The Exchanges

Q Yeah. Can you break that down for us? I saw a couple of founders tweeting about that, but I didn't really have full context.

A Yeah, so I went to this, uh, investor event in Dallas, Texas in December, uh, that Bellerion Ventures put on, and Rick Perry was the special guest speaker there, and I almost think I was in the room when this got decided, but like Rick Perry's advice to all the vision people in the room was, you should join up with Texas, you should sue the NRC, because there's some clause in the, in the regulations that say the NRC has the right to regulate Large power plants, but it's not clear they have the right to regulate some of these smaller things because the risks associated with them are so much less if something goes wrong. And so I, I like, you know, Brent Kugelmess was in there from last energy and he's like one of the main parties to the lawsuit. So I feel like I was like a, an observer to history when the fission community decided we're gonna sue the NRC and get them out of Texas. And, uh, and, and if they, if they succeed and Texas becomes the regulator for SMRs, uh, that's going to be really exciting in, in terms of their ability to innovate and go faster.

AI assessment note: “Rick Perry's advice to all the vision people in the room was... sue the NRC”

Q information as quickly and condensed as possible. Other people might, might prefer an agent that speaks slower, And really draws things out and gives the full context and then lets them answer and, and kind of ping pong back and forth that way. Uh, is there any movement towards like that level of fine tuning that you think might happen in the future? Is it already happening? I don't know.

A Yeah, it's already happening. So I think that there's kind of two, uh, different flavors of this, um, and you can kind of combine them together, uh, over time. So the first one is that if you've interacted with like the, the real time models, like the ones that are So there's open AI real time API. There's Gemini multimodal live API. Um, there's a few others coming out as well. And, um, for all of these models, uh, that natively understand, uh, audio, um, you can actually tell them to, you know, whisper or slow down or, uh, speed up or act hyper. You can kind of give them an explicit signal of, Um, the style or the way that you want them to communicate with you. So that's already available and possible with these models. Um, then there's another part of it, which I think will come, uh, in the next, you know, year or two, uh, where the model will implicitly be intelligent enough to pick up on what your pacing is or your state of mind is just based on the way that you're talking and expressing yourself. Uh, and also if we weave computer vision into it, It might see you and understand from visuals, okay, this person is stressed, or this person seems like they're in a hurry, or this person is calm and like relaxed, and you can kind of tell that by your body language and by the way you're speaking, and it can automatically adjust you in the same way a human would be able to.

AI assessment note: “Yeah, it's already happening. So I think that there's kind of two”

Q if you go back like five years ago, people were really excited about voice, but then you had these sort of like high profile, uh, companies that didn't quite work, but you know, maybe in many ways they were just too early. So What are you seeing on the, on the startup side in terms of new companies being formed specifically to leverage voice and LiveKit and the underlying models?

A Yeah, we're seeing, uh, probably about, um, you know, we're doing like thousands and thousands, many, several thousands of, um, of like signups to the, to the cloud product or commercial product. And, um, most of those, the vast majority are, are startups and growing companies. And out of those, probably around 75% or so, 80% of those signups are, uh, are voice AI companies that are building voice agents. And so the way that I kind of see The market, uh, segmented to a degree is that you have the large AI labs and they have popular, you know, consumer apps. And a lot of them are building kind of open-ended voice agents that you can talk to about kind of anything, right? They're assistants. You can, they do question answering. They do therapy for mental health, all kinds of language learning in the case of speak. Um, and then on the other side, you have these kind of pockets that are what I call kind of voice native systems. And those are really anything you pick up the telephone, uh, To, you know, when you call someone on the other end, call a business, there's someone that answers that line and they're either, you know, doing patient intake at a hospital or they're doing loan qualification or insurance eligibility checking. Um, there's a lot of these kind of business process flows where, um, these are, there are pockets that are like really large in nature. So customer support…

AI assessment note: “probably around 75% or so, 80% of those signups are, uh, are voice AI companies”

Q information as quickly and condensed as possible. Other people might, might prefer an agent that speaks slower, And really draws things out and gives the full context and then lets them answer and, and kind of ping pong back and forth that way. Uh, is there any movement towards like that level of fine tuning that you think might happen in the future? Is it already happening? I don't know.

