Y Combinator CEO on Founder Psychology in the Age of AI

1. 0:00 – 0:26

   Intro

   1. GTGarry Tan0:00

      We may never achieve a utopia, but it is worthy and worth it to attempt. Everything that's awesome in my life is kind of a cult. We had to go over to the Windows team. They wouldn't reply to our emails. They wouldn't fix our bugs, and we had to go over there with a baseball bat. An org like Microsoft can't.

   2. AAAnish Acharya0:18

      Yes.

   3. GTGarry Tan0:19

      But a startup can.

   4. AAAnish Acharya0:20

      Yes.

   5. GTGarry Tan0:20

      And every-

   6. AAAnish Acharya0:21

      Yes

   7. GTGarry Tan0:21

      ... startup must.

2. 0:26 – 4:45

   The 2003 Bleak Moment: Passing on Palantir & Chasing What Was Hot

   1. AAAnish Acharya0:26

      Garry, so good to have you, brother. Thank you so much for joining us today.

   2. GTGarry Tan0:30

      Thanks for having me.

   3. AAAnish Acharya0:31

      Amazing, dude. Well, we're gonna cover a bunch of the basics, and then we're gonna get into the advanced topics to make sure that we, uh, kinda match the, the curiosity, ambition of our audience.

   4. GTGarry Tan0:39

      That sounds great.

   5. AAAnish Acharya0:40

      Does that work? Awesome.

   6. GTGarry Tan0:41

      Yeah.

   7. AAAnish Acharya0:41

      Well, dude, maybe talk to me a bit about the beginning of your journey, 2003. You know, it feels like what's under-discussed is founder psychology and then sort of Silicon Valley culture and how the two are connected. What did it feel like when you first got in the game back in 2003 coming out of Stanford?

   8. GTGarry Tan0:55

      Yeah. I guess I wanted to work at startups, but, uh, 2003 was funny in that, like, Web 1.0 had just ended.

   9. AAAnish Acharya1:02

      Mm-hmm.

   10. GTGarry Tan1:02

       Like, the, you know, NASDAQ had just crashed, and, uh, actually there were no jobs that I could even find in the Bay Area other than, you know, working at Gap IT.

   11. AAAnish Acharya1:12

       [laughs] Rough.

   12. GTGarry Tan1:13

       And so the two job offers I had were, uh, Microsoft for Windows Mobile, which I was pretty excited about, or Expedia.

   13. AAAnish Acharya1:19

       Okay.

   14. GTGarry Tan1:19

       Yeah, I, you know, I had student loans, and I, you know, while I wanted to start a company, I just, you know, felt like I needed to get a job, and then actually there were no jobs in the Bay. So it was, like, kinda hard to believe actually. Uh, that moment was so bleak, and then right before Web 2.0.

   15. AAAnish Acharya1:37

       It felt like it might've been the end of tech, huh?

   16. GTGarry Tan1:39

       Right. A little bit. Yeah, that's, that was the vibe. And then, you know, at the end of the day, like, that's the boom, that's the bust. Like, these things sort of happen over and over again.

   17. AAAnish Acharya1:47

       So at the time, did it feel high status to work in tech or on startups?

   18. GTGarry Tan1:51

       Not at all. I mean, you know, it's sort of lost to the sands of time, but, um, I was still excited about tech, though. Um, you know, I had done a lot of web programming.

   19. AAAnish Acharya2:01

       Yeah.

   20. GTGarry Tan2:01

       And then one of my regrets, actually, was, um, I was sort of chasing what was hot instead of what was clearly the thing I knew. Mark Zuckerberg was probably doing Facemash, um-

   21. AAAnish Acharya2:12

       Yes

   22. GTGarry Tan2:12

       ... around then-

   23. AAAnish Acharya2:13

       Yes

   24. GTGarry Tan2:13

       ... around 2003. And then Facebook didn't even exist yet.

   25. AAAnish Acharya2:17

       Yes.

   26. GTGarry Tan2:17

       Um, and I was getting out of web programming at exactly the wrong time.

   27. AAAnish Acharya2:22

       Because it felt like that was the smart thing to do.

   28. GTGarry Tan2:24

       It was, yeah, it was smart, and-

   29. AAAnish Acharya2:25

       [laughs]

   30. GTGarry Tan2:25

       ... you know, I mean, this is true now, um, you know, in, in front of all these people at Startup School.
3. 4:45 – 10:59

   The Culture of Silicon Valley: Finding the Fringe & Your People

   1. AAAnish Acharya4:45

      all made-

   2. GTGarry Tan4:45

      Yeah. That's right

   3. AAAnish Acharya4:45

      ... along the way.

