Twitter recently announced that it’s going to be building a lot of new things into its product: there’s Twitter Spaces, which is a live audio feature very similar to the red-hot Clubhouse app. There are Twitter Fleets, which look like Snap and Instagram Stories. Twitter just acquired Revue, which is a newsletter product similar to Substack. And it also announced something called Super Follows, which will let people pay creators on Twitter for special content.
It’s a lot, especially after Twitter spent so long seemingly not adding any features at all.
On today’s episode of Decoder, I talk with Kayvon Beykpour, the head of consumer product at Twitter. He’s responsible for deciding what tools Twitter actually builds for people to express themselves. Twitter’s product is incredibly important — it is a flashpoint of interest and controversy for politicians and regulators around the world, and it often seems like politicians in our own government don’t know the difference between Twitter and real life.
Beykpour and I talked a lot about what it took to reset the team toward growth, how he decides what to prioritize, and what the timelines for success look like on different projects. And of course, we also talked about paying for tweets and competing with Clubhouse.
Okay, Kayvon Beykpour, head of consumer product at Twitter. Here we go.
It’s been a while since we chatted. I’m excited to have you on the show.
Almost two years. It was October 2019 the last time [we talked]. Me, you, and Casey.
Yeah, it was me and Casey Newton. I think you had just become the head of product at Twitter. You’ve now made it almost three years. That might be the longest anyone has been the head of product at Twitter, so congratulations.
I feel like I’ve been in the role long enough that people don’t remind me every week how long it’s been since I’ve been the head of product.
There was a sign at Twitter HQ that was “days since last head of product,” but you’ve been there a long time.
The last time we spoke, we were talking about accelerating Twitter’s product cadence, doing all this stuff. [Then] there was a pandemic. But just a couple of weeks ago, you had an analyst event, [with] lots of product announcements coming out of Twitter. Walk me through just the highlights of that, and then I really want to talk about how you decided to prioritize what to build and how you accelerated the product roadmap, because it feels like it’s going a lot faster than it was before.
So, while we’ve definitely picked up the pace, and we announced some new stuff at our Analyst Day, and also reminded folks and showed a longer preview of what they can expect, all the stuff we announced fall into high-level categories of work that have been focus areas for us for the last couple years.
I think even when we chatted, we had a few high-level areas of our product strategy that we’re focused on. One is health: how do we protect the health of the public conversation? The other is conversations: how do we incentivize people and create the tools and capabilities to inspire people to start and participate in conversations on the platform? And the other is what we call interests: how do we connect people to the people and content that they’re interested in? [That’s] fundamentally the main reason why people have been coming to Twitter for the last 15 years, but [it’s] really critical for us to build new, powerful ways for people to do that.
So health, conversations, and interests have been our big rocks for the last two and a half years. We’ve been taking bigger and bigger swings in each of those areas. At our Analyst Day event, we went in-depth on a few of them. Within “interest,” for example, last year we launched a product called Topics, which we got started — very nascent with our work there, but we’ve really accelerated. Today there’s 6,000 topics that people can follow. And it’s very simple. Rather than just following people on Twitter, you can follow a specific topic, and Twitter does the work of recommending the best content or tweets about that topic. So you don’t have to know exactly who to search for.
Again, we didn’t release that at our Analyst Day event, we just gave an update on it and shared some of the substantial progress. In Q3 of this last year, we announced that there are 70 million people that have followed topics. And then just yesterday, we announced that there now are over 100 million people that have followed topics. So a pretty good clip of growth. And we’re seeing really promising signs that Topics is a really useful way for people to connect to their interests.
A couple other things we announced at our Analyst Day events: one was a product that we’re really excited about right now called Spaces, which again is very connected to our work in conversations. Spaces is a new way for people to talk about what’s happening on Twitter, except instead of using 140 characters or 280 characters or a video, you’re using your human voice to connect with other people and have real, audio-based conversations. So: a new effort, but connected to a longstanding priority.
Some newer things that we announced that people hadn’t heard from us include Super Follows. So I’m assuming you might want to talk about this, but this is one that we hadn’t really talked to the world about, or been particularly transparent about, until our Analyst Day event. And we’re excited about that one because it’s really starting to stitch together a bunch of the other capabilities we’re working on.
I won’t go into too much depth until you ask me about it, but that’s probably one of the newest things that we announced for the first time at Analyst Day. And then the rest of it was really just telling a cohesive story to the world around — what is our strategy? Why is that our strategy, and what are some of the product investments we’re working on that ladder into each of them?
I absolutely am going to ask you about Super Follows, but let me try to recast the three areas of focus that you’re describing.
So you have health, which to me, is basically just the overwhelming content moderation problem that every social platform faces. And you’ve got to build tools and strategies there. You have conversations, which is: you need to make a bunch of tools to let people create, [to] actually make content for the platform, as opposed to just consuming it. And then you have interests, which is: find the stuff that other people made that you might like.
Every social platform has that set of problems. I want to take them in turn. A lot of the features you have described that are most interesting to me are [related to] how do you incentivize people to create media. To me, the great story of the internet is that we put a camera and microphone in everybody’s hand, [and] everybody has become a filmmaker. Everybody has become a storyteller. Twitter is one of the best distribution networks for that kind of work.
But I’m looking at something like Spaces, for example. We’ll just start there. It’s an interaction framework that looks a lot like Clubhouse. People are making that comparison very directly. Fleets is now at the top of my timeline. There is the wave of Stories products that hit Snapchat [and] Instagram. Fleets looks a lot like Stories. How are you deciding which of those interaction paradigms are fundamentals that you want to bring to Twitter as a fundamental? And which ones are, “oh man, that’s a great feature. We’ve got to get there first, or fast follow to make sure that a different rival social graph doesn’t form that takes our audience away.”
It’s a good question. And contrary to [what] popular belief might be, we try to never approach it from the standpoint of “here’s a capability, let’s copy it.” We try and approach everything from first principles and, really, customer understanding.
One of the things that [Twitter CEO] Jack [Dorsey] really emphasized when he came back to the company, was morphing our product development process to respect the “jobs to be done” framework a lot more. I don’t know how much you’ve heard about “jobs to be done,” but fundamentally, it just comes down to imbuing customer understanding into the product development process. And so for us, all of the work we do starts from the standpoint of: what are our customers trying to hire us for? What are they firing us for? And how do we build product solutions that have those dimensions in mind?
