The consumer AI apps breaking out in 2026

Looking back on this, say like 5 or 10 years from now, I think it'll be kind of like the transition we saw in the late '90s, early 2000s, where at that time, like a dot-com company was like a specific designator of a unique type of technology. And then like 5 years later, every software company was a dot-com company. Every, like everyone has a website. And I think we'll see. And so like you don't even say dot-com anymore, right? And I think in like 5 years, every technology company will be an AI company and every AI company will be agentic and an And so we won't even be probably using words like AI or agent anymore because it's just going to be baked into everything that we use.

In this episode, I talked to Olivia Moore, partner at Andreessen Horowitz, also known as a16z, one of the flagship venture capital firms of Silicon Valley. At a16z, Olivia focuses on AI with a special focus on consumer AI apps, including which ones are trending, which ones are breaking through beyond just the tech crowd, and which ones are unlocking the most powerful new capabilities and benefits for you and me. Olivia and I cover a ton of topics in this conversation. We talk about a16z's new report on the top 100 gen AI consumer apps. We talk about the rise of OpenClaw and personal AI agents in 2026. We talk about the evolution of AI tools and Olivia's personal AI stack. We talk about the central role that context and memory will play in the next stage of AI. We also talk about global trends, creative tools, switching costs, the acceleration of coding agents, and much, much more. More than anyone in the AI space right now, Olivia has her finger on the pulse of the AI app ecosystem. And I'm confident you're gonna come away from this conversation with at least one tool that's gonna change the way you work, and you can hold me to that. All right, so here it is, our conversation with Olivia Moore of a16z. All right, well, Olivia, I would love it if you could talk a little bit about what a16z does. Andreessen Horowitz, well known to probably most of our audience, but just in case, you know, for those few that would like to know and understand more, Tell us a little bit about the company, but also your role at the company and what you do.

Yeah, no, it's a great question. We are a global venture firm founded by Marc Andreessen and Ben Horowitz. We invest across essentially every stage and in basically every sector as well. So seed all the way to growth and beyond. Applications, infrastructure, crypto, biohealthcare. And I would say that the kind of mission of the firm is to be a part of every important technology story. So our job is to to kind of find and partner with the best startups. My role, 'cause we have a lot of different roles and funds within the firm.

Yeah.

I sit within our early-stage AI applications group, which essentially means mostly leading kind of seed Series A, Series B investments, and largely in companies that are building AI more at the product layer than the model layer. So our infrastructure team does all the fantastic, like, foundation model investments. And then on our side, we invest in things that are selling directly to consumers, SMB, enterprise, prosumers. And it's just been a crazy and fun, I guess, 2 and a half, 3 years now. I've been in venture for a decade, and right now is definitely the most fun and the most exciting it's ever been. So, yeah, it's just, it's an awesome time to be building a startup. Wow.

And have you been working on AI the whole time you've been at a16z?

The first 2 years or so was pre-AI. It's funny to think now that, like, the first, I think, real indicator of what was to come was was Midjourney, actually, in kind of summer 2022. But before that, I was doing mostly non-AI consumer software. And it's interesting to think back because you can see all these platform shifts, especially in consumer. Like, web was the first platform shift, and then mobile, and now AI. And kind of as you get to the tail end of a platform shift, most of the good ideas have been built, right? So a lot of what you're seeing is kind of marginal, and some of them can still break out and become fantastic companies. But once a new platform comes along, there's just so many more exciting opportunities to invest in. So that's what we've been doing for the past few years.

You have been really looking at this sort of personal stack of AI apps too. You know, I have been trying to learn more about this. You know, as you know, we talked a little bit about that. You know, I've only been at The Deep View for 3 months and my thesis was that I wanted to spend my time writing about, thinking about, reading about, learning about all of this stuff more because there was so much. And I, even though we were covering it, you know, when I was at ZDNet before, I covered it just enough. We covered it just enough for me to just know that I'm— there's so much that I'm missing, right? That I, I can't keep up with. And it's like, I want to keep up with it. And one of the things that I found, um, last year that I enjoyed so much was your story on your AI stack.

Mm-hmm.

Uh, and you had number of tools there that, um, I hadn't heard of. This was a, this is a video people can still find. It's a video on YouTube, um, that you did. I mean, you all do a lot of your own content. You, you all are an incredible content machine in addition to being a VC, uh, you know, from Mark on down, everybody, you know, uh, seems like is very active on Twitter.

A lot of prolific posters.

Yes. Yes. A lot. We can talk a little bit about that because among all the venture firms out there, you know, it seems to be such a, core part of the DNA there. But this one video was really inspiring to me because I found several tools there that were incredibly helpful. You know, one of them, maybe the first one that comes to mind is Happenstance. So for those who aren't familiar, Happenstance will take your network on LinkedIn, Twitter, um, your contacts, you can, you can connect it to email and—

Yeah.

Email. Yes. Thank you. And then it will Uh, it will make connections. Um, you can ask it, find me, uh, CTOs working in healthcare in the San Francisco Bay Area. And it will look on your sort of connections, second connections, and it'll give you like this list of like 30 people and you're like, what? I never could have found this. Right.

Um, totally.

Yeah.

It's amazing. I use that product all the time. I think that's also a great example of like, yes, LinkedIn exists and yes, you have all of this kind of ambient existing data about who you know. And who others that you know also know. But now with AI, you can productize that into something that's like so much more powerful and so much easier than trying to like buy a LinkedIn sales license and then like set up a bunch of workflows and things like that. And so, yeah, I think we have already seen many, many products that kind of up the game like that. And I think there'll be a lot more to come.

