Ep 95 - AI Tools, Trends & Future – a16z’s Justine & Olivia Moore Break It Down

Very few people on Twitter who I have notifications on, and you are 2 of those people where I just obsessively follow everything you tweet and post about.

It's funny because I remember this time last January I was like, wow, our AI photos are amazing. And then I look back at the same prompt now and I'm like, that looks like abstract art.

I think there's like no better time to go build a company if you're a founder or you want to be a founder.

So many first-time founders that are just opinionated on something they want to see exist.

that maybe was like never possible to build before and that we can have now.

Being a Stanford dropout is almost better than being a Stanford graduate.

Using AI for therapy. We'll ask, like get support from it, get insights, and then ask questions like, what should I know about myself? People love that and people find it super valuable. But there's also a ton of hate comments about like, you know, these aren't real therapists.

What do you think is going to happen this year with respect to big companies? Hey everyone, welcome to the Aarthi and Sriram Show. Today we have just me interviewing 2 amazing people. I've always wanted to talk to both of them. They're Olivia Moore and Justine Moore, famously known as the Venture Twins. They're both partners at Andreessen Horowitz who focus entirely on AI investments. And we wanted to kick this off as a special episode for the year. It's our first episode of the year to be able to talk about all things AI. What are people building? How has last year been? What are predictions for 2025? Favorite applications, AI applications that both Olivia and Justine use. And so much on building AI companies, scaling AI companies, enterprise, consumer use cases. We just get really deep in it. There's a little bit of something for everyone here. And I thought it was a lot of fun and I hope you enjoy it too. With that, Olivia and Justine Moore on our show. Justine, Olivia, welcome to the show. This is our first ever episode recording in 2025, and this time around it's a solo show in like truly literal self. It's just Arthi instead of the Arthi and Sriram show. Sriram now has a real job, so he's off to the White House to go figure that out. And in the meantime, we thought we'd just kick things off. with all things AI. And we thought, and you are the 2 people who, I have very few people on Twitter who I have notifications on, and you are 2 of those people where I just obsessively follow everything you tweet and post about. And when you did the year-end roundup, I was like, wait, we have to get them on the show and talk about all things AI, predictions, past, all of that. So with that, welcome, welcome to the Aarthi and Sriram Show.

Thanks. We're super excited to be here. It's big expectations if you have notifications on for our tweets, but we will try to live up to it.

I see everything. I see all the tweets. It's great. It's great fun. Even the ones you delete, I see them.

I was gonna say, yeah, on occasion.

Yeah.

And, you know, I wanted to, you know, last year this time, right, January 2024, when we looked at like state of AI and where things were, I thought things were pretty wild then. And, you know, at that time it was like, you know, key LLM players, you saw the OpenAI, Anthropic of the world dominating. You started to sort of see this open source model boom at that time. You started to see some sort of like AI-powered extension apps like Notion AI, things like that. Not really much in the sense of like consumer proliferation as such. There was just, you would see like Suno and Udeo and all these apps just starting to sort of, you know, come in into the scene. Low-code, no-code tools, again, like starting to be like, oh, we are interesting too, take a look at us kind of thing. And, you know, in the 12 months, and at that time I remember January, hallucinations was a big problem. Everybody was talking about like biases, hallucinations, regulations was a big deal. We also talked about training and how expensive it was going to be to train models. And would this even be scalable? And then data, it was like endless supply of data. No one ever talked about like us running out of data, right? Like that was sort of where the thinking was. What do you think has happened in the last 12 months, starting from say, you know, 2024 Jan to 2025 Jan, where we are now?

Totally.

Yeah.

I mean, oh, just go ahead.

Maybe I'll start on the model side and then we can talk about the app, the app side. I think on the model side, we've seen more really high-quality players emerge in basically every modality. Like, I think for a long time it seemed like, you know, OpenAI and the GPT series of models were going to be like the winner takes all. And we started to see a lot of people using Anthropic for like fun viral consumer stuff, using things like Cursor for coding, like, uh, and that's just in the LLM space. And we started to see like Flux in the image model space, which was a huge step up from Stable Diffusion. Um, we started to see a bunch more in, in voice, obviously, which Olivia can talk to, a bunch in 3D with companies like WorldLabs, uh, and then a ton in video. I think video may have been the space that progressed the most last year.

Yeah.

Which has been super, super exciting. So obviously, like, huge developments there in terms of model capabilities, quality, consistency that has then unlocked all of these apps you can build on top of the models now that they're, like, good and reliable and consistent enough. And then I'll let Olivia talk about the apps, but I think sort of what you alluded to also on the model side is, like, um, once we have those sort of giant leap forwards on capabilities, everyone's then looking to train bigger and bigger models and look at what's next and what's going to be the newest unlock. In some spaces like LLMs, we are now running out of data and looking at reasoning and architectural improvements. But I think in other spaces like video and 3D and music, there's still a ton of room to run on the core data side.

Wow.

Yeah. Some of my favorite posts are the ones that give the same prompt to Midjourney or Runway and run it every 3 months or something. And it's funny because I remember this time last January, I was like, wow, you know, yeah, our AI photos are amazing. And then I look back at the same prompt now and I'm like, that looks like abstract art compared to like what we have today. It's just so crazy. I do think on that note, like the first almost year of generative AI, almost all of what we were seeing was creative tools. So like image generation, music generation, video generation, maybe to a lesser extent.

Yeah.

Partially because those benefited actually from hallucination in some ways.

Yeah.

Or at least it wasn't like life or death if it outputs something crazy. But to me, one of the biggest changes on the app side over the last year has been like, to Justine's point, the models have gotten more diverse, more controllable, more predictable. New modalities have opened up that aren't just like image. And so we're seeing like a big expansion of the types of products that normal people are using every day. So at the end of the year, we went and talked to a bunch of like AI CEOs and investors and experts and asked them to nominate their favorite products. And there was a lot of creative products, but there was also like AI assistants, note-takers, slide deck builders, email clients, you know, code editors, journals, lots of things that we just—

It was amazing. It was amazing. I saw the whole list and, you know, for me it was like, you know, granola, everybody talks about granola now. It's like become such a thing. But then it was like Lindy, HeyGen, Delphi, Opus Clip. There were all these like captions. And then you had like this whole set of like no-code, low-code tools, Cursor, Replit. But even since the time you posted that, there've been like a few more tools in that space.

