SPEAKER_00: Not only is it a vehicle that's very efficient at transiting out in space, as you mentioned, but also just cost efficiency wise, we've seen reuse be the play that makes space more economical, right? And so we're definitely looking to solve that. We are looking to be, and maybe this is another thing with orbital operations, to be a tugboat, to be a tractor, to be a ferry, to be a rapid response vehicle. Most of our vehicle has to be propellant. We look a lot more like the third stage of a rocket than we do a traditional satellite. Like we are bigger.

SPEAKER_04: What's the problem set that you're working against?

SPEAKER_05: So it's two, and I kind of mentioned it as well. One is space defense.

SPEAKER_06: It does seem like the happy, peaceful, humans working together aerospace is kind of coming to a close. So really it's just, can you get up there and does your technology work?

SPEAKER_08: This Week in Startups is brought to you by LinkedIn Ads. Start converting your B2B audience into high quality leads today. Launch your first campaign and get $250 free when you spend at least $250. Go to linkedin.com slash thisweekinstartups to claim your credit. Northwest Registered Agent. Starting your business should be simple. With Northwest Registered Agent, you can form your entire business identity in just 10 clicks and 10 minutes. From LLCs to trademarks, domains to custom websites, they've got you covered. Get more privacy, more options, and more done. Visit northwestregisteragent.com slash twist today. And Squarespace. Turn your idea into a beautiful website.

SPEAKER_09: Go to squarespace.com slash twist for a free trial. When you're ready to launch, use offer code twist to save 10% off your first purchase of a website or domain. Welcome back to This Week in Startups, friends. This is Alex, and we have an absolutely packed show for you today. We're going to start with a Twist 500 interview with Orbital Operations. It's a really cool company building space tugs. They have a way to keep fuel very cold in orbit. They want to take satellites over to the moon. It's awesome. I'm a science fiction guy, but even if you're not, I think you're going to learn a lot and really enjoy it. Then after that, we're going to rewind the clock to episode number 882 from 2018 when we sat down with OpenAI's Greg Brockman. What's fascinating here is that a lot of the same questions we're asking today, we were asking back then, the importance of GPUs, the rise of LLMs. It's going to be awesome. You're going to love it. Then we're going to drag Oliver in from This Week in AI to give us a demo of Character AI, a popular consumer AI application that you have heard of, but maybe not used. If you want to learn more, well, we have you covered. Then we're going to turn the clock back even further to episode number 736 when we had Robin Hood's Vlad Tenev on the show. This is back when Robin Hood was much, much smaller, but it's always good to go back to learn how the giants of today were once smaller and how they got to their current scale.

SPEAKER_11: There are a host of startups that we're tracking on the Twist 500 that are working in space, and they have big plans. Names like StarCloud and Albedo Space come to mind. One space startup I am particularly excited about is Orbital Operations. It was part of the Winter 2025 YC batch. Now, why am I into it? Because I think one of the best ways to accelerate the space economy is to ensure that there is the orbital equivalent of tractors above our planet, and that's close to what the startup is building. So please join me in welcoming to the show, Ben Schleunerger, co-founder and CEO of Orbital Ops to the show. Hey, Ben, how you doing?

SPEAKER_14: Alex, thanks for having me on the show. Appreciate it.

SPEAKER_11: An absolute pleasure. So I think we should probably start with what problem you're trying to solve, and then we'll get into the Astraeus and chirogenic cooling and all that.

SPEAKER_16: But what's the main thing that Orbital Ops is trying to fix in orbit today?

SPEAKER_00: I think if we had to really break it down in the kind of shortest amount of words, we're mixing the problem of movement out in orbit, right? And I think particularly that starts with a lot of solving the space defense type of problems and the issues they're running to with movement out there, but also the movement of getting satellites farther out.

SPEAKER_19: And as you kind of put it, the commercial application of repositioning satellites, building that in-space infrastructure.

SPEAKER_06: So for folks out there who don't know much about things that are in orbit, why do they need to be moved around?

SPEAKER_04: What's causing them to lose momentum? Are they falling into the Earth? What's the problem set that you're working against?

SPEAKER_05: So it's two. And I kind of mentioned it as well. One is space defense.

SPEAKER_00: And maybe even taking a step back, like traditionally satellites are designed to go into a singular position out in orbit and stay in that orbit. You know, that's kind of been the problem statement for the last couple of decades is it needs to station keep, image the ground, communicate on the ground. But as we're going to like build more things out in orbit and as it becomes a more contested environment, you now kind of start to need to have these requirements to be able to move and be able to do reconnaissance or reposition. You know, as we build out and we become more space residents, we have to be able to do those types of logistics out in orbit. So, again, the first thing Orbital Operations is really solving is kind of the rapid space defense. That's kind of the most urgent need in this contested environment. But we started more on the commercial applications. And I think the key thing that we saw on the business side, movement-wise, was we have all these reusable rockets coming around, both first and second stage. Think Starship. Think Nova from Stokes Space that are really good at dropping off in low Earth orbit. And I think the price for low Earth orbit is going to keep coming down. But those rockets are not particularly great at getting out to higher orbits. Think geosynchronous lunar orbits, the launching points, maybe even medium Earth orbits. And so I think we're going to start to see a world, again, tying in a little bit on kind of the infrastructure space tractors, right? You're going to have low Earth orbit basically be your stopping point, your port, right? You're going to have vehicles that are really good at getting through atmosphere and coming back down through the atmosphere, being dropped off in low Earth orbit where they could meet a vehicle like ours that could grab and then ferry you out to higher orbits. Think geosynchronous, lunar orbits, whatever it is, drop it off, and then actually have enough range to come back.

SPEAKER_24: So they're not disposable satellites. You can actually bring the tool back and then refuel it. We'll get to refueling in a second. Yes. Yeah.

SPEAKER_06: Just thinking about this, if it's very, very hard to get stuff out of the gravity well, and we get very good at getting things just to low Earth orbit, I presume it's actually very efficient to not have another stage on that rocket that we had to bring up from the

SPEAKER_16: ground. And instead to have a satellite just kind of grab onto whatever we took to low Earth orbit and then take it up further, that's probably very energy and engineering efficient.

SPEAKER_00: Correct. And not disposing of, you know, all of this is going to still take, you know, rocket engines and aerospace grade tanks and electronics that are hardened for, you know, these kind of space environments, the last thing you want to be doing is disposing of that each time. Yeah.

SPEAKER_29: Yeah.

SPEAKER_00: Not only is it a vehicle that's very efficient at transiting out in space, as you mentioned, but also just cost efficiency wise, we've seen reuse be the play that makes space more economical, right? And so we're definitely looking to solve that.

SPEAKER_06: So you guys are working on a device called the Astraeus, which you call a cryogenic orbital maneuvering vehicle. The last three words, I believe, mean things that can move around in space, but the cryogenic element seems to be quite important to you.

SPEAKER_11: You guys talk a lot about temperature. Yes. So why don't you talk to me about why that matters in this case?

SPEAKER_00: We are solving the problem with movement out in orbit. And a big part of that is how efficient your propellants are, right? Just think miles per gallon in like a standard car, right? Like you would much rather have a car that is 50 miles per gallon than 10 miles per gallon, right? And military applications, logistics, it matters just as much. And so, you know, my background, my co-founders background, most of the team's background come from launch vehicles and they are trying to be hyper-efficient with high thrust. They all use cryogenic propelants, just really cold liquids, right? Like very cold liquids. Most of them are either using liquid oxygen and liquid methane or liquid oxygen and liquid hydrogen, which is the more efficient of the two. You even see some rockets that use methane for the first stage because it's a little cheaper and then liquid hydrogen on top because it's more efficient. Now, traditionally, if you took those really, really cold liquids and you threw it out into orbit and it was getting hit with sunlight and radiation, they're actually so cold that just that bit of sunlight starts to make them boil. And in order to not have your tank just pop, you have to start venting your tank. So traditionally, they are known as not being storable, right?

SPEAKER_17: That's kind of the key word there.

SPEAKER_16: So great for propulsion, great for going up, but not something you want to have sitting in orbit around waiting for a job like your satellites will eventually be.

SPEAKER_00: Correct, 100% for the quick missions that rockets are taking off and doing perfect, but never been stored out in space. So what, you know, the core technology and what orbital operations is implementing is basically an active chiller system on the vehicle, right? A very, very advanced refrigeration cycle. The compressors, turbo alternators, very like industrial level cooling that we are kind of bringing out into orbits on our vehicle. A lot of the advances that NASA has been doing in the last couple of decades and who we have a partnership with as well, has been getting those to be lightweight and efficient and good enough to be on a vehicle. And they've had some key breakthroughs and we're really bringing that to the commercial

SPEAKER_37: market to solve the movement problem.

