user visits the feed, and then it sends back, basically a list of post URLs, which then hydrate into the full experience. We did that because we wanted to make it so that the there's true algorithmic choice in this thing, because that's something that matters quite a bit. And so that's an example of composition where somebody else builds some entirely separate service, but it's able to work collaboratively with ours to drive something which feels. We call it feels first party. We call it third party as first party, right? I love that. Yeah, yeah. I don't think RSS was in the same energy, right? Yeah. And the so we've done that so kind of across the board, everything about this is designed to have that level of mixing and matching, and you can pull out an individual piece and reuse it in some way, in terms of how it renders out into the user experience. Another major one is the moderation system does this as well. I think we want to keep working on this, because we could build it out more. But we call this one stackable moderation, where the exact same moderation service that we run can be set up by anybody else and subscribed to by users to help continue to refine down what they're dealing with every day and help to reduce the noise. And I think actually we can go even further with that below. Let me try to do that. So I think that the that approach kind of naturally came out of the engineering culture that we have, yeah, but then we're happy about it, because it allows us to, with the feeds and with the moderation, start to have some of the practical value show up of the whole conceptual notion of an open internet and a decent delegating out power. Yeah, that it could actually be a part of the application itself. Yeah, so that it's not an abstract thing that's just sort of, somebody tells you this app is decentralized, and you go, Cool, okay, and then you never think about it again, that there could actually be immediate benefits to people. Composition really enables that.
Mike McCue:
Yeah, there's like, a Lego block kind of mentality, right? Yeah, you guys make Lego blocks. You allow other people to make Lego blocks. They can build things into bigger Lego blocks, absolutely, you know, and that that's actually kind of amazing, because the level of innovation that I think is possible because of that is something we haven't even begun to, you know, see yet, it's a there's a lot of really cool things that can be done with that.
Paul Frazee:
I think so too. Yeah, I think 10 years out in particular, like I'm really looking forward if the if the tools that enable that composition continue to hold up, then, like finding even more ways to have this, this mix and match, could really turn into something interesting.
Mike McCue:
Yeah. Well, one of the things that some folks are just a little bit hesitant on as it relates to AT Proto. Is the, you know, how federated is it? When will it be federated? Also ownership of it by a company, versus, you know, or W3C, blessing of it, yeah. What are some of your you know? How do you address some of those questions that people ask?
Paul Frazee:
Yeah, so, the design of the technology is federated. So I'll kind of start with where we're at, and then in a positive sense, and then we'll talk about, like, what we're pressing on. So first of all, 99% of everything is open source. The one bit that's not is we have kind of an internal database that we built that's not open source, but there's an alternative version that's available. So all of it is open source. The specs up, so there's everything to work from, and it's the specs and the source code are licensed in such a way that other people can can take it and run on their own. And then in terms of other people running infrastructure, we have already got people doing self hosting of their data, and that's active in the network now, and we are starting to see people start to build alternative applications. Remember what I was saying before about the hosting versus the distribution so we've got some good initial progress on the hosting side, and then the distribution side is what we're now needing to press on to get more active operators of it. The barrier at this point is less the technology and more the operational cost of running an alternative application, because, because you're saying, hey, a big network, yeah, the model that we have is this, you know, by default, is this idea that you're trying to pull in everything, aggregate everything, so that you can give that big network experience. And so what we're pushing on now is starting to build out, or work with the community, to start to have partial views of the entire network so that it costs less to run.
Mike McCue:
That's awesome.
Paul Frazee:
Your instance of that, yeah, is that gradient matters so much? Like, we're at that stage now where there's a lot of a lot of this is still community oriented, where it's a lot of individuals that just have a wild hair and want to jump in and do something and for them, like, if you're asking them to spend more than 100 to $300.03 100 still pushing it a month on setting up the server, that's asking a lot. So you want to be able to set up a nice scenario where these independent hosters can set up their own version of not just the hosting, but also the application. And to do that, you just got to shrink down how much the network that you see, yeah, and you do that by looking at the follow graphs of the users and so…
Mike McCue:
Right, that's awesome,
Paul Frazee:
Yeah, well, okay, yeah, you guys just did this.
