Adam Jacob: Hello, welcome to AI Frontiers, where industry experts run down all the AI news for engineers. It's Monday, November second. Look at how confidently I said the date. so everybody grab a snack and let's get started. I'm Adam Jacob. I'm the CEO of Swamp Club. I'm joined this week by my Swamp co-founders, Paul Stack. Paul Stack: Hello, I'm here in sunny California. You can see it on my left shoulder and a weird reflection. Adam Jacob: Nick Steinmates, Nick: What's up? Adam Jacob: and longtime friend, operations legend, entrepreneur, head of engineering at Railway, Mark Imbriac. Mark Imbriaco: Hey guys, great to see you again. Adam Jacob: It's great to have you. Mark, we invoked Mark at toward the end of last week's podcast because Sean was talking about how much he wished that Mark had written the write up for a railway because Sean, including me, has just has read so many great incident reports that you have written. And so we were part of why part of why we thought of you kind of immediately was that like, man, just so much love for you and for and for how good you are at the things that you do. So we're pleased that you're here. and yeah, you are one of the greatest of all time. Mark Imbriaco: a dubious honor, right? To be around things that break a lot. Nick: Absolutely. Paul Stack: Jesus. Nick: They Adam Jacob: That's one way to look at it. Nick: just call you chaos agent? Adam Jacob: Another way to look at it is you keep putting yourself in high profile things that require real can real professionalism. And you deliver on those things. But or dubious honor, you choose. Okay, let's let's go ahead and kick it off. top of the list this week was Reddits are open source. Banning all all AI projects from even having a conversation. so the deal here was if it uses AI, if it's made by AI or intended for AI, then it is no longer welcome at our open source. The quote is: AI as it exists today is incongruent to this open source community. meanwhile, they also banned all posts not written in English, because you wouldn't be able to tell if the author had used AI in order to write English. And I once I once I once gave a talk toward the end of Oscon's existence about the war for the soul of open source. And this was when they were building, you know, when MongoDB had switched to being source available and everybody was saying that because of Amazon, we were never gonna have any more open source or whatever. I feel like we're just always we're always bemoaning the possible death of open source from one vector or another. But this one just gets straight to how I feel most about all of these topics, which is We made a mistake when we started subtracting the values of how we relate to people from open source. when we turned it into just a thing about how we collaborate instead of instead of being a thing that was rooted in some kind of value, like we saw in free software, what happened is you can write things with a straight face that say you can't be here if you used AI because you want to communicate in English and you don't speak the right language. And like I just don't even recognize what a community would be. that does that thing or says those sorts of things. in the name of like whatever, your fear of AI. Mark, how do you feel about our open source banning AI projects? Mark Imbriaco: I I guess they're all gonna have to switch from Linux, right? Because Linus is all over AI these days. So I they're just not gonna make it. it's it's the the dumbest of all possible takes in 2026 to stick your head in the sand and pretend like we're gonna go back to the way that we did before. And you know, it it's really frustrating because I was at the first three Pearl conferences, the third one became OSCON. Adam Jacob: Course. Mark Imbriaco: and the it was about building cool shit together. It was not about whatever the heck is going on in our open source now. I it just doesn't make sense. I don't know where the hacker ethos went, but they need to find it again in that community. Adam Jacob: Yeah. They definitely do 'cause we've lost it. Keeb, what do you think? Nick: Well, yeah, so I just looked up the stats 'cause I was interested. Reddit is global rank eight in terms of visitors and rank six if you're just thinking about the United States. so it's like still one of the most popular websites in the world today. I've been on Reddit since two thousand five, you know. so it's a I've had a long storied history with that website. And yeah, I think they've they've jumped the shark just like most of the other sort of community social media projects that are that are out there. Adam Jacob: mean, there was an it's not Reddit's fault, it's the mods of our open source, right? It's not like Reddit banned AI. It's Nick: I it's i y yes, but also it turns out if you do the analysis, like, you know, if you're that inside of Reddit there is a very strong power structure of mods, and it turns out that it's quite incestuous. And so, you know, it's a certain worldview that gets highly proliferated among the moderators Adam Jacob: yeah. Nick: that is, you know, on that website. And so anyway, I agree with Mark, the those people are dead in the water. I think it's the complete wrong frame and wrong way of thinking. And also I'm not surprised. you know, I don't think people go to Reddit for the comments section anymore. I think you can also, you know, tall the tall the haters or lovers of Red of Reddit out there. There's obviously boutique communities inside of Reddit where you can find your people and all that's glorious. And like big subreddits are bad and all the rest. But like, yeah, I just I think this is This is just people virtue signaling for the sake of it. Adam Jacob: Okay. Paul Stack: I don't I don't like Reddit. I don't like Hacker News. I don't really care too much about those communities. There you go. I I t I take that back. Adam Jacob: this is the reason the podcast exists. Paul Stack: I care about the communities. That's not that's I'm being I'm being weird about that. But like I I I kinda feel that like a bunch of those like sites are a little toxic and I just like try and stay away from it as much as possible. So I d I I Adam Jacob: Yeah. Just Paul Stack: don't think I even have a Reddit account. That's how weird that is. So Adam Jacob: Yeah, this is why Keeb and I read the news for you, right? Because and and why the podcast exists in general is because we're like, Paul Stack: Exactly. That's why it's my favorite podcast night. Yeah. Yeah, exactly. Adam Jacob: hey, it's our job to go read our open source so that you don't have to. Paul doesn't have to go on Reddit. Paul Stack: Paul doesn't need to go on Reddit and deal with like yeah. Adam Jacob: Yeah, exactly. All right, well, let's go, Nick: Busy shipping. Adam Jacob: let's go business. Let's spend some time in business land. So Apollo, let me get the right name of it. Apollo Global Management posted interesting paper on the macroeconomic. Friend essentially saying, despite the huge investment and widespread hype around AI, it has delivered zero financial payoff to corporate America outside of tech. Basically, revenue per worker is rising for the Magnificent Seven. And for the entire rest of the S P, so the S P 493, which I'd never seen put that way, but I think is a delightful way to distinguish the Magnificent Seven from everybody else, is that is that they're flat, no signs. Of AI boosting productivity growth. and I think what's interesting to me about that is that what what I'm seeing certainly in our work is that when you go out into the enterprise, yeah, there's a lot of AI investment that's happened. It's struggling to return a positive ROI. We're just now starting to build systems like Swamp, like other things, that are starting to see how you could start to deliver positive ROI with this technology. and so one way to read that