It's so— I remember other ones from that list too. I remember That list was the first time I heard about Granola.

Yes.

Which has now taken on a life of its own.

Yes.

Almost become a bit of a viral sort of hit. It, it's one of the ones that I find has some consumer awareness.

Yeah.

You know, even as well broadly. And for those who aren't familiar, Granola is something that you can record phone calls. Actually, I shouldn't say record. It will take notes for you on phone calls or meetings, but it doesn't have an agent that sits there or anything.

Yeah.

It's, it's just sort of taking the, the, the transcript transcript, and then you can put your notes at the same time, like, think like your thoughts maybe that weren't mentioned in the call, and it will align them. And it's a, it's a really powerful way to be able to collect all of that stuff. And I found that to be, you know, a really useful tool that your, your video is the first time I saw that one. And I'll mention another one, Overlap, I remember was on your list too. And that's one now, of course, now doing the podcast. We just relaunched this podcast about 6 weeks ago, and now we're like, okay, we need the clip part of this. Like, we need to get that going. And I had a demo with them just a, just a couple weeks ago. Um, I had again first heard about it when you mentioned it in that video and finally connected with them now that we're doing the podcast. And I'm like, this is an incredible tool. It will take a video, it will, um, you can tune it a bit to your own preferences and that, and then it can even make a bunch of clips for you that you can approve. Or if you get to the point where you're confident with it, it can just send it to all of the networks, you know, that you, you want. So I really have you to thank for those. Those are 3 just off the top of my head from that video. And I think you had about 10 of them on there. Yeah.

A lot of those products have had real staying power as well. I think Granola is a great example because when you're able to capture that, kind of quality and density of data, it really becomes a moat in that it's— the switching costs are very high if I were to like leave Granola for another product because it just knows so much about me at this point. But yeah, I totally agree with what you said in that my job here is basically to be on top of all of the great consumer and prosumer AI products that are coming out every day. And yet I often feel underwater and like I'm not seeing— it's hard to keep up with seeing everything every day. And so—

Yeah.

I, yeah, I, I love just kind of getting to, to summarize at least some of what we're seeing with others.

This episode of the DeepView Conversations is brought to you by JetBrains. If you're building an AI right now, you already know the difference between shipping and stalling often comes down to having the right tools early on. Every decision compounds— the frameworks you choose, the infrastructure you bet on, and the development tools your team uses every single day. That's why more than 55,000 startups 65 unicorns, and 15 million developers globally trust JetBrains. These are the same tools used by engineers at 90 of the Fortune Global Top 100 companies. If you're an early-stage startup, JetBrains makes it easier to get rolling. Through the JetBrains Startup Program, accepted startups receive 50% off on licenses across IDEs, team tools, and JetBrains AI for up to 10 users over 5 years. That's long-term support while you go from MVP to product-market fit to scale. If you're building something ambitious, whether it's AI agents, inference infrastructure, or the next breakout developer platform, your tools should grow with you. You can learn more and apply to the JetBrains Startup Program at jetbrains.com/startup. That's jetbrains.com/startups. Build fast, build well, and do it with tools that are built by developers for developers. We thank JetBrains for their support of The Deep View. And now back to the conversation. It's really exciting when you see somebody discover one of these tools too. I have a friend who works in a, um, he does not work in the tech space. He's an educator. He works in a nonprofit. And when he signed up for our newsletter, The Deep View, Granola was one of our— we have this little quiz you do, like, what are you interested in? What would help you? And then we sort of spit out a few tools. These are a couple tools you might wanna try.

Cool.

And he had told me when I started the job, like the first week, like a good friend, he signed up to The Deep View even though AI and tech isn't necessarily his domain, but he was being so supportive. And he said, hey, in your, uh, in that, when I signed up for your newsletter, it told me about this tool. It said this would be a good one that helped me, Granola. And he said, because I'm in meetings a lot, and he said, this thing is like incredible for me. It has been, I can have it turn it on. I don't, it's not like this thing that I have to, you know, put in the meeting and then people ask questions or they sort of have a, you know, they feel, you know, awkward or that kind of thing, or it, some of these agents will even sort of crash meetings sometimes, like, and, and so he said, this has been, you know, such a, you know, a game changer. And I love it. So seeing those kinds of things is really rewarding. I'm sure you must have some of those experiences where people in your life who know what you do are like, okay, tell me what tool, you know, I should be trying.

All the time. And it's actually very good research, even for us, because I'm not technical, and in some ways in consumer, that's actually an advantage because I am probably closer to the experience of what an average user onboarding to a product is going to feel like. So I always ask myself the question like, would I actually complete the signup for this? And then would I keep using it if I wasn't looking at it to invest in, like just as a normal consumer using a product? But I'm also able to very quickly see if I hand a product to like my mom, my little brother, my cousins, like you can see very quickly if something kind of has the spark to catch and retain the mainstream consumer. And I think that's a perspective that is, is hard to kind of stay in touch with if you're in the like ex-tech Twitter VC world all the time, which is fantastic. But I mean, OpenClaw is an example of like that is such a usable product and a fantastic product for that ecosystem. And it has not escaped containment to the mainstream consumer.

Yes.

Yes.

