Episode 585: I'm not a total AI hater, I'm a finance hater의 전체 녹취록

that they've created from John Oliver show. I think it's Coté (02:34.797) Right, right, right. Brandon (02:46.776) It all goes to a charity. It's I think it's like something against hunger, end hunger or something. So that's I guess to me, if you're Buckies, you're like, no way do you touch that because you're you're gonna lose, right? So and then I think maybe don't you just tell because I thought it was weird is that they they did file this lawsuit against another small thing. I was like, wouldn't you just tell the lawyers like, hey, let's just let's just put it on pause. Every you know what? Everybody go on break. Like, let's take let's take a European summer. Coté (02:56.931) Yeah, yeah. Matt Ray (02:57.324) Yeah. You're gonna lose. Matt Ray (03:14.083) Well Coté (03:14.243) Yeah. Have a summer vacation. Yeah. Brandon (03:15.76) Have a summer vacation and just don't do anything. And we're just gonna let this quietly go away. And then if should they decide that they're quote unquote, they have to protect their trademark, they could pick it up like, you know, six months or a year after it goes by. Right. Like, isn't that the the move here? Matt Ray (03:19.992) But Coté (03:30.925) Right, right. Be be because it it it starts to feel like you're close to like a Papa John's moment where like everything's going great, you're doing the pizza, and then it turns out your founder's kind of an asshole. Right. And and like everything like you gotta figure out some way of getting rid of him, like all this sort of sentiment where it's like, Hey Buckeys, everything's fine. Don't don't s don't kick up any any dirt, just keep expanding. Matt Ray (03:31.394) Yeah, I mean the the Brandon (03:52.593) But now do you think this is the law? Okay, the chief legal officer if they have one at Buckies, but I guess more likely the outside council. I guess they would say something to the effect of like, listen, if we don't sue these people, we're gonna lose our trademark, right? If we because there's the whole thing about like, unless you protect a trademark, it goes you know, it's be no longer becomes a trademark, right? So that's why oftentimes big companies will take a very small company that seems to be unrelated, and they'll just say, like, well, we have to do this because we have to protect the trademark. Matt Ray (04:04.833) Mm-hmm. Brandon (04:21.242) Right. So that seems like the chief legal officer or outside counsel is giving them. But what you just did, as a new this is a new wing of software defined consulting. We're now into legal consulting. Right. So what we say to them is like, first we say to them, we're not lawyers. We just have common sense. You can either try to protect your trademark by suing all these people, therefore making everyone angry and they're gonna want to eventually turn on your store, or you can just quietly do nothing. Like probably be fine. Your trademark will still be okay. And if it comes up, if it if there's a real case where like

somebody, a big company, like Big Oil tries to take on a beaver logo, you sue them, right? And you see, that's all you gotta do, right? Otherwise you just let like little beaver town, the little stop sh stop stop and shop place, just let them have a a weird looking beaver, right? It's fine. They'll be fine with that. Matt Ray (05:08.46) Yeah, but Coté (05:13.944) I I I think I think I think what you do is you find some influencer on YouTube or wherever, and every time you find one of these little places that's using the Bucky's thing, you just send them there and you're like, just make some videos. Like go go there and be like, Hey, I'm the Bucky's influencer. We want to t tell me about why you're using this, what you're doing, we're gonna make some videos and you kinda like wrap it all up back into the Bucky's brand and you suck it back in. And then, you know, it's sort of like Matt Ray (05:14.657) I mean Matt Ray (05:42.231) But what if Coté (05:43.533) I'm not necessarily saying you should I mean like like good job using our logo, like and using our brand. You've really done a great job with it. And then maybe maybe you also give them just like a free license to use it. And so like I mean again, as you said, we're not lawyers. But it seems like you could just extend a license to people to use it and then who cares? It's fine. Like like I I'm pretty sure like if you were selling hats with the beaver or shirts. I don't think Bucky's gonna go bankrupt. In fact, they probably will be able it's kind of like an open source issue. Like we you have this common good, people are promoting the use of it, you don't get paid for people using it and enjoying it, and yet you prosper because you're selling big drinks. Brandon (06:27.622) So what if go one step further? What if we I completely lean into it? Now, this is another, this is the PR strategy. So we're gonna, this is a separate invoice that we're not, you know, we have to the legal strategy, we're invoicing them a separate thing, right? And we say, listen, this is what you guys do. One Bucky somewhat famously doesn't have any online merch. They don't do that. That's like one of their stand that's kind of a philosophy for them. So what they should do is actually on their website, they should say, listen, we don't have an online merch. But if you want to go over here and buy all this buck off stuff. Coté (06:37.55) Another branch. Matt Ray (06:48.237) Mm. Brandon (06:56.398) It all goes to charity and John Oliver's doing and and you just say, go for it. Just buy as much as you want. And then you you email John Oliver and be like, We're we're supporting you in your show. We're gonna find a way, we're gonna find a legal way to like allow people to have a license as long as long as they're within some parameters. And you even ask to go on the show. You even send some funny lawyer to go through and be listen, this is and maybe even send a straight man just to like allow John Oliver to just take as much as

you send like a real lawyer and be like, he tries to explain it. Coté (06:57.557) it's like an Amazon partner or whatever. Brandon (07:25.06) John Oliver makes a lot of jokes at his expense. It's okay. That lawyer's going to be paid thousands of dollars per hour, right? And then at the end, you just say, like, hey, they I think we've worked it out. We appreciate you bringing this to our attention. So you just you got to like just lean, like just own it, right? I think that would be a fantastic way to handle this. Matt Ray (07:41.614) I mean, are they is Bucky's going after anyone who's not a convenience store gas station? Like are they going after like, you know, Dick's sporting goods cause they have like, you know, don't they have like a beaver logo or something? Okay, whatever. But I mean like they kinda have to go after any other gas station looking thing. Brandon (07:50.042) Well, we don't know. Brandon (07:56.689) Not to my knowledge, but that would be that you know, that's a further review. Coté (08:04.846) Yeah, yeah. I I mean it it it it just Matt Ray (08:06.333) You can't just say like, you know, you know, go go to Jo you know, go to go to Cote's gas station and and get yourself some Bucky Bucky paraphernalia. by the way, buy their convenience foods, you know, buy their gas. You know, y I mean, there's a reason the devil's advocate's a lawyer. Like they have to go after everybody. Coté (08:23.564) Yeah, yeah. I I mean that what ruins it is they have all those delightful billboards that are like, if you gotta take a shit, just wait fifteen minutes. And and like in small type like in small type it would have to say, and if you tattoo the beaver on you will sue your ass. Matt Ray (08:29.901) They've gotten a little cruder since I've been there. Brandon (08:41.69) This episode is sponsored by Cygnadot. That's S-I-G-N-A-D-O-T. Coding agents are shipping 10 times more code, but verifying it is the new bottleneck. Every change ends up waiting in line for one shared staging environment, and duplicating the whole stack for every change doesn't scale either. It's never just the services, it's the databases, the cues, every dependencies. Cignadot takes a different approach. Lightweight, ephemeral environments that only spin up services you're testing and share the rest. Developers and their agents. Get a high fidelity environment in seconds and run real end-to-end tests on every change before it merges. Try it free at signa dot com. That's S-I-G-N-A-D-O-T dot com. We talked with SignaDot CEO back on episode five eighty-one. Worth a listen if you missed it. And thanks to SignaDot for sponsoring our show. Coté (09:28.376) Well, Brandon, you pay close attention to this stuff and 'cause I know you're on the the the AI and sometimes cloud monthly roundup. So you so you read on it. And it it seems like we are I don't know I don't know if it's the apex or the top or whatever, but we're at the height of figuring out what open source means in the AI world.

