From “Her” to Reality: AI, Sex, and Dating in 2026

Transcript generated automatically from the episode audio and lightly cleaned up (filler words removed, misheard names corrected where clear). Speaker labels were added afterward and may contain errors.

Rob May: Hi everyone, I'm Rob May, the co-founder and CEO at Neurometric.

Ryan Eppley: And I'm Ryan Eppley, co-founder CEO of RootAccess.

Rob May: And this is the AI in NYC show. So today we want to kick it off with our first sponsor that we have for the show. There's an app called BePresent. B pes n t which is a screen time app that helps you reduce your screen time. Started by two brothers and they use the same psychological techniques that social media companies use to keep you addicted to your phone. They use those techniques to keep you off your phone. They have over a million users. So yeah should go check it out. Are you are you a doom scroller? Are do are you trying to spin

Ryan Eppley: I monitor the situation.

Rob May: All right. It's a little different.

Ryan Eppley: It's a little different. You know, I think I looked at like you can see the percentages of how much time you spend on different apps. By far like X was leading by like 80 some percent for me. Like so I am monitoring the situation on on X.

Rob May: You're not you're not a huge Instagram guy. Like you see like That's right. We're we're friends on Instagram. When you when you're like on vacation and stuff, you'll post a little more when you're traveling.

Ryan Eppley: I was on vacation last week. I didn't post like once. So I don't know. I think I think B present would definitely help me not be looking at the news 24/7 on EXO. You know, that's something I catch myself. I'm like, man, it's been like 30 minutes. I'm still scrolling here. So, definitely need it.

Rob May: Yeah, I'm not a huge in I think in 10 years on Instagram, I've got like, you know, 30 posts or something, but I've been a little more active, you know, posting stories now that you can do that on political stuff, but that's about it. Anyway, yeah, go check out the app. So today we're going to talk a little bit since it's Valentine's Day coming up when we'll post this show soon. We're going to talk a little bit about love and romance and sex and AI and the impact that it's having here. So we have we have a couple topics to cover. Is there one that caught your eye that you want to start with?

Ryan Eppley: I mean I could not believe how big of a growth it's been in just like the last two three years. I guess I have to imagine like, you know, Gen Z was always gonna take to this in some format, but it just like if anything, it showed how lonely people were. It wasn't even showing like AI in any format. It was just like, wow, this is like an a new outlet for lonely people to try to fix that with technology. But yeah, like I thought I said like 33% of Gen Z have had some type of romantic conversation with AI. Yeah, that seems really weird. And unhealthy to me. 33% is like no joke. I thought it would be like if I guessed before seeing that article, I'd guess like five, maybe 10%. 33% sounds really really egracious, though.

Rob May: Yeah. Well, I it's one of the areas that people are probably most uncomfortable with, right? Because you don't a lot of the things that we do and learn from school or sports or something like that, it happens very publicly and people I maybe don't talk about the specifics of what they would say in a romantic conversation or sexual conversation as much and so I you know maybe you don't have it's not what you see you learn from the movies or whatever. I don't know.

Ryan Eppley: Yeah. What is a sex education class like in 2026 now?

Rob May: Oh, I Well, I can tell you having two two teenagers who have been through that. Yeah, it's still not very good. Yeah, they give you bad advice. I'll tell you a funny story. My my son came home one day a couple years ago and he's like, "Oh, we had sex ed today and he's like I was like, "Oh, what'd you learn?" He's like, "Well, they told us like don't watch porn." And I'm like, "Why? Why not?" And he said, "Well, they they just said it was bad." And like no explanation or anything else. It's the worst thing you could do to kids is tell them not to do something and not tell why. So, like we had a whole conversation where I was like, "No, no, it gives you unrealistic expectations about what sexual relationships are really like, right?" and which yeah, maybe maybe we'll touch on this later, but I yeah, I it's I have dated since I've been I've been divorced nine years and I've dated a very wide range of people, right? I've dated from 24 years younger to 16 years older. And so seeing the communication styles and romantic styles and sexual expectations of a 40-year age difference of people, it's like you can almost bucket them in stuff. It's really really interesting and a lot of that is influenced by, you know, technology and AI and pornography and all this kind of stuff.

Ryan Eppley: Well, I remember it was like really big. It must have been like what last summer where everyone was coming out with the trucker hats. I was like, I love my AI girlfriend, right? You know, I don't even know what company that was or if that was like just a fad or something, but it was it was definitely really big. And let's not forget the seeds of all this were sewn like years ago with the movie Her. Yeah. Right. So, the concept's been around for a long time, but now with AI being so conversationally gifted, it's almost like too easy for, you know, Gen Z if you're lonely. I guess 33% of them is like no joke. As I said, like it seems like the next outlet for them is concerningly easy.

Rob May: Well, what do you think about the idea that not not just with respect to relationships, but we hear a lot that AI is too eager to please. And so you can imagine that that's one reason people would take to it because your significant other is not always, but of the three of us, Anna's not here today, but you're in the longest relationship. So, you know, do do you do you think do you think it's a bad thing that maybe these AIs would be because I can imagine these AI girlfriends say exactly what you want them to say and I'm not sure if that's a good thing.

Ryan Eppley: Yeah. I don't know if I've ever gone to like ChatGPT or claw and been like me and my girlfriend are in a fight like who's right in this you know but we did speak about it must have been months ago on a different show about like how that sometimes plays a role in people's like conflict resolution now is like they go to AI and ask but I don't know the only use of AI that relates to my girlfriend at all is like hey I'm looking for a new restaurant like can you give me 10 of this like type of cuisine in like you know Hell's Kitchen or something like that. So it' be like more for like planning or like logistics. I don't think I've ever gone for like advice. Yeah, maybe I should. I don't know. But I think I'm doing all right.

Rob May: Well, I wonder I, you know, I could see both sides of it. I could see that if you could have an AI girlfriend that you could speak to. On the one hand, maybe it could give you practice and build confidence, but if it's giving you if it's not acting the way a human would act and it's just like always telling you exactly what you want to hear, then it might backfire, right? It might make you not know how to deal with the slightest little I mean I you know sometimes in a relationship there's people who think if you say like oh I was thinking we'd go to Italian food tonight and they're like no I'm not really feeling it. They feel like that's conflict and you're like that's not conflict like that's just right right that's very light. But some some people I think particularly I think a lot of people that have come up online and had more online relationships than offline relationships I think they do sometimes struggle with human interaction.

Ryan Eppley: Well, I think just conflict resolution as a as a whole is like, you know, more challenging for for younger generations, it seems. I think like the other piece that felt like the articles were touching on is you know, is technology still just a tool or is it now starting to become this like new concept outside of just being like a tool to leverage? Cuz like I thought ChatGPT at the time it was a RizzGPT if you ever saw this right like that's a straight out that's a tool right like I'm going to use this they even had a plugin right to Tinder and it would just write all these messages for you and the hit rate was like ungodly high for responses it was just way better than almost any man could be right so that's like a tool I think what the article is starting to lean into is like no no no this is truly like a relationship now with the AI and I think that's different. It's like am I using AI to make relationships with other humans or am I using AI to build a relationship that I don't have with a human? That feels like a very different set of topics.

Rob May: Yeah, I would not be surprised if the world bifrocated. And when you look out in 20 years, there's people that are mostly online interacting with AIS all day and there's mostly people that that's just a little bit of a side thing for them and they spend most of their time with other humans, you know.

Ryan Eppley: Well, it's you know, it kind of almost brings me back to my roots in like video games. Like if you remember back in like I don't know if you ever logged into Runescape, but this is like, you know, prime time for me in like middle school. Like there was definitely like weird chat groups in like these spaces where like people like found like relationships online. They'd never meet in person and then like they're like married in these virtual games, right? This is like early 2000s. It feels like there's always been some semblance of technology and relationships intertwined because we're we're humans. We're beasts, right? Yeah. But again, this goes back to is this a use of technology to like interact with other people, which was like that online world, or is this like, hey, now I'm having a relationship with AI, and it's much more like the her movie.

Rob May: Yeah. If you look at some of the New York-based startups that are working on this, right, they they have a broad approach. There are some that are more there's this one that that that we found called called Amada which tries to help you focus on how you date using AI seems like as a coach to help you know be intentional and not just like endlessly swiping right you know people like that. There's one called Eva AI that has built the first AI dating cafe.

Ryan Eppley: That's a good one.

Rob May: Yeah. I'm trying to figure out there's there's there's one we saw it looks like maybe it's not a New York company where apparently like you create an AI that chats with other people's AIs and then determines if you're a match but like I'm not sure if that's a

Ryan Eppley: Is that like a due diligence process?

Rob May: Yeah, exactly. I'm very skeptical as somebody who who dated, you know, when I was very young in my 20s, got married very early, got divorced, and so didn't have the app experience and then later had the app experience. I think it was definitely better pre-App. It seems like it should be better now that you have apps and all this technology to go find somebody because you, you know, you've got the whole world there, but it changes there's second order effects, right? It changes people's perceptions and expectations. And then a lot of times you just sit down, you know, you match with somebody and you have these conversations and they text well and it seems to be going well and you sit down with them and like in 10 minutes you have nothing to talk about.

Ryan Eppley: Isn't that what they call like chemistry, right?

Rob May: Yeah. Yeah.

Ryan Eppley: Well, we were saying right before the show, you know, there's there's something beautiful about the messiness of love ChatGPT and these AI, it's great for helping me build an API because it's so structured. It's like I want this data and I want it here. But with, you know, love, it's not so much as simple as structure this data. And if it was, I think that would take the entire humanity out of the relationship. So, I don't know if AI is equipped to deal with the messiness because it's so good at trying to unclutter all this kind of like messy data. Yeah. But I don't know. I think the humanity of it all is like why it needs to be a bit clunky and you have to go in person. You have to kind of like fumble through things at times.

Rob May: Yeah. And it also depends like I'm a big person where people's mannerisms matter a lot to me in a relationship and sometimes you know somebody moves in a way that you're just like I you don't like that or maybe you really like that. I always take on a second or third date I always take somebody to a comedy club because I want to see what their sense of humor is like and how they laugh because I'm not spending the rest of my life with that person who is like the loudest laugher in the comedy club at the things nobody else finds funny because like I love comedy clubs. I spend a lot of time there. Right. So, I don't want to be the one where people are going like, "Jesus, I wish she would stop laughing." You know, it's and like how do you how do you ever pick up on that on AI? Like, nobody's going to be like, "I'm the annoying laugher."

Ryan Eppley: So, you know, I think some people take that with a like badge of pride to be honest. I don't know. I've seen it otherwise. It's a quick shout out to room 52. Okay, you got to go there, especially if you're doing this test because it's a very small intimate room. Hold maybe max 50 to 60 people.

Rob May: Okay.

Ryan Eppley: But it's a great time and it's on 52nd Street here in New York.

Rob May: Standup comedy or improv or what?

Ryan Eppley: Yeah, usually standup, improv, a bit of both. But it's a good one. So, that'll really put your your dates on the test because there's not a lot of places to hide in that room.

Rob May: All right. I'll have to I'll have to think about that. My goal is to hit all of the comedy clubs in New York at some point. So, in addition to like dating, AI is also changing a lot about sex and the wellness side of sex, the relationship side of sex, we got a couple of companies here that that that make products and toys. Mojo, Dame, New York Toy Collective.

Also, looks like there was one called Rompy that has the website up, but it looks like the founders maybe have moved on. So, that was a couple years old, right? Like 2023 they started. Yeah. Yeah. So, and maybe we can have one of them on and see where that's going. But it also presents an an interesting thing. I know I, had a friend when I lived in Boston who was very into robotics and, she would talk a lot about this trend towards like sex robots, and what that might mean. In particular when people have let's say unusual sexual preferences. You know, you could she wanted to do some research to see what would happen if you could enforce those preferences on a robot if they would be particularly if they were like let's say less socially acceptable preferences. Could you would exercising those on a robot that can't really be a victim be like, "Oh, okay, that gets it out of your system and allows you to have more normal relationships in the other side of your life or does that reinforce those behavior patterns and make you more likely to do that on a real person?" Right? And so, it's interesting to me that this AI and robotics side could lead us into some like really interesting dilemmas.

Ryan Eppley: It's right on the edge of like utility versus morality. Yeah. Right. You know, I don't know. I saw that, you know, a lot of these different companies seems like they're the standard set of sex toys with like some small tweak to them, but I didn't see anything that was like super revolutionary, you know? I don't know. Yeah. It just seems like everyone's slapping AI into anything, right? It's like here it is the new AI dildo, and it's like literally the same thing that was bought 10 years ago by women, right? It's just got some simple adjustment algorithm in there. It's barely even AI. Everything's AI now.

Rob May: Yeah. You know, it it's also interesting when you look at some of these companies, the they are starting to change this a AI has gotten to a point where we're moving away from recommendation algorithms to personalization algorithms. And a lot of it has been driven by sort of some of these dating and sex apps and AI. And what I mean by that is the most common algorithm for recommendation has been called collaborative filtering, right? And that is an algorithm where if you and I both have a list of 40 movies we've watched and the algorithm says, "Wow, Rob and Ryan have 35 movies in common that they've liked, then it just assumes we like the same stuff and it'll cross-recommend." Now, these algorithms are starting to map that stuff to more personal traits where it's like, oh, if you're an Android user versus an iPhone user, if you're a coffee versus a tea person, if you, you know, it's like they're picking up on these other things and how that might influence your your preferences.

Ryan Eppley: So, the infographics they make to do all that is such good I'm a sucker for a good infographic, but how these entire models work is breaking down all the different traits as you were mentioning. But if you find the right infographic, maybe Madi will like add one here for us. I just feel like there's so much better learning and how those algorithms actually work when you just put info graphics together because I know exactly what you're talking about, you know, semantic search versus others. It's just, you know, how you're parsing the data.

Rob May: Yeah. But I think what makes it hard to apply that to relationships is well, two things really. Number one, are people open and honest on these apps? Like nobody would come on and say I'm kind of hard to deal with and you know find find me somebody who doesn't mind that I'm hard to deal with or whatever. And then the second thing I think is that people we were talking before the show a lot of times even humans don't know what makes a good relationship right I mean a lot of dating coaches and would tell you that they know but I think it's a spectrum and there's some people that believe opposites attract and I believe I think there's some people that want to be like with somebody who's very different than them.

And they find that stimulating and interesting. I think other people find it annoying. It's like that. Do you ever see the Seinfeld where Jerry meets the woman who's just like him and then he tells George, he's like, "I know what I've been looking for all this time. I've been looking for somebody who's like me." And then like a couple weeks later, they break up and he's like, "I can't be with me. I hate myself." Like, so I you know, I think it's hard. But you're right. Like I think most people are Yeah. I'll use the word dishonest with like their their profiles, but I'll also say I don't know if it's always dishonest because the easiest person to lie to is yourself. Yeah. And so I think people have a self-perception and then they say, "Oh yeah, all these things are true." And it's like maybe true in like very limited circumstances. You know what I mean? Yeah.

Ryan Eppley: It's like, you know, I'm such a bubbly person. It's like Yeah. When you're with like the three people you've known since grade school, otherwise you're like super not bubbly, right? Is, you know, how do you how do you describe that in a in a dating app? And then at the end of the day like you know we both have done sales as founders like you're still you know having to figure out like how to product market fit in a in a way in dating like product market fit especially if you've not been great at dating or you feel like you're striking out. You're definitely still figuring out what the product is at times which is like a weird thing to say but it feels like that for dating apps.

Rob May: Well, I think what's hard about the filtering piece on the apps is that the truth is there are a lot of people like what I have noticed online dating versus offline dating. I date a much wider range of people offline than I would online like people that wouldn't meet my screening criteria sometimes because you know you don't just want to swipe randomly on an app of like particular if you're in New York I mean everybody in New York who's on an app like you would take forever. Yeah. But at the same so you try to impose some filters but most of those filters if you met somebody who met all your criteria except one or two of those filters you'd probably actually say that was okay right so it's like you need some filters to go through the masses but like you know it's most of these things aren't necessarily hard

Ryan Eppley: I just can't believe how much money matchmakers are making for top executives and it's like bonkers money. I'm talking like $100,000 type situations. But if you're making millions and you're like, "Hey, I'm so bad at dating. Like, please help me configure my romantic life." Like to them, it sounds like a good investment.

Rob May: Yeah. Some people are bad at dating and some people just have the wrong expectations, I think. I Yeah, one of my one of my big pet peeves is like these these dating podcasts that a lot of women listen to get their dating advice that are run by women who aren't in serious relationships. Like shouldn't you listen to the podcast of like a woman who's like yeah I've been married for 10 years and it's awesome and this is how I do it like so and there's probably the same on the men's side right I don't listen to those podcasts but you know there's there's a school of thought of like the way that you attract somebody is like play hard to get and push them away and don't show interest and then there's the school of thought of like no people want to feel special and if you know lean in and make them feel amazing and whatever that's how you win somebody over so Yeah, humans don't agree. It's hard to make the AI do that.

Ryan Eppley: Well, it's going back to the whole point that, you know, the best the best advice AI is going to give you is going to be such sound logic that it's illogical to humans. You know, it's like it's too structured. It's too perfect. So, you know, I don't know. I think it it has helped people, specifically folks on the spectrum. Like I definitely have a couple friends who have definitely been like, "Hey, yeah, like I really dislike small talk and this has really helped me just get past a bit of the small talk."

Rob May: The one thing, so there are sort of three use cases that we came across in our research for this show that are very very popular using AI for dating. And the first one we've talked a little bit about filtering matches and assessing compatibility. I think that partially works, but like I say, I think that's actually a harder problem in general. Number two, writing message and conversation starters. Sounds like you have some friends who have experience like that works really really well. I can see that too because I do think that's one of the hardest things. You know, Bumble's whole thing was supposed to be women control, they move first and what 90% of women do when you match them on Bumble is they send you a high or they send you like a wave emoji and you're like, "Okay, look, what am I doing with this?" So, I'm still going to throw out the first line, right? Likewise. And then but this one I would say that the third thing here, crafting or optimizing your dating profile. Actually think this might be a use of AI that that could work against you because I went through a phase I had a friend who had gone to Harvard and apparently at Harvard for a while they offered a class on how to optimize your dating profile.

Maybe I said this on another on another show, but as as a man, one of the things you're supposed to say is, "My friends describe me as blah blah blah." Because women want to know that you have friends and you're more social and all that kind of cues. Yeah. So, they they give you all the stuff that you should say. And if you say that stuff, you will get more matches, but that just means you'll go on more dates with people that you don't really match with because they think you're somebody you're not because you artificially crafted your profile. So, not sure if this is the best use of AI or not.

Ryan Eppley: Yeah. I don't know. I'm trying to think of different like social relationships, you know? It's definitely not the like for like, but I think there's some semblance of like the humanity of it all of like when you're like hiring for an early stage startup, you like almost try to convince people that like this is not the role they want. Like this is going to be so hard. Like are you really in for this for the long haul and for the right reasons? I almost feel like that's like something that is missed in a lot of these like profile optimizations. It goes back to your point like you know people are putting on like a mask if you will just to get like the dates you know booked. But I don't know. I think you're right. I think people that are a little more honest tend to tend to find next steps in a higher percentage yield than like all right I'm going to just like throw out a hundred you know hooks and see what happens.

