Garry Tan says this YC batch has more defense companies than all of last year, and the era of cost-plus procurement is finally ending

Speaker 2: Right away.

Speaker 1: As he comes in. Gary, how are you doing? Welcome to the show. He's now here. It's demo day. Watch out. It's flashbang. Great to see you. Great to see you. How are you doing? Take us through it.

Speaker 2: How's the day been? Too long.

Speaker 14: Oh, it's awesome. I mean, it's always the big show here in San Francisco. Yes. Mean, it's a beautiful gorgeous day and, you know, who doesn't like to go check out a nuclear reactor in our back parking lot? No uranium, don't worry.

Speaker 1: Woah. This is like the second or third nuclear company though. Right?

Speaker 14: I mean, I think we need we kinda need all of them.

Speaker 2: Right? Let's do all

Speaker 14: of them.

Speaker 2: What's stopping them from going critical in the in the parking lot?

Speaker 14: Oh, well, luckily, no uranium. No uranium. No geyser counter needed. You're all good.

Speaker 1: That's great. Is how much is hard tech booming? Do you have any, like, stats? It feels like every year, there's a slight slightly more people taking the the hardware isn't that hard. They say hardware is hard, but I'm gonna try it. And it feels like that's been a trend that's been growing steadily. Am I am I correct to read that into the trends over demo days over the last couple years?

Speaker 14: Yeah, man. I gotta call out my buddy Brett Gibson over at Initialized Capital. I mean, right when CodeGen first started happening, Brett was my first friend who said, you know, obviously, you gotta, you know, fund the bangers and it's not like they stopped doing b to b SaaS.

Speaker 9: Yeah.

Speaker 14: But, you know, Initializ was one of the first people to do the pivot. Yeah. To say, know what? Like, hard tech is the way it's gonna go and, you know, things like flock safety were sort of the first things we did that really really became huge. And, you know, computer vision is in. And then now with AI, like, everything is open. It's so much easier. I mean Yeah. You can do mechanical engineering. You could do electrical engineering. If you don't understand it, you can build it. You can find every vendor. Like, the supply chain management can happen right there. All in OpenClaw now. Sure. I mean, there's I mean, we're just in a new age.

Speaker 2: Yeah. I mean, we've even seen people like like optimizing their home Wi Fi speed Oh, yeah. Like, you know, different

Speaker 1: It's crazy. A lot of the hard tech boom, I feel like, has been driven by Elon, Palmer Lucky, just sort of breaking the glass on, like, it's possible for this to happen. And that's led to a lot of, like, defense tech startups. But I'm interested in the intersection. What are you seeing, and how much how much conversation is there around hardware and AI, data centers, stuff that's in the token path but not in the direct path, not one shot able by a by a frontier model?

Speaker 14: Oh my god. I mean, honestly, you should probably get Ash. Ashton Kutcher's here today. You should go talk to him. He's got a whole thesis around AI, solar, you know. He's actually pretty bearish on fusion, which is interesting. You should ask him about that.

Speaker 5: Okay.

Speaker 14: Yeah. I think

Speaker 2: That's that's news. That's gonna Yeah. Ashton Kutcher, bearish on Fusion. We love Ashton.

Speaker 14: He is like really He went all the way like, he met everyone. So Yeah. I mean, Ashton has been an incredible investor, especially for Hard Tech. But anyway, I mean, I think what we've been seeing, there's, like, more there's more defense here in this batch than the entire last year.

Speaker 2: Yeah.

Speaker 14: And then

Speaker 1: But there's also more geopolitical conflict. So there's more energy to purchase and the government's sort of changed a lot of the ways stuff is procured. And so the old adage Yeah.