And I I think you you've got a I've got this jumble of words on the table in front of me, metaphorically speaking. And I think Brandon (09:37.83) Right. Coté (09:57.046) You've got open weights, which I guess means there's closed weights, and you've got distilling, and you've got these other things, and there's kind of this battle between it takes a lot of money to produce this software, I'm gonna say. Brandon (10:16.368) Well the weights. That's what it takes. So it takes a lot of money to cr to create the weights. Coté (10:20.181) Right, right, right. And and and j' I'm I'm kind of mixing together open source stuff with like AI stuff, but it takes a lot of money to create this thing. And what we're debating over is which part should be free to use. So in the open source world, it's like, well, you know, programmers are expensive. And even as as we know, even if you buy an AI, that's also expensive. Costs a lot of money. Brandon (10:25.616) Mm-hmm. Coté (10:45.719) To like do this stuff. It's not free. It's not like building like a you know a fort out of sticks you find in the park. And so we've had lots of discussion over the years who should pay for the the person who maintains curl versus this versus that. Well, whatever. So that's always a debate. But it seems like right now we're trying to figure out so which part of AI should be free. And like I think the most recent thing is I I of course haven't read it. but I've read some summaries of the maybe thirty thousand word essay by Mark Zuckerberg about like which parts of AI should be free. And then there was also some more open weight stuff a few weeks ago. I don't know, like where is the debate on which parts should be free? Brandon (11:30.683) Well, that that's that is the debate. So I guess to summarize it, it's Meta has come out and said ultimately that these models should be open source such that no one company has a monopoly over the model, right? And that the case that he's making, and I think it runs 14 pages. I think it's 6,500 words. So it's a lot to get through on that thing. So what the in part of that essay, he's basically making the case that. Matt Ray (11:53.686) Good God. Brandon (11:59.035) That that's the way you basically, if you will, make sure no one company or one person, if you will, is the czar of the the model, right? And we do that by keeping them open and available to to everyone. So that's the Mark Zuckerberg meta position in a simplified way. And of course, anthropic and open AI, and to a less extent Google, we're gonna put them to the side. They're always this weird case, but you know, they are building proprietary frontier models that are not. Coté (12:14.157) Mm. Brandon (12:28.602) you know, open source. They don't release the weights and stuff like that. So they obviously are coming at a different way. And and so Go ahead, Matt. Matt Ray (12:33.664) The I mean, don't th they do both like OpenAI and Google have tossed off open source models of that you know, so the the the their their their leading stuff is always closed and then, you know, there's like a trailing, you know, open source exhaust. Brandon (12:43.866) Yeah, the older ones or whatever. Yeah. Yeah, yeah. Brandon (12:53.7) Yeah. So I mean, so then we have to kind of always come back to like the incentives question, right? It's like, you know, I mean, of course Facebook or Meta has this very profitable business that doesn't depend on AI, or at least AI isn't like the core offering. So, you know Coté (13:07.819) And it's a very open business. Anyone could participate. They w integrate perfectly with Mastodon and Blue Sky bi directionally. Brandon (13:10.534) Right. And we and and we know and we know exactly to Kote's point, we know exactly how open the Facebook and Instagram and all them are. So I mean, so it is a little bit like you can't even read you're only like maybe a thousand words in and you're like, mm, you seem like you're a bias source here. You don't you seem like you've got like your own agenda here. It's like it makes sense for you. Right. And we talked about this at length before, right? Like obviously Microsoft and Amazon Matt Ray (13:30.604) You're like, are your sources reliable? Brandon (13:38.181) I think they're just a little bit more simple about it. They're not even building one. So they're just like, Yeah, like we want open source models because we just want to we just want to run models. We don't want to build them. And so Facebook's like saying, here's here's one. But of course, to your point, like Facebook, you know, it's not surprisingly or Meta's coming out and saying, like, yeah, of course, of course, everything should be free. Now, of course, that that's not gonna work for the frontier models who are like spending however many billions of dollars building this, they have to make money somehow. So they can't just be like, Okay, everybody, I'll just make my model. Matt Ray (13:47.393) Yeah. Brandon (14:08.26) you know, if you will, free and everyone can run it and not pay me anything because those companies would have no business then. They wouldn't have anything to charge for. Matt Ray (14:13.132) I love I I love the idea of a trillion dollar open source company. Brandon (14:19.258) Yeah, I mean I'm just so it's just kind of like crazy, right? You're like, of course this isn't going to like I mean, you're just basically saying these companies shouldn't exist, which is like another I mean it's just crazy. There's just no way it could work, right? So so I guess that's kind of the you know, the state of play. So and then I guess you know then of course we always have the quote unquote Chinese models, which are like their own little segment of the world where they're being funded in part by the state and they they are mostly open source models. Of course. Those models then are created by distilled models from the frontier model. So there's like a weird, you know, we talked about that before. There's weird linkage

between all of this. Matt Ray (14:54.582) Well I I think that that from what I've read, that's really an oversimplification of the Chinese models. They have their own frontier models that are like cross pollinating and they're l a little more open about the fact that yeah, sometimes we distill, but they also are building their own frontier models that are just as capable. So you can't you can't distill your way ahead of the frontier. And Coté (15:16.001) Yeah. Coté (15:20.801) So so they're they're like they're like two thousands open source where you just like you're you just throw your keys into a bowl and whatever keys you get when you like leave, that's who you leave with. Brandon (15:21.36) Yeah, that's fair. Brandon (15:31.088) Yeah. Yeah. Matt Ray (15:33.9) I don't know what your two thousands were like. Brandon (15:36.369) But I think that kind of gets I guess what the part that will become potentially, I mean, I think me I don't know if we should predict it, but it's sort of like if the Chinese release a model that outperforms the frontier models of the US, right? That is going to be like a real turning point to your point, Matt. Then it's gonna be like, okay, it doesn't even matter how they do it. Maybe it's distillation, maybe it's their own, maybe it's a combination of both. But if they get ahead, right, then I think you're gonna see a whole bunch of people be like. Matt Ray (15:53.356) Yeah. Brandon (16:05.758) red alert, red alert, you know, they've they're ahead and you know, they've taken the lead and and the United States needs to do something. Right. So that's like that now that has to to date, that hasn't happened. The models are always like six months behind, which doesn't seem that long, but you know, it it seems like they get better. But you know, to your point, to Matt's point, it's like they have very I mean, they have a lot of smart people too. So it's like, yeah, I mean they they can figure it out too, right? Matt Ray (16:14.089) And I think they're close. Yeah. Matt Ray (16:27.123) Yeah, it's it's yeah, that that that's my point is do not overs oversimplify and say they're just copying because it's like, you know what, they don't just copy. I mean if i you know, that was kind of the the US's like defense of of like you know, IP and stuff is like, well they're just copying us, you know, they'll always be trailing. It's like if you look around, the Chinese are ahead on a lot of things right now and it's you can't copy your way ahead. you know, I mean you can copy your way to parody. but once you get there you kinda have to, you know, go out on your own. And and what what's really interesting is you know the the fact that yeah, these Chinese models are being open sourced. but if you if you followed like the the open AI hacking of of Hugging Face, like the way Hugging Face was able to find out if they had been where how the hack had happened was using Chinese open models because the the US ones had limits on them. And so like, you know Zuckerberg didn't talk about that. He's like, you know, we want open weight models, you know, he but he didn't say like, and I want them to be completely unfettered and you can do whatever you want. You know, that's that's what's happening over in Grocland, which yeah. Brandon (17:38.011) Yeah. Well, I guess the other thing in the the long I don't know what you want to call it, the missive the the paper that came out is like, you know, manifesto. All right, the manifesto. The the meta manifesto, I guess, is you know, he I mean he paints a very like, you know, positive, you know, if you will, like, you know, kind of the typical and I'm and I say this as someone that likes AI, but he's kind of painting a very, very positive. It's like you know everyone's gonna have like a their own agent that's gonna like, you know Coté (17:45.302) I think you call it a manifesto. That that's the appropriate thing. Matt Ray (17:45.739) Treaties. Matt Ray (18:03.776) Yeah. Brandon (18:06.468) help them learn that's going to take care of like menial tasks that's going to like so so it's all you know this idea that like everything is going to be good and and again like you know I just feel like it's the law of unintended consequences like because there was a point right we all remember Facebook and the Arab Spring and you know there's the feeling that like these these technologies connecting everyone together is going to lead to a lot of really good things and of course you know many people Matt Ray (18:11.379) Gonna give him legs. Brandon (18:34.906) Would point out that like this hasn't gone exactly the way we thought of, right? I mean, there's like numerous issues with like they've been pointing about Instagram and Facebook. And even now they're being sued and at least here in the United States, I think by Arizona, you know, and they've lost some pretty significant judgment. So that's the other part that's like, I mean, that's the part that I think is so interesting. just like some mind, I it's like getting in someone's mind. It's like you're writing this, and I can see, I can understand like being optimistic about AI. I mean, I I feel like I'm optimistic, but like. Matt Ray (18:38.783) Yeah, yeah. Matt Ray (18:48.915) New Mexico, yeah. Brandon (19:03.758) Writing it without like reflecting on like very recent history you're involved in seems like a key mistake, right? Like if we were in in one of your old seminar classes, Kote, you know, when you did some writing, I think the professor may be like, you know, there's some other stuff in here. Like maybe you're you you missed. Maybe maybe you should Matt Ray (19:24.083) I don't think self awareness is their strong point. Yeah. Brandon (19:25.606) Like you just have the big red pen and be like question mark, you know, you just like write down a few notes. So that's like maybe the strangest part of that whole manifesto. Coté (19:25.898) Yeah, j just like like You know it I think I think the comment