Mike McCue:
That's a good idea.
Paul Frazee:
I just discovered, is that Flipboard just recently did this, and very excited about this.
Mike McCue:
And works very well.
Paul Frazee:
Yeah, good. Yeah. Yeah. The intuition on this is just like, you know, every user, like, we have a whole lot of people in Japan, and I just don't interact with the people in Japan, so if our server didn't sync them, I probably wouldn't know, yeah. So there's everybody has their just sort of window of the world based on their follow graph and their interests that they are going to come into contact with. And if you could shrink down the sync set to just that you hit like the cost optimality, right? Some of the informal benchmarks we did, we're pretty sure that $100 you can, at minimum, hit, like a million of the users in the network pulling in their activity. We'll love to see that the higher that gets, the better, right, for that cost, and then that will really healthy independent operation. And then, you know, part of what I would hope is that you see more of these independent operators, but then, over time, they continue to grow, but they're on that right gradient where they're putting in more money based on how much you how many users they have and how committed they want to be. And then they eventually get all the way up to running their own instance.
Mike McCue:
Yeah, community on demand, yeah.
Paul Frazee:
So it's a process, but, like, pretty optimistic about the direction it's going.
Mike McCue:
Well, you guys have done completely, you know, I have, in my opinion, you've struck a very good balance between, you know, how much do you centralize versus decentralized. How do you think about the architecture, short term, long term, and I've been incredibly impressed with the way you guys have been approaching it. We, as a developer, we've been integrated ActivityPub and AT Proto double stack, as they say, two stack, and it's totally worth it, because it gets you access to, like the broader social web, there are people who are like, Barack Obama posts on Threads. It shows up on Mastodon. I bridge my, you know, I can, and I can bridge that over to Bluesky. Can see it on Bluesky. It's awesome, yeah, and and, and then with Bluesky, you know, you have AOC who's posting, you know, as she's fantastic to follow on Bluesky. So, you know. And the cool thing is, ultimately, people who just search for Obama or AOC and hit the Follow button, they don't have to know where is the person posting from. Maybe one day they posting on Bluesky. The next day they posting on Threads. The next day they posting on Mastodon. Doesn't matter, yeah, so I, I'm remember, I'm a big believer, and let's open it up to some questions here in the audience.
Questioner:
Hey, what's up? Paul, what's up, Mike, so it's Alex in the back. Hey, so there's so many different questions I'd love to ask, like how and why did ALF take off early in the Bluesky community, or maybe shining a spotlight on like Peter Wong and his fund, or as well as Mark Cuban and the opportunities that that brings in terms of development, in terms of the fediverse. But the thing I really wanted to dig on, since you brought it up earlier, was about self-sovereign identity. And I wanted to kind of look at like your point of view on how much is ICANN kind of the final boss of the internet?. And you know, you're able to bring in domains and use those as a base of your identity. I do as well in the in Bluesky, which is brilliant, but for those of you who are thinking in that direction of self-sovereign identity and how we have this like one centralized entity behind it all. Any advice you would give to teams that are pushing in that direction and trying to get a little bit lower in the OSI model? Is there any advice you'd give to those kind of teams?
