‘Speech should be open to all’: Bluesky COO Rose Wang on the platform’s radical vision for social media

Conceived of by Twitter cofounder Jack Dorsey in 2019 and spun out of Twitter in 2022, just before Elon Musk took over, Bluesky came out of beta in February 2024 and found an eager, if relatively small audience. The social media platform—which, at first glance, looks remarkably similar to X—had amassed roughly 13 million users by the end of October. Then Donald Trump was reelected, with an assist from Elon Musk. Now, the site has more than 21 million users.

Bluesky is built on a decentralized protocol, which allows it to offer users appealing new features and flexibility—and offer the world a compelling alternative to other social media platforms, which are defined by their opaque and all-powerful algorithms. While Musk has complete control over X, which he appears to have used to Trump’s advantage, Bluesky hands that power over to users and outside developers.

Users have nearly two dozen options to control their content moderation settings, with the ability to hide—or see—everything from posts with rude language to ones that contain explicit sexual content or threats of violence. They can mute certain words or topics, and hide replies and reposts from their feeds. Just as significantly, unlike on X and Facebook, the posts that show up on Bluesky’s are not dictated by an all-powerful algorithm run by a centralized group of executives. Outside developers can make their own algorithmic feeds and moderation settings, and offer them to users. 

We spoke to Bluesky’s chief operating officer Rose Wang, who has been with the company since 2021, about the company’s radical vision for content moderation and control of the feed, how the platform works with developers, and what the past two weeks of hypergrowth have been like.

Bluesky has surpassed 20 million users in the span of about two weeks. That’s more than double the users it had just a couple months ago. Were you ready for that influx?

I think “ready” means different things. We were ready on the infrastructure side. Were we ready as a team to grow a million users a day last week? I think that mentally and emotionally we were just trying to keep up with what was happening. We had all these plans to launch different features. All those plans and all my projections are basically out the door. In a moment of hyperscale, you actually just need every single person to be rowing in the same direction at the same pace.

How big is the team?

The team is about 20 people.

So you have a million users per person.

Somebody was like, “Have we thought about feudalism? 20 million users, 1 million users per team member?” But [seriously] it’s been really exciting for us and I think [right now] the only goal we have is keeping the website online.

Do you think that if Trump had lost the election, Bluesky would have had the same growth?

It’s hard to predict a counterfactual. What I would say is that I think the reason why people are coming to Bluesky is because we saw overnight how a platform that’s controlled by a small group of people can basically change its culture and change its bipartisanship overnight. Before [it became] X, Twitter was a global platform for breaking news and conversation no matter who you were. Over time and very quickly, that changed to where it became a megaphone for one party. I think the influx of new users is actually people pushing against that more than anything else. This space [X] that we thought belonged to us, we realized actually doesn’t. We don’t have much power and the only way we can be heard is to petition the decision makers. If the decision makers don’t agree with you, then you really don’t have much choice. 

That’s why I think people are flocking to Bluesky. They’re like, wait, this is basically a world that we don’t want to live in offline. Why would we want [to live in] that world online? People are coming to Bluesky because it’s much more about user choice. 

Bluesky’s interface reminds me a lot of X. Is that intentional?

Let me tell you the story of Bluesky because that might explain why it looks like Twitter on the front end. We were a company that was kicked off by Jack Dorsey in a tweet in 2019 where he said, “I want to put Twitter on an open protocol.” What he really wanted to do was turn Twitter into [what Bluesky is], where people had choice. [They] could choose their feeds and could own their identities. The way to do that though was that you actually had to build the underlying plumbing, the protocol for that to happen. So we were a company set up to build that open protocol for Twitter as our customer. What happened though, is that when Elon Musk decided to buy Twitter, there was no longer that goal. I believe back then his goal was profitability. So Bluesky spun out to become its own platform.

When a new product arises, there’s only a certain amount of novelty budget you can spend for users to learn something new. We had to decide where to spend that novelty budget. And through the familiarity of the interface, we can introduce new concepts like multiple feeds, and different moderation services. So in that way it was an intentional choice to make it look a little bit more like Twitter. It may look the same, but the underlying protocols are totally different.

Does the open protocol you’ve built give you different functionality than X?

This is the beginning. This is us introducing choice through one app. But in the future we’re going to break out of the app and there’s going to be lots of other experiences in the Bluesky ecosystem. 

One of the things that already can be done is you can take the comment section of your Bluesky post and use that as your WordPress blog comment section. That is already taking one social experience and breaking it out of the box. Why do you want that? Because most comment sections of blogs are cesspools of toxic posts because they’re unmoderated. But in your Bluesky feed, everyone that we’ve blocked or you’ve blocked in Bluesky automatically gets blocked in your comments section. All the moderation work that we’ve done automatically gets transferred. This is now thinking about social outside of these four walls. That’s very much what Bluesky is building. Lots of these elements like algorithms and moderation can be split up.

What does decentralization mean to Bluesky?

When we talk about these traditional social companies, they’re actually these centralized systems where you have a small group of people who decide who gets to be on there, what post gets to be on there. 

Bluesky is not just owned by us. We don’t control the server. It means anything we can do, anyone else can do because there’s not a centralized governance structure. Anyone else can build a feed, anyone else can build a moderation service. 

How many custom feeds made by outside developers exist on Blusky at the moment?