   4. GTGarry Tan4:46

      Yeah.

   5. AAAnish Acharya4:47

      I mean, maybe talk about the culture of Silicon Valley today. You know, on one hand, maybe it was despondent 20 years ago. It's crazy to think that was just 23 years ago that you're talking about. Um, now you think we would all be incredibly elated, yet there's this undercurrent, maybe it's the LARPing thing that you mentioned, you know, it's being outside of the light cone, the permanent underclass. Where does that come from? Do you think there's any truth in it? Talk through that.

   6. GTGarry Tan5:11

      I mean, I get it. Like, there's just more competition than ever. And then, um, I think maybe it's just that we're spending too much time in distribution. I think Chris Dixon talks about this all the time-

   7. AAAnish Acharya5:25

      Yeah, yeah

   8. GTGarry Tan5:25

      ... where, um, you know, you're in the idea maze, but, you know, the cool parts of the idea maze are, like, the weird fringe parts, actually.

   9. AAAnish Acharya5:31

      Yeah.

   10. GTGarry Tan5:31

       It's always like, uh, what is a toy? Like, and that's the pattern in Silicon Valley over and over again that bears repeating, right?

   11. AAAnish Acharya5:38

       Yes.

   12. GTGarry Tan5:39

       Like, the idea that you could have a computer on every desk in every home-

   13. AAAnish Acharya5:43

       Yeah

   14. GTGarry Tan5:43

       ... uh, today is like, of course, like, that's what's come to pass.

   15. AAAnish Acharya5:47

       Yes.

   16. GTGarry Tan5:47

       But the people who created that thing, they were fringe weirdo punks, actually.

   17. AAAnish Acharya5:51

       Totally. Yes.

   18. GTGarry Tan5:53

       Uh, and it's-

   19. AAAnish Acharya5:53

       They were outsiders.

   20. GTGarry Tan5:54

       Yeah.

   21. AAAnish Acharya5:54

       I mean, Homebrew Computer Club was not where you were getting laid.

   22. GTGarry Tan5:56

       Yeah.

   23. AAAnish Acharya5:57

       That's right. [laughs]

   24. GTGarry Tan5:58

       [laughs] Yeah. You, you just were, like, following this weird obsession that-

   25. AAAnish Acharya6:02

       Yes

   26. GTGarry Tan6:02

       ... you know, actually I want my own computer.

   27. AAAnish Acharya6:05

       And I think Chris has said this too, I think a good measure of how interesting the thing is, is can you find the edge?

   28. GTGarry Tan6:10

       Mm-hmm.

   29. AAAnish Acharya6:10

       And for the most interesting things, you typically can't. Even if you find the edge of the conversation, you can't find the intellectual edge, which I think is a very good sign. I mean, I had this growing up in the '80s, maybe you had this too. There was sort of one way to be cool in grade school-

   30. GTGarry Tan6:24

       Mm-hmm
4. 10:59 – 13:54

   What YC Gets Right: A Birthright for Tech Outsiders

   1. AAAnish Acharya10:59

      right? The lessons are when your instincts are pulling you to work with a set of people. So, so maybe talk a little bit about, you know, we talk a lot about how Silicon Valley, the magic is it takes outsiders and turns them into insiders, and I feel like YC is the ultimate distillation of that. You know, you go from being two founders, perhaps from a foreign country, to two people with YC on their LinkedIn, and everybody now knows what that means.

   2. GTGarry Tan11:21

      Totally.

   3. AAAnish Acharya11:21

      So talk about how that happens. Is it a process? Is it a mindset? And how has it changed over the years?

   4. GTGarry Tan11:27

      Um, I mean, I have to credit Paul Graham and Jessica Livingston and the original founders, including Trevor Blackwell and Robert Morris. I mean, uh, what Paul and Jessica figured out was that there are lots and lots of really smart people who do not get access, um, to Silicon Valley. I mean, I think in the past you had to, like, go to the right schools, you had to know the right people, sort of, and if you didn't, you'd come to Silicon Valley, and you'd try to get invited to the right parties. There was, like, a social network that you had to crack.

   5. AAAnish Acharya11:59

      [laughs]

   6. GTGarry Tan11:59

      Um, and then, uh, at YC we have a pejorative for that whole scene, which it's, like, we call them scenesters, actually.

   7. AAAnish Acharya12:07

      Okay, okay.

   8. GTGarry Tan12:07

      Uh, which is, I mean, it's sort of a nod to, like, what you used to have to do. Um, and some people still believe that's, like, how you do it.

   9. AAAnish Acharya12:14

      Mm-hmm.

   10. GTGarry Tan12:15

       All they did was say, "You know what? Like, here's a, here's a website. Here's 12 questions." Later, we added, like, a one-minute video.