Paul Frazee:
Yeah, all right, so what Alex is asking about here, we use domain names for handles, and the ICANN is the organization that manages domain names throughout the world. So it is, is it a nonprofit? I don't actually know how that is structured legally, but it is a multinational governance system that I think so far has done a pretty good job of upholding its mandate and staying relatively focused on just respecting the core of what domain names are for. The risk with any kind of organizational system is that an organization could fail at what its core thing is. Now, thankfully, I don't think there's any indication that ICANN is a problem, but I think probably what you're asking about, Alex is whether or not there are alternative technical solutions to the same thing, broadly speaking, that boils down to, could you use a blockchain to try to solve the problem? I know Alex, so I know. So we don't use I'll start by saying that right now, Bluesky doesn't use any blockchain technologies. However, in the realm of, like, across the board, the evaluation that we're using with AT Proto for anything is just, does it actually do the job that's needed to be done for this? Is it actually a feasible approach to how this can be done? And in terms of, like, putting your posts on the internet, not a great fit, in my opinion, yeah, it just you're not getting the cost profile is insane for that per transaction, you're spending a lot of money, and it's very inefficient. The throughput is terrible for usernames. It might make sense, but it would have to come from a point of view entirely, from the expected outcomes working backwards, which means that it needs to be able to hit reliable costs that are not going to be subject to a lot of speculation, that it's going to have very fast resolution and reliable resolution, things like that. So it's really an engineering problem more than anything, I think there's a decent chance that it could be done, but it's also not necessarily something that we're pressing on right now. But that's also one of the nice things about the DID system. It's designed to open up to other approaches, so that can be done at the DID level, or it can be done to the username level.
Questioner:
Hi, Paul. Nice to have you here. It's really cool to pick your brain. You have an extensive history beyond just working on AT Proto and Bluesky. Like prior to that, you worked on the Beaker Browser, you worked on Secure Scuttlebutt some and to some degree, like some of the lessons from Secure Scuttlebutt have gone on to influence Nostr and other approaches to distributed communication systems, for lack of a better word, I was wondering if maybe you could speak to some of the lessons you've learned along the way, and how that has informed your current approach to building AT Proto?
Paul Frazee:
Definitely, yeah, So when the Bluesky team first got together, we actually all came from a similar heritage of the peer to peer world. And there was a lot of things that we all were feeling at the same time having spent time working in these technologies, which was essentially, it can't, there cannot not be any steps backwards in how the technology works. Another way that I was putting it is that it can't be rocket science to implement a comment section right which Secure Scuttlebutt and Beaker were conceptually and technologically really very interesting, but they tended to live entirely on user devices. They were peer to peer. It was like, if anybody here knows BitTorrent, that was how those technologies kind of fundamentally work. So no servers were involved at all. And you had problems with scale. You had problems with account synchronization across devices. I'll put it that way. It's key sync, is what I'm talking about. You had problems with reliable availability of information, and at some point you have to recognize that interesting technology also needs to yield good affordances, both for developers and for users. And that was what we walked in with was like that just can't compromise on that. And the way we ended up squaring that was taking a lot of that technology that we had worked on and moving it into the server stack, because the peer to peer technology, when it comes to this account portability thing, that's what it's designed to do. It's designed to abstract away the individual devices, but but if you move it into the server, then you're able to have the high scale and the reliability of server infrastructure. And so that's the in terms of the technology point of view, was the main sort of lesson carry over from that. But I actually wrote up a post mortem on the Beaker Browser, which, if you want to read a little bit more about that, just Google postmortem of Beaker, and it's like on the GitHub, and there's a couple of other things you might find interesting.
Moderator:
We have time just for one more question.
Questioner:
So I'm asking this to the guy wearing wings, but I often think about the federated world as good guys being on the side of angels and you're creating these little universes that talk to each other and don't talk to others. And you know, one of the things of it in terms of the social engineering of this is thinking about universes that do talk to each other, and we are becoming farther and farther apart in terms of communication in our country overall. Have you thought about how people will talk to other universes versus just talking to themselves in the echo chamber?
