Where Does Crypto Meet AI? w/ Gerrit Hall and Taariq Lewis

Taariq | Seren (00:00.565) this Riverside thing, It's groovy. Rex Kirshner (00:01.838) Yeah, cool. So sorry, I'll just like intro us in and then I'll ask you to introduce yourself. Let's see like editing. cut this all out. Taariq | Seren (00:07.721) Okay. Yeah. Rex Kirshner (00:13.134) All right, Tarek, Garrett, welcome to the Signaling Theory Podcast. Gerrit Hall (00:18.085) Rex, good to be here. Taariq | Seren (00:18.293) Great to be here. Yeah. Rex Kirshner (00:20.888) So Garrett, are a old timer here. So we'll give this to Tarek first and I'd love for you to just introduce yourself to the audience. And man, we met in crypto world. It feels like a whole different age or lifetime ago. So would love to hear what you're working on in this next venture. Gerrit Hall (00:48.861) Tariq, that's for you. They don't care about me. Taariq | Seren (00:50.195) It's for me? right. Okay, so sorry. And I apologize, everything is freezing on my computer and I'm just like, what is going on here? Okay. So like, I'm literally like telling Codex like, dude, why is my computer running slow? Okay. So listen, by the way, anytime I meet with you guys, like I come up with these like procrastinations and they're hot and then they die. The last time I was talking to you guys, was like, L2s, L2s are the future. L2s will change our lives. Rex Kirshner (01:08.323) Yeah. Rex Kirshner (01:16.161) Yeah Rex Kirshner (01:22.616) Yeah. Taariq | Seren (01:23.588) are dead. Rex Kirshner (01:25.57) Yeah. Taariq | Seren (01:26.441) You know, L2s, we barely knew them. So yeah, I've been in crypto since 13. I started as a Bitcoin devs with Andreas Antonopoulos, worked on various products like MasterCoin, counterparty. If you know Polymarket, then counterparty was the Polymarket before there was a Polymarket because we did betting on the Bitcoin blockchain using a fork of MasterCoin, which became Ethereum. So I am unfortunately too old. I need to go into retirement. But it's so good to be back. And I really, really super pumped to talk about transitioning from crypto into AI. then, know, essentially using the lessons we learned from crypto in AI, because they're so obvious, like, it's so funny how everything we're doing in AI is like, yeah, we did that in crypto, we learned that lesson. And now it's, it's, it has a broader audience and, and, and a wider, wider appeal. Rex Kirshner (02:26.446) Yeah, man, totally agree. So why don't you just like jump in and tell us a little bit what you're doing with Serene? Like what is the transition story from crypto to doing this database stuff? Like where do you see the future going with it? Like why are you betting this big here? Like just talk to us about what you're building. Taariq | Seren (02:44.957) Yeah, so we've been, so my blockchain project we've been working on is called Paloma, palomachain.com. And shout out to all the 12 validators that are still validating that network. We love you and we're coming back. And in 2023, we launched something called Paloma Lightnode. And the objective of the Lightnode was to help us essentially allow people to secure the network without having to use a major validator software. And in that we were like, hey, maybe they can use it for AI. Maybe people can essentially allow, use their lightnotes for LLMs to take requests from folks. We couldn't get it to work because at the time the LLMs weren't really useful. Most of the LLMs in crypto were essentially Twitter bots, tweeting and YouTube bots doing YouTube videos. Those were all great. And we spent a lot of time, we did LLM driven NFTs. We had a whole NFT. offering that was all AI based. So at the end of 2023, I think we were in the summer of 2024, summer 2024. When we did the light node, we did another thing we said, hey, why don't we do AI agents for phone calls? Like we would, your agent would got a phone number, it would call you, it would call on your behalf, people would call it. And the technology just wasn't there, it was very hard. We were using Lang chain, we were using Twilio services. And everything was very, very hard to put together because we didn't have AI programming. We didn't have bio coding at the time. This was before Cloud Code came out. Cloud Code came out in 2025. And so I think in September of 2025, we got a hold of Sonnet and Cloud Code. And we were at Sonnet 3.0 at the time. And that really changed the game for us because we decided let's see what we could vibe code. So we started vibe coding essentially database programs. And it's very funny, we were very afraid to vibe code anything in blockchain because we thought that the LLMs were so bad, they would mess up the code and we couldn't afford to have the code mess up. But from September all the way to December of 2025, we started a new company called Seren. And Seren's goal was to essentially be a fork of Neon. And if you don't know Neon, Neon is the database that powers a lot of the vibe coding platforms. Taariq | Seren (05:14.486) and their biggest competitor is SuperBase. if you know any of the Vibe coding platforms, they're all powered by either Neon or SuperBase. So we forked Neon's open source code and three people pulled a $1 billion company's platform into existence in three months as a multi-tenant hosted database. And so that told us that software is over. Once we could do that, it's done. I we went pitching venture interviews And people couldn't understand how could just one or two guys create, you know, essentially what was complex software. took years and years to build, create that in three months. And then we were trying to sell it for 50 million. And I think a lot of people just don't understand, you know, we were like, well, this is what we do in crypto, right? We fork other people's software, right? And we build it and then we sell it. So people were like, wait, you forked it? We're like, yeah, that's what we do. So the crypto culture of forking, you know, you know, you know, what we call Rex Kirshner (06:03.854) Mm-hmm. Taariq | Seren (06:15.042) implementations is one of the things that got us into AI. We copied somebody else's code and tried to run it. And I think at the end of the year, we realized that all software is going to zero. All software is commoditized. There's no defense against ongoing and continuously proving LLM. I think at one point in December, was doing smart contract security debugging by just running everything. doing through Claude and