A Yeah, it's already happening. So I think that there's kind of two, uh, different flavors of this, um, and you can kind of combine them together, uh, over time. So the first one is that if you've interacted with like the, the real time models, like the ones that are So there's open AI real time API. There's Gemini multimodal live API. Um, there's a few others coming out as well. And, um, for all of these models, uh, that natively understand, uh, audio, um, you can actually tell them to, you know, whisper or slow down or, uh, speed up or act hyper. You can kind of give them an explicit signal of, Um, the style or the way that you want them to communicate with you. So that's already available and possible with these models. Um, then there's another part of it, which I think will come, uh, in the next, you know, year or two, uh, where the model will implicitly be intelligent enough to pick up on what your pacing is or your state of mind is just based on the way that you're talking and expressing yourself. Uh, and also if we weave computer vision into it, It might see you and understand from visuals, okay, this person is stressed, or this person seems like they're in a hurry, or this person is calm and like relaxed, and you can kind of tell that by your body language, and by the way you're speaking, and it can automatically adjust you in the same way a human would be able to.

AI assessment note: “Yeah, it's already happening. So I think that there's kind of two, uh, different flavors”

Q which one do you guys fit in? I think you're in the third, right? Changed the way the part is made, uh, developed a way to make stampings, complex geometry form sheet metal parts without dies. And so, uh, can you break us down? Like what is actually going on when this massive robotic arm is, uh, like pushing into this metal? Explain why this is important, how it works.

A Yeah. So we wanted to build a system. We call it the RoboCraftsman, right? And the idea was like, where do we get started? Um, um, and we started sheet forming. Sheet forming is the largest metal processing sector today. I think it's like a two hundred eighty billion dollar industry. You know, most of the metal parts you see day-to-day are sheet metal parts. Like, you know, you're sitting in your car, you know, you're in a sea of sheet metal. Every other car body is sheet metal, for sheet metal parts. Or aircrafts are basically sheet metal cans. Um, but today it takes a very long time to get your first batch of parts in sheet metal work, right? You know, you have to go make dyes, put them in giant stamping presses, like four-story tall buildings. Um, and then stamp your, your, you know, your parts out. Um, so our first kind of application of being sheet forming, um, we have two robots that form, um, you know, start from a flat sheet of metal between two robots. They have these giant fingers that are super strong, but they can basically push and pull on the metal that form it into a very complex shape without the need for any, any dives or tooling. Um, basically, you know, you get your, from idea to, to the first part in matter of hours, right? Um, it's similar to how a potter forms a clay bowl with their fingers. They're coming into the sheet, deform it, um, and, and, and shape…

AI assessment note: “we have two robots that form... start from a flat sheet of metal”

Q Uh, you've been in the game for a long time. Can you tell me the story of your first investment ever?

A Yeah. Um, our box group is named after our first investment. Uh, so, uh, it's like a combination. So the first investment's a company called boxy. Uh, it was a, uh, Roku competitor. Um, uh, we were in their seed round as a adorable check. Um, I didn't want to write my name as an angel investor. So I created an LLC called box group to invest in boxy. Which made me feel bigger. Uh, the box was like a cool nightclub in the city at the time. It felt like a cool word. Uh, I've looked with Aaron Levy. Uh, he registered box.net like two months before I did box group, which I'm bitter about. So at some point he'll sue me and it's all over.

AI assessment note: “So the first investment's a company called boxy. Uh, it was a, uh, Roku competitor.”

Q robot. It's more just like a machine. Then there's, you know, two axis gantries. I'm seeing a lot of robotic arms, a lot of six axis stuff. Uh, and then you get into the humanoid world. Uh, where's the biggest opportunity in terms of, or where's, where's the most need to deploy and like what type of robot is the most important to just roll out in America right now?

A Um, yeah, I think like articulated six axis robot arms are still the vast majority of use cases, right? Um, I think If you look at what humans are doing in these factories, like they're typically using their arms, not other parts of their body. So I think it's just an indicator of like that, that's the most common task, whether you're talking about, uh, assembly, whether you're talking about, uh, quality control, you need kind of vision, obviously, uh, whether you're talking about, um, packaging, whether you're talking about even kind of primary food handling and sorting in the metal fabrication world, you, you need a lot of machine tending, which is, you know, picking up a blank part and putting it into a CNC machine. Uh, we have a bunch of robots that are doing that. All of those are just kind of articulated arms, uh, that can do it. So I think that, you know, there are use cases for the humanoid form factor to kind of do new, new types of tasks. Uh, we also have some gantry robots that we've deployed that do certain types of applications where it's like very high payload, um, uh, work, but, uh, you know, the, to be honest, like the form factor is kind of irrelevant because in all of these types of robots that you're talking about, you have a couple of motors and some motion control, right? Like it's actually, Uh, if you can abstract away the software enough, like all of thos…

AI assessment note: “articulated six axis robot arms are still the vast majority of use cases”

Q we hitting a pre-training wall? Are we hitting a data wall? Uh, RL's very hot right now. Um, where do you think the foundation model labs need to go and, and what are you specifically excited about? I imagine that maybe a co-generation model, not super important to your business, but something that's more knowledgeable and hallucinates neck less about facts, extremely valuable. So what do you want to see?

