Patrick Collison — Why Silicon Valley's most talented should leave

We discuss: * what it takes to process $1 trillion/year * how to build multi-decade APIs, companies, and relationships * what's next for Stripe (increasing the GDP of the internet is quite an open ended prompt, and the Collison brothers are just getting started). Plus the amazing stuff they're doing at Arc Institute https://arcinstitute.org/, the financial infrastructure for AI agents, playing devil's advocate against progress studies, and much more. Watch on YouTube https://youtu.be/WU-lBOAS1VQ. Listen on Apple Podcasts https://podcasts.apple.com/us/podcast/patrick-collison-stripe-ceo-craft-beauty-the-future/id1516093381?i=1000646268840, Spotify https://open.spotify.com/episode/0CN45zb7saXUjlVuUtmbEZ?si=kHwGCph2Sg2M9w5HSglzGw, or any other podcast platform. Read the full transcript here https://twitter.com/dwarkesh_sp/status/1760342775193809154. Follow me on Twitter https://twitter.com/dwarkesh_sp for updates on future episodes. Timestamps (00:00:00) - Advice for 20-30 year olds (00:12:12) - Progress studies (00:22:21) - Arc Institute (00:34:27) - AI & Fast Grants (00:43:46) - Stripe history (00:55:44) - Stripe Climate (01:01:39) - Beauty & APIs (01:11:51) - Financial innards (01:28:16) - Stripe culture & future (01:41:56) - Virtues of big businesses (01:51:41) - John This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.dwarkesh.com https://www.dwarkesh.com?utm_medium=podcast&utm_campaign=CTA_1

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Dwarkesh Patel?

  Okay, today I have the pleasure of speaking with Patrick Collison, CEO of Stripe. Patrick, first question, you have an excellent compilation of advice on your blog for people 10 to 20, and you say there that once you turn 35, you'll write some for people in their 20s. What advice do you have for us now, the people in our 20s now? When's it coming?

• 0:22Speaker 01

  I haven't really thought about that.

• 0:26Speaker 01

  The one I've been wondering about recently is, you know, I said for that advice for people in their teens, they should go to San Francisco. And I wonder for people in their 20s if they shouldn't go to San Francisco. And I mean, Gleb, and I think there's a significant set of people who should in fact go to San Francisco. But the thing that I wonder about is, there is a set of career paths that I think some set of people ought to pursue and would derive most fulfillment from pursuing, and that are really valuable for the world

• 1:10Speaker 01

  if pursued, that require accumulating a lot of expertise and really studying a domain in tremendous depth. And I think San Francisco valorizes, and look, this is also San Francisco's great virtue, San Francisco valorizes a kind of striking out on your own iconoclastically dismissing the received wisdom and the founding archetypes and lore of the Steve Jobs and the Bill Gates and all the rest. And I'm way less successful than those people, but like to some extent Stripe, in as much as it fits a pattern, is an instance of that pattern. And look, that's great, and I'm kind of happy that that phenomenon exists

• 1:55Speaker 01

  in the world. But I don't think that, the world needs lots of other things, right? And I don't think San Francisco particularly, I mean, I'm again using San Francisco as a kind of metonym for a cultural orientation, but I think that San Francisco doesn't really encourage, yeah, the

• 2:18Speaker 01

  The pursuit of really deep, technical depth. And we're here, we're recording this in South San Francisco. And, you know, South San Francisco is most noteworthy in the, in the corporate world for, of course, being the, the headquarters of Genentech. And, you know, Genentech was, was co-founded by Bob Swanson and Herb Boyer. And, you know, they, they produced cheap insulin for the first time with recombinant DNA. Like, Herb Boyer couldn't have done that, like, at age 23. Herb Boyer first had to accumulate all of the knowledge and the skills required to, you know, be able to invent that over the course of a multi-decade career.

• 3:01Speaker 01

  And then, I don't know what age he was when he finally went and invented it, but he was not in his twenties. And, and like, I feel San Francisco perhaps doesn't, doesn't culturally encourage one to become Herb Boyer. Or, or yesterday, at the time of recording this podcast, Patrick Su, one of the co-founders of ARC, which, and maybe we'll speak about, you know, later in the show, this is a biomedical research organization we, we started a few years ago. He announced these, this, this new phenomenon of bridge editing, which is a new recombinase where you can insert DNA into a, into a genome.

• 3:42Speaker 01

  And, you know, it's pretty early, but it, it might turn out to be, you know, quite consequential. And, you know, in order to do something like that, you have to study for a long time and just acquire a lot of, you know, basic and not so basic technical skills. And so, anyway, the thing, and I don't quite know how to synthesize it yet, but as I think about advice for, for people in their twenties, look, I'm not going to normatively pretend or presume to, to know kind of in, like, in which direction one should go in one's life. And obviously, there are successful examples of basically every strategy. And, you know, I'm really glad you're doing what you're doing. At what age? Twenty-three. Twenty-three. So that's- A podcast isn't. I'm not a genetic recombinant DNA here.

• 4:27Speaker 01

  I think information dissemination is a really valuable thing in the world and causes like the guy who who last I heard was in the in the lead for Nat's Scroll Prize.

• 4:44Dwarkesh Patel?

  Yeah.

• 4:46Speaker 01

  Nash told me learned about the Scroll Prize, you know, listening to your podcast. Right. So I think kind of increasing the catalytic surface area of, you know, certain kinds of information is a valuable thing in the world. So I'm very glad, you know, you're you're putting the podcast. Anyway, I don't presume to know what people should do with their lives, obviously, but but I wonder in as much as I was trying to give advice, and especially maybe if they're reading my advice and not someone else's advice, maybe they're kind of thinking about, you know, career paths that look sort of directionally like mine. I think I think my advice might be, I mean, maybe you should do something like what what I did, or I'm trying to do, but but there are other paths as well. And I think a lot of really important invention in the world and a lot of the things that I'm most happy

• 5:29Speaker 01

  are happening actually require a very different trajectory. And I think there are counter counterfactual versions of my life where, you know, I, you know, pursued that path and well, who knows how well it would have worked.

• 5:40Dwarkesh Patel?

  Hmm.

• 5:42Speaker 01

  And sorry, last point in this. And San Francisco is just very status oriented, I feel in this way. I mean, maybe status oriented is everything is status oriented. So that's kind of tautological, but maybe maybe really what I'm saying is I feel San Francisco,

• 5:56Speaker 01

  the entrepreneurs are are held in excessively high regard, in my view. And look, I guess I like entrepreneurs. And, you know, I think, I think, look, entrepreneurs as an aggregate group in the world, you know, all the all the companies built in Stripe, you know, I think I think are I think are great. But but there's just a strange version of it in San Francisco that I think is.

• 6:23Speaker 01

  Should not be should not be people's only fixation.

• 6:27Dwarkesh Patel?

  Yeah, what I like about this and what I like about you is just like you have this sort of a sense of like contrarianism of like the things people are expecting to hear from you in a given moment, you just like really want to just tell them the opposite of I don't even know like when I feel like when EA was a little bit more popular, you're like, here's the problems, here's why progress is important. And when it was down in its depths, like, hey, guys, pay attention. But okay, on this particular piece of advice.

• 6:49Speaker 01

  Michael Nielsen says that, you know, every field in science has, you know, way too many adherents or way too few. But like, the market is almost never in sort of the right equilibrium. And I think something like that might be, I mean, I think the reflexive contrarian way, I'll say that I think reflexive reflexive contrarianism, for the sake of it, is also tired. And, you know, if you if you if you're just contrarian to the prevailing mood, then, you know, you're you're just following the prevailing mood, but with a sign bit inversion or something. So I don't endorse that either. But but I think that I think the herd is a really powerful phenomenon. Actually, one of the learnings of my adult's life has been that it's, I mean, everyone knows and kind of says or frequently hears

• 7:33Speaker 01

  that you should be very wary of following the prevailing tides and moods and winds and everything. But, you know, it's freaking hard to do in practice.

• 7:41Dwarkesh Patel?

  Yeah. So what practically does that look like to hone your craft in any of these disciplines that take a long time? I mean, you know, you've spoken and tweeted about like, some of the problems with modern universities, is it still is that still the defectopath? Do you want to be the great biologist that archives or something?

• 7:57Speaker 01

  Yeah, I mean, well, you know, in many domains, I don't know, right? So in like hardware, which is not a small domain, most things in the world involve stuff and things. And I just have no facility with or experience with, you know, doing things in hardware. And so if you wanted to become like a super skilled practitioner there, you know, what's the what's the best career path? I don't know. Like, maybe it's to drop out and join, you know, SpaceX or something. I'm not I'm not necessarily endorsing just pursuing the most establishment and credential oriented path. I think people should try to find the

• 8:36Speaker 01

  gradient of maximal learning in whatever it is they care most about. And, and, yeah, the question then is, you know, what that is. For biology, look, not that I'm a biologist, but it is very clear that In order to do really good work, there are a lot of bench skills and, well, there are a lot of bench skills one has to acquire, and then there just is a lot of actual specific knowledge where the body, well, any kind of, any kind of life wasn't designed with neat fundamental principles the way that maybe physics was. A lot of it is obviously evolved

• 9:19Speaker 01

  and contingent and messy and complicated and all the rest. And so there is a lot of just specific factual stuff to learn. And I think for those two reasons, I think there are very few, there are very few successful pure autodidacts in biology where you at some point, in virtually every case that I'm aware of, have to have had direct experience with, in and with a top lab where you're seeing how people actually do it in practice. And actually, maybe this also ties back to what we were, some of what we were discussing previously, where your question about the founders and what they learned from each

• 10:01Speaker 01

  other and so on. I think there's an interesting book, Apprentice to Genius, that follows, I think it's three or four, it's three generations of scientists. So someone who mentored somebody else, who in turn mentored another scientist, and they're all extremely successful. And the book is kind of this, this reflection on, I mean, this description of what they all did, but also this kind of reflection on, well, like, what is it that was transferred? And, for example, one thing it describes is, well, one of the most important and subtle questions in science is problem selection, like, just how do you choose what to work on, right? And no one tells you what to do. And you do have to answer this question

• 10:45Speaker 01

  multiple times. Like, with a company, in some sense, you just have to decide, like, once, and then it's kind of, well, maybe it's an iterative process from there. Whereas in science, you're frequently pursuing, you know, completing new problems. And of course, you need to choose something that's sufficiently important and hard that it would be important if you succeeded, but also that it's not so intractable that you can't actually make any progress. And so the book describes how This is part of what the mentees describe that they learn from their mentors. Another thing they talk about is just learning about high standards and what high standards actually are. And when I talk to people in other domains, this is so frequently the thing that I hear

• 11:30Speaker 01

  from them, that when they worked with X person, or Y organization, or in Z environment, or whatever, that they learned what great actually is, and that just permanently changed their sense for what their own standard for their work ought to be. And so, and so maybe one version of what people in their twenties should do is, is, you know, get some ideas to domains you're interested in or care about, but then figure out where can you learn the highest standards? Where are the highest standards embodied? And where can you go and experience that firsthand?

• 12:11Dwarkesh Patel?

  Before we get back to Stripe and Arc Institute and everything, I want to touch on the Parker Studies stuff for a second. So there's a view that says, listen, if we improve the NIH 10% or whatever percent, are we really making a dent in the fact that ideas are getting harder to find over time? And how much of a difference do institutions make anyways? Just like if it's just about a number of researchers and how many people in your society you can put into research? It's not like Singapore can have a much more effective scientific institution that lets it compete with America in science or something like that. What's wrong with that intuition?

• 12:42Speaker 01

  Noah Smith and others have talked about, I can't remember the term he used, something like money-ism. He had a funny phrase. But sort of this idea that we assume there is some kind of constant elasticity between investment in some particular outcome, like building a semiconductor factory in Arizona, or a new bridge, or whatever, and the outcome of the factory or the bridge. And one, the conversion rate between those inputs and the output is not a...

