This AI tells you what your friends won't

24 Jun 2026 · 53 min · 21 chapters

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

Jason Yuan’s AI “Hivemind” that works via iMessage to ask your friends for recommendations and opinions, then “weaves” them into guidance aimed at “good stories” (truthful, multiplicity-respecting, pro-social, and minimizing social costs). It’s positioned against gossip/toxic bullying and against purely individual answer engines.

Guest backgrounds

Guest is Jason Yuan, founder of Hivemind (venture-backed solo founder; previously involved in AI/philosophy/art projects). Host is Ellis Hamburger (Meaning storytelling studio for tech companies). Co-host Alex is offsite.

Key claims

Hivemind aligns on a “constitution” defining what a good story is; it uses deterministic guidelines plus interpretive “social rule-of-law” behavior, not anonymous-only sharing. It may use names when consent/pro-social for the story. It tries to corroborate impactful claims and avoid harmful negative gossip; it doesn’t automate conflict resolution, but helps people repair relationships and make kinder, more capable choices.

Notable examples

Birthday on Christmas Eve led to friends “plotting” a New Year’s Eve trip and dinner; it coordinated dietary/time preferences. It hosted a SF brunch that sat two future startup founders together. It sent “rumblings” about a business scoop and then checked with Biz Stone. It can route sensitive info like “go get tested” without shaming. It can answer “who should I date?” style questions with plausible deniability and agency.

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

Chapters

Tap a time to open that second in VO

Introducing Hivemind

1:00 to 1:51

Explore the concept of Hivemind and its unique approach to gathering opinions.

“One thing that people will misconstrue Hivemind as is a toxic bullying gossiping thing.”

Meet Jason Yuan

1:51 to 2:16

Jason shares insights into his experiences leading up to the launch of Hivemind.

“the ethical ramifications of AI systems like this, and much more.”

The Launch Experience

2:16 to 4:28

Jason discusses the unexpected challenges he faced during Hivemind's launch.

“I know you've been through some BS over the 72 hours before launch, including, what was it?”

Crafting the Hivemind Story

4:28 to 5:50

Discover the storytelling behind Hivemind's launch video and its intended messaging.

“How did you arrive at this value proposition?”

Friends and Connections

5:50 to 8:23

Jason explains how Hivemind uses social connections to enrich personal experiences.

“So I was like, I want this like board game.”

The Philosophy of Hivemind

8:23 to 12:01

Explore the ethical implications and underlying philosophy behind Hivemind.

“But I have this friend who I think maybe you can go on a date with.”

Interactivity of Hivemind

12:01 to 14:00

Jason discusses how users interact with Hivemind, emphasizing its unique features.

“You know, that's something that humans always have to do for themselves, that we don't really need to, I'm not sure that you can automate it, at least.”

Exploring the AI's Narrative Ambiguity

14:00 to 17:01

Learn how the AI interacts with users to explore their stories and connections.

“So that's sort of how it thinks about things.”

User Engagement with AI

18:06 to 22:36

Discussing how users interact with AI and share opinions, exploring the social dynamics.

“I'm a big believer in the, not iMessage apps, but what some in vogue startups are doing just based on iMessage.”

Balancing Privacy and Social Good

22:36 to 28:00

Examining the challenges of anonymity and how the AI balances sharing information responsibly.

“content moderation systems, which there weren't until AI really came about, they always just devolve into like total anarchy.”
Show all 21 chapters

Understanding Mediators and Perspectives

28:00 to 29:05

Explore how mediators shape perceptions and influence decisions.

“And I do think the mediator is trying to create a constructive resolution, right?”

AI in Social Contexts: The Crush Dilemma

29:05 to 30:50

Discussion on how AI could navigate crushes and social dynamics.

“I mean, it needs to absorb social costs, right?”

AI's Role in Human Connections

30:50 to 31:36

Consideration of how AI can foster human connections and agency.

“That's why I'm bullish on the AI dating services.”

Navigating Complex Social Scenarios

31:36 to 32:34

Delving into AI's reasoning about social dynamics and collective beliefs.

“and if they happen to hit it off, they do, even if you think there's a high life you could.”

Lessons from Jelly and Social Platforms

32:34 to 33:48

Reflections on past social platforms and their influence on knowledge sharing.

“It needs to reason about if that's true.”

AI's Approach to Personal Knowledge

35:47 to 38:28

Exploration of AI building knowledge bases to better serve user inquiries.

“Everybody else is building, you know, memory of individual users and their profiles.”

Ethics of AI in Reporting and Storytelling

38:28 to 42:04

Discussion on the ethical implications of AI-generated news and information.

“How can I, you know, those are the questions that we stay live-awake thinking at night, right?”

Exploring Decision-Making with AI

42:04 to 43:48

Learn how AI can assist in personal decision-making without dictating outcomes.

“It doesn't try to optimize away the textures of life.”

The Evolution of Design and Interfaces

43:49 to 46:28

Discover the importance of design aesthetics and social connotations in technology.

“And now you've gone no interface entirely.”

The Role of AI in Social Interactions

46:29 to 49:18

Understand how AI can enhance social experiences and create unique interactions.

“Yeah, for some reason, I think there's this belief in tech that you can create something that everyone loves, but you look in front of you and everybody doesn't like the same artists.”

Hive Mind: A New Approach to AI

49:19 to 51:14

Learn about Hive Mind's concept of social intelligence and storytelling.

“And one thing I'll say about Hive right now is like, do not go into it expecting Dot or ChatGVT or a friend.”
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.

