Building the Star Trek computer with Huxe's Raiza Martin

24:17And he tells me it's AI. And this is in 2022, right? Pre-Chat GPT, even pre-Lambda, as far as what the public consciousness was aware of. and I remember thinking, AI, it's just like a word we've thrown around for years, but it doesn't mean anything, right? Especially, like, for me, where I was like, AI means, like, Star Trek, right? It's like, and until it's up, like, Star Trek quality, like, don't call me. Like, I really have no interest whatsoever. But he was like, no, it's actually pretty good. So he did the initial demos of Lambda. He showed me, we had, like, some pretty rudimentary apps inside of Google, and I was like, wow, I don't really know what this is, but it seems really interesting, and so I studied really hard like before I even got the job and by studying I mean I spent like two days like reading as much as I could and that was like the extent of it but then I pinged Josh and I was like whatever it is I'm in right I was like hire me I was like I'm the guy like whatever it is you think you're gonna build with this it's me and uh he's crazy he really did hire me and so I was the first PM he hired uh into labs and I launched AI Test Kitchen uh launched Notebook LM previously known as Project Tailwind.

25:30But yeah, I just have had just like the craziest time over the last three years, like building cool shit. How many things at Google have been called labs over the last 20 years? That's my question. Well, it's got to be like at least a dozen. It's like we're introducing a new labs division within the existing labs division. I mean, it's such a popular word and it's such a cool word, right? Because a lab just sort of inspires this like scientific exploratory juice where you're like, I get to do fun shit. I work in a lab. And I think what's really cool is maybe not so cool. When we started Google Labs, we discovered there was previously a Google Labs already that had been started and killed.

26:10So there was a little bit of confusion where they were like, oh, is it the same thing? I'm like, I don't even know what the old thing was. Right. And then when when the new Google Labs got started, everybody at Google wanted to have a labs, too. There was like search labs, photos labs, YouTube labs. And then it was like, well, what is the difference between all the labs? And I honestly think like if you paused for a second to think about the answer, like you would fall down a rabbit hole so deep that you would become unproductive. And for sure, like I saw people sort of like thinking about this philosophical problem because Google gives you the space to think about whatever you want.

26:45And I was like, I am not interested in this sort of problem whatsoever. And so I just did not participate in that convo as much as I could.

26:53Alex Heath:I'm so curious about the origin of Notebook LM and how you guys decided to focus on that as a product. It's one of the rare consumer viral breakout hit products from zero to one that Google has done in a long time. And I'd love to know, like, how did you guys come up with it? And then what was it like experiencing that traction that happened so fast? Thank you for saying that it happened so fast, but it had actually been around for like two and a half years before it blew up. And so we were really like grinding in the darkness for a long time. You know, not that dark, but it was like relatively unknown.

27:31As far as the origin of it, I mean, I think one of the things that's most magical about Notebook LM is that we started the project before ChatGPT. And so the use cases were not actually that obvious. But I think there was sort of a really interesting prevailing thought that we had, which was, look, there's probably something interesting you could do with a chatbot that you could just talk to about anything. Right. And that's sort of like the varietal that we have Chachi, BT, Gemini, Claude, etc. today. And that's how a lot of them started. But then there's a second group of thinking or type of thinking, which is what I think we fell into, which was, no, I actually don't want to talk about arbitrary random things.

28:13I want to talk about a collection of things that I really care about for some reason or another. And this sounded really vague in the beginning. But I think I told Ellis this. The reason why it was so compelling to me was because I had done this crazy thing where, you know, if you remember, I had joined Google and I didn't have a college degree. And I had this like chip on my shoulder. I decided now that I had Google money, finally, I was going to go to college. I was like, there's a couple of ways to solve this chip on my shoulder. One way is I can certainly just get my degree so that I can't be found out.

28:48And so I went back to college. And so by the time we had started Notebook LM, this particular project, I was like, what seems like what these LLMs are good at are words. And you know what has a lot of words in it are textbooks and slides and all of these class notes. and what I would really, really like is to create a container of all of this stuff and I just want to talk to it. And at the time that we, you know, when we were building it, it was just not possible. It was super brittle. It was really hard. But I was like, we're at Google. Like it's only going to get better. And so I was betting on the future.

29:25We built the product like with this specific vision in mind, which is actually, it's kind of funny because it was like kind of hard in the beginning to convince people that this was an interesting thing to work on. just like from a product perspective. But it was true that as the technology got better and better and better, my grades got higher and higher and higher. No, I'm just kidding. I mean that jokingly, but it was true. It did actually start to perform better and better and better as the context windows got better, the model got better. There's just a lot that eventually, by the time it blew up with audio overviews, the app itself was pretty robust.

