Sold a Company at 16, Raised $3M at 19 | Dhravya Shah, Supermemory

I don't want to wake up every day and argue with someone about, okay, this is what we should do next. My first 2 companies actually died because of a co-founder breakup where one of the co-founders went rogue because of money or power or whatever, and we had to sell the business off minutes before they destroyed everything. When I talk to enterprises, all they need is sense of control and sense of ownership. The benefit of SuperMemory is the fact that it's an infrastructure horizontal product, because then I can give them maximum sense of ownership and sense of control where they truly own the data, they truly can control it. It's a It's a journey where no one is there to keep you accountable for these things. If you just don't do it, there's no one coming and holding you about it.

What's the case for solo founding?

Being a solo founder is the best way to build something that's true to yourself. That only happens when you have extreme conviction about what you're building. All of these solo companies are literally representations of the founder as a company.

Your story is one of not necessarily rushing into building a company, But following sort of this thread in this interest area of yours to something that eventually became a company. So why don't we start there? How did Supermemory become a company?

I used to read a lot of these startup books and essentially all of them would say, okay, you should have a vision of the world of what it will be in the future. And then you kind of build something to make it true. At Cloudflare, and like a little before I was at Cloudflare, you know, open source models started becoming a big thing back then. And there was only one function calling model, which was the Noose Hermes model. And like I was literally put in a position to work only with the small models because we were doing inference for open source models at the time at Cloudflare. And I had this side project for collecting all personal context because I essentially knew that personal context will become a very important thing in the future where everyone will have their own AI agent just like everyone has their own mobile phones. And in that world, the most important thing will be context about you. And originally this started off, okay, like the way to collect context is bookmarks, notes, things you save, things you look at on the internet. And then I kind of expanded more and tried to do more and more with Apple Shortcuts and stuff like that. And this was all as a side project to collect all context about me and then provide a way for agents to look at it and reason through it and then help me, help understand me and help me think about things better. Then as I would work on this more and more and I open sourced it, and I think in the first 2 weeks of open sourcing this project that I was working on, it was called AnyContext at the time. It absolutely blew up. I had this tweet which essentially said, okay, this is the architecture diagram of any context. That tweet got like 500,000 views. It was just an image. And what was interesting is I was still a college student back then, was just excited about this industry that no one else seems to be excited about. And I had 100,000 people trying out my app. In different ways, and I had no money. I was actually negative bank balance, like quite literally. I was borrowing money from my friends to just survive. And it made sense for me to kind of optimize the architecture and keep it free while providing the best personal context agent out there. So that became AnyContext. Then I optimized everything. I made another tweet saying something like, okay, my entire infrastructure for 100,000 users can run on $5 per month. Here's how. That went viral, so I became kind of like a micro influencer in the field of context infrastructure. And I was going and helping out startups build this out, uh, for free. Um, so someone flew me to New York, someone flew me to SF, and I would come to all these places. I would go into offices and I would help them do that. And that was kind of like the inception of it, and it started to make more and more sense that infrastructure For agents, especially for agents, context will be as important as inference for agents. So yeah.

So it's a really interesting sort of story and we'll talk more about kind of what that actually— how you kind of went from that part. We'll put a bookmark there to like where you are today, which is crazy as well. But it's just when you hear that, you don't hear somebody saying, I was trying to figure out an idea to turn into a startup that I could raise money for. And I think it's really important because obviously great companies can come from somebody who's actively trying to figure out a great company to start. I'm not saying that that's impossible, but at the same time, it feels like that's sort of the default assumption is that people say, I wanna start a venture-backed company and I wanna go and build a venture-backed company. And my suspicion is that at the very early days of any context, if you had said, for whatever reason, I wanna start a venture-backed company, maybe you would've just come up with some other random idea before coming up with AnyContext. So can you share a little bit more about what was happening in your world at the time? I guess you were at ASU at this point. That's right. There are some great people who have come out of it and done really interesting things, but at least in terms of most people's points of view, they're not thinking about that's the place where startups get started, really. So can you share a little bit more about what your environment was like at this point while you were working on AnyContext or even right before that?

Yeah, um, I was working on like a lot of different random projects because I had like a lot of, you know, things that excited me in technology. Um, and I was also kind of lost, so I was just like doing whatever I, I could. And when I launched the first version of AnyContext, a bunch of people even at ASU were disregarding it as, dude, like, okay, like, what are you doing? It's a note-taking app, like, it's fine. Um, and at the time it was, so they were right. Um, but I was never around a lot of people that were in the startup ecosystem, in the startup world. And I still remember when, uh, the first version was blowing up on GitHub, it like in 2 or 3 days it got like thousands of stars. So a bunch of VCs were reaching out to me and I was taking these meetings and they were like, okay, are you raising money? And I'm like, bro, like this makes no sense to me. Uh, you're telling me I can raise $2 million right now? I don't think that is the case. Um, and, uh, that was never even a question or a goal for me. The goal was to build this version of the world that I think must be true. Um, so like, that's where— that's, that's my background. And, uh, I knew very few startup founders And no one would tell me about Fundraise. Actually, I came to know about it when you told me about it. So yeah, I was completely out of touch with all of that.

So maybe just a little bit more of the backstory here. You get to ASU, and how did you actually get there? And then you somehow get involved with Cloudflare. What's the story of getting from India, your hometown, to getting to ASU and then you know, somehow getting a job, like, you know, working at least part-time while you were in school at Cloudflare with Dane. Like, that's, that's a very unusual thing to have happen.

Yeah, uh, I think I just got lucky in most of the, uh, parts. But when I was in India, I was building 2 companies. One was a social media automation tool of some sort, and one was a Discord bot hosting platform, which is actually very cool because I was in infrastructure. It was kind of like sandboxes in that era. It was exciting work. Both of them got acquired and that kind of opened the doors to come into the US for me. I came to the US as kind of a lost student. I did not quite know what I would be doing. In fact, I didn't even tell any of my peers that I code. I was completely normal and people found out through Twitter and stuff like that, but people did not know. Um, that I'm doing anything. And I would just sit in the library, work on things that excite me. Um, and I built a bunch of things in pretty much every market you can imagine. Like, I built something in consumer, I built a database that went pretty viral, I built a meme platform to make and edit memes. Like, pretty much everything you can imagine, all the standard ideas. I built a note-taking app like Notion. All of that stuff. And I was just like, okay, uh, I was launching these things and none of these were startup ideas, you know. I never approached them as startup ideas. I just approached them as something cool that I made.