Codex, like people would hire me and I would say, yeah, yeah, my security team, you know, we're really smart guys. You know, we'll take a look at your contract. What I would do is just have Claude run it through it and be like, hey, find bugs and I would ship it as mine. So when we found that that was really easy and the cost for writing software just collapsing, you know, we really took the idea that Seren's future would be really about stable coins. And people say stable coins, head scratch. How is a stable coin important? So what we do is we believe in our view, and you'll see me post on LinkedIn, I'll say it, if software is going to zero and those prices are going to zero, then the only thing that matters is your ability to buy tokens and your ability to buy time. And the way you're going to buy tokens is you need people to pay you tokens. But they can't pay you tokens by giving you tokens against term of service, but they can give you money. And if you are a kid in Vietnam, and you have time to work on prompts, you want somebody in America to pay you, they can't send you cash. They might be able to, but they can't send it as easily as if they had a stablecoin and they could pay you. And then when you think about it, that human is not going to pay you, most likely their agent will pay you. And if their agent is paying you to get more tokens to work for them, then again, this agentic commerce and agentic model starts looking very much like everything we've talked about in crypto, which is agents will have stablecoin balances. They will hire humans to do work. Those humans will do the work, get paid in stable coins, and then those humans will use the stable coins to buy more tokens so they can do more work. So we see really the future of AI really hinging on a very important pillar, which is called stable coins. And so, you know, just like I last time I saw you guys, said, L2s, I'm not going to say stable coins because stable coins essentially allow for a very smooth transaction process between agents. Taariq | Seren (08:43.809) and humans. And when we started it, we worked on X4.2. We did a lot of work on X4.2. Shout out to the Coinbase team, Kevin Lefeu, Namel, Dal, they all were very helpful to help us onboard X4.2. X4.2 scan, you're going to see a lot of X4.2 transactions. But what we noticed is that again, in X4.2, people were still thinking crypto. We're still thinking crypto, crypto, crypto, like crypto to crypto payments. But where stablecoins work is where people don't need to know it's stablecoins. They just need to know that their AI can either get money, pay them money. or the AI can transact with its own money. And so the default for effective transactions is stable coin, but the default use case is not crypto. So for us in Seren, what we're learning is that people are using our stable coin, not for doing crypto transactions, but doing simple things like I need to find a job. we have a stable coin called SerenBucks and they use SerenBucks. Most of our users don't know that SerenBucks is a stable coin, nor do they care. but they end up using it. They want to buy more compute for tokens, they will buy the stablecoin. They're like, yeah, I'm buying credits. But when we present to them and say, yeah, you're buying credits, but really the credits are one to one to a dollar. And so today we are now talking about how now we have people asking us, hey, can I send my Seren bucks over to my friend? And for us encrypted as old hat, we're like, well, that's That's the wallet. That's so simple. This is nothing. We did this like 10 years already. People are not realizing the rest of the AI world is coming and the way crypto now creeps in further, you know, we saw stable coins as a way for crypto to creep into the financial services companies and some of the big ones, the institutions, but now we're seeing crypto and stable coins as a way to creep into regular people who will be displaced by AI. So we like to say if you're a knowledge worker and your career was built on knowing things, Taariq | Seren (10:48.927) your future career will be built on acquiring stable coins to buy more tokens to essentially generate more stable coins and essentially use those stable coins to invest in. Here we go. Crypto or equities or wherever you want to invest in. And so this, you know, I always like to say I'm always ahead of the curve and seeing where these things are going to go. And I'm not always the guy to profit from it, but I'm really excited to say that yes, stable coins has a really, really important part to play in in the new AI agentic ecosystem. And I think people are just getting aware of it. And I was telling somebody in the future, every AI agent will have their own stable coin. Right. And people are like, Whoa, think about it. Like every AI agent, doesn't matter who the AI agent is in a world that is truly agentic to agentic. Any AI agent can mint their own stable coin. And if they can mint their own stable coin, then they can create incentives. They can create pools and everybody in crypto is like, we did that before. That's all high. You're like, yeah, dude. What is old is new. And what is new is cool. So the, the, the road ahead for crypto is really, really amazing. and the massive amount of growth opportunity, I think a lot of people not seeing. But the way we're seeing it is this is all going to come down to displacement of knowledge workers forcing them to find more ways to generate income and that income being delivered by their AI agents in the form of stable coins. And that's Seren. Rex Kirshner (12:16.376) So, mean, like definitely like very cool, very ambitious, but like, gotta be honest with you, man, like you paint a pretty bleak future, right? Or like the future of people that today do knowledge work, which is like, let's be real, most Americans, at least most Americans that are doing well, like they don't do anything anymore and they really just need to hoard capital so that it can grow on itself. And so like, does that mean that like you get your money and then... Taariq | Seren (12:29.929) Mm-mm. Yeah. Taariq | Seren (12:38.975) Yes. Rex Kirshner (12:46.114) Like, you're done? Taariq | Seren (12:48.915) No, so great question. So in a world in which, so there are over approximately three to five million people employed in the software industry in the US. So yeah, the bad news is their jobs are going. They may fight it, they may delay it, but their jobs are going. And the way it's gonna