Patrick Collison: "What If You Succeed?" Transcript, AI Summary & Key Points

00:07 Okay, Patrick, thanks so much for being here. Welcome to Startup School. >> Great to be here. Um, Harge and I first met 20 years ago and um uh he um we started a company together. Sorry, am I giving away the introduction? >> Yeah, I I thought this was my interview, but sure. Keep going. you're doing >> well. We started a company together uh many many years ago and uh I learned a huge amount from Harge.

00:31 Uh so it's it's really fun to do this. >> All right. It's um well actually I mean speaking of that so when I think when I first met you 20 somethingish years ago um at the time your most impressive achievement uh I would argue was Chroma your dialect of lisp. >> Any lisp programmers here? >> Oh wow. Okay. There was um I think I heard one whoop which is more than I expected.

00:52 Um, but yeah, I I really like list when I was in high school. >> Yeah. So, what I was going to ask is um a prolific 16-year-old today could presumably just like prompt claw to write their di their lift dialect. Um, would you would you advise them to not do that and still still do it? Is there is there any value in such things? >> I don't know. I wonder a lot.

01:11 Um, yeah. Yeah, like obviously on the one hand, uh it used to be really fun to write all this assembly and machine code and to optimize your instructions and ma layout and memory and everything and now we don't have to do that anymore. Compilers do it for us. We don't mourn it too much. And so maybe in the same way we shouldn't mourn source code and we should just transcend the plane of uh instructions to Claudius at all.

01:33 But um but emotionally I miss it. Um how about I think just like as I've been hanging out here um uh with these students like the maybe the question behind it is many of them are just wondering what should they be learning at college like what is sort of in this sort of AI world like how much um how much should they be trying to learn and derive from first principles and how much should they just outsource to the to the AI >> right um I mean my model of this is um is cash um you know with a ch not an sh where Jeff Dean

02:12 has this uh famous set of numbers that every programmer should know uh bandwidths and latencies and just kind of relevant constants you try to reason about as you as you build systems and obviously when you're thinking of building any system or distributed system or whatever like all lookups and all you know relevant bandwidths between different um components are are are very different right uh and you know retrieving something from L1 cache is very different to retrieving from RAM is very from retrieving across the

02:38 network or whatever. And I think it's like that with knowledge where fine, yes, you can ask the agent or something to compute something for you or to look something up for you, whatever. That's a hell of a lot slower than knowing it in cognitive L1 cache. And you can have way more round trips in your brain than you can, you know, muttering through, you know, super whisper or typing it out or whatever.

03:00 And so I think even granting the full capabilities of the of the models I still I still think there's a a pretty like I think for for a long time to come uh neuronal lookups will be will be much faster. Um and and then look if you look in revealed preference uh at what uh companies themselves are doing whether they're companies like Stripe or the labs or what have you um there still seems to be an enormous premium on cognitive ability.

03:29 And so I wouldn't I I think um renouncing that before there's evidence that we've saturated uh those benefits would be premature. >> Um I mean are there are there specific things that maybe you personally either personally or as CEO of Stripe um you still you purposefully choose to sort of do yourself and like retrieve from your own cash? Um even though like the agents who probably do a reasonably good job.

03:56 Um I still I still write myself like I I um I I don't I I both philosophically but also uh specifically substantively uh dislike the writing of the models. I mean it's very interesting right because these can prove the Jacobian conjecture you know whatever and so clearly they're capable of these monumental feats. Um, but somehow I still haven't read the LLM essay that I found super compelling.

04:30 Now, maybe it's just very hard to like RL them in that domain because the, you know, the utility function or something is kind of hard to define. Um but yeah um I I think writing is a pretty interpersonal communication and writing I think are still very fundamental and sort of being able to reason sensibly in the multi-dimensional space of reality and in some kind of indescribable way.

04:50 I feel like the model is still kind of deficient at that. And so I've never I've yet to send, you know, every tool is now trying to prompt me with, you know, pre-written uh suggestions, whether it's, you know, Gmail or, uh, apparently WhatsApp just rolled this out. Um, and I think I've still sent zero of those in my life. Cool. Um, how about, so if you talk about the Stripe story, the early days in particular a little bit.

