🟣 The Collison Brothers LIVE on TBPN | John Collison, Patrick Collison, Henri Stern, Jeff Weinstein, & Zach Abrams

You're watching TVPN.

We are here live at Stripe Sessions. We're joined by John and Patrick Collison. Thank you for having us. How is it going so far?

It's incredible. We have 10,000 people here.

10,000 people here.

Yeah, yeah. And it is just the center. The economy is re-platforming around AI, and I think that's the theme of the conference.

Which economy? Is that the global economy of which you control 1.6%?

That is the— we talk about the internet economy, but you know yourself, everything is in the internet economy these days. There's very few companies that are not interested at being at the center of the internet economy. So it's increasingly the whole thing.

When we spoke back a couple of weeks ago, which in Singularity time is a couple of months ago, or exactly, yeah, we said that we were seeing a significant surge in new business creation. That has roughly doubled again since then. And so as of—

And wait, are you tracking that through Stripe Atlas specifically or through just new companies that are onboarding to Stripe as a payment processor? Both.

Both. So March was an all-time record for new incorporations with Stripe Atlas. And then in Q1 overall, for just new businesses joining Stripe, that was up 71% year over year. We showed the chart at the beginning of our keynote yesterday. It is parabolic. It is exponential. I'm running out of mathematical terms, but it's really very striking.

And do you like the— I guess my collar's all ruffled. There we go.

Someone's ruffled.

Yeah. And do you like the idea that this is driven by AI? This is driven by democratization of more tools, smaller companies, that there's more sort of like lifestyle businesses popping up?

So initially, I thought when I saw these charts that it's It's about new vibe-coded businesses that are maybe more abundant in number, but in any given instance, maybe not as significant or not growing as quickly or whatever. What we're actually finding is not only are there so many more businesses, but the average revenue per business is also increasing. And so it's not just more new tiny ones. Somehow the median business is also doing better.

Yeah, I think even the term lifestyle business there's like a bit pejorative. It betrays a worldview, whereas what we're seeing is there's tons of business now making more than $1 million a year. Like, that number is really inflecting. And in general, like, we think about it in, ah, you got the little collar service. I mean, I just want to be doing my collar myself. And so you also have to look at it almost in Kossian terms, where what we're seeing is more small—

I'm going to have my citizenship revoked for drinking this.

I just want to clarify that these are full-size pints.

These are full-size pints. We're just very large humans.

I just want to clarify that it's Guinness 00, which is very good, but given that it's 11 AM, it's a little early.

Yeah.

Um, yeah, but what are you seeing at Coast?

Oh, just, yeah, we're seeing thanks to AI you now have more small businesses able to do million dollars plus of revenue and be really impactful. And so we think you'll probably see smaller firms have a big impact.

And, and what's driving this is you think AI is making every little step of entrepreneurship slightly easier, like slightly easier to incorporate, slightly easier to create a contract or to hire somebody, slightly easier to create an ad. And then you add all that together.

It's the merging of all the functions. It's the fact that you can do your design and your PM and your engineering all yourself as one really effective person. Sam Altman was here yesterday on stage, and he was talking about the return of the idea guy. The fact that— exactly, yeah, yeah, yeah. The fact that if you, uh, the long-maligned idea guy, but now if you have really good ideas You can actually go execute them, and you have so much more capability to do so. And again, we are seeing that borne out in the Stripe data, where I think this phenomenon shows up as 71% year-over-year growth in businesses launching.

So I mean, it feels like you could do nothing and win. Stripe's in the right place at the right time, but you're probably not taking—

We are not doing nothing, I'll tell you that.

So tell me what you're actually doing. How are you accelerating these businesses? Like, what are you changing about Stripe? What products are you adding? What functionalities are you building?

So all the commerce is going agentic. We think that people will— like people will still be doing stuff, but they will do it through their agent. So what does that mean on the consumer side of things? Like we've all had the experience of, you know, you're in ChatGPT and you're recent— like I was trying to buy a case for my bike and I was going back and forth and I was, you know, asking questions about, well, how do I know hardware, soft things like that. But anyway, at the end of that, you just want to, uh, be able to hit buy rather than like— no one wants to fill out web forms, not a value-add activity. And so we just announced that Google is joining our agent-to-commerce suite. So now we have Google, OpenAI, Microsoft, and Meta all working with us on this agent-to-commerce modality where you can just buy there. That is the consumer side of things. Do you want to talk about the developer side of things?

Well, give you 3 other things. So first, we're seeing a change across the economy and business models, uh, where with inference costs, with tokens and so forth, and so many businesses, you know, in some way taking advantage of AI in their products, we're seeing a shift from SaaS, from seat-based business models to usage-based business models. Uh, and I think this is, uh, this is pretty profound. So we're We acquired a company called Metronome back a couple of months ago, and we're now integrating that cohesively throughout the Stripe platform. We demoed yesterday something that I think is very cool, and is the intersection between crypto/stablecoins and usage-based billing, where basically you can have streaming payments, where you're paying for each token as it's delivered.

In real time.

Yeah, exactly. Today there's a kind of credit extension that implicitly happens with any token billing, which you're billed at the end of the month and so forth, which is actually pretty tricky because if you've token thieves, if you've some failed billing after somebody racked up a million-dollar credit card tab or something, you know, that's a real problem for the provider. So anyway, we're doing a lot around usage-based billing in this economy-wide structural transformation. Then secondly, to this fraud point, we're seeing, I mean, concomitant with the enormous increase in new business creation, we're also seeing a lot of fraud. And there's a lot of fraud attempts. Yeah. So, you know, when you— you can't resell access to software. And so if you got illicit access to some SaaS tool 2 years ago, there wasn't much you could do with that to monetize it, so to speak. There wasn't much incentive to steal SaaS, but you can resell tokens. It's very lucrative to do so. And so all these new AI companies and all these new AI products, growth is not really a problem for them. They have a surfeit of growth, but there are also a lot of people and token pilferers looking to—

ne'er-do-wells.

Exactly.

Looking to run away with these, these, you know, this precious value.

And walk us through how that would work, how that works in practice. Is that somebody spinning up a bunch of free accounts and then like hitting them with just kind of the bare minimum?

Is the motivation distillation or actual reselling?

Both.

Really? Both? Yeah.

So there's definitely just straight resell, but there's also more sophisticated attacks and distillation. So anyway, we've seen that 1 in 6 new accounts across this ecosystem are fraudulent. And for many providers, it's actually difficult now to offer free trials because the illicit accounts that sign up for these free trials, they drive so much expense that it's hard to make the overall economics work out. So anyway, we're extending Stripe Radar to use the intelligence of the Stripe network to help any of these companies not only score payments, but score customer signups, to score trials. And we're turning Stripe Radar into this sort of X-ray vision for the AI economy to try to prevent these new and more sophisticated attacks. I think this is really important in that we're not going to have an internet with self-serve access.

Yeah. So, so, uh, just to repeat it back to you, the, the— if, if somebody signs up and they can basically pay for tokens streaming with stablecoins, that would enable, like, let's say, a big AI company to be able to offer a self-serve product and not worry about someone coming in running up a big token bill and then saying, like, see you, I'm not even in the country.

That's exactly right. But we're also going to try to protect the regular signup flows as well, and they're kind of more straightforward consumer access.

If you run any physical store, you have to worry about shrinkage and, you know, No one wants a direction of travel where, you know, you go to a CVS and everything is locked up behind a pane of glass. You know, good, you know, upstanding people suffer from, you know, the small percentage of crime that happens. It's the exact same in the AI land where similarly they are selling a product that has, you know, real value and real resale value. Exactly. Real costs associated. And so therefore they need to prevent fraud, not just for financial loss for themselves, but also so they can continue to offer a great experience to all the good users.

I want to go way deeper, but I know you have a hard stop. Do you have something else to say?

Well, I was going to say one thing before I have to scurry away, which is, you know, so we're— we have this usage-based billing stuff. We have the agent-to-commerce stuff that John just described. We have this new fraud functionality. And then the other thing is Stripe started out, of course, as a payments company. But we've, you know, over the last couple of years, Stripe has been becoming a broader programmable financial services platform. And that's a bit of a mouthful. What it means in practice is that not only should you be able to get paid with Stripe, but you should be able to hold money in Stripe, be able to pay people around the world, be able to orchestrate your entire financial services business. So to be an operating system for global money movement. And so we announced Stripe Treasury, where businesses can hold balances in Stripe. They can dispatch money to more than 150 countries, the instant free Stripe-to-Stripe account transfers and so forth. And so rather than Stripe be this sort of conduit where money ephemerally passes through. Stripe is becoming a home. And this is very valuable, we think, because Stripe's expertise in building programmable tools and APIs and CLI functionality and all the rest, this is obviously much more important in the agentics era.

Was— sorry to cut you off. Was this in the seed pitch deck?

None of this was in the seed pitch deck.

