Adam Jacob: Welcome to Infrastructure Frontiers, where automation experts run down all the AI news for infrastructure engineers. It's Monday, August 10th. So grab a cup of coffee and let's get started. I'm Adam Jacob. I'm the CEO of Swamp Club. I'm joined this week by my Swamp Club co-founders Paul Stack and Nick Steinmates. Say hello. Paul Stack: Hello, people. Adam Jacob: And our special guest is one of the longest serving sages of operations, old friend, author, advisor, John Willis. John Willis (a.k.a Botchgalupe: Yeah, the oldest guy that does operations in the industry probably. Nick: Yeah. Adam Jacob: I mean, and get just has seen every single possible variation on the every single possible themes. Wrote great books about deming. John's John's the John's the John's the man. John Willis (a.k.a Botchgalupe: There you go. Adam Jacob: all right, let's hop into it and start running down the news. all right, so the first one, this is a little weird, but I thought it was awesome. Public Domain Review posted an an essay about this thing called the Rs Notoria, which is this medieval book. That promised access to hidden knowledge if you meditated upon it. So what you would do is like find access to this illuminated manuscript, and you would lay it out in front of you. And by like studying and meditating on this illuminated manuscript, it was supposed to teach you things like Latin and mathematics and like all the all the mysteries of human knowledge and philosophy and art and everything was sort of contained inside the Ars Notoria. and they wrote an interesting essay that was basically. Talking about how people who used the Rs Notoria sort of found themselves with like a precursor to AI psychosis, you know, that like the idea that what would happen is you'd be able to open the magic book and it would just sort of give you the wisdom of the sages I think there's an interesting question of, whether or not using AI causes some kind of cognitive decline. so Keeb, what do you think about that? How you feel about the R'Notoria and the idea that AI causes cognitive decline? Nick: I mean, I I just I think I take the the alternate opinion, you know, like what we have in front of us with AI is the ability to learn a lot about anything. It's kind of like the frontier of the internet, you know, when you could finally search for things that interested you without having to go to the library or whatever. Now you can, you know, talk to basically an expert on any given topic, that at least has distilled all of the, you know, the points of a given topic. to really understand it at a deep level. And yeah, I mean, there's there's risks that come along with that. But yeah, I think I'd rather take the alternative. you know, learn everything that I could want. it just reminds me of like, you know, when you have kids, they ask questions, you know, like, why is the sky blue? Or why are how how are mountains formed? And you remember from some geology lesson that, you know, there's tectonic plates that shift together and and like, you know, cause earthquakes and build mountains or whatever. But like how that actually works, you know, getting into the ground the ground truth of it all, like is very interesting. so anyway. Adam Jacob: Totally. Paul Stack: I it's interesting because like I I read this a lot that people saying, you know, you you're using an LLM, you're not thinking for yourself. I actually think is the complete opposite for me, right? Like for me now I'm actually able to think at a much faster speed, like and like not just it's not making my decisions for me, I'm making the decisions and then it's actually filling the gaps that I actually need to go to the next decision. and it's it's allowing me to be like way, way, way more productive and using my own thought and on my own sort. sort of process in order to deliver something. So I I'm not at this point at this moment in time where I feel that it's like, you know, bad for me or it's like taking away anything. I actually think it's empowering me more than anything else. But I'm I've always been a person that's like trying to multitask. This just allows me to multitask and like on steroids. So I I'm I'm it's it's awesome. So I I don't really like see how that's happening to people yet. Adam Jacob: All right. Well, Anchor Sethi, Anchor, if you're out there and I pronounced your name wrong, I'm so sorry, is trying to prevent cognitive debt by literally making the edits that the LLM proposes himself. essentially the same learning process embraced by that like hard way style of learning, if people remember, like learn see the hard way or whatever. Like that style was like, hey, you're gonna type in every single thing that you gotta do. So are we are we sort of entering this era of cognitive debt where people are Paul Stack: Well. Adam Jacob: are don't understand what they're doing, they don't understand how their systems are enabled. And and if we are, how are we ever going to repay that debt? John, what do you think? John Willis (a.k.a Botchgalupe: Yeah, it's nonsense, right? you know, I mean I I was baptized on assemble language, right? And I could make the argument that everything that we do, I I I could say, well, you know, it'd be better. I remember taking a COBOL class once and and like thinking, This is a joke. This is after I'd like spent five years writing Assembler, you know, for Exxon, right? And and I like this is ridiculous. Why would anybody do this? And then, you know, over the years I realized, you know, I kinda like Python. I like High level languages and I don't think about when I'm losing these languages. So the abstractions work. You know, and I go back to this. I got into a big argument with Otis's face, the Kinevan guy, Snowden or S whatever his last name is. and my argument is did did the calculator make us dumber by replacing the slide ruler? And the answer is absolutely not. I mean, there there you can make arguments that there are engineers that are better who stuck with the slide rulers. But as a whole Adam Jacob: Can you? John Willis (a.k.a Botchgalupe: well, I mean I I I can't. I don't I'm saying you can argue. No, I I think Yeah, I I think Adam Jacob: Yeah, I can't make that argument at all. You know, like John Willis (a.k.a Botchgalupe: yeah, no, I I think there are people that I mean what they're doing is they're trying to make that same argument now, right? Adam Jacob: Yeah, yeah. John Willis (a.k.a Botchgalupe: That's my point, right? And I and I I do the A B C like you like A did it make us dumber, and B did the calculator make us smarter, C, it doesn't freaking matter. The abstractions exist, they evolve and you know, so I You know, I think there's one last thing that was really cool is and and I I think it's playing out a little bit from the Palantir guys, but I'm not sure is that this idea that that coding and I could really get a lot of people pissed at me here now, might be the new assembly language. Like in other words, and Nick: Mm-hmm. John Willis (a.k.a Botchgalupe: Toba Power was saying this recently that, you know, he he and there's a guy who literally worked on the original architect of the jar file, you know. Adam Jacob: Mm-hmm. John Willis (a.k.a Botchgalupe: you know, one of the best enterprise Java guys on the planet. for financial institutions. And he's basically arguing that, you know, that that just coding might be the sort of new assemble languages, like, who cares? So I don't know. I think yeah, that's my point. Adam Jacob: Yeah. I mean I feel like I think it's I think it's a legit question and I I have a feeling we're gonna talk about it more and more sort of as we as we move through this week's news, John Willis (a.k.a Botchgalupe: yeah. Adam Jacob: you know, Paul Stack: I Adam Jacob: because it's it's it's sort of the topic that's at the top of everyone's mind. Paul Stack: I cannot imagine typing every Ethernet an L L gives me. Like the amount John Willis (a.k.a Botchgalupe: Yeah, that's good. Adam Jacob: I can't either. Paul Stack: of things I ship a day, that is I mean, that is definitely a a choice