Is AI the Sole Future of Consumer Tech, What's Missing in Social & Will Glasses Be The Consumer Platform with Bryan Kim, Partner at Andreessen Horowitz

27:22What they reported, AI companies are growing faster than SaaS. And so what that means is that there's a real use case, there's real money that is being made in sort of the net new companies from monetization, engagement, willingness to pay. I'd love to dive into moats. So from what I understand, hugging face, that's more on the infrastructure side for AI. Is that right? That's right. That's right. The founders started as a consumer company that pivoted to Infront. Yeah. Okay. So how then do you think about moats when it comes to consumer AI companies? What does that actually look like for you when it comes to building big consumer AI companies themselves?

28:05Yeah, I think there are a couple of things. Like you said, there is a classical sort of marketplace dynamics and there are companies you can sort of squint and see, oh, you can really corner a certain set of supply, if you will. Let's call that net new models that are coming out for Hugging Face or net new image models that are coming out for Civit. And if it all lives in a certain place, then if that's the use case, you just go there. So I think that's like one example of classical marketplace dynamics that we're starting to see in sort of AI companies. The other one I would say is sort of workflow almost.

28:40And this sort of weighs a little bit into the prosumer world. But because these products are so magical and useful, what ends up happening is people start to do a lot more on it. And when you do a lot more on it, let's say I started creating most of my videos or images through a certain product, specific product. And what it does is it starts to build out, one, my library and my history of all my body of work, if you will, part one. Part two is if the company is thoughtful, which a lot of our companies are, are thinking through, how do I actually make it easier to export that? How do I make it easier to distribute that?

29:16How do I make it easier for, you know, customers to do what they actually want to do, whether it's distribute, monetize or create? And the more parts of that workflow they own, the more essential the product becomes to their lives. And therefore, it becomes really, really hard to rip it apart, rip it out of their life. So I think, you know, incremental ownership of the larger workflow of what they're doing is part two of how I think about sort of moat. And third, of course, there is this idea of memory or idea of experience that you have built. So that sort of maybe applies to companion use cases or mental health use case or any other use cases where the more you interact with something and the more it's harder to replace it.

30:05So Mike, I'll give you a quick example where maybe this is a strange example, but I write my journal in Apple Note app every day. It's, you know, two to three pages long. What I do with that is I copy it and I paste it in ChatGPT. And I say, remember this. This is my journal. And so at this point, ChatGPT has like a 500 pages journal on me and my deepest and darkest thoughts and secrets. And I ask you questions like, hey, I had this happen. What do you think? Use examples of things I told you already. And that's like sort of a very difficult thing to replicate because now it has 500 pages of documents on me where I'm asking specific questions about myself and it's giving me very specific example that I already gave.

30:48I'm like, oh, like that this must, you know, be similar to the situation like two months ago when you did X or when you went through X. That's incredible. And it starts drawing all this parallel. And like, that's something that I'm not willing to, you know, change. So I think there's a habituation and memory aspect of AI companies that is slightly different from prior products. Yeah, that makes a lot of sense. I mean, obviously, if you're building like a digital twin of yourself, then trying to replicate that digital twin into another, like the actual switching costs of actually trying to take a digital twin and put it on another platform, that would be terribly time consuming.

31:29mean so so why would you actually want to do that that's what that those switching costs and so you actually do have a moat there do you do you think that he talks me a little bit because we've obviously seen these huge rounds when it comes to ai companies um seems like honestly they're like in their own world just in terms of how much um how much money they're um they're raising he talks a little bit about some of the expenses um that that ai companies are kind of the reason why they have to raise raise such large rounds and i know that you know l lm's massively expensive and takes off and takes a lot of money in order to build them.

32:02Does every consumer AI company need to actually have their own LM eventually? I would just do a very quick one-liner answer on that. No, no, I don't think they do. In fact, competing to try to build your own foundational model is very expensive, like you said. The compute cost, inference cost to try to do that, not to mention potentially a data cleansing and data aggregation costs for licensing costs can also be very high. So my answer is no. I think there are incredible open source models that are near basically state of the art that are coming out, whether it's from Lama or Mistral, what have you.