Take something like Fleets, for example. We didn’t start it from the standpoint of, let’s copy Stories. We started it from the standpoint of “why are people not tweeting?” It turns out that some of the reasons they’re not tweeting is they don’t feel safe. They don’t feel safe because what they tweet is subject to public scrutiny. Tweets are on the public record, which is terrifying. Anyone can respond to them, and they’re such a popularity contest of likes and retweets and impressions and all of the social mechanics that we’ve built in the product that actually work quite well for a certain purpose, but can work against you if you’re just trying to have a conversation and feel intimidated by that. And so we go through and enumerate all the reasons why someone might be hiring us or why someone might be firing us.
In the case of tweets, this is a very common observation we have. A lot of people are terrified to tweet. A lot of people don’t use Twitter for the creation side at all. They’re here to consume. Unlocking that, finding, peeling the onion back and understanding the reasons is how we start, period. And Fleets felt like one — of many — but one product solution that can address some slice of that problem statement. It’s ephemeral, and it takes away the public replies. It takes away all the engagements. And to your point, there happens to be a format for that, that customers are familiar with. So we won’t be afraid of leveraging prior art in that sense, but we would be flying blind if we just started willy-nilly copying stuff. That’s not how we’re approaching it.
With Spaces, it’s interesting because… fundamentally, I think people for the last 15 years have been using Twitter to talk, and to talk about their interests. It’s just so basic, but in a way, I think the rise of this audio renaissance that’s happening right now is interesting because it’s taking technology that has fundamentally existed for quite some time and putting a user experience around it and a fidelity around it that allows people to engage in that same job of just having serendipitous conversations with people, but doing it in a way that is synchronous rather than asynchronous, and powered by the human voice rather than text.
And this is particularly a special and passionate area for me because it’s not too different from how we were approaching Periscope. The interesting thing is that when we started Periscope, our goal was to build the closest thing to teleportation, like, [to] see through someone else’s eyes. What we found is that 99 percent of the usage was not someone broadcasting what’s happening on the streets of Istanbul during a protest. It’s people hanging out and talking.
And so, really taking that to heart and building an experience within Twitter that can be a layer of fabric that lets people talk with their voice; that’s how we’re approaching it. There’s some mechanics that we’ve been inspired from other startups that are doing this, and then there’s some things that we’re making our own. That, fundamentally, the way we build Spaces in our product that will allow it to shine. It’s going to look and feel very different than how some other platform will be able to tackle it. So therein lies the challenge and the opportunity for building a great product.
One of the things that has really struck me is this is a lot of new products, a lot of new challenges. What were the top three things that you had to do in your role as head of consumer product, to get to this velocity of product development and feature announcement?
This won’t be in any particular order, but when I joined the company, one of the first things that I felt was a reticence and an uneasiness around taking big bets. And I think there’s a lot of reasons for that. It’s hard to pinpoint one. Certainly, churn in leadership is one. After a revolving door of heads of product, people stop taking any product strategy particularly seriously, because it’s like, okay, well, let’s wait until that strategy changes. And so I think there was a little bit… while no one would say that explicitly, there was this reticence to commit to long-term speculative bets, because it was rare for them to be able to be seen through.
So that’s one that I think was somewhat ingrained in the culture. Which is difficult. It’s difficult for PMs, engineers, designers who really want to push and evolve the product to come up against an organizational resistance that is not tuned to take big, speculative bets.
And really, unwinding that, I think, has been the biggest unlock that we’ve had as a company. Today, when we contemplate solving more ambitious customer problems and in turn, postulating more ambitious product solutions, we don’t get the “no, but” as much. Every once in a while, there’s pessimism around, “oh, can we pull that off?” But we get way more of a “yes, and” vibe, and a willingness, and a patience for terrifying, ambitious bets. Whereas three years ago, any idea you would come up with, there was just a lot of pessimism around — that’s going to take a long time to build. Which is true, we’ve got a lot of infrastructure debt to work through. And also like: that’ll never ship, that would never get approved, our customers will freak out.
Which by the way, related to your question, one of the interesting things about Twitter, which is both a great thing and can be a challenging thing sometimes, is that people use our product to love and hate on our product. And I think it’s one of the most amazing things about Twitter. We get the gift of feedback on a minute-by-minute basis. And that includes getting shit from our customers when we get it wrong. And that can make it very intimidating and scary for teams to build stuff. Because God forbid, “rest in peace, Twitter” trends. And we’ve had a lot of launches end with #RestInPeaceTwitter trending. Which I think, a few years ago, was something that would stop us from taking swings in the first place.
In my opinion, if we’re not seeing #RestInPeaceTwitter trend a few times a year, we’re not taking big enough swings. And so just as a team, shifting our mindset to get comfortable with that and lean into it and embrace it, has been a really important challenge for us to overcome. We’re not completely over the fence yet — it’s amazing that customers are starting to notice that we’re taking bigger swings, but I wouldn’t say we’re completely out of the woods yet. It’s an ongoing journey that we’re really, really focused on.
Aside from that, hiring. Any leader’s job is to hire a great team. And we’ve really reshaped our leadership bench, across not just product, but design and research and engineering. We have new leaders who are filled with the energy and passion of making Twitter live up to its potential. And we think that the potential is so much higher than what the product is today. The product is obviously extremely important and plays a critical role in the world, but there’s still so much potential. I think it takes a collective team to have that irrational passion, to go through a lot of the slog that sometimes you have to go through to get there.
One of the things I always think about is leadership timelines. I’ll give a really silly example. If you’re a news writer at The Verge, your decision-making timeline is 30 minutes. Some news happens, you write it, you put a headline on it, you get it out in the world, you move on to the next thing. You’re operating in those increments. As I’ve moved up in my career, my decision-making timeline has extended into a year, two years, in some cases.
You’re describing a two-year-plus process to reboot a product culture. Is that the right timeline? Is that what we should always expect? Is that the timeline you were expecting, or did you have to make some changes along the way to get to where you are now?
I can’t say I had much precision in terms of how much time I would expect it would take. There’s a lot that I thought I knew about the job going into the job, because I’d been at Twitter for three years and change prior to having this role, which gave me a lot of context in terms of our strengths, our weaknesses, our opportunities. I still feel like I’m learning on the job. I generally am an impatient person. I think it’s one of my best and worst qualities. So I try to be constantly dissatisfied about everything about our work and pace, while at the same time being able to appreciate progress that we’ve made. I’m really proud of how much we’ve morphed the culture, but I still think we have a lot to do.