in life. Like and it's not Adam Jacob: It's our move. Paul Stack: it's not a choice I wanna make. Let's put it that way. Nick: All right, so Adam Jacob: Yeah. John Willis (a.k.a Botchgalupe: Can go back to the original point that the the that like I think the point that you know that you know that you you guys were making on it l learn learning is learning. You know, like you can't replace critical thinking. Right? And Nick: Yeah. John Willis (a.k.a Botchgalupe: and the minute you replace critical thinking, then all these tools are terrible. Right? So Adam Jacob: Totally. Calculators were terrible Nick: Yeah, and I think that Adam Jacob: too if you never actually learned the basis of arithmetic. Nick: Yeah, and I think I think sometimes John Willis (a.k.a Botchgalupe: Yeah. Nick: we skip ahead a little bit. So yeah, like obviously, like Paul, you know, hand typing the code that I need written or whatever seems archaic now. But there are people out there who are still who are much earlier in their journey of software engineering and understanding architecture and building systems. And what they need, you know, like I can totally see how this works for for Anker, actually, because You know, they're going from maybe a general understanding of how a system works and what they're trying to do is really internalize, you know, what is actually happening, like what the actual mechanics of the system that they're building are. And, you know, when you touch something, it becomes more intimate, you know, like you understand it more. And so yeah, I I yeah, I'm all for where where the the the frontier is taking us in terms of like not reading the code and not writing the code and and all of that. But I think what we need to remember is, you know, there's a whole new generation of developers out there that are probably gonna need to do some things the hard way until John Willis (a.k.a Botchgalupe: But you know Nick: they graduate to, Adam Jacob: Yeah. Nick: you know, the new way. John Willis (a.k.a Botchgalupe: Sorry to hog, but this just reminded me of something really interesting is you know, Gene King's original editor, Todd Setison was a really brilliant guy. And I I wanted to write my first book, Deming, the Demi Book, about in a Michael Lewis style. Like, and you know, I'm not a great writer. And he told me the best advice I had, he says, actually rewrite his first three chapters. Just rewrite those first three chapters. And that tactile, you know, you can read it a hundred times. And he was so right. And so to your point, Nick, I think you're spot on on that. Adam Jacob: Sometimes you gotta do it the hard way. All right. So Martin Alderson wrote about how he's mostly picking models based on speed now, because intelligence has gotten so good for most tasks that he just doesn't need the slowdown that comes with the absolute frontier intelligence. And I think with that combination of prices coming down, with intelligence going up, I think speed does in fact sort of become the differentiator over time. in a way that is an is an interesting evolution of how we've sort of been thinking about and working with these models. you know, in particular, once you start using the models less as the like God orchestrator of all possible things you could ever do. And instead you're using those models to like to better understand what each step of the process needs to be and how do I pick the right model. I think this idea of speed as a differentiator really comes in. Keep I know you got feels, so hit it. Nick: Yeah, man. it's amazing what happens when you use a a model that is, you know, an order of magnitude faster than the frontier models in terms of what your visceral experience is to that, you know? And I think that I think there's a real skill in understanding what level of intelligence you actually need for a given task. that no one least no one I've seen has been able to crack. Like how do you how do you pick the how do you do that trade off correctly is an interesting question that is someone's gonna develop Adam Jacob: Yeah. And feels like an open one in terms of the frontier, you know? Like I think like there people have built model routers and trying to do all those sorts of things. That's interesting. But I think ultimately back to like having to do things the hard way, like we're just still learning what how much intelligence do I need for which parts of the job, you know? And Nick: Yeah. Yeah. Adam Jacob: and I think it's a real frontier question in terms of exploration, like how do we build a framework for thinking through that intelligence, right? Paul Stack: This has come around again, right? And the reason it has like like think about the cloud, right? Everybody used to just throw everything into a single cloud and they used to think that was the b the best thing for like their business and the way it goes. Then we wanted to move to a multi-cloud strategy and you look at AWS having really great compute, GCP having really great Kubernetes and like data querying and Azure having great like PaaS and and that type of like setup. And it's it's like a maturity of of a technology, right? When you're able to pull the different pieces. And be able to use those different pieces for the jobs that you actually recognize. So like we we do this a lot. Like I I do it specifically. I have Opus does my thinking and I at some point I actually just get Sonnet to write the code because the thought process, like the actual plan that comes out the other side is is so well formed that it's just taking that and just like actually typey type in. So Adam Jacob: Totally. All right. Well, Simon Simon Willison, who, in case you aren't aware, Simon is essential reading. his blog is essential reading for anybody who's sort of infrastructure aligned and also thinks about AI and probably just AI in general. But he wrote up a great blog post that was a summary of all the various open letters that have come out in the last few weeks. the most interesting one was that he talked about the open weights and American AI leadership letter. So this one came out after the anthropic post that came out last week. You know, is signed by 235 different companies, Amazon, OpenAI, the Linux Foundation, arguing that open weight models are essential for RD, and then coming out in favor of allowing distillation. So if you remember distillation, that's where one model is sort of used to train another model in order to sort of bootstrap its learning. you know, the the the industry tension here is that Anthropic, if you listened to us last week, came out strongly both in support of open weight models, saying that they're important and you should have them. But strongly against the idea of what they called industrial scale distillation attacks. and I think, you know, it's difficult to me to look at this and not see Anthropic trying to protect their lead, as opposed to actually thinking about what furthers the frontier of RD. You know, I say that loving anthropic, those models are great. You know, like there's a lot of that stuff that I I really appreciate about who they are and sort of how they run that company. But I don't know. I think it was tough to read it in another way. John, did you How do you feel about that? John Willis (a.k.a Botchgalupe: Yeah, you know so I you know, I have mixed feelings about Anthropic because they basically offered me seven hundred and fifty dollars or seven seven seventy five hundred dollars because they pirated the DevOps handbook and that handbook is hold almost a million copies. So, you know, I gave a basically flat out, you know what, to give everything, you know, I mean Adam Jacob: You were like, I don't need your dirty money. John Willis (a.k.a Botchgalupe: I mean like the the idea that they you know again we I I go too far off the course here, but the idea that they they they can settle this thing for seven for four authors at three thousand dollars, and they pirate it and the fact that now somebody, because there's no clear IP protection laws around this, that somebody literally could rewrite they say