A Uh, I think RL is the, uh, place where most investments are gonna go to, um, especially with, with models like O-three that are able to do tool calls pretty natively, uh, rather than being prompt engineered to do that, or, like, for example, before O-three, the way we built, like, our agents, is there would be one model that gave, came up with a plan for the query, another model that would execute the plan by converting the plan to, like, Smaller queries, filtering links, calling searches, and then another summarization model that actually takes all the results of the planner and the router and, like, summarize things. Now it's all, like, one single model. Uh, that's great. Like, that, that, that, that means, like, you have to, like, rebuild it. You have to throw away lines of code and rewrite it. But we've been doing this, uh, since the beginning. Like, people think, like, perplexes remain a stagnant code base or something. It's always changed as soon as, like, models became more capable. But the interesting thing about the native tool call kind of models is that if we want like a sonar version of the product that runs on our own, uh, setup, uh, we also need to start doing post training beyond just, uh, training for instruction following and summarization, but, uh, like also like two calls and, and, and completing tasks using RO. Uh, and, and we hope to collect all this intere…

AI assessment note: “I think RL is the, uh, place where most investments are gonna go to”

Q Can you talk about how you're thinking about your own models going forward versus leveraging the existing models? You guys have Sonar and then R, uh, and I'm curious how you're, where you're looking to, uh, focus going forward.

A Um, I think we'll continue to, uh, keep the same strategy, which is, um, have a version of our product that can run with our own models, but not hinder or disrupt the user, uh, from this best experience that we can provide to them. If we're not able to do it with our own model, we'll just use other people's models. We have no problems with that. Uh, my, our belief is that nobody's gonna have a lead in, uh, them having the best AI model for more than a few weeks. The, the pace at which the field is moving like anthropic did the 3.7 sonnet. Uh, and then, uh, within a few weeks, like Google did Gemini 2.5 pro, which was way better. Grok three came out and then opening. I has O three and O four. There's always some debate on like, what is the best model, but they're all looking the same and they're all good for a certain specific set of things.

AI assessment note: “If we're not able to do it with our own model, we'll just use other”

Q of funds that, you know, like KP, that there's now funds that are, like, effectively, like, four or five times larger, and then these sort of crossover funds that are getting involved, and are you seeing more Like, do traditional venture funds like KP ever try to sell into these rounds that are intended more for, ah, employee liquidity? Is that a potential future if companies are just staying private?

A Yeah, I think we're, we're just now in the early innings of this. Um, but we are a hundred percent starting to see it. We, we certainly have not done like a, a continuation vehicle. Um, is, is kind of like the most common term for this, like a CV where you can sell essentially like a slice of your fund. Um, cause you know, commonly venture funds have like a tenure life. And then sometimes you can have an option for like two years of extension. And then after that, you're kind of supposed to be, have the capital returned and like everyone can move on with their lives. Um, hopefully, hopefully much richer for it. Um, and, uh, and, you know, SpaceX has been a private company for what? 20 years now. Um, you know, over 20 years. Um, and so I think the, the most common vehicle for this would be a con, uh, a con, uh, a CV, um, where you sell either like a bundle of company, like a slice of your holdings or like a slice of your fund. And you're starting to see that like several funds have done that. I think, I think NEA Um, has, has, has done it in the last year or two, and you're seeing more funds do it. We certainly haven't done that before, but I think as an asset class, we're going to have to start coming to grips with the fact that the liquidity timeline for these amazing companies is very different than it used to be. Um, and it's actually, um, really flipped on its head because …

AI assessment note: “we're just now in the early innings of this. Um, but we are a hundred percent starting to see it.”

Q not Back in like 20, it must be 2008 or something like that, and, and you'd always have to, oh, I'm buying something, control T, new tab, search coupons, and Honey just baked that right into the browser. Made a ton of sense. Um, and so, and so, uh, what was the first dollar into the business? Like, how, how did you make the first dollar of revenue, I guess?

