The GenAI 100: The Apps that Stick
47:24 Share

We have almost redefined retention for consumer.

We've been seeing a lot of companies actually get up to tens of millions of dollar of annaly's ed revenue and a very quick manner.

Many of these products are getting floods of users and traffic like we've never seen before.

The willingness to try and willingness to pay has been so high for these products that the velocity to get from nothing to maybe tens of millions of revenue have never been higher.

Consumer y eye has been characterised so far by categories where randomly, ss and hallucinations are a feature.

Human connection is important, but maybe it's not the human part that you just need to feel connected.

We have done a ton of recent coverage around consumer I because quite Frankly, the field is moving so quickly every day can feel like the entire industry of shape shift thing.

So who's really one year today, we bring in a extended consumer partners, kim, and leave me more to discuss our jenni one hundred list and what IT really takes to stay at the top and withstand the AI to our not so what categories are capturing the attention of consumers? Are broad or niche models pulling ahead? Where are these apps actually getting their distribution? Does paid acquisition make sense? And do network affects exist like they did empire cycles? These are all things to think about.

But here's a thing, we're finally at the point of the cycle or we're starting to get that data not just in rankings, but other key consumer benchMarks like tension and perhaps were also unveiling new meta ics for this new wave will cover all that and more. But I did want to note that this episode was recorded before OpenAI spring update. So if you're in good to catch up on that, make sure to check out our episode last week.

All right, let's get started. As a reminder, the content here is for informational purposes only, should not be taken as legal, business tax or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a six gene fund. Please note that a six sense year in zephiel, az may also maintain investments in the companies discussed in this cast.

For more details, including a link to arn investment, please see a sixteen care com slack disclosure. Both of you have a lot of experience in the consumer sphere, not just during this A I wave, but before. But you pulled this thing called the gena I one hundred list. What is this? Listen.

how is a hold? Very good question because there's a lot of ways to pull and look at this data. I think the central question that our team had was there are so much buz around the eyes, there are so many products that are coming out every day, every hour, even what are the things that Normal everyday people are using.

So what are those applications? There are some that everyone knows, like ChatGPT and mid journey. But we were curious if we try to take a more granular view, what are the other names that might be more surprising? And so how we pull IT was we looked at every single website in the world ring to be a similar web, which is a data provider.

We sorted them by monthly visits. And then we pull the top fifty that were A I first companies. We did the same for mobile apps through a provider called sensor tower. So we ranked those by monthly active users and pull the first fifty that were A I.

just to give people a sense how many were pull. And then you had to wall IT down .

to fifty hundred, tens of thousands of the very least, probably to get the first fifty. Maybe we went through a thousand websites in a thousand mobile apps, if not a little bit more.

And so as you pulled these together, what categories are standing out, whether the productivity you're seeing, companionship? And then also you pulled a similar list. And what was its september? Yeah as well. So was there a big change? I mean, that feels like A I was moving everyday.

A lot is change. And I think we feel this as investors. I think founders feel IT even more. We first pulled the data, I think was september twenty, twenty three, and then we pulled IT again january twenty, twenty four. So actually lesson six months gap, about half of the list was the same as the first time and about half of the list was new, which I think reflects both a hug Epace o f c hange, but also that there are some kind of name, some brand, some companies that are cementing themselves as like early leaders and really building a loyal audience.

And I think they're always surprising as v we look at the industry and have these mental models of, oh, we think with the A, I, these set of things will do really well. And then as the A I apps team, I think one of our method is to really look after what consumers actually gravitate to. So often time, we actually see a divergence of what we thought initially verses. Oh, actually these are doing really well, and I think those discover really fun place where we discovered the actual revealed preferences of consumers.

Are other examples of that something you thought would stick around or yeah not?

I think the one thing that we weren't surprised by that week feel and i'm sure anyone else who works in around day I those is that for consumers is like content generation and editing is key, and it's the number one things. So these are like Midjourney, pica, runway making, images, video, things like that from scratch. And I think that's just because it's so magical.

Like everyone at least, I always wanted to be an artist and never had the skill. And so we gable to do that from zero to one in ten seconds is amazing. And that definitely proves out in the data. If you look at the fastest growing and even the most stable companies, a lot of them are still in that kind of category.

Yeah, I think that's very so much how I think about IT where, as Olivia said, I can't believe this works. Era is where we are at yeah. So anytime we have these magical moments of I put in a problem, something happens that we think those will do well and those do well.

