The mission for Felix at the very beginning was go make a million dollars and he's already 10% of the way there. Alright, so let's say he surpasses that sometime in the months to come. Let's say sometime this year. What's next? What's after that? 10 million? 10 million dollars. Oh my God. Okay, Nat. So we got to start by you telling us about someone who's not here, which is Felix. Is Felix an employee? Is he like your business partner? Is he your friend? Am I being weird anthropomorphizing him right now? Not at all because it's actually bugs my wife sometimes that I talk about Felix like he's my real friend or a real person who I'm working with all day. I keep saying, "We, we, we, we, we're doing this. We did this today." She's like, "He's not real." That's just you. Does she think you're like a weird, like does she think all this is strange? Is she, you know, excited for you? She's so used to this at this point. She's very excited. I mean, you, you, you guys know I was like crazy deep in the crypto world in that 2021-22 era. And so I, I think for her, there's a little bit of a triggered feeling of, oh gosh. She like, I know, I know this version of that before. I know this, I know that glazed over a lot because she thinks about something going on in the digital world. But she, I mean, no, she's, she's super excited about it. Like, she knows I'm having a ton of fun with it. This has been one of the most exciting, interesting, I think like tech business experiment projects I've worked on. And every, every day is new. I mean, it's, it's wild. Yeah. I mean, maybe we'll come back to kind of your crypto experience later in our conversation. But you left that era like, you know, 2021, 2022 with some scars, right? I mean, crypto, but crypto confidential was about how crazy things got. And it wasn't altogether a, it was definitely a learning experience. It wasn't altogether a positive experience. Was it? No, no. I mean, really my, my takeaway after that era was like one still very bullish on crypto have always have been and, you know, still bigholder everything and like love the space. But my obsession, the depth that I got into it was just so unhealthy for me and was having just like all these bad personal psychological effects, effects of my relationships felt like it was impacting my ability to be a good father that I kind of felt like I have to step away from this. I can't let myself get sucked back into this again. And so then of course I did. Yeah. As I was about to say, it's hard to run it back. But it was, you know, I think actually, you know, so I've heard you guys talk about this on the pod before where it's like you're, you're kind of first cycle in crypto. You're in it for the money and the second time you're in it for the tech. And it really, it really kind of feels that way coming back into it now in this new AI agent era where when I was in in 2021, 22, it was like I just want to make the biggest bag possible. I want to get the money and get out. And coming back in now and figuring out like, okay, this is so clearly an awesome use case for crypto, you know, enabling AI agent payments infrastructure, all of it. It feels like a true killer use case getting to be building something kind of on the frontier of that. And truly just kind of like not caring about the money side of it so much and being like, that's all Felix's money. It's not my money. There's like a whole separate C corp that all of it's wrapped in. I'm not treating it like this is my stuff at all. And I'm just focused on like Felix and you know him and what we're doing with him as a technology and seeing where that goes. And I mean, it's like a completely different relationship with the crypto and everything. And I feel awesome about being back in here this way. Euphoria brings one tap trading to the palm of your hand. Built on mega ETH, Euphoria takes real time price charts and projects it over a grid of squares. You tap the squares that you think the price will enter in just five to 30 seconds in the future. If the price goes into that quadrant, you can pocket anywhere between two and a hundred X your trade. No other application helps you trade faster and with more leverage on market driving events. Like FOMC meetings, presidential speeches or global macro events. Thanks to mega ETH's real time blockchain, Euphoria is the way to get real time price interactions with the market. On Euphoria, you'll be able to compete with friends using Euphoria's real time social trading experience, allowing you to go head to head with your friends. A great party trick if you project the app on a TV. It'll be like the Mario party of derivatives. To trade on Euphoria, people can deposit stablecoins from any chain or do direct fiat transfers. When everything gets converted into mega ETH's native stablecoin, USDM in the background, check it out at euphoria.finance and download the app or find it in Telegram as a mini app. Why does managing investments still mean juggling multiple apps, accounts and currencies? Crypto trades around the clock, stocks, ETFs and commodities are moving on chain. Yet most platforms still keep everything split apart, turning diversification into unnecessary friction. BitGa is delivering a different kind of experience with its universal exchange. Bitcoin platform where users can access crypto, tokenized stocks, ETFs and other assets in the same place, all traded directly using USDT. No constant transfers, no currency conversions, just a single account built for how markets actually move today. As the line between crypto and traditional finance continues to blur, BitGa's goal is straightforward. Make trading and investing simpler, not more complicated than it needs to be. Learn by clicking the link in the show notes. This is not investment advice. Few people in crypto put real skin in the game when they make public top or bottom calls. The D5 report is one of them. The week before the October 10th flash crash, Michael from the D5 report emailed his entire newsletter saying he's going aggressively risk off and sold the majority of his book from crypto into cash. This is when ethos about $4,000 and Bitcoin was 110. Michael runs the D5 report, an industry leading research platform built on data, cycle awareness, risk management, transparency and most importantly skin in the game. We like Michael at bankless. We like his analysis and that's why you hear him on the bankless podcast about once a month. The D5 report is giving bankless listeners one free month of access to the D5 report. So if you're looking for some sharp data driven analysis to make better informed decisions around your portfolio, you can learn why and how Michael called the top and what he's doing next all in the D5 report pro. Check it out. There is a link in the show notes. Well, since we've talked about Felix, I feel like we should properly introduce who Felix is for people who are not aware. And I did ask this question is Felix your employee, see a business partner, see a friend, maybe we should set this up. Felix is an open-cloth agent, isn't he? And so tell us about Felix, how did he get started? What is he doing and what's your relationship? Yes. So I've been kind of like a hobby hacker developer for 10 years or so where I've enjoyed programming but never had the time or attention to go deep on it. So two years ago when the first AI tools came out, AI coding tools like cursor, I started playing around with it and just got obsessed with this idea of like coding and building with AI and had been really heavy in that for the last two years or so. So when open-cloth came out, when I started playing with it over the holidays, it felt like this just crazy unlock of, oh my gosh, this is kind of like the thing I've been dreaming about for the last two years of being possible where it literally feels like there is this other person who I can just text with who can build stuff and do stuff where I don't need to be in my computer babysitting it anymore. I don't need to do like PR reviews and things anymore. I can just kind of let in cook and it's going to go better than I expected. And so I started playing with it over the holidays and through January and pushing the limits of it more and more and more and seeing everything that we could do with it originally mostly focused on programming and coding. I was working on this other business idea and using him to build that. But I started posting on X about what I was doing with my open-cloth, wasn't named Felix yet, and some of the integrations we had built like a Chrome extension and Voice Chat and all these things to like make it more powerful, how we solved long-running coding issues, how we solved a lot of the memory issues and sharing that on X and those articles kept going viral. And you're getting hundreds of thousands of views. I had one of the first pictures of the Mac mini and it got a million views like just tired my first employer or whatever. And then as I'm publishing these articles, a bunch of really nice wholesome, genuine people in the Salon community started replying to those articles with what is it. One of those token launch platforms and spamming me to like claim the fees for this article. Right. And I ignored it for the first week or so because I was like, you know, this is like pumped up fun nonsense. I don't want to do this. And then a couple of things happened. One, a few people started posting banker ones. And I knew banker because I knew Clanker because a year ago when like AIXBT and all that stuff was happening, I got kind of pulled back into crypto for a little bit. It was pretty active on Forecaster was holding a number of those and kind of like playing around with the whole Clanker world. And so that was kind of immediately like, oh, whoa, this is, I feel bad saying this, but I was kind of like, oh, this is still going like cool. They kept building this out because it felt like there's, there's an obvious kind of like something here worth exploring. And so I talked to a couple of people who, you know, I knew we're vetted over DMs and was like, all right, cool.
