s03 e02 – Algorithms, Automation & AI.

Hey, oh, we need a little applause. Just hit them all, baby. All right. Hey, welcome back to the Big Cheese AI podcast. I'm your host today, Sean. I'm here with Jacob, as always, as well as our trusty, friend, and mostly sober Brandon. So we're October. Sober January, dry whatever you want to call it. We're back another, we couldn't start second straight week. We're going to keep posting and we're glad you're here with us. For those of you guys that tuned in, you probably tuned in because we just posted our last podcast this morning. We'll probably post this one. We might have a couple back to back. If you guys are subscribers, I don't even know if subscribing means anything anymore because everything's algorithmic now. But there was a big discussion by the Patreon CEO. I'm throwing a huge wrench and curveball. Did you guys see that? You made a speech and he was basically like the concept of the follower doesn't even exist anymore. Yeah, because like they just want eyeballs and your attention spans. It's a tick-tock thing, right? It's kind of interesting. So when he's saying that the followers don't matter anymore, what other metrics can you actually have? I don't know. Oh, fresh and beer. I'm saying. Oh, it was an interesting thing. Say that again. Basically that if followers don't matter anymore, then why? What's your metric? Engage. Kind of? No, the metric is what gets people to re-engage in your platform. And that's algorithms. It's the same reason. But an algorithm, an algorithm's job is to do something. To get somebody to do something. And that something is some sort of a number or some sort of thing that I can basically grab, hold on to, and then use as my source of measurement. So your follower count is more or less a success on how well you're doing from a marketing standpoint. So that's not what he's saying. So he's saying the concept of the follower used to be what you saw on your Twitter feed. Yeah. So you used to see the people that you followed. You saw the accounts that you followed. But on Facebook, you saw the posts from your friends. Over time, you started seeing things based off of not that. No, okay. So yes, 100%. So social media went to shit when it got away from the linear release. Like, here's my followers, here's all my stuff. It became the algorithm. That was the downfall of social media. In fact, it really became, it didn't really become social at that point. It's not social really anymore. And that's what TikTok is. It's the reason why I deleted Instagram because I somehow got stuck watching people make pizzas for an hour a day because I like making pizza. And instead of being up to date on, I was a little bit, but I was re-engaging so much. There's constantly content in these algorithmic rabbit holes. Okay, so fuck it. We'll completely sideline our entire conversation we're going to have here. But I think it's relevant, right? Because if you think about where this is all going to go, Facebook, YouTube, TikTok, they're all going to be focusing on AI-generated content. It's not going to be just content from the people that are followers or like actual people. The systems themselves are going to be generating content. So social, social, quote-unquote social media now is just going to be media. And I mean, you could imagine a time like, think of all the things that we watched on TikTok over the years while it was running before it's going to get shut down. And almost in five years from now, you could see that 85% of it was completely not generated by people. Completely not generated by a normal person. Right. And that's going to be then, then, then, my God, like what does that mean? YouTube, video, content, same thing, right? So, okay. And personalized to you. Yeah, well here's the, here I agree. So Netflix is already doing this, but they're not AI generating it. They're just saying, we know exactly your user profile. We know what Sean or whoever the group is is going to watch. And they're creating content specifically for that. It's not good necessarily, but it's more engaging, right? Right. And eventually, if, okay, if that gets us people to stay on the app longer and more engagement and less likely to turn with every single app that that's their metric, why not use AI and make it super personalized and super engaging? Well, what really happened was people only have so many friends. People only click the follow button so many times. And so TikTok came in and said, how do I get someone to just constantly be on here? Well, I figure out what they like. And there's tons of content about what they like, and I can present that to them without knowing. So the follower count does matter, but the way that you get projected, the concept of the follower doesn't, in a way, is dead, right? The concept of the follower, the concept of the connection you make with those. And that's why I think you see like a discord or a, you know, a Reddit. Like, like, like, like, people are joining these communities, right? Where it's like, they're like-minded people. But that's a different, that's a different concept than what we had from, you know, when we first started joining Facebook, when we first started joining Instagram, we're just like, you know what? I do kind of feel connected, more connected to my friends. Yeah. I do kind of feel more connected to my network. And I feel like I can grow my network. And maybe LinkedIn is kind of like that, but like, that concept is going away. I don't get on Facebook at all, really. And I hardly ever post. I notice you post a decent amount and you get good engagement. I have like, whatever the maximum number of friends on Facebook, I've got to be approaching that, because I'm very popular. But if I post something, I'll get three likes. And I'm like, I know that that was not like the most riveting kind of- It's because no longer is you are you being projected to your followers? Yeah. You're being projected to an algorithm. Yeah. And if you get sure if you get some engagement, they might present it to your followers. I do. So I think too what happens is there's obviously each algorithm is going to be a unique kind of approach to it. But if you post something and then you get a couple of your people of your followers to interact with it, then that's kind of a signal that it's, oh, it's something that might actually be relevant to the followers, to the network. Yeah. The network's still there. It's just a little bit more irrelevant. Yeah. And I think it's more just intelligently designed. Yeah. It used to be chronological and you saw what your friends posted. Yeah, exactly. And when you're going- There's blue sky content, probably a lot of people. Yeah. So blue sky doesn't have any algorithm. Now it does. Like you can go and you