30:01How was the reception within Google for a runaway PR success like that? Oh, my gosh. I mean, I think it's interesting because I think about where Google was when this had happened. I think like it felt a little bit like it's a giant company, but it felt like we were like on our back foot somehow. Right. Like you could read pretty much anywhere and people were just like, oh, Google got beat. Right. They invented the transformer. But so what? And it was like really negative. And I think like working at Google, it felt really weird because you were this like trillion dollar company, but you felt somehow like a product underdog.

30:44Right. It was like people had largely sort of written us off as like not being an interesting product place. and I just didn't buy it. I was like, you know, something is going to happen soon, right? And if I had a say in this, like, we're going to do something interesting soon. And I would say that there was like this sort of frenzy in between when ChatGPT happened and when, you know, Notebook LM blew up. There was like this frenzy in the middle of just like we have to win, right? There is like an energy of like we have to do something. But it was a little chaotic because everybody was just trying stuff.

31:20And then there's like layers of politics on top of it. There's a whole cast of characters that are constantly like moving things around. And like for me, the notebook LM team was tiny. Like my whole thing was like, I just need to insulate this like core group of people and keep them focused. Right. And it's something. And I just believe like something interesting will happen. If you do something for long enough, something interesting will happen. And I was like, it's just my job to just like shield this team and just keep going. and I think um when notebook LM did blow up the thing that was most surprising was how I felt like the narrative changed I and I it was really really to me one of like the most culturally like shocking things where I felt like there was the narrative of like Google's like this like dumb product organization whatever they can't build new things it felt really prevalent to wow they're back they're doing cool things turns out they can build new things and I think like it started to to really pick up like in terms of like the momentum like ai studio right especially like with logan like tweeting so much it's just like wow they're actually like cool people nano banana jewels like look at all this stuff that they're able to ship really quickly and so i feel like if i had to sort of think about the difference between these two times where google was not cool google becoming cool there was like a very clear moment there where i'm like oh we helped kickstart that movement

32:41Alex Heath:You make Google cool. Put that on your business card. Can't wait. Can't wait to put that on my LinkedIn. I feel like Google being cool oscillates maybe at like an increasing speed these days. Yeah. Yeah. I mean, I think like the meme, right, is like everybody just takes turns. It's like, it's opening eye today. It's Google tomorrow. It's anthropic the next day. Grok even? Like I don't. Wait, so also you're telling me that 20 % time is actually real? Yeah. Well, I mean, it's not really 20%. I think that's the most ridiculous thing about it. Because at the time I was leading AI Test Kitchen and this was so funny.

33:22Like, I can't believe like Josh tricked me into this. He was like, and while you're working on AI Test Kitchen, he's like, why don't you build a new business for Google? I'm like, wow, sounds great. Sounds like a cool opportunity, right? I'm going to go build something. But it was very hard because shipping a test kitchen was hard. And then you had to think about this like totally new thing on the side with the same team that you had. And so we really did have to pull volunteers from like literally like every corner of the earth that you could find inside of Google that you, you know, a person that you could get jazzed about this like really foreign concept of like upload a document and talk to it.

34:02Especially because like in the beginning, there were not a lot of documents you could upload. But yeah, I mean, we were able to pull through. And I think like to the credit of I think leadership at Google, we did all of these demos and they were like, all right, keep going. Seems weird. Seems good. Keep going. And yeah, we did. it is a testament i feel like you could spend your your free time on working toward your next performance review or building the new or building the new revenue stream for for google and i think we know which one most people choose well i'll be very honest that i did not ever write my performance reviews i think like there was a very clear period in time where i just completely stopped i was like you know i really do not like doing this and i'm only committed to doing things that I like and so I stopped doing it.

34:51Alex Heath:There are tens of thousands of big tech employees who agree with you and also do not like their performance reviews. It's funny. My manager, like there was like one cycle where she was like, you are so overdue on like your performance review. She's like, how about I write it? I was like, smash an idea. Yeah, I think you should. I mean, I think you have another winning product there, Riza. Oh, yeah. AI for performance reviews. Yeah.

35:21Alex Heath:Support for this show is brought to you by Solonis. Ever feel like you're being told to wave this big magic AI wand and everything will just be better? Sure, AI can write emails, summarize some documents, and even churn out a business plan in a few seconds. But what about when it comes to AI taking on your big business opportunities, your big problems? It will tell you that's a great question. Then scrape the public domain and give you some generic recommendations. Not helpful when you've got a spike in demand and you're trying to reroute stock through a canal that just won't unblock. Solonis gives AI the context it needs to know how your unique business runs and how to improve it.

35:57Alex Heath:It's sadly not a magic wand, but it does lead to some pretty enchanting outcomes. AI needs context. Solonis provides it. Learn more about how the context model gives enterprise AI operational clarity at CELONIS.com slash context. Support for this show comes from Indeed. When the pressure's on and you need to hire the right person for the job, Indeed Sponsored Jobs has your back. Sponsored jobs posted directly on Indeed are 95 % more likely to report a hire than non-sponsored jobs. Join the 3.3 million employers worldwide that use Indeed to connect with quality talent that fits their needs. Spend less time searching and more time actually interviewing candidates who check all your boxes.

36:40Less stress, less time, more results. When you need the right person to cut through the chaos, this is a job for Indeed-sponsored jobs. And listeners of this show will get a$75 sponsored job credit to help your job get the premium status it deserves at Indeed.com slash podcast. Just go to Indeed.com slash podcast right now and support the show by saying you heard about Indeed on this podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need to hire? This is a job for Indeed-sponsored jobs.

37:13Alex Heath:so you're riding high you've made google cool and then you decide to leave and do a startup is that what happened yes yes i actually um i think it was so unbelievable to many people that they thought i was unwell and i didn't realize it until later that that people thought i might have not been okay but they you know i'll never forget when i when i quit um josh was like hey, he's like, maybe you need a break. He's like, maybe you just need a vacation. Maybe you're tired. And I was like, Josh, what are you talking about? I was like, this is the biggest moment in my life. I'm like taking this thing that I worked so hard for and I am leaving it behind because I see the promise of something else outside.

37:54I was like, I've never thought clearer in my life. And it was actually quite emotional for me, you know, because of like what I had shared earlier, which was I didn't come from a particularly wealthy background. I immigrated to the States. I didn't get to go to college. And I found myself at Google making more than I ever had in my life. And I was just suddenly ready to walk away from it. And so there was actually one night I woke up. It was like at 4 a.m. And I just burst into tears. I was just like crying. And this was before I quit. I was about to quit. And I was crying. And my husband was just so startled.

38:32He was like, are you OK? Does something hurt? Do you have to go to the hospital? And I was like, I'm so emotional about this moment because I realized that my dream was to go start a company. And I was like, I have never felt that the right time was now more than ever. But I also have never felt so poor in my life. I was like, it's just so, it's so crazy that I like am imagining walking away from this thing that I built for me and my family. And my husband was looking at me and he's just like, listen to this. He's like, listen, he's like, you know, they're going to try to give you like money to stay.

39:12Right. But he's like, how much do you think you're worth? How much are your ideas worth? I was like, what? He's like, what's the number? He's like, how much is your idea worth? I was like, I don't know. He's like, OK, is it$10 million? I was like, I don't know. he's like is it a hundred million and I'm like what are you asking me I'm like bawling my eyes out you're asking me a number and he like grabs me he's like listen girl he's like you are not a 10 million dollar girl you're not even a hundred million dollar girl he's like you're a one billion dollar idea girl he's like you go out there you do it and he's like and I got you and I'm like what where did he get this at like four in the morning but it was to me like just one of the most impactful moments of my life where I was like, here I am, you know, thinking about it this way.

39:55And I live with this person who sees me in this like totally different way. And he was like, you go get it. He's like, you go quit your job. He's like, don't go into this afraid. And I was like, man, wow. Yeah. I think he's coming on the show next. What a guy. He's a very, he's a very private, like very introverted guy. That's why like The pep talk came out of nowhere.

40:17Alex Heath:Sounds like the goat to me. Yeah, it was pretty good. I so identify with that as someone who just left the best job I'd ever had to start my own thing and to start my own company. And Ellis has started his own company. And yeah, I feel like a lot of people can identify with that roller coaster you just described. Congrats to all. Yeah, but you have to have something that you're working towards. You have to have something that is so motivating for you to leave that safety net, right? And so I'm curious, like, what was that? Besides, like, I want to start a company. What was the, like, product insight or the thing that you were like, this has to exist?

40:53So I think it's crazy because the idea that we left with is not even the idea that we've built. And I tell this to everybody, which is, look, you know, especially if you're coming out of a large company or you're working like a role, the idea that you have is going to change. and I think the thing you have to have faith in is that you are going to be able to navigate it. Like don't bet on your idea, but bet on yourself, right? Bet on your ability to navigate whatever the thing is that you're doing. And that has proven true for us. But the specific idea that we actually quit with, it feels funny to describe it now, but I remember at the time I was like, you know what?

41:34Somebody needs to give these LLMs computers so that they can actually do interesting things. and I was like why don't we build a platform to basically spin up a bunch of LLMs or agents as we call them now give them computers give them autonomy give them access to slack and messages and phone and have them go do interesting things on your behalf and they just talk with you just like regular people do and that was actually the that was the idea that we quit with and then when we quit we were like well wait a second you know let's take three days just study what's out there Take a beat, be like real human beings and see what is it that we want to build.

42:12And so the idea morphed many, many times before we got to where we are today. That's something every VC loves to hear, right? The founders that are just kind of obsessed with a specific problem space and they're going to get there somehow, some way. I guess for you, it's like been since childhood that you're after this. They're like, yeah, I'm putting money on that girl. I mean, I have loved like technology since I was like the beginning of time. And I honestly like I remember building computers with my dad, like even though we didn't have much money, my dad was like, we could buy parts and we could like build our own computer.

42:51and I've just like been that way my whole life and I think it's so funny there is a publication wrote to me they were like hey you know congrats on your fundraise do you want us to publish the deck that you used to raise money because like that's like a thing we do we share with people how you know people are raising money these days and I had to write back and I was like there was no deck like I didn't have a deck like I just went I met with VCs and I was like I did have a moment where I was like, should I have one? But it was very similar to like how I felt when I was at Google, where I was like, well, I could do it the way that I think other people do it, or I could just do it the way that I think I want to do it.

43:29And the way that I just wanted to do it was I wanted to show up and be like, look, here I am. I have an idea. The idea could be kind of cool. But the thing that's actually really cool is it's me and a group of people that I know can build something interesting. So it's like you can bet on the people or you can bet on the idea, right? Which kind of person are you? Let's shift gears to Hux, where we are today, how would you describe what it is? Wow. Well, today I would describe it as a personalized audio app that creates podcasts or radio stations based on the things you care about. It makes content from your emails, your calendar, your interests, and pretty much just any query that you type in.

44:08I've been using it for the last several days, and I've been pretty impressed. I feel like it's doing things on a number of different levels. it's one thing to get, like, the voice and cadence right. It's another to, like, create a nice program of your day. Morning, Alice. Sliding in just before that Hux session kicks off. Looks like today's paced, but pretty focused. Yeah, that back-to-back access recording block is coming up fast. We'll keep this tight so you can get prepped without scrambling. Some major waves in AI to get to later. Think huge chip deals and a big push on new developer tools.

44:42Also, music charts are shaken up again. if you need a breather from the tech headlines. But I feel like for most people, this is actually going to be their first experience with like agentic AI, like searching for things for them. You know what I mean? Yeah, yeah. No, that's totally right. And I think we were joking about this the other day where we were like, you know, the funny part about building something that works really well and is really interesting and is magical for folks is that on the surface, it has to be so simple, right? Like where you don't have any cognitive load about what the thing is.

45:14For the record, I don't think we're completely there yet, but we try our best to make it really simple. It's just like make a podcast or a radio station about anything. But underneath it, it is very technically complex to create a low to no latency audio platform about anything. And the promise is that, you know, you care about all these things, but maybe you care to listen to the topic, to something about the topic multiple times a day. right so it's like how do you at multiple times of the day tell the user hey something has changed like let's say like open ai right you could go days where there's like not that much interesting news about them but then one day there's like a very interesting news story whether it's like sora coming out or dev day happening right and suddenly there's a bunch of things you might want to listen to so i think being able to enable this kind of like i know this person is interested in open AI.

46:08I'm going to try to make sure I understand what do they know so far? How do I give them the most recent thing? How do I make it so that that content is actually personalized to their context? Like that stuff is pretty hard. And there's like a lot of complexity under the hood that goes into it. But I think that when you get it right, it's pretty magical. Where I was creating a podcast about the Mariners, which is like really funny because I'm not super into sports, but I'm like really into the Mariners and it somehow knows this about me. So it was just like going through like the different sort of like results of like, okay, here's like what happened with the game.