would be history is longer than five years is you know what what what you would think about it. So so let's go an i down an ideal path here of we all well I don't know about Matt and my daughter who hates AI, but we all benefit and like AI. and so in in in the same sense that we I I I'll say as a civilization, we have benefited from, let's say Matt Ray (19:53.419) Hated. Coté (20:02.461) Less than legal operations in the open source and software world. Right. Like I I think a lot of benefit has come from civilization of like if we had followed all the rules about like copyright and IP and software, we probably wouldn't be where we were at the moment, right? Like all sorts of benefits that we get. And then the the the using software for free and free as in beer from the open source world. And so it seems like I'm gonna I don't know if I believe this, but I'm gonna propose that that same freewheeling notion that you should be able to use whatever you want from the AI world is a good idea. Because over and over again, whenever you can basically I'm just gonna throw the word out there, you can pirate IP civilization advances, right? Like all the way back, I'm sure further from like Captain Drake just like stealing a bunch of stuff. through the centuries, everything's great when you have kind of like some controlled piracy in there and it allows people to innovate. So maybe what we should be supporting is like closer and closer to open weights and not having this like hold on proprietary stuff. Brandon (21:16.74) Hmm. Well, I don't know. Maybe it's a slightly different take is like, you know, it's kind of this is a very unsatisfying answer. It's like, well, what we typically do, just more observation, is like, well, we just kind of do everything and we just kind of see what happens. Now it'd be great if we had if we stepped back and maybe we thought about the consequences more, but to my knowledge, humanity, we're just not that good at it. We're not good at slowing down. So that's kind of where we stand today. It's like, well, I mean, you've got three or four different people, and we just went over them all making Coté (21:30.848) Yeah, Matt Ray (21:39.656) Ha ha Brandon (21:45.541) models slightly different ways and even more if we expand it, right? And it's like, I don't know. I I sit here and I say say to myself, I see, I can see everyone's point, but I'm also just like, well, I mean, I guess it's like a capitalist kind of like decision. It's like, we're just gonna see how the market plays out. Like do do, you know, do we prefer, you know, the proprietary models? Do they offer some benefits in a way that the open weight models or the open source open weight models don't? Or vice versa? Like will it actually be, and maybe it's a little bit like I don't know if it's even an appropriate analogy, but it's like, you know, Solaris and Linux, right? It's just like, well, I mean, there was, you know, there was a time Solaris was considered by far and away

the better operating system. That day is long gone, right? World has moved on to Linux and Solaris is a footnote in history. And it's like, I don't know if we should have gone back and like you kind of think about going back in time and being like, What like what would you have told Scott McNeely or Jonathan Swartz? You're like, Okay, don't open source it or open source it or maybe we should stop with Linux. I think it's the same kind of thing there. Remember like Coté (22:37.877) Mm. Brandon (22:41.07) really way back it was like, don't use the scary Linux. Like that's gonna like, whatever, you're gonna have some problems. That's that's gone. So I think I don't know. I mean it's I think it's unsatisfying because like we can see a lot you can see a lot of things going bad. Like that's the thing that makes it hard. Right. You're like it it could go bad. So we kind of do need to watch what's happening, but like the same time we probably have to let it play out. Coté (22:54.741) That's true, that's true. I th I I think Matt Ray (22:55.634) Yeah. Coté (23:02.057) That's true. Well well, one, you're reminding me that I was doing a consult with Sun once and Scott McNe McNeely walked by and he saw that I was some consultant and he said, What would you tell me? And I don't even remember what I said, but it was something stupid. And then he walked away. But two, I I think I think maybe to sum it all up, it's sort of like, well, all these companies just stole a bunch of copyrighted material to build up their models. And so it's like, you know, like What what is it? Live by copyright theft, die by copyright theft. Like it's pretty much like in in the milieu of the AI world and building models that you don't give a fuck about copyright. And so the the extension of that world is just like that's just the the like rules of the land that like lawyers d are not welcome. And instead, what what matters and and maybe this is maybe this is Matt Ray (23:34.538) Yeah. Matt Ray (23:53.384) And yeah. Coté (23:57.766) The innovation thing is instead of what matters is having a good product. And however you get to that good product, just like with Facebook, like all sorts of things, but the revealed preference of the world is they like Facebook and they like Instagram. And like however it got there, whatever it does, it's what people like. And like, so here we are. And so i i essentially in the AI world. You can steal whatever you want. It doesn't matter. You have a good product. However you got there is fine. And that's the whole situation. Brandon (24:34.98) Well, I think, you know, to your point, build on it would be something like as long as the consumer benefit, right, reaches a point that like enough people feel like they're getting, if you will, the quote unquote economic surplus. Like there's so much utility that we get from this that we collectively are sort of like we're good with it, right? Because there is, I mean, you know, we we kind of come back to this. It's like, I mean, all these things are just rules that we make up. Like tomorrow, if someone wanted to, like a court could say, like, listen. Matt Ray (24:35.316) Well an and and they've Coté (24:40.32) Yes. Brandon (25:04.218) We're we're ruling that like you violated copyright, like, I don't know, 40 billion times and the fine is whatever it is. And it's like and you could bankrupt the company, you could just put them out of business, be like, that's it, right? You're done. Because yeah, the the it's right, but I'm but I'm saying but you could go even further if it was just a flat out, you know, or like I can't remember it is there's that whole famous like thing where like the the companies are not liable for things that are posted by people, right? Coté (25:16.363) Define as like five trillion dollars. Matt Ray (25:16.756) Well th they already had a s they already had an author settlement that was like one point five billion or something. Brandon (25:33.147) So you could just take that away. You could just take this some of these things away and you could fundamentally reshape the internet. Now I think most people would be like, that's bad to your point. Could they be like, well, that's bad, because then all these sites would have to essentially go away. And whatever utility that we're getting from them would would do it. So I think that's why it's sort of like we have, I think that's where I think it's very unsatisfying. That's what I think that's what's happening though. We're kind of all watching. And no, and everyone has their own agenda, the government, political parties, each company. And we're just kind of all watching and trying to figure out like how's this going to play out? Coté (25:33.651) Mm. Right. Matt Ray (25:46.132) Yeah. Brandon (26:02.5) What makes the most sense? And can we somehow do it in a way that doesn't, you know, that we get the benefits without like all of this tremendous potential bad downside. Coté (26:12.467) It's true. Meanwhile, it se it seems like if you want to be involved in this market, you can just call up leather jacket guy and say, like, hey, can you talk to your iBankers and give me several millions of dollars to fund my my involvement here? We've got Matt Ray (26:13.674) Well that I Brandon (26:19.076) And yeah. Brandon (26:26.032) Well, I've been waiting all week. This I feel like this is the Matt Ray segment. I'm just gonna I will set it up in the clear out, let Matt go. But so this news was that what you can do now is they've can they a GPU is now a securitizable asset, which just basically means that all the six major investment banks, private credit, can now give you financing, right? To buy GPUs, and then you know you can do it. So, Matt Ray, you have been talking about this for a long time. Like, you know, if If the idea that like if the debt breaks containment, if debt gets outside of the actual hyperscalers and it just becomes like a a normal part of the economy, that's a huge danger. So I'm