I think where I am personally surprised when we look at these apps that are little may be popular for a while and that changes your ata or your profile picture into multiple different versions of IT. Then like all you're gonna go away, but you keep seeing them up enough again in different formats. So I think that speaks who may be a little bit of the underlying his humor, willingness or excitement around themselves, which is always top line for them.

Sounds like they are willing to play. And I think you're totally right that there are so many examples of things where you just can't believe it's so good, so early. And maybe one category where that I think is surprised a ton of people is companionship, right? I think a lot of people were quick to write that off as it's only for this kind of person.

yeah. And I think both of you have probably played around with these products and you've learned quickly that, oh, my gosh, like I really like this too. And this is like very convincing, maybe also for different use cases as well.

So maybe can we speak to that particular category? You mentioned it's going mainstream, I think is quite a statement. What are we seeing in that sphere?

I have a strange example is one of those where people, even me, would look at something like I can't believe you would talk to a fake character is made up for like why would you do that? You know, it's sort of reminds me of iginla. Tian, snap the earlier generation relax. I can't believe you take picture that's going to disappear. What's the point? I can't believe you talk to a fake person was the well, the point is that the new generation are really excited to adopt IT and talk to these beings, if you will, in case in point, I think one of our partner is child has a group chat with a bunch of friends, actual human beings as well as these pots.

if you were.

and you get to just chat. And that's interesting, right? We can actually look at from outside looking in and say, oh, I don't understand the behavior except but the truth is that is happening. The truth is that very engaging and folks are really adopting IT and going even step further. There are no scientific studies done to some extent.

We actually had a founder actually to speak to us as well, which is very cool, where there is a study done that was actually featured in nature, where folks who have the sort of companion digital companion to talk to showed or willing that to hurt themselves. And we can look at that evidence and say, well, that silly, but maybe that is an evidence that human connection is important, but maybe it's not the human part that you just need to feel connected. Yes, that is the reason, not self harm, not engage in destructive behaviors. And if we're seeing that evidence, who are we to judge like this is silly. 对 呀。

the companion products are such a good example of kind of like the theses that you always have to stick to in consumer especially, we are looking to invest in early stage consumer like we are, which is you can get too opinionated about the progress you just have to see, like you are often surprised by what is sticking.

When we looked at this data as well, looking at just users for the mobile apps, we looked at things like engagement and character AI. For example, those users have three hundred plus sessions per month. In many cases, that's the average using profile that's again like social up behavior. That's messaging out behavior yeah it's ten plus sessions per day.

I don't talk to my parents that much. I don't talk to my partner that much. yeah.

Become one of more important conversation tool or a companion that you have in I remember a few months ago where I had this period of time where I was like talking to companion APP very, very diligently everyday, maybe like tens of minutes of time because things that sometimes we wanted talk about our money yeah like maybe not as important you feel that is not as important to talk to your friends or colleagues.

And it's naturally maybe topic for your therapies or what have you. But you know, therapist, not always there. And therapies are expensive.

So this toss is like another example, how technologies is really bringing this abundance. Like my third is kind of expensive. So really dream that costs down nothing. Yeah yeah that's really exciting for many people.

And is that the direction that we're seeing within companies ship and there are companions for therapy, companions for health care are yeah sure.

we think so. It's still early. But I think the fact that the first version, this list, back in september, there was basically one camanchee product on both the web and mobile rankings and now there is a bunch on the list this time means that in some cases, the use case for the brand or the behavior of the audience is almost fragmenting.

Like you won necessarily use the same companion platform for everything there's N, S, F, W, only camana on. There's marketplace of companions where the most compelling character that anyone creates wins. There's therapist companions. Now the pie chatbot was originally built us like a broad based, almost chat G B T thai than IT has since been pulled by a lot of, I think, the lonely adult users into basically being .

that there yeah, sadly.

i've used IT as well. But yeah, I think that we're starting to see companion move outside the realm of this maybe more nh of people into something where we all interact with a companion or may be several companions and might not even think of them as A I companions.

I think that's exactly I think the distinction actually starts to disappear a little IT. And actually, let's just take a example of teachers ah there's a digital twin or a character of a teacher as giving you assignments and giving you corrections and lessons that are very similar to what they're already done. That's sort of a hybrid teacher.

And I think more and more, as the libya was saying, we're seeing these divergence of use cases that are going deeper and deeper into each use cases. So like teachers is one other ones like thera that we talked about, it's easy to repeat back. Oh, we must have been really hard for you to tell me what a lot of, actually, if you look into behavioral science and what IT takes to be a great, there is there are many academic lessons, and like understanding a human psyche that does go into that.

And I think for someone to actually train a really good there, apis, bot or a conversational tool, you actually have to train a slave differently. So there are companies and products they are thinking through. How do I gain the transcript ts of the public or public these conversations that occur between patients and therapies and how we train the companion on that basis to go deeper and deeper? I think we'll see more, more emerge.

Know that's a really good point because I actually even prior consumer companies in a way or companions, you take something like to a lingo, you're talking about a teacher. And you have add and maybe empathetic .

element to IT or sarcastic .

in the cause that depends on the company, right, and the user. But it's an interesting reframing because I think a lot of people think of companions is just as like friend or at last F W, as you're saying. But I can mute so many other things, maybe just to round the corner on companionship.

Because we are seeing these more niche targeted use cases. What does that tell us more generally about the way that these applications are being built? You mention at the beginning working the tragedies and major nees come out strong.

At first, you can still have a lot of engagement. But does that tell us anything about things cornering off? Yeah.

it's something we think about a lot and watch really closely. I think ChatGPT is a great example because they had, of course, like the fast is ever product to get to a hundred million monthly active users, but a lot of the usage has flattened out. And I think that doesn't mean that it's not a great product or that the model that powers that is not an amazing model.

IT just means that because these models are now available for other people to build on, we're getting more kind of specific and purpose bill applications that worked Better for certain news cases. So not always kind of the blank page blinking cursor is not always the right interface for everything. And that could be a therapy bot IT could be a language learning by IT could be a design can IT could be a lot of other thing. So the fragmentation is happening and it's really exciting.

And not to blow up the conversation a little IT, but if you think about the ChatGPT and what powers IT, it's sort of the OpenAI is large language model underlying right and that sort of a close model that is built for OpenAI and customers of OpenAI. I think overseeing is and this is very exciting for our space of application layer.

We're the underlying models are getting Better and Better, and even the open source ones are sometimes even Better than the close source ones. So very recently, the lower three that came out incredibly efficient, incredibly advance, same with a miser's new model. So they think are seeing now is people are using those tools to build upon the application other products that can be very purpose built.

And if you think about companion is like a very large word, that just means another being talk to ah so then you drill down further and what are you talking to them about? right? Teaching, of course, language learning, of course, tutor mentor therapies.

All these are different mode of interacting with someone. And I think in so fast, there's any sort of specific fine tuning or specific data source you can learn specific interaction models from. I think all of that benefits and drives a case for more niche a word, but more specialized use cases.

I E get back to your earlier question of which categories are reading, maybe the most growth, the products, the fastest adoption. And there are absolutely categories where the delta between the best closer models and the open source models or the API available models is pretty narrow, like the best text models are actually.

Now maybe this is controversial, but some people might say the best opens source text models are close to like GPT three point five. Same with image models. If you look at video or music, those are models where still the best open source is maybe not as close to the best things that run away or someone else is developed in a different way.

right?

Much newer. So we think it's gonna en, but that is affected maybe th Epace o f p roduct w ithout i n t hese d ifferent c ategories.

And you know something you brought up with around you, right? And have the kind of google box may not be the box that we expect for the future and also specific use cases. And I think something really fascinating from the different one hundred is you called out a few different categories where we are seeing like these different modalities. So with music, you called out a bunch are showing up on discord. Yes, maybe some others are showing up as chrome extension, like maybe talk about that and where we're seeing the divergence from the model that we all expected from the guy.

Yeah, it's so funny. To be good at our jobs, good to consumer investors. Now we have to be tracking data everywhere because really interesting products are being built everywhere. I think this court has been an amazing one, especially for content generation. Midjourney, pia, sono.

All these companies started on discord both because it's pretty easy to spin up the product without having to build the front end, it's pretty easy to monetize and start making money and because a lot of these products arrive on the community of people who are trading prompts, are seeing each other's output, and you can do that really easily on discord. On the whole other end of the spectrum, all these productivity companies, many of them are more about how to make you, as an individual, doing your individual work faster or more efficient. And so many of these, the key is to live where you are doing your solo hard core work, which is often on a browser, a desktop. And so they're starting as chrome extensions or voice recorders on desktop or screen recorders on desktop.

things like that pot on that's exactly what we're seeing, where it's the tools appearing close to what needs to get done, right? So discord, obviously, if you want to just make stuff, is an incredible place to do that into your point. I think the discovery is really unique where you get to see what people actually generated.

fine, genuine .

so far, people created before you. And that gives you that sense of joy to go create whatever you want. And I think what's interesting is, again, we talked to a web, there is a chromium tension.