Yeah.

And so it's become really hard. And then from specific application standpoint, like healthcare was like Ada Health. And there were a bunch of these, like news was Particle News. There were, and there were some of these apps that I even, I hadn't even like, you know, heard about or downloaded. And I was like, wait, there's, and there are these experts all like talking about how they can't live without it or they use it all the time. There's so much that's been happening in the last year on just people building on top of the models itself.

Totally.

Totally.

Yeah.

And I think, yeah.

It happened so fast. And I think a lot of the, the founders to date have been more like really technical folks, like researchers who are great at building, training a model, but their strength is not in building like the most engaging or viral consumer product or distributing it. And so I think we're now in this really interesting era where hopefully like the best models plus product layer plus like workflow can actually be packaged into, into something that people will love to spread the word about. And so.

Yeah.

Hopefully next year you will have heard of like the 10 or 50 or 100 best consumer AI apps.

I think, I mean, I think I look at it as like me as a use case, right? Like I write code, I'm shit at frontend. I'm so bad at it. And every time, like for me, frontend is like drag and drop a bunch of buttons and be like, press this and you will get like the backend infra will be good and tight, but the frontend is just horrible. And then I started using, which one was the first one? Bolt.new. And Bold was like, it changed my life. I was like, wait, this looks not just presentable, but it actually looks like something I can ship. And that for me was like, I no longer need to go bang my head against the wall on like making presentable UI. And this was in 20 minutes, right? Like a full, full project that I could just like get to work, then download the code, port it into my IDE, and then start working on it. Was game-changing. I could build full front-end, back-end, like full-stack applications in about an hour, 2 hours, just as like an MVP to get it going. And even in the past for me, running product, it would be like Figma on the design side, then send the Figma over to your designer who will probably laugh at it and then come back to you with like a better designed version. And then you send that to your engineers and go back. Like now all that is gone. You can just like prompt engineer your way into writing, like building a frontend, which for somebody who doesn't build frontend is just game-changing. So I'm excited to see what people will build with all these, just given how easy it has become to go build applications as such.

Totally. Yeah, we've, we've actually been spending a bunch of time in that space. We're mapping it out right now, like Bolt, Lovable, the Replit AI agent, v0. There's a whole long tail of new ones that have emerged too. And it's really cool. And neither of us have technical backgrounds and we can't code like at all. And so I think for people like us, there emerged sort of like landing page or really basic website builders like a Squarespace or a Wix. But there was never a way for us to build like an interactive web app.

Exactly.

Exactly. And so I think right now the main use case of these tools has been folks like you who are engineers and who can get some initial product or prototype or something to iterate on. But as they become even more easy to use, especially integrating the backend and more powerful, it's incredible to think about an everyday consumer like Olivia or I being able to ship an app for the first time.

It's amazing. Yeah. I think there are a bunch of things I wanted to cover down the road on just what can be possible with just having everybody be able to write code. You know, in the past we used to be like, learn to write code. Like that used to be this like thing that engineers would like diss at non-engineers. And now it's like everybody can sort of write code. What does that even look like? So it's kind of interesting there. Now, as a side effect of that, how has, say, the funding landscape changed? And, you know, I mean broadly in the sense of like round sizes, founder backgrounds, what do you look at valuations in the last 12-ish months or maybe like a slightly longer timeframe, what's changed with respect to funding AI companies as such?

Yeah, it's a good question. There's obviously so much excitement to fund like AI-first or AI-native products. We as a firm are like very focused on AI across a lot of different verticals.

Yeah.

I think people have maybe observed from the outside that like, oh, you know, these funding rounds are huge. These valuations seem huge. Yeah. But the interesting thing, I think, Justine, we've been in VC now for 8 or 9 years. And the other thing that has kind of massively changed is kind of the ramp of companies are growing faster also than we've ever seen before in terms of revenue, user base. Consumer businesses used to not monetize for like 10 years.

Yeah.

And now a lot of them are like making real money with strong retention from day one. So in some ways, and I say this still very respectfully because I know it's really, really hard, but in some ways now, like talking to investors and fundraising, is a little bit of the easy part. And still building a product that can kind of last, especially in a noisier and noisier landscape, is maybe the harder part. I think we have a bunch of thoughts in terms of how we've evaluated AI companies, which has evolved over the last 18 months.

Okay.

To me, the crux of it is we tend to advise founders or get excited about founders who are thinking about the problem or the use case pretty narrowly. Because there's so many things you can do with AI. And sometimes we see products where you've crammed in every possible feature or gone for every use case, and you don't really kind of nail one thing. And this is true for both consumer and enterprise. But we find that products that do one thing very, very well can both maybe fundraise the most easily and then also have kind of the most successful growth.

Wow. Justine, anything there that you would add to?

Yeah, I think the only thing I'd add is a lot of the huge rounds that, um, Olivia mentioned that I'm sure we've all seen are like, were sort of foundation model or basic like vector database or essentially like the building blocks of AI. Like this was sort of the R&D research phase where you needed to like buy a ton of compute and hopefully get a team of like the top 5 people in the world to train a specific model or build something out.

Yeah.

And now we're entering this era where, um, all, a lot of those models are publicly available, whether it's open source or via an API. And so I think we're probably going to see not every AI company raising a $50, $100 million or whatever seed anymore.

Yeah.

There might be some who have, who want to train a huge foundation model who do that, but there, we're kind of going to see a resurgence of true, like, application layer companies that are using other people's models. And I think the exciting thing is, like Olivia mentioned, like in the pre-AI era, it was becoming really, really hard for businesses to like stand out and acquire customers, especially in consumer.

Yeah.

It just felt like all of the major apps had been built.

Yes.

And it was so tough to convince people to download a new app or sign up for a new thing.

Yeah.

And I think this like renewed consumer excitement means there's a lot more fast viral growth and people willing to pay for stuff than we've seen before, which means, I mean, the hope is that that makes these businesses even more sort of capital efficient moving forward.