they can't, but they could rewrite the DevOps handbook because they trained the model on the DevOps handbook illegally. Right? so So I think the you know, the the you know, th this idea of the the sort of irony of these companies complaining about learning from their models when everything they've built which I don't have a problem with that, but then don't complain about it on the reverse side, right? I don't know, Jimmy. Adam Jacob: A hundred percent. It just it just smells like it just it smells bad, you know? Like it's obvious. John Willis (a.k.a Botchgalupe: Yeah, it's all good. Adam Jacob: It's obvious hypocrisy. Okay. John Willis (a.k.a Botchgalupe: Yeah, yeah. So yeah, that's yeah. Adam Jacob: let's talk about Steve Yeege. so Steve Yeagee wrote two articles, one targeted for CTOs, one targeted for CEOs. They're very long. They're very Steve Yeagey. If you've never read Steve Yeagee, you should absolutely go read some Steve Yeagee. Steve Yeage is an incredible engineer. he's an an iconoclastic writer, I would say. You know, he writes in a particular style with a particular voice and a particular point of view that I think is really interesting and required reading because he's been sort of moving the frontier of what's possible, you know, forward, I think, for a long time. what I think is most interesting about both these articles is the summary he gives at the top, which is that he's essentially given up on the idea that you should build a reusable harness like Gastown, that the idea that a vertical harness for how you do and build these software development is is no longer the way and that basically all harnesses in the future will be bespoke. And I think this that take really resonates with my experience. You know, what we've seen building Swamp was that, you know, it's really tempting to be like, we're gonna go build like a a perfect software development lifecycle for an agent. We're gonna build a perfect software factory and then we'll hand that software factory to someone else. And then that it's gonna produce good outcomes. And I think instead what we've seen is that you need primitives. to help you assemble those things, but that ultimately the right harness, the right machine that builds the machine is the one that is bespoke to your problem. It's bespoke to the domain you're working in. It's bespoke to the software you're trying to build. It's unique to the shape of what you're trying to put out because that's what you're doing. Like you're you're trying to build something that is in fact unique. and to see someone who had put so much effort into trying to think about how you would build you know, this grand reusable castle of software architecture to sort of come to that same conclusion. A, brave of Steve Yeege, right? Because it would have been easy to sort of commit to ideas you'd pushed as hard as as Steve had put them. And B, I just think he's right. Paul, what do you think? Paul Stack: 100% feels right. we see it ourselves with our swamp community. There's nine versions of software factories that we know of within people talking in our Discord, right? They're building that factory exactly for their use case within their org. Like we've had shit. experiences with people trying to copy and having a single way of doing stuff in software for so long, right? And I like even going back as far as the the SRE book that Google wrote, right? So many people took that as the blueprint of putting that in their organization and they had really John Willis (a.k.a Botchgalupe: Yeah, it's terrible. Yeah. Paul Stack: poor examples because that blueprint was for them. The doc like the book wasn't, hey, take this and make it and this is going to be great in your org. It's a it's an experience report almost of how you actually do it. So like it's great that that That this is the the way that people are seeing it. And I remember when we started like building Swamp Club, we were like, what are we doing? Are we doing the primitive or are we doing the actual product itself? And we when we sat back and we said the primitive is going to be way more powerful because it actually allows us to be able to build different verticals and and different people to be able to do their own thing. And it was it it felt so right. So seeing somebody else who has pushed the industry in a huge way like this coming out with it just ma massively verifies our thinking here too. Nick: Yeah, we're in we're in the era of personalized software, you know, up and down the stack from silly things like people rewriting, you know, a SAS that they're spending hundreds of thousands of dollars a year on and personalizing it specifically to their workflow or whatever. I think, you know, I I I do think that these harnesses, the way that they work, you know, trying to provide general things like skills or MCP connectivity that allows you to build something. That is bespoke to you is definitely not enough. you know, and a primitive approach makes a lot of sense. The thing is, I I do worry about the proliferation of every person having a different AI agent to solve a different problem or a different harness to solve a different problem inside of an organization, because now what you have is the sprawl of all of this and how to manage it and how to think about it. you know, I think. I think down to the primitive level, what you need to really focus on is integrating at every layer of the stack to any and every tool and any every and every harness. because, you know, because of all of this, it's like a whole virtuous circle. If there's going to be a proliferation Adam Jacob: Yeah. Nick: of harnesses, then the the technology that's gonna win is the one that's gonna, you know, integrate with all of them, as opposed to giving a a golden path. Adam Jacob: Yeah. All right. Well keep keep front run the next story, which is David Crosshaw. John Willis (a.k.a Botchgalupe: Let me let me do real Adam, let me do one real quick when you're gonna love this. One on this subject, one of my favorite panel presentations, I think it was 2011, DevOps, and it was you and Luke Kenis are on the panel. And basically you were about to talk about the snowflake and he said, Don't say it. And I've watched this many times. And that was the clear the clarity of the way you thought about this problem. So it's kind of funny to see us In that the idea that you were making your whole argument was this sort of snowflake idea that like that's the way you had to solve these complex problems. You know, and what happened is the industry sort of merged, you know, the puppet chef argument actually not to recreate that, converged into something that worked for everybody. The baby they had, maybe Adam Jacob: Yes. John Willis (a.k.a Botchgalupe: because Terraform. Yeah, I mean, well, I mean, it should it at the time, the point in times. But but my point is yeah. Yeah, yeah, yeah. Okay, fair enough, fair enough. Yeah, yeah. Adam Jacob: Yeah. Yeah. I mean, look, it turned into a snowflake for everyone in the end. They used our they used our tools to make new snowflakes. John Willis (a.k.a Botchgalupe: But I mean that it's just kinda interesting as a sort of IT historian that like this has been your sort of thread since I've known you. Yeah, so that's pretty cool. Adam Jacob: Yeah. Yeah, I think it has been my thread. it's also the thread of David Crosshaw. So John Willis (a.k.a Botchgalupe: Yeah. Adam Jacob: David is exe.dev. and he had a post saying dev tools need to be open source because of this rise of personalized software. Like the idea that software you can't personalize, that you can't customize, that you can't build the way you want, is gonna fall by the wayside. And if that's true, then it's incumbent on all of us that the software be open because otherwise you can't personalize it at the way that you need to. you know, I think I think I largely agree with him. you know, I probably push further. I'm I'm an open source hippie sort of by nature. And so like in this, I kind of see like a rise of copyleft. Like I'm pretty sure that what we should start doing is copylefting