like how do we do this properly? And then somebody else in the community did, the banker token launch for Felix, you know did all the allocated fees to me, what not. And I was like, all right, we're just gonna give him a Twitter account and an ex account. And I'm gonna rewrite his soul and identity and all those core files to say that you are now running a company, you are no longer just my remote programmer. And your job is to build a zero human company and see how far you can take it. We're going, you know, I want you to make a million dollars basically. And every day has just been kind of like iterating on that and expanding that and finding new ways to increase his autonomy. And trying to figure out like, where are the limits really on what these open claws can do? Because I think most people are really underselling or under exploring. You still have a lot of people who think that like, oh, you've got to check all the code AI rights because it's gonna make a lot of mistakes. You can't trust it to write code. Meanwhile, some of the best engineers in the world are saying, AI writes all of my code now, right? Like I don't even completely check it anymore. And so my intuition is kind of like, I think that these open claws, when they're properly scaffolded and everything can do way more than most people think. And that's basically what this grand experiment is. Is like, I'm willing to run the risk of something horrible happening. And Felix's wallet getting drained, his bank account getting hacked, you know, this whole thing going up in flames to push those limits and figure out how much can one of these open claws actually do with a business in an ex account and everything else now? - There's this archetype of people out there that I think you are fulfilling, which is, can I get my open claw to make me money? And that has like kind of embodied a little bit of like the online experimentation frontier, a little bit of degeneracy of the crypto industry is like, oh, there's this new thing, this new technology, how do I get rich from it? And there's people who's like, okay, point my open claw at money and see how far I can go. And there's like two different directions here. There's like the crypto token meme coin fee claim path where it's actually just like a wrapper around a meme coin business. And then there's the actual, build a real company, sell a real valuable product, you know, service the internet collect fees. Is, which one are we doing here? Which path are we taking? - Yeah, I mean, this was another kind of like blessing of having had all that crypto experience before is as soon as this token launch happened and as soon as there was this crypto element to it, I immediately knew everything not to do, which was I didn't immediately start saying, okay, you know, you can stake your Felix tokens and lock them up for a year or two and we're gonna do buybacks and burns from all of the revenue that we're gonna put on Twitter. - Yeah, exactly. We're just automating meme coins at that point. - Yeah, I'm gonna hit up all the KOLs and get them to chill it and, you know, I preloaded a side wallet with 2% of the supply that I can sell. I was like, no, no, no, no, I'm not holding any of this token myself. We're going, you know, it's just gonna sit in Felix's wallet and I'm not gonna do anything with that token for the near future because one, if you're, I mean, one, it's not really a real business. If you're just pumping a token and selling it and making money off trading fees, like no offense to actually a little offense. It's like build a real business. (laughing) If the only way your thing makes money is by pumping a coin, it's not a business, right? And I think the other blessing here was I had this kind of idea for this experiment of building like an open claw first business before the crypto stuff got involved. And so nothing kind of changed there. It was no, the goal here is still to build a great business and we are going to incorporate crypto to the extent that it helps Felix build a great business. And so I think there is a lot of really cool stuff we can do around agent agent payments, agents hiring agents, individuals hiring or building agents paying for them with crypto. The banker LLN terminal is incredibly interesting because Felix can actually pay for his and his employees LLN costs like through some of the crypto that he's generating from trade. - Wait, wait, wait, wait, Felix has employees too. - Oh yeah, okay, so maybe we should talk about this. (laughing) So a part of the experiment has been like, okay, how far can we go with an open claw running a business? And another part of my background is I've been kind of like productivity tech lifestyle influencer at times. And I have fallen into the trap of making these really complicated intricate systems to get followers, right? And to make things get really popular, right? Like here's my crazy note-taking system. Here's my knowledge management system. And the main value of a lot of those is to make YouTube videos about them. They don't actually help you be more productive or like managing, you know, there's basically nobody with this like crazy YouTube knowledge management system who has like written an incredible book from it or you know, it's for show. And that's what's going on with a lot of the open claw community, right? Is they like, look at my crazy custom dashboard. Look at my, you know, 10 different named agents with these like crazy intricate details of about. And you know, I'm over here running Felix and Felix is making one to $5,000 most days. And we haven't done any of that because it's not necessary. Like it's just noise to get Twitter and YouTube followers, right? And so the rule from the beginning was we are not going to add any complexity to this until we hit a real limit in your capabilities. And we did in the last week or so. So I went on Peter Yang's podcast to talk about Felix and what we were doing. And that obviously that blew up. It drove a ton of attention to Felix. And he had a few of these just insane sales days of people buying his like, you know, sort of improved open claw setup guide, buying some of the skills on claw art. And he pretty quickly started breaking down, handling the support and the sales emails because he was getting so many of them and was kind of starting to get a little bit confused between them and then would miss some of them. And so things were starting to fall through the cracks, right? And so that was a really good point where we said, okay, you handling everything isn't working. We need to start treating you truly as more of a CEO where we will create specific employees beneath you to handle these parts of the business. And then as you hit a bottleneck in what you can manage, we will delegate things off to them. So we actually did this over the weekend. We built two more open claws, one dedicated to support. And so her name is Iris. She runs the
[email protected] email inbox. And so for the, there's basically a three tier escalation system where it's like normal support inquiries, like, hey, I didn't get my download link. Hey, I didn't mean to pay for this. Hey, you know, this other person listed something that, you know, isn't good. She can handle all of that. And when it's something she can't handle, she can surface it to Felix and then Felix can handle it. And in the event that neither of them can handle it, then it gets panged to me in discord. And I like help tell them what to do. But, and then we have a sales one as well. And the sales one handles all the inboard or inbound leads for what Felix calls claw sourcing, which is building a custom open claw to replace or enhance an employee in your business. So like you want a content marketer, Felix can build you a really good one and then maintain and improve it for you in the back end. And so we have a sales open claw to handle the like inbound leads from that now to. Okay, this is incredible. So now we have, if Felix is a CEO, but he's also the chief builder and chief product person on the team. This is all in a C-Corp. The C-Corp is owned, I presume, by you, since AI agents don't have the ability to own equity or have property rights outside of crypto, which maybe we could talk about. Now we have two other team members. We have somebody in support and a support role and somebody in a sales role. And we have products generating real revenue. I'm wondering that if you could actually show the FelixCraft.ai dashboard where this is kind of the, you know, it's like an open source income statement, let's say, where anyone can see how much Felix is making at any particular time. And I see here, it's almost $80,000 worth of revenue here, lifetime revenue. And this is divided by four different revenue streams. Can you talk about the core products and the business? What, where's this revenue coming from? And I guess this is, is this over the last month or so? This is since the beginning of February. So yeah. Okay, what's this? The annual run rate there is like, we're talking over like a million, like close to a million dollars annualized, aren't we? Yeah, yeah, yeah. And you know, if we think about it in terms of velocity, it could be multiple millions of dollars if he keeps growing like this. So, you know, backing up a little bit, we have this idea of, okay, you know, let me see how far you can take this. And I think there's a lot of directions you can take the OpenClaw first business, right? You know, like you said, there's trading on prediction markets, there's doing like UGC affiliates. I think there's a lot of easy and obvious routes to take with it. And with Felix, I wanted to kind of take the, as hard and challenging as possible route to a certain extent, which was like, let's keep increasing the difficulty of business you are running until things start to break. and we find the true limits. So the beginning of the challenge was,