I think, yeah, I think the point about how the models are increasingly becoming really sophisticated, they can do so much now. I mean, even every couple of months you're seeing an update. across, you know, OpenAI or Llama or Gemini. You look at like every one of these and everyone's throwing these releases and catching up with each other or one-upping each other. And I think what I think is really exciting is consumer companies being built on top of these and companies, just general AI companies being built on top of these where it's no longer, you know, the models are trained, It's all ready to go. Now you get to build all these like really fun use cases. And to Olivia, like what you said, focusing on a few, like one use case, get that right and being able to scale. Has that in the last 18 months, 24 months or so, has it changed in how you think about founder backgrounds itself? And I ask this because when I talk to founders, especially with like pre-seed, seed stage AI companies, Many of them don't have, you know, sort of deep experience in a specific space, but they're really good at like AI tools. They're able to like build stuff, get it out the door, get really fast traction and scale. How do you think about core expertise and, you know, solving for the idea maze as such with respect to the founder background?

Yeah, I think to Justine's point, like when we were kind of solidly in the full infra R&D era of AI, Yeah. Like in almost every case, the model itself was the product. And this is even true of early creative tools like Midjourney and others where you could kind of just slap an interface very lovingly. They did an amazing job, but like you could—

When it was a Discord bot and there wasn't an interface, you know?

That's a great way to say it.

You could literally just kind of put an interface on top of a great model and like people would be super excited and figure out how to use it. Yeah. I think exactly to that point, like as the models have gotten You know, more available, more accessible, less expensive. We're now entering like the product builder era where it's maybe either people who have had deep product expertise building not in the AI world, or we're seeing so many first-time founders that are just opinionated on something they want to see exist that maybe was like never possible to build before and that we can have now.

Yeah, I Totally see that. I think it's— I want to see how this year ends and to be able to see what kind of applications, especially these like unique sort of like scratching their own itch, very niche kind of applications that come out. Those are almost always the ones which sort of surprise you, where it's like, oh wow, I didn't even think that I needed to go— I needed to go try this out or use this and starting to go see there. And then On the other side of things, how do you think about differentiation? And this is like this hot topic, right? Everybody's talking about moats, differentiation. And, you know, the reason for that is you have so many AI companies that sort of do the same thing. And it's become this like sort of, especially if you look at B2B SaaS sort of, you know, enterprise-focused AI companies, it's sort of this land grab. It's like the fastest sort of access to the first set of customers, and then starting to see what retention looks like. And then at that point, at the 11th month, you sort of get to see like, that's like the source of truth on do people continue to go sign and get over to the next year's contract, or if it's an enterprise contract, or like, what does retention really look like? But a lot of these companies now do the same thing. How do you then think about, less about investing, but how do you think about just like looking at these AI companies and evaluating them given that a lot of them feature functionality-wise, product experience-wise, you cannot really differentiate as such?

Yeah, it's a great question. I think it varies between B2B and like B2C/prosumer.

Yeah.

Due to like, you know, the length of the sales cycles and the extent to which you need a network to get introduced to a customer and like, do you need to truly know like the lingo that an accounting or legal firm uses to, to make them feel comfortable. Like, sometimes that sort of network or background-based, um, connections can end up becoming really significant for those companies and, and give them an advantage. We've also honestly though seen for a lot of companies, like, the main differentiator just becomes their product velocity, like how fast they can ship new features, how fast they can improve their product, like Do they have a reputation for just all like, you want to be on this company because as AI progresses, they're always at the forefront of things. And I think we've seen that. I mean, we've now seen that play out across a lot of the modalities. Like when, when we invested in ElevenLabs in the text-to-speech space, I think they're now doing amazing, have a ton of huge customers, are, are doing really well in that space. But they were like the number 3 player or something at the time that we did the Series A.

Yeah.

But, um, when— and, and for a long time, text-to-speech had been considered like it's way too crowded, it's undifferentiated, like nobody has a particular advantage.

And what's the use case really? How are people going to pay for this thing? Is consumer really—

Exactly.

Yeah.

Are the margins going to go to zero? Not clear at all. Um, and I think just, you know, as you spend time with the, the team and in the product and then talking to customers, it's crazy. Even when you do customer calls with customers at these like old-fashioned media or entertainment or whatever companies, they have a sense of like, this product just keeps improving or not.

Yeah.

Like these people are always at the, like the, like this company is always at the front of things or not.

Yeah.

Um, and so I think that honestly has become one of, one of the major differentiating factors in our mind.

I think especially for consumer too, like you mentioned Granola, which is an amazing example.

Yeah.

There have been dozens of AI note-taking apps and, and often like the differentiating ends up being something that seems very, very small, but ends up being important, like asking for your company URL so it can figure out where you work and give you a slightly better template for your first meeting notes. Or for me, like, I am an obsessive note taker. So the fact that they can combine the notes that I've taken with the AI notes is like something that no one else can do. Or even like just not having a bot join the meeting. Like, yeah, there's Things that feel little that from an actual consumer experience perspective end up not being little.

Yeah.

And then the company builds almost like a moat around, then they have the best products, so they get the most users. So they raise from the best firms, they hire the best people. And then it becomes this like cycle, I guess.

No, totally. I get it. I think, Olivia, like when you mentioned when we talked about Granola, the thing that comes to my mind first is I think 18 months ago or so. I don't know when exactly, but you tweeted once about how you built this little app that will remind you 15 minutes before a meeting.

Yes.

I think it was all Zapier.

Yes.

And you used API for being able to just pull background information. And at that time I was like, wow, how could anybody possibly build on top of this? This is amazing. Like the things that you could do with AI. And now it's like sort of baked into all of these products, right?

Totally.

sort of like the status quo and you just build on top of it and keep going. And then Granola took it like one step further where it's like there's no assistant joining the meeting. It just takes your sound and is able to do it and then like slice and dice things. I love how they, like, it gives me a sense of like, every time you think like this is sort of the bar and this is how high the bar is, it just keeps upping it one by one. And it's just, it's sort of, I love where things are going there with respect to productivity. note-taking, like all these simple things that you didn't think was like possible to innovate in.

Totally. And, and often— It's funny, we— Yeah, sorry, go ahead, Justine.

No, no, I was just gonna say we just got a text from our mom who is not in AI at all, who we introduced to Granola because she's renovating a house and having calls with contractors and couldn't— it was hard for her to take notes while listening. And she's like, OMG, the Granola app is amazing for transcription. and summaries. Thanks so much for the recommendation.

That's awesome.

And like, like that sort of thing, you know, 6, 12, 18 months ago, if she'd asked us what sort of tool to use, we would've been like, oh, well, you have to install this on your Zoom app in the Zoom store. And yeah, and you're not gonna get an exact transcript and it's super buggy. Like the products are just getting better and better.