I think in a lot of these areas, when you talk about speed, our ability to identify a problem we want to solve, and actually ship a product solution — which I think is, ultimately, the right end-to-end timestamp to look at — there’s a lot more than just culture and process and people. It’s also infrastructure and technology. At our scale, we need to have really robust systems and a modern technology stack that allows us to build on. And this is an area where I think we’re still really investing in. We’re not out of the woods.
I don’t know how much of the Analyst Day you watched, but this is one of our top company objectives, investing in our own development velocity, making a huge, multibillion-dollar commitment in leveraging public cloud solutions through AWS [Amazon Web Services] and GCP [Google Cloud Platform]. This is its own multi-year journey that I think, fundamentally, will help us unlock even more velocity when you combine it with the cultural changes, and having the clarity of strategy of knowing what we actually want to build. So it’s going to take a village.
So, I think my expectation of [the] time horizon is constantly modulating. In areas of culture and people, we need to move as quickly as possible, and in areas of execution where we have clarity in our strategy, we’re trying to think in weeks, not months or quarters. And then in other areas of a sort of broad scale re-architecture, for lack of a better word, we do have to think in years. It’s like, we’re thinking about juggling on-premise data centers with public cloud migrations, and these things require quarters, if not years, worth of long-term thinking.
So this leads me into the question that I ask almost every executive who comes on the show. What is your decision-making framework, and how has it evolved as you’ve had this role?
I think probably the biggest point of evolution for me has been knowing when not to be involved in a decision. It’s something that I’ve personally had to grow quite a bit on, just because I’ve run a lot of teams, I’ve started two companies, and those companies have grown to be somewhat sizable, but never at the scale that I’m operating now. And so it’s always like any leader is challenged, to figure out when to be in the weeds and when to hire great people and defer decision-making to them.
So that’s probably the biggest umbrella of personal growth and development for me, as the team has gotten larger and as my scope has gotten larger. I think ultimately the highest-leverage thing I can do besides hiring is prioritization. Helping, both me making decisions and also making sure our team is making decisions around what to focus on and what not to focus on. Deciding what we think is a high-ROI bet versus what we think is a low-ROI bet is ultimately the most impactful thing we can do, because it’s the thing we have the most control over. And so that’s still, today, the most important thing I can do and the most important form of impact I can have besides just hiring.
I’m sure people listening know, ROI is return on investment. I’m assuming by investment in that framework you mean, we’re going to spend time building something. What do you mean by return? Is it, active users are going up? Is it, revenue is going up? What is the other half of that equation for you?
That’s a good question. And the reason why this is particularly important for us is — as you or any other user of Twitter can attest — there are far more ideas for how we can build and improve the product than we have time or people or resources to actually work on at any given point in time. Which is a good problem to have.
We have a service that almost 200 million people use every day. There’s a lot of things we can do to improve it. And so we have to be super selective with what we focus on. Ultimately the measure of success we look at, to understand whether we’re helping or hurting the service, is growing daily active usage. That’s a very lagging indicator. So internally we look at, are we solving customer problems? And we have specific metrics that tell us whether we’re helping or hurting customer problems in specific dimensions.
So in the health apparatus, that means we have any number of metrics, whether it’s the number of impressions on tweets that are deemed unhealthy or toxic, or whether it’s the amount of abuse reports that we’re seeing on a daily basis. In other areas of the product, it could mean engagement, or tweets created, or Spaces created, or — it sort of just depends on the problem area.
But when we think about ROI, the investment part is easy. The return part is basically, how much impact are we having against the metrics that matter most? And how do we — and this is the hard part — how do we compare unlike metrics together? Because having yay impact on some problem space in health, which is measured in the absence of malicious activity, is different than measuring yay impact on some feature that results in more daily active users.
So that is actually one of the harder things. And we have a bunch of different approaches that we use, whether it’s having a portfolio balance or whether it’s literally stack ranking and prioritizing on an objective basis rather than on a metric basis. There’s a bunch of different things we’ve tried. It’s an evolving art form, I would say. But ultimately that’s the hard part of the job: figuring out what are the right priorities, and how do we prioritize against them?
Two more management questions then I’m definitely going to ask a million questions about the features you announced.
Casey actually brought this up to me when I asked him if he had questions for you: It seems like your pace of development has actually sped up during the pandemic, which is not true for most companies that went fully remote. Twitter is fully remote. That seems remarkable. Is that perception accurate? And if so, how did you manage that?
I do think our pace of development has sped up. I’m trying to think about how the pandemic specifically could have impacted that. I think of it more — and this is obviously just through the lens of Twitter, not comparing it to other companies who’ve probably had their own formula here — but we have been on a multi-year journey to speed up our development. And so I would like to think that… we’ve had our own hiccups, obviously, but we’re reaping the rewards of that investment over the last few years in our process, our culture, our hiring, our infrastructure work. And so the pandemic, I would probably say it slowed us down, it slowed that journey down for a little bit, as everyone was adjusting to the new normal. I wouldn’t say that it was a net accelerant by any means.
People, our team included, but people even outside of Twitter, have a lot on their minds, and it’s really stressful when you’re trying to do your job with your kids at home without child care. It’s been a taxing time for everyone, so I don’t think it was a net accelerant. What we’ve done through the pandemic, beyond just continuing to accelerate, obviously, is we did make some focusing decisions around, “Hey, these two projects that were kind of in our periphery and articulated as part of our long-term strategy” [shifted to] “this is the forefront now and all this other shit’s going to pause.” We made a bunch of those decisions that I think helped narrow the aperture, which accelerated some projects and paused or slowed down other projects.
And frankly, energy and momentum is infectious, right? As you see the company and other teams build quickly and you see customers noticing that, it inspires and motivates everyone else to want to live up to that. And so we’ve seen a bunch of that, I think, and we’re seeing it right now with Spaces.
Spaces, we love that our customers love it, but there’s a real kind of energy and movement internally at the company. I’ve been at Twitter for six years now, and I’ve never seen the level of energy and embrace around any singular project at Twitter, ever, in my time, at the risk of hyperbole. We’ve had a few of those moments with these big projects — Topics being another one in the last year — that have, I think, helped the team get through the intensity of what’s going on in the world, especially when we see that the product continues to be used and be really instrumental at a time where the world needs to communicate and learn from others about what’s happening.
