Episode 2: Inspiring the Next Generation of Founders, the Latent Space

Hello, everybody, and welcome to Friends in Code Spaces, where we lean into the theory that IRL relationships can actually be fun and in a podcast. So stuff happens in real life. Alessio, I run into you in real life, and I love it. And what I love about that is you are always smiling, you're always, you share so much knowledge, everybody knows you, everybody's like, hey, high five, high five, and I never get the chance to sit and talk with you. And I have learned, oh my gosh, I have learned so much from you and the content that you create, as well. So welcome to the show. Thank you. You know, I mean, you always have cool stuff going on when I meet you. I think one of the first times we met you had like a robot that talked with an LLM. So it's just as fun for me to meet you. Awesome. So I think one of the misconceptions that people have, like maybe you have, maybe people think that I'm the robot guy. I'm in different code spaces, right? I end up in different code spaces. Lately, I've been in the AI code space. I know that you're in the AI code space as well. Is that how you would define your current code space? That sounds accurate. I would say, you know, AI is such a horizontal technology in a way that like it's almost limiting to call it its own thing. It like permeates everything. But that has been definitely my focus the last year and a half, two years. Okay. Awesome. Awesome. So for those of you who do not know, which I know there's like six people that probably don't know who you are. This is Alessio from the co-host of the Latent Space Podcast and the partner and CTO of Decibel Partners. For me, I'm putting the Latent Space Podcast first because I am such a huge fan of the work that you all do. And I know that you know that. But so excited to have you here today. We're going to play some games. We're going to get together and play some games. So you this first game that we're going to play is called Context Window. So I built something that's going to help us get through this. So I asked an LLM, and you know that LLMs might not have all the training knowledge that they need. What they know about you, what they know about Decibel Partners, what they know about the Latent Space Podcast. And not much. It didn't know much. But you know what? It's interested to ask you some questions. So let's give you a little look. And how's your eyesight, Alessio? Can you see that? It's good. Yeah. I mean, it's quite close. I couldn't see that. So doesn't know much from the knowledge cut off about you, which is OK. I think that's totally fine because we're going to fill the context window. So the next time we're going to, this is going to get transcribed. We're going to push that out there and it's going to know more about you. You specifically. I'm sure it has a lot of your content. It's out there. The AI is trying to grow and improve itself because I told it to in a system message that it's trying to get better. So it really thought about the topics that it wanted to ask you about. And so the part of the game that we get to play here, it's looking forward to learning more about Decibel Partners areas of expertise and their clients and that sort of thing, which was what it was interested in. But if you don't want to talk about that, you don't need to. Because I gave it some context too. I gave it your little talk about what what's going on here. Give a little additional context about about the podcast. So these are your topics that you get to choose from. So what would you like to what would you like to choose there? Oh man. Let's start with Decibel. I think there's a small world story. OK. OK. At Decibel all we do is invest in early stage technical founders. OK. And all of us used to work at these large multistage firms and always felt that pre-product market fit. It's really hard for the average VC to understand a technical product. So we built a new firm from the ground up to do that. And my partner John was actually an early investor in Coppler. So last year at RSA, our founder's dinner, Michelle was our guest speaker. Matthew Prince was a speaker at our AI Pioneers event last fall. So there's kind of like this Decibel -Coppler bond that has been there for a while. So I'm kind of disappointed in the model that it didn't make that connection. They should have, probably should have. It's a Cloudflare hosted model too. So that's really we should we will work on that. All right. I'll put that through. This should probably happen automatically. But like I said, we're in a beta version here. But here here's somebody I don't know. Who should we have talk here? I think Callum is going to give us a little going to give you this question. How does Decibel Partners expertise intersect with the