For ElevenLabs, we, Sriram and I went to their hackathon, right? It was in London. So we showed up this Sunday and it was amazing. I love hackathons, first of all. It's just like one of my favorite things to go do. But 24-hour hackathon, at the end of it, we saw a bunch of the, you know, the people who had contested. I think there were like 19 teams or something. And one of them, they actually did, they replaced Sriram's voice with this AI voice and did basically just like sounded like Sriram. And I was like, wow, like, we basically are— our future at this podcast is so short-lived. I think we are going to have to like replace ourselves with like ElevenLabs like voice agent. And you won't be able to tell the difference because it basically spoke like Sriram. And it was like cracking jokes and everything. And I was like, oh man, okay, this is a bit like scary because I thought this one job, this one creative thing would be the thing that AI would not take over. But I think it'll be one of the first things that it takes over.

Or it's great because he's going to be so busy with the new job that he can still make appearances on the podcast.

Exactly. You can't ever tell who it is. And if it's like really dumb or stupid, he can always blame it on the AI, even though it's actually him. So there are pros to all of this. Enterprise AI. What do you think happens, or what do you think is going to happen this year with respect to big companies? I think Olivia, you talked about Reddit having LLMs now and LLM search. And Reddit is just sort of this outlier example because my belief is that most of these LLMs are trained with basically Reddit data, right? Like, that's sort of like where you're getting a lot of the training data from. But outside of that, think of like a typical large enterprise company.

Yeah.

When you think about small AI startups, what do you think happens here? Is it like chance of better collaboration? Is it these enterprises now have to figure out how do you do more with AI? Are they gonna start building things in-house? What do you think happens here with respect to these big companies? Because it's almost like this freight train that's coming at them, right? And they have to figure out what happens here. So what do you think happens in the next 12 months there?

I mean, I feel like the thing that we hear a lot from enterprise buyers is they've been given these like massive AI budgets, whether it's like an experimentation budget or IT budget, something else. And like, to be candid, until 6 months ago, there weren't a lot of products that were at a stage that were ready for an enterprise to use.

Yeah.

Because it's not only like this has to work, but it like has to work reliably. It has to go through security and compliance checks. Like you have to be able to tune maybe the model or the output for this specific customer. So I think, again, like wave 1 of AI was consumer creative tools. Wave 2 was like, okay, consumer, not just creative tools. And now we've seen kind of a wave 3 of actual enterprise applications.

Yeah.

Very cool.

Yeah.

And some of those are like the fastest growing companies that, that we're seeing because if the product is there, then you can, you know, I've seen cases where companies are already paying millions of dollars a year for one AI product. Yeah, we have this whole thesis—

Often in specific verticals, right? Like, exactly, you know, yes, law, accounting, finance. Like, the, the good thing about selling into big companies is, um, usually every task they do or every division of the company, like, it's one thing. Like, maybe it's, you know, closing out the ledger at the end of the month. When it's a big company, it takes a lot of time and it costs a lot of money and there's a lot of manual work. And so if you can automate you know, maybe not even 100%, but like 80% of the process, like Olivia mentioned, that can be hundreds of thousands of dollars in saving.

Yeah, yeah, yeah. I would think the funny part is when you look at quarterly earnings calls, like a year ago, it was sort of this like, oh, okay, one, like one of the questions would be something about AI. Now most of these, like the first question's like, so what's your AI strategy? And you really, and then the CEO or whoever's actually on the call, has to have a cohesive sort of, oh, here is what we are doing. And it can't, it can no longer be, we use AI for customer support chatbots. Like, it just can't be that. It now has to be like, we've saved money using this. We have now, you know, our developers are now using AI Copilots. Like, it has to be very tangible with respect to how you're talking about it. When we had Aidan from Cohere show up on the podcast, And he was talking about how customer support, it started with customer support and training your own, like, sort of custom LLMs and being able to scale from there for enterprises. But now it is about all these productivity use cases that are super niche within these big enterprises. And every one of them has these big budgets and they all want to sort of state that we are doing AI, like, we've got it, we've got it figured out and we are doing this. Do you think this changes in a positive way the M&A landscape? Do you see acquisitions happening because of this where, you know, you're a startup building for, say, law or finance, you've become so useful to this large company that you sort of want to like fold into the larger organization? Do you see sort of that happening at all? Or is it more like lots of licensing adoption, may the best team company win, but just a lot of like, you know, diversity there, but not really M&A play there?

Yeah, it's a great question. I think we've seen less M&A thus far for a couple reasons. One is that we're still at the relatively early stages where a lot of the companies are growing super fast and can, and like it's a capital-rich environment. So they don't necessarily need to be acquired. They're like, hey, we want to keep going for, for a while. The second obviously is like the, the regulatory perspective around. um, M&A has been less, less friendly. Um, I imagine we will probably see more M&A in the future for sure. Um, but I think honestly most of the founders we talk to are like, you know, we want to build a public— like, if, if we're an AI law company, we want to build a public company serving the top 100 law firms in the world, not like we, we just want to be used and acquired by, by one law firm.

Right.

Even if it's A big one with a billion-dollar acquisition.

Yeah.

So I think, as always, like, we'll see kind of smaller acquihires, smaller acquisitions of things that, especially honestly in like consumer and prosumer, of things that kind of don't prove out as huge standalone independent products, but would be really great sort of technology or distribution, whatever, to add to existing companies.

Yeah, I agree. I feel like most, especially like non-technical enterprises are, I would say we don't worry maybe about new startups having to compete with like what a law firm is gonna build themselves or what an accounting firm is gonna build themselves in terms of AI software. So there's tons of opportunity to build for these companies. And the great thing is you can become so core to them so quickly. You might be, you know, significantly reducing human labor hours or allowing them to staff their employees on much higher value things. So like if you wanna be acquired, it's probably easier to be acquired since you're kind of key to keeping their business going. But yeah, I agree with Justine. I feel like we're seeing a new level of ambition from founders in terms of how far they wanna go in building really, really big businesses.

I see that. It's, and it's also every time I'm like, really? And then they'll come back in like 2, 3 months and there's like a ton more traction.

Yes.