There are about 50,000 feeds on Bluesky. I think our team has created maybe a few hundred and the rest is just user created. The feeds on Bluesky are like browser tabs.

Users can customize their feed to block certain terms or people. Does Bluesky do any type of moderation beyond that?

We absolutely moderate. The way that we think about moderation is similar to governance, which is like there’s multiple layers, not just one system or one layer that’s governing. It’s just like in the U.S. where you have the Supreme Court, you have federal courts, and then you have local courts. We want to set that type of system up, which is giving people more control over their local jurisdictions. Our terms of service and community guidelines on the Bluesky app is the Supreme Court. That’s governed by our head of trust and safety, Aaron Rodericks, who used to be the head of election integrity at Twitter. There is no tolerance of hate speech or misinformation, for example.

There can be edge cases though. How do you deal with those?

There’s other types of behavior, like rudeness, AI-generated images, or movie spoilers. Policies on social media weren’t fast enough to keep up with what was happening on the ground. So we took a page from the Reddit playbook and gave community moderators some tools so that they can govern their own spaces. [On Reddit], most of those tools are quite manual and you can’t really transfer them from one subreddit to another. 

The way that we’ve made it even more powerful is that we have given folks programmatic usage of moderation tools. [A user] can actually, across the board, look at posts that have movie spoilers and label that for the entire network. Then you, as a user, can say “I don’t want to see movie spoilers,” and you can subscribe to that label and hide those posts. The types of tools that we’ve given to everyday users can cover edge cases or the needs of diverse communities that a centralized organization just doesn’t have the speed or efficiency to get to.

Is Aaron Rodericks the person who took Trump off the platform?

It wasn’t one person. I don’t know who made all the decisions, but it was a series of folks. And then Jack [Dorsey] talked about the fact that it’s not up to the platforms to make these decisions, which is why he wanted an open protocol. We don’t necessarily get to decide who gets to be online or not. That’s not really our job. Our job is more deciding reachability how far their post can go. So it’s really about freedom of speech with a limit of reach. 

How do you limit reach?

We don’t limit speech as long as it doesn’t violate the laws of the country. But if it violates our terms of service, we won’t connect with you and those posts won’t come up [but other apps on the Bluesky protocol might surface them]. It’s a lot like how the web works where there’s probably a lot of dark corners of the web, but Google doesn’t index them so you can’t search them.

If I posted something offensive right now, how would it get flagged?

This is where AI is extremely useful at the moderation level. We have triggers and we’re catching those at publish. Then after that, we have users report different posts or accounts. That’s just the central moderation that is standard across most social apps. But then we have what we call moderation labelers [third-party services that can tag accounts or content, either for informational purposes or for moderation] that any user can go create. 

Do you see people posting different content on Bluesky than they do on Twitter?

The vibe is very different. And the reason I say that is first of all, on Twitter it’s something like 1% of users were posting most of the content. On Bluesky, it’s around 30%. When the Brazilians came on [to Bluesky] after Twitter pulled out of the country, it was like 50% of our users were posting. The reputation we’re gaining is that you can have a thoughtful conversation on the platform. Part of the reason that’s happening is that starting over is a good thing.

Do you think it can maintain that kind of culture as it grows so fast?

That’s part of the reason we’re really glad that we were able to grow more slowly over the last year through beta because we were able to develop and nurture a set of power users who can help evangelize and help us really tell the culture and reinforce the culture. But that’s also partly where the moderation labels come from. On Bluesky, you can actually label a post or an account as rude. [Being rude] doesn’t go against your terms of service or community guidelines, but if there’s actually an external consequence for you saying something rude and then you getting that label, it might deter you.

You raised a $15 million Series A in October. I have to believe you’re burning through that pretty quickly. Do you plan on raising more?

All my cost projections have kind of flown out the door. When you have a million users joining every day for seven days, costs go up, especially infrastructure costs. Right now, what’s very important for us is that we have global moderation coverage 24/7. Spending has gone up in those areas significantly. 

That said, we announced that we’re launching a subscription model most likely before the end of the year, and so that’ll be our first monetization [push]. We will never put core functionalities like speech behind a paywall. Speech should be open to all. It’s more like offering custom aesthetics or more video uploads or higher resolution images that users will be able to pay for. Subscriptions will help us a ton and give us a little bit more time. But even though it is going faster, $15 million is still quite a bit of money. 

I was sad when I lost my blue check on Twitter, and I won’t pay for one now. How do you approach verification?

When the option is check mark or no check mark, it’s binary: It creates a hierarchy where you’re either cool or you’re not.

What we want to do is open up the system in a way that’s a lot more equitable—and do it in a way that doesn’t require a central verification system. If you are impersonating someone, we label that so there’s a negative signal there. Then instead of using handles the way that most other platforms do, we allow users to use their own domains because you own your domain. So this whole idea of going to a new platform, trying to reserve a handle that is controlled by the platform is no longer there. You’ll see that The New York Times is @newyorktimes.com, so we don’t need to verify because only the New York Times can have that domain

Do you have a feed that’s a guilty pleasure?

One of my favorite feeds is the Moss Feed. It’s literally pictures of moss because sometimes you just want a calm feed.

https://www.fastcompany.com/91233939/bluesky-rose-wang-radical-vision-social-media-twitter-alternative?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss