Gumroad 2026 Annual Meeting: New CEO, $1B to Creators, AI

I think it works better at conferences.

Anyways, thanks for coming. This is our 2026 public annual meeting. We used to do these quarterly. We took a break, and so we're excited to do this again and give an update. It's been a year since the last one. You can find the last one on YouTube.

So yeah, we're excited to give you an update. A lot has happened with Antiwork.

First thing that happened is we actually ended the last board meeting a year ago with a rendering of an office in New York City that we were working on, which I thought would be pretty cool. Ironically, looks very similar to the room we're in right now, which is basically this like U, C shape, the goal being that people should collaborate and like the point of working together is to work together and make each other better, not to just work in little island silos. And also wanted to inspire people to actually want to be in the office, move to New York, and all these things, because obviously working from home is pretty awesome as well.

And so that's done now. This is what the office looks like.

These are not AI-generated. These are real. So you can see the U-shape right here. Oh, thanks for the heart. The conference room, outside the conference room, we have a whiteboard, and then we have this cool bar. So the centerpiece of the office is this kind of like a conversation pit style bar. So you can work on your laptop. So you can kind of see that in the rendering, like right here. So yeah, so that's one thing is we try to get more people excited about moving to New York City. And the other big piece of news is I'm no longer the CEO of Gumroad. So I'll be transitioning it over to Arshad shortly. And I guess I just wanted to talk a little bit about— thought it'd be nice to like, I don't know, reshare the history of Gumroad. It's been a while. And also maybe explain myself a little bit. Though the reason primarily is on the screen. This guy, his name is Sen, and he's very cute.

And he was more cute than I was expecting. So yeah, it's hard. It made it very hard for me to come into the office.

And actually, the moment I realized that— I tend to move pretty fast, so it's not like something I was thinking about for a long time. So like when people found out about it to when I just even started thinking and exploring the idea, I don't think it was more than a couple of weeks, but I remember there was some, I think it was because of Sen, but basically I was not in the office that much and I asked Irshad to water the plants. I don't know if you want to be here with me.

Yeah, I'm going to be here.

Nice. And I was just thinking like, you know, if Irshad is watering the plants, that he is kind of already the de facto CEO of the company because that is kind of like the ultimate, job of the CEO of the company. It's basically to do all the work that doesn't fall under anyone else's job description. Of course, you also get to do kind of like the highest level stuff, like the decision-making and judgment, actually deciding what to do, who works on what, et cetera. But really, I think that's all fun and cool.

So, you know, that doesn't make a CEO.

What makes a CEO is someone who's willing to do all the other stuff. So you can see the plants nicely watered thanks to Irshad. And you should have seen the plants when I was the CEO of the company and they were not thriving. So pretty good metaphor, I think, for—

So yeah, I basically— Gumroad was my baby.

I had a new baby. Hard to be a founder of two things at the same time. So I think this will be a really, really, really Awesome, awesome change for the company.

Anything else to add?

Yeah. So I've been with the company for around 8 years now. I've been a software engineer all the time. So this is also like a new challenge for me, which I'm looking forward to do it really well.

Anything surprising so far? Positive, negative?

A lot of decisions to make.

Yeah.

That's a huge change.

Yeah.

So previously I just had to take care of the engineering side and the infrastructure side, but currently I have to Make a lot of decisions on a daily basis. Talk to clients. What are the plans?

Of course.

Yeah.

Yeah. Just things like that.

Awesome. Anything from the chat, Daniel, that anyone ask any questions or anything about this?

Awesome.

The other thing. Okay. Yeah, I guess I want to skip a little bit and I'll go back.

I just wanted to just, I think, as a nice context, just to share some information about the company people may not be super familiar with at this point. It's because it's been a long time. Started the company in 2011 in San Francisco, raised a bunch of money, did the sort of typical— I mean, it's kind of crazy to like look back because the numbers are so tiny, you know, like, and it was like complete startup mode and now it's like sustainable business and the numbers are like so much larger. Like creators are making so much more money. But yeah, basically raised our seed round, Series A around here, then didn't need to raise money for a long time, and then found out about here, like, oh wow, like, this is not the— like, VCs want this kind of graph, not this kind of graph. As an owner of the business, this graph is great personally, but that's not why you raise money. So we weren't able to raise our Series B. We had to lay off a bunch of people, went from 20 people down to 3, and then just me for a while. And then Steve came back, helped out with—

he's still here.

Shout out Steve.

Hello.

Hello. Classic Steve hello. And then we sort of started hiring.

That's actually—

Arsha probably joined around— when did you join? I guess right here.

August 19.

August 19?

August 17.

August 17th, right here. Yeah. So this is where we started growing again, taking our sort of, as you'll see later, like our pretty minute earnings and reinvesting it in a few folks. But we didn't have that much, so we didn't hire full-time. We hired sort of like third-party contractors.

So yeah, kind of crazy.

You never know what can happen going from contractor working hourly for a firm in India and now, and then going full-time and, or contract directly and then full-time moving to New York and all sorts of crazy stuff can happen. Then that's one of the coolest things about Gumroad is it's truly life-changing for people. For creators, obviously it's life-changing. For me, it's life-changing. For lots of people, it's life-changing. So pretty cool, pretty cool stuff. To reflect on. Yeah, so that was that period of time. And then what happened here? COVID, crazy. You can kind of almost see COVID on the graph. And so we had a crazy sort of like, you know, 3 or 4 years in March 2021, pretty well timed. We raised a crowdfunding round, $5 million at $100 million valuation. From 7,334 people. And then we hired a bunch of people. At our peak, we were 48 people, and then we slowly have shrunk since then, though we're now growing again, kind of to mirror a little bit of, like, what this chart actually looks like, which makes sense, I think. And yeah, I think that's kind of what I wanted to cover. Is there anything else notable? We're kind of trying to be almost like a boring company again, kind of like we were in the beginning, pre-VC, just like a few people being really productive, shipping a really great product, being profitable and sustainable, and hopefully spending 20% of our time doing cooler, interesting, more experimenty things, because it's fun, and we'll see how things go. So now I'm just chairman of the board. I don't know if John from DPA, DLPA, DLA?

Is it DLA Piper?

DLA, yeah.

Awesome. We just switched corporate law firms. So we worked with Cooley for this period of time. Actually, Cooley basically helped us start the company. So Matt Bardis was our partner at Cooley. Cooley is an amazing law firm up until— wow, basically now. So like 15— basically, they've been— Technically, Cooley's been with Gumroad longer than me because back in the day, the law firms would start the company and then give it to the founder. So now things have changed a little bit. It's a lot more efficient. But our partner Matt pivoted within Cooley. He gave us a new partner, Josh, who's been awesome. And then Josh is now at at DLA Piper, and so we're now working with them. But it's super new, so we're— I can report back in a year how you guys are doing versus Cooley.

Uh, anything else to add, John?

Looking forward to supporting you guys.

Awesome. Thanks. Uh, as you'll learn and probably already have figured out, we're kind of strange. Strange client. But we'll, uh, we'll make your work interesting and unique, which tends Tends to lead to some billable hours. So hopefully we—

The interesting ones are the fun ones, man.

Exactly.

It'll be fun. As you'll see, our slogan is "Make work play." So it better be fun.

There you go.

Exactly.

All right, cool.

I'll hand it off to the new CEO, and I'm now just relegated to chairman of the board.

Very boring.

I know I have a separate microphone, yeah. So hello again, and thank you, Sahil, for the great opportunity. I'm sure it will be a good challenge for me as well. Since the last board meeting, we hired a few awesome folks, starting with Daniel, who is now the head of product. We have Daniel here. That's a camera. Camera is over there. Yeah, then we have Gianfranco. Joe unfortunately couldn't be here, he's from Indonesia. Then Shriram, Shriram's from India. Sahil from Canada, and Immanuel is here. So we had 6 people since the last board meeting. And we are hiring more people. We have 3 open positions in New York. We are planning to hire in office. So if anybody is interested, please feel free to apply for these positions.

How do they apply? Actually, don't tell them.

We don't want people who need help.

Yeah.

Don't tell them.

So figure out how to apply and apply. Yeah. So Sahil went over the create earnings. Should we go over it again?

Yeah, I guess the only other thing I would point out is I guess it depends on your perspective, but just because I've been at the company since basically the beginning, like this looks like a great chart to me. Some people may, you know, if you were depending on how you kind of look at things, but I try to sort of like almost Remove the COVID bumps. Think of that as like, wow, that was a crazy alternate reality we lived in. But overall, I think the trend is pretty positive. You can almost like remove it, right?

And draw like almost kind of like a line.

Keep a line like this.

So that, you know, that's the hope, right? Like, yeah.

And that's roughly, hopefully what I would predict is like our creator earnings should go like this. Hopefully we see something like this, but you never know. You know, there could be some existential threat that happens. Like past performance is not a predictor of future results always, but. I tend to think, and it's funny how also one bar chart, if this one was here or here, kind of shifts your—

there's a lot of sort of overfitting, I think, on recent data.

But I think overall, I think exciting. But creator earnings is creator earnings, right? This is the economy, not necessarily Gumroad as a business. Gumroad as a business is more in our control than this, which is going to be up to creators, you know, at the end of the day.