It was fun. It's play.

Yeah, because I think, um, it was kind of my way of doing art. You know how people make paintings? So I would just make like something that excites me, um, even if it's shit, even it's like just a drawing. But I would just put it out there and everything was open source as well. And I would never put anything under a paywall. And a bunch of people took interest and people were like, okay, this guy is building and is really excited about what he's doing. And I kind of built an audience doing that, which really helped me when I launched something that I was really excited about, which was AnyContext. And I made this tweet saying like, you know, here's the code for something that probably is worth a million dollars or something like that. And that was the first tweet about Super Memory. And everyone was like, you know, cheering me on. And then half the people were like, no, this is not. But that only happened because I was doing so many random things for fun. And even this was for fun. So people were excited about me as a builder that created a lot of value in anything I say because people really think of me as someone who's just in for the love of the game.

What's your advice for people who are maybe kind of feeling this push or this pull towards trying to jump straight to the startup? It's like, I think there are probably a lot of people who are maybe in the position that you were in when you just got to college. But they're feeling this pull, maybe it's their environment, maybe it's what they see on X, maybe it's sort of this whole wave with AI feeling like things are just moving so quickly. They feel this pull towards like, I should try and figure out something that's making lots of money or that's raising a bunch of money and doing some sort of high status thing with VCs. What's your advice to them? Because you did something very different and I don't see that many people do it. I think it's a shame that people don't do it because I think that ultimately, as you're saying, like you're creating your art and then ultimately through that, maybe you come up with like the actual thing that like becomes like your, your maybe like longer-term work. Um, but what's, what's your advice to them if they're, if they're sort of going through this?

Um, so my first summer at Cloudflare when I was interning there, I remember a VC reached out to me and they were like, we are starting a residency program. If you can fly out here tomorrow, we'll give you X amount of money and place to live and all of these things. And I remember I asked Dane what I should do. And so Dane gave me this— like, Dane is the CTO of Cloudflare. I was really close with him. I was working directly with him because one of these projects led to me joining Cloudflare at some point. And he essentially said you should either raise from tier 1 or bootstrap, or you essentially— you should make something that makes so much sense that you don't even have to ask this question. Um, and so, you know, like, I think that when, especially as a solo founder, when you're building something that makes complete sense, there is no lack of capital in the world to make it big. But to get to that initial inflection point, that is the hard work. And if you raise before that inflection point, it'll actually get you in more trouble because you'll have to figure out, and you also not know who to raise from. So I would just say the best way to think about whether you should go with the VC route or not is Will money actually change the way I approach solving this problem, or could I do more things with it, or am I doing it for the wrong reasons? Do I have the distribution? Do I have the technical stuff needed to do it? Um, which I thought, like, when I was raising, I was in a very good position in all of these. If I did not raise, it would have been fine. Um, so So yeah, that's how I think about it.

It's interesting too, because in a way, I think that those are the exact right questions to be asking. And sometimes I feel like people are not even caring about those things. The reason they're thinking about raising or hoping to raise is because they feel like that will give them some level of validation versus what you're saying, which is the fundraise is sort of, it doesn't even matter if you fundraise because you already have all of these things. The fundraise is sort of a result of all of these other things. And I think that sometimes people almost hope that the fundraise solves all of the other things. Yeah. When it's sort of the opposite is true. I'd be really curious to hear a little bit more about kind of, you've got a bunch of people reaching out to you at various points of the journey. And I think that that probably happens to some people who are doing interesting things in public. VCs, it's kind of their job to be kind of like homing missiles to try and figure out where they can put money. But that also kind of pushes people maybe to start companies for the wrong reason or at the wrong time. Yeah. So I'm curious what Dane said at that point when you sort of had that one offer and sort of how you kind of considered when it actually made sense to really build a company, because I think that that would be helpful for other people to hear because they might experience similar things.

Yeah, so the first offer I got from any VC was in 2024. It was the starting of summer 2024, and since from starting of summer 2024 to summer 2025, I might have gotten on like hundreds of calls, and I did not realize I'm wasting my time back then. I was just like, okay, I want to show them what I'm doing. Um, because I had heard all of these cool names like a16z and like all of these amazing names, and I wanted them to see the exciting work that is going on here. And I think everyone, like, you know, I would just get on call, I was like, no, I'm not raising. And for an entire 9 months, I essentially said no to everyone, um, until I was at a point where I kind of had crystallized exactly what Super Memory will be and what steps I can take to get there. Because back then I was actually doing a consumer app, and if I would raise for that, I would actually be in a messed up situation right now because the VCs I would raise from would be for consumer. It would be hard to pivot. Um, there would be like, you know, it's hard to change a ship that's big ship. So it's actually good that I was small and said so many nos. And when I had a crystallized view, I just raised from the best angels and VCs that were perfect for my company and what I was doing.

And maybe you could share a little bit more actually about that, because there's this really cool journey of going from all of these sort of passion projects, interesting sort of things that you just put out there just to see what would happen and because you were interested in making them, through to Anycontext. through to what eventually became Supermemory. And then Supermemory even was started as this sort of consumer, sort of almost like a personal knowledge management, personal knowledge graph type product, to then you realized, oh wait, this is actually maybe something that I should be building more from the infrastructure side of things. How did you make that decision? And maybe talk more about how you actually made that decision in conjunction with deciding you were going all in on building a business? Because I think that oftentimes people stop with like, they take the first thing and they just stop there and they're like, that's the thing. But they don't do that necessary sort of like evolution of it to where you are now, which I think is, as you were saying, a much better place to be. Maybe if you could take people through that so that they could maybe learn to look for that themselves.