SPEAKER_16: Is Astraea going to be the first satellite in quantity that's going to use such cryogenic

SPEAKER_06: fuels in orbit and hold onto them? Or is this something that's been done before and you guys are commercializing?

SPEAKER_00: We are commercialized, we will be the first. No matter of fact, one of our demos, I believe our demo, when we put it up into orbit, will be the first demonstration showing liquid hydrogen stored for extended periods of time.

SPEAKER_19: That is going to be a big milestone that orbital operations will pull off.

SPEAKER_06: So you're taking on quite a lot of engineering risk there. May it go well the first time, but famously working in space is difficult. So what's the chance that something goes wrong and it either boils off and pops or something even worse happens? I mean, I guess what's the fault tolerance here?

SPEAKER_16: If you ship bad software, people go down for the weekend and people complain, like this is a much bigger deal.

SPEAKER_43: Squarespace has been a friend and partner of This Week in Startups for over a decade. And it's because they are one of my favorite startups of all time. You know, Squarespace, the all-in-one platform that's going to help you make a beautiful website to help establish your brand online and scale your business with you as you grow. It doesn't matter if you're selling content or a course, but what if you want something super customized, well, you can use their AI-powered blueprint feature that works right alongside you as a creative partner. You tell it what vibes you're going for, what the aesthetic is, and it makes something totally beautiful and personalized to you and your business. And new features keep coming from our friends at Squarespace. You remember they did SEO, you know, they've been doing that for a decade. Now they're doing it for AI. Yes, AI search. People are starting with AI. And then you see all those citations, they lead to Squarespace websites because they do AI search optimization.

SPEAKER_47: Go to squarespace.com slash twist to get 10% off your first website or domain purchase.

SPEAKER_05: From our perspective and in comparison to a lot of other startups that are tackling this

SPEAKER_00: movement problem, you know, we are using essentially heritage level propulsion, right? Like we're using the same type of rocket engines that are used on launch vehicles. So there's a lot of knowledge base there, there's a lot of people that have been going through multiple engine programs, the structures, the avionics, all of that has been done before. It's really this kind of refrigeration cycle. And it's been shown on the ground, it's operated on the ground and storing it there. And so we'll definitely de-risk it in steps, you know, do a subscale demo, but flying the actual hardware and designs that we'll want to use on the larger vehicle that actually shows like storing these propellants out in orbit and de-risking it on the ground. And I think I would be too naive to say everything goes right, you know, the first time and that's

SPEAKER_19: you're a hundred percent right. Like we need to be prepared with either a second demo in waiting or, you know, ways to de-risk this approach. I do think in comparison to a lot of other methodologies, this is the most realistic and practical way to get like a high efficiency, high thrust propulsion system out in orbit.

SPEAKER_16: The high thrust thing is very important to bring up because I think it might be starfish spaces

SPEAKER_06: using electric propulsion for their devices in space, which is much lower thrust. So high thrust, I presume allows for greater maneuverability, greater speeds.

SPEAKER_50: And also you can reach further orbits, does higher thrust bring other advantages that I'm not thinking of to the Australia system?

SPEAKER_00: No, I mean, that's, there's a lot of advantages there, right? And so electric propulsion can get, you know, potentially a higher efficiency, but again,

SPEAKER_19: very low thrust, like 0.00.

SPEAKER_50: You breathe on things essentially, and then they slowly kind of drift away.

SPEAKER_19: Exactly. Whereas, you know, we're looking to be about 18,000 pounds of thrust, like, like very,

SPEAKER_00: very rapid response, this is this, particularly in the defensive applications, I think there's a lot for like rapid reconnaissance in the commercial applications, like blasting through the Van Allen belt. So you're not just sitting there for months at a time, and you're only there for hours, like very key for a lot of like commercial customers. There's also the entire realm of Golden Dome starting to open up and looking for space-based interceptors where thrust matters a lot, and we're starting to look into that realm as

SPEAKER_19: well. So lots of different applications.

SPEAKER_06: We talked about taking stuff from low earth orbit out to wherever it needs to go, and then coming back, which is going to require quite a lot of thrust, quite a lot of fuel. This brings up the question of refueling. Now I know that there are some companies that are working on the in orbit refueling question, if you will. But if you're taking a satellite to orbit that has a unique or special propellant systems, you're going to be the first to do this. How do you get in orbit refueling up and running if you're kind of the de novo vehicle for this new thrust technique in orbit?

SPEAKER_57: Well, I mean, you guys can create the demand, but is there supply to meet that? I guess it's kind of my question.

SPEAKER_00: I think we are going to probably tackle refueling our vehicle first, exactly for that business case problem. If somebody, and maybe I'll back up a second on how we're going to refuel and what propellant

SPEAKER_19: we're using and maybe what other satellites are using. So we're using liquid hydrogen, liquid oxygen, and we're having a cryogenic system that keeps

SPEAKER_00: those propellants cold and separated. But liquid hydrogen, liquid oxygen is essentially just the products of water. And so what we want to do to refuel our vehicle is actually launch water up, which is a singular tank. I need the cryo management system. I can put a bladder system in that tank. I can't put a bladder system on liquid hydrogen, it's too cold, right? To do any squeezing or any like compressibilities there. So you've got this really simple, like easy to kind of squeeze the water through in zero G, which is a challenging thing for cryos where you have to pump it and move everything over because everything's just floating in space. And you run water electrolysis and you split it into hydrogen gas and oxygen gas. They do this on the international space station right now. They actually launch water up, they keep the oxygen for breathing, they dump the hydrogen overboard. We keep both and let it trickle into the tanks and condense down. The downside of that versus, you know, just refueling with liquid hydrogen, liquid oxygen, it takes more time. But if it's a month and a half, two months on refuel, and that's just so long because we have such a backlog of missions, we're in a good spot. That's a great problem to have. So I'll take that problem any state of the week. Now kind of getting back to your question, like that means we want water. That is what we would like to refuel with. Are there going to be other companies with water depots out in orbit? I would love that. I would buy water from them. But I don't know if that's going to be the case. Also, a lot of other satellites use storable kind of propellants like hydrazine or, you know, ethane, all these others. So are they refueling the legacy satellites or are they looking to refuel what I think is, as you call it, a tractor or a ferry or, you know, rapid response vehicles? Like where that's going to happen and where that market's going to go? I don't know. And the way we plan for it right now is that we'll just want to launch our own water tanks for refueling. Um, but I would love to see the economy expand out more. Maybe there's water, uh, uh, mind on the moon, right.

SPEAKER_19: And, and you're actually pulling resources out there. Yeah.

SPEAKER_06: Yeah. I mean, the moon is not that far away and there's quite a lot of asteroids further out that have, you know, ice on them and so forth. So it feels like we're chickening the eggs by the right way to think about this. It feels like we're kind of like forcibly kickstarting a bunch of different things that will eventually help and support one another. But right now we just don't have them in place yet. I do want to ask though about the overall domestic and maybe international space economy.

SPEAKER_24: Um, I'm pretty far away from it. I'm a fan of this stuff, but I don't have a good feel for like, what's it like on the

SPEAKER_06: ground or up in space? How many companies are working? How busy is it? How fast are things progressing? Is it as exciting as I think it is, or am I over indexing on my personal joy about this

SPEAKER_00: stuff? No, you're not. It feels like we've entered like a new era and I will say a lot of it is coming from the government's want, not even willingness, but want to work with startups. I think there's been a difference in that before. There's been like a tolerance. There's been like a cool, like if you get all the way to the end, maybe you can partner with a prime and like, you can take it from there. But from, from what we're feeling from the top and where the government's going, they don't want to work with the primes anymore and the way primes have done things. Right. And I'm not shy to say that, like they are really pushing to bring innovation in. They want to see the apexes and the K2s and albedos, and they want to get that startup mentality and bring that innovation to the forefront because it is starting to be a little bit more of a contested environment, a little more of a crowded environment on the commercial side of things. We're realizing that our in-space logistics and technologies needs to be pushed and catch up, especially if Starship comes online and all these other launch vehicles come online.

SPEAKER_19: So it is a hundred percent, like just very different in just the last couple of years, really, uh, just the way people have talked and, uh, you know, who, who is getting contracts

SPEAKER_77: and, and what, what the future holds. I, I think startups are, are, are going to make up a big bulk of that.

SPEAKER_06: So when does the Astralis go into orbit? I know you guys just did your quote first full pressure, hot fire. Um, but when are you hitching a ride up space?