Speaker 2: The Carps app. Software singularity Yeah. And then hardware Who

Speaker 14: am gonna hand it up to? Who? So it's Emil Michael and Oh, it's our friends who are actually in the Trump administration Sure. Who are like changing everything like the DIU, like it's a new department of war. Like this is never you know, I think that there is actually an awareness that cost plus you know, cost plus was originally designed to be something that created more innovation. And, you know, here we are fifty, eighty years later, and what was needed for fifty or eighty years ago, it's the wrong thing now. You have the defense primes that are just not actually innovating. They can't. Like, it's sort of structural. There's sort of these, you know, it's the equivalent of big tech bureaucracy. Like, I think, you know, I think of iPhone, how stupid Siri is and how terrible it is. Or even like Alexa over at Amazon. Like unbelievable. Shot fired. How incompetent. Yeah. I mean, these are very competent people put into a system that is structurally unable to deliver technology that is actually what people really, really want. And so when you talk about consumer technology, like, you know, it's not that big a deal. Like, you know, my life is 20% worse, 10% worse because I can't use a Siri that actually uses large language models properly, you know, years after it was designed. And I usually you know, when I go to DC, I like to tell politicians that. It's like, hey, like, here's this thing that doesn't make sense in society. But, you know, it's consumer tech. It's it's no one's gonna really like probably very few people have died because of it. But when I really what makes me mad though is when you think about like there are service members who are going to die in in service of our country

Speaker 1: Mhmm.

Speaker 14: To defend liberty, and they're giving their lives because defense primes are not doing the job that they're supposed to do. Sure. Right? They're not innovating. They're not bringing these new technologies to fore, and it's structural. And it's not like individual people trying to do it. It's structural. Right? Like Yeah. Just as the people who work on Siri like, actually, we funded a number of them who have come in D NYC. Delightful, wonderful people. Like, the people who work at these companies are incredible. Yeah. I never want to attack engineers working on this stuff. It's actually structure.

Speaker 1: Yeah.

Speaker 14: Like, you have big tech and you have, you know, big anything.

Speaker 1: Yeah.

Speaker 14: Right? It becomes structural where here a thing that should happen, it does not happen. Why? You know, I think cost plus is one of those Yeah. Like, it's just systematically, instead of trying to create something that's better, it's like how do we maximize shareholder value? Yeah. You know, I get it, but also, why are we doing it this way? And so, you know, credit to the Department of War. Like, they are actually, for the first time in, like, decades easily. Like, they're actually open to a team of 10 or 20 people, whether it's, you know, in Boston, in Austin, even down the street in San Francisco, in in Gundo. Shout out to our Gundo bros out there.

Speaker 1: Yeah. We talked to Dan Durskel about just, like, having range days where any startup can come and pitch and they'll just buy the best product right there. It feels like it's an entirely new day in terms of procurement. Switching gears to software is like, I could prompt that with one with one prompt. I could build that or it's just directly in the path of the labs. It's so easy to build. Is that the new, like, what if Google builds it? Like, I I I have a suspicion that we might have over rotated on the on the, oh, like, it's easy to build that particular product or vibe code that particular product. And it's like, yeah, but you didn't. And then this team did and they went and did all the other hard things to get customers and understand the user and actually build a great product. And so, yes, you could clone it, but they have momentum. And so maybe they won't be a trillion dollar business, but there might be some beautiful businesses that are accelerated and they look like SaaS and they look like they could be cloned, but there's something secret there that's still going on. How do you think about that? Do you push back or do you agree?

Speaker 14: No. I think that, you know what, like, being in San Francisco and being friends with people at the labs Yeah. And being able to see, like, the inside, like, you know, being able to see the inside of, like, you know, what Codex and ClaudeCode is going to be incredible. Right? And then these are all 200 IQ people building 200 IQ models. And, you know, it's gonna keep going. I mean, it's it's gonna be, you know, true AGI. The takeoff is happening. Sure. And the thing you have to realize is like everywhere else, well, those 200 IQ people can't really interface the way that they need to. It's with the 150 IQ people and the one thirty IQ people and the 100 IQ people and humanity.

Speaker 1: Room temp. Need a room temp guy to build a great business. No. You talk to a lot of entrepreneurs and it's like, yeah, this person just worked really hard and found an interesting niche and just deliver it.

Speaker 2: Down to my level. Room temp.

Speaker 14: But like That's great. We're Yeah. Joking, but I on the, you know, on the other hand, like, the world actually cannot interface with these. Like, the models will be too complex.

Speaker 1: Totally.

Speaker 14: The UI will be too complex. Like, you know, the more complicated you make any given solution, the less likely someone's actually gonna be able to take advantage of it and implement it and use it. Yeah. And so, you know, that's where all of the value is in business. It's like, how do you take this new tech that nobody has, that very few people understand

Speaker 1: Yeah.