But I imagine like you're in the fund, you're telling people we're in the funds flow at some point or another. We won't just pass this back to their bank account will actually be able to unlock the value of the—

That wasn't in the seed pitch deck. And in general, the secret to being a Silicon Valley founder is you have to get lucky for many, many years in a row. And then you go back and you pretend that that was your plan all along. You have to retcon. Exactly. You do a lot of retconning. It's a key skill. That's very important. However, what I will say is interesting is we did focus from the very beginning on developers. And we thought developers were very important and underserved. I mean, Stripe's original name was /dev/payments. And payments for developers was— you can go back in the Wayback Machine. That was the tagline on the, on the website. I think what's interesting now, obviously, is that the, you know, all the agentic coding tools need something with well-documented APIs, with everything being available on an API basis. And so I think we have always thought that companies should take developers more seriously. I think we're now seeing just industry-wide companies that took developers seriously are now really benefiting from this AI moment because they built their stuff for the agents.

Also, everyone's a developer now.

Exactly.

And agents feel like they are developer-shaped, at least.

Agents want good DX developer experience even more than humans do. And so if you don't have it, I mean, it's a real problem.

Not going to make sense.

I think they're going to kill me if we keep you any longer.

Okay.

We'll let you get out of here. We did want you to sign this plaque. Can you explain to us what this is and what it stands for? Are you familiar? This is the original print that went into—

Wait, how did you get that?

That's ours. It was given to us and we want you to sign it. You thieves, thank you for the strike. They recommend it here and here. We have a Museum of Business where we try and accumulate signed objects of historical significance.

And this is the new Poor Charlie's Almanac paper cover.

Hey, this is the paperback, paper cover of Poor Charlie's Almanac.

It's heavy. This is how we made the original hardcover, if you have it, of Poor Charlie's Almanac.

Right there, and we'll have Patrick's sign right over there. And this will live on the wall right here.

The literal gold master.

Yes, Museum of Business, and we will let you get out of here. And we will go way deeper. Stay with us. We are going to continue our interview with John Collison from Stripe at Stripe Sessions here in San Francisco. Thank you so much, Patrick, for taking the time to come chat with us.

Great to have you.

Really appreciate having you on the show. It's always a pleasure.

Cheers.

And congratulations. We do want to have you— can we have you hit the gong? What is the headline number?

You need a gong?

Let's have you hit the gong.

Is it 1.6?

Is it 1.6? Is that the headline?

That is the 2025 number. Stay tuned for the 2026 number.

And smack that gong. There we go.

Powerful. Powerful hit. Powerful hit.

I see why you get that.

You can tell he's been lifting.

Yes, yes, yes. Well, thank you. Have a great rest of Stripe Sessions. We really appreciate you coming on the show. I want to just shift over to another Stripe Press book. This is Maintenance of Everything by Stewart Brand. And in here, the general thesis is— maintenance of everything. I mean, I'd love to know what you took away from this, but maintenance is what keeps everything going. It's what keeps life going. He gives a ton of really interesting anecdotes about cars and repairs and weaponry and sailing. And there's so many interesting anecdotes. I'm wondering about your thoughts on maintenance in the age of AI and what you feel and see in this. Like we're in this— maybe there's like this job replacement thing. AI does everything, but then there's so much that goes on in the maintenance acceleration of technical debt. And so I was wondering if you just reflect on like Stewart Brand's work in the age of AI.

So Stewart Brand is obviously amazing and, you know, is maybe famous for the Whole Earth Catalog originally, which, you know, Steve Jobs talked about in that famous commencement speech. But Stewart Brand has been around for the entirety of Silicon Valley history. Like, if you go back and read the Tom Wolfe book, The Electric Kool-Aid Acid Test, I don't know if you guys are big Tom Wolfe fans, but it's talking about like the early LSD culture of the 1960s in the Bay Area. They're having these parties. Stewart Brand is in the book. He's like there at those early parties and everything like that. Yeah, when like the Grateful Dead and everything like that. And so he's just been like there for everything all along. He was the manning the computer at the mother of all demos, the Engelbart famous demo.

That was where the original mouse was demoed.

Exactly, where the mouse was shown.

Exactly, yeah. Among a ton of other things.

Exactly. And Stewart Brand is still— he just published this book. And so Stewart Brand is so intertwined with the history of technology and Silicon Valley's view of technology. And obviously what this book is getting at, and also the Whole Earth Catalog, is this passion for tools and taking care of your tools and kind of thinking about tools. And it's interesting you bring that up in the context of AI, because one thing that I get excited about in the current moment is this idea of individual empowerment. And, you know, you could have different visions of AI. And I definitely have an adverse reaction to this, you know, very passive. You know, I think the term slop is, you know, the term we use for this, just, you know, being a passive recipient of AI-generated stuff. And no one wants the, you know, Brave New World, Soma kind of vision. And instead, if you look at something like, like OpenClaw, I get very excited about that because you are giving computing power and the ability to do stuff with programming and tinkering with technology. It's the Homebrew Computer Club all over again. It's this tinkering energy that's back. And so I just get really excited about how I think tools and passion for tools.

Negativity there's been towards the entire like Mac Mini, Open Claw movement when it's like, well, for how, how, how long, how long has Silicon Valley been like I miss the days of the hacker era. And then the hacker era comes back. And then people—

It's bonkers. I agree. And I was reflecting on this. I think there's something interesting where people maybe have an adverse reaction to early adopter enthusiast culture. The early OpenClaw enthusiast group really reminds me of the early Bitcoin enthusiast group, who also got people's backs up in a way. Like, you remember that famous New York Times article with the guys with the sweaters, like those guys are presumably big into OpenClaw, uh, right now. Uh, so there's something about the— maybe because people tend to overhype stuff, you know, they say my OpenClaw does this, like maybe it doesn't quite, you know, or something like that. But the tinkering is back, and I think that's really cool. And so I agree with you, it's crazy that, um, you know, people have this skepticism because it's easily like the most fun you can have with computers.

Yeah, people on— people on X, you know, see an OpenClaw meetup in someplace like Tokyo and they're just like mocking it, right?

Yeah.

And it's like, isn't this, isn't this, isn't this like core to the culture?

Exactly. Yeah, I'm totally with you.

That is the optimistic take on open-claw tinkering. I love that. That's what I believe. But there is the other side, which is the not just the tools, but the tool-shaped objects. You've heard of this criticism that you wind up— I've had this exact experience where I've gone to Codex and built a whole thing and been like, that could have just been a prompt. And like, I didn't actually need to have it instantiate a Python script. Like, I could have— and I think there is a risk in some enterprises of having 25 people vibe coding dashboards when you could have just had one. There is some managerial deduping and overhead in management. And I'm wondering how you've grappled with that side of AI.

Yeah, I mean, a few thoughts here. One is, I think it's just awesome that software is becoming so accessible. I think we should really celebrate that. Like, I'm writing much more software, even just from a kind of a fun perspective, because because it is so fun and it's this kind of addictive feeling. I think we're also very clearly in a transitional state where it just doesn't feel to me that the way we do things in 2026, where, you know, you need to prompt us to, you know, make no mistakes, or, you know, you saw the caveman token efficiency thing where it's like if you talk like a caveman and I ask you to talk like it, it just uses fewer tokens. And so, but gets the same results, you know, it's as effective. And so if you want to kind of economize in the token budget, you talk to—

like, every image prompt was no six fingers.

Yeah.

And I haven't seen an AI image with six fingers in a year.

Exactly.

It just— we just crossed that chasm.

Exactly. You know, if you're doing anything with financial data, you know, you need to prompt it to do not make up any numbers. And then it's much better at not making up numbers.

Right.

So where we are in '26 is clearly not where we're going to be. The models are getting much better. But I think right now we're in this interesting transitional period. There is this software fatigue, as you say, where, you know, from software maintenance, things like that. I think we'll, we'll get there. And like all these things, you know, there's puts and takes. People are talking a lot about the offensive side of security, which is true with, you know, the very powerful new models. But of course, you also have the defensive side of things where if you ask your, you know, 5.5 to audit, go take a look at this application, take a look at this, just like go putter around this server and give me a security audit and suggestions to lock it down. It'll do a great job of that. And so again, I think we just haven't figured out what the end state is.

Yeah, yeah. It does feel like as we move up the layers of abstraction, we go from, you know, pair programming alongside an AI to being more product manager to almost winding up more in a build versus buy mentality.

Yes.

And yes, you could vibe code that, but if it already exists or it's open sourced already or there's something you can pull off the shelf and it's going to be cheaper than the token cost, cheaper than the maintenance, like maybe—

But again, we'll end up in a new equilibrium where, you know, the— I think exporting your data will become more important because you'll get back to the Unix philosophy of You have individual tools that are much easier to interoperate between each other. You can kind of pipe your data conceptually between applications. If I can also bring this back to payments for a moment, a very important angle on all this stuff. You guys have probably had the experience of when you are doing something in Codex, when you're vibe coding away, you are the adult that needs to reach for things on the high shelves for it. And so it is much better at writing code than you. It's able to architect an application. So what do you need to do? You need to go sign up for an API key for a PDF of a website that I can't access. Exactly. Log in and please get me API keys is like something you end up doing a lot. So one of our biggest announcements this year was something called Stripe Projects where you are able to agentically do stuff in the wider world. And so you can hook up your Codex to Stripe Projects. And if you want to demo this, if you want to deploy a website, it can go sign up for Vercel for you.

This is huge.

