speaker-0: Welcome to the Based Academy, live from the back rooms. We are your hosts, Sylvan Franklin, who's not here, he's biking to a coffee shop right now because he doesn't have internet. And Speak Machine. That's me. Our guest today is a low-level penguin enjoyer from the land of carnivorous plants, kangaroos, and cold ones. Matt from Houghworld. Yay!

speaker-1: Hey everybody. Good to be here, man. Thanks for having me on. This is actually like my first sort of live podcasting that I've done with any other creator. So we'll see how we go. We'll set the expectations low, chat. We'll set the expectations low. Yeah. And then expect nothing, and then you'll never be disappointed. That's how I live my life.

speaker-0: Expect nothing. That's that's a pretty b that's a pretty actually ⁓ let me put a pin that in a second 'cause I wanna I wanna get back to that. Because that's actually a really good point. ⁓ I started live streaming like a week ago. I was in Finland. It was around midsummer, which because there's no sun, there's basically no sun in Finland, they celebr like anything to do with the sun is is like a big thing, they celebrate it. So like midsummer is a huge thing. And I just on a whim, I just turned on my my cap my my camera and I just started live streaming in the forest. And like Like thirty-five people showed up, like just w without any just no no nothing, just live within like ten minutes, like thirty-five people are watching me rant in the in the finished woods. And it felt it was very exciting. It was like it was like, you know, I could get like thirty thousand views on a video. It's like that's pretty good. But it's it's 'cause it's so abstracted away. You don't you don't feel it. But when it's like thirty five people with your eyeballs like staring at you, it it feels entirely it feels entirely different. ⁓ so I totally get that. But the the expectations, I think that's a huge one. I was talking to to my dad about this. I s I said to him that I think I'm one of the happiest people alive because I have such I have such low expectations. Like everything that I do, I I expect nothing to come of it. When I started YouTube, at least at least this channel that I that I have, I had zero I was expecting to get like five views. I was like, yeah. You know, maybe some autistic person somewhere wants to watch me do a Vim config or something. That might be pretty nice. ⁓ and then I make a video. I made a video, I think my my big my first big first video was Computer Programming Is Finally Dying. It was something like I think it was called something like that. And they got like twenty thousand views like pretty much immediately. It actu I actually got featured on a fire ship randomly. Like here yes, people like went crazy. They they were commenting So he was talking about how programming is 'cause of course programming isn't dead, right? It like a bit of a clickbaity kind of thing. And ⁓ he he he says he s it was like a joke, like and then people started talking about how programming is dead and it shows a screenshot of like YouTube search and my thumbnail is my video is just there. I f it felt crazy to like, ⁓ wow. Okay, I'm I'm on a I'm on a corporate owned channel. That's really that's really

speaker-1: Yeah, yeah, yeah. You're in the VC money now. Dude, I I Speaking about the start of your channel, I actually do have a question for you about that. Because the first time I came across your channel and I saw the name, it it seems like almost philosophical. Like speak machine. It seems almost philosophical in nature. Is there like a an origin story for that or why you chose it?

speaker-0: They've got their sites. Yes. ⁓ I like I like the fact that you had the pause and but there's there's supposed to be a pause of course 'cause it's a it's ⁓ in the in the name. ⁓ yes, ⁓ I wanted so the point was computer, reveal your secrets to me. Right. That's essentially the meaning. And it was supposed to be speak computer originally, but it if it doesn't flow quite it's like speak computer doesn't it doesn't quite flow right. So I ended up going for speak machine. That is the that is the meaning behind it. It's It's it's supposed to be like learning about computers and h and their sequence and how they work. And then six months later I'm I'm ranting in the woods in Finland and talking about philosophy.

speaker-1: Well, it really what you said there really like resonates with me because when I think about especially low level programming, it's almost like I'm having a conversation with the computer. So it kind of is like the machine and me are speaking to each other, you know. So when I yeah, like the first time I came across your channel and I saw the name, it it kinda just like the neurons just went fire and boo blah blah blah blah and like, Yeah, that's cool. That's cool.

speaker-0: ⁓ wow, hello.

speaker-1: No.

speaker-0: I like, is is there a plane taking off or something? Yeah.

speaker-2: ⁓ no, you can hear that?

speaker-0: Yes. Don't worry, we c we can remove that in post. We can fix it in post, it's fine.

speaker-1: Is this ⁓ if if this like podcast online platform is supposed to have noise cancelling, it's not working very well?

speaker-2: I like the ⁓ the haircut as well, Christian.

speaker-0: Bro, I let's not even I re I regret it so bad. I was like, I I I can be a bit impulsive sometimes and I was thinking I c I kinda I was thinking 'cause like, cause my hair is really white. I'm 34, but my hair is white. It's just genetics, whatever, right? Six foot four, handsome, smart, charismatic, intelligent. You can't have it all. You have to have something that's like, know. And I thought I think it kind of ages me to have To have ⁓ white hair and then I sh I buzzed it all off and then I went, ⁓ no. No. ⁓ I have a good head shape, so it's fine, but it's like I can do Kingpin ⁓ cosplay. That's that's pretty good. Or Baron Harkonen. We could do like a Vim Vimothy slash Baron Harkonnen thing. ⁓ getting back on track. So ⁓ Matt, you were talking about low level.

speaker-1: Yeah yeah. That's one of the things that really gets me into it is like when you're dealing with so many levels of abstraction in the higher level thing, yeah, it feels like there's so much b well, because there is. There's so much in between you and the actual machine. So it's hard, like imagine if you were talking to somebody in a different language and everything had to go through Google Translate, right? It's kinda like that. But if you both speak the same language, you can get along much better. And that's kind of how I think about low-level programming. It's like When you are actually dealing with things directly, especially in embedded programming, because in in embedded, you're changing the hardware registers directly in order to get it to do different things. That's so much different to the feeling of when you're like programming in Python or JavaScript, right?

speaker-0: Well actually speaking of ⁓ this is gonna look kind of dodgy for a second. I can't believe I didn't have this ready. ⁓ man. I've got like my computer sitting on

speaker-2: Janky our podcast is

speaker-1: I thought I was coming onto a professional studio like

speaker-0: I know, I know, I'm so sorry.

speaker-2: ⁓ it's part of the experience. It's human.

speaker-1: This is like a hazing ritual or something.

speaker-2: You were leaving all of this in, by the way.

speaker-1: I hope I hope so. I hope so. Yeah. ⁓ my god.

speaker-0: So in my hands I have a box, right? And within this box I

speaker-1: Ooh, nice, that's a STM thirty two board.

speaker-0: Yes indeed.

speaker-2: You just recognize that off s off rip like no

speaker-1: Logan isn't worth it, what should I say? Yeah. Yeah, yeah, yeah. He he gave me an easy one.

speaker-0: Very recognizable. That's true, that's true. I I can look it up in a second. I I don't have ⁓ 'cause I I bought it just before ⁓ just before heading out to to start travelling.

speaker-2: Background, could ask a question? Like, could we get kind of philosophical here? Like, do you think of ⁓ like the APIs that you use to program computers? Like, do you ever see those outside of a programming context, like in the real world ever? Because I know for me, like I have switched over to to being really interested in human behavior. And like I now see like people as having APIs, but I wonder if you've ever had an experience.

speaker-1: What like if you're talking to a girl you've gotta hit the like flirting API and you don't have authorization you don't have authorization sometimes. It's like

speaker-2: There are a couple of ways to bypass the auth, but ⁓ yeah, you you you can't do that in an unethical way.

speaker-1: Dude, I actually Okay, okay I I can make a thought about this because this is maybe analogous but not exactly the same thing. I have thought of ⁓ for example, I don't know if you guys ever played like Nintendo DS and you ever had like cheats on the Nintendo DS where you could just like press some buttons and like change the behavior of the game ⁓ with like an action replay or something like that. I often wondered if like having free will. And exercising free will is kind of like having cheats for the video game because there's all of these like social norms that you're not supposed to violate or like things that you're not supposed to do. But if you if you have like the willpower to overcome that, you just like press the buttons and you just break the game. Like you could just go talk to some random person on the street. Why don't you do that? Well, it's because it's there's this like social norm thing about it. You might look weird, you might do this, you might do that. But if you got cheats, you just press the button, you go do it. There's no problem.

speaker-0: Man, we've hit a gold mine. Like less let's stick with this. This is this is a gold mine of a conversation. ⁓ Sylvan, were you about to say something? 'cause you you unmuted yourself. Were you about to say something?

speaker-2: Well yeah, just a real quick comment. Yeah, I I relate to that absolutely as like a as a cult leader. Everyone says, No, you can't be a cult leader, you can't do that. That's not an allowed profession and now it's on my LinkedIn.

speaker-0: ⁓

speaker-1: biggest flex is if you can leave that on your LinkedIn and still get hired. That's the that's the biggest flex. Yes.

speaker-0: See, I I had this realization, I think I was I was like ten or something. ⁓ I realized because there was this point in my development where I realized that I I really wasn't like all the other kids. And I know it's a meme, like I'm I'm not like the other girls, but like literally there was this magical ⁓ so nice nice hat, nice nice double hat, Soviet. ⁓ there was this magical moment where I realized that not everyone thinks the same. And there's this magical territory between how you think and how other people think. And it's all explorable. Like everyone just kinda sits in their little he's just riding his bike in the

speaker-1: I'm trying focus on what you're saying and I'm seeing this in the background like

speaker-0: Premium content right here.

speaker-1: You guys plan this, right? This whole thing's a setup.