05:18 Uh, you were at MIT, then you left to start Stripe. How did you think about that decision? And obviously we're in a stadium full of college students. How should they think about it? How do you how do they know if it's the right decision for them to uh leave college early and go start a company versus stay? >> Yeah. Well, I I think I have the slightly unusual distinction of having dropped out of college twice to start a company.

05:43 So, um so maybe one thing to know is that it's not totally trapdoor. Uh you can uh you can drop out and and in fact return. Uh so um uh I dropped out after my freshman semester um to start this company uh uh with uh with hard uh that was super fun and then after a couple years of that went back did another year uh at MIT and uh and then dropped out again to to start stripe.

06:07 Um, and you know, I um when I went to college, probably like a lot of people here, uh I um I had this vision uh of my life involving becoming an academic and I really liked physics and I thought, you know, I'll do all this physics stuff. It's so cool. I'd read all the Fineman books, you know, all of this. Um and I guess I um well, growing up in Ireland, I hadn't realiz I hadn't thought much about the possibility of startups.

06:35 Uh hello to the other Irish folks here. Uh and um and I mean way back then in the sort of you know pre-Camrian era startups were definitely much less know well known even on campus and so forth. You know when I was dropping out people thought it was super weird. Um I think um you know overall um if you enjoy college I I would actually you know I I think there's no harm in in finishing.

07:04 I I felt this real sense of urgency which I think in hindsight was a bit unnecessary. Um if you but if you don't enjoy college just you know whatever it's not your your thing. It's not what captivates you. You don't really want to learn all the physics things or whatever. uh they're you know I think a lot of parents think that dropping out is very risky and will impugn your reputation for the rest of your life and so forth and as far as I can tell nobody has ever cared.

07:32 So I I both think you don't need to but also the cost of doing so are dimminimous. >> What was the urgency you were feeling? >> The urgency >> Yeah. to to go out and do do something. >> I don't know. Life is short, right? Um, and I I I all I mean it was a general kind of haste. Uh, I think you know a lot of us um I'm I'm sure many of the people here you know you kind of get into this mode of speedr running high school and then you know once you get to college it's like obviously I want to speedrun that as well and do

08:01 all the things. So there's a bit of that a bit of um Mark and Gre also talks about a version of this. I thought that a bunch of the opportunities uh in startups and in Silicon Valley and so forth were ephemeral and fleeting and if we didn't build it then that you know it wouldn't be possible to do it in three or four years and maybe all the opportunities will be gone.

08:20 You know in hindsight I think that um that was a poor intuition. Uh it's been pretty robustly and reliably the case over many decades that Silicon Valley has a surfet of opportunities. Um yeah I think it was mainly those two things. Do you think it's um I mean this is a very common thing that we hear when we talk to students now is that they are part of the reason they want to drop out on mass it seems at this point is the worry that actually now is the moment that there's sort of I think the meme going around is that

08:46 if you don't sort of uh drop out and start a company and make lots of money you're going to be trapped in the permanent underclass. So is that um should everyone here be worried about being stuck in the permanent underclass I guess is the question. Um I think um humanity has always had um a an affinity for these millinarian sort of models of how uh everything is um you know everything will soon come to an end uh and be this uh this sort of permanent transformation of society and so forth.

09:17 There's a great book, The Winged Gospel. People thought that after the um the invention of aviation that it was just like civilization was just enter and humanity as a species were entering a new era and nothing is going to be the same. And yeah, obviously aviation was was a pretty big deal, but uh I I don't think it was sort of quite the um the sociological rewriting uh that some of the uh uh you know, excitable proponents of the time imagined.