And it can pay for it. And so now what was previously a multi-step process where it could kind of do some version— they could like orchestrate Vercel but it couldn't sign up for it. Now it can pay for Vercel, it can pay for Cloudflare, it can pay for Browserbase if you want to, you know, do, uh, uh, headless web browsing and things like that. It can do all this stuff. And so we think the agents are going to get so much more powerful because we're, we're giving them the ability to actually go buy what they need in the wider world. And then it can do way more end to end. And so that's Stripe Projects, which is kind of the second side of Agent to Commerce is your, your agent needs to be able to go do stuff on your behalf. And you're exactly right. Agent to Coding agents are not just for coding, they're for doing stuff. You know, we demoed today, write a research report for me, buy some data in the process of doing that research report. But again, that's not— I happen to use a coding agent for it. It is not a coding task. It's a, it's a research task.

Yeah, yeah. This happened. I, On the show, I used Codex, I vibe-coded a beautiful web page for some joke we were joking about, and I was like, it's on my localhost, I need to send it to Jordi. And actually deploying it was the harder part, which is supposed to be so easy. And that's—

Stripe Projects is the only way you can agentically buy a domain. Like, you don't need to have anything. You don't need to have an account with any of these providers. You can just go buy a domain. It's super cool.

And I feel like that's really important.

How is recruiting— how has recruiting changed over the last few years? Because I feel like so much of of what has made Stripe such an incredible company is convincing the smartest, most ambitious people in the world to work in payments, a place that they historically maybe wouldn't have been excited to work. And then now I imagine the job of any person that could get a job at Stripe can probably also get an offer at OpenAI, Anthropic, you know, some of these other labs. And so I imagine you find yourself having conversations with folks and helping them make a decision. Decision.

Yes. So I mean, so many places you could go there. One, as you say, I think we have always encouraged people to think bigger than just payments. And I think if Stripe was just about payments, fewer people would have been interested in joining. But I think they understand that. Going back to tools and the importance of them, I think people understand that the infrastructure that we give people for entrepreneurship actually changes the outcomes. We're moving the supply-demand curve. And so we can encourage there to be more entrepreneurship. Again, Patrick talks about— 71% year-over-year growth in the number of businesses being created on Stripe. That's not all Stripe, but we definitely contribute to it by making the whole entrepreneurship journey easier. And then when those companies scale, you know, OpenAI sells ChatGPT, I think, in more countries than it otherwise would have because all that infrastructure is just super available and on rails. And so it exists at kind of the small getting started end of the spectrum. And for, you know, the biggest companies in the world, the Amazons and Hertz and Metas and everyone like that that we work with on, on projects like this, you probably saw that, like the stablecoin thing. We answer them. OK, so that's one part of it. It's like you have to genuinely have a mission that's interesting. And I think we believed and believe that this stuff actually really matters and you can change the end equilibrium.

That's one.

Second thing is that it has always been the most competitive time I can remember to compete for talent. It's just like I can never remember a time when it wasn't the case.

It was easy when you were competing with Mark Zuckerberg. Yeah. And then it was easy when you were competing against Elon Musk.

No man steps in the same river twice, because he's not the same man, and it's not the same river. I remember when we were recruiting Greg Brockman, who's our second employee, who we were going to have speak yesterday, but we had Sam instead, because we have a lot going on this week. But I remember recruiting Greg to join Stripe as our second employee, flying across the country to Boston to go meet with him, because why would you join this no-name company that you've ever heard of? And that's the kind of stuff you do then. Always the most competitive time that it's ever been with recruiting. And this is certainly no exception. And, you know, people have a lot of great options. That kind of gets to— I think, you know, people are talking about the death of software engineering. And honestly, the closer people are to AI, the more they're like, oh, you know, software engineering is cooked. You know, it's got 2 years left. I think those people are all high, honestly. Like, um, Just, you have, you have so much, you have, you are able to work with AI so much better if you understand the underlying concepts, and you're so much better prepared for that. And I'd almost analogize it to, remember in the '90s people had this view that like, okay, you know, the cloud, or like a proto-cloud back then, is coming along, the network is the computer, all this vision. And people tried to have these leaky abstraction, or like abstract, you know, they thought they had a beautiful abstraction where you would have, uh, you know, a folder that just happened to be on the network in the cloud. And, uh, you know, you still have that on the Mac today, you know, your iCloud folder and things like that. And same on Microsoft Windows. Those never worked that well because it turns out the network exists and, you know, it can sometimes be down in a way that your hard drive wasn't and things like that. And so it was always important to understand disk I/O and RAM swap and all these kind of things, even if you had useful helper abstractions over them.

Sure.

I think with AI, similarly, to be really effective working with AI You need to understand what it's doing and what's going on underneath the hood. You need to understand that the model and then the memory and then you have your harnesses and you've got your skills and everything like that and how they all interact. And so I think all these people calling the death of software engineering, yeah, are smoking something.

Yeah. I mean, there's also just the software engineers are dynamic and intelligent and creative and can operate at various levels of abstraction. And there have been transitions from—

If you want to make the argument that AI will transform every aspect of the economy, but that engineering, like, yes, core understanding of engineering and software engineering will no longer be valuable. That's, that's— and, and, and, and obviously some, some of the people that are, are more bearish on, on the practice of software engineering have, um, you know, uptime issues. And like Stripe is a category where you guys can't really afford to have, um, yes, you know, uh, reliability issues.

There's two categories of people I would be super bullish on right now I think will do incredibly well over the coming 10 to 20 years. Firstly, high-agency people. We know this at Stripe. The people who are like, I've been talking to customers. I know exactly what we should do. We got to go fix this. But the people who have that pep in their step and they want to go make Stripe better, they are now so much more empowered thanks to AI. The second is double majors. I think if you understand software and understand finance, or if you understand software and understand marketing, you now can go massively improve the entire marketing funnel for your company. And one person can do what would have taken 20 people dredging through all these systems previously.

Famous Paul Graham quote about that too. Yeah. Typically an entrepreneurship team, a founding team, has a collection of like 5 or 6 skills between 2 founders, 3 founders.

Charlie Munger talked about the importance of being multidisciplinary and multidisciplinary thinking, and implied was that like he thinks getting a functional understanding of many disciplines is not that hard. You can just go read the books. Now, you know, you can talk to your AI about it. And so I think multidisciplinary thinkers are going to do incredibly well.

Yeah. Do we have a minute to talk about stablecoins?

We have 1 minute.

Yeah.

Okay. 1 minute. The Economist in Buttonwood. There is— there's a lot of doom and gloom about the stablecoin winter. I put a ton of money in stablecoins. Didn't make a dime. That's— they're saying it's too stable, but they are saying that the market grew 30% in the 6 months up until late October. But total stablecoin assets have barely budged since. Are we in a stablecoin winter broadly? Are you seeing that adoption is slowing?

This is The Economist not processing that there can be a correction, there can be a correction in crypto asset prices and yet more adoption than ever in actual the utility of stablecoins. And the two things are not, not, not the same.

We're always trying to point out to people like the difference between price curves and the underlying activity. And like, those are very different things. And, you know, we're talking about that in the SaaSpocalypse in the context of, you know, SaaS growth is—

yeah, yeah, companies are still growing top lines.

Exactly.

Top line growth is really good and everything like that. And so kind of a similar story with stablecoins where we see volumes continue to grow. And again, we had this in a lot of our demos with agentic pay-per-use payments or streaming payments or things like that. You actually have to use something like Tempo. You have to use stablecoins because economically speaking, just like what else are you going to do? And so we are as bullish on stablecoins as we've ever been.

Fantastic. Well, thank you so much for taking the time to come on the show.

Great hanging out. Thanks for having us. Cheeky Pint.

Yes, there you go. You guys are the perfect setup.

Uh, have a great rest of Stripe Sessions.

We'll grab your headphones as well.

And we will bring on our next guest.

Alrighty, tap out.

Cheers.

Goodbye. Uh, and we will bring on Henry next. Welcome back to the show. Last time we were in person, uh, we'll let you put those headphones on. I think last time you were in person, uh, didn't you help us translate something in French? Something random Do you remember that?

I do. This is my second audition.

Yes. It was about— It's fantastic. You can sit here.

It was about Meta's acquisitions.

Oh, yeah.

That's right. Yeah, yeah. Because we were interviewed for French TV.

Exactly.

And we didn't know what they were saying. And you helped us understand it. That was fun. What a wild time. I mean, I guess it's only getting wilder. Have you been struggling with any of the talent wars? Or have you been recruiting for your team? Are you growing your organization now?

All of the above. I mean, so when we were acquired— this is coming up on 9 months—

we were about 30 people.

At Privy. We're about 60 now. And then Stripe generally is doing so much more.

The soundboard made it incredible.

And then Stripe is doing a lot more in the space. So generally, I think the talent wars are real. But also the reality is twofold. First, you had to be a little bit damaged to go into crypto and stablecoins to begin with. And so we're seeing these people who are still coming in. And then actually, we're seeing a lot more merging of the two, which is to say, yesterday was a good demo on streaming micropayments. Like there's a lot more that I think stablecoins and like, you know, internet native money is going to enable for agentic use cases.

It's still very much in its infancy, but it's like, yeah, we just spoke with the Collisons around, you know, all this issue around AI fraud, people signing up for a bunch of free accounts to get access to tokens or people signing up for an account, running up a bill and then basically stiffing the provider.

Yeah.

