Three years ago, Meta launched Threads as a Twitter alternative, staffed with a bare-bones team. Now the company says its text-based social network has 500 million users — more than Twitter ever reported when it was a public company (In 2024, Elon Musk claimed his rebrand of the service had brought in 600 million users.) Another marker of Threads' growth from side-project to Real Thing is that Instagram boss Adam Mosseri, who had been supervising Threads, has handed control of the service to longtime Meta product guy Connor Hayes. I caught up with Hayes at Meta's headquarters this week to talk about Threads' evolution, its stance on AI content — and how Meta is using the strategies it used to grow Threads as it pushes out Muse, its buzzy AI agent. Peter Kafka: When Threads launched in 2023, you were actively courting Twitter users who didn't love what was happening under Elon Musk's ownership. But Twitter is still around. Bluesky has its own user base. Who is a Threads user today? Connor Hayes: We assumed in the very beginning that growing Threads would be predicated on capturing people from these other services. We've actually been able to grow the market. It turns out there were a lot of people that were using Facebook and Instagram every day who are willing to use an app built around this type of text format that hadn't been using alternatives like Twitter before. In the beginning, it was purely, "I'm looking for an alternative." And it felt like a safer, more positive conversational environment. Now it's starting to grow into its own thing that's built around these communities. We're trying to go really deep on categories that people are super passionate about, in a way that I think Reddit has done a good job with in the past. We're trying to get into this space that's a hybrid of what Twitter was, what messaging is and what Reddit has been. A big way you've grown is by promoting Threads to Instagram and Facebook users. Now Meta is using that same playbook to promote Muse. Is promoting a new AI agent to Instagram users different than telling them about a new messaging app? The biggest difference is that Threads is built around user-generated content. The engine that powers our growth today are promotions we do on Facebook and Instagram, where we know that your account is linked and we can recommend you content from Threads on Instagram while you're scrolling. Of course, you're going to be interested in that, because we found a way to match it. For something like Muse, the big challenge is: How do we use what we know about you and your interests and promote this app to you, without having a base of user-generated content that can kind of do that work for us? So we're looking at a bunch of things, like if you're interested in soccer, what are the use cases around Muse that could be really interesting or productive for that? Can we build units that show off those use cases, rather than using specific pieces of content that we found from the app? What we're trying to do with these promotional units across the apps is not just say, "Hey, there's an AI agent," but "Hey, there's an AI agent that can do this specific thing for you." Whereas with Threads, I think the hook might just be, "Oh, there's interesting content here. I'm willing to come back tomorrow because I want more of that." The hook for Muse is a little bit further down the funnel. All the social networks are grappling with how they want to deal with AI content. Do you guys have a policy or a worldview about me using AI to create all or some of what I make or what I post there? On Threads, we use the same sort of approach that the other Meta apps do when it comes to labeling what we detect to be AI-generated content. But we understand that there's going to be a bunch of people that want to use AI to either better their content or have a better presence on the platform. Substack, for example, came out and said, "We don't want to be the next LinkedIn," meaning LinkedIn is full of AI slop and that it wants to actively police for it. But it sounds like that's not where you guys are at. Actively police, no. But I think that the best version of the product is going to come from human-generated content. My sense is that human-generated content is going to win out when it comes to driving conversations.