Speaker 14: And then bring it to them in a way that actually helps their lives. And Yeah. You know, I mean, obviously, on a five or ten year basis with super intelligence becoming what it is, you know, all bets could be off Yeah. On like a long enough time frame. But certainly for the next like eighteen, twenty four, thirty six months, that's not happening. And for now, hey, man, I we're gonna fight for the little guy and for, you know, the person who has nothing sitting in front of a recycled MacBook Pro to be able to make anything, to make prosperity happen. Like I love that. That's what that's what this logo is. Right?

Speaker 2: I love it. Yeah.

Speaker 14: That's what that's what we're about. Like, this

Speaker 1: is Yeah.

Speaker 14: You know, as long as this is a flag that flies, like, I I will, to my dying day, will fight

Speaker 2: I love it.

Speaker 14: For the startups and for startup founders and for people to be able to create something from nothing.

Speaker 2: Yeah.

Speaker 14: And, you know, I think it's gonna be a while before we really have to worry about, like, are there gonna be jobs and things like that.

Speaker 1: No. I completely agree.

Speaker 2: What out of the current, like, request for start ups is there any category that comes to mind? Are you working on a new on a new request for startups list?

Speaker 14: Yeah. Absolute I mean, we're always doing that. I don't know. I'm I'm trying to convince Ashton to do the, you know, the call for solar and re industrialize America with me, so maybe you'll see that. Joe Gebia, chief design officer Oh, of The United States, he he, you know, sent me the one about agriculture, you know, where I I think that we're realizing that, like, YC has a real convening power, and people really shouldn't have to just go and work on an you know, nothing wrong with b to b SaaS and we love it. Yeah. But also, like, you know, if you've got something in you that isn't that, like, we can actually try to harness, like, all that brain power Totally. And all that AI power into basically everything. I mean, that's my the the number one philosophy that I think people have not absorbed yet enough and that I want everyone to hear and I will shout from every hill is that it's time to boil the oceans, guys. Like, everyone else is out there trying to figure out like, oh, let me like lay off all these people and like, the machines are better than humans. I was like, you know, we live in a broken, completely just it's like a backward world. Right? Like, if we could accelerate technology fast enough, it would solve a lot more problems, people would have a lot more prosperity. And now, like, the biggest one of the biggest things, the blocker is not even money. It hasn't been money for some time. It's definitely not right now. Like, it's just pure agency. Like, do you have an idea? Like, can you prompt it?

Speaker 1: Like, do you actually do

Speaker 14: you actually know how to use this stuff? What do you want? Like, what do you want to manifest in the world? And if we do that, and we do that like 10 or a 100 times or a thousand times more than we have been, and we have you know, and then going back to what we were saying earlier, like, it's actually about the systems. Right? A startup is the defining system where either you do it or you don't. Like, either you have a customer that loves it or a customer that like doesn't buy it. And like most of the time, frankly, like people love to bag on startups, it's because those startups don't make something people want. Right? That's why our t shirt is literally make something people want. It's not more complicated than that. We actually give away the secrets on our freaking t shirts. A great t shirt that you get on your first day at YC. But like, it's surprising how hard it is to do that. And then but if we did that, you know, instead of trying to replace jobs, instead of trying to say like, you know, I could like lower my cost by 90%. Like, the thing is, that's why I love markets. Like, if you have you know, just to take it take like a big tech swing again, when you have a moat that is so unassailable, that is basically like regulatory capture. Like, you basically if the government says like, actually, Apple, you can just not have multiple app stores, then you know what the government just granted? The government is okay with a monopoly. Mhmm. And when you have a monopoly, you have no choice, and then that's what we have. Like Mhmm. We do not have choice about what messenger we use. Like, why is iMessage so stupid? If you had LLMs embedded in that, you would have so, you know, we could actually literally save time by having that level inside our phone. But I guarantee Personally,

Speaker 2: I like when I'm searching for like a very specific message that I know I sent and I can't find it. Personally, I just

Speaker 1: Where where is the bar for big tech now? Where where where is the bar for big tech? Because SpaceX is now a $3,000,000,000,000 company. If you're just hanging out of just 1,000,000,000,000, are you little tech now? Are are you a $100,000,000,000 little tech company?

Speaker 14: Toddler. I mean,

Speaker 1: where where where is the line?