46:46Here's like the record. Okay. But let me explain why this is important. And it was pretty cool. I was like, oh, it's not like your typical sports broadcast. There's something about radio. I feel like I saw an infographic the other day that was looking at like the time spent for audio per day And like a solid third of it is still like AM FM radio. And I think in this moment where everybody is constantly pulling things on demand, like it's pretty damn nice to have someone just like read to you. Like I'm signed up to the L.A. NPR KCRW. And it's just like, here's five news stories from today. And I'm like, wait, not a million that I feel guilty about not reading, but just five, please.

47:25I think that's the interesting part of this moment we're in, which is I think about this all the time where we've never had this choice before, right? Where I'm like, hey, Dev Day, let's just take Dev Day. I really care about what happened there. I do not want to listen to an hour and a half keynote, right? It's like maybe I would on a different day, but today I need the five-minute rundown. But then maybe there's a different topic that I actually care a lot about, which is like, I want to know about like the history of Star Trek. Like how, like what was Gene Roddenberry thinking? Like I've actually, I have no idea.

48:01Right. So maybe I want to listen to like a 40 minute thing about that. But I think like the flexibility of a couple of different things, which is like how long I want to listen for what the specific topics are, what the angle even is, how is this even relevant to me? Like, I think those are the things that are possible today with AI that was like not really scalable or possible in like the pre-AI times. And so I think like there's like this very interesting question of one, do people care? Right. Is this actually like an interesting problem? And I think the second question of like, what are the correct knobs to give to people?

48:36Because it's like, you don't want to build a thing that has like a thousand buttons, right? And someone's always fiddling it. I think you could probably land on this like middle path where it's like, it's just magical for 95 % of people and the 5 % that want to fiddle, like we have the correct controls.

48:51Alex Heath:I really appreciated how simple the interface was. I'm still getting my head around the idea of AI podcasts, AI radio. I think because I'm such a podcast power user and I'm drowning in human created podcasts that I need to listen to, it feels like listening to an AI podcast feels just strange to me. But I will say that the daily overview thing was Dev Day was a perfect example. We're recording this the day after the open eye Dev Day. And I was I I was there and like I had a packed calendar. It was unusual for me. And I was traveling that day and I had like I was doing a day trip up from L.A. and I was thinking of you actually in this conversation a lot at Dev Day because I went to this Q &A with Sam Altman after the keynote and he was talking about how he thinks that AI will be primarily used through voice and he's like so into voice as the main input for AI and I think that probably has to do with the devices he's working on with Johnny Ive and obviously ChatGPT has advanced voice mode and all that but it does feel like voice is a really unique way, powerful way to interact with this stuff.

49:55I agree. I think that that's the next horizon for us. And I actually think that the initial step is, I think that it's easy for people who have this kind of like internal monologue to make the leap, right? Like if you already have an internal monologue, I think it's easy for you to imagine that you could have like an AI voice assistant, like you watch the movie Her and you're like, for sure, right? I get it. But then I think for most people um the movie her is just a movie it's just an interesting sci-fi thing because i think a lot of people don't think about the world as this like interactive interface for technology right it's super nerdy i think most people just live life and that's it and i think that the key to this is is actually passive content that you make interactive and i think that's the bet we're making with Hux, which is how can I give the user so much value with like nothing, like they don't have to do anything, right?

50:56And so like when you use Hux, like you open the app and you just press play, right? You don't have to ask a question. You don't have to start a chat. You don't even have to say, here's how I'm feeling today. And as much as possible, we try to constrain the amount of things that there are to press on the screen at any given time for this reason, because I think that this is like the next page of AI or technology in general, which is like, don't make it hard for the user to get something useful out of it. And I think if you lay that foundation down really, really well, you unlock the next step, which is you delight them with interactivity, right?

51:33And you sort of teach the user that like, hey, you know what? The world can be this like interactive, like technological interface where you're listening to something interesting. Maybe you have a question, We can't do that today. We can't interrupt podcasts today. We can't interrupt radio stations. But what if you could? And maybe it actually says something like, hey, Ellis, this is very interesting for you, this specific thing. And you're like, why would it say that? So I think there's an opportunity to build passive content that invites the user to participate in this totally novel next generation behavior.

52:05Alex Heath:As a professional podcaster for a long time, I would have loved to have been able to go back and interrupt myself listening back to myself and be like, stop saying um stop stop doing filler words i'd love to go back and redo my pods if you can invent that i'll pay you a lot of money i mean an um remover is pretty easy a lot of people in tech talk about like what is the thorough put of the different modalities and you're like what are you talking about and then they're like well what's the io potential of visual versus video versus audio and it's like dude like it's not about that it's just about the time spent yeah like parts of your day like and that's I think why you're seeing I mean what what is it like three to four hours per day of audio time spent like up there with video time spent I feel like uh for that same reason and it's not that it's like oh the death of reading or this or that I mean it's just looking at the opportunities and creating experiences for them yeah I mean I think largely um you know when we started this particular incarnation of the app, one of the things I was talking to the team about is like, it's kind of crazy that when you build anything, you have to compete for the user's eyeballs.