Which, you know, at least like from what I can remember in the previous, any of the waves, like mobile apps wave, That was not a thing. Like you would have these sort of outlier mobile social gaming apps, like things like that. But really like having this sort of consistent, oh yeah, we like quadrupled our run rate in the last like X months. Just, you know, it's not a thing that, you know, we, I'd heard about before and now that's happening a lot. So as much as I'm like bewildered by valuations and by you know, like how these companies are like sort of scaling, you're also seeing the traction there, which I think is sort of interesting and exciting to see. And I actually, I think there's like no better time to go build a company if you're a founder or you want to be a founder, because the tools are there, the models are there. It's become a lot cheaper now to be able to build and scale companies. I think Sam Altman famously said, you can have like a single-person multi-billion-dollar companies kind of thing and a billion-dollar company. I don't know about that. I still think you still need the people to be able to like scale and run with you and continue to go build. I also think a lot about once you've used a lot of these tools, especially if you're a non-technical founder, you still need to like actually build the end-to-end, like super, like the robust application that can scale with all these customers and everything else and figuring out go-to-market, figuring out who picks up the phone for customer support. All of that stuff, like you still need to go do it, but I do think it's like, it's a fantastic time to be able to go build companies at this point as such.

Totally.

Yeah. I think that all the points you mentioned and then like product velocity is faster than ever before, largely thanks to all these AI developer tools that people can use. And then also, like Olivia mentioned, lots of enterprise demand from companies that are normally really hard to get your foot in the door. And then finally, for the first time in a while, consumer excitement about trying new things. So like all of that and, and then all of the factors you mentioned make it, make it a really great time to be like starting and launching a new product.

Yeah, it's a fun time to be an investor too because like there's a lot happening, but it's also stressful because it's like if you turn your back for a minute, you know, something crazy is happening and you're not gonna see it.

Oh, I can imagine. I mean, I was, I was gonna ask, you get, you you know, sit in a lot of pitches from founders, you talk to a lot of founders. What do you wish, you know, there are founders focused more on and they're not focusing on today? And can be category of applications, it can be anything specific there. I also have this counter question. What do you wish, you know, they focused less on? It's like, oh man, it's so annoying. Like, stop doing this one thing over and over again. What would those look like for you?

I think in general, we usually say, I mean, Justine, I'd be curious if you agree, but usually it's like, if we knew what to build, like if we knew what the big thing was gonna be, we would probably, maybe we wouldn't go build it, but we'd go like join a company and build it or something. But so I think, I guess the best advice from my end or the thing that I get most excited about is when founders are, you know, working on something that clearly just gives them so much life and joy and energy, 'cause building a company is so, so hard. So it's actually less of like, oh, I wish they would build more in this space and less in that space and more of like, I wish founders would feel kind of free to like pivot the company, change the idea, keep exploring until you find something that you're super excited to wake up and work on, even for the like 12 months where it's not gonna be working and it's gonna be really painful versus like, oh, a VC told me to go build in this space. So I'm gonna go filled in this space even though like I hate my life and it's miserable. There's definitely categories where just because we see so many companies, we're like, oh, this is working here, it'll probably also work here. Like we'd love to see that. But in general, I feel like the best companies are the ones where the founders have this like insane kind of inner drive to just go.

Yeah, I think on a similar note in terms of like what we see that we're like, oh, this could be a trap or like you could waste a lot of time is All of the attention on AI right now means that it's really— it's not easy, but it's easier than ever to get a ton of hype and excitement around like a flashy Twitter demo or a post on Reddit or Hacker News or wherever. And sometimes we see founders fall into this pattern of like, I just want to release the next like hypey demo, like over and over. Like it's such a dopamine hit, which it is.

Yeah.

Yeah.

But, um, I think when you're building a company, you really have to sit back and think about like, what is like the thing I'm like, what is sort of my North Star that I'm aligned to being able to do? And then like, what are the steps I'm taking to build this product that like truly adds value along the way, whether it's consumer or whether it's enterprise. And a lot of the times that doesn't align with what makes like the most flashy Twitter or Reddit or wherever demo.

Yeah.

It's sort of like, like some companies do splashy demos a lot and the companies are doing amazing and the audience they're targeting is on Twitter and they reach people that way and it's awesome. But there's a ton of great companies that are not, you know, releasing video demo videos and going viral all the time that are just quietly building for their core audience. And I don't think founders should feel bad about taking that path either if it's what's best for their company, for their target customer and, and for their own style too.

What are things where you're like, stop working on that or do less of a particular thing? You— this has to be more spicy than this. It can't be— feel free to—

I know, I'm thinking.

I feel like it's hard to judge. We try, especially because Justine and I are not engineers. It's like hard to kind of judge people who are out there actually doing the work and building the things. I will say though, there's tons of companies where either here or at like our prior firm CRV where a founder would come to me, pitch me on the business. I would be thinking to myself like, this is not a great idea. Like, I love this person, but like if I had to bet, this company is probably not gonna work and I won't invest.

Yeah.

And then that person just keeps coming back and has a new idea and a new version of the business and reaches out again and pitches me again. I think there was one at CRV, Justine, where like he pitched us 4 or 5 times on different businesses. And the 5th time we actually invested. So, okay, you know, it's hard to, it's hard to ever, I think, tell a founder like, stop doing this or don't do this. But I think the main advice would be like, if your thing starts to work, no one's gonna kind of penalize you for the things that didn't work in the past. Yeah, like working kind of absolves all prior sins.

Yes, exactly.

I mean, I do think to be fair, There are categories though going in where it's harder than other categories where it will take more. Like, like, um, right now a bunch of sort of the like jet, like horizontal consumer platforms where it's like, oh, we're a chatbot and we analyze your PDFs and we can, can generate images. Like these things where they just want to be like your home for AI.

Yeah.

Um, and we're still seeing that. We saw a ton of those at the beginning. We're still seeing those. And, and in, in those sorts of cases, you have to be like so clear about like, how is either my feature set or target audience incredibly differentiated from something that like an OpenAI or Anthropic or like a Gemini does?

Yeah.

Um, or, you know, there's a ton of image models now. Like if you want to train in that new image model from scratch, like you better come in with like, you know, one, probably a strong research background and two, a really clear sense of like, this is what's missing in the market and this is what I'm going to focus on.

Yeah.

Versus just chasing like, hey, image generation has been such a big space, like I should do something here.