Yeah, I think there's like a very beautiful way in which nature evolves. You know how like flowers grow and they go through different phases and whatever. When I started out with Super Memory, I was very confident about the fact that this could be big. And I remember when, uh, you know, like when Perplexity would launch something or Google would launch something, I would be like, okay, Super Memory is actually better in this, this, this way, so I should be much bigger. But then I realized that it's very hard to compete because it's not just about product, it's about distribution, It's about positioning. And I was just like one guy doing this. I was kind of delusional for the first 3-4 months, uh, because I was like getting some attention from it. And you know, things were going good. And there's like the inventor's dilemma, you know, you— things are going well, why not? Why change it at all? Then one day I kind of realized that sure, things are going on well here, but again, like, what is this version of the world in which super memory becomes a default way to do things? And I could not really figure out what that would be. Like, would I— like, you know, uh, why would someone not use Notion as their personal context or something like that, right? And I could not figure out how I could be uniquely positioned in my own skills, in my product, in my positioning to do this. And I could only figure out after, you know, like kind of doing, doing small iterations and big iterations, like the entire journey from the entire pivot has also been in very small iterations actually. And I would say we are going through one more pivot right now, um, from just doing memory to context cloud in general. So I think like this evolution is super important at all times and getting away from the inventor's dilemma is the way to win. I think there's like the Vercel CTO quote, which is iteration velocity is the answer to all problems in your life. So yeah.

If you're enjoying this conversation, then perhaps you're starting your own company. And if so, you should apply for the Solo Founders Program. I've spent the last 7 years working with startups and founders who are figuring things out. And I'll tell you, the ones that make the most progress are the ones who make sure that they have relationships that can help them go the distance. The Solo Founders Program is about bringing 10 solo founders together so that they can build solo together versus building solo alone. I'll work with you, we'll invest $100,000, and we'll make sure that you make the most progress that you possibly can in 3 months. If that sounds interesting, you should learn more and apply at solofounders.com.

Maybe you could walk through post-decision to be a B2B-focused company. Maybe you can walk through what's happened since and to where we are today with the big announcement. I think that the first thing that you announced on the B2B side of things, if I remember correctly, coincided with the move-in day for SFP.

Yes,

that's right. So what was happening on that day? And what's happened since?

Yeah, so the day I moved into SFP, I launched a router that automatically edits your requests going out to LLMs to be more personalized and have more context about the user that the LLM is talking to. It was a good idea, but I realized that no one wants a router on top of the routers they're already using. it makes no sense for, you know, most companies to buy in because they would have to send their API keys through me, and I'm not a trusted player yet. So then I had to, you know, kind of do some research, like real research, to prove the value of this router. And I did publish the research and stuff like that. That went really well. But the output from that research was the fact that, okay, Um, you know, people want better context and context management, but people don't want to do it in this format. So then I put— made like an SDK for people to use. And the SDK back then was very simple. It was kind of just retrieval, um, and you would just put in documents, you would search documents, something like that, right? And I realized that, okay, people do retrieval, but why would someone use me for retrieval, not the 1,000 other people who do retrieval. So I introduced this new way of doing things with the model context, but the actual technology was nothing interesting. It was just something that anyone can do. Then we moved more to the memory stuff, and we were only doing memory for a while, which is forming it into a graph, and then the graph evolves, and then we provide that context to the model in some way. When we only did memory, people had a bunch of problems around cost, around quality, around the things that they could do to make it perfect for their use case. And so I had to put in pretty much all configurations they would need to do into the system to make this work for them. And slowly I brought back the retrieval. So it was retrieval and memory and then People started talking about file systems being the way to do it. And a lot of people come in with their own opinions about, okay, here's how I really need to do it. I don't know what technology to build it on. But we already had a file system thing, so we introduced file systems. We introduced a bunch of other products that people typically do, like for extremely personal agents, there's profiles and stuff like that. And we kind of had to be this horizontalized non-opinionated stack, which then later we now call the Context Cloud. And we got here one by one by perfecting each product along the way. Now that we have the Context Cloud, people love it. My entire calendar is filled with people trying to use it and stuff like that. But when I talk to enterprises, all they need is sense of control. and sense of ownership because they want to not only own the data, but they wanted to make it governed and exactly for them and stuff like that. In order to get to that level of control, they need to try it out because we know that Super Memory provides that control, but they need to make sure that they can run a POC, they can run some tests that make it perfect for them. Now tomorrow we are launching a self-hostable version of Super Memory. which anyone can use to run the entire stack locally. And it's kind of like we hate money, but we think that it will show you the power of SuperMemory, and it'll be a very useful tool for us to, you know, show people how powerful the cloud is. And then, you know, that'll be one more way to use us.

Now you said tomorrow, but God willing, we're Putting this podcast live tomorrow, so it's if you're listening to this now, it's happening today. Again, again, we're hoping to turn this thing around and get this out so that people who are listening to this are listening to it right as as this this launch happens.

We don't like making money. Today we are announcing the most requested feature in super memory history. Something that everyone has been asking for for so long, we had to do it. You can now run SuperMemory completely locally. The cool part about it is that it includes everything: an embedding model, a graph database, a vector database, all the things you need for context. The stuff we run on real SuperMemory Cloud is now run locally, and all you need to do to run it is npx supermemory local. No setup required. And as soon as you do that, it will open up instructions for using it as an API or through Claude Code, OpenCLaw, whatever you use. Wait, I forgot to mention the best part. You can run it everywhere on any operating system. Linux, on your Mac Mini, Windows, Railway, Hedgner, anywhere your agents run, SuperMemory Local can run. So yeah, Super Memory Local is here. It's fully yours. No API key, nothing. I'm excited to see what you build with it.

Maybe you can share a little bit more about the philosophy around open sourcing the binary. And then we can talk a little bit more about sort of the learnings to date and how that could be something that other people could apply to when they're going through and building their startup.

Yeah. So when you, like, you know, in my version of the world in the future, my basic point is inference will be run everywhere, locally on device. It will be on people's own premises. It'll be on thousands of inference providers, on the model labs, inference everywhere. And the data either has to be on a third-party system, the context itself needs to be on third party, or the first party. We got the third party covered. SuperMemory is the system that can hold the context that works with every model and can be used in different ways for any use case. But the first party, we did not get right before, which now we will because you can just run SuperMemory locally with your local models and you own the entire stack end-to-end. if models can be run locally, why can memory not be run locally as well, uh, completely self-contained? So that's the philosophy of the binary. And it also helps us have an easy ramp into the enterprises because now the way to use and own your own data is to just run a command in your CLI, and then you can reach out to us when you want to run it at scale. So yeah, that's kind of The thing that we are going for.