53:20It's like a weird business to be in, right? I am trying to steal your attention from something else that you are doing. And audio is really special in the sense that you can participate in it when you're doing something else. Maybe you're taking a walk, maybe you're riding your bike maybe you're working out maybe you're doing your chores and actually like a lot of our users they whenever they send us feedback they tell us the thing they're doing and it's like this really magical organic thing where someone will be like i was walking my dog when hucks did this or i was driving to work when hucks did this and it's like kind of cute where i'm like we're just people we're just people doing stuff and if we we sort of think about this it's like hey what's your place in a person's life and it's like how can you make that place really good like that that's like our singular motivation is like hey if you listen to it for like 15 minutes during your commute how do you make it like a pretty good 15 minutes where you were like wow I'm glad that I did that and I think that's why this particular race or this particular game of like audio is actually like a really meaningful one where you're like how creating meaningful and interesting audio content is not that easy.

54:31But it's like a pretty, it's a pretty special place. Yeah. If you unlock it, if you unlock it, it's pretty good. And I think about this, like I listen to podcasts a lot. I listen to music a lot. I listen to Hux a lot. And I feel good. Every time I do these things, I feel pretty good. It seems like in some ways, while we might talk about it as radio or podcasts or whatnot, it seems like you're kind of going to be at the center of this like holy grail moment of AI that pulls things for you and gives them to you, whether it's audio or feed. And you tweeted the other day that OpenAI's new Pulse feature has some striking similarities to Hox.

55:07Can you tell us about that? I feel like they cloned your old website as opposed to your new website. Yeah, yeah. It was interesting, I think, because when I first saw it, I saw the video first, and I was like, oh, my gosh. I was like, it's like the old website. But then I kept poking around, and I was like, some things were really similar like a framing of um you know it's all about you versus like it's all it's all you I was like okay it's like super close we looked at sort of like you know no scrolling listen and and it's kind of like personalized feed instead proactive versus like doom scrolling there was just like multiple things that made me think okay right if it was like one design element like that's that's okay that's cool but I think because it was so close I kind of felt like, well, you know, it's tough, right?

55:55Like, I don't work at Google anymore. Like, I'm the little guy. And OpenAI is this behemoth who has basically taken, like, our same message and they're saying, this is the future. So on one hand, I was, like, really jazzed. I was like, wow, this is, like, really cool. Like, we launched it before they did and they said the same thing. And it seems like many people are excited about it. so in general it's like very cool validation for the space and the direction we're in on the other hand it's super scary it's like oh man like do i really want to take them on um but i think in practical terms this happened all the time when we were working on notebook lm2 i think in practical terms i think you see a lot of this in emerging products and product ideas and product paradigms which is like maybe you don't do it on purpose right like you don't literally look at a website and you like copy it maybe you don't um but maybe you're influenced by it down the road right and it comes out in your writing it comes out in your design and then it starts to look really similar because everybody is kind of like trying to find inspiration for something that doesn't exist um what i think is true though is that what differentiates products is basically how long you go at the problem for and how differentiated your insights are And I think that's kind of a weird thing to say, but I said it all the time when we were on Notebook, which is like these things will happen where people will crop up.

57:20They will do the same thing you are doing. But the thing that you have to remember is that even though it's interesting to look at what they are doing, it's what you are doing that actually makes the difference, right? Like for your life, for your customers' lives. And the truth is that you just have to keep going. You have to keep going at it. I mean, I think even in the two weeks since we've launched, I think we have learned so much about what users want and what the user experience is that I'm just like, wow, like that makes me feel good about the direction we're going in. And I don't particularly feel the need to stop just because, you know, we might be doing something similar to OpenAI.

57:55Alex Heath:I asked Sam Altman at the Dev Day Q &A yesterday about Pulse, and he said it was his favorite feature. and for those who don't know it's gated right now to the 200 a month pro tier of chat gpt but it's the first time that chat gpt will use your context to suggest things to you and push things to you like like hawks like like more of an assistant uh and i think they see it as a huge opportunity to make chat gpt less um insurmountable less overwhelming to people who don't know how to prompt it, right? Because that's the thing with AI right now is to really take advantage of it. You have to know how to extract from it, but there's something magical when it can just bring things to you and learn from you.