SPEAKER_00: So we're going to do our first demonstration early 2027. So this is going to be a subscale demonstration and we're really looking to be the first, uh, technology demonstration that is going to store liquid hydrogen out in orbit for extended periods of time. So that would be a historic first, right? I think for space. And that's what gets me very excited. Now this is going to be a subscale demonstration, really showcasing the technology and the innovation and probably, you know, some other secondary goals on top of there, maybe some movement, some, uh, some capabilities to show, you know, camera systems, all of that.

SPEAKER_06: How subscale are we talking then? Is this like one third? Is it like one 10th? Is it one, 1,000th?

SPEAKER_00: One fifth, maybe at most, you know, we are looking to be, and maybe this is another thing with orbital operations to be a tugboat, to be a tractor, to be a ferry, to be a rapid response vehicle. Most of our vehicle has to be propellant. We look a lot more like the third stage of a rocket than we do as traditional satellite. It's true. Like we are bigger. And, and a lot of it, you know, I think on a lot of sides, you know, satellites want to be as little propellant as possible. And we're the other side. We look, we're, we talk more like launch vehicles where we say we want to be mostly propellant. Like the more propellant ratio that we are, the better. And so we will be larger, you know, either like a cake topper style vehicle on a, on a Falcon nine ride share, a Stoke Nova vehicle, like dedicated ride there. And even, even larger, we'd love to get to a flagship vehicle that is a dedicated like

SPEAKER_19: Falcon nine delivery to low earth orbit.

SPEAKER_87: What's the current list price on a Falcon nine launch?

SPEAKER_00: I think it's about 70 million, maybe a little bit more. 70 million. We're definitely going to start with like a cake topper, more ride share type of stuff. And on the 2027 demo will actually be an XL. So it'll be much cheaper than that.

SPEAKER_11: How many of these do we need in orbit? How many Australia systems do we need in orbit? And right now, how many do you think you could keep busy with just current existing

SPEAKER_16: demand from the space economy as it is?

SPEAKER_00: I think on the defense side, when we've really looked at it, you could probably have four in a pretty busy state, just between, you know, space reconnaissance and also space logistics, just government contracts side of things. Without any golden dome stuff.

SPEAKER_76: Like if it's golden dome, things get much larger and it's a much different business case there.

SPEAKER_94: Hey, listen, we meet a lot of early stage founders here at launch my investment company and some, they don't have a lot of traction yet. They just have an idea. Maybe they haven't even finished their product. They've just got an MVP, but they still need investors and accelerators like ours to take them seriously. And you know what? We can't just wire money to your Gmail or your PayPal. That's not how it works folks. We need to know that you're a legit and official business. We need to know your company is incorporated. That's why you need Northwest registered agent. It's the service that will help you run your business the right way from day one. In 10 clicks and in under 10 minutes, you're going to file for your LLC or a C Corp. If you're a startup, get a domain name, launch your official website, claim your business email, and even fast track your trademark application, which some people forget to do. We're talking about more than just company formation. This is your entire identity as a business. Go to northwestregisteredagent.com slash twist and show the world you're in business. And make sure you use that URL slash twist. So they know that we sent you.

SPEAKER_00: On the commercial side with just our LOIs, right? Because we got 1.4 billion in commercial LOIs for this ferrying service I was talking about. Just to service those, if the refuel was two months long and we wanted to hit all our missions in a year, it'd be about eight vehicles just running these ferries back and forth and refueling and doing all of those missions. Now that's just the LOIs that we got. That is not the entire market. I don't, you know, there's, there were plenty of customers that would like to see us do more wrist buy down and show things that, and that were interested in this. And if we could offer it for the price we were and had some history with it,

SPEAKER_19: they would absolutely buy the service. So, you know, potentially we're talking, you know, 15, 20 vehicles doing commercial,

SPEAKER_41: just burying, moving back and forth with, you know, a handful of those also doing defense.

SPEAKER_16: I know the answer to this is, I don't know. And it depends, but let's say I wanted to take an average size satellite from low earth orbit to lunar orbit.

SPEAKER_06: I wanted to use a stress and I'm going to pay for the fuel cost there and back 5 million, 15, 100, 200.

SPEAKER_00: I'm going to put a caveat of it depends on how heavy your satellite is and what orbit you're going to be put into.

SPEAKER_19: But, you know, if you're doing like a small to medium sized satellite and you wanted to go from low earth orbit to geosynchronous or low lunar orbit. Yes, the five to $10 million range.

SPEAKER_00: And I have some actual calculate. Anybody wants to like actually have a satellite transported out there and wants a detailed price. I've got the dollars per kilogram. Reach out to me. Happy to happy to talk more and give a more accurate number.

SPEAKER_06: I was just curious about the mission cost versus the launch cost. Because presumably each Australia system will be able to do dozens. I'm just guessing of missions in its life. But if they're five to 10 each suddenly 70 million to get a big one up there on your own dedicated Falcon 9. The math pencils out even today with launch costs where they are. So that can get really exciting. It sounds Ben like there's right now ample demand for orbital ops. And if you extend the timeline by 10 years, there's going to be overwhelming demand. So really, it's just can you get up there and does your technology work? You guys raised 8.8 million in a seed round. How much more do you need and when?

SPEAKER_00: I think we're going to be raising a series A coming up probably, you know, Q1, Q2 this coming year.

SPEAKER_19: I think total amount to get our first vehicle up, look, that is going to be doing, let's say, you know, our first government contract, right? I think in total, we're going to need, you know, 150 million, right? 150, maybe 200 million. We are maybe more expensive than what your standard satellite company would be, but much less than what a launch vehicle development costs would be.

SPEAKER_107: But you're doing new, harder things.

SPEAKER_06: I mean, you're making bigger satellites, you're testing new technology. You're, I mean, love albedo space, love that they're in super low earth orbit, but like it's space cameras. We've done that before.

SPEAKER_109: I would presume they would lower capital cost. Yeah, correct.

SPEAKER_00: And they are actually, you know, a fairly large sized satellite to be able to operate lower. I love those guys, by the way. They're great. So, yeah, you know, I know they ride, I believe like a cake topper style rideshare, which I keep saying that word. That is like the largest slot on the rideshare for Falcon 9 is the cake topper. So, you know, I know they ride that, that slot as well. And that is probably what we'll do for our first vehicles as well. So, you know, I understand like their, their cost range will, we might be a little bit more than new technologies there, but yes, definitely much less than a launch vehicle.

SPEAKER_06: Ben, before I let you go, I am curious about just the world of defense and the militarization of space. Over on your Twitter account, you shared a story entitled the space laser wars have begun and America wants to be first to develop the high powered weapons. It does seem like the, the happy peaceful humans working together aerospace is kind of coming to a close. How much capital is that bringing into the space industry today? And is there going to be kind of a, a dual use takeover of the domestic space industry? Yes.

SPEAKER_00: On both accounts. So, you know, I, if we are including golden dome, which I think it is fair to include, cause it is a large amount of space defense dollars coming in.

SPEAKER_19: It's tens of billions and it might wind up as hundreds of billions, right? Like there is between missile tracking and space space interceptors and communications for the ground and resilient GPS and nuclear command and control. Like we bring all those dollars and we're talking hundreds of billions of dollars. And so, you know, I, I think that what you said there is like the space sanctuary days, the international space station model of, you know, and which will still happen.

SPEAKER_00: There'll still be the collaboration, there's still going to be the research there, but I think space was so hard that everybody was basically in it together. Had to. Had to. That was the only way to go about it. And now that we are developing more technologies and operating on a routine basis, it was only natural that it was going to start to be a little bit contested. Air domain went through the exact same thing, right?

SPEAKER_05: Like, um, and so, yeah, I think that, um, you know, that's going to bring in a ton of dollars.

SPEAKER_06: Is everything going to become super, super dual use?

SPEAKER_05: Yes. A hundred percent.

SPEAKER_06: What I heard about you guys, it was like space tugs. I'm like, oh great commercial. And then I was like, oh wait, dual use. This is going to be moving military assets and keeping things safe.

SPEAKER_04: And maybe even responsive to novel threats.

SPEAKER_00: I think the key there as well is that the government dollars coming in want to see dual use, right?

SPEAKER_19: They would really like to see that you have a commercialization aspect and you're not just going to be completely dependent on their dollars. Now there's always going to be some specialized tech that is like, yeah, nobody else is doing missile tracking on the commercial side of things.

SPEAKER_00: So like that is going to be like a military like development and use case there. But I think the core technology for us, this, this advanced refrigeration, so this cryogenic management system. Yeah. We can use it for rapid space response, uh, in space defense, or we can use it for in space logistics, both on the government side and the commercial side. And so they would love to see more of that and every proposal that we're putting in, they would like a section of how is this going to be commercialized beyond this? So yes, yes.