A The first dollar of revenue didn't come until 2015. So, let's launch the product 20 12, 2015. Um, we pieced together kind of a rolling seed round of anybody that would take a flyer on this company with a couple 100,000 users and maybe interesting, but haven't really figured it out. And we use that to hire somebody with a background in affiliate marketing. And we had, we being George and I, co-founders of the company had thought that this might make sense. Cause after all retail me, not just mentioned and other companies, that's how they monetize this. Uh, when you click on that retail may not link, it's actually affiliate tagging you at that moment in time. And so you go to do that coupon search and whether the coupon works or not, it's, uh, they're getting an affiliate commission on that. And that's why there's the, they hide the code and open multiple tabs and do all that stuff. So that was how it was. And we're like, Hey, we could probably monetize that way. Uh, we, George and I reached out to the affiliate world and. They, they basically were like, you guys are a toolbar. You probably are up to no good. Go away. And so we did. Um, we didn't realize that there had been some bad actor toolbars in the space, uh, immediately prior to that, which was, uh, it's like a surprise to learn, but we went back to just building a cool product for consumers. And then, um, when we hired so…

AI assessment note: “The first dollar of revenue didn't come until 2015.”

Q post. Uh, out of the air, you know, and read it on the plane, right? So it's like, how do you, how do you get people to sort of, like, understand, uh, the, the change that's happening now, and, and the coming change, and does that, and does, like, kind of helping people understand that excite you, or do you not feel like you even have the clarity, uh, yourself?

A I definitely feel like I do not have the clarity myself. And my, my take on this too, is I'm not sure anybody does because the history of every big innovation is looking back in hindsight, virtually nobody got it right. And even the most diehard optimists underappreciated how much change it was going to bring. My favorite example of this is when the Wright brothers created their plane, they virtually, they only marketed it to the U S army because they themselves, the Wright brothers did not see much commercial use for the airplane. They knew you could strap a machine gun on it and drop bombs out of it. And then the army would like it for that. But the idea that the Wright brothers themselves foresaw Delta airlines, like traveling, like absolutely not, not in the slightest. That was true for cars. That was true for computers. It was true for the early internet that even the pioneers who are the crazy wild maniacs underappreciated how much change it was going to bring because every new invention, it's not what you create. It's what other people like manipulate Like, like what you do. I remember one example, it was, um, for Photoshop, a lot of the, the, the, the engineers at Adobe, when they make new tools for Photoshop, they have no clue what artists are going to do with those tools, but they're like, let's just find every way to manipulate an image and someone else will figure o…

AI assessment note: “I definitely feel like I do not have the clarity myself.”

Q Is this kind of, like, scaling? Are we, are we compute bound here, or is it algorithm bound?

A Right now, I think ARC AGI two that we just introduced this week on Monday, uh, is pretty direct evidence that we need new ideas still to get to AGI. Um, we shared, I don't know if we can throw it up or we can put in the shows or something. Um, there's a chart I shared on Twitter that shows the scaling performance of even frontier AI systems like O and pro and N three against the version one of the dataset that we've been using for the last five years. And then version two that we just introduced this week. Um, and basically it resets the baseline back to zero percent language models, uh, pure LLM systems are scoring like zero percent now, uh, again on, uh, on arc B two, um, single COT systems like R one and O one score like one percent. And, uh, the sort of estimates we have right now for the frontier systems that are actually adding pretty sophisticated, you know, uh, synthesis engines on top, like a one pro and three single digit percentages on their efficient compute settings. So, um, I think there's a very interesting point that like, You know, we, we kind of moved from this regime where people were claiming, oh, we're just going to scale up language models, right? More data, more parameters. We're getting to AGI and people realize, I think that's not quite the story now. And then there's a new story that's emerged over the last like five months, which is, oh, we're going …

AI assessment note: “pretty direct evidence that we need new ideas still to get to AGI”

Q acts and it gets bad and there's just all these balancing. And then also we do the mice studies, but does that really transfer? And you know, what happens when someone's like, you know, vaping pink liquid for, you know, 50 years, like we don't really know. And we're kind of figuring it out right now. Um, how, what, what gives you confidence about the, the low impact of this?

A So, you know, I, I really sympathize with people that are worried about this because it sounds on his face like crazy to be modifying the weather and to be dispersing chemicals. But, um, to your point about like vaping stuff for 50 years and not having longitudinal data on the health outcomes, like cloud seeding, even though Rainmaker is very innovative, even though a lot of new things have been done in academia that we're implementing from like the last five years, cloud seeding is 80 years old, dude. This was invented in the United States in 1945. GE has the first patent from 1946. People have done 80 or 30 or 50 years, depending on the watershed, studies on the concentration of silver iodide that ends up in the water and the soil. And after decades of operation, you only see parts per trillion of this stuff. Like, you need super sophisticated instrumentation to even detect it in the first place. Um, and so there, there's been no negligible health impacts found to, uh, either people or agriculture or the environment. Um, and those studies, they've been done. And like, should Rainmaker replicate those? Should we continue Continue to do it just to prove it out further. Totally agree, but like, it's totally safe, and if you think about the, like, LB-fifty, like the, the lethal dose of silver iodide, it's less than salt. It's less than table salt. It's milligrams per kilogram, uh…

AI assessment note: “cloud seeding is 80 years old... studies on the concentration of silver iodide”

Q So anyway, uh, can we move on to the Cygnus spacecraft? Can you break down? I, I'd never even heard this word term before. Can you give me a little bit of history and then what's going on with the news?