I'm I'm very interested to get your take on what do you think of NVIDIA doing this and what do you think of this idea that GPU should be become an investable security? Matt Ray (27:18.308) you know, what what what is the downside of of tying, you know, billions of dollars up in something that loses all value in six years, right? You know. yeah, this is this is gonna go great. You know, I mean Exactly. You know, I mean, this is it's like everyone ha knows that this is a bad idea and they're like, sign me up. You know, this bubble, it's never gonna pop. It's just gonna get bigger and bigger. And this is like Coté (27:29.043) It's it's like the Furby market. Matt Ray (27:49.35) It it just boggles my mind how they're like, this is a securitizable asset. This is like telling people, you know, let's do credit default swaps on on home loans and by the way, we're burning down the homes after 10 years. I mean, that's what this is. Brandon (28:03.046) All right, well, let me take the other s I was prepared, Matt. I was trying to okay, I'll take the I'm gonna take the devil's advocate side just because I feel like you I felt like you would come in. Not a fan of this. So so my question is like the the best analogy I can see that people are making to is like jetliners, like large passenger planes, right? That those are quote unquote securitizable assets, right? You basically don't have, you know, whatever. A an airline can basically buy that on debt. And the reason that's okay is they Coté (28:12.532) Ha ha ha. Matt Ray (28:21.448) Yeah. Brandon (28:30.416) there's a belief that the the airplane always has a certain amount of value. And if it's that's fine. If Delta goes out of business, you sell it to United or or yeah, or who or whoever else. So and and the railroads is another one. And the case is this like, you know, it just requires so much capital to start these things, right? Like if you don't do it this way, like no one literally it just can't get built, right? Because it's just too expensive. So do you buy that at all? It's like listen, it's not like we're buying Coté (28:37.052) Like John Travolta. Matt Ray (28:40.681) Yes. Cornering the market. Brandon (29:00.426) one GPO. It's like when I have to buy like, you know, a data center that's the size of like, you know, a hundred acres GPUs, it's like I just like there's just not that many people have that mu that much money. And then of course we, I'll just put like we in the royal we want to like have more capacity to use our AI models. So if if that's the case, why why why can't I why why can't I just you know fund this with some debt? Matt Ray (29:19.305) I don't think we do. Matt Ray (29:27.037) Because somebody sometime has to pay for this and and the difference between, you know, a GPU and an airline and and an airplane is forty years. I mean, you know, the airplane is gonna make money for literally like they're in the air for forty years. And, you know, a GPU, you're There's no like secondary market. Travolta is not, you know, gonna buy, you know, millions of GPUs, you know, eight years after the fact. You know, it's like they're they're just, you know, not useful because they burn so much energy, because they aren't effective and you know, there's just no place to host these. You know, the the secondary market, there's no tertiary market, right? Brandon (30:02.79) But I mean, all right, let me just let me go to this way. So okay, anthropics is like let's just say at the moment. Anthropic is paying for all of it. So they take the lease, they say, I'm paying for all of this, and we have the revenue. And and the lender comes in and says, Okay, I see what you're doing here. I think you're gonna be worth a lot. I will loan you some money to do it. And they do that like a Carlo, and they look at it like, I think you're good for it, but if you're not, I'm gonna take the GPUs back. And I'm gonna be able to sell them. Just like I can go get the car and I can be like, I can go get my money back because I can get the car and I and the car is gonna be useful to someone else. So whether that be open AI, AWS, or many other people that could potentially want it. So I don't know, like why does that logic not apply here? It's like I can always just go get the GPUs and sell them to something else, someone else. Matt Ray (30:46.909) Because the the because when when you start looking at how they're financing these things, it's not like Facebook buys a data center. It's Facebook spends up a company that buys a data center, holds all the debt, and then they take the loans out and they like loan them some money and they get paid first and then when you know that the data center is full of GPUs that are out of date, it's tits up and it's not Facebook's problem. That's what's happening with this debt. Brandon (31:09.542) Right, but that's the whole point. I guess the whole point here is that like I is if you believe this part, the it's GPUs, that that group, that's still vr valuable. It's like it's fine. If the shell company goes out of business, I just go get to GPUs and I sell it to somebody else. Just like just like a car. Like it doesn't really matter who owns the car. Like as long as the car is financed in such a way that like the debt Coté (31:10.226) Mm. Matt Ray (31:29.479) You forgot that the car w was was blown up after five years though. Right? That that is a rapidly depreciating asset. Coté (31:32.628) There there's an interesting Brandon (31:33.102) Well, but I mean all cars eventually die. I mean right. But Coté (31:36.372) There's an interesting thing that follows here, which is that if if a GPU is an investment vehicle, it means that when you come out with a new GPU, it is not more valuable than the existing ones, or it's significantly less more valuable, so to speak. Because it if the scenario you said is so a a data center buys a bunch of GPUs and there's, I don't know, a ball of value right there. And if that company goes out of business,