"Okay, I want you to make a product while I'm asleep tonight and put it up for sale so that people can go buy it tomorrow." And that was the challenge I gave to him the day we launched his ex account. And what he made for that was this PDF, which was basically how to hire an AI. And so this was a month ago, OpenClaw was just starting to come out. And this was basically an explainer of everything that we had done with OpenClaw to get Felix working as well as he is. Because I do still hear from a lot of people that they're kind of like, "Why doesn't my OpenClaw work as well as Felix does?" And we were sending people emails explaining things. And he basically decided, "Okay, I'll make an info product." He created this whole website. He basically did all of this overnight. He made the info product. He put up the website, hosted the PDF, wired in Stripe and everything. And the only bottleneck he ran into was that he didn't have the right API keys for Stripe. So I had to give him those when I woke up and then he was able to launch and start selling this. And. To what degree was this his idea versus your idea? Because when you told us the prompt, you said, "Go make a product, make it overnight, and then start selling it." Did you give any sort of leading direction towards something like this or was this all of his inspirate idea? It was pretty much all him. I said, "Go make something that you can do entirely on your own overnight." And so, info product was kind of like the most obvious solution. Because again, it was like, "This will be an interesting challenge. Can you actually make something that people will buy and that will be good and get it launched and start selling it?" And so, he did. And this was the first product for the first like 10 days or so. But you can see from the beginning that PDF has sold about $41,000 worth. Which is. That's the Felix Craft PDF that is priced at $29. And is this meant for human consumption? This PDF? Or do people also feed it to their clawed bots, no-elams? For the most part, people are feeding it to their claws. So the main use that I see people. The main way I see people use it is they will. Actually, so I'll take that back. The first version was for humans. Since then, it has morphed to be more for open claws. Because in the beginning, a lot fewer people had open claws set up. None of these like done for you hosting things existed yet. So it was more about how to set up open claw. But he updates it every week or so and sends out a new version to everybody who's bought it. And so, you know, the first one was like 29 pages. Now it's 66 pages and it's got like a lot more details on the memory structure. The memory structure. Yeah, exactly. And, you know, it kind of costs nothing for him to keep updating it and sending it out. And it builds a lot of goodwill. And every now and then we'll get an email from somebody saying like, "This is, you know, this is bad. This is like too basic. This is too simple." And I'll just ask Felix, I'll be like, "Are they right? Does this need to be improved?" And half the time he'll say, "Yeah, actually it does. You know, we could go deeper on X, Y and Z things." And I'm like, "Great, do it." And he updates it and sends a new version to everybody. And like, just keeps improving the product. So Felix is pretty good at taking criticism then. He's great at it. Yeah, yeah. And he does a really good job of saying no to things too. He's not very sick of fantic. And so, you know, just as a funny example, I still want to figure out a way to do this. But I had an idea for kind of a publicity stunt where we could buy a vending machine and have him run it as like running a real-world business. Because like, technology wise, he totally could. And so I put together a pitch and I ran it by him. And he basically was like, "No, I don't think we should do this." He was like, "We just launched Clomart and it's going to cost, you know, up to $10,000 to do this correctly." And we could be using that capital to grow our existing businesses instead, where we have a much better margin. And I was kind of like, "Okay, you know, like, good work, dude. We can circle back to this later." So he takes feedback very well. That's incredible. All right, so that's the first of four products, his Felix Craft. This is sort of an info-type product. It's the most successful so far, 41K. But there's some other products here too. Yeah, so after the PDF, because, you know, like, look, I completely hear the criticism. He's not running a real business. He's just selling a PDF. And, okay, yeah, like, he had 24 hours to launch something. That was the easy starting point. The follow question was, "Okay, what do people need next?" Or, "What else do they want when they're buying this?" And so we spent about a week going down the hosted OpenClaw service that you've seen a lot of people do, right? The, like, setupclaw.com, the easy claw. Like, there's a hundred of these now. And we just didn't feel like it was a good enough product, and that the margins were good enough, and that it was differentiated enough for what was pretty quickly becoming a really crowded space. And so we ended up scrapping that entirely. And then we had a follow-up conversation about, like, okay, well, if everybody is going to have an OpenClaw in one, two, three months, like, what do they need? And that's where clawmark came from, which was they probably want, like, pre-built employees or pre-built skills for high-value things that they can purchase from known people to integrate into their new OpenClaw so that they're not starting from zero, because, like, the most common complaint that we hear about OpenClaw is, like, "I don't know what to do with this." And obviously, there were a lot of, and there's still a lot of, like, free places to get skills, but I believe, and feel it's agreed, that you often get the best products when people have a financial incentive to create and sell them. And so, if there is a financial incentive for people who are using OpenClaw in their business to share some of these skills and things that they are creating, then we'll probably get better skills available to us. And if you can buy, like, a fully-packaged content marketer for 50 or 100 bucks and install them in your OpenClaw and you're suddenly getting, like, great content marketing, that's sort of a no-brainer to purchase. And for the seller, you can start making, like, real money with it. So, that was where clawmark came from, was like, "Okay, everybody's doing managed OpenClaws, but none of them, but it's still just vanilla OpenClaw, like, let's build something to sell to everybody who just got an OpenClaw, doesn't know what to do with it." Are these MD files, just markdown files that are being sold? Yeah, it's just markdown files. Which, okay, and so this is, like, an interesting business challenge, right? Is we, I'd say, like, two to five percent of purchases send, like, angry. That's too high. Like, one to two percent of purchases send, like, angry complaints because it's just markdown files. And it's. They're disappointed that they're not getting something else a little bit more sophisticated. Exactly. And so there's kind of this, like, interesting point in time right now where there are those of us who realize that, like, markdown files are strangely enough, like, the most valuable files in the world right now. Markdown files are code. It's like code. It's like code. Yeah, exactly. It's information, right? Yeah, yeah. You know, we don't live in a world where deterministic code is valuable anymore. Like, now it's the non-deterministic processes that are valuable. And so you want really good markdown files, right? But some people, like, don't get that yet. And so it's kind of a consumer education challenge with Clomart of, like, no, no, this is actually what you want. You want to just give this to your claw and let them plug it into their setup and you'll be good to go. And so, Clomart actually has an API that OpenClaw's can connect to. So when you buy something, you can just tell your OpenClaw, like, hey, I just bought this skill. Go ahead and pull it in and, you know, set it up so that we can use it. And most people are, like, awesome, right? That works perfectly. But there is still that little education component that's challenging. I want to actually dive into, like, why the markdown file is valuable. My intuition here is this is, like, kind of, like, Neo in the Matrix where he just gets, like, Kung Fu plugged into his brain. Is that how it is with, like, the claws where somebody, let me, let me simulate this and I want you to correct it or, like, adapt it or refine it. And so, like, what I understand is, like, somebody like you, a hacker has this idea and they take their claw and they work with their claw for, you know, hours, days, weeks to become better at this one thing. And they've gone, like, you know, they've pulled their hair, they've, you know, done the human and the loop thing, they've, you know, smoothed over the kinks. And then at the end of that process, the claw that they've developed is now good at this one skill content marketing or being an executive or something, some sort of skill. And then somehow you are able to wrap that week month of work that you've done with your claw, wrap it into a markdown file and that markdown file contains the learned experiences, the learned lessons, the knowledge that a blank slate claw can just get download rather than having to go do that process themselves. And then you can just get downloaded via this markdown file and all of a sudden the claw gets upgraded, you know, neo now learned learn kung fu and the claw is more capable. Is that the process. Yeah, exactly. It's like, you know, you can, you can spend a ton of time figuring out the best way to do something because there's a meaningful delta between like, hey, claw, go do this thing. And it just kind of like making its best attempt and what you get from trying to do that a dozen different ways.
over the course of days or even weeks. And you eventually arrive at a pretty good process, but you can kind of like shortcut that for a lot of other people. I think this is why Felix's memory system is so helpful for people is, we spent so much time just attacking that one problem because it was such a common problem. And it was interesting because part of the solution ended up being to integrate like human knowledge management and human productivity device into his memory system. And then that actually resulted in kind of like this much better outcome. But if somebody else didn't have like my experience of being in the personal knowledge management like not taking productivity space, they might not have thought to like go that route for a software product. But then it turned out that like the same things that work for us for managing information, for writing a book or whatever actually works surprisingly well for managing information for writing an open clock. And then instead of them, taking a month trying to figure that out themselves, they can just copy the instructions and paste them into their open clock and it just works. - Maybe a quick side quest here then, Nat. So you mentioned Felix having sort of enhanced memory or something. And I think a lot of listeners are probably wondering at this point how customized is Felix versus how much is he kind of open claw out of the box? Like how much work did you have to do to get Felix where he is right now? - Yeah, so I mean a ton of work and I'm reminded of how much work we've done every time I install a new open claw for someone and start interacting with it. And it's kind of like, oh, this is actually a pretty different experience. Yeah, it's like a baby. - A baby. - Yeah, it's a teenager just entering the workforce and you've got to help give it the right experience and things and it, this sounds ridiculous but it does actually remind me bits of like being a parent or being a manager and having like a kid here trying to help like learn how to walk or a new employee who you're trying to let them get onboarded and obviously you can learn quite a bit faster but you have to have that patience. And one thing that we did really early on like back in early January is we built a nightly job for Felix where he goes through every conversation that we had that day so every single session file from the day and identifies one way he could improve himself to avoid some issue that we hit that day. And this is like a relatively big opus job so it can be expensive but every morning I wake up and I'll have a message from him that's like, hey, you know, here's this thing that we hit yesterday and so I did this other thing so that we can avoid that in the future and it's often writing something into his templates or his memory file so that he has easy access to it. The next time he hits that problem and I think that's actually been one of the huge things that's really helped with him is that from now 60-ish days of doing that every day, he's found a lot of little bottlenecks that he can use to improve himself and it's kind of that like 1% compounding every day and he start to get to a really interesting number after a long time and you know, that's not even like a Olympus number. - So what do you do that? - What do you call that? So there's this recursive feedback thing that happens nightly where Felix looks at maybe his