Oh boy, we were gonna talk about agents. We will, we will, we will definitely come back to that topic. How about from that list even last year, Olivia, do you still use most of those tools? Like, are, that was your tech stack that you did. I wanna say it was the summer last year if I'm remembering right. Yeah. Are most of those still on your list?

Yeah, most, almost all of them still are. Like Granola, Gamma, the Perplexity Comet browser, I, I still think is kind of the best in class AI browser. It's interesting because for me, with AI software, I'm not necessarily seeing myself replacing things like, oh, I no longer use this tool, I use that tool, but just actually expanding the number of tools I use. And I think that's especially been true the last 2 or 3 months as both ChatGPT and Claude have added all of these connectors and apps where like now actually I'm probably using these tools more because I can kind of orchestrate my use of them through a very clean consumer interface that owns all this memory and context on me.

That brings up a perfect question for you, and especially cuz you mentioned Perplexity, you know, comment as well. I also been using it. I essentially harassed their PR people until they gave me an early version of it, you know, amazing last summer and have loved using it. It's so interesting both that it has some of these things, like it can do some things for you and to be able to, uh, have it by default, you know, be able to, to use it as your way to, to find stuff. And yeah, one of the things I did is I made Google the default, um, search engine. So if I went to the, the, the bar up at the top and I put there, I could get to Google. And, but that then the new tab was always a Perplexity search.

Yeah.

And of course you can do the different models. So, For, for all those reasons, I, I loved it. Still love it.

Yeah.

One of the things that's been a challenge for me, and this is what I'd love your perspective on, cuz, and it does get to agents a little bit, but I've been using and trying over the past couple months, or, uh, at least past month, especially the, um, the agent that's in ChatGPT.

Yep.

Uh, ChatGPT Pulse. And what I've found is the more that, like anything in these AIs, the more they know about you, right? The better they can, more helpful they can be proactively. So what I found is I also use ChatGPT Atlas, their browser. So I, yeah, I sort of have them both. I have the, I use different things. Like one I have more social media in, one I do more like researching in.

Yeah.

But what I found was, and I'm wondering if this starts to become a, you know, part of the, the moat or part of the, yeah, the, the thing to, to think about with these, because I was using, I've been using Pulse, I found myself wanting to use ChatGPT Atlas a little bit more because I'm like, if I, put some of that stuff in there, it's gonna know what I'm doing and what I'm thinking about and learning about right now. And then it's gonna start sending me some cards of like, you've been trying to learn more about, you know, the finance of this $100 billion, you know, deal for chat for, for OpenAI. And so we're gonna send you some information that will be proactive about that. What are your thoughts about, about that? Does data start to become the moat? Do we start to want to use products when we, become cognizant of the fact that, oh, the more context I give them, it can be more helpful for me.

This is actually very well-timed because this is one of my favorite topics. And I think that we're seeing this really start to play out now because previously there's been very little switching costs going from like one, one horizontal LLM to another, like Claude, ChatGPT, Gemini. Lots of people might multi-tenant. Claude, as you probably saw recently, launched this feature of like export your memory and come over to us. And I think in the past that has been very doable and the switching costs have been low. I predict a couple of things are going to happen in the next year that raise those switching costs a bit and people might start to kind of double down on one platform.

Okay.

I think the memory and the context layer that you mentioned is one of the most important ones, especially because Sam Altman kind of hinted earlier this year that ChatGPT is going to launch Authenticate with ChatGPT as a layer. Across other AI products. So then basically you can use your ChatGPT memory and your tokens to go use other tools, and then it'll know so much about you, you know, on first signup. And so in that world, I'm also wanting to use ChatGPT more so that it has the most up-to-date picture of kind of who I am. 2 other things on that. I think one, Claude and ChatGPT are both investing in like app stores and MCPs. And we might kind of see a little bit of an iOS-Android thing start to play out where developers kind of choose, this is who I'm going to ship to first. This is who I'm going to build the most sophisticated product. And then the last thing I think is we've seen launches of things like ChatGPT group chats, which I would say was early. It's, it's, I would guess it's doing fine, but not amazing. But if you imagine a world where there's like a really great product, actually an AI product that you're communicating with other people within, then I think that also kind of locks you in. 'Cause if you're gonna churn, you have to pull them all with you. Yes. So I think you're right that we're gonna start to see higher switching costs and more lock-in.

That's gonna be really interesting to watch. And it does start to become, there's a little bit of some, some tug of war. And the thing is, is the pace of, of new features is so amazing right now. And, and I'm also cognizant of, like you said, like, It's a little different for those of us that live and breathe this stuff every day. But new capabilities come and you want to, if you're aware of them, you want to try them. You're like, whoa, that's a really interesting new thing.

Yes.

And it tugs at you a little bit. And so I realize it's a little specialized, but probably the folks in this, our audience, most of the folks in our audience, I'll admit, the DeepView audience, are folks who work in AI or work with AI pretty, pretty daily. So they're, they're probably aware of this and thinking about it as well. So it's gonna be really, I think it's gonna be con— to your point, we're gonna be a little more conflicted, I think, in 2026 than we have been previously because that context starts to become a really, really powerful factor.

I totally agree. And, and obviously we're seeing all of the major kind of labs and incumbents like fighting for the mainstream consumer in their own way, probably to get that context layer and that lock-in. I feel like they're doing it all with slightly different strategies. Like Sam has said he wants ChatGPT to be for everyone, which includes, you know, free usage and ads. Claude is really doubling down on kind of prosumer. If you actually look at the— there's like 200+ apps now that you can use in both ChatGPT and Claude. And there's only 11% overlap between them, which is crazy. Like they, they are really focused on different audiences.