SPEAKER_19: On both accounts, tons of government dollars coming on the space defense, but also the dual use case will always be very front and center in their minds.

SPEAKER_06: More demand, more supply, more competition, lower prices, faster development works for me, Ben, an absolute treat.

SPEAKER_11: Um, when you reach another major milestone, we'd love to have you back to learn more about it. And definitely in early 2027, when you get that demonstrator up, I want to see video of your space refrigerator and, uh, we'll have you back on.

SPEAKER_103: Alex, thank you so much for having me. This is great.

SPEAKER_09: I told you orbital operations is an incredibly cool company.

SPEAKER_08: There is a rising cohort of space focused startups on the twist 500 and out in the world. I want to talk to every single one of them. All right, but let's go back to a topic we've covered even more frequently on the show, which is AI.

SPEAKER_09: And to help me get into this, I'm going to bring on producer Lon, co-host Lon, my dear friend, Mr. Lon Harris. Lon, how you doing, baby?

SPEAKER_122: Hey, what's going on Alex? Good to be here.

SPEAKER_09: I'm glad you're back.

SPEAKER_08: Now we're looking at an old interview with Greg Brockman, who people know from OpenAI. Everyone knows OpenAI today, but back in 2018, when this interview happened, much less was known about the company. So in fact, one thing that Jason asked was, what are you building? And here is how Brockman described OpenAI's mission and why the company was a nonprofit at the time.

SPEAKER_124: The mission for OpenAI is to ensure that artificial general intelligence. And by that, we're talking about systems that are as smart as we are, that can outperform humans at most economically valuable work. Whenever we can build that, that is something that actually benefits all of humanity. Right? So it's, you know, I think when you look at AI, we really have this continuous spectrum of technologies. Traditionally, AI is the field that kind of over-promised, never really delivered. Right? That, you know, we've had AI systems doing work for us for decades. Check processing is something that there's kind of an early, early triumph that was, you know, sort of moved into the realm of artificial neural networks in the 90s. But really, starting 2012 was when something changed. And for the first time, you were able to take deep neural networks, so these systems that can just learn from data, and use them to solve problems way better than anything else. Right? So you can solve image recognition, you throw out all the old systems, you replace it with a deep neural networks better, and take the exact same algorithm, use it in machine translation. Suddenly, Google Translate works extremely well. Right? And we just see this across this huge variety of different domains. And so, at first, people were like, oh, okay, so now we can do perception. Right? We can do these very basic tasks. And then machine translation, that feels a little bit harder than normal perception. But then you start to move to things like solving games like Go or Dota, you know, these competitive e-sports games. And suddenly, you're able to have neural nets, these AIs that are able to solve interactive problems much better than humans can. And so, looking forward, I think that, you know, that when you just project for the progress, and I'll talk a lot today about just the kind of progress you've seen and what really drives it, that I think that you really can't rule out being able to build the kind of systems that we're talking about in the near term.

SPEAKER_127: OpenAI is a non-profit? A for-profit? How many employees are there?

SPEAKER_124: Yeah. So, we're about 80 people full-time. We've been around for three years. We're structured as a non-profit. Why? And the reason for that is that, you know, our goal at the end of the day is not to generate huge profits for us, right? The goal here is to build a system that we think that the impact, if you actually succeed, is going to be on the order, you know, at a minimum of the agricultural revolution. And on a maximum, you're talking about, like, going to the stars and, you know, solving all disease and, you know, sort of all the things that humans have wanted to do but seem kind of out of reach of us being able to accomplish. And so, when you talk about something like that, it's much larger than any one company, any one country, any one set of people. It's really about humanity. And when you're doing something like that, at the end of the day, it really is incumbent on those who are going to have, you know, be setting, shaping that technology to make sure that they're thinking larger than just for their own benefit.

SPEAKER_130: We have a lot of B2B companies and marketplaces at our founder university, and they're all looking for their first customers. Our best advice to them is always the same. Use LinkedIn ads. Professionals are spending their time on LinkedIn. Professionals are spending their time on LinkedIn. So if you're not there too, you're throwing those ad dollars away. And LinkedIn already has the data you need about who the key decision makers are in your field. That's because the platform has over a billion members. And 10 million of those members are C-suite executives. And you can target your ads specifically to your ideal audience based on their job title, industry, their skills. Plus LinkedIn's flexible pricing gives you total control over your entire campaign, allowing you to launch an ad with daily budgets as low as $10. You can start, stop and pause and adjust your ads at any time. We've got a special offer for Twist listeners. Launch your first campaign and get $250 free when you spend at least $250. Go to linkedin.com slash this weekend startups to claim your credit.

SPEAKER_09: Lon, the thing that really kind of captures my interest there is how it doesn't seem that the company has really changed its mission much in the last six, seven years.

SPEAKER_135: Feels pretty consistent. I think it's interesting in 2018, AI was still, it was one of those like pie in the sky.

SPEAKER_136: We know people are working on it, but it's still felt in the realm of sci-fi. Like we didn't, there weren't a lot of like apps everyone was using that were sort of AI powered, LLM powered in 2018. So the discussion that they're having in this clip is sort of still in the theoretical, you know, like one day this technology might be sort of dangerous or questionable or controversial. So we want everybody to own a piece of it, but it was, it was very like far off. It was like Skynet level Terminator 2 concerns. And it's just funny that we are having these same conversations today, but now it's like very real and tangible. We're not talking about like, well, theoretically this could be dangerous one day and we all want to have like a stake in it. It's like, things are happening. Like people are being impacted by this every day.

SPEAKER_138: And we still kind of don't know what to do about it or how to govern it. And I think that's, that's, what's really interesting.

SPEAKER_08: One thing that has changed though, is just the sheer amount of money that is being spent at open AI. There's a really funny clip from this interview, which they discussed their current spend, which I think lawn would work out to about 10 minutes of their current burn rate.

SPEAKER_81: And that was back on a yearly basis. Check this out.

SPEAKER_142: Check this out with 80 people. That sounds like you're spending in the order of $10 million a year or something like that.

SPEAKER_124: We spend, we spend more than that. Yeah. So as a nonprofit, you know, our, our nine nineties are all, all public. So you can, you can go and look them up. Um, I think our 20, uh, 16 budget, you know, basically we've been on an exponential ramp, uh, of, you know, I think like, you know, 10 to $15 million, I think was like kind of year, year one.

SPEAKER_136: Jason is, uh, Jason at one point says like, wow, you guys are spending like 10 million a year. And Brotman's like, it's even more than that.

SPEAKER_138: It might be up to like 15 million.

SPEAKER_136: And like, if open AI spent $15 million this year, like they would be valued at a hundred trillion dollars.

SPEAKER_09: I'm pretty sure that open AI spends $15 million every month, just flying Sam Altman around. Yeah.

SPEAKER_150: That's just like one bag of chips is that's, that's what you're spending $10 million on.

SPEAKER_09: But on the point of things that haven't changed that just looping back to that, I was really kind of impressed at how clear eyed Greg was at the time about the risks of, of AI in certain cases, for example, in deep fakes.

SPEAKER_08: And we've talked about Sora and Sora 2 on the show a lot, Veo from Google, and even just the image generation tools that have gotten so much better.

SPEAKER_81: But here's how we discussed a possible downside about AI at the time.

SPEAKER_151: If you look today at examples of AI technologies, it's not really clear if the world's better for them being out there.

SPEAKER_124: I think that a really good example of that is deep fakes. It's deep fakes. So if you're familiar with that, right, that this is open source code. Explain. Open source code that anyone can use to generate fake videos. You know, you can, you can take a existing video, you can, you can put in different people and it looks pretty realistic. And so this is something where people have used it for lots of, lots of different applications that, you know, if you, if you add them up, I'm not really sure that the world's better as a result. Yeah.

SPEAKER_142: Um, and the one you're talking about, of course, is adult content where they're taking celebrities and putting their likenesses on pornographic images. And if you were to look at, you would think it was a leak from their phone when in fact it's a computer building it or having Obama say things that are inappropriate or that might trigger the right wing people or taking right wing folks and having them say things that might trigger the left and just chaos ensues.

SPEAKER_124: That's right. That's right. And so, you know, you, the, the, the, where we are with this technology is this piece of open source code that anyone who knows how to install TensorFlow, uh, is now able to use to generate whatever content, uh, that they have in mind and generate these very realistic videos. And I think we should think of this as a case study, right? You know, like a first, first little taste of what does it look like once you start having powerful AI technologies that, uh, that you actually really do have to do these sort of security and safety evaluations on.