Three years ago, Meta launched Threads as a Twitter alternative, staffed with a bare-bones team. Now the company says its text-based social network has 500 million users — more than Twitter ever reported when it was a public company (In 2024, Elon Musk claimed his rebrand of the service had brought in 600 million users.) Another marker of Threads' growth from side-project to Real Thing is that Instagram boss Adam Mosseri, who had been supervising Threads, has handed control of the service to longtime Meta product guy Connor Hayes. I caught up with Hayes at Meta's headquarters this week to talk about Threads' evolution, its stance on AI content — and how Meta is using the strategies it used to grow Threads as it pushes out Muse, its buzzy AI agent. Peter Kafka: When Threads launched in 2023, you were actively courting Twitter users who didn't love what was happening under Elon Musk's ownership. But Twitter is still around. Bluesky has its own user base. Who is a Threads user today? Connor Hayes: We assumed in the very beginning that growing Threads would be predicated on capturing people from these other services. We've actually been able to grow the market. It turns out there were a lot of people that were using Facebook and Instagram every day who are willing to use an app built around this type of text format that hadn't been using alternatives like Twitter before. In the beginning, it was purely, "I'm looking for an alternative." And it felt like a safer, more positive conversational environment. Now it's starting to grow into its own thing that's built around these communities. We're trying to go really deep on categories that people are super passionate about, in a way that I think Reddit has done a good job with in the past. We're trying to get into this space that's a hybrid of what Twitter was, what messaging is and what Reddit has been. A big way you've grown is by promoting Threads to Instagram and Facebook users. Now Meta is using that same playbook to promote Muse. Is promoting a new AI agent to Instagram users different than telling them about a new messaging app? The biggest difference is that Threads is built around user-generated content. The engine that powers our growth today are promotions we do on Facebook and Instagram, where we know that your account is linked and we can recommend you content from Threads on Instagram while you're scrolling. Of course, you're going to be interested in that, because we found a way to match it. For something like Muse, the big challenge is: How do we use what we know about you and your interests and promote this app to you, without having a base of user-generated content that can kind of do that work for us? So we're looking at a bunch of things, like if you're interested in soccer, what are the use cases around Muse that could be really interesting or productive for that? Can we build units that show off those use cases, rather than using specific pieces of content that we found from the app? What we're trying to do with these promotional units across the apps is not just say, "Hey, there's an AI agent," but "Hey, there's an AI agent that can do this specific thing for you." Whereas with Threads, I think the hook might just be, "Oh, there's interesting content here. I'm willing to come back tomorrow because I want more of that." The hook for Muse is a little bit further down the funnel. All the social networks are grappling with how they want to deal with AI content. Do you guys have a policy or a worldview about me using AI to create all or some of what I make or what I post there? On Threads, we use the same sort of approach that the other Meta apps do when it comes to labeling what we detect to be AI-generated content. But we understand that there's going to be a bunch of people that want to use AI to either better their content or have a better presence on the platform. Substack, for example, came out and said, "We don't want to be the next LinkedIn," meaning LinkedIn is full of AI slop and that it wants to actively police for it. But it sounds like that's not where you guys are at. Actively police, no. But I think that the best version of the product is going to come from human-generated content. My sense is that human-generated content is going to win out when it comes to driving conversations.

Three years ago, Meta launched Threads as a Twitter alternative, staffed with a bare-bones team. Now the company says its text-based social network has 500 million users — more than Twitter ever reported when it was a public company (In 2024, Elon Musk claimed his rebrand of the service had brought in 600 million users.) Another marker of Threads' growth from side-project to Real Thing is that Instagram boss Adam Mosseri, who had been supervising Threads, has handed control of the service to longtime Meta product guy Connor Hayes. I caught up with Hayes at Meta's headquarters this week to talk about Threads' evolution, its stance on AI content — and how Meta is using the strategies it used to grow Threads as it pushes out Muse, its buzzy AI agent. Peter Kafka: When Threads launched in 2023, you were actively courting Twitter users who didn't love what was happening under Elon Musk's ownership. But Twitter is still around. Bluesky has its own user base. Who is a Threads user today? Connor Hayes: We assumed in the very beginning that growing Threads would be predicated on capturing people from these other services. We've actually been able to grow the market. It turns out there were a lot of people that were using Facebook and Instagram every day who are willing to use an app built around this type of text format that hadn't been using alternatives like Twitter before. In the beginning, it was purely, "I'm looking for an alternative." And it felt like a safer, more positive conversational environment. Now it's starting to grow into its own thing that's built around these communities. We're trying to go really deep on categories that people are super passionate about, in a way that I think Reddit has done a good job with in the past. We're trying to get into this space that's a hybrid of what Twitter was, what messaging is and what Reddit has been. A big way you've grown is by promoting Threads to Instagram and Facebook users. Now Meta is using that same playbook to promote Muse. Is promoting a new AI agent to Instagram users different than telling them about a new messaging app? The biggest difference is that Threads is built around