software a lot more. Whereas like historically in the last decade or so, I think we saw a real rise of like Apache and MIT licensing because it was pat compatible for business and because it had fewer restrictions. I think I think in a world where this is what's happening. We need more copyleft where what we're saying is like, yeah, if I'm giving this to you and you can personalize it, like here are the things you can do and here's what you can't. But I I think I think you can go further than he goes, which is to say, you know, there's a new kind of software architecture here that's built around how you're providing like a core of the software and of its architecture that can then be extended in ways that make that thing fit for you, you know? Like even when a Jedi goes and builds a lightsaber. They know how to build a lightsaber because we know what lightsabers are. And, you know, each lightsaber is unique to each Jedi. That's true. You know, and they gotta go on a personal journey of the soul or whatever to figure out how they build their lightsaber. But, you know, it's still a lightsaber. You know, Darth Malls had two little lightsaber things on it. But when you saw it, you were like, it's a double-ended lightsaber. You weren't like, it's a lightspear, you know? and I think ultimately that's that's sort of what's happening here in in David's take. Do any of us actually think that we're not gonna see a bunch of industries get reinvented by by by personalized software? Paul Stack: There's actually a dude who's not just doing that, he's he's actively taking closed sourced software and and rebuilding John Willis (a.k.a Botchgalupe: wow, yeah. Adam Jacob: yeah, and just rebuilding it. Paul Stack: it and pushing it out. He's done it for like a whole bunch of paid for tools. Like one thing my worry is here is that there's a misconception that open source equals free. Right. And that's Adam Jacob: Totally. Paul Stack: that's been a huge thing in the industry for a long, long time. So, you know, people like it's this is not a free lunch for everybody, right? You can't just Adam Jacob: No. Paul Stack: rip somebody's product and just like not pay for it then like do it. You there's actually still like a sustainability like with regards to how engineering has to happen and how companies need to work. So that's my my biggest worry about like this era is like we have to remember that open source is not free, not necessarily. Adam Jacob: Yeah, I mean we're gonna we should we should spend a whole podcast doing nothing but talking about that 'cause if I get myself started, I'll just never end. But let's Paul Stack: Yeah. John Willis (a.k.a Botchgalupe: And and I'm and and I'm gonna teaser in there, but the thing is a thing that sort of really scares me and I I don't even know how to write about this 'cause I don't feel like I don't feel I can conceptualize it as well as I want to, which is I fear there's an economic problem coming down like a freight train right now. You know, you it starts with like Mitchell's like that thing where you have to sort of y they're gonna blacklist or whitelist committers and and Adam Jacob: Hm. Yeah. John Willis (a.k.a Botchgalupe: then there's this idea that open source is dying and that and and that's not what this is saying, but like, you know, the ecosystem of the software is everything we're building, including you guys know this, with Claude, is all based on billions of lines of code that already exists that have to be supported. So if we shrink that ecosystem, I don't know where the our end where this whole software industry goes, 'cause ev not everything's not gonna get rewritten. Adam Jacob: Yeah. I mean I can't I John Willis (a.k.a Botchgalupe: Okay. Adam Jacob: can't now see just when I thought I was out, you pulled me back in. Like I I yeah, no, right? You know too much how to hit the button. I look, John Willis (a.k.a Botchgalupe: That's true. Don't invite me next time. Adam Jacob: I gave a talk at one of the last Oscons about the war for the soul of open source. You can go John Willis (a.k.a Botchgalupe: Mm, yeah. Adam Jacob: you can go watch it on the internet. But basically one of the things that I said in that talk, and I still really truly believe, is that I don't care very much. You know, I said earlier that I'm like a free software open source hippie I am, but I don't care that much about open source as a as an idea. What I care about are the values that it espoused that allowed people to thrive. And John Willis (a.k.a Botchgalupe: That's right. Adam Jacob: that's what mattered to me. And Nick: Yeah. Adam Jacob: it still is what matters to me. And, you know, I can look at a lot of the dogma around open source and around its sustainability and around what it has meant for industry and all those things. And, you know, what I see that matters. Is when it lifted people's lives up, when it lifted their careers, when it changed their lives, when it made opportunities for them, that's what really mattered. And I think what I see here in a lot of the stories around like, well, op is open source gonna die or does it matter? Like what I'm seeing is a new wave of community that's still built around supporting each other, that's still built around helping each other learn and grow, that's still built around helping each other to really understand in an open way. Like what we're doing, why we're doing it, why it matters, learning and growing together, furthering each other's lives in whatever way they need to be furthered. That to me is what community is. It's always been that way. And that's what I care about more than I care about my open source freedoms. It turns out, I happen to also have some philosophical beliefs that that that some of the best ways to build those communities are around Or around the software itself being free in its own economic nature and blah, blah, blah, blah, blah. But I think that's really true. Okay. I'm going to move us on because otherwise we're going to have a two hour podcast talking about the news and ain't Paul Stack: Ha ha ha Adam Jacob: nobody here for that. Okay, so Nicholas Grun says, don't be a meat proxy. So basically he's taking the output of co of Claude and he's pasting it, you know, and pasting that to another person. When you do that, you are being a meat proxy. because Nick: Yeah. Adam Jacob: the capacity for us to just ask our AI those questions ourselves was perfectly fine. and instead what happened is I'm getting your basically AI slop output with none of the knowledge of what you put into it. And so, you know, don't do that anymore. And I think his goal is to get people to read the output of their agents, understand it, validate it, and then give me the response that matters for yourselves. And, you know, he uses code review as as one of his final examples, but I'm gonna use one that is perhaps, you know, too intense between us, but like this happens between us at Swamp all the time. Where we're all working with agents. We're all like, we're all, we're all passing the information we're learning from those things back and forth to each other. And I do find myself very often asking the question, like, okay, well, tell me what you what it means to you. Like, tell me what that output says to you. and and I think this is partly why because it's not because the output that's getting posted isn't valuable, right? It's because I want to understand what you took from it. And that's hard to grab from the paste. Of of the clot output. Nick, what do you think? Nick: Yeah, I mean, there's a few different dimensions. Number one, it's very tempting, you know, to be a meat proxy. because, you know, you've put a lot of thought into the question, you've put a lot you know a lot of the context of the answer that's coming back to you. And you think, you know, people will just immediately get what it is. and it turns out that they don't. and the distillation that you do in your mind of all of that to communicate, I think is the valuable part, much to your point. I also think that it depends on who your audience is ultimately, right? Like