Bluesky Is Building The Decentralized Social Media Jack Dorsey Wants, Even If He Doesn’t Realize It

Say That Again

from the directionally-right,-specifically-wrong dept

There was a bit of news in the world of decentralized social media over the past few weeks. It kicked off with the announcement that Jack Dorsey had left the board of Bluesky. This was followed by an interview Jack gave to Mike Solana where he explained his thinking on all of this. There was also a flurry of talk claiming (misleadingly) that Jack had endorsed ExTwitter.

As with many things related to Jack, I think a lot of what happened and what he’s saying has been misinterpreted (by people across a wide variety of ideological viewpoints). A lot of what he said is the same stuff he’s been saying for years and is actually quite sensible.

But, the one area where I do think he’s wrong is in some of his commentary about Bluesky, which surprised me a bit. Because I think he greatly misrepresents what is happening with Bluesky and why I still find it to be the most interesting experiment going on these days in social media.

I do think some of his views of Bluesky are colored by his experience at Twitter over the few years leading up to his stepping down from that company (a few months before the Elon saga began).

I want to go through parts of his interview with Solana, but I think understanding where Jack is coming from is actually really important, because when laid out clearly, it explains it colors his perspective on these things. Basically from the start, Twitter was pretty much the most permissive in allowing all kinds of speech on its platform with an extremely light touch toward moderation, but not no moderation.

As everyone who creates a speech platform learns at some point, you need some form of moderation. Otherwise, your platform gets filled with (1) spam, (2) scams/fraud, (3) illegal material.

However, there is a step up from there towards other moderation issues. Those who stop at the point of moderating just those three things quickly discover two other challenging issues. First, your site gets overrun with jerks, assholes, and trolls who make life miserable for all your other users. As we’ve discussed in the past, so much moderation is really “would y’all just stop being so awful to each other?”

Then, secondly, you have the Nazi bar problem. Many people get confused by this. It’s not just saying that a site is overrun by Nazis. It’s saying that in tacitly blessing the appearance of proverbial (or real!) Nazis, a site is blessing the space as a Nazi hangout, and then it gets that reputation among others, who realize maybe they’d prefer to hang out elsewhere. That leads to new challenges, often limiting growth of both users and business model options.

There’s a ton of nuance and challenges in figuring out how to draw the lines. You can easily see how sites can go too far in one direction or another, creating larger problems. Sites can (and do!) overblock in many cases. And sites can (and do!) underblock in many cases. And much of that is subjective anyway. What appears as overblocking to some may appear as underblocking to others. This is the old impossibility theorem at work.

This also means that if you’re “the decider” in these situations, everyone’s going to be mad at you. This is the nature of running a public platform. This was some of the thinking behind my protocols, not platforms paper. It was a look at whether or not we could, maybe, move away from the world in which we had one single “decider” to get mad at, and allow for a lot more experimentation.

The truth is that Jack had to deal with at least some of that. He was never really “the decider” at Twitter on moderation calls. The company had others who had to make most of the tough calls along the way. But Jack often got blamed for those decisions. And I think that getting blamed for that sort of stuff got to him. He’s made it clear over the years (including in this interview) that he would prefer not to have done much moderation at all. But the realities of the business say you have to for all the reasons listed above. He mostly delegated those calls so that it wasn’t on him to decide. But, he was still the face of the company getting yelled at for all of those decisions, and he really seemed to dislike that (which to some degree is understandable).

That appeared to be a big part of what appealed to Jack about my original paper. He had built a tool that was designed (successfully) to enable all sorts of speech and was celebrated for that fact. But over the last few years, he was increasingly being beat up on all sides, in part because of people (on all sides) were upset with moderation choices. And no, it wasn’t just “censoring conservatives” (a thing that didn’t really happen). Across the political spectrum, you had complaints about what was left up and what was taken down.

Jack, like others, reasonably thought: maybe it’s not great that a single person in a company gets blamed for all these decisions (that seems to have been a bigger concern than the fact that there was someone who could make those decisions in the first place, but it’s easy to merge those two concepts).

Yes, some people will say “suck it up, if you run such a platform, that’s part of your responsibility.” But it’s really important to impress upon you how deeply impossible this is, and how damaging it feels to be in a role where you’re just trying to enable a community to exist, but which requires some level of moderation to keep the community from destroying itself. Yet everyone is blaming you for basically everything.