And, and how stablecoins could effectively be a solve for that, which is like, hey, if you're, if you're, if you're going to just sign up and use this tool, that's fine. We don't necessarily know who you are. We don't need to KYC to use like a, you know, a language model. But you might need to use stablecoins to pay as you go, just to pay as you go. So there's no risk for sure.

I think a lot about it in terms of like, you know, tier 1, tier 2, tier 3 ISPs where you'll call between, call it Verizon and AT&T, and then at the end of the month you'll settle the bill across the tier 1s. Like, I think there's just such an interesting design space for managing token spend payments and so on, using programmable money in the era we're coming into.

Or have you been tracking the showdown between the stablecoin community and the banks? This was in the— this was in this Economist article as well. And I want to know if you think that this is a true collision course, if there will really be a showdown. The concept here, as put in Buttonwood in The Economist, is The prospect of stablecoin growth spooked America's banks. This is because a loophole in the law lets stablecoins offer deposit-like yield without bank-like regulatory burden. The American Bankers Association warned that a $2 trillion market for stablecoins, if it was all dollar pegged, could lead to a 10% decline in bank deposits that would raise average funding costs for lenders by about a quarter of a percentage point, they said, crimping their profits. And driving up interest rates on all manner of loans for everyone else. Is this a real issue? Is there a logical solution? Is there going to be some sort of unification where the stablecoin community and the bank—

and did we not get a— did they not get a solution with the Genius Act?

I think it's— yeah, please, I'll follow you.

Oh, well, I mean, in here it says bankers and stablecoin issuers have since been locked in a fight over the final shape of a follow-up bill, the Clarity Act. And so That's where we are right now. But I want to know where you think it goes from here.

So Polymarket, last I checked, has it around 45%.

Okay.

The clarity bill will make it out of committee and actually pass. And I think the reality is Genius was basically as buttoned down as it could be on this specific issue. And this is why clarity needs to exist to set standards and procedures for actually doing this. I think it's a really—

if you were such a genius, should have provided More clarity.

It's very good. But this is, this is why, this is why the show exists.

Exactly.

I see the crowd doesn't like it. Crowd says it flops.

It flopped.

But for those of you viewing, I see this as the acoustic set versus the UltraDome. This is you guys coming in unplugged. Totally. But, but I know I think it's an essential issue. And I think right now you're seeing also a twin thing around tokenized deposits versus stablecoins and like two different formats for how digital assets, I guess, can impact finance and fintech generally. Um, the reality is I think more competition—

say more about tokenized deposits. I can imagine what that is, which is like I go to a bank, I give them an actual dollar, and you get it, but then they're keeping the dollar there, they're lending against it.

So it is, uh, it is— maybe they, maybe they give me some yield even, and they'll run all of the same infra in the back that exists today, but you have a digital representation of your deposit that you can use. And so in a sense, it is, uh, it is akin to being handed a coat check ticket.

Yeah.

And you get to walk around with your ticket and at some point you can redeem it at the coat check or another coat check potentially versus the stablecoin where you're actually holding the asset itself. And I think those are the two pathways that are getting a lot of exploration in the space. Ultimately, I think, you know, a lot of these rails will merge. And I think more importantly, the question is one of like competition for the consumer. The consumer should have more ways of actually earning on their, on their assets. And so So the idea that we need, you know, greater clarity, as it were, with issuers and institutions that are putting out stablecoins is so absolutely clear. But I think we should view it first and foremost from the question of like, what benefits the companies that are using stablecoins or, you know, the internet generally? What benefits the users? Not how different is this from the current infrastructure that we have? Because that is, you know, not a way to build new systems in the world.

Yeah. And then I guess there is also the sort of the like additive outcome where, yes, you get to $2 trillion in stablecoin assets, but it's for new use cases that are not competing with like the traditional credit card lending, mortgage lending, asset-backed loan industry. And so it's, it's not putting that downward pressure necessarily on the traditional banking system. And so I'm sure the banking industry is sort of like cautiously optimistic. This is all like net new dollars that are being created and used around the globe and for different purposes. But if it's— but if it does directly come after, well, this company used to just use credit cards for everything and now they're not paying us, then that maybe has some negative effects.

And I think the two things you'll see as run-ons is basically how does stablecoin deployment evolve worldwide? To what extent is it a Western thing versus the rest of the world thing? That probably has impact on, you know, USDT versus USDC market share. Like, there's a lot of interesting run-ons to this exact question. The second thing is at least this is why I'm excited to be at Stripe. And, you know, Stripe isn't doing any crypto things for crypto's sake.

Sure.

They are doing, you know, how do we build better rails for our users no matter what the rail is under the hood?

Yeah. And it's less ideological and more pragmatic first principles.

This is a very good tool. Exactly. And we should, we should put it to use in ways that, that betters the experiences. This is where I'm interested in some of— I mean, DeFi has had a very complicated few months with basically a lot of exploits and things like that. And this is still an industry in its infancy. And yet But we're seeing a very interesting merging of the rails with, for example, vault curators on Morpho, folks like Apollo and others taking part in it. And so we're seeing, again, just a merging of traditional finance and online finance in a way that I think is going to continue to—

Recent exploits, is that more social engineering for the most part? This is not AI-led?

In those specific cases, my understanding is that is not AI-led. I think it's very much social engineering, you know, bad setups in various places that have led to unfortunate outcomes. And I think it's just the space maturing. I think we are seeing, as a side note, this is very interesting. So we use at Privy a number of basically AI tools to do like constant pen testing, to review every PR that goes out, to do a lot of this. And we're also seeing quite a bit of, uh, just better, uh, uh, bug bounties coming in, uh, much more sophisticated. But the baseline basically on the bug bounty has gotten much better, and you're now able to string together like more complex attack vectors that you couldn't do before. So that evolution is like mega clear. And one of our engineers, Andrew McPherson, uh, used, uh, ChatGPT to find an exploit in React, uh, and it was a zero-day in React, and The fact that React is a 15-year-old library put out by Meta, it's open source, it's been beaten to death because it basically backs every product on the web. The fact that we're finding these now is insane.

And do you have more clarity on like a zero-day? Does that allow someone to potentially get into someone's computer, their browser, or is it more just like cause the browser to crash in a certain— or it'd be strung together? Because there's like a wide range of what you can do once you exploit something. Can just like, you know, take the service offline, which is bad, or you can like copy on the website into— yeah, that— get into the database, steal— like, like there's a wide range of like negative outcomes.

Yeah, I'm not actually sure what the, what the ramification of his work was, but you know, there's a very good economic market for buying zero days, and the price will depend on what it allows you to do.

That makes sense. Um, has there been— have you been tracking any movements in the stablecoin community to do interesting non-dollar-backed stablecoins. I'm thinking of like Peter Thiel back in like '99 sort of proposed like a Vanguard Total Index, like a VTI-backed stablecoin where you have $1, but the dollar is backed by basically a massive basket of assets that includes gold, oil, real estate, a slice of stocks and bonds. And so it's sort of more diversified than even the US dollar could be, and maybe it's less stable relative to the dollar, but it's— you're bought in on the global economy.

So I think first we're going to see a rise of non-dollar stablecoins as, you know, more people embrace it, especially, you know, and that would just be foreign currencies. Yeah, foreign currencies.

Yeah.

Second, I've heard of a few projects that are trying to basically have an inflation stablecoin, which is the stable actually grows with inflation so that you're on a purchasing power parity like stablecoin. Right. Exactly.

Are the famous assets that you can buy from the Treasury. Treasury inflation protection something securities. I forget.

That one. But that's the idea of one of them. I think they're really hard to instrument and get quite right, which is why I don't think they've taken off just yet. And then I think the third piece is— this is not a stablecoin, but it was a project I heard about recently I thought was very interesting. There's a company called Pearl that is finding a way to do fast matrix multiplication and back a blockchain on this so you can issue tokens based on how much compute you've got ready to do useful inference.

That's interesting.

And so there was a question for me of like, what other useful work could you back the currency on?

Yeah, that, I mean, we've talked to Prime Intellect a few times. They've been doing early on some crypto token-backed compute. And interestingly, the 3D rendering community, I don't know if you're familiar with the Render token, but basically, you know, most 3D artists, if they're rendering CGI for a movie, TV show, something like that, animation, they will typically have an on-premise rack of GPUs that they use to render the graphics or the final output. And the Render Network was an effort that I think was pretty successful to basically, well, I'm not rendering my project, I'll render yours and I get tokens that I can put back on the network because each frame is deterministic and separate. So you can render, you know, if there's 24 frames a second, you can render thousands of frames across across a whole bunch of different machines. They all will look the same when they come together. And so that was the first example of like, oh, a crypto project that's being done for useful work marshaling compute around. And now you can imagine like 1,000 times more ways to—

These are hallowed times. While crypto prices are down and there's so much interesting thing happening in tech, you can actually build very useful things in a way that's relevant.

Prediction time.

When we look back on 2026, what would be your price for USD C price target for 2030.

This is John's favorite joke. Um, but, uh, I hope it's $1.

It's a disastrous outcome.

Um, uh, what do you think the story will be around the intersection of, of, uh, AI and crypto, specifically agents and, and, um, stables?

You know, it's a great question.

I, because everyone, everyone keeps saying it, right? But I feel like we need to have like massive of, it's very clear that when an AI tool has product market fit, adoption is like extremely rapid, right? You just immediately hockey stick. And so I'm like sort of like waiting for a hockey stick moment like that at the intersection of stables and agents.