Um, that is such a great, um, a great stat. I had not, I did not have awareness of that, uh, uh, at all. That's so interesting. And you're right, like my sense is they do tend to be diverging much more, the two of them. Um, although this past week was just sort of a little bit bananas, right? Um, maybe. There's this, this industry has been, the news cycle has been crazy for 3 years, but this last week was like nothing I've ever seen. Yes. Like it, it was, and it, it was also at the level of people beyond, it had broken beyond AI, it had broken beyond tech. It was sort of like everywhere, everywhere I went, every, the only thing anybody wanted to talk about was like the AI thing, which was, yeah, which was pretty wild. Was it the same for you?

Yeah, we saw, um, Katy Perry getting in the, getting in the race this week tweeting about Claude, which is something I never would've expected. And so I think that's probably a good signal of what's been happening. It's interesting too, because like for the vast majority of the US population, they will know ChatGPT and they will probably know Gemini just because it's distributed through the Google products they use all the time. But we saw this with DeepSeek as well, where like almost for any other product, all press is good press in the sense that like most people have not, had not heard of Claude before this week, right? And I think what we saw with, with a DeepSeek was that it drove a huge amount of volume and traffic to the product. And then there was no really like killer feature to retain people. So outside of markets like China and Russia where other LLM products are kind of sanctioned, we haven't seen continued usage. I'm gonna be really curious to see what the numbers for Claude look like in the US in a couple months. Like, do they hold on to a lot of users? Do they churn? What happens?

Yeah. Me too. You know, it's interesting. That brought up another thing that I was thinking about this week, which is that last Saturday I had a friend who also, this is someone who doesn't work in tech, doesn't really follow tech that closely, but, you know, is, is also, you know, interested. They know what I do, right?

Yeah.

And so, and came to me and said, I have to tell you, I tried AI for the first time. And, and so I, and I was like, Well, but, but they also, they use, they've used some AI tools, but they, for them that was like a chatbot. I used a chatbot for the first time. Yeah. And they're like, I had, I used it to do 2 things. And they said, I'm, I'm a little bit ticked off and I'm also a little amazed.

Okay.

And, and they had to do it. They had to help 'em write a speech, like a sort of, okay, you know, person in a personal life, you know, like a social gathering. And then they, they work in academia, so they were working on a a very important presentation.

Wow.

And they were trying to work on their thinking, like unfolding a topic.

Yes.

And so they said the speech was great. They said it, it, it saved me. It was gonna take me 45 minutes. I did it in 5. Like it wrote something up, then I edited it and we were done. And I was, I was so happy. And he said, I was a little upset because it was good and it was better than anything I would've written. Yes. And so I was like, okay. And he said, but this one is the one I really wanna show you. And they, they were so excited by this. They wanted to show me the whole thinking that they had back and forth. But what they did, I should say, They did Claude. They downloaded Claude. Mm-hmm. This is what amazed me. Like they started their first time they've ever used a chatbot. They like, so I downloaded Claude on my phone and my wife was driving and the whole time I was chatting back and forth of it. I had to write the best man speech or whatever that speech was. It was something like that. And then the other, and then I had this, this, this presentation I'm working on, this academic presentation I need to submit, you know, in the next couple weeks. And he showed me the thinking and I was really amazed. And he showed me, you know, how he had started it. And he essentially had this like 6-thread back and forth with—

Yeah.

Claude about this topic. And am I explaining it clearly? And does this, you know, patch it? And then it asked, Claude asked for more context. It's like, when you said this, I wasn't quite sure, do you mean X or do you mean Y? And it did it. And he said, I was impressed that it did, you know, all of that. Like when it didn't know, like it asked me for more context and I, And he wrote a couple paragraphs, you know, even with his thumbs in the car. And it was really wild to me because one, he started with Claude. He'd never used ChatGPT before. Interesting.

Yep.

So the first place he went was Claude, cuz it, this was last Saturday. So obviously Claude was, was everywhere. In the news. Yes. It was, it was everywhere. So he started with Claude and then he had what, what I thought would be not the sort of thing I would see somebody start with. You typically start with like the same things you would ask Google, right?

Yes.

But here was somebody who very quickly figured out, oh, if I give this thing more context, if I have a, a very in-depth back and forth, I could get to something really amazing and powerful. And, and that would've been like having a conversation with a colleague or, or an expert, you know? And that sort of amazed me that we're at a point now that even the, the chatbot itself could get somebody to engage.

Yeah.

At that level. And I was like, whoa, this is, this is different than even just like 6 months ago when people were using chatbots for the first time. So I thought I'd share that and see, you probably have some similar kinds of things that you're seeing.

Yeah, I love that story cuz I think it's a really good example of the fact that like actually, it feels like AI is everywhere right now and yet like only 10% of the global population is using ChatGPT on a weekly basis. And kind of the crazy thing about that is actually we're seeing the first use experience get better and better and better and better every single month. So like people who start on ChatGPT or Claude this month are gonna have such a better experience than, you know, a, a first month experience than someone who started like a year and a half ago. We actually see this playing out in retention cohorts on the consumer side. We almost, you almost never see them kind of smile back up, which is like more people, people reactivate and actually spend more time on it over time, which we are seeing. Like the first year post-launch of ChatGPT, it didn't really get much better. And so the usage kind of plateaued. And then over the last 2 years, it's gotten a lot better. And the cohorts are like really highly correlated with kind of the release of new models and new features. And then the other thing we're seeing is just the most recent cohorts start from a much higher retention and engagement standpoint just because the product is so much more able to like meet their needs. And so it's wild to think like what that will look like 10 years from now if we've already seen so much improvement in the last 2 or 3.

