WEBVTT - Garry Tan Decides Who Gets to Build the Future 0:00:09.690 --> 0:00:14.030 The Builders is a production of Kaleidoscope and iHeart, presented 0:00:14.190 --> 0:00:20.020 by Ramp. I'm Walter Isaacson. Welcome to The Builders, my 0:00:20.060 --> 0:00:26.360 series of conversations with the most consequential minds shaping our future. Today, 0:00:26.720 --> 0:00:32.970 I'm speaking with Gary Tan. Y Combinator has become one 0:00:32.990 --> 0:00:37.490 of the most powerful and important brands in Silicon Valley. 0:00:38.229 --> 0:00:45.110 The companies that emerged from YC include Airbnb, Stripe, DoorDash, 0:00:45.510 --> 0:00:51.830 and Dropbox. Together, they have transformed how we travel, spend money, work, 0:00:52.070 --> 0:00:57.770 and communicate. Now, Gary Tan is deciding which builders come next. 0:00:58.570 --> 0:01:02.630 Gary first entered Y Combinator as a founder. He became 0:01:02.670 --> 0:01:07.500 an early investor in Coinbase and Instacart. And today, He 0:01:07.620 --> 0:01:13.450 leads YC itself. And so in some ways, he's helping 0:01:13.630 --> 0:01:18.370 shape those who shape the future. Here's my interview with 0:01:18.450 --> 0:01:29.459 Gary Tan. Gary Tan, welcome to The Builders. 0:01:30.300 --> 0:01:30.919 Thanks for having me. 0:01:31.580 --> 0:01:35.270 As a leader of Y Combinator, you have to pick 0:01:35.660 --> 0:01:38.880 founders who are going to make it. Do you look 0:01:38.920 --> 0:01:42.020 for one particular thing or are there like 20 different 0:01:42.080 --> 0:01:43.959 models of founders that can work? 0:01:44.819 --> 0:01:47.380 Well, so the wild thing about Y Combinator is that 0:01:47.540 --> 0:01:52.160 we get 80, maybe 90,000 applications a year now for 0:01:52.400 --> 0:01:55.500 about 800 spots. So it is less than a 1% 0:01:55.500 --> 0:01:59.650 acceptance rate. Maybe 8 to 10% of those people would 0:01:59.670 --> 0:02:03.570 get an interview, sometimes less now. And the wild thing 0:02:03.630 --> 0:02:06.010 is we actually spend only 10 minutes with them. for 0:02:06.030 --> 0:02:09.030 that final stage. But the interesting thing is, you know, 0:02:09.110 --> 0:02:11.950 Paul Graham, the founder of YC, figured out that you 0:02:11.990 --> 0:02:14.250 didn't need half an hour. You don't need an hour. 0:02:14.800 --> 0:02:17.840 You could probably actually do it in about five minutes, 0:02:17.960 --> 0:02:21.520 but five minutes feels rude. So that's how YC ended 0:02:21.580 --> 0:02:22.320 up at 10 minutes. 0:02:22.580 --> 0:02:25.000 So what did Paul look for and what do you 0:02:25.080 --> 0:02:26.960 look for in those 10 minutes? 0:02:27.600 --> 0:02:30.860 I mean, it sounds almost too simple, but 10 minutes 0:02:30.919 --> 0:02:34.620 is a wonderful time. constraint, kind of like how 140 0:02:34.620 --> 0:02:36.660 characters for Twitter way back in the day was a 0:02:36.700 --> 0:02:40.760 wonderful constraint. If you can say something very simply that 0:02:40.919 --> 0:02:43.900 hits us, like, you know, here's a user, here's a 0:02:43.960 --> 0:02:48.090 real person who has a problem, and here's how the 0:02:48.120 --> 0:02:51.050 technology that we build actually solves that. We just want 0:02:51.070 --> 0:02:53.270 to know what it is and why it's awesome. 0:02:53.830 --> 0:02:54.530 And then if we have a. 0:02:57.430 --> 0:03:00.190 That founder is the person to go and build that thing. 0:03:00.490 --> 0:03:02.370 And you can do all of that in 10 minutes, 0:03:02.530 --> 0:03:05.950 and it's an up-down decision. Something like 10% to 15% 0:03:05.950 --> 0:03:10.190 of all self-made billionaires as of present time, age 50 0:03:10.190 --> 0:03:12.790 and under, are YC alums. 