speaker-0: ⁓ y yes. Y yes we did. This was entirely w were that smart with that good F. W bro Yeah, you yeah, I think I think it's good.

speaker-1: Is are you okay?

speaker-2: You you you can't just have like normal talking, kids won't pay attention to that these days.

speaker-1: This is like Saboy Self. ⁓ I think we're gonna get along just fine.

speaker-0: Yeah, I think we're

speaker-2: Go on, go on.

speaker-0: That's when I realized that there is there is a lot more to to the to to the world that we realise, that we that we think about. People never quite people never really leave the boundaries of the of the game of the playable game area.

speaker-2: Would you say that was easier or harder than reading the entire Vim manual?

speaker-0: Much harder. Much harder. Also if ⁓ Matt if you don't know, I have a video on my channel where I read the entire Vim manual. It's eleven hours long. It took sixty-nine

speaker-1: That's enjoyable. Wait, really? That's crazy.

speaker-0: It because I I cause I had to read it and then edit the whole thing because it's cause it's it's high because I couldn't just do it in one session, I'd make mistakes. So it took like sixty-nine hours of so it it's enjoyable to read, not so enjoyable to vocalize and edit and then render the whole thing. I I I literally my computer got so hot. Luckily I was in Norway in winter, like I had my computer rendering overnight next to the open window. That's not even a joke, by the way. Try rendering an eleven hour video with effects and like you'll you'll see you'll see what that how that goes.

speaker-1: I respect the dedication on that. That's crazy. That's crazy.

speaker-2: I I just wanted to ask Matt, like what ⁓ what kind of low level projects do you work on?

speaker-1: ⁓ it depends. Like there's okay. I think because I revealed that I got a this big fancy job, right? People are gonna find my LinkedIn anyway. So I might as well tell you guys about my like what I was doing before that, which was a really cool low level job. Is probably one of the most Australian jobs you could probably think of. Have you guys seen on like Instagram there are those like flying surfboards? They're they're like a surfboard with a mast and a set of wings on the bottom. And some of them have an electric motor on the bottom where you can just like pull the trigger and it'll the electric motor will go and the thing will come up out of the water. I I wrote I used to work for a company that that made those. ⁓ and I wrote the firmware that runs on them. Well, I didn't do all of it. There's the guy, ⁓ I know Sylvan you've talked lot about having like mentors and stuff. My mentor at this previous company was like cracked out of his mind. Like he could do all the electric motor control stuff, the Bluetooth stuff.

speaker-2: Dude, that shit's fire.

speaker-1: All the way, like we had an app, he could do the app stuff and the electronic stuff, like the PCB design. He was just across all of it. And when I saw him, I was like, this is who I want to be when I grow up, kind of thing. But yeah, I learned heaps from him and I did a lot of really cool stuff for that company. And it's yeah, it's it's a lot of just ⁓ you reach for your STM32 board. It's kind of on that kind of level. It's like you're programming microcontrollers, you've got a whole bunch of different little computers, because they're just little computers, right? That have to like talk together and do things in sync to make something happen. And ⁓ yeah, that was my last job. So as a firmware engineer there.

speaker-0: The Sokobasmia project. So I know you had the one where you changed the ⁓ the colors ⁓ of your of your light. That was that was ⁓ that was some really cool stuff. Like how did you how did you get into low level?

speaker-1: This is a really interesting question. ⁓ and it actually goes back to my sort of like Joker origin story, right? And I I wanna hear what your guys' joker origin stories are as well. So you'll you'll have to tell me later on. But ⁓ when I was a kid, this okay. This might be really weird coming from a guy who now works at Microsoft and previously was like a super big Linux guy. But yeah, I know, crazy reveal, right? Crazy. No, dude, but okay, when I was like 12 Yeah, yeah, yeah, yeah, yeah. I'm sorry, I'm leaving the chat right now. When I was twelve, I went to an Apple store, right? And I played with one of the iMacs there. And for whatever reason, my like twelve year old brain just like latched onto that operating system. And I was just super obsessed with Mac OS X at the time. ⁓ and so what I did when I got home was I Googled as a joke, how do I run Mac OS X on a PC? And I was just expecting it to be like because nobody I thought nobody can do that. It's like an Apple computer, right? And what surprised me is there was this big community of people doing this thing called Hackintosh. And so when I was like 12, 13, I got really into Hackintosh stuff. And to do Hackintosh stuff, you need to learn a decent amount about how Mac OS X works as an operating system. And playing around with that, and you have to like download and install custom drivers. They call kernel kernel extensions and stuff. And that kind of introduced me to what a kernel is and ⁓ all that kind of thing. So That was kind of the starting point where I got into like operating systems and low level stuff because I tried so hard to get this thing to work on my shitty little laptop and I like got a half working and I was very happy with that.

speaker-0: It's interesting how ⁓ 'cause I mean I I ⁓ I also I also had a Hackintosh. ⁓ like 'cause I I really wanted a Mac but I couldn't afford one. So th so that was that was the lows of the logic. Like I have a PC, I love Mac OS. I c I I was using them at university, so that's how I got introduced to them. And I literally made a Hackintosh because I couldn't afford to buy a Mac and here I am now flying to two different countries to buy a l a MacBook. We can get into that a bit later. ⁓ but

speaker-1: I sick.

speaker-2: Wait, I I already ha I I I also have like a a Mac desperation story. Like it's it's i it's not as bad as that, but like I when I was gonna get my first laptop when I was like I wanna get into coding at I don't know, thirteen or fourteen or something, like all my friends were like, You should get a Windows laptop so you can like game with us, you know, and they're significantly cheaper. ⁓ but I was like, No, I I need to get a MacBook, like the the The aura of it is just like, dude, it you can't resist it. Their marketing is too good. And so like, yeah, I did like, I don't know, probably a couple of hundred hours of yard work just to be able to afford this like shitty Intel i five laptop that was like way below the Windows standard. But then once silicon came out, like they just they're still crushing it.

speaker-1: Yeah, dude, for in my opinion, it's kind of flipped a little bit because when I was doing the Hackintosh stuff, I got into it maybe like twenty thirteen, twenty twelve. In that kind of era, in my opinion, like Mac OS X was one of the what the best operating system out of all the big three in terms of what I was looking for. And Apple was like fully on top in terms of software. But honestly, I think and but the comparison was their hardware sucked. That's why you a Hackintosh. Because their software was great, but the hardware sucked. In my opinion at least, and I might catch a bit of flame from Mac users who are in this chat, but I think it's flipped now. Like their hardware is off the charts, crazy. But I'm just not a fan of what they're doing with their software. So that's why once support for like I I did a Hackintosh thing. My Hackintosh was like kind of working for a long time, but it had all sorts of problems. So I just switched back to Windows seven for gaming. And then when Windows seven went end of life, I just switched fully to Linux.

speaker-0: see I I I I don't think I don't think you should get any hate for that. I think that's pretty legit. I think I think that

speaker-2: Yeah, I agree. Good take.

speaker-0: I don't think anyone makes better laptops than Apple in terms of hardware. Like they're just superior in every they're everyone's trying to copy them, but they just it's just not the same thing. I had an M1, now I have an M4. My dream is to have like this M4 just run Linux bare metal. And I know about Asaki. I'm aw I'm aware of Asaki. It's just not if you want to do this professional stuff, you can't use Asaki. Cause well, and you can't. Like as a video editor, I was thinking, should I get like a a normal laptop and just install like Arch on it and just use that? But I don't wanna have to call my c my clients and say, sorry, my Arch installer is broken. Can I get back to you in like a week? Like that's just not that's just not an option, right? And so I have to get a a MacBook. I Mac OS is going in the wrong direction. ⁓ it's it's ⁓ it's not it's not doing very well. But the hardware is so good, I just had to I just had to get another MacBook. There was just there was just no way around it. Is that a hundred percent on that?

speaker-1: I'm gonna sound like a total shill here, but I got one of the I got a brand new this. I looked up the retail value on this thing. It's like ⁓ 3,800 Australian dollars, which is probably around like 3,000 US or something. It's for work, they sent it to me. It's a brand new, like one of the Qualcomm Snapdragon Surface Pros. And it's actually, I was surprised because I haven't used Windows Arm in literally ever. And everything I heard about Windows Arm was a sucked. I'm gonna sound like a total shill here, but I'm really enjoying using this. Like genuinely, I'm really enjoying using this laptop. It's really quiet. And this is probably like because it's ARM based, you know, more than anything else. But it's really slim. It's got all the ports that I want. It's very quiet. It feels like a MacBook. It feels like an ARM MacBook. And the software like Windows on ARM has come a number of like quite a way. So that's what I'd say on that. You can put Linux on them, I think, but obviously I can't, because it's a corporate laptop.

speaker-2: Is anyone else getting kinda like freaked out by how good the software is getting? Like even the the things I don't like about Mac OS, like just there's like an an update every day and like every agent harness has an update every day and every like everything is just moving so fast now.

speaker-1: Do you guys think that's a good thing or a bad thing? Because okay, here's a question for you actually. Would you say over the last ten years, as a general concept, not pointing at any specific company or companies here, has software gotten better or worse in the last decade? What's your opinion on that?

speaker-2: I would say like it's got well, kind of both. You know, that's a really complicated question because like some people have done really well and other people have done a lot worse. And in my opinion, that's like less of a software issue and more of a people and regulations issue. I know for the medical industry at least, like that's kind of where I'm involved in the software world. Like the big three companies are all just in this stupid deadlock of like suing anyone else who gets within range of them and like It's too complicated for anyone else to tackle it. And I I would say that's the same thing for some of the bigger companies like Windows as well. They're like the way they've structured how people work together is just like so bad that they've kind of started to stagnate. But like by and large, I think software is getting a lot better.