happen, it's not gonna be a collapse of the work for those people. What will happen is, I wrote about it today, is first thing will happen is the salaries will get capped. And it's already happening. As soon as AIs were brought into the enterprise, and given to workers, they were not given raises to use them. And LLMs demand attention. they demand focus, they demand time, but you do not get paid. So when you're working at Meta and you have to show your boss that you've been using LLM, you can't say, guys, guys, you know, didn't have this before, you know, could you please give me a raise? Meta will say, no, this is your new job. You have to manage the LLM. So it's already happening where the income expectation for the knowledge worker who has been given AI tools is now capped, which is essentially means they're getting a salary cut, right? We just haven't seen it like it hasn't been explicitly like we're gonna pay you less money, but what's gonna happen? Rex Kirshner (14:01.454) I'm sorry, I guess I don't follow there. what data are you seeing that AI technology is creating salary caps? If anything, like I think it's causing the best knowledge workers to rise to the top to get paid more and more. And the problem is, it just, well, look at the guys at OpenAI. They just got a hundred million dollar salaries. Taariq | Seren (14:18.249) But they can't get paid more and more. It's impossible. That's the one. No, those those salaries I would say are indicative of OpenAI's participation in the bubble. They are not reflective of what's going to happen in the rest of the economy for people who are in knowledge work. So don't ever compare the regular guy working at LinkedIn with the guy who's a, know, know, AI, OpenAI has what, 8000 employees, you know, who's going to be 8001 salary employees, not comparable. What's going to happen is you're simply going to. So the people who are using LLMs and AIs in their jobs are being more productive, but they're not being paid for the additional productivity they're squeezing out of their jobs. They're not getting raises, right? So because then they would be across the board. And the reason why you can't get raises is because your competitors have AIs. And if your competitors have AIs and they're writing the same software to compete in your market with your products, which before didn't have this competition, and now you are writing software to compete in their market, the margins of software will compress. Because if everybody now has access to token generation, that has access to reasoning models that can create software that is a production level quality, the prices must go to zero. You can't force more revenue out of that, Rex, because you have an oversupply of software. This is economics 101. When supply goes up, price has to go down. And if price goes down, competition goes down because people exit the system. So engineers are gonna quit their jobs. They're gonna say, you know what, screw this. Taariq | Seren (15:53.736) First of all, I'm not having fun writing software anymore because I'm not creating art. I'm not creating real work. I'm just pressing a button and this replicator system is throwing out code that works. Thirdly, I can't make my salary high anymore because they have this guy named Simon who keeps writing shit code and he keeps staying hired. He doesn't get fired. He writes shit code. He hallucinates, but we keep hiring him and we keep giving him more work to do. He is very, very cheap. He is cheaper than me. He's currently 200 bucks a month. I cost at least 20 grand a month. So how is it that they're going to keep hiring more of me and less of him? They're going to hire more of him and they're going to hire less of me. Or they're going to tell me you're going to have to take a pay cut or your pay won't advance. We'll do something else for you. So we argue what's going to happen is people are to start getting compensated in tokens. They're going to say, listen, Rex, I know your salary was 500K, right? These are a million. But here's what we're going to do. We're going to give you $10,000 worth of tokens every month. How you like them apples. And what are you going to say, Rex? Are you going to say no or you going to say yes? Rex Kirshner (17:05.792) I mean, if I liked the job, I'd stay. If not, I'd look for a new job. Taariq | Seren (17:09.245) You have nowhere else to get that many tokens at the job. So remember what you can do with tokens. With tokens, you are Superman. Rex Kirshner (17:12.827) I mean... Taariq | Seren (17:16.881) you can create any reality that you envision. So it is our view that your boss will keep you in the job making less money because they're going to give you more tokens. And if you get more tokens, then you can do anything. You can build another blockchain. You can build a company. You can, you know, create a spaceship. Like you can literally do anything that you imagine if you are given enough tokens. Because right now, Rex, I can tell you, you don't have enough tokens. You don't. You need more tokens and you could find ways to make money with those tokens, right? Like you, you'll be saying, Hey, Tark, you know, I found a way to sell my videos online and I have an AI agent that goes out and you know, gets, you know, charges people for ad revenue. All these things are now open to the knowledge worker, which they didn't have before because they were stuck in one job and they didn't have an agent that could work while they were at the office. So our argument is that in this new world, salaries are going to get capped. You're going to get paid less, the other way that companies will keep you is they're going to give you a lot more tokens than you can simply get from a subscription from Claude or Kodex. Rex Kirshner (18:29.302) yeah, I mean, I think I understand where you're coming from. just think that, you know, like when I, when I was working at anhyzer bush, right in 2015, the fact was, was that like our financial systems were in magnetic tape on mainframes that were created in the seventies. And like, even then the cost of displacing that and creating a better solution from the implementation side was trivial. Taariq | Seren (18:49.204) Yes. Rex Kirshner (18:58.956) Right? Like you could easily go to Pivotal and get like four developers and rebuild a financial system from scratch. That was never the problem. so like, as Gary here, like he knows that I'm like pretty bearish on what AI is going to do to jobs. But when I think about