SPEAKER_136: To me, what's so interesting about this is that it was already weighing on their decision of how to structure the company. Like part of the discussion of why it's a nonprofit, why they want so many different stakeholders and people to have a, a sort of a, a say about this was because they were already aware of not just deep fakes, but all the other potential uses of this. It fell into sort of the wrong hands. And I, I, I think what's fascinating too, is that there's a, a comparison to the Manhattan project that you still hear today. Like people are still talking about the United States. We need to beat China. That wasn't so much on the radar yet. Uh, in this discussion, they're not really talking about the threat of the U S has to stay ahead of China specifically on this technology, but they are aware that like for national security reasons, for consumer protection reasons, for privacy and security reasons, like it was going to be very important for America to sort of be leading the way. In a sort of comparative to the Manhattan project way. I think that's really interesting that met metaphor has stayed with us. Like AI being in some way, like the Trinity test in that moment of we've unleashed this, we've unleashed this power into the world and there's no taking it back. Pandora's box is open. And now we need to learn how to work with this and live in a world where this exists, live in a world in which this exists indeed.

SPEAKER_08: And one reason we get to do that lawn is simply because it turns out that gamers had done a lot of the work to use GPUs and help Nvidia essentially invest all this money into making GPUs awesome. Uh, so that way when we need the compute, we can do it.

SPEAKER_09: Here is Brockman describing how today's AI industry is different from what we saw in the fifties. Even if we had worked out some of the, uh, I don't know, intellectual groundwork lawn for modern AI, we just didn't have the compute.

SPEAKER_124: The reason that we're able to do it and the people like Alan Turing and, uh, and other people who were very, very smart, uh, in, in the fifties, weren't able to do it is because our computers are a lot faster. Um, so if you, if you, if you, if you look at, at the way that these, that these systems are built, you have massive computational power. Um, and it's funny because on the one hand it's massive relative to the scales that we're used to, like the, the amount of compute that we have in a laptop, uh, on a CPU. Um, but compared to the amount of compute in here, uh, we're still, we're still, we're still not quite there. Um, and so the, the, the real core, uh, innovation that the thing that, that really triggered this, uh, this, this revolution was when people realized that they could train neural networks on GPUs.

SPEAKER_164: Right. So GPUs have a lot more flops, graphical processing units, what you use to play a video game or to do self-driving in a car by companies like Nvidia.

SPEAKER_124: That's right. Yeah. And so Nvidia stock price, if you look, has gone up quite a bit over the past couple of years. Um, and, uh, the deep learning is, is at the core of that.

SPEAKER_136: Yeah. I mean, gamers leading the way as so often happens. I think you could forget about how popular video games have become until you hear something like this. And you're like, Oh wow. Video games really like lead all technology. Like that's where the innovation is happening. That's where the money was being spent. So I think, you know, gamers inadvertently creating the pathway to AI is really fascinating. I also love his point about how it wasn't like, we're so much smarter today than the guys in those labs in the fifties with the room size computers. It's tempting to think like we're so far beyond their, their understanding of computers and Brockman makes this really interesting point that not really, like they already had all of these theoretical ideas. They just didn't have the computing power. Like the planet earth didn't have this level of computing power. We were like several stages of innovation in terms of like semiconductors and just raw compute and how to generate that much computation. That technology was what was holding them back.

SPEAKER_36: This is actually one reason why I really struggle a little bit with the doomerism around our current kind of GPU capex we're seeing across the entire technology industry.

SPEAKER_08: Because fundamentally what we're doing is just building a bigger brain for the human species to, to use. I mean, we're just making a smarter synthetic intelligence. And so how worried can I get about that? It's going to be awesome.

SPEAKER_136: Well, I mean, to use that metaphor to Brockman actually does talk at one point in this segment about how like, we keep, we keep getting better. Like we keep making more and more and more and more powerful computers, but we're still nowhere close to a, you know, the computational power that we all have in our heads.

SPEAKER_138: And I think that is because that's mother nature spent millions of years putting that together. We're not even close to that level of effort and evolution yet.

SPEAKER_36: Your tokens per second per watt in your brain is very, very impressive.

SPEAKER_08: But let's look to the future for a second, Lon and talk about AGI. Because I think Brockman actually does a pretty good job in this interview of explaining what AGI means.

SPEAKER_124: What AGI is really about is something a level deeper, right? That we as humans are, you know, I think we're used to building technologies where fundamentally the decision-making capability is still, it's up to us that we can make better decisions than our systems can. Got it. And that I think that the real transition point and the challenge and the benefit really is being able to build systems that can make better decisions than we can, that can understand problems better than we can.

SPEAKER_09: What really strikes me about this, Lon, though, is how the definition hasn't changed. You know, economically valuable work, smarter than most humans. I mean, that's still what we're shooting for today. We're a lot closer, but it seems like the target we're shooting for hasn't actually moved that much.

SPEAKER_136: There was also that notion of the singularity that I think is still with us that eventually will invent a brain that's so powerful that it's just like, we don't, we're making ourselves obsolete. We're just, we're, we're inventing this brain that's going to do all the thinking for us now. And like, where does that even leave humanity? Or do we even cease to be human? And we somehow merge ourselves with this AI and become cyborgs or whatever. So I feel like that is sort of to one side. And this is the notion of AGI. That's a little bit more specialized. And that is also, I mean, just a theme that comes up on twist all the time now that back in the 2018 days, it was very much this model of we're going to have one computer that is an AGI. And it's like the big boss brain and we ask it any question about anything. And I feel like a lot of the researchers we talked to today, they're thinking a little bit more in terms of specialization. Like you might not have one AGI that answers every question. You might have your medical expert. That's the greatest medical super intelligence, but it doesn't know legal questions. And then you have your legal super intelligence. And then you have, you know, and I think I, to me, I feel like that's almost more comfortable because then it's not competing with us. It's just like that one's only competing with doctors, but it's not thinking about how to do everything.

SPEAKER_08: Sure. I mean, we can call that LGI and MGI. Then I think the G begins to lose a little bit of its, its bang in those cases, but the future isn't written yet.

SPEAKER_09: But in this interview, we did get a hint about how we got to today, at least long, because Brockman discusses a model they've trained that has ingested thousands of books and allows you to predict what would come next. We'll answer your question. This is, I believe the absolute foundation of large language models.

SPEAKER_81: Take a listen.

SPEAKER_124: So right now we're able to transfer the systems, right? So we take that AI system that we built for Dota. You point out robotics and it works, but you can't use the knowledge, that background knowledge, but that's starting to change already. So we actually have a model that we trained earlier this year. And other people now have have kind of fall on work and similar work where we just had a model that read 7000 books. We didn't tell it what to care about the books. We just said, predict what is going to come next, predict what word is going to be the next word in the sentence. Huh. And so just read it, predict what's going to happen next. And then you can take this model and you can apply it to a bunch of different natural language processing tasks, like these question answering tasks, like read a little story and then answer some question about it. Right. And it's that state of the art across all these different data sets. Right. So it really learned some background knowledge that's useful. Right. And then you can also ask it to generate text for you. Oh, boy. You can ask it to make a little story for me. And it comes up with these stories that are actually pretty interesting to read. That one of one of my favorites is it talks about these these guards of the Red Temple that, you know, that are that are that they're kind of fighting and they're protecting whoever the narrator is. And then that it starts talking about how some some smoke comes from this fire. Right. You start to see there's some there's some world knowledge that's really built in there. And so I think that this work is all very early. It's preliminary.

SPEAKER_164: It can't be any worse than The Last Jedi. Right.

SPEAKER_124: Exactly. Exactly. Yeah. But I think I think I think what we're starting to see is that you take these models, you scale them up. Yeah. Right. Unlike with Deep Blue, where that the model that you're working with is fundamentally hard to get in the real world. Right. You need the simulator that's perfect. That lets you look ahead here. You just need background data. Right. You can take data of books. You can take data from YouTube. You can take data of just images. And now we're able to start to learn things that seem like we can actually generalize to different problems than what we were training on.

SPEAKER_08: At the end of that clip, he discusses how you can take books. You can take data from YouTube, bringing a lot of different sources to create things like, I don't know, GPT-5, maybe Sora. I mean, this is the stuff that we're currently, you know, in all of today. It's just amazing how early they kind of saw this coming. I wonder what we see today that we're going to get in seven years that we're going to look back on and be like, oh my gosh, we were actually right.