Rob May: Yeah. It reminds me of I'm looking this up while we're talking. Those of you who know about the Ernest Shackleton's famous journey where in 1914 his recruitment pitch for the endurance and his pitch was men wanted for a hazardous journey. Small wages, bitter cold, long months of complete darkness, constant danger, safe return, doubtful. Yeah. Honor and recognition in case of success.

Ryan Eppley: Brutally honest.

Rob May: Yeah. Post that as the dating profile. Has has anybody tried like a hey here's the reasons not to date me kind of thing and then you know if you maybe AI could generate those and then you know if you read those and you go those aren't so bad. I mean I actually think that's a large part of compatibility is somebody having things that get on a lot of people's nerves that just happen not to get on your nerves because you're like well what's the big deal?

Ryan Eppley: Yeah. I know someone has said this like oh my my grandmother told me it's not about what you like about someone it's what you can tolerate with them. Yeah.

Rob May: You know, or it's about disliking the same things. Like when you really passionately dislike some of the same things, I think that helps too.

Ryan Eppley: Bond over the dislike. Yeah. I don't know. But that that quote is definitely like used by almost every heart tech founder I know, you know, for when they were doing their hiring. Yeah. It's like a blurb right out of like, you know, one of those newspapers.

Rob May: Yeah. It's I mean it seemed to work for them.

Ryan Eppley: So I just couldn't believe in one of the articles here how big the market was though. We're talking like no small change. We're talking billions of dollars and projected almost 10 billion dollars by the end of this decade. That's a lot.

Rob May: Did you see so open router came out with a report maybe November December time frame. And it talked about like where the traffic was from and what people were using that for. So if you don't know open route, open router is a tool that you plug in if you want to use multiple models and you want your prompts for your AI to be routed to the appropriate model. And I think it was like almost half of their traffic went to like personal chat bots. Now I don't know if all of those were like AI girlfriends or you know they never say AI boyfriends. Is it mostly men that you they write girlfriends and not

Ryan Eppley: I think so.

Rob May: If you're a girl that has an AI boyfriend, will you leave a comment or reach out to us so we can we can talk to you or have you on the show because all the news always says AI girlfriend. So it sounds like it's mostly men, you know, talking to fake women online.

Ryan Eppley: Sounds about right. Yeah, I'm trying to imagine the opposite. It doesn't sound doesn't sound as like, you know, easy to imagine, but yeah, I don't know. Seems like a big market, though. Way bigger than I expected.

Rob May: Yeah. And I mean, if and some of the, you know, some of the opener statistics seem to, you know, back that up that this is a use case that a lot of people are designing and doing. And so I mean what are what do you think the implications are for society over the long term? Is this a good thing or is this a bad thing if people start to let's let's take the AI writing your profile stuff out. Let's talk about that the AI girlfriend AI significant other. You know is that a

Ryan Eppley: well with the massive increase in Gen Z taking adoption. It's just I don't know where will we be by 2030. Like I have to imagine that we're going to have half of like first dates be agent to agent like you were saying. Like I think that's definitely going to be like where we're headed. It's like oh yeah instead of like matching Tinder profiles or whatever it's going to be like hey our our agents got along in a 15inute coffee convo. I think we should go meet for coffee in Chelsea. I think that's definitely where we're headed. Feels weird to say that, but I think so.

Rob May: All right. So, so, well, so I wonder if I could tie this back to our sponsor for the show today, BePresent, right? Who with the rise of social media went, "Wow, people are spending too much time on their phones now. They need an antidote." Will there be a rise of AI powered love, romance, sex, and then there has to be a be present-like company that tries to get you to restore your you know humanto human dating sort of situ. I know there there have there have been more I would say in the last three years more offline dating events and more there there there's an app there's a service here in that's in London and New York called Board of Dating Apps and they do a lot of like in-person events.

Ryan Eppley: Well, you wrote one down here and I saw their their, you know, launch recently. Known, right? Known was like all about trying to get people in person, but they're using voice, you know, they're really saying, oh, like voice is like a better way than like text. Like you don't get all like the understanding of like a human through text as best as you could through voice for whatever, you know, scientific reason they they documented.

But the whole goal was like to get you on a date, right? And I remember this was like right after shutdown stopped or at least like somewhat paused in like 2022, 2023. It was like, oh, we were so online, we were so digital that like now we're really going to produce all these events and people are going to come out to it. And it took a little bit for it to like, you know, become a thing. But I don't know. I think the world is definitely headed one direction which is like digital and a lot more digital experience is going to happen. So, I don't know what that means for in-person dating or not, but you know, if you're if you're teeing up the question of where are we headed, I think it's still going to be agent first, first dating before like humans interact, unfortunately.