just kind of the scenario you're sketching out, I think maybe that's one possibility, but I just look at how business actually works. And I think that. Taariq | Seren (19:01.173) hurt. Taariq | Seren (19:05.311) Correct. Rex Kirshner (19:26.124) Really what we're going to see is that like people like the three of us who understand how to use these tools, like we are going to do really well. We're going to become much more productive. We're going to generate a lot more money. That money is going to generate money. It's great. And like the reality is, is that our upside, our salaries are uncapped, but like new graduates. Nah, why would you ever hire entry level again? Taariq | Seren (19:36.501) Yes. Taariq | Seren (19:49.345) Yeah. Yeah, that's a beautiful point. I want to lean into that. I have not found an answer for the 10x people are going to go accelerate and then everybody else is going to get left behind. I keep saying that we got to give them tokens. Like I keep coming back to what's going to happen is UBI or food stamps will be token stamps. People will get tokens to help bring them up speed so they don't get left behind. we're going to give tokens to young people and say, use these tokens to learn the job and to bring yourself up because the AI can teach you what you need to know. I'm very much in the camp that, you know, in this, and of course the, the, the AI bubble may burst and if the AI bubble may burst, then all bets are off target. You know, if tokens are now expensive because the AI companies can no longer afford to subsidize people to use them, what's your argument? I'm like, yeah, I don't have much of one, right? Because if Anthropic truly has to charge us the cost of using tokens, many of us will not be able to afford it. So this AI bubble can only grow for so long before it collapses on itself. But again, we've been in crypto. We've seen this before. like, hey, we know bubbles. We know Ives. In crypto, every token collapses at some point. Gerrit Hall (21:11.207) We know what. Yeah, exactly. The fact that we call them tokens is just so funny. We've seen this. Taariq | Seren (21:16.853) So we were like, yeah, we just we've seen this party. So we know the collapse is coming. We know the bubble will pop But in this moment now my argument is that we have weak week, know The market stays irrational longer than we are rational So what we need to do is I tell everybody get as many tokens as you can So I you know, you'll see my post on LinkedIn I tell knowledge workers you have to go to your boss and ask them for two subscriptions You want a Claude subscription max and you want a codec subscription max? and want those two subscriptions to do whatever you want with personally. And that is how you tell your boss, I will be loyal to you and I will stay in my job and you don't have to pay me more money. And recent how it's just put out a post yesterday on the two choices for public and private companies. I don't know if you guys read it. They said there are two choices. You either 40 X your group, you either 10 X growth with AI or you 40 X your marginal revenue or gross margins with AI. And in that article, they said it's going to be two things. The startup and the public company are going to fire people for the 10x growth and then use AI just to grow. And then the other company is going to fire people. It's kind of funny when you read this, it's going to be firing people. And the other one is the way they increase the margin is you fire people and you replace current systems with AI. This is a 16 Z's blog post from yesterday, essentially saying that you have, they said there is no middle. You have two choices. You either go for growth with AI or you go for increased marginal revenue with AI other than your finance. Rex Kirshner (22:51.436) Yeah, but man, like I think this is a 16 Z right? Like this is the shovel salesman saying you should buy shovels for the gold rush, right? Like Mark Andreessen very famously wrote that essay that software will eat the world and like wrote this investment case that everything will converge to software. like, if you want to just like be on the side of VCs, like you can tell the story of the world that that's how things played out. But like, man, take a look around, like the Taariq | Seren (22:57.353) You Taariq | Seren (23:04.437) Correct. Taariq | Seren (23:16.169) Right. Rex Kirshner (23:19.086) problems that are reaching into our lives right now. Gas prices, war, like immigration issues. Literally none of this has anything to do with software. Taariq | Seren (23:23.381) Not software issues. Well done. Well said. I have to agree. No rational person can argue the point. know, I will, you know, again, as an, you know, I can't change gas prices. I can't change immigration. I can't change politics. But what I can do is change the work I do at the day. I can choose where I go to work. I can choose what work I do. I can choose where to interview. These things are in my power. I can choose what to learn. And so Rex Kirshner (23:51.884) Yeah. Rex Kirshner (23:56.045) Yeah. Taariq | Seren (23:57.144) you know, I hear you, but at the same time, like, you know, even though the problems we we we first that frustrate us are are are terrible. The more scary thing is, will you have a job? You know, will your company still hold on to you? You know, I, know, and I in this moment, I'm like, why wait to find out if you are a knowledge worker? You don't want to say I'll take, you know, I'll take the bet that I won't be displaced or my salary won't be capped. You know, you have to to take the Andy Grove approach, know, God bless his soul, I am paranoid and I do not know when my boss will replace me with the AI. I mean, I'll tell you right now, our two best customers for Seren are businesses that are doing this. They're using our technology, they're using Seren to pay their employees Seren bucks, the stable coin, in exchange for the employees teaching the AI their job. Like we have two business companies that doing this, and they're not paying us trivial amount of money to do this. So it's happening. Yes, it is happening. And what we say to them is that if you want employees to spend more time training the AI, Rex Kirshner (24:58.764) Yeah, it's the business model of Accenture basically. Taariq | Seren (25:15.497) You can't pay them cash because cash is useless. What you got to give them is more tokens because they can put the tokens to work. They can have their kids, they can put their