speaker-0: Interesting. Interesting. I I ⁓ so I think that there's a lot more software. That's obviously the case. It's exploded in how much software there is. I think that software in general is getting worse in terms of in terms of quality and stability. They're introducing features for the sake of introducing them, whether anyone wants to use them or not. Like I don't know anyone who uses Apple intelligence. I've literally never met a single person who who uses it. ⁓ my god, no, no, okay, I know one person now.

speaker-1: Interesting. ⁓

speaker-0: th there's so many like now with AI, new feature, AI, new feature. You can can YouTube's like, hey, I can summarize all the comments you got on this video. Do you want me to summarize all the comments you got on this video? No, I'd rather just read people's comments. Why would I want to summarize if they bothered to watch my video? I'm gonna bother to read the comments. I find myself looking for software all the time. That's like all the software that I can use. Just because it it just feels a lot more reliable. But yeah, my take is that software is getting worse as as an overall trend.

speaker-1: It's very interesting that we got two points here.

speaker-2: I would say actually like what like there's there's a lot of really, really dumb stuff going on because of like the the ⁓ kind of CEOs who are not technical like grabbing these features and like spamming them everywhere. And that's caused like a lot of dumb software. But you also have like really good developers like Jonathan Blow, Casey Muratori, ⁓ Ryan Fleury, like the people who really go Deep on these things, like being massively accelerated by agents at this point. There was a time when vibe coding sucked, and it's vibe coding still sucks, right? But agentic engineering, in my opinion, doesn't suck. Like if you were a good developer, it's a massive productivity boost to have like well-tuned agent harnesses working for you. ⁓ it's just like it it took people kind of a while, and it's still taking a while to kind of find. how these AIs are actually gonna work. And this is like kind of leads into a broader philosophy I have on AI, which is you can't produce anything better than you would be able to produce manually with AI. So I I think of AI as like this force multiplier of your cognitive output. And if you have bad ideas, you're just gonna shoot bigger bad ideas. If you have really good ideas, like if you're Jonathan Blow, you're gonna be able to move way faster, in my opinion.

speaker-1: Yeah, I generally agree with that. I would say in terms of when I look at how like productive I am or whatever, when I'm talking about maybe making stuff for YouTube or doing my own projects or something like that, I'm actually not so concerned with how fast I get things done. I'm concerned I've I've this is my sort of philosophy on how I'm using AI at the moment, and I'm curious to hear you guys as as well. I'm doing this thing I I like to call diff maxing. Which is when I look at a diff that I'm about to check into my repo, what is I'm trying to optimize and maximize how much dopamine I get by looking at that diff. Meaning like if I ⁓ let's so let's say AI uses a spectrum, right? There's like full manual code and there's like full vibe code. Either way, at the end of the day, I'm looking at a diff that I'm about to check into my project. And once you it's interesting that you mentioned people like John Blow, Casey Muratori, et cetera. I have this Take that I want to hear you guys' opinion on as well, which is that if you want to be a truly great programmer, you have to be opinionated about programming. Because I think fundamentally, if you are opinionated about something, it means you care enough to have for the so for example, even Uncle Bob, who is the clean code guy, I don't agree with his ideas. But in terms of great, great doesn't necessarily mean like good, right? It just means like significant or or whatever. He's certainly significant. And he has very strong opinions about how things should be done. And Even if I disagree with his opinions, I think that's respectable because he cares enough about what he's doing to come up with those opinions in the first place. And so when it comes back to using things like AI and AI agents, I have my own opinions that I've formed about how programming should be done, how code should look, how it should fit into the overall system that I'm trying to design, and what paradigms I want to use. And I can look at a piece of code or like look at a diff, and I just like check that out with my own opinions, and I'm like, Do I like the way that this is going? If I'm doing my whole like prediction stuff, I'm predicting in the future, is this gonna cause problems for me based on my experience, et cetera, et cetera? So I'm not necessarily focused on how fast I'm doing things. I'm thinking about like, okay, if I coded this fully manually without AI, or if I fully vibe coded it, where in this spectrum do I get the maximum amount of satisfaction from looking at that diff based on my own personal feelings about how how programming should be done. So I'm purely focused on what is the optimum place where I can get the most ⁓ quality based on my own judgments.

speaker-0: Context, I'm a ⁓ recreational programmer, as as they might say. I don't I don't work as a programmer. I mean, I guess we should also do like like the the origin story here, because just it just adds more context, right? I started learning how to program when I was like seven. My parents had a Windows computer, Windows 98, best best Windows, by the way, ⁓ in my in my opinion. I was enamored with it. There was something so beautiful about it that I couldn't explain. And my parents went, ⁓ he likes computers. So they got me a tutor, very, very thankful for this. ⁓ I actually had two different tutors. I started, ⁓ I did Pascal Visual Visual Basic, I did some C. We started programming on paper. Like we would we would lay out the lot, the logic. Just looking at Sylvan. We would lay out the logic ⁓ of the program on paper before we ever typed typed anything. Then I ⁓ I was told that to give become a good programmer, I w I needed to be really good in maths. And I had a terrible maths teacher and I thought I hated maths. So I went I went up going on this, I went on this rabbit hole of doing music and film and photography and blah, blah, blah. And then a few years ago, I had a channel and I hired a programmer to help me automate some back-end stuff for my channel. And then he just abandoned the project halfway through. He was like, Yeah, I'm I've got stuff to do bye. So I was left with like a thousand lines of Python code. And I thought, okay, I know how to program, I know algorithms, I know a bunch of stuff. Let me just W I I watched a six hour Bro code video on Python and then I just spent the weekend going through

speaker-2: ⁓ I remember Bro Code. Yeah, I l I learned I learned a lot from him.

speaker-0: It was a magical moment. And then I spent the weekend just reading through this, like looking at all these functions and figuring out what everything does. And there was this beautiful moment where I got it. And it just it just like in one day I learned about APIs and how like a lot of a lot of stuff I just didn't know about before. And it all clicked. And I realized, ⁓ my God, this is I forgot how fun this is. I forgot how incredibly fun this is. I ended up

speaker-1: And

speaker-0: thinking, do you know what? Let me just get back into programming. I've been doing it on and off for years anyway. I always kept go going back to it. 2013, Java and Eclipse, I know I was doing Java. I was doing Java for some reason. Not even professionally. I just I just liked a bit of Java. And now I'm here, right? So I personally don't use AI for programming at all. given that I'm in a position where I don't need to L just like how Casey Moratori says, he doesn't use it because he enjoys programming. Like if you if you like playing drums, you just play drums. You don't you don't use a drum machine because you enjoy playing drums. ⁓ just like how you said, like for your projects, you don't you know you just do them manually, or I assume that's what you meant, because well, you enjoy programming. So there's there's no reason to to t the the how efficient you can be at things that you enjoy is irrelevant. Like if you enjoy riding your bike, You don't really care about writing it as efficiently as possible. You just care about enjoyment maxing ⁓ this essentially. So for learning and for improving, I don't use it at all. ⁓ I've never tried ⁓ I think I tried claude code as like a curiosity to see what it does. I did I tried Chat Gipity for programming early on. I will occasionally use ChatGipity f to generate like some bash code. If I if I have like a one use one time thing that I gotta use once. I need to use bash. Sure, give me the code. Damn it, it's broken. Just do this, bro. ⁓ yes, yes, of course, certainly. And then it just like a very obvious fix that for some reason I just can't figure out. And then eventually I get some code, I use it, I discard it, I'm done. That's what I use AI for. However, I if I get hired as a developer, I'm probably gonna have to start using it to some degree. But my my use of it is almost

speaker-2: Zero. I can I can give my ⁓ my coding background story as well. So ⁓ I I'm I'm actually like an artist. I'm not a programmer at all at heart. Like I'm trying to cram the ⁓ the foot into like a super misshapen tiny shoe there. Like when I was in high school, I learned like the first thing I learned was ⁓ I think React and then I very quickly switched to Svelte, but like I did not care about state, I did not care about using Well that was it was what everyone was doing. You know, like I mean I I got I got rid of React like real quick, like like frame one I dropped it. ⁓ but yeah, I I was spending more time like I memorized every Tailwind utility class. ⁓ like I would just try to stylize websites, that's what I was into. And ⁓ like every project I made was essentially just a front end. Like I I really did not like databases. ⁓ or concurrency or distributed systems or anything like that. ⁓ what was that? Did my

speaker-0: Apple intelligence.

speaker-1: Guys, we found the windows.

speaker-2: Weird. Yeah, that is weird. Why did they come up?

speaker-0: It's listening. It knows.

