In this episode of The Next Next, host Jason Jacobs interviews Sahil Lavingia, founder and CEO of Gumroad, a platform that helps creators sell products directly to consumers. Sahil discusses his journey from being employee number two at Pinterest to founding Gumroad and then launching Antiwork in 2025, a company aimed at promoting flexible work arrangements. They dive into Sahil's approach to building ambitious companies, the transformative impact of AI on business processes, and how Sahil is now bringing operations back in person in New York while utilizing AI for automation. Other topics include the evolution of work culture, decentralized decision-making, and the future of venture capital and software development. Jason also shares his personal learning journey towards building companies with flexibility and maintaining a balance with family life.
In this episode of The Next Next, host Jason Jacobs interviews Sahil Lavingia, founder and CEO of Gumroad and the newly established company, Antiwork. They discuss Sahil's journey from being employee #2 at Pinterest to founding Gumroad, and now venturing into promoting flexible work arrangements and leveraging AI to automate business processes. Sahil shares his insights on the transformative impact of AI in the software industry, his approach to building and scaling companies, and his vision for the future of work. The conversation also delves into how AI is altering traditional employment structures and what this means for venture capital, enterprise software, and individual productivity. Tune in for an enriching discussion on embracing change, optimizing for efficiency, and staying adaptable in a rapidly evolving tech landscape.
00:00 Introduction to Sahil Lavingia
00:31 Sahil's Early Career and Gumroad's Evolution
00:59 The Shift to AI and New Business Models
03:58 Personal Reflections and Work-Life Balance
10:00 The Impact of AI on Software Development
21:04 Future of Work and AI Integration
36:48 Project-Based Work and AI Efficiency
40:22 AI's Superiority in Software Engineering
40:59 The Efficiency of AI in Learning and Documentation
42:17 Understanding Tokens and AI Context Windows
44:09 The Future of AI in Software Development
44:25 AI's Impact on Different Professions
47:38 The Role of AI in Sales and Marketing
53:37 The Evolution of Software and Open Source
55:05 The Future of Venture Capital and Software
01:00:01 Building Efficient Workflows with AI
01:14:29 Concluding Thoughts and Future Directions
Jason Jacobs: Today on The Next Next, our guest is Sahil Lavingia. Sahil is an entrepreneur, designer, and writer best known as the founder and CEO of Gumroad, which is a platform that enables creators to sell products directly to consumers.
In 2025, he launched Antiwork, which is a company that promotes flexible work arrangements and challenges traditional employment structures. Now, I was excited for this one because Sahil is a fascinating guy. He's a writer and published author. He was an entrepreneur from a very young age. He was actually employee number two at Pinterest before he left in 2011 to start Gumroad.
And then with Gumroad, he started down the traditional Venture path, then retreated and got really small, really asynchronous and really remote, with how he was building gum road. Now he's entering a new phase where he's bringing things back in person in New York, and he's also leaning hard into. AI to automate as much of their business as possible.
We have a long form discussion, we cover a lot of ground and it really got my gears firing and I hope it does for you as well. But before we get started.
I'm Jason Jacobs, and this is the next, next, it's not really a show. It's more of a learning journey to explore how founders can build ambitious companies while being present for family and not compromising flexibility and control, and also how emerging AI tools can assist with that. Each week, we bring on guests who are at the tip of the spear on redefining how ambitious companies get built.
And selfishly, the goal is for this to help me better understand how to do that myself. While bringing all of you along for the ride. Not sure where this is going to go, but it's going to be fun.
Okay, Sahil Lavingia, welcome to the show.
Sahil Lavingia: Hey, Jason, how's it going? Thanks for having me.
Jason Jacobs: It's going well. Yeah, it's been a few years since we connected and I don't know if you remember this or know this, but you were a key inspiration for me in using the rolling fund structure and actually in helping me write the initial memo too. And that was a huge part of MCJ.
So thank you for that.
Sahil Lavingia: That's awesome. You're welcome. Yeah I remember that very well. That was an exciting time in the early days of COVID, I think feels like a different era.
Jason Jacobs: Yeah. And for me at the time I was an active angel and I was doing it publicly and there were a bunch of people that were bugging me to raise a fund, but I didn't know if I wanted to be VC and I also didn't like that if you went and raised a normal fund, you have to be in this quiet period when I was fully in public.
And then I saw you tweeting about. Active solicitation. I'm like, how can you do that? And you were like, oh, it's just rolling fund structure.
Sahil Lavingia: Yeah, I definitely got some few DMs that were like, you can't do this. And it's no, you can just things change every once in a while, right? Some, you just pick. Some people don't realize that, but the laws can change. And the 506 C I think fund structure made it possible to tweet or like publicly solicit, which, if you're building an audience online, makes so much crazy to think like you have to do all this stuff in silence, so it made total sense to me.
Now people know because of people like you that we can, yes, you can raise a fund and talk about it as you do it. And then, yeah, just makes it just, it just feels like the way things should be right. That's an, like a software first mentality to raising money and deploying capital and doing all that kind of fun stuff.
Jason Jacobs: You're a fascinating guy. I knew that. But then as I prep for this discussion, I think my fascination only grew. I didn't give you much context on My journey So I'll give you a bit of context on what I'm up to and why I Asked you to come on the show and then we can get into it.
Sound good.
Sahil Lavingia: Sounds great.
Jason Jacobs: Yeah, so I've always been an entrepreneur and I'm still an entrepreneur. I mean building stuff From you know bringing new things into the world is what I love to do. I find myself at this juncture just super involved in my kids lives. They're almost 13 and almost 10.
And I've found that as they've gotten older, they need more parenting, not. Less parenting and I feel really fortunate that I'm in a position where I can be so involved in their lives And I don't want anything to get in the way of that. But at the same time, I'm not retired. I'm full of ambition I love bringing new things into the world but I can't build the way that I always have before, where it consumed my entire life.
And I also just, as I've gotten older, I'm 48, like I, I just have a lower tolerance for doing shit I don't want to do. And so for example, not being in control of my own schedule having outside stakeholders that You know will be disappointed in me If the wind takes me to the left or the right because i'm following my you know intellectual interest and it's different than the direction I thought I was going in Or things like that.
And so it's I still want to build, but I gotta find a way to, stay smaller, stay in more control, have more flexibility without compromising ambition and without compromising presence with my family. So that's one thread. And then the other thread is what is happening with AI?
I'm new to this, but this is crazy. And yeah, there's a lot of grift and noise and hype and oversaturation. But at the same time, like we're at the early innings of what seems will be profound implications for how we live and work. And I wonder how that's going to impact how startups get built and funded.
And I wonder how it's going to impact how someone like me might be able to build different. So similar to the public journey I went on six years ago with MCJ, I'm heading on a new. Public journey now so that's one Thread and then the other thread in parallel is that as I've been out having these discussions.
I've uncovered that gosh There's a lot of other people like me experienced founders that need to build different that are wandering around and like their lives are funny and nobody's telling those stories, no one's speaking to them, and so I also just as like a satirical counterpart, want to build a show called Fogey Founders that is almost like Silicon Valley for the middle age, or, like what Silicon Valley would be like American Pie for, or like Back to the Future when Biff was was washing cars as an old man, and I want to, do that mobile first and iterative, starting with characters on TikTok and Insta and fleshing them out and building followings and weaving them together into skits and then Having it evolve into a show over time without any gatekeepers So those are the two things i'm up to and yeah, and you've built different and you're leaning hard into AI And you also didn't start out building different, you started out building the normal way, right?
And or the normal to Silicon Valley. And gosh, it's like there's so many parallels and so many things I want to talk about. I can't wait to dive into this stuff with you. So that's some context.
Sahil Lavingia: Wow. That's a lot of concepts and that's freaking awesome to hear. I think it's bodes well for the future.
Jason Jacobs: I'm pretty fried up about it. My wife doesn't understand it. My my, my partners from MCJ I don't think understand it. The hockey dads at the rink definitely don't understand it. There's not many people I think that, that understand it, but I don't care. Like I'm so excited and it just feels, it feels right.
So with that, maybe, I alluded to some of it but maybe just, give the compressed version of you.
Sahil Lavingia: Yeah. The compressed version gets longer every year, which is
Jason Jacobs: what happens.
Sahil Lavingia: I started out as a, I guess like a. A designer, a product designer, I wanted to make stuff like from a pretty early age and Photoshop was like my avenue to doing that. So I started out as like a web designer, designing websites.
I got very quickly frustrated with that, so I learned how to code, HTML and CSS. Then the iPhone came out in 2007, I think, and I started building iPhone apps
Jason Jacobs: The iPhone was 2007. The app store was 2008.
Sahil Lavingia: Yeah, and the app store really, really changed things for me because it felt like I could build something, I could design something, build something and then show somebody and they could use the thing and they could even give me money if they wanted to.