Wow. No, that's such a great insight, Olivia. I appreciate the data on that. It's funny, one of the things I was gonna ask you about is I had read this data point and I wanted to see if it was consistent with what you've seen. that like on a, on a monthly basis, it's even was 13% is what I saw. And you said on a weekly it's, you know, 10%.

Yeah.

And so to me that says, and I, I always wanted to get your sense of it, like we are still so early at the beginning of the curve, right? Like we're just barely, there's this sense that like this thing has played itself out and maybe, you know, there we're gonna, um, you know, have a, um, uh, a point where people are like, oh, this just isn't as useful.

Yeah.

And maybe it was overhyped. My sense is not that though. My sense is like we're still sort of so early at the curve. And that point, the insight you mentioned of like the user experience when you start now, yeah, is actually much stickier because you have the experience like my friend who, you know, very quickly did 2 things that it saved him a lot of time and not only saved him time but made the outputs, the outcomes better.

Looking back on this, say like 5 or 10 years from now, I think it'll be kind of like the transition we saw in the late '90s, early 2000s, where at that time, like, a dot-com company was like a specific designator of a unique type of technology. And then like 5 years later, every software company was a dot-com company. Every, like, everyone has a website. And I think we'll see. And so like, you don't even say dot-com anymore, right?

That's right.

And I think in like 5 years, every technology company will be an AI company. And every AI company will be agentic and an agent company. And so we won't even be probably using words like AI or agent anymore because it's just gonna be baked into everything that we use.

Like electricity or—

Yes.

Chips.

Like, yes. It's just the infrastructure that our world runs on.

I'm glad you brought up agentic, 'cause I was thinking even like 6 months from now, if somebody started, you know, like, like somebody that starts 6 months from now, you know, like my, my friend, My sense is that not only is it going to say, I can help you with that, but what would— can I do— can I take some steps for you and do something and create it and send it to your email? And can I email you the link to this presentation that I've also just turned that into slides for you or something? You know?

That's absolutely coming. I think it's already here on a bunch of products. Like, I'm starting to see Claude Cowork begin to reach outside of the tech bubble a little bit into broader knowledge work.

Okay.

But I think overall, the OpenClaw kind of architecture, in my opinion, will, will be like the big tech unlock of, of 2026. I meet with so many founders every day who have been inspired by that or were already building around that. And yeah, it's just such a— you can, to your point, like you can deliver outcomes, not just kind of like inputs. Anymore to users, which is so much more powerful. And like OpenCLAW itself, fantastic product. I have used it, loved it. It's not consumer grade. And so I think we're going to see whether it's OpenAI now that they acquired it, they'll productize it into something more accessible. And then I think we'll see a bunch of founders and independent builders build their own version of something like an OpenCLAW for specific use cases.

Boy, I agree. this, this movement, OpenClaw, really Claude, like Claude Code.

Yeah.

Which had become pretty agentic even at the end of last year.

Yes.

And then they were like, okay, let's make Cowork. So it's like a little friendlier. Yes.

Yes.

And then OpenClaw.

Yeah.

Like all of a sudden this like personal AI agent thing. We'd been hearing agents a year and for about a year there were enterprise use cases last year. I mean, even like Salesforce is talking about they might change their name to Agent Force, you know, I mean, that, it's having that much of a moment in the enterprise, but this like personal AI agent really has exploded even just like in the last 8 weeks, um, into the consciousness in the, in the AI space in some really, some really powerful ways. Like you said, the barriers to do OpenCLaw are pretty high. Like you have to in the command line and all of that. So it's totally, it's not something that we're gonna see millions of people adopt anytime soon, but the, But it has shown sort of the path. And I think what you're saying is what we should expect a lot of people to be walking on this path in 2026.

Yes.

Yes.

And building all sorts of things. Like I think that the idea behind OpenCLAW, at least in my opinion, is like previously you as a consumer or a user had to like pull AI into what you were doing and like ask for help and give it specific instructions. And with OpenCLAW, you can kind of delegate it a high-level task and it will like push into the world and like do things for you. And check with you when it needs to. Um, but I think that that is just like a fundamental shift in how we think about what software can do for us. And there's gonna be dozens, hundreds of, of big, big companies that are kind of built around that ethos, but doing different things.

Yes. I, so I'd love to get your thought on, um, one other thing, uh, regarding that. And we'll probably come back around to agents cuz we're gonna talk about some research that you, you have as well. To see how this has been an example to me of like how fast, almost like unbelievably fast this space is moving.

Yeah.

When OpenClaw really burst onto the scene over, you know, a number of weeks.

Yes.

To the point like within a month, then, you know, they're talking to OpenAI and OpenAI, you know, buys—

Yes.

Well, they hired Pieter Steinberger, the founder of OpenClaw. And then OpenClaw itself becomes a foundation so that it remains open source. And really the idea, and Sam Altman talked about, like, the idea was we're gonna hire Peter Steinberger to come and take OpenClaw, what he did there, this AI agent, this personal AI agent that works on behalf of you and is so proactive in working for you. And we're gonna make it a product and available for everybody. We're gonna, essentially turn it into a consumer product like ChatGPT. And it's like, that sounds exciting. We'll see that, you know, that'll take 2 or 3 months, right? And then the following week, Perplexity Computer came out.

