Bản chép lời đầy đủ của Episode 586: Watermark Away, Baby

like it just time goes by and it's just like, Yep, we all use it. And I think I mentioned on a previous episode that I think they popularized it for paying. I think Coté (05:06.283) Vision Pro. Brandon (05:26.961) They popularized it, I think, for metro and subway rides. I think that is very, very common. And at least in the United States, they've now popularized it. Not popularized, they're trying to push the US identification through. And then you said passports. So, so it is, it's sort of like as it just it grows and grows and grows. So I think Jennifer, you know, we should we should have like a small place at LegacyConf for her, right? She should come out and just be like, here's how it's done, people. Here's how you take old boring system that everyone's frustrated by. Coté (05:29.739) Ex yeah, yeah, exactly. Coté (05:39.297) Yeah, yeah. Brandon (05:55.707) And here's how you make it super popular. And and of course I'm sure she made like a billion dollars in personal equity doing that. Coté (05:57.57) Yeah. And you know, you you've you've you filled in some good analysis always from from my my my screeds. And one, yes, I'm talking about NFC and being in the payment network and all of that. And then like I remember my old Dell friend Mark Cathcart. I remember he was he was a CTO at Dell. you know, I don't know why he he was a CTO at Dell. He's a great great guy. And I remember one day he walked over and he's like, he had this great British accent, which I'm not gonna impersonate, but he's like I've I've figuring out what to do with NFC with servers. And he showed me like his his phone and he was like, I don't and this is maybe in twenty ten or something. And we were both kind of like, I don't know what you would do with that. And he's like, Yeah, you know, maybe you you can walk up to a server and touch your phone to it and it'll identify itself and tell you things. And you know, just like like so yeah, so NFC has been around for a long time, which gets to the second point, which I think is and you know, not a not a unique insight for us, but I think Brandon (06:50.279) Yeah, tell you what it is, right? Coté (07:00.149) I think business wise and strategy wise, Apple's specialty is bringing things to market and a broad market, right? Like most of notoriously, they haven't really innovated that much stuff from scratch, right? Like their whole operating system, like all sorts of stuff. As I think Aaron was saying on a podcast you won recently, they use tons of open source and on and on and on, right? Obviously, they didn't invent the mobile phone or the tablet. Or any of these things. So, but what they're really good at is, you know, here we are. Now we all use it. And it's Apple stuff. So special sauce. They're they're they're doing it. Good, good stuff. Now speaking of special sauce, I think someone has been trying to inject their special sauce into the created content that the robots are making. We got we got lots of conversation over the past week. Well, at least I had to use the disclaimer with someone I was like, everyone is saying, and I was like, Well, Brandon (07:51.515) Yep. Coté (07:59.98) I mean, hold on. Not everyone, like the four people who talk about Apple stuff that I read. But you know, there's there's much conversation about anthropic putting in textual watermarks. And maybe, maybe you've done the research to see how this actually works and you can fill in the details. But it it seems well, maybe you should do this. It's my understanding is when it generates text, it's trying to do something favorable for being. Brandon (08:13.777) Yeah. Coté (08:29.809) figure it out as AI generated later, which I think AI text generation already does that because you usually know when it's written by an AI. And then I don't know, like it doesn't make sense that there would be invisible Unicode characters because then you would just give it yeah, that seems like nonsense. Brandon (08:38.769) Right. Brandon (08:44.293) No, no, no. None of that. It's all mathematics. Yeah. I'll give you the the quick t ten set explanation. It's basically when when AI is generating the next word or the next token, if you've for all those that's done their Claude Code training or probably any of the AIs, it's like what you can do when for the next co token you can say temperature. And temperature is just sort of like how creative do you want to be? Do you wanna like always do you want to be basically if the word is like, you know, good, say good. Coté (09:05.463) Yeah, yeah. Brandon (09:12.989) But if it, you if you could also say, well, don't just say good all the time. Something like sometimes say exceptional, sometimes say, you know, whatever you want. And you you control that through what they call the temperature. So that's just telling the AIs like, you know, be more like took take all the possible words and if you will, distribute the probabilities more even across them and pick one, right? And that's basically what they're doing. And so and for specific words that will be unnoticeable to you. Coté (09:33.877) Mm. Brandon (09:38.0) Right. It's basically gonna use a variable or a setting for the temperature. Then AI can figure out as like, I see. The temperature setting for all of these words is whatever X. So I know this was an AI thing, right? So that's basically how it does. That's why there's nothing to do with the text. So it's all just basically the math. And so they're just controlling it. And they're controlling it for like, as they say, it will be kind of undetectable to the human. So we want like you reading it won't be to be like, I see. Every, you know, fifth word, the temperature's Coté (09:51.637) Okay. Yeah, yeah, yeah, yeah. Coté (10:04.609) And that's Brandon (10:07.001) is I can tell that every fifth word the temperature is set to like a value of point six instead of point eight. Like you won't know that. Coté (10:11.699) Right, right, right. So so t to to rephrase it, if I had

if I had three sentences that were a hundred percent Claude generated, Claude knows what its default temperature settings are. So it can look at that text and almost kind of reprocess is the wrong word, but reprocess it and be like, yeah, these are the words that at our default temperatures would have been chosen here. So so there's a point nine, you know, there's a ninety percent chance that you know a human that that we Brandon (10:16.882) Mm-hmm. Brandon (10:33.607) Right. Coté (10:40.801) Let me phrase this nuancely, that an AI would have written this. Now maybe there's some wacky human who would have written this too, but this is the way that an AI would have written this this stuff. Brandon (10:49.115) Right. But the way they do that is they that's why they they say they only quote watermark, it has to be a certain length, right? And then it's like, sure. Like it's one of these possibilities, sure, it's always possible a human being would have written the same thing. But like once you get beyond a few words, it's like statistically very unlikely that someone would be like Yeah. Yes. Right. But it's like it's probably you probably don't have that much time. So Coté (10:54.796) Mm. Coté (11:02.911) It's like a it's a it's a new monkeys thing that you could have given given enough monkeys and enough time, they could write exactly what Claude would output. Coté (11:15.561) Yeah, yeah. Yeah. Well, you know, I I think this brings up all sorts of like all sorts of interesting topics. And and the, you know, of course, I assume it's driven by like I think this is two weeks ago, there an EU policy came into effect that I I know there's some new it doesn't apply to like personal usage, but in certain uses, I think commercial and government uses of AI generated stuff, you have to identify that it's AI generated stuff. So sure. which is interesting. I I, of course, and then maybe to the next point, like when I read that two weeks ago, as as I'm sure all the Cloud crazy people would do, I immediately went to Claude and I was like, Can you do this with my stuff? Any anytime I generate an AI thing, I'm sure, I'm sure this regulation specifies a JSON sidecar or something that I put in the metadata of an HTML page that I publish. So when you go to incomprehensiblemedia.com, which is pretty much a hundred percent generated, I want you to tag it with this stuff. Brandon (12:00.286) Ha ha ha. Coté (12:15.281) And and it was like, as as so often happens with Claude, it was like, slow down. Like one, this doesn't apply to you, right? it only applies to these certain things. And then two, the regulation is actually not that specific. It doesn't say how you identify it, what you should do, what standard. They do, however, give you some possible logos you could put on things, which which was a bit of a disappointment. But it was, it was just like, it doesn't actually say. Brandon (12:19.848) Don't worry. Brandon (12:39.862) okay. Mm-hmm. Coté (12:45.697) You need to