0:03:13.490 --> 0:03:15.750 Let me ask you a real quick question then. Suppose 0:03:15.880 --> 0:03:18.930 Elon Musk walked in and you had 10 minutes with him. 0:03:19.280 --> 0:03:22.380 What do you think the percentage chance is that you 0:03:22.400 --> 0:03:23.540 would actually fund him? 0:03:24.480 --> 0:03:28.109 Oh, 100%. I mean, in current form, 100%. No, no, no, no. 0:03:28.190 --> 0:03:31.269 I'm saying you saw him. You didn't know he was successful. 0:03:31.669 --> 0:03:35.280 He was a guy who'd just been, say, ousted from PayPal. 0:03:35.660 --> 0:03:37.740 I mean, I had a managing partner meeting the other day, 0:03:37.860 --> 0:03:40.560 and that's actually one of the example questions we ask 0:03:40.620 --> 0:03:44.040 ourselves to figure out whether we're on the right track, right? Like, 0:03:44.140 --> 0:03:48.110 by definition, if YC is operating properly... Like if an 0:03:48.230 --> 0:03:51.670 Elon or a Bill Gates or a Steve Jobs or, 0:03:51.890 --> 0:03:54.590 you know, a Drew Houston from Dropbox, a Brian Chesky 0:03:54.630 --> 0:03:58.790 from Airbnb, Brian Armstrong from Coinbase, Tony Hsu from DoorDash. 0:03:58.870 --> 0:04:02.630 Like if someone like that walks into our office, can 0:04:02.670 --> 0:04:04.790 we say yes to them in 10 minutes? And if 0:04:04.840 --> 0:04:08.080 that answer is no, actually, I'd have to rewrite everything. 0:04:08.220 --> 0:04:10.830 I would have to retool everything, right? That is actually 0:04:11.010 --> 0:04:13.190 almost the only question that really matters when it comes 0:04:13.240 --> 0:04:14.820 to whether YC is doing well or not. 0:04:15.120 --> 0:04:18.440 And if you had walked in years ago, would you 0:04:18.480 --> 0:04:19.320 have accepted you? 0:04:19.339 --> 0:04:22.480 I think so. 0:04:23.320 --> 0:04:26.200 My startup, Posturus, it was dead simple blogs by email. 0:04:26.520 --> 0:04:28.990 And I mean, I had a pretty banger demo. So 0:04:30.350 --> 0:04:33.430 we had something that I think Paul and Jessica and 0:04:33.490 --> 0:04:37.490 Trevor Blackwell, the founders of YC, really got immediately. They 0:04:37.510 --> 0:04:39.430 didn't grill us at all. Like we came on with 0:04:39.510 --> 0:04:42.549 a really good demo and they said, well, that makes sense. 0:04:42.810 --> 0:04:45.440 And then we got funded. So sometimes it's that simple though. 0:04:45.860 --> 0:04:49.880 And I was creating that company, running that company for 0:04:49.920 --> 0:04:55.039 a while. How fundamental was that to your ability to 0:04:55.120 --> 0:05:00.590 be at Y Combinator and to be making venture capital decisions? 0:05:01.810 --> 0:05:04.420 Oh, I think it's quite important. I mean, There's a 0:05:04.440 --> 0:05:07.020 Jay-Z line that I love. One of his songs, he goes, 0:05:07.430 --> 0:05:09.110 everybody wants to tell you how to do it and 0:05:09.120 --> 0:05:12.809 they never did it. And so, I mean, the entire 0:05:12.930 --> 0:05:16.430 act of being an investor in technology is actually bewildering 0:05:16.470 --> 0:05:18.830 and strange if you really think about it too long, 0:05:18.930 --> 0:05:23.250 because what we're doing is selling money. Like tech has 0:05:23.310 --> 0:05:27.230 now become such a fundamental force in the financial world 0:05:27.490 --> 0:05:31.400 of society that that there's way more money that wants 0:05:31.440 --> 0:05:34.080 to go into tech than good opportunities. 0:05:34.740 --> 0:05:36.580 And that's really what YC is. 0:05:36.640 --> 0:05:39.409 And if you look at most venture capital, it's crazy 0:05:39.450 --> 0:05:42.110 that it's grown so much. The S & P 500 0:05:42.110 --> 0:05:45.089 is almost entirely driven by tech. And then the wild 0:05:45.130 --> 0:05:48.410 thing that Paul Graham figured out and Jessica Livingston figured 0:05:48.450 --> 0:05:51.190 out when they first started was, when they started in 2005, 0:05:51.190 --> 0:05:54.640 the type of people who would get funding, it would 0:05:54.680 --> 0:05:58.160 be your buddy at Stanford Business School, your golf buddy, 0:05:58.339 --> 0:06:01.800 like maybe someone who your son went to school with 0:06:01.960 --> 0:06:04.670 at like some prep school at Harvard Westlake or what, 0:06:04.700 --> 0:06:07.490 you know, I mean, it was like access, right? 