themes and trends Alessio has discussed on latent space? And are there any common threads that tie the company's work together across different industries and domains? I would say latent space, really the core behind it is creating this AI engineering practice. That's something that SWIX, my co-host, kind of pioneered. And it's all about finding things that people understand are important, but don't quite know how to talk about, you know, like I think AI is in that space where today is slightly getting more mature. But when we started the podcast a year ago, everybody was like, oh, yeah, language models. And it's like, how do you use them? It's like, oh, I don't use them. It's like, but they're important, you know? So the goal of the podcast was like, how do you take things that are technically interesting and bridge the gap between what's commercially viable? So the work that we do at Decibel is often taking founders that are at the forefront of these technological waves and help them build great businesses. And that's, you know, Cloudflare is a great example. You know, like Matthew was like manually collecting domains that he was putting on the list. And then Cloudflare now it's like a, you know, a massive company that is like super successful. And what we want to do with latent space is inspire the next generation of founders. It's like, hey, you love AI. You maybe don't quite know how to make it into a business, but you understand that the technology can change a space that you're interested in. And we work with founders to bridge that gap. Can you talk a little bit for those of us who haven't been through a VC flow like that? Can you talk a little bit about what do you mean by that, by health? Like, what does that what does that practically look like? Yeah. So, well, a lot of times, especially in ads capital. So I led the seat for Jeremy Howard's new research lab. Jeremy, obviously, is a great tech, technological. How do you say technologician? You know, I grew up in Italy, so I'm just making it up. Dude, that word that you just made, it's like a magician of technology. Yeah, he is kind of a magician of technology. Technologist, I think is what you were looking for. But I'm going to give Jeffrey, I think I think he is a technologician. I like that. And so the investment there is like, how do we take the ideas? Jeremy was eager to test AI. He actually helped a lot of people that are deep learning engineers today learn. How do you take this technology that is being almost kept behind closed doors at OpenAI and Gemini and DeepMind and all these labs and move it to a place where most enterprises can leverage it? You know, so the work that we do as VCs, one, obviously fund the company before you can actually make any money. And then how do you connect people on the enterprise side that need to solve these business problems, but they themselves cannot be in touch with the early stage community? So we kind of bridge this gap between the two. And then we do the usual work of, you know, helping with recruiting. Obviously, Jeremy is like, you know, everybody knows Jeremy. Everybody wants to work with him. But with a lot of founders is how do you build the company? How do you hire for roles that you've never hired before? So you're a brilliant engineer. How do you interview a VP sales? That's not easy for somebody to figure out. So that's a lot of the work. And then helping the founder go down the right commercial route. You know, I think a lot of times VCs that are less interested in the technology are more focused on monetization, especially with technologies like AI. It's more about the adoption and the data flywheel. So as an investor, you have to understand what really matters in the market. Awesome. Yeah, yeah, yeah. And you're bringing wisdom. You're bringing wisdom so that they sell. That's awesome. That's really cool. I hadn't thought about that before, actually. Thank you. Thanks for that. Oh, yeah. No, this content is actually helpful. I hope. Can you choose another one? Let's choose another one. Let's do late in space. Let's do it. Let's do it. We can't not touch late in space. We've got two questions here, actually. I think I like the second one. Yeah, I like the second one. So how do you see late in space evolving in the future? I feel like you are constantly evolving that show, which is awesome. And are there any new formats, topics or themes that you're considering exploring to further the podcast mission of exploring the intersection of AI and engineering? There are a lot. Swiggs and I always think about this. So when we started, the only format we had was a one on one founder interview. So somebody will come on and tell us about what they're working on and whatnot. Then we decided to bring on researchers. Then we decided to bring on folks