use what what is it? There's some AI thing that's like CPS or CPA three point seven that's an actual like how to do hashes and all this kind of stuff that you can find in Gemini images and other stuff. but yeah, I w I was like, here's another thing I can get Claude to do for me, and it'd be great. Which I think goes to the next thing is it's sort of like, you know, I'm all for USB C, so I think the EU should have a very strong hand. In American technology product development. Because otherwise, we would not have USB C. We'd have a bunch of other bullshit cables that we have to keep track of, right? Yeah. Just like whatever. The EU fixed our cable problem. I I think I think that's undeniable. and so great. But in this case, I don't really like, I think maybe even in some follow-up thing, John Gruber wrote this is like, I don't really know what comes next. Like Brandon (13:24.903) Lightning. Mm-hmm. Coté (13:45.599) You identify something as AI written and then what? Like you're like you're like, it's AI written. Brandon (13:50.759) Well But there seems to be like always it's always interesting when anytime something taught touches like writing, because obviously by definition, the people that write the most, journalists and stuff, it's like it's it's like really hitting home to them, right? Like, cause today, no one almost universally, I I'm I don't know, some incredibly large percentage of source code is now AI generated. And like, like it's not, it's there's you don't need to watermark any of it. It's just like the assumption is like this is. Coté (14:07.295) Yeah, yeah, yeah. that's it, yes, yes. Brandon (14:21.564) probably AI generated, right? Like I mean, and nobody is like to me Coté (14:24.811) That's a good point. I I wonder if if code is watermarked. That that would that would be great. Brandon (14:27.912) But it I but no one is buy I don't it's almost like what if it isn't watermarked, it's like, why did you write a line of code that AI did not like check? At least do something with, right? It's almost like what's wrong with you? Like, did you not run a unit test? Are you crazy? Are you just like, you know, are you swimming without a life jacket? Like why would you not like at least have the robot like look at your stuff? So Coté (14:37.077) Right, right, right, right. Coté (14:41.484) Yeah. Brandon (14:48.026) so it's interesting to take that perspective where it's it's almost completely the opposite. Like if it isn't watermarked and you wrote it by yourself, it's almost like, hmm, we got to be suspect of that. Like that may not, that may and so it's interesting to me that the writers are all up in arms over this. And I thought, like, you know, Ben Thompson and Gruber, like they're like, you know, the tech writers of the tech writers. They're they're always talking about how important is like every word on their site was written only by them and it was never touched by AI. And sometimes I of I often think like both of you. Coté (14:56.478) Yeah. Coté (15:10.029) Craft. Brandon (15:16.648) Who I like, let me say one, I like, I read them, I listen to them, and I support their work. But at b at many times I thought to myself, each could use an editor. Each one of them could use an editor. And and so it's so interesting that they're like, they're just like so purest about it. It's like, really, I think it could just make your writing better, right? And and it doesn't and they to me, it's like, and this is what I was kind of thinking going forward. It's like you're talking about you know, how do you identify it? But it's almost like I want to put like a footnote in my email, probably AOI generated be Coté (15:25.271) Yeah. Brandon (15:45.958) Just assume AI wrote a lot of this. Or if not, it had a lot to do with this. I mean, almost everything I'm writing that has some types going on. It's like at a minimum, it's going to be proofread. But sometimes I will sit there and I will draft it for a while. I'll be like, I don't like this. I like this. Coté (15:49.355) Yeah, yeah, yeah, yeah. Coté (15:59.648) Yeah. I mean I mean I I know I know the people who have like data center exhaust fans blowing into their ear through like a window in in America are gonna like hate me for saying it, but just like like I I I like gladly disclaim everything that I do with AI cause just like 'cause it's like it's great. I wouldn't be giving it to you. Like Brandon (16:18.834) Right. It makes it better, right? That's and I think that's the point they're trying to get at. Like these water and I think they're they're I don't know. I think the writers are very concerned that like, well, if someone 'cause there's gonna be a lot of like, you know, someone's gonna go through and like, I don't know, someone's gonna go to an Atlantic article and be like, We found a section that was AI generated and there's gonna be some scandal at the Atlantic, be like so and so writer, like look used AI. And it's like to me, it's like great. Like, especially Coté (16:40.533) Yeah, I mean there there has been. Yeah, and and yeah, and and and like, you know, obviously if an AI makes something up, that's a problem. Just like there was that what was it like the series of little pills guy who wrote a book and he was on Oprah and then it turns out he made up a bunch of shit. And it's just like he didn't use AI to write that, but you know, it was bad. Like and so like clearly if it makes stuff up, that's a problem. But that's more that's not AI's fault. That's like the author's fault for being lazy is the wrong word, but for like not checking. Brandon (16:57.118) Yeah. Yeah, it's just weird. Yeah, just fabrication. Yeah. Coté (17:14.079) Right. And like, and so it seems like, yeah, and and then you know, that that's that's the other point that we this weird is so there's another angle on this, like when I hear the, you know, the two, the, the five or two people who talk about it, the the the the writers complaining about it is