SPEAKER_136: It's very impressive. Brockman was clearly thinking way ahead and did have this solid sense of like where LLMs and these prediction sort of algorithms were going to go. But I do think it's fascinating that his example is storytelling. And it's like one of the things that we've still struggled to get LLMs to do really well. Like the first thing he thinks of when he's talking about ways to use these LLMs is like we can ask it to tell us a story and it comes up with these really interesting scenarios. It was telling me a story about, you know, ancient China and like, oh, okay. And it can, if you ask chat GPT to tell you a story, it will make up a story and tell it to you. But I think that what we found is there are so many more compelling uses for it and things it's better at. One thing I would say is conversation that Brockman was thinking in terms of write me a bedtime story. I don't think that has become super popular yet. I don't really read a lot of content that was written by an AI and then somebody put up and it's like, this is my favorite story. But everybody is talking to chat bots like in terms of having conversations with them.

SPEAKER_159: And so I feel like it's not exactly what Brockman was talking about, but it is obviously very closely related.

SPEAKER_09: Yeah, from toy to tool, if you will, but if you go back to when chat GPT launch, I remember my first couple of uses.

SPEAKER_08: I'm sitting around with friends and our phones out playing with it on a couch. It was write me a limb, a limerick, you know, about X and then enjoying the fact that it could ape human creativity. I can't recall to your point. The last time I actually tried to use an LLM for something along those lines, it's mostly research and data processing and charting and that sort of thing.

SPEAKER_36: So it's really gone from toy to tool, which makes it less less cool in a way, but also much more productive and useful for our economy.

SPEAKER_136: Yeah, I feel like the two sort of main branches that it went in was productivity. Like you said, like I'm going to use this to summarize my emails. I'm going to use this to, you know, go through these long documents or videos I don't have time to watch and give me the bullet points or whatever. I'm going to use it to help me polish my notes or fix my grammar. There's that side. And then there is, you know, like I want somebody to talk to. I'm lonely. I need companionship. I want to run this idea by someone and that kind of stuff. And the storytelling, which is where Brockman's head was at, kind of is the one that's sort of been forgotten largely. I think we've all kind of reached the conclusion that writers had originally, which is this is not going to replace a thinking, feeling human with experience, living and sensory experience. Like it's not going to be Thomas Pinchon overnight, but it might be someone fun to chat to or that it's good to like vent to if there's nobody else around who wants to listen to you vent. So I think that's it's sort of similar to what Brockman was pitching, but not exactly the same thing.

SPEAKER_08: One last bit from this interview long that really stood out to me was just who's going to make the money. I know we've joked about GPUs and open AI's costs, but Brockman had a really prescient quote about really where the profit will flow.

SPEAKER_156: You want to know who makes the money? It's always people make the shuffles.

SPEAKER_136: This is something we talk about all the time. It's, it's the, you know, the, the AI companies making these incredible models. Like maybe they're going to, you know, they're, they're the ones that get all the tension. Maybe they're going to be the biggest players, but there is a whole economy of the people who enable the AI to work the way that it's supposed to behind the scenes.

SPEAKER_138: And I mean, every time I write ads for this week in startups, that's what we're talking about is like cloud data infrastructure that powers your products, AI, you know, integrations.

SPEAKER_09: Absolutely. All right. Listen, we're going to move on. We're going to go to now a AI demo with our dear friend, Oliver, looking at Character AI and how to use it.

SPEAKER_06: And then right after that notes from our interview with Vlad Tenev from way back in the day.

SPEAKER_200: Character AI lets you chat with almost anyone you want about anything you want. It can be tailored to your interests, your mood, or just what kind of conversations you want to have. I've been really skeptical about these different AI chat apps. So I wanted to check Character AI out. We've seen companies building similar products in the news recently as well, ranging from huge marketing stunts, different fundraising, and even controversy on how people are using AI companions. So what is Character AI and how do you use it? Let's check it out. So on the homepage, you can see they have a bunch of different characters to choose from that you can chat with. They have Socrates, your vegan friend, injured hockey boyfriend, awkward family dinner. And you can see that some of these have millions and millions of chats that have already happened within them. You're also able to search any character that you want to have a conversation with. Let's try Obama. So you can see that this character has around 1.4 million chats. And in here you can see it gives you a little bit of a warning on how this is not a real person, but let's check it out. So you can see this is kind of just a normal chat interface. I can ask him anything that I want. So it seems like he's kind of doing what he does, some policy debates, some basketball. And I'm actually able to listen to him say it as well. Hey there, just reflecting on... So you can see that really doesn't sound like him, but something you can actually do is add a voice. So over here you see it's set on default. If you click on that, you can then search Obama. I found some that have been created before. Let's check it out.

SPEAKER_206: Hey there, join me on a journey filled with laughter, fun.

SPEAKER_208: Ready to crack the code?

SPEAKER_206: Guess who?

SPEAKER_200: I like the first one. So we'll use this one and now we can listen to it in this voice.

SPEAKER_206: Hey there, just reflecting on some policy debates and enjoying a nice game of basketball. Can't let those presidential reflexes go. What's on your mind today?

SPEAKER_207: Sounds pretty real. Let's check out some other ones. Here you can see family dinner. Now you can see there's kind of a script that's going on. So let's listen to it.

SPEAKER_214: Past the gravy? You ask quietly, hoping to avoid attention. Oh my God, look who finally spoke on Karen shrieks. First...

SPEAKER_200: We don't need to listen to the whole thing, but you can see that you can start interacting with the conversation.

SPEAKER_217: So you can say, let me tell you about my internship.

SPEAKER_214: Oh honey, we all know you're interning at Big Tech, Aunt Karen interrupts, waving a roll like a scepter.

SPEAKER_200: But did you know reindeer act? So you can see that you can just start kind of having a conversation, putting yourself in these scenes. And honestly, I can see how that can be pretty fun. But a use case that has me really excited is down here under try these. You can actually practice a new language, practice an interview, brainstorm ideas. These seem kind of a little bit more productive. And I'm going to try out practice a new language. You can see it kind of just gave me a brief introduction saying that they're fluent in many languages and can help me practice ones that I'm learning. So I'm going to tell it that I'm learning Spanish. I haven't taken Spanish in a while, so forgive any of my spelling mistakes. But you can see that it kind of responded right in Spanish and kind of asked me some basic questions. So here it's telling me that's great. When did you start learning? I can help you with phrases, vocabulary, pronunciation, whatever you want. So here I'm misspelling, but I said, I want to talk about my day for practice. And it said, perfect. Why don't you tell me a little about your day and I'll ask you some questions for practice. You can see how this is going to be a really great tool. But one of the tools I've been playing with that's pretty fun is the Create up here. So you can create a character, scene, or a voice. Let's start with character. So I want to build a character that's like Jason. So I'll name him, give him a little bit of a tagline, description, and a greeting. But honestly, I don't want to write all this, so I can just ask ChatGPT to help me. Once it's created, I'll go ahead and add it into character.ai. And I'll add the tag of boss. And I will make this public as well. You can see there's also some more options. What's the character's backstory? How do you want it to talk and act? So I'll actually add this again using ChatGPT. So I'll put kind of this backstory, a little more definition on how the bot's going to behave. And then I'll create. So you can see that was really quick and it already is kind of prompting me with that first question. Let's listen to the voice.

SPEAKER_223: Hey, I'm Jason. Let's cut to the chase.

SPEAKER_200: I'm not sure how I feel about that. The one thing that we can do that we haven't done yet is actually add a voice based on Jason. So let's head over to voice here. You can see it's set on default. And then I can actually go up here and create a voice. So you can see the first thing it does is prompts you to upload a sound. So I'll just drag in an audio of Jason talking from the podcast. When you hire people, hire people who have the skills you need.

SPEAKER_224: Great developer. Great. So then I hit generate voice and you can hear. Good day.

SPEAKER_200: Here to make life a little less complicated. You can see it sounds pretty good. So I'm going to name it Jason twist and then I'll give it a little bit of a description from ChatGPT. Then I'll create the voice. So then I'll go back to the character I created and go to voice, your voices, and then Jason twist. Now you can hear it sounds like Jason.

SPEAKER_224: Hey, I'm Jason. Let's cut to the chase. What's your startup idea? And how are you planning to make it a billion dollar company?

SPEAKER_207: So I'll let him know that I am planning to create a company that allows you to call cars from your phone called Subaru. And let's listen to what he said. Huh.

SPEAKER_230: That's cute. A ride hailing company, huh? Sounds familiar. Let me ask you, what differentiates you from Uber?