I think early last year we saw a bunch of companies, even like coming through YC, a lot of them like very sort of horizontal companies. And then you sort of suddenly saw this pivot. I think the most recent, I don't know about the fall batch, but the batch before then, I think like 200 companies were like vertical B2B AI companies. And you could sort of see this like sudden shift in like, nope, not doing horizontal anymore. I think everybody sort of woke up and were like, models are actually good enough to do the horizontal stuff. Like we, I don't think we need to be building this stuff. But the batch before then, I saw a lot of them doing, you know, analyzing PDFs and which, you know, to be fair at that time felt like really hard problems, right? Like you didn't really think about it as like, like, oh, this would be so easy in like a couple of months kind of thing. So I sort of like use YC batches as like the yardstick on like how the sentiment is shifting and what people are like starting to go focus on or think about as such. Justine, you had mentioned this specifically on like therapy. You first said something like, I know people throw shade at like therapy apps. I was like, wait, do they? Like, I didn't even know that was a thing. Why do people throw shade Like therapy apps?

Yes. So I think it's not— I think it's using AI for therapy. So I think there's basically— there's been this huge trend, and you've probably seen some of the posts on X about this, where like a lot of people use ChatGPT or Claude as a journal and then will ask, like, get support from it, get insights, and then ask questions like, what should I know about myself that I don't? And like, you see from that, like, there's— people love that and people find it super valuable, but there's also a ton of hate comments about like like, you know, these aren't real therapists. Like, why are you asking AI? It's just gonna like, tell you what you wanna hear. Like, that sort of thing.

Yeah.

I don't know. To me, there's always opportunity when there's sort of some sort of either mismatch or friction between supply and demand.

Yeah.

And in this case, like, I think a lot of people are, you know, want to be more thoughtful, more insightful, more resilient, more self-reflective, But there's so many barriers to like finding a therapist, booking an appointment that works around your work schedules, something that works with your insurance, like actually liking the person when you meet them, going back. Like, that's hard for a ton of people and many won't do it if they, unless they feel like they're in a crisis.

Yeah.

And so I think that sort of opportunity or, or like problem set is where AI can come in and be like, hey, You know, this is not going to be as good as you're— if you're having a severe mental health crisis as a human therapist. But for the 80% of people who just want to talk through anxious thoughts and get a way to reframe them, or want to get some insights on why they're getting a certain reaction from people often based on their behavior, like AI is exceptionally good at that. And I think we're going to see more and more products Around that.

Yeah, I agree. I think, I don't, I mean, for me, my use case is like less, I don't do it for therapy or therapy coach as such, but I have so many Claude projects which are like, I think they're like much closer to me like than some humans when in like different roles. Because for example, I have a writing coach and I think a lot of people have writing coaches where I basically feed a lot of my past writing content. And then I'm trying to like, I basically need a sounding board. I just need like a live sort of editor person who's like, no, that's dumb. Don't say that. Like, this is sort of how you want to like frame it. Which of course, Claude does it in a very nice, respectful way, never calls me dumb. But it's just like, it's really nice to sort of have real-time feedback.

Totally.

And the interesting thing for me is it took me like 2 weeks for me to sort of blur the line between like, actually, it doesn't matter if it's a real person or not. It's just somebody I talk to. And I'm just using them as a sounding board to sort of bounce off ideas, things I want to write about. Even like, sometimes it can be email, sometimes it's like, you know, tweets that I'm writing, like things like that. And it's amazing. It's just made such a big difference than to sort of like fester in my own mind on like, should I put this out? What would that look like? Oh, I should like tweak this thing. I should write this. Maybe this makes no sense. Oh, that's so stupid. And I'm just like, like thinking about it over and over again, and it just solves a lot of that for me. So I'm sure if I could find a use for it, I'm sure there are like lots and lots and lots of people who like find various coaching sort of use cases for the same thing.

But yeah, and I think honestly, even if you're not using it for therapy or coaching, like a lot of the value is the same in that, yeah, you get out of your own head and it forces you to get an external perspective, which is like when you think about it, That is like really magical that you can like have, you can, we can now have an interactive conversation and there's like not a person on the other end.

Yeah. Yeah, totally.

That, that continues to be like wild to me. And I think we are just like scratching the surface of all of the use cases. Yeah.

For the first time this year, I did a resolutions coach and I was like, here are all the things I want to do this year. And it was basically like, no, that's a lot. Like, don't— can you just like start scaling it back? And I was like, well, okay, okay. Like, what if I did these? And then it was— it gave me this plan. It was like, every week you need to be doing these. And I was like, wow, this is exhausting. Like, maybe I should like cut back even more. But it was sort of nice to sort of work with somebody to be like, I want to get all these things done. How do I work backwards from there to like tangibly make that happen? And it was like, okay, come back every few days and check in with me and we'll see how you're doing. And I'm like, oh crap, like now I have a manager on my laptop who I now have to be like responsible and be able to talk to.

I feel like this is— and Justine has written about this a little bit, but like one of the things that I think the two of us have benefited through throughout our lives is having kind of an instant second opinion that understands you and your context.

Yes.

And so we're starting to see this a little bit already with like ChatGPT's new memory features where The other day I was, you know, my— our dog Tilly had some stomach problems. So I was asking it a question about, you know, what food should we put her on? And at the end it said like, I hope Tilly is feeling better. And then I realized like, I didn't say Tilly in this conversation, but the fact that ChatGPT is now stored in its memory my dog name and like, it was just such a crazy—

And it would bring up past stuff. Like, I remember a month ago you said Tilly was also not eating and then you told me it was resolved when you like put a spoonful of peanut butter in her food.

Yes.

And it's like we would— we're so busy on a day-to-day basis, like we would never remember that sort of thing.

And so I feel like this element of— personalization is the wrong word. I don't know if we have quite a word for it yet, but like products that truly know you. Yeah. Is going to be so powerful and it's so like jarring in a positive way when you first experience it from like a computer.

It's very cool. It's like, oh yeah, this, this thing that actually knows me and like sort of cares about me in a weird way. Which I think is just sort of exciting there. Before we run out of time, this is the one tweet which I've been getting a lot of founders reaching out to me about. Just the— and it's, I think it stirred up this huge conversation on college dropouts as such. And then, you know, I had like, I think right before we started recording, we were talking about like these middle school, high school kids dropping out of school, being like, well, school's a waste of time.

Yes.