0:06:07.790 --> 0:06:10.190 As it was for a few millennia. 0:06:10.330 --> 0:06:11.650 Time immemorial, right? 0:06:11.890 --> 0:06:13.589 But do you all think you broke the back of 0:06:13.630 --> 0:06:16.849 that or is being a startup junkie or a startup 0:06:16.930 --> 0:06:19.890 entrepreneur something that you still have to be pretty privileged 0:06:19.910 --> 0:06:21.300 and pretty connected to do? 0:06:22.180 --> 0:06:23.580 I mean, it would be a mistake to say that 0:06:23.600 --> 0:06:27.010 we were actually a meritocracy, but I will say and 0:06:27.170 --> 0:06:30.230 affirm loudly that we aspire to be. 0:06:30.610 --> 0:06:33.330 We will never be. You know, I think that the 0:06:33.470 --> 0:06:35.570 woke age of sort of like five. 0:06:35.410 --> 0:06:37.890 Or 10 years ago, we sort of lost sight of that, 0:06:38.110 --> 0:06:41.700 that like meritocracy could never be achieved, but it is 0:06:41.740 --> 0:06:43.599 a virtue to attempt to achieve it. 0:06:44.279 --> 0:06:46.800 Well, the reason I ask about that in terms of 0:06:46.839 --> 0:06:50.360 a startup culture, which is what you very much nurture, 0:06:50.940 --> 0:06:54.480 is that it used to be somebody coming out of college, 0:06:54.560 --> 0:06:57.680 take a Steve case, for example, could go work at 0:06:57.720 --> 0:07:03.380 a company like a Procter & Gamble and learn a business. Nowadays, 0:07:03.460 --> 0:07:06.800 it's very hard to sort of get a steady job. 0:07:07.180 --> 0:07:10.220 And the economy for young people tends to depend more 0:07:10.260 --> 0:07:13.670 and more on startups. And yet it seems to me 0:07:14.490 --> 0:07:17.790 that unless you got a safety net, unless you're somebody 0:07:17.810 --> 0:07:20.950 who doesn't have to make sure your next meal can 0:07:20.990 --> 0:07:24.360 be paid for, you can't be in that game, and 0:07:24.420 --> 0:07:25.980 it leaves a lot of people out. 0:07:26.560 --> 0:07:30.440 I mean, that's why YC gives $ 500, 000, for instance. When 0:07:30.490 --> 0:07:33.490 I did YC, it was only $ 18, 000. Like today, you 0:07:33.530 --> 0:07:37.130 hear about seed funds and Stanford kids raising $ 30 million 0:07:37.190 --> 0:07:40.930 on $ 300 million post-money evaluation. It's like literally, like it's 0:07:40.970 --> 0:07:44.420 just an idea and one person, and the company is 0:07:44.460 --> 0:07:47.760 somehow worth $ 300 million. But when we were starting many 0:07:47.800 --> 0:07:50.080 years ago, It was a very small amount of money. 0:07:50.260 --> 0:07:52.590 You had to show that you had something that was great. 0:07:53.310 --> 0:07:54.489 And so that's why. 0:07:54.490 --> 0:07:58.730 YC didn't stay. And we easily could have just stayed 0:07:58.770 --> 0:08:02.530 at $ 12, 000. 0:08:00.070 --> 0:08:01.890 Or $ 15, 000. 0:08:00.890 --> 0:08:03.330 But I think there is a recognition that if we 0:08:03.370 --> 0:08:06.090 want to be a thing that allows people to participate, 0:08:06.790 --> 0:08:09.370 you don't have to have rich parents. The half a 0:08:09.390 --> 0:08:12.390 million dollars getting into YC, you could just have an 0:08:12.450 --> 0:08:15.230 idea and you could come in and meet us for 0:08:15.230 --> 0:08:15.790 10 minutes. 0:08:17.140 --> 0:08:20.160 You were one of the earliest employees at Palantir. You 0:08:20.200 --> 0:08:23.430 got to see how Peter Thiel and Alex Karp in 0:08:23.490 --> 0:08:27.630 particular really worked. What did you learn from their style? 0:08:28.390 --> 0:08:30.850 I realize now, like, if I hadn't rage quit, I 0:08:30.890 --> 0:08:33.570 would have been, you know, sort of a peer to 0:08:33.630 --> 0:08:36.720 some of the, you know, real legends who really built Palantir. 0:08:36.800 --> 0:08:38.500 You know, I would go on walks with Karp and 0:08:39.160 --> 0:08:41.300 he would sort of, I mean, he was a philosopher. 