It's really exciting. I think that one thing that's really notable about you and Supermemory is both the speed at which you iterate, and it feels like you're constantly launching, which it doesn't just feel like it, you are constantly launching, but also your willingness and openness to continue to push forward with new things and kind of either throw things out or move beyond them. Because I remember the day you moved into SFP when you made this announcement and launched this version, it got an absurd amount of interest and traction. Yeah. And I think a lot of people would actually take that and they would say, okay, that's validation. We're good. We're locked in on this. And you were always, and continue to be, pushing forward to figure out what is the thing that's right for for now, because ultimately what we built then was good for then, but we want to keep moving forward. We want to move the industry forward. We should talk about evals at some point and sort of the system that you've come up with there. But ultimately, you don't feel very married to the present. You're always pushing to the future. I'm curious, sort of, if you have any reflections on that, the philosophy there, because it seems very right, but at the same time, it seems to be a thing that a lot of people tend to struggle with, especially when they get the validation of all the likes and all the— Yeah. Because I think you got millions of impressions on that post when you moved in SFP, and I think everybody was kind of— and it was also in the middle of your fundraise. So can you share a little bit more about the philosophy there and how how you don't end up getting sort of like maybe overly married to things because you get that sort of external validation.

Yeah, so the story is actually very interesting, right? I was in the middle of my fundraise pivoting to B2B at the same time. And in my fundraise deck, there were literally a slide which was like, we kind of do B2C still, but we also do B2B. And that day we launched the, the router, which got 500K impressions, lots of organic views, everything organic in one day, and millions later. Um, and I still remember I actually thought this is it, like this is the way to do things, etc. And again, like I never thought that I need to do anything more to make this successful. But sometimes you have to realize, like step back and realize that, okay, this thing that I'm doing is not received the same way I'm imagining it. So we still have users for the router. We could not discontinue it, but it's completely buried, right? Like people don't know how they can use it. We also had to shut down the original version of the consumer app and we had to kind of get to a better like, you know, build a new vision of the world of what the consumer app should look like, should be, which is a completely background context engine instead of like having a lot of things on the— you have to do on the app itself. The important part to realize here is not just the numbers, it's the vibes, which is if I feel like this is blowing up right now and this is not the future, I'll just end it. But also if I feel like something is the future, which is people will be using file systems, I will be building it.

Even if it doesn't get as much of an initial pop.

Yeah, exactly. So right now, for example, very few people use our Soup Memory file system, but we think that it is, you know, an ideal version of where context will be. And, um, yeah, it's again, it's It's kind of a philosophy-based approach. Um, yeah, like there's also a story, you know, there was IBM that was absolutely blowing up in the world, was the greatest company in the world, and they had the inventor's dilemma of like, okay, we built custom machines for each enterprise in the world, why should we do anything else, right? We are perfect. And then the next person who led IBM, which is son of the founder He was like, no, no, this makes no sense. We should kill everything. He killed the company for 2 years. He built the IBM 360, which then became the next big thing that made IBM the best company in the world again. So yeah.

I think it might be interesting to talk a little bit about team at this point because, um, you know, you've got a great team, um, and solo founders are not solo forever. In fact, like, in some ways they need to bring on teammates earlier than co-founded companies because when you have co-founders, that's your initial team, right? So what was it like sort of bringing on the first people? What was the dynamic of that? What are things that you did right? What are things that you did wrong with hiring?

I think my philosophy has always been bring true builders, even if not non-technical, they should be truly you know, invested in the craft of the way things are done, fun to work with, and someone who— with extremely high autonomy. Um, and people like that are quite easy to find in the open source community because they are true builders, because they're doing things for free. That's where I found some of my earlier employees. Like, there's Mahesh, there's Shreyansh, all of them essentially just started contributing. We also hired 2 more that was just contributing to our open source work. And I think that's like a great magnet for people. And how I think about building the team is also when they do come in, you should give them all the tools necessary for them to be kind of the lead, and they should feel like this is something that they also own. I have a lot of team members right now. You know, I've done a lot of things right, like I have some amazing employees. Some of the things I did not do right before was sometimes you would give all the resources to someone and you would not kind of realize that they are not doing the right work, or they are not being nice, or not fun to work with. And in that case, you need to get rid of them as soon as possible because that brings the entire team dynamic down. And I did do that as well. Um, I think in the course of Soap Memory, there have been 30 people who have worked with us. Um, and now we are a team of 7 core members. Um, so yeah, it's, it's been interesting. And pretty much all the 7 of them have been here for a year. Hmm. So it's incredible. Yeah.

When you were fundraising, what was that experience like once you decided that you were going to fundraise? Because I think I remember, if I recall correctly, we were talking after ODF, we were getting ready to launch the first cohort of SFP, and you and I were sort of riffing on sort of like the fundraising narrative. But I think that at that point you were still on, I think you just said earlier about this, you had a slide for talking about sort of the B2C, But a large part of the narrative was starting to shift into, and then we realized a lot of people actually wanted this sort of stuff that we were building that allowed us to do the B2C stuff. They actually wanted it for their own applications. Yeah. But could you share a little bit more about what the actual fundraise was like in terms of what you learned throughout that, and then also what you learned about how to tell a story?

I got really lucky in my fundraise as well, to be honest, because I was very online. Um, and I think that's actually one of my superpowers, that I am very honest and true in my online self.

Yeah.

And, um, that led to a few people back in the day, like, you know, some really good angels to be connected with me and my journey and what I was building. So I had a few messages like when you're raising, let me know. I'll be happy to put in money. That was actually very helpful because it gave me a validation, but I never replied to those messages. I was like, I'm not raising right now, but I love that you said that. Yeah. Um, and when I did start raising, like, uh, the point when I decided to do it was, um, like, first of all, the entire world kind of started changing, um, Claude was starting to becoming big back then. At the same time, it was kind of obvious that ChatGPT is not the only big way to use AI. So from a consumer angle, that, that was amazing, right? Like, oh, now there's 2, 2 major or 3 major ways to do it. And the thing that triggered it was I was in Vegas and one of the VCs was so excited that actually flew to Vegas And, uh, I mean, I met them at a conference and I also had an a16z meeting at the same exact time, um, online. And both of them went incredibly well, but not well in terms of, okay, we will invest in you, but in the— in well in terms of this is so obviously the future, this is the way it should be. And my thinking is crystallized into my narrative, the way I'm going to do it. And I think Julian helped me a lot with that because when I was in college, I kind of was doing whatever made sense at the moment. I never thought of it as a company. But when I got into SFP, it started making sense that this will be a company. and the story around it. So yeah.