article, I think, is it's all doom. And I think the other way to read it is that, right, well, that's because it starts with technology people who figure out how exactly is it that we're going to use this technology and then how are we going to use that to create positive ROI? And in the technology companies, it doesn't start as a positive ROI conversation in the first place. It starts as a what can we build and what's possible and what's the frontier and those sorts of things, which then become markets you discover, which then returns ROI. Paul, what do you think? no AI growth at all in the magnificent four hundred and ninety three. Paul Stack: I I can't see how there's no growth at all. sure that the big seven are are like a big thing, but like there are definitely like sectors within organizations that are helping drive those organizations forward. You know, maybe there's no overall positive growth, but like it's not that these companies have their head buried in the sand and they're not gaining anything from the technology that's actually out there at this moment. Then I also look at it in a different lens. It's like there's a bunch of money to be made. dealing with those companies. This falls into why Gartner is saying that what, seventy percent of all AI projects or AI incentives are going to be killed by this time next year because of the fact that, if there isn't that positive growth and there isn't that positive ROI that's there. But it just shows that it's those ecosystems, those companies are ripe for restructuring and how you can actually be, correctly built and revisit how they actually build their processes on top of AI rather than trying to retrofit into a current setup and a current structure. Adam Jacob: We think he Nick: I was watching it's a little tangential, but I was watching a speech given by Elon Musk, like twelve days before the first shutdown before the p the first bankruptcy threat of Tesla. And the case that he made to everyone in the audience who was listening as he was trying to drum up some dollars was like the reason why the first road Tesla wo roadsters are a hundred and eight thousand dollars. And not the mass market $30,000 Tesla that you see driving around the roads all all day, now. is precisely the because the technology is incredibly expensive to get started before you get production lines optimized and all of that. And I think that's exactly what you're saying in the article. It's basically saying VC backed companies are spending ninety percent of the dollars or whatever. It's like it's all just like, you know, wishy-washy in terms of of who's spending the money and who's getting the value. I think we all believe as people who will represent startups sitting here, like that's the right, you know, that's the leading indicator of the rest of the market. We're building the future, Adam Jacob: Totally. Nick: you know, and what we're building the future for is technologists right now. And you know, have like Harvey or whatever who's trying to do it for law. You have breakthroughs that are happening in mathematics as a right as a result of the large firms investing in them. But the tooling, the UX, the whole process around it, is not optimized for that yet. And I think, there's one lens to look at this, which is like you should be bearish on AI. And I would flip it to be the exact opposite. You should be completely bullish. The fact that they've made two trillion dollar companies by solving a slice of the world's problems and all you can see is unlimited upside on the other ones if they can get the UX side of it right. Is that is this not just an advertisement for the anthropic IPO in disguise? and so anyway. Paul Stack: you know took my s you you're taking my conspiracy theories away. Come on. Adam Jacob: Mark, what do you think? Mark Imbriaco: Yeah, I agree with with Paul and Nick. I think this is this is where you you trot out my favorite William Gibson quote, the future is here, it's just not evenly distributed. very, very much the case here. I think obviously the tech companies, the startups of the world making maximal use out of AI, being very productive with it. I know all of us would say that there's a dramatic increase in productivity in our businesses as a result of AI. And I think if you look through the micro lens at these big organizations, you'll find the same thing. You just don't find it evenly distributed across them. You find pockets where Adam Jacob: Yeah. Mark Imbriaco: there's the people who are able to deal with the lack of affordances and the the rough tooling that exists for anything if you're not a software developer and still make value out of it. And until we get to that we sort of get over that UX hump where that just becomes hard to hold wrong. right now, we were we were in this place last year. If you think back almost exactly this time last year, Opus four point five landed and everything changed. We went Adam Jacob: Totally. Mark Imbriaco: from a world where it was hard to hold the tools correctly and get maximum value out of it to a place where it was kind of hard to hold them wrong. It was hard not to get value out of it. Didn't it didn't matter like how badly you use it, you're still getting value out of it. and we're not quite in that place for for people who aren't, you know, nerds like us. but we're getting close and tools things like Muse and Grokbot and all of these things I think are the sort of leading edge of what's possible. Adam Jacob: Yeah, for sure. And I think that's the way you would look at that profit margin chart. And what you would see in that profit margin chart that stayed flat is that there's massive opportunity once we unlock how to build that tech how to bring that technology to those people, that there's just a humongous opportunity that remains in actually unlocking those margins. Mark Imbriaco: Skill issue. Adam Jacob: yeah. I mean, basically Mark Imbriaco: Mm-hmm. Adam Jacob: it's a skill issue. All right, let's talk about Anthropics S one because how can you not? So I think Nick: There's no self plug there really quick to say and we're building that platform and you should buy it. But anyway. Paul Stack: I was gonna say swamp exists, absolutely. Yeah. Adam Jacob: Well look, you know, that's why y'all are you know, that's why y'all that's why we're all here. okay, so for a lot of you you perhaps don't read S ons for fun. but I ca I do. I know I know I bet Mark probably does to some degree. Certainly Keeb does. I don't know if Paul has has ever like read an S one for joy. I will say that the anthropic S one is absolutely nuts. And so, you know, sometimes the reason that people read S1s is because they're trying to learn about other people's businesses so that they can better understand how to build their own. I think it's impossible to look at this one and learn how to build anything because it's so out of shape. It's so wild. So let me just walk you through some of the numbers. So top line revenue for 2023 at Anthropic was $41 million. top line revenue 2025 was $4.5 billion. Top line revenue projected for the six months ending June thirtieth, twenty twenty six, sixteen billion. so you know, cr it that's insane. So a three year a three year run from 41 million to sixteen point two billion, that's just like, what are we even doing here? You know, like that's a that's that's a run like essentially no one has ever done. people have long been saying that what's happening here is the tokens are being subsidized, they'll never be able to turn a profit. you know, what you see here when you actually dig into the operating loss is that the vast majority of the loss, so they took a $72 billion operating loss. That was mostly sorry, they took an 80 an eight an $80 billion operating loss. $72 billion of that was basically markups in convertible notes and derivative liabilities. which means