0:08:41.380 --> 0:08:43.540 You know, he was not a technologist, I would say, 0:08:43.740 --> 0:08:47.530 but he was incredibly knowledgeable about the world. One time, 0:08:47.550 --> 0:08:52.229 I was sent to DC to do some demos for customers. 0:08:52.910 --> 0:08:56.090 And I remember being like a 25-year-old, and I was 0:08:56.130 --> 0:09:00.370 in my ripped jeans and band t-shirt and getting sent 0:09:00.410 --> 0:09:05.410 to stay at the Mandarin Oriental in DC. And we 0:09:05.429 --> 0:09:09.650 would wonder, Alex, why are we staying at the Mandarin Oriental? 0:09:09.690 --> 0:09:10.810 Look at us. We're just kids. 0:09:10.870 --> 0:09:12.570 We don't have to do that. 0:09:12.929 --> 0:09:15.030 Just set me up in the Motel 6. We're a startup. 0:09:16.780 --> 0:09:19.590 Alex looked at me and my other early engineers for 0:09:19.620 --> 0:09:23.710 the Palantir Finance product, and he said, guys, you're going 0:09:23.750 --> 0:09:28.510 to go and talk to such-and-such general or such-and-such VP 0:09:28.730 --> 0:09:31.450 of some Fortune 500, and you're going. 0:09:31.370 --> 0:09:33.590 To show this incredible software that you've built. 0:09:34.130 --> 0:09:36.090 And then as you go out, you're going to make 0:09:36.150 --> 0:09:38.130 small talk, and they're going to ask, well, how long 0:09:38.150 --> 0:09:39.640 are you around and where are you staying? And if 0:09:39.679 --> 0:09:42.460 you say Motel 6, they're going to think that they're 0:09:42.480 --> 0:09:47.439 buying Motel 6 software, and Palantir is... Mandarin Oriental software. 0:09:47.780 --> 0:09:50.119 And I'm like, okay, now I get it. 0:09:50.660 --> 0:09:52.480 And so, you know, yeah, I don't know. 0:09:52.679 --> 0:09:54.500 I think embedded in that story is just a lot 0:09:54.559 --> 0:09:57.010 of different things. And I learned a lot from Alex 0:09:57.050 --> 0:09:57.370 from that. 0:09:58.050 --> 0:10:01.630 Don't go anywhere. There's a lot more. right after this break. 0:10:10.320 --> 0:10:14.720 I know you've been fighting this antitrust battle. How important 0:10:14.780 --> 0:10:18.320 is that in order to keep innovation and startup culture? 0:10:18.980 --> 0:10:21.589 I think, you know, True free market enterprise has never 0:10:21.630 --> 0:10:25.130 been tried. I think that in tech in particular, we 0:10:25.170 --> 0:10:28.300 have very short memories. I would argue that Google itself, 0:10:28.420 --> 0:10:31.199 one of the big tech companies, was able to exist 0:10:31.380 --> 0:10:34.600 and thrive and become a trillion-dollar company because of the 0:10:34.700 --> 0:10:38.070 actions of the DOJ with the consent decree on Microsoft. 0:10:38.120 --> 0:10:41.349 Microsoft was ready to take over the web, and it 0:10:41.530 --> 0:10:45.970 used its platform advantages in Windows to default install and 0:10:46.030 --> 0:10:47.910 take over the browser market. 0:10:48.210 --> 0:10:50.670 And do you think Google is now doing the same thing? 0:10:50.690 --> 0:10:52.370 I would argue yes. 0:10:52.730 --> 0:10:55.980 The Delaware C Corp sort of actually insists on it. 0:10:56.110 --> 0:11:00.179 Like the governance structure itself actually insists on you're supposed 0:11:00.220 --> 0:11:02.620 to do your duty and your duty is to maximize 0:11:02.640 --> 0:11:06.260 shareholder value. So people are following the letter of the 0:11:06.300 --> 0:11:09.460 law and the spirit of like their Delaware C Corp 0:11:09.480 --> 0:11:12.450 to close platforms. Like it's less and less percentage of 0:11:12.500 --> 0:11:14.689 the screen real estate if you do a Google search 0:11:15.270 --> 0:11:17.469 is even non-Google properties, right? 0:11:18.190 --> 0:11:19.410 And so that's self-preferencing. 0:11:20.230 --> 0:11:23.240 So a lot of your tech bro friends are very 0:11:23.400 --> 0:11:27.660 much against antitrust enforcement. How do you have a startup 0:11:27.760 --> 0:11:31.380 culture if you're going to allow big companies to self-preference? 