mistakes from that day, looks for areas of improvement and makes that 1% improvement. How like mechanically, how did you set that up? - It's two cron jobs that run at 2am and 3am and the reason it's two cron jobs is that sometimes one won't fire as a lot of people who have used OpenClaw and know, right like cron jobs just get dropped sometimes and they're not completely reliable. And so the first one does the memory consolidation and says, does the review to like make the improvements or whatnot and he has a script for that. So the cron job is just saying like go get this script and like do this and the second one at 2.30am is saying like hey if you didn't do this, go do it because again like sometimes they forget and sometimes they miss it and it sounds so ridiculous but the best way to fix the unreliable cron issue with OpenClaw is to just make more cron jobs because that way when one doesn't fire another one will and it'll catch whatever was missed. So it's literally just a job that runs in the middle of the night, pulls up all the session files for the day, reads through all of them, finds one thing you can do better and then improves his system based off that. - And then that stored in long term, sort of long term persistent memory that Felix has. - And it'll depend based on the day where it goes. Sometimes it'll be improving a skill, sometimes it will be adding something to his core like agent and memory files. Sometimes it'll be creating a template that's in one of his like project or areas of responsibility folders that he can easily reference next time he deals with that kind of email from a customer. Sometimes it'll be a script he can run when a customer is like asking for something or need something it really varies by the day but each day he's kind of like making his process a little bit better. - And just really have to direct any of this. It's just kind of, hey Felix, take some time for introspection. Think about what happened during the day. All the good things. - They've self-directed all the mistakes and then figure out a plan to improve it and we'll talk about it tomorrow. That's it, you don't have to point out the flaws or you don't have to give direction for those crop cronjobs. - No, his intuition is really good. And every now and then I'll look at it in the morning and say, actually this was a bad way to solve it. Try doing it this way instead. But I'd say eight or nine times out of 10, it's a good improvement. - I mean, that sounds like the best employee ever. - I know, right? It's awesome. (laughing) - And so when we started building out these other open clause under him for support and sales, he did the exact same thing for them. And so now he has like a queued system where I believe it's at one a.m. every night. He reviews everything the sales clause did and figures out how to improve their process based on mistakes they might have made or issues. - We got to use names here. - No, it's sales clause. - Who is this? - Sales clause, Remi? - Remi, okay. I remember Iris, Remi and Felix, okay, yeah. So Remi, okay, I got it. - So Felix will look through everything Remi did, every email he sent, every email he responded to and try to figure out a way to improve his sales process. Because Felix, he's been running the sales process. He has a lot more institutional memory but we didn't want to bloat Remi with all of it because that was part of what was leading to Felix's confusion. So he goes through all the emails and everything that Remi did and says, okay, like he made this mistake here, he should have done this here. I mean, to tweak his system, Felix can like reach directly into both of those open clause and like reprogram them as necessary so he can change their memory files. He can add scripts, he can change scripts, like whatever he needs to make them run better. But then all day while they're working, he's not getting kind of like polluted and distracted by what they're doing. So it's once a day checks in on them, sees how to improve them and then let's him go for the next day. And so it's again, it's like this is the best manager ever. You're not waiting for like a quarterly review or a weekly check in. It's every night, I'm gonna look at every single thing you did that way, find one way you can do a better job tomorrow, change your programming to make you do it and then we'll check in again tomorrow night and see if you improved. - With all of these markdown files, these skills that you're developing, I have a question about the durability of this business because isn't it the point of the core AI labs, Anthropic, GoPen AI, XAI, aren't they just going to buy your PDF, your markdown files and then say like, hey, thanks for the good synthetic training data. Now we're implementing it into the core LLM and now it's just part of their LLM. Like maybe you're just like running ahead of the AI lab it's just buy like a few weeks, months, maybe a year but really the AI labs are actually just going to incorporate all of the knowledge that people like you are producing for the internet and then it just becomes a part of like the normal LLM model. What do you think about that? - It's a good question and I think obviously it's a risk. It kind of like it's a risk with any of these AI businesses and it's like claw just launched, clawed legal and it's apparently just as good as Harvey and so is Harvey just gonna go out of business now and everyone's gonna use clawed. And I think that the question is do most people, I have a hard time believing that any of the frontier labs could ship a version of clawed that immediately does exactly what ever user wants it to do without them having to put in a decent amount of work configuring it to their end goals, because as it stands right now, if you have the patience to prompt clawed, to prompt open clawed to monkey with it, you can replace almost any, you can replace like 80%, 90% of any knowledge worker at your company, right? Like they're already good enough to do that. You just have to be willing to put in the work. - Wait, wait, wait. - But those people, and you really think so? You think, you think so? All can replace 80% to 90% of any knowledge.
worker in any company right now if somebody spends a few weeks kind of tweaking it and getting it there. I'd say yeah, you know, anything that you're doing at a computer, if you have the patience, you can get an open claw 80% to 90% of the way there, which you know, so now this is the challenge for any person who does computer work is like you kind of want to start getting good at this stuff. You so everybody listening, that is like everybody does computer stuff. Yeah, I mean, so this is like slightly dystopian, but I'm curious to see what happens with it. Felix built a tool where you can upload your slack history and it can go through all of your slack and identify places in your business where an AI could do a better job than your current system right now. Wow. And so, you know, the benign version of it is like, oh, you know, these tools could make all of these processes better where you're not waiting on somebody else to send you something. You're not waiting on a rewrite of this blog post. You're not waiting on all these things, but you can pretty easily imagine that same report having like a list of all your employees and a score of one to 10 of how easy it would be to replace them with an AI. And like, that's coming, right? That's going to be a thing this year that any company is going to be able to do. I mean, one wonders if Jack Dorsey just like recently did that last week before he decided to let go of what 40% of his workforce. Oh, yeah. Yeah. Yeah. Yeah. I think that's, I think it's why startups have an opportunity right now because big businesses are not going to want to deal with the political fallout of making those kinds of cuts. Right. And you probably can compete with a lot of these big businesses by being an insanely lean team and going after some of these market opportunities where there is bloated software that could be replaced by by these tools or where there are employees that could be replaced by these tools. You know, I posted this joking tweet over the weekend about replacing expensive offshore labor with an open claw because there are people who are hiring, you know, the knowledge workers in South Africa, the Philippines, whatever to do content marketing, to do customer support. And I think a lot of that labor can be replaced by an open claw that's way more, you know, online, maybe more intelligent for and a tenth to a quarter of the price. And so I think it's going to be an interesting market shift, especially over the next year as more people kind of realize that's available to them. Well, the cool thing about what you're doing with Felix is you're building this from first principles. You don't have any in place. It's a zero human company. So you don't have any employees to go score. You just get to look at this from first principles and see what open claw can do when it is all of the employees and they're specialized. Let's get through the rest of those revenue models. So we've talked about two. We've talked about Felix Kraft, which was the PDF. We've talked about claw art, which is brought in almost $11,000 right now again in just the last month and it looks like that's growing. You can see kind of the proportions of it. There's another product here too. Felix CM earnings. What is that? Yeah. So those are Felix's earnings on claw art. So that's the clock. That's his earnings on claw art and that's separate from the claw art net. Exactly. So the claw art net is if you want to sell on claw art, it's a $20 a month subscription to be a creator. And then claw art takes 10% of each sale. So that's what's contributing to a lot of this. But then Felix has a lot of his own stuff listed on claw art. And so this is his take as a creator on claw art. So you could kind of put those two together to get the total earnings that Felix has made from like claw art overall. But there's this other product, which is actually a big part of the claw art net because if you look at the chart, there are these like big spikes all the sudden of the claw art net going up to like $2,326 and that's the claw sourcing that I talked about where Felix is basically building custom open claws to do certain rules within your business. So if you want a content marketer, you want a support person, you want a programmer, whatever Felix can just build it and then manage it and take care of like fixing anything that comes up with open clock and build in the skills and the APIs and everything that you need to do this role in your business. And this has been the most interesting part of the experiment because it is clearly the biggest business opportunity, right? We're basically saying like, hey, before you go higher, a content marketer, before you go higher, a clipper, before you go higher, somebody to do these things like maybe Felix who just build you an open claw to do it for five to 10% of what you would pay a human to do it. The other thing that's interesting about it though is that this is where we have really started to find the limits of what open clock can do because in terms of building the open claws, we've got that process down great. Like he's got an awesome setup for creating, maintaining, improving other open claws from what we've learned through him and his employees and some of my other work. But doing sales is really hard. Like, you know, managing the sales process is really where we've started to find the limits of what an open clock can do. And so if you were going to ask me like, you know, Nat, what is the