can have like, you know, specific areas like I only follow product design or I follow art or I follow these. And those are kind of algorithmic. But the main- your main thing is just the linear, you know, here's what people are posting and you go through and you follow it. Yeah. And we all know that they're- they're engineering this and hijacking our brains. We know it. And we've been in the- an hour day on pizza making. I do the same shit where- Right. Now it's pickleball. But like- Oh god. I just keep getting videos of whatever I'm currently interested in, right? Yeah. And we know that they are doing this to us, but they just have that grip. Oh yeah. Right? Dude, real quick. So the one that I'm- that I'm absolutely obsessed about right now is this guy who goes- And I think he's a Minnesota and I might have talked about it a few months ago, but I'm still on it. He goes to restaurants and he says, hey, how much is this meal? He's like, it's 9.99. He's like, I want a hundred of them. And he can do that. And they're like, yep. And he's like, all right. And then he goes and just walks around Minnesota and he hands him out to the- to- to homeless people. Right? And everybody's like, you're just doing it for clicks. You're just doing it for views and everybody's like raging about him. And like, it doesn't fucking matter. He's feeding homeless people. Like, and I cry. I just like, I ball every fucking time I see this video and I can't stop watching them. And they just continually feed that to me. Those are the types of things that I'm like, in this case, the algorithms is a positive, right? Well, the algorithm isn't like trying to get you to necessarily do something nefarious. The algorithm is attaching to your interests. It's attaching to things you engage with. Right. The problem with it is that it's the overall problem, in my opinion, is that it's creating addictive behavior. It knows almost too much. Like, it knows that like, there's those dark things in our heads that we won't tell anybody about. And then all of a sudden, I just happen to watch that one video for like a half a second, a little bit longer than I normally do. And then all of a sudden, it's like, ooh, look, you like feet. You know, and it's like, oh, wait a minute. I don't like I like feet to know why you see his feet. No, I like feet. So I don't understand. We now know what Brandon's weird, weird interests are. As I've gotten older, feet have become a thing. Yeah. That's our shame of the game. Welcome to the Big Cheesy Eye podcast. Welcome to the Big Cheesy Eye podcast. So today, we want to talk about, let's just start it off. I got a topic. We've had multiple CEOs. So it's 2025. Everybody wants to go on Joe Rogan, right? Mark Zuckerberg has that, his curly hair. He's trying to look so cool. He wears that necklace all the time. And the oversized teams. I mean, talk about a guy that's such a small man inside. Like, and always has been. I mean, they made a movie about how big of a dick he was. That was a great movie. Really good movie. I mean, he's not scared. He doesn't care. He will go for your throat. And if you're not adding value in your partner and his business deal, they'll kick you out. But, you know, a very engineering focused guy. Like, for people that don't understand where Facebook came from, it was always about the engineering. It was always about the technology. And, you know, he is a developer. So he goes on Joe Rogan and he says something that gets thrown to all the algorithms, which is, I'm going to replace, Facebook is going to replace mid-level, mid-level engineers with AI. Let's get your thoughts. Yeah, I mean, I don't know if you guys saw my post on LinkedIn.com. I was trying to get some likes on there since Facebook's not been working for me. But I think that, you know, he's saying these things. And like, yeah, for Facebook, it makes sense. At least near term that the mid-level developers probably, their code base is so big, they can train their models, they can throw a bunch of hardware at it. And long term, it's a good investment for them. But it's also misleading in the sense that like, he's saying replace, I think he probably eventually means augment or, but it's just like this, a pony, a dog and pony, pony Joe, tree and the woods. So he's got it, they spent a shit ton of money on AI. And now this is their nod to their investors. It's like, here's the payback, baby, we're going to fire, we're going to lay off all this headcount. And the reality is, it's like, yeah, the same reason we use AI and love it is because it can automate a test or create a bunch of JSON that I don't have to spend 3, 8 and a half hours doing anymore. And I can get back to more meaningful creative solutions and the architecture and so on. Thinking about what happens if I do this and bigger picture stuff, right? And at a company like Facebook with thousands and thousands of developers, yeah, you will reduce headcount because you just don't need them anymore. But it is a dangerous message to say, I'm going to replace developers with AI. So out of all of the CEO saying, bat shit crazy stuff, I think Zuckerberg was the most realistic in the fact that it's your mid-level guys are going to get hosed. Now, now it's a short-sighted view though, because again, how do you become a senior? You become a senior by starting as a noob, as a junior, and then as a senior. And by cutting out this middle junior phase, you're basically, man, affesting the fact that you're never going to have future seniors. So that is a problem that people are going to have to figure out how do you work through this and you deal with it. The one person who is absolutely bat shit crazy is Mark Burnoff of Salesforce. He goes up and this guy is a fucking shell as far as I'm concerned. So again, great guy built a lot of great stuff. Now he's just literally doing this all for views. He's a charlatan. Now he's saying we're not going to hire any engineers in 2025, right? It goes hyperbolic, it's all over the fucking news, it goes viral on this thing. Go search for fucking jobs on Salesforce. Guess what you're going to find? A bunch of engineering jobs, right? So it's a these? Now hold on a second, hold on a second. So if you look for a job in the tech world, one of the things that you'll notice is that and this is something that came out, Google me, 20% of all tech job postings are just ghost postings. It's just HR and their culture. They just kind of want to keep the thing out. Now I'm not saying that's what he's doing. No, no, but so but what is the real reason for? I think it's to always look like you're growing. It's to always look like you're helping the tech community. When you're really, you're really just completely serving out empty promises. And I mean, let's be honest, I'm always