Sara Beykpour: The Next Wave of AI’s Impact On Journalism

9 May 2024 · 50 min

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

Podcast Notes: Generative Now | AI Builders on Creating the Future

Episode Title

Sara Beykpour: The Next Wave of AI’s Impact On Journalism

Episode Summary In this episode, host Michael Mignano speaks with Sara Beykpour, CEO and Co-founder of Particle, an AI-driven news platform aimed at reforming the journalism landscape. They discuss Sara's extensive experience in creating innovative media platforms, including Twitter, Vine, Secret, and Periscope, as well as the challenges within the news industry and how AI can contribute to developing sustainable business models for publishers.

---

Key Highlights

Guests and Background

  • Sara Beykpour
  • Former Senior Director of Product Management at Twitter.
  • Co-founder of Particle, an AI-driven news platform.

Major Topics Discussed

  1. Sara's Career Path
  2. Journey through Twitter and innovations like Vine, Secret, and Periscope.
  3. Reflections on her experiences in Silicon Valley's tech startup ecosystem.
  1. The Evolution of Social Media and News Consumption
  2. The transition from platforms like Vine to Periscope and their impact on content creation.
  3. Twitter's role in shaping the news landscape.
  1. Current Challenges in Journalism
  2. The disconnect between consumer needs and publisher business models.
  3. The rise of subscription fatigue among consumers.
  1. The Role of AI in Journalism
  2. Particle utilizes AI to summarize news articles, providing tailored content to users.
  3. AI as a tool to enhance the reading experience while ensuring content accuracy.
  1. Business Models for Sustainable Journalism
  2. Exploring new models that could support publishers, including revenue sharing based on content usage.
  3. Discussion on aligning publisher incentives with consumer needs.

Episode Chapters

  • (00:00) Introduction to Sara Beykpour
  • (00:58) Sara's Journey Through Silicon Valley's Startups
  • (02:43) The Early Days at Twitter and the Birth of Vine
  • (06:42) Vine's Impact, Decline, and Shutdown
  • (12:21) From Vine to Secret (Anonymous Social Media)
  • (14:45) The Periscope Era: Live Streaming's Rise
  • (21:47) Exploring the Evolution of Social Media and News Consumption
  • (22:26) The Impact of Twitter on News and Journalism
  • (27:57) Challenges and Innovations in News Consumption
  • (32:14) The Rise of Generative AI and Its Effects on News Publishing
  • (33:26) Leveraging AI in Particle for a New News Experience
  • (46:15) Exploring New Business Models for Journalism with AI
  • (49:28) Opportunities With Particle

---

Key Takeaways

Sara Beykpour's Insights

  • Innovation in Media: Sara has been pivotal in multiple groundbreaking media apps that shaped user interaction and engagement.
  • Changing Landscape of News: The traditional news model is struggling, and AI can play a crucial role in its transformation.

Challenges in Journalism

  • Consumer Experience: The current news consumption experience is fragmented, leading to widespread frustration among consumers.
  • Publisher Models: Many publishers are facing economic pressures due to outdated business models that focus heavily on clicks rather than quality content.

The Future of AI in News

  • Particle's Approach:
  • Personalization: AI enhances user experience by summarizing news and providing varied perspectives.
  • Engagement: Allows users to ask questions and encourages interaction.
  • Content Accuracy: Incorporates reality checks to verify AI-generated summaries.

New Business Models

  • RevShare Models: Proposing new revenue-sharing frameworks that compensate publishers based on actual content usage rather than traffic metrics.

---

Conclusion The episode offers valuable insights into the rapidly evolving intersection of AI and journalism. Sara's experience, combined with her vision for Particle, presents a compelling case for how technology can redefine news consumption and improve the sustainability of journalism.

---

Stay in Touch

  • Website: [Lightspeed Venture Partners](http://www.lsvp.com/)
  • Social Media:
  • Twitter: [@lightspeedvp](https://twitter.com/lightspeedvp)
  • LinkedIn: [Lightspeed Venture Partners](https://www.linkedin.com/company/lightspeed-venture-partners/)
  • Instagram: [Lightspeed Venture Partners](https://www.instagram.com/lightspeedventurepartners/)
  • Podcast Subscription: [Generative Now](http://generativenow.co/)

---

These notes encapsulate the key discussions and insights from the episode, providing an organized reference for understanding the critical themes and conversations about the future of AI in journalism.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:05Hey, everyone, and welcome to Generative Now. I am Michael Magnato. I'm a partner at Lightspeed And this week on the podcast, I'm talking with Sarah Bakepore. Sarah has been at the forefront of creating novel media platforms throughout her entire career. She built and launched apps for Vine, Secret, Periscope, and Twitter. These were truly groundbreaking products that set the stage for today's media landscape. Now, Sarah is the co-founder and CEO of Particle, an AI-driven news platform designed to both summarize stories of interest to the user, as well as allow deep dives on topics readers want to explore further.

0:41I talked with Sarah about her time riding the wave of social media startups, why she started Particle, and how AI could help make new business models work for publishers. Check it out. Hey, Sarah. Hey, Mike. How's it going? Good. Thanks for doing this. Yeah, thanks for having me. So obviously, we're going to talk all about Particle and AI, but I think it'd be really fun to talk about Vine. Sure, let's do it. And Periscope and Secret and all of the amazing companies that you were at. It feels like when I think about your career, I feel like all of the most explosive moments in consumer mobile apps, like for whatever reason, you've been there.