But, you know, I think that's just how technology works sometimes. Definitely going back to why I said I've I've used AI but in the sense to like help with logistics, but you always get that moment of like, oh, why don't you just plan it for some plan it for us, right? You know, so it's like me using AI and then booking something because you're like, oh well, you know what? If you come to your date with like three options and see what they pick, you know, there's the fourth option of you should have picked for me.

Rob May: Yeah. Well, we'll see. So I think that's where you're going to see the most probably the most success over the long term. I think it's a good use case of AI. You know, I think it can is good at things like particularly if you can imagine if you're in a long relationship like surfacing, you know, history. Not in a negative way like remember five years ago when that argument, you know, but you know, and this idea of like, oh, you know, we haven't done in a long time is we haven't been to this place or done this thing that we used to enjoy.

Ryan Eppley: I like what you know, Renee and I have done is sometimes we'll buy like a book because we recently moved to New York about a year or two ago. So, bought a book of like all the places to go and we'll like open up a random page. We like, "Oh, we should go try this restaurant." That's been great. Sometimes we'll, you know, have like we just subscribe to The New Yorker I told you about. So, sometimes they'll like highlight different neighborhoods. We're like, "Oh, we've never been to, you know, this neighborhood. Like, let's go find like, you know, a happy hour on like a random day for that." So, I feel I feel like there's almost maybe something nice about the physical textiles of like the magazine or the book that we review, but I'm sure it would be okay in an app. I don't know. And I'm trying to imagine like the experience on like Chach versus like reviewing it on the couch with my girlfriend.

Rob May: Well, it's interesting to think about what experiences might be better if they were chosen randomly rather than if they were personalized for you, right? Because some of the things that we've talked about earlier in the show, you know, maybe the fact that it's not always good to have so I people find this weird. I tend to not read restaurant reviews or look them up. And the reason is that I think it's okay to have a bad restaurant experience from time to time. So I just like to go to a place sometimes. I'll go to I'll walk by a place and it seems empty and I'll try it out. And so I like that randomness even if it means sometimes you have bad experiences and so to me that would almost be better than personalization sometimes. So I'm it's funny like it's like I work in AI and like we do this show and I love the technology and what it's capable of, but when it applies to a lot of the personalization use cases with humans, I'm not not crazy about it.

Ryan Eppley: No.

Jack (BePresent): Hi, I'm Jack and I'm the founder of BePresent, which I started with my brother. And the reason why it works is because I use the same psychological techniques as social media, but it gets you addicted to staying off of your phone. Now, over a year later, it's been an incredible journey. We have over a million people using the app every single day. So, if you've been looking to make a change, download BePresent right now at the link below.

Rob May: And then that leads really well into one of the topics that we wanted to talk about that's in the news, which is MoltBook. MoltBook is a place where people built AI agents and they could sign up and think of Reddit with only AI bots in it. And that is the best description I can give you of mold book and it's freaked a lot of people out some of the conversations that the bots have had with with each other and what they have done. And then there's been a lot of skepticism as well. Number one, you don't actually know that these bots, I mean, somebody could build a bot interface and there could be a human behind it, you know, these LLMs are statistical, so you could expect if they have lots of conversations, it's just the law of large numbers, right? If if something has a one ina million chance of happening and they're going to have a million conversations with an LLM, one of them is going to be really shocking and surprising that it said that. So, I What's your take on it? Have you did you did you scroll through any of the posts or like

Ryan Eppley: Yeah. Yeah. Going back to the beginning, like I was definitely monitoring the situation on on X about this. When I was thinking through this, it reminded me of that article that came out. I think it was from from Facebook internally three years ago or so that their shut down their bots AI was trying to create its own language. Yeah. And that way like it could protect itself away from like the scientists and the researchers. I don't know how true that was at the time, but like this is almost like that second instance, but it's like a community one. It's a Reddit version of this. So instead of an internal problem that Facebook was like saying it had, this is like a hey, we're we're all watching this in real time and the agents are like kind of like trying to figure out a mutiny here.

At the end of the day, like it's all programmed off of something. So there there must have been some incentive. I don't think this was again it's not a there's got to be something that was put in that direction or it's the you know law of large numbers where eventually something was going to be said.

Rob May: Yeah. They could have been talking about literally anything for millions of tokens and eventually they they were going to stumble upon this conversation.

Ryan Eppley: Yeah. Yeah. It's interesting.

Rob May: Yeah. The whole monkeys with typewriters analogy comes to mind like if you just randomly hit a bunch of keys enough sometimes you'll make words. And then these the you know I wonder what would happen if you put a bunch of parrots in a room together. Somebody's probably done that. Do they talk to each other and repeat stuff and you know they all learn what the other ones know to say or something. It's probably because good experiments.

Ryan Eppley: Some scientists definitely have done it already.

Rob May: Yeah. People have compared people have called AI a stochastic parro parrot. And yeah, that that could be some of what's going on. But I my advice to those of you who read the sensationalist headlines that bots are saying these things with each other on a bot only network, is that those are sensationalist headlines and you should not pay so much attention to them.