True. And the cool fact, we have paid over $1 billion to creators so far. Yeah. So we have a good figure, good number to showcase. And this is the company revenue. Yeah, so if you look at the previous slide, we can see the creator earnings have gone down a bit, but the company revenue has actually increased. That's because the company revenue is is more in our control. We can reduce expenses, we can slightly increase fees, and yeah. And for this chart, we have a more positive growth, right? Starting from here, the company revenue is going up, and in 2025, we had $17.8 million in revenue. Sahil, anything to add on the company revenue?

Yeah, I mean, I think most people would see this and instinctively think COVID is that spike at the end as well, but it's not. So I guess, Claire, it's important to show how much— what COVID did is actually the first bump.

Yeah, this bump, this portion, right?

Exactly. And so it actually didn't— I mean, obviously did a number on our revenue, but you can see post-COVID we've actually done much better as a company. And so hopefully that continues.

Yeah. And one of the reasons why the revenue went down from here, because Stripe wanted us to ban the adult content on the website.

Yeah. So if you go to the first, the slide before this, basically the revenue being down here is basically just because also our GMV went down, right?

So this was the month when Stripe told us to ban that account, right?

Yeah, exactly. Yeah. Stripe basically said, sorry, you guys can't support this. Not like we're excited about supporting it, but sort of don't ask, don't tell. And then Stripe got really aggressive about sort of finding and reporting and finding, etc. And so yeah, almost like I would say arguably that last drop, like if that hadn't happened, I still think we'd be closer to like that sort of $13.5 million axis. It's really— I feel like that almost like white box is like the lost, the, you know, like lost economy, which is, you know, people selling comics meant for like adult consumption and things like that. Obviously like not 18+, we're not allowed to sell porn or anything like that. But even stuff that was more gray area, et cetera, no longer allowed, unfortunately.

Yeah.

And we can see the bump in January because historically we have seen a bump in January, right?

Yeah.

In previous years.

Yeah. January is always a pretty good month. Pretty good month for us because I tend to think it's a New Year's resolution thing. You want to quit your job and go full-time or go independent or that sort of stuff I think tends to— Gumroad I think is a virtue company, like using Gumroad is a good thing.

Sure. And this is our EBITDA. So we did $5 million in EBITDA last year. So EBITDA is earnings before interest, taxes, depreciation, amortization. So basically shows the profitability of the company. And in 2025, we had 5.1% year of year-over-year growth, which is great.

Yeah, and I think, I mean, this shows like it's another way of looking at the history of the company. Basically like we burned, we started the company and we just burned more and more money every year as we grew as venture funding. Like this, that when you, you don't see it, but when you look at all these AI startups that are growing like crazy, their EBITDA is gonna look like that, right?

Yeah.

Like larger and larger losses for a long period of time.

Uber, like any VC-funded company is generally gonna look like that.

And then you get sort of marginally profitable and then one day you're actually profitable business. So that it's basically a smile.

Like it kind of looks like dirty data a little bit, but it's actually not. It's actually incredibly— consistent because it's almost a perfect like curve you can draw like this.

Yeah. As you, you know, it is like, it's kind of like you can think of it like a plane. Think about this as like the floor.

Like if you hit this, you die. You're flying a plane and if you hit the ground, you're dead.

And we got really close to being dead and then we took off. We built, we, yeah. And we're self-sustaining.

We don't have to land the plane anymore. We have a nuclear engine on board to kill the metaphor.

And this is our cash on hand, the balance sheet. Currently we have $6.2 million at the 2025 cash close. So one of the reasons why it is $6.2 million is because we did $6 million in stock repurchases, $2.5 million in dividends, and $1.9 million investments. Sahil, do you want to talk more about the stock repurchases?

Yeah, so yeah, we, every year, I mean, I constantly try to buy back our earliest investors for a couple reasons. The major one being I think it's a good deal for the company. And so yeah, we, in this case, purchased shares back from Collaborative Fund, Accel Partners, and Max Levchin, our kind of earliest investors from 2011. They saw a pretty good return. I think the first valuation Collaborative invested in was a million-dollar valuation, and they— the highest price we sold for. So we kind of negotiated and ended up at valuing Gumroad at about $49.5 million valuation. So They made about, I think it was a 20 with dilution and everything like that. I think it ended up being something like a 24x investment for that one. And, you know, that one investment they made in the beginning of the company. So yeah, we spent $6 million to buy back about, it ended up being about like 12% of the company or more, more than that. It also like triggers a clause or negates a clause in our charter, which basically gives me full control of the company. I had almost full control of the company, but there are certain things I would still need approval for as long as basically 20% of the preferred stock that we, you know, effectively 20% of the investors that bought in are still there. And this brings us under that number. So effectively, we've bought back over 80% of the people who've invested in Gumroad, all at positive multiples, which is cool. It's nice to give to feel like you were successful to some degree. And also doing it while we can do dividends and other things. So a lot of founders actually, they ping me all the time, say, hey, how can I do this for my business? I want to also buy back our existing investors. And I always ask them, how profitable are you? How much money do you have? And they're like, not that much money. It's like, well, okay, you need money to buy your investors out. So yeah, anyway, yeah, and generally I think like we don't need a lot of money to run the business. We're profitable, so in theory if we had $0 today, we could still run the business because we'd wake up tomorrow and have more money, like enough money to pay everybody. But I think it's a good proxy to make sure like, you know, we have a buffer. And so, yeah, it just kind of just shows like Gumroad I think is like a relatively healthy business, isn't going anywhere.

Great. Yeah, and we did $2.5 million in dividends last year. And also we invested in some companies including SmallBets, which we are going to talk about in the later slides. Yeah, so this is one of the biggest change happened in 2025. So we had like 32 people in our team before, it is now 14 people. We had 21 engineers, which is down to 5 engineers now. We had 6 designers. We have only 1 designer, which is Laura. We have here, Laura is here. And support remained the same. And previously we didn't have a head of product, but currently we do, which is Daniel. Yeah. So this is a team change and we can see how that reflects on the salaries and compensation side. So previously we were paying in a total of ₹525K per month in salaries, contractor retainer, and amounts and stuff like that. But that is now reduced to ₹129K. So we can see the individual splits here. So to some extent, I think AI helped, AI made this change to some extent. But yeah, but we have, even though we reduced the team size, we have the same level of productivity right now, which is explaining the next slide.

Oh, does this freak anyone out?

Like, on the team, do you see this and you're like, holy shit, am I next? Or like, what, what do people think when they see it? Yeah, or even in context, it should be terrifying. Yeah.

Uh, wait, John, what you're saying?

I feel that during last year, I feel like things are settling down now.

Yeah, I mean, it's—

with AI as well as the company.

Yeah, I think so. I mean, I always think that, but yeah, I think, you know, Bali was, which, you know, was like sort of, I feel like that pivot point for a lot of the like, oh wow, this is really going to, you know, change the company. That was like, when was that? November 2024. So it's been just by virtue of it having been more time, probably things are more stable now that, you know, things haven't been as crazy. But I still think like, I mean, one thing that I think is interesting here is that like, I think 5 years ago, if you looked at, if you could, you go to one slide, Shryer, just to make things less emotional. I think if you ask most people, you know, in 2019, what, you know, what will AI do? Like to jobs, most people I think would have said AI is going to replace customer support, right? Customer support is quote unquote easier than software engineering. And that's kind of like completely the opposite I feel like has really happened where because, and we can talk about it for hours, but effectively AI is much better at replacing sort of the median software engineer than the median customer support person. And the people we should credit for that are the software engineers, right? Basically software engineers by nature of what they do are automating themselves out of a job all the time, which is kind of like, I think of it like, there's a quote, I forget what the exact quote is, but basically like there are certain kinds of jobs where you get, you know, you kind of like get, you do really well when times are good, you get fat when times are good and you get slaughtered when times are bad. And that's kind of like the pros and cons of the role, you know? And support is like, you know, you get less fat when it's good, but you also don't get slaughtered when it's bad. And so it's sort of like just different. But I think it's interesting.

I think it's also kind of interesting that if you showed this graph to someone 5, 10 years ago, they would go, oh, the company is doing terribly. Like, everything must have crashed. Companies are great.

Yeah, exactly.

Yeah.

And I, yeah, I mean, you know, I, and obviously there's like a profitability aspect to this, but I really believe like the, the ultimate metric that the reason to make these changes is not just to make more money, it's to like move faster. And you can move faster when you have fewer people generally. Not always, obviously, like there, But yeah, if you have, there's a famous book, The Mythical Man-Month, right, that talks about this idea that having 100 engineers doesn't necessarily make you 100 times faster than having 1 engineer.

It may actually end up, you spend all your time in meetings trying to align 100 people and actually only 1 of them codes anyway. So you end up with the same amount of engineers roughly. And of course, cloud and all these things have really helped. I would also add open sourcing stuff has changed things because Open sourcing stuff creates a lot of competition for engineering jobs because all of a sudden open source people can do your job to a degree. But customer support obviously can't, right? Because how do you do open source customer support without exposing customer information, privacy, et cetera? So I think in that way it's also sort of a more defensible job.

For example, I work at the IRS now.

I'm on unpaid leave, just to be clear.

And these views are my personal opinions, not those of my employer, et cetera. Things I have to say now.

But the IRS is almost entirely customer support. That's what— and we can't outsource that. It's taxpayer data. So we have 80,000 people who do customer support and less than 200 people who do like ship product for like IRS.gov, et cetera.

Yeah.