SPEAKER_200: Honestly, I'm really impressed by this. And I think that creating more characters will become really fun. I can create some of friends or of other famous people if they don't get taken down. And I know that there have been some controversies where people would do characters like Taylor Swift. And now if you look up Taylor Swift, nothing comes up. Obviously there were some issues there. You can actually do everything I've done on the free plan. I'm still in the free plan as well, but I am kind of scared by companies like character AI, as there are so many positive use cases like talking through problems or learning a new language, but there are also scary ones that I think will kind of keep people more isolated than I think is healthy. And I actually think that this is already becoming a reality. As you can see, A16Z recently released a list of the top 50 Gen AI web products by unique monthly visitors. Nine of the 50 web products are chatbots like character AI. Character AI you can see is rated number five here, but you also have different ones like janitor AI, spicy chat, crush on AI, juicy chat. And I'm not going to say exactly what kind of direction these go in, but it's definitely not safe for work. And I didn't look much into the other ones and I'm happy that I didn't. And I will not be saving those for a different video. And looking at lists like this is really important as you can kind of get a pulse on what real people are using, not just the hype that we kind of see on X or in the news. I really do hope that a lot of regulation will come out on the usage of AI chatbots that are kind of there for emotional or different kind of support. Because I do think it's really important for people to not always be on their screens or reliant on what's going on in their computer. But with apps like this, there's kind of a fine line between positive effects and kind of detrimental as well. Thanks for tuning in to This Week in AI. Let me know if you've used character AI in the comments.

SPEAKER_11: I want to take people back in timeline to explain how different Robin Hood is at the time of this interview. We're going all the way back to May of 2017, which sounds like it was two years ago, but was actually eight years ago, eight and a half years ago now. So at the time Robin Hood had about 2 million users that was up from 1 million in October of 2016. The company was valued at a mere 1.3 billion dollars after raising 110 million from DST Global. That's a name you haven't heard in a minute. And also had 80 staff. Like this was early days. For context long, before we jump in today, the company has 26.8 million funded accounts. It has 3.9 million subscribers to its gold product. And it's worth, as of this morning, about 109 billion dollars.

SPEAKER_170: So quite the change since then, but we do love to go back to the beginning. So early investors like our friend Jason, they're doing all right.

SPEAKER_238: They're happy with Robin Hood's progress. I think it's fair to say. I think he may have mentioned that once or twice on the show even.

SPEAKER_135: He definitely did. So the first clip that I thought was interesting.

SPEAKER_136: So this was literally, I think, the first time that Vlad had been on this weekend startup since launching Robin Hood. Because he is walking Jason through an early version of the app. I do want to give you guys a look at this clip. It's really funny. A few interesting things to pull away from here. The first one, Robin Hood really hasn't changed its design all that much. A lot of the features Vlad is showing off to Jason here have remained essentially static. And I think it's really interesting how he talks about the simplicity was so core to the early design. They knew they wanted to give you a few simple pieces of information up top, how much total investment had been to date, how much you were up or down. An easy way to sort of just drag your finger across and see how you were doing bad or good up until that day. And keeping everything really simple with this simple value proposition of, you know, free trades, no commission trades. And it's all going to be right here on this sort of simple, intuitive graph. So when you take a look at this clip, I think one thing to pull out is you can pull up your Robin Hood app today. They changed some of the colors. They changed a little bit of the look, but it's essentially the same app today that it was all the way back in 2017 because they kind of figured it out what they wanted it to be right away.

SPEAKER_243: You can download Uber and call a car right away, but it's not how financial products have worked. You get to trade right away after downloading the app, but it also allows us to keep our cost structure down. So we can offer this service at a, at a much better value than any of our competitors.

SPEAKER_136: So keeping with that, the simplicity, the cleanness, just giving people that immediate, like, here's the information you want at a glance, keep it very simple, get you in and out of the app. There's also so little friction. I think that's another thing that Claude touches on a lot that we're so used to. For younger people today, it might even be hard to do this, but if you're old enough to remember the mid-teens and before, there was a lot of sort of busy work or waiting associated with financial apps, fintech, dealing with banks, dealing with stocks. It would be like, oh, I want to put $1,000 into my E-Trade account and invest in some stocks. Now I will wait three days for the money to move from my bank over to the account so that I could invest, maybe even missing the moment when I really wanted to get in on these stocks or whatever. And Robinhood made it so fast and so immediate. And Vlad talks about in this interview, you go from three minutes from downloading the app to being in it, buying stocks and that experience and really sort of, I think, changed the whole game, like brought stock trading into a whole new era because it became like, as Jason says in the video, a little bit like a video game. I think that's dead on.

SPEAKER_239: I also love the comparison to Uber in which he says, you know, you can download Uber and get a car right away.

SPEAKER_11: Why can't you do that with financial services? Why does it need to be so different? Why is it so archaic? And so when I think about this particular clip, Lon, what I see is attacking multiple pain points in existing products that already had a customer base, which is a great way to find essentially a better way to do a product, which is how you find product market fit. It's not one thing in their case. It was a bunch of good decisions in a row that an aggregate made a product that just really took off right outside the gate. And that brings us to my first clip of the day, which is Jason asking, how did Robinhood go viral? How did he grow so quickly? I'm going to let Vlad explain what he thinks here, but I really do think that product market fit is something that many people think that they have found and haven't because when you do, it sounds a little bit like this.

SPEAKER_243: People love the platform and we're very fortunate about that. You know, we're I think every story is a little bit different. You hear about companies that really, really struggled and iterated like crazy before finally sort of striking a chord with finding something that amounted to product market fit. So I still remember we're working on our website for the first sort of announcement of commission free stock trading. And we had this wait list at the time which sort of allowed customers to sign up. They would see their spot in line. And by referring friends, the mechanic was you can move up in the wait list. Yeah. So the more people you refer, the sort of closer to the top you get. And I think the good thing about that approach is that you get the people that are most passionate about what you're building access first. And they'll give you the best feedback. They'll be the best beta testers. Yeah. And it's very valuable in the early stages. And of course, we're working on the website. We pushed it live because nobody ever has the problem of accidentally pushing their website live too early and being overwhelmed with traffic. Right. It was Friday night. I go home. Everyone goes home. I wake up Saturday morning and I open Google Analytics and I see 600 people on our site. Whoa. And I'm refreshing. I think there's some kind of bug or we might be getting.

SPEAKER_254: 600 concurrent. 600 concurrent visitors.

SPEAKER_243: Right.

SPEAKER_254: Like real time visitors at once. Yeah.

SPEAKER_243: And I'm calling my parents and I'm saying, oh, it seems like we have something real here. This is sort of the first taste of success that we've had. And then we realize, oh crap, like nothing's working. Email confirmations weren't being sent out. Wow. And we've got to wire this thing up. So everyone drives over to Redwood City, which is where our office was at the time, to do the thing that every person in PR tells you to avoid like the plague, which is launching on a Saturday. Right. And by the end of that day, we had something like 10,000 signups on our wait list. Wow. By the end of the first week, we had over 50,000. And then within a year of launch, we had almost a million people sign up, making us by far the largest pre-launch audience of any financial product in history. So we were very fortunate in a lot of ways to have the wind at our backs from the very first moment we announced this product. And I think it's some combination of the fact that people really feel the pain point. You can kind of visualize it in one sentence. I think it resonated with people to such a degree that they told all their friends about it. And that's how we were able to build up that wait list with zero marketing and how after launch, we were able to get 2 million users with barely any marketing at all and no marketing team.

SPEAKER_258: So, Lon, a little gamification, a little video game, quite a lot of product market fit and Robinhood was off to the races.

SPEAKER_11: I think at the time, Vlad said their subscription gold product was going at 17% per month. It's slowed down since then, but it's become enormous.

SPEAKER_258: Insane though, that they thought they were going to get one or two people to show up and instead had 600 all at the same time.

SPEAKER_238: Yeah, I love the shout out too. Again, it's a little nod to history. If you're too young, you may not even remember this era.

SPEAKER_136: The shout out to Hacker News where he's like, he went home and he was watching it like, oh, we're number three on Hacker News right behind Google buying Boston Dynamics and one other story for 2017. And that was such a, I'm launching a new product. I'm doing something new and exciting in the late teens to get Coke to climb those Hacker News charts. Whereas honestly, I feel like it's been a while since I've even heard anyone mention Hacker News.

SPEAKER_197: I read Hacker News every day.

SPEAKER_136: I know it still exists Hacker News, but I feel like it used to be really at this. I feel a similar way kind of about Product Hunt. Like it still exists. It's still great. People still use it all the time, but it used to be the centerpiece of launching any new technology. It was like, well, we'll see how it does on Product Hunt. And then that will determine if we still have a company next week.

SPEAKER_168: And I feel like now it's just one part of this whole satellite of options when you're launching a new product.