speaker-2: Yeah, yeah. There the the coincidences, the synchronicity has been on the rise. I don't know if it's been for you guys, but yeah. Things weird coincidences. Yeah. ⁓ but yeah, I'm I'm definitely more like the ⁓ tortured artist kind of ⁓ like I'm not a I'm not a John Blow type of guy. I'm like a Carl Young type guy. And ⁓ so the most technical project I ever built, I think, was this fractal ray marcher and I say I built, but like it was really my roommate who did like all of the ⁓ the C and OpenGL stuff. Like I sat there with him. I kinda understood the like rendering pipeline, kinda, but like I was just there to change the colors and like make cool looking fractals. So now I'm like obsessed with this language types, which is like basically just like if if JavaScript didn't exist like it's like well, it's like a scripting language meets like a styling language, but there's like Literally no public space. There's literally no like it it just outputs to PDFs, PNGs, like SVGs, right. Like that's the compile target. And I use harnesses and agents quite a lot because I don't like the logical stuff. Like I I learned the ⁓ the functional bindings, you know, the like the map, rep, ⁓ reduce kind of Haskell y programming style so that I could use types, but really I just want to make cool things. So I think it's like a brilliant translation layer between English language and graphics output. So I have like a template where I describe to the agent like, you know, this is I have a bunch of different types of documents. Like this is a thumbnail, this is a text overlay, this is like a credit thing. This is like I'll have my ⁓ like the people that The boost my server, I I take a like a typed slide and I compile like a series of PNGs that show like the boosters like progressively being rolled out. And then I overlay that with like ⁓ the BBC Sounds archive has like every everything that's been on the British broadcasting network. And I put like this ⁓ this typewriter effect ⁓ and I like cut that up and replay it and like I don't wanna think about the logic here. I just tell the agent, like, all right, use FFmpeg, snip up the audio, use image magic to like color this, like lay it all together, and then like if I don't like it, it's kinda like a I feel like a potter, you know, and like spinning it into existence. Like the implementation details, once in a while I'll go in and check with NeoVim, but most of the time I don't really care. I just want something that looks cool because it's like not it's not prod code holding up ⁓ anything important.

speaker-1: I think that's fair. And actually like even before the AI agent stuff came out, people would have that like, ⁓ if it's just gonna be used once, does it does the implementation detail really matter? Like my first job, I was pretty much just like a script monkey where I would take Excel spreadsheet in one format, transform it into Excel spreadsheet in another format, and like low-key today that job wouldn't exist because of agents. But ⁓ at the time it's it's kinda like that. And all these scripts that I would write to do this thing. It didn't matter how fast they ran. Like, you know, it's all in Python. If it takes five seconds to run, it could have taken like 500 milliseconds, but does it really matter? Because the goal is not the implementation. The goal is what do the document you get at the end of that. And I think correct me if I'm wrong, but that's kind of a similar thing to what you're doing here, where it's like at the end of the day, your output is not necessarily anything to do with the code. Like if you're designing a production software system, then The implementation detail matters a lot because in that instance, the experience that your users have is directly based on their interactions with your software. But if you're doing something to get like a visual output, I mean like LaTeX could be another example of that, where does it really matter how bad the person's LaTeX code is if their PDF document at the end of the day looks good?

speaker-0: Man, we should have like a little bell. Whenever some whenever a guest says the word baste, could we we can just ring it. We can just ring it. Also can I just say shout out to Matt. He made his bed for the for the interview. Like look at his bed in the background and look how look how tidy it looks.

speaker-1: Dude, I'm not gonna lie, I actually did. I actually did. Cause I was like wondering, does this does this platform that I'm being invited to have like a blur effect or not? You know, I clean my room just for this. Every other day of the week. It's cooked.

speaker-0: Man, that's so fun I I I that's so funny, I appreciate that. That's so cool.

speaker-1: No guys, I make my bed every day, I promise. I swear. Yes. First thing when I wake up. First thing when I wake up.

speaker-2: Should return real quick to the an interesting thing you said about the Excel sheet being like automated away because I know quite a lot of people at like different companies that my friends work for that are like getting laid off now because a lot of the kind of menial tasks are being automated away. And it seems like I'm I'm haven't looked at any economic statistics or anything, but like this is kind of a catastrophe because ⁓ it There there have been fake jobs forever, right? That's been kind of a a a thing with the ⁓ the boomer economy. But like the veil is being ripped away right now pretty brutally. And like we're gonna need ⁓ a way to like pay those people a living, you know, because otherwise like there's gonna be riots in the streets. Like it's kind of ⁓ it it's a an an interesting conundrum.

speaker-1: Are you talking about fake jobs specifically in programming or just in general?

speaker-2: Well, in general, but I'm more talking about like white collar jobs, I guess like translating a file format or like a lot of the jobs I've heard of is like you're taking like copying an Excel sheet and sending an email and like now just a ⁓ a competent engineer can that's like a thirty second prompt.

speaker-1: Well, I can tell you about the reason why I made the comment I made about my first like programming job would be like kind of automated now, is because back then I was working in traffic engineering and we had a lot of data, like geospatial data. So that's working with like XY coordinates that you would plot on like Google Maps or they call it a GIS. ⁓ so if anybody has ever used QGIS, it's a great open source like mapping tool and and whatnot. But basically There's like all this spreadsheet that comes in. It was working like transport. So the government transport had all this data and we just needed to transform it in different ways and plot it on maps and stuff and and do whatnot. But it was essentially just like take the Excel spreadsheet from one format and then put it in another one. And there were so like this was not a something you could just do manually because there were like hundreds of these spreadsheets in the same format and they all had to be converted into this other format. And it was a whole big thing. So My job from the get-go, this was before AI. So this was before Agents. It was like 2020, 2022. And that was my job. My job was to write Python scripts to do the transformations. So they would give me a new format, a new like Excel format, and give me an output format and say, write a script that takes this data and merges it into this format. And maybe sometimes I'd do some like mathematical operations on the data or whatever, you know, as you do. But the problem was each of these. Excel spreadsheet file formats was just like slightly different to the point where I couldn't just reuse my old script. I had to like kind like the core was the exact same, but the format of the Excel spreech was just different enough that you couldn't copy paste, you kind of just had to do it again. So it was like the corporate hellscape of like my job was every day I will do the same thing, just like slightly different. And I these days, like if it was with agents, right, I could do one script that does the thing. And then every time I had I got a new format in, I could just point Claude at the other script and I said, hey, I've already written a script that does most of this, and this is the way that I like to have it implemented. So just like take that and change it such that it it inputs this format instead and outputs in that format. And maybe I wouldn't have not had a job, but I would have had much less work to do because that company actually did have like I was so good at writing these scripts, the company ran out of work to give me on multiple occasions. So Let's say I run out of work on Friday to do. I probably would have ran out of work on Monday by like eleven AM if agents existed back then. So I don't know. Like but that kind of thing arguably is that real ⁓ the is that real programming, dude? Should that be a job? I don't know. That's what I'm

speaker-2: It is.

speaker-0: So you're a seven X engineer is what you're telling us. Well, I I guess f actually no five X engineer 'cause you're you're you're done with your work on a Monday instead of Friday.

speaker-1: Hey, on my resume, it says the estimate for the manual time that I saved is forty hours a week. So I saved a full engineer salary, guys. That's what my resume says.

speaker-0: By charging men to their salary to do so.

speaker-1: Yeah, exactly, exactly, exactly. There's that's the it comes back to the I think it like X K C D meme or something along those lines of like, ⁓ there's this thirty minute task that I'll spend five hours automating, trust me, it'll be great.

speaker-2: It's interesting though, like there's programmers are like they we we kind of optimize like compulsively. And like if we're given a task, even if it's kind of mundane, you know, we'll we'll eventually get to a point where we've automated it away. It's kind of just like this infinite, ⁓ how do we keep like making the process faster and faster and faster? And I guess it's like the real divider there is like ⁓ not everyone maybe at at a programming job like was truly doing that optimizing thing the whole time. Like they were given the task and they were okay with that. Whereas like I think the mark of a true programmer is they're never really satisfied with the system. You're always like figuring out a way to improve it.

speaker-1: That's really good point. But one of the things that I would say is, and this is where I think the problem is. Because when I took this job, I was a first-year uni student. You know, like I was fresh, I had not no work experience in the field. And so doing this like super repetitive programming task where I just like learn how to do stuff with code, that's a good experience for me, you know? And nowadays, and I think this is the the fear of everybody that everybody has, because we can like say, ⁓ well, that's not real programming, blah, blah, blah, blah, blah. But Junior engineers or like people that are just trying to get their first job, that's the kind of stuff that they get hired to do. So I think that's as well like really rough, is what people are seeing is, well, let's say that job disappears. If I'm just starting my uni now, I couldn't have taken the same path as I took when I started out. Because that job just doesn't exist. I have to try find something else. Which I think is a much a much scarier thing. Because yeah, we can go on and on about how like, ⁓ well, that's not real programming. Real programmers should be doing this. But you don't learn to do that optimization if ⁓ or like learn what that iterative process is if you never get a chance in the first place, you know? 'Cause that job did teach me a lot about how to write like by the end of it, I was writing these scripts to an absurdly optimized level. ⁓ and if I just didn't have the experience to be able to do that, I don't know, maybe I wouldn't maybe I wouldn't have had the same pathway.