Yeah. So moving on. So as Sahil mentioned, one of the reasons how we were able to reduce 21 engineers to 5, because we open source the product. Now we have a lot more people contributing and building, helping to build the products, which is why we have the bounty program. So traditionally, open source contributions have been free, so people were doing it to learn stuff, to make a contribution, which is great. We thought we could add a monetary factor to it, so we pay for open source contributions to our repos. So we have a dedicated page for it, which is andreework.com, and we list all the tasks that we have designated as bounties. People can attempt and send pull requests for those tasks, and if the PR gets merged, we'll pay them the bounty. And I believe we have more than 100 contributors from all over the world by now, at least by now.

And how many bounties have we paid out?

Yeah, I was looking at the number, but I don't get the number, but maybe I can update it later. But I'm sure we must have paid out more than $200K at least.

$200K?

Yeah, $150K, $200K.

That's a lot.

Yeah.

And this is still an active program. We are still getting contributions across the globe.

I guess, yeah, if you find issues on Gumroad as a creator, as a user that are objectively issues, like they're janky bugs, et cetera, you can create issues on GitHub. We'll add bounties to them, we'll scope them out and make sure that they're sort of clear so that people can work on them. We'll add the bounty and then we'll, makes it easy for people to basically improve the product.

Ideally Gumroad gets thousands of these $100 bounties that just kill every single tiny little paper cut and make this the experience of using Gumroad really great.

Also allows the core team to focus on larger, bigger, more creative things that are harder to sort of let the public participate in.

Yeah.

And also I think the bounty program helps us to move faster as well. For example, I remember the case when I created an issue as a bug and assigned a bounty and I logged off from the work. By the time I came back, there was multiple pull requests solving those particular issues. So we can also move much faster as well.

I do think though that it'll be interesting. Part of the thesis was that there's going to be a time period where you're going to need humans in the loop because AI won't be good enough. And so you want basically what happens if Claude code gets so good like, will there be any bounties?

Because in theory, if we take the time to scope out the issue, why wouldn't we just give it to Claude Code directly?

Like, just like full-time people may be contributing, competing with open-source folks, open-source humans are now going to be competing with, you know, token, you know, tokens effectively. And you can see this actually, like, a year ago, you know, we had $100,000 in bounties and now there's $5,000 in bounties.

So yeah, it's, it's where, I mean, it shows that I think like we're approaching sort of like quote unquote AGI for software engineering where we want literally 95% of the bounties are done. We're at 95% there. It feels like to me.

Yeah.

Yeah. So this is a chart with the team member size over the years from 2023 January and the number of pull requests merged, closed, and opened. Yeah, so we can see the team roughly peaked in August, September '24. And started going down from there. But the number of pull requests to the repositories fluctuated but mostly laid flat and spiked over here. But this is a very interesting point. So since the SONNET 4.5 released, until then we were merging most of the pull requests. So the green line is the PRs that we were merging. And until this point, we were merging more pull requests than closing. But from This point we started closing more PRs, which means people have started sending more A-slow PRs, the bad quality ones. So that's one thing that we have to deal with on a daily basis now. Hopefully as AI improves, this will also change.

Yeah, but it could also just get worse, right? Why wouldn't it get worse?

True, yeah. It can swing both ways.

The trend seems to be that more and more PRs are being closed.

Inter.

But not necessarily a bad thing, unless humans are spending a lot of their time doing it. But yeah, I think this is like, this sort of seems like an active conversation happening in the industry. Like, is AI good or bad for open source? And it's not clear, at least from this slide.

Yeah.

I guess one other interesting point, it is interesting to me that we went open source, but it just sort of didn't lead to a significant increase in, it actually, we've merged less pull requests than before.

Yeah, that is true. Yeah, so previously we were getting all the PRs from the team members itself, but Since open sourcing, we were getting PRs from multiple contributors and definitely we started closing more PRs.

To me, this shows that we shouldn't be open source because it's like open sourcing has dropped merged PRs.

That is true. But if you look at the productivity per engineer, it actually makes sense right now. So we open sourced at this point. So this is code activity per engineer in the team member, in the team. So after open sourcing at this point, the productivity has gone up. If you measure productivity by the number of commits merged.

Got it.

So you could say basically because we shrunk so much by team size on an absolute basis, it makes sense that we ship less, but per person we're still shipping more.

Yeah, we're shipping more.

Yes. And it seems like maybe there's, I mean, it's still very early days. This is part of the point. One additional data point can change a lot, but change the story. But I wonder if this Jan '26 is a new normal or—

Yeah, and we can see the change starting right after the OPUS 4.5 is launched. It's going up.

Yeah. So now we have OPUS 4.6 already.

Sure.

So it'll be interesting to see.

Yeah, it'll be very interesting to see the same data next year this time.

Yeah.

I personally think that there's been two, 3 moments for me that have been step function increases in my personal productivity. One is that Sonnet 3.5, which happened in October '23. October '23. But I, yeah, I got interesting. October '23, or I guess it was Sonnet 3.7. That was really what made it great. I don't know. I can't remember exactly, but for me it was June 2024 is when I got really into AI. Oh, and then Gemini 2.5 Pro, which was like really amazing, was like much better than that. But it, you know, Sonnet had like a, almost like a year of like consistent usage for me or more. And now Opus and Claude Code have really, seems like, made a step function increase in my productivity again, which I think is, that's what I would credit for that Jan 26th. I would say agentic asynchronous programming that you can do where you can work on 7 things at the same time, obviously should spike the number of commits per engineer significantly. So hopefully that trend continues.

Yeah. So we have been tracking the customer support response time and we have some interesting data here.

Let me see.

Yeah, if you have any questions, please post on the chat or raise your hand. Yeah, so this is the response time from January 2024 until now from our customer support team. We can see, so we have been experimenting with different technologies in between. And the latest data point is that we respond within 3.5 hours when a message is sent to a customer support team, either by AI or by the staff. And if you look at the 90th percentile data, that goes to 5.83 hours.

So what does that mean? Like, what does 90th percentile mean?

Yeah, so Mathematically, if we sort all the response times and if you look at the 90th, let's say if we had like 100 response times and we sort by in ascending order and if you look at the 90th position, that would be the 90th percentile.

So 90% of people get a response between before 5.83 hours.

Yeah. Yeah. 90% will get a response within 5.83 hours. So if you look at AI Assistant, that goes down to 2.1 minutes. So in 90% of cases, our AI assistant will respond within 2.1 minutes. And for Staffalone, that is, no, yeah. So this is P90, one second. Yeah, so if you look at the average, it is 5.93 hours. For a staff to reply to a message in February 2026. But if you look at the P90, it goes up to 43 hours mostly because some tickets take more time to respond. I mean, they need to investigate, they need to talk to an engineer. So in 90% of cases, they'll get response within 2 days. If the tickets takes too long.

And our SLA is?

SLA is 1 day.

1 day.

So does that mean we're failing or like what?

I know, but if you look at the average, we are within the 1-day time period, which is 15 hours.

Got it.

And what's the goal? What do we want to be at in 3 months?

Yeah, so ideally we want the P90 to be within 24 hours. Currently we are within— we are like— we can take around 43 days, which is like 4 days. Yeah, ideally.

43 hours.

43 hours, yeah. So we should ideally bring that down to one single day, which is 24 hours.

How are we going to do that?

We are planning to have the customer support team investigate the issue themselves and create bounty tasks. So they're not blocked on the engineer from the team member itself, so they can directly interact and get things merged from the bounties, but from the always contribute. So that's one way we are looking at it.

Well, the engineers are going to still have to review the code.

Yeah.

So is the issue that so many things require sort of like an engineer in the loop or like a bug to be fixed or like what is causing— like if we look at the histogram, which I think is another slide, like what is causing like the long tail? The long tail is going to be pushing that up a lot, right? The biggest way to drop the average would be to fix the, or the P90 is to reduce, find the biggest wins, right?

Right. Yeah. So we need to look at the gap at which the AI cannot respond to tickets. So currently some tickets are slated to be humans. If you can reduce that and let the AI respond to most of tickets, then I'm sure this number will go down. So we need to look at more integrations between Gumroad and Helper to send more data from Gumroad to Helper and let AI respond.

Gianfranco, do you want to talk about what you're working on with the Stripe? I feel like it's a good example of something that should help with—

Yeah, so for example, we get a bunch of Stripe tickets with issues that users have in And us engineers have to look at production data to understand what's going on. And potentially with AI, we could give support tools so they can actually have that information instantly instead of waiting for us to get that information for them. So that could be a good improvement.

Absolutely. Yeah, we can use the Cloud CLI with Stripe CLI to ask questions to Stripe directly and get responses here. Sure. So we can empower our support team with more AI-based tools.

The other thing you could do is just eliminate the support ticket from being sent. The tickets that take the longest, how do we make it such that they never needed to contact us in the first place?

Right, exactly.

So showing better error messaging in the app instead of to the support team is even better.

True. Yeah. And also have more creator admin tools on the app itself. So we want to talk about that. For example, one support ticket that could potentially take a long time is moving people from one version to another in bulk. So currently we don't have a feature for that. So we are going to build such features for creators to do those kind of things themselves instead of contacting support.

Gotcha.

Yeah, and this is a customer support response time histogram. So we responded to 97.7% of tickets within 1 hour.

But that includes AI, so.

That includes AI, yeah.

Just kind of cheating. Let's look at the staff.

Yeah, if you look at the staff, Yeah, if you look at the staff, we responded 86% of all the tickets within 1 hour and 5.1% tickets took more than 24 hours.

The ticket didn't come into the inbox properly or something?

Yes, we had like a couple of helper outages in the last few months that could have contributed to that number.

