S12E6: Data Ownership in the AI Era with Anna Kazlauskas, Founder of Vana

I think there's really this spirit of hacker culture at MIT, which is just build whatever you want. We used to go to actually the dumpsters at MIT where the labs would get rid of their hardware and their computers. And so I would go there to get GPUs to like mine Ethereum with and just build a bunch of stuff. I just had computer parts all over my room. The world's your playground. There are lots of smart people around who can help you.

Welcome to The Room Podcast.

I'm Claudia Laurie.

The founder.

And I'm Madison McIlwain, the funder. We are on a mission to open the door to the room where it happens, sharing the stories of some of the world's most iconic builders.

Authentic storytelling is core to our community. It all started as two best friends who met in the heart of San Francisco as product managers in 2018. When I left Uber to start my company, Madison was our first yes. It was these conversations that birthed The Room Podcast.

And nearly 5 years later, our guests have built businesses with a combined enterprise value of over $100 billion. You'll recognize conversations with your favorite technology founders, from Perplexity to Vercel to Zillow to Cloudflare, all across our 110-episode catalog. The best way for you to get to know our community, both online and offline, is by subscribing to the podcast and our newsletter, which you can find at theroompodcast.com.

Perkins Coie supports the most innovative entrepreneurs and investors in fast-moving and high-growth sectors, addressing their myriad of legal needs. But the firm doesn't just provide end-to-end legal and business counseling to its startup clients. It also facilitates introductions to key advisors and sources of capital. Perkins Coie's interactive website, Startup Percolator, offers access to programs, resources, and rich dynamic content. Designed to assist entrepreneurs on their startup journey. To learn more, go to startuppercolator.com and perkinscoie.com.

This podcast is brought to you by Mercury, the banking platform businesses like The Room Podcast use to simplify their finances. It's time banking did more than hold your money. Now it can. With Mercury, you can pay bills in seconds, close the books faster, and even send invoices. Not only does Mercury do away with a patchwork of tools, it eliminates guesswork, giving you complete and accurate visibility into your business's finances, all from one account. Apply in minutes at mercury.com.

Welcome back to The Room podcast. Today we're excited to take you inside the room with Anna Kozlowska, co-founder and CEO of Vana, a crypto-native platform building a new paradigm for data ownership and user-owned AI. Anna's journey began at MIT, where she studied computer science and conducted AI research at CSAIL before founding her startup Iambic through Y Combinator. Her work in crypto at Celo combined with her engineering background ultimately led her to start Vana with a bold mission: to give individuals control over their data and a stake in the AI systems their data helps power. Vana's pioneering a new category called DataFi, where users can collectively own, manage, and monetize their data through on-chain tools and data DAOs. Since launch, the platform has spun up over 300 data DAOs and raised $25 million from funds like Paradigm, Coinbase Ventures, and Polychain Capital. In this episode, Anna takes us through her founding story, what it means to build at the intersection of crypto and AI, and how Vana is navigating both regulatory UX and perception hurdles that come with both. We also explore how user-owned AI could reshape incentives online and the talent it takes to get there. Madison and I were struck by Anna's clarity and long-term conviction, and we're thrilled to share this conversation with you. Let's open the door.

Anna, thank you so much for joining us in The Room today. We're really excited to have you, and we always like to start at the beginning with our guests. So to start us off, we're curious, where did you grow up and how has that shaped your view of the world?

I was born in Montreal, Canada. My mom's from the Philippines and my dad's from Lithuania. Very global background. Moved from Montreal to Sweden as a kid, and then to Minnesota, which is mostly where I grew up until I went to school in Boston at MIT. So, yeah, mostly cold places, lots of different perspectives, just coming from different places, a bunch of different languages. In terms of how that shaped my worldview, like, there's a lot we could analyze. I don't know. Like, I think that— I do think that some of the technical work felt more universal. I was born in Montreal, sort of spoke French, actually went to school in Minnesota in French as well. But really liked math type stuff. It felt like a more fundamental truth. You could do it in any language. It was true of programming as well, which I got into in middle school, just programming graphing calculators. Those are the geographies where I grew up.

Thank you for sharing that. I love that idea of a universal truth that is math and being a language that no matter what you speak, you can actually kind of understand and make new friends and intuitively get to know people. I definitely felt like math as a young girl was definitely something I thrived in and loved and Well, I wouldn't say it's core to my identity today. Like I could totally see that as a safe space for you as a young kid and learning along the way. And so I guess as you grew in that interest, and of course we'll get to what you did at MIT, but before even that moment, how did these parts of who you were inform this idea of becoming a founder? Did you always think you were gonna become one?