speaker-0: I think that a lot of people are in trouble, ⁓ in the sense that I I don't know how much room there is gonna be in the future for people who just aren't ⁓ intelligent. ⁓ I like I don't know what they're gonna do. Because it appears to me that I mean I mean automation has always been the direction that we've been going into, even if AI didn't happen. As as programmers, right? What we do is we automate things and with we we have like a a This desire to make things very, very efficient. And that's being ⁓ that's being used to make things faster and and more efficient. That's always been a direction. Even if we never got LLMs, that would that's what would happen anyway. And so I think that the question that Sylvan was hinting at was, what are what are most people gonna do? Like if ⁓ you know, even if AI stays where it is right now and it doesn't get like I don't think that AI getting really, really good is is an inevitability. I think that We're already starting to hit ⁓ diminishing returns. We can you know we can improve the software around the technology. We can we can do agents and harnesses and and markdown files to tell so we can do a soul.md to give it a soul. That's how it works, of course. But we can make the software around them better, but at least for now, the the so the the technology seems to have kind of slowed down. And especially if it relies on hardware, which is what's been doing, that's how we got the improvements. recently is because of hardware improvements. If that also slows down, which is looking like hardware is also slowing down, which is why low level is becoming more important. We can get into that a second. ⁓ if that happens, AI doesn't have to keep getting better. But even if it doesn't, automation will continue. And so I guess the question, Matt, would be what do you think would happen in that context? Do you think it's going to be UBI? Do you think ⁓ People are gonna be just given ⁓ fake jobs to do at at scale. ⁓ what do you think is gonna happen?

speaker-1: Well, dude, okay, this is something that makes me really sad about the situation that AI as a whole is in. ⁓ and of course, like as obviously the disclaimer is this is my own personal opinion. Like it's not affiliated with the company that I work for in any. I can't believe I have to say that now, but ⁓ that is that is what that is. There's there's two paths that we can kinda take when it comes to AI and the way that people use it. There's like if you guys have seen

speaker-2: A pastry I'll be right back.

speaker-1: ⁓ good. If you've seen like the two paths meme where it's like one is like the super happy castle and the other is like the like lightning sort of destruction castle or whatever. ⁓ but the the dystopian path, I think we're on the at the moment, it seems we're on the dystopian path because what people wanna seem to be doing is getting up on stage and saying things like, ⁓ you know, in six to twelve months programmers are gonna be replaced or like it's gonna be a white collar bloodbath. And they've been saying that for the last four years. And just seeing that, it makes me really sad because I'm really excited about the technology. From from a pure like back in 2022, when this stuff was first coming out, there was the stable diffusion models and the ⁓ like Lama was getting released. Everyone was really excited about Lama because it was open, ⁓ like you could download it and self host and whatever. I was running that like locally back in 2022, before the hype was like super real and whatnot. I was always really excited about this technology. ⁓ and I think the technology is really useful and I actually very much enjoy using it. But I think there's a a reason why, for example, there was a poll recently that said AI is now ⁓ less popular than, for example, the war in Iran. Yeah. I think there's a reason for that, and it's that there's so much fear mongering and dishonesty going on about what the technology can do, what it will do. The sad thing to me, and we're seeing like AI content flood the internet, and really low quality AI content flood the internet. And I think that drives a lot of the opinion as well. My utopian point of view, like I think the best way people could use AI is always with like the human experience as the center point. And it it's kind of like what Sylvan was saying before about it being like an amplifier. I fully subscribe to that point of view. I think, for example, and ⁓ I wanna know what your opinion on this is, ⁓ Christian, because I know you're writing novels and stuff. But if I was a novelist, the way that I would see AI is not the case where you just prompt and say, Hey, write me a couple of pages on this. That's a dystopian view. The utopian view is more like, well, the human writes a page and it's all coming from the person. And then the person asks AI for refinement or as an interactive notebook or as an interactive thesaurus, where you you've written your page, you've got your soul out onto the page, you've said what you want it to say, and you use AI to like literally, like I was saying before, you diff max and you just like dopamine max. How do I get this to where I personally look at it and get the most satisfaction? And say like this is the this is the best thing that I could produce. It says exactly what I want it to say. I think that's true of art. I think that's true of coding. I think that's true of anything. Where you just like you have the human, it's me. I'm here, I'm doing this thing, and the AI is gonna help me get get it to be the best that I can possibly do. But I think because people are really lazy, they're gonna do the thing where it's like, ⁓ I wanna write an arc article on the internet. I just say, hey chat, write me an article about this. So it's like inverted. You know what I mean?

speaker-2: Yeah. I have I have two things to respond that I'll let you you speak, Christian. So the the last point I think I absolutely agree with, and I think ⁓ there's the AI is like alarmingly good at procedural problems, and in fact, I think we'll only see it get better at procedural problems. I I don't really try to ever speculate about the trajectory of things beyond like vague like I I don't even really know because it's just there's so many brilliant people working on so many things, it's like hard to even say what will happen. ⁓ but what I do think is that beyond procedural stuff, we have like stories. ⁓ and humans think in stories and we have for a very, very long time. Like you get like the hard sciences, right? You got like math, ⁓ physics, chemistry, ⁓ physiology, biology, behavior, anthropology is like kind of your spectrum of hardness of science. But then on the other side, you've got like the kind of softness of science where you've got like or or human knowledge, you've got like stories Pictures, images, religion. It's like that side of things you can't really procedurally generate no matter how hard you try. I don't think you can really well, I don't know. I d I don't want to say that, you know, because I might be proven wrong, but and then I guess the other thing is like I was full team dystopia for like last winter. I was like, ⁓ no. And now I'm full team Utopia because like I think ⁓ Well, you you said, right, like, what about the people that are and I I don't wanna you know, there's like lots of kind of derogatory terms for people, but like normie or like the the best way I can say it is like uneducated, because I think people all have potential. ⁓ there's there's obviously like innate variance in how quickly people can pick up the patterns to the world, but I believe everyone could be trained ⁓ to like flourish to do ⁓ Like this is just my behavioral ⁓ background talking, but like intelligence is kind of a construct, right? We can't actually measure that like empirically. ⁓ it's just kind of a vague, like, blob like thing. And I ⁓ in my opinion, all of psychology should be deprecated in favor of behavior. But like ⁓ what we can measure is your behavi how you behave your outputs. Like, and we can see this person is more creative ⁓ based on the variance in their outputs. But we can actually maximize that. Creativity is like an operant, meaning it's a function of its consequences. So like if you take someone and you like give them a crayon and they like draw something, at first it sucks, but you're like, good job, you know, draw another thing. And like you keep giving them bigger and bigger challenges. You can really train people to do like basically anything. ⁓ and it's kind of just like the environment that has led people to be so uneducated. It's because our education system is like eighty years out of date. with learning theory and our learning technology. ⁓ like we can learn really, really fast now. We can teach people at like a hundred times the rate we we used to be able to in the past. I've seen it firsthand, like it's been done to me. And ⁓ that technology I think just has to be spread to more people and then we'll have a utopia.

speaker-0: I think it's really cool. You know, when we started this podcast, I was a bit worried about I thought we would have like the same takes on everything and that I don't think that's good. I think we should have we we should diverge in in some ways. I dis I disagree on this point. I think that the environment I think that the environment is a component of intelligence. However, I think that it's it's mostly d inherited from from your parents. I think that generally speaking, smart people have smart kids. And I I I don't I don't think that anyone that just anyone could be trained to be a good mathematician or a good artist or a good programmer or a good anything. Regarding the the novel the novels and the and the writing, I think the question here is what is the product? So if you look at programming and writing science fiction or any kind of writing, if for programming the the product is the output of the software. I am autistically interested in code, but that's just because I'm an autist. Most people aren't. And even for me, like in software that I'm not interested in, as long as it works, like we're using Riverside to record this, right? As long as it works, I don't really care how the website is put together. I'm loosely interested about how how the back end works and how it's connected, but as long as it works, I'm fine, right? And so how the code what the code looks like, how it's logically structured, is kind of irrelevant and it's a hundred percent irrelevant for the vast majority of people. But for for writing, this is where it diverges a bit, for writing if you're writing purely factual or historical information. So if you're writing about something that happened in the past, if you're writing stuff about philosophy or anything that is supposed to be knowledge, you can't I wouldn't personally do it, but you could use AI. And in fact, there wouldn't be any reason to read those books because you just ask the LLM directly, right? In that context, how it's written doesn't really matter. If I want to learn about ⁓ carnivorous plants in Australia, I can just look up the information. The more condensed it is, the better. However, when it comes to fiction, that's where the product is how you feel. When I write a something, I want to make the reader feel a certain way. I enjoy the feeling of crafting that and imagining putting myself in the in the in the mind of the reader and imagining how they would experience this based on all the context of the story. And my concern would be that by using AI, I would be introducing patterns that I may not be immediately aware of. I might be introducing elements that might make my work homogeneous, my my work and the work of others homogeneous to a point where I would lose an edge that I might have just by writing myself. And so unlike programming, where the code doesn't matter for most people, with fiction, the code absolutely matters.

speaker-1: That is a very interesting point because ⁓ let's say that you're working on like the scope of one page and the AI is making suggestions to you and you go, ⁓ okay, this one looks all right. That's for one page, but then you're right, like as if you do that for like every page, you might tend towards it's like the frog in boiling water thing, where the frog doesn't realize that it's getting boiled ⁓ if you like slowly crank the temperature right. But then by the end of it, you're right, like it's it's different than you thought. That's a really interesting point.

speaker-0: And also how would you even like there's all this all this all these studies around how people are struggl like if you want to learn with AI, it's a bad idea because you're not actually learning anything, you're just being exposed to information and people are having people have written they've done these tests where you you have people write with AI and write without AI. People with who wrote who wrote with AI can't remember the contents of the essay. And writing a book is already kind of it's already quite difficult in terms of keeping everything in in your in your in context, right? Because there's so much context. And I'm writing in the same universe. So all of my stories, I already have a bunch of them already planned out. They take place in the exact same universe just on slightly different timelines. And so I'd have to rely on this machine to accurately remember the context of all this stuff. And I would somehow have like it would I mean, not only would it have to remember it, it would be influenc influencing it or maybe even driving it entirely. If it's coming up with suggestions, what it's doing is it's writing the lore of the world that I'm trying to build. And so I'll end up at a point where do I even own this world anymore if I have to ask the machine, hey, sorry, what what did you say happened it a thousand years ago in the story? Like it it wouldn't feel like it's mine. It would feel like an entirely economic activity at that point. It would feel like I don't care about ⁓ how beautiful the the the writing is, I don't care about how ⁓ how ⁓ cool the story is, is it making me money? And if that was the approach that I would have, then I would probably just use AI ⁓ to to write it. But given that my approach is I want to feel really big brained and cool. And I want people to look at me and go, hey, that that guy Christian, he's pretty smart. And if I make money from it, that's pretty cool as well. That that's my that's my my take on it.