1:22Like you were there. And I don't know if that's luck. I don't know if you're, you know, one of the most sought after people in all of tech, like? Tell us about some of these experiences. I think a lot of it was right place at the right time. I went to school and I grew up in Toronto and I went to the University of Waterloo. And then my now brother-in-law interned at Google and kind of came back from that experience and was like, Sarah, you got to get yourself to Silicon Valley. And I was lucky enough to get a couple of internships at Google in my final years at Waterloo and just fell in love with California and Silicon Valley and San Francisco.

1:59When I think about where, you know, finding Twitter, being able to work at some of these awesome consumer companies, it really felt like right place, right time. It wouldn't have happened if I was anywhere else. And I look back on that and think how lucky I am to have found what, you know, what really speaks to me and in the place that where it's all happening. So what was the sequence of these opportunities? So you're in California. Yeah, there's we can rattle through them again. There's obviously Twitter. You spent a long time at Twitter. We're going to talk about that. You're at Vine. You're at Secret when that when that like totally took over tech for a couple of months or a year or whatever.

2:37Yeah. And then Periscope, which was also super explosive. So walk us through. Sure. The journey. So in 2009, I started at Twitter. OK. The company was about 75 people. It was a startup. And it started in what year? Oh, seven. Oh, seven. Oh, six. Oh, seven. It was really early. Yes. Yeah. It was startup life. We didn't have air conditioning. It was very hot. And everyone fit in a couple of rooms. And pretty soon after that, we moved to a larger office. But, you know, those early days were, you know, real, real startup days. I worked as a tools engineer. Okay. Building support tools for support agents, trust and safety agents.

3:18And I was, you know, a year out of college at that point. Even back then, Twitter's doing trust and safety. Oh, absolutely. Two years in, it was already that big that it needed that level of moderation. For sure. And there was a sizable number of agents and team. And at that time, it was all, at least I think, if I remember correctly, it was local. These were not outsourced teams. These were people, employees of the company that were doing trust and safety. Is this when people were still texting to use Twitter? Because I remember when I first got it, I had to text a number to tweet. Is that right?

3:5240404. Yeah. Yeah, I have that tattooed to my... No, I'm just kidding. There was a texting SMS interface, and that lasted actually a very, very long time. But when I joined, we had a website. It was before... It was not a Web 2.0 website. It was static pages. It was probably 2010 that we moved to a more Web 2.0 asynchronous JSON. What about an app? I guess you probably didn't have an app then, right? Wow. We had an API. And so there were third party apps on Blackberry and other phones. Twitterific. Twitterific. What were some of these? Seismic. And then eventually, Lauren Richter created the first real iOS client that we acquired.

4:40That was a long time ago. Early days, working on tools. I moved after a couple of years, I moved to mobile and started working on Twitter for Android. And that was also not an acquisition, but it was a demo app that Google had created and gave to us and said. And they did that with all of the popular social apps at the time, Foursquare. So Foursquare, Twitter, they all looked the same in those early days because Google wrote them all. Back up, explain this. So Google wants companies to adopt Android. Yes. So they build you apps. Exactly. Wow. And then they handed it to us. Wow. That's fascinating.

5:14It was fascinating. And a really interesting model, you know, given that they were building an open source OS and it was like they were really positioning themselves kind of differently from Apple. So we had taken that Android app and started to kind of work on it. And I joined, I think, about a year after we had kind of taken that app. And I was the second Android engineer. And we hired another one. And for a year, it was just three of us. So did that for two years ish. And then one day, Dick Costolo, who's the CEO at the time, comes up to me and he's like, what do you think about moving to New York for a bit?

5:56I was like, say more. He's like, well, we're acquiring a company called Vine. And they only have an iOS app and they need an Android app. So they're in New York. And this is, of course, before remote was like even a possibility. Right. And this was, you know, end of 2012, basically. And he was like. You're young, you're single, go to New York, have a grand time, build me an Android app. And I did. So I moved to New York in March of 2013, started the Android team. And Vine had not launched yet, right? Nobody, nobody in the world other than Twitter really knew what Vine was. And investors, I guess, a lot of VCs in the app.

6:39But yeah. Yeah. And, you know, I'm sure people who are watching or listening, I'm sure most people remember Vine, but maybe just to recap for them what Vine was. To the young kids today, Vine is kind of like TikTok if you kind of went backwards about a decade. um uh vine was six second uh videos uh that ran on a loop it had a really unique interface for you to record where you just tap and hold to record so you could create these cuts um and it really unlocked a lot of creativity the constraints of vine six seconds only tap to hold created uh or allowed for the creation of very uh neat and interesting things people would do stop motion and magic tricks and music and all that kind of stuff.

7:27My memory of the landscape of Vine is that for a lot of people, it was the first time they laughed from something on their phone. A lot of people experienced comedy on their phone for the first time. And I think when people talk today about bringing back Vine, which every year someone says, let's bring back Vine. I think Elon just tweeted about it a few days ago. I think what people miss is that feeling of joy that they had never experienced on the internet before, let alone their mobile phones. People are nostalgic for that. And I don't think actually anyone wants Vine. So I agree with you, but I also feel like the way I hear people talk about TikTok, they often express a similar feeling and that it makes them feel joy.

8:15They laugh. It's funny. And maybe, you know, not coincidentally, you know, you mentioned when you were talking about Vine, and how it was almost like before TikTok, before TikTok existed, a lot of, if I, if I'm correct, I think a lot of the users that became very influential on Vine went on to become really influential YouTubers and TikTokers. And so it almost created this, this new generation and this new form of creativity that kind of does still persist to this day. Totally. It's crazy. So many influencers launched their careers on, on, on Vine and, and, um, went on from there. Um, But, you know, it's an old tech.