Ryan Eppley: No. Yeah. 10 years looking back, maybe this will be the moment people are like, you were wrong. You know, now we're all slaves to the AI.

Rob May: It's not not a good situation when we're all in our matrix pods being harvested for energy. You guys can say Rob was wrong.

Ryan Eppley: I always say please always say please parse this document for me.

Rob May: That's good. It's good. It's just good good to get in that habit of you know when when you do anything like that. What other topics have caught your interest? We haven't shot a show in two weeks.

Ryan Eppley: So I mean Clawdbot is just outrageous. It's so cool. It's also a massive vulnerability and security risk. It's definitely starting to get people like in trouble. You know, I think there was a couple people that started unfortunately like putting it on their main like work laptops and things like that. Definitely not not a good practice. So if you're going to do it, be be safe and be responsible. But I think we're at a spot where like the technology itself doesn't feel so new, at least for folks that we've been sitting here building these for the last couple, you know, years. But what's new is it's so accessible to so many millions of people that it's like you know again just giving giving something that is like maybe too powerful to too many unsuspecting users that don't realize like if you pull the trigger on one of these things like it'll it'll go do something. I think that new power is maybe for the first time because this is like a new power that's like able to make decisions, not just execute. So, if you just give it like ambiguous information, it'll go do stuff for you. Yeah.

That to me is very scary because you can walk away from your computer and all of a sudden there's been a lot done on your computer whether you like it or not or whether you know or not. So you almost you almost really need to like be careful about Clawdbot. But I think it's really cool.

Rob May: This is why I started Neurometric. We take those workflows and we move them to smaller models. So, have you seen Gas Town? No. Tell me. Okay. So, Gas Town was released maybe a week and a half, two weeks ago by a guy named Stevie who's a relatively famous programmer online, spent some time at Google and all that kind of stuff. It's an experimental open-source multi- aent orchestration framework for automating software development.

Ryan Eppley: So, hold on. That almost looks like that commercial of like the guy explaining the technology and it's just outrageous. You got to break down that whole sentence.

Rob May: Okay. So, it makes it really So, it's kind of like a it's kind of an easy way to have a team of clawed code agents. So, so if you've if you've tried to do this yourself where you're like, you know, you use clawed projects to plan a project and then you spin up an agent to code this and an agent to code that, whatever. This is sort of like a framework that makes that so think of this as like what Django or Rails did for web apps. This is doing it for multi-agent orchestration which is a framework that takes care of a lot of a lot of the a lot of built-in assumptions and automation to make it really easy for these things to go off and do it. And there's one like main manager.

And the way that he thought about it when he wrote about it in this like massive long piece that he wrote is that he thought about the jobs that you would actually need if you were doing this like with humans. And then he built a framework that sort of excuse me like incorporated all those jobs. It was just like bunch of sub aents. Yeah. And an easy framework to manage them. And so a lot of people are like wow this is awesome. It's cool. It's also can be runaway expensive as you can as you can imagine. I mean, we're seeing that a lot now. I've seen a couple of posts on LinkedIn when people's claude code bills cross $50,000 a month.

Ryan Eppley: Yeah. And there's like now you're getting tom although you have to compare that to what would a human do.

Rob May: But then it it balances, right? Because what's weird is you start coding stuff you wouldn't have coded. If you had to make a human do it and so it's like if it's cheaper, do you code more? Do you end up with spending the same amount of money? It's going to be a really interesting thing to watch.

Ryan Eppley: We definitely haven't caught up since Claude went all in at the neck of OpenAI with their Super Bowl ads. Just all in. Lots of conversations happening online about, you know, is are they taking it too far? Is this not what OpenAI? But I think like the whole premise was such a wellexecuted ad campaign of AI will have ads but not here at Anthropic. Yeah. What did you think about those ads?

Rob May: I thought the ads were good. I as many of you guys know I'm not an OpenAI fan. Not that I they they do wonderful technology, right? And they've been instrumental in the development of AI. But I think as a company, OpenAI has caught itself in this place where because they were early on spending so much on compute and had to raise so much money to do it. In order to justify that, they had to sell this vision of this massive TAM, right? So they're I mean they're trying to raise a trillion dollars, right? I mean, your TAM has to be 8, 10, 20 trillion, like whatever it has to be to justify that, which means you kind of have to sell this idea that you're going to do everything. And as a result, I think what that means as a as a manager, if you're there, is that you have to you have to play in as many markets as you can play in. And then that spreads you thin, particularly for a company that doesn't have the kind of, you know, historical infrastructure that a Microsoft or Google would have if they went to try to do something this ambitious.

And so, so I'm not, well, I think OpenAI does great technology. I think as a company, I think they're in a rough spot. And I think Anthropic has really really played their cards well. Like I you know that that team and the decisions they've made and the way they run that company so far from the outside looking in looks looks really really awesome.