like Logan Kilpatrick, who was the head of developer relations at OpenAI. And then we're like, OK, there's the people, but then there are like the facts. Right. So like papers. So we started doing a weekly paper club in our Discord. So every Wednesday at noon, there's a paper club. We discuss one paper. Sometimes the authors come by. That's not recorded, actually, today, but it's more because it's quite chaotic sometimes. Just like a lot of people and like people ask questions and whatnot. And then we started doing on Fridays, AI in action group. So folks that are deploying AI in production, they do like a short talk on a specific topic that might be RAG, that might be evals, that might be all these different things. So we try to branch out from how do you learn, which is the podcast, to how do you like stay in front of the cutting edge that might not be applicable today. But you want to understand in two years, what are we going to see that is going to change the way it works? Yeah. And then the AI in action is I just need to ship this thing next week. How do I do it? Can you help me? Yeah. So that has been on the community side. As far as content goes, you know, we started doing these live from conference podcast. So we had a New Reps recap. That was about six hours of content. So there's a lot to recap. SWIX is actually at ICLR right now. So we'll probably do an ICLR recap. Then we have the AI Engineer Summit. There's an AI Engineer World's Fair in June. There's a lot going on in latent space. It's like a lot of interest, obviously. And we had, you know, almost two million people in the first year come to latent space on the newsletter and the podcast and the YouTube channel. So it's been quite the community. Yeah. Yeah. Yeah. I mean, I love I love what you are building and how how helpful it is. I think that like I spent a lot of time on the BART here and you spent a lot of time with you. You might not have known I was with you, but you were with me. I know sometimes at times people stop me at like events without looking me in the face. Like they're behind me. Yeah. And they're like, oh, hey, I know. Oh, you're Alessio. No, I heard the voice. You're the guy from the podcast. It's like, oh, yeah. OK. Great to meet you. That's awesome. That's awesome. Actually, I want to talk. I think we're I think we're ready to go to the next game, I think. I think we're ready to go to the next game. I'm ready for anything. You're ready for anything. Remember what you said. So you've seen the the you know what Hot Ones is. Mm hmm. Yeah. We're gonna do something called Cold Ones. OK. OK. Somebody's walking in. Somebody's walking in. Producer Peter's coming with a thing. Now, do you know what this is? You just said earlier that you were. Yeah. Yeah. You had. I grew up in Italy, but I've seen these on The Simpsons. You've had it on The Simpsons. Exactly. Pina Colada in red. So your choice. Do you want red or pina colada? Red. Red. You're going red. So you haven't had one of these before. No, I actually don't think I have. If you were you said Rome. What's the the the the when it's so hot, you have to have a treat as a kid that you want to chase. Is there like an ice cream truck or what's the thing that you do? Is it Italian ice? Is it called ice? Yeah. So when you're. That's funny. When you're. So if it's hot, you're at the beach in Italy. Yeah. Because there's not much to work on anyway. And we have this like a carts with like big blocks of ice. Yeah. And you just run up and do like shaved ice and then they put syrups. OK. So it kind of looks like this. I think we're going to. It's probably much healthier. Let's. Oh, I mean. Let's let's cheers. Cheers. And I want you to have your. This is your first Slurpee. It is. OK. Let's get it on camera. Let's do it. Let's do it. Hmm. Cherry? Possibly. Yeah. Well, definitely not actual cherry. It's red. So what I'm trying to bring here is a feeling of nostalgia. So let's go back to that ice that you had on the on the beach there. And I want you to think about the first time that you were around a place where you're like, maybe I like coding. Maybe I like technology in this way. Um, yeah, there are a couple of things. So I grew up in a family with not a lot. And my first computer was something that my mom's office was going to throw away. She was a, you know, like a cashier. Yeah. And it was like this old iMac G3, you know, the ones with the. Yeah, gorgeous. The color back. Yeah. No, it was cool when it came out. But when I got it was like 15 years old. But it was really cool to just have something that you can like create things. Yeah. You know, and then what were you creating at the time? I would just like, you know, nothing really like using like, you know, drawing software and whatnot. And I was like, you know, computers