Speaker 14: Honestly, it's like after people are like $10,000,000,000 valuation, I'm like, God bless you. Okay. Please don't please don't, you know Steep them all everyone. For the next yeah. Please don't Steep them all.

Speaker 1: Next generation. Decacorn is the graduation.

Speaker 14: That's my experience.

Speaker 1: Yeah. No. That makes sense.

Speaker 14: Become a Deckar Connor, you start hiring these Google PMs who come in and, you know, the same people who closed off the platforms on App Store, things like that. You start hiring those PMs and then it's over, guys, you know.

Speaker 1: Oh, wow.

Speaker 14: Which is fun. Like, as long as the thing is like, as long as there are markets, as long as there's choice.

Speaker 10: Yeah. Of course.

Speaker 14: There's gonna be a check on that. Right? And so that's that's my fractal. I'm like, hey, man.

Speaker 2: No. I've always been drooling the

Speaker 14: oceans. Agenda.

Speaker 1: I think it's

Speaker 14: Let's make so many more people. I mean, it's you wanna get political? I don't know. I I have a thing about that.

Speaker 1: Not particularly.

Speaker 14: Okay. No politics

Speaker 8: today. No

Speaker 14: politics today. I I No politics today.

Speaker 1: I would rather talk about gaming

Speaker 2: You're running. That's it. You're running. For president. For president of The United

Speaker 14: I'm Canadian. Sorry, guys.

Speaker 1: Oh, interesting. I I I wanna I I didn't know that for some reason.

Speaker 14: I was born in Winnipeg. I'm naturalized though. Yeah. You know, God bless America.

Speaker 1: Talk I I what what I actually wanna know about is G Stack. I wanna know Yeah. Of course. About G Stack in the context of LIC. I feel like YC, the whole goal, the whole system is built to allow the founders to talk to their customers, to build something people want. And that's why there's book face with investor reviews because you could see a founder getting distracted and being like, I need to optimize my fundraise, and so I should go build a database and interview everyone about what funds are the best. They don't need to because it's been done for every y c company And that's gone through the is there a lens where I can view some of the research, some of the work, some of the experiments that you're doing around vibe coding, around g stack as let me sort of like do the unhobbling but fall on the sword of like optimizing systems and experimenting so that that burden doesn't go on to every single y c founder that's token maxing

Speaker 3: Yeah. In the

Speaker 14: Absolutely. Okay. So g stack was sort of my project from like January through March. I'm still maintaining it

Speaker 1: Yeah.

Speaker 14: Fixing it. But like, my new thing is g brain, which I Yeah. Want it to be like the Postgres for agents.

Speaker 1: Oh, okay.

Speaker 14: So the thing I realized

Speaker 9: Yes.

Speaker 14: Is that like, you know, when an agent you know, a human can only keep like seven plus or minus three things in their head. Right? Like a phone number is about it. Right?

Speaker 2: That's generous.

Speaker 14: But yeah.

Speaker 1: I can keep hundreds hundreds of things, but yes, I understand

Speaker 14: That's very nice. Some of us are special. John, you're very special. And but a computer with an LLM, it can keep about three Harry Potter books in their head. And so That doesn't sound like person would say

Speaker 9: it that way.

Speaker 14: I mean, it's I mean, like a stack of

Speaker 1: better when you say like it it it can contain like 75 A million. Parrot ish problems or like some

Speaker 14: That's right.

Speaker 1: Complex math. Harry Potter and But like yes, I understand. The the the context window is is significant and and compared to a human.

Speaker 14: Yeah. And then when you think about like what most computer systems are, like, actually, you should think of like the Library of Alexandria.

Speaker 1: Yeah.

Speaker 14: Right? Like thousands of books, tens of maybe millions of books. Like actually, it's even bigger than that. It's like the whole Internet. Right? And then that's what like sort of like Karpathy's knowledge wikis are, people talking about company brains.

Speaker 1: Yeah.

Speaker 14: Like, you could basically take all of the relevant info about customers everything that you've ever any any person that like any one of the company's ever even met. Like, that's what a CRM is. Yeah. You can have that in like a 100,000 or a million markdown files that like comprises everything that that business is. Yeah. So that's basically what g Brain can do. Yeah.