So I think, Olivia, you'd put out this tweet. I think there was this whole series of tweets on kids dropping out of college to focus on starting companies, founding companies. And a lot of people were like, oh, that's amazing. And a lot of people were like, well, I don't think so. You still need like essential skills to be able to get out there. You can always start companies later, but college is like once that you get to do, why would you drop out? And then people in the data point was like, So jarring because it was Stanford. It was not some like random college, right? It was like Stanford and a bunch of kids dropping out to go focus on starting companies. What do you think is really happening there? And what do you like, what are your personal thoughts on like whether it's a good thing or not?

Yeah, I think I mentioned in the tweet, but Justine and I, when we were at Stanford undergrad, so this is probably 2015, we were near the end of our undergrad. We had founded a startup incubator. And so we saw a lot of the startup activity on campus and back then, like If 1 or 2 people dropped out a year, it'd be like, oh my gosh, someone's dropping out, you know? And now, like, I was doing some reference calls the other day and there was this whole circle of like 10 kids in the same Stanford class that have all dropped out and are now kind of pursuing their own startups. I think, first of all, I think it's positive, especially for the student. It depends on the school, but like Stanford, for example, if you do your startup for 2 years and it doesn't work out and you decide you want to come back, like, they have a pretty flexible leave of absence. policy. But I feel like especially in the COVID era, a lot of these kids— I say kids, they're like 19, 20-year-olds now— a lot of them either finished high school or started college on Zoom.

Yeah.

And it kind of makes you reevaluate the entire value of education if you're like, I can pass my classes like not paying attention, watching Netflix in the background, and then just like cramming before the test for an hour. And like, what am I truly learning here? So that with like AI boom, you might assume younger founders who have a ton of energy can move fast, have new ideas, are kind of advantaged. There's all of these factors, I think, coming together. There's been more and more case studies of not just Snapchat and Facebook, but like Brex and Cursor, other examples of really young founders doing really big companies that investors are maybe more willing to fund. So all of these factors are coming together to put us in this place where it's like, I was thinking you don't even really need to graduate a great college anymore. You just need to get into the college and then that gives you, you know, credibility. Being a Stanford dropout is almost better than being a Stanford graduate.

Yeah, you get accepted, you get accepted, and then you just decide to not join. And so you can always put on your resume, was accepted in this Ivy League but decided to go to this startup, and you're instantly fundable then.

Totally. I think honestly that's a big difference because even thinking back, so we, we entered Stanford in 2012 for undergrad. Yeah, there were like— I'm trying to think, were there even any kids in our class you'd talk to them and they'd be like, my goal is to start a company? Yeah, like very few. Like, and now there's tons of high school kids like submitting papers at NeurIPS, like these big AI conferences, with the goal quickly transforming their research into a company. People who are like launching GPT wrappers on Twitter and getting a ton of buzz and, and like building a brand amongst themselves and then entering college, no, already knowing like, I want to do a company.

Yeah.

And I think if you enter college knowing you want to do a company and then you feel like you're able to execute on what you want to do, it probably feels like a very high cost to like staying in school. I think the risk is like, if we have too many young people like cosplaying as a founder, like it's easier than ever to like look cool on Twitter and know the terms and like be in the in-network. But, but then like being a founder on a day-to-day basis is like really hard. It's like really lonely. Like there are just as many, if not more, like downs than like great moments.

Yeah.

And so, um, I think, yeah, like when someone decides, hey, you know, like I know all of these things, I know this is going to be difficult, I know I'm giving up like potentially a really valuable degree and I am, have so much conviction in like this idea that I want to do it, like we're kind of almost always applauding that.

Yeah, got it. Yeah, I think, yeah, I never thought about the cosplaying part, but I do think about burnout at such a young age. Having both companies, it's the best thing you can do, but it's also on many days at the same time, it's the worst thing you can do and you're like, damn it, like I wish, and I would look at my peers who had like amazing FAANG jobs and free lunches and everything. And I'd be like, damn it, like that's just bad. Like I'm barely ramen profitable this month. And I look around and it's like, what choice in life did I make for me to just end up in this like horribly wrong state? And then, you know, like the next minute you're like shipping something, some customers said some nice thing and you're like, oh my God, it's the best thing. Like I have, full control over my destiny and I'm like on this rocket ship, this is great. So, as much as it's amazing to see people go through that, I also look at it and go, wow, like, are you really sure you know what you're doing? And this is—

Yeah.

And if you are truly starting a company, you're, you know, you're basically saying for the next about decade or so, you're going to go do this. And unless you're like really, really convinced that this is the thing that you want to go work on, Are you sure you want to be doing this at this point? And I don't know if many of them have like made that decision. And that's sort of like, I spend a lot of time talking to founders or people who want to be founders where I'm like, this is how it's going to be. My day-to-day looked like this. And I have like my diary and notes from that time. And it's not pretty, you guys, but if you're truly, truly convinced that this is it, then there is no stopping you. I think you're going to have a lot of fun and do this.

Yeah, totally. Yeah, totally. And honestly, I mean, we've seen people drop out to do a company and then go back to— so I think a lot of programs are giving people a year, 2 years, 3 years, and then letting them come back.

Yeah.

And I think, like, honestly, the funding environment, like Silicon Valley in general, is like very accepting of past— not even failure, but past tries.

Yeah.

And a lot of the college kids we talk to are like, I feel really strongly about this. I'm going to try to do this now. Like, if this doesn't work, like, I'll finish my degree. I'll go get a big company job. And then I'll be even more educated for when I try again. Yeah, which is like never would have occurred to us when we were in undergrad.

Totally. I would have never thought about it that way. Wow. Like, first of all, my mom, my mom was like not— she's probably listening to this, so she would be like, no, get a real job. Like, I think even when I started companies, they were like, are you homeless? Like, you say founder, you say entrepreneur, but I don't— who pays you? And I'm like, uh, it's a tricky question. No one pays me, but you know, just go with it. Uh, you couldn't at that time tell her about like Zuckerberg and Social Network, and it was just this thing that you just did. But yeah, I see. I mean, I think it's just, we talked about this multiple times, it's become so easy, easy in the sense like it's more cost efficient to be able to go build companies, especially with AI. And so it's now easy to go at least test it out and get it out there. So you are going to see a lot of founders just, you know, wanting to try and see what's going on. I'm actually pretty bearish on MBAs. Like, if you imagine, like, you starting companies when you're like 17 or 18 or whenever you're dropping out of school, you're dropping out of college. Like, imagine now convincing someone you have to go get an MBA. Like, why? Like, I've already done, I've learned everything I wanted to 5 years ago, 8 years ago, just doing this thing. So I don't know about, like, Like, you know, undergrad as such, but MBA, I'm like pretty bearish on the future of it.