09:42 So uh I um uh you know it's it's hard to predict anything especially the future uh but uh I would um I would take the under on this being the last couple of years uh to uh to create a company. >> Fair enough. Um so going back to the Stripe story um Stripe ostensibly seems like a good idea like even on day one it's internet's a big deal, money is a big deal like combine those two things presuming is that how it went when you went told people you wanted to start stripe did everyone just say hey this is a great this is

10:10 obviously a good idea. It was kind of funny. It was um it was so some something we learned from YC um uh was uh that the importance of focusing on very concrete easy to explain uh customer problems like it's it's very easy to um to hallucinate or to you know imagine some customer problem that's uh not actually something viscerally felt by a person who would pay money.

10:35 Uh and so over the course of in part working on Octomatic together, we sort of encountered this issue of it being really annoying to deal with uh the move movement of money or payments whatever on the internet. Um and on the one hand it seemed like a an obviously good idea in the sense that nobody liked the existing ways of doing so. Um and uh they were broadly extremely unpopular and kind of antiquated and legacy and you had to like fill out all this paperwork and go to the bank in person and the paperwork was in

11:01 Latin and just like it was all bad. Um, but then the flip side is, uh, it just seems kind of ridiculous that two kids would start a financial services business. Um, and fintech didn't exist as a sector at the time. Like the word literally didn't exist. Uh and so it's just kind of you know we felt like the proverbial squirrels you know in a trench coat trying to masquerade as a sort of a real business uh or as you know serious adults but obviously knowing nothing coming in about the about the space and and certainly a

11:35 lot of people we met and pitched or banks or partners or whatever that we talked to. I mean, you they didn't literally laugh us out of the room, but I you you kind of see them looking for the button to like call security uh under the desk and have them haul us out because it just seemed so improbable. So anyway, I'd say like it both seemed like an obviously good idea in that people really wanted this, but also a bad idea and that nobody took it seriously.

11:56 Um but I I think that I think the fact that it was ultimately the fact that it was grounded in such a concrete actual real user problem saved us. Um you actually speaking of that how did you you had to in order to actually build the product you had to get a banking partner and um do things that a typical software company did not have to do as two young founders like how did you manage to convince a a bank to trust you in the end?

12:20 >> Yeah. Um well actually this is not an answer uh to your question uh but um just a thing that strikes me as I sit here is the reason we decided to start stripe um is because so John and I were in college together. He was in his freshman year and we went to startup school uh in 2009 which was held in Berkeley. Uh and we um we thought it was pretty cool.

12:45 Um and so we went and we got sushi afterwards in Petrero and we were walking back from sushi and we're like you know we'd kind of been kicking around this idea for um payments thing or like we've been thinking about the space and it was walking back that evening after startup school uh that we decided to start stripe. Um, I remember literally where we were in the road and I remember who we said to each other which was, "Yeah, you know, we might as well because it probably won't be that hard."

13:12 >> Okay, so moral of the story is go get sushi and promo tonight and you might start the next stripe. >> Um, and yes, be beware of sort of these these ultimate yak shaves. We thought we could do it on the side while in college, you know, take a couple months and that was almost 17 years ago. Um, at the time I remember you were also unusual in that you took sort of longer to do a big public launch and especially within the YC world the motto is very much sort of launch early, launch quickly, be out there and iterate.

13:42 Um, could you maybe talk us through a little bit about that? Why did you do it that way? >> Yeah, so um we started working on Stripe um kind of seriously in the uh well we started working the week after that school but um we're kind of in college wasn't full-time. started working full-time the summer of 2010. We launched publicly September 2011. So almost uh two years after like the first lines of code after the repo was started.

14:07 And yeah, waiting two years to launch seems I mean you know if we're going to YC meetings you know every every week I think we'd have been you know bludgeoned on the head. Um I think um look in many domains that probably is the wrong thing to do. um in our domain to kind of your last question because we had to do so much stuff around security and partners and money movement and infrastructure and reliability and all the things it just we didn't feel like we could scale a really good self-s serve experience without

14:35 getting a lot of the kind of the preconditions um and the infrastructure in place. Um the I think this the thing that saved us um and meant that it wasn't a total walk in the wilderness is we had production users almost from the very beginning. So first lines of code um in fall of09 we got our first live production user uh in um in January of 2010. So like two months into working or whatever and it did very little like it was very laral and incomplete.