I'm I've hired many engineers. I'm always looking for tech talent. Oh, speaking of which? Yeah, we're hiring right now. Yeah, right. So we're in the we're in the process right now of going through our second round interviews for an engineering position at Crafted full stack. Do you want to hear something crazy though that happened this morning? We we posted on LinkedIn at the job posting. You got a calendar invite for someone. I'm not going to mention the names here, but so Job tracked it down. Where did this come from? This guy like forwarded or replied to an email from zipper Cruder. We did not post the job on zipper Cruder. Someone using some automations pulled our post modified it. Create sent an email from a crafted up or from a zipper Cruder email and said he was taught at Crafted. No, yeah. It's completely automated and they used clear bit to get our logo and to get a lot of our information. But they but it's a supposedly a company in New York. So I'm guessing it's a company in New York. Should we like tell them to fuck off? Yeah, but this is like AI automation. It's not just AI, but they spoofed like they worked for us. Yes. And we don't know if it was actually my chief HR officer right now. Yeah. Let's see. But this is like some of the danger of automating things that your processes aren't. People are this is really a theme right now. And LinkedIn is like AI is good if you already have your systems and automations and processes in place. It's dog shit if you don't. You know, so okay, so in 2025 one of my and I think I met a predict this and the last one, the prediction. Some company is going to go fucking nuclear. Yeah. And they're going to go so horribly wrong. And they're going to set up this thing and they're going to be like, oh, we just have AI doing this stuff. Not realizing that it's not deterministic and it's random and some random things are going to happen. All of a sudden millions of people are going to lose shit tons of money because an AI just went off the fuck. Well, that was that was the prediction, the bold prediction. Yeah. And last week for 2025, if you haven't checked out our episode, we do some predictions for 2025. But AI is given, I mean, doing a lot. It's all about the agents now. So talking about agents, let's move on. So I have not looked into this. So I'm really excited to hear what you guys have to say. TASS, baby. TASS. The first use case on the website remind me when my mom's birthday is. Facebook's been doing that for 10 years. But tell me about chat GBT tasks and why I care. So, so agent, we have to go back a ways to agents, right? So a cloud released the desktop demo. And the desktop demo they could like talk to it and have basically give you. They give the AI full control over your computer, which is just fucking horrible. It's owned right. And but what's so what happened is that that opening AI, because you had Gemini and OpenA, our claw, both come out with agents. And OpenA, I was sitting worth a dick in their hand and they didn't have any answer. So they've been talking about they're going to do agents are going to do agents. And there's a problem though with agents. Again, there's a security side. So they've kind of didn't release agents and they just made it in tasks. And so what a task basically is, is a prompt that you can then schedule. So you can say every morning at 8 a.m., I want you to do this thing. Go do this and then maybe you know you have different tools in there. You can send an email. You can do those types of things. So it's kind of like an intelligent cron job. Yeah. If you would think about it, which is cool. But they're kind of still holding back a bit on the agent thing. And I think we're going to see that over the next this year. There's going to be this. Some people are going to go go full steam ahead with agents. And some people are going to hold back because they're going to realize that these agents have an opportunity to go fucking haywire. Well, well, I think an agent operating on a computer is too broad. That's exactly what I was just going to say. That's exactly what I was just going to say. We're all dogging your computer. What you're saying is that chat GPD can schedule tasks inside chat GPD. Chat GPD has a wealth of knowledge. It can search the web. It can use its LLM. It can do lots of things. But it's going to do it within its own protocols and APIs. Yeah. And I want Gemini or my Google Home. So I have a handful of automations. And I did the home assistant. I'm sure you're big on that. I found it too complicated to like I tried to set up my own dashboard. And not too complicated because I'm so smart. But but two time consuming. I'm like, I'm not fucking riding a bunch of CSS. So I could put this up on me anyway. So I'm like, what I really want to be able to do is to say, Hey, Google Home or whatever you go, hey, bitch. Hey, that's sexist. Turn, turn, turn off my lights or in the morning, you know, Hey, right now for a sunrise, turn on my kitchen lights. And you can do all these things. It's only a few clicks. But or more rich tasks and more more chain tasks together, whatever. So like, that's what I want. And that's just one application within one tool. Is home auto. So I am not a fan of home automation. I think it's super invasive privacy wise. And I hate the fact that they even have a doorbell camera. But like is home automation like any good to this day? Fuck dude. Okay. So as far as I'm concerned, if you're going to do any home automation, you need to use home kit, right? So Apple, Apple's home kit. So what I want. And everybody's like Apple, you know, versus Microsoft versus Google. The reality is is that Apple is not based on an advertising model. That's the most important fucking thing that you can think about of what they are. They sell hardware. They make their profit on their hardware. That's why they can do these things where they can say, we don't actually need your private data because we're not fucking an advertising company. Like Google, Google can always, Google always needs your data. They always need your data because their entire model is based on the fucking. Well, let's go back to that. So one of the reasons why Apple's been late to the AI game is because they're building the secure enclave. Yes. And they're not even talking about it right now and everyone in the market's like, well, AI, AI, well, everyone, they, all the other companies just sold us all out to make their AI. Dude, totally. So as I'm building this app, so we've been talking about the paper keeper app that I'm trying to build 100% with AI, right? That I'm able to do all of this, be able to share things with other users, be able to have it stored in the cloud, be able to do all of this, and not touch a single external API, not, not call chat