I think there's also this aspect of like, you were in college and like, you know, some people are very anti-college and they're like, oh, drop out or don't even go, right? But I don't think you're anti-college or I think you had a good experience from all we've talked about with ASU. You have lots of friends there. So you're not an anti-college person. So for you, dropping out wasn't necessarily a badge of honor or even something that you felt like you needed to do. And then of course, there was the visa, right? You needed to get a visa. And I know Noemie helped you a lot with the visa, but could you share a little bit more about sort of— and I think this is useful for other people to hear because there are probably a lot of people who are in similar situations where it's like both they're still in school and maybe they also have a visa that they need to apply for, like an O-1. Can you share a little bit more about what that was like and that decision just to help other people think through it?

Yeah, in college I was actually doing pretty well. I was top of the class for most of the time I was there. I had tons of fun, went to tons of parties, lived with Americans in my first year. It was probably the best time, and I would actually not leave college if it did not make so much sense. Um, that was never the goal, that has never been a thought, and even when I was racing, it was like I actually dropped out kind of before, like I made a decision to drop out before I even got the visa because it was that obvious, you know. I had also kind of dropped out from my dream of getting into the most prestigious place in IIT and the typical route. And even that was not like, it was kind of a leap of, you know, faith. I did not know what I was doing, but that turned out well in high school. So I always had this belief in myself that, you know, whatever it is, I will figure it out. I also had some cushioning because I was in Cloudflare before. I had done some work there. I knew that if I, you know, if I, um, don't make it, um, At the time, my thought was, uh, it's going to be fine. Now that's no longer an option, obviously. Um, but yeah, like, that is all the considerations that were in my head at the time. I remember also talking to my parents. They were essentially saying like, no, no, you should not drop out, you should have like this certificate. And, um, my brother was also like, okay, you should not drop out. All of these VCs are here to, you know, like money launderer. Like, it just— it was just so insane that someone's giving me millions of dollars to work on this idea full-time that everyone I asked was saying no. Um, but in my head, I had so much belief that I will do it that I just decided to do it anyways, um, despite the visa constraint. So yeah.

That's awesome. Um, So, if I were to sort of highlight part of that, it's that you felt such a tremendous pull on this being the right thing for you to spend time on, not VCs wanna give me money, or it would be really cool to tell people that I was starting a company. Yeah. It was because you'd spent enough time on this as something that wasn't a company, But it eventually evolved into you realizing that this was the right thing for you to be spending time on and building as a company. Yeah. I think that's really important because oftentimes people don't go through that process. And I think it's often a failure mode for people, and they ultimately maybe drop out for the wrong reasons. But it seems like this was definitely the right thing for you at the time.

Yeah. And I'm actually glad I dropped out, and I still think For context, it had been 2 and a half years since the very first iteration of any context that I built. So I was working on this at that point for 2, 3 years now as just something that I'm very excited about. I had dived really deep into the tech. I had dived really deep into how this would look like, how I should approach this problem, and a lot of that is not true anymore because no one knows what they're doing. And even in AI, even in— especially in memory, no one knows the right answer to it. And at the time, I was also, um, thinking about, you know, how important it is to be in college and have the exploration, um, and how powerful that time is. It's like you're a free agent, you can do whatever you want. So what would you do? It's a very exciting time. I think everyone should do that in college.

And I want to take some time towards the end here where we talk about various people in your journey after you decided you wanted to make this company. So maybe talk about some people at SFP, talk about some of your investors. Before we do that, I think that one thing that I've noticed as sort of an outsider to the memory, like AI memory context space, is just how much FUD there is. Yeah. Could you share a little bit more about kind of— I'm trying to say it in a nice way, but just like the bullshit, essentially, that's out there about memory? And also sort of what evals actually should look like in this world for benchmarking. And also, what bad benchmarking looks like.

The reason why, as you mentioned, there's a lot of bullshit in this space specifically is because the barrier to entry is probably the lowest. Because think about inference or databases or all of these. They're really hard to start with. And anyone doing it is definitely, you know, extremely smart and motivated and is doing it for no reason but something that they know is correct. And it's also very deterministic to measure how these things perform. In memory, it's not that deterministic. A lot of people almost have AI psychosis where they think of something and then the AI builds something for them. And then the AI runs its whole own benchmark and it turns out to be the best thing. And people tweet about how their thing is the best. But the thing that people should really think about in— so there's benchmarks and non-benchmarks. Inside of benchmarks, you cannot just run one thing. You need to look at multiple different benchmarks. For example, there was MemPalace recently. It only measured against LongMemEval. And their code, like their benchmarking code, was completely different from their actual code. But the actual code literally optimized for the wordings inside of LongMemEval, and it built a graph. Like, they literally looked at LongMemEval and they built the system for that benchmark. Because it's— I think it's called the Goldhardt's Law. When you have one goal that you're optimizing towards, the goal is no longer relevant. Hmm. The second problem is that the setup itself looks very different, um, for different people evaluating it. Because it's not just about how you remember and store things, it's also about how you bring it back into the, the system, the context. How many tokens are you ingesting back, for example? Or is it a tool call or a hook? Um, is it, um, you know, in what, what model is used to look at the context and answer it? And then what judge is looking at it and measuring the quality, even if you use the same judge, it is not deterministic. So in this completely non-deterministic system, you should run like Beam, you should run LongMemEval, sure, on like a deterministic or as deterministic it can get environment. And we have built a tool for that. It's called MemoryBench. That will help you run a deterministic test, show you all the numbers related to it, like latency, quality, speed in terms of the answering model reasoning, but also the tokens given to the answering model and also recall at 5, recall at 15, like pretty much all numbers. And then when you post these numbers, you should be honest about it, which we do in all of our materials. But also when you Think about what's not in the benchmarks. It gets even more interesting. Because if I say something as simple as hi, what should a memory system do? Will it go and look up who is this user, which will not get any more information about me? So real systems are not built to perform well on benchmarks. Think about the best memory stuff you know, you know, poke, you know, like, uh, ChatGPT, Claude— these things perform really bad on these benchmarks. The reason is that they're built on something that's closer to the user, for the user, and not for showing some numbers on Twitter. And so that's what we, we try to do at all times. Like, that's what our products do. Like, Profile is for deep personalization. Uh, we have a playground that will help you make it perfect for your use case. We help the enterprises set up and benchmark it on their own data, on what they care about. And all of that is really important. And I think people don't think about that stuff a lot.