This is my last management question, then I really do want to start talking about Spaces and Super Follows.
Twitter’s important to the world. We just came through, obviously, this very contentious election. We came through an insurrection at the Capitol. I saw today House Republicans are demanding a record of every content moderation Twitter has ever made because they say you’re censoring Americans.
That level of focus in the culture, how do you manage against that? Because Twitter is very important to a very small, but very powerful, group of people. A very influential group of people. And it seems like as you create new ways to create, you bring more people onto the platform, you show them more things, that level of influence just goes up in a way that — it’s possible that regulators around the world are basically going to make you stop doing things you want to do.
I think this aspect of Twitter, again, we were sort of talking about earlier, but this aspect of Twitter is one of the most unique things about it. You sort of — working at Twitter, you ride the roller coaster of whatever is happening in the world. And you feel all the highs, you feel all the lows and there’s an intensity to that that is really difficult sometimes, but it’s also really powerful and inspiring.
Did your team stop working every time Trump tweeted? I just imagined that grinding you to a halt all the time.
No. I think we’ve gotten pretty good at being resilient enough that a single tweet doesn’t cause people to stop working. We have a lot of people on our policy enforcement team that are really good at their jobs and I have a lot of respect and admiration for the level of stress that they hold in their jobs. But I think largely… the engine of Twitter does not grind to a halt when a single person tweets, but it can make it emotionally stressful to be on the job.
I’d be lying if I said that that wasn’t true. But again, you feel all the highs and you feel all the lows. You take it, when something really tragic happens in the world, or if someone does something really nefarious on the platform, it’s hard to not take that personally as something you’re responsible for. And I would like to think that that helps motivate us to do our best to build the product.
But it can also really suck sometimes. Politics and regulation is one very small slice of that, honestly. It’s not just that. But it certainly plays into it.
Let’s talk about all these new features. One thing I would link in common to everything you’ve announced — [you] bought Revue, which is a newsletter platform. You are doing Super Follows, which will let people pay for tweets from a certain — do you call them creators? Do you call them “tweeters?” What’s your preferred nomenclature?
I usually refer to them as creators. Obviously that term is so broad, because creators can be journalists, creators could be musicians, creators could be professional publishers. We’re still looking for a name that perhaps is more all-encompassing, but creators works. And by the way, it’s not just tweets. Super Follows will let your biggest fans subscribe to you and get access to exclusive content, which could be tweets, could be DMs. It could be subscriber-only Spaces. It could be subscriber-only newsletters. It could be subscriber-only Fleets. Like, think of it as stitching together all of the current and new and upcoming forms of content that someone can create on the platform, and really having this new subscriber layer and a new community, essentially, they can create that content for.
“Subscriber-only Fleets” is a real phrase and I’m going to hold that close to my heart.
You’re building all these tools that basically let the most popular people on Twitter, or people with a passionate following, monetize that graph in different ways. Either you’re going to kick over to Revue and you’re going to charge people for a newsletter on Revue, which looks a lot like Substack. You’re going to do subscriber-only Fleets and people are just going to pay you directly. There’s a million other ways to monetize.
That has been kind of the missing piece of the puzzle for most of the dominant platforms. I think YouTube was the gold standard for creator platforms for the longest time, because it had a built-out revenue model. And that was still mostly just AdSense and a slice of ad revenue.
You want a lot of people to use Twitter, you want them to find their interests, and now all these little corners of Twitter are going to be behind various kinds of paywalls. How are you thinking about that holistically? If the best people on Twitter, or the most interesting people on Twitter, have this incentive to keep people from seeing their stuff, that affects the value of the service as a whole. So how are you thinking about that balance?
I don’t think the incentive is exclusively to keep all content behind a paywall. I actually think that there will continue to be an incentive for creators to sort of juggle — which, everyone’s juggle and formula is going to be a little bit different — but juggle what content they create and make available for all, and what content they create just for those subscribers. At the end of the day, you have a tremendous amount of incentive to create some form of content for all to see, because that’s how you build an audience. And then some percentage of your super fans will totally be willing to pay for content that is exclusive. You already see this mechanic with newsletters. Some newsletters are free, some issues or posts are behind the paywall. And I think a similar mechanic will exist.
And, by the way, a lot of super fans, they just want to support their favorite creators. You see this with Patreon. I think Patreon is one of the most interesting companies in this space, frankly. And it’s not always about exclusivity. It’s about recognition and patronage, and just being able to support the creators who you appreciate the most and want to make sure they can sustain their craft. And so I think every creator will have their own twist on this, and will have their own sort of ratio of how they want to mix their content.
Maybe some creators will continue creating all their content publicly, but they just want to be able to have private DM conversations with their subscribers. Or once they go create their newsletter, or drop their video, or release their merch or whatever publicly for all, they go have an audio conversation where only their subscribers can really talk about it. So I think there’s so many permutations that we imagine, that I don’t think it’s about taking the content that exists and putting it behind the paywall. I think that this new layer will incentivize a new layer of content to exist that doesn’t actually exist today.
One of the hard problems here is Twitter is an app on a phone. I use desktop Twitter, but I’m guessing the vast majority of people that are interacting with Twitter are doing it with an app on iOS or Android. If you want to enable a payment inside of an app on those platforms, you have to give a cut to Apple or Google. I’m assuming Twitter is going to want a small slice of whatever payment, because they’re enabling the interaction. You’ve probably got a payment processor in the mix. That’s a lot of cuts.
Are you going to talk to Apple and Google about reducing the 30 percent cut for Super Follows? Because that’s a lot for a creator to bear.
Yeah. I mean, we haven’t announced or decided any of the specific numbers in terms of what the cuts are. But one important piece of context — and then we can jam on this specific question, I’m not dodging it — but the important piece of context here is, we are not… for Super Follows, our goal is not for Twitter to make money. Our goal is for creators to make money. I think Twitter may incidentally participate in the transaction in some way to sort of cover our cost, but our goal isn’t to maximize revenue.
This work, just going back into the management organizational stuff, ladders up into our conversations work. And the goal of the conversations work is to incentivize conversations, to incentivize people to create conversations. So it’s just an important, subtle — but very meaningful for us — aspect of our positioning that we’d love to put as much money into the creator’s pocket as possible.