are kind of cool, but whatever. And then I got my first PSP, you know, the PlayStation. Oh, yeah. And we could, you know, couldn't really afford more than one game per year. But there were just like so many good games. Yeah. I'm like, I need more than one game. Yeah. So I go online. It's like, how do you download PSP games for free? And I'm like, and it's like this whole world of like, you know, homebrew software and whatnot. So that was like the first time when I was like, oh, man, if you learn how to program these things, you can like just get anything you want. Yeah. As a kid, that was like getting a video game. Yeah. Do you think that's true still? Yeah. Yeah. If you learn how to code, you can get people, you know, like you can create agents to do things for you that you wouldn't need a person to do before. Yeah. So that has always been a pretty important lesson for me. OK. You know, sometimes you need to understand how to leverage the systems to get the outcomes that you want. Yeah. Yeah. Sometimes people are just looking for the system to do the thing. Yeah. You know, versus building it. It takes more work. Yeah. You can do things that are for you. So you let's. Yeah. I know. I can see how people get hooked. Wow. That one tasted like suntan oil. Yum. Let me try it later. Sounds great. So. So you're playing on the computer. You're thinking about coding. Did you get into a coding space? Did you. Did you ever code? Yeah. Yeah. Yeah. Yeah. I mean, I. So I started doing the PSP stuff. I was probably like, I don't know, 15 years old. Yeah. I think. And then I went to study computer science and I dropped out of school to start an open hardware company. OK. I learned to code and I learned how to design boards. Oh, nice. That was I wouldn't recommend it to anybody. Don't do a hardware company. As a VC, I still give that advice. And then I ended up leading engineering at a bunch of startups. And that's always been like for me, like coding has been part of the DNA. What was that first code that you did? Um, the first code was definitely getting a PSP to jailbreak. OK, so that was definitely the start. I'll never forget it. And it was really my first job, like doing it for people at school. So you took money to. Yeah. Oh, for sure. Freelance. It's like I already cannot afford things. I need to share this with others. That's incredible. Yeah. So if Sony's listening, don't come after me. You already came after George Hotz that came on our podcast. Don't come after me. It's like the schedule limitation. I'm sure it's over. All right. So you are you have all the wisdom that you have now. You're you're giving this to other people. If you could go back to PSP days, Alessio, what was the piece of advice that you give right then to to to you there? Well, what advice would you give? I would say, yeah, I would say my advice to my previous self was probably be more public about the things that you do. OK, I think that's something that I learned later in life. It's like sometimes just by sharing what you think is interesting, you can attract people to you that are of similar things. And that's how Zwickz and I met, you know, just like on Twitter. Actually, I met Zwickz in person for like two years. Oh, really? After I met him. And it's he's also a big promoter of this whole like, you know, learning public. Yeah, yeah, for sure. And yeah, yeah, that would be really it. And building a community of people that are like minded. Sometimes you just got to put the back signal out there. Like in Rome, it's like nobody cares about computers. You know, it's like people are like play soccer, go to the beach. It's like go at gelato. Yeah, there's less there's less of a technical community. OK, so do you remember what you and Zwickz were talking about? So he ran this thing called the DevTools Angels. OK. And I forget. I think he was I think he was an angel and like a company that I was looking at or something like that. And then we connected on Twitter and then we have this latest. Well, now it's called Latent Space Discord, but it used to be a DevInvest Discord. OK, it was basically anybody who's interested in talking about investing. There's also a developer. OK, it was like a kind of like a private space. It was like invite only at the time. OK, do it. So we started spending a bunch of time on there together. Funny enough, we have like a lot of Coppola emojis because everybody loves Coppola in the Discord. So whenever whenever the Discord, the Coppola price moves, everybody in the Discord is like, you know. Oh, you got a lot of investors. We have like the stonks with like, you know, the Net logo. And yeah, that's how that's how we met. And obviously now we're good. Yeah. Yeah. Cool. And then how did you decide to get a good call? You know, sugar, sugar addiction. Yeah. How