8:50You're a VC. So, you know, if someone were to be like, I'm starting a new video sharing app, you might have a lot of skepticism in today's world. And that's kind of how I feel when people say they want to bring back Vine. I'm like, well, it was a moment in time and it's over. And if you wanted to do something novel in the video space, do something novel in the video space. I don't know that we need to go back to that. Well, when it launched, I mean, it was really, really explosive. I mean, I can't even imagine what that day must have been like for you and the team. I remember getting it. I was working at Aviary here in the city and it just took over our office and it completely distracted us from work that entire week.

9:28Everyone publishing and posting. Hadn't joined the Vine team yet. I knew I was joining, but I hadn't joined yet. And so I was still in San Francisco doing exactly what you were doing. It was that moment where you're like, video on the internet is going to change. Video online on mobile is going to change and you could just feel it. And of course, it was hilarious. Uh, my moments with Vine were when we launched the Android app. Which is a huge unlock for any mobile startup because it's like half or more than half of the world, right? Yeah. I was given a very strict deadline on this app. Now this is all, you know, water under the bridge.

10:02The, the direction I was given was you must launch Vine on Android before Instagram launches their video product. And we did. I think we beat them by about two weeks. Nice. And then the next big milestone in my Vine memory is the day Instagram video launched. Wow. It changed the company? It changed everything. Wow. It was significant. We saw the impact to the numbers immediately. Everyone has their own story of Vine. Everyone has their own history with it and their own relationship to it. In my history, that was the day the writing was on the wall. Wow. Wow. Okay. And it took years for it just to, you know, shut down eventually.

10:48But for me, that is, I knew it then. I remember when it got shut down, a lot of people were saying it was a huge mistake. And, you know, obviously, like you said, people to this day want to bring it back. What do you think? Irony of ironies, I led the shutdown. Oh, wow. Not in that, like, I wanted to shut it down and it was my decision. But in the, if I need to shut down, Sarah, you're in charge of making sure that all the boxes are checked. and GDPR is complied with and all of that kind of thing. So I actually led the kind of logistical shutdown of mine. I think it was the right move. It wasn't growing.

11:25It was expensive. We did not have the je ne sais quoi that a consumer app requires to keep going and growing and building. And it wasn't there anymore. And so we went through leadership changes. We, you know, some teams were working on different things. And I'm sure folks on the team will be like, well, we didn't get enough resources or the leadership team didn't believe in us or I disagreed with the direction. I'm sure everyone has their own story of what happened. Yeah. And I think that's probably why you hear so many stories about what happened. And my story is, is that the bottom line wasn't going to work for the company.

12:00The company was public at this time. We had other things going on. It was a rough time, couple of years, you know, 2015, 2016. I digress on that because this was after I came back, right? I left after Vine. Yeah. I came back and among other things. You left to join Secret? I left to join Secret. Okay. Yeah. It lets you post anonymous messages. It was anonymous. And then people could react to them. Yes. Yeah. But what was really the really interesting angle was that you uploaded your contact book so you could see if the anonymous post was from someone you knew or from a friend of a friend. Right.

12:36And that is what made it salacious. Yeah. It was just like, you know. Who posted this? Like, which of my friends said that? Yeah, yep. But it was also very, very explosive when it launched. It was explosive when it launched. I'll start to sound like a broken record because it was an unlaunched iOS app that only some people had access to. And the founders were looking for an Android engineer. So I had been at Twitter now for four and a half years.

13:05And I thought, you know, company had gone public. worked on a bunch of teams. I was like, it's time to do something different. I was still in my, you know, mid-20s, was kind of okay jumping into a startup, taking a risk. I met with the founders, David Vitale, Chris Bader, and Jude, and just kind of fell in love with that team. We got on so well, and they were building something that had not yet launched, but a lot of people in Silicon Valley were in. And it was just really exciting. So I joined them to build their Android app in early 2014. For the second time in a row, from March to June, I built and launched an Android app.

13:50Wow. Second year in a row. And that was really awesome, really fun. I mean, that team was extremely high caliber, extremely high caliber engineering design team. It was the startup dream. I had, you know, I was not married, had no children. I was coding all day, every day. And then like, just like going out to bars at night in San Francisco, and then just like doing it again the next day. It was the dream. The app launched, had the most massive spike of downloads, usage, all of that you could possibly imagine. And then started to plateau. Started to what? Plateau. Yeah. And these anonymous apps, they always seem to implode a bit, right?

14:31There's there's sort of a problem with the format. The moderation is difficult. Yeah. It was easy for that to get out of control. Yeah. Especially when it's anonymous. Periscope. Right. So a year later, Secret was kind of on its way down. It was kind of clear that it wasn't going to work. And to this day, I commend David and Chris for deciding to wind it down instead of, you know, trying to revive this thing that was no longer alive. a new app emerged that had not launched yet called Periscope. Wow. Once again, unlaunched. Unlaunched. Third year in a row. Third year in a row. And they had an iOS app and they needed an Android engineer.

15:17I got put in touch from a couple of mutual friends to Kayvon, the founder and CEO of Periscope. I met with a team. And at the time, you know, Periscope was a live streaming app before live streaming ever became a thing. So I first discovered Periscope while on a date. Someone I had met took me out to a bar. He's like, hey, have you heard of this app called Periscope? And I was like, no. And so he started a broadcast and he created some funny title like Broadcasting My Date. And I looked at it and, you know, people were joining and there were these parts floating. And I was like, this is incredible.