speaker-1: It's interesting that that's the way that you're describing science fiction writing is almost analogous to how I view personal coding projects, actually. Because ⁓ it's the same thing like you're talking about writing the law. When I'm building my own personal project and getting into the head of it and wanting to feel big brained and learn something, right? And the feeling of it being my own thing, that's exactly the same way that I look at AI contributions in that sense as well.

speaker-0: They are very analogous. ⁓ I actually spoke about this in a video at some point randomly, I don't remember which video, but programming and writing fiction are very analogous because you start off with a loose idea, something very, very loose, something on a planet or something, something some conflict, right? And then you have all these disparate pieces all over the place. And you you you have to figure out the logic. How does this work? And you lay out the logic. You have a loose structure of how it works and then you start filling in the details. It's so similar to programming, it's actually not even funny.

speaker-2: I'll I I'll kinda say that I think everything is actually just programming. It's all programming. I think there's a ⁓ so there's certain like activities we do, like art and like interpersonal communication and writing fiction, which have this kind of fuzzy veil atop them that they're like, ⁓ this isn't a procedural thing. But the more I think about it, like deep down it is kind of a procedural thing because like

speaker-0: Go for

speaker-2: I was I was thinking about this yesterday, like what is creativity and art but just optimization with a lipstick on it? It's just like optimization but like we call it you know, we we have like a certain vibe atop it and like that may seem crazy, you know, but like you could take a piece of art and you could do this now and like for instance like a YouTube video and you could ⁓ tweak you could parameterize it into discrete little things. How do we tune each of the colors? How do we tune each of the little sounds? And we could do that for like stratified groups of people as well. This type of person is gonna respond well to like a squeezing sponge sound. This group of people is gonna like this kind of thing, you know? And then you do that across as many people as possible, throw like a CNN on that, and it can be optimized. Like it looked like it was art on the surface, but it's just like a algorithmic thing. And I would say the exact same thing could be done with fiction. I know that's that's a hot take, but like ⁓ I think we're just not we're not close to that yet. We would have to parameterize and break down. We'd have to dive into like story science really deeply ⁓ to optimize like what is the perfect story. But it it kind of gets back to that question, right, of it's like why why do you do it? ⁓ and it's like playing the drums versus using a drum machine. Like at a certain point Like if everything could be handled by a super intelligent AI that was, you know, not just not just like w capable of any procedural thing, but like we cracked the code on generating infinite unique stories, infinite unique ideas as well. Like this thing could just churn out more than any human output ever could. Like at a certain point you have to spend your time or you'd just get depressed and stop living, you know?

speaker-0: Can say something on that? Because I want to touch on what you just said 10 seconds ago about generating unique ⁓ infinite ideas. Do we even want that? Hear me out, right? Hear me out. Right now, what you could do is you could go to Chadji and you can say, generate me endless look stories in the universe of Lord of the Rings. You could do this, but I would argue that no one cares.

speaker-2: I I would I would argue back actually that no, you can't generate infinite unique stories of the Lord of the Rings. I think any output of a LLM currently is finite. It's contained within the human infosphere. Because all all of our language or verbal behavior is just like stuff people have said in the past being rehashed. You could get a bunch of combinations of that. Yeah, exactly. You'd have to observe reality and like

speaker-0: Yeah, to do infinite.

speaker-2: Yeah, invent words essentially.

speaker-0: you have to have like Welsh words with like seven L's and like just make up ⁓ like a new Welsh language. But my my point was entirely just to finish it, my point was I don't think anyone cares about made up artificial s stories because you could do that right now. You can you can you can do that literally right now. And yet I from what I under from what I know, people aren't using it. I think that there's something about the human touch. I think that people These stories weigh more in our minds because they're made by humans.

speaker-1: I think there was a there was a really like a very interesting thought that we could discuss maybe is is I remember in twenty twenty two when the stable diffusion models were just getting good and you know it wasn't just like Will Smith eating spaghetti, it's like, ⁓ we can actually make something. There was a big news headline. Yeah, exactly. Exactly. Will Smith is making spaghetti code as we speak with Claude. ⁓ No, but there was there was this big headline that

speaker-0: He's now making Spaghetti.

speaker-1: Some dude entered an AI generated photo into like a Texas art competition or something like that. And it won. Like it won the competition. Yeah, yeah, yeah, yeah, yeah, yeah. ⁓ and that was big news back then. And everyone was saying, like, ⁓ artists are finished, it's over, you know, like pack it up, go home. But I think people have come to realize like, cause we don't talk about artists being done for anymore. Cause who genuinely, like when you look at a piece of art, not only is it the human touch that you're looking at. But also I think people appreciate the time and effort that it takes to create that piece. So for example, my brother's girlfriend is a very talented visual artist. And we have a a Corgi, like a dog, and she did a painting of this dog. And it was really, really good. Like the detail was immaculate and it was it was great. Now what you could do is take a photo of our dog, put it through ChatGPT, and say, Hey, can you do this in an oil painting style? But who cares? Like she made this thing, it took her a really long time. The product is immaculate and she made it. And there's all of that other emotion that goes into where whenever we look at this piece of art that we have on our wall, it's like there's so much more that goes into that and you appreciate not only the human touch and you appreciate all of the time and effort that somebody took to create that piece as well. So

speaker-0: Remember like Expedition thirty three got into hot water a little while ago because they used AI placeholders for some posters in the video game. And then it turned out that ⁓ sorry, it was just placeholder, and then they they replaced it with the actual art. They forgot to replace it. I think the reason why f why people find AI use in in a corporate setting to be icky, but like AI use in a corporate setting to do with art, why they find it icky is because they know that The artist would be a reverse centaur. They would be forced to to kind of like use AI as much as possible to optimize money, money, optimize profits, and that's it. I don't think that people would have an issue with it if the artist was a centaur. In other words, if the artist was given a tool, an AI tool, it's like, listen, bro, or like your your brother's girlfriend, right? They would still call her bro. Listen, bro, they're talking to her. You we trust you that you will be judicious in your use with this AI tool. Here's a tool. It can help you in certain conditions. It can we we trust you as a a talented artist to use this tool in the way that you think would be best. If there was no pressure, no economic pressure, no use this or you're fired, if it was just the tool, I don't think anyone would have that much of an issue with it ⁓ issue with it. I think that's where a lot of the ickiness from ⁓ that has to do with AI art comes from of like of like this is inauthentic, it's entirely profit driven, there's no soul, it's just you it's it's a purely money seeking activity, and that's not what I want when I look at art.

speaker-2: Yeah, I I big time agree. There's ⁓ it's not my idea, but Neil Postman, ⁓ who's like a cultural critic, he has said like we shape our tools and then our tools shape us. And this thing of like the concept of a harness is nothing new. In fact, I would say every tool is actually a harness. You know, like ⁓ when we invented the like stand-up saddle ⁓ for horseback riding, like that reshaped ⁓ com like mounted combat because people can fire bows and arrows and ride their horses at the same time. And you could say that like I'm not a historian, this is probably completely inaccurate, but like that

speaker-0: It sounds cool, so I I'll believe it.

speaker-2: The Mongol Empire, right, to ⁓ be much more effective and then they they spread everywhere and that changed the course of history. ⁓ and it's the same thing, like our our tools are continually shaping us. You know, I have my magical titanium brick, I can summon like any of my accessories from anywhere. Like I can I can call whatever I want. I can look up any piece of information, and it's the same thing with like art and and thinking and like are the b the line, like the the tool line is getting fuzzier and fuzzier. Like they're shaping us more and more and more. We're getting closer and closer to kind of the c the core of what makes us human, the like story generating machine or whatever that's in here. And eventually what's going to happen is like BCIs and like people want to plug the machine right into the brain, right? We're we're already like if you're if you're using I don't want to give them free publicity actually. If you're using a TTS, pay us If you're

speaker-0: I'll censor the name of the of the brand so they can have to pay us. Yeah, yeah, yeah.

speaker-2: I feel like

speaker-1: Insert TTS brand here.

speaker-2: I d ⁓ I do know someone at they might they might do it. There's like five

speaker-0: I'll have censor that as well. Thanks for giving me more work to do. I appreciate it so

speaker-2: It was right next to the other thing. But yeah, you know, like when you're talking, you've got your T T S and you've got your like five agent harnesses. It's like already your brain is plugged into the machine. And I'm already kind of like augmented human in terms of like the art I make, right, is like a harness, it's a human harness. Like the types plus kind of thing. ⁓ I can't God damn it. There's so many brands. The marketing is like sneaking into our language.

speaker-1: That shows how effective their marketing is. They've cooked you. You're like a sleeper agent now.

speaker-2: Yeah. I know. It sucks. I I we only we have a rule on this podcast we can only recommend open source tools that that are not for profit.

speaker-1: Pretty base.