32:44So I don't think they need to build their own foundational model. However, what I would say, Mike, is that people, the slight difference that I see now in the age of AI versus prior is that before, you know, these tricks is the wrong word, this like sort of ecosystem and technological baseline of mobile apps, fairly well understood. And there was like some baseline. Therefore, the game, if you will, was more around how do I understand nuanced information? How do I understand UI and UX to really deliver and capture the heart and mind of consumers? I think that shifted a little bit in that every, even this week, right?

33:27Pika dropped a new model. BFL dropped a new model. OpenAI had a dev day. Google LM is going viral. There's all these new things that are happening that if you're not actually very technically savvy and understand the underlying shift in the model level and technical level innovations, it's really hard for someone to want to build on top of it, right? If you're not aware of it, it's like hard to be on top of it. So what I actually am observing these days is founders with deep appreciation of the current innovation and the underlying foundational model layer, appreciation of that, and know how to translate those state-of-the-art into their customer problems are really, really important.

34:19versus, oh, I'm going to go build my own model because I have this preconceived notion of what it's going to do. And I'm going to incur tens of millions of capital to actually go train this thing, which, frankly, in two weeks, a larger, you know, larger hyperscale company can release and you just may not know about. Got it. So, yeah. So I guess it's staying in some level in your own realm of genius in terms of what actually the use case that you're doing and build it out in also a lot cheaper of a way, right? Because building your own model is terribly expensive. What do you look for when it comes to actually founders building AI?

35:00Do they have to have deep technical ability? Do they have to really kind of understand as well distribution and marketing? How do you evaluate teams? I would, you know, in the age of AI, I would say three things. It comes down to three things for me. One is, do they have an appreciation, an understanding of the underlying tech? And again, this doesn't mean they know how to build models or that they're researchers, right? It just means they know what's happening and they are, you know, really fluent in the recent development of the technological shift that can affect their product. So part one is appreciation of the underlying tech.

35:38Two is velocity. Things are changing so fast, so fast these days. So I think, I'm not sure if I would say gone are the days, but it's a lot more, the probability of success goes a lot higher, becomes a lot higher when the founder is able to move very quickly because the underlying stack, the underlying tectonic shift of the platform are changing so quickly that people who are more likely to succeed are the ones that understand that and move really quickly. Every two weeks, every three weeks, new product, out, out, out, out. And not everything will work, but that velocity, when things are changing so fast, matching that velocity is so important.

36:21So I became a lot more appreciative of founders who have a tendency to ship faster than to really sit and think of all the nuances to Instagram. And the last one I would say is, you know, you have to understand your customer base really well and understand their problem set. So those three things were a combo of under like appreciation of underlying tech, velocity and understanding the customer problem set. These three things is what I tend to look for in founders these days. What's the earliest that you would actually enter a company? Not not talking about money wise, but more so about not even not even how much they've actually monetized.

37:00It might not have monetized, but in terms of number of users and, and even just understanding that there is there that, that, that they are actually solving, you know, for you to get to conviction, understand that they're actually solving a real problem in that, and that customers are maybe realizing that they are or, or consumers are. Yeah, that's a great question. Oftentimes that this, this depends on partners. So I'll sort of give you my N equals one, like my own personal preference where I tend to enter is, you know, I think I've sort of grown this humility that I don't know anything, that I can have all these opinions about consumer behavior.

37:40Oh, this should go well, this will work, et cetera. And I think time and time again, there are developments that I see that I don't understand. I'm like, huh, like, why is that a thing? And I think it goes, you know, to draw an analogy, I think, you know, the older generation when Snap came out and photo disappeared, photos was always meant to be commemorative like why why would you take a photo that disappears that doesn't make any sense similarly when i now think about you know some of our you know the children of my partners they they now have this behavior where they're doing school work and they're on discord or chat groups what have you and it's all their friends and they have two to three bots in there just hanging hanging about talking back and forth they're they're you know contributing to the conversation.