• 13:21Speaker 01

  Cosmological constant, like maybe any of these things could be done for a half or a tenth or, you know, whatever of the cost. But two, there's, you know, there are even deeper questions as to like, you know, is it possible at all? Or, you know, what else would have to change for it to be possible? And what are the other constraints? Like by kind of just talking about these things in funding and dollar terms, you're kind of making the implicit assumption that the only relevant constraint is the financial one, where in practice, maybe it's permits or it's, you know, labor shortages or, you know, it's, it's, it's other things. Anyway, in the context of the NIH and science and R and D, I'm really skeptical of, of this, kind of this, this same approach being brought to bear where, you know, we can, where

• 14:05Speaker 01

  we can, we can just talk about the amount that we're spending on R and D and think that that's implicitly a useful measure of the, of the output. And, you know, and to a fairly close approximation, there were around 1% as many practicing professional scientists in the US pre-World War II as post-World War II. And, or say even 1950, and, you know, the other kind of epi-phenomena, you know, in papers or patents and so forth, you know, it tends to, you know, follow pretty similar ratios. And, you know, we, we got a lot of like pretty good stuff in the, in the, in the first half

• 14:48Speaker 01

  of the century. And, you know, despite increasing the amount that we spend by, you know, between two and maybe, maybe, I mean, slightly more than two orders of magnitude, not quite three, it's, it's just, it's not clear to me that there is like a direct linear relationship. And so, you know, when, when analyzing the NIH or how we should pursue any of this stuff, I'm inclined to try to get way more, I guess, concrete and tactile and try to think, okay, like, what would success here look like at, well, what is happening today at the micro scale and what are the, like, what are the actual problems? And then what could success look like at the micro scale?

• 15:33Speaker 01

  And then what might it look like to scale that up? And just to give one kind of pointy example of that, we ran a survey of the FAST grants, grant recipients after FAST grants, asking about their, their normal work and not, not about anything to do with FAST grants itself. And we asked them if they had flexible funding, that is to say, if they could spend their research dollars, their, their current research dollars, however they wanted. So not if they like had more research dollars, just if they could spend, if they could direct their current dollars, however they wanted, how much their associated research program would change? And we give them three options, not much, a little and a lot. Seventy-nine percent said a lot. So four out of five said that their research agenda would change a lot if, if, if this constraint was removed.

• 16:18Speaker 01

  And so, like, should the NIH funding level be, you know, X or 1.1X or 1.2X or whatever? That seems to me like a bad way to analyze this question as compared to, you know, for example, perhaps, how bound and constrained should an NIH grantee be in choosing their, their research agenda? You know, maybe, I mean, if, if their judgment was way better than, than that of the committees, not saying it is, but maybe it is, who knows? And maybe there's a 5X improvement, you know, to be generated just by making that one, one switch. So, so yeah, I'm just, I'm very skeptical of these, of these financially oriented frameworks.

• 16:54Dwarkesh Patel?

  Hmm. I mean, maybe the financial is not the right word for it, but just like trying to map inputs to outputs is the framing which you're using to compare the pre-World War II

• 17:03Speaker 01

  Yes.

• 17:04Dwarkesh Patel?

  inputs to what's happening now. And if it was particular to the scientific institutions, you'd expect, for example, that things that are disconnected from like the NIH specific structures.

• 17:14Speaker 01

  Yes.

• 17:15Dwarkesh Patel?

  I mean, obviously, you, you've talked a lot about the, it is getting harder to find paper.

• 17:19Speaker 01

  Right.

• 17:20Dwarkesh Patel?

  And, you know, you know, like, sector through sector, it's not like NIH is funding more. A lot of progress, right? But even there, you see you need exponentially more researchers to keep up the same level of progress. So it does seem important to have these level effects that are one time in the case of something like COVID, where like, yeah, we need that level effect right now. But when if we're framing it in terms of like hundreds of years from now, we're going to be, this is going to be the thing that increases growth rates. Which is a sort of framing that is also supplied when talking about these progress, these things. Does that make sense in that context when like all these sectors are seeing these slowdowns, which seem consistent with just like, yeah, this is how the economy and science progresses over time?

• 17:57Speaker 01

  I don't know is the short answer. I think it's really puzzling. I think the, yeah, like the the constancy of of US GDP growth is, I think, just like one of the weirdest things. I don't know if we've got explanation for it. But but also, I don't think that it's, or sorry, an obvious thing to do would be to shrug and say, okay, well, just it's it's it's over determined or something. And that's just like how countries work. But you can look at other countries where it's obviously manifestly not the case. And so, you know, what is it that's weird and special about the US? The thing that I wonder, right, in a lot of these cases is you could get many of the observed system phenomenon characteristics if we like, if we weren't actually adding productive capacity.

• 18:40Speaker 01

  Like, that's a simple way to explain a lot of it, in that if you're just adding exponentially more unproductive capacity, then then on a stylized level, a lot of this stuff would just fall out of it. Now, I'm not saying that we're necessarily doing that.

• 18:56Speaker 01

  But it could be that, you know, maybe we're maybe we're making them.

• 19:02Speaker 01

  Well, there's lots of ways where that could be the that could be what's effectively going on. Even if it's not the case that like the marginal people or things or organizations themselves are bad, it's just kind of somehow how the how the components interact. But the the fact that you could get, yeah, these these exponentially diminishing returns through the addition of ever more nonproductive capacity makes me makes me not persuaded that the low hanging case is necessarily true. And give some weight to the prospect that, yeah, it is, it's fundamentally structural, cultural or organizational. And, you know, just to give a

• 19:46Speaker 01

  Sort of a micro example there and it's a it's it's a very basic and an obvious one, but you know, I think it's interesting to compare the SpaceX R and D budget and the NASA R and D budget and and like to actually look at those two time series together. And you know, it's, we were just returning to the kind of the financial point again, but I it it seems pretty clear that as NASA has that the trajectory of NASA's efficacy has has not fully followed the trajectory of its inputs.

• 20:29Dwarkesh Patel?

  Yeah, yeah. Although the the point about the marginal inputs we put into science have not been as highly effectively used or as high quality as what was before, like the 1X is a much higher quality 1X than 100X. That it's not clear what you do to fix that. Like if it's just a case of there's like a limited amount of John von Neumanns in your society that are like part of the pre-World War II 1X, it's you know, it's not like we can just put 100X more John von Neumann type physicists into into science.

• 20:56Speaker 01

  If the binding constraint is, yeah, the number of John von Neumanns, then yes, that's I mean, that's that's bad news, I guess. Like, there's not a lot we can kind of do on the margin. But I'm not sure that it is. Like, I think the I guess I keep going back to the the cultural and the sociological point where so Gertie and Carl Corey, they ran a lab at the University of Washington, St. Louis, and six of their students, if I recall correctly, went on to win Nobel prizes. And they had a well-known lab and they got like good students, but they weren't the most prestigious lab in the world. It's not like they got to, you know, cherry pick every year, like the single most promising person. And so, you know, something was going on there. And there's a book about it and

• 21:41Speaker 01

  it tries to get into this a little bit. And I don't know that I can figure out quite what it was. And there was also some good fortune where they got into molecular biology at a good time. But but I think there are these kind of hopeful data points where, you know, again, they were obviously extremely brilliant. People, but I think that the thing that distinguished them and their students was not that they were these, you know, seven sigma Martians.

• 22:05Speaker 01

  I think rather that they they found organizational structures that and cultural practices that really worked. I think those are at least in principle more replicable. Now, you might still say, OK, fine, in theory, but you know, how do you actually do that? And I think that's, I think that's the big open question.

• 22:21Dwarkesh Patel?

  OK, I think that's a great, that's a great point to talk about our institute. So, yeah, I mean, I think you just kind of answered this question, basically, but it's not exactly like biology research is, it's something that society has neglected. So what's the theory of change here? Is it just such a story similar to Stripe in that if you get the right people, even though, you know, there's like tens of billions of dollars of biology funding, getting the right people and the right culture and right dedication is what it takes?

• 22:45Speaker 01

  Even though there are lots of scientists and lots of, lots of universities, and there's a lot of homogeneity today in, in how science and in particular, how biomedical science is pursued, where and kind of basic, you know, research in an academic context, you know, before there's any commercialization or prospect of it in sight. And, and I don't know that the model is necessarily a bad one. Certainly, you know, we're not particularly claiming that it's a bad one, but sort of the, the construct of universities, labs, P I, a principal investigator running the lab, that person applies for grants primarily to the NIH, maybe supplemented by, you know,

• 23:23Speaker 01

  other, other sources and, and grants reviewed by committees with, with pretty study sections, as they call them, with kind of pretty rigid scoring criteria and so on. Like, that's the structure. And it just seems suboptimal to me. I mean, homogeneity is bad in basically any ecosystem, especially ecosystems where you're,

• 23:49Speaker 01

  you know, where you're, where you're producing, or excuse me, where you're seeking tail outcomes. And we thought that for a variety of reasons, like, well, from first principles, that other models should be possible and that like, we had, we had specific ideas as to how, you know, one particular model might be a Might be a good idea and, you know, complimentary to status quo in very short terms. What's different about ARC is, one, scientists are funded to pursue, scientists are funded themselves to pursue whatever they want. So it's curiosity-driven research, whereas NIH grants are given for projects. Second, we build a lot of in-house infrastructure so that scientists can draw upon other platforms and other capabilities that they don't have to go and, you know, build and maintain themselves.

• 24:33Speaker 01

  Whereas, again, in the kind of standard university economic context, scientists, you know, would virtually always have to do that in-house. And because of the natural skill constraints on any given lab, you know, that effectively circumscribes the ambition of a possible research program.

• 24:50Speaker 01

  And then, thirdly, we try to provide career paths for people to remain in science, you know, if they don't want to become principal investigators, where the university structure kind of commingles the training purpose of academia with the execution, where the people who are doing the work are typically the grad students and the post-docs, who are both themselves, at least nominally, on the career path of themselves eventually becoming principal investigators. And, you know, there are lots of people who, for all sorts of different, very valid reasons, like, love science and love the pursuit of research,

• 25:32Speaker 01

  but don't want to be a manager, running a lab, choosing their own research programs, and dealing with all of the overhead and typically grant applications that are concomitant with that. And so with ARC, we have a real emphasis on hiring scientists who have finished their post-docs, finished grad school, and just like, that's what they want to do in their lives. And there, again, isn't really a career path for them today. And one of the things that's actually really exciting about the discovery that we mentioned that came out yesterday, this new bridge editing technology, is that work was led by one of these senior scientists, who'd finished his post-doc. And, you know, it's not clear to me that he wanted to go and to become a PI, but he loved science and he's He's an amazing researcher, clearly.

• 26:18Speaker 01

  And so he's able to go and to have that have that career at ARC. And, you know, and in addition, the, you know, the the the prospect of these mobile elements being, you know, usable in this way for this like genomic insertion, whatever. That's a pretty speculative, you know, out there thing. And, you know, had he applied to the NIH to go and pursue that? I mean, he didn't, so I don't know what the outcome would have been. But Jennifer Doudna's work was, if I recall correctly, funded by DARPA. Because, you know, her CRISPR NIH applications were rejected. And of course, Carolyn Kuriko's NIH applications, you know, for for mRNA vaccine work were famously rejected.

• 27:03Speaker 01

  So so it at least seems very plausible that it wouldn't have it wouldn't have worked out.

• 27:09Speaker 01

  And and so, look, it's all these things are random. And, you know, I can't make any definitive claims about, you know, the what would have counterfactually happened. But it seems plausible to me that this thing announced yesterday wouldn't have happened or would have been less likely to happen in, you know, in a different environment.