did you how did you decide to start that podcast? It's funny. It's kind of a meme. Sean was like on two podcasts in a week or something. And I posted him in the Discord and I was like, oh, you're such a podcaster. And he was like, ah, it's like, oh, maybe maybe we should do a podcast at some point. I told them and he was like, oh, maybe we should do one recapping stuff because I think it was like two months before we were at OpenAI. And someone was there. It was like a hack day. And he gave this talk about the models are getting better and whatnot. And there were maybe like 40 people. So it was like quite a small. Yeah. And we were like, nobody else knows about this. You know, like in our even in our developer circles. Yeah. Like most people were not tuned to like LLMs and all of these things. So that's why our first guest was Logan from OpenAI. It's like come tell people what this actually does. Yeah, it works. And then, yeah, just kind of started from there. We've done it now every every week for a year and three months, something like that. So it's hard to do every week. We were like, man, every week. Like, is there enough stuff to talk about? Yeah. And now it's like, yeah, there's enough stuff for sure. Absolutely. It's been fun. Yeah. Cool. Awesome. So our our producer, Peter, did some digging. OK. And we'd like to talk about your ex. OK. But by that, I mean, Twitter. Yeah. Yeah. Yeah. Yeah. Did you did you have a little heartbeat? No. Well, I mean, I've been with my wife for my almost 10 years. So I was like, man, you're going way back. I don't know where you'll find that. But let me skip to it real quick. All right. So here we go. This is what is this? What is Fana Hova? What is it? Tell me about that. Fana is like, you know, shorter of my of my last name. OK, so it's like a nickname. Yeah. Yeah. My friends and my family always. OK. But it's how we stay in. Right. It's for the latter. So you're never going to get it. Yeah. And I was a huge Jay-Z fan growing up. OK. And actually, I started at a website called Rap Genius dot com explaining lyrics. So I was like, OK, what nickname should I use? Yeah. Fana Hova. And also it's like Fana Hova. It's like Fana Hova, you know, that's kind of like a double entendre. OK. And I was like, well, I might as well just use the same everywhere. But so few people recognize it. But when they recognize it, they're like, let's go. And behind me on my Zoom background, I have like a bunch of like Jay-Z vinyls. Oh, nice. OK. People can get there. People. Yeah. If they get on. Yeah. Awesome. And now and now the context, Wendell, we'll know about you and Jay-Z's relationship. Yeah, it's going to be great. So here's a great it's a little great post about you from Forbes 30 under 30, a post that we found here. Congrats on that. That's huge. Yeah. Super huge. And you talked a little bit about this already. We talked about you and your Italian eyes there. Does that stir up any any. How'd that feel? How'd that feel for you? No, it was great. Well, actually, I didn't know I was going to be on the list. Yeah. And I woke up and my friends in New York had texted me about it. Yeah. I was like, oh, congrats. I was like, but what? Yeah. I was like, and then I went on Twitter and I saw it. No, we're just proud of it. Mostly for my family. Yeah. It's just to send the link to my mom. It's like, hey, look, this is what I'm because like there's no venture capital in Italy. Yeah. So they're like, yeah, you're what is your capital? Sounds great. I'm very proud of you. Whatever that means. Yeah, exactly. Exactly. And I think like Forbes is obviously like, you know, worldwide. Yeah. Awesome. So my dad obviously printed it and my dad prints everything. So he's just like a serial printer. But this one was like a highlight for him. It was great. Awesome. Well, congrats. Congrats on that. We have this shirt from you here. And I'm actually interested. I know this is Brev. Dude, so great. Dude, that is so great. He runs brev.dev. And printed the shirt. This SF is dead. I feel like there is a period where SF is dead. And I would like to go back in time a little bit before. And then we'll talk a little bit after. But like, do you do you think that SF died? And do you think it's back or it's on its way back? Where what is this? What is this? What is this instill? I moved to SF middle of 21. OK. SF was pretty dead. Yeah, everything was pretty dead. 21. Yeah, RIP. RIP 21. And I. Yeah, at the time, it was really that actually got a. We had an office in South Park and my previous firm. And it was nobody there. I was like, I will go there. It's like Excel was empty and like CodeSuite was all the other firms. Like nobody was showing up. I was like, is anybody still here? It's like, what's going on? And obviously, part of it was like, you know, SF's kind