I didn't really know what being a founder was, right? Cuz I grew up in Minnesota, so I like knew I really liked building things and taking things apart, and I knew I liked programming. I think that my mom had heard of Silicon Valley and had told me like, okay, there's, I think you might like it, but I was super far out. And I actually, at the time, I don't think I really realized what was possible through a for-profit entity. 'Cause my dad is a professor and my mom's a writer and like none of us really come from that, I don't know, business side. And I honestly didn't know that many people who worked at companies. I just didn't really have it on my radar. I did know that I loved building things. I would always take things apart, put them together, code different things. And so that builder aspect I knew from super early on. I actually thought I wanted to be chair of the Fed because I was like really into central banks. And so yeah, I had a picture of Janet Yellen in my high school bedroom. I like worked at the Fed in high school. The Fed is central to it all. They control currency that controls the economy. And that was more of a lens through which I looked at things like I, started reading The Economist pretty early, which is how I got into economics and then started to read more research. But I didn't really realize the more founder path just because I wasn't that aware of it.

I mean, exposure is so much of what we know. And so your early curiosity brought you to MIT, which in many ways is the establishment for building. There is just so much ecosystem and culture at MIT around being a builder. So tell us a little bit about that. You became an engineer studying computer science. How did you go from just building in general to building in a tech-forward way?

I was always interested in this intersection of economics and computer science. And so I got to MIT to study that. And pretty soon after, one of my dormmates was really into cryptocurrency and Ethereum. And so he was actually mining in his room. And to give you a picture of this guy, like very free-spirited, really aggressively believes in freedom, never used any fiat currency, only crypto, never wore shoes, only ate soylent, very radical in the sense of what, what can digital freedom look like? So through that, got exposed to the early MIT Bitcoin Club, which was like 5 people at the time. I started mining Ethereum and was just really fascinated with this idea of being able to build a system which is jointly owned by a bunch of people and you don't need a central entity, whether it's a company or a central bank or whatever that point of centralization would be to actually coordinate stuff. So started learning about these different systems people were building, met the first entrepreneurs I got to know who were, 'cause I think I lived in a weird dorm where a lot of people would drop out and march to the beat of their own drum. And so, got to meet folks who had done that and started to see that there were different paths you could take. I think there's really the spirit of hacker culture at MIT, which is just build whatever you want. Like we used to go to actually the dumpsters at MIT where the labs would get rid of their hardware and their computers. And so, I would go there to get GPUs to like mine Ethereum with and just build a bunch of stuff. I just had computer parts all over my room and sort of this like view of almost whatever you wanna do, just build it. Like the world's your playground. There are lots of smart people around who can help you. And so, that was the culture that, that I was happy to be immersed in.

And that led to CCL, which is MIT's AI lab. What research did you ultimately do during that time?

At the time it was called natural language processing. Now people think of it as LLMs and AI and the paper that had come out right around that time time, which was 2017, is the Attention is All You Need paper, which is the foundation of modern AI. And the big insight there was this new way of being able to have an AI look at a bunch of stuff at once and contextualize it and know what's important versus not. It wasn't yet able to generate text, but it was able to analyze text. So a lot of what the research I was doing was related to is building these like AI-friendly representations of text. For the purposes of being able to analyze a bunch of data and understand. It was actually related to some economics research. So we were studying these transcripts of villages in rural India, and we were analyzing how the way people spoke changed after an intervention and using these vector embeddings to map the differences.

Super interesting grounding us in 2017 lore of what was NLP at the time, and now what we all call AI and LLMs. It almost feels like it happened slowly and then all at once in terms of now what we take for granted as just being what computers can do. But at the time, even that was still unknown, as you mentioned, the inability to generate text. It could just read and slightly infer and make differences. So—

Yeah, that was like clearly the big unlock, right? The question was, when can this actually be generative? I would always ask my professor, Regina Barzilay, who's like this pioneer in natural language processing. I'd be like, how many years out until these models work for generative stuff? Email her every few years. And I feel like in 2017, she was estimating like 5 years out, which was actually like spot on if you think about the ChatGPT moment. And the thing that I was really noticing in doing that research is data is really what matters. I just always wanted better data, and I was like, oh, I wish I could train on this, I wish I could train on that. But often data is living within the walled gardens of some corporation. And so if you want to interact with it, you probably have to go work at that company. And then it's like, well, I don't necessarily want to join this company just for the purposes of being able to train AI on their data. I was starting to see that data was going to be a really big component of AI in a world where AI started to do more and more economically valuable work.