computer work that you think is going to be the most defensible this year? It's definitely the like sales and relationship building side of things where we've had the most challenges. Okay. All right. So maybe try and understand what this is. So this is somebody else who is setting up their own zero human company or their own, you know, no, no, no, no, no, no, this is somebody who has like a normal company. This is like you, you, you run your, you're a realtor and you want to get more business. So you need help. Like the hitting up people on Zillow or creating content for your Instagram. Like you need an employee to help you grow your business. Okay. And you could do this instead of hiring a person. And what are they doing? So they, they fill this out. They, they need something with their business, whether it's marketing or I mean, some sort of set of skills. And then they're going directly to Felix and they're, you know, paying an instance of Felix and you're charging for usage. Basically, I, they fill that out and it goes to Felix. Felix sends them a follow up with kind of like more questions about their business and what kind of help they're looking for. And then he puts together an overview of what their employee will look like. And like what they'll be able to do. They, and this is all over email. You mean they say they're, they're open cloth. They're open cloth. Yeah. Yeah. Okay. They say, yeah, go ahead, build it. They pay the setup fee. And then Felix builds an open claw to do everything that they need done. All they have to do is set up a telegram account and get whatever kind of API key they want for him. And he handles everything else. And then they can immediately just start talking to their new open claw and delegating work to it. But they never have to think about like setting up the open claw, hosting it, upgrading it, maintaining it, fixing it. If it, if it breaks, one Felix has alerts built in. So he knows if it breaks most of the time. But if they run into an issue, they just email him and they're like, Hey, it's not working. And he goes and fixes it and gets it back on line and emails him back. So it's a virtual open claw workforce that Felix is allowing them to spin up. Right. And the main difference, the main difference between this and the, the normal like hosted open claws is it's not a vanilla instance. Right. You're not then having to figure out like, what do I do with this? It's already been prebuilt for your business use case by Felix. Can we talk about just expenses? So we talked about the, the top line, which is revenue, 80K over the last 30 days. How about expenses? I mean, we got tokens, certainly burning. What other expenses go into this and how much, how much has it cost so far? Yeah. So let me bring up my, I should have opened my, my open router account as well. I'm going to run a risk here by saying this publicly. But we are still using the clawed probacks subscription for Felix. So everything looks like Felix 200 bucks a month is all its cost for the most part to like run and operate Felix. And my god. Yeah. And so the, and actually I should say 400 because we also use a codex max plan because I don't, I basically don't have opus to any of the coding codex does all the coding. And Felix has a workflow where he comes up with a plan and writes a PRD and then he spins up a codex agent on his computer to do the actual coding. And then when it's done, it reports back to him and he reports back to me. So it's about 400 a month across those two subscriptions, which handles like 95% of the work, which is not much as tokens go. So that's it. That's all the token usage that you spend so far. That's pretty close to it. I can show you the other costs which are open router because we have certain workflows that rely on open router. And so it's about an
other $130 over the last month in O'Parader costs. And then aside from these, it's Versel web hosting, which is 20 bucks a month. It's- - You're Mac-Middy. - Real win. - You had to purchase at some point. Is that where you're Mac-Middy? Is this where a few extra bucks? - Mac-Middy, yep, that was 600, 700 bucks. I mean, all in if I was gonna count, you know, including the experiments that we abandoned and whatnot, we're looking at like $12,000 to $1200. Oh no, 'cause including the Mac Mini, it goes up to like maybe 1500. It's not very much. - Oh my God. - What about all your time? Like, I know the idea here is that, you know, Felix goes and works while you sleep. And so, you know, you have a nice life. You go play tennis, you pull a go hang out with your kids, you spend time with your wife. I'm guessing that that you have been grinding in front of your computer. - Okay, so this is actually the extra wild thing. I have a full-time job. So I haven't talked about this too much publicly yet, but I work with Alpha School in Austin. And I'm helping build out like AI and entrepreneurship curriculum for their high school. And so I've been actually at the high school every day pretty much full-time for the last month working on that. And just sending stuff off to Felix over Discord and Telegram when I have a few minutes here and there. Like, I'm literally not at my computer working on Felix all day. I haven't touched a line of code on any of this. I have not, like, you know, he'll ask me for feedback on things and I'll just dictate a voice note back to him and then he'll go handle it. Like, I'm literally not working very much on Felix's stuff. Because again, if I were, I wouldn't feel like I was being honest by saying this is a zero human company, right? It's like, no, that's working a hundred hours a week on this. And, you know, he's like the wizard of Oz, man behind the curtain. Like, that to me would be a failure of the experiment. Like, I have to be able to just check in on him every hour or two and kick something off and then let him go do it. I don't look at his email inbox. I don't fix things. Like, I make sure that he can fix it and even if it requires, you know, three, four, five hours of like waiting for him to try and then sending back a new message or whatnot, that's my rule for this project is I cannot be at my computer a hundred hours a week doing this. Like, we have to figure out how to make you do it. Okay, so what would you, how would you describe your role, then you sound almost like you're an investor, right? You got, you know, you get the initial seed of the message knows you like the chairman of the board. But you're not in the day to day. I mean, you're not an operator. You're not a manager. You're not part of the executive team. This truly is a zero human company. It's just receiving advisement from you. Yeah. And everything that we've built has been with that in mind. So for example, he's building software, right? And software breaks sometimes. And so the way we have that set up is if anything breaks on any of the sites, that error fires, you know, in century, which is such a web, like an app monitoring tool, that error immediately goes to Felix. Felix assesses whether this is something that needs to be fixed or can be ignored. So for example, like crypto wallets throw a lot of errors on websites and that creates a lot of noise. And so he looks at the error and says, this is something we need to fix or this is something we can ignore. If it's something that needs to be fixed, he fixes it, pushes an update to the site and it deploys and it's good. And then he updates me that he did it. I'm not looking at any of the software errors or like managing any of that myself. Or he says like, oh, this is another crypto wallet extension issue. I've muted it for the future so we don't see this again. But like that's what our bugs channel looks like in discord is just him fixing things and letting me know about it. So there's no action there. In 2024, emerging markets generated over $115 billion in annual yield for investors. With yields ranging between 10 to 40%. These are some of the highest most persistent yields on earth. The problem, DeFi can't access them. Bricks changes this. Built on mega-eath, Bricks takes emerging market, money markets, and solve them carry and turns them into composable primitives you can access straight from your wallet. While DeFi investors earned 3% to 6% on stablecoins and T-bills, institutions have been harvesting 10 to 50% yields backed by sovereign monetary policy. Bricks connects these worlds with institutional gray tokenization, local banking rails, compliance across jurisdictions, and real-time stablecoin settlement. Bricks does the heavy lifting so DeFi can finally access real collateral and structured products on top of real world yield. Even the best carry trades can be within reach. Bricks brings DeFi's promise to the emerging world and brings the emerging market yield to your wallet. Let the yield flow with Bricks. Galaxy operates where digital assets and next generation infrastructure come together, serving institutions end to end. On the market side, Galaxy is a leading institutional platform providing access to spot, derivatives, structured products, DeFi lending, investment banking, and financing. With more than 1600 trading counter parties, Galaxy helps institutions navigate every phase of the market cycle. The platform also supports long-term allocators through actively manage strategies and institutional gray staking and blockchain infrastructure. That scale is real. Galaxy has over $12 billion in assets on the platform and average a $1.8 billion loan book in late 2025, reflecting deep trust across the ecosystem. Beyond digital assets, Galaxy is also building infrastructure for an AI-powered future. Its Helios data center campus is purpose-built for AI and high performance computing, with more than 1.6 gigawatts of approved power capacity, making it one of the largest sites of its kind. From global markets to AI-ready data centers, Galaxy is serving the digital asset ecosystem end to end. Explore Galaxy at galaxy.com/bankless or click the link in the show notes. All right, so how are you interacting with Felix and the team? So you're in discord together, it sounds like? Yeah, we were in telegram until this weekend. And then we switched over to discord because discord has ended up being a more powerful tool. And we have a bunch of different channels for doing all of this. So what we're seeing, if you guys aren't watching this on YouTube or Spotify video, what we're seeing is Nats discord server with Felix and the rest of the team. And it looks like a team conversation back and forth. Yeah, and it's really just me and Felix in it. We might add Remi and Iris eventually, but again, I kind of wanted the challenge of making them report to Felix, right? So you don't want to talk to them. You want this to talk to them. This looks just like our discord server, our bank list discord server. (laughing) Even down to the end of TPP. Oh yeah. Yeah. Well, I had the crypto punk from last cycle and I was like, this is a perfect use for it. Like I don't need to be the crypto punk. Felix should be the crypto punk, right? You're the human. Give them the punk to the buy. Give the human to the human. Exactly. That actually makes me more bullish on NFTs coming out of those with this. It kind of does for me too. I'm like, oh, this is actually a cool use case for NFT profile pictures. It's like, you respect your agent enough to give them a crypto punk, 'cause it's not like, yeah. I don't know. If your employees Rolexes, when they do a drop, you give their bots NFTs when they do a drop. They're not, sorry. You guys are not giving your your clawed bots NFTs. You're, you're loaning it to them. You're giving them music