A Yeah, I mean, I brought it up because I think it was like two weeks ago that we chatted about like the, you know, astronauts being stranded, and then, you know, sort of SpaceX going to, you know, bring them back down. Um, this past week, it was revealed that the Cygnus spacecraft, which is a spacecraft, I'm pretty sure, made by Northrop Grumman, um, that originally was awarded basically commercial, uh, cargo contract, the ISAC. So this was like back in like, 2010, 11. It was basically, you had Cargo Dragon, which was done by SpaceX, and then the Cygnus spacecraft done by Northrop Grumman. Um, to basically deliver cargo, so like food, water, et cetera, to the International Space Station. It generally performed, you know, well, it's kind of a crazy vehicle in that it's only one-time use. Like, it doesn't even have, like, a shield or any way to, like, come back to it, so you literally, like, ship it up to the ISS, it drops off its things, and then it just, like, burns up in orbit, so it's, like, the epitome of just old-school, you know, sort of aerospace. Um, they didn't even, like, attempt to make it economical from the, like, fundamental design phase, but, like, it's, you know, delivered plenty of cargo to the ISS at this point. Um, this past week, um, They dropped it. Um, and so NASA was like, look like, yes, this has been tested, et cetera, but like, we don't know what happene…

AI assessment note: “Cygnus spacecraft, which is a spacecraft, I'm pretty sure, made by Northrop Grumman”

Q and for companies that are in defense at all, uh, are, are there the actual machines themselves that then sit on the factory floor? I imagine a lot of the best ones are built and designed and manufactured, you know, everything like in China. So how do you, how do you think about sort of the supply chain for the autonomous factories that you guys are building and sort of.

A You know, it's a, it's a very hard question that the country's got to answer. So the machinery market, so we invented in Air Force invented CNC machine. Um, and we no longer make CNC machines in the US. And I think you saw this article about, you know, Haas white labeling some China ones and selling to the Russians. Um, but the main, the main manufacturing markets where we buy from is China's number one. We don't buy for them because they got spyware all over this stuff. And then it's, you know, Germany, South Korea, Japan are the best machine makers. And it's kind of like the car market, you know, you can buy an Italian one and it behaves like an Alfa Romeo and it's really fast, but it's going to, it's, you're going to be in maintenance hell forever. And the German ones are really stable and the Korean ones are quite good. So. You know, if the balloon goes up with the conflict, uh, you know, can we even buy the amount of capital equipment to put in the factories because we don't produce that capital equipment in the U S anymore. So, you know, part of my job is. Grow super fast and buy all the capex from our, from our allied partners and make sure it's in the US. But, uh, yeah, we no longer make the underlying machines in the country anymore.

AI assessment note: “part of my job is. Grow super fast and buy all the capex from our allied partners”

Q of bounce around. They're using cursor one week, and then cloud code the next week, and then they're bouncing to this one. Um, what are you seeing, and what do you think's happening, and do you think there's any, uh, any merit to that fear in the market, or on X right now, that maybe the AI revenues that we're seeing are not as sticky as previous SaaS installation cycles?

A It's certainly suspicious. Uh, I mean, it, anyone who tells you differently is sort of, and I was a little worried about our growth. Like, is it, is it going to go away? What's going on? Like, this is like a little too fast. Um, and I think, uh, you know, we really start focusing on the, on the metrics that really matter, which is the revenue quality retention, things that more advanced companies are thinking about, like net dollar retention. We actually have above hundred percent net dollar attention, even for our consumer plan, which is probably unheard of. But, um, uh, you know, I, I think that, uh, I think maybe ARR is becoming a bit of a vanity metric, like how fast you can grow the ARR. I think you want to make sure you're providing an awesome service and also make sure that people just want to stay on your platform. I think, especially on the VS Code Wars, the switching costs being so low, Is, is going to be a challenge for them. You can go from Copilot to Windsurf to Cursor in like five minutes. Same project. Quite like it's one click. You can open the project. And so they're going to have to find a way to differentiate. The great thing about Replit, we provide this really great convenience. And so if ain't broke, don't fix it. You deploy something on Replit and people, people really tend to stay.