At the same time, Jack was also dealing with an extremely dysfunctional board at Twitter. He had activist investors threatening to fire him if he didn’t enshittify the platform by squeezing more money out of its users, even as he took steps towards moving Twitter to a protocol. Jack, for all his faults, did seem to want Twitter to actually be a good platform, more than just one that sucked money out of people.

I believe Jack viewed the original plans for Bluesky as something of a lifeline, separating out the protocol layer, and allowing there to be competitive moderation services/interfaces for interacting with content on Twitter. In an early conversation I had with him regarding what Twitter would do in such a world, he suggested they could still beat the competition by focusing on “conversational health.” This would still involve moderation, but just at the Twitter service level, rather than the protocol level.

This is not to say that people should sympathize with Jack. Realistically, the people who actually were making the tough calls at Twitter and trying to balance all of these factors deserve way more sympathy. But Jack was getting yelled at the most over this stuff, and I think that really got to him. He was being blamed for not doing the impossible. To outsiders, the solution always seems “easy.” Stop the bad stuff, promote the good stuff. But, the “easy” solution that outsiders always seem to think will “work” is probably not dealing with the realities and competing pressures of managing a global community like Twitter. Understanding the realities, pressures, and (impossible) trade-offs at least helps understand the perspective, and why it can get exhausting to be constantly blamed from all sides from not magically “solving” things.

That takes us to the interview with Solana, where Jack more or less says all of this in expressing his desire to support decentralized social media protocols, first through Bluesky, and then later through nostr. As I said a few months back, even before Bluesky had launched, Jack had told me that he thought nostr was closer to my vision of decentralized protocols than Bluesky was going to be. And, as such, I was kind of surprised that he remained on the Bluesky board as long as he did. But his explanation here isn’t all that surprising.

He makes it clear that he looked on protocols as a better approach for all the reasons discussed above (and which Jack has spoken about multiple times before, including when he first announced the Bluesky project). He viewed it as a way of separating some of the tough moderation calls from being pointed back at a single CEO.

We were doing something similar to what we did at Square at the time, which was fund a bunch of open source developers to work on the Bitcoin protocol, because it directly benefited everything Square was doing in terms of money movement.

I wanted to do something similar with Twitter, because it was the only way to get out of a lot of the issues we were seeing around the decisions we had to make on accounts, and the pressures we had as a public company based entirely on a brand advertising model. The only way to do it was to remove the protocol layer from Twitter and make it something we didn’t control.

So what if we created a team that was independent to us, that built a protocol that Twitter could use, and then build on top of? Then we wouldn’t have the same liabilities, because the protocol would be an open standard, like HTTP or SMTP. Twitter would become the interface, and we could build a valuable business by competing to be the best view on top of this massive corpus of conversation that’s happening in real time.

So it took us about two years to interview people [who would build the protocol]. We actually looked at Nostr — I think the team even talked with fiatjaf [Nostr’s creator] — early on, but for whatever reason decided to pass. I wasn’t really privy to a lot of that conversation, or more likely, I wasn’t paying enough attention.

We eventually landed on Jay [Graber]. She seemed great, and we decided to fund her. Around that time, I was also planning my exit [from Twitter], and Parag [Agarwal] was going to take over. And when Elon made the offer to buy the company, I think she had this general fear of — what do we do? Like, is there any way that the funding could be taken back? We gave them $14 million to work on the protocol.

Again, all of that is accurate, and is completely consistent with everything Jack has said from the day he announced Bluesky. But, also, some of it is just unrealistic. Part of the nature of the Impossibility Theory is that no matter what, even in a decentralized system, it will be impossible to do moderation well, and people are still going to hunt down someone to blame. That is human nature.

Thus, I think part of Jack’s negative reaction to Bluesky was that he saw that the blame and demands were still coming. But that’s inevitable. That’s how this is always going to work, even in a decentralized system. He prefers nostr because he knows that while he’s still supporting nostr people know there’s no clear person to blame, and they know that they can’t yell at him about the moderation failures (and, yes, nostr has a ton of moderation failures, with spam and scams). But, as we’re seeing with nostr, it also creates some very real limitations, especially with regards to user adoption and growth.

Indeed, initially, Bluesky was supposed to be like that, but Twitter was still going to be the main service component above the original Bluesky protocol. Jack would still be getting the blame for running the service.

However, the Elon situation changed things.