58:44Alex Heath:And I think that's like what you're doing with Hux and OpenAI seemed very, very excited in that direction. So it's really interesting and you can monetize it in a lot of interesting ways. Like I'm curious, like how you think about that? But is it for a product like this, do you just do subscriptions or do you think that at a certain scale, there's an opportunity to do like really good high touch advertising in an interface like this? I think that's probably what OpenAI will do actually for Pulse is kind of what they were hinting. So I'd be curious to hear how you feel about that. I mean, I think we're really, really early still.

59:21but I'll tell you like my early thoughts on it which is I think that in order to keep most products free like you'll have to do advertising I think that's just kind of how these types of products work but you know the stage we're in is we're just trying to validate is this interesting to people do they want this is it useful and I think when we get to the the part about monetization I think those are the only two avenues which is like you could do a subscription you could do ads You could do both. But the way I think about ads is if you have a system that actually knows you uniquely well, I think you could solve a really important problem for both advertisers and people, which is I don't really want ads for things that I don't want to buy.

1:00:05And I know that advertisers don't want to spend their money showing ads to people who aren't going to do anything about it either. and I think about this sort of like two-sided problem where it's like okay look let's say you're like blue bottle and you want to advertise right I think it's not terribly difficult with how Hux has been built to just know like okay this person's going to be in Palo Alto they have a meeting on University Ave right this person drinks a lot of cold brew because I see their receipts all the time it's not crazy to be like Riza you have a meeting blah blah blah here's an ad and just disclose it to the user and just say it's an ad.

1:00:44Blue bottle's right there. They have cold brew. They have a single origin blend. They have a multi-origin blend. It costs this much, but here's a 25 % coupon. And here's the bathroom code. And here's the bathroom code. Which we pulled from Yelp. So I think there's something interesting to be done. I think we're kind of far away from doing it. But yeah, I mean, lots of things to think about there. How soon is the quality of the voice going to be indistinguishable from humans, do you think? I mean, I think you see these studies that come out and they're like, oh, voice AI is not distinguishable from humans, right?

1:01:23Most people can't tell already. And so I see those studies and I think about it and I'm like, well, I wonder what those contexts are. You know, whether it's like customer support or maybe when you make reservations and you call for things like that. but I think when you're producing long format content I think that there's like an interesting thing that happens which is you can probably tell it's not a human but the quality is good enough where it probably doesn't bother you anymore so I think it's probably not so much that you know you can't tell it's person versus not a person but I think it comes down to like is it tolerable enough for this particular media that you're consuming and that's really what what we measure which is like a lot of people are put off by ai voices because it's like really weird it's like still very like dystopian to think about there's like things pretending to be people but i try to think about it from a perspective of like human voices have a particular quality to it where we can listen to them right like it like mentally we're just like yeah i listen to whatever this person is saying and i'm okay with it whereas like my kids call them clankers right they're like i'm not listening they actually say this all the time they're like i'm not listening to a clanker I'm like, whoa, that's super funny.

1:02:38It's like, how do we get to a - How clankers in this household? My kids joke about it all the time. They're like, mom, do you think like in 10 years, I'm going to be like introducing you to my clanker boyfriend and you're going to be like, oh, Zoe, why'd you date a clanker? You know? But I think that's the question, which is like, are AI voices good enough to be listened to? Yeah. I mean, none of the AI voices are going to be as good It is the 93.1 Jack FM voice here in Los Angeles. Have you heard this guy? No. He's incredible. He's like, we play what we want, Jack FM.

1:03:13Alex Heath:Yeah. Yeah. Highly recommend. We've been dancing around this, Ryza, but behind all this is the fear I have that you're going to put us out of the job, Ellis and I. And so I'd like to know your advice for human podcasters, measly human podcasters in this moment, it because it is kind of crazy when you see how things are progressing and it's like well if someone can just get a hyper personalized podcast on any topic it's like where do I fit into that yeah yeah no I mean I think this is like a serious question and it's a serious question that I asked when we worked on notebook lm and I'll say that like my my perspective on this hasn't changed too much like it's it's changed a little bit but with notebook one of the things we discovered is that people weren't trying to replace existing content.