kids through school with the tokens. They can figure out new ways to make money, side hustles with the tokens. They can figure out new ways to be useful to the company with tokens. They can do things with the tokens that they couldn't do. But if you don't give them tokens, they will be resentful, they will resist, and they will sabotage. Rex Kirshner (25:42.327) I mean, man, that's, get pretty powerful and pretty bleak. mean, I know that like, I just, I don't feel like tokens are as valuable as, you believe they are at least today. Right? right now I like, I'll admit this, hopefully I don't get banned because of it, but like, have two Claude max plans because I use so much compute, but like, that's enough for me. You know, like that's like, Taariq | Seren (26:04.831) This is awesome! This is awesome! Dude! Rex Kirshner (26:09.472) I can't max out the second one, I can max out the first one. Taariq | Seren (26:12.085) That's awesome! You are proving my point! You have two Claude Max plans! That is amazing! I don't even have two Claude Max plans. Rex Kirshner (26:20.3) Well, yeah, but I guess what I'm saying is like, don't really think like the tokens aren't the valuable thing here, right? Like if I give my CloudMax plan to my mom, she is not capable of creating a side hustle from it, right? Like that's just not where the technology is right now. Taariq | Seren (26:37.107) I want to speak to Mrs. Kirchner. Where is she? Could get her on the phone. This is unacceptable. I want to speak to mom. But I agree. I agree. It's not this is again, this is not an immediate like, boom, this all happens all at once. Everything. Rex Kirshner (26:40.59) I mean... Rex Kirshner (26:45.624) But you know what I'm saying? Rex Kirshner (26:53.656) So how do you reconcile this with this feeling that we are in a bubble and it's gonna come crashing down? Like you're building for this future, but you also are anticipating a crash. Like how do you think about that? Taariq | Seren (27:04.373) So I take the position that yes, there is a crash coming because we're in crypto. We know what they look like. We are in a bubble. The costs for the APIs have been, for the LLMs have been subsidized by the subscriptions and it can't continue for long. However, the LLMs improvements can continue. and if the LLMs will continue to improve, what may happen is just like in crypto, we'll have a slowdown, but then we'll have another ramp up again of the train has already left the station. There will be more and more integration and more and more automation of knowledge work. So the automation of knowledge work is not going to stop. has already left the station. like, again, I will say about in smart contract security, I love Fire Pan, but I'm like, guys, I always tell everybody, formal, very peculiar, is not going to be a barrier to entry. If somebody can just write a prompt and say, do the formal verification on my smart contract, right? The barrier to entry will be the relationships, the networking, the human contact, things that the LLMs can't sort of interact with, but those are just temporary barriers. What's going to happen is you're to have to reduce your price because Tarek just went and prompted two smart contract securities and he charged $2,000 for the work. whereas somebody was charging 50 grand. Again, if other people can do these things and use these tools and compete on price, then I argue that it is gonna be very hard for software prices to stay high or things to continue as they were. So I'm just saying. Gerrit Hall (28:42.025) Good night. Yeah, can I push back a little bit in your concept that software is going to absolute zero as a price? Because I get it that it's directionally true that software prices is definitely going down. But anyone who's used these tools knows that there still is a finite amount of developer hours on the planet. And yeah, it's not as easy if you just go in and say like, hey, Claude, build me a robust enterprise sales app like Taariq | Seren (28:51.241) Yes. Gerrit Hall (29:13.243) one shot at make no mistakes. It doesn't work like that, right? When you get in the, the first 90 % is quick, the last 10 % still takes a long time to get it good, get it working. So you can build fantastic app with, where maybe you needed 10 developers before, maybe you can get away with two and cut the eight that weren't doing much. But there still is some amount of scarcity. And the obvious point of this is like, Why does Claude and OpenAI, Anthropic as companies, like, why do they use Xenefits or like the existing like HR software? They don't just say, vibe code me HR software, because like that was not the tough part. The tough part was like all the random business relationships that got built out and encoded and like all like random quarter cases so they don't get sued when they like hire employees in 50 states with 50 laws. Like there, there still is like some amount of like scarcity. It's not, it's not. It's definitely a grid that's coming down, but there is some scarcity in software still. Rex Kirshner (30:14.798) Yeah. also, sorry. Yeah. And the other thing I'd say too is like, we've seen this phenomenon in basically all tech, right? Where like margin basically or price crashes down to zero and everyone thinks like, Oh my God, this is going to, you know, like bandwidth is so cheap. How's there ever going to be a cable company again? And it's like, Oh, it turns out when bandwidth got 99 % cheaper, demand increased by 10,000 X and that Taariq | Seren (30:14.928) I agree. Rex Kirshner (30:44.066) like grew the amount of revenue that was available. And, you know, that's ultimately where like our current tech giants came out of. So like, what do you think about that idea that the cost of software is driving to zero? It doesn't mean that there's going to be no revenue. It means demand is like going to follow. Taariq | Seren (31:01.333) I want to say that yes, because now, you know, software prices go down, demand goes up. I absolutely agree. They're just not going up for the SaaS companies. They're going up for the open claw implementations. They're going up for open AI. They're going up for Anthropic. Those are the people that are sucking in the increasing demand for that software. And again, if your customers are leaving you, even if it's 5%, even if it's 2%, that will impact your gross margins. That will impact how much you can hire and how much you can It will impact how much you can borrow. again, you're going to have