Paul Frazee:
Yeah. So you're talking about the kind of the cultural universes that people Yes, yeah, yeah. I mean, this is something that we talk about a lot, and it is that's challenging beyond technological questions, yeah. So, yeah. So what constitutes wisdom here is an open question. I think to some degree, I am less inclined towards the make sure that that everybody gets a platform all the time sort of thing. I think that respect for people and having a positive connection with each other is quite valuable. In fact, I think one of the things that leads to a lot of conflict between people is just bad interactions across these universes where they develop an even lesser opinion of each other whenever they do come into contact, so that the experience that they have with each other in terms of actually building crossing that bridge, you got to be mindful of like it's not enough just to get them into the same space, because then they'll just argue. Right? So how can we get people into a time with each other where they build sympathy and empathy for one another and actually come to like each other. And so to some degrees, I try to think of it less in terms of breaking out of filter bubbles, and more in terms of encouraging just positive and convivial experiences with each other and reducing that base level of toxicity that they or eventually the term I like, is torment that the internet is just constantly subjecting you to. That is my best guess at how to move forward on that. But fingers crossed.
Mike McCue:
I'll just add a real life example. So I have a troll pen pal on Bluesky. And so, yeah, I posted something about, you know, the Elon Nazi salute. And this guy, you know, who is, clearly, you know, an Elon supporter right wing. Very, very right wing, post it back. That's not, that was not a Nazi salute. Like, look, here's all sorts of other examples of, like, Obama giving a Nazi salute and so on, right and and what was really interesting, though, is like, this guy, you know, claims in his profile, like, I'm here to, like, actually talk to people and have real conversations, right? So I was like, All right, let's have a real conversation. And, you know, every now and then, I'll actually reply to him, and he'll reply back, and we'll, like, go back and forth, and we'll, you know, it's not like I convinced him, you know, of anything, but at least I showed him the respect of like, hey, you know, you have your opinion, and I've got my opinion. And, you know, I do think that was actually kind of cool. I feel like there's a little bit of an obligation that people have sometimes to like, not just be like, you don't know what you're talking about, right? And I do think that there's you sometimes you don't want to feed the trolls, but I will say that like, it's kind of cool. It's kind of cool to be able to actually engage with someone like that. If I ever decide to, I can always block them, sure, right? But this person's been actually reasonably respectful back. And that's that's a start.
Paul Frazee:
That's nice outcome.
Mike McCue:
It is nice, because also on on Bluesky and on Mastodon, there are ways to protect yourself, right? So I feel a little bit more willing to engage with people that I otherwise wouldn't like I would never do this on Twitter. That's just insane, right? Yeah, because all the other trolls have come swarming, right? But I know if that were to happen on either Mastodon or on Bluesky, I could easily, you know, just block those people and just never think about it again.
Paul Frazee:
Both universes are putting a lot of thought into that side of it, you know. And actually a lot of the tangible effects of the more delegated out moderation is to find ways to have that slightly more hands on, kind of cool down. Yeah, you know, and yeah, we're kind of constantly looking for new ways to take the temperature down on that sort of stuff, so that, you know the I mean, in this case, what you're saying is that you felt comfortable even taking a shot at it. Yeah, that's a good step in the right direction.
Mike McCue:
Yeah, it is really great. And it's a testament to the work that you guys have done, that the Mastodon team has done, the Threads team has also put a lot of energy into this. And I have to say, I'm like, I'm it gives me hope you know that you can, if you know how to use the tools, you can actually engage with people and feel safe and maybe try to start to bridge some divides. So I think that's a great question, and a great question to end on too. We're sticking around, so feel free to come talk to us. We've got another we'll be doing another discussion here soon. And then the feniverse party tonight. So thank you again for the good questions and for taking the time to be here. We'll see you again soon. Thank you everybody.
Well, thanks so much for listening! You can find Paul on Bluesky: @pfrazee.com.
You can follow Mike on Bluesky @mmccue.bsky.social and on other federated networks.
Post-production on this episode comes from Kaleidoscope.
Lastly, if you’re interested in this podcast, you might want to check out Surf, a new product from Flipboard. Surf the social web and create your own custom feeds. Learn more at surf.social.
Until next time, we’ll see you in the fediverse!