   11. AAAnish Acharya12:22

       Yeah.

   12. GTGarry Tan12:23

       And, uh, people on the other end who are builders are going to try to give you a fair shake. And so today we have, uh, you know, 16 equal partners, including myself. And all of us, like, went through YC and had successful outcomes and, um, I think there's something valuable to that, that just, you know, welcome to s- welcome to Silicon Valley. We just, um, celebrated, uh, 7,000 people coming to San Francisco, the majority of them, the very first time they came. It's, you know, our startup school every single year. Um-

   13. AAAnish Acharya12:54

       It was so incredible

   14. GTGarry Tan12:55

       ... we kind of call it, like, a birthright, but for tech. [laughs]

   15. AAAnish Acharya12:57

       It looked like a Roman coliseum, dude.

   16. GTGarry Tan12:59

       Yeah, yeah, yeah, yeah.

   17. AAAnish Acharya13:00

       [laughs] It was amazing.

   18. GTGarry Tan13:00

       Yeah. So I, you know, I think that that's the coolest thing about Silicon Valley.

   19. AAAnish Acharya13:04

       Yeah.

   20. GTGarry Tan13:04

       Like, you sort of... It sort of doesn't matter where you're from or who you are.

   21. AAAnish Acharya13:08

       Yeah.

   22. GTGarry Tan13:08

       It just actually matters what your ideas are, and can you execute? The on- only other thing that I'd say, like, YC people get that a lot of people don't get is, um, a community of people who-

   23. AAAnish Acharya13:19

       Right

   24. GTGarry Tan13:19

       ... you can be super real with.

   25. AAAnish Acharya13:20

       Right.

   26. GTGarry Tan13:20

       It's like if you go to a TechCrunch event, you look to the left, you look to the right, and it's like-

   27. AAAnish Acharya13:25

       Yeah

   28. GTGarry Tan13:25

       ... are you really gonna be able to tell them anything real? Like, you're gonna hold your cards pretty close, and, uh, you know, you're gonna say, "How you doing?" And they're gonna say, "Killing it."

   29. AAAnish Acharya13:34

       Yeah.

   30. GTGarry Tan13:34

       And then they're gonna ask back.
5. 13:54 – 21:05

   Solo Founders, Vibe Coding & Founders Being 400x Themselves

   1. AAAnish Acharya13:54

      you know?

   2. GTGarry Tan13:54

      Yeah.

   3. AAAnish Acharya13:54

      Talk about the historical upside sort of surprises of, like, founder types, business types, and the downside surprises. What have you learned through YC?

   4. GTGarry Tan14:03

      I mean, what's cr- we're sort of in the middle of a big transformation now, I think.

   5. AAAnish Acharya14:07

      Okay.

   6. GTGarry Tan14:07

      Um, for the longest time, we really, really believed that you needed co-founders, and I think that everything else being equal, having co-founders is net really, really good.

   7. AAAnish Acharya14:16

      Totally.

   8. GTGarry Tan14:17

      Um, I think some of this is, like, the beginning of a cult.

   9. AAAnish Acharya14:20

      Yeah.

   10. GTGarry Tan14:20

       Like, you know, a cult where there's only one person, and that person can't convince anyone else to join them is, like, you know-

   11. AAAnish Acharya14:28

       Worse than-

   12. GTGarry Tan14:28

       ... worse than a cult.

   13. AAAnish Acharya14:29

       Yes. [laughs]

   14. GTGarry Tan14:29

       It's just, like, a crazy person, actually.

   15. AAAnish Acharya14:30

       Right, right, right, right. [laughs]

   16. GTGarry Tan14:31

       Um, I think this is, like, an old Derek Sivers TED Talk.

   17. AAAnish Acharya14:34

       Okay.

   18. GTGarry Tan14:34

       Right?

   19. AAAnish Acharya14:34

       Okay.

   20. GTGarry Tan14:34

       Where it's like, uh, yeah, one person dancing in a field is a crazy person.

   21. AAAnish Acharya14:39

       Yes.

   22. GTGarry Tan14:39

       But, um, you know, it's really the second person-

   23. AAAnish Acharya14:42

       [laughs]

   24. GTGarry Tan14:43

       ... who, like, gets up-

   25. AAAnish Acharya14:44

       That's right

   26. GTGarry Tan14:45

       ... and joins that person. Like, then you j- what you find is l- you know, suddenly 10 people or they're-

   27. AAAnish Acharya14:51

       It's a party

   28. GTGarry Tan14:51

       ... all dancing. Then it's a party.