to go to the public markets and say to them, yeah, we can borrow more money or you can buy our stock. And the public market is going to say, why am I going to do that when you are continuously losing customers to LLMs? It is going to be a hard argument to say, no, no, our upside. know, Chamath had a nice post about it where he says, you know, capital markets are going to look at this and start pricing, repricing equity in terms of, you know, multiples, multiples no longer. It 40 to 50. It might be two to five. We're going to have smaller companies with less revenue, with less, quote unquote, Silicon Valley type upsides, because again, it is simply the nature, you know, people looking out can't predict that your LLM Influence industry will be able to drive growth around a small number of winners, as opposed to now a large distribution of players and competitors and self-serve users who don't have to buy Xenefits because they just built their rudimentary version and it works. And it works on open claw and it's agent to agent. It's crazy. But you know, this is the nature of disruptive innovation. you know, Garrett went to MIT. I know he was in the, you know, innovation class at Sloan. So I know he got an A in that one. But Gerrit Hall (33:04.039) Yeah, I was, you took Christian Catalini's class, right? Like the one, no, no, was Brynjolfsson, Eric Brynjolfsson, where he talks about bits versus atoms. And he talks a lot about the concept of when information is free, but it's still never free, right? Like the first page of Google is different from the second page of Google. Like if it takes five seconds, one second, half a second to load the website, there is costs. So I think you're... Taariq | Seren (33:11.829) You know, I love him. Yes. Taariq | Seren (33:22.229) Agreed. Agreed. Agreed. Gerrit Hall (33:30.995) Directionally correct that like yes software is getting cheaper and that's going to scatter the playing field But I don't think it's like fair to say that it goes to free because they're just become something else. Is it like Do you get it coded in an hour versus a minute versus a day? Like those are very different concepts, right? Taariq | Seren (33:45.686) Correct. Yeah. Yeah. The non-competitive. Do you have the regulatory licenses? Do you have the brand? All these things will become competitive modes. I got to jump now for the investor call. I love you guys. This is awesome. Rex Kirshner (33:57.645) No, please, man. Thank you so much for joining us and yeah, man, talk soon. Good luck. Yeah. Bye. All right. Rex Kirshner (34:16.174) Alright, we can take a breather for a second. Dude, that was a little bit going off the rails. Gerrit Hall (34:18.109) Yeah Gerrit Hall (34:23.593) That's what you get when you get Tarek, right? You get spicy hot takes. And yeah, he likes to say he's always like a few years ahead of the curve, right? So whatever he was pitching five years ago was what's like hot today. And when he's pitching today will be hot five years from now. Or I don't know, maybe he's getting synchronicity with our time. Rex Kirshner (34:41.89) Maybe. Rex Kirshner (34:45.536) Yeah, I don't know. Yeah, mean, I think he's more, and obviously, like, I wish this wasn't recorded when I was saying this, but like, I think he's more, like, four weeks behind the curve. Like, you know, the best example is the whole L2 thing. He came on and was like yelling about how L2s are the things, after that was smart. Gerrit Hall (35:00.307) That's right. Gerrit Hall (35:08.264) Yeah, I guess so. guess so. Well, let's see if we can get them on for future weeks. I think it'd be fun to get them back. Rex Kirshner (35:11.894) Yeah, All right. Let me queue us back up and then we'll go for 15 more minutes if that works for you. Gerrit Hall (35:19.549) Yeah, got about 15. Rex Kirshner (35:21.198) Yeah, cool. All right. So we just lost Tarek there, but, um, very interesting conversation. And, uh, we'd love to like kind of continue to pick his brain because like, dude, I don't know if I agree with everything that he was saying, but I do think he's willing to like envision and lean into like the disruptive nature of this technology more than, uh, most of the conversations we have. Gerrit Hall (35:46.215) yeah. Yeah. Tarek is a, like he, he knows how to be at the intersection of technology and business in like an amazing way, right? Like, cause he's out there talking with customers. He's launching businesses. he's hustling and getting funding. so yeah, he always brings a great perspective. We've got to him on future weeks. Rex Kirshner (36:05.622) Yeah, for sure. So man, he actually referenced your latest venture. And I don't think we've talked about that on the pod. So you want to share with us what you're working on? And I think it's a very interesting use of AI because, well, I'll let you go ahead. Gerrit Hall (36:24.295) Yeah, absolutely. So I'm like pretty thrilled because like for a few weeks of the show, I've been talking about AI and I feel like I hadn't quite been public about like my next venture was so I couldn't quite like talk too much about it, but now it's official. Like it's all out. I just did a big long interview with DL research today where I got to like lay out the thesis for this and, know, fire pan as well as do a retrospective on my time at curve. you know, talked about a lot of topics, but Yes, like the official launch announcement came out last week. I am one third of the FirePan founding team. FirePan is bringing continuous AI smart contract monitoring. And what this means is if you look at all the hacks that we're seeing in DeFi, first of all, it's kind of remarkable because there's no one left in DeFi, right? And yet we're still seeing hacks accelerating, which is crazy. If we look at like the types of contracts that are getting hacked, like these are not contracts that just like YOLOed on the mainnet. In most cases, 2025, every single smart, like major smart contract hack that you write about came from an audited smart contract. Like these are companies that did what they're supposed to do. They spent usually six figures plus getting their smart contracts audited, usually by multiple auditors. Like there's the hack that happened just earlier this week on USDR. I think it had something like 18 audits. So the security situation for smart contracts and DeFi is utterly dire. And the DL Llama interview piece, they were asking me, why did you take now to step out? And one of the answers that I gave was quite