There are transaction costs, be it the platforms or otherwise, that will need to be involved in that transaction. This is distinct, by the way, from other things we might do in the future, like a Twitter subscription. You’ve heard us talk about this. And the goal of that is very different. It’s not about empowering creators, necessarily. It’s about providing premium features for power users that let them do things that they can’t do with Twitter today. And so that’s a subtly different goal.
But on the creator side, I think the platforms and the app stores, be it Apple or Google, they have these tried and true workflows and frameworks for developers to leverage when it comes to facilitating digital goods and in-app payments, and that comes with a cost, but it also comes with a lot of benefits that increase conversion and improve simplicity and decrease fraud. And so there are definitely pros and cons there.
But we’re not tying this to one platform. Ultimately this will work on iOS, Android, web. And so we have to have a sort of a framework and a workflow that allows all of that to work cohesively well together.
Let me push on it a little bit. I’m a prolific tweeter. I say, “Okay, I’ve got 100,000 Twitter followers. I’m going to charge 10 bucks a month. I convert X percentage of them to Super Followers. Now I’m making real money. I’m making a couple thousand dollars a month.” Except I’ve got to give some amount back to Twitter, even if you’re not trying to make a profit, I’ve got to give some of that back to Twitter, because that’s how they’re covering their cost. There’s a payment processor in this mix somewhere that’s going to take — or maybe you’re going to be the payment processor. I’m told Jack Dorsey owns a payments company.
What? [Laughs]
So there’s a good synergy there. Then 30 percent of that $10, I’m giving away $3 to Apple and they’ve given me nothing as a creator. I get no value out of that. Maybe Twitter gets some value out of that, but me as a creator, I’m just giving away $3 of my $10 to a giant platform company. Are you in a position to advocate for a lesser amount there?
The way I would think of this in the context of Super Follows where we’re building this layer that didn’t exist before — even if that $10 comes down to $7 because of a 30 percent fee, that’s still $7 more than you’ve been able to make on Twitter than before.
So don’t get me wrong, I would love for that to be $9 instead of $7, but at the end of the day, that’s not something that we have direct influence over on one platform. So it’s not a focus for us right now. Our focus is to build the best possible experience that is good enough that people aren’t going to think about the cut. They’re using this to have conversations and establish community and get recognition from their favorite creators. I think if we do that right, then good things will happen.
Do you think there’s a way to build this feature that gets around that rule?
We’re not in the business of getting around platform rules.
You guys make a lot of platform rules. I imagine you have a good incentive to —
Listen, I get that there’s a lot of controversy around this, and some of it is for good reason. I think that there’s a lot of positive things about these. This isn’t just a highway tax. There’s a lot of cost and effort involved in building these ecosystems that allow you to accept payments, and there’s a lot of fraud or risk involved in the whole customer service flow around refunds. A lot of that is taken off of your plate.
Now, if you’re a bigger company that has a lot of those processes, then you don’t want to pay someone else for that stuff. But we don’t want to innovate around the payment flow. We want to innovate around letting creators make money from their audience. So that’s why this isn’t a big thing for us. We’d rather minimize as much effort as we can around the table stakes of facilitating payments, so that we can spend our energy on innovating for creators and for super fans of the creators.
Let’s talk about Spaces for a second. How long ago did you conceptualize Spaces?
Well, I would like to say that the beginning of when we knew audio was special was, we had this feature in Periscope, which we called Hydra — internally, I don’t think we actually ever even called it that externally — but basically when you’re live, you can pull in guests and have a zero-latency live conversation with them while everyone else watches. Casey has used this many times. I’ve been a guest in his Hydra broadcasts and fundamentally, the infrastructure that powers that is literally what Spaces is, just with a new skin.
I’m sort of kicking myself now because when we released that feature two and a half years ago, we were like, “Wow, this is way more interesting than any other Periscope broadcast,” because it starts to become this multiplayer experience where the broadcaster is having this interesting conversation with someone, whether it’s a friend or someone who dropped in as part of the community or whatever, and we all get to watch. Rather than being this one-way broadcast, now it’s a conversation that you’re listening to, kind of like a podcast.
At the time, we were going through stuff and live video was not a focus. So we ended up deprecating or putting a sunset plan together for Periscope, but we knew there was something interesting there. We just didn’t move on it.
Fast forward to about a year ago, we really started investing in audio and thinking about how we can enable audio as a new form factor for conversation on Twitter. The same team that’s driving Spaces today really started focusing on that. The first product that they built was what we call voice tweets, which we put in market late last year on iOS. It lets you record your voice and tweet it out basically.
Around the same time, they were thinking about the sort of conversational experience, and this is when audio really started heating up and Clubhouse was getting a lot of traction. So we had a long and winding road to refocus back on the sort of multi-person conversational format for audio. But the team that’s building Spaces now has had their heart in this for quite some time. Obviously, hindsight is 20-20. We found much more customer success and impact and excitement around the multi-party experience than we did with the voice tweets experience. We still see lots of really awesome use cases there, but we’ve shifted all of our focus to Spaces now.
One of the things that strikes me about social apps right at this second is, on the one hand, TikTok is an explosive success and the tool it presents to the user is an extraordinarily powerful video editor. The best TikToks I see are really well-crafted videos that use a lot of elements and bring them together that is just a very — in its way — a very high-production kind of storytelling. All the way on the other end of the spectrum, there are Spaces and Clubhouses where I just show up, I’ve had one too many, and someone is going to let me talk to a thousand people. Where on the spectrum do you think Twitter needs to be?
I think that this medium is much more close to the latter side of the spectrum than it is the former.
It’s quite intentionally not a highly-produced sort of editing environment. Anything live, I believe, is going to immediately put you on that side of the spectrum. That has pros and cons. But one thing that’s worth noting is while a lot of these audio conversations, a lot of the audio conversations that drive headlines and people talk about, tend to be these really large ones where you have few speakers and many listeners. That will happen when we GA [release for general availability] this product. If Bill Gates and Elon Musk have a conversation in a Space, there’s going to be a lot of people listening and very few people speaking.
But 99 percent of these Spaces, and we’ve already seen this in our beta, are not that archetype. That’s an important archetype and an interesting one. Most are 10-person rooms, less; people just hanging out and talking about their interests, whatever it might be.