of like COVID restrictions were in for a long time after other cities and whatnot. And I think a lot of the talent was also like, you know, let me just go somewhere else. But I think luckily, I kind of brought a lot of people back just because it's really hard to not be a part of it. You know, like if you really if you're like a founder, you're like at the cutting edge of it. It's it's hard to like be in Miami. Yeah. You know, I was in Miami last month for a wedding. It's like I love Miami, not for work. Right. You know, so like when you're like outside of it, it's just like a different vibe. Yeah. And I think like the the shirt, I love it so much because it's like a, you know, it makes fun of itself. But it's also like, you know, a skeleton like surfing. Yeah. It's like it's not it's not about tech. You know, it's just about, you know, like the vibes. Yeah. We're dead. Yeah. The vibes are back. Yeah. Yeah. Yeah. Like easily, easily six meetups. And I don't even know how. Yeah, I know. I think last year we're late in space. We've done thirty five events. Geez. Yeah. So you're part of the you're part of it. That's my fault. I'm sorry. So you were in my Spotify to rap. Nice. How does it feel, man? Like that's that's I mean, like not not because it's me, but like just that. Like how many hours did you say? How many do you know? How many minutes? So I've listened to thirteen hundred minutes of you. Let me just do a quick math. I would say probably like. One point five million minutes, two million minutes. Wow. Late in space. How's that feel? A little crazy one. I'm like, why do you even care? Some episode, some episode like we did this episode with Quentin from Luther. Yeah. Yeah. About the mathematics of triangular lamps is basically like how do you calculate, you know, the memory that you need and whatnot. And we released it. I was like, OK, maybe like two or three dozen researchers will care. And then it was on the front page of Acronews for like hours. And it was like a very widely listened episode. So, yeah, to me, the learning is always like even if you don't think people care, just do it. Yeah. Because somebody might find it interesting. Yeah. And like you will really make a difference for them. Yeah. And whenever I ask people, you know, what's your favorite episode? Like folks have a lot of different answers. You know, there's not like, oh, these are the only episodes that people care about. Right. These ones are like flops. Yeah. It's like very varied. Yeah. So I love how approachable you make it. I don't know how you do it. I don't know how you both do it. And I don't if there's a secret to it, like you should let everybody know. But you are like it's like, oh, I've never thought about Vision AI things before. And then you're like, oh, here's and then you are asking these like really rich questions. And you're like, oh, I understand all of Vision AI. Like, I don't I don't even know how you do that. I don't know. But it's really, really impressive. So if people have not listened, Latent Space podcast is get some more million minutes on there. Yeah. No, no. We're ready. Yeah. We had a big spike when Elon Musk followed us on a lot of bots, too. But yeah, yeah, he comes with an army clone army. Do you want us to take away the Slurpee from you? Because we have another game coming up and we need. We need a prop. I'll take it. OK. All right. Here you go. Cheers. Thanks, producer Peter. Cherry. Yeah, I knew it. I knew it. I knew your flavor. Yeah. So many are big gulps. All right. So we're going to put you up. Hold on a second. We've got a we've got a little Cloudflare playground, AI playground here. And we're going to put you up against Llama 3. Let's do a Llama 3. Llama 3. Let's do it. And I'm going to give you both the same prompt so that that prompt is you are incredibly thoughtful, which I think is very true of you. So I don't need it. I don't need to do a Llama. Yeah. And you must answer the question from the user in five words or less where I am the user. And I want you to make sure it's your best representation of your ideas as possible, which you probably always would do anyway. But, you know, you know, try harder. Yeah, exactly. So a little a little try harder there. And I'm going to let me grab this. This is definitely a beta show. Here we go. And the first question that I want to ask, if I can find it again, is what is the telltale sign of when a startup is worth investing in in five words or less? A telltale sign. And you kind of just your finger writing. And what we're going to do is we're going to just take a look at both of them, but we'll see what you say first. All right. So if you can hold that up and show the camera what you said there, right here. Yeah. Terrible handwriting. So founder has purity of motivation. Wow. That's