user-generated content. The engine that powers our growth today are promotions we do on Facebook and Instagram, where we know that your account is linked and we can recommend you content from Threads on Instagram while you're scrolling. Of course, you're going to be interested in that, because we found a way to match it. For something like Muse, the big challenge is: How do we use what we know about you and your interests and promote this app to you, without having a base of user-generated content that can kind of do that work for us? So we're looking at a bunch of things, like if you're interested in soccer, what are the use cases around Muse that could be really interesting or productive for that? Can we build units that show off those use cases, rather than using specific pieces of content that we found from the app? What we're trying to do with these promotional units across the apps is not just say, "Hey, there's an AI agent," but "Hey, there's an AI agent that can do this specific thing for you." Whereas with Threads, I think the hook might just be, "Oh, there's interesting content here. I'm willing to come back tomorrow because I want more of that." The hook for Muse is a little bit further down the funnel. All the social networks are grappling with how they want to deal with AI content. Do you guys have a policy or a worldview about me using AI to create all or some of what I make or what I post there? On Threads, we use the same sort of approach that the other Meta apps do when it comes to labeling what we detect to be AI-generated content. But we understand that there's going to be a bunch of people that want to use AI to either better their content or have a better presence on the platform. Substack, for example, came out and said, "We don't want to be the next LinkedIn," meaning LinkedIn is full of AI slop and that it wants to actively police for it. But it sounds like that's not where you guys are at. Actively police, no. But I think that the best version of the product is going to come from human-generated content. My sense is that human-generated content is going to win out when it comes to driving conversations.

Three years ago, Meta launched Threads as a Twitter alternative, staffed with a bare-bones team. Now the company says its text-based social network has 500 million users — more than Twitter ever reported when it was a public company (In 2024, Elon Musk claimed his rebrand of the service had brought in 600 million users.) Another marker of Threads' growth from side-project to Real Thing is that Instagram boss Adam Mosseri, who had been supervising Threads, has handed control of the service to longtime Meta product guy Connor Hayes. I caught up with Hayes at Meta's headquarters this week to talk about Threads' evolution, its stance on AI content — and how Meta is using the strategies it used to grow Threads as it pushes out Muse, its buzzy AI agent. Peter Kafka: When Threads launched in 2023, you were actively courting Twitter users who didn't love what was happening under Elon Musk's ownership. But Twitter is still around. Bluesky has its own user base. Who is a Threads user today? Connor Hayes: We assumed in the very beginning that growing Threads would be predicated on capturing people from these other services. We've actually been able to grow the market. It turns out there were a lot of people that were using Facebook and Instagram every day who are willing to use an app built around this type of text format that hadn't been using alternatives like Twitter before. In the beginning, it was purely, "I'm looking for an alternative." And it felt like a safer, more positive conversational environment. Now it's starting to grow into its own thing that's built around these communities. We're trying to go really deep on categories that people are super passionate about, in a way that I think Reddit has done a good job with in the past. We're trying to get into this space that's a hybrid of what Twitter was, what messaging is and what Reddit has been. A big way you've grown is by promoting Threads to Instagram and Facebook users. Now Meta is using that same playbook to promote Muse. Is promoting a new AI agent to Instagram users different than telling them about a new messaging app? The biggest difference is that Threads is built around user-generated content. The engine that powers our growth today are promotions we do on Facebook and Instagram, where we know that your account is linked and we can recommend you content from Threads on Instagram while you're scrolling. Of course, you're going to be interested in that, because we found a way to match it. For something like Muse, the big challenge is: How do we use what we know about you and your interests and promote this app to you, without having a base of user-generated content that can kind of do that work for us? So we're looking at a bunch of things, like if you're interested in soccer, what are the use cases around Muse that could be really interesting or productive for that? Can we build units that show off those use cases, rather than using specific pieces of content that we found from the app? What we're trying to do with these promotional units across the apps is not just say, "Hey, there's an AI agent," but "Hey, there's an AI agent that can do this specific thing for you." Whereas with Threads, I think the hook might just be, "Oh, there's interesting content here. I'm willing to come back tomorrow because I want more of that." The hook for Muse is a little bit further down the funnel. All the social networks are grappling with how they want to deal with AI content. Do you guys have a policy or a worldview about me using AI to create all or some of what I make or what I post there? On Threads, we use the same sort of approach that the other Meta apps do when it comes to labeling what we detect to be AI-generated content. But we understand that there's going to be a bunch of people that want to use AI to either better their content or have a better presence on the platform. Substack, for example, came out and said, "We don't want to be the next LinkedIn," meaning LinkedIn is full of AI slop and that it wants to actively police for it. But it sounds like that's not where you guys are at. Actively police, no. But I think that the best version of the product is going to come from human-generated content. My sense is that human-generated content is going to win out when it comes to driving conversations.