So how did this insight around the power of data, which by the way, totally agree, led to Iambic, your YC-backed startup in batch 2018? Walk us through that insight to this company.

It was very much a 20-year-old kid being like, oh, okay. Like they gave me this job of sorting documents at the World Bank. So the other interns were like sorting through all these documents and I was like, actually AI could do that pretty easily. And so I like trained a relatively simple model to do that. And any AI model is good data. That matters a lot more than anything else. And so, yeah, in this case, how do you teach an AI model to be able to categorize a document? They do a lot of that in Washington, DC, just sorting documents, adding up how much do we spend on this, how much do we spend on that. Trained an AI model that was like actually working really well for that kind of stuff. And then honestly accidentally started this company where then I was selling document sorting software to government agencies. Ended up Dropping outta school, going through YC, it was like my first intro into Silicon Valley. Like I had literally never worked at a company before. I had only done research at central banks and at CSAIL and was very much a researcher, but really liked it. What I saw is that if you have a for-profit mechanism behind something, you can scale it a lot. And there were a lot of really smart people in YC who I got to learn from, and it was just my crash course into Silicon Valley, right? I dropped out and I was living here in San Francisco in a house with 12 other founders. I think of them as almost my startup older brothers. They were just helping me and all my friends were in school. I just dropped out, moved out here and was like, okay, like, what is this thing? I remember there were even, there was some event at a bar and I like wasn't 21 yet for YC and I was like, okay, like I actually can't go in here. So yeah, I mean, I think that this is really awesome spirit of generosity and mentorship in Silicon Valley and a bunch of folks helped me land on my feet and figure everything out. One thing I realized in going through Y Combinator and kind of learning about just how involved a company is, as I was like, whoa, this is not a summer project. This is a 5 to 10 year journey that you're dedicating your life and like every possible resource towards. And I was like, wait, I don't care about sorting documents. Why am I doing this? And so I was working with the wrong person at the time. And actually YC was super understanding. I actually don't wanna spend 10 years building this. This is probably not what we should be doing. But I guess going to the data question that you were asking, like One of the main things we were running into is how do you get the right data? So I was actually buying data on Amazon Mechanical Turk, getting people to label documents and stuff like that, 'cause I always needed this better training data. And I was surprised how bad some of the tooling was and also how expensive it was where I was like, hey, I feel like some of my like relatives in the Philippines could do this for cheaper, right? I was just like, I think that there's definitely some kind of inefficiency here. And that's what got me super interested in this data problem and being on the other side of it, which is as a machine learning person training AI models, how do I get the right data?

Oh my gosh. Incredibly cool story to just hear about your early YC experience and sounds like very much that classic startup story with the 12 people in the house pulling on a thread that laid a foundation for kind of what you're building now. As you mentioned, prior to YC, you had never worked at another company, and so After iambic, you joined Celo, which I believe was working on making blockchain work for mobile. Tell us a little bit more about how that experience shaped your understanding of decentralization and data ownership, which then led to Vana.

I had actually spent time my freshman year at MIT explaining blockchain and Bitcoin to regulators. So I had gone to the Ivory Coast. I met someone from Zimbabwe who showed me a picture of literally wheelbarrows full of money. There was like hyperinflation, and so it reached that point. And I was like, wow, these guys really need Bitcoin. Some of us, we like want Bitcoin or we like want crypto for the purposes of self-sovereignty and freedom. This use case in emerging markets where you really don't have a stable store of value is an amazing zero to one. Previously you had no stable store of value. With crypto you do, whether it is Bitcoin or like a stablecoin, which is pegged to the US dollar. And the thesis with Celo was really like, How do you make blockchains accessible to people who need it most in emerging markets? So how do you make that available on a low-end Android phone? It's not that good at interacting with a blockchain, right? Like at the time, especially there was a lot of technical stuff you had to run on your device, et cetera, to be able to actually interact with these chains. It's gotten a lot better over time, but the view is like, how do you build global Venmo almost that's just built on top of crypto rails? So that anyone in an emerging market can like send, receive, save money. What I saw there, well, one, I learned how to like build a new blockchain, which is actually a massive undertaking. You've heard the term like Layer 1, that basically is referring to kind of like a new standalone blockchain where it's a bunch of different nodes that are keeping state of the network and finding consensus. And so, I got to work with like an amazing team. I joined really early. I was a third engineer. We were working in the Tenderloin, WeWork, uh, in San Francisco super early on. And it was just great cuz the founders were really experienced entrepreneurs. They had sold their last company to GoDaddy. One of the founders was like an MIT professor. And so, I felt like it was a good group for me to learn from because I'm a little more academic, research-leaning, impact-focused. So, lots of learnings along the way. Getting to see that scale from 5 people to like 100 people. What even is that? How do things play out? Little tiny details that you might not think matter that much related to hiring or related to onboarding. It actually like really matters and plays out. And that was a really great team to get to build alongside for me. Many years.