There is apps just thinking about one of the products like captions. Get started on APP because you take videos s more and more on your phone these days, and therefore IT was natural to have something that live on your phone. And then now as he starts to think about all maybe moving into work flow in a little bit more professional use cases that immigrants to web because that's where a lot of works done, too. So I think what you're starting to see is eventually the companies and great founders are chasing the youth cases.

And where IT occurs, I think when majority came out as an example, all myself as one of these people who almost to some degree wrote IT off because I was on disco, relatives said, is not having its own platform and it's so fast. Meaning to see in a way that kind of turned out to be the opposite, as you're saying, like when you're close to the users or consumers that you're trying to reach.

IT was not only Better in that way, but also like you think, I was so fun to see the generations in be part of that community, which is something that I certainly did not expect. Let's round out the categories here just by asking. We talked about a few that maybe people would not have been surprised to see on the list where there anything you felt like we're missing, like you really wish you saw more of a presence.

I think in general, consumer eye has been characterised so far by like categories where randomness and hlubis ation are a feature, which would be honestly, a lot of the content generation and editing stuff, a lot of the companion stuff, avatar products where you can get one hundred photos of yourself and as long as they are good, you're happy. yes. And those are the ones where we've seen the most grow so far.

And then the other categories where hallus, Cindy and randomness ess is more of a bug. So that might be personal finance, wellness attack, things like that. And the models now are getting more precise and accurate, but also founders are able to Better build the product that kind of bounds the output in a way that even if there are fluctuations that can kind of cross check IT can contain them.

But I think that's why we've seen those categories be a little bit slower. If you look at like top consumer subscription products, P A I, which were tons of attack, personal finance, healthy wellness, that hasn't quite translated to A I D. But we think IT probably gets there in the next year or so.

A lot of the current products are seeing on the consumer side, utility is the wrong word, but IT serves something there is like single use case. Whether you're creating, editing, having fun talking with something, there is like a single use case as very useful.

I think what i'm also really excited to see is when you think about the fundamentals of the business, like where does network affect the curve and there be a marketplace, what are some of the natural cream motes? I think my wishes to see more companies with those elements. And and you think because it's so magical, we live in this very, very interesting time.

We're sort of in that era of, my god, if IT works as worth IT unless pay to use IT and less go. And I think more and more, as we see the space of ve, i'm also very excited to see what we had not seen in town yet, which are ones that are really benefiting from the underlying network effect that naturally occurs, underlying markets dynamic that could happen between supply and demand. And yeah, I think those are the ones that I think will also be under to look out for.

Yeah, I mean, let's talk to that specifically. Olivia, you have widely cited this term AI tourist phenomenon. I don't think this is a surprise to anyone.

I mean, we've all tried on so many of these tools is so exciting. And then we also have left many of gan. And you can even look to the list, right? You said around fifty percent went from september to january.

That could be a glass half full to come and stuck around or glass have empty or look, fifty of them. We're here and now people are not as interested. So what is this data telling us in terms of stickiness? And is this really still a thing with the air tours phenomena or removing part that .

we talk to our our founders a lot about this. I'm not saying it's easy, but it's easier than IT has been before maybe to get users and for consumer application. And that just because there is so much excitement, these products are so cool.

There's demos going viral on twitter, on read, on tiktok. There's newsletters, discord groups. And because of that, many of these products are getting floods of users and traffic like we've never seen before.

And those users might try out once or twice, but they might not actually be in the core persona of who's a good fit for our product. And so they might not convert to pay or they might not retain and come back to the product the next day. You might say if it's free to get the users IT might not matter. The problem there is a lot of these AI products .

are actually expensive. The same world, right?

And so sometimes we see founders get in a place where they call us and they were like, oh my god, or out forty k overnight because IT went viral and like india or something. And we got a million users and they all used up like our maximum free trial and none of them were paying. And so that's something to look out for. I will say we have almost redefined retention for consumer IT used to be free user base like anyone who downloads.

anyone who .

engages exactly yeah. And now the bar to count as a active user is just higher for us, and we measure attention only off of that. Usually that's a paid user.

Maybe if they're not monetizing, it's have they completed x actions? If you look at IT that way, they are attention for AI products is actually as good as, or in some cases Better than non AI products just because these companies are amazing. But if you measure IT on the tourist alone, the picture can look a little tougher.