there's almost and I I understand, I I make stuff. There's this I think the scen there's two scenarios that you would be afraid of with this watermarking stuff. One is like, I am using AI and Brandon (17:14.451) Right. Brandon (17:25.854) Mm-hmm. Coté (17:43.52) It's changing my words around. in not not because it's better, but because it's watermarkable. So I'm fine with it changing my words around for copy editing or even having it go past, you know, have me review it. but if it's changing words around not because it makes it better, but because it makes it traceable, then it might be tricking me into accepting worse words, right? And and so so we got that. Now there will be another one that I think there's a concern is. Brandon (17:50.334) Mm-hmm. Coté (18:12.619) Well, if I do use AI to write something, now everyone's gonna think it's a hundred percent AI generated if the watermark passes it. Whereas they won't appreciate well, appreciate's the wrong word. I won't get whatever benefits I get by someone thinking that I wrote it, whether that's the the f the satisfaction that I have or being a good craftsperson or the paying of it, whatever it may be. And I think like like with the first one, I would apply the principle of like. I don't know, you read it and liked it. Right? Like, like, like it's it and and then this the second one, I would I the second one gets to another thing, which I'm kind of thinking about as I'm listening, is like, I mean, despite how long we've talked about it, this is all just kind of bullshit. Like who the fuck cares? Like you could you could watermark stuff here and there, and like it doesn't matter. Like Brandon (18:44.818) Yeah. Well Brandon (19:02.462) Well that's the real question, though. Is it going forward, it is it true that people are not gonna care? Because I think that I think if you said to like, if we could just tell all the people that are worried about it, maybe you said like maybe, maybe as you said, it's only five people, but as they said, listen, going forward, this is gonna be the case. Everything's gonna be watermarket marked, but no one's gonna care. It will not affect like how people perceive your writing either way. If you promised them that, then I don't think they would care. And I think what people are worried about is like, because I think this like it's like almost like an emotional thing. It's like Coté (19:22.922) Exactly. Coté (19:32.246) Yeah, yeah, yeah. Brandon (19:32.398) I'm good at words. If I don't, if people don't believe all the words are coming from just me, then my value is not as good. And whereas I think you and I are just saying it's like, no, it's like we we the reason we're reading you is whatever. You have some combination of good ideas, writing, maybe you're popular, maybe it's just whatever. There's lots of reasons people get read. And it's like it doesn't affect to me, it doesn't affect and I think music is a good example. Like There's so much music now that's used like auto-tune, people use singers use like professional songwriters, people that make beats, people that make lyrics, like all of that comes together. But at the end of the day, like I like if the song is good, right? I listen to it and I never think to myself, like, this was like an auto-tuned produ made by this producer, whatever. I'm like, that song's really good. I just listen to it. I don't care. And if the next song is like you've 100% wrote it and it's all you and everything's good and it's not that good. Coté (20:21.078) Yeah. Brandon (20:27.686) I don't listen to it. You know, and it's like I don't it doesn't matter to me at all. And I think that's like really the problem. Coté (20:28.833) No, I I I think I I think I I mean I think I think that's my my a my AI filter, so to speak, is the same as my everything else filter, which is like, did I like it? Yes, no. If the answer is yes, it was great. And I think maybe what we need, I don't think I think there was a moment where what's his last name? Ted wrote the famous book about the aliens and the big rock that comes. he he he wrote a couple of pieces that were Brandon (20:41.544) Yeah. Coté (20:56.529) Very thoughtful anti-AI pieces. Anti-A is even a little strong. They they were good. I I don't know if he's written something for a while, but we need maybe someone like a Paul Ford could be this person for us, Brandon. But but we need he he I think he got close with that that last thing that he wrote. but like we need someone who's like the Kinji Alt Lopez of AI content. And you know, if if if the listeners don't know, Kinji, he has he you know, worked at the America's Test Kitchen. He wrote some of the best cookbooks ever. and he's got a great YouTube channel. And he's the kind of he takes an engineering mindset to cooking. And he'll he'll he'll basically be like, he kind of popular speaking of Apple, he brought to market the idea of reverse earing a steak and really like tested it out and popularized it. and so he'll do things like I really like like he his most recent cookbook was about walks. Brandon (21:42.878) Okay. All right. Mm-hmm. Coté (21:52.908) And so he spent three years, maybe more, like buying all sorts of walks, figuring out, going to places and like engineering like an engineer or a developer would. How do you cook stuff in a walk? Like he cooks it 50 times, blah, blah, blah. And then he just makes a video to to explain it. Like, this is the best way to make mopu tofu and blah blah blah. So he's like, he's in this category where he's not like a gourmet chef, right? He's not like a Michelin five-star, or is it three stars? Whatever, like person. But like he knows how to you can tell he like really knows the craft and knows how to cook good. And then all of a sudden he'll have an Instagram where he's like, I'm here at Costco and this is maybe the most amazing hot dog I've ever had. So I don't know he's done that with Costco, but he'll go to these places and it'll just be

like, Yeah, I mean, you go to this place and you get the the hot dog and you gotta make sure to put like the Heinz mayonnaise on it with like this this stuff, and it's fantastic. Like he he likes things because he likes them, not because of how they were made. Right. Brandon (22:52.69) Right. And it it's just sort it's I guess it's output, you know. And it's like what's we always talk about in like on digital transformation. Like we just we just care about what's the business objective. Right. And I think the same thing, and maybe that's the takeaway from like all this reading and all the making music is like at the end of the day, if something is good and it's helpful and it's useful for me, I'm it's fine. I don't care what tools you made from it. And I think this that's my message to all the writers out there. It's like, listen, before you had a human editor, I've had a human editor and sometimes you kind of get lost. I Kote, you've experienced like maybe get lost in three paragraphs. And you're like, they just put a big red X and they're like, here's one sentence. I think you should just say it this way. And you're like, done. You're like, that you're right. Thank you. Like you freed me from like my I was lost. I was in, I was, I was out at sea. And you like, and it's like, well, who wrote that? Was that my ideas? But they they they said, hey, say it this way. And I and I kept their edit. And it's like, that's the same thing. I feel like AI is so good at that. So so I think fear not. Don't worry about the watermarking. And then this is gonna be my public announcement. Just Coté (23:33.963) Yeah, yeah. Brandon (23:50.066) Just assume everything that I've touched, it yes, it has a watermark. You don't even have to test it. I'll just tell you right now, I it's what everything I in fact the watermarks marks don't even exist now. Just assume they did. I've everything I'm writing is is going to be watermarked from this moment going forward. Coté (23:57.686) It's it's pre watermarked. Coté (24:04.639) My my entire ouvoir is a watermark. That's that's that's that's that's what I have. Well, you know, I think I think what you arrive at here is the the technology is not the problem. The people are are the problem and how they perceive it and do it. And then, you know, I had I wanted to get your thoughts on this. I had an interview over on the software defined interviews podcast with one of my old co-workers, Brian Ross. And I think, you know, Whitney and I, I think we we got we got kind of close. To the mystery of, as you were saying, digital transformation, which is a dismissive phrase to be like transforming and changing how an organization works, right? And like what like you've got an organization doing their IT one way, and then you have transformation occur and then they're doing things in a different way. That's that's what you would expect. Hopefully a better way. But but what is evident is you usually have an organization doing things in one way, they have a transformation moment. Brandon (24:42.323) Mm-hmm. Brandon (24:53.672) Right. Coté (25:02.369) And now they're doing things the same way. And like it's really hard to figure out what people actually do differently in those two areas, right? And and I don't as as and and you know, the the thing that I get from people like Brian and others is it often feels like just a lot of one-on-one conversations. Back to that people problem of just like, we just gotta convince the people, e this individual, this person, this person to kind of just line them up. Like I I was talking with another person earlier this week and they they were talking about some situation where they they they had been asked to do some big transformation thing. And it was making me remember a lot of transformation strategy I've encountered over the years where the board, the CEO, they approve a plan and then and then you're all excited to go work on this transformation. And then day one, the boss comes to you and says, All right, now what you have to do is go to visit the eight country heads. in Amia and convince them to do this. And and and you're like, what? And then he's like, they don't say this part, but the the the executive would hand it to you is like, and by the way, in three months, I want you to go ahead and schedule a meeting to tell me why we should be doing this. Brandon (26:02.878) Right. Brandon (26:19.102) But I think so the thing I came away from, and I guess I want to try to answer some of these questions because I was thinking a lot about it. And so the interview you did with Brian Ross, and he's in the title of the episode, at least the first part is the The Chief Therapy Officer. And he he kind of tells a story that I think many of us have heard before. And it's a good story. People should go listen to it about like just what you said. It's like you gotta go in, you listen to everyone's problems, you let everyone kind of get their issues out, and then you hopefully then can then propose some change, right? And that's that's really how it comes to. And so Coté (26:40.525) The therapy. Brandon (26:48.082) Gonna go back and say, like, that's what people want to happen, right? But like this is maybe too pragmatic, or maybe it's a little bit controversial as to say is like there may be a reality where it's like, unless you can say why there's a need to change right now, there will be no change. And so an example would be something like you need some external, oftentimes it's like an external event, right? Coté (27:08.661) Mm. Coté (27:15.253) An exogenous event. Brandon (27:17.052) Right. Something's coming in. Like obviously something comes in where it's like we we must change. Like you had a catastrophe at work, right? Like some type of very bad thing happened to your IT systems. And if it doesn't change, then you know your business will you go out of business. Right. Or a lot of these things are like legislative. Like you now are exposed to some