And I didn't have to wrap my head around all of this as like a 16, 17 year old kid at the time. It was. Pretty eye opening for me. And so I got really into that. I built my first job was I was at Pinterest. I was the second employee of Pinterest where I built design and built Pinterest for iPhone.
Then I got laid off to invest in a, my stock started a company called gum road which was like one of the first places where creators could come and sell digital content. I called it bit. ly with a credit card form built in, but at this point, people probably don't even know what bit. ly is, but
Jason Jacobs: I
Sahil Lavingia: a really easy way to sell digital content on the internet in 2011. And then for the last, I guess now 14 years, I've been doing that just shipping that I've had a few different sort of. Side tracks, I got really into creative writing and painting for a few years. I still do that but I got really into it after I would, I, Gumroad failed to raise our series B in 2015.
Jason Jacobs: You even moved to Utah to follow was it Utah that you went to, to take the course from the author that you love
Sahil Lavingia: yeah, Brandon Sanderson. Exactly. So I moved to Utah in 2017 and took this science fiction fantasy writing class with Brandon, which was amazing. Learned how hard writing, yeah. is at least without AI helping you. And that guy's a monster in a good way, he just is incredibly productive And then while I was there, I started learning how to paint and bought some oil paints, did some plein air painting, figure drawing like 10, 20, 30 hours a week, got really into that.
And then in 2020 when COVID happened we both together, separately, but together launched rolling funds on AngelList. I started investing in stuff and. And part of a lot of these things, honestly, the last, those two sidetracks was because like software was just boring in a way, like I felt like crypto was a kind of hope for a lot of us because it was like, Oh, maybe this is like a new sort of reinvigorating event in which everyone get re excited about consumer internet and software.
And just wasn't really that I think it has some applications, but not for like consumer use cases. And then I really is the thing that really got me back into building gumroad. And all of a sudden, I felt like I could actually have new ideas and be more ambitious. And just like gumroad was one of the first, there was like this resetting of the starting line for software.
With AI, it it's like we moved from knives to guns or something. And it doesn't matter how good you were with a knife. Like we're now using guns. And if it feels like that kind of moment happened, in the last like year or two. And so got reinvigorated, got this office. We now are in Dumbo in New York, Brooklyn, New York, and hire.
We now have six
Jason Jacobs: I didn't know you were in New York now.
Sahil Lavingia: Yeah. I moved to New York in 2020, late 2023 and love it. And I just think it's super key. If we want to build the future that we do it. In person have a really tight feedback loop collaborate a lot and sort of push each other on our thinking and productivity and AI can do all the solitary work for us over time.
So it more and more ironically, I think of running the company a little bit people like to make fun of companies like Google and Facebook Oh, they spend all their time in meetings and they don't get any work done. And software engineers spend 35 hours out of 40, just like.
Trying to align interests and resources and things like that. And I almost think that that's how like a small company is now, because effectively we all have. infinite resources in the sense of we all have Devin and cursor and effectively these things act as like thousands of little junior and mid level engineers for us.
And so we actually, even if we're only five should be spending all of our time, like we each had a department, equivalent, and we need to make sure that those resources are like allocated in the right direction. And, Elon has a great bit where he says the productivity of a company is the vector sum of all its, all of its people.
And that basically means that the more alignment you have, right. The more you're headed in the right direction. If you have a bunch of great people, but we're all working on like random things, like it doesn't really matter that we can even potentially cancel each other out in our efforts.
So I think it's becoming even more important to basically synchronize and make sure that we're moving in the same direction and we're not wasting resources and time. And the other side of this is that if you're not. Working in this way, someone else will, right? So it's not like business is interesting because it's like a sport.
It's not just about being good at software. Like you're inherently competitive, right? You're competing with other humans. It doesn't matter how good AI gets. Like you're still competing with humans that are using AI. And that's the goal. The goal is to be number one, the goal is to win. And you have to recognize like what is the shape of the winner?
How are they working? How are they thinking? And you have to think and work. Like them, if you want to win. And that's the fun that to me, that's the fun of the game. I would not be that interested in playing a game if the best I could do was. 80th place, right? Like I would rather just put all my money in the S& P 500 and hang out with my friends and family.
But the idea of going to college, the idea of working, the idea of learning, it's about trying to contribute to humanity and like the sort of like the corpus of knowledge that we have and that requires winning in a sense. So yeah, that's what I, it's funny cause I'm, I feel like I'm back to where I was in a way where I feel like this is becoming, it's new, but also trad in a way, like it's the way Building used to be like, but it's also it's different because, no one's ever worked like in this exact way, pre 2024.
So excited.
Jason Jacobs: why, I listened to it was probably a few years ago, but you were talking on some shows and what you were saying at the time was that you essentially you Wake up and get the crap that you least want to do out of the way first, and then that frees up your time for the rest of the day, to either do deep work or just do stuff that you love and you worked asynchronously, you tried to have no meetings and stuff like that, it sounds like, I might be hearing wrong, but it's almost Is it a 180 from where you were at that time?
Sahil Lavingia: It's almost perfect. 180. I feel like you could say it's a 180 or you could say that it's in the same, depends on what layer of abstraction you're operating at, right? You could say it's perfectly in line with what I was trying to do, which is to try to be as efficient as possible.
The sort of world we live in today is completely different. I would say that we were so successful at that way of working that we did all the things. That way of working allowed us to do, and there was nothing left to do, if I wanted to work with no meetings and completely asynchronously, what I would do is just sell the business.
Then I would have no meetings, no deadlines, no employees, and, that's the outcome I could have had, but I want to keep building stuff, and I'm now You know, in a sense going 180, but really, I think I'm just doing the same thing I've always been doing, which is trying to optimize for learning and building new stuff and trying to get really good at it and participating just like I think we're participating in the future of remote work that we were on the bleeding edge.
I was telling my team, actually the positive side of me is that you get to be on the bleeding edge which means that, like, when good things happen, you get to be first to the good stuff, but also that means when things change, you're also going to like. And the bleeding edge requires some bleeding right every once in a while.
And at the end of the day, like I run a tech company, I think that's the goal of a tech company to embrace change and to do new things, not just to be successful once and then just get to do that infinitely. The brilliance of software is anything that needs to be done over and over again.
You can automate. Therefore eventually you run out of that stuff and then you have to be creative. And I would also say just to be frank, like part of the no meetings, no deadlines, no, the kind of like the asynchronous work is I just didn't enjoy it that much. It was boring, and so I was trying to automate all of it and be as outside of that as possible.
But now with AI. And being able to have new ideas and use V0 and cursor and Devin to actually ship my ideas instead of have to have a team ship my ideas like. It's fun again. And like, when it's fun, I want to be in the office. I want to be around people like I don't want to like, I don't think of my friends and family as people I want to like, work asynchronously with and have no meetings with right?
I want to be with them. I don't want to be in a long distance relationship with the people I love. So it's I would say also, frankly, that maybe that was a sign of maybe, it was just boring and I was optimizing for a different set of variables.
Jason Jacobs: I mean i've heard you talk about waves before and I think about waves as well and when you figure out what wave to ride there's some of it that has to do with you and your life stage and circumstances and then there's Others of it that has to do with what's going on in the world and timing and think things like that one thing that I wonder about even now as I'm setting out to work different is how much is Society changing and the underlying infrastructure changing and how much is it me being in a different life stage and having different?
criteria, so how do you Think about that when you think about like the seasons, because it seems like you go through different seasons like I do, how much of it is about external versus internal.
Sahil Lavingia: think that there are some, there's parts that change externally. I think most of it is probably internal. That's my guess. And I think there's even though the technology is like speeding up the pace of development or so it seems like there's some level of consistency in history.
Like I, I think that, for example, 4 years seems to be like a recurring theme in my life. If I look at there were 4 years where I was like, was like a venture backed startup trying to like, be a billion dollar company. Then there was these 4 years were, we were like, I was doing other stuff and I was running it as a kind of an efficient business.
Like lifestyle business. Some may say then there was like, the 4 years after that, where it was like, hey, we're now trying to grow, but in this different kind of way during COVID and all this kind of stuff, we raised crowdfunding. I was doing the rolling fund trying to eat my cake and have a two and then now we're in this like sort of new era, and and college degrees are four years.
I think there's some sort of physical law or something that exists that sort of it takes four years to process and learn and get really good at a certain thing before you then go on to the next four years of that season. And maybe that also just leads that may also be an external thing more than an internal thing where you have just, I actually believe that things could happen much faster than they do, but I think society.
Slows things down. Generally, presidents also have four year terms. I think there's like a sort of there's a human brain can only deal with so much change. For a certain amount of time. So we need to move slower. I think every once in a while you have someone who can just break that physical law and say, actually, we're going to have 40 years of change but generally, I think humans would prefer, like a technology adoption curve.