The way you build a product experience for that use case is very different than the way you build a product experience for the massive space where there’s an interview happening or a panel discussion with a ton of listeners. We think both will happen on Twitter, and particularly for Twitter, we know all of the influencers across sports, entertainment, news, politics — they’re all on Twitter. So we want those Spaces to happen, but we know that to build the best, most durable product experience here, real people need to be able to have fun and have great, healthy discourse that happens through a medium like this with very different dynamics. You have a small room with a single-digit number of people or dozens of people at most.
So that’s what I’m really excited about. And frankly, that’s something that we kind of fucked up with Periscope. We had the same dynamic where both of those archetypes existed, but we didn’t make any one of them particularly great. The experience kind of broke down when there was a broadcast with 300,000 viewers. It just felt like chaos. And the experience felt a little bit lonely if you had 10 people, or if you had 30 followers. What are the odds you can get one of your 30 followers? So getting that right is really top of mind for us.
One of the challenges with Spaces, with all these social audio products, is moderation. So when you open Spaces right now, it tells you, “someone reports this, we’re going to record it.” But then that also is— there’s no long-tail audience there. [If] Bill Gates and Elon Musk had a conversation on Twitter right now, you can see it happening. Then maybe if you missed it, you can [still] go see it on Twitter. Maybe BuzzFeed will aggregate it. Someone will do something with it. There’s a long tail of people getting to consume that. Spaces right now, doesn’t have recording, it doesn’t have an archive, but you also need that stuff in order to moderate it well. Where’s the push and pull of that stuff?
I think there’s a couple of layers of this. One, just as a baseline, we have our policy and enforcement apparatus that needs to be well-integrated here, and sort of up to snuff for the medium. So that’s where we do record right now, for purposes of health enforcement. There’s a bunch of tools and automation we can build to make that as efficient and sophisticated as possible, given that there are going to be many Spaces live at any given point in time. That’s very difficult. You can’t solve this problem alone with it, but it is important nevertheless, and the good news is we’re not starting from scratch here. We’ve built much of this for the last six years with live video. There are some nuances that are different about not having a video stream to accompany your enforcement techniques, but there’s a lot of shared attributes in moderation here.
The second layer, and I believe the most important one, is actually building native product features and native functionality imbued into the product that helps all of the actors within this conversation: the hosts, the speakers, the delegated admins that the host might have, and the listeners, be able to take part in optimizing for health. This sort of decentralization, I think is really, really important, not as the sole measure of health enforcement, but ultimately, a lot of what’s going to keep these conversations safe is making sure the right people have the right control.
The host should be able to decide who’s in, who’s out. The host should be able to empower some of their speakers or their listeners to be able to aid in the moderation process, because if you’re hosting a conversation with 20 people, let alone if you have 3,000 people in a Space, you’re already juggling a lot of shit. You need some backup, and this is product functionality that can help the conversation feel not overwhelming, help the conversation feel, ultimately, safe from unwanted interactions, and can feed into the sort of health enforcement apparatus as well.
So that dimension is the one where I think there’s a lot of innovation to be had. That’s the one where we’re trying to lead from the forefront. In the next three weeks, a lot of the features that we’re going to be adding fall into this category, and it’s sort of in the spirit of scaling moderation. And again, as a base layer, we obviously need to do everything we can to have clear policies, the right policies, [and] enforce against those policies, but we also would be kidding ourselves if we thought we could be the police officers for all live audio conversations happening globally at any given point in time. We need to be smarter than just that. So we’re approaching it as holistically as we can.
This was a great Periscope feature where — you’d be watching a Periscope, you would see the comment stream float by, and then you, as a user, a comment would get flagged to you and it would ask you if this comment needs to be moderated in some way. That makes sense when it’s text. How would you build something like that for audio? Are you just going to ask people, is this speaker being rude?
So we call that the jury moderation feature, which to your point, I think is arguably easier to do in text because you’re scoping the deliberation that you’re asking the jury to make, to be a comment. But by the way, that can be really hard sometimes because the context of when the comment was made really matters. So we sort of learned the hard way that there’s a lot required to get that right. I do believe a system like that can be effective if pointed at the right direction. I think one of the missing elements here, though, you’re not just asking a listener, “Hey, is this person rude?” A lot of what’s tricky about the health space is that what people deem as healthy or unhealthy is very subjective. You might be really bothered by something that I am not bothered by.
I think one of the challenges with Twitter is that who owns the space — using the term space loosely here — has been very amorphous in the past, and you’ve seen us actually try and make this way more intentional, where you own your conversational space. When you tweet something, you can hide replies to your tweets. It’s not just a local mute or a local block. It impacts what other people can see. You can now start tweets with conversation controls that prevent other people that you choose from not being able to reply. Likewise, a Space, literal Space in this case, is an even more deterministic version of that. It’s your Space and you ought to be able to define the social norms of that Space, and we ought to be able to build capabilities that help your community, whether that’s speakers or listeners, enforce those social norms.
What I just described is a lot easier to say than to do in the product, but I believe that if this product takes off like we think it will, it actually provides a lot of really interesting potential for us to build a decentralized moderation capability that’s, frankly, a lot harder to do with the other parts of the scaffolding of the Twitter product. Now, add on top of that, the challenges of audio as a medium and synchronicity is a lot harder because there’s a time element, but SpaceX can send a ship to space and bring it down. These are solvable problems. We just have to focus on it.
Sometimes I think content moderation might be harder than launching a rocket and landing it. I’ll be honest with you.
Do you have plans to let people broadcast their archive recordings? It just feels like if you’re recording them for moderation purposes, there’s a value there that can be extracted. Or are you thinking this is fundamentally an ephemeral product?
We’re not religious about it being fundamentally ephemeral. We’re starting that way because we think it’s simpler and we need to focus. I think building “recording” — and I’ll tell you why I put that in air quotes later — building “recording” the right way is tricky. It’s not as simple as just, “Hey, here’s a recording.”
I think most people don’t want to listen to a 67-minute recorded conversation that serendipitously evolved based on who came in the room. We also learned that the hard way with Periscope. We started ephemera — well, it was actually not ephemeral. The recordings were 24 hours. Then it became indefinite as a default, but most people do not want to play back a conversation unless it’s a highly-produced podcast like this one, where the person who crafted it did so with the recording in mind.