speaker-0: If they if they pay us still, they still have to pay us. Can I just point something out? I've noticed something in the in the I've noticed something in the source of this reality. No one refers to their phone anymore. Everyone says the magical box in my pocket. And when they do that, they always grab it and they show it. Have you noticed this? Like I've seen this so many times. People will say, ⁓ I just have this magical thing in my pocket that does the thing. No one says their phone. Isn't that

speaker-2: Yeah.

speaker-0: Like, I don't know if you've noticed this I've seen this on like ⁓ many, many times people do this.

speaker-2: We're merging with the machine. What

speaker-0: I mean, we are literally cyborgs already. So yeah, we're not we're not ⁓ we're not that far off. I think you have a good point though. I think because like, I mean, for example, I don't have an issue with an artist using Photoshop, right? Even though I know that Photoshop is influencing the art, it has to. I've used Photoshop, I know it does. I accept the use of Photoshop because I know that the artist still has control over what's actually happening and it's similar enough to the actual physical process. Although I know that they are being influenced by the software itself. So it is a it is like a back and forth sort of loop that's happening with the software. But I still accept it, as long as I know that they're they're not pressured into doing certain things just for the sake of money maxing.

speaker-1: And see that's yeah. That's I think where we can take this back to coding so that we keep the viewer retention up. You know what I'm saying? But ⁓ Where we where we take this back to coding is Sylvan, I actually think a lot of the things that you said probably like a minute ago are exactly the same way that people are really ⁓ scared, but also there are lots like on Reddit, if you go to Reddit, God forbid you go to Reddit, but there's like a lot of ⁓ people who are very anti AI in

speaker-2: Yeah.

speaker-1: programming and in the workplace and there's a lot of negativity about it. I think because of the way that it's framed, I think if people were framing it honestly, exactly how you were saying, you would prefer AI tools be framed for artists, which is like, we trust you as a professional, as someone with good taste, as someone with good judgment who's gone to school for this thing and you know about this thing, to use this judiciously and to use it to where you believe it is correct. Whereas what we have is a dystopian hellscape where everyone's like, Token max or you're fired. But that's winding back. Like, I don't know if guys have seen all the ⁓ like for example, the Uber CTO has said he's because they blew through their AI budget in like a month and a half, they're gonna start winding that back. And I heard rumors, I don't know if this is true, I heard rumors that Meta has stopped their token leaderboard. Apparently the morale at Meta has never been lower, by the way, which I thought was also pretty funny. But I think it just comes back to like people are having something forced on them and saying. No, you do it this way. It's going to be better. Trust me, bro. I d I definitely went to school for this thing. Yeah, I went to school for business. You can trust me. This will make your job better. It's not like you went to school for computer science and you learnt all of this stuff. No, I'm telling you as a business major, trust me. This will make us money. Well

speaker-0: This is the whole Centaur and Reverse Centaur thing. So a Centaur is a human who uses AI to to in their work, where or the machine to do their work. It's the head of a human on a machine body. A reverse centaur is the opposite. It's a human body being puppeted by a by a machine, which is a nightmare scenario to be in. ⁓ if you're a programmer and you're allowed to use it's like, look, you can use AI. We trust that you know what you're doing. We trust that you will use it sparingly and that you will use it like a scalpel, like a scalpel whenever you need to. Then those are centaurs, those people who love using AI at work. A lot of programmers are reverse centaurs. Use AI now faster, 10x, 100x, ⁓ where is this a hundred a thousand X, bro? A hundred X is so yesterday. That's where you're being puppeted by the air by the machine and you're struggling to keep up. That's the difference between people who who enjoy and who don't enjoy using AI at work. It's the ones who are allowed to use it judiciously as people who know what they're doing in their field.

speaker-2: A lot of opinions on the ⁓ the fear-based marketing and fear-based ⁓ like methods of coercion in general. I think there should be legislature against that. The reason they use it is because fear is the most contagious ⁓ and strongest human emotion there is, and they weaponize it in advertising and marketing, and almost all of our politicians kind of are operating on fear-based principles. ⁓ and in fact YouTube is filled with like psyops and propaganda that are just literal fear bait. And it's just a very, very inhumane way to treat humans. It's like treating humans like they're cattle, like they're ⁓ they're being factory farmed for their ⁓ their their money and their time.

speaker-1: Do you agree with a lot of what you've said there? And I think at the end of the day, that is the the amount of damage that that has done to the population's mental state is absurd. Like if you're on any of these, I don't know, again, God forbid you go to Reddit, but like R slash CS careers or something like that, just the amount of Doomerism, and even on YouTube, I get recommended people who are doing like Doomer AI videos where they're like, dude, I can't get a job, I can't get a job. AI's gonna automate everything, what's the point? What's the point? The amount of damage that that has collectively done to an entire generation of aspirational people, dude, it's it's horrifying. It's absolutely horrifying. And I don't blame like I wanna stay positive and I try to stay positive and just like cut all of that out, but I don't blame anybody who fools for it because it's everywhere and they just push it so hard. It's like if you are a young person Who doesn't have a lot of life experience and you're very uncertain about your future. And all you hear when you you you listen to like tech podcasts is like, programming is over. It's done, guys. Pack it up. Go home. I don't recommend you learn computer science. I it's gonna be done in five years. In six to twelve months, we're gonna automate all programming jobs. And then like a month and a half ago, they backflip and they're like, ⁓ hold up, guys, it hasn't been as bad as we thought. Maybe we should slow down a little bit. The amount of damage that they've done to just like People's ⁓ morale in the last four years is absurd. I don't like it.

speaker-2: I would say as well, if I can cut in, is ⁓ it's not just that, it's like subliminal marketing in general should be heavily regulated. It's not just the AI industry, it's everyone in a marketing department who kind of does something that's like doesn't seem ethical, but they go ahead with it anyway for the bottom line. Like I mean TikTok, like short form content as an industry, people who use marketing to get you to buy like garbage plastic. devices and like just hyper consumerism fueled by ⁓ there's I I've looked into this a little bit. There's like a bunch of tactics that people use, like nostalgia, ⁓ for some reason, I don't know the scientific reason, really increases people's willingness to part with their money. ⁓

speaker-1: Why do you think I bought this Minecraft t shirt? Nostalgia, dude, it's the they got me. They got me.

speaker-0: It got me.

speaker-2: Yeah. Yeah. Marketing as a field has come a long, long, long way. It's it's scary what they can trick you into doing and buying. I mean, I'm God, there's brands everywhere.

speaker-1: You don't recommend that's a uni hat, right? You don't recommend it? Bad uni?

speaker-2: ⁓ well, I mean, it's like one of those environmental things, you know, where it's like I blame the environment and not individual people in it. Like they're greedy shit bags. Like everyone in the entire ts town hates these guys. They charge way too much. It's a bloated bureaucracy that offers a horrible education. Like they're an absolute joke. ⁓ and they basically they're a they're a diploma mill, essentially. Like they they exploit like ⁓ Yeah, brutally roasting University of Vermont. Like their their business model is the ⁓ children of hedge fund like executives and dentists and doctors that are like, you know, like it's a it's an okay education if you really go and seek it out. You know, you can go and talk to them. But the problem is like for the average person, that's not the case. They're just like sitting in a a a lecture essentially playing like a YouTube video to you. Like it's not worth the out of state tuition is, I don't know, forty thousand a year or something absurd. It's like very clearly just a cash c company. It's not about education.

speaker-1: That's how much my entire degree cost. In Australia, it's pretty good. That's how much I paid for the whole like five and a half years of study. ⁓

speaker-0: Mine was a third of that, by the way. A third of yours, Matt.

speaker-1: Dude, I wish I could just Europe Max right now. That'd be crazy.

speaker-2: Europe is so based. You guys are awesome.

speaker-0: Do you what? This is such a beautiful segue. because hope, right? I think hope is very important. I think that peop you know, people trying to to ⁓ to ⁓ do max to make money, I I'm not a huge fan of that. But I think this is an opportunity to ⁓ to talk about something really cool. So, Matt, I think you finished university like late last year, right? Okay. ⁓ we can go into the teaching assistance stuff in a second as well. I think that's I think that's very interesting.

speaker-1: Yeah.

speaker-0: But ⁓ you interviewed at Microsoft and you actually you actually work there now, which I know we joke around with Microsoft. It's actually a huge a huge accomplishment. Actually, ⁓ unironically, like ⁓ well done. I wanna learn more about the process. I wanna learn I know that you you weren't too big into too big into lead co, which I I I agree with you on that. I I wanna learn about the process. So how how did you apply, how how did the process go? And like tell us whatever you whatever you know that you can talk about.

speaker-1: Thank you. Well I'll tell you this. I'll tell you this. Regarding the process itself, in the video I didn't talk about how many applications I did because I didn't want to come off as boastly. I made one application and I got in off of one application. So that's different yeah.

speaker-0: Is this the same one? 'Cause you made that video like five months ago, I think, on your second channel. So it's the same pro okay, so same process.