• 27:25Dwarkesh Patel?

  When we think forward 10 years or 20 years, the specific line where you of research, where you understand the effects of genetic architecture on different traits, and also you can edit, invert, insert whatever the DNA arbitrarily, you know, you're salt like a little cell anemia. You've done the obvious things. What does that lead to? What are you excited about?

• 27:50Speaker 01

  Well, the thing that I think is really interesting about it is is using it as a new kind of telescope. By which I mean, you know, when people hear about CRISPR, there's

• 28:02Speaker 01

  there's an obvious excitement and legitimate excitement around, you know, using this to cure things, you know, directly in the body, using it as a kind of therapeutic. But you can also use CRISPR to to try to figure out what's going on in cells and in cell cultures in a in a kind of structured way. And so. The body is interesting in that it has this, you know, this, this switchboard of like, like the DJs, I guess, at the, you know, with those fancy mixing sets, you know, of 20,000 genes. And with CRISPR, you can systematically go and sort of perturb each gene one by one, like mashing all the keys in sequence and try to figure out, well, you know, what the

• 28:45Speaker 01

  effects of perturbing this versus that are. And, you know, if you do that in a cell culture where, you know, you can subject the cells to some stressor or some treatment or, you know, whatever, you can kind of see different perturbations affect different cell outcomes. Or you can just kind of use it for synthetic data generation more broadly, where, you know, you could perform all these perturbations and then sequence and kind of see what's happening in the cells and so forth. And, you know, single cell sequencing has come a long way. Anyway, point is, there's a lot you can do with all this gene editing stuff for, for discovery and for, and for data generation in the broadest sense. And, and, you know, that's really compelling because, you know, for, for a lot of diseases,

• 29:30Speaker 01

  they're complex in, in the sort of fields jargon, meaning, I mean, yes, they're complex in the colloquial sense, but they're specifically complex in that they're not, they're not infectious, like not just like some pathogen getting into you. And they're not monogenic, like, you know, Huntington's, where it's like one specific mutation. Instead, it's like some combination of environmental factors, but like maybe some genetic factors as well. And, you know, it's, it's somewhere in between. And by figuring out, and that includes most autoimmune diseases, most cancers, to some extent cardiovascular disease and neurodegenerative disease, like the big ones we haven't yet solved. And, and You know, so then coming back to these functional genomics technologies, what's interesting,

• 30:15Speaker 01

  I think, is trying to figure out the how it is that the genetic component of those diseases happens and works and so on. And even if that's only a small contributor, it can potentially shine light on just like what the general pathway is. And so the question would be, and look, this is speculative, none of this has actually happened. But by figuring out the genetic interactions between genes and, say, Alzheimer's, can you can you figure out the, you know, how the how Alzheimer's arises, which we don't understand today? And then, once you understand how Alzheimer's arises, maybe you can use kind of conventional technologies and targeting to, you know, figure out how to how to inhibit that or

• 30:55Speaker 01

  to or to sort of, you know, modulate those pathways. And and so, yeah, that's what we're really excited about from a from a functional genomics standpoint. And there's kind of an AI angle as well that, you know, we could talk about if you want.

• 31:09Dwarkesh Patel?

  Hmm. How do you think about the the dual use possibilities of biotech? I am sympathetic with the idea that like, if you think of prior technologies, just like Google search, or even like just the computer itself, right? There's you could like forecast in advance, like, oh, this has all this dual use stuff. But for some reason, just like, history has been kind to us, and maybe we should just the the Chester's meta fence here is like, keep doing science. But with biotech, there's, I mean, we don't have any specifics here, but there's like specific things you can think of with this specific technology, where you could imagine some nefarious things. How do you think about that? Just why not just like focus, let's say, on ameliorating the risks first or something like that?

• 31:49Speaker 01

  Well, I don't think that I don't think the binding constraint on harmful use of biotechnology or bioweapons today is pure biological capabilities. Like, if some set of incredibly capable, intelligent people wanted to want to, you know, cause tremendous harm with with, well, presumably with pathogens, but with something biological, you know, they they wouldn't necessarily to invent anything new, they would just need to apply currently known techniques in in kind of a malevolently directed fashion. And I think there are some concerns and some risks there with respect to with respect to things that

• 32:29Speaker 01

  don't invent new technologies, but do make them more accessible. And so, I mean, I think the sort of, I think the question of is, you know,

• 32:41Speaker 01

  How would the effect on the world be if there was a sufficiently sophisticated LLM that it could help anybody synthesize and disperse smallpox? I don't know that the laws of physics prohibit such an LLM existing. I presume they don't. And would the world be fine if such an LLM was widely distributed? Like, maybe, but maybe not, right? So I think there is that kind of threat factor. But my point is, I don't think knowledge at the frontier of biology is the relevant margin here. And if we take seriously what this is, I mean, we don't need crazy AI risks to motivate this.

• 33:24Speaker 01

  The world is perfectly capable of originating really severe pandemics and pathogens itself, plus all the other diseases that are not pathogenic. So we got other problems. But whether we care about the possible dual-use harms you just mentioned, or we just care about the things that already exist, to ameliorate both of those, we do need enhancement of our capabilities. Like, there are a lot of biological problems that we don't today know how to solve. And so I think in that respect, if one were to do what you're proposing and try to advance the defensive side of this, I don't know that what one would do would necessarily be

• 34:05Speaker 01

  that different, because there are just fundamental capabilities that we would presumably need to have and that we don't today have. And by trying to solve current human diseases, I think you're probably also pursuing something pretty close to the best steps to solve the potential diseases that malicious actors could cause in the future.

• 34:27Dwarkesh Patel?

  That makes sense. I mean, zooming out from bio risk in particular, just like, how are you thinking about AI these days?

• 34:35Speaker 01

  Well, I think everyone has to be sort of high perplexity in the sense that, I mean, the verdict that one might have given at the beginning, you know, we're recording this here, pretty close to the beginning of 2024, the verdict one might have given at the beginning of, you know, 23, 22, 21, you know, back, say, the last eight years, those would all, I think, have looked pretty different. I mean, maybe Gwern might have scored the best from 2019 or something onwards, but broadly speaking, it's been pretty difficult, I think, to forecast. And so I think the basic position to a first order has to be one of, you know,

• 35:17Speaker 01

  some degree of humility. I think, as your blog post identifies, the big question right now is, you know, to what degree, to what degree scaling laws hold. And I guess, if they hold, then, you know, what exactly is it that we're, well, asymptoting is maybe a presumptuous word, but it's not an asymptote, but like, what is it that we're approaching? You know, it's not, we don't necessarily know the shape of that thing, whatever it is. And, yeah, I think, I think it's a lot of, yeah, I think, I think how one should feel needs to be, or ought to be very sensitive to like, the exact parameters of, you know,

• 35:57Speaker 01

  those curves. And I just don't think anyone, anyone knows what, like, the true value of those parameters actually are. So, you know, it's clearly going to be important, is already important today, and it has a pretty central bearing on both Stripe and Arc. And we'll see.

• 36:16Dwarkesh Patel?

  Yeah. I wonder if the meta lesson here, and I totally agree with that sort of general sentiment, but I wonder if the meta lesson that we got from COVID, for example, and with things like FAST grants was. You obviously can't predict these things in advance, but the most important thing, even in addition to these specific sort of countermeasures trying to come up in advance, is like when the thing is happening, having competent individuals who can synthesize and organize information and also having these like new initiatives and institutions to get the right thing done.

• 36:48Speaker 01

  Yes, the adaptability premium is probably going to go way up for the next decade.

• 36:51Dwarkesh Patel?

  Yeah, and with that in mind, and I know you have already a couple of day jobs. But yeah, I feel like something like Fast Grant's, like when the time comes down to it, like, I don't know, you like, it should be like, you know, you'd be like, one of the top people you could think of in terms of having expertise and respect in a wide range of domains and competency as a leader. I don't know, just like keep it in the back of your mind, or maybe in the middle of your mind, given how far we are into the transition.

• 37:15Speaker 01

  Well, Fast Grant's was was was three beloved squirrels in a trench coat, or I guess, well, I was one of the squirrels, so I don't know, I'm self full of it. But it was also Tyler Cohen, who's an amazing person and a great friend, and then my wife, who's also one of Arc's co-founders. And so Fast Grant's was not this like giant impressive that would qualify me for anything at all.

• 37:42Dwarkesh Patel?

  But it doesn't have to be giant, right, to have that kind of bigger impact?

• 37:45Speaker 01

  Um, yeah, I guess as an objective matter, that's true. I mean, look, I, John and I try to be very self-aware of the limits of our expertise, which are, which are, you know, which are very proximate to us. And I'm sure if something like that was necessary, there'd be. I mean, look at Operation Warp Speed. They chose a super effective domain expert, Monsef Slawi, to run that, and it was just like monstrously successful, like truly remarkable. And I don't know who the Monsef Slawi of, I guess it would depend whatever the problem in question is, but I think my recommendation would be like, figure out who Monsef is and like, go hire Monsef. And I

• 38:28Speaker 01

  I think is extremely unlikely. I think anybody who deemed me the monsef of that thing is probably mistaken.

• 38:35Dwarkesh Patel?

  I think you're being too humble. But just seeing on fast grants, now we have the retrospective of how effective the fast grants recipients were compared to the other grants that were given out by, let's say, the NIH or NSF. To your knowledge, what has been the reaction of these institutions to the discrepancy between the speed and effectiveness of fast grants? Have they analyzed their protocols and what happened during COVID? Is there any sort of retrospective there on their part?

• 39:02Speaker 01

  Not to my knowledge, but I don't want that to sound like an indictment. Like, maybe they've done a lot of reflection and I just don't know about it. I don't think I would know about it, even if it had happened. So I don't know. I mean, look, most, well, I don't know anything about the response at CDC or FDA or NIH or NSF or any of the relevant organizations or their international equivalents. And so none of what I'm saying should be taken as specifically, not only not critical of them, but not even a comment to them. I just don't know what they did. But in general, organizations are not awesome

• 39:47Speaker 01

  at self-reflection. I think, I assume as a default prior that some of the dynamics we discussed at the beginning of this are rooted there, where none of the people who started those organizations are there today. And so what exactly are the incentives of those leaders? And, you know, I haven't, yeah, it's not clear to me who would have the incentive to really take stock in a fully objective and self-critical way to figure out what was done well and what was done poorly.

• 40:29Dwarkesh Patel?

  I promise not to be too myopic about AI, but one more question. Long term, we can't forecast, maybe even medium term we can't. But near term, it looks like we might have things that look like AI agents and they might need to trade. What does the financial infrastructure for AI agents look like?

• 40:45Speaker 01

  Yeah, I think that's a really interesting question. And I think

• 40:51Speaker 01

  I think automated or autonomous transactions, I mean, they already exist in to some extent today. I mean, you know, lots of lots of services have usage based billing, right? And a lot of the expenses being incurred are autonomously incurred. Like no, no human is pushing a button when Stripe does most of what it does with cloud computing and incurs, you know, some, some, some class with some cloud service. So it's, it's in some kind of extremely primitive way happening today. And I assume it will follow some gradient where some of those, some of those decisions are either directly or indirectly being made by an LLM or some LLM equivalent or, you know, whatever.

• 41:33Speaker 01

  And I think there'll be some, yeah, some, some almost unnoticed smoothly smooth continuum up to, you know, very considerable degrees of autonomy. But it's not that we're going to like wake up some month and be like, oh my God, you know, suddenly the, the bots have been unleashed. And I think there'll be interesting questions there around, I mean, this, this will now sound very kind of parochial and, and kind of, and maybe getting excessively tactical or something. But I think we're very interesting questions around the legality of those in terms of like, are these, are these treated as the responsibility of the, of the owner? Or is there, is there any degree of kind of independence granted? How does liability work? How, yeah, which, which rails are best suited?