it's not actually cash. It's like it's like stuff that we have to pay out, but it's mostly paper money that's just gonna convert. the side effect of that is that the the actual loss was about eight billion dollars. and if you keep going, and you go all the way down to their consolidated balance sheet, you know, they've got 130 billion in cash assets right now. They've got 130 billion in working capital. They have 140 billion in assets against 117 billion in liabilities. Then you go into the like non-gap financial measures. We won't have to go into definitions there, but like, you know, again, what you start to look at is their operating margin. The operating margin for the six months ending June 30th, 2026, negative 4%. So Everybody who said that what was happening here was that they were they were subsidizing tokens, you were throwing money in the fire, they were never gonna turn a profit. They're all wrong. they've been wrong the whole time. you know, the argument around sort of, hey, but what's gonna happen to their looming data center commitments and all these balloons that they have to do, all that stuff's real. I think if you read the S one, what becomes obvious is those are all contracts that are beneficial to everyone involved. And if what happens is they can't convert. or buy at those numbers, what they're gonna do is change the deal on both sides because it's good for everybody. And so it's not like they're gonna be stuck holding the bag. and so yeah, I think essentially Anthropic is a monstrous company, the likes of which I don't know, I've certainly I've never seen. Paul, what do you think? Paul Stack: Wait, I just read, just on Friday, definitely not today, I just read that their credit rating's gonna be marked as high junk because of their spending commitments. so there's Adam Jacob: Sure. Paul Stack: which is which is kinda and that's like gonna add fuel to the fire of people saying, they're like as soon as the IPO they're gonna absolutely tank in value and it's just all smoke and mirrors and a ruse, right? Like Adam Jacob: Yeah. Paul Stack: that's there's there's just a lot of weirdness. I d like the numbers are astonishing. They're absolutely astonishing. Adam Jacob: Look, there's a at at that number there look, there is no more credit to extend to Anthropic. What what credit would you extend? They have yeah, they've they've Yeah, not only have they taken all the credit, they've invented Paul Stack: I'll I'll give him a couple of hundred dollars. I don't mind. yeah. They literally have all the money. Yeah. Yeah. Yeah. Mark Imbriaco: They have it all. Yeah. Adam Jacob: NVIDIA invented new credit vehicles in order to extend yet more credit into the Z go. So of course the credit is junk. anyway, of course it's junk. Paul Stack: Yeah. But it it's like for me it it's it's astounding numbers. It's like absolutely fascinating numbers. Like when especially when you like read about their committed spend, like through the next X amount of years and like how much they're actually growing. Like a fun statistic is supposedly that the word billion, Is in that S1 240 times, just because of the fact like w Adam Jacob: Yeah. Nick: Mm-hmm. Paul Stack: when you when you actually are a company who are dealing in that level, like and that's just like how you actually refer to the numbers, like fair play to them. Honestly. Like seriously. Like if you can build a like a a business of this size at this like level and this quick, I mean, there's something like it's it's what everybody aspires to, right? Like When companies started being billion dollar companies, that's what people aspire to. Now we're aspiring to trillion dollar companies. it's like Jesus, it's like it's unbelievable. It really is. Adam Jacob: But here's a trick for entrepreneurs. Do not say that you're the next anthropic. Paul Stack: Yeah. What did we s what did we say that the word Adam Jacob: Ain't nobody the next anthropic. Paul Stack: was? Like it was a it was a unicorn when it's a billion. What did we what did we come up with and say it was Adam Jacob: I don't know. They call them Decakorns, but I don't even know what this thing is. Is it a Pegasus? Paul Stack: Deck accourts, Jesus. Adam Jacob: Is it a is it a is it an archangel? What's happening, Nick? Nick: I I think I think part of the People are trying to still quantify or explain the anomaly that is anthropic in open AI, and there's just no words for it. And so they just go to finding an angle of attack that somehow brings some normalcy to this majestic, fantastic, amazing, you know, finally a company that lives up to all of those words in terms of aspiration and impact or whatever. And it's just nothing short of magical what's happened. And so good for them. And also, yeah. You know, is Adam Jacob: Well. Nick: this the sea change of of trillion dollar companies and they're just at the the front end of it and twenty years from now we'll be all be looking at it back and everyone's a trillion dollar company or whatever, the maximum wealth theory. Or is this just a complete anomaly? Is it Paul Stack: Hey, I'm here for it. We're running a company. I'm fucking here for it. If this is the the start of trillion dollar companies, I'm just saying. Yeah. Adam Jacob: Maximum wealth theory. What do you think, Mark? Mark Imbriaco: I I think everybody's overthinking it, right? at the end of the day, it's to me very simple. There is unlimited demand for intelligence. Unlimited Adam Jacob: Yes. Mark Imbriaco: demand. When the demand is infinite, like no one gives a shit about power law distributions or anything else. Like none of it matters because the the like it's infinite. The the space, the total addressable market, we can't quantify because it's everything. Adam Jacob: And it's going to expand. Mark Imbriaco: it's only going to expand. so when the addressable market is is infinite and you've got basically two players who are consuming the vast majority of it who are so far ahead of everyone else, it's hard to imagine a valuation that doesn't end in trillion. nothing else makes sense. The demand is only going to keep going hyperbolic and the the yeah it's just it's generational. I think kind of crazy that two of them popped up at the same time. Adam Jacob: Totally. Mark Imbriaco: of course there's a whole lot of incestuous movement Adam Jacob: Yes. Mark Imbriaco: between the two. So it's not super surprising, but crazy that we get to live in this time where everything is changing. yeah. So Adam Jacob: Yeah. Mark Imbriaco: I I I think it's not that crazy. I think the there's a lot of cope from people who who want Adam Jacob: For sure. Mark Imbriaco: it to be not real. there's a lot of yeah. Adam Jacob: Absolute silence though. on X, man, the day that S1 dropped, all those people, radio, nothing, nothing, not Mark Imbriaco: I think the Adam Jacob: a word. Mark Imbriaco: the only thing that can stop them though is physics. can they physically get enough compute? Can they can Adam Jacob: No. Mark Imbriaco: we get enough power? Can we like can we expand the the substrate that they need in order to be able to continue to grow? There's a real question there, that I don't think anyone has a satisfactory answer to. Adam Jacob: No, for sure. And but I think the worst that case that happens there is the rest of us slow down. and they continue to print money, like a magic money printer. And you know what I mean? Like even the downside there doesn't feel like it's actual downside for them. 