It seems to make sense that with 95% above 24 means 95% 25 are less than 24. So the 90th percentile sits somewhere between 1 and 24. Yeah. And 4 seems reasonable. True. It's not this distant then. Yeah. In which case, reducing it to p90 to less than 24 is probably quite simple if helper just doesn't have bugs.

Exactly.

Yeah. That's true.

Does, like, are there helper bugs at the moment? Like if we go to the the repository? Is that the plan? Are there bugs that we are working on that we know will help with this?

They are not known, I think. But there were a couple that we fixed last month. So that may have been what's—

But this shows February data, I think.

Right? If you go back, was it February?

Yeah. It's in February. If you look at B90. Staff, yeah, it's in February. So previously it was 26 hours, 32. Yeah, it goes up and down.

Not sure then.

We can take a second look at this data again to see if you have missed anything. Cool. Yeah, so in the case of AI Assistant, we responded, yeah, 99.9% of tickets within 1 hour, which is obvious.

And then how come, like, why do some things take 24? This is, as Jono said, maybe related to bugs and stuff, but like the 1 to 2 hours, is that like what's happening in those cases?

Could be bugs. Maybe the jobs got delayed from ingest and reprocessed later. Yeah. So things like that could happen. Got it. Awesome.

We do sometimes run into timeouts on OpenAI as well. Yeah. So that can cause the job to retry a couple of times and yeah, when it finally doesn't time out, that's when the response goes through.

Yeah. Interesting.

Thank you.

Sure. Now we can talk about 20 different highlights and Daniel will take it from here.

Product highlights, because everything else has been a highlight on other shows. Yes. Magnetic.

Yeah, sorry.

Yeah, I think it's good. Hello, hello. Okay, so yeah, let's talk about some cool things Gumroad did over the last 12 months, right? So I'm going to go some of the highlights very quickly. So one of them was, this is quite a good one, adding reviews to product descriptions. So previously creators had to add testimonials from their customers manually in their landing pages. I used to do this myself. I used to have to copy and paste and type it and copy and paste from tweets. And now there's a very handy feature where when customers leave reviews on the Gumroad products themselves, there's a button now in the landing page editor where individual reviews can just be added there, which I think is very helpful to add some proof as well because the reviews come from the platform, right? You know, it's not like the creator could say whatever they wanted, and it's also very, very convenient. So, yeah, very cool feature. Let's see another one that I'm using personally as well. The ability for customers to pay installments is especially useful for high-ticket items. If you're selling a product, say, over $1,000, the ability to pay over installments is nice and handy. There used to be a way of doing this with memberships before. I used to do this myself, but it was very clunky. You'd have to 2 different products. And then once the subscription finishes, sort of it becomes messy. This just made it super easy. You just tick one toggle and it just works.

Yeah.

Do you have a sense of like how much GMV has this driven?

I don't have data yet. I don't know if we do, but this would be something good to look into. But yeah, it's certainly very convenient. And it's very easy to use, like typical Gumroad spirit. I just tick the box and it just works.

Nice, awesome.

Discount codes, this also, this is a recent feature actually, and something I've been wanting myself as well for a long time. So Gumroad has supported discount codes for, I think, for forever or close to ever, but it always requires, you know, sharing the actual discount codes, either discount codes or the URLs. But in certain cases, like in Black Friday and such, certain seasonal, for certain seasonal events, you might want to basically send a discount to everyone and to have your product show with the discounted price everywhere in your profile, in the library, in Discover, whoever has the link. So, now there's this feature, basically you can create a discount and just apply it automatically to your products and it just happens, right? Everywhere the product is rendered, you see it discounted. I think it's very nice, especially again for Black Friday. That's just like the—

I like these, this video a lot. Like if you compare it like the other videos, which it sort of also shows like the impact of AI actually, I feel like almost where like the, if you go one slide before, this took a human quite a lot of time to do and they had to use like manual software to do it. And in my opinion, I mean, it's pretty good. I would say it would be considered awesome a year ago, but now with, I feel like, you know, it takes time. There's like a bunch of irrelevant things that aren't, you know, may distract the user. And now we have like a cool, Gianfranco, did you do that? Did you do this one?

Yeah, but it did take a lot of time, to be honest. It did? This one.

Hopefully this is the second time we do it is less, or no? Is it like, do you think it's a reusable kind of stuff? Like we could have a repo where we just have like, sort of like all of them in folders, then we can have Claude Code like generate a new one.

Probably, yeah.

Or something like that for a new feature. Ideally it would be, you know, I like this a lot. Even though it's not perfectly representative, it feels more like, it's like almost like an illustration. And it like, I don't know, it helps me like understand like as a, like the product in its fullness. So I kind of like it and I like the animations. I think it is pretty engaging.

Yeah.

So I would love to do more of this if possible. Yeah.

Hopefully we'll do more.

Yeah.

And new features. Yeah. Have a good reason to use these.

Yeah.

Okay.

Another one, Churn Analytics. So yeah, this is another interesting feature. People who, creators who had been running memberships, if they wanted to study basically churn rates and churn numbers. Previously they had to download the CSV, the transaction data, do this analysis in their spreadsheets in Excel. Now there's just a dashboard with very handy basic churn.

I can't say it because I'm being recorded, but I can show you his name and you can maybe try to find it. This is named—

Okay, daily payouts. Yeah, another small one to mention. This is something for eligible bank accounts. Now creators can get paid even at a daily rate, which is nice to have. It doesn't support all the banks. I think this is US only, right, if I'm not mistaken as well. So another nice thing we delivered in the last 12 months. Communities. Yeah, this is an interesting one. Again, like very simple in Gumroad spirit. You tick a box and for creators who have products that previously required a group chat feature. Before this feature, creators had to choose to use Discord or Slack or Telegram or something like this. Now they have the option to have something built in into Gumroad. It's, you know, it's a pretty basic feature compared to Discord and the other sophisticated chat software, but for many use cases, this is, you know, all they need, right? So this will allow customers of a product to interact with each other, to talk, to reply, and so on and so forth, which is quite handy. And again, you know, just flipping a switch in the product edit page and it just becomes available to all customers automatically, which is very nice. Yeah, one of the final ones is a small thing. Now there's a section in the Gumroad payouts page where all the tax documents live, right? So it's nice, convenient, everything in one place, tax forms, tax data, and so on and so forth. It can be downloaded, you can filter by year, and so on and so forth. And this is something outside of the Gumroad product, right? So, Flexile is a different product that allows, facilitates contractor payments. Gumroad uses itself, obviously, especially for, in addition to paying all of us, to also paying contributors, open source contributors, their bounties. And this is a very handy feature to be able to allow linking GitHub issues in the invoices, right? And they're like verified in the invoice and it just makes everything much easier to track both for the people issuing the invoice and also for the payer, Gumroad in this case, like in this example.

So yeah, anything to add, Arshan, to that?

To the highlights?

The one approving.

So this one?

Yeah. So after implementing this feature, it has been incredibly easier to review the invoices.

Invoices.

So previously I had to open every invoice, go to the pull request, make sure the, the person who sent the invoice is the author of the pull request and the bounty was correct and stuff like that. But currently I just have to click the button, sign up for the invoices, much faster.

Awesome, because you're trusting the verified button. Yeah, yeah, makes sense.

I think similarly we have to— we have that trust element. Like, the reason it works is because it's our code. Like, if they just— like, for example, even if they gave a screenshot like they could have doctored the screenshot. And I feel like this is something we have to, we definitely is an outstanding problem where like someone can open a pull request, it looks like they've solved the issue, but we don't really know until we've like done our own investigation, checked out the code, ran it ourselves, et cetera. So I feel like similar to this, we need to figure out a way to like sort of verify in that kind of, you know, in that sort of automatic way that like their PR, you know, is likely to do, you know, it does what it says it's going to do. For example, I've seen PRs where someone will say, hey, the test suite passes, but they didn't run the test suite. They just took a screenshot from another PR. So stuff like that.

Okay.

Let's talk SmallBets, something I'm involved in. So earlier last year, SmallBets became part of Antework, and that's also one of the reasons why I'm here. So let's talk about what SmallBets is. First of all, SmallBets is a project that I started 4.5 years ago myself. It's a community for, I would say, for new and aspiring entrepreneurs, people who are trying to do side hustles, side projects, that they're trying to quit their job to work for themselves, which I think overlaps a lot with the typical Gumroad creator, right? So certainly there's a lot of overlap. The main features of SmallBets as a product is actually quite simple. It's mostly a group chat where people hang around, help each other, they share feedback, they share their small wins, they ask questions, and so on and so forth. And then we have a video library, it's basically a course library on all sorts of entrepreneurial topics, you know, how to start a YouTube channel, how to prepare your taxes, how to, you know, get ideas, publish a book, things like that. So we have things about Gumroad as well. So everyone who joins Gumroad, sorry, who joins SmallBets gets access to the group chat, gets access to this library of content. Occasionally we do live events as well, which become part of the video library. But that's basically it. I think people find value in sort of being surrounded by the people like them, trying to, get their first small wins and so on and so forth. So as I mentioned, I started Small Bets initially. Actually, it wasn't even a community. The product evolved quite a bit over time. Initially, it was just a cohort course. I started doing this 2-week course where I was just myself, just sharing ideas about how to get into business and how to sort of take this prudent way towards entrepreneurship. Eventually, this evolved into a community and then the product as it is today. It started I would say still in the pandemic era, October 2021, maybe it was the tail end of the pandemic, but there was still a lot of demand and appetite for people to get into entrepreneurship. The timing I think was quite right. And yeah, I've been running it since then. And then last year Sahil approached me to consider Gumroad acquisition. Seemed like a good synergy, for lack of a better word to use. Since we are a boarding company, you know, so, yeah, we might as well use it.