Yeah. That's a bit slower than like the sort of what could actually be possible, right? Like it took time to go from an iPhone to having everyone having a smartphone, right? From 2007 to maybe COVID right until everyone really got one or cars or TV or radio like these technology waves like go from the front page of the New York Times to like everyone's bedrooms like over a 15 year period of time, or maybe you could say for seasons of 4 years or something like this, like Carlotta Perez.
She has that bit where, things go from these like technological disruptions where they first get like society has to like, learn about them. Then they have to adopt them. Then there has to be like a graduated change into the new system. Yeah. And then once everyone is in the new system, then you can make another evolution that requires everyone to be on this new sort of this new place, right?
Where you have like in design, you have the iPhone basically look like skeuomorphic, right? It was like trying to teach everyone like, Hey, this is like how documents work digitally. And it looks like a document. And then once everyone is in that new state, then you can actually get rid of all the design and say, actually, now everyone knows.
You know that this is a document. You don't need like a floppy disk with a save button. You can just auto save, right? You can just start to graduate people. But it takes time. I think that's one of the big lessons I learned when I was trying to build my company is that even if you believed something could happen in two years, it'll still take 16,
things take time. There are physical laws that slow things down. There's natural laws that slow things down. There's like just society's own incentives that, people are not going to want to see change. Even running the business, like it's not as obvious to me that just because the AI comes out and makes software engineers 10 times more productive, our software engineers are going to be 10 times more productive, right?
That's going to take time and incentives. And sometimes to me, like twisting an arm. Easier to do in person.
Jason Jacobs: So if you've been consciously or subconsciously operating in four year cycles, where, what cycle are you in now and how far into it are you?
Sahil Lavingia: Yeah, that's a good question for me to think about. I wish I could pause and think about this for an hour and come back to you, but reader can, I guess the listener can, but I think we're
Jason Jacobs: The 200 dollar ChatGPT can.
Sahil Lavingia: Yeah, exactly. Deep research come back, process this stuff for me.
Yeah, I think we're early in this new cycle that probably started, let's say, in, 2023 or late 2022 with let's say chat GBT and like the end of COVID. And it was this realization that people were sinking into that we're really going to get basically this just like the iPhone moment.
I think we're going to see this like drastic change in how people interact with software. And what software means and interact with each other through software. I still think we're probably like a year and a bit into it. We decided to get rid of most of our remote staff or convert them to project base and make a lot of changes, get people in the office, incentivize that all day, all these, basically get rid of remote work by default
Jason Jacobs: And didn't you? Didn't you have workers in different countries all over the
Sahil Lavingia: Yeah we're working on seven different O 1 visas right now. Like it's a big shift for sure. We just got our first approval, which is pretty exciting. In six days. The cool thing is if you want to live here
Jason Jacobs: you use an AI platform to help facilitate that?
Sahil Lavingia: we used a a new age immigration firm, let's say I'm sure they are using a lot of AI to generate, one of the best use cases for AI is like dealing with governments that have onerous forms and things that you got to do, that are a bit. Not, a bit meaningless, but also like basically existed as a way to regulate demand that is now being, overturned.
But yeah, I would say we're maybe 25 percent into this wave of effectively like realizing I think the sort of like pithy summary is that like typing is no longer the bottleneck. I think we like. Yeah. Or have been building software, this entire industry has been operating on this idea that like you had an idea and the bottleneck to shipping that idea to customers was typing, right?
It was basically front end, back end, full stack, iOS, mobile engineers, like actually typing, like literally okay. I can whiteboard this to do list app, right? And then it's going to take three days for someone to actually type it out, right? It's a few 100 lines of code. It's going to break.
It's going to need tests going to need all these things. Now you can just say, hey Replit or Bolt or Cursor or Lovable or whatever. ChatGPT, I need a to do list app that, I have ADHD and so I want automated reminders every three hours if I haven't made an update on one of these tasks and I want it to be sent to my Slack account and this is my Slack API key.
And within, 15, 20 minutes, you may still run into bugs, right? The typing part is now gone. Like it's no longer the bottleneck. I tell everyone that no one should be typing like code like you should really be outsourcing that. Just like you don't type assembly, right?
You should probably shouldn't be typing. Python or JavaScript or TypeScript or Ruby, you should be typing English and then the AI is translating that because the code itself was not meant for humans to type. It was meant for like computers to know what to do. It's a very different but that I think has changed a lot of priors.
Imagine an entire industry built on the premise. The typing was the bottleneck and now that's no longer the case. Like a lot of things need to change. You get rid of that bottleneck and all these other bottlenecks start to show up and you can get rid of those. , a lot of businesses were a lot of languages we chose to use.
A lot of tech stacks we chose to use a lot of people. We decided to hire a lot of different, even sales cycles and, whether we closed source or open source, the code big. Decisions were made based on this premise that like typing was the bottleneck. And that's gone.
For example, if typing was the bottleneck, you could hire someone remote and then they could just go off for five days and type a lot and then they'd come back. And that was like the efficient, right decision to make. But if typing is free. And then all of a sudden, what's really critical is making sure the instructions are the correct set of instructions, right?
That if you, it's you slightly tilted rocket to Mars. It's going to end up somewhere else, right? Like, all of that autonomy requires a lot better precision. And the launch has to be so finely executed because. Then, 100, 000 hours of work are going to be done by the A.
And then it's going to come back and it's going to perfectly execute what you said you wanted it to do. But what you said you wanted it to do may be different from what you actually wanted it to do. And so the skill then becomes prompting, is the term, but basically getting really good.
And just like I was saying before this is what Google trains people to do. It's like they were Sort of treating their workforce of tens of thousands of junior mid, 1, 2, 3, 4, 5, 6 level engineers. It's like that workforce. And then they have a layer on top, which is really what Google is.
It's really like the few 100 people that are making the product decisions, right? And prioritizing really effectively the people who are hiring and firing. Those are the people who are really in charge of these companies. And I think people are realizing that the people below them are really just like the equivalent of bricklayers, right?
And the real estate construction industry, the real estate developers are the ones in charge of the zoning of what gets built of how things look like the people funding the capital, right? The debt providers, the equity providers, the investors, the developers, the entrepreneurs, and then. Really, no one goes down the street and says that construction worker has a ton of leverage in the system.
I think that realization is what's coming to the software industry, which is not always sexy and fun to talk about, which is also why I think it's not insider information in a way that's it's not as publicly talked about. The public doesn't necessarily benefit from it in all the ways that entrepreneurs and investors and capital owners may.
Jason Jacobs: So these tools started coming about and you had this awakening and you got. Refilled with energy and excitement about the future, and so it's oh, instead of just trying to minimize my work and get it done, first thing in the morning so it can free me up to do whatever I want, now work is becoming more front and center of, it's back in the zone of delight and moving out of the zone of a void, right?
And and couple areas I want to go with that, but I want to take a quick pause just to mention that your your rant just now helped validate me grabbing on a whim a domain school of, schoolofprompt. ai, almost like School of Rock. I'm still learning what prompting is, but it's just hey I think what you said the instructions you give it are becoming more and more important, and that's a skill set and learning a language in itself.
And, Who's teaching that language, and what are the best practices for that language to be taught? We can shelve that, but but if you ever want to talk more about that I do have the domain. But what I want to come back to is, so it's okay, you're getting refilled, and it's excited to go and build stuff again.
Two things I want to drill down on. One, the old way, the last four year cycle, where did Gumroad fit in your Portfolio of pursuits, right? And then does that portfolio consolidate coming into this new era and then, so yeah, I want to talk about like portfolio construction and then as a followup, then I want to dig in on okay, the gum road in the old mode and then the gum road in the new mode, cause you mentioned that you're reinvigorated to dive into gum road.
So what does that mean in practice in terms of where it's going, what got you excited and how to get there?
Sahil Lavingia: Yeah, I think that it's Gumroad, it is exciting because we get to ship a lot more stuff for Gumroad, like our customers should be pretty happy and excited with us because we have a lot more going on our
Jason Jacobs: Customers being the creators that are building stuff.
Sahil Lavingia: exactly the creators that are building stuff on Gumroad selling digital products and courses and memberships like we're building community like a mini slack within Gumroad, we decided not to build that feature because it would have taken a long time and been a huge burden probably would have taken six months or a year to really build.
We now think we can do that in a quarter with 1 engineers time, right? So the pace of being able to ship these features, let's say it's 4 times faster. Our 2025 roadmap really just goes to the end of the quarter, right? Because we can just get so much
Jason Jacobs: But were you not excited about these features before it became easier to type?
Sahil Lavingia: I was, but it just wasn't the core issues I think with Gumroad is that we were just competing with new startups that were offering these products and services and had the energy and youth and venture funding to be able to build better products faster. The problem is when typing is the bottleneck, like the.