Speaker 1: It's like People misunderstand that like YC has been building software to help batch participating companies for over a decade at this point, I Probably longer, maybe fifteen years. And so that is in

Speaker 14: the building that in.

Speaker 1: Yeah. Exactly.

Speaker 14: I mean, actually, I think, you know, now we have all these tools that are gonna be built directly into YC Bookface.

Speaker 1: So Yeah.

Speaker 14: You know, you're gonna wanna be a part of the community.

Speaker 1: Yeah. I've been seeing it's been getting better. Like, the the pace of the pace of software development on Bookface and YC's internal tools is accelerating. It's been Yeah. Refreshing to see that that it's people often have a window into it and they see like, oh, there's this demo, but there is an internal product that is evolving and getting better, which So is

Speaker 14: so the magic moment for is basically like being able to take any book that exists in your entire business and then making sure the three books that really matter for the thing you're trying to do are loaded. Yeah. And then that's basically ASI. You don't have to write software anymore. You can just straight up use Hermes Agent or OpenClaw plus this. I am trying to make it work with a Codex and Okay. You know, CoWork. But honestly, I think that those are like Hondas. And honestly, I mean, open man. Okay. It's still the Ferrari, but, you know.

Speaker 2: Open. Yeah. That's good.

Speaker 1: Yep. Very, very fun. What what are the trends that you're seeing in applications? Are applications increasing in the age of AI? More people wanting to do startups? Or are people coming from big tech?

Speaker 2: Applications as in like people

Speaker 1: Applications too wisely. People trying to start companies. Like feels like in all this anxiety of like big tech layoffs and stuff, like the flip side of that is that there's more entrepreneurial opportunity because we

Speaker 2: just pointing people to the YC website Yeah. When they're when they're maybe they're not a software developer and they decide to build their first app in the last six months Yep. Which there's been many. I'm just like, go follow this even if you're not gonna apply. Like, this is this is this is what works.

Speaker 14: Yeah. Did you see PG came out and said it? He's like, if you're nontechnical, just stop calling yourself that.

Speaker 1: Yeah.

Speaker 14: Because you should I mean, why why limit yourself, actually?

Speaker 1: I completely agree.

Speaker 14: I completely agree. I don't know. I think we're seeing

Speaker 1: Non technical was always like because obviously, like, it's not a binary. Like like, I learned Python and a little bit of c plus plus and I could, like, sort of deploy a website. But, like, I'm, like, levels below, like, a true, like, you know, incredible computer scientist who's, like, creating Python. And but I would just sort of like be like, yeah, okay. I'm technical because that's good, but it's like there's obviously levels to these things. And the more you can like, you know, get away from

Speaker 14: that If sit you sit in front of Cloud Code

Speaker 1: Yeah.

Speaker 14: I guarantee you you could fix some bugs in Python and submit a PR that would be accepted. Yeah. Everyone that wasn't true a year ago.

Speaker 2: Personally, I think John could solve continual learning.

Speaker 4: That would be great.

Speaker 14: You could do it. You just it's just a matter of will.

Speaker 1: I got a podcast to run. I'm I'm trying to be low agency, low taste right now.

Speaker 2: Low taste.

Speaker 1: Low taste. Anyway, thank you, Garry. Congratulations on demo day.

Speaker 2: Yeah. Great to see you. Catch up.

Speaker 1: We'll talk to you soon.

Speaker 14: Catch you guys.

Speaker 1: Have a

Speaker 2: good rest of your day.

Speaker 1: You see

Speaker 2: I like the all black.

Speaker 1: All black. I like the all black. He's looking good. Did you see the Rolex open a new boutique on the Swiss Alps in inside of a former telecommunications tower? It looks like a Modern Warfare two map. Look at this building. This is an incredible place to hold a boutique. I guess you have to hike up here. But look inside. This is a Rolex boutique. You go in there. They show you some watches. But this looks like Modern Warfare too.

Speaker 2: Carter says, imagine being told they have nothing in stock here. You both know very well they have watches in stock. Yes. What actual resort is it on?

Speaker 1: I don't know if this is a resort. I think this is a a telecommunications tower. But I gotta know more about this. We gotta we gotta get someone to call in from this particular boutique because this is an absolutely wild wild wild scene. But what a great like brand activation.