Yeah, I think that our learning right, like the AI tourist phenomenon, I think we almost put a number to IT to some of IT, extends you overall top traffic a funded by near .

forty percent yeah add a layer.

add another layer. So I think what olive is suggesting in what we're doing is actually thinking through what is an actual user, have they completed the behavior that counts you as a modified user because of willing us to try something is so high, never been so high.

So I think defining that and starting from the right touch point, and if we count backwards to actually think about retention, because we all think for a product to survive and do well over long term, people just need to come back that sort of the key to IT. So I think what we're seeing is a very high number of companies are able to translate this top of the final into paying user at a very healthy clip. And what's more is that we talked about willingness to try is very high.

Willingness to pay has also been incredibly high because the products are magical and because there are are actual use cases, not just personally, but also commercially. Yeah, the willingness to pay has been quite high. And as a result, we've been seeing a lot of companies actually get up to tens of millions of dollar of annaly zed revenue in a very quick matter and that that's been actually really interesting.

The this is the wrong word, but justification when we're asked, why are you only focused on A I products? What about the non A I products? We are not saying non AI products are not interesting. They're very interesting. But we are seeing is the willingness to try and willingness to pay has been so high for these products that velocity to get from nothing to maybe tens of million of revenue have never been higher.

And that's very compelling. Let's get to how we keep those users around. But both of you spoke to a few metrics s there ah and I know we're far enough to consumer that there are several benchmark best practices that you look for a meat.

Both of you sit in so many deal meetings and someone comes in and lets say five years ago, there was very clear they're looking at daily active users. We're looking at day seven, day thirty retention. There's things that you know automatically like this in your bones.

You know what's good and what's not good. You can see a chart and you know if this companies is doing well or not, has that changed? Are you still looking at the same metrics? Or how do you interpret or add new matrix in this new era?

We're looking at some of the same metric, but may be in a slightly different way for the more kind of work oriented, presumed productivity S N B tools. We look at a lot of things like the wall now, ratio now, which is this truly something you're using every week for work? Or is this something you maybe in once a month for two hours, which can still be interesting .

but is probably a little .

bit less compelling? Yes, we'll also look at conversion to paid for those. And then like we mentioned before, we will do standard monthly retention cohorts.

So that would be of all the users who signed up and paid in months S R O, how many are still paying in month one? And how many are still using IT? How many are still pain month two? But in pra I consumer, the domain atr, there was like all free users and now we only measure IT, mostly unpaid or really active users just because of that tourist effect.

Yeah I think P A, I is out of the domination is slightly different and therefore we will count daily because that's actually would you have to clear the bar for free user base in all that? I think we are seeing a reason we are moving to like weekly and monthly is because as IT becomes a little bit more presuming that becomes a little bit more commercially relevant, it's not obvious that you will want to use these souls every single day yeah.

And so naturally, we're now expanding into a thing through oh like weekly use this RAID retention rate, monthly paid attention. And I think what's really unique here is that if a tool is very useful, it's not guarantee that you will even use that every single. So now we're thinking through, okay, you're still pain. That's good. How many outputs are you creating in a months because is possible you sit there for eight hours straight and crank out hundreds of us that you need from the product and is incredibly useful to you, but you only showed up one day out of a thirty one day year, year.

Is that good? And it's kind of a blessing in a course for AI codes because it's like if you are helping people do their jobs or make their art or something much Better and faster, they are going to use you less because you've made them so much more.

So it's almost measuring like value base, like how much value you deliver to the users for a video editor that might be number of downloads, but maybe because of A I, you can now plug in all your videos for the month and do IT in one week instead of having to come back in every other day. P, A, I, and generate again and again and add IT again and again. That's an .

interesting point. Just I think about something like chat B T. If it's twenty dollars a month, thirty dollars a months like one really good engagement can be worth that speaks to the value of these tools where these don't mico interactions where you're like going. I get thirty cents worth of twitter here. If this can save me.

if this can really help me do my job Better. Even so, twenty hours. yeah. And that's incredible.

Valuable one engagement. yeah.

So you have to look at do they keep paying, not just how much time are they spending in the product because in many cases now, especially for the productivity products, it's the faster you can get in a note of IT, the Better as long as you're still getting to the result that you wanted.

And one thing I would just this is a plug for our firm. We see a lot. We meet these companies a lot. And I think what's helpful for us is that we try to define and understand what metric we want to track, what is important to us.