I think it should be a choice. If you think that the conversation was worth playing back, you ought to be able to do that. I personally am a little bit more bullish on two things. One, obviously the host should be able to save it and do whatever they want. Maybe you host a Space, you save it, then want to go edit it. You should be able to do that. So that, we will build that for sure.
I also think that the notion of letting the audience pick sound bites and share them as clips could be really, really powerful. Now, the challenge with that is you have a sort of a really challenging consent issue because you have the host’s intent in mind of, does the host want this conversation to be preserved or shared? [Then] there’s the speakers, who are a different actor than the host. Their consent is really important.
So that’s really challenging to get right. It’s one of the reasons why we haven’t just turned this on. We want to be thoughtful around how we do that, and if we can build an experience that balances those motivations and empowers the right people. But I think assuming that can be solved, just the way that Twitter works fundamentally, I think a recording concept could be very powerful here. It could be a very, very easy way for people to catch the most important parts of the conversation they missed. It can drive tune-in to conversations. I think it’s actually a very powerful aspect of how Twitter already works, but we just need to be thoughtful around how we build it.
Do you think of Clubhouse as the competitor here? There’s a lot of heat on Clubhouse. They have some different ideas about how that product works. Is social audio a market and you’re going to take all the marketshare? Or are you feeling differently about it?
I don’t think social audio is going to be winner-takes-all. I think there’s going to be a lot of interesting startups and big companies that build this. I am a huge fan of the Clubhouse team. I’m inspired by what they’ve done. They’re not going to be the only ones. Facebook has a playbook here. I’m sure they’re already working on something. Spotify is making big investments in audio. I think we’re going to see lots of people building on what is now a mechanic that will forever change how people communicate on digital platforms, which is a great thing. There’s going to be great competition.
I think for us, this is fundamentally Twitter. People have been coming to Twitter to talk about what’s happening for 15 years. It’s imperative that we build this into the product in a way that feels super natural and cohesive and allows people to have a new form factor in which they can talk about what’s happening and find interesting conversations. So that’s our focus, we are excited by the energy and the ecosystem as well. And I think there’ll be lots of people who have different rifts on it.
All these new products you’re building — one big theme here is you want really interesting people to use Twitter. A lot of them already are. You want them to have additional tools, maybe that’s a subscription product that they pay for. You want them to have a range of monetization options — that’s Super Follows, maybe that’s Revue. That just seems like a new revenue apparatus for Twitter, even if your goal isn’t purely revenue, that’s just a new business model for Twitter to be in.
How are you thinking about bringing along the existing user base to use all those new tools and new opportunities versus who you want to go out and get? Because that it seems like it might be richer, to go get some new people to use Twitter in new ways.
I think both of those pursuits are really interesting values. I think that we hope that with a lot of the capabilities that we’re building, it will inspire and motivate our existing customers to use them. Frankly, a lot of the work that we’ve done in the last year and a half has been sort of in that vein. We talked to our customers, we’re trying to understand how we can better serve them, and we released these features that motivate them to be able to use the features because it solves new problems for them. I think what we’re disproportionately doing more of now, that we haven’t really taken on in the past, is introducing use cases that can attract new customers.
Revue is a good example of this. You could hack Twitter to be a good longform publishing tool; threads and tweetstorms are kind of a good example of this. But largely, if you want to publish longform content, you would go elsewhere. And so we are trying to be intentional around introducing those new use cases, trying to diversify the tools that our existing customers have that they can use, but also inspire new people to get value out of the distribution potential that they can get from Twitter. So it’s not one or the other, it’s definitely both. And we’ve come from a place where we’ve been hyperfocused on making the existing products we have work better, and we have some wind in our sails that we can sort of open the aperture a little bit and start to explore use cases that have been further out of reach.
I ask because we spend a lot of time talking to policy makers, and we hear about big tech so much, and Twitter is always on that list. [You said] Facebook has a playbook: they’re going to spin up a team that’ll copy five apps, the three that worked they’ll launch in Brazil and move on. They’re big enough to just do that.
Twitter is much smaller than these other companies from a userbase perspective, from a revenue perspective. Just coming back to that ROI conversation: is it a better return to get more new people to use Twitter or to get more revenue, interactions, whatever it is, from the existing user base more often?
I think this is going to seem like a cop-out answer, but both. We have almost 200 million people who use Twitter every single day and we need to make the product better for them. There are areas where we don’t serve their needs and I believe if we do serve their needs, they will get more value out of Twitter.
We have even more customers who use the product on a monthly basis that aren’t daily users, and therein lies an opportunity to serve them better. And then on top of that, we have over 2 million people that are either brand new to Twitter or sort of returning to an account after not being here for 30 days or more. 2 million of them, every single day, come to our front step. That’s a lot of people who are trying to use Twitter that are new. So we can clearly be serving them better too.
So that’s why, when we started our Analyst Day presentation, I said, we believe that Twitter can serve billions of people every single day. We’re very far from serving billions of people every single day, but we genuinely think that the market of people who are interested in following what’s happening and connecting with their interests is billions — it’s everyone. Who isn’t interested in that? So our job is to really close that gap and to find all the things, large and small, that can help make Twitter more valuable to customers.
I asked Adam Mosseri this when he was on — it was right after the election, a little bit more time has passed now — but obviously, Trump is banned from Twitter. Is he going to stay banned?
I’m not the expert on this, but he is banned from Twitter.
Okay. There’s been a couple interesting moves toward decentralization, from Twitter’s perspective, in content moderation and in the platform itself. You launched Birdwatch, which is a community moderation system. How’s that going?
It’s super early days, but very promising early results. We’re just in pilot mode right now, it’s English-only. So I really want to emphasize: still lots more to learn, but really promising across a few different dimensions. Birdwatch is a decentralized moderation tool, so basically a community of people can annotate tweets and enforce them either against our policies or against policies that we may not have. They can provide annotations around why tweets may be misleading and explain why.
What we’ve been really pleasantly surprised by is good faith actors, speed, and quality of response. Probably the thing that we were most keen to learn is, were people just going to have one-liners like, “This is not correct,” or are they going to explain why something might be misleading? And I think we’ve been really pleasantly surprised by the quality and robustness of people’s annotations. And it’s just been very, very promising. As well as the speed at which they’re annotating tweets, despite the fact that this is a pilot that is only visible on a separate micro-site, so you would think that there’s actually not much incentive for people to be quick to annotate these tweets. So again, lots to figure out here, this is not going to be an overnight product solution, but I think that empowering a community to help aid in, certainly, addressing misleading information, let alone other policy areas within health, I think it’s very promising.