So then fast forwarding to 2021, you start Vanna, but before we jump into that story, how did you decide that, okay, I wanna start something again and I am along for the ride for a 5 to 10 year journey?

Yeah, there was never a moment where I was like, it's time. I think that I had this in my mind for a while of, so just the idea of data ownership, to me it's almost as important as physical property ownership, probably just because I spend so much of my time online in the same way that, You wouldn't want people to be able to take your physical belongings from you. And so even back at MIT, actually, like Solid Project, which is started by Tim Berners-Lee and is this approach to data ownership I tracked for a while and was interested in. I'd gotten really interested in Urbit as well for a period of time, which is a similar approach to data ownership. But then the question was like, okay, tactically, how do you do that? And at the time, me and my co-founder Arthur, who goes by Art, we honestly still didn't have that good of an answer to like, how do you do this tactically? Like, how do you solve data ownership and along the way build better AI models from it? But we had like a very early version, which was, I guess, data labeling and paying people in a stablecoin and being able to connect your data to a wallet and then use that. I was building it and I was like, okay, like I wanna build more of this. And so I probably can't keep being a fellow. And so then I just started building that.

Often that's how the best companies are built, natural progression for something you're so passionate about. So let's talk a little bit about Vana. It's creating an alternative to centralized AI monopolies using crypto principles. So think ownership incentives, decentralization to build an AI model that users collectively own and benefit from. You've teased this out a little bit, but tell us, why should users care?

From an end user perspective, if you think about maybe 3 to 5 years out, we have super powerful AI, right? You can imagine making both an AI of yourself that can do the things that you're really good at, whether that's writing or thinking or being an economist, whatever it is. That AI model, if it's trained on your data, it's really important that you actually own it. That AI model could be earning wages on your behalf. It could actually be doing a lot of economic activity. And if you think about a world where humans get most of their information from AI, then whoever decides what data goes into AI is basically deciding truth. And one good example of this is back when Google Bard had put this filter on where like if you asked it for a medieval king, it would show you like an Indian woman, right? And I think a lot of people were like, oh, it's this crazy wokeness of the 2020s, what's going on here? But like if we didn't have another source of truth, that would just become true for us. And so whoever is deciding what that AI actually believes and is trained on is like really shaping the future of society and knowledge and information. And yeah, I think it's very important that is done through a democratic process. And then as a user, I think that often people think about data privacy as today, maybe just like an ideological choice, right? It's just a preference. And I think in an AI world, it becomes much more of a significant economic piece. A lot of people are worried about AI taking their jobs. What happens if OpenAI creates AGI and then none of us have work anymore? But if you have the right attribution in place, everybody has already helped to make ChatGPT what it is by posting data on the internet. So if you have a good attribution, then everyone just owns the AI model that they help to create. So one of the guiding use cases we keep in mind is how do you get to 100 million users who contributed their data? To be able to train a better AI model than any single company can. And the reason why you can do this in a decentralized context is that ultimately AI just comes down to data. If you want the most data, the approach today is that you get it from either the public internet, so you can scrape everything. If you take a whole public internet, it's around 15 trillion words, and that's what leading AI models are trained on today. So Llama 3, around 15 trillion words. Now we're like, okay, we wanna make better AI as an industry. Where do we get more data? We've already used up the whole public internet. So the approach has been buy it from companies. Reddit's earning like $200 million selling their data. Apple's buying data for $50 million. And what a lot of people don't realize is that they actually legally own all their data, right? So in the same way that when you park your car in a parking lot, the parking lot doesn't own your car. It's just sitting there. They can't like randomly take your wheels. They can't decide not to give you your car back. When you put your data in a platform, it's just like that. Like it's legally your data and you can always go and get it. And so the approach that we take at Vana is by having users take their data from across these different platforms, you're actually able to cut across the walled gardens of each of these individual big tech companies and build a larger dataset than any single company has, and therefore a better AI model, right? And so then you really have this path towards extremely powerful, super intelligent AI that is owned by everybody who contributed their data.