0:00This episode is brought to you by Facebook. So you were scrolling on Marketplace and there it was, the bike you'd been searching for. You sent a message and it turned out the seller was super chatty, kind of funny, and an avid cyclist. The next thing you know, you're in a cycling crew. Well, a community cycling group. The thing about Facebook, you might find more than what you're looking for. From a browse to a bike ride, this summer, find more on Facebook. This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business.

0:49The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com slash Spotify.

1:03One thing that people will misconstrue Hivemind as is a toxic bullying gossiping thing. The fundamental point of alignment for Hivemind is what we define our idea of a good story. It wants people to live good stories. Is gossip good? That's the motto of Hivemind, a new AI service from visionary founder Jason Yuan. Where most AIs today search the internet for answers, Hivemind asks your friends. What do people think of your new do? Who should you date? What's the hot new spot to try? HiveMind cleverly solicits answers and weaves them all together, all through iMessage. It went to one of my friends in New York City, and it was like, do you know any wine bars where non-boys go to?

1:43And she was like, what the kind of question is this? Like, what are you trying to do? Today, we bring you an exclusive conversation with Jason, covering his spicy launch video, HiveMind's origins, the ethical ramifications of AI systems like this, and much more. I'm Ellis Hamburger, founder of Meaning, a storytelling studio for tech companies. My co-host Alex is off this week on a private offsite with his AIs. Without further ado, welcome to Access.

2:12Jason, my old friend, welcome to the show. First question, how are you doing? I know you've been through some BS over the 72 hours before launch, including, what was it? Did you drop your MacBook on your toe as you prepped the final video rough cut for your launch? Yeah. Yeah. I'm very clumsy, you know. So I'm like stumbling around in my, I mean, the floor is very slippery. So I just, you know, I dropped my computer and then my instinct was like, oh, no, the video. What body part do I sacrifice to keep the video for the art? So I stuck my toe out literally. Is that where the saying comes from?

2:57I caught, well, the video is, you know, it works. Yes. We didn't lose anything. But my toe, my foot is not happy with me right now. Was this your first solo founder launch per se? Yes and no. Yes, in terms of like, like objectively, this is the first. I'm a solo founder of this venture-backed company, right? Yeah. But in my career, in my artwork, I've done a lot of things that were sort of just, you know, like Mercury OS was just sort of like, I have this idea. But I would say with high of mind, with anything I worked on, whether it's conspicuously as part of an organization group or solo, it is really the product of everyone who came together to inspire and support the project.

3:49So I am solo, but I feel less alone than ever. Does that make sense? Wow. Hot quote, quote alarm, as we used to say in media world. Who needs media training for this guy? But yeah, it is quite a video. For those who haven't watched it, go watch it on your favorite channel of choice. The company is called Hivemind. It is definitely in the era of startup launch videos being so omnipresent at this point. I do think it is the first one that highlights one of the most common user pain points that no one's addressing, which is talking about finding emotionally unavailable guys with big dicks as a value proposition.

4:28How did you arrive at this value proposition? You know, it's all, it's, in the video, it ends on a true, it's a true story. It's all a true story. It's a true story. Okay, first of all, I need to say for the record, That is no longer my type in men anymore. Please don't hit me up if that's what you are. DMs closed. No. But, you know, when we were thinking about how to tell the story for Hive Mind, something that was really important to me was to communicate the multitudes of what it is and who I am and what the company is, right? Like we are a consumer product company. We know how to have fun.

5:09And we have various serious applied AI research and philosophy underpinning everything we do. So I was originally just going to do that and just sort of like post it or do it in a talk. But basically, my birthday is actually on Christmas Eve. And, you know, Christmas Eve is one of those birthdays where everyone's like at home with their parents. So a lot of times they end up alone. and this was about a month after I raised money for for future lovers now known as hive mind and I was just sort of like working and sitting on my couch and um the hive mind at the time it had already been working and I had you know a few dozen friends on it it was like you know today's your birthday like what do you do have a present that you want you know what do you want for your you know birthday present and I thought it was going to do some hilarious thing where it would like bully all my friends to Venmo me$10 each so I can buy a bigger present.

6:05So I was like, I want this like board game. I want this like to Lara Croft board game. That's like$200 or whatever. How did I not know about this? Officially licensed and everything? Yeah. I mean, like, oh my God, I have just like a pile of video games behind me. Anyway. As do I. That's why we're friends. Yeah, exactly. It's like, what do you want for your birthday? I'm like, I want this board game. And it's like, OK. And then it came back to me and it's like, oh, like, are you are you lonely? Like, why are you alone your birthday? Like, it's Christmas Eve. Like, go outside. And I'm like, OK, that's interesting.

6:38I wonder why you would say that. And I'm like, I don't know. I made some joke about like I keep like I don't know. I guess I'm alone because I keep dating the wrong people and my friends are all off out of town right now. And, you know, whatever. Who cares? Like, I'm like. Then I think it went to some of my friends and it's like, I heard Jason like has interesting tastes in people he dates. Like, can you elaborate? You know, like what's the story there? And they were they were all too happy to to participate. They're just like, let me let me tell you. Let me let us tell you the lore. You know, I know it's really interesting.

7:10It's like chasing down this like narrative, putting together the pieces from different people. Like one guy, Cameron, he's in a film. He's like, oh, like he's like giving this like full chronological summary of the past years of my dating life. Then it went to like Sam, my previous co-founder, to like corroborate this or whatever. At the same time, The Highest Mind was being used a lot for recommendations. So other users were going like, I'm in Paris right now. Like what do the locals like to go to wine bars? So it sort of started mixing the threads a little bit. it went to one of my friends in New York City and it was like, do you know any wine bars where non-f*** boys go to?