   29. AAAnish Acharya14:52

       [laughs]

   30. GTGarry Tan14:52

       Yeah.
6. 21:05 – 23:03

   Business Loops: Skillifying Every Task Into a Markdown File

   1. AAAnish Acharya21:05

      Amazing.

   2. GTGarry Tan21:05

      That's the hope, anyway.

   3. AAAnish Acharya21:06

      Yeah. That is the hope.

   4. GTGarry Tan21:07

      Yeah.

   5. AAAnish Acharya21:08

      Talk a little bit about loops. I think there are kind of coding loops, um, which are well-discussed, and now the graphs thing, which is sort of, you know, uh, both funny and, and also real, but it feels like the most interesting loops are business loops.

   6. GTGarry Tan21:21

      Oh, yeah.

   7. AAAnish Acharya21:21

      Like, how can you actually have a loop where the output is changing the way the business itself operates? I know that Tom gave this great talk about how to organize your company in that way.

   8. GTGarry Tan21:29

      Oh, big brain. Yeah.

   9. AAAnish Acharya21:29

      Talk about how YC is organizing around loops and how you're advising your companies to.

   10. GTGarry Tan21:33

       Yeah, absolutely. I mean, I mess around with OpenClaw. I created GBrain, which is kinda like SOTA retrieval.

   11. AAAnish Acharya21:41

       Mm-hmm.

   12. GTGarry Tan21:41

       Um, I basically distil- I-- earlier this year, I created garryslist.org, which is sort of like actually an agentic newsroom.

   13. AAAnish Acharya21:47

       Mm-hmm.

   14. GTGarry Tan21:47

       And then I had to create, like, really, really good RAG around, like, you know, uh, personal deep research just to be able to put together some of those articles.

   15. AAAnish Acharya21:56

       Mm-hmm.

   16. GTGarry Tan21:56

       Uh, and then all I did was, like, I had OpenClaw look at that repo, extract out, like, the agentic RAG part, especially around, like, um, like, after-the-fact, uh, graph generation.

   17. AAAnish Acharya22:08

       Mm-hmm.

   18. GTGarry Tan22:08

       Um, and then it just ended up being SOTA even against, like, MemoryPalace and a bunch of other open source.

   19. AAAnish Acharya22:14

       Incredible.

   20. GTGarry Tan22:15

       Uh, and then what I realized is that just allows the agent to know what million tokens to keep in its head for any given task.

   21. AAAnish Acharya22:23

       Yes.

   22. GTGarry Tan22:24

       And so, um, what's interesting is, like, I think there are a couple things at work right now, like, um, the frontier model labs that are y- you know, letting you use ChatGPT or Claude. I think that they're still pretty constrained on cost and pretty constrained on, um, amount of compute they wanna give you.

   23. AAAnish Acharya22:44

       Mm-hmm.

   24. GTGarry Tan22:44

       And so if you really want to Token Max, you actually have to use something like Hermes Agent or OpenClaw.

   25. AAAnish Acharya22:50

       Mm-hmm.

   26. GTGarry Tan22:50

       Um, and then you have to, like, tune it all the way up. Like, you're just like, "Give me-- Like, let me load a million tokens or 800,000 tokens in-

   27. AAAnish Acharya22:59

       Yeah

   28. GTGarry Tan22:59

       ... to any given request, and, like, have that be in your, like, sol.md."

7. 23:03 – 28:09

   Token Maxing: How to Live in 2028 Today

   1. AAAnish Acharya23:03

      Yeah.

   2. GTGarry Tan23:03

      But when you do that, like, I think that you basically get to live in, uh, 2028. Like, you know, it costs, I don't know, 50 or $100,000 a year to, like, use the agents at full strength, like full 150 IQ-

   3. AAAnish Acharya23:18

      For every request, for every interaction

   4. GTGarry Tan23:19

      ... like, for every request.

   5. AAAnish Acharya23:20

      Mm-hmm.

   6. GTGarry Tan23:20

      Um, but you'll feel it more or less immediately.

   7. AAAnish Acharya23:23

      Yeah.

   8. GTGarry Tan23:23

      Uh, it just costs, like, a crazy amount. But, um, for a CEO or for a founder, uh, it actually makes a lot of sense to do that, and you have to give yourself permission to Token Max in that way. Um, but, you know, and then what you get is, like, I get-- you get to live in 2028 today.

   9. AAAnish Acharya23:40

      Amazing.

   10. GTGarry Tan23:41

       Um, but I think, like, going back to what founders can do, um, if you live that way, uh, not only do you wanna do that task, you wanna, like, skillify it.

   11. AAAnish Acharya23:51

       Mm-hmm.