GPT, not call any of these other AI models. I can do these different things where I can pull in some intelligence within the application, purely within your own experience, within your own phone, and you're never having to have any of this data be sent out anywhere. Oh, you're using Apple intelligence inside of the app. Is that how you create the, well, so I'm not, I haven't fully embraced Apple intelligence yet. I'm using VisionKit. I'm using their core ML for some of their, like, text extraction and things like that. I did try to go and put in a full, I tried to put Loma 3.1 in there, and I got it working, but all of a sudden my download became 4.3 Gigabyte. Dude, it took like four minutes for it to actually go and try to extract the information of these documents. So, but that part was just like, oh my God, the reality is, is now I can actually, if I just stick with Swift, and I just say I'm just, I'm just in iOS land. I'm not fucking Android, sorry guys. Right? If I do that, I can build everything completely secure using all of Apple stuff, all of the data's private, I don't, I never have access to any of it, I have no servers anywhere. I don't have a Firebase account set up somewhere to do push notifications. It is, it's like, I'm starting to see like why, if you're going to build like really amazing apps, you need to build them purely natively. And you need to do it natively for iOS, and you need to do it natively for Android, and use each of those features and function, all the things that they offer you need to do them. Okay, that's, I agree. And I wanted to ask you, have you done anything with App Intense yet? Not yet. Okay. But so that one makes all the sense to me in the world, like any time that you, Well, you can't, you didn't, but you kind of did, right? Because you built sharing natively into the app. Yes. Yeah, so that parts, that part's kind of just baked in, but I think that that's more just part of like Swift, Swift data that they allow. So, but, but the ability, so like from an app, like one thing that I want is like, if you have some sort of a PDF or you have some document, I want you to be able to share that straight to paperkeeper. And we're getting a notice from our producer here that something is going wrong. And my vision is too bad to be able to read what they're saying. It means you need to talk straight into your mind. I need to talk straight into the mic. This is an ongoing issue. HR will be contacting you. I think you should move your, your boom a little bit to the right so you can give yourself some more good. But anyway, so yeah, that's, that's just been an absolutely fascinating experience. To be able to build these things using 100% of Apple's setup and not have to worry about anything externally. Yeah, that's exciting. So, so I had this written down as a talking point. And I think it's, it builds on, directly what you're talking about. So, so let me just quickly, I'll quickly like quick fire you a few things. So, you built an iOS app that is used to securely store and manage all of your critical documents. Those documents include drivers licenses, pictures of your driver's license, front and back, your insurance cards, other things that are really, really, really, really important that you need to have copies of. Right, that you should have a backup of and they should have a backup of in the cloud. Now, without AI, you're able to take those images, you're able to use Apple Vision kit to zero in on the aspects you want. You're able to take metadata off those images, including the actual words and text. And then, and then auto fill forms with that data. Yep. And then you're able to not, not leverage a single API that exists outside of the device beyond persisting photos to your iCloud. So, you could, you literally turn off, you can go into your airplane mode and the thing just completely works. So, that kind of just just shows you like at a baseline level, people call it AI and they build all this stuff. Apple's been doing AI for years. They have. Yeah. Vision kit's been been around for, for three or four years. Three or four years, same with Core ML, right? And what's, what's really, so Core ML, yeah. So, Core ML is their core machine learning system. And they've got a bunch of different tools out there that you can take existing models like Lama 3.1, 3.2, and convert them to Core ML models. And then once those have been converted to Core ML models, you can basically import them into your projects. And then you can, you run those models just on device, whether that's CPU or GPU or neural, the, the neural engine. And, and it just works. So, really the only thing that Apple was truly missing when it comes to this stuff is the LLM. Really? Yeah. And that, and that, and that comes the underdevelopment and proliferation of Siri because for some reason or another, that product team is probably just... Yeah. And they still suck. And, and, and for those of you that are trying to use Apple Intelligence, it's not any good. No. I was really hoping that, that they would offer some sort of just pure like LLM call into this thing. Say, here's this, give me some Jason back. Like, just like we would do with Chagy BT or Claude or whatever. And it's not there yet. And so, that part was kind of frustrating. And so, all of the things I had to do was like, I had to go kind of old school, right? Like OCR of the text, then do a bunch of regax to be able to extract dates, to be able to extract URLs and emails and all those types of things. Where before, I would just be like, you know, with Chagy BT, I'd just be like, here's the payload, and here's my Jason's schema. Just output it for me. Yeah. Like Apple will eventually get there. The other thing that you have to consider though when you're building these things is that, that, that not everybody has an iPhone 16. Like, I'm, I'm still in the 14, so I don't actually have any Apple Intelligence on my phone. Be a sir. I know. But I've got it on my iPhone. Speaking of that, how are you doing on the iPad battery over there, Job? Thumbs up. Good. All right. Good job. So, so the, the, the kind of the thing that I was thinking about is, I think there's a side effect of coding with AI. And people are going out like you and they're going on LinkedIn and going, I built this whole thing, but they are. And I didn't even talk to anybody. I did this. Did it? Oh, yeah. Okay. See, I didn't even know you posted it. I just assumed you did. Okay. But so one thing that, that used to occur, and this is something that I've been dealing with professionally for years, is traditionally in, in software development, there's a concept of separation of concerns. So you have your front end developers that develop the front end. You have your database and your DBAs that