It seems like one thing that people really don't understand, at least the people who are building memory companies, is just how much the ability for the end business, the end developer to actually make decisions, to really be able to choose, to have sort of the autonomy to structure things themselves. It feels very much like a lot of the systems that are there are effectively hardcoded, and there's very little sort of ability for the teams that are evaluating using these systems to actually do anything on their own. And I think that More recently, you've described it almost as like what you're building is sort of the contrast to that, which is like these Lego blocks. Yeah. That you can arrange, you can choose to use some of the Lego blocks, you can choose to not use some of the other ones. Can you share a little bit more about your philosophy around sort of giving people, giving developers, giving companies the ability to have like much more flexibility and autonomy in how they actually develop these memory systems using your using your sort of pieces?

Yeah, so I remember talking to you about this back when I was thinking about it, and you mentioned this very cool thing, which is it's a Lego set with no instructions on what's perfect, because what's perfect you will decide and you will figure out for your use case. And designing it is kind of hard because this set has a type of piece which is configurable in itself. For example, in memory, one thing is, okay, what should the model learn? What should the ontology be? And how you want to structure it in the data. Second is how do you bring it back to the harness? So a lot of people think, okay, tool calls are the right approach or MCP is the right approach. A lot of people think hooks are the right approach. But let's say you're building a business agent. Right? You would want it to have some knowledge of the business and what's going on right now in the business inside Slack or whatever. You would also want it to know the documents, so you have to put retrieval into it as well. But you don't— and then you need to have memory, but you don't know what you did not learn, right? Mm-hmm. Because there's no way for any system to perfectly determine what the user will bring up in the future. So we have this thing called hybrid mode where if a memory does not show up, the document will show up. So all of these different configurations are possible with SuperMemory. Let's say you have tons of files and you want— you have some determinism in it, like it's kind of semi-organized files. But it's also not organized at the same time. So you would attach a file system to our agent and the file— the agent can go figure it out on its own. So there's like so many ways to do it and we enable every single way to do it in the cheapest, best format that's perfect for any use case.

What's the thing that really sort of gets customers excited about that?

Whenever I'm in sales calls, it's always interesting, right? Like they come in with some perfect way of doing it in their minds, and then they try to bring it up like, okay, can I do this? Can I do this? I have different users, I have different teams, the teams have different access controls in them. Can I build in my agent in a way where it follows these access controls? And for each of these questions, the most exciting thing is like whenever they have a question, I already have the next slide is about that thing. The most exciting thing recently has definitely been some of the research we have done in specifically in bringing this back to the harness. That has been very, very interesting because when we tell people that, okay, we learn everything, there's a model internally that is determining what is important that you can configure for your use case. and it's stored in our data structure, they get very excited because they know that they cannot build their own model, um, or they could not fine-tune a model because they don't have the data to do that. And we have it, and that's how we make it really cheap. So yeah, like all of these things that kind of show them the complexity of the problem then makes sense why, um, they should be building on top of our infrastructure instead of just bare metal. And

they also have that flexibility of being able to use some of the building blocks or using them in a certain way.

Exactly.

That's interesting. Okay, so let's pivot this to talking about co-founders and solo founders. You are a solo founder, but it's not necessarily because you thought that you were going to be a solo founder. What's the story? Did you try and have co-founders? Like, share more about sort of how you ended up becoming a solo founder.

Yeah, um, my first 2 companies I had co-founders, and EpicHost was an interesting story as well because it actually died because of a co-founder breakup, um, where one of the co-founders went rogue because of money or power or whatever, and then they essentially tried to destroy the business. And we had to sell the business off minutes before they destroyed everything. So, you know, like I have had insane, like extreme experiences, but also have had like kind of, you know, lukewarm experiences where I trialed with a bunch of people also for super memory. And even before this, for like various different things I was doing, and there was always either lack of compatibility Um, there was a time when I was trying to build a startup in college. Like, I was like, bro, I should build this as a startup. It was a healthcare thing that I should never have thought of. Um, but I still remember, like, I was not as excited about it, um, or the opposite person was not excited about it, or, um, there was some compatibility issue. One more thing that would happen is different ways of work, which is I like to be very hands-on in the engineering as well as the sales. But, um, you know, there's different ways in which people build the companies. Um, one more thing was, um, you know, the vision of the future and the directionality. I don't want to wake up every day and argue with someone about, okay, this is what we should do next. There is a lot of easy money everywhere in the world. Um, I don't want to do that. You know, like, I want to go for the, for the thing that is kind of hard to do, and it'll take years to get there. We won't make any money, or we might die, we might be on the streets, um, but we have to do this. Um, it's very hard to get someone as excited about it, um, especially when you're changing paths all the time. Um, you have to get the opposite person aligned, which is very hard to do. So yeah.

I could imagine, you know, at multiple points over even just the last year with Supermemory, if you had a co-founder And obviously you have these incredible teammates, but it's a different dynamic with teammates necessarily than it is with co-founders, where you maybe have to feel this need to really convince a co-founder in a different way than maybe convince a teammate. And I imagine that that could have potentially gotten you stuck at various points when you knew that you needed to keep moving, keep progressing, but maybe somebody else wanted to stick with the other thing. Kind of what we were talking about where you got some viral traction or some viral interest, and maybe one person, if you had a co-founder, they would misread that as sort of like, this is the optimal stopping point, we just continue on this. Maybe you can share a little bit about what it's like being a part of Solo Founders Program, but particularly maybe what you've learned from some of the other people in Solo Founders Program, because I don't want to make this like an advertisement for SFP, but I think it would be really interesting for you to share maybe what you've learned from people like Chai or Philip, people who are, who are also building companies solo, because I think there is something really special about being around other, other founders. And even just being in San Francisco, I think there's a lot of value. But maybe share a little bit about kind of what being around other solo founders in SFP has been like.