   12. GTGarry Tan23:52

       Like, you, you do some, like, feat of strength, and then you turn it into a markdown file plus code plus tests-

   13. AAAnish Acharya23:58

       Mm-hmm

   14. GTGarry Tan23:59

       ... that can be reused and, like, put into a cron job.

   15. AAAnish Acharya24:02

       Yes.

   16. GTGarry Tan24:02

       And so, uh, I think we're just seeing, like, across the board, um, a markdown file is an employee. [chuckles]

   17. AAAnish Acharya24:11

       Yeah.

   18. GTGarry Tan24:11

       And it's an employee that, uh, will do the job perfectly every single time.

   19. AAAnish Acharya24:15

       Yeah.

   20. GTGarry Tan24:15

       And it'll do it, like, as many times as you want.

   21. AAAnish Acharya24:17

       Yeah.

   22. GTGarry Tan24:18

       And then, you know, at that point, all you're doing is, like, take any business process that you need in your company-

   23. AAAnish Acharya24:24

       Mm-hmm

   24. GTGarry Tan24:24

       ... and you just do it once perfectly. And the first time you try to do it, it's gonna be bad.

   25. AAAnish Acharya24:28

       Expensive.

   26. GTGarry Tan24:29

       Yeah.

   27. AAAnish Acharya24:29

       Yes, many iterations.

   28. GTGarry Tan24:30

       Yeah. But then at the end, you can-- you-- as long as you can- Tell very, very quickly and in very simple ways, like, "Hey, agent, fix this, fix that."

   29. AAAnish Acharya24:40

       Yeah.

   30. GTGarry Tan24:40

       Like, "This was wrong." Um, your- the actual trace and the actual, uh, history-
8. 28:09 – 33:10

   Constructive Conflict & Pedro's Meeting-Transcript Agent

   1. AAAnish Acharya28:09

      So one of the things I'm really interested in is conflict, right? Most people at an enterprise would say conflict is a bad thing, but I actually think that there's this sort of productive, constructive conflict that at its best is the search for truth.

   2. GTGarry Tan28:22

      Yeah.

   3. AAAnish Acharya28:22

      These systems allow us to do things like explore conflicting approaches-

   4. GTGarry Tan28:26

      Okay

   5. AAAnish Acharya28:26

      ... and totally unbundle it from the human emotion.

   6. GTGarry Tan28:28

      Yeah.

   7. AAAnish Acharya28:29

      And I think that's actually very important because this is why the sort of the decentralization in an organization or an economy works well-

   8. GTGarry Tan28:36

      Oh, yeah

   9. AAAnish Acharya28:36

      ... because every idea gets tried, you know?

   10. GTGarry Tan28:38

       Yeah. Uh, do you know, uh, Pedro from Brex is, like, the most aggressive person I've met who does this with his exec team?

   11. AAAnish Acharya28:45

       No.

   12. GTGarry Tan28:46

       So, which is cool. Like, he uses-- Like, he actually was one of the people who, like, really got me into OpenClaw.

   13. AAAnish Acharya28:51

       Okay.

   14. GTGarry Tan28:51

       Um, because he just came in, and, you know, he had lunch with us at YC.

   15. AAAnish Acharya28:56

       Yeah.

   16. GTGarry Tan28:56

       And he just walked us through, um... You know, the, the vibe around OpenClaw was that it was too unsafe and-

   17. AAAnish Acharya29:02

       Right

   18. GTGarry Tan29:02

       ... dangerous, and like delete your whole computer, and stuff like that.

   19. AAAnish Acharya29:05

       Oh.

   20. GTGarry Tan29:05

       Uh, what he ended up doing was, like, building an entire open source layer, which he actually has open source, called Crab Trap-

   21. AAAnish Acharya29:12

       Okay

   22. GTGarry Tan29:12

       ... that is, like, an agent that watches all the network traffic and all the things that, um, OpenClaw does.

   23. AAAnish Acharya29:18

       Nice.

   24. GTGarry Tan29:19

       And then that's how he squares the circle. Like, Brex is one of the most regulated, you know, fintechs-

   25. AAAnish Acharya29:24

       Of course

   26. GTGarry Tan29:24

       ... of course.

   27. AAAnish Acharya29:24

       Yeah.

   28. GTGarry Tan29:25

       Um, and then the upshot of what he got is that, um, he has his agent look at the meeting transcripts from, like, all of his direct reports, and he, like, can see that, like, two levels down.

   29. AAAnish Acharya29:38

       Yeah.

   30. GTGarry Tan29:38

       And then the thing that really jumped out at me is, like, he can use, uh, meetings that he's not in-
9. 33:10 – 39:08

   The Torture of the White-Collar Job & Life Above the API Line

   1. AAAnish Acharya33:10

      Maybe another way to say this is just this is the least happy you'll ever be.

