How Bluesky Is Seizing the Opportunity of Twitter’s Decline With Custom Algorithms

How Bluesky Is Seizing the Opportunity of Twitter’s Decline With Custom Algorithms

Twitter is in freefall — ad revenue is down more than half since last year, owner Elon Musk is mucking around and firing people left and right, and users (including me) have noticed a steep decline in the quality of their feeds. Bluesky, a near-clone of Twitter built on a decentralized protocol with the support of Twitter co-founder Jack Dorsey, is ascending as its forebear declines.

The company is gaining users, but not too many — a new user can only join the social network with an invite code from an existing user. It’s also innovating where Twitter failed, building in bulwarks against toxicity from the start.

To get a better understanding of how Bluesky is approaching this unique moment, Gizmodo spoke to Paul Frazee, a product developer and protocol engineer at the small company. Frazee has hosted livestreams of his coding as he builds what could be Bluesky’s differentiating feature — custom algorithms

This interview has been lightly edited for length and clarity.

Gizmodo: Where did the idea for custom algorithms come from? They seem to be a realisation of Bluesky’s decentralized ethos — is that right?

Paul Frazee: It is right that it’s a major realisation of Bluesky’s decentralized ethos. The team knew pretty early on in the development of the protocol that we wanted to have custom algorithms. Decentralization through federating and hosting is interesting, but often it’s not actually meaningful to users, so really early on we started poking at what we could substantively do that would change the experience of social media and actually decentralize the parts that matter to people.

Algorithmic choice isn’t a new idea, but we aimed for it early on. It’s an example of why decentralization is better in practice too — because it can do stuff like this.

G: How do the custom feeds/algorithms work? I know an early version has debuted on the social network, how is that test going?

PF: The test is going really well. The way they work now is fundamentally how they’re going to work in the long run. It’ll get more sophisticated over time but the concepts are all there and working how they should.

We have a motto that “third-party is first-party,” so there could be these third-party services that feel like they’re natural and just a part of the application. A majority of the custom feeds on the app now are completely off of our servers and they’re on someone else’s infrastructure. We just talk to their server and then we show the feed in the user interface. That’s a pattern that we hope to repeat in a lot of different use cases, and algorithms are just the first.

How are you envisioning Bluesky’s custom algos as being different from Twitter’s?

The biggest difference is that most social media companies right now try to intuit what users prefer based on their behaviour, and we’re instead offering an open marketplace of options. You have more of a direct hand in what you’re getting as an end user, as opposed to having to figure out how to train the algorithm. There are pros and cons, because sometimes the algorithms that intuit what people want can be really great at it. But other times, you sit there thinking how long you can look at something before you mistrain the algorithm. Like, you might be like, “Oh no I’ve watched this video for too long and now it thinks that I’m into building sand castles,” when no, that was just one video you wanted to watch, and now you have algorithm anxiety.

We think that not only is the opportunity for choice interesting, but having a completely open marketplace just makes it possible for people to come up with entirely new ideas for how algorithms can be done over time.

How is Bluesky making these feeds less toxic than those of other social media companies?

I’d like to think that custom algorithms would fundamentally push things to be less toxic but I also don’t really think that necessarily follows. It could, but I just don’t know — that’s an untested theory. I think that the opportunity for less toxicity will probably come more from the way that we apply open moderation where people can opt-in to filters on top of the algorithms. That will give users more control over what vibe they’re tuned into at the moment.

What opportunities for innovation do you see in Twitter’s continued decline? Seems like the current state of the industry is akin to when a big tree falls in the jungle.

The opportunity for innovation was there even if Twitter stayed the exact same as before. I’m pleased that there is more interest in an alternative than before, but we always, from the beginning, never assumed that would be the case. We sat down and said, “How do we make a better experience than what Twitter did?” And the real opportunity is in creating something that’s a fundamentally more positive experience. You get more of what you want, less of what you don’t want, and it’s less emotionally chaotic. That more pleasant experience is the most interesting thing to me.

It’s weird that we have this love/hate relationship with social media. Sometimes we love it, and sometimes we’re like, “Hey, maybe this really sucks.” That suggests that there’s something fundamentally unhealthy with the structure. So to me, all of the innovations that are possible all need to be pointed at making something that makes people happier and more connected with people and harmonious experiences. And that just needs to be the goal.

Where does Bluesky see opportunities for innovation outside of the custom algos feature? Opportunities for growth outside of Twitter’s defecting user base?

There seems to be an opportunity for technology that is able to handle a wider variety of social use cases, but we don’t know for sure because it’s an untested theory. For example, it seems feasible that this could run the gamut of the FB or Instagram use case. Certainly getting podcasting or direct messaging in there is all possible. It’s a strongly compelling idea to have this all exist in one shared system where things like identity and reputation can easily transfer over to another social experience. There’s some interesting opportunity to do really the exact opposite of what all the other platforms are doing – this is actually an open space for lots of things to happen.

Twitter has never been for everybody. People who love Twitter don’t love it for its toxicity, so I want to make it better for people who love the product. And then for the people who always bounced on it, I would like to find what was missing for them and make it something they enjoy. It may just be that the public performative experience of microblogging is not for everybody. For them, it might be more in the DMs or private groups, but there’s no reason why the plurality of these social experiences can’t be brought together.

And one last question, a bit more fun than serious, that we’re asking everyone in the series: Are the following five things overhyped, under-hyped, or appropriately hyped, and why?

AI chatbots

Appropriately hyped. Clearly they’re doing things that didn’t seem possible not that long ago and they’re showing at least some use cases for people right away. I feel like it’s a good bet that they’ll continue to find interesting things they can do with it, so I think the hype is exactly right.


If you asked me two years ago, I’d say overhyped, now I’d say appropriately hyped. The problem with cryptocurrencies is they don’t scale, they’ve never been able to function as a payments mechanism because the transaction fees were too high and the settlement time on some of the networks was too long. If those weren’t problems, then the UX would have been an issue. If the tech is not able to find a price point that’s competitive with other payment systems, then it’s a useless technology.


I don’t have a strong vibe on drones.

Portugal (as destination for digital nomads working remotely, though the honeymoon does seem to be ending)

I wasn’t super tuned into the Lisbon thing. It seemed to be pretty tied to crypto. I don’t really know.

Embracing failure

The problem with embracing failure is that it almost poses failure like a goal. The point is to learn from it. I do post-mortems on projects and I publish what I learn. I’m very focused on making sure I come away from a failure having learned multiple things and not repeating those mistakes. In the same way that we should be open to receiving criticism, we should be open to learning lessons from failure. But if you’re not engaging with a project with the goal of a win condition, then you’re wasting your time and everyone else’s time. You should embrace failure to be more effective as a person and as a force in the world.

The Cheapest NBN 50 Plans

It’s the most popular NBN speed in Australia for a reason. Here are the cheapest plans available.

At Gizmodo, we independently select and write about stuff we love and think you'll like too. We have affiliate and advertising partnerships, which means we may collect a share of sales or other compensation from the links on this page. BTW – prices are accurate and items in stock at the time of posting.