So here's my prediction. First, we're seeing so much activity, like 35% of new Privy signups today are for people building agentic wallets. So they're using the wallet, they're like giving agents, for example, like some allowance. And then they're like, you know, they have an ability to claw back assets if the agent doesn't spend it. They can control spend in a very like particular way. So we're seeing so much tinkering happening here that I think is going to lead to emergent behaviors that are really interesting. My prediction is first, great use cases are going to happen in the background. There'll be a lot more about like business model innovation, things like what you were talking about earlier, which is how do we gate, you know, payments in a streaming sense so that I can get paid on a, you know, per token basis much more easily.

Yeah, exactly. No risk.

Yeah. So I think that's where we'll see the first few. And then over time, we'll see a much greater pickup of basically agents actually running their own wallets. And that'll be across both fiat and stablecoin rails. Mm-hmm.

Very cool.

Red button or blue button? Would you push? Are you familiar with this?

I have no idea what you're talking about.

Locked in. Locked in.

I don't have time for these things.

That's incredibly bullish. It's a thought experiment. You can go check it out online. Anyway, thank you so much for coming on.

Great to see you.

This is fantastic.

Thank you.

Uh, we'll talk to you soon. Pop those headphones off, put them down right there, and we're good to go. Jeff, welcome to the show. We are here live at Stripe Sessions in San Francisco. Good to see you.

Great to see you.

Throw a headset on, sit down, introduce yourself again. Yeah, you can sit here, here with that headset.

Hello, hello.

Uh, introduce yourself for the, the fans, the viewers, the chat.

Howdy. I'm Jeff.

Stripe.

On the product team.

Jeff Weinstein. Good to see y'all.

Good to see you, Jeff.

This is the first time I ever met you in person. We've been talking on Zoom for 10 years.

Have we met in person?

Wait, is that actually true?

I don't think I've ever met you actually in person. It's like one of these things where you've met so many times through the computer, it's like true.

That's crazy. That's crazy. No, I, yeah, I assumed we had met multiple times.

Crazy.

Great to see you in person.

Yeah, this is amazing what y'all have done.

Uh, how have you, do you ever visit Delaware? Do— when you show up in Delaware, do people just start— you know, just start cheering when you walk out of the airport? You're basically like an athlete that, you know, wins a World Cup. They're coming home and they, they do like a parade, I'm assuming.

I got to meet—

so, so before we get into this, Jeff, uh, Jeff runs Atlas. He created Atlas. You're responsible for like a quarter of all new C-corps.

Atlas is now incorporating about 26 6% of the new—

Gong moment?

I think it's a gong moment. I think it's a gong moment.

Oh man, this is a life honor.

But there we go.

Oh yeah, it's got a really good ring to it.

26% of all new C-corps, right? That's the focus? Correct.

Yeah. And today we're celebrating our 100,000th new Atlas business was incorporated recently, and we brought the founder, and she runs this incredible business for AI software to help power plant owners run batteries in a better way is like fantastic. So it's been a great thing. I've been recently focused on all this crazy agentic stuff we're doing.

Yeah. What, what is actually changing about Atlas? Because I've used it several times. I think I probably used it 5 years ago or something. I don't even remember how old it is.

We used it for TBPN.

We used it for TBPN a couple years ago.

Fantastic.

And it was like fine. And I didn't really have any product requests.

It's not, it's not fine.

No, no.

I just mean like, it was so beautifully like that, that like the point, the point is that you don't think about it.

Exactly.

You're like, that is how it should have worked.

Exactly. And I'm always fascinated by products that are like, like, okay, yeah, thumbs up, it did my job.

Like, yeah, I'm good.

What, what, what about it?

I'm gonna make you go use LegalZoom.

Hey, that's a, it's one of these products you just want it to work like electricity under the—

exactly, exactly.

So I have—

how do you keep it there?

So, so I want to ask you about, is there an— do you think there will be an incorporation apocalypse? Because theoretically I could just have— I could tell a smart agent that can use my computer, that can use the internet, Hey, do you like go— I need to create a C corp in Delaware.

It turns out that the agents want to use Atlas.

They'll use Atlas.

And, uh, you know, not, not to preview too much, but coming up, Atlas CLI.

Okay.

There we go.

And another exciting—

generate me 10,000 C corps. Generate me 1 million C corps.

I think that would be a lot of money, but yeah.

Vibe, Vibe Incorporation coming. Yeah. Another exciting Atlas update is now that you can do full safe fundraising through Atlas. And coming soon, you'll be able to move the money from investors directly into your treasury account and just have these amazing little fundraising links to be able to quickly raise money along the way. So we think it's another way we can help entrepreneurs get going. Though my recent focus has been all this great AI stuff we're doing to help commerce businesses succeed in the next wave of the internet, all the Lynx CLI work. So you can have your robot help you shop along the way safely across the internet. Machine payments protocol to have robots send stablecoin to each other in small bits. Yeah, it's been a really whole new world.

See the, the customer segmentation, customer target, customer ramp. Are you rerunning the Stripe playbook? Like Stripe was adopted by a lot of startups. It was very— it sold through developers, which was very unorthodox at the time. You used to sell through the CFO. Now you're selling through the CTO or just some developer who chooses Stripe. Are you trying to be the, the incorporation platform for every tech company, every startup, every software small business, and then you're going to go into restaurants, you're going to go into Fortune 500, the next Fortune 500 companies. I mean, I guess it doesn't quite map to, uh, to Stripe because, uh, the big companies are— don't necessarily need Atlas if they're already up and running. But, uh, how do you think about, uh, like landing and then expanding?

Yeah, I think that we're trying to get all the technology businesses in the world to start on Stripe. Start with, you know, just as a project, use Stripe projects to be able to configure your entire stack. I mean, do you remember having to go and make 100 little accounts for to test a database here, test the CDN provider? Oh, should I use this particular thing for email? And then you have to go, you know, type in password and be like, no, you need two dollar signs or one dollar sign, and you gotta go get the secret keys and like put in the wrong place, and then you gotta rotate. There's like a whole mess. Constantly. And so we're excited to take, make basically, continue to make everything programmable, making corporation programmable, make selling online programmable, and now we're making provisioning entire new stacks. So we think we're gonna see, you know, we did 100,000 Atlas startups over about 8-ish years. I can imagine doing the next 100,000 in, you know, in another year, another 2 years. We're just seeing the curve of new businesses grow so quickly. I think it's tools like this agentic provisioning of an entire stack quickly which is going to really ignite the next wave of founders to go from just Vibe coding to like an entire Vibe business.

How much attention are you paying to how Atlas shows up in LLM results? Because there's a lot of companies, Profound is one, James, the founder, has been on the show that help you optimize that, understand it. I think it's helpful to just create a product that's loved by everyone that uses it and and ultimately gets talked a lot about sort of naturally. Is that something that you're putting time directly into, or is good awareness in LLMs a byproduct of just building a great product?

You have to make your business legible to the agents. You need, you know, if you want your checkout page to have bright colors, like that's not what the agent wants. If you want your homepage to have a crazy JavaScript thing that like loads only in certain browsers with the GPU turned way up, that's also not that's not what agents want. Agents want perfect information about what your offer is, what the tool is, what the SKU is, what the service is, what the shoe is, and we're basically helping businesses just make their entire checkouts, their entire SKU offering legible, and Atlas is actually part of that too. You know, it itself is a business. Atlas actually runs on Stripe, and so we, you know, make sure that the agents have access to the full, LLM-formatted Markdown text. We bake it into all of the places where agents can self-serve and learn about our products. And then we make everything at Stripe a programmatic primitive such an agent can use it directly. And so we've seen that the CLI, the command line interface for interacting with Stripe, which had, you know, not like a side project for Stripe, but kind of an advanced tool for developers who never want to touch the dashboard, like sort of a terminal purist. This has become this incredibly fast-growing part of Stripe. Stripe because the agents are like, "Yes, I love the fact you have perfect information about all the things I can do on Stripe. Of course, I'm going to do so," versus try to—

Computer use around the website.

Yeah, it's a little clunky. At some point, it feels almost like a Dilbert style of using a computer, whereas the agents can just use it directly.

Can you take me through maybe like day in the life? Today is probably pretty different than like 2 or 3 years ago. I'm sure AI seeped in a bunch of places, but maybe if you go back to like, you know, a little bit earlier, like how has your day-to-day changed over your career and what is it like now?

We're really just following what customers want and they're coming to us. I mean, I've worked at Stripe for 8 years. They are coming to us with a veracity of new needs. You know, before it was like, hey, let's kind of put a meeting on the calendar next week to talk about how to grow your business in the next country. Now they're, I need to launch in 50 countries this afternoon.

Yeah.