0:11:32.020 --> 0:11:35.570 Well, I mean, my view is that we shouldn't allow self-preferencing. 0:11:35.809 --> 0:11:38.949 And then my super pro startup stance is we should 0:11:38.990 --> 0:11:42.929 stop self-preferencing. We should require open markets. Like I think 0:11:42.970 --> 0:11:46.530 the DMA in Europe, for instance, is a great example 0:11:46.620 --> 0:11:50.350 of forcing big tech to open its own platforms and 0:11:50.700 --> 0:11:54.220 not just data transparency and privacy, but also portability and 0:11:54.260 --> 0:11:57.780 reverse engineering. These are sort of things that allow for 0:11:58.280 --> 0:12:01.840 open markets and open access to startups. And if you 0:12:01.880 --> 0:12:04.559 ask me, is big tech too powerful? I say absolutely. 0:12:05.160 --> 0:12:07.540 And the answer and the antidote is little tech. I 0:12:07.580 --> 0:12:11.880 think it's not that unlike robber baron times in that 0:12:12.460 --> 0:12:15.150 if you own the railroads and there was a monopoly 0:12:15.170 --> 0:12:18.740 on the railroad... you could set the price on literally 0:12:18.820 --> 0:12:22.530 everything else. And so one thing would lead to another thing, 0:12:22.650 --> 0:12:26.290 and you would have multiple monopolies across both commodities and 0:12:26.750 --> 0:12:28.670 the railroad the commodities travel over. 0:12:28.929 --> 0:12:31.030 Well, wait, let's look at AI. I'm going to push 0:12:31.070 --> 0:12:35.490 back a bit. We get a whole lot of new companies, Anthropic, OpenAI, 0:12:35.550 --> 0:12:39.640 they're all popping up. Somehow Apple wasn't able to self-preference 0:12:40.040 --> 0:12:43.360 and embed its own system in its own phones. Even 0:12:43.440 --> 0:12:46.160 Google is trying to do that where it almost forces 0:12:46.220 --> 0:12:50.240 you to use Gemini when you're using Google search. And 0:12:50.280 --> 0:12:52.300 yet these other companies are doing well. 0:12:53.140 --> 0:12:53.800 Yeah, that's right. 0:12:53.900 --> 0:12:57.490 I mean, I think what's funny about the AI companies 0:12:57.520 --> 0:12:59.590 is like they're almost certainly, I mean, I think you 0:12:59.630 --> 0:13:01.890 can fairly say that they are the new big tech. 0:13:02.790 --> 0:13:06.570 And so I think I have concerns there as well, right? 0:13:07.230 --> 0:13:09.870 Critics of some of the AI companies would say that 0:13:10.410 --> 0:13:13.750 It's possible that as the AI wars progress, there might 0:13:13.809 --> 0:13:17.780 not be open access to the latest and best frontier 0:13:17.820 --> 0:13:20.420 models that those companies may gatekeep. 0:13:20.880 --> 0:13:24.280 You think they should be open, that all the AI 0:13:24.320 --> 0:13:25.300 models should be open? 0:13:25.900 --> 0:13:28.410 I mean, that would be certainly the thing that supports 0:13:28.470 --> 0:13:31.390 startups and the ability for new people to create new 0:13:31.490 --> 0:13:34.990 companies and, frankly, new wealth and to give consumers choice. 0:13:35.150 --> 0:13:37.770 So what is the role that Y Combinator wants to 0:13:37.850 --> 0:13:38.680 play in AI? 0:13:40.200 --> 0:13:42.760 I mean, what's funny is I think of YC almost 0:13:42.840 --> 0:13:45.840 as like, you know, we're not a union, but we 0:13:45.880 --> 0:13:49.660 have union-like characteristics. There's not really a group out there 0:13:49.740 --> 0:13:55.850 that can speak full-throatedly about policies and, frankly, actions that 0:13:56.210 --> 0:14:00.010 both governments and companies should take to make sure that 0:14:00.390 --> 0:14:03.250 the golden goose can still exist. And, you know, whether 0:14:03.290 --> 0:14:08.210 that's the billionaire tax in California or antitrust laws in D.C. 0:14:08.429 --> 0:14:09.089 Or in Europe. 