38:24Mike, why? I can tell you right now that I don't fully understand why. Like, I don't really get it. But it doesn't matter. It's happening. That's what's happening. People will do what they do. So my job is to observe. So Mike, when you ask, when is the earliest you play? I like to look for some statistically significant number. and that means if you're a very extroverted founder you know maybe you have 100 friends and you ask all your 100 friends to use the product and when you're raising a fund ask them to say log in every day please spend 30 minutes each day so i can like juice my metrics what i tend to think of like this like non-gamable statistically significant number is probably in the low hundreds that you know product is no longer used by just your closest friends And it's sort of out there.

39:17It includes random people that would act more like normal customers. And then from then on, I try to tease out how people are using it. So it doesn't need to be thousands or hundreds of thousands of user base, but it does have to be some number where I feel it's representative of a larger group of people that you're trying to address. And that's when I would typically play. That's still pretty early. um so how how as well when when you have some of these companies raise you know pretty pretty significant rounds and of course um and some of them as you pointed out you maybe don't need to actually build your own llm actually maybe it doesn't make sense at all to build your own llm where where do the big expenses then lie when it comes to actually scaling some of these um uh on on the AI companies?

40:12Where did some of the basic expensive come from? A lot of the AI companies, they're structured. This is like old CFO in me coming out. I used to be a CFO before A6CZ. It all comes down to some extent, like a margin structure. A lot of the AI companies that we're talking about have a naturally high margin structure, 80, 90%. So what that means is that most of the cost is sort of similar to the traditional cost structure of SaaS companies, whether it's compute, cloud, or people. One of the largest cost center for our companies is really building out the team that can ship the product. It's the product people.

40:54It depends on the use case, but you may have research folks, engineers. Oftentimes, we do see companies that are sort of earmark some amount of capital to really distribute the product to grow through acquisition or paid acquisition motion. But the truth, Mike, is that most of the, I don't want to say most of the companies, a lot of the companies that I see and are really big fans of, a lot of them haven't touched the actual capital rates. Because AI actually, the product that these companies are delivering are very, very useful. People pay for it, high margin. And therefore, more often than not, what I'm seeing right now, Mike, is that these companies are operating at a way where they're purposely running thin to cash flow break even.

41:47So they're not making money most of the times, but they're not really drawing down the capital that they raise. Right. And that's the type of thing that we see where they're thinking about company building in a very enduring way to say, OK, here's my margin structure. I'm near I'm almost net burn zero. I can I can sort of invest more in growth. I can invest more in people based on the growth potential and what I want to get to in the next stages of the innovation curve. And these are typically the companies and even in the AI world that we get most excited about. I want to shift gears a little bit into an area that I know is very deep in a very deep love to you, consumer social.

42:32I know that, you know, Andrew Chen tweeted out and also wrote an article about how there won't be these big horizontal ad based consumer social apps anymore until maybe there's a new platform shift, which I definitely want to also talk to you about that. But and that and that now we're now we're seeing a shift that if you that the future of social is more geared towards vertical subscription apps and how that's maybe the only way to win. What are your thoughts on this in terms of the future of social as it regards to some of the up and coming emerging companies? So part one, I think it's certainly more difficult, right?

43:10Because in this world where large products like Instagram or Snap or Pinterest or Twitter or TikTok has sort of carved out their own areas, it's like fairly harder for a net new company to enter. And it's also a factor of look like time spent on on mobile are limited. I think we're all sort of spending almost a maximum of time that we would spend on, you know, mobile mobile products and, you know, not spending time on something just for spending time on a net new thing just means you're spending less time on the other things that are also really, really useful and fun. So what I think it is harder, but two, I would look at it this way.

43:49And this is how I look at the social scene, if you will. Social products are not one thing. They're largely two categories in my mind. One is interest graph products, where you don't need to be connected to people. Like TikTok, I don't know anybody, and it's still a really useful and amazing product. Discord, don't need to know people. If you have friends, that's great too, but you go into where you want to. Same as Reddit. Twitter, I don't know most of these people, but I have these interest graph products that are eminently useful to me. So the other, I would say, is, you know, what I would call friend graph products.