'Cause yeah. Paul Stack: It's funny, s Sam Altman said today, he on Friday, Sam Altman said, Hey, maybe people shouldn't have RAM in their personal computers anymore and it should be all available in data centers You know, like Adam Jacob: Yeah. Mark Imbriaco: Yeah, that's a great way to make friends. Paul Stack: Yeah, right? I was like, You're not taking my RAM from me? Like, no way. Adam Jacob: Yeah. And I think, Mark, an inch you work for Railway and Railways had an explosive growth in the last several years, you know. And like, you know, a company that grew at the rate railway has grown ten years ago, twenty years ago, would have been absolutely unheard of. You know, were also at Heroku. Railway's grown way faster than Heroku. and and so like there is something that's shifted, I think. in in the market and in the way we think about growth and in the way in the way we think about companies and I think it like it's it's changed, you know? Mark Imbriaco: I think you have to factor AI into your worldview about everything if you wanna understand anything about what's happening in the world. yeah. Unlimited demand and unlimited creation. It's pretty awesome. Adam Jacob: It's pretty great. yeah, if you're an engineer, there's never been a better time. stressful, but awesome. okay, so Caitlin Lessie. Caitlin, I'm so sorry if I got your name wrong. There's no that's the thing I'm learning as I do this podcast is that there's not like a magic machine that tells me exactly how to pronounce people's names correctly. but they wrote a blog post called New Problems for Engineers to Solve. I thought this was a good breakdown. Of the problems created by the multi-agent architecture. So if you think about, we've talked about this in Swamp Club land in terms of like you should be building AI systems out of repeatable behaviors, in part because of the problems that I think we see in both ROI, but also just in engineering behind the multi-agent architecture that sort of emerged out of this wave, this first wave of of how we build AI systems. They point out infrastructure problems, identity problems, security issues, monitoring problems. All of this is caused by the shape of that architecture and that there's not yet really great answers to each of them. they frame it up as an opportunity. I think it is an opportunity, but it's mostly, I think, an opportunity to just skip the multi-agent architecture altogether because it's a silly one. Mark, what what do you think about multi-agent architecture and the n all the new problems it crops up for the world? Mark Imbriaco: I think that the the point that infrastructure needs to change for an agentic world is completely accurate. It's interesting though. I think all of these conversations are talking about the changes that need to happen. It's the this is wild. I I sort of am just thinking about it this way, but all of the changes are on the development side, not on the production side. There's way fewer things that need to change once you've got the thing built and you want to get it deployed compared Adam Jacob: Yes. Mark Imbriaco: to sort of the development loop. There's a bit of an asterisk there. We have a take on this at Railway that production is part of the development process these days. But Adam Jacob: It is. Mark Imbriaco: taking that aside, I think the the thing that's missing is primitives that are that that give you the things that you need in order to be able to iterate quickly. time travel is a thing that you want. you want to be able to to fork and snapshot and do that on every layer all the way through the stack as many times as you want. And There's a bunch of these primitives that existed for a long time, but they're not exposed. They're not they're not built for the scale that we need to work with agents these days, right? it's like, great, my file system supports snapshots. Okay, great. What if I have 20,000 of them? What if I have a million of them? is it how's Adam Jacob: Yeah. Yeah. Mark Imbriaco: that gonna work? because I could tomorrow have twenty thousand of them and a week I could have a million. so I think all the all the primitives just weren't built for the scale. And we see it with GitHub, for example. GitHub had to basically refactor their entire infrastructure. in order to be able to deal with the explosion of growth. Adam Jacob: Yeah. Mark Imbriaco: it's insane. Adam Jacob: And still unclear whether I want it. one I I think the other piece of the new primitives, obviously we agree because we're in the we're in the platform of primitives business, but I think, in general, it's so clear you need a bunch more new primitives that just need to be way weirder than the current crop. Keb, what do you think? Nick: Yeah, I don't know. Yeah, I don't know. Adam Jacob: Yeah, yeah. Nick: I think Mark is right. That's it's crazy. that you just have to rethink everything from ground principles. And also, it's I was just talking to a candidate interested in applying who isn't s isn't quite so technical, but is thinking about like how do they maximize their success inside of this new world. And like there's a very clear difference. There's a very clear difference emerging about people who are building systems ready for AI to attack, you know, ready for AI users, and then the users themselves bringing themselves to the problem. And what happens is like I think usage goes off across the board as a result of that, for everyone, and it gets really interesting. And so anyway, I'm like deep into thought about the implications of that, both from a system builder standpoint and How do you scale to meet that demand? And also just a real understanding that yeah, demand doesn't depress, it only increases from here as we go to the moon. and what that means is, you know, is interesting. So anyway. Adam Jacob: Yeah, for sure. I think let's move on. I think Google continues the trend of announcing models before you can play with them because they're just too dangerous. Gemini for Carbon is the latest of those, of that list. you know, whatever. The numbers look promising. Paul Stack: Yeah. Adam Jacob: and and how I feel about it is that no matter what, if you have access to the compute, I feel like you can build competitive frontier models. And at relative speed, because, Google appears to. I think it's weird that they continue to to basically be third place with a bullet. I feel like they're third place with a bullet in cloud too. where it was like, who do I love the most? And you're like, well, AWS does best. And then you're like, well, Microsoft's pretty good. And but I don't nobody likes the interface, but nobody ever got fired or whatever. And then Google's down below being like, but we're the smartest engineers. And I feel that way about all of these model releases from Google where it's like, but we're super smart too, and you can have it soon because it's also very dangerous to use our models without, proper Nick: Ha ha. Adam Jacob: security clearance. yeah. Anyone got thoughts on Gemini four carbon? Paul Stack: my thoughts on new model release weeks. We said last week's podcast there was none. We got through a whole week without no model releases. This week Adam Jacob: The whole week with no new model releases. Paul Stack: we have a model release. Or we may not have a model release. It's Schrodinger's model at this moment in time, right? Because nobody actually knows if it's a real release or not. But fair play to them because they're they're trying to keep up in the frontier. They're trying to show Adam Jacob: Yeah. Paul Stack: that they're a serious power as a a frontier model offering, but it's it's really difficult to take their model releases as serious as anthropics and open AI when when you can't actually get your hands on them easily in a like in a a readily available way that's like it allows you to be able to test them. And I know people who have been using some of the earlier versions of Gemini and they're really happy with them. And I've used Gemini with anti-gravity, two two Gemini two one pro and it was it was great. And I'd love to be able to try the new ones, but they've just taken it completely away from the ability to use it. Mark Imbriaco: I don't know, man. I think the maybe unpopular opinion, but they might