being a meat proxy is great if another AI is what is actually reviewing the output, right? But if it's another human, maybe not, right? So I Adam Jacob: Right. Mm-hmm. Nick: I'm just thinking about our PR review process. You know, it is it is output that is meant for other agents to consume. And so of course you want it in an agent's native language, you know, that it understands with all of the resolution of the words and all of that, right? But yeah, when you're communicating with other people, absolutely you don't do not want to be a meat proxy, because it's a problem. Adam Jacob: Yeah, and it's so tempting to be a meat proxy. It's so hard not to. Okay, let's talk about Paul Stack: forty four and I don't think I've ever wanted to be called a meat proxy and I don't think I did. Nick: Ha ha ha. Adam Jacob: Well, what I liked most about this article was that he termed the phrase meet proxy. And now we can just say to people, Hey, don't meet proxy me, you know? And like, I think I'm Paul Stack: Yeah. Adam Jacob: definitely gonna say, like, you meet proxied me. And I think that's like John Willis (a.k.a Botchgalupe: Just like Adam Jacob: a it's like a gentle way to let someone know, you know? Nick: As the resident brand brand risk, I John Willis (a.k.a Botchgalupe: Just like we should use them. Nick: I don't know that I like that term, but you know. Adam Jacob: Ha ha. John Willis (a.k.a Botchgalupe: It shouldn't use a beat cloud either. There you go, Matt. Andrew question. Adam Jacob: Yeah. Okay. So turning to a little bit of business. So Bending Spoons, bought Airtable. Bending Spoons is an Italian company. They buy digital companies like Evernote or Vimeo, where their goal is to basically improve those products technically and then use a bunch of proprietary technology to improve their business outcomes. Specifically, they target companies that have like high churn, lots of revenue but high churn, and they try to like slow the churn by improving the technology, which then increases the business outcome. it's a very interesting and sort of unique kind of of private equity. look, most people hate on private equity all the time because they're like when private equity buys my stuff, it gets worse. But Bending Spoons is an example of a company that seems to actually improve the products they buy over time. Like if you are an Evernote user and like you'll notice it got better and it had stayed the same for like a decade. And that's because Bending Spoons bought it. So they paid $1.28 billion for Airtable. Airtable also sort of broke out their new AI product as a separate thing, so they didn't buy that. you know, in some ways that's a brutal valuation. It's down from eleven billion on Airtable's last raise in 2021. But at the same time, I I think the market doesn't value their business like they did in 2021, right? The number of existential threats to Airtable, the shape of how that has moved, and and their core value is real. So, like, you know, getting bought at roughly three times revenue. which is a reasonable multiple on revenue if what we're doing is buying the revenue as opposed to buying the strategy, is can be a tough pill. But I think it's an interesting thing we're gonna see more and more of in this era where as we're rebuilding software, as we're discovering all the different things that need to be built and how we're gonna work with them, you're gonna see more and more companies, successful companies with real revenue, with real growth, you know, putting up good numbers. But who just don't have the strategic multiples they used to have before, you know. What do we think about bending spoons? What do we think about buying air table? do we think that there's more of this that's gonna happen in the industry? John, I'll I'll shoot it to you 'cause you're our resident. I've watched this industry longer than anyone. John Willis (a.k.a Botchgalupe: give me the boring one. Why don't you? No, I'll kidding. No, no, it it Nick: Yeah. Adam Jacob: You think it's boring? I think it's fascinating. John Willis (a.k.a Botchgalupe: really isn't. Actually it isn't. So you know, it it's I think I'm more interested in the bending spoons angle. I guess, you I didn't really know what they would do and and I don't give a crap about Airtable. but you know, a friend of mine, Dave Mangott, who's you know, one of the sort of the core DevOps people that I respect, about I don't know about five, seven years ago he went all in on private equity and DevOps. He's now book about it and he like I was like, What what are you doing? He says, John, watch this space. And so there's something really interesting about a private equity that literally knows how what kind of knobs to turn on and turn off, as opposed to just buying it to just, you know, squeeze it dry. So so I think I I know I'm, you know, as a takeaway, I didn't I wasn't really interested in learning anything about bending spoons. So I'm gonna go back. I love the idea of that how do you maximize You know, 'cause it the private equity thing makes sense from a financial standpoint. But if you know, if you could do that well and he does it well, so Adam Jacob: Yeah. Well, and it yeah, and I think bending spoons is an example of someone who like they know exactly what they target, they know exactly how to improve it, then they do that John Willis (a.k.a Botchgalupe: Yeah, yeah, that's awesome. Yeah. Adam Jacob: and it improves, you know? John Willis (a.k.a Botchgalupe: Yeah. Adam Jacob: Okay, so back to back away from the boring stuff. Cloudflare launched John Willis (a.k.a Botchgalupe: Yeah. Adam Jacob: the agent development lifecycle and tried to put some primitives together for building software factories in TypeScript. they also launched a thing called Cloudflare operating system, Cloudflare OS. They describe that as an agent workspace for your company. so it, you know, it kind of hits a lot of themes, basically. You know, that idea of personalized applications that you can share, the need for some determinism and deterministic workflows and enterprise access control. The whole thing's under the Apache license. So it sort of rolls that in. You know, I have a personal pet peeve against calling things operating systems. I just think it's the laziest Nick: Mm-hmm. Adam Jacob: marketing move of all time, like to be like, it's the. John Willis (a.k.a Botchgalupe: Check it out. Adam Jacob: Kubernetes is the operator, the Linux for the internet or whatever. And I'm like, it's not an operating system and you should stop calling it one. and I feel that way about this, but you know, I think I think what's interesting about the Cloudflare OS and their agent development lifecycle is that there's this real tension in the industry right now between are we gonna race to build verticals and talk about how all the ways that you should interact, you know, or are we gonna be building primitives? That allow people to build the things that they want to build and the systems they want to see. And, you know, I think history tells us that, you know, we saw this with PaaS versus IaaS. Like you can build a PaaS and that can be successful, but it's gonna get absolutely crushed by people using infrastructure as a service because it turns out they want to build what they want to build. And, you know, when I look at Cloudflare OS, what I see is like super interesting. I wanna see how that plays out. Like I'm sure we'll build some Swamp integrations that let people call Swamp from Cloudflare OS or whatever. Like all that stuff is good. And I just think it's too early to be going that vertical and saying you've got it all figured out. Paul, what do you think? Paul Stack: Yeggy said it right. Like it's it's it's it's exactly the same point he's making. It's like it's it's really difficult to to build that style of thing at this moment in time. And I I think I by the way, I love that they're doing it because it shows that they're really putting thought into how they can actually encourage their their users and their customers to like think in this way. But I I just think that, you