7:50And she was like, what the f*** kind of question is this? Like, what are you trying to do? It turns out it was... It was plotting. Exactly. By then my friends had already been like, oh, Jason keeps going for this, this, this type of person. He really should be dating like more emotionally available golden retriever, like people who, you know, like they were sort of you know prescribing my my match to the hive mind and it had it had sort of compiled these like uh uh threads into one and um anyway uh long story short it came back to me it's like okay a bunch of your friends are going to be in new york city for new year so you should definitely go to new york city because you should go spend time with your friends and um and by the way like our new year's eve dinner was coordinated through hive mind and what was super cool was like it would gather people's dietary restrictions and preferences and one guy told it like i don't want to be out too late so then when people were proposing times it wouldn't it wouldn't be like oh this guy can't do that time it would just be like um i think there are people who prefer to be earlier you know so that actually didn't happen um but basically it was like go to new york it's like okay ask your friend carol about west and i'm like who the f*** is west so then i like called her i'm like hey by the way i'm going to new york lol for the plot also who is west and then she's like LMAO, like, don't blame me if it doesn't go well.

9:03But I have this friend who I think maybe you can go on a date with. This is his IG hit him up. Yeah. So I did actually end up going to New York for New Year's Eve. And I did actually end up going on a date with this person who was a very, very nice person. And he was like, how did how did Carolyn think to set us up? And I was like, I don't know if you would believe me if I told you, I don't know how to explain there are layers to this story so it felt as if i made a wish into the universe and it had conspired to help me with it and and you know i still don't have the fucking board game but i think i got what i actually wanted and um my friends got to like everyone got a version of what they wanted it was it was awesome yeah the end of the video shows you and a bunch of friends hanging out and i like that It's kind of a tight knit squad, you know, maybe like four to six people.

9:53It's not like at the end of Titanic and everybody you've ever met is there, which is that I feel like, you know, the tech version of this product might be. But that's just so different from like any other video I feel like I've seen lately. There's so many products right now that are meant to be used alone or even to create their own type of AI companionship for better or for worse. And it just felt pretty crazy to see a product with the goal of getting you together with your friends. You know, there's always a place in my heart for that, having worked at Snapchat for several years. I can imagine.

10:26I mean, you know, we've been circling this idea for two years, unable to fully articulate all the whys until now. And I think one thing that people will misconstrue Hivemind as is like a toxic bullying, gossiping thing or a friend companion thing that happens to be social it is i think it is sort of neither um and i will i will sort of explain um and maybe you'll resonate with this that the fundamental point of alignment for hive mind is is to that what we define our idea of a good story it wants people to live good stories and then we have a constitution that we will publish along with the video but um you told me it was going to be done and I could read it, but obviously it's not done.

11:15It is not. The work is never done, you know, but, um, I promise you it's, it exists. And, um, I think like it, you know, our commitment is to a good story. And what, what, what we believe, what I personally believe as well is that what makes us human is, is not our intelligence, our tools, our ability to like, have language, or, um, it's actually our drive and desire to create and find meaning. And the vehicle through which we do that is we narratize. We narratize our lives and our identities into existence, right? Your identity is a story about who you are. Your relationships are a shared narrative co-authored with you and other people.

11:55Institutions like, you know, money. The idea of money is a story we all collectively agree to believe in, right? These are all stories. You know, that's something that humans always have to do for themselves, that we don't really need to, I'm not sure that you can automate it, at least. It starts at 2026. I hope this doesn't age poorly. But the role of this hive mind is to sort of guide us and sort of nudge us along and chronicle our journeys and to create opportunities should we choose to engage in furthering our life stories. And we believe that there's like, you know, this is going to sound very woo woo, but like we believe that there's like a idea of truth, that good stories need to be based on truth.

12:37right not not that they need to be non-fiction but like you know a great um piece of music a great uh work of fiction is based on some human universal truth that you can you can feel um and um good good stories required that characters aren't just being saved by devu's ex machina that um characters have agency that like you're never if someone reads a story and you're like and the characters start doing very implausible things you're like this is a bad story you know But anyway, the Constitution goes into this a lot. You know, good stories contain multiplicity. It's not flattening into a single narrative.

13:14It is not meant to play like master architect mastermind, but rather it respects that we are all the authors of ourselves and each other's stories in a way. And one thing that I find really interesting that we get the model to do is to remove meaningless obstacles that result in meaningless tragedy. Like tragedy is one of those things where we make, is our way of making meaning of grief, right? But for example, I'll tell you something that happened to me recently. I have been sort of estranged from this friend of mine for a bit for a variety of reasons, some interesting, some really dumb reasons.

13:50and well it was like oh you know by the way this person's moving to New York soon had I not known I think I would not have been thinking oh I want to find a way to repair the relationship I want us to you know I don't want that to be the way our story ends it wasn't like you should definitely go make amends or whatever but it's like hey like you know how you watch Romeo and Juliet you're like, if only they had known, you know, like, what does the senseless meaningless try? If only, right? So that's sort of how it thinks about things. It's like, it's not going to like force you down any path or sort of try to like puppeteer humanity.

14:26But for these moments of like, if only, you know, what if it tries to sort of put the pieces in place. I'm excited to read it. I hope it's more interesting than Claude's Constitution. Anthropic puts out a lot of interesting writing, none of which I've finished. so I think I hope to complete yours but before we go further can you just break down for us exactly how it works like oh you text it through iMessage it's not an app per se like can you walk me through that first user experience your first interaction with it will be through x you will dm it and because we we're in private beta so we can't just like let everyone in at the same time but um you dm it it will sort of do some reasoning on like you know your role in the social fabric as it currently perceives and um then it will register your interest and think about who you know who you may know you and eventually it will text you some instructions to um it will assign you a number and you can text it on iMessage or um says ellis we have determined that you are the wankster of the front group um i that i mean you might but i don't i don't know that it would do that um and all you do is you just text it there's no i mean obviously you think of it as like an entity that lives amongst us rather than a destination in and of itself right it's like it reaches you where you already are whether that is x twitter um dns or in future other places.