The thing about my Protocols paper was that it was designed to convince someone like Jack to go down this kind of path. I believe in the power of protocols, but the challenge was always going to be how to get users to embrace such a system. The easiest path was to have a platform with an existing audience embrace it and bring the users.

The other possibility — building something brand new that was just so good people would flock to it — is just incredibly difficult.

There was a third possibility, though, that I definitely didn’t expect: Elon buying Twitter and repeatedly making a mess of things, driving users away in droves, causing many people to seek alternatives.

In this chaos, Bluesky became a sort of weird hybrid approach. It started out building for that first scenario (Twitter’s gonna bring the audience) and ended up having to do the second (build something new and hope the users come) all because of the third scenario (Elon bringing complete chaos to the ecosystem) suddenly opened up a new opportunity for the second scenario.

In other words, in the past, Bluesky was supposed to be the protocol, with Twitter being one platform using Bluesky’s protocol. But when Elon killed everything, Bluesky also had to step up and replace the Twitter part — the service part — itself, offering a reference app built on the protocol.

Reading Jack’s interview, that’s the part that made him disillusioned with Bluesky. Bluesky’s Jay Graber recognized, smartly, that having Bluesky set up as a public benefit corporation separate from Twitter enabled it to do some important things. This included pivoting to building out a service that could take people fleeing Twitter, while also setting it up to be (hopefully) more sustainable long term. It allowed Bluesky to neatly detach from Twitter right after Elon took over and canceled the contract that Twitter previously had with the Bluesky team.

It also created echoes of things Jack didn’t like. Just the fact that there was a “board” at all made him worried about the dysfunctional board that Twitter had, which caused him all sorts of problems. And as Bluesky launched and was growing, it faced some of those early speedrun issues, as users showed up and demanded specific moderation choices and tools (faster than Bluesky was able to build them).

But it was making those choices at the platform level, while continuing to build the underlying protocol.

That has created some real challenges for the Bluesky team. I believe in their original thinking, they weren’t going to have to deal with the most thorny moderation challenges directly. That would be passed off to the service level: Twitter (and then hopefully others who might embrace the protocol). But, in this world, Bluesky also became that layer.

And the fact that most users don’t separate out the protocol and the service layer meant that now Bluesky was under a lot more pressure as if the protocol and the service were one and the same (in part, because at least for the time being, they kinda are). Also, in building things out, I think Bluesky (probably correctly) realized that passing off moderation to the service layer rather than the protocol layer does not solve for all harms, and there are times when other solutions may need to come into play.

It seems all that just gave Jack flashbacks to all the problems at Twitter:

In Jay’s case, she decided she wanted to set up a completely different entity, a B Corp. That accelerated even more when Elon made the acquisition offer, and it very quickly turned into more of a survival thing, where she felt she needed to build a company, and build a model around it, get VCs into it, get a board, issue stock, and all these things. That was the first time I felt like, whoa, this isn’t going in a direction I’m really happy with, or that wasn’t the intention. This was supposed to be an open source protocol that Twitter could eventually utilize.

And then, as you know, Elon backed off [on the acquisition], and that disaster happened [laughs], until he finally bought it, which was the worst timeline ever. But throughout all that, it became more and more evident that Bluesky had a lot of great ideas. And they’re ideas I believe in. I think the internet needs a decentralized protocol for social media. I think Elon needs it. I think X needs it. I think it removes liability for the company, to separate those layers.

But what happened is, people started seeing Bluesky as something to run to, away from Twitter. It’s the thing that’s not Twitter, and therefore it’s great. And Bluesky saw this exodus of people from Twitter show up, and it was a very, very common crowd.

This tool was designed such that it had, you know, it was a base level protocol. It had a reference app on top. It was designed to be controlled by the people. I think the greatest idea — which we need — is an algorithm store, where you choose how you see all the conversations. But little by little, they started asking Jay and the team for moderation tools, and to kick people off. And unfortunately they followed through with it.

That was the second moment I thought, uh, nope. This is literally repeating all the mistakes we made as a company. This is not a protocol that’s truly decentralized. It’s another app. It’s another app that’s just kind of following in Twitter’s footsteps, but for a different part of the population.

Everything we wanted around decentralization, everything we wanted in terms of an open source protocol, suddenly became a company with VCs and a board. That’s not what I wanted, that’s not what I intended to help create.