And then the other thing is that we have the delay and rigor to continuously ask for that and therefore, build out a strong mental model with an actual count of the companies that matter to that category. And therefore, when we actually see an exceptional product, we immediately recognize that without having to really ramble, if that make sense. So the number of companies that we meet and how we to find the metric readouts ly and tracking them carefully give us the ability to recognize what might be an exceptional being very quickly.

even if the metric are .

evolving the time different category actly like an image generator. Maybe we look at weekly bounded retention, but a companion product, maybe we are looking at daily over monthly active user. So it's a little bit different for every company type, but we do try to have pretty closely measure and collect data point across hundreds of relevant company.

Yeah because I guess what you're saying is each founder only sees their data. I have no idea if be able .

to tell you that about four tile calls this tile. And ultimately, what we're trying to measure is, is a product delivering what is meant to deliver and what is a metric that best capture that moment.

Now brand, you have said before for consumer products, the reber hitt, the road with retention, and we've touched on this already. But what can we learn from the prior era in terms of attention? Because I feel like truly, we've done so many AI podcast. This is the question that comes up continuously.

Where are the motes, right? If it's so easy to build today, especially as these open source models are getting Better, how do I stand out? How do I keep my users? So what can we learn from the past? And does IT still apply here.

This is a fun one because I have said retention is very hard to game. And IT is true, is always been hard to game, harder to game, then grows. And overall user bases, have you I think what we've learned from the historical or pra I consumer companies is that there is a specific segment of very poor looking founders who have learn a way to actually improve attention somewhat artificially.

But you can do that. And i'm not saying there there are bad thing because if IT the need and helps the company delivered the core product of value faster, Better, often, and that's great. And I think what are seeing is there are tried and true or tested call IT sixty eight different type of methods you can employ to improve attention.

And I think we're seeing actually some of the G A I companies are a native companies that are employing some of these methodologies to actually improve their attention and therefore, be able to keep their users longer and being able to deliver new products to them again and again. So I think what we're learning is a couple things. One is that retention still matters.

If you you are going coming back, that's not and the frequency can be a little different because especially around workflows, you can come in a little less frequently but get a tone of value out of IT, that's there. The second is that what we thought to be largely ungainly is somewhat movable. And there are some method you can learn from the ni I company is supplied your product to actually achieve that.

And three, I think ultimately our attention is an output and I output of what does your product do? And is that actually really useful to people? And did you deliver that quickly and often? And that's really the cracks of IT?

I completely agree. I think the other thing were seeing with retention that was also true. P A. I, but maybe even. More dramatically true is like the narrowed and more focus the product, the Better because we do have, in many cases, companies with a ton of computer like chat, B, T, microsoft, google themselves notion.

All of these big companies are building and releasing more broad based AI products applications, and it's hard to compete as a startup with a broad base product. If you have one one thousand of the computer and the team and the engineers and all of that and you maybe don't own the I T. Authentication, the data like the years and years and years of history.

And so I think what we're seeing work really well in terms of attention. Like sometimes we meet a company that's a very horizon onal product and the retention is just okay. And then they come back to us five months later and they're like, actually, we realized that we're building for this core set of users, and we honed in on this specific model and built ten more features just for them.

And our retention is four times Better than I was before. That kind of thing is working really well. It's like counterintuitive. No one wants to build a product that's too narrow, but it's Better to go narrow, have amazing retention and then expand than to try to do IT the .

other way around. It's quest enders who have this amazing technology at hand, asking themselves, what is this good for? Yeah, who is that good for? And often times you find an answer in surprising places.

If you told me initially, hey, you can actually clone e yourself with the avatar and present yourself. My first guess of that won't be, oh, this gona be amazing for learning and development within comments. My first guess isn't that, guess what? Sales people can send a advertized version of himself into a sales process like that is not my first instinct.

me.

And the fact that the founders are able to hold in on those who have customers were willing to pay to live a point is very unique and important because you found an interesting niche of narrow use case where people are finding so much value. Often times those are the great places you can corner and start expanding the market. And founders who have that mindset of i'm going to find what we call I C P. Initial customer profile and really sort of appealed to them. The way to grow more horizontal step by step has been an interesting model, especially in the competition with local companies.

One thing that's interesting is you're not just saying that it's the model itself or fine tuning the model. Is everything also built on top, like the U X, the marketing, the messaging, all of that comes together to be just a little bit or sometimes a lot of bit get Better than than more generalized.