15:56This is an experience that we haven't had before. and I at the end was like you need to introduce me to the founder of this app because oh wow I you know this is this is something I would be interested in working on so he introduced me at the same time another mutual friend of ours introduced um introduced me to came on and um uh met the team and they were like we need an android engineer and I was like well for the third year in a row let's do this I uh you know went back to secret told them that I was going to leave, you know, did all of that. It was really bittersweet for me because I really loved working on that team.

16:36But I knew that we had to move on. So I went back to Kayvon and his co-founder, Joe, and they took me out to my favorite bar in San Francisco, which is Smuggler's Cove. And they're like, okay, we would love for you to join us. We have to tell you something before you join and that it's that we've sold the company to Twitter. And I was like, wait, what? I'm pretty sure I cursed. And I was like, I just left there. And then I calmed down a little bit. I called Dick Costolo right after that meeting because I had seen him that previous weekend. It was the Superbowl, if I remember correctly. And that whole weekend I was like talking about this company Periscope.

17:18And I was like, Dick, you should look at this app and this company. This is really cool. I'm thinking of going to work there. And he was like, uh-huh. Yeah. Okay. Okay. Did he already know about it at the time or? He had already bought the company. Oh, he had already bought it. Where I thought you were going was you had sort of tipped him off. Oh, no. And then he went and acquired the company right after you told him. He had already acquired. Oh, wow. Good poker face, I guess. I know. Yeah, great poker face. He was like, I know I didn't say anything because I'm not supposed to, but now that you know, you should come back to Twitter and we'll figure all that.

17:51Were you bummed that it wasn't going to be a startup anymore? Or like, did you reconsider joining? Well, they had this plan where they were going to stay separate from Twitter. Yeah. And so we were going to have our own office and it was going to stay separate. We're going to operate like a startup and all that. The other thing was that the acquisition was supposed to remain a secret for the foreseeable future. and but I was kind of I was rejoining Twitter and so we had to come up with a way to let me rejoin the corporate software without tipping anyone off that I had rejoined the company because you know that would raise questions like oh Sarah back at Twitter or like what's going on that must have been complicated so I yeah I rejoined Twitter under a fake name what yes and This is real?

18:41This is real. I rejoined under a pseudonym. The person that came and did my onboarding, Alex Stillings, who was one of the most awesome IT guys. He came to my desk at Periscope to do my onboarding personally and told me, like, put your real name here, put your fake name there. And so I reentered the system with no face and a fake name. What was the name? The name was Shana York. Wow. And to this day, I still get mail under Shana York sometimes. because, you know, for a while that was a name that, you know, had a mailing address and everything. I mean, I'm assuming at some point this leaks, like you said, there wasn't a lot of remote work going on back then.

19:26What happens? How did people find out that you were actually back and that Twitter had acquired Periscope? So for three weeks, it was really funny because, you know, I didn't go into Twitter HQ. I just stayed at the Periscope office. But sometimes people would come visit the Periscope office. So I remember the day Mike Davidson, who was one of the designers at Twitter, came to see the Periscope team and say hello and saw me sitting there. And he was like, what? Yeah. Why are you here? Anyway, three weeks later, a journalist, and I can't remember who it was, got the scoop that Twitter had acquired Periscope.

19:59Got it. So they had to make an announcement about it. And at that point, there was no reason to keep it a secret anymore. So I tweeted about it and everything was out of the bag. Did it launch around the same time? I feel like I remember, again, like Vine, it was sort of Twitter has acquired a company and here's the product. It was not long after that that it launched on iOS. Yeah. And I remember also having like a really interesting first use case with the kind of like your story. I was I was still at Aviary and we were in a different office, though. And I remember that there was an explosion of a building on the Lower East Side.

20:35and you could see it from the office. And I had never used Periscope before, but I just downloaded. I took it, I fired it up. And literally within a minute, I had like tens of thousands of people watching it. And I had like journalists reaching out to me, asking me if they could use the footage on like the nightly news and stuff. I was like, this is crazy. This is an experience, like you said. It was really fascinating. Really fascinating. It really changed the way you can connect online. with other people in a way that we definitely take for granted now because we have short-term memories. And now, of course, you can live stream with somebody else and all that.

21:13But yeah, it was super novel. I think it's really interesting. And obviously, we're going to talk about Particle, your new company and product in a little bit. But I think it's really interesting that these three experiences and really also Twitter, not only have you been sort of at the forefront of these new moments in consumer tech, but they've been around media formats, right? Six second videos, anonymous postings, you know, live video stream with interactive hearts and chat and stuff. But they've been these novel formats. It's it's super interesting. I don't know if it's coincidental. When I look back on it, it feels like it can't be coincidental.

21:52But at the time, it does feel coincidental because I was just kind of going to where the next thing was. And again, that's being in the right place at the right time around the right people. You know, finding the next company that you're going to work at because you're on some rando date is not going to happen anywhere outside of Silicon Valley. I mean, maybe it'll happen in New York, but, you know, that it just matters so much. So, of course, you ended up being at Twitter for a long time after that. I stayed like six and a half, seven years. Yeah, and leading product for a number of really, really big initiatives.

22:25I think this leads into Particle nicely. Obviously, Twitter is a place for media. It's a place for news consumption. It's a place not only for consumers of news, but also publishers of the news. So much of the media, journalists, publishers rely on Twitter for publishing. Talk about that experience. Talk about your experience with news and journalism and media throughout the course of your six years at Twitter. Like you said, Twitter has been at the forefront and the intersection of news and social and what's happening. And journalists are such a huge part of that. Most of my exposure to media and journalism at Twitter was through working with the press.