Three years ago, Meta launched Threads as a Twitter alternative, staffed with a bare-bones team. Now the company says its text-based social network has 500 million users — more than Twitter ever reported when it was a public company (In 2024, Elon Musk claimed his rebrand of the service had brought in 600 million users.) Another marker of Threads' growth from side-project to Real Thing is that Instagram boss Adam Mosseri, who had been supervising Threads, has handed control of the service to longtime Meta product guy Connor Hayes. I caught up with Hayes at Meta's headquarters this week to talk about Threads' evolution, its stance on AI content — and how Meta is using the strategies it used to grow Threads as it pushes out Muse, its buzzy AI agent. Peter Kafka: When Threads launched in 2023, you were actively courting Twitter users who didn't love what was happening under Elon Musk's ownership. But Twitter is still around. Bluesky has its own user base. Who is a Threads user today? Connor Hayes: We assumed in the very beginning that growing Threads would be predicated on capturing people from these other services. We've actually been able to grow the market. It turns out there were a lot of people that were using Facebook and Instagram every day who are willing to use an app built around this type of text format that hadn't been using alternatives like Twitter before. In the beginning, it was purely, "I'm looking for an alternative." And it felt like a safer, more positive conversational environment. Now it's starting to grow into its own thing that's built around these communities. We're trying to go really deep on categories that people are super passionate about, in a way that I think Reddit has done a good job with in the past. We're trying to get into this space that's a hybrid of what Twitter was, what messaging is and what Reddit has been. A big way you've grown is by promoting Threads to Instagram and Facebook users. Now Meta is using that same playbook to promote Muse. Is promoting a new AI agent to Instagram users different than telling them about a new messaging app? The biggest difference is that Threads is built around user-generated content. The engine that powers our growth today are promotions we do on Facebook and Instagram, where we know that your account is linked and we can recommend you content from Threads on Instagram while you're scrolling. Of course, you're going to be interested in that, because we found a way to match it. For something like Muse, the big challenge is: How do we use what we know about you and your interests and promote this app to you, without having a base of user-generated content that can kind of do that work for us? So we're looking at a bunch of things, like if you're interested in soccer, what are the use cases around Muse that could be really interesting or productive for that? Can we build units that show off those use cases, rather than using specific pieces of content that we found from the app? What we're trying to do with these promotional units across the apps is not just say, "Hey, there's an AI agent," but "Hey, there's an AI agent that can do this specific thing for you." Whereas with Threads, I think the hook might just be, "Oh, there's interesting content here. I'm willing to come back tomorrow because I want more of that." The hook for Muse is a little bit further down the funnel. All the social networks are grappling with how they want to deal with AI content. Do you guys have a policy or a worldview about me using AI to create all or some of what I make or what I post there? On Threads, we use the same sort of approach that the other Meta apps do when it comes to labeling what we detect to be AI-generated content. But we understand that there's going to be a bunch of people that want to use AI to either better their content or have a better presence on the platform. Substack, for example, came out and said, "We don't want to be the next LinkedIn," meaning LinkedIn is full of AI slop and that it wants to actively police for it. But it sounds like that's not where you guys are at. Actively police, no. But I think that the best version of the product is going to come from human-generated content. My sense is that human-generated content is going to win out when it comes to driving conversations.