The winners are the ones who are like more well capitalized, right? Versus today. That's no longer true. I think the winners are the ones who have higher quality design ideas and product ideas, which I think I have leveraged with more so than maybe a 18 year old Y Combinator founder. And so I think it levels the playing field in a way.
Jason Jacobs: Were you, so in that mode of prior, it's like you could have done it if you got more capitalized, but it wasn't worth the price to you, so you would just concede?
Sahil Lavingia: Basically, yeah, it's like we were just happy with, second place and trying to be really good at second place instead of trying to be trying to read,
Jason Jacobs: Because your freedom was more important?
Sahil Lavingia: Yeah. And I could do other stuff with my time that could potentially be number one.
Jason Jacobs: Now that these tools are coming along, all of a sudden it seems like your freedom's not as important, though. Explain that. Is
Sahil Lavingia: is still pretty important. I can leave the office anytime I want. But I think I was just looking at the way I was spending my time and I felt like a lot of it was. I was free to go on a walk and listen to an audio book. And why was I doing that? It was to maximize my learning and enjoyment.
And, I would go to the gym. I still want to make sure I still do those things that are so important to me. But ultimately I think learning like the best way to learn is often to build. It is often to not just consume content, but to create content and create
Jason Jacobs: it that these new tools are enabling better learning, which is moving the Gumroad work up higher on the learning value list?
Sahil Lavingia: Yeah, and I also think that we don't have to, like a lot of what I'm spending my time with is not necessarily just building Gumroad, but actually like taking the resources we have and building new products. For example, the thing I'm working on at Gumroad is actually a new, completely new product called gum.
new. Which is basically if I were to build Gumroad from scratch, but today, how would it look like? And that's what gum. new is going to be. It's going to be a really easy way to build and launch landing pages to sell digital products and collect emails, which is basically what people use the Internet for.
And then there's other products. So I actually like renamed the holding company, the umbrella company behind gumroad anti work. And if you got anti work, you can see, but we have five different projects that we're working on. One and basically each one of these is a thing that automates something that was not inherently fun and creative about working on gumroad the business.
So for example, like Handling customer support tickets or doing content moderation or doing QA or approving invoices and dealing with payroll and equity and board consents and option grants and all of that kind of stuff. There's all of these components to running Gumroad. And if I can just what I was doing in 2020, like how can I automate all the stuff I didn't enjoy doing so I could move on to the stuff that's fun and creative and full of learning.
And often what's full of learning. I think the other thing that's different about this era is that sometimes if I want to learn about real estate development, like I can just go read a book, right? There's periods where like the craft is. Is proliferated and there's like a, there's a, you can go to school for it.
But in the beginning of an industry, in the beginning of the technology wave, like really the only way to learn about it is to do it. You can't really read a book on like cursor versus Devon versus AI prompting and all this stuff, like really hasn't been like the papers, the replicated papers, haven't yet been written about this stuff.
Like we're still on the cutting edge. And I think sometimes the fastest way to learn about it is to just. Start working on a project and the beauty of where I was I have a company. It's profitable. I can have a team of engineers and designers help me build this stuff. And I think it's risky.
I think I was getting bored with just there was no risk in my life. I was just You know, like the average day was like very high, but there was like a lack of, I think, like social engagement, a lack of connection with other people. Like I was doing almost like I was too stoic.
I was like the sort of then monk at the top of the mountain. Eventually I don't know. I just got bored, of that. And it's like, how long can tweet I Nevolisms before you want to hang out with people and actually get in the arena, quote, unquote and fight and do jujitsu and there's this, it's almost like a new kind of philosophy.
You have the philosopher that just sits at the top of the mountain, and then you have the philosopher that runs a business and does jujitsu, right? I think you can do all of the things, I think that's part of the power of technology is you can have a clone of yourself, do all the stuff, then spin up a new version of yourself to do a bunch of new stuff.
Then use AI to automate all that stuff. And then all of a sudden you have three or four different silos, and it wasn't like I was just doing nothing, I still was doing like, I was. I wrote a book in 2021. I was running the VC firm, the rolling fund. I was running gum road. I was tweeting bangers.
Hopefully, I was trying to do a bunch of different things. And then this next era of life, I want to run this business. I want to have new product lines. I want to acquire new companies. I want to use AI really well. I'm also having a kid in May. So I want to juggle, I want to juggle, I think
Jason Jacobs: Are you doing less outside of anti work now?
Sahil Lavingia: I'm probably doing less schmoozing, right? I think I was spending a lot of my time just like hanging out. Like socializing, going to lunches and dinners and still doing some of that, but I try to minimize that. And I try to like, really focused on my time.
A lot of my, my thinking is like, how will life, how will I act post newborn? And how can I start acting like that now? And post newborn, I probably won't be like, I will have to have a really good reason for going to a dinner or waking up early to go do something or like being outside of the, the three block radius of my house and office.
And so I have to just justify a lot of those. Decisions more I think just like when you're young, I was more of a yes person. I would just say yes to everything. And now I'm trying to be a little bit more focused on that. Partly it's because if I say no to things, I can be in the office and actually ship stuff, right?
And the sort of leverage that shows up is a lot greater than just typing really fast.
Jason Jacobs: A typical venture backed company, and granted, they're not all the same, but if you had a stereotype, it's like the value prop is call it, 80 percent of market salary, depending on stage, but not full market salary, so somewhat competitive, but less than you'd make if you went to work at Google or Facebook, or Meta.
And then some equity, and that equity, may or not, be worth something, but if you outperform, it could be worth a lot. What was the value prop in the async typist distributed phase of Gumroad for the team? And how is that evolving as you enter this next phase of more consolidated in person more ambitious, more creative if at all.
Sahil Lavingia: Yeah. I'm trying, I'm still trying to figure this one out, honestly. I think it's important. I'm one of the big sort of Complexities of running a business is, you don't want to just serve yourself, right? And you don't want to just build a company that's purely about serving all of your own interests.
That's part of it. You want to hire people who are like inherently aligned with what you may want to be doing anyway. But at the end of the day, like people have different priorities in life and are in different stages, and if you want to build a company. You can also just choose not to hire anyone. But if you're making that decision, I think it's important to consider that.
The way I currently think about it is that
Jason Jacobs: How did you? How did you when you were in the remote phase?
Sahil Lavingia: So in the remote thing, it was about flexibility, right? It was about, Hey, you can make a good income to your point. Like 80, 80, 80, 90
Jason Jacobs: It was all hourly consulting, right?
Sahil Lavingia: It was basically all hourly, people get paid between 100, 200 an hour, they get to choose their equity split.
So they get to earn based on what they want, like a percentage of the company. And then that entitles them to distributions, profit sharing, dividends, et cetera. So it's like there's upside, in the case of upside and we run the business quite efficiently and eventually it'll all be automated.
So there is like a sort of a, like I think last year was about over a 10 percent cash field. So it's like a pretty solid Return on investment relative to any other investment
Jason Jacobs: Like a bonus paid out as a dividend.
Sahil Lavingia: Exactly. Exactly. So it's long term capital gains taxed. And now I think it's about learning.
It's basically you want to come in and actually push your thinking. It's not about getting paid 80, 80, 90 percent market. So hopefully we still do that, but getting paid 90 percent market in New York is very different than getting paid 80, 90 percent in
Jason Jacobs: it still hourly?
Sahil Lavingia: we're converting people to monthly retainers and full time salaries.
So you can still work if you want to work 20 hours a week, for example, you can still do that. But I prefer it be like, calculated at the monthly rate. That way you don't have to do time tracking. You're more incentivized to actually, indulge AI. For example, the issue with hourly is that inherently, if you can use AI to get your work done 10 times faster, you get paid 10 times less, right?
Like that. The incentives are not aligned in that way. All the remote folks are converting to project based. So we're going to basically take all the work that people do remote and say, Hey, let's scope it out. And then let's like. Basically, effectively, people would have to bid on these projects and say, Hey, this thing I think I can get down for five grand if they can do it in one hour using AI, then boom, they get five grand for one hour's worth of work.
And I think it'll be much like the gig economy. I think they've all talked about this on the Joe Rogan podcast a long time ago, but he called it basically high quality gig work. Ironically, he talks about how it may take 10, 15, 20, 25 years to happen. And it's here today, I think. Or will be in the next few years with AI, but effectively AI will scale people's ability to get work done so fast that the work will almost all work will be gig work basically, just for example, if you want a new logo for your website, you don't generally pay hourly because truthfully, the person who's really good at doing that can get it done in three hours.
You pay 25, 000 because then they can do a bunch of other random crap, employ their friends to do some stuff, build a nice little pretty website. But really the creative exercise of designing your new logo, like the Citibank logo, it's like a napkin sketch, right? It took either 15 minutes or 30 years, right?