0:14:09.590 --> 0:14:13.010 we're going to start talking about it and just having 0:14:13.050 --> 0:14:14.590 a seat at the table. One of the things I 0:14:14.610 --> 0:14:19.530 discovered as I started examining politics, most tech people are 0:14:19.570 --> 0:14:22.630 not interested in politics, but politics sure is interested in us. 0:14:23.230 --> 0:14:24.930 And so I decided. 0:14:24.630 --> 0:14:26.710 Like three or four years ago, like, oh, well, I 0:14:26.730 --> 0:14:29.830 should probably spend some time and reciprocate. Like, let me 0:14:29.850 --> 0:14:32.950 understand what's going on. And what I discovered was pretty wild. 0:14:33.130 --> 0:14:36.630 I mean, millions of dollars in AstroTurf money, going to 0:14:36.710 --> 0:14:40.910 groups that would take around basically almost hired actors. They 0:14:40.930 --> 0:14:44.630 would claim to speak for small software companies or, you know, 0:14:44.670 --> 0:14:45.190 little tech. 0:14:45.910 --> 0:14:48.650 And they were just like paid chills, actually. 0:14:49.050 --> 0:14:51.310 And they would go to the halls of power, whether 0:14:51.350 --> 0:14:55.510 it's Sacramento or D.C. And they would say, no, no, 0:14:55.610 --> 0:14:58.630 like this thing that actually mainly helps big tech. We, 0:14:58.690 --> 0:15:01.090 the little guy, really wants that. And I was like, oh, wow, 0:15:01.130 --> 0:15:03.510 this is crazy. Like no real startups actually go out 0:15:03.570 --> 0:15:06.790 there and talk to real people in D.C. And I 0:15:06.830 --> 0:15:09.810 saw a system that was broken, and I wanted to 0:15:09.870 --> 0:15:12.580 fix it. And so that, you know, we do very 0:15:12.640 --> 0:15:15.540 little things. Like, we have a tiny budget. It's, you know, 0:15:16.000 --> 0:15:19.340 probably one one-hundredth that of, like, even any given single 0:15:19.380 --> 0:15:22.820 big tech firm spends on policy. We have one lone 0:15:22.860 --> 0:15:26.680 full-time hire in D.C. in a man named Luther Lowe. 0:15:27.500 --> 0:15:29.490 And we run it on a shoestring. But on the 0:15:29.510 --> 0:15:33.230 other hand, like, we're people-powered. I see that like a 0:15:33.290 --> 0:15:37.530 repeating pattern, like there are broken things in society. There 0:15:37.550 --> 0:15:40.590 are systems out there that, you know, have money, like 0:15:40.650 --> 0:15:42.620 just trace the money and like, you know, where's the 0:15:42.650 --> 0:15:44.520 money coming from and who benefits, right? 0:15:45.060 --> 0:15:46.240 And then the heartening thing. 0:15:46.680 --> 0:15:48.100 You know, I think a lot of people talk about 0:15:48.140 --> 0:15:50.540 social media in a very negative way. They think, you know, 0:15:50.600 --> 0:15:53.290 it was Facebook that ruined democracy. And it's like, I 0:15:53.330 --> 0:15:55.729 don't really, I don't think I agree. Like, I think 0:15:55.770 --> 0:15:58.950 that It's the little guy not realizing he has a voice. 0:15:59.070 --> 0:16:01.489 He and she do have a voice. And then, you know, 0:16:02.330 --> 0:16:05.370 when I look at YC, like, our job one is 0:16:05.410 --> 0:16:08.530 to find the next great founders. But I think right 0:16:08.590 --> 0:16:10.670 under that, it's like, how do we make sure that 0:16:10.710 --> 0:16:13.410 those people can have access to markets? How do we 0:16:13.890 --> 0:16:16.360 help people who would not have a voice, would not 0:16:16.390 --> 0:16:18.620 be able to go to D.C.? How do we give 0:16:18.640 --> 0:16:21.580 them that megaphone? And then YC itself now has a megaphone. 0:16:21.600 --> 0:16:24.280 You know, we have more than 2 million subscribers on YouTube. 