legislation. You must be compliant with it. If you're not compliant, you will go out of business. Right. Your competitor is now doing something in a way that Coté (27:42.273) Yep, exactly. Brandon (27:46.598) gives them such a cost advantage that if you do not do it, you will go out of business. Right. all of these things, like without something like that. And you I used to work a lot in migrations, right? And I this is and and and modernization. And, you know, the calls that I th always thought went well, but nothing happened was like people are like, man, we we've found this new technology and we're thinking about Looks like it could be a huge savings for us. We we just discovered containers or virtualization. Like we think we should maybe virtualize everything because we have this old legacy. And they'd be like, and we have this great call and we talk about it, we talk about the technology, and then nothing would happen. Right. Now, the call that you get on to, and the first thing the person says like, Brandon, the data center that we're currently in is being decommissioned on August first. All the machines are coming out. They've already sold the real estate. The real estate's been sold, it's it's going to this other group. We have to get this done. And it's like, you know what? Coté (28:22.677) Mm. Brandon (28:40.274) those product like when you're on that, that you have like you're not talking about technology. You're talking about schedules, how we're getting it done. And you're like, like stuff is happening. You know what I mean? Like that's the call where like the salesperson's excited. I'm on there. Like we're all like geared up. Like we know we have to get something done. Right. And that's kind of like I guess where I kind of think of digital transportation. And that's what I thought about when Brian was talking. It's like, you know, maybe this is too like, I don't know, it's almost I don't want to say negative, but it's like if you go into a place and nobody has Coté (28:40.941) That gets done. Brandon (29:09.192) There are none of this. There's none of that. There's no compelling event. You're probably not gonna be successful in transformation. Just kind of like therapy. If you're just like a generally happy person and life is pretty good and you're like, I don't know. I'm just like had a bad day. I'm a little re a little depressed. It's probably like the therapist is probably like, why don't you just go on a walk? You know, like it's not like there's not gonna be like a huge, like your life is like kind of fine. You just kinda have it, you're just having a moment. You're probably getting through it versus like someone that's had like Coté (29:18.827) Yeah, yeah. Brandon (29:39.068) I'm having a life crisis. Like I can't do this. I can't go to work. I can't. I'm having some type of like the therapist probably gonna give you some like pretty specific advice that you're gonna have to look yourself in the face and like maybe take some action. Right. And like and and so I I don't know. So so maybe that's just a question back to you. It's like Coté (29:51.233) Yeah, yeah. I I like I like that this this is like this is like Brandon's theory of successful transformation. Is like people only change when they have to. Like it the Brandon (30:03.292) Right. And if you and I guess this is the whole thing. And it's like, and if people don't have to change, and you can identify that I'm not saying I I don't know, maybe this is too real on the pre-sale side. It's like, if you walk into an organization, you're like, I've spent a month in here, right? Like it's all the forward deployed people. I've been in here for a month. Business is pretty good, not great, pretty good. People are making, you know, good money, not great money. The leaders, they're here. They're not no one's getting changed. No executives losing their job. And you look around, you're like, That's there's probably not good conditions for change here. There's probably it's like you can do a project, maybe do a project, and maybe it's just more like you're just you can maybe sit more positively. It's like I can spend a month or two educating these people on some new technology that eventually maybe it will become really important to them. But right now, probably not much is gonna happen, is kind of my and like you could probably say that at the very, very beginning of the project, right? And this is where I think maybe Brian was saying it's like when things fail, like this would be a good follow-up question for him or anyone that does this. It's like Coté (30:37.612) Yeah. Brandon (31:02.95) When did you probably know that it's unlikely we're gonna have some significant business outcome here? Like that would be like an interesting question for anyone that kind of works in this world. I don't know. What what are your thoughts? Coté (31:05.771) Mm. Coté (31:15.317) Yeah, well, I mean two. I think well, I mean two things. One, I think there probably is an enterprise s software sales, probably everything revolves around the renewal calendar for your your top five infrastructure. Like whenever whether whether that's reserved instances you have or some license or whatever whatever that that that gets laundered as in words, right? That whatever your renewal cycle is, that's Brandon (31:32.211) Mm-hmm. Coté (31:44.076) Because that's what your your your your vendors are gonna be writing stuff on. Like if you're on a one, three, five year renewal, they're not doing jack shit thirty percent into it. They're just, you know, showing up, maybe get someone to come talk to you. Yeah. Sending you some pins, you know, whatever. But all Brandon (31:55.56) Hanging out, right? Mm-hmm. Brandon (31:59.838) And the renewal, to your point, like a renewal where the customer's like, the bill's too big, that could be an event, right? You know. Coté (32:02.721) That that's an event. Yeah, yeah, yeah. But but you've got to have so number one in in the

software defined talk transformation consultancy is you need to identify an event, an external event. And and I'm gonna I'm gonna use some phrasing that I discovered with my with Claude playing DD recently. I was like, we keep doing this adventure today and nothing ever happens. And it was like, here's what I'm gonna do. Whenever you meet an NPC. Brandon (32:26.791) Yeah. Coté (32:31.251) I need to figure out that if if your characters walk away from the NPC and nothing happens, that encounter shouldn't have happened. So like there needs to be like consequences that are at play. And in the same way that like if you walk in, do a transformation thing and like nothing happens if you don't transform, you didn't need to transform anything. Right? Like there's nothing. So you need like that that kind of a thing. And now the second thing it's making me think is. That kind of theory of things would explain a lot of industry-wide initiatives in open source that seem to not work out well. Like to get, I don't know. I mean, I think it's true. You could look at OpenStack and be like, this was more of a cool idea rather than something that needed to happen. And there's a lot of like hindsight and 2020 stuff. But like having actually worked on that stuff, it was there was not as much demand for that. As it was for just AWS. Like an OpenStack sale would basically be like, instead of using AWS, why don't you use this? And I think as as the as they say, the market proved out, they were like, how about how about I use AWS? And and then similarly, like I you know, like, like, like, I think in the past couple of years, like the CNCF has been really big into like. Brandon (33:36.936) Mm-hmm. Brandon (33:44.126) Yeah. Coté (33:57.198) Using Kubernetes to run inference versus models and stuff. And I kind of think like why? Like, I mean, I get I get technologically, sure. Like, whatever, right? Like you like, it makes total sense. You have, you have this, you can have all sorts of discussion about like exact match, but like you have you've got a a a file that describes how you want your shit configured, and you've got a system that does that. Wonderful, right? but Brandon (34:06.109) Right. Coté (34:27.245) I think the people running inference and training are d they're good. Yeah. They they they don't need anything. They're they're fine. Like they they gotta all all they need is more of what they have, not a different way of doing something. Brandon (34:31.336) They're happy. Brandon (34:36.199) Mm-hmm. Brandon (34:40.67) So I wonder one of the things I was asking is like the terminology I wanted to get your take on this as well was, you know, I've sort of posed it as like mostly changes are kind of like, I don't know, scary, negative, but maybe is is the consulting way to say it's like, listen, if we're gonna make this change, we then need to identify what are the new incentives that will exist for the people that are involved in this change. And if there are no new