frankly, I believe that there's an existential risk to DeFi right now. I took all my money out of DeFi. I cannot stomach, like from what I've seen, both like working and just reading the head working on fire pan and reading the headlines. Like I cannot have a dollar in DeFi responsibly right now, because like what we're seeing is that the attack surface is tremendous. hackers can point like it's, it's efficient for them to point a ton of compute power at, all the major DeFi protocols, find the weak edges and drain the money. Gerrit Hall (38:48.137) In contrast like the like surface of DeFi is too broad for like if I'm want to say like I'm on just go defend all of DeFi I can't do that. I can't say Claude analyze every smart contract that's been deployed and find the exploit right like it's It's a huge attack surface, but the hackers are like getting very very good and if you look at the hacks they're like increasing in sophistication and I don't know if this is true or not, but like Rex Kirshner (39:00.024) you Gerrit Hall (39:14.941) I have to think when things like Balancer have been around for years with no issues, and then all of a sudden, overnight, they find a really obscure zero day that nobody had seen before. You can't tell me that there wasn't someone at Lazarus Group that was sitting there plugging in a clod or some sort of sophisticated frontier model at these things and really fuzzing it intently and finding the attack surfaces. AI is going to destroy this. It's crazy. Go ahead. Rex Kirshner (39:40.202) One man, here's the thing, like. Rex Kirshner (39:45.081) Well, here's the thing, like you don't, it doesn't even need to be Lazarus group, right? Like the barrier to entry for these types of attacks has dropped so dramatically. Like really you just need to have like the persistence to continue to put, and you know, some knowledge of asking the right questions, but like, I don't know if there's any reason to believe these hacks are coming from Lazarus groups. They might be coming from just like some teenager in Indonesia, but I definitely agree with you that they're coming, they're LLM. if not completely driven, at least assisted. Gerrit Hall (40:18.985) Yeah, exactly. this is with the, you know, we're at, like, I'd still say like in early days in terms of like the sophistication of LLMs with coding. Um, like, you you've gotten to play with Claude and Codex and Gemini and like every new version they release gets really, really good. And, uh, and, uh, like the problems that I had with like Claude code a few months ago kind of have disappeared with the newest versions that came out. Like it's getting very good. there's still going to be like tremendous advances, which aren't going to make a difference for most people, for most tasks. Like if you're like building a website, like for signaling theory, like you probably don't need, you know, the next versions of Grok to like build out the website. But if you're doing like complex tasks, like folding proteins or fuzzing smart contracts, like there are real gains that are going to be found and had, that are going to like move the needle. like my policy is everything in DeFi is getting hacked. Everything period. get your money out on the sidelines, at least for now, and see what's gonna happen, right? It's like when you see that there's some crazy macroeconomic event, like you take your money out of crypto just for a bit to sit in the sidelines and just wait it out and stay safe and see what happens. Maybe it's okay, and you miss out on a few months of yield farming or something, but you could lose everything right now. I seriously believe that the security threat of Rex Kirshner (41:21.518) You Gerrit Hall (41:45.863) what hackers can do with AI just means sit in the sidelines for crypto for just a little bit and stay safe and wait till the dust settles. Like wait six months, wait a year, I don't know how long, but like just see what happens and if it starts to look good, if it starts to look like security gets improved and it's safe, then you can scale back in and buy like wade through the rubble and see what survives. Anyway, that's. Rex Kirshner (42:09.538) Yeah, man, I definitely understand where you're coming from. I the combination of just how brutal the last cycle was and how it's clear that there's LLM assisted hacking that's accelerating, like I totally get it. Definitely bleak though. Gerrit Hall (42:30.741) At least that's where, yeah, well, that's where like, mean, at the end of the day, have to be an optimist about these things, right? Like, because I don't think that this, like, yes, it's bleak, but I don't think it means it's the end of crypto period, right? I don't think it means it's the end of DeFi period. Like, I do think that these things can survive. It's just going to be like a very tough thing. And I also think that like, what I can do is I can try and build FirePan. Like if the enemies, the hackers have... advanced AI weapons. The solution to that is that people who want to launch in DeFi have to have advanced AI defense. You have to be like taking the same tools that they're using. You have to have continuous smart contract monitoring because the audit that you got six months ago is great. But in the past six months, how many new attack vectors have been exposed? Like either someone on your team is sitting there reading through all these attacks and like checking it, or you just have it plugged in automatically. Rex Kirshner (43:05.773) Yeah. Gerrit Hall (43:27.335) you say, fire pan, watch this repository. Every time new hacks like get in the fire pan, scouts are gonna be updating. We're gonna be updating all of our defenses against it. So you'll all of sudden get a warning that like, hey, you know, the same thing that happened to balancer could happen to you. You might want to like take a look. You might wanna get your money out. Rex Kirshner (43:47.767) Yeah, no, man. I think we used to talk about this earlier when we were first transitioning this show more out of crypto and more into AI tools. But there was this idea that people, used to hear from pretty seasoned, experienced developers that anything that came out of AI coding tools was total slop and every company is going to fire. like five junior devs and then realize that they need to hire like two senior devs to like fix all of the garbage that came out. And like, I don't know, man, maybe that was true like in an era before I was really plugged in, but I just, at