0:16:24.520 --> 0:16:28.100 We can sort of speak in aggregates. And I'm really 0:16:28.130 --> 0:16:31.740 excited about that because I think that that's us trying 0:16:31.800 --> 0:16:34.000 to speak truth to power. I don't want to deny 0:16:34.040 --> 0:16:36.460 that YC is also a power in and of itself, 0:16:37.140 --> 0:16:39.860 but the world is full of systems. These are all 0:16:39.900 --> 0:16:43.200 systems that were created and constructed by, as Steve said it, 0:16:43.400 --> 0:16:47.380 people no smarter or better than you or I. And 0:16:47.580 --> 0:16:49.440 we can mold it and change it. And once you 0:16:49.460 --> 0:16:51.180 realize that, you'll never be the same. 0:16:52.040 --> 0:16:56.210 You were born in Canada, Chinese background. Tell me how 0:16:56.310 --> 0:17:00.530 important you think immigration is to the United States. 0:17:01.510 --> 0:17:04.970 I mean, I shout to everyone who will hear that 0:17:05.170 --> 0:17:08.490 about 50% of all the startups that really make it 0:17:08.670 --> 0:17:12.500 have at least one immigrant co-founder. And what we realize 0:17:12.540 --> 0:17:15.379 is that's one of the things that makes America incredibly unique. 0:17:15.940 --> 0:17:18.760 People from around the world want to come here. And 0:17:18.780 --> 0:17:20.600 in tech, they want to come and build the future. 0:17:21.140 --> 0:17:26.030 That's certainly my story. My mom's family came over from Burma. refugees, 0:17:26.830 --> 0:17:29.950 and my dad came here from Singapore. 0:17:30.150 --> 0:17:33.070 Well, you talk about your dad, and we biographers tend 0:17:33.090 --> 0:17:37.470 to believe it's sometimes about dad, especially when somebody's successful. 0:17:37.710 --> 0:17:39.659 You had a pretty rough time, though. I mean, your 0:17:39.700 --> 0:17:44.180 father was at times unemployed, at times an alcoholic. Tell 0:17:44.220 --> 0:17:46.160 me how you had to deal with all that. 0:17:47.100 --> 0:17:48.840 I mean, it really made me who I am today. 0:17:48.859 --> 0:17:51.960 I figured out that I can actually put myself back 0:17:52.180 --> 0:17:58.520 into let's say 1992, I'm 11 years old. And basically, 0:17:58.540 --> 0:18:00.400 I don't know when my dad would come home, like 0:18:00.440 --> 0:18:03.340 maybe around five or six, but I wouldn't know how 0:18:03.380 --> 0:18:06.620 many beers he would have already had. And I realized 0:18:06.640 --> 0:18:09.830 there's sort of this scanner that happens where if you 0:18:09.869 --> 0:18:13.310 don't know what father is coming home that day, then 0:18:14.070 --> 0:18:18.750 you get very attuned to other people and their interstates. 0:18:18.930 --> 0:18:21.520 And like, how to sort of react to that. And 0:18:21.540 --> 0:18:24.570 then it's interesting because now I realize like that's a 0:18:24.590 --> 0:18:26.389 scanner that I use all the time. Like I can 0:18:26.430 --> 0:18:29.450 sort of sit down with someone and it doesn't happen 0:18:29.490 --> 0:18:32.490 every time, but it happens relatively easily for me where 0:18:32.950 --> 0:18:35.310 I can sit down and just kind of feel what's 0:18:35.330 --> 0:18:37.770 going on for other people as someone who sort of 0:18:37.850 --> 0:18:41.990 spends time with hundreds of startup founders every single year. 0:18:42.070 --> 0:18:45.060 Like that's actually a pretty powerful superpower of. 0:18:46.100 --> 0:18:48.560 Let me talk about what may be the next step 0:18:48.720 --> 0:18:53.330 in AI, which is AI agents. And you've said that 0:18:53.390 --> 0:18:56.429 you don't want to fund companies that look at AI 0:18:56.550 --> 0:18:59.189 agents as a feature. You want them to look at 0:18:59.230 --> 0:19:02.790 it as the core operating system. Explain how you think 0:19:02.830 --> 0:19:06.630 that changes what we think about when we talk about AI. 0:19:07.520 --> 0:19:08.359 Yeah, absolutely. 