really. Wow. Llama tree. Yeah. Let's see what they did. Let's see what they say. Strong team and validation. What do you what do you think? It's not going to make it. I'm not going to make it Llama tree. I'm sorry, Zuck. Can we talk? Can we talk? Metaventures is not going to work for Llama tree. Maybe for 100B. Maybe that's what they're doing for 100B. Yeah. I would say startups are like so painful. They take so long and they're most likely not going to work. And what I look for in founders is like, yeah, purity of motivation. I call it and I got that from a old colleague of mine. It's basically like a founder that just has to solve that mission no matter if it's like the hot thing, no matter if it's like working. No matter if people agree with him. It's something that you see usually just in the best founders. You know, most most, you know, venture is like this idea of a power law. So a normal distribution is like most outcomes are in the middle. They're kind of average in venture. Most of the money is made in just a small number of companies that are the outliers. And then everything else is worth nothing. OK. So you're looking for people that are like just exceptionally passionate about what they do. And they're just going to stick with it even when it's like not working, because most likely it's not going to work. And when it doesn't work, it just doesn't work. There's no like, oh, you've done a good enough job. Right. You know, it's like, no, no, you're going to go out of business. Yeah. You know, yeah. It's something that is not easy to suss out because most founders are somewhat good salesmen, you know, like of their idea and whatnot. If they're able to raise money. Yeah. So for me, it's more about digging into like, why do you actually care? Like what what in your life has kind of led you to like be here in this moment? You know, like I meet a lot of founders that are like, oh, yeah, my crypto company didn't work. Here's my company. I'm like, why? Yeah. Well, not. Yeah. It's like you make money. It's like, well, it's not that easy. You know, so usually those those companies usually don't work. Right. But a lot of times, I meet with founders that are like, hey, I've actually been in AI for a long time, but never quite felt that technology was ready for primetime. And now I think we cross the threshold where it's ready for like large production work, you know, workloads and all of that. So. Awesome. Yeah, that's that's important. And you see that if it when it gets bad, they're going to stick with it because it's it's. Yeah. Yeah. Yeah. I've seen I've seen a lot of I was investor in a company that did a previous firm software for restaurants. OK. And then COVID hit. It's like all the restaurants are closed. It's like, oh, shit. Yeah. I don't know if I can curse, but go for it. Yeah. And then it's like, OK, how do we make this like a delivery software now? Right. You know, it's like, how do we kind of like. Yeah. But it's like if you don't care, you're like, oh, well, I got unlucky. Yeah. Go ahead. Yeah. It's like, yeah. Throw in the towel. Right. To do something else. Right. Yeah. Awesome. Awesome. That's a great. Let's let's get to it. All right. I'm a one zero. Let's go. Get out of here. This is augmenting, right? Like you can use it if you want to. So this is this is an interesting one. Who makes the best podcast guests? Five words or less. Five words or less. Be thoughtful, too, please. Thanks. I'm doing four four words. I just need four words. Oh, yeah. That's all I need. What'd you get? I got honest, witty and storyteller. Yeah. You could even hyphenate that. You could even. Yeah. Yeah, exactly. Yeah. All right. Let's see what it says. Tim Ferriss and Joe Rogan. I don't know. I don't know. Joe Rogan on Latentspace. Somebody told me I'm the Joe Rogan of AI. Oh, really? I don't know if that was a compliment, but I don't know. I don't know. I just ignored it. I didn't acknowledge it. I was like, OK, no smoking pot on Latentspace, but you never know. I'm going to give it one more guess. Christopher Hitch. I don't even know who Christopher Hitch is. Not me either. I don't know what we're doing. I don't know where this is coming from. All right. So you've got three. You have three new guests for your show. Exactly. Tim Ferriss. I definitely think he won that one. So it's 2-0. Exactly. It's 2-0, man. And I know Peter Axe was like, producer Peter's like, you should talk to him about basketball. And I'm sorry, I don't know anything about basketball. So I can't. All right, Peter, you're in. Get Craig out of the chair. Yeah, he said something about the Knicks and I'm like, dude, no, I'm not going to get anything out of it. So, OK, here. This they bet it better know, right? What are the next big three big things in AI? In five words or less. Yeah, you're