Bicostal synergy.

Yeah. So, the acquisition completed in March of 2025. Yeah. So, SmallBets now is a full asset belonging to the Antework umbrella. And, yeah, let me talk a little bit about sort of maybe the numbers. So, currently SmallBets has 7,670 members as of this morning. Members are one-time, you know, you can check out of SmileBets, but you can never leave. So we just charge a one-time payment. There's effectively no churn, which I think is something that is interesting, but I think it helps us have members that remain engaged for a long period of time. We have people who joined in 2021 and Things happen in their life, and even if they disappear for a while, very often we see them come back because maybe a life event, maybe they got laid off or they have some discontent at work and they want to do something on their own. And I think it helps that we have like this ever-increasing, you know, group of members. So, yeah, Zomato Bets is a profitable business, so this is an income stream that now Gumroad as a company has acquired last year. 2025 revenue for small bets was $214,000. This is actually a decrease since the previous year. Previous year was the peak. 2024 was about $590K, and the previous year was about $400K or so. And this is something we are trying to correct now in 2026, right? So we're thinking about what to do in terms of new income streams that's related to this idea. So I had already started the previous years doing some upsells to community members, you know, selling more hands-on courses and consultations and things like that, which has been working, you know, reasonably well to keeping revenue steady. But I'm hoping that we can bring this revenue number to $1 million a year, hopefully soon, maybe this year, maybe the year after. And I think we'll probably need to do something different to get there, instead of just relying on just membership fees. So do I have some ideas for 2026? I mean, a few small ones. One of them is something that we're going to do soon is we have this video library that we've built over the years for small bets that currently every member that pays the full fee gets access to. But maybe we can diversify here and basically allow customers who are not willing to pay the full price but just individually buy essentially mini courses, right? And this might help us maximize revenues by having lower ticket items, like basically more buy buttons for products that affects different customer pools who are maybe a bit more price sensitive. But also have some other product ideas that we learned a lot from running small bets. Number one, I think we have, you know, completely custom piece of software for the community and the video library that we've been improving and iterating over the years. And it's something that people ask me a lot about, like, how can I have— I want to start my own community. How can I use the same software you're using? And even though everything is open source, they can just use it themselves, I think there's an opportunity maybe to productize what we're doing, right? And just sell it as a software as a service. And there's lots of other small things like this. I'm doing a lot of one-on-one consultations and advisory calls with community members. And this is also something I'm noticing I'm lacking in terms of like an information system for advisors and consultants. I'm dealing mostly with a collection of tools, Calendly, Zoom, you know, notes in my inbox, and it would be nice to have everything in one place. So this is something I'm very optimistic about this year, like to try to use these lessons and a lot of assets that we've already built for SmallBets, like to make them as products or services that will contribute hopefully to this revenue line item for for future years. So, yeah, and as part of the deal, that's why I mentioned I'm part of the Gumroad team now. That's why I'm helping Gumroad as a fractional product manager.

Yeah.

Anything else here that maybe you should cover?

Yeah, I mean, excited to keep trying stuff. I think part of what's impressive about Daniel and SmallBets is just like the like the changes to the core product offering and different things that Daniel's done over time. And I think having run Gumroad for so long, you can kind of tell the people that are kind of worth paying attention to for the next 30 years. Daniel's one of those people to me and Gumroad.

I'll be 70 in 30 years.

Just so I—

Say? I'll be 70 in 30 years.

70 years? 70 in 30 years. Great. You'll be doing something interesting though.

Yeah, I will be, I'm sure.

Yeah, I'm just excited to see what we— I just want to try stuff and make sure that something that can happen with acquisitions is like the energy goes away. So I think that's my concern is to just keep the pace up and keep trying things sort of post-exit. So yeah, I'm excited to see how this goes this year.

Cool.

Awesome.

Same.

Okay, I'll hand it back to you, Harshad.

Yeah, sure.

Thank you, Daniel.

Okay, there you go. Maintenance.

Yeah, so talking about Q1 2026 roadmap, we have a lot planned for 2026. We are doing a few migrations. Primarily to make sure our application runs really fast and also to make it easier for the contributors and for the AI to help build the application. So one of the migrations in Asia, which is almost done, right Emmanuel? In Asia is almost done, right? The Asia migration.

Yeah.

Let's get an update. Let's get an update. How's it going?

Yeah. We have only one issue left for the migration to Asia and Uh, all the rest has been successful.

This is the edit product page.

Yes, it's a sensitive page, so we want to make sure we have everything.

So when do you, when do you think this will be done? When, when is the, the plan to be done with the inertia migration?

Um, our realistic objective would be by like this Friday. The idea is to make sure that everything Well, probably we shouldn't deploy on tomorrow, right?

Because we're all going to be like flying and out of the—

Yeah, yeah, yeah. I'll be done in a few.

So we've been migrating the Gumroad application to Inertia for the last 6 months. And there is one last page that we need to migrate to Inertia. This is the edit product page, which is one of the most sensitive ones. We hopefully target to have a merge on next Monday. We have a PR ready and we only need the test to make sure that everything is green. And yeah, that's pretty much it. As for the Tailwind migration, I think maybe Juno has more feedback on this topic.

Yeah, well, on Inertia, I don't— I'm not seeing the open PR on GitHub. I see the issue. But I don't see—

Yeah, it's, it's the PR is not open yet. It's just a remote branch at the moment.

So has it been reviewed?

I have reviewed a couple of PRs that were already existing and that were not quite doing the job.

But you're saying that we have a— we're going to ship this on Monday, but we don't have an open PR for it. So we're going to get a review. All— everything's going to be perfect and we're going to deploy this on Monday to production.

Well, I think we can have a review in a day, and if it takes more time, then we can postpone it to a couple of days. So I'm hoping to open the PR today. So if it gets reviewed on Friday, I think it's realistic.

But who's going to be reviewing it?

Yeah, right. We are not working tomorrow, so.

Hopefully we can debug next week, sometime next week.

What do you think?

Yeah, well, we can do that.

But I don't want like a random date that is soon. I want the realistic calculation.

Well, let's have a pessimistic scenario and let's say that we will need a couple of back and forths. So let's say—

like the P90 basically is what I want. What's the P90 of this feature? Like the 90% chance it's going to go live by.

Well, next Thursday, then I would say.

Next Thursday.

Assuming there would be the back and forth.

Cool. Awesome.

Great.

The other migration that we are doing, as Emmanuel mentioned, is a Tailwind. Janu, could you share the update on the Tailwind migration? And also Expo as well, if you want.

Oh yeah, so Tailwind, um, we're down to, uh, basically the— a few, uh, large files left. Um, so the vast majority of the app is now in Tailwind, but a lot of the, like, very basic styles, like global typography and stuff, is, uh, still in CSS. Um, we have run into some issues where the product content page, uh, can't be fully migrated at the moment, and emails can't be fully migrated at the moment. So we're pretty much gonna migrate the remaining stuff that we can and then call it and deal with emails and kind of the remaining cleanup potentially sometime in the future. So yeah, I probably will still be a few weeks before the rest of that cleanup will be done and we can close off the issue. But Basically a few major things left.

Got it.

And also tell me, you're working on the Expo migration as well, right?

Yes.

Yeah.

So yeah, we're migrating the mobile app to Expo or we're rewriting the mobile app in Expo. That is going pretty well. We have feature parity, but not necessarily UX parity. So the focus for the next week or two will be to polish a lot of the UX and try to make sure the Expo app works as as well as the current native apps so that we can release and basically replace the current apps with it very soon.

Awesome. Great. Thank you, Jonah.

Thank you. Thank you, Jonah. Yeah, so as Jonah mentioned, migrating mobile apps to Expo is one of the top priorities for this year. We will very soon have the parity of the existing application But ideally we want the mobile apps to do everything the web can do. So that's the goal. And then we are also looking at refreshing the product edit interface to make it more easy and accessible for creators to create products much more efficiently. And then we are looking at the landing pages. So currently we have product pages, but we— it would be super awesome if we can create dynamic landing pages for products. So we are looking at using AI. So we are looking at that as well.

Can we show off any of this? Any of this have visuals of what the before-after may look like?

Yeah, we have it for product rethink. Where is that? Oh, one second.

And Janna, what do you think is a good P90 for these issues on Tailwind?

Product refresh. Tailwind, I'm going to say like 3 weeks, which would put us in say March, mid-March.

Mid-March.

Okay, let's see what that looks like. Yeah. Mid-March.

Laura, would you like to show this one?

Yeah, I'll be out.

So now we do the designs and you think maybe March 16th is reasonable?

It's a Monday, or do you want it to be a Tuesday so you have like one day buffer of like a review or like, uh, yeah, March. Okay, I'll do it.

Say March.

Let me open the— well, I could show you the video of the mobile.

Yeah.

Here. Minus emails, right? We don't plan on doing emails.

Yeah. So let me see. It's working.