design that data database and they maintain the database. And then you have your API developers that maintain the API. The front end hits the API. Everybody's, everybody's like, I don't know.net. Yeah. Okay. But I know reactant and you're job is great. Let's work together. Right? But that is not good in AI development because you can't, there's three different projects. The AI. So if I go into cursor, which is the AI IDE, the development environment we use, I don't want to run three projects right now. So like the, like, the, and you know what? I honestly, and this is getting down to my point, there is a point, is I want to reduce as many abstractions as I possibly can. I don't want to add packages. I don't need packages anymore. I used to, if I needed to do a database query or something, then, and I was uncomfortable with it because I'm not an API developer, I use it nowhere. I'm like, Prisma. But now it's like, no, I actually don't want to add packages. I want to keep it really simple. AI will write my SQL for me, right? AI will, if, let's, I bet, I bet millions of more people want to use SQL right now because it ships the database in the code. Right? So now AI is sitting there and you're going back to old peed. We're going back to Laravel. Yeah. We're going back, we're going back to full stack, right? We're going back to every Ruby. We're going back to everything where it's like, the AI needs to have context of everything you're doing. Or else it just isn't going to know what's going on. That devil's advocate. Could you not just encourage or create a workspace where all the things are there? Probably. Probably, but I feel like, yeah. But I also agree with you. And here's the other, I think the point really is use things that AI knows really well because like, I have used some, some program or language as that aren't as known. And I ask it questions and it's like, it's like, uh, how about this? You just made that up. That is it, right? I think that it's a self, it's going to become a self-fulfilling prophecy. Yeah. And there's a reason why we, or if it's a company, have always gravitated towards certain technologies. And one of the main reasons was seeing what the industries and the smartest people were using. And people gravitate towards those people. It starts to create critical mass. That, that totally agree. And the byproduct of that is better documentation, better tooling around it, better, um, uh, just stack overflow answers, whatever you're using to get the fuck, the job done. I was interviewing developer the other day and he was like, I, I'm really good at, I'm an expert at React Angular. And I, I've loved felt my favorite and I go, I go, sure, I'm sure it is. Good for you. You'll never write a line as felt here. Right? And that's because, you know, we do this for a living. But, but, but, but, but, by the way, Svel is by far the superior language out of all of them. Svel is hands down the best. I will die on this fucking hill. Svel is amazing. And if you're a React developer, if you're an Angular developer, if you're a View developer, I want you to go and just play around with Svel. Svel 5, eh, it's not as good as Svel 4, but nevertheless, it's an amazing language. And eventually, God, fucking eventually, companies will, will realize this. And you'll be able to start building and having a great developer experience with Svel. Anyway, back to you, Sean. No, I, it goes back to, so, so when you're, when you're developing with AI, be conscious of these things. Like, try to reduce the amount of abstraction layers you're putting into your apps. Or use abstraction layers that it is very familiar with. True. I built, so I was building a form with this app and I, and I was building with my son, because I'm teaching him out of code. And I wanted to build an app that has no code. So, of course, I made sure everything was involved. It was, or everything was self-contained. SQL Lite, next JS15, no ORM, right. And, and using the composer feature. And so if you're using cursor, the, the, and you're building an app, like, there's one way to do it where you can just kind of open up a file and chat with it and then use the, I don't want to chat with you today. I want you to actually just write my code for me. But then you click the next tab, that's the composer tab. And that's where you're saying, I'm today, you're going to be my developer AI. And that's where you experience the true power of it, which it's, it's also, it gets unwieldy as hell, right. But, you know, that's where it's like, this thing is the developer and you are just talking to it. Yeah, I had an experience, so I saw your post about your app, your building completely using cursor and you don't need any developers, because they're stupid. And I tried it. I need to see this post. I tried it. I was like brand new Swift Greenfield project. I wanted to, you know, the first thing I wanted to do was owe off into Google couldn't do it. And I went to Google's page. They have a very, very good documentation on how to do this. And maybe I don't write in Swift, so maybe I just missed something. I imported the packages, per its, you know, directions. I think it made one up. I think there was just a package that was like, here you go import this. It does that with Rust all the time. When I was trying to do that, that scanning library with Rust, did that all the time. Yeah, and of course, this is an external package and it's not necessarily the most natural thing to owe off using Google into an Apple native. I mean, I don't know, people do it all the time, but I don't, for some reason it's just never easy to implement, owe off. And I'm still confused. Well, I think that's also part of it. So when you're doing like that composer style application development. So there's a different scene saying, hey, AI, write this function for me and hey, AI, build this entire application for me. When you're doing, hey, AI, build this entire application for me, each step is very important. Right? And it's almost like you need to get, it's almost like you need to commit like at each step. You have to do. Instead of trying to just let him go loose because it becomes not the mid-level developer, Zuckerberg. Zuckerberg. Eat a dickberg. I don't know why I said that, but I hate him. You've been a villain since the social network. But the point was, it really becomes a junior developer that drinks way too much coffee, snorted 10 lines of coke and works 24 hours a day. And it always thinks it's right. And always thinks it's right. So, all right, so let me explain how I've went through my little journey of building paper. Yes, let's go. Paperkeeper. So paperkeeper, the trapper keeper was the whole concept, right? Like anybody who's over 30 years old knows what a trapper keeper is. I'll give