   2. GTGarry Tan33:13

      Yeah.

   3. AAAnish Acharya33:13

      You know? Because I think it, like, one of the other implications of what you're describing is just the ergonomics of the sort of white-collar job today. You know, you think about ev- you've worked at big companies. Like, you think of just the kind of morass that you had-

   4. GTGarry Tan33:26

      Yeah

   5. AAAnish Acharya33:26

      ... to get through.

   6. GTGarry Tan33:27

      It's torture.

   7. AAAnish Acharya33:27

      You think of how lossy the communication is, like the entire management science sort of managerial class, like maybe that was a 100-year distraction from what was founder mode, you know, in, in, in the Industrial Revolution, all the way through Henry Ford, all the way through some of the early Silicon Valley founders, and then everybody just kind of hit this tidal wave of management.

   8. GTGarry Tan33:50

      Oh, yeah.

   9. AAAnish Acharya33:50

      Now we have this new model. It can lead to dramatic productivity improvements, like three days a week of really intentional work, maybe one day.

   10. GTGarry Tan33:58

       Yeah.

   11. AAAnish Acharya33:58

       And the other days you're touching grass, and I think the time that you do spend is just so much less burdened by this stuff that you always felt like you had to wade through.

   12. GTGarry Tan34:07

       Yeah, that's right. I mean, if you look at the typical Mag Seven day-to-day, it's insane.

   13. AAAnish Acharya34:13

       Yes.

   14. GTGarry Tan34:13

       Like, I remember when I was 22 working on Windows Mobile, uh, we were working on, you know, these crazy devices that were, like, years before the iPhone.

   15. AAAnish Acharya34:21

       Yeah.

   16. GTGarry Tan34:22

       And, um, we needed integration, for instance, from the Windows team, which was like, you know, on the other side of the highway in Redmond.

   17. AAAnish Acharya34:29

       Right.

   18. GTGarry Tan34:30

       And then I remember-

   19. AAAnish Acharya34:31

       That's right

   20. GTGarry Tan34:31

       ... you know, me and my sort of PM mentor, uh, we had to go over to the, to the Windows team, and, like, they wouldn't reply to our emails. They wouldn't fix our bugs. They wouldn't even, like, mark it as won't fix. They just ignored us.

   21. AAAnish Acharya34:44

       Yeah.

   22. GTGarry Tan34:44

       And we had to go over there, uh, with a baseball bat.

   23. AAAnish Acharya34:47

       Yeah.

   24. GTGarry Tan34:47

       Like, I don't know why he brought one. Like, it's wasn't like he was-

   25. AAAnish Acharya34:50

       You kidding?

   26. GTGarry Tan34:50

       ... gonna, like, rough anyone up. It was just, like, this vague threat or something. [laughs]

   27. AAAnish Acharya34:54

       Incredible.

   28. GTGarry Tan34:54

       It was actually hilarious, and I think about that all the time. It's like what kind of, like, bureaucratic hell, uh, is it where, like, two perfectly smart people who are making, like, good tech money-

   29. AAAnish Acharya35:07

       Yeah

   30. GTGarry Tan35:08

       ... you know, are put into a situation where, like, there's a fiefdom over here. Like, literally there's, like, seven levels between us and that person.
10. 39:08 – 41:33

    The Real White Pill: Everything Is Slower Than You Think

    1. GTGarry Tan39:08

       think will happen

    2. AAAnish Acharya39:08

       Which is so interesting 'cause what's traditionally seen as a hierarchical society came up with this very bottom-up-

    3. GTGarry Tan39:13

       Beautiful

    4. AAAnish Acharya39:13

       ... way of working which led to, like, arguably a 40-year generational run-

    5. GTGarry Tan39:17

       Oh, yeah

    6. AAAnish Acharya39:17

       ... because of process power

    7. GTGarry Tan39:18

       Yeah, which is beautiful

    8. AAAnish Acharya39:21

       How, how many... As you think about kind of our evolution towards this vision, how many of these problems do you think are intelligence-bound, like we need smarter models, versus there will be other attributes of models that will unlock success?