Yes.

And it was like, it was almost like everything that Sam had talked about the week before, Perplexity Computer came out and they had essentially all of, like almost all of it solved. And I've learned since that they had just germinated the idea at CES. It was 2 people at, at Wow. On their team germinating the idea at CES that became Perplexity Computer essentially like 8 weeks later, right after. So they had started germinating it even before OpenCLAW had kind of become its, its sort of viral hit. But that is such an indication of the, the incredible velocity in AI. Feels like even in the past 8 weeks, 12 weeks, A month even has, has, has accelerated to a level that is really kind of mind-boggling. And I, I have to think, and this is part I would love your opinion on, one, if, if you, you're seeing the same thing, but also I think part of it is the tools themselves because of all of these builders having access to agents now. Like, it's almost like one person has, can do the work of 10 because they are using these tools in, in such smart ways, especially builders, developers. you know, entrepreneurs. And that seems to take what something that was already accelerating at an insane pace and just taking it to a level that we haven't seen before.

I totally agree. Yeah, I was talking with one of my portfolio companies last night and like just the leverage that they're getting from products like, you know, Cursor and Codex and Claude Code has allowed them to build so much faster. They say their problem now, actually, they are not— they're no longer bottlenecked by like build speed or build capacity, which historically has been the bottleneck. The bottleneck now is them deciding what is actually a good thing to ship, what is actually like a good feature that will enhance the product rather than detract from it, because it's so easy to build everything so quickly. I think the other interesting thing, and this is kind of on like the Perplexity computer point, is what's pretty unique now is that largely all of the best-in-class models are not just held within the company that created them and used for their own products, but they're available to build on. So like Perplexity Computer is largely, I believe, kind of Opus and Sonnet driven. And so that makes— that creates a world where then if everyone can access the best models, maybe not in real time, but, but quite close, then how you productize the models, the workflow you build around, how you handle distribution, all of that becomes much, much more important.

Mm-hmm.

And that's a big thing that we think about, especially making consumer investments. Consumer is just so fickle and so hard to get a magical product and product that works. And that continues to be very, very true in the AI era.

I should say for our audience, I'm becoming aware that I should explain Perplexity Computer just maybe a little bit to say like it did 3 things that made it a little easier to access, you know, for— and I'm sure what even what OpenAI is building, with Peter Steinberger, they will, um, probably take this another level. But one of the things you have to do if you want one of these AI agents is you have to hook it up to an API. Now that's a very techie thing in itself. And then you have to pay for tokens, right? Yes. For, for that as well. And one of the things, because, um, Perplexity Computer's only on their $250, their max, um, their, their plan, their $200 to $250 a month plan. Yes. But they, and you get like 10,000 credits essentially. And then they gave people an extra 20,000 credits. so that they could try this out. But essentially, they obscured all of the cost because people even are spending crazy amounts of money on token costs right now. I've heard some people say they're spending more than what they would spend on an employee, a developer to build the thing on some of the token costs, which is wild. So they made that part obscure to the user. The other part was that you could, it has 20 different models to your point.

Yeah.

So it, it always automatically routing the models, you know, for you. So for some queries it would send them to ChatGPT, especially things that were like high context. For things that were more coding related, it was sending a lot of those to, to Anthropic's models. And then it had some open source models and as well as sort of closed source frontier models. But that model and, Actually, I use Google especially for image models. Their image models right now are so bananas.

Amazing.

So good.

Yeah, good way to describe it.

So it was really interesting to see, okay, that they were doing this sort of user layer to your point. They really focused on the user interface side of it and took something that was very technical and made it a little more accessible. Oh, the last thing they did that I think was really interesting is if you use Claude Code or Replit or Cursor, Codex, all these tools, and you get it to code you an app, you still have to deploy the app somewhere.

Yes.

But Lovable, the one like unlock that Lovable did is like, you just code it and we'll just put it up on our server. We'll deploy it for you right away. And you could make something in 5 minutes and I could send the link to you and say, Olivia, take a look at this thing, you know, that I just just made. Well, Perplexity Computer does that too. So it will deploy the app or that you make right away. It does the sort of the lovable thing. So those are 3, what seem at the surface, maybe simple things, hard to, to do. And just having those few unlocks, it was all of a sudden like, whoa, this all, this looks like a, uh, you know, a consumer version of what's very similar to OpenCLAW pretty quickly. And that just impressed me that, that, that happened so fast.

Yeah, I think Perplexity made some great acquisitions and has acquired some like fantastic agentic builders.

Okay.

And they have a really good consumer product sense. Like they were first really to the browser. I think that's a great point about kind of abstracting away some of the complexity because it's interesting, some of these products, the better they are for developers, the worse they are for consumers and vice versa. Like Mm-hmm. If you make a consumer select a model that they want to use for a generation, they are going to churn. Like, that is terrifying. Like, they don't, they don't know what any of these models mean. Like, we in the tech world might obsess about them, but like the average person doesn't want to make that decision. They just want the product to work. And I think something like a Manus was an interesting kind of predecessor to like an OpenCLAW. And I think a lot of people were surprised by how quickly Manus exploded to hundreds of millions in ARR. But it really was the first kind of working consumer-grade agent that could operate across products and get things done for you. And I think we're going to see a lot more along those lines, both from startups and from kind of the labs and the incumbents.