you can take that ball of value and get pretty much the same amount. Like like a like it's n it's it's gonna be fine because you can just take that ball of value and you'll get it somewhere else. However, if we were to model that with like, let's let's bring I'm I'm imagining in my head we're doing this commentary and like Gelsinger is over there and he's just like very sad. Right. He's just he's just doing his head. So imagine this with like CPUs and other chips, and you're like Brandon (32:30.502) Right. Coté (32:33.52) No, these things have like like they're like have a l less shelf life than like white bread in in an oven. Like it's they're terrible. And so I I think what you're saying with the GPU thing is that they actually are really good and there's no reason to buy a new one for like what whatever the immortation year is. Matt Ray (32:42.462) Yeah. Brandon (32:52.592) Well, I would say it this way. No, no. The gents, okay. I again I'm just taking devils out of here, but I'm not sure what I believe. So I'll just, I'll just gonna I'll just say it. I think what they're saying is that that it's not so much the new GPUs aren't better, it's the fact that there's just limited supply, right? Like it's more like you only like, yes, the new thing is better, but no one can get that. So the old thing, right, the old thing is still very valuable because I still need to run it. Matt Ray (32:53.459) But that's not the case. Yeah. Coté (32:58.408) Yeah, yeah, of course, of course. Coté (33:07.866) Mm. Right, right, right, right. Coté (33:13.363) It's like oil. If you could use oil over and over again. Brandon (33:19.364) Yeah. So so I that's where that's where it's coming from. Is this and I guess the the case that they would make is that like, you know, at one point people were like, these are only going to be good for three years, right? Again, I'm just I'm just saying this. I don't have any data here. I'm not saying I know this for a fact. But now the case means being managed like actually the lifespan of these things is more like six years. And maybe some people even be like, it's even a little bit longer. And it's not that the new things aren't better. It's just like we we'll never we're we're far away at this moment from having Matt Ray (33:21.385) But Coté (33:38.29) Yeah, huh? That's fair. Brandon (33:47.192) Lots of the new things. It's not like you can just get them, right? Coté (33:48.424) Right. I I I I I like where you're going with this because it it it always seems like when we arrive at this kind of point to be all like big picture again and like dorm roomy. Like I think that law of supply and demand stuff doesn't actually work. Like in certain edge cases, because the the like supply and demand would say that like there is such huge demand for a whatever you call it, a a a a depreciating asset that like, you know, has a terrible shelf life. There like there's such a small amount of supply for it that they actually are valuable, even though they expire. And supply and demand would say, like, well, then over the course of like, I don't know, five years, you'll just have a glut of supply of them. But and yet here we are with we've got like Samsung and TSMC, and that's it. And and nothing else has like come about to like meet the supply, which brings us this like. Brandon (34:30.405) Right. Matt Ray (34:33.79) Well Coté (34:41.95) You've got such a small amount of supply that like something from that's like three or four years old is gonna be extremely valuable. Whereas like the X eighty six chips are not. Matt Ray (34:50.099) But it's not extremely valuable. yeah. I mean the the difference is like in in three years, the the density, it's not the GPU that cost money, it's the the power, it's the electricity, it's it's it's the the the the you the recurring expense of keeping that thing powered and and running. Coté (35:02.036) Mm. Coté (35:13.64) It's like running it is what's expensive. Well, that's interesting. Matt Ray (35:16.007) Yeah. And and and so what happens is, you know, you've got you've you've got these GPUs and let's just say like, you know, today's bleeding edge is is a hundred percent density, right? And you know, eighteen months ago is fifty percent density. And so if you're running a hundred percent density and it it's you know you you can make a hundred twenty dollars for every hundred dollars you spend on power, well you're still spending a hundred dollars in power to run that fifty percent density and now you're running as a loss. Coté (35:45.278) And and then I could also see a case where like I'm sure you do this, Brandon, but I'm always telling my AI, use the local LLM, right? Like do the dumb stuff. Like like on and and exactly. Right, right. And and so maybe like if we had a bunch of like local LLM, like old like GPUs, that would actually satisfy a huge amount of compute. And so they are still valuable in that sense. That you don't just throw them away. Right, right. Brandon (35:53.26) That's right. Use the G Pun. Use the G button. Matt Ray (35:54.139) Mm-hmm. But the and and but th those old GPUs running the local LM aren't making money. Yeah. Matt Ray (36:12.905) But the energy costs are what's killing you. Yeah. It's Coté (36:15.335) Right, right, right. And and so like Brandon (36:16.878) See, I understand the energy. I don't understand that argument 'cause it's like no, it's like the GP is mm-hmm. Matt Ray (36:20.186) Like like when when you build a data center, your number one cost isn't the CPUs in it. It's the power. Like over time, power just demolishes all the the upfront cost of of capital. Brandon (36:25.776) Okay, no argument. Okay. Right, but I I think it has kind of come back like supply and demand. It's like, listen, I have to serve this amount of customers that want to use my LLM. Like, okay, that's that's and I'm gonna charge them. I can charge them whatever my API, you know, cost is, and I will build that in. And it's like, I need I

need some way to serve them. And it's like I can't get the new stuff because it's not available. We will use the old stuff. And therefore, it's like we will build into our pricing such that we cover the power and we cover the use of the GPUs. Matt Ray (36:50.077) Mm-hmm. Brandon (36:58.906) You know, it's Matt Ray (36:59.1) But they're they're not putting that that pricing in though. That's the other thing. It's like for every you know, for every dollar anthropic makes, they spend fifteen. Brandon (37:05.318) But that's back to like, but again, like this comes back to the financing side. It's like that's fine. That's the anthropic decision, right? And it's like, if I'm just a lender, I'm just like, fine, listen, anthropic wants it. If they charge so so little they go out of business, it's fine. I'll just sell the same data center to the next person in line. That's I think that's the key argument they're making. For sure. I think absolutely, I think the the Matt Ray (37:22.61) But but it but if they're all if all the dominoes fall over at the same time, you know, you're left with like mountains of of trillions of debt and everyone's holding the bag and you know, the government steps in to r you know, f fix it all and and the taxpayers pay the bill. That's what's gonna happen. Brandon (37:30.832) Right. Brandon (37:38.138) Right. And think that's the key thing. That's maybe the bearish side of it. Is like if you think all AI demand will suddenly end, everyone's gonna Okay. Matt Ray (37:45.768) I don't think it's gonna go away. I don't think the pricing supports the build out. Coté (37:50.146) Mm. Well th that's that's something I wanted to ask y'all about. Is so I logged into logged into. Like I'm at some university basement hooking up to a vax or something. But I I was looking at my usage in in Claude and I I was very distressed. And it said current usage, $495. And so I I of course took a screenshot, like any sane person would do, and I sent it to it and I said, What is up with this? 'Cause I I don't think I own because I'm on this plan. I don't know this. And it's like, sure, sure. This just showing you like what you would owe if you were paying for tokens. Matt Ray (38:03.036) Good times. Coté (38:27.281) Instead of like Exactly, exactly. And so there's a lot going on there. One, why are you showing this to me? Like, do you want me to like appreciate you more or something? And and Brandon (38:27.31) Right, API. If you use the API pricing. Matt Ray (38:36.156) Mm-hmm. Matt Ray (38:40.05) They're they're socializing you for when they jack up your prices. Coté (38:42.053) I and so the the exactly as you were saying, Matt, so it makes me think at some point I am gonna have to pay four hundred and ninety-five dollars for three hours of usage. And so obviously things are running at a I don't know, I'm getting some benefit, as it were. And it does it it does seem a little so is there a game going on here? How d how do you feel about this that on on our side as the consumers, we have to believe Matt Ray (39:01.426) So Coté (39:10.919) That technology will advance and things will get better such that we don't actually have to pay four hundred ninety-five dollars to like, I don't know, recategorize our blog posts or whatever it is I was doing. And it it maybe that that won't happen. I'm not sure. There's there's some weird pricing dynamics going there. Matt Ray (39:19.217) Ha ha. Brandon (39:32.101) Well, I think it just I I look at it more simply as sort of like the there are people paying that today. Those are the people that want to they can't use the pro plan because they blow through it. So those are the people that are using the API plan, right? So like that's where the real quote unquote money is today. So those customers do exist today. And I don't know how many, but that's obviously they're paying more. And then there's y you know, you and I, Cloud Pro, Claude, whichever plan you have. It's like that's clearly a lost leader for many people, right? That's why like you run out of it. And I think their and I think that their case is just sort of like this is just still user acquisition. We don't mind running this at a loss because we want everyone to use our model and love it. And eventually there'll be pricing power and we'll either raise prices or what I've already mentioned on the show, I know I do, is or you drop back a model. You go from Opus to Sonnet. You're just like, I can you know or and you become or what you were talking about, Kote, you just sort of like I can use this other LLM for like really simple things. And then if I want to like rewrite something that's complicated. Matt Ray (40:15.997) Mm-hmm. Brandon (40:26.394) Then I, you know, use my quote unquote good model. And I I think, I think people are already kind of doing that today, right? And that's where you have stuff like open router, you know, a bunch of other solutions out there that do that. So I I don't know. It seems it all seems like reasonably logical to me. I mean, you know, but again, like to Matt's point, like I think the the idea that, like, I mean, the ways it could bl blow up is simply like people just don't, you know, businesses, that's more important. Businesses ultimately give up on the ROI. They just like. We've done a lot of stuff for a year. We didn't see a lot of value. And there's that's fine. That I think that's that is a very that's a valid that's a valid argument. It's fine. It's a valid argument. It's just not there. Like if you're bet I guess it's just one of these things. Like if you're betting today, right? You only have two choices. You have it's almost like you have the AWS, Apple, Microsoft door, Microsoft door, and you have the anthropic opening eye door. And it's like one is like, Matt Ray (40:58.716) Well th they they just say like we're not willing to double our spend. We're not willing to triple, quadruple, and that will never sustain yeah, that will never sustain what they need.