44:24And these usually start as a friend product where you're connecting with people you know, Facebook, Snapchat, Instagram all started like that. The thing, Mike, that I tend to think about in this friend graph category, what I would call like interest graph category, like there's TikTok of the world and it's just very, very hard to compete. It's so addictive and so fun. And so it's harder. On the friend graph side, look, I think the truth is if you look at the average number of friends on each of these platforms, whether it's Instagram or Facebook or Snap, I can pretty convincingly tell you that this is way, way, way larger than your actual close friend group.

45:08Right. so you know for someone on instagram of like 500 or 1000 or 1500 friends like is that really you know the the close friend and the graph that you have in there they're you know these platform making changes to add that green uh circle instead of a you know close friend uh circle versus not and to like really help that but the truth to the matter in my my mic is that there is no close friend graph product still that people use and people crave connection they crave a sense of belonging they crave being able to share and express their feelings and sort of communicate with their friends and if you think about where this is happening nowadays it's like really messaging apps they're happening on iMessages group chats what have you but just because it's happening on group chats doesn't mean that there's a close friend graph product that really deserve to be uh the crown jewel of like the new social category so i am still of the mind mike that the close friend graph throne is up for grabs it's up for grabs and there there's a world where that can be built i remember there there were like a couple that popped up that i mean I'm sure I'm sure way more than a couple, but, but, but that, that, that was that we're in that space, but weren't able to make it, but that's, that's actually really interesting.

46:35So, so there's still opportunity you think from a horizontal play that for like a close friend graph social product. That's my view. Yeah. I mean, that, that makes sense. I mean, right, right now how I typically communicate with a lot of my friends, it's all through like, you know, WhatsApp, it's all, it's all whatsapp groups and that sort of thing and so and so do you actually have you know a way that's actually where you actually have real profiles of people that you actually you know that are actually it's like you know the early facebook what they what that used to be um when it came out do you could you actually have that again so um uh because facebook right now doesn't really solve i would say that um yeah i mean my my sort of understanding of the technical or technological world is like history rhymes and it has like some sort of cadence so like maybe we go back to a place where there is like a like a myspace where you have your you know identity or your interest as a as a profile don't know like i again like similar to what we talked about not going to pretend i understand what what it's going to look like but i do think there's a lot of vectors that are interesting for the separate product to arise what what do you think will be of course in the past two weeks we saw um which i know meta had been working on for for a while and uh in partnership with with ray-ban but release their their ai release orion their ai glasses do you think well what do you think will be the next computing platform maybe after the iphone does it would it be glasses would it be goggles something a bit thicker and uh and and maybe not as much ease of use.

48:14Would it be something in that realm? Or are you thinking that it could be something completely different? Man, I think maybe this is a controversial take. There are more of mobile devices than humans out there. And when we say next device, next platform, I'm very, very excited for that and for what that could be. However, I also understand the time and effort it may take to replace, again, seven, eight billions of these things out there. So I think the distribution is so, so ubiquitous that I still think the main platform will continue to be mobile. Again, glass is a very interesting form factor.

48:57So are watches. So are all these wearables. Would it get to, you know, penetration that we've seen with mobile? Would it, you know, over time function completely separate from that? I'm of the view today, Mike, that it's going to take, I don't know, maybe a decade plus for that to change. So do you think in terms of glasses, it might be more of like an iPad where it's like a nice to have instead of a main computing platform? It's funny. The glasses, I don't know if you're a fan of this. I'm a nerd. So I grew up in Dragon Ball and there's a scout. There's a thing called Scout that like enable you to see people's energy level and like HP, which is kind of fun.

49:41I think about that a lot when I think of glasses where not everyone wears glasses. It is very, very useful, but not everyone wears glasses, just like not everyone wears watches. I think it enables very unique experiences and over time, whether it's spatial or visual in front of your eyes or being able to real-time translate in different languages, et cetera, I think they can be very, very amazing specific use cases. So I think over time, the way I think about it is that it'll sort of settle in couple use cases that are really, really helpful. So I tend to think of the watch as like two things.