this point, I think we're deluding ourselves if we like don't acknowledge that AI tools can do things better than most humans. And right, like maybe there is that guy worth a hundred million dollars that OpenAI is hiring. First all, that's going to be extremely specialized knowledge. what these tools is offer beyond professional level competence at anything that you ask them to do. I still think, Tarek made an interesting point where he's like, I'm afraid of, or maybe he was afraid of using AI tools to develop smart contracts, right? And like, I totally understand that, where that's coming from. But on the other hand, I think about all the developers I've met in my five years in crypto. do I think that even a single one of them is more competent than an LLM? Especially if you know how to kind of guide it along. It's hard for me to believe that. Gerrit Hall (45:31.561) I mean yes and no I think that for like a basic smart contract like if I say like hey like Let's like mint an NFT that sends royalties for this episode to everyone who was on it a smart contract like that that's probably something an LLM could do if you're talking about something that like Michael Egorov launches Lama land which has like this like or or curve USD which is like this like completely novel like way I think that that that there's still some extreme edge cases which where I think that like the... Rex Kirshner (46:02.242) You don't think you could, if you had that idea, you don't think you could talk an LLM through it? Gerrit Hall (46:08.265) Not without massive customer losses. Like I think it could get there. It can get there. It's just like, there is like, I do tend to think that like there is, there is at the top of like any field, like the top 0.1 % or top 0.01%. I don't know how many decimal points you need to get out to that becomes really tough to train an LLM on. Cause there's not enough training data, right? Like if you're talking about like, Rex Kirshner (46:12.514) I guess, you know, yeah. Rex Kirshner (46:32.91) Dude, but that's, that's like almost exactly my point, which is sure. Like those people exist. The hundred million dollar engineer exists. They do something very specific at the edge. Right. But like in general, we're not talking about those people. We're not dealing with those people. Like those people by their very definition are rare and like not who's developing most smart contracts. And so like, I'm not taking what I, I'm not saying what I take Michael Agarov or Opus. Gerrit Hall (46:44.979) Mm-hmm. Gerrit Hall (46:57.577) yeah. Rex Kirshner (47:02.296) What I'm saying is would I take the like median solidity dev or opus? Like, I don't know, man. I'm taking opus. Gerrit Hall (47:09.545) Oh yes, I'd actually agree on that front. Yes, yes. Like if you're saying like make me another Perpdex, like there's a million Perpdexes out there it can read through and like I think it could do pretty good. I would still say that you will want to like get the code built and then like same as like with smart contract development, like the development of the smart contract is the first 5 % of it. the extensive work that you do to secure it is the other 95%, but I'd also trust Opus to do that too. To the extent that it could be as good as the median smart contract. But remember, I also think the median smart contract is getting hacked. I do tend to believe that, I don't know, one of the things I was researching the other day is like, Rex Kirshner (47:44.716) Yeah. Gerrit Hall (48:04.701) What happened to copy protection from like Commodore 64 programs? Like 30 years ago, they had copy protection on software and it was good enough, right? Like on the Commodore 64, I could have a copy of like, I'm just, where in the world is Carmen San Diego? And they had like a sophisticated enough key that was basically uncrackable with the technology of 1990. Flash forward to 2025, I was looking, has that been hacked? And yes, of course, that's entirely cracked. Advances in computing, advances in all of this, all of technology, just means that's easy to crack. You can't tell me that 100 years from now, you've got quantum computing, you've got artificial super intelligence, and whatever comes next on the tech tree, you cannot tell me that they're not just gonna look over all the smart contracts and be like, oh yeah, I can spend four hours and crack that. Rex Kirshner (49:01.922) Yeah, no, dude, mean, of course, right? But I just don't know how relevant that is because like we could trivially crack the enigma code that the Germans use. Like I could literally do it on my phone right now, right? But who cares? Gerrit Hall (49:07.335) Right. Gerrit Hall (49:15.333) Right, right, right. And that's true. As long as it's uncrackable for 10 years, let's say, that's enough to build an empire off of. But we're not seeing that. The median time between when a DeFi developer launches a smart contract and when their DeFi protocol gets hacked nowadays is one year, two years maybe, and it's getting shorter. It's dropping to six months or three months, and that's not enough time. Rex Kirshner (49:43.48) For sure. No, man, I think at the end of the day, like I very much agree with you. And the point that I was trying to get to was I think that these models are better than most developers. Like really, I think they're better than at least 90, probably higher than that percent of developers. And so I just very much like get behind what you're trying to do with Fire Pan, which is to say, like, let's not. Gerrit Hall (49:57.042) Agreed. Gerrit Hall (50:03.805) Agreed. Rex Kirshner (50:11.49) be in our ivory tower, is like AI models develop slop and like there's no way we would let them get involved in something as important as smart contract auditing. Like I think that's exactly the wrong approach and the right approach is to do exactly what you're doing, which is leverage them. Gerrit Hall (50:27.665) I mean, to be honest, that is the approach of a lot of the smart contract auditing firms. think it'll change very quickly. I think they were going to realize that the kind of their model, like having top tier auditors is good and has a place. I also just think that like the AI technology is there. Why wouldn't you use it? Rex Kirshner (50:48.14) Yeah, yeah. All right, cool, man. Well, I know you got places to be, so I won't hold you that much longer, but Garret, congratulations on Fire Pan, and we'll talk soon. Gerrit Hall (50:58.097) I think you're certain to care.