• 42:18Speaker 01

  What kind of transaction velocities we're talking about here? Because if it's, you know, a billion transactions a second, then the properties of that system should look very different to, is it, you know, one giant clearing transaction every day? And again, if we just use the analogy of the, the usage based services, those tend to incur liabilities, you know, in, in, in kind of tiny increments, but then to kind of to settle like on a monthly basis when you pay your bill. So maybe these agent transactions will have, will have that, will have that character. So I think there were a lot, excuse me, a lot of practical applied questions. But, but I think what you're saying around these autonomous transactions conceivably being an important dimension is, is very true and real and is, you know, one of the ways in which

• 43:01Speaker 01

  the, the, one of the interesting ways in which the economy, you know, might change and expand over the next decade. And, and I think it's possible that, that crypto plays some role here where, you know, it's,

• 43:15Speaker 01

  if. Yeah, we we take KYC and AML very seriously for humans and we want to know the human that is associated with some particular financial activity. Obviously, that's a murkier question in the context of some AI agent. And if we, in some blurry sense, look at crypto as the part of financial services that is de facto exempt from AML by design, then yeah, maybe that plays a role.

• 43:45Dwarkesh Patel?

  How long before Stripe was founded do you think a product like Stripe could have been invented?

• 43:51Speaker 01

  Hmm, that's a good question.

• 43:54Speaker 01

  Well, in, you know, depending on what exactly you define Stripe as being, I think conceivably decades earlier, in that, I mean, on some level PayPal is a kind of Stripe. And there are many payments companies before PayPal. And you could go all the way back to cash registers or something, right? So it depends on, yeah, these definitional questions. I mean, the particular secular tailwinds that we benefited from around the rise of app stores and the on-demand economy and maybe the startup boom post YC and after the financial crisis, those particular tailwinds were idiosyncratic and specific to Stripe.

• 44:38Speaker 01

  And I guess the GFC was away to nine and Stripe was founded in 2010. And so, you know, in as much as you define those as being core, then not that much earlier. But mostly my story of Stripe is one of market inefficiency. And I do wonder, yeah, why much of this didn't happen sooner.

• 44:57Dwarkesh Patel?

  Yeah, I always find it really interesting when there's these cases where it wasn't even the case that like, well, it could have been started sooner, but there was nobody in the market. There were like many people in the market and they weren't just like random people. There were technology companies headquartered in San Francisco who were in the market.

• 45:12Dwarkesh Patel?

  Do you have some explanation for why it didn't occur to them?

• 45:15Speaker 01

  I'm hesitant to generalize too much because, well, I only have maybe n equals one experience, and so I think it's dangerous to over extrapolate from that. Maybe n equals two now with ARC as a very different kind of organization, but an organization nonetheless.

• 45:30Dwarkesh Patel?

  Or if you include all the features of Stripe, n equals like 10, 20 something.

• 45:33Speaker 01

  Okay, so yes, depending on your definition, maybe there's some kind of samples out there. I guess my general view is most products and most businesses.

• 45:43Speaker 01

  Things can just be done much better. And I think moats are typically kind of overrated. And I mean, the payments are a great example of a domain where on a logical basis, you would say that there are so many sources of defensibility, where there's the network effects of the account holders, and there's the data network effect slash economy of scale for fraud and so forth, and there are regulatory moats and barriers, and, and, and. And yet, not only does Stripe exist, but there are lots of other, I mean, there's a whole fintech ecosystem today, right? So, yeah, I think it gets down to kind of deep questions of what are the, what's the binding constraint on just the number of effective organizations that exist in the

• 46:28Speaker 01

  world? And for any given sector, why is it that number of companies, rather than twice that number of companies, and so on? I think it's about motivation and ideas and people's willingness and determination to organize talent and so forth. But these kinds of more sociocultural explanations, rather than, I mean, Hamilton-Hellmer is probably the leading scholar of some of the sources of defensibility for businesses. He has this niche, but very well-known in the niche book called Seven Powers. And it kind of attempts to disaggregate all the various sources of market power in this respect.

• 47:10Speaker 01

  And I think that is true and important in so far as it goes. But nonetheless, it's kind of strange to me that nobody had done Stripe before Stripe.

• 47:19Dwarkesh Patel?

  When you think about the fact that moats are overrated and just like doing the thing is underrated, does that, like, what is Stripe's mode in that context? Does that make you think differently about Stripe's mode?

• 47:32Speaker 01

  Yes, one. And I guess I do think that one can have

• 47:45Speaker 01

  Organizational and cultural moats. And maybe this contradicts what I was just saying, or maybe it's consistent with it in the sense that it's a kind of cultural explanation. And I think that inasmuch as we have a moat, it's because we have a very good understanding of our domain and a set of people who actually care about solving the problems and who are continually paranoid at the prospect that we might be forgetting something important instead of trying to figure out what the important thing that could supplant Stripe's approaches is and making sure that we build those first and so forth. I think organizations that are, I mean, there's, you're familiar with conquest's laws and

• 48:27Speaker 01

  there's conquest's third law, I guess, which is that one should model organizations as if they're run by a cabal of their enemies. And, you know, obviously it's, or presumably it's tongue in cheek, but, you know, it's interesting to try to think, I'd like, well, kind of, what is the kernel of truth in that? And why would it be there? And I think what's going on is that, I think, most organizations, when they start out, are actually trying to achieve their stated goals. Like, somebody started the organization for a reason, and probably it was for the stated reason. But then, over time, you know, that person and that set of people who initially populate the organization depart, and some set of new people come to, you know, take their place. And there's multiple versions of that, this generational turnover on a continuous basis.

• 49:12Speaker 01

  But say for the fifth generation, like, why are they there? And to what degree do kind of their particular, specific, local incentives align with the nominal, originally stated goals of the organization? And I think there can be a lot of misalignment there, right? Where they're following a local path and conceivably even the leader of the organization, not even through any, like, fault of their own, per se, necessarily, just that they're human and they have their own incentives. And again, the original kind of constitutional incentives of the organization might be quite different. And so, I think this phenomenon is kind of a fact of life, and I think these kinds of explanations, for me, are

• 49:54Speaker 01

  Are much more explanatory in trying to figure out sort of why some of these things either happen or don't. And to your question, like in as much as Stripe has a moat, what is it? I think it's that, I mean, others can judge to what degree it's actually manifested and rooted in practice. I think it is, but you know, I'm a biased observer, but I think it would be that people at Stripe really care about solving the problems that we say we are trying to solve.

• 50:19Dwarkesh Patel?

  Hmm. Yeah, the, the, the point about the misalignment over generations or over time is interesting. And actually, do you have examples of institutions which have for decades or hundreds of years managed to keep their original, not only mission statement, but the organizational competence? Because you think of tech companies, even the oldest tech companies have not been around that long, right? And they're some of the biggest tech companies in the world and the median age of the corporations like famously low. What is a good examples here?

• 50:48Speaker 01

  I think some of the explanations around the, the effects of shareholder capitalism and sort of the idea that shareholder capitalism as a mechanism does, in fact, have, have, you know, has some consequence with respect to the incentives of organizations and their long-term fates. I think those theories have some credibility and I think it is very plausible that shareholder capitalism even attenuates the, the, the duration of some of these organizations. I'm not saying that's definitively true, but I find it incredible, the, the idea that it is. It's not clear to me that that's necessarily bad, even if it is true, right? In that, are we on the side of the humans or of the kind of aggregate innovation in the world? Or on the side of like the corporations, you know, qua legal entities?

• 51:33Speaker 01

  And, you know, yeah, it's not clear to me the answer, there should be the third. At the same time, or maybe in fact consistent with that, you know, if you look at, say, Europe or, you know, some other places like, you know, in Denmark.

• 51:48Speaker 01

  There's for reasons related to the tax code there, a lot of organizations are either controlled by or very substantially held by nonprofit foundations. And so Novo Nordisk, for example, the GLP company, but Maersk, the shipping company, I believe also Lego, a lot of these corporations are controlled by, and again, usually have a lot of their stock held by foundations. That has the secondary effect in many cases where they actually do embed in a legally binding constitution their mission. And so I'm not an expert on Novo Nordisk, but I happened to get a book about it over Thanksgiving. And actually,

• 52:33Speaker 01

  there's also a book on the Danish industrial foundations. But it's enshrined in their constitution that they have to make insulin broadly available really cheaply, or at least cheaply in Scandinavian countries. And then I think they're allowed to charge market prices elsewhere. And I think that, and then the rest of their profits, they have to, they're again legally obligated to reinvest in R and D. Is that somehow causal in the fact that they then invented one of the most remarkable pharmacological discoveries of the last 20 years in these GLP-1 agonists? I mean, plausibly. And so I think, yeah,

• 53:14Speaker 01

  I think these questions around the why it is that the median age of organizations and corporations is what it is are definitely interesting. And I suspect it's a somewhat contingent aspect of how we've chosen to organize large corporations in the US today.

• 53:31Dwarkesh Patel?

  The thing you're mentioning about this firm seems very similar to the export-led growth in Asian.

• 53:36Speaker 01

  Totally.

• 53:37Dwarkesh Patel?

  Right? You have tariffs. This one company, your task is making the cars, so you better make the cars good. You have no competition, but you had to invent the best car in the world.

• 53:44Speaker 01

  Yes, yes, yes. And I think, I mean, You know, we are all fans of Smith and Ricardo and all these characters. And even they, I think, are sort of less dogmatically attached to free trade than perhaps people today interpret them as being. But I think people like Friedrich List and those other, not quite contemporaries, but quasi contemporaries are maybe on a relative basis underrated. And I do think, I mean, in as much as you believe the kind of sociological, cultural skill, whatever, even vague alignment, not in the AI sense, but in the more interpersonal sense, in as much as you think these are important and explanatory, then yeah, I think

• 54:28Speaker 01

  you end up thinking about some of the things you just raised.

• 54:31Dwarkesh Patel?

  That's really interesting to hear you say that because if you think about Stripe's mission, right, it's to facilitate global trade, to make sure that some firm from India can compete with any firm in Nigeria, whatever. So like the room for you to have this sort of learning curve where you're less efficient than the global competition should be less if Stripe exists, right?

• 54:50Speaker 01

  Yes.

• 54:51Dwarkesh Patel?

  Isn't Stripe the anti-List company?

• 54:53Speaker 01

  Well, it depends which version of List. And I mean, and to be clear, I'm not specifically endorsing these tariffs and trade barriers. I think the history associated with them is checkered at best. Look, I think it's possible that if you have a like specific sector where you have, where you have, you know, clear goals and a credible path to actually achieving some substantial degree of success there, and then probably some more kind of conjoined propositions, then maybe some degree of activist trade policy might be on net the beneficial thing to do.

• 55:36Speaker 01

  I don't think that that describes most sectors in most countries at most times. And yeah.

• 55:43Dwarkesh Patel?

  That's so interesting. I'm trying to, I think there's an interesting thread here and how it relates to Stripe Climate in that you're, I don't know, like subsidizing these learning curves that these East Asian countries did for their own internal companies. I mean, you haven't picked out like a specific company that's going to necessarily be the key of carbon sequestration, but yeah, how do you think about this?

• 56:04Speaker 01

  Well, maybe a way to unify the two points, and I'll speak about Stripe Climate in a second, is that I think, I guess it's Say's Law about, you know, demand creating supply. And inasmuch as Stripe aggregates more and more global demand, I guess part of the, I don't know, it seems too self-aggrandizing to call it, you know, the theory of Stripe, but you know, some vague hunch in Stripe is that that aggregation of demand can have important expansionary effects with respect to the ensuing supply. And yes, Stripe Climate is some version of this, you know, hypothesis is applied on a much smaller scale than Stripe itself, but still, you know, real and well, we'll see, maybe important. And the basic idea, just for folks who aren't familiar, which I assume is most of your audience.