1:04:03They were just creating content that didn't exist. And it was really kind of funny to be like, people were creating podcasts out of their resumes, or they were creating podcasts out of like, you know, personally, I got this like 100 slide deck from like the workspace folks when I was at Google. And I was like, there's just no way I'm going to read the thing. Right? There's just like no chance, right? And there's no way that Ellis and Alex are going to make a podcast about this hundred slide workspace deck. But now I have the ability to turn it into one. And so I think like largely that is still true.

1:04:35People want content that just doesn't exist and is not feasibly going to be made. But I think one thing that is true that has changed, you know, since since we launched Notebook LM is I think it's kind of like the eyeballs perspective, which is you are just competing for a limited amount of space in a user's day. And I think that's very real. And we think about Hux as like something you add to your routine versus something that displaces something in your routine. But that's like overly optimistic. You know, of course, it's like replacing something, especially if you're already like a listener.

1:05:12And so we actually ask people all the time. We're like, what did you replace with Hux? What were you doing? And I think for the most part, many people actually don't say that they were listening to something else. And this to me is like very interesting, which indicates either we are reaching a subset of people that were not podcast listeners to begin with or like music listeners. But I think we'll probably see the results of this change over time. And so my my sort of TLDR on here is that for the first time, I think we have technology that competes with like the audio space. I think we're going to discover how that changes.

1:05:49Personally, I still listen to the same amount of like podcasts that I do. I just listen to Hawks is like a totally different time.

1:05:56Alex Heath:All right. We're safe for now, Alice. Yeah, I feel like, I mean, this is kind of my broader take on like AI's intersection with news and whatnot. I think it's like if you're like a link blog or a link newsletter that aggregates a very, I don't know, predictable subset of tech news. Like, yeah, I mean, that's also just not that differentiated or interesting even like of a skill set, you know. Meanwhile, if you are bringing in all elements of your experience, opinion, expression, even if you're an artist, like that's the thing that AI really can't do. And so I feel like it's just pushing us in that direction.

1:06:33We need to think about how to help you self-actualize into the most differentiated version of yourself, Alex. I think that that's how we ensure your future is safe.

1:06:41Alex Heath:Yeah, this is going to be an important process we'll do together on the show. What was a product you tried lately that blew your mind? Oh, I really like Sora. I really like the Sora. And I was so surprised because I'm not a video consumer, number one. Like, I don't even use TikTok, which is embarrassing. But I am also definitely not a video creator. But I was really impressed with the novelty of the Cameo feature. and the thing that was so startling to me was that I wanted to see myself in the videos, which I have never had any such desire to do. But the very first one I made was me in Arrakis, like in Dune, with like a sandworm.

1:07:25I was like, that is so goopy. It is so cool. And it was delightful. And I probably used it for 24 hours and I didn't use it again. But I think for the 24 hours that I did, it was quite magical.

1:07:37Alex Heath:Ryza, we really have enjoyed having this convo with you, and we really appreciate you coming on the show. Thank you. Thanks for inviting me. This has been really fun. Thanks for asking me about my favorite Star Trek captain. To be continued, we'll do favorite species of aliens next round. All right. Species A472 is my all-time favorite. Cardassians, for sure.

1:08:03Well, that was a fun chat. and Alex, you now have a personal podcast to listen to as you scoop your dog's poop every day. Is that like literally the only free time you have left that has been unmonetized by some company?

1:08:17Alex Heath:Pretty much. That's my Huck's time. We're new here. So please follow, like, subscribe. You know what to do on all your favorite platforms of choice at AccessPod, including YouTube, where we have the full video of the show. And if you like this episode, please leave us a five-star review or thumbs up and share it with a friend. You can find Alex's well-sourced newsletter at sources.news slash sources. No, really, it's just sources.news. And you could find me at Hamburger on Twitter and at meaning.company. Our producers are Shiraz Dume and Rishi Rajagopalan. We will see you next week. Bye.

1:09:03Alex Heath:Support for the show is brought to you by Salonis. Ever feel like you're being promised AI that will magically solve any problem your business might have? Sure, AI can chat and summarize. But what about big business issues? The ones affecting your unique company? You need the Salonis context model, which gives AI operational clarity so agents can reason correctly, decide sensibly, and act reliably. AI needs context. Salonis provides it. Meet the model at C-E-L-O-N-I-S dot com slash context. How many third-party vendors does your company use? 20? 200? Thanks to AI, someone on your team probably added three more this week.

1:09:44And your security lead has no idea. Traditional third-party risk management can't keep up. Vanta gives you continuous coverage across every vendor automatically so you actually know what's in your stack and what to do about it. AI on, risk off. Vanta.com slash TPRM