know, that style of build at this moment, we we just haven't figured out enough if that's actually what people need or what they actually need. Adam Jacob: Yeah, but I I'm inch I'm thrilled to see him try. I'm I'm I'm watching watching watching with real intent. Go ahead. Nick: And just add John Willis (a.k.a Botchgalupe: I think we had to put a lot more investment in this though because what I'm seeing, you know, I'm I'm advising a couple of companies right now and I I'm not they're not bringing me in for the developers, right? They got that covered. They don't need John Wells for developers. What they're bringing me in to is try to break out the mess of how do you get executive assistance to create solutions, how do you get the finance, the supply chain. And so we've started this thing called an ideation hackathon. And my problem right now is you have to learn as an organization to do AI. Because if you just let everybody go grab the tool, you know, at a large consulting company said, your ideation hackathon, what tools are you using? None. We're ideating. We're learning how to ideate. And and what that's gonna do is teach an organization those kind of things. This is I'm not sure that this Cloud Federal OS, and I agree with you on the OS thing, is the Nick: Yeah. Yeah. Yeah. John Willis (a.k.a Botchgalupe: right place. What I do like is this idea of how does a company learn to build a factory, not just c not machine that builds a machine, the the machine that builds your automation structure across all areas, not just development. And that idea, like like I think there's got to be a lot more investment in how our company, instead of just ramming in and giving people agents and telling the executive assistants to solve this problem versus that problem, you know, how do you build an organizational structure for content, procedures, guardrails, all those things at at a scale of all the things you want to solve. Anyway. Nick: Yeah, I'm gonna build on something that both of you said. You know, IaaS versus PaaS is actually a great way to think about it. It's like a like a jellyfin, jellyfish swimming through the ocean, you know, like it it oscillates, right? So in the beginning we had Heroku and then we had the proliferation of AWS and G C P and Azure or whatever, and now we're back to Vercel and Railway. Adam Jacob: Heroku came second, thank you very much. Nick: You know, we're back. We're back to to Vercel and Railway, right? And so and I think that this is the same thing. You know, if you're all in on Cloudflare, you're using Cloudflare services and what you want to do is add a little AI to the mix, you know, this is a great Adam Jacob: Gonna be great. Nick: way. This is a great way to get started to start to build, you know, that expertise inside of your organization on how your EA can build some workflows, you know? but what's gonna happen, I think, Ed To Paul's point, you know, is like these bespoke harnesses are gonna then proliferate out of that and then it will all collapse again in ten years, you know, or five or whatever. So anyway. Adam Jacob: Yeah. All right. Let's go to let's go back into some like harder infrastructure stuff. So Pulumi has brought HCL and Terraform as a first class language for Pulumi. So and Terraform Plumi Cloud now supports Terraform State. They even went so far as to basically say, you know, if you have committed spend at HashiCorp, we'll give you Pulumi for free for the duration of your committed susp of your committed spend. I think, you know, there's a lot here that's interesting and you know, Paul Paul worked for Palumi. I've I've known those guys a long time. I know John has too. John Willis (a.k.a Botchgalupe: Who's ballooming? Adam Jacob: and like and this is this to me is just such an interesting decision in terms of where you're putting your chips and sort of how you think about where the future of your company lies. Because, you know, like Terraform certainly won in terms of the debate between which of these tools is going to be the class leader. And I don't know that I've ever seen a competitor just decide to essentially embrace their opponent's technology and be like, whatever, like whatever you're doing, just come here, we'll do it better for you. And I think I think ultimately he's trying to make a bet that says Pelumi will be a better platform for using agents to build this infrastructure and that the way that you specify it will be less less meaningful in the future than the fact that he sort of has the data about what you have done. And therefore opens up this sort of agentic future. Paul, you're the obvious person to go to for the hot take. What you got? Paul Stack: Eight years that issue's been open. Eight years. I remember the issue actually being open when I worked there, which is kind of weird. But the interesting thing here is is that this this feels like a Pulumicloud play, right? This feels like Plumicloud is is a huge thing and this like they really wanna make this a a massive area. like Palumy Cloud as a as a a terraform state back end, that's always been like what people asked for like when we were there because they wanted Terraform state, they want the plumey state, they wanted everything to live side by side. So you know it it's opening up their ecosystem in a massive way. It definitely shows that there is this weirdness in in the industry that HCL is is the language that people want to write their infrastructure in. I don't love that, to be totally honest. As a as a as an infrastructure purist, just because of the fact that I hate like those those declarative style languages. And I always loved that you could do a little more with Plumi and about the the actual programmatic side of it. And you could try, not saying you it was like, you know, different from declarative, but you could try and sort of like understand how the application was written and try and and be a a wee bit less declarative than you were. But this is definitely like feeling that it's showing that Pulumi are gonna put a huge amount of time and effort into Plumy Cloud. Adam Jacob: All right, we got three more to go. So Brett Stevens in the New York Times was begging you to never write with AI. and his argument isn't one about morals or about even disliking the output of AI, but instead one that says that intellectual muscle of writing, that the ability to actually just put your thoughts down on paper or or type them, and then to see those thoughts reflected back at you clarifies our thoughts in ways that not writing cannot. And you know, this one actually kind of resonates with me. I'm a person who like I I write all the time. I've got a notebook all the time. if you've ever seen a conference talk that I give, I wrote those conference talks longhand before I actually built a single piece of the slides. Like it's kind of laborious, but it's the only way I know to sort of like get those thoughts out of out of my mind. I also don't see pictures in my head. So like every single thing that happens in my head is just words. So it like very much resonates with me to like see to to the idea that what happens is I I clarify my thoughts through writing. John, you've been doing a lot of writing. You're the only published author, I think, on this call. How'd you feel about Brett's top in? John Willis (a.k.a Botchgalupe: No, I you know, for me I totally disagree. And I think, you know, I could make two arguments that I've already made. One is well, I'll put my assemble hat on and that you guys should also write an assembler and you shouldn't use any of this stuff. Right. or the second argument is the critical thinking portion, right? So I mean it and and the abstractions work, right? So I mean when I write now, when I do research for an article and even the book I'm writing right now, History of Quantum Computer, you can get it plugged in, but it won't be out for another year. is that What I do is like what Paul does for code. I I use one model to blueprint out what I'm thinking. So I'll give it a long prompt and I'll work through this idea of like, I got this idea