Like depending on your perspective of what went into that logo. And so I think more and more things are going to shift to project based. You can even see this in lawyers, like lawyers charge like 1, 500 an hour. We took some of our lawyers make 1, 800 an hour. The reason is because it's just like a hack around.
The work should really just be project based. It's just that the work is so different and flexible all the time that you can't, the sort of time it would take to scope the work would be more than the work itself. And so you just end up with a stupidly high hourly rate. And then you apply that to all these other industries that it doesn't really make sense to do.
So I think I basically think all software. will be project based. It'll be gig work, not too different from what on Upwork, except the bar, the quality, the scope to scale for what you will get will be just far greater. And that's,
Jason Jacobs: I'm not technical, but do you lose although with AI, maybe I'm becoming more tech, but that's another whole discussion. I'm trying, do you the whole familiar, I think of, in my home if I just have a plumber come and it's project based, I lose from the continuity of having the same plumber come again and again who really understands all the intricacies and nuances of my plumbing, right?
Is it, does it work similarly with code?
Sahil Lavingia: I think the nice thing with code is you have x ray vision, right? Like part of what, why you want the same plumber, the same architect, the same electrician is that you can't see through the walls. And so there's things that aren't obvious to a new person, but in theory. If you have all the schematics, if there's no tech debt, quote unquote then you could have a completely new plumber every single time and they should be able to solve the problem.
Jason Jacobs: AI making cleaner or messier codebases?
Sahil Lavingia: I personally think it's making cleaner code because I'm the one using the ai, right? Like I personally think that AI writes better code than I write. It may not make the best architectural decisions that I may make, but on a line by line basis, like I think anyone who thinks they write better code than AI is basically just.
Wrong. For, we can talk about why they are saying that. I think they're reasons, but basically AI is better. It's better at following instructions. It's has more energy. It knows more about every different kind of language. The idea that it's writing messy, quote unquote, like it doesn't write messy code, I've never seen AI write messy code.
It may write the wrong kind of code, but will it write messy code? No, it's generally a smart, like I, I do think if you asked it, for example, if you take any like sort of atomic technical interview question and you ask AI versus a human, I would put AI in the 99. 9th percentile, right?
Like binary tree search. Bubble sorting, any sort of constraint, worded word problem, right? AI will solve, right? It would be like an LSAT, like LA AI is going to be, is much better at doing the LSAT than most humans, right? Like the AI will get, I don't know what O three mini pro high gets on the offset, but my guess is like a one 76 or something like that, like almost a perfect score, right?
And it should, it knows more, it has a bigger brain, it doesn't need to sleep or eat, doesn't get hangry, right? Like on a first priority basis, like it will be better than any software engineer on earth. Like for example, if it makes a mistake, which it may make, I can tell it that it made a mistake.
It'll remember that it made that mistake and then it will never make that mistake again. Versus humans I hire who are very good, who pay a lot of money, I may have to tell three times. Before they really internalize, Oh yeah, I should not do that. Versus the, just a little, imagine you had a list of things, every single thing that that plumber, every plumber you've ever hired had to note everything that they learned that could probably fit, it says a thousand little bullet points.
And then every new plumber that came in, you just gave the thousand bullet points and said, Hey, by the way, these are all of the learnings of past plumbers. So make sure you don't, these things the problem is it would take a human, like Hours to read those thousand bullet points and internalize.
It would be like studying for the LSAT, right? But an AI, you can just give it the thousand and boom, it knows it, it will factor all of those things in. And so the question becomes are you willing to do that? Are you willing to document your code base? And most people are not frankly, right?
So I'm the one generally doing a lot of that stuff. There's like an upfront cost to doing it, but I've done it all.
Jason Jacobs: done through typing?
Sahil Lavingia: It's AI generated, but there is some level of typing the prompting. It's still prompted, right? So I can, I still have to do some research and figure out, okay, how do I co cantonate the file, the repo? Then how can I take that repo? I now have to find a model that is big enough. As a context window, that's 2 million tokens.
Gumroad is 2 million tokens, right? An average Word document, like a page, is 250 tokens, right? A million, 2 million tokens is, is a, hundreds of
Jason Jacobs: Right when we get off this, I need to make a little note that I need to go to chat GPT and learn all about tokens because this is a new concept for me.
Sahil Lavingia: Yeah, AI thinks, right now at least, it thinks in tokens, right? A token is roughly about 4 characters. 4 English characters. And 2 million tokens is about 250, 000 words. Or sorry, 500, 000 words, right? 500, 000 words, like an average line of code is maybe five to 10 words or less probably four words.
And you have 500 W it's about 120, 000 lines of code. This doesn't include all the tests for the government repo. Cause those are tests. Not just the functionality of the code base, but that is about 2 million gumroad, this big code base, pretty large product can fit, I can copy paste about 80 to 90 percent of the gumroad code into Gemini pro Google's Gemini pro model is the only one that I think has 2 million to context window and I can say, Hey, summarize this into a word doc effectively, and then I can put that word doc into the code base.
And then have Devin look at that word doc as it's generating new code and say, by the way, just you would train a new engineer, right? Like a new engineer has to do this. They have to go understand some of the code and then in their head, have a cache, have a little mini summary word doc that's somewhere stored in here.
I can tell you, okay, yeah, Gumroad, for example, the back end is Ruby on Rails. There's a routes. rb file that generates the, that has all the routes. And when you visit a URL, it hits this, then that points to this. And then that renders a controller and that controller renders a method. And then method renders a view to the view has all these instance variables in it that come from the controller.
And each of these may talk to a service or may talk to a model. A model may talk to a presenter. All of that I can put into a Word doc that AI generated for me. And then Devin. You which may use cloth 3. 5 sauna, which only had 200, 000 lines. So we can't 200, 000 tokens. So we can't actually do gumroad.
But it can if it can take us, a 10th of the code base and then the summary and get pretty close. But this is all by the way, these are all bottlenecks that will go away in the next 2 years. These are all hacks that eventually will be able to get rid of the hacks.
Two years from now, every model will have sort of 2 million context window and you won't even need the summary. It would be able to generate it on the fly.
Jason Jacobs: one of the things I'm sorting through for myself is that I can see how AI is making someone like you who comes from a kind of hybrid design and coding background to swing a much bigger bat if you will because as you said what's most important is moving from just the typing or the bricklaying in that analogy, but to the actual development and craft and taste and things like that.
For someone like me who grew up in enterprise software startups and there was an invisible wall halfway through the layout of the office and on one side was sales and marketing and on one side was engineering and product and they didn't talk very much, right?
And I, although I've been a, CEO, overseeing whole companies, even in those companies, I was heavily slanted more towards the, that left side of the office, for better or for worse, I probably shouldn't have been, but that's just been my bent. Historically, and so for someone like me who's setting out to better understand how to, work different and swing a bigger bat, if you will, I'm not that craftsman, I'm not that tastemaker, so do I try to become one?
Or does that mean that I am I will gain more benefit from partnering earlier than someone like you?
Sahil Lavingia: Yeah, I don't know, honestly. I would like to believe that anyone can become, a really good craftsperson, but it does take that upfront investment that doesn't go away, right? Just like you said, you still have to learn how to prompt the AI. The way I think about it is it's a simple multiplication.
Equation, which is let's say times a hundred, right? So if you're a zero at something like AI will not make you a one, right? But if you're a one at something, it can make you a hundred. If you're a 300 at something, it can make you a three 30, 000, right? Like it can give you a it can allow you to clone yourself, but as I, I really think it doesn't.
Replace the thinking aspect of what I do. It replaces the typing aspect of what I do. So all I can do is type a lot faster. There's nothing like inherently that it unlocks for me that I wouldn't have been able to unlock myself. It just speeds up that unlocking a lot. If you're a designer, you can design a lot faster.
If you're an engineer, you can engineer a lot faster. If you're a really good salesperson, you can probably do sales a lot faster. If you're really good at marketing, you can do marketing a lot faster. But if you're not like, I'm not that good at sales. AI until it's we have AGI effectively.
It's not going to help me do sales any better, right? Because if I'm not good at discriminating, if I don't know what a good sales email and a bad sales, it doesn't matter if I can generate like a thousand of them, I'm still going to make I think of it like a cooking competition, right?
A master chef is going to be a thousand amateur chefs. It doesn't matter how many times. An amateur chef tries to compete with Bobby Flay, like Bobby Flay will still win. There's something in his head where he can know what, he can discriminate all of the ideas. And his head much faster than a beginner could try and fail and try and fail and try and fail.
The only way to beat someone like Bobby Flay is to actually become good, right? Actually get to that level of good. And AI may prevent that actually, right? Because someone who just is I'm good enough. I can just go to McDonald's. I don't need to ever become a good chef. And then they never do.
And then that's fine. They still get the median quality is improved and their food intake, but they don't ever become a chef.