Yeah, living in the house was probably the best decision ever, not only from a perspective of being around other founders But being around people who can push you towards the right direction, but just by them existing. For example, Chai was my roommate, uh, is my roommate right now, but he was like in the room next door to me. And I would go on walks to Dandelion's Coffee with him, and he would tell me all about formal verification and how how people build their architectures, and he would be super excited about it. I was like, bro, what is this? Like, you know, I don't get any of it. And he would keep sending me links of him excited, him being excited about what he's building. And that kind of leads to osmosis of learning, of like, that's an engineering side of thing where you learn that, okay, I should probably look into this. to build really good reliable infrastructure. And I did. And now we have like really good infrastructure that pretty much never goes down and it's pretty cool. So that's one thing which is help in learning and osmosis of learning. Second is people like Ryan who are just extremely relentless and teach you discipline. I would sleep on the nook in, in the office sometimes just after work. And I would wake up to his voice, him recording like, okay, today is day 34 of building this company, and I did this yesterday, I'm doing this today. It was like he's in the army, you know? It was like he's an astronaut in space, but he would have that level of discipline. Um,

this is pretty early in the morning.

Yeah, this was in at like 5:30 AM. It's extremely cold and No one should be working at this time, but he would. So that taught me a lot of discipline in my life as well. And he has also been killing it recently because of his discipline. He also has world record for most number of pull-ups, by the way. So that's cool. Then people like Philip who are unhinged at everything, like, you know, default talented, default good. At everything they do, amazing people to hang out with. Like Namanya, you know, like there's so many things that my life was fulfilled from by being in that house, and still is because I can still walk up to them and I can be like, "Bro, like this is what I'm thinking. What should I do?" We talk to till like two a.m. downstairs, just talking about the world, talking about startups, something that's annoying me, something that. Is I can help to others and it's been an amazing experience.

What's a really random tangent here, but I just realized that you and Philip, and we probably talked about this quite a while ago now, but you and Philip both had hosting businesses.

Yeah.

Hosting businesses are great sort of like bootstrap businesses. Like they were good ways to get started, but they ultimately weren't the things that either of you ended up working on. Did you guys talk about that or like what's, Like, what's something that you've learned from Fillup?

Yeah, um, by the way, I think the hosting business is an interesting one because it needs— like, there's a high barrier to entry to build it, and, uh, it needs very good business sense, um, because there's a lot of unit economics that you have to do. Like, everything is, you know, okay, this margin, this percentage, everything is something like that. Which is, by the way, why our company also did not work out, because I was not perfect at it. But when you're building something that you want others to also do, you build ways to enable others to do it as well. So that leads to hosting businesses, I think. I have spoken to Philip about a lot of things. One interesting thing is his brain is not built to stop. He's like Iron Man where he was in a cave working on something and then he comes out of the cave with guns everywhere, kind of that guy. Uh, now he sells to public companies and like, you know, all of these large government agencies. But he was— he's also someone who's a true builder and he just does it for the fun. He's also very, um, he is extremely disciplined about his values. He's like, okay, everything must be open source, it's fine if I don't make money. And his values is what has gotten him here. I, I have never seen anyone with that deep values about building everything for free. And I know Ryan jokes about it, but I think it's a great thing to have those values in life. So yeah.

It's certainly how most people have heard about Docmost. And for those who are listening, Philip is the founder of Docmost. It's sort of the best option for on-prem wiki software for governments or big businesses that really want to have control over their data. But he started out as an open-source project. And interestingly, similar to you, how you had an idea with Anycontext that became Supermemory, he was actually doing hosting for open-source sort of wiki software, and people were paying for it, and that was a good business. But then he saw that all of the open-source wiki software was kind of crap. Yeah. So he's like, why am I just doing hosting for this crap? Why don't I actually build the good open source wiki software and then charge for the on-prem with all of the different logins and all of the things that enterprises need? And he went and he built that and it's doing extraordinarily well because of that. Yeah. Maybe you could share, and maybe this is the last thing before we get to the final 2 questions, talk to me about your investors, what you've learned from them.

Yeah, I think I have some of the best strategic investors that anyone can get. I got very lucky. So, and there's various different stories as well about like, you know, how I met them. For example, I was introduced to Joshua Browder through Sudarshan, and I lived with Joshua Browder like days before moving to SFP. I was you know, crashing at his place because I didn't have any other place in SF. And I remember like he would, he would come and he would be like extremely direct at, okay, um, you know, figuring out exactly what I'm going for. And he would keep asking me questions until I have the direct answer. He's extremely direct. And then when I did have a direct answer, he would come to me and he would be like, okay, can I put in more money? Can I put in? So he's like so convicted. Um, like I would be at 1 AM, you know, doing something, and he would come up to me and he was like, do you want 200K more?

This is the trick of having somebody stay at your house.

Yeah, I mean, like, that shows extreme conviction. And he would sign the safe right there. He would wire the money right there. It was insane. Uh, it's like 1 year of runway, um, in 1 night, you know. But then that chain of events also led me to Suza Ventures, which is Shahir. And Shahir is also a guy who knew nothing about memory and context and this world, dived extremely deep in, and he knows pretty much everything I know about the market now. And that is something that most VCs would not do with that much conviction. Sudarshan also keeps calling me every week asking if I need any help with sales etc., etc. One of my VCs sat with me on the sales call, and then after the call they were giving me feedback about how I should improve. Um, I share research with, uh, one of my angel investors, and they give me feedback on the research I do. So all of these things are very rare, and I feel very lucky to get that stuff. Also, solo founders like you, I just walk up and ask like directional questions about my life, the company, all of these things. Like, it's very hard, uh, to talk about these things, like to think of your investors not as investors but actual people who are along with you on a journey, um, to build something great. And, uh, instead of just people in suits putting in money, you know, that's what I imagined them to be. But it's all people in jackets and, uh, and shorts, uh, helping me out.