Annotating tweets looks like labels — something you rolled out during the election. Lots of conversation back and forth over whether they’re really effective. Do you think they’re really effective? They are a moderation product solution.
I honestly think that the jury is out. I think that there are applications of labels that are effective. For example, one of our forms of labels also disables engagements, so if a tweet is particularly egregious and violates some of our policies, we not only label it, but we disable engagements around it so that it cannot be amplified. And that is extremely powerful, and is effective in sort of diminishing the visibility and the spread without context, at least, of that content. I think our other labels, we’re actively still learning from which applications of them are effective and which aren’t.
One aspect of this that I think is critical is, through our research we have a hypothesis that people will trust these assessments more if it comes from a diverse community of voices, rather than us as a singular social platform. And that’s one of the hypotheses that we’re really exploring with Birdwatch, it’s not Twitter saying, “This is misleading information,” it’s a community, a jury of your peers that are saying, “This is misleading and here’s the rationale for it.”
If I made the comparison to Wikipedia, would that be directionally correct? You’re building a sort of Wikipedia backend of labels that people participate in?
I think that’s a very apt analogy. We looked at Wikipedia as a source of inspiration for how you can build a set of mechanics that can allow a community to thrive. Wikipedia also built the community, but they more built the mechanics that allow the community to thrive. This is sort of like self-healing, self-corrective nature of a platform that is fundamentally open — by virtue of being open, it certainly opens itself to risk vectors, but can sort of self-heal and allow an incredible corpus of knowledge to flourish, which is Wikipedia.
I think the same challenges exist in a world where you have decentralized moderation; anyone could label a tweet; a lot of things could go wrong there. And so our job is to build the right mechanics within Birdwatch and the right incentive systems that allow quality annotations to flourish and garbage to not flourish. That’s the hard part.
We’ve talked about a lot of new features, a lot of them are in the app. Spaces is iOS-only, you’ve started testing it on Android. Those are a lot of proprietary features, you’re going to probably build some revenue stuff around it. But you also run TweetDeck, you also support third-party clients. There has been some work in the background at Twitter to decentralize the entire platform in some way. How does that all play together?
So, a bunch of different things there. TweetDeck is another Twitter client that we own, it plays a very important role, I believe, in the world. It’s got a small but mighty user base of power users and many of them—
The Verge’s newsroom runs on TweetDeck.
Totally. And we haven’t given TweetDeck a lot of love recently. That’s about to change; we’ve been working on a pretty big overhaul from the ground up of TweetDeck, and it’s something that we’re excited to share publicly sometime this year. And so that’s just an example of a Twitter-owned and operated service that we will continue investing in. We also, over the last five years, I think, haven’t given a lot of love to our developer ecosystem. A bunch of reasons for that, some missteps that we’d taken in the past, then also sort of prioritization. We are also changing that; in the last year and a half we’ve really stepped up both our commitment and follow-through on just innovating around the API again, getting the API back to parity from our own internal APIs that we use to build functionality.
I think we’ve got a lot of trust to earn back with developers, since we’ve made a lot of mistakes in the past, but it’s something that we’re actively investing in. We hope we’ll allow developers to build really awesome stuff around the Twitter ecosystem. One of the reasons why Twitter is where it is today is because of developers doing cool shit that we would’ve never thought to do. And so that’s something that we’re trying to do more of, not step away from. More to come on that as well.
How do you balance — Spaces, for example. Are you going to let third-party clients access Spaces?
We will at some point. It’s not an active priority today, and that’s mostly just focus. We did this with Periscope too, we opened up the producer API, which lets people use any RTMP stream and go live with it. And I expect that there could be lots of interesting use cases with Spaces as well. Not just on the creating a Space side, but I think there’s lots of really amazing developer tools I can imagine around managing a Space, particularly large Spaces, I can imagine whole control centers being built around it. Complex moderation capabilities that people get through a web dashboard that Twitter might not have in its native tools or analytics, that help you understand how people are going in and out of your Spaces and how long they’re staying, and speakers, and what topics are driving the most tune-in, when stuff gets shared. I would love for us to build some of these things, but I know that developers who have expertise and passion in these areas will be able to do a better job, frankly.
Like anything else we were talking about at the beginning, it comes down to prioritization and us kind of understanding people’s needs. Developers are one important cohort and we want to make sure we build for their needs, but we’re moving very quick right now, so it’s not our immediate focus. But we’ll definitely get there.
That’s always the balance. You’re moving fast; it’s hard to open up to developers. There has been some talk about decentralizing Twitter entirely, having it as a protocol. There was just a white paper. Is that real?
It is real. You were talking about time horizons and timescales earlier, and Jack, as our fearless leader, his job is to think in the longest possible timescale. This is an area that we are investing in and the way we’re investing in it is, sort of sharing our ambition and belief that Bluesky, which is what we’re calling this ambition for a protocol layer, is something that we do want to create. And it’s something that we hope that Twitter can become a client of in the future. And we want to help build a structure and apparatus and hire the lead to be able to create this protocol. There’s still a lot of questions and work, and it’s super early. This is not something that will be in place in the next year or two. This is long-term — super ambitious and speculative, that is super interesting — but it’s on a multi-year timeframe. So we’ll see.
Last question. What should people be looking out for next from Twitter?
Oh man. People should look out for us to keep moving quickly, keep trying things. May not be the edit button tomorrow, but we are going to be taking swings—
Oh wait, that was the actual last question, it’s up there on the list. Am I getting an edit button?
Wow. You just totally, “screw the last question” —
You went there. I didn’t even open that door, you opened it and walked right through it.
Well, what do you think?
I think you’re going to tease me with it for the rest of my life and that’s just the dance we’re in.
No, no, what do you want, is what I’m asking.
I want to be able to fix typos on my tweets. I routinely tweet the wrong link, this is a common thing that I do, and I have to delete and start over. So I would just like to be able to edit a tweet in a 60-second window.
Yeah. I think that’s a very reasonable ask, and I think you should be able to do that at some point, for sure.
Confirmed?
Not confirmed. But there’s no philosophical reason for us not to do that; it’s purely just prioritization.
Yeah, with the highest priority as trolling me personally, constantly. It’s great.
You and Casey, you’re right up on my list of people to troll.