Jason Jacobs: Here's what I'm wrestling with though there's different types of expertise, so if you take if you take the example of let's say, automating certain mundane, time consuming parts of being a lawyer, or an accountant, right? There's the or sales! We can start with sales, right?
What would a sales person or team do to get more efficient? They would build out a sales ops function of humans, right? And sure those humans could have some software over time But it was like pretty human driven when it came to setting it up, structuring it, analyzing it, etc and so in a world where you want to start automating on the one hand, there's the sales expertise, right?
And so that salesperson knows how to sell and so they can extend the capabilities of themself or compliment or on the other hand, there's the sales ops expertise of that person, building the tools. But then there's also the Sahil expertise of being a part coder, part designer.
What expertise do you need to, you can inform here's the tools I want, but when it actually comes to building them, what skills do you need and how important are the old skill, not the old skills, but like the traditional skills of engineering, design and product management.
Sahil Lavingia: Yeah I think they will remain important. I think this, like the, they will just not be important in order to scale basically. So for example, instead of having a hundred salespeople, you may just have one, and that one person who would have in the past built out an entire organization in order to do sales.
May just say, actually, there's now tooling that I can use to like, basically have a system that sells, sends the emails, schedule the calls, even takes the first zoom call as an AI on video. That's five to 10 minutes to field more inbound questions. And then when they get to the tier two or tier three point, that's when I engage with them and they're already qualified and all of a sudden I can do this all myself and I don't even need a teamed behind me.
And I think that's what startups will software, at least starters will look like. It's just partnerships. There'll be like five people, one designer, one engineer. One salesperson, one marketer, each owned 25 percent of the company. And they all commit to saying, look, we're never going to need a single other person.
Again, will just be project based. If we need a consultant or we use AI. And if you think about real estate development, like you can build a 300, foot tall skyscraper in New York and your company to do so will be a partnership with three people. Someone who raises the money, someone who does all the business stuff and someone who designs the building and deals with all of that, right?
And maybe a product manager. You have four people, they can build an infinitely tall building, basically. And the rest of the work is contracted out, right? And I think that's where software will look a lot like real estate development, where you will just have partnerships. And you would just have an infinitely scalable set of instructions that kind of go out and the role of the partners is to like, just really control the process, control that prompt the system quote unquote prompt, which is anytime work is done.
This is how we do work at the company. And that becomes like the guiding instructions and anything that's creative is done by one of these four partners and everything else is delegated to the most efficient sort of way, which. Eventually we'll just be robots, right? And AI, like it won't even be outsourced to contractors because eventually the contractors, the reason they're contractors, because there's still some thin layer of creativity.
That's Oh, we need to find the, we need to find them. If they don't show up to work, we need to find a new person. There's still a layer of craft that goes into every single layer of. All of these, this massive, stack of what it takes to build a building. Eventually, you'll have robots and AI and like perfect quote unquote instructions.
And then you will have this massive verticalization where all of the sudden you'll have four people and a team of 100 robots that are now building every building in New York, right?
Jason Jacobs: What do you think will happen to venture capital and venture backed startups over time? And that might be a question in terms of what will the implications of AI be? But also just, where do you think it's going more generally?
Sahil Lavingia: Yeah, I think that they have to find a new industry. I think at the end of the day, like software is old, it's mature. It's not a cottage industry. I think venture capital performs best. I remember when I was raising the rolling fund, I had a short call with Andreessen and he was like, look at the end of the day, venture is about investing in the stuff.
No one else will invest in. And the thing is like software is mature enough where you can, if you can evaluate the cost, the benefits, the risk. You can raise debt. You can raise money from angel investors. You can raise money from all the public with crypto. Who knows what the venture should go find stuff that no one else will back, right?
Crypto was that at one point. I would argue that like space, genetics, genomics, robotics. But software itself, like the, just doesn't really need, deep seek, I think blew a lot of people away because I think it's signaled like the end of venture and software, like there's just no need, like the entire industry.
I think should realize that the point of adventure capital is to like fund adventures, like to fund the discovery of America, not to fund like a building getting built in Europe in the whatever 1600s or 1500s, 1400s. And similarly, I think like there are frontiers of knowledge that we have very little information about today and venture can go fund those explorers.
It can go fund people who, they're super young, they're super hungry, they have nothing else to do. And they can put together science and research and try to apply it, which is what computer science, we still refer to computer science as computer science, but most people with a computer science degree aren't scientists.
They've never done anything that required making a hypothesis and seeing if it worked or not. And most likely it led to failure and then writing a paper so that other scientists can replicate it, etc. It's a completely sort of engineering discipline. And so I think we should go back to science, basically.
Venture should be funding science.
Jason Jacobs: So do you think directionally, one implication that this leads me to is that software being provided by vendors may be on its way out or consolidating and there'll be a rise of more companies that just build their own tools in house because it's so much cheaper and more personalized to do
Sahil Lavingia: I hope so. We're open sourcing all of our software at Flexile, all of it. Yeah, basically.
Jason Jacobs: going to be commoditized anyways?
Sahil Lavingia: It's all going to be commoditized and just like Linux and Ruby and Python and JavaScript and TypeScript, like they're all open source, the winners all. You're seeing it now with AI with DeepSeek, right?
Like in the beginning, they're closed source. They're highly creative. The science is being figured out. The science is hard. And so you need to incentivize people. You incentivize people with equity. The equity has to be worth something. So you have to like, Have equity be worth something by selling commercial software, right?
You can't sell commercial software. If it's open source. And so you have to not give it away, but eventually over time, it's not creative. It's about efficiency. It's about replication. It's like what deep seek did. And the incentives that drive people are like fame, notoriety, reputation, right?
Like different things that and so you eventually have the sort of sciences figured out, then you have the commoditization via open source, right? For example, if you want to design and build a building, it's open source, quote, unquote no one's keeping those secrets away from you, but, drug discovery is still closed source because there's still secrets there, right?
There's still science to be had. And, yeah, I think 100%, there's a great article by Chris Paik, The End of Software, that he wrote, which he makes this case. Basically, the cost of generating software will go to zero, and AI will be able to generate all of it for you. And so the cost being able to sell Software is going to be very difficult.
And I still think there'll be software. You can sell video games, music, art, film. These are still pieces of software that people will still buy, but enterprise software, I think, and I think you're totally right. If it's possible for the reason a lot of the software was so successful is because it's offered you to really expensive.
So like you couldn't, like you could use. Deal or gusto, or you could hire someone who costs 400, 000 a year to build your own mini version of custom internally, right? What are you going to do? Like the cost of software has to be so high for you to do the in house stuff. But if the in house goes to zero, if you can just have an AI agent say, Hey, I just need a tool that like helps me collect invoices from freelancers and pay them monthly. And then 20 minutes later, you have a tool, it's completely free. You disconnect your Stripe account or your bank account and you're, it's free. Then like, why would you use all of this stuff?
And so I, that's the bet we're making. Flexile is a Deel competitor. We're never going to win against Deel, frankly. They're too big. They have too many salespeople. But if we open source, maybe we can. The issue with closed source, the reason it eventually gets replaced by open source is because closed source may bleed to the best software, but never the cheapest software.
And so eventually you have a fragmentation. You eventually have a lack of a monopoly, which leads to a lack of a standard, which is then a subpar customer experience, right? Like you the, having two internets would never work, right? It never, it doesn't lead to the best consumer experience.
But if you have open source, then you can actually have the cheapest be the best and eventually have a standard. And so eventually there will be a standard for the, we all use the exact same open. For example, the WordPress, right? Half the internet runs on WordPress. Why? Because it's open source.
All servers run on Linux. Why? Because it's open source, right? Every browser runs JavaScript. Why? Because it's open source. It's not proprietary. Not one company can make a bunch of money off of it. It's better for the world though, right? And so I think the question becomes do you care about making money?
Do you care about making what's best for the world? If you care about making money, you can go do something novel and interesting in science, raise money, keep it closed source. If you're like us and we just want to build cool stuff and help the world get better, we're going to open source everything.
I still think there's a way to monetize, which is basically licensing, which is saying, Hey, Disney, you want to use your own, you want to use Flexile. You want to self host it, just fork our code. Don't even have to talk to us. But if you have over a hundred employees or a thousand employees or 10, 000 employees, maybe you just pay us a licensing fee.
I actually would like to believe that this becomes a new business model. That's much more popular because the reason you wouldn't do that before is because then you'd have to go collect all these licensing fees from people, which then you have to hire lawyers. Who make 1, 500 an hour. So you have the same problem, but on the other side, which is you have in house lawyers are too expensive.
So that business model doesn't make sense. But eventually if lawyers also the cost of lawyers go to zero, then all of a sudden you can say, actually, we're just going to give away all of our software for free, and this is the business model we're pursuing an anti work. We're going to give away all of our software for free.