• 56:49Speaker 01

  So we observed in 2018, I guess, that everyone seems to agree that carbon removal will be very important. And, you know, even if we decarbonize the economy on the kind of time scale that optimistic people, you know, on the most optimistic timeframes, there'll still be, you know, an accumulated stock of carbon that is a problem. It sounded pretty weird. There were virtually no carbon removal companies in the world in 2018, maybe there were two or three or something. No companies had ever purchased from a carbon removal company. These were really sort of science projects. And so we thought, well, you know, somebody's got to start and it might be valuable to, you know, not only transfer some dollars, but to kind of confer some credibility on this sector. Not that Stripe is the world's most credible company, but you know, it's better than nothing.

• 57:34Speaker 01

  And so we started contracting with some of these carbon removal companies. That went pretty well and they seemed kind of appreciative of us. And so we thought somewhat more about this and we then in 2021 formed Frontier, which is an AMC, an advanced market commitment. So inspired by The first AMC, which was a pre-commitment to purchase vaccines for developing world's countries for diseases that, I mean, well, either were kind of market failures where pharma companies hadn't pursued the vaccines, or where just like the profits weren't sufficient to pay for the program. So we took this for carbon removal. We raised a billion dollars. Stripe was the first investor. We're not actually investing, we're just buying. So

• 58:18Speaker 01

  they're the first company to commit. But then we're joined by Shopify, and Alphabet, and Meta, and JP Morgan, and a bunch of other companies. And now there's like a fairly active sector of carbon removal companies. There are, I think Frontier has contracted with forty, between forty and fifty companies, the overwhelming majority of which didn't exist when we started out with this. Actually, we ran an anonymous survey back the end of last year, and we asked them to what degree was the existence of Frontier somewhat causal in their starting the company in the first place. Again, this is an anonymous survey. And I think it was seventy, seventy-four percent of the companies said that Frontier

• 59:02Speaker 01

  played a causal role in their starting the company. So, yeah, I think these inducement effects can be pretty significant.

• 59:10Dwarkesh Patel?

  Yeah, that's huge. Well, what other ideas you've come across where an AMC would be an effective instrument of moving forward, the tech?

• 59:16Speaker 01

  Hmm, that's a good question. We've actually been having some of that discussion internally. Just, it's not that we plan on doing it ourselves necessarily, but just wondering are there, are there people we should share our technology with? Not that it's even technology per se, but share our experience with or something and try to help along. I mean, there's still, I think there's still a lot of stuff in the biomedical field. And, I mean, patents are...

• 59:49Speaker 01

  You know, patents are pretty useful insofar as they go, but you know, there's a lot of there's a lot of innovation that seems like it would be socially beneficial that patents don't provide a way to cover the cost of. And so, you know, there was some excitement a few years ago about Manos, which is a, it's a sugar. And there was one paper that, or maybe a few papers, I can't remember, that suggests that maybe tumors will selectively take up Manos rather than glucose, but they can't actually metabolize it properly and then they just die. And so, you know, maybe this could be an effective, you know, onco

• 1:00:33Speaker 01

  treatment of some sort. But Manos is like, it's a generic sugar. It's been, you know, understood for, I guess, more than a century, and you couldn't patent it, importantly. And so, it's not clear who has the incentive to even, you know, fund the work to test, you know, whether or not this would actually work in practice. And, you know, this is not an endorsement of Manos, but just, there are things of this shape, where there's something where you can clearly see, wow, that that might be very beneficial, but it's not totally clear how the kind of economic structure of the market can make it possible. So, I think there are still a lot of those across the biomedical landscape. I mean, look, there are still a lot of vaccines that, you know, could in principle exist that don't. I mean, Lyme disease, you know, there's no, there's one vaccine that was withdrawn from the market

• 1:01:17Speaker 01

  over some safety concerns that I think were misplaced, but, you know, it's still no vaccine.

• 1:01:21Dwarkesh Patel?

  It's not even that well understood, right? Like, people have chronic Lyme disease, we don't know, like, if it's legit or not.

• 1:01:28Speaker 01

  Exactly. Yes, yes, yes. And, but it's a good question. I mean, look, maybe some of your listeners will know, yeah, will have ideas for fields for, you know, we sorely need an AMC.

• 1:01:38Dwarkesh Patel?

  Hmm. Yeah. I want to go back to Stripe for a second. So, you're famously, you know, appreciative of craft and beauty, but also you appreciate the power of scale and growth. Is there a type of-

• 1:01:51Speaker 01

  And speed.

• 1:01:52Dwarkesh Patel?

  Oh, interesting. Okay, yeah. But is there a type of craft that is just not amenable to speed, growth, scale? If you think like a Japanese chef, he's like learning to cook rice for, you know, a decade, and then he can move on to the sushi or something. Is that just not competitive in the modern world?

• 1:02:08Speaker 01

  Craft, scale and, and speed, I don't know if they're strictly necessarily intention in every case, but they're definitely frequently intentioned. So just yes, I think is kind of one short answer to that. At the same time, a lot of the most successful companies are those that I think are distinguished by the extent to which they, they exhibit appreciation for and like skill in realizing craft and, and beauty. And so LVMH, some of the largest companies in the world, and like that's literally their business. I think Tesla is pretty good at this. I mean, they're good at many things, but including this. Obviously,

• 1:02:52Speaker 01

  there's Apple. I mean, TSMC is a kind of, you know, it's not the Japanese sushi chef you, you mentioned, but it is the, it's the TSMC chip sushi chef in, in Taiwan. And so much, again, tacit knowledge and, you know, difficult to, to, to transfer skills. So I think it might be the case that, that craft and the pursuit of it is, is as important as, as it's ever been. And, and certainly as, as Stripe has gotten larger, I think

• 1:03:32Speaker 01

  we ourselves have come to greater conviction in this where, I think part of what's interesting about these aesthetic qualities is they're generally speaking unquantifiable. I don't know if they're intrinsically unquantifiable, like maybe you could like train a model to do so or, or something, but today they're, they're broadly speaking unquantifiable. And yet they are actually, you know, they, they influence people in significant ways. People very demonstrably care about aesthetics and they care about, you know, if they're a company, they care about the aesthetic characteristics of the products that, you know, they produce, just like on an intuitive level. People know that that's true, but, you know, yeah, it's, it's, it's difficult to manage that at, at an organizational level where there

• 1:04:17Speaker 01

  isn't a P and L associated with it. And if you're screwing it up, you, you don't see a neat time series decline. But over the 14 years of Stripe, You know, we have, I guess, through a kind of, not exactly trial and error, but just by studying cases where things worked well at Stripe and cases where things worked less well and what customers responded well to and so on. It really seems clear to us that even in a domain like ours, where we are selling primarily to businesses, that this is something that's truly important. And also that, in as much as, you know, getting back to what we were discussing previously,

• 1:04:54Speaker 01

  you want, in as much as the sociology and the kind of cultural explanations of defensibility are real, the best people, you know, consider themselves craftsperson in their domain. And they really, above almost all else, want to work with the best other people. And so I think it's, it may almost be true that even if, from a customer-facing standpoint, craft was not valued by the market, you actually might still want to build an organization that indexes very heavily on this because you just want the best people for other reasons. And now, as it happens, I think customers do in fact value it. And I think the evidence is broadly consistent with that.

• 1:05:37Speaker 01

  But, but yeah, I think, I think it's very hard to assemble groups of the best people if you don't take the practice of the work super seriously.

• 1:05:45Dwarkesh Patel?

  Hmm. What kind of beauty or craft or simplicity is more important? Interface or implementation? There's famously that essay that Unix is successful because the implementation is simple and not the interface.

• 1:05:56Speaker 01

  Yeah, I guess the interface is kind of simple, but it, there's a lot of like asterisks and caveats and, you know, edge cases that, yeah, I guess Unix doesn't handle for you.

• 1:06:08Dwarkesh Patel?

  But Stripe does, right?

• 1:06:10Speaker 01

  And presumably it depends what you're, what you're building, right? For TikTok, it's probably more important that their interface is simple and even if their implementation's a mess, you know, that's, that's probably okay. Nothing it is, I've no idea. Whereas for Stripe, yeah, people are.

• 1:06:25Speaker 01

  People are on some level purchasing our architecture or purchasing their ability to do certain things and some set of things rather than some different set of things because of what our architecture makes easy and makes possible. Now, I don't think, I mean, if by interface you mean the visual GUI interface, then maybe we can draw some separation there, but I guess we don't really draw that distinction. Like, we think of the interface to Stripe as being the architecture.

• 1:06:51Dwarkesh Patel?

  Right.

• 1:06:52Speaker 01

  We're selling, no one else seems to agree with me, but I often think of Stripe as similar to Mathematica, where we're selling kind of a self-contained universe to model whatever it is of interest to you and that you care about. And we're providing some primitives and some, yes, kind of interfaces and tools and so forth to enable your modeling, but fundamentally, we're helping you do something in your own terms. And in that sense, I don't think the architecture and the interface are necessarily that separable.

• 1:07:31Dwarkesh Patel?

  That's a really interesting analogy. Although, I mean, if you think of Mathematica, the entry that that's giving you to is just like the platonic objects of math.

• 1:07:41Speaker 01

  Right.

• 1:07:42Dwarkesh Patel?

  Whereas for you guys, it's like, the entry is to like Visa error codes, right?

• 1:07:45Speaker 01

  Right, right, right.

• 1:07:46Dwarkesh Patel?

  Like, the end object is not the platonic...

• 1:07:49Speaker 01

  That's true, though, in both cases, yeah, I think, yes. So the analogy falls down in a few respects. But, look, I mean, the idea of a transaction is pretty fundamental and is roughly as old as the quadratic equation or something. I guess the transaction's older. And Mathematica, especially today, excuse me, today now supports all kinds of, I mean, to a very impressive extent, supports all kinds of like crazy arcane stuff. Like, if you go through the more obscure packages in Mathematica, you can definitely find things that are, I think, much less broadly employed and understood even than Visa error codes or something. But, yes, look, these are not the same. It's more just, I find it to be an interesting source of intuition. And I think what...

• 1:08:34Speaker 01

  Wolfram has done with Mathematica is pretty amazing.

• 1:08:36Dwarkesh Patel?

  Yeah. Another way in which I'm curious how you think about this. One way in which Mathematica maybe differs is if they had to make a change in Mathematica, like, big deal, somebody has to learn new syntax. If you make a change, you know, it's like billions of dollars of

• 1:08:51Speaker 01

  Yeah.

• 1:08:52Dwarkesh Patel?

  transactions don't happen.

• 1:08:53Speaker 01

  Right.

• 1:08:54Dwarkesh Patel?

  Like, what, how does that change the way you think about the initial architecture and just the stakes?

• 1:08:59Speaker 01

  Yeah, it's a good question.

• 1:09:03Speaker 01

  Well, actually, first of all, and just beauty with respect to architecture, that I'll answer that one. So just as a side note, I guess, I think it's interesting that API design in general doesn't get more study as a discipline and as a practice. I think it, I think it plays a significant role in the fate of platforms or can. I'm not saying it's always the determinative thing. And if you get it right, you know, there can be compounding positive benefits, you know, and the converse. And I think it's really striking that with, say, with, you know, mobile app development, which was one of the most dynamic and fast-moving ecosystems of the past 10 or 15 years, that, you know, so many of the objects and the classes,

• 1:09:47Speaker 01

  say, in iOS development are prefixed with NS, less so now with Swift, but, you know, for much of the iPhone's history. And, you know, the NS, of course, refers to the next step, you know, back from next in the 90s. But then when you get API design and architecture right, it can be so enduring over literally multiple decades. And, you know, even the face of what are otherwise kind of frenzied evolutions in everything around it. And Unix, of course, is kind of another example of this, where, yes, Unix has tons of shortcomings, but like, the architecture has basically worked for now, you know, more than I guess around half a century.