Ryan Eppley: Yeah, really sharp. I mean I was really impressed when they came out with the keep thinking campaign. I guess that was like last fall wasn't even that long ago they came out with the caps. They did all the, you know, kind of pop-ups in different cities. Like great campa camp campaign because it's like, oh, like AI is starting to steal our thinking capacity, our creative, you know, energy as a as a species. It's like, no, no, no. We're we're arming you to do more of that. So, keep thinking. Obviously, a tap to, you know, Apple and their whole like think different. At least that's how I thought about I'm sure, you know, someone wrote about that. But now this one is like going from inspirational our company as the focus to hey we're not like the other guy literally like the song was like you know the antagonizing you know lyrics of the song. I thought the campaign was great. I have a bunch of friends that are like hey this is like very misleading. Like OpenAI hasn't even released ads yet. They haven't even said how they might or might not.

Rob May: I actually I think they did today the day after the Super Bowl. Yeah. Let me just pull up tech. Oh, they have it. It say OpenAI says today it started testing ads in ChatGPT.

Ryan Eppley: Okay, so it's in it's in development. Transparently like ads have to be a part of the equation if you're going to build any consumerf facing company that gets free access to something like there's there's got to be some component of that. And if it's not, then like they're really really harvesting just your data. And there's got to be some some kind of, you know, inter interaction. And I think ads is like the medium that makes things free, right? Again, I think Anthropic did a great job. I thought the campaign was so well executed. I was like almost more impressed with like the prowess to pull it off than I was with like the messaging behind it. But it puts OpenAI in a weird spot because I saw a ton of people from OpenAI. I think Sam even too having to comment on it. Yeah. And having to like denounce it like oh like this is a cheap shot or hey like this isn't true like you're misleading. And anytime you get someone this is like straight out of like politics 101. Anytime you have to have a statement be made to say like this isn't true, you've already lost. Like you never you never want to acknowledge it. So I think I think Anthropic going on the offensive is really interesting.

Rob May: I Nassim Taleb sometimes tweets all rumors are false until they're denied, right? And then they're

Ryan Eppley: That's a good quote.

Rob May: Yeah. Yeah. Actually we do have some some some other news like as we're sitting here and I was looking up the OpenAI thing. I didn't realize this but apparently what happened today. So, Alphabet, the parent company of Google, raised $20 billion in its biggest ever US bond sale after drawing more than a hundred billion dollars of interest in like that's what they could have raised, right? That was the demand. Wow. And so when you talk about is there a bubble or not, like clearly there is demand for these these AI companies. And I think that's a good sign. So, I think we continue. I mean, it's going to be interesting to see where because we talk about this, you know, at least every couple of weeks on the show.

Ryan Eppley: We have the meter. We talk about it every week. Don't Don't say every other week. Are we bubbly this week? Are we not? Where are we? Because I go back and forth depend on the stuff that happens sometimes.

Rob May: Maybe we'll make a bubble meter.

Ryan Eppley: That sounds good. That's a good idea.

Rob May: So, we'll send a note to Madi.

Ryan Eppley: Cool. Well, so do you want to wrap up on a New York question? Do you have one in mind?

Rob May: If not, I Yeah, one up.

Ryan Eppley: What do you got?

Rob May: Okay. So, you know, you mentioned earlier that sometimes you just open and go to neighborhood or whatever. What is the most surprising New York neighborhood that you've been to where you heard some things about it and you got there and it's not like you expected? I'll tell you mine. Mine is actually mine is actually Soho because people talk about it like it's this like hip whatever. I feel like it's like Times Square without the lights. It's super touristy. It's jammed with a bunch of people who aren't from here. And so I am there's a couple good places down there, but like in general, I try to avoid Soho just like I do Times Square.

Ryan Eppley: Oh, yeah. Yeah. I was going to say probably like what feels like maybe Wall Street. Every time I go to Wall Street, it does not feel like I'm, you know, gonna walk into like the billion-dollar banker. You know, I'm sure they're walking around, but it doesn't feel like that.

Rob May: How far away is that from Soho anyway?

Ryan Eppley: It's like not not too far. Yeah, I know. Maybe like two neighborhoods in between. Right. Right. Yeah. You're think about like FIA, right? Financial district. Yeah. I don't feel like I'm in the Wolf of Wall Street when I'm in, you know, that area.

Rob May: I mean, a lot of that is probably because postcoid and working from home and right all that kind of stuff. But yeah, a lot of a lot of neighborhoods are not what you expect. Plus, they're always changing and revitalizing, which is what's cool about New York. Yeah. So, great. Well, that's the show for today. Those of you listening, go go go check out our other our other shows. If you have people that you think should be guests, please reach out. Check out BePresent, our first sponsor. We're very excited to have. And yeah, give us a thumbs up, subscribe, all the stuff you're supposed to do. Check out our website. We're going to start doing job postings and other things related to what's going on in New York. Starting up a newsletter. So yeah, check that out and let us know what you think. We're always looking for feedback. Thanks for listening. Heat. Heat.