AI assessment note: “It's certainly suspicious. Uh, I mean, it, anyone who tells you differently”

Q Well, we got to ask about the Vercel drama. We talked to Guillermo earlier this show. What's going on there? What's your take? Can you break it down for us?

A Well, you know, I, uh, someone asked me, um, do you, do you plan on supporting Next.js and Repl.it agents? I said, no, because, uh, we want to bet on something like Vite that is open, like community builds, uh, and the way Next.js has been trending is that it's optimized for Vercel. They disagree, but it is actually not a controversial statement in the larger community. There's a project called Open Next that makes it so that, you know, uh, Next.js is deployable on more platform. Matt, the CEO of Netlify, commented, uh, agreeing with me. The larger community feels like, you know, Next.js is Is perhaps more tied to Vercel than they would like to admit. And I think they overreacted to my comments and, and, and they, and they strized my, my comments. I think it would have been, uh, uh, less of a, a story if they, you know, they didn't overreact to it. And I look, uh, uh, Vercel is an amazing product. It's an amazing brand. They've done a great job managing a lot of open source projects, not just Next.js. And for me, water on their bridge. Like, you know, I, uh, You know, I, I think, you know, emotions get, get high on Twitter and you start calling and, you know, I wish them really the best.

AI assessment note: “the way Next.js has been trending is that it's optimized for Vercel.”

Q that they did and it was sort of the The unintentional effect is that the ad product gets worse for advertisers. Like, why is it, why is it happening? And are your, uh, contacts at Meta saying, Hey, we're act, we know this is not great and we're actively trying to get back to where we were, or is it just get used to it? This is the new normal.

A Totally. Yeah. No, some of the things are out of Meta's control. Like Meta has obviously gotten less efficient post iOS, which was when Tim Cook and Apple really put a lot of restrictions around what data did Meta have access to. Um, at that point you went from, as an example, being able to optimize for a 28 day outcome. So if you're optimizing for a purchase that could happen over 28 days and Meta could take, could attribute that, that purchase to itself. Now that's just a seven day period, right? So now it's like a much smaller period. You have, uh, Fewer or less ability to exclude the customers that you want, and Meta's began weighting these click-based outcomes. So all of those things together means Meta's only able to attribute orders to the platform on a shorter time frame, there's less exclusions, and it's serving to people who have already seen your ads. If you're trying to introduce new people to your brand and actually generate awareness, generate intent, and, uh, drive incremental orders, that's just a harder environment to do it within. And those are like, That's half things that are out of their control. And then part of the things that I mentioned are within Metis control, and I think they are trying to address. Um, so the answer to your question, Jordy, would be a combination of the two. Yep.

AI assessment note: “part of the things that I mentioned are within Metis control, and I think”

Q Um, out of your, out of your portfolio, you know, maybe some more recent investments. What, what do you feel is the, the most under-hyped, uh, Portco, the company that, you know, should be getting, uh, more headlines and attention, but isn't for one reason or another?

A I mean, one that flies relatively under the radar, but I'll tell you why it's not going to be under the radar for very long is a company called chapter, which is based in New York city. And the founder is Kobe. Kobe worked at, uh, Palantir back in the day. The reason it's not going to be under the radar for very long is, um, up until two years ago, Kobe had a very interesting board and a very interesting co-founder. So his co-founder was Vivek. Um, and his board was JD and Peter Thiel. So talk about going from like, oh, we have this great board dynamic to all of a sudden being like my board members, the vice president of the United States. And, you know, Peter's still very involved in this company is doing really meaningful work and AI is going to accelerate it. It is helping this very big problem of senior citizens in the U S like navigate the Medicare, Medicaid healthcare system, pick the right plans. It is Super hard to do. Um, there's all these contingencies and other to understand the flows and prescriptions and what stuff are you on? What state are you in? And so they're making that process easier. It should help healthcare costs in La Ronde. It should help patients. This is really important work, but just talk about up leveling your connections in the world, uh, by the facet of who you chose to be investors. Um, we were the, led the seed there and, um, it'll be a many, m…

AI assessment note: “one that flies relatively under the radar... is a company called chapter”

Q Journalists are underrated. Underrated Bloomberg. Who is this?

A This is Wall Street Journal opinion section. It's great. Uh, California regulators and the Biden administration tried to boost the electric truck market with mandates and subsidies. The Inflation Reduction Act included a 40,000 dollar tax credit for buyers of electric trucks. EPA rule last spring requires that electric truck models make up 25% of long-haul tractor sales by twenty-thirty-two. Man, talk about a bull case for Tesla now. It's crazy. Uh, another electric truck maker, Lordstown Motors, filed for bankruptcy in twenty-twenty-three. Fisker failed last June. Rough time. The British electric bus startup arrival sold its assets. Lots of bad stuff going on in the EV truck market. Power law rules everything around me. That's the story here. Anyway.