You know, you know, I think the, um, the really cool thing is you assembled like great crew, a lot of people. I think more than the average company does for the size round that you raised for angel investors. And I think you did a great job of that because each of these people could help you with something specific. Yeah. Siddharshan, who I think is like, I think he's probably pretty well known on X, but I think most people probably don't know him yet. I think he's great. He's just incredibly high energy. Josh, hopefully, you know, he's a solo founder actually himself. We should have him on the podcast soon. Like John Andrew.

Yeah, he was helping me with sales. He was like, you should hire 2 sales reps, 3 AEs, 3 BDRs. He essentially made like a chessboard of people you should hire for me, which is cool.

And I think that the interesting thing here is like, you know, people, I think that it's easy to say, oh, well, I don't have any connections. It's like, well, the way that you get connections, if you don't have connections, you're an outsider to Silicon Valley is you put stuff out there and you put it out there in public. Yeah. And that's it. You were a kid in India, and then you were a kid at ASU, which is not a thing that people are thinking of when they're thinking of startups. Maybe they will start to do it. I know there's some cool entrepreneurial stuff happening there, but I think that the lesson is put stuff out for play. And then if you put stuff out for play that's just your art, your interest, that will gain you followers and attention, and those people will say, oh wow, he's actually working on a company now. Yeah, should get involved.

No, I, I think it won't even give you followers or attention for the first few years. That should not even be the goal. Like, you know, I see a lot of people being depressed about, oh bro, like, I post a lot of things, I don't get anyone looking at it. But sure, like, because your goal for posting something should not be someone looking at it. It should be to, to show your real self on the internet, and then There will be a network effect that starts at some point, but you should not expect it.

This is like the thing. I mean, I went to school for music and I loved music. And like, very often you'd play to a room with nobody in it.

Yeah.

You know what I mean? So it's the same thing as launching a product that you've been working on on X and getting like no engagement.

Yeah.

Like that happens.

Yeah, it has happened to me a lot of times actually.

You gotta do it not because you're hoping you're— if you play to an empty room, and you're disappointed about that, you quit, you're not going to be able to actually get there ever. So that's really cool. Let's close it out with the classic questions that we ask. The first one is sort of the bear case, like the case against solo founding. And then once we talk about that, we'll end it with the case for solo founding. What's the bear case? Why shouldn't you be a solo founder?

Yeah, I think as a founder, it's like having, you know, in Matrix, there's a lot of darts being thrown at you and you have to avoid all the darts or you have to like, but you cannot avoid the darts because you're a founder. So you have to do like, you have to actually hold them somehow. You have to catch all of them, do all of those things. And if you're not very good at multitasking, like there's so many things to do, you will not get everything done. And especially if you're not able to recruit a good team, it's kind of really hard to, you know, do all the basic things to be successful. And I think like building a team, building a distribution, uh, doing the sales, doing the engineering, all of that at once can be really challenging. And you have to be a different type of person to be able to be okay with that challenge. I think everyone can do it, Uh, it's just that you have to train yourself mentally for it. And it's also a very— it's a journey where no one is there to keep you accountable for these things. So if you just don't do it, there's no one coming and scolding you about it. Um, no one being like, bro, what the— what the hell are you doing? Which was actually— I think, I think I've told Julian and other solo founders about this as well. I think I was terrible at this, um, for the longest time. And now I have more confidence in the fact that I am able to do the sales, do the engineering, do all of these things and keep myself accountable to it. I also do have other solo founders in the cohort that come and shout at me if I don't do

it. Even after?

Even after, yeah. Even right now, um, I can just walk to Chaya and he'll be like, did the deal close? And I'm like, no, bro, I didn't follow up with them. Like, what the hell? Um, so yeah, like, um, I think you should definitely try to hold yourself accountable.

And also put yourself in a position where other people can push you. Exactly. Even if they're not your co-founder. I think a lot of people benefit from it, even if they do have a lot of internal discipline, to be around other people, to have that model is really good.

Yeah. I remember you being like, oh, bro, you're not tracking all your calls. You don't have a CRM. You're not actually doing all the things that— then I was like, man, you're right. Uh, and then I did all of those things right now. So yeah.

And now it seems like— I mean, it seems like thank God you implemented the CRM and stuff a while back because you showed me your calendar. It's like full up with calls for people who want to use this. Um, cool. Let's end with the last thing, which is what's the case for solo founding?

Being a solo founder is the best way to build something that's true to yourself And that only happens when you have extreme conviction about what you're building. It's kind of like if you are doing it with someone else, it's a shared— it's one person's vision that you are kind of inside, or it's your vision that you need more people to do. But it's typically a standard business, which does work out. But if you want to build something big, there must be something that's true to yourself. that you're really excited about, that you're okay to do in your darkest, deepest times, and if no one is by your side. And I think so all the businesses that solo founders do start are kind of that, you know, like it makes sense that they would start those businesses because it's like literally their representation inside a company. Like think about Philip, Ryan, uh, Chai. Like, you know, when I think about these and the companies they are building, it makes so much sense that they are the ones. But also the people who came to the podcast Um, Vercel, for example. All of these solo companies are literally representations of the founder as a company, um, which is really cool to me. So yeah.

That's awesome. Cool. Great way to end it. Uh, thanks for being on, man. It's, it's, uh, it's funny because we obviously hang out, but so it's weird to, it's weird to like record like this, but, uh, really awesome. Like, I feel like we talked about things that, you know, we've known each other for years now. We're still talking about things we haven't talked about before. So this is a great way to do it.

That's amazing.

Awesome. Well, thanks for doing this.

Amazing.

If you enjoyed this conversation, we would love it if you could share Solo Founders with everyone you know, particularly people who are considering starting companies solo or who are already doing it. The entire goal of what we do is to normalize solo founding, encourage more people to solo found instead of end up with co-founders of convenience. If you'd like, and we'd really appreciate it, you can go leave a review on Apple, you can go give us a thumbs up, you can subscribe on the YouTube channel. Anything helps. And of course, if you are starting a company or considering starting a company, you should go look at the Solo Founders Program. We work together with you in San Francisco alongside of a bunch of other solo founders. The idea is it's much better to be solo together than to be solo alone. I'll spend a lot of time helping you, and we also invest $100,000 in your company. So if you're excited to learn more about that,