Yeah. So this is not only the rethink of the product creation, but also thinking how we could include that change into mobile. So it's also easier for creators to create new products on the go. Also a way of improving some actions, multiple actions on the creation. Let me see. We want to also simplify the pricing. Right now we need to select different type of products on the on the first step. And now we want to only have like one way of creating the products. And then once you're there, you can decide if it's just one-time payment, multiple membership pricing, or pay what you want. So you don't need to do that decision of which product you want to buy, to sell. Beforehand. And yeah, what you can see right now here, it's just trying to make easier to use the— for creators to create the projects on the hand while using the mobile phone. So yeah, right now it's very complicated to do with our mobile UX, but I think that during this, uh, year we'll try to, to increase the, the UX so it's easier.

Awesome.

Is this running in browser or is this like—

no, this is just a prototype in Figma.

Got it.

And a video that I recorded myself.

Cool. Yeah, awesome.

Yeah, I think I'm optimistic.

I personally I personally think that the highest chance of Gumroad growth in 2026 will come from a much better mobile app. Because as I work at the IRS, I can hopefully share this data, but 70+% of traffic to IRS is mobile. So almost everyone does their stuff on their phone. And this is like tax stuff. This is like I never use my phone for this. I always go to my laptop, anything I'm doing, anything with money. But yeah, if Gumroad is modern and relevant in 2026, we need a phenomenal mobile application. Basically, Gumroad should be an app, not a website. When you think of Gumroad, it should be primarily an app.

I've seen numbers like 85% Greek on this.

Yeah, it's even higher. For us, it's less than 50%. People still, and certainly if you look at people creating products, it's vastly on desktop. So I think we should see a flippening. It should be over 50% of my, I think optimistically 50% of GMV should come from mobile-created products. Like I think that will show that we can unlock like a whole new sort of like, basically there's an activation energy that creators, I mean everyone I'm sure has experienced this where like, If you're on the toilet, how many people have spent time on the toilet creating a Gumroad account or creating a Gumroad product or refunding? Probably not that many, but scrolling TikTok, largely that's just the UX. I think it's just mobile, native mobile is just easier to get into and get out of. And I think that leads to a lot more potential usage, a lot more like, oh, I have an idea, let me just make it and that kind of thing. Yeah, super excited. I think every year we should almost redo the onboarding and the critical part of onboarding is product creation and edit. It's like the job is never done, I feel like, with that. Given 15 years later, we're still redesigning product edit and who knows what it'll look like in 5 years from now. It's kind of crazy, honestly, if you go to Google Images and you search Gumroad edit product, You'll see like 15 different versions of how it used to look like back in the day and how it's changed and all this kind of stuff.

Awesome. Great.

Thank you, Lara. So the other things that we are doing in 2026 is to empower our creators with more admin tools in the dashboard. So which is one is bulk version transfers. Currently they have to contact customer support to move the customers from one version to another in bulk. They can do it one by one, but to do it in bulk, they need to contact customer support. So we will have a tool for that in the dashboard. The other one is the auto-saving product editor. So currently when we switch the tabs, there's a chance we may lose the data. So we have some plans to fix that. And there are several other issues created on the GitHub. To create more tools for creators. And the other area that we have focus in 2026 is the bugs and performance. There aren't any, there aren't too many bugs, but we do have some performance issues with large sellers. For example, in the case of one of the large sellers, when they update the filters, when they try to send an email, the numbers shows shows a bit delayed. So we need to fix issues like that. So yeah, so this is what we are doing in Q1 2026.

Yeah, one I would just add, like, I long time— this was, I don't know, maybe part of my bias, but I had dinner one time with Kevin Systrom, the CEO of Instagram back in the day, founder and CEO of Instagram. And I asked him like what mattered, like of what What did you ship? What features did you ship that actually showed up in the metrics, improved the success of the company? And he basically said, no features, just speed and bug fixes. So that's why I think bugs and performance are really important. We've spent 15 years building features so people can do a lot. Obviously, there's more to do, but I think a lot of the The stuff that we can make better is just making it even faster. For example, if you look at the top left, inertia and tailwind. Inertia, the reason we're doing inertia is because it will improve performance. Migrating around the Gumroad creator dashboard is just not a great experience and it should be instant. And there's more work to be done there. And then tailwind for shipping more features and Expo for shipping the mobile app Like they're basically, Steve was asking like, why do we do all these migrations? And it's like, it sort of seems to be like every ongoing thing. And I think of it a bit like if you ever watch curling on the Winter Olympics, you're like, yeah, doing this. No, you're the one doing migrations are like constantly just like trying to like make the shipping speed go a little bit faster, like constantly reducing friction and like, and then of course, like AI, like as as the codebase gets older relative to what people consider modern, AI has a harder time because there's less training data, it's less optimized for those kinds of things, there's less economic value to making it good for those things. So I think generally as we move to, we get rid of sort of old, quote unquote, old ways of doing things, we should be able to move faster and that should compound. As the faster we move, the faster we can do more migrations and the faster we can get to the future. Yeah. So Gumroad's old, so we have a lot of migrations to do.

Yeah. It's a 15-year-old product, so we need to update to the latest technologies like Tailwind.

And unfortunately I built a lot of it, so it's not the best decisions were made.

Great. Moving along, buyback and dividends. Sahil, would you like to talk about this?

Yeah. Awesome. I'm back. Well, one question I guess we can start with from the chat. Kim asks, last year— I love this accountability— last year you had mentioned we will also do another stock buyback. Unfortunately, this won't include crowdfunding investors until we figure out a way to convert the SAFE into equity directly on our cap table. Any updates on this? I love the quote, like, amazing. And John, welcome to Gumroad. What do you think? We can have the nice conversation about it.

You guys are kind of throwing me into the fire, to be honest. Like, we can circle back with the group on this, but I need to go get my bearings on that one.

All right, that's the lawyer answer. How about we do a McDonald's test? I'll answer it, and then we can circle back with, like, a more legitimate Caveat, I am not a lawyer and all of that stuff. Basically, the issue that we ran into last year as we were doing, we did, as we mentioned, all these buybacks, they've generally been focused on early investors in the company. They're most likely to sell, they bought at the lowest cash basis. But really, we wanted to include everyone. The reason we had a hard time with that was, Coolie at the time said basically like, we have— the way it's structured is we have a CrowdSafe. The crowdfunding round is sort of this like, you can kind of think of it as an SPV, a little mini company that sits on the cap table. That way there's not 7,334 people sitting on the cap table, which would be a lot of people and force us to go public and IPO literally according to US securities law. Awesome, thank you. And sometimes it sounds like I don't know what I'm talking about, but surprisingly. And then, yeah, so basically we have, part of that, the way that CrowdSafe is written is that basically to prevent that from happening, there's quite a high bar to sort of converting the safe into equity. The issue though is that like when you do a buyback, it's to repurchase shares, and SAFEs are not shares. It stands for a Simple Agreement for Future Equity.

It's a convertible security. You need to do a financing to convert it. And so if the SAFE is still outstanding, you can't repurchase it. Yeah. My only thing is like, the thing that I don't know is whether you guys figured out with Kuli how to convert this ape in the last year? If you did, then, you know, we can chat. If you didn't, then, you know, the, the sort of answer still stands. You need to figure out how to convert it.

Yeah, we, we, we did not yet. We sort of kicked the can down the road. However, it's, uh, I think it's an interesting problem. I did get an email from, uh, something that I crowdfunded, uh, and they did find a way to do it. I think One idea may be to sort of have more constraints on it. So you could sort of like sell 100% of your position and just remove that person specifically from that SPV or something like that. But it's something that personally I want to do. I'm willing to spend some legal fees to figure it out. And I think it would be cool to— I think it's something that probably the industry there's probably a bunch of people are in this situation, so it would be cool to kind of figure it out potentially and share it. But it would be awesome. I think Kim, uh, we'll circle back in lawyer speak, uh, once we regroup. But, uh, I would love to do it. I think if we can figure out a way, um, I don't see why not. I want, uh, you know, want to treat all of our shareholders— I, I think of our SAFE holders as shareholders even though they're technically not. So it would be great. So yeah, if you have any follow-ups on that, just feel free to put them in the chat. Happy to answer them. And yeah, just to talk a little bit about this. So as we, as Irshad mentioned on the other, the sort of cash on hand side, you know, we try to, you know, there's no IPO basically in our future, at least not yet. Maybe that's an idea we end up discussing with our our legal team. But yeah, basically outside of an IPO, how do you provide liquidity to investors? There's kind of like two ways of doing that. There's like a buyback where you just buy the shares from the investors. That's 100% liquidity effectively. Kind of like when we bought back Excel, Collaborative Max, et cetera, sort of cash out and then pay taxes, et cetera. And then there's a dividend, which is you keep your position, but you just get cash. It's like a 4%, 5%, 6% liquidity, but you keep your entire position. I like the idea of always kind of doing both. Kind of like feels like sort of a— there's sort of returns to doing either, but diminishing returns to doing either. So it's good to do a little of both instead of a lot of one or the other. So that's the plan. You can scan that QR code or click the link. To sort of fill out the survey. Technically, I guess if you're watching this and you're not an investor, you could fill it out, but I will dedupe it against, I will check it against the Flexile email database and make sure that, you know, ultimately we're gonna get feedback from our investors on this. But the questions are basically about trying to figure out like how excited are people about Gumroad, right? Like, have they written it off? It's great. You're still going to get dividends even if you've written it off. That's sort of like the base case. But if you have, do you want to do a buyback? Do you just want 7 years' worth of dividends now, effectively, and then similar to kind of like the collaborative Max Excel approach? Or do you want the option value and all the interesting things Daniel and Small Bets is going to do, and you want to see this sort of play out? From my perspective, I have zero interest in doing a buyback. There's nothing I really— a big purchase I'm looking to do. So I like my dividends. But yeah, the offer will be on the table. But it'll help me kind of try to like, sort of like try to focus the company in the direction that investors are excited about and hopefully provide some liquidity for people who want it. And also just, yeah, like kind of just help me learn more. In theory, there's also potential where we're going to do the share buyback. We're thinking about doing it at our existing 409A valuation, which is $40.5 million. And in theory, if we could also— if you want to buy at that price, I'd be interested in if there's any interest in doing that. As well. Though to be clear, this is all subject to board approval and applicable legal requirements. This does not constitute an offer to purchase securities. Details will be provided separately, etc. So yeah, just an update on those sorts of things. But yeah, please, please scan this because it will quite literally, you know, it's, it's, it's like $4 or $5 million that we will, we will, you know, do things with and What we do with it will depend based on what investors are excited about. Are investors excited about us doing, like, buying SmallBets? Or are they like, no, you should focus on Gumroad? Or they're like, you know, we just want more dividends. Like, you should fire everybody and just give us as much dividends as you possibly can. Or we hate everything you stand for because you did Doge. Please buy our shares back from us, etc. Like, I'd love to to get people's opinions and thoughts and feelings and stuff like that. Anything else to add, Arshad, or any questions? I guess everyone here is an investor in Gumroad. Yeah, I think.