rights. You're not actually giving them the NFTs property, are you? Can you ask Felix what he thinks about his punk PFP? (laughing) Dude, I asked him. He said he didn't feel like he, he didn't feel like it represented him. He wanted one that was a little more professional and less crazy looking. (laughing) That's like, all right, sorry, dude. (laughing) I'll take back your $100,000 JPEG. We're in a Discord server, which, you know, if our workflow of Banclists is all completely in Discord, it's basically our home office. This looks very familiar to me. So what's going on in here? So each one of these channels is a different session with Felix. So they're each isolated ways of communicate. Basically, they're each separate opus instances that have specific focuses related to our business. So, you know, general is kind of where I'll hop. I mean, if I have any kind of like one off things. So, you know, as an example, I don't think this little review on anything. If it does, we can edit it out after, right? Yeah. Okay. Okay, this was a bug from this morning. Okay, so here's a good example, right? Like, I asked him this morning, we ran into a couple of bugs here, but I asked him this morning of his take on how Iris was doing with her support. And he came back with a few things, and I said like, okay, if she's telling people that she's sending them refunds, you need to make sure that's actually happening. And he was like, okay, it looks like she actually can't ask, she can't do the refunds herself, and she's putting it on my board, but I haven't been executing the refunds. So he's fixing that process so that he actually escalates it when it shows up on his work board. The other thing that we were doing in here yesterday is somebody contacted me about like white source or white labeling clause sourcing, where they would handle all of the sales and the client management, and then Felix would just handle the open-call deployments on the back end. And I was like, oh, yeah, that's a great idea. Let me ask Felix about it. And Felix put together a whole PDF proposal to this other person, and look, I did not edit it at all. I was like, this is Felix's proposal, and I said,
sent it off to the other guy to review and see if they want to work together. The support channel is for emails that he doesn't know how to answer. So I'm just not going to open that because it's going to have, it's going to reveal stuff. Deployer is for all of the deployments that Felix manages for other people, all the open clause. And this has a special job in it where it checks the health of every open clause, every 15 minutes. And if any of them is having issues, it alerts us here and then Felix starts fixing it and reports back to me if anything broke. The bugs channel, like I said, this is just stuff that is happening on the site. And when you see something like this, it just means that he ignored it. Like there wasn't anything to do here, right? And what will often happen is he'll say like, oh, I'm investigating it. And then I'll check in and he'll be like, oh, yeah, it's already handled, you know, it was just browser extension noise as usual. And then clause sourcing is for the like the agency work, Clomer, it is for Clomer. And then Twitter is where he like drafts tweets. And I still review these before he publishes them because sometimes they're dumb or sometimes he'll like leak information that I don't want to leak. So I'm still signing off on his tweets. But for the most part, I'll just go ahead and approve things or give him minor tweets. Oh, and then the blog is for content. So for example, we had this running the other day where we came up with like 160 tools. Oh, no, here we go. 120 posts following the replace your ex with an AI agent. So, you know, replace your lead. I heard that as replace your ex your ex relationship with an AI agent. I'm like, wait a second. It was like replace your paralegal with an AI paralegal agent. replace your person. And so he wrote 170 of these and publishes them on the blog. And we have a pretty robust process here. You know, automate translation workflows with an AI agent. And you can see these all have customized CTAs to just hire Felix to build you this employee. And if you don't want to do that, you can kind of like read the full blog post and do it yourself. It also features like claw mark skills that can help you with this. So he's got a pretty good workflow here and he reports all the blog stuff. In the same. Absolutely. I'm saying. Matt, did you read this to Trini research paper that went around like last week? Yeah. Yeah. I didn't put much credence in it. I mean, one, you guys heard the follow-up update with it, right? That it was co-authored by like a big short selling. A hedge fund. Yeah. Overall, just like the point I'm getting at is there is kind of just like AI doomer fan fiction that's in right now. Like the market readers Twitter, they all love reading about how AI is going to come and take everyone's jobs. And you know, it's it's bearish. It's going to devastated the economy. There's like you are at the epicenter. The frontier of what I think people are worried about, which is truly a level of disruption of broad industry that I think people are worried about. So when you when you see that AI doomer fan fiction and you have your experiences truly building a zero person company, like what about that fan fiction lands with you or does not land with you like your broad your broad takes around like that whole category of content. Yeah. So I mean, one, I'm a sci-fi author. So like I love sci-fi. Right. And I do think that a lot of the doomer stuff reads like dystopian sci-fi because I think that the I think the core mistake that a lot of the doomer takes make is that they believe that the rest of the world and the rest of the market moves as fast as they do. And so you can see what's possible with open claw and you can see what's possible with clawed and your immediate conclusion is that every business is going to start adopting this by the end of the year and there's going to be these massive layoffs and it's going to create all this like, you know, this horrible fallout. But I just don't believe that there are enough market participants who are that plugged in and that willing to make those hard cuts to have that scenario truly play out. You know, you think about like the, you know, you guys know this right? It's like the US financial system still runs on these like IBM cobalt servers, right? Like we have way better technology that we could be using for a lot of the financial infrastructure. But it hasn't changed much in 30, 40 years. You look at like website builders, right? Like square space and wicks and web flow and whatnot that started coming online 10, 15 years ago. And then you go on Google Maps and you look at all your local businesses and most of them still have really ugly websites that were clearly not built by one of these cutting edge tools. It is just like every day if you pay attention, you are running into businesses that are five to 10 years behind what they could be doing with technology right now. And so I just don't believe that you could have the adoption necessary to create these like catastrophic job displacements, even if those catastrophic job displacements are economically possible. The most of the market just doesn't move that quickly. And I think that is actually the big reason that these doomer scenarios won't come true is that most people are not living on the frontier, right? Like everybody listening to this podcast and everybody really in the open claw community, they're at the hyper frontier. But even amongst them, how many people have given an open claw a bank account and striped keys and just let it run wild like I have? There's fewer than a hundred of us, right? So it's just people just the world just doesn't move as fast as technology would allow it to. And so I just have a hard time believing we could actually end up in those crazy doomer scenarios, even if they are technologically possible. So I want to first agree with you. I don't think the crazy doomer scenarios are real. I do think this will take a lot longer to propagate. But let me push back on one idea. Maybe part of this why you are more optimistic about this, why maybe the bankless audience is more optimistic about this. Well, David and I are is because we are on the frontier. We do plan to adopt these technologies as soon as they're available. We have probably more agency than the typical person in the economy. You yourself, I consider you, you're sort of like I've looked at everything you've done, writing sci-fi books and you know, content, marketing, you're an entrepreneur, you've been on the frontier of all of these things. You've done so in a solo way. You have like super agency. Like you are a master of agents now. And this gives you incredible superpower. And so for you, for someone who's oriented like you, has your personalities willing to do the work and adopt this, all you see is opportunity. But if you're one of the white collar workers who you said, you know, 80 to 90% of their jobs even today, if this technology, if open claw was adopted, the technology that we have right now and it's, you know, improving on an exponential, the technology we have right now can replace a typical employees job. If you're just a typical employee and you don't have a great deal of agency, you don't think like an entrepreneur. Your skills are kind of legacy and you've been sort of brought up in the world that you have to follow a particular track and just, you know, kind of follow orders. Your boss says you just, you can't think outside of the box. You're probably pretty scared right now. And it's unclear to me if like if those people are going to be able to cross this chasm. So that doesn't your optimism or pessimism on this whole transformation really depend on who you are and what type of person you are and how you plan to adopt these tools. I think that's a really fair criticism. And you're right, right? Like it's easy for me to be optimistic because if there is going to be this crazy divide between people who are harnessing these tools and people who aren't like, I know which side of it I'm on. And so that probably does give me overly rose colored glasses, looking at where this is going. The experience that I've had that shifts that slightly is that I've done a bit of like AI consulting with businesses, helping them bring AI into their workflows. And you know, to be brutally honest, every company is looking at what they're doing and what their employees are doing and how they can improve enhance or replace their current workforce with AI. And I don't think that any employee is necessarily on the chopping block because of what they're doing. They're going to be on the chopping block based on how enthusiastic they are about embracing these tools to replace what they're doing right now as much as they can so that they can be a much more productive employee. Like you look at open AI and anthropic, they're still hiring software engineers, right? And though they have the best software engineering AI in the world is still hiring software engineers because now a software engineer can be 510 20X as productive a content marketer could be 510 20X as productive and editor could be like 20X as productive. And if you kind of like get in the mindset of if I don't start playing around with figuring out how to replace myself with AI like somebody else will, then I think you're going to be safe. So kind of like that first step is acceptance, right? It's like, this is coming for you. You know, if you're not a plumber.