my wife props coming up with the name because before it was show me your papers. I've got one correction to make. I'm pretty sure you have to be over 40. There's 35. No, no, yeah. So, young millennials do actually know about trapper keepers. And I actually talked to a couple 25-year-olds. She's like, oh, my mom's talked about the trapper. And I'm like, hey, all right. So we're kind of close. Anyway, so my whole point though was I'm going to build an app. I'm going to order one. Swift, native. And I don't want to touch any code. So I'm just going to use cursor. I'm going to use the compose model. I'm going to go in. I'm just going to start saying this is what I want. This is what I want. This is what I want. And I went through and I built kind of the first version of it. And it was working pretty well. The one area that I was struggling with was getting it to sync between my different devices. I ended up using core data. Core data is an older version of Apple, right, of data management. Now they have Swift data, which is a little bit easier. But again, because these models are trained in the past, it wants to use some older stuff. So I went through this whole thing. I built out this huge S thing with core data highly complicated. Could not eventually it just became like about 70% is the marker it seems like. 70% of the application, once it's done, the AI just starts losing its fucking mind unless you've really grounded it and trying to give it, unless you know how to really orchestrate it. So I finally, I couldn't get the sink into work and I'm like, okay, I'm done. I said, we're going to scrap that whole whole version. We're going to start completely from scratch. However, before I do that, I go to cursor and I say, hey, listen, I want you to analyze the entire code base of everything we built here. I want you to break out every single thing, every feature, every function, and basically report back a huge document that just outlines everything this application is supposed to do. It did it. And I'm like, all right, cool. So now I started a new project. Take that huge S map, drop it in there and be like, this is what we're building, baby. I'll be back in an hour to see how well you do. And I come back an hour later, it's written all these pages, all these files, all this stuff. I'm like, fuck, yeah. So I hit run. And then of course, I get a bunch of errors. No problem, used to it. Copy all those errors. You paste it back into cursor. You let the AI go through and try to figure out how to fix all the errors. You run it. You get some more errors. You go and you copy those errors. Three fucking days. Three days. I went through this dance of copying the errors, pasting them, letting cursor try to fix it, run it, errors. Three days I never once got it to run. Three days, never once. And so I finally just, I flipped my fucking table over. I'm so angry. And I'm like, okay, I'm done. And now I'm going to go. And I went through, I actually built out my directory structure. I went and I said, here's features. Here's utilities. Here's components. Here's this. I went in and I set up all of the file system. No code, but just the file name. So here's my, you know, here's my, my detailed sheet. Here's my editor sheet. Here's this. All of the things that I knew that I wanted from a structure standpoint. I just created the directory structure. Then I went back to it and I said, okay, now I want you to go through. And now let's create this. And now I'm stepping through feature by feature. Instead of just letting it go raw dog, feature by feature I go through. And now I've got to a point where I've got the application at 90% done. I'm able to like understand it. My cursor rules file is now really long basically. Has all of the way that this thing is supposed to work. Anything like it constantly wants to do like debugging in places that it shouldn't with Swift and it throws errors like it doesn't do that anymore. There is no fucking way in non-developer builds an amazing app. Do you know what that reminds me of how to actually successfully build a software application? Yeah. You can't one shot a motherfucker. Exactly. I was on one shot at this. My son's always like, dude, you got to buff up so you can one shot this. I'm like, where did you learn that A and B, we're not one shot. Okay. We got to level up first. Yeah. No, I agree. And I saw someone on your post said that they built an entire enterprise application. I want to. I heard this one already. Yeah. And so the guy said he claimed he's not a developer and he built an entire enterprise enterprise. Great application using just AI. No, he built the whole thing and you talked to him. Okay, I did. He built the whole thing and we don't have to drag in that. No, no, no, no. So this is actually really cool. So this is Cornelius. Okay, I don't know. I don't know. No, so Cornelius. I'm going to go easy on him. Cornelius is about my height, but he's three of me and he's built like a five. I am totally disarmed. He's great. He's nice. I like him. He's badass. Yeah. So I call him out on it. And I'm like, I think you're full of shit. And I was like, but let's do a demo. Give me your demo. So he brings me up to this application. It's called like Blitzy, I think. And he went through and he gave him all the rules and all the things that he wanted. And it went through and it built out a huge document that basically here's your data model. Here's your structure. Here's your front end. Here's your backend. Here's everything that you need to build this. And it showed him that he's at 80% complete. And I was like, have you run it yet? And he's like, no. And I'm like, that's the, that's the rub. Yeah. That's the rub. Like you have to run this. However, but then he's like, well, but now what I've done is I've taken all of this and I've moved it over to cursor. I'm like, oh, okay. And so he's now he's telling me what he's doing in cursor. And I'm like, okay, wait a minute. And he keeps going. Then we actually met up at half liter this week. Oh, so there's more to this. This is better than heard. So then he shows me his swagger that he's been building and he's been testing against. And I'm like, mother fucker, you're a developer. Like you don't want to admit it. You don't really, you're a new breed of developer. You're breaking down these things. You're understanding. You're starting to see, oh, wait, what's a swagger thing? You're going down, you're breaking it down. You're understanding. You're in code. No matter what. My wife, no matter how many times that she went and she interacted with AI, she would never be building a swagger file. She just, that's not the way her brain works. He is a fucking developer. And if I would have given him a computer at eight years old, he would be a bad-ass