0:19:08.440 --> 0:19:10.670 I mean... We're going to get a little sci-fi here, 0:19:10.750 --> 0:19:13.190 but this is something that we're starting to see ourselves, 0:19:13.210 --> 0:19:16.190 you know, rebuilding the way we operate at YC. I 0:19:16.230 --> 0:19:19.070 think the number one thing that people get wrong is 0:19:19.190 --> 0:19:21.139 they just think that AI is just chat. 0:19:21.680 --> 0:19:23.520 And then it is so much more than that. 0:19:23.740 --> 0:19:27.340 We now have an internal AI agent, for instance, named Straylight, 0:19:27.480 --> 0:19:30.780 that kind of has a personality and then it has 0:19:30.820 --> 0:19:34.770 a company brain. Like it actually has the total context 0:19:35.010 --> 0:19:39.320 of everything. Founder we've ever backed, anyone who's ever emailed us, 0:19:39.380 --> 0:19:42.139 like all the applications. The way I think about it 0:19:42.380 --> 0:19:46.310 from an interesting tech perspective is like a human can 0:19:46.369 --> 0:19:49.030 only keep track in their heads like seven plus or 0:19:49.050 --> 0:19:52.070 minus three things. And what an AI agent is, is 0:19:52.210 --> 0:19:54.790 already super intelligent. It can keep, you know, I was 0:19:54.850 --> 0:19:57.600 explaining this to my 11 year old. What an AI 0:19:57.630 --> 0:20:01.130 can do is it can keep a stack of 1000 pages, 0:20:01.340 --> 0:20:04.500 think like three Harry Potter books. You know, it can 0:20:04.780 --> 0:20:07.820 needle in a haystack, any page, any fact, anything that 0:20:07.859 --> 0:20:13.450 happened in those three Harry Potter books. It can just know, synthesize, 0:20:13.830 --> 0:20:17.869 work with. Like the working memory is absolutely gigantic. And, 0:20:18.150 --> 0:20:19.930 you know, the thing is in another year, it won't 0:20:19.970 --> 0:20:22.920 be three Harry Potter books in the brain of an 0:20:23.080 --> 0:20:25.240 AI agent. It'll be five or it'll be 10 or 0:20:25.260 --> 0:20:28.680 it'll be 20. And then the set of things that 0:20:28.760 --> 0:20:30.600 it can operate on is even bigger. 0:20:30.619 --> 0:20:33.320 Well, give me a better example and make this more 0:20:33.380 --> 0:20:37.240 concrete for me, because data retrieval, even from three Harry 0:20:37.280 --> 0:20:40.720 Potter books, That's kind of cool, but that's not, I mean, 0:20:40.740 --> 0:20:44.389 that's just data retrieval. What makes something an agent? 0:20:45.010 --> 0:20:48.449 I mean, I would point to what Peter Steinberger of 0:20:48.690 --> 0:20:52.090 OpenClaw figured out, which is if you use cloud code 0:20:52.310 --> 0:20:56.190 or codecs, I would say right now, they're so worried 0:20:56.230 --> 0:21:00.050 about sort of repeatability and safety that they actually trade 0:21:00.190 --> 0:21:03.440 off all of the beautiful parts of working with an 0:21:03.500 --> 0:21:07.790 AI agent, which like basically can be so similar to 0:21:07.830 --> 0:21:10.560 a human that like it sort of takes on human qualities. 0:21:10.820 --> 0:21:13.000 So tell me what you ask your agent to do. 0:21:13.600 --> 0:21:14.100 Oh gosh. 0:21:14.340 --> 0:21:16.600 I mean, one of the things I did, I, you know, 0:21:16.640 --> 0:21:19.820 it's read all of my email and all of my Dropbox. 0:21:19.840 --> 0:21:22.820 It's read like all of my Apple notes, like, you know, 0:21:23.119 --> 0:21:25.560 even my therapy notes going back 20 years. I mean, 0:21:25.640 --> 0:21:27.820 for instance, like I used it for prep to be 0:21:27.859 --> 0:21:30.550 able to talk to you, not to get too meta 0:21:30.590 --> 0:21:33.770 about it, but in about 15 minutes right before this talk, 0:21:33.930 --> 0:21:37.320 like it was able to you know, download and transcribe 0:21:37.560 --> 0:21:41.320 your top 10 podcasts you've ever done. And then it 0:21:41.440 --> 0:21:44.629 actually diarized them. So, you know, I had an LLM 0:21:44.690 --> 0:21:48.110 go like line by line, timestamp by timestamp to pull 0:21:48.210 --> 0:21:49.490 out like major themes. 0:21:49.930 --> 0:21:51.689 And that was really helpful for me. 0:21:51.750 --> 0:21:53.970 But to be able to do that for like almost 0:21:54.210 --> 0:21:58.260 any thing that you do in your day-to-day life and 0:21:58.300 --> 0:22:01.260 for it to do it like somewhat more autonomously. And 0:22:01.320 --> 0:22:04.419