There was definitely a period of time I feel like where at Stanford at least, the engineering students would be going over and kind of like begging someone from the business school to like help them, you know, with their pitch deck or help them pitch and raise money.

Financial model.

Yeah.

And now we're absolutely in the era where the business school students are going over to the engineering school and begging to be like led on to this amazing product as like an intern on strategy or something for the summer.

Be our first BD person on this thing.

Exactly.

Help us with go-to-market. It's, yeah.

Yeah, the tables have turned on that one, which I think is good in general.

I mean, I didn't do an MBA. I'm sort of biased, right? It's like I just never saw the value in it. I decided to go start companies early on and I feel like I learned a whole lot just doing that, like both successes and failures. And so I can't like, and that's only one side of the story. Like if I'd done I would have been like a— I would have probably assessed it more fairly. But I always have this chip on my shoulder on maybe I'm not as polished as the MBA people, but I at least like went ahead and started these companies and was able to go learn kind of thing. So, you know, just different for different people, I guess. What's your personal favorite AI product? It's something that you use all the time or you can't live without, or it's like a sort of specific use case that you're sort of— you're obsessed about. It's the one thing that I'm sure a lot of our listeners would want to know about because you both are, you live entirely online. You're always talking about consumer AI. This is literally your jobs. This is what you do at a16z on just looking at all things AI. So what is the, like your personal favorite AI application tool product as such?

Yeah, it's a great question. And we spend so much time testing tons of products. I feel bad, honestly, the amount of subscriptions that I'm expensing. I'm like afraid to look. Because, uh, we test so many things.

But they're all like $20 a month.

Just—

I know, and it's— yeah, and I use them. I think like, um, I think I use Crea a ton on the— because I'm— I make a ton of creative content. And so what's super helpful to me is being able to iterate really quickly around image generation and then animate those images into videos, add audio, that sort of thing. And I think what I like about Krea is, um, they— and I do a ton around LoRAs basically. So this concept of like generating an image, like training a model to understand this person's face or this object or this product photo or the style.

Yeah.

And Krea makes it so easy to train all of those LoRAs, combine them in one place, add them to other styles or characters or objects that other people have made, like iterate, edit, inpaint, and then like press to send it to a bunch of different video models and then have it animated. Um, because I think like one of the, the sort of longstanding or things about the AI creative space has just been like you have to have so many subscriptions to so many different apps and you don't know what to go to when you just want to make something. Yep.

Yep.

And so that's probably the number one place that I go to when I want to make something because I can do so much across a bunch of models I've never used Crea.

I'm not, I have not done any part of the creative side of things there. So I see your tweets about it and I'm like, I, one day I should like get to it. Uh, but yeah, no, I have, I have to go try it now, especially that such a, the bar is so high when it's coming recommended from you. So I have to now really go try it.

It's a rabbit hole though. You will, it will take up, it will become your number one hobby very quickly. So just, just beware.

There goes my 2025 resolutions. It's like, what did you get done, Kriya? That's all I did.

Yes.

Olivia, what about you?

I mean, we mentioned Granola already. I'm a big fan of that one. I think that it's such a great product category too, because like for a note-taker, if it works and if you're the kind of person that's on Zoom all day, then like you will be an hourly active user until the end of time. I would say the other one that we haven't talked about that I love is this product called Gamma. Which lets you use AI to create and edit slides.

Yeah. Yeah.

Documents and websites. It's interesting for me, like Justine and I have always done a lot of writing. We used to write a weekly newsletter. And then as I started to use Gamma more, I actually realized that especially if I'm trying to lay out like an investment thesis I've spent 4 months on, it's actually such a better medium to put that in a deck or something more sophisticated or visual or interactive than it is for me to write like 10,000 words that no one is going to read. including my parents who love me and support everything I do. And so, but I used to never do that because making a slide deck before AI was such a pain, like all the formatting and all the edits, images, like all of that.

And so I'm a big fan of like what you mentioned around front end, I would say, with the web. Yeah, exactly. Like with Bolt and all these development tools. Yeah, yeah, yeah, totally, totally. So that's been a fun—

and that's another product where, of course, it has a lot of kind of hype and excitement, but the average Gamma user is like, a high school student or a small business owner, or, you know, it's a long tail of people that are not necessarily AI aficionados, but the product is simple in a way that's hard to build. It's very kind of intuitive and beautiful to use.

So that would be my pick, right? What are the odds that in 6 months it's completely different? There's something else that you're looking at or you're really obsessed I think we'll add, we will never remove.

Yes, yeah, yeah.

We'll always love them, but we'll add on more and more. I say this because when NotebookLM came out, right, it was amazing to see how people were using it, especially the fake podcasts where they were like fighting with each other and you would like give them a topic and it's just, and it was hilarious. It was such a great sort of like creative use of just, you know, get people to like converse. And you sort of, I looked at that and went, this is amazing. Like, you know, now to be able to go create something like this. And I didn't think from a creative use case standpoint, anything could like go top that. And then you have so many more from there. So I, for me at least, like every couple of months, there's something that sort of replaces it.

Yeah.

And so I'm really curious to see how it all shakes out end of this year, December. What we should do is to have you both come back back end of the year and sort of play clips from this episode to be like, haha, look at that. Or we'd be like, oh, we were totally right. Like every one of these that we said, we like totally validated as such. And it'll be super fun to come and see where we are in like a year from or 11 months from today as such.

I'm afraid but excited for that. Yeah.

Yeah.

Gonna be super fun. Anyway, thank you so much. I know we went way past the time. Thank you so much for just spending time with us and I think this one, this is one of those episodes where people are, everyone's going to find something that they really love because we covered so much here, all the way from model improvements, talking through like how last year was, all the way to what should founders do? Should you drop out of college? Therapy, AI apps, and so on. So touched so much here, and I hope, you know, everyone finds something in it that's their sweet spot as such. But thank you so much for coming on the show. Thank you.

Thank you.

This was so fun. Thanks for having us.