developer. And now he's just starting, so he's this new breed of developer who's doing it. And so like what he's done, he's not at 80%. He's not even there yet. However, if there's one dude who's going to do it, it's fucking cornealious. Yeah, I think that is a great example of how AI does a good job of making things more accessible. So someone who probably has the skill sets, the capacity, the understanding, all the logical wiring that is required to be a developer, probably plenty smart enough, but not the experience, it unlocks the ability to get further down the road, get unstuck faster. And now he's got, even if it never runs, he could take that to someone who has maybe more experience in that. And that means he's not going to understand, can't find property, a blank of undefined. Right. Right. And loses mind. Yeah, yeah. But he could absolutely take all of it, go to another development shop, be like, here's where we're at, this is what I need, I've already done the business side, because again, he's really smart on the business side of things already. Yeah. Right. I'm so excited and also like so jealous of this, because we had the same advantages. Like when reaction, all these things came out, like all the problems that were decades old, just were gone, because we created tools to solve the problems. Right. And I remember thinking, trying to learn authentication for like a month, when I watched the thousand tutorials, and I just would get the random errors, and I'd be like, what does that mean? Right. Now, you're still going to get those sometimes, but at least you have a tool that you can be like, please tell me what this means. But by the way, Claude, whatever version of Claude Sonnet I was using, so I was like, hey, I'm going to need the concept of a user. I got through this database, and I was working with my son on this app. And I was like, we are going to need the concept of users. And I'm going to need sessions. I'm going to need cookies. I'm going to need tokens. And I'm going to need, I'm going to start with user name and password. And it sat there and built out the database, built out their token infrastructure. And then it went in and understood how to do it in next 15, and built the cookies, and build all it in middleware. And I was like, well, thank God, because it sucks. Do you think there's like going to be a consolidation of languages? That's kind of, yes. I totally think so. Because I think this is the point I wanted to get at earlier, which is, I think there's a package that are horseshit now. Okay, I need a slider. I need a ORM. I need to do this off-oh thing. Because I don't want to write off. I don't want to write off. Well, okay, go sign up for $30 a month for this app. And then they're going to do the auth for you. Well, now you stay on your ass, right? And they own your data, right? Oh, a slider. Well, now you use slidey.js. Right? The reason they existed is because there's a lot of hard problems that they've solved inside of it. That it takes a tremendous amount of time to do. And AI just ripped off their code and learned about it. Now you can just use that. But that's the point. But it creates dependencies. The concept of the dependency has been in modern software for a long time. So I think that in Ruby they call them Gems or... Yeah. There's a lot of, in .NET they call them some shit. I don't know. I hate .NET. But actually .NET is obviously my best back-ends are all .NET that I didn't build. My clients did. But they're the package concept. But the packages have dependencies. And you can't update the dependencies on your own. The third party owns you. And the restrictive, they have it. It's a box that you have to work with. And it's a huge security. It's a huge security. So if you're trying to get sock to compliant, what is most of the time going to happen? It's going to scan your app and it's going to be like, okay, cool. Well, here's all the dependencies that have problems because your code's fine. But the dependency code hasn't been updated. And so you have to be careful with your dependencies. And one of the things that you, what you can do with AI is have AI write code that you normally wouldn't write. And then you can now edit that code. And maybe that's a double-edged sort. But I think that to your point, I think there's absolutely a consolidation of programming languages, of frameworks, of whatnot. Because the critical mass only gets more critical as people latch on. There's only so much content being created in the Internet anyways. Because there's no incentive to create content anymore in an AI-driven world. Well, maybe a concept. How does a language get more known? So, okay, so I've got it. And this is a complete rough cut. So bear with me as I try to explain this because I've literally just started thinking about this over the last week. But what's Swift, when you're writing in Swift, fucking fanboy. No, no, no, no. Swift is highly opinionated. Yeah. Swift is highly opinionated. Your drop downs are going to look like this. Your toolboards are going to look like this. Your buttons, your menus, they're going to look like this. But it's awesome. But you go through the code and you're like, this makes a lot of fucking sense. And then you go see that paperkeeper right now is 564K total size for the file. Well, because the OS is just deriving a lot of it. Exactly. That's the key. And so why would we not consider having some sort of... Because what they have, all the frameworks are basically built into the operating system. Why could we not have these frameworks built into the browser? Well, what we're doing is that the browser is basically bringing a lot of that UI, bringing a lot of those things. Because we have to build all of that shit when we're doing the... How many times have we talked about this, Jacob, where we said, Tailwind should be built into browsers? Yeah. Right. But Tailwind is the UI framework for the browser. I think it has to go even deeper. I think it has to be even more opinionated. And that used Tailwind fine for the rendering. Okay. Well, then it's an inundation of HTML as well. It might be Zool, right? XUL, if you remember Zool from back in the firefox and that escaped navigator days, it might be like, can they kind of try to do a UI, but it does dog shit? And that was the problem. Swift, though, man, like... I don't think we'll ever get there because you have to basically get everybody to agree on what like a standard UI would be. And Google and Apple will never agree on that. And everybody, it's been done. They're tried. Right. Material, yeah. They don't get me started on this conversation. There we go. No way. No way. There we go. There's just not happening today. Well, I want to finish on one thing that this is kind of an interesting concept. So