So here’s where I think Jack’s understandable concerns about the very existence of a board and the pressures of being a corporation offering a social media service begin to diverge a bit from reality.

Bluesky has continued to build exactly what he has wanted. And it’s almost exactly what Jay promised when she interviewed for the job. I know this because, at Jack and Parag’s request, I sat in on a bunch of the interviews of the various people they were considering to lead Bluesky to provide feedback. And Jay’s pitch matches extraordinarily closely to what Bluesky has become, including the corporate setup.

Indeed, one of the things that struck me about Jay’s original pitch, unlike most others that I remember, was that she included a discussion of how the setup had to be sustainable on its own, and not just as a thing Twitter funded. That turned out to be prescient, but also what has made things work.

Bluesky has built a protocol, ATprotocol, which was built by folks who had experience with the same protocol that inspired Jack’s current favorite, nostr: Secure ScuttleButt. SSB was a really neat decentralized protocol, but it was nearly impossible for the average person to use (trust me, I tried, and I’m even more motivated than your average social media user). Bluesky’s original developer previously worked on SSB. nostr’s creator has admitted that he based many of the ideas on SSB, he just wanted it to work better.

And, while Jack talks up the need for an “algorithm store” where users get to choose their own algorithm, Bluesky has that! Its feeds solution is amazing and there are over 50,000 different feeds you can choose to give you views into the conversation. It’s great, with some upgrades coming soon to make it even more user friendly. Plus, they’ve added in the ability through its composable moderation tools for there to be more customized moderation offerings, which are already creating really unique and user-empowering offerings. So everything that Jack is saying he wanted from Bluesky is there.

It didn’t go in a different direction. It built the fundamentals, as promised.

But the one main difference was that, due to the implosion of Twitter, Bluesky also had to build the service layer. And Jack was really done with taking the blame for moderation decisions at a platform (even if he wasn’t the one actually making those calls). Being blamed for everything sucks. And that was starting to happen with Bluesky, which was not what Jack signed up for.

But it was necessary.

In order to get regular people to use it, Bluesky needs to have a user experience that feels like a centralized provider. One that feels familiar. That doesn’t require them to learn about the underlying infrastructure, or understand what “federated instances” means, or learn how to store a private key securely, or what the fuck a NIP-05 identifier is.

Without Twitter to provide that front end, it made total sense for Bluesky to build that. And, just as Jack had planned for Twitter to still provide an interface on top of Bluesky that promoted “conversational health” while allowing others to surface other aspects of the global conversation, Bluesky chose to do that as well.

Yes, some of that process was messy, especially with some of the demands from users at a very early stage, when the company had limited staff and resources. And, early on, I think the Bluesky team had to come to terms with the fact that it was now taking on that side of things as well, which wasn’t intended. But what has continued to impress me in watching Bluesky as an outsider over the past year and a half, is how true it has stayed to its underlying vision, while also still trying to make its own app-layer usable by people who will never care about the decentralized protocol.

But the interesting thing about where Bluesky has gone over the past year is that, now that it’s building both the protocol and the service layer, the team there is actually thinking deeply about how moderation can work effectively in such a world. How it can use this different structure to actually look for ways to minimize the very real harms that happen in internet communities, but without being as heavy handed and all controlling as a centralized service would be. Jack was trying to offload that because he didn’t like taking the blame for it, but that doesn’t mean the harms aren’t real. And some people do need to think about how to try to minimize them. And to do that in a way that still builds a platform people want to use without having to worry about all the details.

I think this is important. Some (including many people on nostr) argue that users need to understand the power of a decentralized protocol to embrace it, but I disagree. Even if most of the users of a decentralized system don’t know or care about the fact that it’s decentralized, the fact that the underlying protocol is that way and is set up such that others can build and provide services (algorithms, moderation services, interfaces, etc.) means that Bluesky itself has strong, built-in incentives to not enshittify the service.

In some ways, Bluesky is building in the natural antidote to the activist investors that so vexed Jack at Twitter. Bluesky can simply point out that going down the enshittification path of greater and greater user extraction/worsening service just opens up someone else to step in and provide a better competing service on the same protocol. Having it be on the same protocol removes the switching costs that centralized enshittified services rely on to keep users from leaving, allowing them to enshittify. The underlying protocol that Bluesky is built on is a kind of commitment device. The company (and, in large part, its CEO Jay) is going to face tremendous pressures to make Bluesky worse.