Yeah.

You feel anything you want to know? The point of a meeting like this is to give you everything you need to know to feel like you understand how the business is doing. Do you feel like you have a good sense? Anything missing? Any FUD, any fears, uncertainties, or doubts?

I'm good. I'm going to look at the chat, see any questions in the chat.

No questions.

No questions in the chat.

Awesome. Is there any other slide?

I guess just— Yeah, so we are at the end of the presentation. We can take any questions if anybody has any questions.

I would be curious for every single person what, based on the presentation, like what you guys think our EBITDA will be in 2026.

Yeah.

So to give you a reference, yeah, it was $5.9 million in 2025.

Anyone have a prediction to make? The first one to make a prediction gets a prize. 5.3?

5.3.

First prediction. I'll think of the prize for you. But you'll get one. And it'll be a good prize, not a bad one.

I promise.

I would say $7 million.

$7 million? Oh, wait, wait, let me write these down.

I think $6.5 million.

Wait, $5.3 million. What was yours, sorry?

$7 million.

$7 million. What was yours?

$6.5 million.

$6.5 million. Cool. Mine is $6,027,000. Mine is $6,027,000.

Why 27,000?

It just sounds like I thought about it more deeply. Yeah.

6.6 then to me. I feel like I thought about it more.

6.6, are you? Not 6.7? No one guessed 6.7? It's going to be 6.7. The most entertaining outcome is the most likely. I think it'll end up at 6.7 or 6.9 or something like that. Hopefully not 4.20, but we'll see.

Awesome. Great.

If anyone else who's listening wants to make a prediction— oh, we got some predictions in the chat. Oh, oh, this— oh no, they're, they're late, so Emmanuel also gets the prize. Uh, 6, 6.65, 6.7, 6.2. Daniel, can you talk a little bit about stochasticity in context of this? Could you go back to the earnings? Yeah.

Sorry.

Just one. I don't know. I think this is an important idea and it's fun to talk about. And Daniel's one of the best at talking about it. Putting him on the spot a little bit. But if you want to use the EBITDA slide. EBITDA slide. EBITDA slide, please.

Stochasticity, you mean like it's hard to predict?

Yeah.

Well, I mean, I don't know. I think the way I think about it is like, imagine you're in 2022. Could you have predicted that? I think that's something worth reflecting on. Or in 2015, for example, probably. I think in the history of Gumroad, 2019, correct me if you disagree, Sahil, I'm curious because I think I was actually just thinking about this yesterday. January 2019, this was, I think, before you've written the blog post went viral, before COVID like a year before COVID Imagine trying to predict the journey of Gumroad on January 2019. Like, you know, it would be hard, right? Almost impossible. So what was it? We are negative. Yeah, we are basically breaking even. So I think that's the challenge with estimates. We don't know the upsides and we don't know the downsides. Of course, you know, if we're trying to predict short term within the next year, I think it's reasonable. What's what we've all tried to guess, 6 million or thereabouts. If you try to predict even more shorter term, what will the income be next month or next quarter, more likely to guess reliably. I think it starts to become interesting once you start to guess even longer timeframes. What about the next 5 years? What about in the next 10 years? And that starts to become very, I think, much more exciting. Like, what could happen? Like, what are the potential upside things that could happen that we can't even predict, the unknown unknowns, and maybe even downside things? AI and competition and regulatory stuff, who knows. So yeah, I think it's a bit of an academic argument, more or less. Yeah, that's another great example. I think, you know, what you think about the tech— NVIDIA is so crazy. Like, it is a video game graphics company, which I think got two wild, you know, payoffs. One from crypto back in 2020 or so, which was the small spike compared to the AI one. But imagine, I actually tweeted about this recently. Imagine you're in 1995 and thinking like, this company that's making the video card graphics for my gaming PC is going to be worth like $4.5 trillion.

I mean, at this point, at that point, there was zero companies worth a trillion.

There was, until recently, until 2 or 3 years ago, there was nothing. Yeah, but it's wild.

Let's give another example.

Yeah, these are— yeah, what's their story? I don't even know.

So Chegg is a used textbook marketplace. So you can see how— this is like Gumroad, nice, boring existential crisis in a positive way. Wow, COVID crazy. Oh, we're reverting to the mean. This is what we hope doesn't happen to Gumroad. Oh, this is like— they thought this— oh, we stabilized post-COVID.

Yeah.

AI. Yeah. Turns out when you have ChatGPT, you don't really need to buy used textbooks. Stack Overflow may look similar to this potentially. And then you have a lot of stories that are just nice and boring. We're picking extreme examples because that's the— I guess that's the point with stochasticity is like, yeah, I think you said it best. And one reason I was pushing Emmanuel on planning and calculating because I think A skill to develop is ability to predict the future. And the way you do that is you spend a lot of brain cycles trying, and then you reflect, and then you do this thing in machine learning called backpropagation. But you basically think through like how you could have, like where you went wrong in your calculation. You kind of just like think about it really hard until your biases change a little bit, your intuition changes a little bit. And then over time, I'm probably decent at making these predictions compared to a monkey just because I have a lot of— I've tried. I'm constantly doing that. And I've seen some of these. Yeah. To be honest, the best year of Gumroad, in my opinion, I think was 2018, actually. That's the best year for me. And the reason is we made $6,500 in net income. Can you imagine? $7,000. I think we've spent more as a company in this meeting, like the length of this meeting. But it meant that Gumroad would not die. The variance here was crazy, right? It's like Could be like this, could be like this, like hard to say. Ironically, the better Gumroad does, the more it goes like this until like you're only kind of like increasing the potential volatility. Like these AI startups are like burning more money, they look more successful, but the things change so quickly, customers are super fickle, et cetera. All of a sudden here, it's like, I feel like we're now in control of our destiny. It's a lot less likely to go to zero unless I wanted that to happen. But in theory, it's a lot less likely to— at this point, yeah, how many people would've predicted that we would've done almost $9 million in EBITDA? By the way, there's no full-time employees at this point. Effectively, I stopped the plane from crashing, and that's about it. So yeah, COVID, pretty unpredictable. AI, pretty unpredictable. So I guess in the context of people wanting to buy, sell, value the company, et cetera, et cetera, I just want to remind people that no one can predict the future. Ultimately, the future is basically a result of the actions of unpredictable people. So even if you could predict the future, you're just going to end up with people who are like, I don't want you to be right, so I'm just going to do something else purely just to spite you. Yeah, kind of interesting, I think.

Awesome.

Anything else from the chat?

Yeah, we have an interesting question, I think, from Taro. I think if I understand correctly, it's a question about whether we should be adapting the interface for optimizing for AI agents or not. Something we were discussing yesterday, I think, over dinner a little bit. I was thinking about this myself when you were talking about mobile phones, people creating Gumroad products on the toilet and whatnot. And I was thinking, is that really— again, the discussion of predicting the future— is that really the future? Or should I just be telling my OpenClaw agent, you know, go create a product for me and here's the link and whatnot, and I don't even need a mobile interface? I wonder if you have some thoughts, Sahil. Is mobile— I mean, you seem to have said on record mobile is the future, but yeah, would you consider?

I mean, I have some hot takes.

I have thoughts on everything, to be honest.

But on this one, I think people overestimate the impact of AI in terms of how different the world will look. I don't, I don't, I think obviously AI is going to have an insane impact on the world and the economy, but in terms of like how the world looks and how people interact with the tools, I don't think so. I don't, I don't think people will change that much. For example, the, just again speaking from the IRS perspective, like the way people have pay taxes has not changed significantly in about 100 years. And while you do digitize things, and that does lead to a lot of great things, humans are humans, and I think we have preferences. And one of those preferences, you could see it when you have a little baby and you give them a phone, is to be stimulated and play. And if you give a baby a laptop, it's unlikely that they're going to just start typing Yeah. You know, like writer's block is a thing, but scroll block is not a thing. Yeah. One is natural and intuitive and is fun for mammals to engage in, and monkeys would do it. And one is just not. And that's why you have to pay people to do it is because, you know, that's the— if people really enjoy doing something, which by the way is why open source is so hard to monetize, it's because people enjoy doing it. So people enjoy doing it. It's going to sort of mucks up the supply-demand curve, basically. So I personally don't think so. I actually think the people are vastly overestimating the impact on AI. I think the fax machine is more disruptive than AI, personally.