• 1:10:30Speaker 01

  And so, you know, we talk, we're also trying to impress upon people at Stripe the importance of multi-decadal abstractions. And I think people sometimes respond to that, thinking that that's some insanely lofty, kind of implausibly ambitious, I don't know, hyperbole. But no, I think that's actually just what happens when you, when you get this stuff right. And, and, and if in fact you get a rise, you, you can just reap these, or really the people building on your platform can reap these incredible benefits for, for a very long time. To the Mathematica point, they, I know, take backwards compatibility really seriously,

• 1:11:13Speaker 01

  where you can run programs, you know, written 20 years ago unchanged in today's Mathematica. That really raises the stakes in API design for sort of obvious reasons. And we have that same problem ourselves, where when we think about introducing something new, it's not just, does this like exigently address the particular need that's motivating it today, but, you know, do we think we can stand behind this, you know, in, in 2044? And, and how do we think the world might evolve around us such that it all remains coherent? And we certainly don't always get that right, but that's, that's on some of what we're trying to do.

• 1:11:51Dwarkesh Patel?

  Is Visa an example of this? And one might even say that one of the downsides of being able to use implementation for many decades in the future is even if it's, it's self-sustainable and you have this ecosystem and equilibrium set around it, if you, you know, can't modify it, just because of people's locals and incentives, you get stuck in this equilibrium that's worse than it could be otherwise.

• 1:12:12Speaker 01

  I see, I see. I think the card networks generally, Visa and Mastercard, are

• 1:12:19Speaker 01

  Our pretty good equilibrium. Where it's easy to judge today with the world as it exists in 2024, but I think you have to look at the world as it was when they started out and the particular problems that they're solving. I think when you compare the financial landscape in the US or in the Western world to those in other places, it's certainly not clear to me that the US has gotten a bad hand, so to speak, or is somehow stuck in any meaningful way. So, you know, the card networks do a couple of things. They originally, they were designed to replace a store credit and to, I mean, it was a credit

• 1:13:02Speaker 01

  card originally, not a debit card, right? And that was important and the availability of structured consumer credit, I think, is actually a pretty big deal and pretty beneficial and especially beneficial typically for lower income people. And then with the advent, I guess, of jet travel and mass market tourism and so forth, then they helped supply travelers' checks and various worst alternatives like carrying cash around in your little bag. And then with the internet, they were substantially involved in enabling online transactions. And I think that the fact that they got the architecture so right that so much of this, so many of these different use cases were able to be addressed by their core design is just really impressive.

• 1:13:47Speaker 01

  And like the guy who designed all this, D. Hawk, I think is, I mean, he was a remarkable person. And even people complain about interchange. And lest I sound like a defender of the card ecosystem, I mean, Stripe is on the, well, it depends, you could look at multiple places, but like many people will consider Stripe to be on the wrong side of the interchange cost equation in the sense that we're giving away the interchange revenue to other companies. And so I don't think I'm structurally biased in favor of interchange. And yet, I will say, I think it's pretty interesting what interchange made possible where it's a, it's a

• 1:14:28Speaker 01

  Distribution incentive fee where you're paying other entities to go and like do the work of recruiting these customers and convincing them to get a card and you know getting them to maintain the card and to pay it off at the end of the month and like all this stuff. So you're you're paying for that, just the pure distribution. Like there's a person at the end of the flight telling you, hey, sign up for the United Credit Card. Whatever. But like, you know, that's that's what we're just paying for.

• 1:14:50Dwarkesh Patel?

  That guy annoys me.

• 1:14:53Speaker 01

  We'll we'll get to the counterfactuals in a second. So there's there's that. There's you know, paying for the actual credit issuance itself. And then there's the customer support and all the ancillary things around the dispute handling and so forth.

• 1:15:07Speaker 01

  And then I think it is interesting to look at the cases where, for whatever contingent reason, you know, the card networks didn't arise. So Germany is, you know, one of the the classic ones. And like dealing with the, I mean, from our vantage point at least, dealing with the online economy in Germany as compared to the US is so much worse. Like if if Stripe could push a button and have, you know, really broadly adapted cards in Germany, ala the US, we would like push the hell out of that button, right? You can look at China, which on the one hand does have does have

• 1:15:37Speaker 01

  Alipay and and WePay or, you know, WeChat payments are really ubiquitous. And so in that sense, they're they're very digitally enabled from a transactional standpoint. But those products don't tend to be as sophisticated with with consumer credit. And so, yes, the kind of the the transaction fees for transferring money that you in fact already have are, you know, that's super cheap. But I think you need to look at it on kind of a fully loaded basis where, okay, but what about the cost of actually getting the credit to make the purchase in the first place, you know, as a credit card would enable? And I think as you look at these other counterfactuals in other places, one feels a kind of, you know, gratitude for what it is that D Hock and, you know, and Visa and Mastercard and the card networks made possible. And look, I'm not saying they're perfect or anything, but I think that.

• 1:16:22Speaker 01

  I'm most interested in critiques, and I'm not saying again that one can't make them, but just I'm most interested in critiques from people who've really studied the ecosystems of other countries. Because I think it's easy to underestimate what we got in their invention.

• 1:16:38Dwarkesh Patel?

  Yeah, maybe there's a sort of chess against fence kind of thing going on here. If you had to design payments from first principles now, does it make sense that, you know, all these things you mentioned, taking on credit risk, the chance of fraud, disputed adjudication, should that cost like 2%, 3% of each transaction that happens in the economy? What would payments look like if you had to design that from first principles?

• 1:17:01Speaker 01

  Well, we're seeing a live version of this experiment play out for the first time in many years in a number of countries today, where central banks are becoming more active in designing national payment schemes. And so, PICS in Brazil launched in late 2020, I think, but it's, I mean, I'm sure you've heard of UPI, the central bank. UPI was kind of the instigator here, where it's the central bank payment system in India, and it was tied up with Aadhaar and their national identity system and so on. But that inspired a lot of central bankers in other countries to go and build their own UPIs. And so, yeah, PICS in Brazil launched in 2020, and now a significant majority of all Brazilian

• 1:17:44Speaker 01

  adults are like weekly active users of PICS. Again, even though it launched in 2020, so it just had this incredibly rapid adoption curve. You have Swish in Sweden, you've, you know, there's, I mean, across East Asia, Japan, Thailand, Switzerland, you know, central bank after central bank are deciding, hey, you know, we should have our version of this. And so, this is a kind of reinvention of the payment system from scratch.

• 1:18:12Speaker 01

  For kind of hard to understand reasons, yeah, things typically seem like once you layer in the customer support and the consumer protection and the fraud prevention and the anti-money laundering controls and the credit, you know, the things just for some weird reasons seemed asymptotically around, you know, two or three percent. It's important to also note that a lot of the two or three percent, you know, beyond just covering the costs, much of the surplus ends up getting, getting remitted to consumers in the form of rewards, not in every country, but in many countries. And, and if you look at the, the, the public reports from, you know, various banks in

• 1:18:51Speaker 01

  the US, like their interchange revenue where, you know, they're getting these delicious fees on every transaction, you know, as you put it, like a lot of that is going straight back out the door to the consumers themselves and so on. So it's not clear how exactly, once you think about the economics, like if it's going back to the consumer, should you include that as a transaction tax? Or is it just like a weird sort of, a weird circular relationship? I've not seen any evidence to suggest that the, that the two percent or thereabouts is massively inefficient in the, in the scheme of things.

• 1:19:24Speaker 01

  I'm not saying it's the optimal level, maybe one percent would be better, but you know, within some range of one to three percent, it's, it's probably reasonable.

• 1:19:32Speaker 01

  The, I mean, as, as we think about some of these, these, I don't know, these ad velorem fees and, you know, figures, the, I think the place where there's even more change at the moment that I, that we find ourselves thinking more about is actually the changing structure of global tax, where, you know, the, the idea of, there's been a reasonable amount of innovation, I guess, in the, in the tax domain over the last, you know, century where. Income taxes got pretty high, then we value-added taxes and so on.

• 1:20:09Speaker 01

  Like, the new thing, at least in the online context, is jurisdictions remitting, or, excuse me, imposing sales taxes on businesses that don't have any kind of locus in the jurisdiction question. So, you're a podcaster in the Bay Area, and the Dwarkesh merch store will have to pay, the town of Uppsala in Sweden will have a special tax on baseball caps. And you will need to know about that particular tax on baseball caps. And any baseball caps that you are selling to the Uppsulans, you'll have to collect that

• 1:20:54Speaker 01

  amount from the buyer, report to Uppsala, and then eventually figure out how you're going to get that money to Uppsala. And obviously, it's this combinatoric problem of buyer jurisdictions and product types, and then all the different jurisdictions that you have to remit the money to. And those amounts, we're not talking three basis points. The taxes in question are often 5% or 10% or something. So, it's not trivial. And so, just as I think about the funds flows on the international, like, how all that's evolving and unfolding, I think changes in tax law are actually a much bigger deal than anything about the transactional economics.

• 1:21:37Dwarkesh Patel?

  Yeah, yeah. But, by the way, it's not the Dwarkesh Podcast, it's Lunar Society Podcast LLC. Registered on Strive Atlas. So, any merchandise I sell in the future, it'll, Strive will take care of that.

• 1:21:47Speaker 01

  Yes. Okay, well, if there's ever any Stripe complaints, we...

• 1:21:50Dwarkesh Patel?

  It's great. It's been super useful, honestly. Like, it would've been much more difficult to get business operations going.

• 1:21:55Speaker 01

  Do you think, sorry, I know you're supposed to be interviewing me and other guys, but did Stripe play any, like, even on the margins, counterfactual role in you charging for anything? This is the thing I was interested in. Like, when we talk about growing the GDP of the international, it's not like, get the existing GDP onto our rails. It's sort of, where on the margin can we cause there to be economic activity that isn't already occurring? So, yeah, like, you did, in fact, start the podcast before incorporating, but were we causal in any fashion in the merch? Or anything of that nature.

• 1:22:31Dwarkesh Patel?

  I like to the extent that like Substack would not be like a convenient place to get payments from to begin with, that's like definitely a thing. And also, you know, if I do.

• 1:22:40Speaker 01

  You wouldn't charge for the newsletter if Substack hadn't made it super easy.

• 1:22:43Dwarkesh Patel?

  Yeah, and also if I do like an ad or something, just like I wouldn't even know how to begin with getting the money if I didn't already have an LLC through Stripe that with the associated banking account that I can get the money through. So yeah, probably kind of fracture responsible for a lot of the monetization.

• 1:22:59Speaker 01

  That's cool.

• 1:23:00Dwarkesh Patel?

  Yeah, yeah, yeah. Appreciate it.

• 1:23:05Dwarkesh Patel?

  So what are some unexpected compliments to payment processing you see in the future? So, you know, all this stuff, Atlas, identity fraud detection, you know, in retrospect, it might have not have been obvious back then there was a good compliment. Now it does seem that way. What would be like this in five, 10 years?

• 1:23:26Speaker 01

  Honestly, our problem ends up being that too many things, more things than we can possibly pursue look like compliments, right? In that, you know, every business, almost by definition, has revenue. And so we obviously want to help them generate and accept and manage and orchestrate, you know, everything pertaining to that revenue. But once you're, you know, once you're in that flow and you kind of just go through the steps of running a business, yeah, a lot, a lot else looks relevant and somehow connects quite directly. You know, we're, we're, we're not, I mean, when Stripe started out, Stripe seemed like

• 1:24:10Speaker 01

  it definitely wasn't cool. It was, it was sort of the opposite for just a couple of us and we thought that we could make this the superior payments API. And for the vast majority of its history, Stripe has, I think, attracted people who are drawn to unglamorous infrastructure challenges and problems. And, you know, we're not a company that specializes in, you know, making beautiful cars. We, we make roads. And, and I bring all of that up because I think it's relevant to this kind of compliment question where, you know, in our discussions internally, a lot of it and.