incentives or disincentives, I mean, whichever way you want to say it, that's another flag to be like, this is probably not needed. Right. Like at the end of the day, if somebody is is is if you will measured or rated exactly the same, then they're and and it to be fair to everyone, it's like, then why would you change? Like you were probably doing things in a way that you thought was optimal. Your incentives are the same, and you're being, you know, and you don't and nothing changes, then like why would you expect, you know, people to embrace this thing, right? Because they've already, they already think they're doing a good job. Now, if you change their incentives, you're like, Coté (35:10.198) Yeah, yeah. Brandon (35:36.966) I'm only and then we kind of saw this incentives can be bad. Like we saw this with AI, where people are like, you know, we're incentivizing everyone to use a lot of tokens. And it's like that lasted like a month. And people are like, no, no, no. That's a bad incentive, right? So you gotta, and sometimes they're not always good. Sometimes it's bad, right? But you you really need to think about it in a way. And I think, you know, like the classic example would be like, you know, do you work at a place that's like, if you will, I'm gonna rate you on how many like phone calls you complete, right? Well, then people just like get smart about hanging up. They're like, if it's a hard case, you're like, sorry about disconnected, right? But if you got if you got rated on like customer satisfaction, right, or hard let's call it difficult cases solved, then maybe you stay on those calls longer, right? And maybe now, of course, maybe that means other people don't get help. But like, you know, just something like that, right? The whole idea, and you can imagine like with AI going into like a call center and be like, listen, we are changing the c the incentives. Like we're it's not about call volume. If you get a call. Coté (36:05.335) Yeah yeah yeah. Brandon (36:33.928) That means the AI I couldn't do it. And I expect you to be on for a while because it this is by definition a hard problem. Whereas before, maybe you were incentive to be like, you know, tell them to use a website, you know, file a ticket, you know, get off. And and so I don't know. So I just think it's like, and this is where I think maybe we like as the industry on the industry side, and this is what I was thinking about your guests, was just sort of like maybe there should be some more willingness for us to say that up front. Like if we can't identify any of this. It's like I said, maybe and maybe the way to soften it is be like, listen, I'm just here to teach you about this thing for now. Cause I think you should know about it in your toolbox in your toolbox, like AI, like OpenStack. And you should know about these things because maybe down the road it will be useful, or you can think about how it will be useful and then we should reconvene. But if not, you know, let's just like, hey, you know what? Let's call this a day. Let's hit the happy hour early and let's let's have a great weekend. Let's, you know. Coté (37:25.653) Yeah, yeah. Yeah, that's interesting. Like there there is a like in a in a lot of corporate strategy stuff I I I got involved in over the years. I mean, obviously the goal of the the the the big three firms is just like increase share price, you know, eventually. And the I I I get I mean that it that kind of explains why it usually just amounts to firing people or cutting costs. Because really Like you're doing you're you're not asking to increase share price by doing something new or better, even though that may be the result. But like the really you're just saying, like, I just want the share price to go up. And the easiest way to do that is to like make it cheaper to run the business. I don't know. I feel I feel like I've just stated some obvious things there. But the the the point being that you're you're kind of like putting forth a filter. I think in sales you would call this some qualifying. That that should be done. But it's almost like you want to hire the transformation to do this qualification before you even hire them, right? Which consultants love that. They always want to be paid to figure out if they should be paid because no one ever wants to pay them for that, you know, get a free estimate, which is infuriating. But like, and and the question would be just what you asked. The questions are like, is it even possible for us to change? The only way it's going to be possible is if. Brandon (38:23.91) Mm. Brandon (38:37.832) Right. Coté (38:50.955) There's as to use another fancy word, there's an existential crisis. If if you if if if if my character walks away from your strategy and nothing happens, that strategy shouldn't have existed, right? That that that strategy as MPC doesn't need to be there. So there needs to be a crisis. And then two, we also need to make sure that like you can do it through incentives. I I think incentives is is the biggest subset of just like humans deciding to do stuff. Brandon (39:04.03) Exactly. Coté (39:20.433) And like and do you are are the humans actually gonna do something differently? And before you even talk about like what it is you're gonna change to or what you wanna do, you've gotta just assess that and I don't know, and then and then report on that. And it seems like that would save a lot of people a lot of time. But then also for the consultants, it would get them paid for a different type of engagement. So their money is is is still gonna be fine. So the the consultants will be cool. but yeah, I think I think we've got it solved, Brandon. We need to start a business. Brandon (39:51.496) Well do. And I think, you we'll go out on this one. I asked last week, I said something like, Hey, like I want to know of like places where like AI has been used and something's happened. And like this was a story that we talked about a while back

with somebody, I'm gonna leave them anonymous. They said they said Duolingo, like back in the day, Duolingo basically did get rid of a bunch of staff and they now r if you will rewrite their lessons using mostly AI, or at least a lot of it for AI, and they have like a smaller staff that just reviews them. Coté (40:15.095) Mm. Brandon (40:18.428) That way the all the lessons get refreshed instead of being refreshed like once a year, they get refreshed, you know, however often. So that way when you're learning Spanish, you're not just talking about manzanas the entire time was as I was talking about when my my experience with Duolingo. So so I know that was a thing. It's like it is kind of negative because a bunch of people got fired, but it's also a place where like, yeah, they they clearly kind of probably saw that if we don't do something and reduce our cost basis, everyone's just gonna learn Spanish with an AI tool, right? And it's like Coté (40:30.465) Ha ha ha. Coté (40:44.633) Yeah, yeah, yeah. Brandon (40:46.246) And and they are they are kind of facing the existential crisis. So maybe it was bad. I don't know. I I'm not really passing judgment on like, did they treat the people well? Was it good? Was it bad? But like you see an example of like, yeah, I bet you they're all in on AI and I bet you they don't have to explain it very much to like why they did it, right? People kind of get it. Coté (41:02.709) Yeah, well I I I think I think maybe to have a a a cold hearted look at it. Like so so related, I don't know if it was a rumor or I think it's a rumor, but someone some PE firm wants to buy workday or wanted to, depending on when you listen to this, if they do or not. and so so keeping that there on the on the table, like the duolingual thing is interesting because so you know, I I I spend every day now trying to learn Dutch so I can get a permanent residency. And Brandon (41:15.272) Yeah. Coté (41:31.404) Guess what, Brandon? I bet I bet you know what I'm about to say next. I have the AI do all the stuff for me. I got, you know, our our our buddy Whitney, she sold me on Anki cards. So I got an Anki card deck. It's got a completely open interface. And so I just go through and, you know, I got a if you go to Kote.pizza, I got a whole podcast series that you can listen to. I got flashcards. Like I'll be sitting there doing a flashcard and you hook it up to 11 labs, whatever. Like I'll I'll sit every night I'm making a new flashcard, a new sort of thing. And it's like, Brandon (41:36.774) Absolutely. Brandon (41:42.11) Mm-hmm. Brandon (41:49.95) Mm-hmm. Coté (42:01.399) Perfect, we'll see if it's accept the ultimate outcome is I get a permanent residency. But it's perfectly tuned for like how I want to learn, like within minutes. Right. And so, and yes, I'm not just learning about, you know, here here's the the you know the the cli the klein the the little church and the park strat like and so it does seem like in that instance I get it. And