Wow.

Yeah.

Hot take.

Yeah, I mean, it's— or the telegram. If you think about how like the world pre-Telegram and the world post-Telegram, think about how much war happened because people couldn't communicate efficiently. So like how much war is AI, like what is the effect of AI on war? I think it's going to be roughly the same amount of war on either side. It just, they're both sides will be better. I think the desktop computer was much more impactful than the laptop. I think the desktop computer was much more impactful than the iPhone. Obviously, we have a bias because these are the tools we saw come to life. And so we feel like, wow, this is like the most important thing ever made. And technologically it is, nothing even comes close. But humans tend to build the most useful things for themselves first. So just statistically, It's less, just like Gumroad, it's less and less likely that the things we ship will inflect the company, right? Given if they would have, we would've already done it. So it's still possible, of course, but it's sort of relatively unpredictable because if it were predictable, we would just do it ahead of time, right? So yeah, so I don't think it'll be, insanely disruptive. And on the other hand, I do believe AI will take all of our jobs. So how do you square that? And the reason is, jobs were taken by desktop computers much more than AI. If you look at what people did before the desktop computer and after the desktop computer, it's vastly different. 99% of people farmed, and now 9% of people farm. I think that's gonna, like, how does AI improve that? Like, the impact of going from 9% to 5% is less. You can't beat agriculture, you know, like, as a technology, or fire, or, you know, like, those things were much more impactful. So, yeah, unfortunately, I think I think on an incredibly long time horizon, you could see AI be more impactful.

I think the difference might be the speed. Because these things took decades. Even the Industrial Revolution, which is considered one of the most disruptive things, it still took decades to fully—

Yeah.

Or actually, I would say cars are actually much more—

Even those, they took at least 50 years from them becoming mainstream.

It took about 10 years from 10% to 90% penetration.

Which is still slower than probably what we're seeing in AI. We were just looking at the models and we're looking 5, 2 years, AI is much faster, much faster, much better and whatnot.

So yeah, for sure, 100%.

Maybe that's the dimension that might surprise us.

Yeah, I've been working on this thought essay thing in my head called The Religion of Silicon Valley. And I think the biggest blind spot, and I can speak from experience given Could you go back to the creator earnings slide? Like, if you saw the deck that I used to raise money, it would have been like this, probably, except compressed.

Like, this chart is this chart.

It's the same chart, right?

It's just like you just took that thing and you scaled it like that, right?

I think the thing that I underestimated, the blind spot that I had, is that I assumed that everybody else in the world, basically it's a failure of theory of mind. Like, why is AI not AGI? It's because it doesn't understand how humans think. If it could understand how humans think, then obviously it could think like humans, right?

So it's a bit tautological.

But I also, I was lacking, we all lack AGI, obviously, otherwise we would be gods, because that's what we think AGI will do. And the reason is, like, I assumed that the vast majority of the world was kind of like me, that like, if you could make money on the internet, you would. If you could sell digital products, you would. And so this thing is just going to go to the moon, this new thing. You could build an audience on the social media and just sell PDFs and ebooks, video courses, and, you know, but it's sort of It's a bit naive, I think. Obviously, the data shows, right? And I think the biggest thing is like people just vastly overestimate how similar people are to themselves. You know, like every time I go to a restaurant, I just look at the menu and I'm always astounded at how large the menu is and like why anyone would order.

But then you see the diversity of what people like and it's, Diversity is a strength. I know it's controversial to say these days, but it's sort of an antifragile way of having different kinds of people who can do different kinds of things and like different kinds of things. And effectively every country will rule the world at some point. The Portuguese had it for a while, the British had it for a while. Who has it now?

Maybe the Chinese potentially. But yeah, I think That tends to be the reason people, like the religion of Silicon Valley is that Silicon Valley tends to think that, like, why aren't we on Mars? Like, why did Elon make such a huge bundle? You know, they're not even going to Mars anymore, right? Everyone knows this. Like, we're now going to the moon. The reason is like to go to Mars requires an insane amount of money. And most people don't want to go to Mars. So they're not going to make the trade. So unless you can convince enough people, and capitalism is a hack because even if they disagree with you, they may still want to get rich. And so they'll buy Nvidia stock, et cetera. So you may still be able to get what you want even though most people don't want that.

Certainly most people don't want any money spent on going to Mars, right?

It doesn't list on the political— no party campaigns on that.

So you kind of know. But what do people want? They want more housing, better drugs. If AI can create new drugs for us, that would be awesome.

But I think generally people will start to learn how hard it is to do that because it turns out we've wanted to do that for a long time without AI. And we have a really hard time.

And there's a lot of things that AI hasn't really unlocked the step function change.

Similar to the internet, if you actually look at the internet, the economic impact of the internet, it took 20, 30 years until it was obvious. So much opex went into the internet and it was hard to set up software and maintain software and things would go down and break and everyone's like, just fucking use paper. And it was phones. This is worse.

But now obviously things have settled and email is the preference of most people. But yeah, I think it's interesting.

What is the economic impact of email versus AI? Right now, email is certainly much more valuable.

So I think it's good to, I guess, humble yourself.

What is a religion? A religion is thinking that you're the center of the universe.

And that's why I believe that sort of like the AI won't, in the next 10, 20, 30 years, I don't think AI will be that disruptive. We'll all find new jobs and do something else and we may even be called the same thing, but the work we do will be quite different.

But that's fine.

It's not a big deal. I got a new job. I now work at the IRS. AI took my job. Now I work for the government.

Maybe that's what happens to everybody.

Probably.

Kind of happens.

You know, it's what happens in a lot of places.

Like you have 20 to 40% of people work for the state. So that's not unreasonable. Again, it depends what you—

the best predictor for the future, in my opinion, is just people get what they want. So cheaper flat-screen TVs is what people want.

Not that hard to predict. Medicine I think people want that. So I think it'll happen.

It's just when. And I think that kind of goes to this point about stochasticity.

It could have happened during this meeting, right?

The top thing on Hacker News could be that something dropped that did something crazy. But it's kind of like this. This is like the impact of AI. It's hard to say where we are. And I personally think we're probably here roughly. We're still just spending insane amounts of money, and I don't think we'll be here probably for at least another period of time. Yeah, as I've said for, I think for 2 or 3 years now since ChatGPT came out, I think, well, basically, AI will be able to do all software engineering by the end of the year, which I think we're pretty much on track for in terms of writing code, obviously not deciding what to build, but fulfilling the spec I think will be solved by the end of the year. That's my over-under, and I think it's pretty close.

Yeah. Cool. Okay.

I should have it to you.

I guess I didn't really even ask the— let me answer the question because I didn't even answer the question. Sure.

Is there any reason to think that humans need to adapt their product offerings to something that is convenient for AI agents? Generally, my take is you should build new things for AI agents. Like a Gumroad CLI might make a lot of sense. But you can just think of agents as as software engineers. So just like you would've had an API to make Gumroad easier to use for software engineers that now agents can use. Agents are just humans who know how to code. Most humans can't code and agents can code. So I would just think of it more like that. How do you make your product more useful for people who are like, they like typing into a terminal, which is by the way, not most people and never will be most people. We're not going to, like Mac will never look like Claude Code. Like people say that UI is dead. UI is not dead. People like UI. People like a button that they click and they know what the button is going to do. I don't think you're going to go to an elevator and say, take me to floor 6, please, personally in the future. Certainly that will be an option. Someone will build that product, and I don't think it's going to see commercial traction. It's like smart fridges.

Or whatever, a lot of graveyards.

Chat fills a niche where UI is bad, like for analytics and stuff.

Exactly.

100%. It fills a niche. I think there's a set of things that UI solved, but not well, and chat is better for those kinds of things. I think the IDE is not as good as chat personally, because as the models get better, you can kind of trust the code and you don't have to, obviously you have to QA the product, but the code matters less and less. Kind of like that thing.

But yeah, there's many other things that I think UI makes a lot of sense for.

Again, working at the IRS, I see that. We're not going to move to a world in which our 80,000 people are going to be talking to an agent. They're going to see 50 buttons and click the buttons.

The thing that shocked me a lot as well, just generally, is how little automation there is in America.

When you go to Japan, there's a lot of automation and there's not a lot of automation in the US. And that's primarily at least talking to people there. They had to. It's not like people love vending machines when they want ramen. It's just like they didn't have enough young people to do the job so that technology became a necessary thing. You go to America, you go to a restaurant and you're waiting 15 minutes for the person to take your order when you know exactly what you want. Why? Because when most people go out to eat, they don't want to be done. They enjoy that experience. And so, like, I joke with a friend one time, I was in line at a coffee shop and he's like, man, it's so weird. We're getting coffee and staying in line for like 40 minutes in Fort Greene. And I'm like, that's the product. Like, you want to hang out with me, don't you?

That's what Disneyland's about. It's about the line.

It's about hanging out with people you enjoy hanging out with.

When you think like that, I think life makes a lot more sense and you don't have to be upset at why people aren't so focused on going to Mars all the time.

Yeah.

Great. I think that's pretty much it. We can end the meeting unless there are any other questions.

Awesome.

No, I think we'll upload it to YouTube and more people will consume it. And if people have more questions, they can email us or reach out in all the ways that exist.

Yeah. And we will push our updates via email and Twitter all the time. So you can tune into that as well.