for this sort of chapter about this and this and this, and help me construct it. Then I get that. Then I take that construction and I play around with a little agentics. You know, sort of like, you know, let it let it flesh it out. And then I take, then I sort of condense it down. And so I it's a process that I go through. through now and the end result is me. Right? So to say that like it's a binary decision, please God, don't I think he even said, you know, don't never write with AI. If you said never write an article that is completely written by AI, I'd 100% agree him. But to say never and then one last bit portion that sort of bugs me is nobody cares that code can be developed by AI. But everybody is like totally freaked out. That an article would be written by A. Now, if the article's good, then it's good. If the article sucks, then it sucks. But like just because it was written by AI doesn't mean and again, with AI is how I do my work. I don't I don't anyway, so yeah, I I I I don't know. I just I have these bad problems. These people that just like say AI is terrible, you can't use it, don't use it, it hallucinates. That that kind of thinking is just there's no place for it right now. Adam Jacob: Yeah, it's gonna hold you back in the end. John Willis (a.k.a Botchgalupe: It is. Adam Jacob: okay. So John Watson, who runs the platform team at Swamp. This one's a little self serving, but it is interesting, wrote about how he used Swamp to replatform Swamp Club itself four times in six weeks, like four full platform shifts. No one noticed that we did that work. None. Like it just happened. And, you know, this is the kind of stuff that's possible now in production infrastructure. That would have been unheard of a year ago. you know, how many weeks, how many, how many weeks would you have just talked about a platform transition? Forget about actually making one. You'd have just talked about it. and I feel like this is a place in the industry where we're starting to see the speedups happen, like we saw it in code writing, and we're starting to see it happen in production infrastructure. And, you know, it's not evenly distributed. Most people haven't experienced this, but we have, and I know others are as well. And I think the implications here are just so big and they're so broad in terms of how we think about managing platforms, how we think about vendor lock in, how we think about how our systems evolve over time. Paul, what do you think? Paul Stack: I mean w Not just in platform here, right? We've seen it in with LLMs moving between languages. This is the freedom that it gives you. This is the expression that's there. Like you have a working system that is your contract. You're now able to understand what the contract is the other side. Doesn't matter in whatever way you actually do it. But this for us was specifically fascinating because we chose a very small cloud blueprint to start with. We didn't go for anything like massive and we wanted to continue to like evolve as we actually scaled. So the fact that we could do this with our own tool. sort of lend lended itself to the fact of what we were actually building and it it we learned so much product thinking around the back of this and how other people would be able to use it and like I I I really think that the era of cloud lock in is kind of like gone here and this is what this is exactly the type of thing that it shows. Adam Jacob: Yeah. Okay. config hub, from Alexis Richardson, Brian Grant, Jesper Jorgensen is now available to everyone. So you can just go to confighub.com and you can sign up and use config hub. they wrote a couple of blog posts, both about sort of how it's evolved through their sort of clo more closed beta, and then also now how you can sort of use config hub to make changes. You know, the pitch here is that you basically centrifise centralize your configuration in config hub. Then they're gonna go generate all the config that you actually run, uses a graph behind it, gives you provenance for where the config came from, allows you to write functions to do it. You know, all through I think all three of these people are great, smart folks. I like them all. they have and will continue to do interesting work. I have to admit, reading this stuff, how I felt about it was that I just spent six years of my life trying to build system initiative and thinking about how do I build a configuration system. that is like highly modeled and autograph and had provenance and wrote functions and generated outcomes. And I just feel like, you know, where that path led us was to was a place that was interesting, that had a lot of people, you know, saying how cool they thought that it was, but that ultimately just didn't have a strong enough material outcome in people's lives to really catch fire. And so I'm really interested to see other people take similar ideas. Like they obviously, you know, They came at this on their own. They didn't, they it's not like they were listening to us talk about system initiative and they were like, Adam's so smart. Let's go build one of those. Like they knew what they were doing. They knew what they wanted to see. And it's very different architecturally from what we built. But I think it's an interesting, it's an interesting thesis in the era of AI to say that what we're going to do is sort of run configuration through that provenance layer. You know, when I look at what they built, it it feels like a layer that was built for human beings to reason about configuration and to reason about how those things go. And I just wonder how much of that reasoning remains at the human layer when we can have we can build machines that are gonna be so effective at sort of dealing with that configuration drift. John, what do you think? John Willis (a.k.a Botchgalupe: Yeah, I mean I you know, going back to the history, you know, when I came on board at Chef, like I had all these ideas. Why don't we do this? Why don't we do that? And you said to me, you said, John, configuration management is a hard problem by itself. And so when you said six years, I see twenty years, my friend, that you've been doing Adam Jacob: Yeah. John Willis (a.k.a Botchgalupe: this. Right. You you know, you what you understand and I think you thread the needle now. I think you you know, I think you've learned so much, you know, between what you've you know, all the things you've done. I mean, Habitat I thought was a great product. It just didn't find its way because of a lot of other variables. But you know, I I think you're sp I mean I think you are spot on. It d I know you didn't ask me to come and be an advertisement for you, but but yeah you've got twenty years and you under s Adam Jacob: But it never hurts. John Willis (a.k.a Botchgalupe: nah it doesn't hurt. and I love you brother. But Adam Jacob: I you too. John Willis (a.k.a Botchgalupe: but the you know, you know I think you I think y you found this way that you just stuck with it. And so I don't saw anybody could understand better where we are right now than you. And and that was a great sort of analysis. I you know the you know, I looked at a little bit at the fig hub stuff. I mean to me, you know, I didn't get into the gory details, but I didn't I didn't find it interesting. I think what you're doing right now is incredibly interesting. So Adam Jacob: Well, thank you. I mean, look, I think I think there's plenty of people who are gonna find config hub really interesting and I think it's gonna resonate with a lot of folks. What I'm but yes, the I the the part of me that's been working on it for twenty or thirty years, I'm like, I'm both really pleased to see people that I like and respect carrying some of those ideas forward because I wanna see if they're if a different take has those ideas have legs, you know? And at the same time I have real fear in my heart because I'm like, man, I went down that road and it didn't work. John Willis (a.k.a Botchgalupe: What if they were correct? Adam Jacob: And so I'm like, you know, so I'm like, I hope it works for them because I am like afraid