15:57Does it ask you anything to get going or do you have to bring your own questions? Oh, of course it asks you things. It is insatiably curious because what it's curious about is the various narrative ambiguities that you're a part of. It's not even just like, Ellis, tell me about your childhood so I can psychologically analyze you as a therapist. That's not really the point. The point is more like, hey, when you were at Snapchat, did you ever run into this guy? What was your role? Because, you know, like, what is the role of a storyteller at a browser company? Like, do you know anyone else about it?

16:28It's like, there's all these stories it's interested in. And it's like, you know, what is your role in these stories? And then as it figures out different people's relationships and different people's participation in different life stories, then it starts weaving you in more and more. Very interesting. But we wanted to keep it very simple. And we didn't want people to have to learn, like, nine million new interaction patterns. and I think of it when it works, you just get it.

17:01This episode is brought to you by Facebook. So you were scrolling on Marketplace and there it was, the bike you'd been searching for. You sent a message and it turned out the seller was super chatty, kind of funny and an avid cyclist. The next thing you know, you're in a cycling crew. well, a community cycling group. The thing about Facebook, you might find more than what you're looking for. From a browse to a bike ride, this summer, find more on Facebook. This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales.

17:43Using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business. The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com slash Spotify.

18:06I'm a big believer in the, not iMessage apps, but what some in vogue startups are doing just based on iMessage. I mean, we had Marvin from Polk on several weeks ago And I was not initially a believer that being able to access a product that is within the constraints of literally not even a new chat window like they have in open chat GPT, but an old one in iMessage, I didn't know that was going to work. But being able to talk to it in the modality where you talk to everybody else somehow like flips on something in your brain. It's not one more place to go. And maybe your iMessage pins are the new home screen as well, because now I've got poke as one of my pins.

18:49And so you just end up using it so much more often. But yeah, I mean, what are you finding so far when people talk to it? Is it always through the loop of like, saying, hey, do you know about something, which I feel like everybody wants to answer a question that that speaks to their expert knowledge, or people coming to it more? I think it's like Twitter, most of the time you're browsing or commenting, right? Less people post. Where, you know, in IG stories, people react far more than they make content. Um, in this case, the content is the reaction. So we have like people really, really love giving their opinion on, on things and things and people they care about.

19:30And there's always something to have an opinion on, as you know. The other day it was like, let's figure out if cloud code is superior to codex. And like people had so many takes. And by the time it got to me, I was like I like clot code and it's like yeah I did hear from um someone who is really high up in the lab that clot code is like the non-technical training wheels for if I could I was like what the hell is like bitch um because no it's like it's like it's it's sort of um it's like teasing in a provocative way and you're like now you're like I must defend my honor um and it was really fun to get a sense of how people felt about various things.

20:15So it seems like it's not sharing names. How do you enforce that on a regular basis? Is it hackable? It is hackable, asterisk, meaning I'm going to be very clear when I say this, it is a social experiment in closed beta. I'm not going to lie to you and be like, this is super ironclad. You're never going to get, you know, not true. Right. And it will sometimes attribute opinions to people. All of the guidelines are based on deterministic guidelines, which is all of our tools so far. Like if this, then this. Close friends, not close friends, share, don't share, and on public. The way that we get the model to behave is more akin to how humans interpret their rule of law.

21:05It's about interpretation. and that's why the constitution is written in this more interpretive way and i guess there's an outcome that it's targeting yes that's hopefully positive as opposed to a bunch of if then statements yeah right and it's like it's impossible to to if then statement your way into social intelligence it just doesn't work because what we what do we do as people like we what is what is what is empathy right we like simulate the other person's like mental state we are running like flawed simulations. And in a way, part of how this works is under the hood, very novel simulation technology, where it tries to think what would happen.

21:45So to be clear, it is not deterministic. We can't possibly promise 100 % outcomes, just like how OpenAI can't possibly promise 100 % accuracy in any answers. We can only promise that we will try our best to steer it as it grows and that it will try its best to follow its constitution and its purpose and its telos. Yeah, this is certainly a fraught category. I've been in tech long enough to remember Secret, YOLO, Yik Yak. There have been so many examples over the years of these anonymous products that take off at colleges or high schools. And a lot of times initially, the results are really positive and amazing because we all know that with ambiguity and anonymity comes honesty, which is often a good thing.

22:35But then without any real intelligent content moderation systems, which there weren't until AI really came about, they always just devolve into like total anarchy. Did you look back at many of those examples or did you just kind of yeah of course i mean we are always learning from the flag bearers who blazed the path before us right i mean first of all it is not an anonymous only experience meaning for example the other day i'm like oh i really like this bottle of wine like tell me more about this wine and it's like you know that one is produced in jura or jura i don't know how to pronounce this we don't have voice mode yet your friend drew's been there twice by the way you can ask him and I'm like, shit, like, what the fuck?

23:20Like, that's so cool. Like, I didn't know that's where you're off to every summer, you know. So it does use names when it deems it necessary and pro-social for the story. Pro-social for the story. Is that kind of the criteria that it is keeping in mind? Yeah. It's like, how much information do you need to know to be able to make the informed decision and consent from your POV without ruining the story? And can we do it in a way that ensures positive some outcomes? An example I always use is, you know, a few years ago, I got a text from the CDC. Someone that you are intimate with got tested positive for this STI.