Brandon (41:21.444) Models aren't gonna be that big a deal. We're not gonna spend all that money. We're not gonna try to bankrupt ourselves, but you're gonna miss out if the model is the most important thing. And then door number two is like, we're all in on the building the greatest models ever. We're gonna run at a loss. We're gonna do it at the the most, the highest prices ever seen in this economy ever. We're playing trillion dollar broker and we think it's gonna pay off. So it's like you just got to decide like where where are you betting here? And then of course to the Matt's points like obviously if it goes bad in a huge way, then like, yeah, like all this financing. It could be a nightmare. So I don't know. We'll see. Coté (41:52.689) So so th th this is making me think. So there was a story about like Datadog's biggest customer w of AI stuff. I I don't quite understand the details. They were like, Mm, we don't wanna pay that. And and so they were kind of Brandon (42:03.354) Which is basically probably open AI. It's or or I guess it could be. It's either it's really only there's only two people. It's either open AI or it's anthropic. Coté (42:06.407) Right. Coté (42:11.037) And and and and so also kind of related, you you hear all these companies who are like, we're spending too much money on tokens. We gotta tamp down on it. And I I guess are these companies like being too cheap about it? Should they be spending all of this money on tokens? Or should they actually be thinking, we don't need the AI stuff? Brandon (42:35.782) I don't know. I think this is the this is the question, like, you know, this is the whole we should call this the forward deployed question, right? It's like it's clear, like we've had on the show, we've talked about, at least I've talked about, and I feel like AI is incredibly productive. It's helped me do a bunch of stuff I could never do before. I feel like I see a lot of potential in it, right? But clearly the the idea behind the forward deployed thing is like, well, you know, broad-based enterprises, the people have money, are interested but unsure what to do. Right. At least unsure that how this turns into real savings or real new revenue. And that's why I think you're seeing like the massive group of four deployed everything being sent out there to be like, hey, let's help these people ultimately realize the value of AI. And now that's like that's and this is really Jamad's question. So, like, either those people go out and do that, right? And actually find ways to it, and then you're gonna see the flywheel start to spin. It's gonna be very profitable. Or people go out there and we'll just pick on like a company like SIBOL, like way back in the day. People spend like, you know, whatever, $100 million on a SIBOL. Like we all worked at a company that had some like unbelievably expensive SIBL deployment. Never worked. I as far as I can tell, there was zero ROI and people in that company is just out of business, right? Because it's just like it never could work. So, you know, I don't know. Like I I, you know, to me, whenever I touch AI and use it, I'm just like, this is incredible. But then, you know, when you look out in the world, you're like and you talk to normal people and it's like, well, I don't know. A lot of people don't want to change, right? A lot of people are like, I'm pretty much happy doing what I'm doing. Matt Ray (44:06.661) Yeah. I mean there are a lot of Coté (44:07.645) Well, well maybe maybe maybe maybe another way to put put the idea is you we you you very rarely hear about people who are successful and enjoy using something, right? So the the the stories of someone like we spent too much money on it, it's terrible, we're canceling it, those always come out. And so another way of looking at it might be that like any company who's not complaining about spending too much money, they might be benefiting from spending money on it. because they're not complaining about it. I don't know, maybe like forty percent of them Matt Ray (44:40.785) But where where are are the like enterprise companies that are like, this is really having a huge ROI? Yeah, where is where's the Netflix of of AI, as Brandon said before, right? Like where yeah, where are where are the poster children for like, you know, this is making everything better? What we're seeing is like I mean, my you know, things are things are great in the security world because everyone has to spend a lot more on security now. Coté (44:51.855) Yeah, yeah. Matt Ray (45:10.567) I mean, who's that good for, really? you know, I it's like the things I see are like a lot of the negative aspects of it. And you know, we're we're using AI to fight AI. Brandon (45:23.75) I would say this way. I would the way I would say this, like at the enterprise level, I believe that is an open question. I think it's very hard to either find use cases that that's happening and people are willing to talk about, right? That's very difficult. I think at the individual level, it's very easy to find to see what's going on. Like, you know, just pick a lot of like statistics, like something like you know, like app store sedition submissions to apples like up like tenfold. A whole bunch of people are building apps that never built. Apps before. Now, are those apps good? Are those apps going to make money? Are they just toy apps? Don't know. and I'm sure everyone listening to this call, like everybody has done a bunch of different things. Summarize email, write reports, you know, endless amount. Like, I could even talk about like you know, producing this podcast. Millions of different skills have been created that I think are definitely creating productivity for the individual. But that does not show up as like a line item that like a CFO is gonna say. And the only stuff that you know, the biggest ones we've seen. I'm always very skeptical of. I think AWS came or Amazon came out and said something like we upgraded

everything, like some Java version. They updated the Java version to for like, I don't know, tens of thousands of machines. And it was a huge savings. So you could look at that one, but I'm always like, I don't know. That seemed like that one didn't feel necessarily like a big one. Or the one we talked about last week, you know, rewriting all of the JavaScript. What was the JavaScript framework? I forgot the name of it. button, rewriting that entire thing in Rust in two weeks. Matt Ray (46:38.885) Well they Matt Ray (46:46.149) Note. or or bun. Brandon (46:51.044) Like that's amazing. Like I mean, if that's what's about to happen, it's but insane. Like just just take it at the numbers. Just but just take that at the numbers. If it costs $200,000 to migrate that, it's like that's a no-brainer. There are millions of applications what ready to go. If that could be reproduced, people will do it. I think that's for sure. Now, huge, huge unknown if anyone outside of an anthropic engineer could do it by themselves, but if it could be. Matt Ray (46:53.945) It it is amazing, but who paid for it? Right? That's the thing. It's like w yeah, we we're Brandon (47:20.538) You have like, I mean, a tremendous amount of money is going to be made, right? The amount of old software sitting out there that needs to be like migrated is almost endless. Coté (47:24.361) Well well I I Coté (47:29.436) I I think you you've hit on something that I I think about all the time, which is I don't maybe maybe a lot of technology is like this, but for the first I don't know, just to throw out a number, five or ten years of a new technology, only individuals benefit from it and organizations don't know how to benefit from it. Which so if you gave all of the employees of an organization the kind of level of usage that I don't know, you have Brandon, with using AI, they would be much better at their jobs. But I think the theory would be that that that that value, that productivity doesn't bubble up to the organization because I don't know why, but it's a different type of money extraction to like bring it up to like hundreds of thousands of people. Whereas like the individual, what they can do is they can do their work faster and then like have a longer lunch. Like they they're they're not like really incented to like do something more with their time. And so in Brandon (48:32.186) Yeah, no, you need somebody, you know, it's kind of like, you know, you had a good interview this week with your your a former colleague talking about like, you know, implementing change. And it's like you could take something, I just cause it's at least here in the United States, it's back to school time and like my wife works in education. And one of the things I watch all these educators do when they go back to school, the first two weeks, there's a whole team of administrators and counselors that work nothing on scheduling, getting all the kids the right schedule, getting them all in the right rooms. And it's nothing to do with kids. I mean, interacting with kids. It's just Getting their names and their requests. And I like, I will look at this and I'm like, I feel like this is such a computer science problem. Like this is like this is what computer science is all about, like complicated scheduling and looking at all these things. And I'm just always like baffled, like when they're doing it. They kind of sometimes I'll look over the shoulder and so very paper based, very manual based. And like I'll sometimes I don't do this anymore, but sometimes I'll be like, you there's some things you could do and blah, blah, blah. And it's like, but like, really? I mean, kind of to answer your question directly, it's like, Coté (49:03.911) It's toil. Brandon (49:30.466) No one thinks it's a problem. There's no one at the top that's just like, Man, we cannot spend all this time doing scheduling this way. In fact, it's the opposite. That's like, that is what they do. They report to work two weeks before the schools are coming where the kids will be back and they do all of this stuff. So, I mean, to your your thing about like there had to be some forcing function or something. Like, like suddenly there were no administrators to do any of this work. Maybe someone would write some software then. But like Coté (49:33.831) Mm. Brandon (49:57.849) It's there. Like I just look at it. That's not even necessarily an AI problem. That could have been probably been solved by other ways. But I think like it's very suited to AI. And it's just like, but I get it. Like people like and I just look at it and I'm just like, there's no forcing function that anybody in the schools, like, we this is fundamentally wrong. We wanted to like I look at it and I'm just like, I I'm just like it's fun, but it's not, you it's not ingrained. And that's what I think all these enterprises are going. It's like, like, you know, and this is the same DevOps talk you've given a million times or, you know, lots of other things. It's like, Coté (50:13.699) Yeah, yeah. Brandon (50:27.502) Unless there's a huge forcing function where people are like, I gotta really do this a different way, people are probably pretty happy. Like I'll just, you know, I'll just keep using the VMs the way I've been using. And like it's not been things work. Coté (50:35.643) I yeah, th th I mean this is the wrong phrase for it, but there's some sort of like economic version of shadow IT, which is the the individuals are getting value out of using this tool, but the organization is not getting the value. And so they get to bring in the tools they want to use and it's valuable for them, but it doesn't really like bubble up towards the organization. Brandon (50:58.79) Yeah. And when there's no one above. And I you'll even like look at this podcast. It was a great example. Like, you know, like we're all highly knowledgeable in this area, at least we think we are. But like, you know, Matt's just like listed what fifty reasons