23:07and it was it's it's such an interesting relationship because not only are you talking to them because of your product announcements but they also use your product to you know to announce whatever it is that they want to announce whether it's what you're working on what somebody else is working on but now I'm talking to a lot of the corporate side of media publishers and and journalists so it's a definitely a different angle that I'm learning a lot about but the last thing that I worked on at Twitter one of the one of the last things was Twitter Blue okay which is the premium subscription product and we had acquired a company called Scroll that was led by a guy named Tony Hale who's a very good friend as part of Twitter Blue we included a product that allowed you to bypass ads or to have an ad-free reading experience and to push back the paywall a little bit if you were a Twitter Blue subscriber.

24:02And you could also see as a subscriber where your dollars were going, you know, to the publishers that you were reading and all of that. And so that was one of my first real forays into the kind of the business side of journalism and media at Twitter, kind of working with Tony on this project. You mentioned a minute ago that you're sort of, you're building for obviously readers and consumers, but also journalists and publishers. How do you balance that? And also, do you distinguish between sort of building for journalists versus versus building for anyone that wants to tweet? Right. What's the difference between a journalist and just me wanting to tweet, I guess?

24:38How does that define and inform your your roadmap, I would say? So I would say that this is actually, you know, looking back, one of the places where I feel like Twitter really struggled to to define its user because Twitter used a lot of traditional OKRs and metrics. And the fact that only if, you know, a fraction of people tweet and the fact that of the fraction of people that are tweeting, a subset of those are responsible for the content that everyone actually wants to see was difficult to measure. And so to build something for journalists where a lot of people, open up Twitter and they want to see what journals have to say, was often difficult to justify because it didn't directly impact the top line metrics.

25:33Right. And this is a problem that I'm sure many big companies struggle with. And you're smiling, so maybe you have that experience as well. This was difficult at Twitter to kind of get the resources needed to justify some of these power user features that are responsible for creating the content that everybody else wants to consume. I tried to push as much as I could to get these features into Twitter Blue. And the justification I could make was that Twitter Blue, to start, was going to be for power users, journalists or power users on Twitter. And we can justify that not because it's going to hit top line DAU, but it's going to have incremental revenue.

26:18So I found another metric to measure these folks by, and armed with that, was able to kind of get more things through. And how did journalists react to this? It was mixed. I think not all of the features that we were building were necessarily, you know, for the kinds of things that they wanted. I would have loved to have, you know, been able to spend more time there to build features that made their lives a lot easier. How did you see the consumer experience evolve throughout your time at Twitter when it comes to consuming news? Like, what was that like? And maybe what was it like back then versus what it's like today?

Read the full transcript

26:55One of the reasons why Twitter became the place to get your news from is because it was real-time breaking. And Twitter put a lot of effort into making sure that real-time breaking news could be disseminated as faster than an earthquake. That was the joke, but it was actually true. So with that use case and journalists knowing that if they were to tweet something that the people that need to see are going to see it right away, it became a place for people to understand what's going on and for people to get connected with what's going on around them. And that is the ethos, actually, that inspired Particle, which is we want to connect people with the news that they care about that's going on around them.

27:39So that's a large part of the motivation for starting Particle. Yeah. So let's talk about Particle. So you're at Twitter post Periscope acquisition for about six years, building for, I have to imagine, hundreds of millions of users, publishers, journalists, consumers of news, media. And obviously you leave and now you've started a new product that's tackling a similar problem. Yeah. Tell us about Particle. And then obviously I have a million questions for you. Sure. So Particle is a reimagining of how you intake your daily news. And we are trying to provide a personalized, streamlined, but also multi-perspective view of the news and the news that you care about.

28:18We're going all in on news and we're going all in on on publishers and high quality journalism. This is just from my perspective and what I hear from people outside of tech and business is I think there's a sense that news is kind of broken on both sides, the consumer side and and the publisher side. You know, consumers have to go to a bunch of different places to get news. There's, you know, some sites are ad supported, some are paywalled. It's just like a very disjointed experience. Do you agree with that? Like, how would you assess the news consumption experience today? Yeah, the news consumption experience today is hard.

28:55And it's really hard for consumers. And in many ways, I think it's also hard for journalists because the incentives are not aligned. Yeah, talk about that. What do you mean? The incentives for a story and the writing of a story, an article, are aligned with both creating high quality content, but also clicks and ads and views and scrolling down as far as possible. Like you hear people talk about that all the time on Twitter. And, you know, people accusing the media of being biased and, you know, really driven by clicks and ads. Like that's from your perspective, being in the trend is like that's real.

29:28That's real. It's how it's how it works for most news publishers. that the only way they monetize is through traffic. And there's the first click to get that page view, and then there's scrolling all the way down. So you'll notice that not everyone, but a fair number of articles bury the lead as far down as possible so that you scroll all the way and you see as many ads as you can. The incentives are not really aligned between content creators and consumers. And consumers just deal with it. And the, you know, journalists just deal with it. And there are other models, you know, there's paywalled models, subscription models.

30:15But I think consumers are getting frustrated with those experiences as well. I think a lot of consumers don't want to pay every publisher for content. Right. It's expensive. Subscription fatigue. Subscription fatigue. Yeah. And there are attempts to do things like a paper article or you buy some kind of set of credits. Some some there's been companies that have tried to do credits and then you spend credits to get behind paywalls. But I don't think people want to feel metered when they're reading random stuff online. Yeah. And I think it's different if you're you know, you're you're paying for, you know, an audio book or movie.