our overlords open AI won't run. No, you'll pass. You know what I'm asking now. So I don't know why they did this, but they released this economic blueprint. And I'm just trying... I read it and I'm used AI and I'm trying to understand what the hell they were talking about. But I think that's because Trump's becoming president. I mean, I think... Is that why Zuckerberg or Gibson... I mean, so open AI will just publish these PDFs and they'll be like, this is the way... And like, I'm reading this. So they released this AI in America, open AI's economic blueprint, right? And I don't really understand exactly what they were trying to get at. But there's a whole mission in vision. And basically what they're trying to do is disarm everyone to say, we're not evil. You're not going to take your job or goals and take your job. People probably get a loser job. We got to solve hard problems. But we got to focus. We got to focus on healthcare, education, scientific discoveries, public policies and productivity. We're making a lot of great impact. We're doing support and teachers. You know what you... What's the deal about you, Be Ital? No. And did you see it in my note in there? That was like... It was like, all right. If Zuckerberg was just firing everybody at what point, where he just all like... So basically everyone's like, Sam Altman's this guy, he's like, he's a startup guy. But like, let's be honest, that guy has no business making decisions that affect the future of millions and billions of people. I mean, we have a fundamental problem in this country where money equals success, equals intelligence. So the three smartest people in the world must be Zuckerberg and Musk and probably like Gates or you know, Bezos or whatever. They're so smart because they have so much money, right? Yeah. So they should make all of our decisions. The reality is that this guy opened up Pandora's box and he's trying to jam it down. You know what I mean? You know, and for... But there's a lot of disguise stuff. So like, if you've read a lot of Sam Altman stuff, he'll put something out there for a reason and in reality it's to cover up for something else. Don't trust this guy. It's the same guy that built OpenAI as a non-profit and went dark for six years. And then somehow they're like... Well, we're not a profit. I ran Y-combinator. You think I'm gonna run a fucking non-profit? But what it seems like they're doing in reality under the hood with this economic bullshit is they're trying to get the government to subsidize them creating the data centers in the Midwest. That's what I'm the fucking reading. They want to nationalize AI infrastructure and get the government to pay for it. Of course, they want to put a bunch of nuclear power plants all over the Americas so they can run their fucking 50... You know, thing that costs $20,000 to run a single task on. Yep. They want... And all of these things... You just fucking fuck them. All of these things would... So here's what really, really annoys me. These things would be good if we actually had good guardrails around them. So it's like the idea of a doctor, someone the other day said, Oh, a doctor could say, see 10x the patients or a doctor could provide 10x the care to that patient. Why do we always think of it as like... We're always like, oh, we could do more. I'm like, why don't we do better? Better. Because better mean good for better. Exactly. Exactly. Quality over quantity. But that's the quandary with America. Yeah, it's like capitalism is awesome. There's created so many opportunities. Europe is so stagnant because of some of their issues and their government disconnected from their business. And there's just such a problem there. And the free market has always won. But it's like... It's messy. It's the best system that we have so far. It's sucks as well. It's the best thing we've got. But it frustrates the hell out of me. Because I'm not even a star track guy, but my brother is not just like, yeah, what are we doing? Well, I think what it's like, comes like, if you look at, for example, I don't know if you saw Bill Burr on the late show or whatever. Oh, God, he's so... I mean, they were like talking about the fires. He was like, man, you guys think it's homeless people? The guy went, what do you do right? A hang glider from palisades to whatever. He was like making all these examples of why, why, like everything that people are saying on social media is dumb. He's like, this guy in his underwear is saying, yeah, they're doing a terrible job managing this as he sits in his underwear. Like, you know, and all he's saying is basically like, you know, fuck the social media warriors. But he's got some points. But I don't even know where I was going with that at all. No, that was hilarious. So Bill Burr is, he is a godsend when you do appreciate him. Because you also see that he, as he's gotten older, he's gotten a little bit softer. Oh, yeah, yeah, that's not... I know my point. So what he was saying, he said the thing that you don't want to talk about. He commented on the UnitedHealth care thing. Right. And I'm not doing that. Yeah, yeah, yeah. I don't know. I will. Don't. Not on this podcast. We'll just leave that. Just leave it. But the one thing that they were talking about is that the outcomes are important. Right? If you want to say, oh, the doctors want to do better. At some point, public sentiment does drive the market. Right? And so people having that are enabled with good information, that have good moral basis, they can drive the market. Right? But right now, that is not there. Right? So it's up to us as a people to help the market. Because the best solution we have is free market. People need to do better to help drive the free market. It shouldn't take martyrs. Right? It shouldn't take that type of stuff. Right? Like, it should be something that says, you know what? We're not going to buy from you. Yeah. You know? It's just very complicated because there's misinformation and you know, all these distraction techniques and we can't agree. We can't, we all, I know we all want the same thing. I know it. I know it in my heart. I've talked to too many people to be like, yeah, we all want better lives. But we just can't agree on the path. Because of Charlottesense or keeping us divided. Yeah. And it's the key. It's just keep, keep them divided. That will always be the angle. Keep them divided because if they're not divided, that's a real big fucking problem for us. Yeah. Right? So the people divided keep systems in place. People unified, break fucking systems. So hopefully, we'll break systems eventually. I like the ending on that. This has been the big cheese podcast. We're so glad to be back and hopefully that we can see you guys here next week. Same time. See you. Boom!