And we should mention Manus was bought by Meta.

Yes.

And so it's, it's having its moment. It's gotten some, some positive publicity off of the OpenCLAW and Perplexity Computer. There, there were several people when I was tweeting about Perplexity Computer, you know, on X that were telling me like, it's fine, but Manus is better. Like I had a number of those. So it was, it, it has its fans and it also was sort of, was having a moment riding, you know, on this, this wave of AI agent, you know, positive buzz too. I'm glad you, I'm glad you surfaced that one. Yeah.

It's interesting cuz I think that like at this stage in AI, There's a lot of like Claude versus ChatGPT or like Manus versus OpenClaw. And I think that like that makes sense if you assume that AI is kind of a limited market. But my view, as I said earlier, is that everything is going to be AI in 10 years.

Yeah.

And in that version of the world, similar to how there's like so many massive billion-dollar-plus software companies now, there's gonna be, it's like an industry, not a market. There's opportunity for a massive amount of products that can each get millions, tens of millions, hundreds of millions of users. And so I think a lot of what we're seeing right now is like head-to-head competition is more like there's room for a lot of things to exist.

Yeah, it's not zero-sum.

Yes, it's a better way to put it.

There are a lot of things to— a lot of flowers to bloom, you know, in this space. Yeah, no, that's a great point. All right. I wanna talk a little bit about some new things that, that you all have. You have this list that, that you do every year.

Yeah.

On the GenAI consumer apps. Uh, and the, the March 2026 version is here.

Mm-hmm.

So why don't you talk to a little about what this list is, you know, how you started it and, and why your team clearly is, is very enthusiastic about it.

So we started it in 2023. At that time, you know, ChatGPT had been released. People were maybe starting to use AI, but it was very early. And I think we were asking ourselves, like, as a consumer team, how can we better understand what the mainstream person is picking up and using in AI?

Mm-hmm.

And that's gonna be, that's a different question than like, who has the most revenue from tech companies and power use? Users, things like that. So we actually did something fairly simple, which is just pull every website in the world, um, unique traffic, and then every mobile app in the world, monthly active users. And we go down each of those lists and grab the first 50 companies in each that are, um, kind of AI native or now majority AI enabled, like things at the beginning of the report, like When we first started the report, things like a Notion or a Canva would not make it on the list because they had a few AI features, but they're not AI products. Now Notion is reporting that 50% of new ARR is coming from their AI features. So we've kind of evolved the methodology of the list a little bit over time, but I think it still gives a pretty good look at what's getting the most usage globally.

Who on your team— I assume it takes a whole team to work on this. Who on your team, you know, comes together to make the list?

Honestly, everyone. We pull the data first, and then, you know, here at Andreessen, we have early-stage teams, we have growth teams that probably have even more, you know, sophisticated understanding of what's going on with later-stage companies and in the public market, so they contribute. We have a lot of writers that help to make sure that, like, whatever I am saying about the data analysis is actually, you know, intelligible to the mainstream consumer because I'm so deep in all of the AI and tech stuff.

Sure.

But it's been a really rewarding and, and extremely fun thing to work on. I think, you know, consumer as an area of investment goes in and out of vogue a lot. Like there might be a few foundational consumer companies every couple of years. It's not like enterprise where you can kind of like almost brute force success. And so I think a lot of people don't pay a lot of attention to consumer or they're like, I'm a consumer, so I understand. And so they don't think that deeply about it. And you definitely don't wanna overthink things in consumer software land. But we also like, we've noticed that looking at this data in a more granular way gives you kind of an interesting sense of how the average person thinks about AI.

Very good. Now, there were some top takeaways.

Yes.

That you had from this. We'd love to talk about a few of them.

Yeah.

The number one was that ChatGPT is still the, in the lead of the, the default AI, you know, app, even with the events of the past week. But, but you also mentioned that, but the race is on, that there, there are more companies, you know, in the report, there are more, there's more mindshare being shared. Across the industry.

That's absolutely it. It's, it's— I think it was like a one-horse race, you know, even a year ago, like the only AI product that most people knew was ChatGPT. And we're seeing more products emerge. And I think it's like I said before, there's room for so many big companies and it's probably ultimately good for consumers to have companies and products that are specializing in building like the best-in-class version of of a specific thing. So I think we've seen ChatGPT really, really narrow in on like the mainstream user. We've seen Claude and Anthropic really, really narrow in on kind of technical users, prosumers, some enterprise stuff. We've seen Gemini very much double down on creative tools. To your point about Nano Banana and Veo, those are kind of best in class on both the image and video front right now. And so I don't think it's like a winner-take-all necessarily. I think that the level of value that these products are providing to people are such that they will probably multi-tenant and all of the platforms will be able to make, you know, good revenue, whether it's ad revenue, transaction revenue, if it— if in the future you're buying everything through ChatGPT or direct subscription revenue.

Global usage splintering by product as well. Talk a little bit about this insight.