30:55You're kind of strapping in for we just scroll. We just we just scroll. We just bounce around. Right. Yeah. And you don't know what the quality of what you're going to see, you know, it's variable quality. And so I don't think people like to feel metered. So the subscription model is definitely better. But there are too many subscriptions. So it sounds like nobody's happy. Like consumers aren't happy. Publishers aren't happy. By the way, is this is this in any way in your in your opinion, contributing to kind of all the change that's happening on the publisher side? Like, I feel like I always read about newsrooms getting shut down, massive layoffs, consolidation.

31:29Is that a result of these two sides being unhappy in any way? Like, are those connected? I think it has to be connected because if consumers were happy, then some of these business models would work a lot more. Right. So the business models aren't working anymore. It's not working anymore. I think there have been changes as well from Google and Meta that have changed the way people get connected to the news in the first place. And so those business models are not standing up to how they used to be. Now, this is all even before the introduction of generative AI. Right. Right. So things were dicey even before generative AI.

32:07And what generative AI is doing now is providing an abstraction layer. Explain what you mean by that, by an abstraction layer. So with generative AI and with many of the generative AI products right now, you can get the information that you were looking for without fulfilling the promise of traffic. Right. Right. And that is, I think, going to be very disruptive to the industry. How does that work? So I go to a bot, I ask it a question, it delivers me information that maybe it pulled from a news source. I never go to that news source's website and see the ads. Yes. Yes. Got it. Or even if you're using a new search engine like Perplexity and you ask it a question and it pulls a bunch of answers from some sources.

32:55but the answer is good enough that you don't need to go any further. Right, and they do cite the source and they give me an option to go out to the source. But there will always be a tension between how good perplexity is at answering your question and its ability to drive the traffic. Particle is leveraging AI, obviously. So, okay, so how does Particle leverage AI in a way that solves the consumer pain point of not wanting to go to all these different sites and get my information from one place, but also in a way that is net positive to the publishers and their business models. So Particle aggregates article content from many different websites and summarizes the news for you.

33:39And it presents it to you in a way that you can easily skim. You can go as far or not as far as you want to based on the time that you have or the interest. So you're really kind of in control of that experience. And you can also see the source content of where it all came from. So you can go directly to the websites if you'd like. But we are interested in how we can partner with news organizations to come up with a business model that works for them and being able to continue to use that content. And do you think that involves sending the user out to the source website to activate the ads or the paywall?

34:30Is it some sort of monetization that then takes place on Particle that you share in in some way with the publisher? Like, how does that actually do it? I'm going to separate this into kind of two pieces. One is ensuring that the publisher is compensated, which is their, you know, the primary goal. And two is ensuring that their brand is respected. And so in order to make sure that they're compensated, we need to come up with new business models. And I'm very much interested and I'm talking to many publishers right now on how we do that. And my hypothesis here is that it should be correlated to the usage of the content as opposed to scrolling down or clicking or anything like that.

35:11So Particle uses your content. We should figure out how to make a business model work out of that. The second part is the brand. And in a world where we are abstracting information for the user, separating the answer that they're looking for or the summary that they're looking for from the source content, I think publishers wonder, okay, even if I'm getting compensated for it, though, what about my brand? And there are many things that we are thinking about into how we can incorporate brand into Particle so that users know when a particular piece of content has specifically answered their question that we highlight that for you.

35:55And there are other ways that we've been talking to publishers about in terms of how to incorporate their brand within the app. And that's something that we're talking about and we're definitely interested in pursuing. We sort of breezed past the user experience. You mentioned AI summarizes the news. Like, how does that help me as a reader? Why does the bullet point sort of skimmable summary make such a difference for me going to, say, you know, one of the big publishers' websites? I won't name any in particular. First of all, you get a summary, which means that you can get up to speed on the key facts of the story in much less time.

36:33And, you know, we're all busy people and we don't necessarily have time to read anything, to read everything. But, of course, if you want to go read further, you can. The other thing that it does is it allows you to perform AI analysis to go, you know, kind of deeper into the story, whether that's ask a question, have the summary rewritten for you in a way that's going to resonate with you more. So one of the things that we have in the app is an explain like I'm five. But so this is a complex physics story, something about black holes or large hadron colliders or whatever. And you can have the story rewritten in a way that kind of speaks to you more.

37:09So it's it's really tailoring the experience for you. Another thing that's important about particle is that we are sourcing everything from across the spectrum. And so you can understand the bias of the story and the bias of the coverage in a very easy kind of we've done all the work for you and we're just kind of giving you the result way. So you can understand if what you're reading is the result of balanced reporting or if it's one sided. In today's world, a lot of folks who feel like they, you know, they don't know where to get their news from anymore end up going to like a handful of websites.

37:49I had a friend who recently told me that she was like, well, I try to find the story on the New York Times and then maybe I try to find it on CNN to see if there's something else to be said there. But then I feel like I've really pushed myself into the left corner. So I begrudgingly go over to Fox News to try to find the story there and see if there's another perspective that I need to consider. And we're just going to do all of that work for you and present it to you in a very beautiful, easy to consume way. And if you don't want to go further than the summary, you don't have to. But if you are like, actually, this story matters to me in a way that I need to go deeper, we're going to make it super easy for you to find those those sources and, you know, do whatever research you need to do.