AI assessment note: “This is Wall Street Journal opinion section.”

Q Uh, they are making it available to pro users, but why don't you get into what they're actually doing?

A Yeah. So they, in the announcement here, they say they're releasing operator, which is an agent that can go to the web to perform tasks for you. Now, ChatGPT four O has already been able to go to the web and retrieve individual HTML pages, suck them in and use that as context. So it can get more up to date data, but this is a step forward in that the operator It has its own browser and it can look at a webpage and interact with it by typing, clicking, and scrolling. It's currently in research preview, meaning it has limitations. Uh, but this is one of the first agents, which are AIs capable of doing work for you independently. You get a task and it will execute it. And this is fantastic. I think this is gonna be very, very helpful. Um, there's already been the ability to use ChatGPT to write a little bit of code, like little code snippets. So I was doing something where I wanted to, uh, scrape A website and just paginate through there, like, 20 pages of standard formatting, like, you know, list of, uh, of like, I think it was like podcast episodes and I just wanted them all on a spreadsheet basically. And it was able to write the Python code for me, but it didn't really work because it needed like JavaScript and stuff. It was like, it just wasn't, it just wasn't quite there. But with this, it's, it's like the AI has a web browser that can fully render a webpage and just click t…

AI assessment note: “they're releasing operator, which is an agent that can go to the web”

Q Do you want to re-explain the bucket?

A Yeah, yeah. Okay, so, so, so we have, so, so we have, you know, we react to a lot of, A lot of posts from X on this show. We wanted to create some differentiation here. And so, uh, when we see something on the timeline, that's a banger, it's, it's from a friend, it's from a brother, maybe a previous brother of the week, someone really important. It's about a really important topic. We put it in this stack. We print it full size. It's a banger. This is from Max Meyer. You know, it's going to be good. It's important. He's setting the discourse, but every once in a while we see posts that are good, but from people we don't know, Maybe they're just kind of random, timeless. We could just react to them whenever we want. So we print them on these small four by six cards, and these are the bucket poles. We put them in the bucket.

AI assessment note: “we print them on these small four by six cards, and these are the bucket poles”

Q You look like you're having fun, but your, your LinkedIn profile is dead serious. Uh, do you, do you, do you, uh, you know, Like, I'm sure internally as a team, you guys are focused on not letting the hype and the crazy initial traction get to your head. How do you think about running the team when the team is only experienced sort of massive sort of immediate success?

A Totally. And I think one thing that compounds this is that so much of our team is like college dropouts and new grads, right? Cause it was the extension of our initial network at Harvard, Stanford, MIT, et cetera. Um, and so a lot of them haven't Seeing what normal companies look like. This is all that they're used to, and it's easy to get caught up and lost in so much that's happening. But I think the most important thing to stay grounded is just focusing everyone on the numbers that matter long-term, right? The network effects of the business. How do we build up the marketplace? How do we learn from all the performance reviews that we're getting to build our usage data flywheel rather than focusing on a lot of the lagging indicators like Revenue or, um, you know, or more of the customer signals.

AI assessment note: “most important thing to stay grounded is just focusing everyone on the numbers that matter”

Q Could it be something to do with their fee streams?

A Well, someone had a much more, uh, nuanced take from this, which was that, uh, celebrities are becoming, like, multi-platform, multi-disciplinary. There actually are multiple income streams, and so, this was not, I, I gave a quote too that they didn't use, but someone, someone chimed in and said, like, yes, a lot of these people are, a lot of these celebrities are doing, Um, much more complex deals. Uh, starting brands, starting companies, building holding companies, and so they actually wind up interacting with, like, the finance bros much more, much more often, and it's not so much like, oh, you got set up with someone with money. It's much more like, oh, you work on deals together, and you talk about the same things, because when you're a celebrity at a certain level, yeah, you're going on set and filming a movie for a couple weeks.

AI assessment note: “There actually are multiple income streams, and so... celebrities are doing much more complex deals”

Q a weekend, a 100,000 dollars. And that ladder of price discrimination is a hard for a lot of businesses where it's not token based pricing, where it's just as you know, Uh, consumption-based usage of the product, but actually getting the place where, look, if you have a billionaire in your audience who loves what you're doing, how do you actually get an offer to them that's the right size?