be right that it's too dangerous. I think that I think there's a degree of truth in like Adam Jacob: Yeah, for sure. Mark Imbriaco: look at what's happening in the world, things we're gonna get to later in the news. I think the some of these models are pretty scary. So maybe, maybe it's true. I I keep expecting it's funny, like of course I want to dunk on Google just like everybody else does, because they are they should be so f much further ahead than they are. they have all of the advantages. They have infinite compute. They have they had the best brains in the world forever and ever. They have all of the data. why Adam Jacob: All the money. Mark Imbriaco: the hell they're not winning is a mystery to everybody, unless you've ever looked at how they go to market, then it's not so much a mystery anymore. Nick: Well or the sheer arrogance of of Paul Stack: Burn off. Nick: their belief in themselves, you know? So Mark Imbriaco: Yeah, Adam Jacob: Yeah. And I feel like I feel like there's a piece of that that's really true and that we should all love them more than we do. and there's people who really love we got this we got this email from somebody who tried Swamp and then just lit us up about how dumb we were cause it wasn't Kubernetes and the and we should do stuff that was smart like Google. And it's because that person just loves him some Google, he loves the engineering of it, he loves the shape of it. He doesn't work for Google. He just loves the idea of Google. And I think the four of us hanging out here, we're all like, I don't know. We've all they've been mean to all of us on some level, you know? very smart, very smart. Brilliant. People I love. Absolutely. Mark Imbriaco: They're super smart though. I think we can't take that away, right? And we all know some brilliant people at Google and have nothing but respect for the technical ability, which is why it's mystifying Paul Stack: Holy. Mark Imbriaco: that they're in the position they are in the market and Adam Jacob: Yeah. Mark Imbriaco: this space at this time. Adam Jacob: Yes. Yes. I'm more mystified Nick: Well what's Adam Jacob: than I am. I'm not upset. And I d I certainly don't think they're dumb. I'm just like what wh where? Nick: It's so f yeah, it's like the frontier is so far ahead is so far ahead, we just said, that if you're going to second place intelligence, then you just wanna go hit up the Chinese homies or open router or you know, or host and hope for open source. And so like I think what we're finding is that this mediocre middle sort of thing can exist. at least in in the same angle of attack that open AI and Anthropic are winning in. Adam Jacob: Mm. Nick: And so yeah, always trying to play catch up must suck for them. especially Adam Jacob: Gotta be, right? Nick: especially since the whole reason why open AI got funded to begin with was because no one could compete against Google. And as it as it turns out, they absolutely can. because it's not AI like because it's not ad. Spending at the end of the day, it's not advertising at the end of the day. Is that where all of the brain trust inside of Google is actually? Anyway. Mark Imbriaco: I it's interesting, Adam Jacob: Yeah. Mark Imbriaco: right? Maybe this is connected to that report about, big cos not getting value. I think we all say you can't sit in the middle and be successful here. And all the big companies are like, yeah, what about copilot? We like copilot. And and Nick: Yeah, yeah. Adam Jacob: Yeah. Yeah. Mark Imbriaco: we're all like, give me the most expensive, smartest tokens I can get my hands on. Screw the cost. I don't care. Give me, give me the good stuff. and like I think I think there's something to be said for that. Adam Jacob: That it totally there is. And Copilot's a perfect example. I feel much less of a way about Microsoft's AI strategy than I do Google's. And partly it's because I have lower expectations, because I'm a Linux nerd or whatever. But Microsoft has proven over and over and over again that they know how to print money. So you should never discount that that they're gonna do that. But two, it's because it's because that strategy, they're willing to let it lag a little and and be the people who bring it to you in a consumer-y fashion. And but that's not Google's strategy. Google's strategies were trying to lead the frontier. Meta has taken a similar strategy, but I think, come out of the gates a little stronger in their offerings. all right, let's turn our sights to Europe. So Mistral has set out Mistral Large for open source, LaCronc comes to open weights. This is the strongest open weights model that exists right now. competitive with the frontier models. The release continues to be beat the drum of sovereign AI. So this is this is a large model that competes with the Frontier, that was trained in Europe, that that is is built and brought to you by Europe, you know, partly explicitly to create that sovereignty boat. And so I think it's I think it's interesting to see, you know, right now. A lot of that conversation is America and China. I think I think we are starting to see Europe come up in the model race and starting to show up in these conversations. And strategically they're beating that drum around sovereignty, around open weights. Paul, resident European. What do you think? I know. This you have to go to Paul, because it's like, okay, as somebody who has an eur an EU Nick: I like how anytime we talk about sovereignty it's Paul. Paul Stack: No. Mark Imbriaco: Yeah. Adam Jacob: passport, please to please to Nick: So good. Adam Jacob: weigh in. Paul Stack: their score their benchmark scores are pretty good. Right. They beat Adam Jacob: They're great. Paul Stack: they beat Deep Seek Flash, they beat GLM. they it looks like it's better on high mode, of course. Like it's just the way it works. Like, you know, I I'd love I I'm actually gonna try and run this myself, like to figure out what it is I actually like like about it. Like they say it's like even their tagline, which is build for coders, you know what I mean, trusted by defenders. I mean, that's a ballsy like tagline in this day and time in in AI at this moment, right? With what's what's been happening and what's going on. But it's it it's pretty fascinating. Like it continues to be fascinating that the open models are are are pushing like the frontier models in the way that they are. And I think it's it it's really good for us as consumers because it's hell it's gonna really help drive down tokens. Like we had a we had a a conversation at lunchtime today. It's like what happens if you know the cost of tokens goes way up. It's like well we try and get our hands on hardware and we'd run different models and we'd figure it out. That's just the way it works. So the competition sorta puts there you go. I actually said it would be Keep West O one. I actually there you Nick: Mm-hmm. Adam Jacob: Yeah, it's right there. Paul Stack: go. And it it's very much a case of that consumerism, for us, is fantastic. It gives us that that ability to cho for choice. And if all else fails we can go to Google. So that's what happens. Adam Jacob: Are you running open weight models at all at Railway? Mark Imbriaco: No, I think this is funny. When I think about well well, it's like this the meme, the I don't think about you at all meme is how I feel about Mistral. I I just Adam Jacob: Ha ha Mark Imbriaco: like and un they they need to show more than one release that actually does something. And I you know, I think the the macro environment in Europe makes me worry about their ability to compete. I think can they get enough power? Can they get enough data center space? Can they like get air conditioning? can they and and Paul Stack: We don't need air conditioned in Europe. we've got it such a perfect Mark Imbriaco: Here's the other thing, right? Paul Stack: temperate climate. So