• 1:24:55Speaker 01

  The significant majority of this is still about, okay, where are there actual practical shortcomings and limitations in even our core bread and butter? And that's, I mean, payment processing might be a slightly too limited term to use for it. Maybe it's more about just global programmable money orchestration, which, yes, is consumer to business payments, you know, the sort that we're just discussing in, say, the context of your Substack, but it's also business to business payments. It's also payments where there's credit or lending involved. It's also how you hold money. It's how you convert money between different currencies. It's how you represent money that's held by different legal entities

• 1:25:38Speaker 01

  and how we make it possible for even individuals or small businesses to act as kind of micro multinationals and all this kind of stuff. But those problems that we just skimmed over are all, even though they all directly pertain to the movement of money, they're not small. And if, you know, if we could just solve those really, really effectively, then, you know, Stripe will be a very consequential organization and, I think, force in the world. And, you know, I think the counterfactual importance of building some of this stuff as we go to newer markets that are, on a relative basis, more poorly served is actually increasing rather than shrinking. Like, in the US, there were payments companies before Stripe,

• 1:26:21Speaker 01

  and maybe if Stripe had never done its thing, like, eventually you'd have found some way to, you know, monetize a newsletter or something like that. But if you're in, you know, if you're in Albania, you know, it's, you know, the set of options available to you is far more restricted. And so I think that, yeah, the marginal impact as we, as we expand globally increases quite a bit. So, anyway, that's all to say that even though we are interested in and do today pursue some of these direct adjacencies, I think that the core problem of global sort of money orchestration remains, remains a really, it's just big and unsolved problem.

• 1:27:02Dwarkesh Patel?

  Does that look like being a better interface for all these complexities and glossing them over under the seven lines of code? Or does that look like actually replacing the rails and the infrastructure to make all this more efficient and effective?

• 1:27:15Speaker 01

  The former, the former. Like, it's just not that useful to build financial ecosystems that are self-contained, right? A financial island is not that helpful. It's much more valuable to build, I don't know, a financial, this is mixing metaphors, but, you know, a financial AR network or something. But I think we would much prefer that Stripe plugged into every existing system and rail and domestic organization, rather than that we tried to come along and supplant them. And this has been Stripe's strategy very deliberately from the beginning, where, you

• 1:27:58Speaker 01

  know, there were lots of companies when Stripe started out that were trying to do their own thing and go their own way, whereas our belief was, you get these, I mean, it's classic, I guess, Metcalfe Law stuff of, you know, by enhancing the capabilities of an existing ecosystem, you create quite a bit more value.

• 1:28:15Dwarkesh Patel?

  Okay, let's go back to Stripe. Is Stripe a writing culture for the benefit of the writer or the reader?

• 1:28:21Speaker 01

  It can be both.

• 1:28:23Dwarkesh Patel?

  But which one's the more so?

• 1:28:25Speaker 01

  I think there are actually really considerable benefits on both sides, because for the reader, it's not just that it's maybe more efficient to communicate stuff through text, though in many cases it is, but also there's like this intertemporal benefit where, you know, future readers can try to understand the through line and the thought process that, you know, led us to this point. And I think that's, yeah, very considerable. But it's also true that I think that, I mean, I write things and lots of people write things in order to organize one's own thoughts. And, you know, if that kind of ability was taken away from me, I think I'd be meaningfully less effective. So how exactly those balance out is hard to say. Maybe, maybe the

• 1:29:09Speaker 01

  I mean, they're not actually separable. That's my answer. Like, literate cultures are just a different thing. And I don't mean literate in some kind of faux-intellectual way. I just mean, maybe textual cultures is a better term here. Where, you know, Bruno Latour spoke about how, you know, he thinks part of how the printing revolution, like Gutenberg's, caused the scientific revolution was by making knowledge more rigid. Where before, if some observation didn't match, you know, some claim, you can always kind of shrug and be like, well, I guess the person who transcribed that thing,

• 1:29:54Speaker 01

  you know, just like made a mistake or whatever. And so by making things more rigid, it's easier to break them. And, you know, then you can notice discrepancies between, I guess, the theory or the claim or whatever, and, you know, the actual reality. And I think there's some version that organizationally where, I mean, I'm not drawing like that precise parallel, but there are analogous dynamics where the nature of oral cultures and textual cultures are just quite different. And, you know, so the kinds of collaboration that are possible and the kinds of consistency that can be achieved, like, it is just fundamentally different. And, you know, is it, is the, you know, front or rear wheel of the bicycle more valuable?

• 1:30:39Speaker 01

  And I guess theoretically you can be unicycling, but like, as a practical matter, you do just need both.

• 1:30:43Dwarkesh Patel?

  Hmm. I know I said no more AI questions, but on this particular point, it actually seems very legitimate to me that you might expect firms that have a lot of writing to be the first to experience the productivity gains of AI, because there's all this context that the model doesn't have available readily. I don't know if that's something you anticipate.

• 1:31:02Speaker 01

  I think that's probably true. Yeah, I don't know. And if the model is really good, maybe it's real to pick stuff up quickly. But I think most organizations are not recording all of their meetings for a variety of reasons. And if they're not, then yeah, there is this question of what is the corpus? How do you get up to speed? So yeah, my guess is that'll be true.

• 1:31:21Dwarkesh Patel?

  Tell me about the internal LLM you built.

• 1:31:24Speaker 01

  Oh, it's we didn't build an internal LLM. We built we built an internal LLM tool for making it very easy for people to integrate LLMs into into production services, but also into just into their regular workflows, you know, as as humans. So the ability to work directly, I guess, with the LLM as a standard chat agent, you know, as lots of people have built. But then also to integrate that with some of our tools for, you know, querying and accessing data, or maybe most interestingly with sharing prompts across, you know, different people. And so, you know, somebody might discover, I mean, one of my favorite examples actually

• 1:32:08Speaker 01

  is somebody put together a prompt for optimizing SQL queries. And, you know, it doesn't always work, but but sometimes it does. And like, it's very cheap to ask us, you know, got any ideas for optimizing the SQL query? And, you know, sometimes it'll come up with with some good stuff. And and so, yeah, the collaborative abilities there have proven have proven surprisingly kind of high return. And then having just, I mean, lots of organizations have this. We're not claiming that it's very novel or anything, but having kind of a central bus to through which to route all access to these LLMs such that we can, you know, experiment

• 1:32:52Speaker 01

  with different models and have, you know, some degree of, you know, observability into the respective performance trends and and the the usage of different cases and so forth. Like that we have found building a fairly significant amount of production infrastructure around LLMs to be valuable. And now, given the proliferation of LLMs themselves with all of the obvious contenders, this is proving quite valuable because we're able to try to figure out for different use cases, which models, self-aware models, who knows, are most effective. And I don't know what the total number of invocations is, but I think we're making millions of invocations per day now.

• 1:33:36Speaker 01

  Like, there are just dozens of dozens of actual production use cases across Stripe. And all sorts of really... I mean, the financial services ecosystem is in some way a giant analog to digital exercise because humans are analog and intentions and identities and all these things have... There's always some degree of uncertainty around them and some noise, but then transactions are digital, right? And we often find in these analog to digital conversions that LLMs can be a surprisingly interesting augmenting tool.

• 1:34:17Dwarkesh Patel?

  And actually, on that point about the, I don't know, the flexibility and the edge cases in the way humans interact with these systems. I mean, in some sense Stripe is like a really high-stakes bug bounty program, right? If somebody hacks it, not only the financial services, obviously, like money's in play, but if there's a reliability issue, not just because of a hack, but because you deployed the wrong way, a significant percentage of world GDP would grind to halt, at least while it's down. What... I mean, how do you deal with that kind of responsibility? Like, how do you keep the uptime and keep the reliability while deploying fast?

• 1:34:58Speaker 01

  Yeah, this is one of the things we've spent the most time on. And, I mean, back to this point about wanting to be the place with the best people and if you... and the value of focusing on craft so that you can have the best people. In the context of software development, one of the things that developers really hate is, well, actually, two things that developers hate. Slow development cycles and it'll ship in the next release in a month and that kind of thinking. Developers also hate being paged at 2:00 AM for incidents. And so, yeah, given the criticality of the...

• 1:35:37Speaker 01

  Businesses that we serve, which is in rough terms 1% of the global economy. I mean, that's it's not totally clear how to measure this because we're measuring. We're not measuring like GDP is defined as final goods and Stripe is not only selling final goods. And so like in theory, there could be a bit of double counting, but Stripe is mostly selling final goods. Like we're not used for by and large for giant supply chain shipments. So I think maybe there's a mismeasurement of 10 or 20% or something, but but long story short, I think it works out to about 1% of global GDP. It's about a trillion dollars a year.

• 1:36:20Speaker 01

  As you say, that then makes us like really terrified of outages. And so we work so hard to enable fast iteration and development cycles without having outages. And just to kind of put some numbers on it, we deploy production services that are in kind of the core charge flow around 1,000 times a day. Like most of these services are automatically deployed. So when anybody makes any production ready change, it just like goes into production and it's kind of meticulously and carefully orchestrated so that it first is just running some small sliver of traffic and then incrementally more traffic until it's everything.

• 1:37:03Speaker 01

  So about 1,000 deploys per day at roughly or somewhat in excess of five and a half nines, like 99.995% reliability, which which works out to about, I think about 102, 2 and a half minutes of unavailability per year. It's not that we have obviously two and a half contiguous minutes of unavailability, but that's what I kind of that's what it approximates, even though it tends to happen as kind of background radiation throughout the year. And getting to that point, yeah, just takes a huge amount of investment in and in those security properties that are less readily measured, but analogous to those figures. And

• 1:37:49Speaker 01

  I guess Silicon Valley doesn't tend to, and I'm perhaps being now unfair and kind of attributing things to Silicon Valley, but maybe a lot of the tech industry doesn't place a lot of value on process and operational excellence. You know, we kind of culturally value the spontaneous and the creative and the icon of plastic and the path breaking. But building mechanisms that can enable really reliable, enable the very reliable provision of important services at scale and, you know, removing the, the, yeah, sources of variability that can, you know, really cause a bad day for a very large number of people. I don't

• 1:38:33Speaker 01

  think they get quite as much cultural credit. But yeah, we have found that we've adopted all sorts of, you know, for example, we found that, you know, there's this kind of a core feedback loop around, you know, it isn't, none of this sounds like rocket science, but, you know, defining what it is that we care about and then like building automated measuring systems to obviously measure to what degree it's actually happening in practice. And then to sort of try to figure out, well, in the cases where we're not living up to that, like, what is the reason? And then to, you know, to actually intervene and to improve the system so that, you know, that's not happening. And then importantly, to build kind of secondary controls that detect, you know, instances of deviation long before

• 1:39:18Speaker 01

  they actually cause like a production problem or anything, but just, you know, where we understand the behavior of the system in sufficient detail that, you know, we can instrument it in some upstream way. Most of what I said there, I think was well understood by production engineers in, you know, 1930s. So again, I'm not claiming that it's any kind of radical breakthrough, but we have found that the adoption of these practices and like really Kind of tenacious like multi-year form and is yeah, just yields really high returns. And there may be other organizations that like both ship at that rate and can maintain

• 1:39:56Speaker 01

  that sort of developer velocity at this kind of combination of scale and reliability and security. But I don't think there were that many. And I think it's a real testament to yeah, the remarkable folks at Stripe who made it happen.

• 1:40:13Dwarkesh Patel?