even the next level for Duolingo would be like. Brandon (42:04.892) Right. Brandon (42:23.378) Yeah. Coté (42:30.231) Totally customized to one person, which then I think that gets to the workday thing. Is like I was, I was talking with the AI about the workday thing. And I was thinking, it's like an interesting instance because a lot of us use workday and we're like, I would rather I'm gonna put it politely, I would rather not be using workday. Brandon (42:38.246) Mm-hmm. Brandon (42:49.458) Yes. Agreed. Agreed. Coté (42:52.001) And I could see that and and and and and the Diolingo connection is I think both of those businesses will never be in growth mode. Right. They'll always be just they're where they are. They might like grow a little bit more, but really they're they're they're set. They're done. It's kind of like from a completely different area. It's not like Patagonia's gonna get better. Like they're done. Like ev even even like Costco, they're just g they'll grow, but whatever. So like the workday business and the Duolingo business, like nothing interesting new is gonna happen. So at that point, your job is just to like take out costs and retain customers. And so of course that's what a PE firm would do, right? Like that's like and maybe what's inter what'll be interesting to see if they do, because this is what I would think to do is like. I bet we could replace pretty much everything with AI. And not only and not only that, and and pretty much everything is the flippant part of it. I I bet we could improve how things are working here by using AI. We could have the same exact number of staff who, as someone who's used workday for a little while, don't seem to be doing much. at least on on my experience, my the way I use it. We could just give them AI and then they could have. Brandon (44:07.815) Mm-hmm. Coté (44:15.287) Better software. And then even more so, like part of the thesis could also be like, you know, has anyone thought of writing a skill? Brandon (44:24.622) Yeah. Coté (44:25.771) Like th like there's there there is like so I and I bring all this up because this is a slightly different thing than the transformation thing I was saying is there are instances where kind of like with virtualization, right? Like where you actually can take a technology and apply it to a business as it exists and just make the business better if if you apply it without having to change anything. And and I think that Brandon (44:46.354) Right. Yeah. Well, I think that's the that's the thing of it's less transformation. It's more just like it's the more brutal thing that wants to say it out loud. Reduce cost by reducing headcount, by just optimizing exactly what you do. And as soon as you take away the whole strategy tax of like, we have to build something new and we change the incentive to be like, just do what you're doing and make the profit margin. That's usually what pro PE is doing. It's just saying, listen, your profit margin is 30%. I want

it to be 40. And I don't care about like, and then they're not what they're normally talking about, like, they're saying like, Coté (45:07.159) There, yeah, yeah, yeah. Brandon (45:19.218) We don't see revenue growth as a big issue. That's not our number one incentive. Our number one incentive is profitability. So when you give those incentives, things do change. And like workday does seem like the perfect example. Out like, and well, I'll just say it's slightly different. Like outside of something very different happening, without without the company literally probably, you know, starting a new division and trying to do something radically different, that business is going to run exactly how it has for a long period of time. And Coté (45:22.966) Yeah. Coté (45:47.127) That's right. Brandon (45:48.102) And so, yeah, so sure, that seems like an ideal private. And it's the ultimate thing. Cause like, you know, the funniest thing about workday being taken private would be like when they evaluate this would what would be I would love to know. I hope some private equity person will say it's like when they go through and evaluate who is going to stay or leave, do they go to the workday performance reviews to make the decision about who's good and who's bad? My guess is probably not, right? Which is like, which is like very meta, right? Like the pro. Coté (46:08.109) boy. Coté (46:16.597) You're you're you're right. They should be able to perfectly predict with inside workday, like in anything you want about staffing. Brandon (46:18.002) Yeah. Brandon (46:22.342) Right. Probably somebody in workday has a high performance review and is doing very well. That would likely be let go in a private equity restructure because, you know, they're expensive. And it's like, which to me, like then you get into like if you're really getting super discreet, like, well, what is workday? It's like, it's nothing to do with performance review. It's it's a liability management system that HR needs, but it has no Coté (46:48.641) Yeah. Brandon (46:50.502) It doesn't provide anything beyond that, which is like, you know. Coté (46:52.499) Right. And and and then and then the I mean to to rant just a small footnote rant. Like that the the ROI of most what do you call it HCM systems, like like HR systems, is based on not tracking the amount of money individual employees have to spend dealing with that shit. Right. Like if you actually brought this would be like the platform engineering DevOps mentality is like you gotta look at the whole picture and then your ROI is shit, is terrible. But now that maybe I want to suggest another test over in our private equity. division, which is we we have a new model where we take one of our summer interns and we give them Cloud with the Cloud Chrome plug in. And if they can make an enterprise software better with those two tools, we're gonna look at acquiring you and optimizing you. Brandon (47:22.118) Yeah. Brandon (47:41.874) Fantastic. Let's do it. Let's do it. We just first step we gotta get an intern. Coté (47:46.238) well, maybe we'll just sign up for a Gemini account and use anti gravity as as as as our intern. well, let's see. So speaking of of anti gravity things, do we have any bureaucracy this episode? Brandon (48:02.568) See a couple things here. I just want to thank I I'm gonna say this name wrong. I think it's Evgene, Evgeny. Anyway, they gave me some talk about digital transformation, they gave me some feedback on my my long pet project, my milestone iPhone app that lets you track birthdays and anniversaries. And they gave me some fantastic feedback. I've now released a whole bunch of new versions based on that. So I appreciate that. So check it out if you if you want to see. Coté (48:26.093) Brandon (48:29.304) Absolutely state of the art iOS development. And you have a birthday. That means you're that that means you're in the market for my free app. You can check it out. So I wanted to thank them for providing some feedback. That was fun. And of course, if you would like a sticker, as we kind of wrap up summer, it's not really wrapped up, it just means the kids went back to school. But the kids went back to school, that means their water bottles are going to school. That means they're almost surely going to lose them. That means what do they need? They need a software to find talk sticker. They'll be one of one, I bet. In On the playground. And to get one, all you have to do is send your postal address to stickers at software defined talk.com. Happy to send you a sticker anywhere in the world. Coté (49:05.131) Well y that it that I I I use your app unintentionally all the time, Brandon, because it pops up a notification and I'm like, I wanted to know that. It's it's it's great. I you know, I think that's a good that's a good capstone for a lot of what we were talking about, is like I don't I don't know why I don't just have a notification in my calendar. It tries to tell me about people's birthday, but I'm like, I don't wanna I don't wanna know that way. But the but with your app it's great. I I I know when people's birthdays are. In fact Brandon (49:30.502) Just doesn't yeah. Brandon (49:34.43) Appreciate your support, K Ty. Appreciate your support. Coté (49:36.615) I I wanted to go use it the other day but I didn't have my phone with me so I couldn't use it. I wanted to look up how old my mom was. and Brandon (49:43.358) All right. Well then we now that's the new thing. We now support the iPad. That's what everyone was waiting for. We do. We have widgets. We have widgets. We have we have everything that Claude can do. I'm just in here. I'm like, can you do it? Claude, yeah, give me the widgets. Can you do an iPad version? Give me the widgets. Like, give give give me everything. I like it. That's right. Exactly. Coté (49:46.061) Do you have do you have a lock screen widget so I can instantly okay. Okay, I gotta gotta check it out. Coté (49:57.42) Yeah, it's just