nah, it's fine. Adam Jacob: Mean the data center needs it though, right? Mark Imbriaco: They've they've got brilliant people, but they are not the Frontier Labs and they're simultaneously not shameless enough to distill the Frontier. so I I don't know, I it puts them in a bit of a tough spot. Adam Jacob: Right. Too too ethical and too capable. Mark Imbriaco: I mean, let's be honest. That's where all the the like our friends in China, these models didn't come out of nowhere. Adam Jacob: Yeah, yeah. No. Nick: Yeah. Well. Paul Stack: But there will be companies that use them in Europe. that's just Adam Jacob: For sure. Paul Stack: what happens just 'cause of the fact that they there there's a real B in the bonnet about not transmitting data to the US and letting Nick: Imagine Paul Stack: it Nick: imagine trying to build a data center in in Europe. We have a hard enough time in America, anyway, that's gonna be tough. Mark Imbriaco: I hope they succeed, right? Because I think we do need more people at the frontier in this space. I think a world where there's only two powers is not great. so the more the merrier. I wish them success, but I I think the jury's still out. We need to see more. Adam Jacob: For sure. I think all right, let's get weird. one of the fun stories this week was Quen Pooter. So this was basically they use a gated DeltaNet attention layer. I say that as if I know what's going on, as RAM. And then they ported a working 32 bit RISC-V CPU inside of Quen35. and then gave it a fake file system through the same mechanism, and now you can run code. directly inside the LLM without a tool call as part of inference. and it actually runs code. Like you can compile your C code down to to this CPU and run it within the model itself. I just think this is super neat. It's absolutely cursed. You know, like I have I have no idea like why I want it. But but but it's just one of those things where you're like, yeah, sure. Why not? You can do everything that you want to do. it's just numbers in the end. Mark, what do you think? Mark Imbriaco: I love everything about it. for exactly, it's like it's just the right kind of weird to Adam Jacob: Yes. Mark Imbriaco: like to get me excited because it's it's just to your point, it's so weird and cursed. It's like making computers in Minecraft. it's Adam Jacob: Mm-hmm. Mark Imbriaco: got that same vibe and I love everything about it. I no idea Adam Jacob: Yeah, me too. Mark Imbriaco: if it's practical, who cares? Doesn't matter. Like this is this is why it's so great. This time is awesome because you can build shit that was like super interesting but was would have taken way too much effort before in the before times to make it. Now you can just make weird things. And it doesn't take you 20 years toiling away, with your little model train set in the basement. you can just whip out some really crazy shit on a whim. Adam Jacob: Yes. Yeah, absolutely. Nick: Yeah, compiling to LLM is insane. but I used to think that, you know, compiling to browser, aka WASM was also insane. And so, Adam Jacob: Right. Nick: you know, yeah, and my mind immediately went, Mark, to the same exact place. As soon as you added Logic Gates to Minecraft, then what happened is people built computers inside of Minecraft that can play Doom or whatever. It's like like insane that that happens and fun and whimsy. Yeah, whimsy, whimsy, whimsy, whimsy maxing. Insane. Sorry, Paul. Paul Stack: I understood all of the words that you said, but I don't know what order they mean and what I think, 'cause I just let that one go right over my head. That sounds just batshit crazy. It really does. Adam Jacob: Yeah, I I I read the article and they did it, they did describe what a gated delta net attention layer is and why one could use it as RAM. and I feel the same way about it. I feel like when I read books Mike like that Mark Burgess wrote, where it's like, I understood the explanation, but then he writes the formula out as like a physics professor. And I'm just like, yeah, no. You know? Like I just gotta take your word for it on that one, bro. I believe you that the math is the math and you made it math and now it works. But like it was just it immediately was beyond my capacity. not to understand it in theory, but in practice, I'm like, I don't okay. Sure. You know? it's a it had a little bit of a Star Trek sort of hoobledingy vibe. I think another curse story that has nothing to do with AI, but we should talk about anyway. The kerfuffle around MacBook Pros with a touch screen. Mark Imbriaco: Hmm. Adam Jacob: Just everybody freaking out about it for nothing. I am running a Dell XPS that is great. I think Mark owns one of those too, but I'm not sure it's what he's using. It has a touch screen. I don't really touch it. and you know what happened? Because I don't touch the screen, nothing. it just means that it works like a normal screen. So I'm not sure why anybody was mad that now you can touch the screen, even if you don't want to touch the screen, because it's just a screen and you could just use it like one. Paul, I think you hated this. Paul Stack: I hate it. I absolutely, what an absolute I between that and the Jew, Jesus Christ, like, you know, what is going on there at this moment in time? Although people tell me they love the Jew. Friends of mine have it. They're like, you should get a Jew. I'm like, absolutely, I should not get a Jew. Mark Imbriaco: I can't wait until we get to what I'm plugging this week. I can't wait. Paul Stack: no. Like what we need, right, is we need people sitting around their laptops, Adam Jacob: God, I hope it's not too well. Paul Stack: eating their lunch and touching their fucking screens, right? That is a hundred percent what we need in this day and age. Cause Adam Jacob: why'd you go straight to the yucky though? Paul Stack: that's what's gonna happen. Like, I mean, I don't touch my laptop screen right now and it's already it feels disgusting. And like when you can touch it, it's gonna feel worse. It's like a it's a germ bomb waiting to happen. It's terrible. Terrible. Adam Jacob: Yeah. I don't like that. That's awful. Paul Stack: Yeah. Terrible. Nick: iPad sales must be bad. you know, like trying to turn your Paul Stack: Ha ha ha. Nick: laptop into an iPad is kind of insane. And and and also I agree with you, Adam, like if you don't want to fucking touch it, don't touch it, you know? Adam Jacob: It's like a very easy, straightforward fix. Why we had to go on X to talk about it, I'm unclear. Mark Imbriaco: Mm-hmm. Paul Stack: But it's there. It's like the touch bar, right? I didn't want to touch the touch bar. I'm gonna touch the touch bar. This is the same thing. Adam Jacob: Okay, but the touch bar in to defend the touch bar, at least the touch bar was interesting. The touch bar was it was. It it it turned out it turns out no one wanted it and it was worse Paul Stack: Was it? you could put lights on it, right? Yeah. Adam Jacob: at all. I nobody used it, including me. But when you first saw it, you were like, That's cool. And then you moved the slider for volume and you were like, That was awesome. And then that you never touched it again. And, you know, so it failed. But this is just like this isn't even that. This is just like Paul Stack: But but you But they've but they've they've gone through the cycle, right? They took away all of the ports and they're like, You need nothing, no ports, you need one USB C port, and then you Mark Imbriaco: Bavary. Paul Stack: buy one of our special adapters, which is four and a half million dollars. Adam Jacob: Yes. The dongle apocalypse. Paul Stack: Yeah, right. And then they brought it all back. You even have an Ec D card port now in the new MacBook Pros for all of the X D cards I still have, right? You know what I mean? It's like, yeah, they're I don't know. I don't know what's going on there. I really