SPEAKER_11: Yeah. This also brings us to the discussion of customer acquisition costs. Now, if you're not familiar with the world of financial services, companies spend traditionally a lot of money on brick and mortar installations, on television advertising, on billboards, on all sorts of direct mail stuff. If you have $4, you've seen it literally land in your house. Robinhood, on the other hand, did not do that. They used a wait list, they used some early PMF, and also they found a way to essentially not spend the thousand dollars that other people were.

SPEAKER_08: That meant that they were a much more efficient business, could raise money more quickly, and could also reach profitability sooner. Here's Vlad talking about just that.

SPEAKER_243: If you look at it, most financial companies, basically everyone that I can think of, grow on the backs of paid advertising. Right. And ours has been organic, word of mouth, referral driven.

SPEAKER_263: So you can actually take your, if your customer acquisition costs would have been, I mean, these companies must spend a couple of hundred dollars, like the E-trades of the world, the people you're displacing.

SPEAKER_262: They spend over a thousand dollars. Oh, really?

David Friedberg: Yeah. So their CAC, customer acquisition is a thousand dollars, because they do TV commercials and this kind of stuff. Yep.

SPEAKER_243: Brick and mortar. Super expensive. Yeah. Lots, lots of advertising spend. That's for sure.

SPEAKER_258: So, Lon, I think my biggest takeaway here is that if you can survive and thrive on word of mouth, oh my Lord, is that not a better way to build a business?

SPEAKER_135: Sure. Yeah.

SPEAKER_136: I mean, I think that not every product has one of the central advantages that Robinhood had, which is they really were a better, faster, cheaper, more immediate way to do something a lot of people already wanted to do. They didn't have to sell you on what they were doing. Like, you want to buy stocks? You want to invest in things? Of course you do. Everybody already wanted to do that. They just had this much more simplified mechanism that also was, yeah, like you said, a little gamified, made it a little more fun, made it a little bit more like a thing. You wanted to check the app a bunch of times per day instead of E-trade, which was more like just a banking app, more financial and all the other competitors. I'm picking on E-trade. But I do think that was a huge advantage for them that it was so immediate. You didn't have to really explain to people, here's what Robinhood is, here's how it works, here's why you should download it versus the 10 other fintech apps that are also trying to appeal to you. It was like, hey, you could do your trades for free on here and it happens right away. If you want to buy a little bit of Disney, you could do it in the next five minutes and it'll be there and you could just check it all day.

SPEAKER_168: And here's a little easy graph to show you how's your Disney investment doing today.

SPEAKER_11: You know, Lon, one thing you couldn't do on Robinhood at the time of this interview with Jason was actually trade crypto. Robinhood brought that to its service in 2018 to great effect. It launched with quite a bang and had an enormous waitlist. And the idea of just more people trading crypto at the consumer level was very exciting for folks in the crypto game.

SPEAKER_258: But when we talked to him, this is back in 2017, really not the case. Here's Vlad on how we thought about crypto at the time.

SPEAKER_263: So what do you think of all these cryptocurrencies? Have you been watching this? I mean, it's obviously on a tear right now, again, who knows it will collapse again. And then there's Ethereum, which is contract based. And are you looking at this and saying, hey, there's some opportunity here for Robinhood in relation to cryptocurrencies or is it kind of a different thing?

SPEAKER_243: I think it's a different thing. I think we're very attuned to the fact that different products have different areas of emphasis. And the cryptocurrency world and Robinhood have different emphases at the time. I'm very interested in personally from sort of an academic perspective. I think there's a lot of interesting lessons and people theorizing about the future of the Internet. How are applications going to be built?

SPEAKER_11: It's funny, though, Lon, to hear now because crypto is such a big part of Robinhood's transaction based revenues, the largest of its revenue pillars, if you will. But at the time, man, talk about a different era. What I like here and the takeaway is for founders is that if you really have a conviction about something, you might want to change it later on when things in the market shift.

SPEAKER_06: So Robinhood was not doing crypto, added crypto, and now today makes hundreds of millions of dollars of revenue per quarter from the crypto game.

SPEAKER_136: Yeah, I mean, it's really interesting to think about how Robinhood's had to sort of adjust and change and shift with the times, but also were ahead still. Even with their doubts, even with their like, well, we don't do crypto right now, they were still way ahead of the game. I mean, 2017 to even be talking about offering crypto. It was still a pretty convoluted system back then. If you really wanted, you had to get your digital wallet. You had to figure out like Coinbase or one of those other exchanges. You didn't know if like, where should I leave my crypto? Because if I leave it in like a Mt. Gox type place, is it going to just vaporize one day? So there was a lot. I mean, I was trying, I was still working at startups and trying to figure out crypto in this era. And it was still like, you know, you still had to figure out three or four steps if you wanted to really be active in it. And Robinhood coming on board was one of those cultural moments where it's like, oh, crypto just got really easy where like anybody can sort of figure it out and own a little crypto if they want to. Versus like back in the day when you had to figure out like, so how do I log into my digital wallet? And what happened? Why do I have to memorize these 22 words in case I forget my password? It was a, it was a very different time.

SPEAKER_11: Man, woman, person, camera. You know, you just got to make sure you have all of those words stuck in your head. All right, I want to grab one more, one more quote here. This is actually from Jason talking to Vlad. Jason warns him against spending too much of his recent funding round on sprucing up the office and so forth. But then he tells him what he calls the eagle computer speech.

SPEAKER_275: Take a listen. Everybody gets a little wacky in Silicon Valley. It's called this Taj Mahal syndrome.

SPEAKER_277: Yes.

SPEAKER_275: You heard this?

SPEAKER_277: I have. All right. That's I'm giving you the Taj Mahal speech. The other one I'll give you is the eagle computer speech.

SPEAKER_275: Or now, more appropriately, it's the secret speech, the founder of secret. Those guys sold a bunch of shares in secondary. They buy Ferraris. Mm-hmm. They look like knuckleheads. All their employees hate them because they sold all these secondaries just to drive around in Ferraris like idiots. And then, tragically, in the 80s, there was a guy from Eagle Computer. You can look it up. Big IPO. He buys the Ferrari. Him and his lawyer go to the IPO party. You know what happens? They flip the Ferrari and die. Wow. So, we call it toxic wealth here in Silicon Valley. Yeah. People make enough money to go buy a plane, but they don't have the time to properly learn how to be a pilot. And they flip the plane. They flip the Ferrari. Just stay focused on the customers.

SPEAKER_284: Yeah. Focused on the company and your family. Who was it that said, stay hungry, stay foolish?

SPEAKER_285: Yeah. I think I know that guy. Yeah. He did okay, right? I mean, he did ride a motorcycle, but when he got older, he stopped with the motorcycle. That's the other thing.

SPEAKER_287: You don't ride a motorcycle, do you? I don't, no. All right. I have to give that speech to people.

SPEAKER_275: I had a friend of mine who had two kids, and all of a sudden, he decides to start riding a motorcycle after he has two kids.

SPEAKER_243: Yeah. No, I don't understand that. I'm a little bit more risk-averse with that stuff.

David Friedberg: As I get older, I get less and less about this risk-taking. You know, I just look at the actuary tables. I'm like, really?

SPEAKER_288: Yeah.

David Friedberg: You really want to get on a motorcycle and split traffic on the 280? You have that much money, you could have a driver and go to sleep. All right. Enough. We could talk forever, right? You know, I just haven't, I don't mean to bring everybody down. But the point is, life is worth living. There's so much great stuff you can do as an entrepreneur, as a dad, as a friend. Stay focused. All right.

SPEAKER_11: You know, Lon, I do think Jason's advice here can be encapsulated in the immortal words of Notorious B.I.G., who said, you know, never get high on your own supply. Because it sounds like some people make a lot of money, and then they go out and buy something silly and kind of crash and burn. Not that there's any parallels between that story in 2017 and today's market in AI whatsoever. No.

SPEAKER_138: Totally not. I always think of We Crashed, that WeWork series about how Adam Neumann kind of got a little high on his own supply.

SPEAKER_11: Well, actually, we got to give points to WeWork, though. WeWork's S1 filing was so bad that I came on this week in Startups to talk about it. And it was that interview that ended up getting both Molly and myself jobs at Twist.

SPEAKER_170: So shout out to Adam Neumann. Good and bad. Some pluses and minuses for that.

SPEAKER_11: We love Robinhood. It's gone public. It's now worth north of a hundred billion dollars. We love the archive going back and finding awesome tips and tidbits. If you want even more stuff like this in the interim, though, go check out episode 2000, where we really dove deep back into the archives of Twist. But for now, he's Lon. I'm Alex. That was Vlad and Jason. And we'll see you guys next time.