50:21One is that it's a notification device on your watch that enables you to, you know, non-awkwardly check emails or whatever while you're with something. Part one. Part two, health. Like that's what it's for. In my mind, those are the two main things. now for glasses again i i don't expect to know what what is going to be but i think there is an incredible power and what i would call dismantling of power of babel right like in in bible like god created power of babel to like split the humanity and say like you're all you're all gonna have different language and culture and like not gonna understand each other like this type of a the underlying ai technology loaded up into like a form factor that's always on your like ear and mouse level could enable complete, you know, it sort of dismantling of Tower of Babel.

51:08Now I can talk to you in any language. I can speak in Korean and you'll understand and you'll speak back in Roman and I'll understand it and what have you. And so I think there would be specific use cases that arise out of like the Orion type form factor that would be incredibly useful, incredibly well adopted and very, very interesting. And those are the use cases and product that get built on top of it. I'd be very excited to see. Got it. But in terms of actually replacing a mobile, it's more so from a use case to use case or more maybe niche or more particular instead of being ubiquitous. I think that's right.

51:45As an investor, what's one contrarian belief that you hold? And maybe it's this one. But what's one contrarian belief that you hold that's maybe be different to your colleagues at A16Z about the future of consumer technology or AI? Interesting. That's a great question. It used to be contrarian that you have to have a thesis versus a consumer you observe. I don't know. You just got to watch. So that used to be semi-contrarian, I think, more and more accepted that, yeah, we really don't know what's going to happen, so we're just going to watch. Maybe the thing that I held was this.

52:25that in sort of products or consumer, maybe this applies to consumer and everything as well,

52:35I'm always of the mind of prioritizing retention over growth. And what I mean by that, if there is two separate product that looks almost identical, one is growing really well, but really like not great retention. And one is not growing at all, but has incredible retention. I would much rather bet on the latter finding out a way to distribute or the latter finding some hitting a lightning in the bottle and going viral on TikTok, whatever, and actually getting distributed versus the former figuring out a way to make the said product much more retentive. I.e., I will bet on the product that has much more love and retention than strong growth and strong distribution.

53:17That makes sense. I also, on this show, since it's all of consumer, I also have a lot of food and beverage investors as well that come on that will say kind of the same thing when it comes to retail strategy, where they would much prefer investing or looking at brands who might be only in one retail region, but their velocities are insane. They're so powerful rather than, you know, across the US, but their velocities actually aren't that great, even though they have a lot more accounts, a lot more doors. Do you think in, you know, in this kind of age of AI or the beginning of AI, when I say AI, like the LLMs, let's technology, do you think that there's going to be new ways to actually monetize some of these apps?

54:06Or is it going to be very kind of traditional that we've seen so far in consumer tech with it could be through it could be through advertising? It could be a it could be through a subscription. Do you think that there's going to be new kind of ways actually to actually make money through customers with with these AI technology?

54:31So, yes. at the end of the day if you think about make money as where the money comes from whether the customer pays or whether a third-party advertiser pays right i don't know if that changes a whole ton because we're just saying like either you pay or someone else's pay like right right yeah i think that's the same yeah but if you think about what is the character or what is the what is the nature of the of the revenue stream i think one thing that i'm actually quite intrigued about Mike is AI right now is in many cases so useful that it starts to, you know, if you think about the historical thought process around, oh, like let's take SaaS as an example.

55:16I will, you know, charge you based on seats, right? There's some licensing fee and like every per person, I'm going to charge you some dollars and then you're going to pay me that. And I think already we could see a world where actually forget, forget seat, because my product is so, so helpful that you start using this, you'll be a hundred X more productive. So I don't want to charge you per seat because you know, that, that's a, that's not a great, great deal for me. What I do want to charge is, Hey, like if you actually make a lot more money or you have to end up spending a lot less money, I want some, I want some part of that.