You mentioned monetization as a core piece. If you own something, you can make money off of it. And so with Vana, users own their data and they can be compensated for its use. What kind of compensation models are you exploring for users who contribute their data?

When you contribute your data, what you're doing is you're basically joining almost like a labor union for your data. It's like a group of people who come together and pooled their data. And one example is there's a Reddit Data DAO. They have like 140,000 users who've pooled their data all together, and now they've trained an AI model that it's mostly just good at like shitposting 'cause it's trained on Reddit data, right? But you have the end-to-end flow working there. There are some other ones. So there's like a Tesla one that's focused on car data. They sell data to car battery companies to be able to train AI models to see how the car battery degrades over time. I think that each car battery company pays $3 a month per user. So that's one case there you can pay and stables, you can also pay in these dataset-specific tokens where you're basically getting ownership of the dataset overall. And then when that dataset is used, essentially the amount of money that the company is paying to access that data can either be redirected to everybody who's holding a share of the dataset, or you can just burn those data tokens, which is the equivalent thing. That's all done at the DataDAO level. So DAO stands for Decentralized autonomous organization. You can really think of it as like a union, a cooperative. Each of these data cooperatives are led by their own teams. So they're probably around 50 people full-time who just build these data DAOs on VANA. There are 20 of them live today. There's a 23andMe one that's trying to buy 23andMe the company outright, which is actually interesting. There's like a Twitter one. There are some sleep health data ones, biohacker, any dataset. You can think of, you can actually build one of these data unions around, and then you're able to aggregate that dataset, whether it's to train an AI model, to sell the data, or just to be able to negotiate with it. You could actually negotiate with a platform for better terms of service, like in the case of 23andMe, where their goal is to prevent 23andMe from selling their data.

This is an interesting framing. I hadn't really internalized this concept of DAO as a union, and I think people have mixed reviews on their perspectives on unions today, but I think of the inception of unions, like there was like an actual, almost like physical need for unions because of the workers who were not cared for, overlooked by large entities. And I think at the time though, there was a competing theory of thought, which is around like welfare capitalism that's actually coming from the institutions itself, almost extrapolating this analogy of unions to that concept as well. To what extent do you feel like the big tech companies, putting them in the place of the automotive analogy, do you think will they start to like help empower users with their data to compete with this unionization DAO approach?

If you have a situation where data unions are forming and users are asking for more control over their data, will the big tech companies start to give them more control? Over time, yes. And I think we're actually starting to see that, which is very positive. If you wanna build self-sovereign data, then you want users to have more and more control. One example of this is like, it used to be really hard to request your data to get it out of a platform, even though you legally own it, they can make it super inconvenient. And that's getting better and better. So now you can get it within 30 days. In Europe, there are specific cases where you can get it via API access. You can always get the real-time data. I think we'll see that in the US at some point, but I would mention that the incentives for a big tech company are the opposite. They actually don't want users to like realize that they own their data or have access to that data because that's literally what they're making money from. It's like a direct trade-off. And so we think of it as this like data revolution where you actually are shifting how people think about, hey, this is actually mine. I should care about, um, owning this. In terms of historic analogies, another one I think is interesting is actually comparing it to serfdom, when people didn't own their lands. And if you look at when people got good land ownership rights throughout Asia and stuff like that, basically farmers became way more productive if they actually owned their land, because then they could benefit from the fact that like, if I come up with some more productive approach to agriculture, that directly benefits me rather than me being like, ah, I always have to hand this over. So I think that we could see that with data. As well. And we're already starting to see that where people are like, hey, I want my data to be clean and good. And those are some interesting kind of historic examples. Yeah. The union framing, I think it can be helpful because unions are really interesting in the US. I actually didn't know, did you know that UAW, United Auto Workers Union, I think they have a billion-dollar endowment. They're actually an incredibly powerful force. And so, yeah, there are a lot of structures that we can borrow from for the purposes of data. But yeah, the comparison of data as labor, I think, is also very relevant. Data is the input that creates something, and that is what these humans, like, that's what all of us are actually contributing. And so we wanna defend that right.