going to win this one. You're going to you're going to go undefeated, I think. All right. This should all be affinated, but. Well, yeah, you're doing that. OK, this is enough. This is enough. All right. It's really true. So simulation and services as software. OK, nice. Nice. Let's start with simulation. So we actually just did an episode about a simulated AI. It's really so open AI started with reinforcement learning. So reinforcement learning is really you put an agent in a simulated environment basically. And depending on the action that it takes, you said that was good, that was bad. And kind of you use that as a loop. But then what they found is like it's really hard to scale to generic knowledge and send. So they moved on to LLMs. But now you have these LLMs that are great for text completion like we're doing right now. But when you ask them to do a more agentic workflow. It's hard to trust and it's hard to teach it how to do it in a supervised way. You know, you need a lot of people to give feedback on the agent decision. So I think actually the next thing is using this model. So like self -play some of these roles, especially in like more enterprise settings. So, you know, you have a new product you want to launch. How do you sell it? What are like the objections the customer would have? So you would have like, you know, a sales agent talk to like a customer agent trained on like your customer conversations and then have them kind of like play out a lot of scenarios or things like that. And then come back of it with suggestions for the user to actually say, hey, OK, this is how you should think about the downfalls of this. And I gave a talk at NVIDIA GTC about the rise of the full stack employee. I called it. So using the simulation for people in different departments to simulate what others will say. So if you're a salesperson and you want to pitch a contract to a customer, it's like you have like this legal agent that you can like talk to and say, OK, what's like my legal team going to say if I propose like this type of structure and things like that. Same with marketing. It's like, you know, how would they message like the new product? You know, so it kind of removes the limit from folks to ask suggestions from other teams. You know, they go to them once they're like a level above. It's kind of like an initial code review. Yeah. It's like you've done the first step. Yeah. That's clever. Yeah. And then services as software is like the outside in version. So the simulation is like inside in. So you're like already in the company. Services as software is things that you used to have outsourced because they're not really strategic to you. Yeah. Using AI agents to have on demand 24 7. You know, so we're seeing this in security. You know, there's like a big cyber security talent shortage, as everybody knows. A lot of companies just don't have the right folks on team. You're going to have one company that does security agents as a service. So you can have them in your organization kind of help you triage a lot of the problems and slim them down for your actual security practitioners. And we'll see the same with software engineers. You know, you're going to have automated software engineer do some of the initial work. So those are like the two sides of it. That's enough. I know that as for three. But yeah, I only got two right now. I can come up with them. I mean, you get two examples of the two examples inside there. Explainable AI, transfer learning and meta learning. Is that are those even words? I haven't had a meta learning. It's like obviously talking their own book. Yeah. But yeah, I would say, you know, transfer learning is why we're here. Right. Like, you know, it was something that already been there. Explainable AI is like the usual thing. It's like, you know, it's like explainable humans. It's like, where are they? You know, why don't you do what you just do? I don't know. Yeah, just kind of all my previous experience led me here. But I don't know specifically why. OK, OK. So you went you went 3-0. Yeah. Yeah. You 3 -0. Yeah. They need they need the 400B. For sure. For sure. Yeah. That was a sports term. Exactly. This is why the NBA like best of five in the first round. Sometimes it's just over. You don't need you don't need seven games. So thank you so much for being on the show and playing these silly games with us. And we really I love the work that you do. So if you haven't already, everybody at home, check out the Late In Space podcast. Follow Alessio and follow. Join the discord, right? There's a discord. There's an awesome discord about AI. It's such a great guest. Thank you. Thank you for being here. Thank you for making so much content, man. And I learned so much. So thank you. Bring out the slurpees to celebrate. Thanks.