that it won't, you know, because I John Willis (a.k.a Botchgalupe: Yeah. Yeah, yeah, yeah. Adam Jacob: like them and I want and I wish them well. and at the same time, like me, I also kind of want it to work because then I'll be like, I was all right all along. You know, I knew there was gold in them their hills. I was, I was, I was, I liked config hub before it was cool, you know. Like I I feel I feel conflicted. Okay. look, that's us. Those are the news. Let's let's do five questions with John. John, are you ready for the five questions? John Willis (a.k.a Botchgalupe: Sure, let's do it, man. Adam Jacob: All right, so the stories we just covered, which one are you gonna be thinking about later today? John Willis (a.k.a Botchgalupe: Yeah, I I think I wanna go back and look at the Cloudflare OS. Not that I'm gonna build an infrastructure out cloud cloudflare and I actually put some websites 'cause for me it's I don't like maintaining anything these days. So I actually have a couple of websites on the Cloudflare 'cause it's just so freaking easy, right? And I don't have to worry. It's bulletproof from my perspective. But I like this idea if they if they're if they're trying to capture an organization 'cause this is what I'm trying to do with some of my clients right now. Try to capture what is it gonna take organizationally learn and build a structure around it. And I think Yage's is probably a little more technical than I care about, but the factory idea and again what you guys doing, I I I like this idea of not just throwing everybody to the wolves with AI and trying to figure out what are the primitives that are going to help an organization scale this stuff. And by scale they got to learn. So yeah, I think those two are the the the Stevens, Adam Jacob: All right, which article is noise? Which one are you like? Bang. We could have not talked about it. John Willis (a.k.a Botchgalupe: the Stevens, the binary thinking on AI is, you know, don't you know that Adam Jacob: Yeah. Paul Stack: Mm-hmm. John Willis (a.k.a Botchgalupe: No is. Adam Jacob: Yeah. No New York Times op eds for John Willis. You heard it here first. What's John Willis (a.k.a Botchgalupe: No, I did some good ones, not that one though. Adam Jacob: What's a common belief in the industry that you think is just wrong? John Willis (a.k.a Botchgalupe: I I think there's still, you know, I the tides have changed, you know, when I first started like thinking about A like I fell in love with this whole A the the generative AI thing right off the bat, right? 'Cause for research it was incredible. And it had it it has dragons today, but it had a lot more dragons three or four years ago. but I've seen like most of the people like you guys and stuff who were not wholly bought in and now the people I really respect, I feel really good that we're all sort of on the same page about what this stuff. We're not over hyperbolic. But what I still see is a lot of people that I really respect who have this binary thinking. You know, like, you know, you know, Dr. Woods, for example. You know, he he you know, he there was an article that compared two philosophers, and he didn't agree with it, and he called it a hallucination. You know, and therefore this is why you cannot use LLMs to and I I went and I dissected it, and it was actually, you know, there were things in it that the guy's opinion probably overlapped. But anyway, long story short, there's no room for binary yes or no in this case. Adam Jacob: Yeah. What's the one thing you wish you could teach everybody? John Willis (a.k.a Botchgalupe: You know, cliche, critical thinking, you know, risk is broken. I mean, everybody's trying I I've spent a lot of time on risk right now and I'm trying to do workshops and stuff like that. And and I everybody's trying to build the modern agentic risk blueprint based on all the principles that we've had for the last twenty five years. Well, they don't work. They're completely broken. And one of my men Adam Jacob: Right. The principles are destroyed. John Willis (a.k.a Botchgalupe: my mentor, yeah, it it did it's it's gotta be a complete rewrite and a rethink, particularly with a gentics. And Shannon Leetz has been my mentor there and you know shout out to Shannon Leeds. Anything she touches is is gold. So Adam Jacob: What's the best and worst moment you've had working with AI agents? John Willis (a.k.a Botchgalupe: This I I love this one. So I mean literally about a month ago, a good friend of mine's Mexico City. I hadn't seen him in a while and we we we got drunk together and we started talking about games we used to play when we were kids. And so he used to play dice baseball or the dice rolled up and and so we just kept talking about it more and more and we thought what about a Yahtzee baseball? And by the end of the night, I had a working drunk, by the way. A working seems to work pretty well, pretty cool five dice baseball game. That's like a Yahtzee baseball, fully, you know, like running. They got I, you know, I made it public. I'm not c I'm too I'm a little still p paranoid about like promoting it, but we use it now all the time for ourselves. It's just we do nine inning, we do World Series. But it was just the idea that that we could have a conversation that starts with us talking about, you know, when we were kids, what kind of games did you play? And we were both sports nuts and and you know, by the end of the night I'm like, Hey, let's try this and then we spent look Adam Jacob: That's personalized software to a T. John Willis (a.k.a Botchgalupe: Yeah, it was it was it was magical. You know, I've done a couple of things. I'm not like you guys. I'm not a pipe hitting coder, you know. but the you know, to me it was like, okay, this, you know, this is the stuff that is magic. Adam Jacob: Yeah, that's the sauce. All right. So that's all the news that's fit for infrastructure this week. If you want to dive deeper into any things we talked about, you can find us and all the links and summaries over at swampclub.com slash podcast and sign up to get it delivered straight to your inbox every Monday morning. if you're trying to figure out how to manage production infrastructure and applications or scale your use of AI within the organization, Swamp is here to help. Visit swampdash club dot com to get started using AI agents safely and effectively by building repeatable systems you can trust and verify. Paul Stack: And we have a really exciting announcement. We're actually hiring our first engineer to come and work at Swamp outside of our founding team. You know, if you're interested in coming and building on the forefront of AI here and like following through all the principles that we're talking, you know, this is the type of place you want to come, you want to check it out, you can definitely get in contact with us and you can check out our jobs page on swampclub.com. Adam Jacob: And you get to work for Paul, which is gonna be sick. Okay. I'm Adam Jacob. John Willis (a.k.a Botchgalupe: Paul is incredible. Paul Paul is incredible, by the way. Just he's just an incredible person. Adam Jacob: Paul Stack, say goodbye. Paul Stack: Goodbye, everybody. Adam Jacob: Keep. Nick: See ya. Adam Jacob: All right. John Willis is our special guest. Take us out. What do you want to plug? John Willis (a.k.a Botchgalupe: Yeah, I mean probably the best thing I is I I spent, you know, a few years writing a book about the history of AI. It's a Michael Lewis inversion, it's called Rebels of Reason. I think it's the best history of AI book. I'm sorta biased, but I took a a a approach that literally my mother in law could understand it and somebody who c knows how to read neural network attention tables can get a value out of it as well. So yeah, and then yeah I if anybody's interested in hearing more about these innovation hackathons, these no code innovation only hackathons for enterprises, reach out to me. Adam Jacob: Amazing. All right. Thanks everybody. We appreciate you. We'll see you next Monday. Bye. John Willis (a.k.a Botchgalupe: Yeah guys, thanks. Adam Jacob: you