like it's just like be right back. Hold on. Coté (50:05.833) well, there's many conferences coming up, lots of DevOps in September. but also in September, there's We Are Developers North America and they just released a an overview of the topics that they have. And, you know, I went to the one in Berlin, it was great, and I I've been thinking, I I I was looking at my my European conference schedule for for the the fall, and it's a well known thing. That the American tech conference scene, you know, not not what it used to be, but the European conference scene is great. And so I think what what I think is gonna happen here is you're gonna get whatever that expertise that's making things different in Europe, it's coming over here to San Jose to to apply to it. So based on the experience I had, it'll be a it'll be a a huge, nice big conference, like a general purpose tech one. But let me just tell you, Brandon, the topics that they have. They got built building building with AI as a craft. Brandon (50:57.18) Okay, I'm ready. Brandon (51:00.976) Mm-hmm. Yeah, I'm in. I'm in on that one. Coté (51:01.345) That sounds like something I would want to present tour in. Agentix systems in production. Now that's professionally in my nine to five job, sadly, because I work with a lot of Californians sometimes, nine to five and then 630 to 730. But you know, that that's something I would be interested in. Like the you got multi agent architecture, MCP, tool orchestration patterns that hold up all sorts of stuff. That's probably a good topic. Lots of, as they say, real world world stories. And then modern languages and runtimes, always gotta have that. Platforms, pipelines, and developer experience. That's also a track I would probably speak in, going over your platform stuff. Then you got cloud scale and reliability. I think they should have called that track cloud scale and reality. and and it would be it would be case studies about people using that stuff. Then you got data and Brandon (51:50.878) Any any track that has reality into it, we should just go to. We'd be like, Yep, that's we need to we need to touch grass more. Coté (51:54.382) Then you got data analytics and the AI training stack, security and trust, quality and testing in a non deterministic world, engineering leadership in a shifting stack. That's that's a good title. So there's a lot of things there, but that's the point is like I think the scope of the conference is one of the things that makes it worth going to. If you're in San Jose or you find yourself going there on September twenty third to twenty fifth, we might have, I haven't gone and checked. But they gave us twenty-five free tickets, like a like a radio show to give away. But if you go to software define talk.com slash five eight six, then you can see if there's any of those twenty-five left. If there's not, or for whatever reason, if you prefer to spend money, you can use the discount code dev pod fifty. And I think that'll probably give you fifty percent off or fifty dollars off. I don't know. You gotta you it's Brandon (52:26.142) Mm-hmm. Links in the show notes. Yep. Mm-hmm. Brandon (52:47.782) I think so. I mean it's a it's amazing how that works. Yes, I think yes indeed. Coté (52:52.255) It's it's one of those two, something involving the the number fifty. Now, before that, you might want to go to DevOps Days Graz, which is September 4th to 5th. Of course, right before that, you will have been at the Cloud Foundry Summit September 21st to 22nd in Heidelberg, Germany, to see me and speak and many other people from the Cloud Foundry world. You know, it's gonna be totally worth it. You're gonna have to maybe miss the last few talks on September twenty second, so you can get that. You gotta take the train up to Frankfurt. And then you're gonna get a one-way flight probably on Luftanza or you or to go all the way to San Jose, probably San Francisco, and then drive down there so you can roll into We Are Developers on September 23rd. You got it figured out, listener. and then after that, or actually concurrent to that, you have DevOps Rockies, which is September 22nd and 23rd. We also have a discount code for that. Get out your pencil. It's two six D-O-D-S-W-E-D-E-F-T-A-L-K. Which you can find if you go to softwaredefined talk.com slash five eight six. Then there's DevOps Days Dallas, September 28th and 29th, which I wish I was going to. That was always nice. You got DevOps Days Vilnu, September 30th to October 1st. I'll be at DevOps Days Istanbul, which is October 24th, 2026. I'll be giving the keynote there, or a keynote. I think there's a morning and evening keynote. I think one of the team topology people is there. Maybe I'm wrong though. and then the VMware user group is in Orlando, October 20th and 22nd. There's Cloud Native Denmark, where I'll be giving, that's where the Team Topology Guy is also talking. I'll be giving one of the keynotes there, November 19th. And there's Scale 24X in Pasadena, California, another conference I wish I was going to April 1st to 4th, 2027. And then I would be remiss this Saturday, which is as as I alt tab over to my calendar. This Saturday, September 22nd. I will be at AI Connect in Riga, Latvia, giving a talk about the most recent version of my talk about how I play D D with AI and why it tells you everything you need to know when it comes to doing agentic enterprise development. what I learned there. Now, speaking of everything that you need to know, what do you have to recommend this episode, Brandon? Brandon (55:10.59) My recommendation this week is tech text sniper. So it's a very simple utility. And so what you do is you can it's like you're taking a screenshot, but what it does is it automatically just gets all the text. So for some reason, if you it has text you just can't copy and paste on your computer, basically you take a little picture of it, but it automatically takes a picture and then if you will, it takes that

text, puts it on your clipboard for you. So like if you're in the world of AI and you're just like, I just want that text. I don't want, you know. Coté (55:26.39) Mm. Brandon (55:39.006) I don't want the whole screenshot. You can use text sniper to do it. So super simple utility. It's almost one of these utilities that, like, if you're on a Mac, I guess it's like command shift three, four, and five are for screenshots. And so their their shortcut is command shift two. So it's just like, it's like, yeah. Like, do I want a screenshot or do I want the text or whatever? So it's just it's a perfect little utility, almost to the point you're like, this should just be built in. I don't know why this isn't just built in. I think even behind the scenes, I think it probably uses the Apple. Coté (55:50.227) yeah. Coté (55:57.134) Detection. that's good for AI stuff. Brandon (56:07.6) OCR technology. But nonetheless, I think it's ten bucks in the Apple store. So I feel like you just especially with AI today, like you just want you want your Claude, you want your robot to know something, but you can't, and you gotta do a screenshot, but you just really want to know that like one like log error thing that you can't quite, you know, cut and paste. You just grab that and you're like, here, robot Claude, like here's the text. I'm I saved myself like a little bit of tokens. You didn't have to look at the thing. Coté (56:14.049) That's a good idea. Coté (56:23.777) Yeah, yeah. Coté (56:31.751) I'm gonna I'm gonna check that out because there's there I use clean shot pro or X or something, which I think I'm I'm taking I'm taking screenshots all day, every day. Loves love screenshots, scrolling screenshots, everything. And yes, for the for the AI thing. But I found one called Shotter, which misses the last E. And it has a it has a CLI to it. So of course, you know where my mind goes, well, I'm gonna give this to the robot and then and then done. But Brandon (56:36.55) Mm-hmm. That's a good one, I've heard. Mm-hmm. Yeah. Brandon (56:55.197) that's a good one. Mm-hmm. Yeah. Coté (56:59.925) If this thing has a CLI, that would be even better because what you want to do is basically just be like you're typing and you're like, this is what ChatGPT, the desktop app, does. And I have no idea why Cloud doesn't do this, right? Is like, you know, like it'll you just want to go to Cloud and be like, what the fuck's happening on my screen and just have it take a screenshot and tell you. And of course, two years ago, Microsoft fell on their face about building this into the the the computer. Brandon (57:02.952) Ooh. Brandon (57:10.32) Mm-hmm. Yes, I heard this. Mm-hmm. Brandon (57:24.51) into the a into the operating system, everyone got mad. Coté (57:26.935) But I think we've all kind of gotten over that and realized it was a great idea. and so yeah, that would be that would be nice. Well, speaking of, my recommendation, I guess it's a whatever, it's a recommendation is so back