But by committing to an open protocol they’re building, it creates a world that makes it much harder to force the company down that path. That doesn’t mean there won’t still be difficult to impossible choices to make. Because there will be. But the protocol is still there.

And that’s why, even as Jack namechecks my paper here, I think he’s wrong in the conclusion of this paragraph, saying that Bluesky went in another direction. It didn’t:

All that said, I really respect Jay. She was under a lot of pressure to survive and do the things that she did. But directionally, I just don’t align with it. And I’d love to see more effort placed on open protocols akin to Nostr, which hits every single attribute that I was searching for when we originally kicked this idea off. If you go back to my thread, and Mike Masnick’s Protocols, Not Platforms article, it hits every single one of those things, whereas Bluesky ultimately just went another direction.

Bluesky went in the same direction it planned. But it was forced to add on another layer — the service layer that Twitter was supposed to provide — and that was the part that Jack was already sick of.

I like nostr as well. I think it’s cool and has some really cool development happening. Some of the new services that have popped up using nostr are great. But Bluesky has an open protocol, ATprotocol, which actually has many similarities to nostr, but in a manner that hides the technical complexities from users, making it more approachable by the average user.

That’s important! Because the coolest, most elegant protocol in the world is useless without a userbase. And so far, Bluesky is set up in a way that “normie” users can just use it without caring about all of these details. And that has made a difference that even folks endorsing nostr have seen.

Up above, Jack bemoans the “common crowd” that found and enjoys Bluesky, but that’s who you need to build for if you’re not just building a small clubhouse for the technically savvy.

It’s entirely possible that someone will build a nostr client or service that does something similar. I hope that happens, honestly, because I think it would be a good thing. But right now it’s just not there. And if Bluesky followed a similar path, it would just be one of a list of rudderless protocols like nostr, farcaster, and a variety of others that people never remember.

There are some cool things being built on nostr and farcaster (and ActivityPub and Bluesky). I don’t think anyone needs to be particularly tribal about these things. But I think Jack is overcorrecting for his negative experience at Twitter. He thought Bluesky’s role was just to build an open protocol. But the Elon situation necessitated also building a reference app on top of it, which is actually helping to drive the entire ecosystem forward.

There’s a lot more worth commenting on in the interview, but this piece is getting long enough already. I think Jack’s explanation of why he’s embraced nostr (and pushed Bluesky originally) still makes sense, especially given the situation he was in over the last decade or so. Even as many people seem to want to dismiss what he’s saying, he’s right that the approach he’s pushing would be the best for an open internet instead of one controlled by just a few internet giants.

He’s also correct (as he notes later on in the piece) that the traditional advertising business model creates some difficult pressures for companies. He hoped that Elon taking Twitter private would help deal with those, but I think he overlooks how badly Elon miscalculated in creating those new business models.

What he’s missing is that Bluesky is, in fact, building exactly the kind of solution he wants (including one that is exploring other, better, business models). It’s just that they’re building it in a way that the underlying protocol issues aren’t important to the everyday user, but are still there if they ever want to go deep and explore.

That means the incentive structure for a better system is there. The risk of enshittification is diminished. The ability for users to “choose their own algorithm” is there. The ability for others to build their own interfaces, and algorithms, and moderation services, and more is there. It’s all there. Every bit of it.

Yes, it also has a board and venture backing. It also has all sorts of people using it, beyond a crew who revels in the technology infrastructure, which remains a niche audience. Bluesky remains the closest approximation I’ve seen to what I hoped for in my paper. It’s disappointing that Jack doesn’t see it, but I’m glad he kicked it off, and I’m glad that he is still pushing for decentralized social media protocols via nostr.

I’ve seen some people worry that Jack completely disconnecting from Bluesky harms the project, but I don’t think so. It’s true that Jack’s name helped generate a lot of the initial interest (and media coverage). But Jack has been checked out of Bluesky for at least a year (and really a year and a half).

If Bluesky is going to succeed, it needs to shine on its own, separate from Jack. Having some people (falsely) think it was “Jack’s new site” may have generated some initial interest (and some initial backlash!), but Jack publicly cutting the cord means that Bluesky now gets to succeed or fail on its own terms, out of Jack’s shadow. And, for a true decentralized protocol to succeed, that’s probably a good thing.

Filed Under: decentralization, jack dorsey, jay graber, nostr, protocols, protocols not platforms, social media
Companies: bluesky, twitter, x