15:10 Uh and our first production customer was Ross Buché at a company called 28 North. Um and all I could do was charge a card. Uh and so you know Ross would charge the card. Uh and you know then he would ask sort of a reasonable question like you know how do I how can I look at all my charges? Uh and like reasonable request and so you know let's code up a little dashboard here.

15:31 Uh and then be like well I want to refund a payment and you know all right we'll build refund support. Uh and then you know after a couple of weeks he was like so you know at some point do I get my money and like also a reasonable request so let's uh let's build that functionality. So it was very kind of just in time development. Anyway so we we had a a production customer from very early and then we did increase it was in private beta we increased the number of customers every single month you know all the way to that

15:56 public launch. And so every, you know, every week we had actual customer feedback requests, new users coming in. We're learning things from reality as opposed to our own kind of hypothesized or extrapolated uh conception of it. And I I I think if you have, you know, a significant stream like that of of um of grounding, I think it's probably okay to not be like launch.

16:20 No, >> wouldn't you? >> I mean, you're you're an expert YC partner. Do you agree? >> That's a good question. Um yeah I mean it is this is the issue with advice in general is it's sort of so generalized and like the especially in startups the exception proves the rule right so I think those are yeah certainly certainly if um you know your the cost of failure is high um then it almost certainly you have to sort of take longer to like build um maybe a slight tangent but something I'm curious about is related to this

16:53 though is you know we were talking like with with these coding agents, the ability to just like build and produce software cheaply and quickly. Um I I wonder should people be taking more of this path like should people be more ambitious in general with what the version one of the thing that they launch is? Um or you know or is it still fundamentally good product design to start like narrow and focused and then expand out once you know what people want?

17:16 >> Yeah. Um it's a good question. Um I think probably in the era of AI I mean I don't know and you know to some extent YC will will be I think the expert here but um you know this the whole kind of traditional lean startup doctrine of exactly what you say like start out by buying the Google ads or something and and identify this crevice or whatever and and iteratively expand out from it.

17:49 I think you can certainly imagine that that becomes much more competitive and much more um you know aggressively tilled and it's kind of hard to find those those little niches. The internet's a much bigger place than it was 20 years ago when some of those ideas emerged. Whereas taking these really divergent starting points where nobody else uh is uh is uh trying to um occupy that territory is is maybe a more like basically maybe you have to more aggressively decorrelate uh in the era of AI and I think it is interesting

18:23 to think about you know many of the companies that are most successful over the last 10 years so many of them are are very anti-lean startup right uh whether it's you know the labs themselves or Anderil or um yeah, you you can go down the list. A lot of them have this characteristic. So I um yeah, I think maybe a better way of saying it is 20 years ago the whole lean startup thing was uh was almost the only thing to do because of capital available and you didn't have AI that made I don't know spinning up an

18:54 organization with many different potentialities, capabilities so much easier. Whereas now I think you can start these much more aggressive and ambitious things up front. um within sort of YC and probably startup law at this point you're famous for the um at least the Paul Graham term schle blindness the stripe um uh at least on the surface was not like you know involved a lot of schleps like things I presumably weren't like the um intellectually most interesting things to uh to work on um and I always found that

19:20 especially interesting for you because you just mentioned you you had academic interest in physics and um I just clearly like you know a deep intellectual and have very many things that you're interested in as Stripe has sort of grown grown into this in this big company. In what ways sort of you know in what ways um are there sort of like intellectual um rewards that you've you've given up and which ones have you gained?