It's going to be all open source. So anyone can use it. People can use AI agents to fork it. People can use AI agents to contribute back to it. There can be projects they can bid on and even get paid for that work and get dividends in the future earnings, if there are any, then we'll just have a team of robot lawyers go around and see who's using the software and ask for money. That would be our business model. We'll be asking for money. We'll be like musicians, right? Like you can listen to Taylor Swift for free. You can just go on YouTube and type Taylor Swift and listen to all the music. And every once in a while you monetize the head of the curve. You say, Hey, I'm performing a show over here or Hey, Disney, you're using our software.
Could you give us like a hundred grand a year? And then all of a sudden the business model is amazing because you're 100 percent margin, right? We're not hosting it. We're not hiring the software engineers to work on it. The energy to power the servers is not being paid by us
all they have to do is send us a check once a year and we're all good, right?
We're still figuring it out. But I think that's The current approach, the best answer I have for you at the moment.
Jason Jacobs: But I love businesses that philosophically it's almost like the Craigslist model. You just generate so much value that if you ask for and return a little smidge of value back relative to the value that you're providing, people are overjoyed to pay. And in fact, they'd be, eager to pay more than that because it's like still pales in comparison to the value.
And then you can make a nice living and you can feel good because you're just delivering so much value to the world. It's hard to pull off and practice, but that's the dream.
Sahil Lavingia: Yeah. I think the other thing I like about the model is it'll test do we actually like doing what we do? Because if you do, then you'll keep doing it regardless of if the money's coming in or not, you'll continue to make music. But if we stop, then all of a sudden it's a good test of Oh, maybe we just did it for the money.
It's a good filter for recruiting for, if you're willing to work on open source software, you love doing it, I heard a joke that's it's like open engineering is like the only thing people do on the weekends. People contribute to open source software on the weekends.
It's a good task of like people who really enjoy what they do, and that's, and those people are the best at what they do, and it's the people I wanna work with anyway.
Jason Jacobs: So do I have time for a quick free consulting question?
Sahil Lavingia: yeah, a hundred percent Anytime. Anytime here or later,
Jason Jacobs: Great. So I'm heading down this path where I'm having a lot of discussions like this and then I'm publishing a weekly newsletter that's essentially just like an investor update for my non existent company. My investor update for my walkabout, right? It's like public accountability and also just calls out where I need help and who I want to talk to and I get a bunch of goodness that, comes from that, and people can see the progression over time, and it'll resonate with some, and it won't resonate with others, but you can find your people, right?
So as I'm doing that, I'm already seeing that it's okay like the podcast editing process, the script helps me compress it a lot both from cost and from time and control, but it's, there's, I'm still doing some Hunting and pecking and I think there's opportunities to make that more efficient over time the prep process same thing the publishing and promoting process same thing the show notes the transcript The sound bites for marketing clips the posting them There's insights that can be gleaned from the aggregate body of knowledge that's coming from all these discussions and those insights can inform Us, internally, and us is just me to inform our direction, what topics we cover next, etc.
But it could also be content in itself a living, evolving body of what's the zeitgeist say about what's like where all this is heading. And so what I want to start doing is plucking microtasks for example, Take podcast prep can I just, can automatically when an episode gets scheduled, can it just like neatly go out and comb all the podcasts Sahil's been on and all the blog posts he's been writing about and what new products Gumroad did or Anti Work has shipped, et
Sahil Lavingia: Yeah.
Jason Jacobs: And just lay it out on a platter for me and just make my process more efficient to, to get up to speed. And I want to start picking up more processes like that over time and build like an agent army. But I go to, a tool like Repl. it or even Claude or, I'm still chasing my own tail around.
So for starters, I'm going to try maybe some pair programming with like an old runkeeper PM's gonna do pair programming. We're gonna record it. We're just gonna try to help me build my first agent. What advice would you have for me to do this knowing that if left to my own devices, I'd just be going and ripping through having these discussions.
But at the same time, there's value in getting my hands dirty, but it's not my comfy spot or, my zone of excellence. So what would you do if you're me?
Sahil Lavingia: Yeah. One, I would basically have what I do. A lot of people honestly over complicate this kind of stuff. I just have an app, Apple notes and I just have Apple notes for certain things. If there's something I do often, I just keep adding to that Apple note anytime I need to do it, for example, let's say you have podcast prep for me, right?
You can just say, Hey, I did all of this podcast prep for Sawhill. I'm interviewing this new gas next week. This is their name. This is their Twitter URL. And you just go to deep research. I would, is what I would probably use now. You could probably try deep seek R1 and and I actually did it. I said, doing a podcast with solid India, give me what he's been up to recently.
And I'll just, and then, it'll, they'll ask some follow up questions and then it'll take five minutes to 30 minutes and all it's doing is combing the internet, right? And just doing what you would do probably, right? Just Google searching and then clicking like the top 30 links and then reading all of them, processing all of them, and then compiling them.
Quote unquote, a report, right? A research report, just like if you had a grad student or an intern or something like that. And then it would come back to you with a thousand words that you would then read. And yeah I think a lot of people, they overcomplicate things. They think they need a SAS, but really I think the difficulty in this new world for people building software is like, how do you monetize it when it's so easy to just use Apple notes, which is free and then go to chat GPT, which is just really, you're paying for the model that You know, probably not going to train your own model.
And eventually there'll be like, open deep research. Someone is working on open source will win on this. So eventually I really think actually like the sort of like long, long, like the monetization that lasts for software is actually selling the model.
And try GPT do like eventually there'll be an open source version of cursor and the, they're the fastest stop software startup to get to 100 million, like eventually it will go down to zero again unless they find a new, at least on their existing business model, because no one's going to pay for code editing longterm, like five years from now, it's, the quality of crystal will be available completely for free.
If maybe it already is and no one has noticed, but they'll have a coding model that they will have trained using all the data that they're generating. And so I think it's, like a Fred Wilson wrote that Bob puts a long time ago at freemium, right? Where like you have a free product and then you have a paid product.
I think it's like a free product will give you the data and then you use the data to train a model, but your SAS is free. It's the model that costs money to use. I think that's maybe one direction for where I think like internet consumer software could go, right? Chat GPT is quote unquote free to use, but like having access to Oh three mini pro is 20 a month or something.
I think I could see the same thing playing out for a lot of things. But yeah, basically I think it just being self aware about like the things that you ask yourself and writing those things down. I think that's like the same thing I tell like the engineering team here, like you can have Devin write all the code.
And if you think you can't, that's just like a failing on your part to not be able to be self aware about what answers you're trying to get to before you start writing the code and then just write all of those things down. And then Devin can write the code for you. Like at the end of the day, there is a task switch from creative to wrote.
From what you said, I think was like avoid to delight. There's to me, like creative is delight and wrote is avoid. And as soon as you can notice your task, start at really creative and high level abstraction and slowly, for example what's the prompt that I might need to, in order to get the best possible Google results like that may be creative.
But then reading all 30 of them is pretty boring, right? And so you can at some point you can be like, okay, I'm done. I'm done with the part of the job that I'm delighting in, and then I can pass the rest off to AI to finish, right? That's so right now it's doing it, it's read 17 sources so far. It'll be done in three minutes.
Jason Jacobs: So I'm finding that there's a there's like an evolution from Oh, I'm hearing a lot about this stuff. What does it mean to then? It's Oh you can use it for this and you can use it for that. And that's what it's capable of. And then, Oh, like here's some ways, some baby steps of ways I can start to use it.
And then, like trial and error, experimenting with different tools to find workflows that add value. And then it's how do I automate those? And so you just get. Deeper and deeper the way that there was a lot of emphasis placed on learning to code in the internet era what emphasis, where should the emphasis be now to better equip each of us to adapt to this next era and where are the opportunities slash pain points, cause I've been experiencing some of the pain points firsthand in terms of helping people along.
Is that something that you're thinking about with anti work as well?
Sahil Lavingia: Yeah, I think, I don't know. I honestly think that for some reason, the stuff is the same, like it's still product, it's still design. It's still engineering. I still think like a hundred years from now, it'll still be those things. Unless we have whatever artificial super intelligence or something.
But still then sales, marketing brand, connecting with other people, like philosophy, physics, math, English literature. I think that those are fundamentals. Like those are just. Humans will always operate at that level of abstraction. Everything under that or above that ends up being like, what we do in order to participate in capitalism.
But I think there's the base level of basically what you learn in high school, frankly in college that stuff is just critical. The stuff that you learn, regardless of your degree. It's how I would think about it, right? Like the stuff that you're going to learn regardless of if you want to be a philosopher or a psychologist or a mathematician or a scientist or a researcher or an engineer or a designer, right?