Sorry, what is sales tax in all those places? How can I mimic all these fast-growing startups that are going at light speed? And so basically we're taking everything that we built at Stripe and turning it into sort of a managed service. You have managed payments so that you can sell in all these countries. We have the machine payments protocol where you can say, hi, I've been using, you know, I spent 20 years selling to consumers. Now I need to sell through agents, take all the tools we have and make, make things available to agents. So my day-to-day is basically almost a whiplash of like customer meetings where they're coming to us with either a new type of fraud vector they're seeing through AI, they want to expand into new countries, and we're just trying to take the pieces, the programmatic primitive pieces we've built over the last 15 years and turn them into these sort of CLI tools. Is normally we would, we would put out a beta or we put out some waitlist and we'd wait for a couple people to sign up and then we'd email them, then we'd send them the docs, then they'd be on vacation, you know, or like get busy. And then like you'd hear from the next week. Now we put out a beta, they, developers send their agents after our beta page. They will go around our waitlists to use it. They'll give us feedback from their agent logs of how the developer about how the agent was using our developer tools in like minutes. And so the development cycle at Stripe is just increasing, increasing, increasing, increasing because the customer feedback loop from the highest taste developers is so now agentic that we don't even have basically have time for meetings. It's just like putting these betas out there and getting faster.

So you think like less meetings now?

I think it's less meetings.

I was wondering if you were, yeah, if it was like, I imagine that you're a fan of like talk to your customers customers, right? Like, that's a big piece of what you do. That's your— that's very much your job. Of course you have internal meetings, but you're probably talking to a lot of customers. But the shape of the feedback—

I was hoping that with AI you wouldn't have to talk to customers anymore. You would just throw like 10 Slack prompts in there and then just say like, you know, I don't have time.

I don't— we're very fortunate slash cursed with that our customers are not shy. You know, they are extremely happy to hop in Slack with us and just bombard us with their needs, but now they're pasting the context windows from their agents along with them. And so it's sort of this back and forth between our agents and their agents trying to like winnow our way to the best product. And, you know, so again, I think the just like the clock cycle on Stripe has just dramatically increased, which is why you saw so many of these announcements recently. And we're really proud that a lot of this stuff we're doing is just like GA. You can use the Link Wallet today to delegate transactions to your agent in a super safe way. It's very slick. In a push message, oh hey, time for my agent to, you know, buy me that gift. Like, fantastic, thank you, agent. Um, and so we're just, we're putting this all out there and really encourage people to try and get us—

are you forcing yourself to constantly, uh, use agents to buy stuff? Uh, it's like right now I feel like a bunch of people want to, and there's certain instances when, where they can, uh, but it's not exactly easy everywhere.

Yes, it's a two-sided, it's sort of like a three-sided marketplace.

Place.

You know, you've got on the seller side, they need to be able to accept payments from agents. That's why we have the machine payment protocol, or all our APIs are being upgraded so that you can just use, you can, you agents can check on.

And I go through this all the time where I will like see a product that I'm interested in, I'll be on my phone, I will click add to cart, and then I will get to the cart and I'll see they're not on Shopify, and then I'll just churn because I'm like, I don't, I don't have time. I didn't like, I don't have time to go through, fill this out, punch in my credit card. And so I'm really, really excited for this moment where it's like, I want to buy this thing, please buy it for me.

Totally. Yeah, it's— in some ways we have this incredibly modern world in which everything's available at our fingertips, but then it's like unfortunately overly available only on our fingertips, where I'm sitting there typing in 16-digit numbers, like trying to like peck my way through a checkout page. Even the Stripe checkout page, which is like pretty high-performing, I'm still like, I can't believe they don't use LingQ. I can't believe I can't just like have my agent go buy this. So, on the seller side, we're making it incredibly easy for any of the 5 million businesses on Stripe to just accept payments from agents. That's like one side. Another side is on the consumers, they need to be able to have a way to super ergonomically on their phone, slick mobile app and link, say, "Hey, agent, I approve $50 transactions for the next 30 minutes in the following topic. Go, agent, go." bam, with, you know, 2FA, Face ID, whatever, I can now approve an agent and it can safely shop for me on the internet, like giving like a rambunctious intern like a perfect card that can only be spent where I want. And so the third side of the marketplace is agents themselves, whereas we're working with, you know, Copilot and Meta and ChatGPT and Gemini so that baked into their services, they have access to all of the buying and the customer mandates so that we can make agent-to-conversational this super slick, super safe way to buy online. I think we're now finally getting all the infrastructure in place where we can get this flywheel starting to spin. And with the model progress, I think you're going to see amazing new applications in Agenda Commerce.

Is there a sentiment internally that the company got kind of lucky by focusing so intensely on developers and then accidentally getting the benefit of being extremely well built for agents? I guess it's like, because you guys are obviously, you guys are You guys are, well, incredibly smart, incredibly forward-thinking, but when you guys started focusing on the developer experience intensely, you, I don't believe that you were thinking about AI agents a decade later.

This is a surprise for us too. It's a useful externality to our non-religious religious obsession with programmatic tools.

Secular and religious.

Yes, it's—

Religiously secular.

I can't, there's not too much which which is like a faux pas at Stripe, but saying, yes, let's put this incredible piece of financial infrastructure behind a gate that no one can use, and you have to call and go through a maze to get to.

Steak dinner.

Yeah, exactly.

You're not a steak dinner company.

Yeah, we showed how many users are now signing up for Stripe each day, which is in the many tens of thousands a day. And that's a lot of steak dinners to get each of them turned on. So it sort of actually surprises us how few companies—

Humanoid steak.

Yeah, like how few companies in the world actually provide self-serve services. And, you know, I could say that, but I also know that there's 10,000 people at Stripe who are like doing the work to make those things go.

Of course, there's steak dinner every once in a while.

Yeah, and yeah, yeah. I mean, I would like to get a steak dinner personally.

For the 100th Atlas signup.

Yeah, 100,000. How ambitious is Treasury? I imagine you've been getting the customer feedback from companies forever saying, hey, I incorporated, like, why are you pushing me out to these other financial products when you already have all my information? I know you guys can store money, I know you guys can move money, but how ambitious is that product?

Yeah, Stripe Treasury is going to be this incredibly straightforward way for a business to be able to store any currency from any country, move it around the world, and just handle all of your banking and money management activities in one place.

This.

And it— I think this is something that we've been building the infrastructure for, for so long, you know, the ability to do currency conversion instantly or be able to pay out cross-border. But we finally decided that because of stablecoin, now is the time to make this a truly global product from the beginning. Whereas years ago, we would have had to make sort of a country-by-country decision to do a thing like Treasury. But we're going to really global out the gate and get to 160 countries by the end of the year. And so that's really where our ambition comes from is how to make this global infrastructure. And then of course it's going to have a super slick UI. It comes with a metal card, 2% cashback, really super easy to use. And if you use Stripe to earn money through your customers, everything's in one place. All the money settles quickly. It really feels like one of these kind of better together products where where in my head I've written documents that say, hey, this will be a lot easier. And then when you actually go to use it, you're like, oh my goodness, this is like much easier.

Very eloquent way to say we're coming for everyone.

You know, there are many services in the world that we're not going to by accidentally be able to be better than. And I also will note that everything we're building Stripe Treasury on is available for anyone else to build.

Sure.

And so we're going to see some much more specialization, I think, in financial-related tools. And we're excited for Stripe users to be able to do it really quickly.

Okay.

Yeah, yeah, yeah. I mean, I think, I think the agents coming to finance will be a big, a big story of basically the next 2 years because as a somebody who started my first real business while in college and then I was realizing like, wait, okay, so we've signed this like contract and then there's like these invoices, but then like the invoices get sent around and then there's like all multiple stages of approval and it's like, why does it need to get approved if it's already in the contract? And, and, and you go, you go like understanding like the complexity of money movement and how many humans are in the loop to every process. And it can make, it can, it can be great that there's humans in the loop because it prevents fraud and, and, and hopefully prevents errors, but it also creates errors and it also slow slows things down a lot. And so when you think about in the future where these things can be handled by machines, I think it's going to be—

Yeah, I think that's very, you know, it's like dangerously skip permissions and money like tends not to be exactly the joining parameters you want. We want smartly suggest permissions.

Yeah.

You know, like I look forward to the day where I get a notification from my smart robot that says, hey, here's all the ways you can improve your business. Here's the things you should be able to do to close your books. Here's the things you can do to handle the invoice. I'm looking forward to those permissions, you know. I don't want to be—

but even something like, so imagine a business owner receives an invoice from someone. This happens like still, like if you've worked with freelancers enough, you will have experience where a freelancer sends you an invoice. It basically just has a number. It does not have like any account details. You're like, okay. And then you're replying back, okay, how, how can I pay this? Maybe it has an email. Yeah. But a world where you receive the invoice, you just drop it into your own one kind of agent and you say, pay this, I'm approving this already, I have approved this.

And the agent can go follow up and keep everything in person as well.

Yeah, it can go email the person, it can confirm everything, and it can just process it. It's like that historically is something that I've done with having like maybe a part-time, uh, part-time like back office financial operations person where I'm like, this is approved, pay it. And it's like, that doesn't need to be like a, uh, it's, it's not like really high leverage work. It probably doesn't need to be something a business is paying a thousand, couple thousand dollars.

We, we used to sort of joke that every chat app will eventually become like a banking app, but now we act like every banking app is actually now becoming a chat app. Uh, and so I think we're gonna see the human agent interface related to the full lifecycle of your business to become the new modality by which we run our businesses. And you want that to be in a super programmatic, super safe place, uh, that works globally. And that's what Stripe, Stripe Treasury, all this new payment stuff we're doing for Gentex.

I think you guys are culturally very humble, but the more I'm processing every conversation I've had today, I'm like, wow, they're actually coming for everything. They're incredibly well positioned. Everything's going to be programmable. Everything's—

agents are going to be doing everything.