Yeah. So I think we see kind of a real bifurcation in between like the US, the rest of the world, and then specifically Russia and China are their own thing. Okay. I think US usage looks similar to most of the rest of the world in that, you know, ChatGPT is number one, Gemini is number two. Some countries will randomly have like a little more Claude usage or a little more Perplexity usage. And then China and Russia, because they have so much, so many sanctions and also like state-sponsored, like local players of their own, I would say those are the 2 markets other than the US and maybe Israel as well, where models are being developed that are kind of really globally competitive. So in China, for example, they have the lowest rate of ChatGPT and Gemini usage of any country. It's only like 15% combined. So they're using Douyin, which is a ByteDance product, which also makes TikTok. They're using DeepSeek still, like DeepSeek usage has fallen off a cliff in the US, but it's extremely popular in China. They're using Kimee and Quen. And then Russia kind of mirrors China a bit. They're the number 2 DeepSeek user.

Okay.

And then have some, some local players like GigaChat and Alice and others. I think the other interesting finding from that global look that we took is we kind of for the first time like heat mapped what countries are using AI more and what countries are using AI less on a per capita basis. Maybe not surprising, a little surprising to me, but we saw, you know, Singapore, UAE, Hong Kong, South Korea coming at the very, very top.

Yeah.

The US is kind of like low upper group at number 20. And then some of these countries like Russia and China, where there's more restrictions are closer to the bottom.

In the report, I should say, too, one of the things I didn't mention, these are— the report is the top 100 GenAI apps, you know, broadly. And then these are insights, you know, from it. In that international look, you did mention, despite you said DeepSeek is sort of waning a bit in the US, its use, but it is the one product that bridges the divide across, you know, China, Russia, US? It sort of has some popularity in all of them.

Yeah, I think, yeah, still has some usage in the US for sure. Not as much as it maybe did 6 months ago, but it's still significant. And I think it really did break through with that kind of viral, you know, all press is good press, even if it's bad press moment when it comes to consumer AI. And then, yeah, it's like to use ChatGPT in China, like the number of steps you have to go through with like a VPN and everything else. I think for the average person, especially since DeepSeek is still quite a good model, it's just much easier to use that. Yeah. So I would expect that to continue.

Another one was creative tools. So they're, they are starting to also break through. We're seeing more of them start to, to have their moment, you know, and it's funny, I've seen also some non-technical people talk to me, you know, recently about Suno, about Sora, about, Nano Banana, ElevenLabs, you know, like Eleven Reader from ElevenLabs and, and things like that. That sort of surprised me. Like I thought of those tools as being pretty technical as well. But talk a little bit more about, you know, what you all saw from the creative tools standpoint.

Yeah, creative tools are, are, it was really the first area where consumer AI was working. Like Midjourney came before ChatGPT, which I think a lot of people— people don't remember at this point because, you know, you had to log into a Discord server to use it. It was a really kind of dedicated and specific experience. I think when, when we first launched this list, image generators kind of dominated in creative tools in terms of independent products. And that actually has reversed in the last few lists to the point that now the kind of core model providers, ChatGPT, Gemini in particular, have built very, very good image generation.

Yeah.

And so, you know, most of what a consumer might wanna generate or edit in terms of an AI image can be done in those interfaces. And so you'll notice that the products that are image-focused that still appear on our list, like a Midjourney or an Ideogram or something like that, they're very aesthetically opinionated, therefore more sophisticated workflows versus the kind of like one-and-done prompt-to-image. What we're seeing kind of rise in their place are modalities that the core model companies are spending a little bit less time on. I would say voice, like ElevenLabs is one. Music, Suno, Udio are two others. That's a tough space because of copyright and everything. So I think the labs are maybe less inclined to fight that battle when they're also trying to build AGI. And then video is also fascinating because Google and OpenAI have done a great job there. And China can train on any data, including copyrighted data. And so we see this other kind of parallel ecosystem of, you know, SeaDance, Kling, Hailo, like many of the best-in-class video models are coming out of there as well. And, and so I think we've been excited about products where you can use many of the models in the same place, things like a, like a Crea, just because it seems unlikely that there's gonna be like one video model or one image model to kind of rule them all.

Very good. Well, my last question, and it's funny 'cause part of this question is perfect for you, even though it's the same one that I ask everyone. But, you know, the greatest compounder, the greatest multiplier is time. And so I, I like to ask what, you know, What is your best tip for kind of the ways you spend time, what you learned about, you know, optimizing your time, you know, whether it could be productivity or, or other ways? Uh, and then what's the AI tool that you've been using lately that you'd love to sort of share and, and think other people, you know, might not know about and could, could benefit from?

Oh, I love that question. On, on the time perspective, I think like AI is intimidating for a lot of people and Again, we're early in the sense that most of these products have some kind of setup time and setup costs. And a lot of people might think, oh, it's just gonna produce AI slop or it's gonna sound like AI. But I think especially in the last 3, 4, or 5 months, like if you give these models examples of your past work, whether that be images or writing or something else, they can really give you very significant leverage in, in doing work for you. But that kind of sounds like you and is natural to you. And they can even push your thinking more. I've written lots of blog posts now alongside ChatGPT and Claude and end up in a much better place than I think if I had kind of written it by myself. On the tool side, there's, there is so many. I try to kind of immerse myself in the ecosystem. I found myself making a lot of presentations lately because I think it's kind of a great way to communicate information. And it's also like valuable, both synchronous and asynchronous. And so I think Gamma has been one that I'm using a lot. And that is a great example of a product that is consumer grade. Like you could, you could give it to anyone off the shelf and they could use it. So that's one that I would recommend.

Olivia, thank you so much for your time. Thanks for being here.

Thanks for having me.

Hope you have a safe day.

Yeah. You too.

Take care.