  Last day at point, but like actually the fact that you have this huge internal tooling and testing is like, once you get the AI engineers, they can just like push the commits and you have the infrastructure set up that it can be readily evaluated, you know?

• 1:40:28Speaker 01

  Yeah, across the board, I think so much comes back to what has to be true for us, for us actually to be able to build and to kind of take seriously this goal of building, you know, the best software. And it's easy to say that as some lofty, vague, hand wavy aspirational statement. But if you sort of take that seriously as a goal, and if you think that, well, what would you have to measure if you know, you're actually going to pursue it in earnest? And part of the characteristics of organizations that do produce it, I mean, you get down to, well, you know, customers have to really like your stuff. And so, okay, well, how can we, how can we measure that? And how can we systematize

• 1:41:13Speaker 01

  the process of making sure that there aren't repressions there? And so we have this concept of experience journeys, which are sort of pathways through Stripe that we really care are always implemented at a really high quality level. And it has to be true that developers can iterate very quickly. And we just kind of spoke about how to make that happen, you know, and, and, and. And so I feel like, I mean, maybe a kind of a theme through, through everything we've talked about is actually taking the goal seriously. And I feel like a lot of what we do at Stripe is, again, I disclaim any, any genius in it. I think it's just the very earnest, repeated, serious and Long term application of of of taking the goal seriously.

• 1:41:55Dwarkesh Patel?

  A few more Stripe questions.

• 1:41:58Dwarkesh Patel?

  One percent of GDP, global GDP is like, is such a staggering number. When you think about where further growth for Stripe comes from, does it come from the internet economy expanding? Or does it come from Stripe becoming a larger share of the internet economy? And to the extent that Stripe is growing faster than the internet, if like, we consider that the beta in your case, where is that alpha coming from?

• 1:42:21Speaker 01

  That's a good question.

• 1:42:25Speaker 01

  Well, the the customers that Stripe serves are outgrowing the internet, the internet economy as a whole, like, in aggregate. Now, at some point, those have to converge for kind of obvious mathematical reasons. But but you know, we're 14 years in, they haven't converged yet. So I think there's, I think there's a lot of headroom there. And, you know, say Stripe is hunting around a trillion dollars a year. When Stripe started out, the global economy was 60 to 70 trillion-ish. The global economy is now around 100 trillion. And so, you know, we still have quite a bit of headroom before the like, you know, the the amount of activity that is coming out to Stripe is like really budding up against the ceiling of of global economic growth. And of course, it's not like there's no ceiling on global economic growth.

• 1:43:09Speaker 01

  And, you know, for all sorts of of reasons, you know, it it could be vastly higher than it is. And I don't even mean new technologies or AIs or whatever, but just, yeah, obviously, all the kind of basic per capita math you can do around, you know, what if everybody had a, you know, an income on par with the US? And, you know, I think it is. And like, one of the reasons I I am so interested in working on Stripe is I think it's, you know, it's the the the old line, the the Lucas line, about how when you start thinking about differential rates of development in countries, like, it's hard to think about anything else. You know, why does Brazil have the particular income and GDP level that it does? Why does Poland have the level that it does? Why does, why did Ireland have the trajectory that it did, where we went from being the kind of the sick man of Europe

• 1:43:53Speaker 01

  to to now one of the wealthiest countries there? And I feel like Stripe is some applied version of this question in practice, where you're you're kind of building software products, but in some sense, connected to or, you know, touching upon these questions of, well, why aren't there more country? Excuse me, why aren't there more? Why aren't there more companies? And what determines the growth rate of a company? Like, why, you know, when?

• 1:44:20Speaker 01

  When you start the merch store, like, why does it have, you know, X level of buyers rather than, you know, 2X? And I actually think those questions, you know, I think those remain fruitful questions. We actually haven't optimized the meta system of business to any particularly great extent. For the vast majority of business, businesses have been offline, inefficient, you know, analog, everything. And it's really only over the last, you know, two, one to two decades that a significant share of this has been meaningfully digitized.

• 1:45:01Speaker 01

  And the prospects for efficiency gains and optimizations there are still pretty significantly underexplored. And we find incredibly basic things, like just, you know, just extending capital to businesses.

• 1:45:14Dwarkesh Patel?

  Right.

• 1:45:15Speaker 01

  I mean, the reason we do that is not to generate profit from the loans, but because we find that the businesses whom we extend the capital then just grow faster on a, on a sort of persistent, subsequent basis. Or, you know, trying to figure out, how, how does a business decide which countries it sells in? And you'll find even for the smallest business through to some of the largest businesses in the world, that these are very kind of ad hoc and not particularly deeply thought through questions, like, you know, why don't you sell in Mexico and Brazil or wherever? It's like, well, it seemed kind of complicated and we didn't quite get around to it and so forth. And so I think there's, to your question about, like, where, where does the growth come from?

• 1:45:59Speaker 01

  I think that there's still an awful lot of low hanging fruit in just asking some of these incredibly basic questions.

• 1:46:05Dwarkesh Patel?

  So, when we think about the, the way in which Stripe will continue to grow in the future, in some sense, it'll obviously involve a lot of big businesses and, you know, you've now processing a significant amount of Amazon volume, there's other businesses you're doing deals with. First, tell me how you think, it kind of makes sense how an exponentially growing startup would contribute to exponentially growing growth for Stripe? How does, like, does Stripe keep growing in the same trajectory when it's these existing big businesses that you're partnering with? And the second, like, also like the case for why these startups matter is like so compelling, right? Like a new thing is coming into this world, then we should really support it and make sure it happens. Why is it compelling that like Amazon can fulfill orders more efficiently or something?

• 1:46:50Speaker 01

  Yeah, those are very good questions. So on the first one, you're right. It's Stripe is doomed to eventually grow at the rate of the economy.

• 1:47:12Speaker 01

  There is just a question of how long it takes to get there, right? Now, the good news is I think it can be a very long time because there is, as we just discussed, there is so much low-hanging fruit around different optimizations and improvements that are possible. And so I think it would be many decades before that happens, but it's true. That will eventually occur. On the second question about, yeah, what's the like? It's obviously virtuous or compelling or exciting to foster all these nascent startups and to kind of be an anti-incumbency force. But what's the case for supporting established businesses?

• 1:47:54Speaker 01

  I think people misunderstand where a small business typically, not in every case, but at least in the cases where we denote them startups, there is usually an embedded innovation and the innovation is kind of all that the company is. Like they have a new idea and they're going to do something better or different or whatever. And so generally speaking, we like innovation and so we have positive sentiments towards that startup. But there's a lot of innovation that comes from large established businesses. That's not all they do. There's also just running the existing thing. And so maybe it's a smaller share, but the aggregate fraction of innovation that comes from established businesses is really large.

• 1:48:33Speaker 01

  And we have to be cognizant of the cognitive bias of the startups perhaps being somewhat more conspicuous and maybe on a relative basis, the improvements in turbine technology or in fab technology or in insulation technology that come from established businesses.

• 1:48:55Speaker 01

  Or to choose any sector of the economy and a significant fraction of the important inventions that occurred over the last 10 or 20 years will have come from the incumbents. And so I think as a general class, and Tyler, of course, wrote a book on this, I think big business is underrated. And if you look at the survey data, people tend to have very positive sentiments, not only towards startups, but towards small business as a class. Whereas if you, even though they have negative sentiments or relatively negative sentiments towards big business, not that bad on an absolute basis, but not as favorable. I think it's true that established businesses tend to pay better, they tend to be more

• 1:49:37Speaker 01

  efficient, more of the innovation in our economy comes from them, and they produce a lot of consumer surplus. I think the specific case for Stripe working with them is typically they're coming to us not because they want to take the thing that they're already doing and just go to all the work of transposing it to Stripe, but because either they want to do a new thing that they're just not doing today, and so it's associated with some new business line, or some new innovation or invention, or whatever, or they've spotted the opportunity to, I guess, to maybe not produce a new product, but to meaningfully change how they provide an existing one in a fashion that, again, yields consumer surplus.

• 1:50:22Speaker 01

  And that sounds very abstract and theoretical, but in practice, what it tends to mean is they want to take this thing they're selling in this market and sell it in many more markets. Or they've realized that they're selling it in this kind of modality, and they should sell it in other, more convenient ways, like they should sell it on mobile or something. And each of those, if it's successful, if people actually buy it in any significant numbers, I guess we're getting this decentralized signal from the economy that there's now something of value being provided that wasn't here to fore. And of the, as I take stock of the businesses, like the enterprises that...

• 1:50:59Speaker 01

  Are in the process of migrating to Stripe or that did soar the last year, whether it's the large retailers or the large global manufacturing firms or shipping companies, things like this, it typically has one of those two patterns, new product or current product sold to people who weren't buying it before.

• 1:51:21Dwarkesh Patel?

  Yeah, yeah. I mean, if you think about just like the big trends in the society that are needed to solve our big problems, right? Like Moore's Law or the cost of solar or something, these are just, you have marginal improvements over many decades.

• 1:51:30Speaker 01

  Yes.

• 1:51:31Dwarkesh Patel?

  The big tech or big companies are just able to invest a lot of money into doing the R and D.

• 1:51:37Speaker 01

  We're likely to see a bit of improvement, yes, is underrated.

• 1:51:39Dwarkesh Patel?

  Can I ask about John for a second?

• 1:51:41Speaker 01

  Sure.

• 1:51:41Dwarkesh Patel?

  So you guys recently published Poor Charlie's Almanack and subsequently Charlie Munger has passed away.

• 1:51:49Dwarkesh Patel?

  Did Munger ever comment on your relationship? And if or whether it reminded him of his and Buffett's?

• 1:51:56Speaker 01

  Not to me, but he knew John better. And so it's possible that he did to John.

• 1:52:05Speaker 01

  Yeah, I don't know.

• 1:52:07Dwarkesh Patel?

  What have you learned about marriage from John? I mean, this sort of like co-equal, intense, lengthy partnership. It's like the closest thing to that you have is marriage, right?

• 1:52:17Speaker 01

  Well, I'm relatively new to the practice of marriage. So I, you know, it's maybe in a decade I'll be able to kind of extract the generalizable commonalities.

• 1:52:29Speaker 01

  I suppose the general thing I would say is, I think, working with people you're close to is underrated. And I'm doing ARC with Patrick Su and Silvana. Fast Grant was with Tyler and Silvana. Stripe is obviously with John. And actually, John was also, I should mention, instrumentally involved in ARC's formation. Like, it would not have happened without John.

• 1:52:59Speaker 01

  And could give more examples, but I feel like for all the ventures of any significance in my life, they've like not only been with others, but been with other people that I'm very close to and where I had and would like to have an enduring relationship that lives them. And sometimes when here's the advice that you shouldn't work with friends or maybe you shouldn't work with your partner or something like that. And look, I all these things are idiosyncratic and there are there are instances of every possible permutation. But but for me, it's been a really rewarding experience. And yeah, I think John and I

• 1:53:41Speaker 01

  can work together for you never know life, but but you know, I think we'll probably work together for decades. And I for us, it's been a really but both an important source of just meaning and again, fulfillment, but also I think there's a real complementarity. And I think that Stripe would be a less effective company, you know, without either of us. And I just mean from like a bandwidth standpoint or something, but I think we both bring different things to bear.

• 1:54:12Dwarkesh Patel?

  Patrick, I think that's a great place to leave it. Thank you so much for coming on the podcast.

• 1:54:16Speaker 01

  Thank you.

• 1:54:18Dwarkesh Patel?

  Hey, everybody. I hope you enjoyed that episode. As always, the most helpful thing you can do is to share the podcast, send it to people you think might enjoy it, put it in Twitter, your group chats, et cetera. Just blitz the world. Appreciate your listening. I'll see you next time. Cheers.