A Um, I'm a, so, customers are super fractal, right? And so, you can absolutely make the same amount from one percent of your customers as you can from the other 99, and have half your revenue from one. And typically, like, at each level, you have another double. So it's like, if you're at, let's say, a hundred dollar a month price point, there's gonna be another double of revenue at a 500 dollars a month price point, another double of revenue at 2500 dollars a month, another double at, you know, whatever, five or 10 times that is gonna be. And so, Um, I think making offers available to people, that is where business owners like struggle. They feel weird about it and they sell out of their own wallet when it's just like, just make them available and state the facts and tell the truth. So it's like, if we had one big plaque in our marketing team, it's state the facts and tell the truth, but it's the whole truth, which is, Hey, we're going to do this thing. It's absurdly expensive. And it's only for somebody who, if you're reading this and you're like, Oh, that's not absurdly expensive for me. It's for you. And if you read that and you're like, oh my God, I'm offended by that. Good news. We have less offensive prices that are lower. And for that, we do all these other things that are, you're going to get way more people who are going to be doing it with you. But if you're a special…

AI assessment note: “just make them available and state the facts and tell the truth”

Q Um, what's a category of business that you haven't, uh, built or invested in, uh, that you're excited to at some point?

A I'm still looking for the, the magical med spa. Um, I think med spas are super, I love pseudo medical, um, because you have maxing. Yeah, all of it. Well, think about it. Like there's this aging population that has more money than anyone else. They don't want to age. They see the Brian Johnson's. Like it is the zeitgeist of right now of like, I want to live forever. I want to look beautiful and young forever. And they have all the money. And so if you look at the demand side of that, it's huge. And you look at the supply med spas are, they're still wildly understaffed. And I know this because I see business owners every single week. And so I've like, there's a handful of categories right now that I'll meet business owners and they're doing better than they should be. Like there's some business owners I meet.

AI assessment note: “I'm still looking for the, the magical med spa.”

Q Yeah, I mean, isn't it from, like, min maxing?

A Min maxing, and that's what I'm getting at. Like, you shouldn't be token maxing or token minning. You should be token min maxing. And the idea of min maxing is when you're in a video game, you want to get the most resources for the lowest cost. So if you're playing League of Legends, you're going to put just the right amount of, of gold into this particular item or this particular stat point in an RPG, and you're going to try and find the optimal build based on your resources. You have confined resources. And somehow, It feels like these like tech and all these companies went through some sort of hallucinatory, hallucinatory period where they lost sight of that. They should always have been min maxing because that is the goal of business at all times in all things. You would never say like, oh yeah, we're ad maxing this quarter. We're going to buy as many ads as possible. Oh no, it didn't work. We, we spent way too much on ads. They were not effective. We need to admin spend as little as possible on marketing. It's like, no, you always want to be spending. You always want to be getting the most return on ad spend, the highest leverage output.

AI assessment note: “Min maxing, and that's what I'm getting at. Like, you shouldn't be token maxing”

Q Yeah. What about sports? I've seen some, uh, some robots playing tennis. They're going to be playing basketball soon. I don't think I'll be watching robot basketball, but Uh, how do you think live events, sports, basketball will change over the next decade?

A I mean, maybe for the referees, like you've seen in tennis, but that's it. I think in reality, more people will want to go to in real life events than before, because if, you know, if you're just managing agents, looking at output, looking for exceptions, you're going to want some human touch, right? You're going to want to be able to engage. And I think that that's really, really important. Um, and I think that's where sports will grow. I mean, and you're starting to see that now with, You know, what happened with the Olympics, the world baseball classic, you know, became much more popular because people want that disengagement from all the, the, the stress that that's happening right now.

AI assessment note: “I think in reality, more people will want to go to in real life events”

Q You know, the, the humanoid, uh, founders will tell you, you know, how do you solve stairs, right? Like, it doesn't work for wheels. But if the robots are really great, people, you could put, like, a mini robot, uh, uh, elevator, right? Like, if it's on wheels, like, you could just put it, yeah.

A Like, you see, like, the old house dumbwait, um, dumbwaiters, right? Where there's just the thing where you pull it, you put it in there, and you pull it up, and it goes up to the next floor, and somebody opens it up. You're gonna see, you know, a mechanized equivalent to that, right? Where the, it's, it recognizes the little, you know, ant robot that's coming up, and it opens up a little door that's, you know, that leads to the size of whatever it is it needs to carry or whatever, and then it goes up the dumbwaiter, does its thing on the next floor, the next floor, the next floor, and does what it needs to do. I don't think, you know, stairs are an issue at all.

AI assessment note: “I don't think, you know, stairs are an issue at all.”