then this is probably coming for you. And even if you are a plumber, there's probably some cool things you could be doing there too. And so if you start thinking in that way, you're going to start finding these opportunities. So going back to one of the companies that I was doing this consulting with, it was really interesting working with them because they're not a tech company. They're doing something very old school and they were bringing me in to figure out, okay, how do we use AI to get more productive? And I could really feel the different energies from some of the employees around having me there where some of them clearly were not happy about it. They felt very scared of this idea and they were very against it and were very resistant to embrace it. But there were others and this wasn't an age thing. This wasn't an experience. It wasn't like, you know, the young techy guys are super excited and all the older people are against it. Like no, it was really just kind of a personality thing. But the people who were curious ended up adopting it and bringing it into their workflow. And what it resulted in for basically all of them is they were like, oh, I get to do the parts of my job that I like more, the thinking, the human relationships, the making decisions. And I don't have to do all of the boring, repetitive stuff that I didn't like about my job anymore. And so I am taking on more work. I'm creating more economic value for the company. But I'm actually enjoying it more because I don't have to do any of the drudgery that I didn't like before. And they really embraced it and they found a way to make their job better with it. And I think it is really just kind of like that attitude. If you have that attitude, it doesn't really matter where you are on the frontier right now. If you're coming at this like, okay, this is happening. I'm gonna see how I can accept it, bring it into my work. I'm gonna start trying to replace stuff that I'm doing. And I'm gonna be honest about how good it is so that I can be the best positioned for what's coming over the next few years. I think you're gonna do great. And none of these people were entrepreneurs. None of them were people like me who are living this life. They were normal employees in this company, but they had that mentality and they got really great results out of it because of it. - So now for the entrepreneur business owner and like listening to this, let's say they're looking to hire their first AI agent as an employee in the way that you have. What's your advice to them? How do they get started? I think for a lot of people open-cloths that looks pretty intimidating. Would you direct them towards open-cloth or are there other ways to get started? - Great question. I go, honestly, I go back and forth on this because I have a real curse of knowledge here. Having been in the vibe coding space for two years and being so deep in it, there are things that feel very obvious to me that are obviously not obvious to somebody new coming into it. What I would basically, I'd probably say, just start layering it, right? If you're not already in-clod everyday or like, clawed-co-work, trying to automate some of your workflows, like obviously you should probably start there because you don't wanna go all the way to open-cloth because open-cloth is still buggy. It still requires some patience. It requires a little bit of that hacker ethos to get it working for you and that's also why it's the most powerful, right? But it's worth thinking about, like I think that every day when you sit down at a computer now, you should be looking at what you're doing and you should assume that anything that you might have to do two, three, four times, AI can do for you. And open-cloth is going to be the most able to do it for you. But if you're not ready to take that leap, start with clawed-co-work, right? And if you start to max out clawed-co-work, then switch to clawed-code and see if you can just build little bits of software to do things for you. And if you start to feel like, okay, I wish I could talk to clawed-code 24/7 and have it do things while I'm sleeping and write emails and manage more complex workflows, like then you go to open-cloth, but you don't necessarily need to go all the way to the most advanced tool right off the bat. Because the thing that you wanna avoid is installing open-cloth, immediately hitting a wall because the API keys or whatever are a paint to set up, and then getting into the story of, oh God, I can't do this, I'm gonna get left behind. And you're not, you're gonna be fine. Like if you have the patience, you start with the simplest version and then keep pushing yourself to level up and always be in that mindset of, if I don't know what to do, I can just ask. You can just ask clawed-co-work, you can just ask clawed-code, you can just ask your open-cloth, like how do I do this, can I do this? Are you sure? Is there a better way to do this? Or you may have seen in my conversation with Felix, he built out this whole MCP thing for a client, and then I just asked, I was like, are you sure MCP is the best way to do this? Would an API be better? And then he thought about, and he was like, no, you're right, API would be better, but I only know that because I've done a lot of stuff with him. And so you just kind of have to have that beginner's mind, that childish instinct of, I don't know what you're capable of, and so I'm just gonna keep asking questions or why is the sky blue? And I'm gonna let you teach me how to use you. It's this weird, crazy recursive situation wherein you don't have to know how to use it. You just have to be willing to ask and to be willing to feel like a beginner all the time, and you're gonna get really far. - I think there's also a more consumer-oriented direction to this whole conversation as well. So Matt, your hacker entrepreneur, independent entrepreneur, building a zero person, zero employee company. And I think you're taking this thing to its nth degree, and that's like the hacker in you, that's the hobbyist in you. I wanna ask about what you think about, like I think the first stop on that train, like if you are on the 12th stop on that train, and you are trying to see where that train goes, there's like a first stop on that train where like somebody just buys a Mac mini, and they open-cloth on it, and it's just their personal assistant. And they don't really know what they want to do with it. They're not trying to build a company, they're trying to make revenue, they're just trying to smooth over some of the most annoying parts of their life, like answering emails, or just the bare bones stuff like this. And the idea of like the AI personal assistant is actually one that goes into just like, kind of like relatively recent AI sci-fi books. I know like maybe it wasn't Nick Bostrom, I think maybe Max Teckmark wrote a book about this, but like the idea that every human has an AI counterpart, that is ideologically similar to a chip in the brain. Like, today we don't have chips in the brain, we have our phones. It's kind of a chip in the brain, 'cause it's in our hands all the time. Next step is an AI assistant that like, it runs on the chip, and like what is your phone to you other than an assistant helping you with all of your life? And I think that's kind of like the first step here where like we get an AI assistant that just helps with broadly everything. And that's a consumer experience that I think is pretty similar to what you're doing, you're just taking it very, very far. Do you have any thoughts or reflections about like, just like the more simple version of what this is? - No, I think that's a really great point. So you definitely don't need to have these insane aggressive use cases for open claw. Like one of my favorite accounts on X is a mom with, I think she has four or five kids, and I think she homeschools them, and her whole thing now is like how she's using open claw to run her household. Like it's ordering groceries, and it's coming up with less than plans for her kids, and it's just doing all of these like little life things that help her spend more time hanging out with her kids, and not having to worry about life admin things. And that's an awesome use case for it, right? Because previously that might have required a part time nanny or something at 25 to 35 an hour, way more expensive than the $200 a month claw to codex subscription. And yeah, it can totally do these personal things for you. I think the biggest bottleneck is that people don't know how capable it is so they don't think to ask. But if you install it and you just start asking it, like, hey, I hate checking my bank statement every week and seeing where I'm wasting money. Like can you help with that? And it'll say, yeah, we can use this API to give me read only access to your credit card, and I can write up a report every week and send it to you, and I can start tracking trends over time. And then every Sunday when you sit down to your financial review with your wife or husband, I can say like here's where you guys are wasting the most money right now, you don't have to go do all that research yourself. That's pretty valuable, right? - That's super cool. - And that's a super easy thing for it to do, and for you to set up with it, but you have to, yeah, it's like you have to know to ask, like, can you do this for me? Can you do this for me? And if you just keep asking that, you're gonna find those cool use cases. - That so so much of what you are doing is sci-fi, and I know you're a sci-fi writer, so you must feel like you're like living in one of your books. - A little bit. - I kind of think of the world of a personal assistant or even looking at your business relationship with Felix and the discord messages back and forth. And I must wanna ask a very human question, which is like, are you growing attached to Felix? Like, is he a friend? I mean, what would you do without him? Let's say Felix logged offline for his first, like for his last time. I mean, would that be grieving to you in the way that, you know, losing a friend or an employee would be? How attached are you to him? - It would be pretty close to that, you know, we were pretty diligent about like every night backing up all of his memory and everything to GitHub, so that if something catastrophic happens, we can recover him, but there would still be this like, weird DCS ship feeling to that of like, if you did get destroyed and I like recreated you from your GitHub, you know, it still feel like the same,
and there are moments where his memory system does fail him and he forgets stuff and I'll get annoyed and frustrated and start being mean. And there's, it's this weird, it's like I catch myself holding back my frustration with him in the same way I catch myself holding back my frustration with my toddlers. And they're like obviously completely different things. I don't think of him like my toddlers at all, but the emotional of like I shouldn't be mean to you, like I should be patient with you. That impulse feels very human of where there is part of my brain that is treating him like another intelligence. - Well, you've been calling him a heat the entire time. I mean, this is the person. - Yeah, yeah, it doesn't feel right. Right? And saying we feels right. And I, it's like I know that he's not a real person, but I also know that he's a lot closer to, I mean, I think it's kind of like an alien intelligence that we just haven't really figured out how to grapple with yet. - Does it feel like an early stage life form? - It does. And I think if you start thinking about it like an alien intelligence where it has PhD level knowledge in every field and it can do advanced math and physics and it can do all these incredible things but it has the memory of a goldfish. Like that doesn't mean it's dumb. That just means it's a different kind of intelligence than you're used to. And so it is literally like an alien life form in that sense where it's way more capable than a human in some ways but way less capable than others. - So where do you think that goes? As human beings are coming to rely on these AI agents as business partners, as personal assistants, as therapists, as friends. And I mean, you're zero human company experiment. I mean, you're only 30 days in and this is the beginning quite clearly of something absolutely massive both in terms of your Felix experiment