    9. GTGarry Tan39:34

       I mean, honestly, it's all human, uh... Humans are the ones who are gonna, like, slow this down. And then, um, actually, I'm realizing this might be a white pill

    10. AAAnish Acharya39:43

        Okay

    11. GTGarry Tan39:44

        So if you look at, like, the tenor of what's going on just broadly in the discourse about AI, it's all, like, this sort of doom and gloom, like all white-collar jobs are gonna go away. There's gonna be a permanent underclass

    12. AAAnish Acharya39:56

        Yes

    13. GTGarry Tan39:57

        And then, uh, funny enough, what I think is, um, all of the bureaucracy and the slowness and the seven plus or minus two, uh, of, like, every company, every institution, every organization, all the middle managers in the world-

    14. AAAnish Acharya40:13

        Yeah

    15. GTGarry Tan40:14

        ... that is actually the white pill [laughs]

    16. AAAnish Acharya40:15

        100%

    17. GTGarry Tan40:16

        So, you know, these things are gonna be slower than you think-

    18. AAAnish Acharya40:21

        Yes

    19. GTGarry Tan40:21

        ... right? Uh, I looked at, you know, we had 7,000 people. Most of them were 18 to 22 years old

    20. AAAnish Acharya40:27

        Amazing

    21. GTGarry Tan40:28

        And I looked in their faces, and, like, they're all AI native. Like-

    22. AAAnish Acharya40:31

        Yes

    23. GTGarry Tan40:31

        ... they were born more or less, like, you know, be- they became adults in the time of ChatGPT

    24. AAAnish Acharya40:37

        Yes

    25. GTGarry Tan40:38

        Uh, and it's gonna... You know, and for us it's like now we, you know, run shit, which-

    26. AAAnish Acharya40:42

        Yeah

    27. GTGarry Tan40:42

        ... is nice

    28. AAAnish Acharya40:43

        That's great, right?

    29. GTGarry Tan40:43

        Uh, but, you know, imagine us, like, 20 years ago.

    30. AAAnish Acharya40:46

        Yes
11. 41:33 – 44:44

    What the Next Computer Looks Like: Voice, Memory & the Harness Wars

    1. GTGarry Tan41:33

       you know? [laughs]

    2. AAAnish Acharya41:33

       They won't. I love it. [gentle music] Okay, I have two more questions for you. Um, you know, one on this ki- sort of trying to envision the near future, what do you think the next computer looks like? What does a new computer look like?

    3. GTGarry Tan41:46

       Oh, man. I think in the short term it'll probably keep looking like the way it is, but it seems inconceivable that, um, it'll stay that form factor. Like, it's pretty clearly going to be voice

    4. AAAnish Acharya41:57

       Yes

    5. GTGarry Tan41:58

       I think there's a lot around memory. Um-

    6. AAAnish Acharya42:01

       Amazing

    7. GTGarry Tan42:01

       ... it's gonna take a while to... You know, to- we're, we're at this moment where everyone's sort of happy with ChatGPT or Claude Cowork or whatever it is

    8. AAAnish Acharya42:08

       Yeah

    9. GTGarry Tan42:09

       Um, but I don't think they will be in the long term. Like, um, I think you're gonna want a lot more context, and you're gonna want, like, all watched over by machines of loving grace. Like-

    10. AAAnish Acharya42:19

        Yes

    11. GTGarry Tan42:20

        ... you're gonna want this thing that, uh, is benevolent, knows about, like, your hopes, fears, and desires-

    12. AAAnish Acharya42:26

        Yes

    13. GTGarry Tan42:26

        ... and then is sort of, like, relentlessly searching to help you in that way, and that's actually gonna be really exciting. And, you know, the path through that will probably be, like, computer use, memory-

    14. AAAnish Acharya42:37

        Yeah

    15. GTGarry Tan42:37

        ... like, uh, diarization, like ingestion, like-

    16. AAAnish Acharya42:41

        Yes

    17. GTGarry Tan42:41

        ... um, yeah. I mean, I think there's gonna be a lot there. And, uh, you know, the cool thing is it's actually gonna take longer. Like, this isn't a... Like, I don't think people have even talking about this now. My guess is, like, next year, 2027, will be like the harness wars

    18. AAAnish Acharya42:58

        Yes

    19. GTGarry Tan42:58

        You know, and then, uh, but it's really a cost thing. Like, it's just too expensive right now.

    20. AAAnish Acharya43:02

        Yes.

    21. GTGarry Tan43:02

        You know? Um, but it won't be. It, like, there will be a war for a billion consumers-

    22. AAAnish Acharya43:07

        Yes

    23. GTGarry Tan43:08

        ... uh, right when, like, the frontier model compute of today will be, like, maybe 50 or 100 bucks in, like, two or three years.

    24. AAAnish Acharya43:17

        Yeah. Yeah.

    25. GTGarry Tan43:18

        And then that will be when, like, the most vi- the browser wars will be back on.

    26. AAAnish Acharya43:22

        Yes, yes. Well, we're already seeing it. Like, the price per token of Frontier is going to infinity. Like, Mythos is infinity dollars a token. You can't have it, and the price per token of the model that was Frontier a week ago is collapsing to zero.