For the first time in a really long time, we're fundamentally rethinking what our relationship is to data, what our ownership model is with data, and how we're turning data into an earning opportunity. And so with Vana, you're really pioneering a new category here, which is DataFi. In pioneering a new category, often you need financing in order to scale and really realize your vision. And so today Vana has raised $25 million from notable investors, including Paradigm, Coinbase Ventures, Polychain Capital. Who was the first person to say yes to investing in you? And then as a corollary, what's something about fundraising, especially for AI and crypto, that you wish more founders just knew?

I think the first people to invest were just angels I already knew, like the fellow founders actually put in for our initial round. One of them put in a very large, like $350K or maybe even more, but he took a massive bet on us, which I'm super grateful for, cuz I think like super early on it was hard explaining this idea of how central data would be to the AI world, especially if you're also trying to talk about crypto. People are like, what is this? How do these things connect? Like, how does it even work? And so honestly, the people who took the earliest bet were just betting on, they were like, we think Anna will figure this out. Maybe there was some view on, okay, data it seems like a good category. But I think that honestly, I look back at my decks from then, 'cause I've been helping founders building on top of Ona these days raise, and I'm like, I don't know if I should send them this. This is like such a bad deck. Like, I don't know if this is a good reference. And so yeah, those are some of the earliest people. I mean, Polychain was the first institutional fund to take a bet 'cause they led our seed and they had a lot of context on what works from a decentralized perspective and then had seen a few early things in AI play out. And so I think that they were just willing to take that bet. From the perspective of things that could surprise you about fundraising, ultimately you wanna bring like the right people on board. I think that there's a phase where you're just like, I just want anyone, you know, like where you're just here kind of grinding through a raise and you're like, I'll take anything. You actually really do wanna be careful about who you bring on because they're like shaping the future of the company. It's important that the future that you see is similar to the future that they see, cuz you are probably gonna build towards the future you see. So, I think that actually being mindful, and this I think came up because Vana is a decentralized protocol. There's a way that you could build a very large Web2 data market that would be a centralized company, but that's not what I'm building. And so, I think there were cases where there were funds we would talk to that wouldn't necessarily see that. And I think honestly, in most cases, they just ended up passing. It wasn't a question of not choosing to work with them, but investors are just like human beings on their journey. What's their goal? What are they trying to do? How can they help push us forward? They're actually managing someone else's money. What have they told their LPs? Because now I've been helping like some of the funds that invested in Vana. I'll do like founder reference checks with their LPs where I'm helping. And it's helped me see that there's actually another level to this where the venture fund is providing a service to the LPs. And so they have to frame your company in a way that resonates with the LPs who are actually putting the money in, who are, they don't know tech at all. It's like a family office or like endowment of universities. I would just think about the incentive structures, what's going on in the system, and just generally like bringing the right people in. I would really make sure that you are on the same page with the folks that you're bringing on about what the path generally could look like over the next 5 to 10 years. I think the best investors actually help you be like more ambitious. I remember one of the first conversations I had with Paradigm. I had been raising, and a lot of the questions I was getting about data early on were like, how's this gonna work? This could go wrong, this could go wrong. And Paradigm was just like, like, okay, cool. So what happens if everything works? Then what? And I was like, wow, that's awesome. Yeah, let me tell you about that. And so I think that finding people that help you think like that, and it depends on who you are and what you vibe with and what it is, those are the folks you want on board. So yeah, just feel it out and get the right folks in.

Thank you for sharing that story with us and how you've navigated the dynamics and also the importance of humanizing your investors. I think is true. There is this almost like mythical component when you're a first-time founder that becomes unwound found as you become more established. And now it's been a few years since you initially raised for Vana. And if I take a step back, I feel like my time as an investor has been driven by these two hype cycles, crypto and AI. You're really building at the intersection of those. And a huge part of what we love to do with the podcast is tell stories. And sometimes the way you tell your story has to shift based on what an investor is looking for, like looking to hear. I'm curious just how you've navigated with this truism of unlocking data in a decentralized way across whatever data needs might be, whether it be for AI or for crypto or otherwise. How have you continued to constantly evolve your story and tell that both to customers and to investors?