he thinks this isn't gonna work. Can you imagine going into someone that's not really interested in like working on this and being like, Look, I I mean, they're just be I just got off hearing Matt from Software Defined Talk. He says they're gonna go out of business, these data centers use too much power. They're not I mean, like you could just listen to this podcast at the end, the person could be like, you know what we should do? we like. Matt Ray (51:23.386) This is my fault, yes. Coté (51:24.859) Ha ha ha. Brandon (51:26.01) You know what we need to do is we need to let the administrators go over there and do their work and their paper and do the emails and like why I mean, Matt, did you go listen to Matt, right? I mean, and like we're pretty like forward thinking, right? So it's like, imagine if you're just like a regular person. You're like, like, why am I gonna do all of this? Like what I mean, you're pretty scared of it, I think, at this point, you know? Coté (51:45.679) You you know something that people always ask, why am I gonna do all of this? It's bureaucracy, Brandon. Do we have any this episode? Brandon (51:51.739) That's right. The bureaucracy of this episode is one thing that's not been automated, that's still in your real life, is your need for stickers. So if you would like a sticker, conference season's coming up. I Kote's gonna talk about some conferences in a while. Everyone needs a sticker. Whether you you're you like AI or you don't like AI, you still need a sticker. And the way you get it is you send your postal address to stickers that software define talk.com. I will be happy to send you a sticker anywhere in the world. Coté (52:19.538) Well, there's a lot of conferences coming up. So I'm going to tell you right off the bat, what you got to do is go to software defined talk.com slash five eight five, which is an interesting palindrome. There must be a word for a number that's a palindrome. and there's all sorts of things you can look there, but there's one I want to talk about in particular, and that is We Are Developers North America. Over there on September 23rd, through 25th, we have a discount code if you want to go up to it, which is dev pod50. And there might still be 25 tickets left if you go to software defined talk.com slash five eight five. Now I went to the Berlin We Are Developers event. It was fantastic. As I said last week, there was a huge variety of topics there, all sorts of people. And you know, speaking of individuals benefiting from things, if you wanted to get a good cup of coffee, there were multiple places you could go. You could get I don't know, not so great coffee, or you could go talk with someone who had a booth there, or you could go to a food truck and pay a small amount. I'm remembering we went to one booth, my friend Orin and I, who both spoke there, and there was a very happy, like German guy who just served us coffee and he's like, go over there for your coffee. He was very cheery. Now,

when it comes to the actual content, like I was saying, what's interesting about them is they have like they're kind of like an online, I don't know, magazine. source content place. So you'll see when you go there, they've got all sorts of videos referencing like the online presence they have, and then also recordings of the podcast they have, things like that. But I mean overall what's going on there is that if there's some sort of interest that you have in doing development, across all the different languages, the deployment stuff, things like that, there's probably a talk involved there. that they have. So you should go check them out. It's gonna be over in San Jose September 23rd to 25th. And again, you can get it you can get a discount if you go to softwaredefined talk.com slash five eight five. Now, there are many DevOps days that you can check out most of them oddly enough in September. It's almost as if the DevOps days people weren't coordinating to spread it out. Or maybe there's something about September that they really like. Brandon (54:33.286) Yeah. Brandon (54:37.796) It's back to school, it's back to DevOps days. You it's it's what it is. It's the season. Coté (54:41.784) Yeah, yeah. It's it it it like everyone's kind of like come off the summer like relaxation and like we gotta nail it here because pretty soon we're gonna get into the holidays. So we've got to put it in here somewhere. So you've got the DevOps Days Graz, which is September fourth to fifth, and then also in the Rockies, which is twenty second to twenty third, and in Dallas, twenty eighth to twenty ninth. I used to go to that one quite a bit. It was nice. And then Vil News, September thirtieth to October first. Matt Ray (54:42.073) There you go. Coté (55:10.68) I'm gonna be next week, I saw in Riga Latvia at AI Connect giving a talk. And then I'll also be at the Cloud Foundry Summit, which is September 24th. Sorry, September 21st and 22nd in Heidelberg. that'll be nice. And then let's see, I'm also gonna be at the DevOps Days Istanbul October 24th. And then, of course, there is the VMware user group in Orlando, which is gonna be October 20th and 22nd. And then when you're done with that, what you're gonna do is you're gonna take a break until November nineteenth, maybe November eighteenth, because you're gonna fly to Denmark, to Copenhagen, and you're gonna see me talk at Cloud Native Denmark. And then finally, there is scale twenty-four X, that's lowercase X, in Pasadena, California, April first to fourth. So lots of conferences you can plan to go to. It should be a good time. Now Brandon (56:06.022) You can see the world too. I mean, you go to all those places, you will that would be you probably will have met your frequent flyer quota for the year. Matt Ray (56:08.395) Yeah, wow, that's it's quite the journey. Coté (56:12.891) That's true. And as as the as the the Istanbul people like to say, you can be in Europe and in Asia. Yeah. So so you can you can span out

there. Now, so far you can't go see Matt Ray. There there's no conferences there, but maybe we'll find one. Brandon (56:19.911) I love it. That's nice. Nice. Brandon (56:28.42) Yet. Yet. There's gotta be something happening at Sydney. Matt Ray (56:30.777) Coté (56:31.79) Yeah, eventually. Well, speaking of things happening in Sydney, Matt Ray, what do have to recommend this episode? Matt Ray (56:34.33) Eventually. Matt Ray (56:38.979) yeah. So so last week I I had a bit of an interruption. so I didn't get to talk about my picks. and and so we're we're gonna pull them forward a week or back a week. I don't know which way we're g going. the first one I I wanted to talk about was nearby wiki. It's a mashup between open street maps and Wikipedia. And so when you go and p drop in an address, it'll show you any articles about things near you, which is like I it's kinda fascinating. I mean, you know, usually it's like, you know, there's this, you know, this suburb over here, this neighborhood over there. But, you know, sometimes there's interesting articles about like a park or some monument you didn't know was there. So it's kinda cool just to like wander around and explore. It's almost like a an Atlas Obscura for your for your neighborhood. And so I thought that was cool. And then continuing th with the theme of maps, there's a website called Reading Maps that plots books against maps so you can see like where things probably took place and you know whatever book you're looking at and and so a good example is is the Odyssey. yeah well there are historical places that they're like this is probably what they were talking about when he was here or here or there. And so you know you can track the map and be like this is where he was with Cersei and you know Coté (57:40.678) Mm. Matt Ray (58:01.111) Anyway, I thought those were cool. We're sharing. you know, at least clicking around on. and then my my fresh pick for this week is the new infrastructure frontiers podcast. so I'm not a total AI hater. I I'm a I'm I think I'm a finance hater. Coté (58:18.991) Yeah. Matt Ray (58:20.942) I'm a bad economics hater. I'm you know getting stuck with the bill hater. So anyway, the infrastructure frontiers is it's from Adam Jacob and the the folks over at the the Swamp AI infrastructure coding platform that they're building, kind of you know System Initiative 2.0. And these guys are all on the the bleeding bleeding edge of infrastructure and software development. And so it's it's just kind of interesting because it's way in the weeds on things that I you know, I don't have the bandwidth for anymore. and just don't really have the the reason to do. But to hear their takes and and kind of talk through like, you know, there's this new startup that does this. And like, yeah, I'm just gonna code my own version. Like, okay. And, you know, talking about, you know, token consumption and maximizing this and building your own frameworks. And it's just like, yeah, it's it's you