    27. GTGarry Tan43:34

        Yeah.

    28. AAAnish Acharya43:34

        And I actually think to me that is a huge white pill for a lot of reasons. You know, one of the most important of which is just we need more consumer AI software, and in a world where most consumer products are free to try, you just can't do that when you have really high costs of incremental distribution. The magic of software is zero cost incremental distribution. We have these magical primitives. Like, we have computers that are emotive and have emotional texture for the first time in 40 or 50 years.

    29. GTGarry Tan44:00

        Oh, yeah.

    30. AAAnish Acharya44:00

        What a tragedy that we haven't built more ambitious consumer products yet.
12. 44:44 – 51:27

    Local Politics & Why San Francisco Turned

    1. GTGarry Tan44:44

       try it again.

    2. AAAnish Acharya44:44

       They've done a great job.

    3. GTGarry Tan44:45

       Yeah.

    4. AAAnish Acharya44:46

       Okay, last question. Um, you know, we talked a lot about kind of utopian visions for the world and problems that are maybe intelligence-bound. You've also put a lot of energy into sort of local politics and doing things I think are very pro-social, but it seems like more of those problems are human coordination-bound.

    5. GTGarry Tan45:02

       Oh, definitely.

    6. AAAnish Acharya45:02

       Yeah. How do you get energy to do that work? You know, is there a white pill for that aspect of our society? Maybe talk a bit about that.

    7. GTGarry Tan45:09

       I mean, the white pill is just believing that, uh, America still has rule of law and that, um, people in government are trying to help and, like, somewhat earnest. Like, it's not perfect.

    8. AAAnish Acharya45:20

       Mm-hmm.

    9. GTGarry Tan45:20

       Um, I think that the white pill for me recently was even just doing jury duty in San Francisco and to see, like, this orderly process of really a cross-section of San Francisco. Uh, one thing I realized was, like, actually we are a big part of, like, the city. Like, we're 10, 20% of the people are, like, directly working in tech.

    10. AAAnish Acharya45:42

        Yeah.

    11. GTGarry Tan45:42

        And so, um, to see something that was... I mean, the judge was really good. I mean, it was a horrible case but, like, you know, government is not here to, like, ruin people. Like-

    12. AAAnish Acharya45:55

        Yeah

    13. GTGarry Tan45:55

        ... and so, and then at that point, um, we can vote.

    14. AAAnish Acharya46:00

        Yes.

    15. GTGarry Tan46:00

        Like, we actually have a say. Um, and then I think everything is downstream of media at that point. Like, I think the San Francisco moment that got me very, uh, engaged was realizing, you know, Asian Americans are 25% of the voting populace, about 30% of the population, but during COVID, like, we had a district attorney who was turning a blind eye to some of the most heinous crimes-

    16. AAAnish Acharya46:24

        Heinous

    17. GTGarry Tan46:24

        ... against our elders.

    18. AAAnish Acharya46:25

        Mm-hmm.

    19. GTGarry Tan46:26

        And simultaneous to that we had a school board that was openly hostile against Asian American kids who just wanted to learn algebra in public middle school.

    20. AAAnish Acharya46:35

        Mm-hmm.

    21. GTGarry Tan46:35

        Which hit me because, uh, I was that public school kid in Fremont in the East Bay. Like, if I couldn't have taken algebra in middle school, I couldn't take calculus by senior year, I wouldn't have been able to study engineering at Stanford, and I wouldn't be here doing any of the things I get to do today. And I'm like, "Oh, okay. Now, now they've made it personal."

    22. AAAnish Acharya46:54

        Yeah.

    23. GTGarry Tan46:54

        And like, let me understand. Uh, and it turned out to be, like, sort of all the same people, the same people who wanna destroy all data centers.

    24. AAAnish Acharya47:01

        Yeah.

    25. GTGarry Tan47:01

        The same people who, um, you know, don't want to build any housing-

    26. AAAnish Acharya47:05

        Yeah

    27. GTGarry Tan47:05

        ... in, uh, anywhere in California or the world.

    28. AAAnish Acharya47:08

        Yes.

    29. GTGarry Tan47:08

        Like, it's sort of NIMBYs who want to virtue signal, who, um, sort of have these ideologies that, like, don't square with mine. I mean, as a builder you're like, well, the, you know, words have meaning, like, s- supply and demand exists, markets exist, like, incentives exist, and then, you know, I think that government is whe- you know, when people like us don't speak up, uh, when we don't, like, pay attention and, like, hold our media accountable, not just the government-

    30. AAAnish Acharya47:38

        Yes

Episode duration: 51:27