55:51And I don't know how to account for that. Hey, but I want to figure out a way to actually account for how much value I actually delivered to you. And I want to cut out that. If you end up hiring two people instead of two 20 people in a call center, because I gave you my technology, I don't want to charge you for two seats. I want to charge you for the, you know, some fraction of the revenue. It's like, what, what is that? And like, how are people thinking about the sort of outcome based pricing? And that can happen on the consumer side too. Hey, you saved, let's say you're a content creator and you spent 100 hours editing and doing all these things on video.

56:30And all of a sudden through AI, that shaved to an hour. How much would you pay for that? And that subscription potentially, but also you could sort of base on the value you're driving, I could potentially figure out a way to get a cut of that. So I think there is a new way that you could imagine based on how useful, just how useful this technology is that you can monetize in a slightly different way than before. That's really interesting. Actually based on revenues that you're creating or money that you're actually saved and actually monetizing that way, rather than just having a flat fee or, as you say, one additional seat or what have you.

57:13but actually the depth in terms of what the product actually saves you on money-wise. It's really interesting. My final question is, what's one book that's inspired you personally and one book that's inspired you professionally? Oh, different. So one is personal, one is professional. Okay. I love books. This is great. By the way, if you love books, there's a product called Readwise that I religiously use and I forget things, right? So you load books and you highlight the books you like and it gives you like a key um phrases of the book daily so like you're always like refreshed in terms of like the books you like so uh you know if i think about the books that i have there uh on a personal level i don't know inspirational is the right right word but i love this book called the order of time and it's by carlo rovelli who's a physicist i think he's I don't know, like some crazy smart physicist out of Italy.

58:12But he writes like a novelist, and he blends these ideas so well. And one of the reasons why I like it is just, it's about time. And I think that's like the limited resources that we all have here. And, you know, even as like an investor, like figuring out ways to actually use your time wisely is very important. And it just sort of highlights the relativity and the impermanence and the different definition of time versus like the static, look at the clock and we have an hour type sort of way. So it gives me, it's given me like a deep appreciation about the concept of time, if you will. So that, that's my favorite personal book.

58:51Professionally, this is a weird one. it's a weird one um so the book is called when we cease to understand the world it's by a chilean author he also wrote a book called maniac which i think is a follow-up to eniac uh like not a crazy person but actually some you know shortened version of a crazy computer um the reason why i like this book it follows you know the key physicists and like these leading mathematicians and innovators in the history. And it sort of tells you how peculiar and unique they are. They're not normal people. They're not at all. They're not like you're nine to five. They're crazy.

59:36And I think what I really appreciate about that, Mike, is that a lot of the founders that we have unique strengths. And, you know, if they had a sort of very regular, you know, point of view of the world, I think they would do regular things. And so similar going back to like my Alibaba story, we had this like crazy English teacher, Jack Ma, who didn't really speak English that fluently, that wanted to change entire China by doing these large things. They're not normal people. They're dreamers. They're visionaries. And as a result, our interaction with them and how we understand them, how we work with them also tend to be quite unique.

1:00:21And so this has given me an appreciation of the people, venture, like we talked about business, business models, companies, but at the end of the day, it's built by people and the founders. And I view this job is such a unique privilege of working with these unique people, building an amazing, amazing thing that could change the world. And I think that's why this book to me is so meaningful that, you know, I get to work with people that are not normal, exceptional. Love that. I don't think we've had anyone on this show mention these books. So you are truly, you are truly an original, Brian, truly original.

1:01:01So excited to add him to the newsletter and everything. but Brian, this has been such a pleasure thank you so much for your time this has been so fun, thank you Mike for the time thanks for having me and there you have it, it was a pleasure having Brian on the podcast, Brian, thanks again so much for coming on if you're enjoying the show I highly recommend also checking out the newsletter at theconsumerbc.com you'll receive all the latest news in the world of emerging consumer and as well as be the first to know when a new episode drops you can also, wherever you're listening to this podcast whether it's on Spotify, Apple, or YouTube.

1:01:35If you can hit a like or subscribe, that'd be amazing. Thank you for listening.