We succeed exactly when our customers succeed. So I feel— I think I sleep very well at night knowing that our business model maps exactly to our customer success.

Yes.

What else do you have on that thing, by the way? We got a lot of stuff.

Faster mode.

Okay. What about the bottom right one, just for fun? What's the bottom right one?

Bottom right one? This is the gong. Perfect.

Okay, great. That's a good one.

Okay. We have a whole different setup there. We got so much stuff. Do we have time for one more question, you think? Or is Zach ready to rock? What are you thinking?

Oh, there he is.

We got a couple of minutes?

Yeah.

Okay, cool. Hypothetically, like 21-year-old new grad, college grad comes to you and is One maybe wants to join Stripe, other one maybe wants to build a business on Stripe. How are you routing them in 2026? What are you talking them through?

Well, I love that those are the two options. Yeah. So that, that, if that were the case, that would already be a win. There's many amazing companies.

Sure, sure, sure.

I think that people should really, you know, follow their passion. I mean, look at the two of you, what you've built, right? Like, I've got to know you over so many of your awesome new entrepreneurial attempts, and I remember you telling me you're going to do this TV show. And I wasn't like thinking, you know, I wasn't like, no, but I was like, certainly I will never be on that show.

Sure.

Yeah. Stop by that crazy thing.

We're not gonna make a pub and then have you do College GameDay in front of it. That'll be very strange.

Yeah.

So I, I mean, like, look, you, you've, you've followed your, you followed your noses on this topic.

Mm-hmm.

And you're, and you have a good sense of it. And, and if you talk to someone who's, who's got that kind of passion, the tools to build something today are, what all of us, I mean, dreamed of. Imagine having what we have today when we each started all of our startups in the past. And so I honestly, I really encourage them to try it. And I will also say if they're considering joining, I mean, look, Stripe's a 10,000-person company. We're probably in a reasonable high echelon of like low communication costs, but there's 10,000 of us. You gotta have, it's like a higher communication cost than you and your co-founder. But look, the Stripe Projects thing where you can provision all these third parties That got started 7 weeks ago by 6 people.

Wow.

You know, we, we, we put the links, the link for Agents for Wallet, that was 10 weeks ago. And it's, again, it's built on all the stuff we have, but we're, we're benefiting from all of these new tools as well. So I would tell that person, I hope she finds a passion that she has, especially in an area where has a personal interest, uh, and just go at it hard with the new tools. Um, and otherwise, you know, stripe.com/jobs.

What's, what's underrated about working at a big company for a long time?

I'm on year 8. What's underrated?

Overnight success.

You know, you—

Presumably, I mean, you've had how many, I would guess like 100 really talented people have tried to convince you to join their effort or leave or, hey, let me give you like $50 million to go after this thing.

Have that one call me. The other ones, no. But I I—

you're joking, but like you could easily go out and raise like $50 on $250 for some like payments related.

Yeah, I could make a rival TBPN. Yes.

We need more of those.

Interesting that you're not saying I'm going to go build a rival Stripe. That's very bullish for Stripe.

Yeah, yeah. This one seems to have worked out very well for you. So yeah, this one I could spitwise. I think that what's really kept me me very energized and more so is I tell people, like, on your worst day of work, if you're looking forward to your most unhappy customer, then you found your place.

Sure.

And I, like, I want to hear from an entrepreneur somewhere in the world who's got a serious problem with Stripe, and I want to get on the phone with them. And that is just, that's enough of, like, that's enough of a jolt to keep it going.

You're hearing it here first.

If you have a problem with Stripe, DM me. I will give you Jeff's personal phone number.

Cell phone number. Yeah. 410-913-2877.

That is it.

So give me a text. I couldn't possibly get more spam. So let's just add to it, which is another thing I think we could solve with machine payments, but a different day's story.

Good point. Good point. Well, thank you so much for being here, Jeff.

Okay.

So good to see y'all. Thanks for coming to our podcast.

We'll find you after.

Okay.

We'll hang out after.

All right, guys.

Take that off and let's bring in Zach from Bridge. Coming back on the show, second, third time we've talked to him a bunch, right?

Couple times.

Zach, he's bringing the world's smallest cheeky pint. Cheers. Oh, yours is nice and frosty.

Cheers.

Welcome back to the show in person.

Nothing like a non-alcoholic micro pint.

Wait, yours is non-alcoholic? Because I got 0%, but I asked for 0% non-alcoholic beer.

Oh, okay.

So my beer is 100% alcoholic beer, 0% non-alcoholic beer. Maybe, but I still tell everyone I'm drinking 0%.

But you can't tell. There's no— so you have no idea what you're drinking.

I have no idea.

Yeah.

All right, let's start here.

Okay.

The Economist says stablecoins are over.

Fire back.

Oh, man.

In The Economist.

Oh, man. In print. Speaking of things that are over in print.

I love this. It says two stablecoins.

Yeah. Print media says the future is over.

Says the new thing that is coming is over.

Where the shots fired. Why the stable market is fizzling. The stablecoin market is fizzling. They say stablecoins are, to proponents, the respectable face of crypto. I think it's pretty respectable face right here.

I hope so.

In contrast to volatile Bitcoin, let alone speculative meme coins, they should be backed by holdings of Treasury bills or other dollar-denominated assets. On paper, this makes them safer store of value than many fiat currencies. But they say after growing by 30% in the 6 months to late October last year, total stablecoin assets have barely budged. Is this true? Is it over?

Oh man. Yeah, it was really nice knowing you guys.

It's actually my last— Yeah, yeah.

This is the last time you will ever let me on this show again.

So, I mean, there are a bunch of different predictions. Projections in here. Uh, the stablecoin market's worth $300 billion. Some people are calling for $2 trillion over the next, uh, 3 years, 4 years, 5 years. But all of that is very macro level. I want to hear about what you're actually seeing in terms of adoption, new use cases, glimpses of, you know, new areas to deploy stablecoins, and then just what the big market movers are.

Totally. I mean, we I mean, our— so our business, you know, we shared this in the annual letter last year, grew 4x last year. Our business continues to grow enormously. Like, we grew 20% over the last month alone. So, you know, we certainly don't see correction, correction, stablecoins slowing down.

And I do think total supply is, you know, flat, but these things always go through.

Yeah, if you want to look at when, when you have a correction in general digital asset prices, then trading activity at times falls, and then it looks like you have a slowdown in adoption. But then you actually have to look at how people are actually using stablecoins for non-trading-related activities.

Exactly. Like, stablecoin transaction volume is growing enormously. The number of transactions are growing enormously. And we certainly see that, you know, the number of folks building with stablecoins is growing enormously. Like, we have folks from You know, a bunch of banks now are leaning into stablecoins. Pretty much every fintech that you could think of that is meaningful is doing something with stablecoins and probably doing it with us. I mean, we obviously announced yesterday with, with Meta paying out folks in stablecoins. So I do think that this, like the hyperbole around stablecoins is dying down a little bit and that creates the opportunity for some of this, this like FUD. But yeah, you know, at the same time we're like like, you know, there was this thing, I think Henry said it yesterday, which is like funny, like the crypto world where for a long time people were like pointing at every problem and being like crypto solves this. And now they're no longer pointing at every problem and saying crypto solves this, but it actually is solving—

You guys are happy to solve problems with crypto.

Yeah, like very, very quietly under the hood it is solving a number of problems like banks for cross-border payments. We're seeing it for working capital needs. We're seeing it for for just infrastructure for cars and infrastructure for global expansion. So it's certainly not slowing down.

So there's like some very big meaningful categories. The cross-border payments one's very tractable, huge.

Yeah.

Is there a coming intersection of like Vibe Coding and stablecoins that you're seeing glimpses of? I'm just thinking about like—

Yeah.

Like there was this moment when iPhone games became fairly easy for like a kid to build in their bedroom. And you got Flappy Bird, and I think it was $0.99. And I'm wondering if, if micropayments, stablecoins, there's some sort of intersection there where you get Vibe Coding and you get some sort of new flourishing like ecosystem of like, of like smaller transactions. I mean, people don't love micropayments, but there's like, there's something that people I feel like could play with if they're spinning something up in a, you know, in an afternoon.

Totally. I mean, I, you know, one of my reflect— one of the reasons why we started Bridge is like I was reflecting on the fintech space generally. This was like in 2020 or so.

Yeah.

And I realized, or one of the realizations that I had was that the cards were basically just the entirety of fintech. Yeah, everything was like, you know, and people talk about platform shifts, like AI is this big platform shift and mobile was a platform shift. But in financial services, platform shifts take a little bit of a different shape. And cards were this new form factor and cards combined with the internet was like 90% of all—

or you're talking about like credit cards from like the '70s?

Yeah, that—

yeah, yeah, because we rode that boom for so long. There was a time when exactly people would pay by doing like reverse debit, and there were so many hacks on top of the credit card.