38:30I feel like I could see people taking kind of either side of this. You know, I feel like some people want the bias completely stripped. Right. They want it. They want just the facts. Right. They're sort of they're sort of sick of the spin, whatever side it's on. Whereas other people, I feel like they almost want the entertainment value of the news and they want the bias. They want the slant according to whatever sort of the side of the aisle they're on. Can you get both from Particle or is it just about sort of summarizing it down to the most basic facts? So currently we report on the facts that are most kind of present across all of the sources.

39:10So those are the ones that are going to be those are the ones that are going to bubble up to the top. And those are the ones that are going to be most likely to be facts because all of the kind of sources have reported on it. Um, but we are working on ways to bring in kind of individual analysis and voices from individual publications and individual people that will give you the kind of slant or, uh, kind of more personal, uh, uh, perspective that you're, that you might be looking for. So in the works, I feel like you mentioned, you can ask the news questions. I think that's, that's pretty interesting, right?

39:46We talked a few minutes ago about how people are leveraging tools like Perplexity and ChatGPT. So clearly, like people have questions about the world that they live in today. Right. And you're and you're letting them now ask them on sort of more of a micro basis around a particular story. I have to imagine that, like, people often have the same questions. And so I wonder if that's an opportunity to almost like keep the content dynamic and fresh and evolving and sort of catered to kind of what people do. people actually want to know about the story. Totally. One of the aspects of Particle that I think is unique right now is that our questions are public, which means that if you ask a question on a story, everyone else will see it.

40:31And maybe I had the same question and I appreciated that you asked it. So I will give you an upvote. And we're thinking about how we can add the social layer, which is an important part of understanding a story. There's the facts, there's the analysis, this, there's the commentary and, you know, bringing that last piece in around, you know, what are people saying? What are people asking is an important part of understanding the story. And we want to make sure that you feel like you have understood the whole story when you when you scroll past it. Not to sort of spoil you sharing it, but you you have an audio feature as well, which I think is really, really interesting for a couple of different reasons.

41:10Obviously, like people get their news in lots of different modalities. Some people read it, some people listen to it, you know, some people just watch CNBC all day while they're working. It feels like with AI, not only do you have the ability to summarize in a way that really wasn't possible, but you also have the ability to generate content in a way that wasn't previously possible. And so you can, in essence, generate for these other modalities like audio. Like, talk about that part of your strategy, because I think it's super, super interesting. Right now in Particle, we have kind of a V1 of a daily podcast feature.

41:45And we have an AI-generated podcaster who will kind of run you through the day's news. And we regenerate it a few times a day so you can get fresh stories throughout the day. And we're interested in how we can actually personalize that, make it really tailored to your interests, but also make it a little bit interactive as well. So maybe, you know, you're looking, you get in your car, you hit the play button, and you can see the five stories that are coming up. And you're like, actually, I really want to go deep on that one. Make that one one minute. Make that one five minutes. Make this one 15 seconds.

42:19or, you know, I maybe before you even get into the car, you scroll through a particle, you see a bunch of stories. You're like, add this one, add that one, add that one, add that one, and we'll generate a podcast for you on the fly. And because the voices are so good, right? Like the AI is not good enough. I think, you know, I don't know who you're using to power that, but, you know, companies like 11 Labs have gotten so good at this that you can literally just make it on the fly and it sounds like a human. It sounds incredible. The tech for AI generated voices is really incredible. Yeah. There's obviously a challenge with AI right now when you go to chat GPT or any of these products with hallucinations and, you know, really like trusting the accuracy of what's coming out.

42:56Now, that's happening on right now with these products. It's happening on a very sort of self-serve kind of on-demand basis. Like I have a specific question. I'm getting an answer. Okay, let me take this answer. Go cross-check it with some trusted sources to make sure that it's true. I feel like when you now apply this to news and maybe a behavior of consuming like a stream of news all the time every day in this product, the stakes get a lot higher for trustworthiness, right? I'm not going to have time to go check everything. How do you deal with this? So the first version of Particle, in the first version of Particle, we saw what is kind of the acceptable LLM hallucination rate, which is 1 % to 3 % hallucinations.

43:40And when we started generating these summaries and seeing that this would happen at about that rate, my co-founder Marcel and I, we kind of had this moment where we're like, this is not going to be acceptable for news. And so we put a lot of effort into building a pipeline that does what we call reality checks, but that's just an internal name, that re-verifies the outputted content against the source material. And it'll do that. So, okay, 1 % to 3 % hallucination rate times the reality check, another 1 % to 3 % hallucination rate. So you've now brought that percentage down significantly. And we will run the reality check up to three times and automatically verify that that output is correct before failing.

44:42We basically never see it fail more than three times. I think it's like maybe happened once. And if it does fail more than three times, we just discard the story. Oh, wow. You don't run it again. You just get rid of it. Just get rid of it. Wow. So you never have hallucinations, basically. It's zero percent. It's impossible. I would say the kinds of hallucinations that, you know, we kind of have come to accept as an LLM hallucination, we do not see. We sometimes see errors. And errors are not necessarily the same thing as I'm completely hallucinating. um but like um for example a problem that we see right now which we will solve and is solvable is if uh there's a big game coming up a championship game and every news um outlet is predicting the winner um the llm is not great at understanding that it's a prediction and not a statement right so sometimes it will uh state the prediction as a as a statement It's not a hallucination.