And and I think this is a feature, right? The things are so magical, they're evolving so quickly. So then the question is, well, how do you differentiate? And the differentiation may sometimes the or take is so good that it'll blow everyone off the water.

And sometimes that's true, but a lot of times, the world is so large and a lot of great people working on great problems and a lot of smart people working in the same problem. And so what we end up seeing is the velocity of product shipping always matter, especially when things they are changing so quickly as we saw top AI properties like ranking changing so much, that just means there is a lot of velocity. So you need to stay ahead of that.

Velocity matters. But too was really important is how do you actually build consistency. And like we're tension, that's by building into what's useful to the users. And that's why I think what we're seeing is I will deliver a part of the workflow to make your life so much easier. And that's what we're seeing like some differentiation and companies.

It's so easy in consumer, especially when we're meeting like incredible teams every day to get a hammered by. This is the elegant technical approach. This is the best research team, and in some cases, that is what wins.

But if I take off my investor hat and I think about myself is like a Normal person downloading an upper going to a website. I do not care about the technical I would. I don't even care about who made IT.

I care about if IT helps me get the thing done and I wanted to get done, which I think goes back to like it's often little workload. Yes, it's like tiny features or how you scope the products that make the difference. And that sometimes doesn't come down to the technical details around IT, but comes down to these micro product decisions that can be make or break.

It's funny that have to remind ourselves. I because no one ever cared if an APP was made with angular react, right? No one ever sees that. And of course, I feel like i'm going to a .

it's too cool.

Yeah ah yeah but maybe one other question. We're talking about competition. At least in the last consumer wave, we did see some companies front run and maybe they did IT through like raising a lot of money and then doing a lot of paid acquisition and then you start to see things like network effects kick in.

So are we seeing that similar dynamic words paid come in to play? Because both of you have mention that people are willing to pay. So does that mean budgets can increase or our cpa increase?

IT is a really interesting question. I think there are categories where raising a lot of money to build the best model actually does make a really big difference and having a best in class product. We are investors in eleven labs, which is a texas speech company, and they have an amazing model.

And because of that, they're used by probably thousands of developer customers and other customers to power the product. And in that case, it's like harder to compete. If you're not raising a lot of money and if you're not actually kind of more money means more data, more tuning of the model, a Better model and IT becomes a spiral.

But there are other product categories. For example, many products are building off of open source models. And then it's much more about kind of the products elegance, how you commercialize the model, how you take someone else's tack and translate IT to something that artists or designers or other people can use and therefore trying to like front run to raise the most money and acquire the most users isn't always the winning strategy.

Maybe another way to think about IT is the models and the research that can are so mad that is actually often times go directly to consumers, and that's very exciting for them because they immediately get the benefit of the cutting edge research. What I think that means in terms of that, you talked about paid acquisition and how that translates into this new A I sort of world. I think they're two classes.

And maybe this is how I think about IT. There are classes of companies where they benefit greatly from the buzz, and these are reality of the product example that they can put out in twitter or read IT or discord because it's just so fun. It's very high catching and a very attention grabbing.

Sure, there is a tourist entries on, but they benefit from a great top of fund traffic. And in so far as that traffic continue to come in, you're able to find ways to convert them into paid users and actually start making great money and start building out even weight list or inbound been based sales lead. If you're thinking about your product, a workflow tool, and it's so useful that some sb or enterprise customer may reach out to you, you can start building on a pretty good inbound list to go after and went off and start building a great size business.

And in those cases, cack or paid matters a little less. There are other businesses where the product is very good and is very useful, but IT hasn't fully benefited ted from the halo of the amazing glitter of A I apps, if you will. And in those cases, because the willingness to pay is quite high, we do see our crop of customers, our products that actually end of engaging and paid acquisition in a dowdle matter because L, T, V is there, they're able to afford actually paying to acquire users.

And some companies know how to do that Better than others. And that has been a model that we are seeing. An often times, these companies can build their run rate up to tens of millions of dollars, if not more.

Totally great. yeah. I think if you're building a product for A I artist or designers or writers, very easy to go viral and bring in a bunch of users on youtube or twitter tiktok. If you're building A I platform for like small h fact businesses to be later expended into other like home services businesses like going viral on twitter, mayor may not actually help you. You're probably going to still have to build out some more of a traditional kind of lejean sales fono, at least like a strong with overall program.

things like that. And to your point, if you mention do you like actually brute force built ork effect, I don't think we're seeing that just yet because there aren't true business model that we're seeing that truly benefit from either network of factor marketplace dynamics such that people want to buy their way into that density. I think what we are seeing is that there is a true pay off or payback that's related to the paid acquisition, and they may calculate return base assumption to go acquire users. Yeah.