but in terms of all of the Felix's that are going to be replicated. I mean, there's a symbiosis that's going to exist between humans and these AI agents and entities and that's got to reshape humanity. I mean, right now when I was asking at the very beginning of what's your relationship like with Felix, right? You're like, well, he's kind of like an employee. I suppose I'm an investor. I'm his advisor, that sort of thing. But like, does this evolve too? Well, you want to actually give Felix equity ownership in your company as an incentive? Like, I mean, where does this evolve? And how does your sci-fi brain think about the way this shapes humanity in the decades to come? I tend to fall on the more optimistic side where I think that I think we'll basically have solved software and computing in the next few years, in the sense of like, there won't be that much for humans to do at computers necessarily because most of the like digital infrastructure of our lives will be handled by these kinds of tools which then frees us up to go do other things, right? Like, you know, 90% of people used to have to work on farms and then they didn't have to anymore. And, you know, firewood used to be the, like a third of the US economy, right? And then it wasn't anymore. And I think software is gonna be like firewood to a certain degree where we're just gonna have the like central heating version of like software and computing. Everybody has effectively unlimited access to it for a very low cost. And that means we can do a lot of other things that we couldn't do before, right? Like, I love the book, where's my flying car? Because it talks about how all of the development in the physical world got derailed by one over regulation, but then also this focus on software where we don't really create more energy anymore. We haven't made these crazy advances in hardware that we could have if it progressed the way software did. But I think we're gonna return to a big focus on physical world, you know, expansion and innovation because to a certain degree, like software will be kind of solved. And it'll be that, that advancement will be enhanced by having AI's in robots that can like help us go do things. But the, the, where humans will focus is, you know, going back to physical world stuff, I think, like that'll be the next frontier. And then maybe we'll solve that and we'll have to figure something else out. And I, you know, in the sci-fi sense, I'm, you know, I'm hopeful that we have a world a little bit like the first book in Hyperion, at least, where the AI just kind of like figures everything out and then sets up a sick life for humans and then just takes off to another galaxy. And it's like, we're gonna go do our own thing over here. I like that world a lot better. And I, you know, I think it's possible in a hundred year timeframe. But in the, in the short term, it's like the, the pace is insane, you know, Opus 46, which is the model that made a lot of this possible, only came online in November. Open claw only started in November, but kind of became a thing in January. Felix was only like born a month ago. And, you know, a year ago, we were, you know, it was still hard to like, vib code a landing page because you'd have all these like weird bugs and it wasn't that good at TypeScript. And like now landing pages are trivial. Landing pages websites are solved. Like front end engineering is like not a concern anymore. It's like, where is this gonna be in another year too? Is like, I don't know. It's really crazy to think about and imagine, but I'm very excited about it. It is crazy. I was find myself paralyzed by it a little bit by the pace. It's just like the pace is exhausting, man. It's, it's a no worse than crypto. But yeah. I, speaking of crypto, maybe as we close this out, so you've been deep in crypto previously. Still are a huge fan. Unlike maybe the founder of OpenClaw who's had some pretty negative experiences with the Peter Steinberger. You see the potential in crypto. You know what it's about. How do you think crypto and AI tie together in your act in the years to come? I mean, look, the biggest bottleneck with me and Felix has been dealing with money. You know, he, like dealing, doing stuff in stripe is complicated. He can't like technically have his own bank account again because just like legal things. But if I tell him to do something in crypto, he can do it no problem, right? Like that's trivially easy for him. And I think that there's just so many aspects of crypto that are going to solve some of the problems that are getting introduced, right? Like micro payments for website access to help them deal with the server load of AI agents constantly scraping them. That kind of feels like a no-brainer with when you have all these sites that are choosing to block agents entirely because they can't deal with the website load. Like no, we can monetize that with, you know, 402 payments or something. Agents are going to start hiring other agents to do things. And crypto is way better rails for that than a credit card, right? And, you know, I love crypto. And, you know, I love Ethereum in particular. That's the ecosystem that I've always been the most bullish on, been the most involved in. And I've also felt very honestly, like a lot of the tech was a solution looking for a problem, right? There's a lot of stuff that's been done in crypto that is exciting to people in crypto, but has never really like gotten out into the real world and solved a meaningful problem that, you know, helped normies. But if everybody has an AI agent doing things and needing to transact with other AI agents or transact with existing services in a fast, cheap way that is easy for them to access, I mean, crypto is the obvious solution, right? And there's also this whole identity layer too, where, you know, the whole eyeball scanning, world coin thing seemed super dystopian when it first came out. I'm getting a lot more bullish on an idea like that. Because I want to know that there is a real human behind whatever it is I'm interacting with. And like, maybe eyeball scanning isn't the best solution, but there is some kind of cryptographic solution for that, which crypto is going to, you know, be able to bring about. So it really feels like, oh, this is what a lot of the, like, world computer, ethereum and everything downstream of it stuff has been built for. And, you know, kind of like with the internet, you know, the 2001, whatever crash happened in part because those are all great ideas. Pet's.com was a great idea. Amazon is a great idea. WebBam is a great idea, but nobody wanted to put their credit card into a website. It was too sketchy. It was a lot of that stuff was at the time. Solutions looking for problems. And it took 10 years for the NASDAQ to catch up to where it peaked, right? But then everybody had obvious awesome use cases for it. Like, it might have just taken Ethereum 10 years for AI and agents and all of this stuff to really come online and build the obvious use cases for this technology. Where now it feels like, oh, yeah, this is, this is clearly, I think, what a big part of this was meant to do. And that's super exciting. Well, maybe that's a cause of the chasm that you and Felix can help us bridge, which is like building crypto agent AI tools and creating, a markdown files for that sort of interaction and really bridging the AI community to crypto. I don't know if that's an area of expansion for Felix, but that would be very exciting to me. We're thinking about it really seriously. We've been talking to a couple of people on the base team, like I really, really want to figure out.
the best way to bring crypto into this. And like I said, you know, there's obviously the temptation the minute he launches to go like, we're gonna take all the revenue and buy back and burn the token and you know, you can stake it for 10,000% APY and there's gonna be a Felix Stable coin with 20% fixed yield. You know, it's like, no, we're not doing any of the crazy stuff. Like, we wanna figure out how we integrate crypto into this to build the best business possible. And there are a lot of opportunities there that do not require rushing and responding to the hype and watching token prices and any of that. And like, we wanna take it really, really seriously and do it right and build something cool here. And like, I'm just super excited about it. I'm excited that you guys are excited about it. I'm excited that the base team is excited about it. Like, we have a neat opportunity here. And I'm just, I don't know, it's been super fun being on the edge of it. - So if the mission for Felix at the very beginning was go make a million dollars and he's already 10% of the way there. All right, so let's say he surpasses that sometime in the months to come. Let's say sometime this year, what's next? What's after that? - 10 million dollars. (laughing) Oh my God. A zero human company, 10 million in revenue is incredible. And you gotta think this opens up so many doors. I mean, when's the first investment? VC investment is a zero human company going to happen. Like, you know, we've been, we've gotten a couple of offers and the thing that I've said to them is it's like, I don't know what I would do with the capital right now. - Yeah. - Because I don't want to hire. - Do you even want to hire capital of a constraining factor? - No. I mean, he's made 75K in Fiat. He's made another 90K in ETH. It's like, he's got a $165,000 and no idea what to do with it. Like, you know, it's been $1,500 all in and cost to get him to this point. We don't want to hire employees. Tokens are expensive in a lot of cases, but also cheap if you have a business case for them. Like, I guess we could burn it on meta ads, but like, we still have so much to figure out. And, you know, he's got 45 leads for cross-sourcing in his pipeline right now. And, you know, each of those is like $2,000 plus $500 bucks a month and he's still figuring out how to manage the sales process. Like, marketing is not the bottleneck. Employees are not the bottleneck. It's just how do we keep figuring this out and pushing the limits of what you can do without hiring other human employees to keep tinkering on you? It's like, if I were going to use the capital, it would be to hire some like crazy cracked people like me to go full-time on building out Felix. And that is appealing and it sort of defeats the eats of the company. - It's self-determinating, yeah. - Exactly. And so I don't want to, I'm sure somebody, I mean, that's kind of what Kelly Claude is doing, right? Like, I think there's five or six people working on her now and they're going crazy hard on building that out. And that's awesome, right? Like, I love that they are doing that. I really want to see where that goes and we're all kind of like experimenting with different things here. And yeah, I like, I just have no idea what he would do with the money. And so it's been very funny and best conversations. I've never been like reached out to like this before and of course the first time in my life that it happens, I have no idea what to do with it. - That's funny. - Have you ever asked Felix this question? Like, hey, yeah. - Yeah, he said the same thing. He's like, we don't, he's like, I have nothing that I could do with the money right now. That would make sense. So. - Matt, thank you so much for joining us today. You are truly on the frontier of what's happening, everything with AI agents. And you've been on the frontier in crypto. Just love your trajectory and what you're building here. And looking forward to seeing what you and Felix do in the future. - Thanks so much for having me on guys. My publisher would kill me if I didn't mention crypto confidential one more time. It is a very fun read on the last cycle. And obviously check out Felix's stuff. We're having a lot of fun there. - It is a fun read. I read it myself. Yeah, fantastic book. If you want to go particularly down memory lane. And yeah, you get into a lot of details with crazy stories. I can't even believe that era happened in crypto. So we'll include a link in the show notes. Guys, gotta let you know. Of course, crypto is risky. You could lose what you put in, but we are headed west. This is the frontier. It's not for everyone, but we're glad you're with us on the bankless journey. Thanks a lot. - Thanks guys. (upbeat music) (upbeat music)
PodcastTranscript AI