Totally. It was a new platform, it was a new primitive on top of which all these new financial services got built. And cards were like growing meekly for a while, and then the internet came along and they were like uniquely well suited obviously Stripe, this is what we do, but uniquely well-suited to solve payment problems on the internet. And then they just took off. It became reflexive. You know, more people were paying on the internet, so more people had cards, and more people paid in person, so then more people bought stuff on the internet. And then, and then it just took off. But it was this like brand new platform shift in payments tied to an external platform shift. Our, our view is that stablecoins represented another platform shift in payments, and I think AI could be the corresponding one that helps drive drive adoption of stablecoins. And I think that all the payment problems that we're solving now are like fiat payment problems, and there's big markets and so on. But I think the much bigger market will be, you know, the new means of payment that are going to be required in these agentic services. Like, and, you know, just simple things like the incrementality of payments is going to be very different, very small. The velocity of payments is going to be very different, very fast. Uh, the reversibility of transactions, like cards are, you are are not well suited in a situation where you don't want transactions to be reversed or disputed. And so I think there's a lot of benefits to stablecoins, but we're at the very, very early stages. So you're kind of squinting and looking way down the tunnel and trying to predict what will happen, but it feels like it's there.

Can you explain the tension between the traditional banking sector and the stablecoin world?

Oh, man.

What are you saying if there was someone with a bank and they say, I make my money by taking deposits, and I'm heavily regulated. I have to cover those costs with, you know, with interest. And you're going to be able to do it without as many regulations. You're going to outcompete me. Should I jump on the stablecoin boom? Is there a world where we work together? How should the traditional banking system interface with the stablecoin world?

I mean, it is— I would say complicated.

Okay.

Do you spend much time in DC?

I, I went earlier this year and, um, it was actually like mean— it was like very heartening. I don't know if y'all have spent, spent time there, but like I hadn't been to DC since I was in high school and like a high school trip. I grew up in North Carolina, so we could like take the school bus up there. And, uh, I went into, you know, like you're working at like a tech company or something like Stripe and it feels like It's like impossible to understand all the ways in which Stripe is growing. And then when you zoom in, it's like literally just like one developer coming to Stripe and signing up and like, you know, one person selling. And then I went to DC, and DC feels like this incomprehensible, like, you know, complexity of decision-making and stuff. And then we just met with a bunch of folks, and you realize it's just like one person talking to another and like them trying to learn, and then they're going and talking to another person. And they were like genuinely curious. And these folks are like voting, you know, debating stuff on like agriculture and like, you know, airline safety and then coming in and talking to me about like the minutia of like stablecoin regulation. And they were like pretty well-versed. So I think that like one speaks to how high on the docket it got for them and like how important it is. And it has like, you know, I spent time there beginning of the year. We've spent more time recently as like Clarity Act has come about, and we think that, you know, it has certainly put the banks and the more stablecoin-friendly folks at odds. But my experience has been that that is not universally the case. Like, there's a whole group of banks coming together. There's multiple groups of banks coming together to launch stablecoins. There are multiple community banks that we're actually working with on different stablecoin initiatives. The one sort of outlier is, um, is some of the other folks that have more directly, you know, JP Morgan has their own stablecoin and like their own thing that they're trying to push forward. So there is some competition and, and that sort of thing there. But, um, I, I hope that we can all like work together, but it certainly has been contentious at times.

Give me a price prediction. 2030, if I put $1,000 in stablecoins, how high can it go? What do you think is possible?

Oh man, I mean, I think you are going to be be exactly as wealthy then as you are today.

That's fantastic.

With inflation.

With inflation.

Advice to founders that are feeling— that want to start a company today but maybe feel like they're too late as part of this AI boom? Because I feel like when you started Bridge, there had been— there was big stablecoin companies that had been created. Maybe you felt at times that, am I too late? Clearly you were very, you were very early and the timing was perfect because you got to like sit out a terrible bear market and then you, and then you were like, wow, I'm a genius.

That's certainly not what everyone told me. Many people in the midst of it told me I was wasting, wasting my time. Yeah, yeah, I would, I would say, I mean, the lived experience for me, and like, you know, you kind of see this now in some of these hotter spaces, is that folks kind of build what they think others want them to build. Um, like, think about something in AI, you know, you're building this thing for this market that isn't like a thing that you deeply believe in, and so you're, you're building what you think an investor wants to fund or what you think other people in the industry will think is interesting. And what happened for me is like, I was building something that I deeply had conviction should exist. And that is what enabled me to persevere through like some really dark times and build a company that, you know, ended up, you know, hitting the, the sort of timing situation at like the perfect, at the perfect moment. But like literally, you know, if I didn't have that conviction, I would have pivoted, you know, a year before, you know. I mean, FTX— I mean, literally, you know, Terra Luna blew up, then FTX blew up, then SVB went under. I mean, it was like one stablecoin went under, the whole crypto market was done, and then USDC depegged. And so we got—

yeah, that's why it's not funny when you make this joke, John.

Yeah, I mean, we got booted off like 4 banks before we got started.

I assume stablecoins are going to stay stable, but I understand there is risk. There is a new technology. Can you help me understand the interaction of like AI coding agents, AI code review in stablecoin software development. Historically, I have always had the perception that anything in crypto is not just higher stakes but more technically complicated. It's more like, I don't know, writing a compiler than, you know, a front-end website. And so it's high stakes. Uh, LLMs do make mistakes. At the same time, the new models, Mythos 5.5, like like superhuman. And so there's probably a way to integrate that successfully. What works? What are you hearing or seeing that makes sense? Are we still in like the Centaur era where you need a great software developer working alongside AI? Is this something where you can just throw an agent at it and it's, you know, oh, there's no more risk anymore. It's not going to hallucinate.

I mean, I would say for us, it kind of feels like we are are one or two steps behind where it feels like some of the folks are who are in pure software. Where they're like, oh, I'm just going to have this long-running agent vibe code an entire browser and ship it in a week and be fine. We're nowhere close to that. Like, we couldn't— there are big parts of our business that are pure software in a ton of ways. Like, think about transaction monitoring or stuff like that. But the risk from a regulatory reason of having things off and then missing things and then not being able to appropriately justify it because someone vibe-coded a thing, it's just not possible. So the ways in which we use it end up being different. And another very quick example is we're standing up a national trust bank. As part of standing up a national trust bank, there's a bunch of requirements, one of which is that every commit has to be approved. Wait, what? So, so like you literally cannot have—

it's over for fast takeoff. Fast takeoff's cooked. This is the most bearish— still got it, humans. Still got it. Yeah.

FDA for AI.

This was something somebody was saying.

It is like, you know, the, the Jevons paradox of employment. We're just gonna have to hire like more and more and more code reviewers.

Yep. Everyone's gonna get employed. Yeah.

It's an interesting— I, I was talking I was talking to someone, small business owner, who was very, very fired up on VibeCoding. He told me specifically he hadn't slept in 4 nights. I was asking him the product that he was working on. He was like, this is something I've thought about building forever. I was like, well, have you ever thought about rebuilding this sort of vertical SaaS fintech platform that you run your whole business on? And he was like, ah, no, it's $1,200 a year. It's like, they give me so much. It's like, it would never make sense. Sense to— and so it's like the one piece of like core technology that he runs his business on, he even— he, he's so fired up on Vibe Coding, he's like, I'd never rebuild that, would never make sense. And but, but so there's two things, there's like a value equation and then there's like a potential risk kind of like cost equation. And I feel like with stablecoins specifically, the potential in crypto, like the potential cost is so astronomical where like saving a little bit here to then introduce $100 million of risk.

Yeah, but there's a bunch of other things that are amazing. Like for us, we have like re-architected a lot of our infrastructure to be able— like before, one of the hard things in crypto is adding a bunch of different blockchains. A lot of blockchains can kind of be the same, but many are very different. And so it's super tedious to add a bunch of different blockchains. Or the same thing is true of like adding new payment rails. So you might add a US bank, and then you need a redundant US bank, and then you need a European bank, and then you need a bank in Mexico. And all of these integrations kind of look the same. It's like you're getting money in, you're checking the monies in, you're sending money out. And what we have built is the ability to sort of add all of these things very, very quickly. And so we took what used to take 2 weeks to add a new blockchain now It takes, you know, someone actually did it in a meeting the other day of like adding a new stablecoin on a new blockchain all with a bunch of cloud that, you know.

Does that mean we need 10,000 new layer 1s?

That is, it's the Jevons paradox of blockchains.

Because it's easier to use a new chain, we need 100,000 more, a million more.

Yes.

Maybe honestly one for one. And every American should have their own layer one.

We're just maximalist across every dimension.

There was a moment in stablecoins maybe last year where it was like this, like this, like small, like county in New Mexico is going to have their own chain.

That was, that was a stablecoin.

That was Wyoming. Wyoming has its own stablecoin. So we're going to have every state's going to have its own stablecoin. Every state's going to have its own blockchain, maybe even every city. Yeah.

Yeah.

You know, you could have your own blockchain. Yeah. Neighborhood.

There you go.

Yeah, the Mission stablecoin. Yeah, the Presidio stablecoin.

Which one would you buy?

Soma?

I'm going for the Mavericks stablecoin.

The Mavericks stablecoin.

Okay, well, it's just a taste thing.

Anything else?

No, this was great. Thank you for coming on.

So much.

Great to meet you in person.

You can close it out with us. Yeah, leave us 5 stars on Apple Podcasts and Spotify. Sign up for our newsletter at tbpn.com. We'll be back tomorrow in the TVPN UltraDome, 11 AM. Shark Thank you for tuning in and have a great day. Goodbye.

It's been an honor.

Cheers. Goodbye.