to the DND and AI. Now I do like the DNDbeyond.com, the official Hasbro site for the DND universe. And so I have my character sheets in there. And you got this problem, you always want to sync up your character sheets with your AI buddy. Cause all you got is a PDF export and it doesn't actually export everything. It's terrible. Brandon (57:30.375) Yeah. Coté (57:56.502) So I mentioned it earlier. I had this workflow where it would use Chrome. I don't know, it's probably against the terms of service, so sorry. But I would use Chrome and it would like scrape, scrape it and put it back together. And then one day I'm talking to my little robot buddy and it's like, hey, there's an API. And I was like, what? Because I've looked for a D D Beyond API before and they don't have it. But they just I guess the nature of how they've architected it, if you make your character public, which is public in that it's a link that if you know a GWID, you can go find it. So it's not really public. It'll just give you the a full JSON of your character. And so if you're playing D D with D beyond an AI, you should know that there's just a URL you can go to to download a perfectly formatted JSON file of of your character instead of screen scraping. So I'll put I'll put the the literal link to the the the API and maybe that'll solve your problems. Check that out. I guess Brandon (58:27.464) Yeah. Brandon (58:32.017) Mm-hmm. Coté (58:54.155) I guess the moral of story is always ask the AI if there's an API before you do anything else. Because it doesn't actually check all the time. Well, something that is checked all the time is when the show is over, because then we know it's over. now, what you have experienced has been another episode of software defined talk.com. You can actually verify that you've listened to it by asking yourself, did I listen to it? And unless you're in one of the states. Of mentalness where you wouldn't know, you will know. So now you know that you're at the end. Go to software defined talk.com slash five eight six and you can find links to everything we talked about and things we didn't talk about, those discounts. If you want to go to We Are Developers, you may be able to get free tickets or a discount. And we got a great Slack channel you can join and look at things. You know something else that has an API, Slack. So you don't I mean, you do need to sign up for an account. Brandon (59:46.63) Mm-hmm. Coté (59:50.712) But if you can't be what do the British say? You can't be fucked or something. You you can't you can't be bothered to to actually log into Slack and look at it, just hook it up to your robot and you can enjoy it the same way. We don't care. We love AI. And yeah, in fact we might need to add something so we can archive it. But no, that's not true. We we want it to be deleting. anyway, see you next time. Bye bye. Brandon (01:00:03.985) Right? W watermark away, baby. Brandon (01:00:16.766) All Kote, we did it. Well done as always. I was

gonna ask you, you know, building on your your Dungeons and Dragons idea there, is like I, of course, I'm not really into D, but like I love fantasy football. And I was thinking when you're saying I have been using AI to help me with my fantasy football, and I I was thinking that the one of the greatest uses of AI is like all these sites that really are just like game sites that don't really have that much security. Like you just cause this is what I found with fantasy football is like you just give it the like the unique ID of like your team, it can just get everything off the website. And I'm just like, this is like I but I actually think that this is a benefit as a feature, not a bug. It's like this is just fantasy football. I don't care if other people can see my team, but what I do care about is can my robot Claude friend get all of the stuff I have and it can then give me advice about what to do. So it's almost like Coté (01:00:52.94) Mm, yes. Coté (01:01:08.907) Yeah, no I I think the Brandon (01:01:10.77) I'm gonna in a rare rare event, it's like almost the the lack of security around, let's call it like unimportant game data has actually become a huge win for the AI bots to help you play your various whether Sunge and Dragons fantasy football. Coté (01:01:24.781) Absolutely. I think I think, you know, maybe this is a little bit more too much of the the late 2000s open web API thing. But I think maybe it's not so much with no proof of that. But like I think like I joked about this with the workday thing, but I think at some point not being accessible by API is gonna be like a disadvantage for your app, right? Like like I like Brandon (01:01:46.515) Yeah. Coté (01:01:49.846) The fact that I can actually extract out my character from DD Beyond makes me want to stay with DD Beyond. Where I before that, I often think like, do I really need this? Right? Like, cause it's difficult. And so I think in that instance, like it seems scary, but like you should have a UI for humans and a UI for the AIs. And I think probably most of your customers will be like, Yes, it's it's like that's great. That then I can use the stuff. Cause really that's Brandon (01:01:56.157) Yeah. Coté (01:02:19.639) We just you know, I I think I forget if I've talked about it here, but like my my other well, yeah, we had a whole episode. Like I don't think you really need a UI anymore. You just need headless apps. We already have a UI. Brandon (01:02:29.938) Well, maybe that's the whole and that's where we're getting I mean, what you're talking about is like the final level where it's like, no, it's almost API first and then UI second. Because it's like 'cause as people w I mean, as like you and I are just talking about all the time, it's like, yeah, it's just actually better if Claude can just do this for me, you know. It's actually I prefer that interface than I do like you're very, you know, well thought out, you know, point and click, drag this, drag that. Coté (01:02:36.746) Exactly. Coté (01:02:45.261) Yeah. Coté (01:02:51.519) Yeah. Like like I I would I would love when I'm filing my expenses to like have it hook up to work day and be like, Hey, love what you're doing with UI, your UI. Sure you're living a good life. I don't need it. Right. Like I'm gonna make my own UI. Everything's gonna be cool. Whatever it is you're doing, you can keep doing it. I just wanna interact with you differently. Right. Like and and that that would be wonderful. Just so nice. Brandon (01:03:16.562) Well, it's also wonderful. We've had a bunch of people in the public chat watching as we've been streaming this. So it's fantastic. Also, it makes me feel good that the streaming is actually working again. Because last week, I don't know, it didn't work. So so what I did was I did what everybody does when streaming doesn't work. You delete the streaming link and then you just add it back. And magic, it works. Okay, that's what I did. So that's all I did. So I want to thank everyone that's in here. I saw a bunch of good comments about workday. as well as several other things. So thanks for hanging with us while we stream. And of course if you watch this later on YouTube probably, we appreciate you watching the channel. And what what did we always say? subscribe and like. No, it's like and subscribe, I guess. I don't know how we should start saying subscribe and like. We'll just switch it up, you know, and say yeah, you know what? If you can, maybe just have have this summarize. Summarize the transcript via your favorite robot and then just tell us you like the show. That's that works for us as well. Coté (01:03:54.913) We should say subscribe and like, that would be great. Send us your APIs. Coté (01:04:09.685) Yeah, just lo load up Claude and the Cloud Chrome plugin and like and subscribe. Brandon (01:04:15.998) With that, we will talk to you next time.