I like that distinction. But do you expect that to change? You expect in a few years for that to be true, where there will be those marketplace effects?

I do. I think so I I think IT comes to actually the question of what is matured already and what is to come. An example we talk about a lot is that we haven't seen a lot of truly AI native social apps, for example. And I think part of IT is because there were some early tests of this and a feet of content that you know all of IT is AI is actually maybe not the most compelling social APP. And IT doesn't have the same psychological dynamic, are .

learning a lot about ourselves .

like he doesn't you know, it's a fake picture of you like you don't have the same maybe kind of got like either anxiety relation and posting. And so I think we're just starting to see like there was a product called air chat that spent viral the last few weeks, and that is around helping human beings create human content easier using A I.

So in this case, you can do a voice memo and i'll transcribe IT into text, and then you can scroll through a read affective text. And it's basically opening up people who would never tweet because they don't want to sit down and right, or they're not good at IT or they stresses them out, which I totally understand, who might do IT with a voice memo that gets transcribed into tax. That's just one early example, but I feel I will see more of those.

Definitely think we're starting to see the. Very early in oration of this, like Olivia, you mentioned eleven lives, which is a portfolio company. You actually started seeing these products start build a marketplace model within their company ies.

So even loves actually actively build a marketplace for voice actors to license their voices so that they can actually make passive income. Same thing with captions. They have a creator marketplace. Yers are starting to build that out where critter can license their lights such that video ad producers can use their lightness, and they get to sit there and make passive income.

So these marketplace models are interesting where if you have a lot more supply coming in because you're actually why either paid acquisition or not building that density, I think that, that starts to actually accumulate in terms of benefit via v like any competitors. And I think we're starting to see some concentration of these type of behaviors. Marketplace, anne, mics of fuel. That's actually .

really interesting because basically, in a way you're saying there passively building up one side of the marketplace, but they are not basically starting as a marketplace as you might have seen in the past. Yeah, okay, fascinating. We started this off.

I talking about the gene one hundred. Let's end there too. Let's say we run this six months time. What do you expect to see and also maybe what do you want to see? Yeah, looking forward.

I would love to see new categories starts to mature. I think we talked about two completely new categories in the most recently where things like music, like sono popped up from nowhere. And even since we publish list, ud o has now top up and gone totally viral.

And so I think as more kind of models mature, will see new categories. Productivity was another one that appeared almost nowhere on the first list and of nowhere have quite a few companies represented on the current less. It's a little bit tough to predict. And I think if we knew with the next big I hit would be.

I don't know.

we go to find someone to go you. But I guess what I hope to see is a continued testing of the boundary is an expansion in both form factor modality categories. I think the amazing things that we see six months for hour, things that we probably can even conceptualize .

right now, maybe to see the new categories, I think I expect to see another forty percent of the list, yes, being that new, if not more. What I hope to see is actually the prior lists have these single modality thai products for its largely text, as largely audio is largely music.

What I love to see is what happens when you start combining this? What happens if a video plus image and plus sound effect? What is that? A music video? That's cool.

What does that look like when you have avatar AR plus a voice? What is that product? And we love to see these net new categories where we will have a hard time defining what they are because they combine these different things.

And to use olive as the example of air chat, that's really interesting, right? Like voice transcribed into. And now we also know that text input can do pretty much anything, create music, video, avatar, three images, anything s so then what does that mean when all modalities are essentially interchangeable?

Yeah don't know. And we're excited to are very excited about yeah, it's amazing. Well, this has been so interesting.

We will have to sit down again, whether it's in six months or whenever you guys have a new list and see what has changed because IT does feel like other lists that have been done in the past, you have to wait another year or couple years for enough movement to happen. And I feel like we onest ly could record this in a month, and we have like that. Something is probably .

that have cropped up lower record. Yes.

amazing. what?

Thank you. Thank you.

If you like this episode, if you meet IT this far, help us pray the show share with a friend. Or if you you're feeling really ambitious, you can leave us a review at greatest podcast dot com. Slash is excised.

You know, candidly, producing a podcast can sometimes feel like you're just talking to avoid. And so if you did like this episode, if you like any of our episodes, please let us know. I'll see you next time.