Depending on who you're talking to, the story should be very different. If I'm talking to a technical builder who could start a data DAO on Vana versus a non-technical builder, I would say very different things. And the same thing is true if you're talking to, um, the Web2 AI kind of investor who's coming from that background versus Web3. So one thing I used to do more is I would try to convince Web2 investors who weren't that sold on the thesis of crypto, this is why decentralization is great. And I actually have come to the conclusion that I don't really think you can change investors' minds. They're basically just investing along some thesis. You need to either fit into that thesis or you don't. I guess I would frame it along the lines of their thesis, but maintaining truth to what we're doing, which is super important. And I think that there are cases, especially raising from some of the big Web2 AI funds, which we've had conversations with, and I'm friends, you know, those are great people. Often it, it actually might not make sense for them to take a bet on Mana because they like don't necessarily see the value of decentralized networks. And if you haven't spent time, especially in crypto, cuz it's such a different industry, I can be very confusing framing it to anyone, whether they're coming from an AI background or a crypto background. It should be very easy to understand. Like they should be able to go and explain it to like their grandma or like someone, you know, like they, I actually should make it simple. And I think that often when you have these really complicated technologies that people aren't that familiar with, you can get stuck where you're like up in those technical details, but you should really just make sure, okay, this is the 10,000-foot, like bird's-eye view of what we're trying to do. It's really hard. I think it's also hard because as a founder, often you're kind of living in this future that you see as inevitable, but other people don't see that yet. We exist in the present right now, and then you're like building towards this thing that you want and that you see really clearly, and you have to like either talk to people who share a similar view of a future. That's like the easiest pitch to do. Or you have to like paint this view of the future, which can be really hard because is you live so much in that future that you're not even really aware of the other perspective. And I think that was especially true for us 2 years back, 3 years back, where they were like, what are you talk— like, I think that for any raise, you're gonna get a lot of nos. And I think our first raise, I think we probably got 40 nos before we actually started to get something coming together. So yeah, just maintaining that conviction where you're like, I know this is what we're doing, even if nobody gets this right now, or like 1% of people we talk to. I trust that at some point they will see, and at some point we'll get them over here.

Let's touch a little bit on some of those big milestones that have happened since. I think we read that there's been over 300 Data DAOs that have been created on Ivana. Congratulations. Walk us through some of these more product and company-wide milestones that you're excited about as you look into 2025 and beyond.

These days it's a lot about the success of the Data DAOs, them reaching these different milestones. Actually, over the weekend, the car Data DAO hit 5,000 drivers who are contributing data, which isn't like a million people, but a lot of them are Tesla owners. That's actually like a very good kind of subset of the dataset. And they are the first data DAO, which is venture-backed. So that was a big milestone for us of starting to see that people can build new kinds of companies on top of Vana, which then themselves are commercializable, venture-backable. So I think that's one thing that I'm spending a lot of time on in the ecosystem, which is like, how do you help help great founders build really strong companies around this like user-owned data. So I think that there'll probably be something really interesting in like the health category, something in robotics, I think. So those are two that I would call out that I'm particularly excited about. You mentioned this earlier, this idea of DataFi, right? You might be familiar with DeFi, which is like decentralized finance, being able to, yeah, trade like different tokens and build different financial primitives in a decentralized context, like lending and loans and yield, those sorts of of things. What's cool about VANA is you can like build that on top of data as the underlying asset in the same way that companies will buy like oil futures today, right? If you buy a bunch of oil, you kind of want to de-risk that. I think in the future we could see like data futures or like these different, almost like an index fund for data. That's still super early, but is a sort of space which is data fi coming alive as actually like a financial asset class. That I think is really interesting. Insurance products to be built on top of that, yield products, et cetera. And then the last thing I would mention is data portability on VANA. So when you connect your data to a DataDAO, you're both granting access to this union group to make decisions over your data, but you also just have a copy of your data, which is connected to your wallet, which you control. Then you can log into any application, say a health app, and then all of your sleep data is there. All of your message data is there in a privacy-preserving way. So you can build really cool applications. And so from a developer perspective, I think that we'll see a lot of personalized AI type cases, use cases possible on top of Bana too.

Super helpful. My brain is going in so many different directions cuz you're unlocking all of these other concepts that I hadn't even thought about in terms of your point of a developer ecosystem that empowers this to the extent that the concept of decentralization is a crypto-native concept and it has to intersect in to your point, more of a Web2 evolution of AI, like what degree of technical feasibility from the crypto-native perspective are you seeing these builders? How are you thinking about the broader quote unquote Web3 ecosystem having evolved since what I would consider like the peak of the hype cycle, like 2 years ago?

Actually, a lot of the people who are building Datadas or Anvana are not necessarily coming from a crypto background, so they're not necessarily Web3 developers or Solidity developers. Some of them are more like just data people who love data data and are like, oh, like I wanna build around this dataset, or I wanna build around that one. Also, like we've tried to make it relatively easy where if you don't know how to write smart contracts, you can actually still build a data DAO. A lot of the involved work is in Python. Maybe also a year out, none of that will even matter cuz AI can just make it very easy.