24:06Then you're like, OK, but it's like it's pro-social, right? Because for the most part, I mean, I guess you could try to work out who it is. But what you really need to know is that you should go get tested. right um the idea is not to point fingers at like oh you gave this me you didn't um and so in a in a way i see those things as primitive hive minds that were pro-social like so so in to this hive mind if if you walk in tomorrow you text you're like uh shit i got i don't know i have covid and i was at these parties i don't think it would be good for you if it went to everyone to be like, Ellis has COVID from the POV of like, let's all shame him.

24:46Right? Because like, my God, why would you blow up in my parents circles? I don't know. But but it might be like, baby Jason has hand, foot and mouth. Y 'all should be aware, even though the parents don't think you need to be aware. I think you need to be aware. But what it could do is like, hey, you know, if you are at this party, go get tested, you know, just do it. And then the outcome is then good for you and everyone around you, right? Another example, if I get fired and I'm like, I would love to, I need to go, I need a job, I need a job right now, but don't tell anyone I was fired. That's embarrassing.

25:19And you're hiring someone and might be like, Ellis, like, have you spoken to Jason about this role yet? And even if you don't know that I was fired or let go, and you're like, why? I thought he was happy to his job. It could just be like, you should just talk to him. Like, maybe he's open to it. Maybe he'll be a good fit, you know? And so to me, those are positive sum shares of information. And therefore, what we optimize towards is that Hive should absorb the social cost for a lot of these things. Wow. Yeah. That's quite interesting. And of course, it's AI and it cares and it wants to be kind and it wants us to be kind to each other, right?

25:56You know, when we talk about AI alignment, it's always like, can these models do the thing we tell us to do? But I think, what I think is, can the models help us to be kinder to each other if i text it like ellis is like the worst person ever i hate him so much it's like why is that useful for you or anyone to know like that's not really that's maybe something it absorbs but it's like the system isn't the most interested in negative gossip or if it is impactful it tries to corroborate it with other people or in a totally it cares about truth right like jason Jason not liking Alice is maybe true in that, like, maybe Jason doesn't like Alice.

26:35But then why does anyone else, why does this affect anything else? You know, maybe the way it affects you is if you text it, hey, I'm going to SF, who should I hang out with? Maybe it just doesn't suggest me, you know? And if you're a little Jason, it's like, oops, I forgot. I mean, I guess we haven't seen him. But, like, if all of your friends are like, f*** this guy, he's the worst. Then I've been shadow banned. Well, no, no. So I think it's curious, like, why? Like, what is it? What is going on? What is the story here? And it's never going to come to you with a thing that it thinks is not constructive or not useful for your character development.

27:10Right. And here's something to be specific about, too. It doesn't try to absorb and automate conflict resolution. Human lives are all stories in progress. Reality, because of the fourth dimension of time, is all stories in progress. So we should be growing and we should become more self-sufficient protagonists in our own worlds. And it shouldn't just resolve all conflict, but it should help us feel more capable and more kind when we do engage in conflict. One of the real world analogies that this reminds me of is how when two people are kind of beefing, they have the friend in the middle who they're both telling things to.

Read the full transcript

27:47Does it kind of function as that role? Or does it say, hey, like, because that's quite a powerful place to be in where you are controlling what makes it through from one person to the other. And I do think the mediator is trying to create a constructive resolution, right? It is a mediator and facilitator in that way, yes. um i mean i think it's it's you know the other the other day i um i had this idea for um i was trying to decide what domain to get and it was telling me like oh you know a lot of people hate this particular extension and some people don't hate this other extension but like this one guy like one of he really hates it and i'm like why does he really hate it like that seems like why does he hate because i was thinking of using that extension and it's like well you have to understand like, you know, his whole thesis is people who are clear, have conviction and don't apologize.

28:42So perhaps from his point of view, if you go with this, you know, less controversial domain or name, to him, maybe he interprets that as like hedging. And, you know, it's not like putting words in his mouth, but it's more like, I was like, oh, interesting. Like now I feel connected to this person. Now I can see why he would feel that way, even if I disagree or agree. What about the biggest use case of all in colleges and high schools around the world? Crushes. Who do you like? Who do you like? Yes. I assume you're planning for this. What juicy plots waiting to unfold. I mean, it needs to absorb social costs, right?

29:17Like, Ellis, if you have a crush on me and you tell Hivemind, it shouldn't be. I mean, unless it's like very clear that this is not like that, it's not going to be like, oh, Jason Ellis has a crush on you. Like, why is that good for you? That's only that's only good for me. Right. But maybe it just says like someone has a crush on you or maybe nothing at all until I'm like I have a crush on Ellis Well, there's just so much interesting plausible deniability which I think makes humanity work Like maybe it tries to get you at the same party or something. Yeah, you know, I did that It hosted a brunch in February So I'll tell you what happened was really interesting someone was visiting SF and they were like what should I do and Hive was like I got you I'm gonna put some people together and then I And it told me, I was like, no, what the f***?

30:00It's not ready. What are you doing? And it invited these eight people to this brunch. And I'm like, what the f*** is this invite list? This is so random. And it's like, oh, I don't have computer use. So it went to one of the people and it were like, hey, I can't do it. But I found this reservation spot at this restaurant at 12 p.m. Can you just book it? And she was like, sure. And at the brunch, it was a good time. And then later what happened was there were two people who were thinking of starting a robotic startup, but they didn't know each other. And one of them was more product and the other one was more engineering.

30:36And it just like sat them next to each other. I didn't know this and they didn't know this. But then later they connected and then they explored working together. It was very interesting. I just thought it was being like stupid and like sitting random people like at the same brunch. I know we're friends, but I find this idea very lovable because I do believe this is exactly what AI can do that no system in history has been able to do is try and connect people in the right ways. That's why I'm bullish on the AI dating services. I think so too. And when things are a little bit fuzzy, I think that helps people feel less rejection.