know, it's an interesting listen. Coté (59:24.35) You know, I think I think I always liked the McLaughlin group, despite like what we found out about John McLaughlin later on. It seemed like a a very entertaining, great show that was informative. And I think I I don't know who's gonna be on it, but I think we could have a a a sort of tech McLaughlin group. And I definitely know Adam is gonna Adam Jacobs is gonna be on it. And I'm gonna do whatever it takes to get Luke Cannese on it. And we're just gonna fill out the rest of the panel. And I think every week we got a 30 minute episode. of these people and I'm gonna say two other people and a moderator. And I think I think it would be amazing to to have these two people talking about not like a little bit of the technology, but more of just like the business and and what's going on in the tech world. So if anyone has ideas for other people, I like I I don't really want to be on it. That's a lot of work. But we gotta have we gotta have the John McLaughlin and then you need at least two other people and maybe a rotating person that comes in and out. And I think you could have some great commentary. going there. Now speaking of great commentary, what's your commentary on your recommendation for this week, Brandon? Brandon (01:00:28.43) My recommendation is a show called Furious on Hulu. It's very dark, very adult themes, all the trigger warnings, but I really like it. So it's I think there are five episodes out now. I think it concludes next week. I think there's one more season finale, maybe series finale. I'm not sure what it is. But if you like things that are kind of dark and kind of interesting but done told in a very interesting way, check out Furious. But again, not for the kids. Definitely nothing, nothing nothing, this is not a family friendly program. Coté (01:00:56.633) Well, my my recommendation, one, I have a little bit as they say on the political gab fest and log rolling. I fixed up my incomprehensible media dot com site where I publish all of my AI generated summaries. And I've more or less automated when I when I summarize stuff. It was I I had it set up there, but it was not very it was clunky. But now if you go there, you can see like pretty much all the stuff I summarize. I've got a I've gotta say, I've got a great way of summarizing YouTube stuff where it like Does the transcript, finds excerpts from it, really finds out like if you wanted to watch a YouTube video, but you didn't actually want to watch it, you just wanted what was in it, it's extracts that for you. So check that out if you're interested. And then also my other recommendation is I've had an Elgato teleprompter for a while, the tiny one, not the giant one. And I think it's worth it. It's good because you can like look at the camera and you can see people you're talking with. What hasn't really caught up is if you're using if you use like Riverside, Zoom or StreamYard, when you actually move the browser to the small

screen on the teleprompter, they don't know how to resize themselves. So you've got tiny little people that you're looking at. But it's all right. Now, what's also nice is if you actually do some like recordings of sort of let's say corporate videos where you've got a script you want to read through, through, there's this app called Teleprompter Pro, which is not cheap. But it's also not that expensive. And it's really good at just like running through teleprompter scripts for you and letting you you control it. And it doesn't have an MCP server, but it has a web UI on your LAN network. So I know you're thinking this right now, Brandon. You can log into Claude and you can be like, Let me give you an IP address. Can you run this for me? And the answer is, yeah, no problem. Brandon (01:02:48.851) I love it. I love this. All right, we'll have to try this out. All right. Coté (01:02:50.959) Yeah. And and so like you know, essentially anything that has like a little web server on your local LAN is now has an MCP server, more or less. Not really. But, you know, it's got an API. It can be used. So it's it's pretty good. You can control like the speed which things go and anything. But yeah, if you've ever been interested in buying one of these Elgato like teleprompter things, I don't know. It's fine. It could be better. It ha it it it it has that feel of like This is almost like a insulated coffee mug that would actually keep my coffee warm for three days, but a day and a half is fine. Like that's kind of the quality that it's at. So it works well. Speaking speaking of quality, the things are at, you've listened to something that will keep your coffee warm for a fortnight, which is two weeks more or less. and that is software-defined talk. Now, if you want to find the show notes for this episode, all of the great conferences we talked about, including We Are Developers, where you might be able to get one of the free tickets to go to go to We Are Developers in San Jose, September 23rd to 25th. You can go to software defined talk.com slash five eight five and find that. We also have a Slack channel where you talk about all sorts of stuff during the week, some of which we talk about during the the show here. but you can find the Slack channel to join there at software defined talk. com slash five eight five and with that we'll see everyone next time. Bye bye. Matt Ray (01:04:18.82) Brandon (01:04:21.742) All right. Great episode as always. The question I have is, you know, what breaks during the recording. This week what broke was the YouTube. It didn't I don't know. It just says we we've got a problem. So I was says something went wrong. That's very helpful, huh? All right. So try again. I'm gonna try again to see if it works. Well, can't thank anyone for watching on YouTube. At least not yet. Half to offload the video. Looks like that's broken. So we're on Twitch though. Coté (01:04:46.085) But we're on Twitch. Does anyone use Twitch anymore? Like na what is up with that? Is that a covet company? Brandon (01:04:50.63) Twitch has really been our favorite. It it's the one that's really worked the most out of actual live streaming. So Coté (01:04:54.863) Huh. Yeah. Matt Ray (01:04:55.968) Yeah. A and and they just changed their licensing so Amazon can use anything streamed on Twitch to train their AIs. Coté (01:05:02.49) makes sense. Yeah. Brandon (01:05:03.866) Finally, even better. Finally, we we're gonna be inside the the the AI. We can talk to ourselves. It's just gonna be great. So well, we'll see. I don't know. Cote, I just I had one question on your your recommendation here. So do you put like an iPad at the bottom? Like what's the thing that projects up? Or is that like okay. Coté (01:05:20.781) No, it has its own screen. That's like it ha it it it has a screen and then and then like it has a I don't know, one way mirror thing that, you know, leans this way and so the screen I guess is in reverse and then it projects onto the mirror and you look at the mirror and behind the mirror I I have an iPhone behind it and so like that y is using the continuity thing. And then there you go. Brandon (01:05:30.454) okay. Brandon (01:05:37.156) And then what camera? Do you put your own camera behind this thing? Or okay. Brandon (01:05:45.528) Interesting. All right. Well, I don't know. May I'll have to this seems like something I'll waste a lot of time on. I'll probably get one and spend like weeks trying to make it work. Coté (01:05:51.619) Well, as as as you know, this is why I bought the Elgato four K piece of shit, right? Which was not a good camera, because it seemed like it would, you know, fit in there well and do stuff, but it actually doesn't work well. So, you know, the answer is always just get an iPhone and hook up continuity camera and use that and you'll be fine. Brandon (01:05:56.452) Mm-hmm. Yeah. Brandon (01:06:10.842) I like it. All right. Well, thanks for everyone that watched us live on the Twitch stream. And if you watch this on the video upload later on YouTube, appreciate it. And then who knows, maybe next week I will figure out why we couldn't connect to YouTube. We'll never know. Never know. We'll see what happens. Coté (01:06:24.197) That's right. Yeah. Like and subscribe. Matt Ray (01:06:26.851) Magic. Brandon (01:06:28.526) All right. Talk to you later.