19:44 >> Yeah, I I mean look in any company there's a bunch of stuff that's um not that rewarding or in and of itself all that interesting like set setting up payroll. No one sort of starts a company so that you can uh you can set up payroll. uh and certainly building business financial services there's all sorts of you know more arcane and extensive uh versions of that um I think that um I actually feel extremely lucky with stripe um and in this respect and uh I think this is something I don't know if you need to think

20:13 about it that much upfront but I think one should think about it maybe before you raise a significant amount of money um you know you always worry naturally about possibility of failure and you know what'll happen if you fail and how to mitigate and avoid failure and all those things. I think you need to ask the uh the sort of converse of that uh what if you succeed and you know you raise money and you have customers and you have employees and a whole thing like are you going to be are you going to enjoy that?

20:44 Are you going to want to work on that for 10 years for 17 years for 30 years? I mean Larry Ellison at Oracle is going for I mean I guess it'll be a half century soon right? Um, so, so you know, what if you succeed? And in the case of Stripe, I really love it because, you know, we're working with the world's most interesting and innovative companies.

21:07 uh like we're uh 25% of all Delaware in corporations are started with strike uh via Atlas and then we get to partner with them and work with them and hear from them and get their feedback and get the request and everything you know through the entirety of the journey up to being the Shopifies and the Open AIs and the you know all all the um uh the uh the standout successes.

21:28 Um oh and actually speaking of Atlas uh we're giving free Atlas incorporation to everybody at Startup School. So um If you are uh struck by the uh the urge to found something uh you know over dinner this evening as we were uh just email startup school atstripe.com and we will get you your link uh for free antlas. Um but uh but yeah I you know I think PG latched on to something where yeah there all these kind of menial tasks but but in the kind of totality of Stripe I find it so interesting like every business is a kind

22:05 of applied theory on how some aspect of the world works or how some market works or how some you know h how if it's a new company with a new um a new model it's kind of a contrarian thesis on some counterfactual just like it's it's I've never met a Stripe customer and thought that's boring. Um so so it's actually the business as a whole has been the opposite of uh of the blindness um instinct >> and and you have a particularly unique in um perspective on this because you work with the big model um providers or big lab

22:38 companies and you work with all of the fast growing AI startups on the ground. Uh something that came up a lot here yesterday uh honestly comes up within the batches too is people are just worried about um is my idea going to get sort of trampled by the the big uh lab providers and I'm giving your perspective I'm just curious like how how should people think about that?

22:58 >> Yeah. Um >> yeah again predictions are hard and certainly the labs are very competent capable organizations. Um and maybe to separate a little bit, will rapidly improving AI capabilities do this or will the labs specifically themselves do this? Um I think in general the track record of like no organization if we go back 20 years you know there's some of the sense with Google like you know when we were doing Octmatic the question was always for our company and every other company you know what if Google does this and

23:35 Google seemed kind of omnipotent and had this uh immense number of incredibly talented people and essentially infinite access to capital and server just all the things and just human organizations are complic licated and it's very hard to have um to manage to aggressively prosecute 100 different priorities and to deal with all the issues and interference that arises among them and so forth and so you know Google has done incredibly well in a bunch of specific places but it's not like Google has done all the things even

24:09 if in some kind of basic material sense Google maybe you know had that ability so I'd say that the kind of the track record that of that is um is uh is checkered and and in general I think that fear has been overstated. No, I think there is a more specific thing of just like models themselves. Forget the labs. Even even if the labs aren't specifically ambitious about expanding their scope just like literally LLMs uh will will obiate a bunch of or agent capabilities will obviousate a bunch of of you know specific

24:37 verticals or tasks or something you know hard to say obviously contingent on one's forecast it's the model capabilities themselves uh but you know in certain cases I'm sure that will happen and you know in certain domains it has already happened. Looking at the stripe data, one thing I will say that I think is gerine to people here, um there are many more businesses getting started now than there were a year ago, like as little as a year ago.

25:01 Um way, way more than we're getting started, you know, 5 years ago. Uh and actually the relative change between last year and this year is pretty much the largest relative change we've seen in any given year. So for example from 19 from 2019 to 2020 we saw a big jump you know understandable during co so you know um February to April of 2020 or whatever uh you know I I think the growth rate inflected to maybe 50% or thereabouts uh year-over-year in terms of new businesses getting started um as I speak the number of new

25:34 businesses starting on Stripe is up around it's a bit under but around 2x year-over-year um which again is the largest relative jump uh we've seen. Um and you might think, okay, fine, you know, there's way more vibecoded kind of lightweight slop, you know, whatever. Like fine, there's more things, but like are they actually succeeding? Um but actually the median business uh is doing better this year than a year ago.