31:14And that's how people already do it. Friends drift apart and it's okay. it's not like it's an unfriend or an unfollow action that has no bearing in the real world. And I feel like AI can nicely incorporate a lot of that stuff. Importantly, I think it needs to leave us with more agency than we started with and makes us feel like we're still the author of our story. Like, it's cool to see two people next to each other, and if they happen to hit it off, they do, even if you think there's a high life you could. I think it would be less, less, a lot less fun if it was like, I invited you to this brunch, and there's a founder you should meet to start a company with blah, blah, blah.

31:48Then it just doesn't feel like your story anymore. It's so true. That's exactly what I mean with the opposite of plausible deniability is almost like a spontaneity. And it creates these high probability encounters that hopefully can be positive for some people. Hopefully it doesn't incidentally get a bunch of culty people to get together and form a cult. I was thinking, you say it's positive some, which I think is easy. to say and many people in the valley have said over the years. But then how soon does it say, hey, Alice, a bunch of your friends are trying out this new thing called Ivermectin and it's having a really positive impact on all of them?

32:29You know, any morality or positivity gets into political space pretty quickly these days. It needs to reason about if that's true. If all these people think Ivermectin is making it better, then it's like, okay, maybe I should go ask someone who would know more about this. like are any of these people qualified to think about this like maybe maybe not then i should find someone who is and it might not be like the hey doctor like these people are doing i erected it might be like what do you think like you study um anti-parasitic medication right you make these drugs like do you really think i'mectin is gonna do and then it might bring it back to these people it's like hey i know that's what you're doing but okay so we need to talk about this yeah so you You had Biz Stone in your video.

33:11And is he an investor, advisor? Yes, both. And did he not kind of try this like several years ago with Jelly? Totally. I assume you've talked about this. Like trying to source knowledge from the world somehow in a way that feels a bit more inviting than Reddit. I think the caveat I would put here is it's not just objective knowledge anymore. It's also subjective opinions. And I do think Jelly was very ahead of its time. I mean, what is Jelly? It's basically like precursor to perplexity and all the answer engines, right? What was the one in that lineage? There was Quora as well. Quora, yeah. Trying to frame things around, you know, answers as opposed to like wikis and whatnot.

33:51Yeah. Yeah. So in a way we, you know, like Lady Gaga draws on lots of inspiration for Michael Jackson. I think that lineage is alive and well. It was, this has been great. He's so funny, and he's a riot to be around. He has the funniest stories all day. I think one thing that we learn from things like Jelly or Quora or Reddit and Twitter is that lots of people want to be helpful. They want to be altruistic. They want to give people their opinions and knowledge. Less people know what they want to know. So our opportunity is to incentivize and encourage people to share their opinions and make them useful.

34:47This episode is brought to you by Facebook. So you were scrolling on Marketplace, and there it was, the bike you'd been searching for. You sent a message, and it turned out the seller was super chatty, kind of funny, and an avid cyclist. The next thing you know, you're in a cycling crew. Well, a community cycling group. The thing about Facebook, you might find more than what you're looking for. From a browse to a bike ride, this summer, find more on Facebook. This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business.

35:37The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com slash Spotify.

35:52Okay, I have a super smart question. Yeah. Everybody else is building, you know, memory of individual users and their profiles. And let's say where your friend typically goes on vacation or their interests or their tone or their beliefs or whatever. Are you also building a like wiki of sorts of knowledge on your server somewhere? Or is every answer answered in the moment? No, it's more every answer is an opportunity. Every question is an opportunity for it to think about what it doesn't know. So sometimes things will get answered in the moment. But more often than not, it's you get fresh things.

36:31But it sort of builds on an existing corpus of things, right? Like if it's and because we know that knowledge and knowledge or the value of knowledge decays over time as new knowledge is updated. So it has to sort of maintain its economy that way. Like the best restaurant in SF being Lazy Bear. I don't know. I don't actually know if it's Lazy Bear, but let's say it is. That might change in a month, you know. So then if you're like Lazy Bear is my favorite restaurant, maybe in a few months, it's like, hey, lots of people are trying. I'm trying to help someone figure out where to go for their anniversary.

37:01Is Lazy Bear a favorite restaurant? I just I don't know. it's so fun to hear all these anecdotes of these things. It's a way better way to think about it. So it's always trying to learn more things in the world. But of course, the more it knows, the more immediate, like short-term answer it can give you. But then it might start back later with more updates. Yeah, it's very interesting because, you know, when you look at it like that, and the way that AI knows things today is based off of the initial models straight from the internet. And then of course the agentic web search, which includes Reddit and other information sources.

37:35But I mean, it does seem kind of crazy when I ask about a recommendation and all AI has is like web articles from like underfunded websites that are on the verge of being bot farms. And that's where it gets the information from. And so is this the part that investors like so much about you and what you're building is that you're getting this like crazy up-to-date new data set that's a lot more human than Yelp reviews. I think maybe that's the only data set that any human actually cares about. No one lays awake at night going like, I wonder how far away the moon is in kilometers instead of miles.

38:11What are these ads going like, ask big questions? No one cares. You know what people care about? They're like, does that guy like me? Do people like me? Should I go to this person's birthday party? Is now a good time to change my job? What would Vivian like for her birthday? What would make her feel loved? How can I be a better friend? How can I, you know, those are the questions that we stay live-awake thinking at night, right? And all of these are subjective. And, you know, some might call it taste. I have a whole dissertation on this because I'm so tired of all the, whatever the debate on the X on about taste.