Calculus, for example, I think is essential. Everyone should learn calculus. To some degree, you should learn derivatives. You should learn integrals if you're, you should learn color theory. Everyone should learn to paint, right? It's to be non optional. We should get rid of some, people are like, oh, everyone should learn to code in high school.
It's I think you should learn like math, but like coding, I don't know. It's not essential. I've seen people who are good at math, learn how to code in three days because yeah. You realize, Matt, coding is just doing math in a recursive way, right? There's a lot of equal signs in coding, right?
That's all it is like you're just doing basically algebra. Yeah, I would say that it's not that different. I actually think it'll be the kind of same stuff that's always been important. And. Really, it all comes down to just like understanding yourself and if you understand more of how you work, you'll understand how other people work and what they care about.
And you'll be successful in all of the ways that matter in life, which is connecting with other people, finding a place in society where you feel you're adding value and earning enough of an income to like benefit from the other things that people have. By the way, I got the report back if you're curious about it.
Jason Jacobs: I'd love to see it.
Sahil Lavingia: I'll send it to you after, but yeah, basically it says like Gumroad. Gumroad. Is profitable we're focusing on this thing called anti work, which is renaming the company, which is about eliminating the unfund parts of work. We have 2 new projects flexile and helper. We're doing, I'm doing some investing via capital.
I wrote a book, I recently wrote an essay called God node, which talks about founder led decision making. And, oh, current focus and shifting direction is about, yeah, basically creating tools. To remove the tedious parts of work for himself and others but consistent about doing more with less and enabling others to do the same in their creative businesses.
It's not amazing. At the end of the day, like AI, I think of it like a really good mid twit, right? It's like the mid twit me it's really good at creating like, research at like a hundred IQ level. But being creative, being weird, being eccentric, having a point of view and a strong opinion, making spelling errors, on purpose students will always be better at those sorts of things.
Jason Jacobs: You say it'll be more of the same, but it, at least at Gumroad or anti work, it seems to be evolving quite substantially how you work and you guys are easy to do it or are able to do it because you're small and nimble and in control and don't have a lot of decision makers.
But if you're Goldman Sachs, or you're Pricewaterhouse, or you're you know, like a big established firm who's done things the same way for a long time, right? It's the first inclination is, oh, we need to do more of this, and then it becomes like a bolt on that's off to the side, that's like a, it becomes a sustainability initiative, right?
Or it's not actually gonna transform the core, but hey, it's gonna, it's almost yeah, our sustainability stuff comes out of our marketing budget, that's the 5%, right? Not the 80%, right? Do you think these firms will evolve or do you think these firms will be displaced?
Sahil Lavingia: I think mostly they'll be displaced. I think it's very difficult. I think it's really about who, who makes the decisions in the company and who owns the company. And if those are the same person or the same group of people, you may have an incentive to really change. But truthfully, most of these organizations are not owned by the founders.
They're not owned by the people making the decisions at this point. They're like the federal government pre doge or whatever, right? There's just like a level of the people getting paid or not the people. The owners and so there's not a huge incentive to really it's hard. It's like an insane amount of creative willpower in order to like make these changes and there's a lot of risk associated with them.
And a lot of people may not like you very much, at least temporarily. So I think. Really, there has to be an incentive to drive that. And at the end of the day, the biggest incentive is you're a new company and you want to displace the old big one, right? Like it's hard to compete against that. So eventually I think it'll be displaced, right?
I think but it'll take a lot longer than people think because there's just so much institutional inertia.
Jason Jacobs: but I'm going to ask that same question again a different way and see if I get a different answer, which is, let's say there's a high up at one of these. Established companies that have done things the same way for a long time or a number of them from different industries that hear this episode and listen to all the things you just said, and they say I've already been feeling this way, and I'm so anxious about it, and I don't understand what's happening, but I know it's not going to be good for me.
I want to stop this train wreck from happening in slow motion I want to right the ship who do they call, what do they do do they just put their tail, do they just brace themselves for impact or go leave and start something from zero or can transformation happen and who I don't know.
Who's equipped out there? Is it McKinsey? Who's equipped out there to help them navigate through this transition? I know
Sahil Lavingia: Yeah,
Jason Jacobs: will say that it's them. But who is it really?
Sahil Lavingia: yeah, maybe it is the, the Deloitte's of the world or whatnot. But I think and they're at the end of the day, they hire really smart people and they pay them really well. So it may well be them. I've even thought about maybe anti work should eventually become like more of a consulting firm and teach
Jason Jacobs: You could help me. That's why I keep pushing you on this. I'll pay, yeah. I'm finding I, I've got needs and I bet I'm not alone,
Sahil Lavingia: yeah, I think at the end of the day I would love to believe that, a really smart, hardworking person at one of these companies can build leverage for themselves and say, Hey, I want to like, do this. I want to embrace this change. And we can take this function and make it a lot better.
And you just have to convince it. The higher ups, right? You just have to convince the board, or the CEO or the combination of the 2 in order to make this change and get the lever, Elon was not able to do this by himself. He had to convince the president and the chief of staff on the head of the own be And I think you have to have a track record of doing that.
I think I would build a prototype. It's Hey, on the side, I did this thing where I know we have an organization over here that has 10 people that does this, but by the way, I was able to build a little prototype and we should really scale this thing up and can you cut, do what you want with it, but eventually, you could become the VP of AI at this company.
I think it's possible. It's certainly possible. I think eventually, you get both. You get the displacement. You'll eventually have like new companies that will scale and compete, but also you will have. Innovation that happens at these companies, because eventually they will feel the pain and change and iterate and pay for the software or the consulting in order to change in these kinds of ways.
Or you'll have the old people leave or retire and you'll have the new wave come in and just operate like this by default. Eventually I think you end up with both scenarios. It's never like one way or the other. I think it requires just like a lot of interpersonal skills, right? Like it just requires a lot of figuring out like the incentives and does their skills, frankly, I don't know even if I have them frankly, right?
I think I'm good at running a company because I can make decisions. And I was like, go from Pinterest because I was probably a bit too, bullish or, I was a bit too transparent about what I wanted to do with my time and life. And, I think. It's not a skill I have to censor myself too well and manage like people's psyches and whatever is whatever politics, I guess it would be called, but
Jason Jacobs: Me neither.
Sahil Lavingia: but maybe that's what I should do.
I definitely thought about that. Maybe that's a skill I need to learn. Maybe I should sell the business and try to do this at a larger company. Maybe that's like the right approach to having a large impact or I don't know if it is, but I think it's definitely I would learn a lot trying that.
Jason Jacobs: We're pushing almost an hour and a half, so I don't want to keep you, but I will ask is there anything I didn't ask that I should have?
Sahil Lavingia: No, I don't think so. It'd be good to hang out in person sometime, so let me know if you're in New York and I'll let you know when I'm in. I think you're in Boston.
Jason Jacobs: I'm in Boston. Yeah.
Sahil Lavingia: Yeah, I'll let you know next time I'm in Boston, and we'll be fun to chat more.
Jason Jacobs: Yeah. And we'll, yeah, keep this dialogue going for sure. I have no doubt you're going to be Doing interesting things, learning interesting things, building interesting things and hopefully I will as well. And wouldn't be surprised if they intersect like they did with the rolling funds back in mid 2020.
What about for anyone listening how can we be helpful to you? Or is there anyone you want to hear from? Anything you're trying to solve? Anything you want to put out to the audience to, absorb and or mobilize for.
Sahil Lavingia: Yeah, just check out my I have a YouTube channel. Actually, I've been trying to publish some videos on this kind of stuff. I have a video on like PMing with AI. That's cool. So if you just google YouTube Sawhill Lavingia.
Jason Jacobs: I tried to take your, I tried to pay for your course, but your course on this stuff
Sahil Lavingia: Oh, yeah, I took it. Oh,
Jason Jacobs: but it wasn't a, yeah, it wasn't available
Sahil Lavingia: yeah, I just decided I'm going to make all this stuff free.
I just don't feel like I want to charge people for it.
Jason Jacobs: Oh, so is it up?
Sahil Lavingia: I haven't published that specifically yet. But eventually
Jason Jacobs: I want to take that thing.
Sahil Lavingia: Yeah. Awesome.
Jason Jacobs: Thanks for coming on. Best of luck. I'm looking forward to watching and cheering you on and to revisiting this. I'm sure at some point we'll have you back on the show for an update. Maybe in four years. Yeah.
Sahil Lavingia: exactly.
Jason Jacobs: Thank you for tuning into the next, next, if you enjoyed it, you can subscribe from your favorite podcast player. In addition to the podcast, which typically publishes weekly, there's also a weekly newsletter on sub stack at the next, next dot sub stack. com. That's essentially for weekly accountability of the ground.
I'm covering areas I'm tackling next and where I could use some help as well. And it's a great area to foster discussion and dialogue around the topics that we cover on the show. Thanks for tuning in. See you next week!