don't. It's very strange. Adam Jacob: Well, I can tell you what's going on at lunchtime. Yeah, they're gonna make plenty of money and we're all gonna buy them and people are gonna be like, touchscreens are so back. Paul Stack: Well they're making billions, so there you go. So yeah. Yeah. Adam Jacob: DoorDash has bought Instacart. This is only AI news because you have to eat while the robots work, I think. But but yeah, massive consolidation Paul Stack: Touch your screen. Adam Jacob: in the in the North American food delivery market. fun fact, we once had a business plan to restart Cosmo.com. the OG one. This is one of the many businesses that failed to start that I tried to start. that was a good idea but died. rest in peace. That business plan. but I just thought this one was interesting because the sort of seeing how hard it is to bring companies up, to make them go public in this era, to get them into the market, like the two large players in that market consolidating into one, you know, I think bodes actually bodes well, for sort of the market in general. yeah. And now your groceries and snacks will all arrive from one source. Anyone have anything they wanna say about DoorDash acquiring Instacart? Anybody feel away about it other than me? I was kind of assuming that I was gonna put this story out there and it was gonna like, Yep, Doordash bought Instacart and I was gonna be like, Yep. Nick: I I mean Mark Imbriaco: Much it. Paul Stack: What's gonna happen on the YouTube is there's gonna be tumbleweed goes across the screen while we wait there. So Adam Jacob: Yeah, yeah, yeah. Nick: What's I mean what's funny, what's like fascinating is that you know Instacart's delivery service was Uber Eats before this. and you know there's actually not a lot of overlap in terms of restaurants and grocery with respect to either two companies. And so back Adam Jacob: It's a good it's a good blend actually. Nick: back to Venn diagrams that we talk about in terms of partnerships or whatever, like you know It seems like a good consolidation to go tackle that market together. And so Adam Jacob: First or peanut butter and chocolate. Nick: you can see how it comes together. You can see why it's interesting. we all gotta eat, Adam Jacob: Yep. And also boring. Nick: so it's all relevant or you know. So Yeah. Adam Jacob: Okay, let's talk about another weird thing. This one was the Chinese spy that was cozying up to to Jensen at NVIDIA. Keeb, what do you think? Chinese espionage. Nick: Who forgot to give them the talk? Mark Imbriaco: Ha ha ha. Nick: We we we Adam Jacob: Which I was a couple weeks ago. Nick: we covered this, you know. If you have a hot lady hanging around you all of a sudden after you've made your Adam Jacob: And you're not sure why. Nick: after you made your lots of monies, and you're not quite sure why or what the motivation is, like you know. And also who's surprised? Like Paul Stack: What was it, a San Francisco six? Yeah. Nick: who's surprised at the end of the day that anyone, you know, Any nationality is trying to cozy up to people that have interesting information, state secrets, you know, trying to get some chips on the down low, whatever it might be, whatever the motivation ultimately ends up being. I don't think anyone's surprised. And also, like I appreciate the hustle at the end of the day, like way to innovate, you know. Adam Jacob: Yeah, I love how much you have a a respect for the Chinese espionage machine, You're just like yeah, you're just like, you know. Nick: I just have a love for the game that is life, you know? I also grew up in in just an era where like you know, you just it's like the true hacker ethos where you just like try to hack the system in some way. And anytime you can see that in any applicabil applicable part of life, I just you know, I always respect the hustle. No, you know, even if it's a little dubious or devious or whatever. Adam Jacob: What do you think, Mark? Do you respect the Chinese espionage hustle? Nick: Which does not mean that I condone it. I don't want the Chinese to steal all of our technology or whatever, And also Paul Stack: New swamp, you don't want new swamp? Nick: I also do want to expand Swamp into China. Alibaba open source homies, give me a call. Let's make something happen. So anyway. Mark Imbriaco: it's f I think it is it proves the thing that we all already know, which is the weak link in these systems is the human. it's Adam Jacob: Yeah. Mark Imbriaco: like it's the the the old joke that I'm not gonna break your the encryption on your the 72 character password on your encrypted drive. I'm just gonna hit you in the head with a brick and kidnap you and make you unlock it. like it's the same thing, but a little bit con gentler and and a little bit more obvious even. yeah, it doesn't surprise me in the slightest. Nick: Yeah, well. Paul Stack: Unless your password is the number four. Just saying. Mark Imbriaco: Ha ha ha. Paul Stack: Right, yeah. Then it's yeah. Adam Jacob: Yeah, then nobody has to hit you with a brick. Nick: It yeah, I I I have a feeling this transaction was a little bit more pleasurable. Yeah. Mark Imbriaco: I that's the Adam Jacob: I Mark Imbriaco: like that's the upside of your Jensen. Adam Jacob: there's the re the report is very clear that they have no evidence of the nature of this close relationship. For those Nick: Mm-hmm. Adam Jacob: listening who can't see me, I was making air quotes because that's what was in the article. Air quotes, close relationship. but yeah, no no definitions. He's been married to the same woman since like high school. he's like so you know, let's hope for everybody that it like everybody's cool, you know? okay, let's talk about Binance. Binance got hacked. Six point three billion dollars with it be. You know, I've never been a never been a crypto guy. I I like looked at the crypto thing and was like, this is a casino and nothing useful can ever be built on this thing. That was my position all along. I don't hold any crypto. I don't not mad at people who do, you know, like it's cool, put your money where you want it. But you know what's hard to do. Hack JP Morgan Chase for $6.3 billion. It's just tough. It's just tough. you know, like you're not doing it. And sorry if that curses JP Morgan Chase. yeah, there's plenty of plenty of ways that that that the banks can get you. But it's not gonna Nick: I think we call those subprime loans or something like that, but anyway. Adam Jacob: be that someone hacked you and stole a key in 20 minutes. In 20 minutes they stole 6.3 billion dollars keeb. Respect for the game. Come on now. Give us the love. Paul Stack: No ho Nick: Listen, I I I I saw the tweet today or Friday or however we're talking about it where like, you know, CZ, the head of Binance said, Yeah, we have this covered. you know, like our anyway. Also, if you're keeping crypto on an exchange, you're doing it wrong. Everyone said that all along. So if you're losing money as a result of s of them getting hacked, that's on you. Just as much as it is on them. and yeah, I mean, it we're not talking about crypto anymore, we're talking about AI now. You know, that that part of the our lives has has gone and came and went and now everyone Adam Jacob: Man, you and you say that, but I was in New York not that long ago and I was in a cab and the cab driver was like, You're from San Francisco, and I'm like, Yeah, and he was like, You got crypto? And I was like, No. And he's like, man, you gotta get in on crypto. I got all my shit in crypto. And he spent this hour Nick: Yeah, that's Paul Stack: Jesus. Adam Jacob: and a half long drive into Manhattan. Traffic was awful, just just chatting me up about all of his crypto shenanigans. And like, you know. I think about that guy and losing s losing his six point three billion dollars or whatever and I'm like no. Nick: Yeah, I lo I love that for him and