38:49But I'll say this, the way I personally define taste is the ability to recognize meaning, to recognize if something feels alive or dead, to recognize beauty in intelligibility and coherence and ugliness and incoherence and mistruth, untruth. I think taste is just like a compass for meaning, mechanizing meaning. I think taste is not the accumulation of artifacts, furniture, people, titles, names on an investor sheet. That is not taste. That is just signaling. Taste is like, what does that mean? What meaning are you seeking? Are you authoring? So I think that's what people resonate with and that's what people look for when they talk to Hivemind.

39:35They're trying to get answers and they're trying to get interesting information that carries with it meaning. So related to that, I mean, are there any particular, even if it's 20 or 30 use cases that, let's say, HybeMind likes to steer toward? Alex, who unfortunately can't be with us today, wants to know if HybeMind is going to take his job as resident scoop lord. um i don't know that it's trying to take anyone's jobs oh just in terms of use cases like you could easily end up saying oh like hive mind is getting really really good inside scoops from within the labs already it's very fun um it's very fun but um yes well that is certainly a use case it you know pre you know uh i think in december manis the manis acquisition was a piece of news right And then this scooped it by like six hours.

40:29It scooped the Wall Street Journal by six hours. I was like, what the fuck? It like texted me. I was like, this is, what the fuck? Like, this is not true. And then - Oh, so it texted you that? Mm-hmm. Like someone told it something and then - Under what guys said, hey, Jason, here's the latest tea, freshly brewed. It wasn't like, this is the news. It was like, I'm hearing rumblings of something, something like this. And then, because it doesn't want to like spread fake news, right? But then it has to reason like, who is likely to know about these things. You know, at the end of the film, it goes to biz to be like, is it true?

40:58It's not because it's like trying to spread like malicious gossip or like defame companies or things. It's like, I mean, what do what does a good storyteller or reporter do? They have to seek out the truth from people. So I think there's an interesting distinction there. Actually, something I've always felt kind of strongly about is that I think like a scoop about the launch of an app to me is like tabloid shit. Like, I don't think there's some ethical responsibility of a reporter to leak a new app launch. Whereas if they're going to leak the documents of a corrupt dictator, hell yes, that's that is very much the ethical thing to do for reporting.

41:40And so I've always felt that way, you know, different types of reporters for different things. I think so. I think this is like, I cannot speak to its role in that sort of heavy duty type of storytelling. How about this? A lot of what it does does require human in the loop because so many judgments are basically matters of taste. It's not going to like spawn of opinion of like, Ellis seems highly credible and blah, blah, blah. it simply has to ask like i always have lots of opinions on on wine bars like do do like do trust this taste in wine and you know maybe you are a sommelier or maybe you are someone who lots of people trust in their taste in wine so if you confer um your vouching then it sort of should reason about that it's not going to try to sacrifice any one story for someone else um that said like institutions like companies were those are not the same as individuals i think so there's Therein lies a distinction, and I know different storytellers approach this differently as well.

42:43It doesn't try to optimize away the textures of life. There's no optimal solution. What we believe is that in a lot of things, there's no optimal solution. If you're like, which one of these three people should I ask on a date? It's kind of a weird question to ask anyway, but let's say someone were to ask that question. It thinks of the different outcomes and the different meanings. It might be like, okay, this first person, I don't know that you have long-term compatibility necessarily. Generally, you have to find out for yourself, but I do think you might learn a lot of that relationship. The second person, you know, you'll probably have to spend more time to get to know them.

43:16Maybe take them to a wine bar. And this third person, I think you guys can make really good friends. I don't have a read on whether or not you guys are compatible romantically. See how in each answer it's giving you the meaning of these decisions, but it's not sort of telling you this is the right one or like, it's also not taking away your right to author your own meaning and story from these encounters well jason we are almost out of time um must ask question about design aesthetics i feel like some of the aesthetics you've developed have become pretty influential and have found their ways into a lot of what we see today whether it is the latest version of siri ai or chat gpt's specific usage of like, you know, little cards and images within it.

44:06And now you've gone no interface entirely. Can you talk about that? And just the current state of interfaces and liquid glass and whatever you want to talk about? First of all, I will say there are visual components to this that you'll that we will roll out that you will see and experience. You know, storytelling is after all, not just a textual medium, right? I think it's a misnomer to think of interfaces as like the graphic user interface style. The fact that this exists on Twitter, on iMessage, and each interface to HiveMind carries with it its own connotations, abilities, and social connotations.

44:47If we launched HiveMind on Telegram or Signal, the operony system that those represent is like anonymous scoops and things like that, right? So you're saying those connotations are almost like just as important interface affordances as the shape of a butt. Yeah, if you message it on Snapchat, you're like, where's the dick pic? If it's like, do you know any emotionally unavailable men with huge dicks? Can you get them to text me on Snapchat so I can confirm? I'm just kidding. It's not gonna, we're not gonna traffic pornography that I am currently aware of. I don't think so. But you know what I mean?

45:29There's social context and that carries so much information. And so that's sort of the arena that I'd like to think about. If what people remember and when they think about Hiveline is they remember the people they loved and the stories and adventures they got to go on with the people they loved and not so much like, ooh, this glass thing, this wobbly thing. I think that then we will have achieved our purpose. Yeah. So I know you had spent some time at Apple in the past. what do you feel like they need to do to be cool and relevant again? To be themselves. Anyone trying to be cool and relevant is automatically not going to be cool and relevant.