26:03 Um and so and then if we kind of um stratify it and look at the probability that any given business will reach some revenue threshold a million $5 million $10 million whatever um those all seem to be getting better. uh business are uh the time to revenue for new companies in corporate with Atlas is declining and so by all the kind of objective metrics we can look at uh it seems to be a better time than ever to start a business now again things can change I don't know what the world's going to look like in 5 years but

26:37 you know speaking today on July 26th or whatever it is uh of of 26 um I think it is the stripe data would suggest it's there's never been a better time >> um I mean We see the exact same thing in the YC batches. Companies are just able to grow faster than ever. Um certainly within the batch it used when back in again the old days when Arj and I were first starting out like getting to a million dollars of revenue like run rate revenue was a big deal like people would know about that company.

27:03 They be like you know I heard that X company got to a million dollars of revenue and now I mean that that's >> yeah that should you should be by your first month it feels like should um that's an exaggeration for everyone here. Um uh but I mean certainly within sort of that sort of like the YC uh part of the life cycle like day 0 to 90 it's really being driven by I would say enterprises willing to buy from startups which is the new thing so you can sign these new contracts um within like the batch.

27:32 Um you have the data as the companies keep growing. I'm curious are there other factors that are driving these sort of um uh inflicted growth curves from like 1 to 10 and 10 to 100. I I think it's really the dynamic you just mentioned uh which is businesses uh businesses everywhere are more um springloaded uh to adapt and to try new things and they have a real terror of being left behind with archaic and antiquated ways of operating.

28:07 And so in normal times, you're a new startup. You've, you know, some mechanism for doing whatever and you pitch the CIO or the CTO or the whoever at some company and they kind of don't want to talk to you because, you know, your thing is not validated. Maybe you won't be around in two years, you know, all all the kind of obvious objections. But now people know that, well, the risk of the status quo is actually extremely high.

28:30 And so even if there's risk in doing all the new things, well, this path also looks pretty dangerous. And so I really think there's never been a better time for startups to to sell um and to have their products get adopted at you know pretty meaningful scale right out of the gate. uh a lot of YC companies in recent times have demonstrated this but uh I think it's a it's a really pervasive uh dynamic and and there's a bit of it I think also I mean Stripe is not a consumer company obviously but you know I think there's

28:55 some version of this on the consumer side where I think consumers I mean are also pretty I mean consumers have complicated views on AI and maybe they don't want the data centers uh but uh people are very intrigued by the products and I think there is a kind of they're kind of beguiled by them and there's a a predisposition and an openness to experimenting with the new.

29:18 >> Um maybe just more broadly something I'm curious about is again with this the data you have at Stripe um has anything you've seen in that data um changed a belief you have about AI broadly say over the last 12 months? I mean there's a fear uh that AI is going to be this um hegemonic centralizing totalizing force where a small number of companies gobble up a very large share of the economy and many companies at the forefront of AI um have done incredibly well and I think will continue to do uh incredibly well for

29:58 sure. But based on what we can see at Stripe, the hunger and the intensity with which other companies are either getting started taking advantage of these new capabilities or existing companies are retooling. I don't worry about the centralization in the same way. Uh I think there I think there are going to be many thousands of winners. Um and again we try not to offer any definitive prognostications cuz the future is not predetermined but based on the the trend lines we can see I think we are heading towards a um a

30:34 more decentralized world and one with more broad-based prosperity. >> Cool. All right. Well, I think that is all we have time for today. So um thanks so much Patrick. Thank you for having me. And um >> it would be remiss of me not to say that Stripe would not exist without YC. Oh, >> cool. All right. Thank you so much.