for those people and I'm like a NFT holder. I'm like, I I've got some crypto and and all of that and also yeah, making it your whole identity just like anything is kinda weird. and so Adam Jacob: What do you got, Mark? Mark Imbriaco: I I I everything you said about JP Morgan is exactly right. I think the the shocking part about this is not that they got hacked, it's that the the the attack got out so money so much money in such a short period of time. Unreal, Adam Jacob: Totally. Mark Imbriaco: the speed at which that much money moved. And okay, not actually money until they they're they're able to to convert it to to actual money. But still, like the value that was lost so quickly is just mind-boggling. I saw it and We we all we were talking about this before the show a little bit, about how everybody's getting attacked and how how the the the the world is a scary place. This just highlights it. And I saw the news last night and I was just in shock. Paul Stack: I have so many questions. How did nobody notice? That six point five billion. what are where are the safeguards? Where are the alarms? what sort of systems do they have internally that that Nick: Mm. Paul Stack: didn't have any level of triggers? Jesus, if somebody does something in Swamp that's like re increases their request from ten a second to a four hundred a second, keep knows before they know. you know what I mean? It's like you there's th that just seems that there's a a a weird level of accountability there. Adam Jacob: Yeah. it's it's Nick: That is true. Well, there is this thing that if you know Adam Jacob: it's what's built into the system. Nick: twenty four magic words, then you can steal any amount of money that's sitting inside of any crypto wallet. And so who knows what the details are. The volume of data or the volume of of dollars, it could have been just as much ten dollars as it is five billion, it turns out, if what happened is you gained custody over someone's wallet, And so anyway. Who knows what happened? We're gonna find out more. Yeah. Adam Jacob: And CZ and and Binance holds your wallet. And so there you go. Get Binance, get the wallet. And look, the the Wired reported that this is that multiple crypto tracing firms think this was the North Korean hacker group trader, trader trader, who's sort of behind this heist. So, like, that's the other, like the international piece of this is like North Korea just stole six point three billion dollars from let's call it Paul Stack: The US? Adam Jacob: their Chinese friends. Nick: Yeah. Adam Jacob: you know, like and like that's an interesting geopolitical maneuver there too. all right. That was a big week. lots of weird news kind of moved us around, but but that's the news. So it's time now for five questions. You ready, Mark? Okay, of the stories we just covered, which story are you gonna be thinking about later on today? Mark Imbriaco: It's the Quinn Pewter one, obviously. it's just so fun. Adam Jacob: Right. I think you're gonna go run a Quen Peter. Mark Imbriaco: I'm gonna go I I'm gonna go read that article and be just as baffled as you were. Adam Jacob: Yeah, it's gonna be great. which article do you think was just noise? You're like, there's no reason we talked about this. Mark Imbriaco: Yeah, it means the Doordash and Uber Eats, like who I mean and an Instacart, who cares? Adam Jacob: Who cares? people who order their groceries. Mark Imbriaco: Mm. Adam Jacob: what's a common belief in the industry that you just think is wrong? Mark Imbriaco: I think it's the idea that junior engineers are a are a dying breed. I think the I think that's a I think that's wrong. I think it's wrong. And we'll talk about why in the next question. Adam Jacob: what's the one thing you wish you could teach everyone? Mark Imbriaco: Uh-huh. That that's like this is the best time in the history of the world to not know things and to have good questions that you're willing to ask. So like the junior people can bootstrap so unbelievably quickly. I think it's just it's a great time to be eager and young and to have elastic brains and all of those things. Adam Jacob: Totally. I we were writing some job descriptions and we kept writing some variation of that into the job descriptions, you know? Like, do you want to just rip at this? You know, do you want to like figure it all out? And, you know, we were talking about it in marketing, we were talking about it in engineering, like I totally agree with you. what's the best and worst moment you've had working with AI agents? Mark Imbriaco: I don't know. The worst moment was the entirety of Opus V until 5.5 because it was just so frustrating to talk to. The the w it just made me angry every time I had to interact with it. And maybe the best moment was when five five came out and talked to me like a like like four five did. I don't know. I think the best moment was the first time I got it to loop for more than an hour doing something. And I was like, I gave it good instructions, I gave it a high-level architecture, and I gave it some. some guidelines about sort of how to to decide when it was done and to watch this thing just rip doing crazy benchmarking and going deep on a weird technical idea I had and it just went for like an hour and a half and spit something out at the end that was amazing. Adam Jacob: Yeah. Sick. All right. Well, that's all the news for this week. If you want to dive deeper into any of the things we talked about, you can find all the links and the summaries over at Swamp Dash Club.com slash podcast. You can sign up to get it delivered straight to your inbox every Monday. Morningish. if Paul Stack: Hush. Adam Jacob: if you're trying to figure out how to build AI systems that are repeatable, token efficient, and solve real problems, Swamp is here to help. Visit swampdash club.com to learn more. Mark Take us out. What do you want to plug? Mark Imbriaco: obviously I would plug Railway, but now that now that I'm here, I should talk about the iPhone Duo and and and Rex from Super Logical running Nick: Ha ha Paul Stack: no no. Mark Imbriaco: B Top on my framework desktop. It's amazing. And if you got your hands on one of these and put Rex on it, Paul, you would change your mind very, very quickly. it's very cool. I love it more than I thought I would. it's it's just fun. Paul Stack: No, looking at it, now I want one. no. Adam Jacob: Okay, but tell us about Rex. Mark Imbriaco: but j yeah, Adam Jacob: The people must know. Mark Imbriaco: just look. This thing is running tail scale connected to my framework desktop, running a full terminal window with with Btop running it just because it's colorful and fun. Adam Jacob: Yeah. Mark Imbriaco: and the thing is just glorious. I I go downstairs and I'm sitting in bed and I want to prompt my thing that's running on my desktop, and I just do. I don't have to pull my laptop out. I can just whenever I want. Adam Jacob: Yeah, that's pretty good. Nick: Just don't open and close it seventy Paul Stack: I think I see love in Keebe's eyes Nick: five thousand Paul Stack: there. Nick: times. Adam Jacob: What was that key? Nick: I said just don't open and close it seventy five thousand times, it turns out. Adam Jacob: Which will take your lifetime. Mark Imbriaco: CNET built a thing that did it 139,000 times and it looks just like mine does right now. Nick: Okay, all right. So that Adam Jacob: Yeah. Yeah. Nick: that person did FUD and I fell for it. That's what you're telling me? Mark Imbriaco: And and listen, I bought Apple Care and if it does that, I'll throw it on the ground and go get Adam Jacob: No Mark Imbriaco: another one. Adam Jacob: the only thing that has folded it that many times is a machine. No human was seventy five thousand times. they'd still be doing it. all right. That was a good show. Mark, thank you so much for coming. folks, we're gonna see you again next week. Bye.