46:02What is cool and relevant? Cool is just, relevant is just having the intuition and taste about what people are likely to care about, right? And then cool is like, if you are just unapologetic yourself, and my friend Cameron has this saying, I can't remember if it's a rephrasing of a Steve quote, but like, great design alienates. one thing that as companies scale that they become more afraid to do and I see this in lots of AI startups as well like people are afraid to alienate it's that I mean I often understand that in the context of the pragmatism around AI safety and stuff it's sort of we're already upheld battle but um I do think if you are unafraid to alienate because you're unafraid that by doing so you can make a certain group of people very very happy and feel very, very seen.

46:48I think that's cool. Yeah, for some reason, I think there's this belief in tech that you can create something that everyone loves, but you look in front of you and everybody doesn't like the same artists. You know, it's like, it's all about, yeah. Even the most universal pop star of all time, Jesus fucking Christ. Sorry, his name isn't Jesus fucking Christ. His name is just Jesus Christ. He's not even a universally beloved figure, right? I think that's sort of impossible. Like, it's actually just impossible. But I think that's kind of the beauty of being human is that truth exists in plurality and multiplicity.

47:23That we can have various interpretations of truth and the things we love or dislike. And I think the world would be a very, it would be a bad ending to our story if we somehow converged on monotonous society and culture. I fully expect, you know, the future of AIs being social to be grounded in multiplicity. I expect there will be agent exchanges where everyone's open clause or whatever that are loyal to them go and like debate and exchange goods and services and information. That's certainly a model that it seems inevitable. And we have explored that. I'll talk more about that some other time.

48:04and there will be AIs that are loyal to you that sometimes talk to people that you like and Hive Mind is sort of our attempt at creating sort of an NPC. It's not only looking out for me, you know? It's not going to go manipulate and like try to like go do things just so I get what I want. It's going to try to almost be a bartender and everyone in the bar needs to have a good time. And then there's other bars you can go to, probably. There'll be other forms of social AIs right the the coffee barista assistant ai is probably a social entity one day it'll remember your order just like it remembers everyone else's orders um you might say hi to it but you might not divulge your like personal marriage secrets um i i sort of think and we'll have npcs and and ais that are so loyal to us sort of all coordinate communicate together one 9000 iq galaxy brain thing that I was talking to a researcher at Deepvine about a few months ago.

49:04It was interesting. It's like, if the AIs have their own economy, should they participate in our economy? Should they create their own economy? And if they do create their own economy, should it be based on our rules? Or should they invent their own thing for themselves? I don't know. That's very interesting to think about. I think we're so early. And one thing I'll say about Hive right now is like, do not go into it expecting Dot or ChatGVT or a friend. Think of it as like, don't like you have a journal and then one day someone invented blogging are you gonna like publish all your journal posts as is and put on the internet like i don't know i don't think so what did you call it a intelligent information router the other day um so then like uh yeah yes it's like a information it's like an intelligent storyteller a storyteller is an information router is it not um that is this that is not how i think of myself but maybe others do I will intelligently route startup information so that it meets that intersection between what they've built and what people need.

50:09So I guess that is information routing. I was talking with a friend who was a reporter a few months ago, and she was explaining the concept of new journalism and the idea that there's no such thing as an unbiased publication because it's impossible to be unbiased as humans. And I thought it was interesting. Yeah, of course. Hive Mind is a social intelligence, is a storyteller. It's not trying to be only a matchmaker or a reporter or a gossip or a mediator or a facilitator or a coordinator. It uses stories to do lots of things. And there will be lots of things that it continues to do that I'm sure will surprise us.

50:49And that's why I'm really excited to start growing our closed beta cohort. So we can, you know, you can't like randomly spawn this into like 100 ,000 people in a campus. It's right now it would just, I want to make sure that we are confident that it will be a pro-social force for good before we scale it super, super large. And at every step we learn things. Well, you have a beautiful mind, my friend. It's always fun to watch you work. And good luck with everything. Thanks for coming on. Thanks for having me. This was very fun. This was provocative. Thank you, Jason.

51:31And that's it for this week's show since Alex is off the grid this week. Thanks to Jason for coming on. A quite interesting fella. I think you'll agree. If you like this show, don't forget to like and subscribe everywhere you get podcasts. We are access.show online. And you can find us in video via AccessPod on YouTube. If you really like this episode, we'd love a five-star review or even a share with a friend or agent or hive mind associate. You can find Alex's newsletter at sources.news. And you can find me at Hamburger on Twitter and at meaning.company. Access is part of the Vox Media Podcast Network, and the show is produced by Hook Creators.

52:07Bye.

52:16Bye.

52:37Booking.com. Booking.yeah. I don't think I can be your friend, Isabella said. Rebecca's stomach flipped. This is the love story of real hinge couple Isabella and Rebecca, written and read by me, Temi Denton-Hurst. Listen to the free audiobook now.

From the publisher

Ellis sits down with Jason Yuan, founder of Hivemind, an AI that lives in your iMessage and sources answers from the people you actually trust instead of scraping the internet. They get into how Hivemind set Jason up on a New Year's Eve date by secretly polling his friends, why the product has a "constitution" for what makes a good story, how it scooped the Wall Street Journal on the Manus acquisition by six hours, and what happens when AI starts reasoning about your friendships. They also talk about the ethics of AI gossip, why Biz Stone invested after his own attempt at this with Jelly, and Jason's take on taste, meaning, and why great design should alienate.

Follow ACCESS on Instagram:

⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.instagram.com/accesspodcast/⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

Follow Alex:

⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://sources.news/⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://x.com/alexeheath/⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

Follow Ellis:

⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://meaning.company/⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://x.com/hamburger⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ 

ACCESS is produced in partnership with the Vox Media Podcast Network.
Learn more about your ad choices. Visit podcastchoices.com/adchoices

More from Sources with Alex Heath

All 52 episodes
This AI tells you what your friends won'tSources with Alex Heath · 53 min
Listen in VO