Like a Replit for your data DAO creation.

Yeah, that would be great. We should actually build that. I like that. One thing I would call out is I think that in Silicon Valley, if you look at 5 years ago, what was required to build a company, it was much more of a technical undertaking, right? A lot more of the risk was technical. And these days a lot more is like distribution, go-to-market. And that's something that we're seeing with the data DAOs and some of the founders that are building on Vana too, where the technical piece is really important, but the go-to-market piece is also very important. How are you gonna get users contributing their data? How are you gonna get companies and researchers actually accessing that data? That distribution piece has become more and more important for successful founders to think about. It still can be helpful to be technical, right? Like you can think about things things in more of a systems-based lens, but it's easy to forget.

I know we're coming up on the end of our conversation here, and so we've spoken about a ton of interesting concepts that Vana is pioneering and really an entire ecosystem that others are building on top of Vana. And so, we've spoken a little bit about the past for Vana and the founding story. What is something that you are most excited for that Vana will be launching potentially later this year?

The example I mentioned earlier of a foundation model where you have 100 million users contribute their data, that can sometimes sound very like pie in the sky, like, oh, when's this gonna happen? But there is a project that will make that very tactical where we're actually doing full pre-training of building a foundation model across a bunch of the different data DAOs that I'm really excited about. And the reason why that's interesting is that it actually shows that if you have people who contribute this data, you really can build a meaningful alternative to centralized AI. And it really is putting that test in place. Is this path even possible? That's one thing I'm really excited about. We're building this better version of data access where right now, if you wanna access data from a given DataDAO, you have to talk to each DataDAO. So if you wanna access data from 5 different DataDAOs, you have to like go put forward a proposal for all 5 of them. And so we're working on ways to make that more programmatic so that as a developer, you can just access the data that you need without needing to go through that. The reason why that's interesting is I think it's really nice for individual developers to be able to come in, train an AI model. That's like a little more of like a technical product point. And then maybe one other thing I'd call out is we're working on like a unified interface to contribute to a bunch of the different data DAOs. Right now, you have to go to all these different UIs, and so you're hopping between different applications that people have built. And we were seeing that people are replicating work, each building out similar things. So, we're working on something to make that more streamlined. It feels like a data wallet. You can just put your data in there, and it makes it a lot easier.

These are incredibly exciting things that are coming up. We have to have you on the podcast next year because I think these days roadmaps that seem like pie in the sky, as you mentioned, are actually very much happening quickly and quicker than we expect. So, that's incredibly cool. Anna, we're gonna ask you our hero question for the podcast. Who is a woman in your life that has had a profound profound impact on you and your career?

You know, so I don't know Janet Yellen personally. I've literally never met her. I should really send her a note or something. But Janet Yellen. Yeah, yeah, totally. Janet Yellen. Like, so grateful. Like, she set an amazing example for me. Maybe Regina Barzilay, who's one of the people who got me into AI at CSAIL. So, she was really influential as well.

Thank you so much for being on the podcast, Anna, and telling the story. We're excited and it's been such a pleasure.

Awesome. Yeah.

Thank you so much for joining us at The Room Podcast. If you want more from The Room every week, subscribe to our newsletter at theroompodcast.com/newsletter. We'll be back next week with a new episode and inspirational guest.

Tuesday, 10 AM Eastern, 7 AM Pacific.

See you in the room. Perkins Coie supports the most innovative entrepreneurs and investors in fast-moving and high-growth sectors. Entrepreneurs addressing their myriad of legal needs. But the firm doesn't just provide end-to-end legal and business counseling to its startup clients. It also facilitates introductions to key advisors and sources of capital. Perkins Coie's interactive website, Startup Percolator, offers access to programs, resources, and rich dynamic content designed to assist entrepreneurs on their startup journey. To learn more, go to startuppercolator.com. And perkinscoie.com.

This podcast is brought to you by Mercury, the banking platform businesses like The Room Podcast use to simplify their finances. It's time banking did more than hold your money. Now it can. With Mercury, you can pay bills in seconds, close the books faster, and even send invoices. Not only does Mercury do away with a patchwork of tools, it eliminates guesswork. Giving you complete and accurate visibility into your business's finances, all from one account. Apply in minutes at mercury.com.