45:41We gave it all of this information that said Team A is going to win. Right. So it says the next one. Right. Right. But the game hasn't happened yet. But the game hasn't happened yet. Right. There is prompt engineering to fix that. That's not a hallucination problem, but it is still an error. Yeah. From strictly LLM hallucination perspectives, we do not see it. You talked a little bit earlier about finding new business models that support journalism and media and publishers in a way that they're going to be really positive around this technology. I think it's probably pretty obvious to many that the media industry and publishers are, I understand, they're always a little bit concerned around new technologies and platform shifts.

46:23They've gotten burned in the past by other very, very large tech companies, which will remain unnamed on this podcast. How do you navigate that with AI? Like, how do you how do you make it such that publishers are excited about the potential of AI and companies like Particle and not like, nope, we've seen this movie before. Like, we're not we're not doing it. We're going to we're sticking with our strategy to go direct. I think there are publishers that are that have that in mind and are still, you know, are still sticking to that strategy. And I think that's going to be, you know, depending on who the publisher is, that's going to be tough because consumers don't want to pay everyone.

47:02And so, you know, we'll see how some of that goes. I think a lot of publishers and a lot of the ones that I've been talking to are getting very excited about the future with AI. They know that AI is here and it's not going away. And we need to figure out how to make new business models work. And I think there's been a lot more appetite for that in recent months. And it's trending in a positive direction. What's an example of a new business model that could work here? So a new business model would be around RevShare. So something like paying for content used. Based on consumption? Based on consumption, either all the way down to the user consumption or the app like Particle can just count every time they use something.

47:49And we can correlate it to eyeballs, but we don't have to. Right. I feel like there are lots of other things that could be summarized for me. Yeah. You're obviously so focused on news right now. You started the conversation there. But where else can this go? Do you plan to summarize other things for me? We're really interested in summarizing other formats like podcasts, for example. There's a lot of podcast content and there's definitely not enough time in the universe to listen to it all. and wouldn't it be great if you knew that, you know, 15 second segment about something that you really care about in this podcast and here's the timestamp and you can go directly to it.

48:29So we're thinking about how we might kind of bring in podcasting into the source of data. We're thinking about how we summarize all sorts of things across the internet to enhance the experience. Yeah. I mean, I could imagine, I don't know if you want to get into the enterprise, but like, I can imagine you summarizing my email or like an important document or paper that I need to read or I don't know. I mean, they're probably like an unlimited number of things that you could help me learn faster. Yeah. I think we'll see a lot of companies summarizing a lot of stuff in the next couple of years without trying to boil the ocean.

49:05We think that the particle experience that we're starting to create is going to be a great way to consume lots of different kinds of content. Awesome. Uh, are you hiring? Yes, we are hiring. We are hiring engineers, uh, across the U S we're based in, um, in San Francisco, just north of San Francisco. Uh, but, uh, hiring all over the U S and we've got, um, you know, mostly remote employees, what roles and like, where can people find out about them? Um, you can find out about them, soon on Particle.News. And we're hiring back-end engineers, SREs, and full-stack engineering. Awesome. Sarah, thank you so much.

49:45This has been awesome. Thanks for having me, Mike. Yeah, have to do it again sometime. Appreciate it. Yeah. Thank you so much for listening to Generative Now. If you like what you heard, please do us a favor and rate and review the podcast on Spotify and Apple Podcasts. That really does help. And if you want to learn more, Follow Lightspeed at LightspeedVP on YouTube, Twitter, or LinkedIn. Generative Now is produced by Lightspeed in partnership with Pod People. I'm Michael McNano, and we will be back next week with another conversation. Thanks so much.

From the publisher

One could argue that the model for the news business is broken. So can AI help fix it? This week on Generative Now, Lightspeed Partner and host Michael Mignano talks to Sara Beykpour, CEO and Co-founder of Particle. Particle is an AI-driven news platform designed to help people learn exactly what they want to know about what's going on in the world in the most efficient way possible. Michael and Sara talk about the central product role she played for groundbreaking products like Twitter, Vine, Secret, and Periscope, as well as why she started Particle and how AI could help make new business models work for publishers.

Sara has been on the forefront of creating novel media platforms throughout her entire career. She has built and launched three apps that set the stage for today’s multimedia social landscape online: Vine, Secret, and Periscope. She previously worked as Twitter’s Senior Director of Product Management, leading Twitter Blue, Twitter Video, and conversations. 


Episode Chapters

(00:00) Introduction to Sara Beykpour

(00:58) Sara's Journey Through Silicon Valley's Startups

(02:43) The Early Days at Twitter and the Birth of Vine

(06:42) Vine's Impact, Decline, and Shutdown

(12:21) From Vine to Secret (Anonymous Social Media)

(14:45) The Periscope Era: Live Streaming's Rise

(21:47) Exploring the Evolution of Social Media and News Consumption

(22:26) The Impact of Twitter on News and Journalism

(27:57) Challenges and Innovations in News Consumption

(32:14) The Rise of Generative AI and Its Effects on News Publishing

(33:26) Leveraging AI in Particle for a New News Experience

(46:15) Exploring New Business Models for Journalism with AI

(49:28) Opportunities With Particle


Stay in touch:

The content here does not constitute tax, legal, business or investment advice or an offer to provide such advice, should not be construed as advocating the purchase or sale of any security or investment or a recommendation of any company, and is not an offer, or solicitation of an offer, for the purchase or sale of any security or investment product. For more details please see lsvp.com/legal.

More from Generative Now | AI Builders on Creating the Future

All 90 episodes
Sara Beykpour: The Next Wave of AI’s Impact On JournalismGenerative Now | AI Builders on Creating the Future · 50 min
Listen in VO