speaker-1: One application. In fact, I applied for so when I graduated uni, I already had a job and I was pretty happy there. But you know, know your worth, Queen. I applied to a couple pl I applied to four four jobs. ⁓ two of them were at ⁓ Boeing, because I have a friend who works there and he's like, ⁓ you can come work with me at Boeing. ⁓ one of them was at Amazon and one of them was at Microsoft. Boeing, the both of the Boeing ones rejected me, but I got a call back from Microsoft and Amazon. I declined Amazon and I just did Microsoft. So I made one application to Microsoft and I got in. Which seems to be not the experience of many people. That's why I didn't really want to bring it up in my original video because I hear people are doing like lots and lots of applications. So it's not something I wanna like boast about. But that was it. I did a LinkedIn call to apply, one application. The whole interview process I did my I did a r recruit a phone call. Actually I missed the first recru recruit a phone call. Because I was on holiday and I got a call from a Malaysian number and I was just like, I thought it was a scam call, so I just didn't pick it up. And then they kept they called me a week later, and I'm like, I guess I'll pick this up. And it turns out it's a recruiter. And like, hey, I'm here from Microsoft about your application. Meanwhile, like the week before, I like dodged this call because I thought it was a scam. And I'm like, ⁓ my God. So I was like, ⁓ my God, I'm so sorry about that. so that was the recruiter phone screen. They just we just had like a a 30 to 40-minute call where we chatted about my past experience, about my resume, just to get a feel for if I was a good fit for the team. Because the way that it worked for me, at least, was I got talked to by a recruiter, and then all of my interviews were conducted by people on the project that I was on. I don't know if that's a universal experience or not, but it was for me. Like everybody that I had, like a technical or system designed interview, was actually on the project that I was on. ⁓ And I mentioned this in my video as well. I think it's a really good idea to Google the project that you're applying for before you do your interviews. ⁓ especially because the one that I'm on is fully open source. Like you can just go to GitHub and look at it. So I just like before my interviews, I was just reading about this project on GitHub. Number one, to make sure it was a good fit for me. And number two, so that I could actually demonstrate that I applied for this role because I'm really interested in this project. So I I wonder I've thought about why it was that I had such good you know, callback with my resume and I think it's because I was like looking for jobs. There were other postings that Microsoft had at that time that I could have applied for. But I chose this one specifically because I looked at it and like I said in my video, I just think like this one is me and I applied for that one.

speaker-0: I think you spoke about this in the video, how people apply to like random jobs and they they never hear back. But if you apply to the things you're actually interested in, your C your C V is actually for that thing, then you're more likely to actually get ⁓ responses.

speaker-1: For sure. Because they told me in my interview, like your resume is just like perfect for this role. So it's it's part luck as well, you know. The LinkedIn thing said like over a hundred applicants as well. So they would have to do some kind of like AI filtering or like ATS stuff on that. So there's for sure a luck component to it as well. Like you've just got to get lucky with the filters too.

speaker-2: This advice is really good and I've heard it from professional HR people and ⁓ and hiring teams as well. Don't cast a wide net. Go really specific. Like research the role and like a a mentor gave the example to me. Like, if you're kind of into software engineering but not super into it and you're into firefighting, like look for fire stations that need software automation and like really sell yourself that you're into firefighting. ⁓ it's just it's gonna be Like they they get such a sea of these applications that are all just super generic. Like people just spam buzzwords like people say cloud architect now and they've never like they don't know any of the Kubernetes flags or anything like that. Like they just put all these buzzwords and it's meaningless. I mean, you you can also go like next level. You can be a little bit devious and evil. Like the recruiter I know, ⁓ who is she's kind of a mentor to me, connected to a mentor figure of mine. Like she color keys. ⁓ her resume to the company's branding. and I mean you can get even you can get even worse. Like I would use OSINT tools to like find information about people so I could use like ingratiation, which is a social engineering technique, to like I mean you don't you don't have to go that evil, but

speaker-1: Dude, if wherever company you work at in the future sees this recording, you might be in a bit of trouble.

speaker-2: ⁓ dude, no, that's the thing. I'm so charming, like I can tell people I'm a cult leader and it doesn't even matter.

speaker-1: It's on his LinkedIn, remember?

speaker-0: You know, ⁓ I've actually ⁓ I've actually done the last one. So ⁓ when I used to apply to ⁓ I haven't applied to freelance gigs in a while 'cause I've just been working with the same clients for for years. But when I was applying for freelance positions, I would actually do a bit of OSENT research to figure out the name of the person who's gonna see my application. And I've literally had instances where they were like, How did you 'cause I said 'cause it said it didn't say hello or to whom it may be to whom it may concern. It said, hello Matt. Right? Like it it would directly and p they would say, Ha that was really cool. Like we we know that like how did you find the the person that was gonna read this, right? So that that that bit actually you should do a bunch of I mean, you should do a bunch of research before you apply anywhere, really. I don't I don't I don't know why you wouldn't. It it shows that you're interested, it shows that you care. Even even before you start working there, it shows that you you're interested and you care and you're resource for. Like if I if I give you a problem I don't have to hold your hand. You're like an an autonomous fleshy robot that can go around. You're a hum billion ex engineer, human agent. Let's go. It's already it's already copyrighted.

speaker-1: Yeah.

speaker-2: Copyright, copyright, copyright that.

speaker-1: Guys, our new startup is gonna be building here ⁓ shit, sorry, my lights are on timer. Yeah, no that was an evil hit eleven PM where I live. So my lights are let me just fix that real quick. I'm so sorry guys.

speaker-0: For a second back. ⁓ just turned red for a second, they turned evil.

speaker-2: No, that that was cool. Right, right. That was a vibe.

speaker-0: In the nineteen forties now.

speaker-1: Okay, no, we'll go back, we'll go back, we'll go back. We're evil. We're evil.

speaker-2: ⁓ yeah. ⁓

speaker-1: Yeah, our new startup is gonna be building human agents. I think that's really gonna be the future. So we'll go to YC, we'll go to San Francisco, we'll get it done. Human agents.

speaker-0: Maybe maybe not. Maybe not.

speaker-1: How do you manufacture a human agent?

speaker-0: I think I know. I think I know exactly how you manufacture a human agent.

speaker-1: I don't know. Yeah.

speaker-0: Cabbage something about cabbages. Yeah yeah yeah. ⁓

speaker-1: ⁓ there's this bird that drops them off outside your headquarters.

speaker-0: It's

speaker-1: Actually a draw that's

speaker-0: ⁓ actually speaking this is a perfect perfect segue. Speaking of having kids, right? Imagine this isn't gonna be weird, I promise. Imagine that I'm I'm ten. I'm it's it hi, it's me, it's me, little Christian. Hello, I'm ten. And I notice you, Matt. ⁓ we're we're hanging out, Matt, and I notice that you have this in your hand, right? And I go ⁓ my god, I forgot. I have the I have the the the board in my hand, the ⁓ the SDM board in my hand.

speaker-1: You can't see your video by the way. Yep, yep, yep, yep.

speaker-0: So I'm holding in my the audience should be able to see this. I'm holding in my hand, I'm holding the SDM board. And you seem really preoccupied and really interested in it. And I'm like, hi, I'm 10 year old Christian. Hello. And I'm like, why is this? And I'm asking you, like, why is this board so cool? Why do you like it? Sell me on on doing ⁓ embedded programming.

speaker-1: Well, first of all, my lawyer has advised me not to talk to random ten year olds on the street, but Yes. No, if if we're being serious for

speaker-0: W we're cousins, so it's okay.

speaker-1: Yeah. Yeah. ⁓ look, if we're being serious for a minute, I think a lot of people really want to understand how things work. And I think that a really if you think about what abstraction is, especially in computer science, abstraction is necessary to let us be able to do more. So for example, nobody is coding in opcodes and binary anymore for good reason because like it's just an inefficient way to get things done. So in order to make that more efficient, we built assembly language. That was considered high level at some point. And then assembly language became too cumbersome for people to do, and so we built more. And you can think about it like this: like in one line of Python, how much back-end stuff goes into one line of Python? Whereas to do a task, for example, what I was doing at that company was writing like Excel scripts. It was so fast for me to just write out the Python to do that. If I had to write the equivalent thing in C, it would take me literally forever. So I think abstraction has very good use in computer science. ⁓ There's the phrase like standing on shoulders of giants, for instance. And that's a phrase for a good reason, I think. But at the end of the day, I think a lot of people just want to know how stuff works. And that was definitely what it was like for me. And so a lot of people, I think, are not satisfied by typing print into the Python console. It's like, okay, well, how do let's go let's go a little bit deeper than that. How does the print function actually get called? What takes that text and interprets it? And then decides to print something out to the screen. How are the characters going to be represented on the screen? Let's go a little bit lower than that. It's like, how does the graphics and windowing system work to even show characters on the display in the first place? So there's like so many layers that you can just, even from one simple thing, like print hello world, you can just go deep down to a final fundamental level and understand how that works. And I don't think anything is a better teaching instrument for that than embedded systems. Because for example, if you think about a desktop computer with Linux on it, how are you going to write a program that directly interfaces with the registers of your CPU? You can do it, but it's like the CPU is so complicated and it's got multiple cores and multiple threads. It's crazy work to try and write something for a desktop CPU. But in the embedded world, the CPUs are still simple enough that you can understand actually how they work at a fundamental level. And so you start there. If that's your entry point, you get familiar with it, you're like, okay, so this is what memory mapped.io is, this is how registers work. ⁓ this is what hardware acceleration looks like when I have dedicated hardware to do things like floating point operations, for instance. And you can scale that up from there. So actually we're twinning. I've got I've got a STM board on my ⁓ on my desk as well. But once you it's just a an easy entry point to this entire world.

speaker-0: Nice.

speaker-1: And so what you can do after that is you can look at something like the Intel intrinsic intrinsics guide, which is basically just like, okay, what ⁓ baked in hardware features can I use on my Intel processor? And you can actually call them from C. And for example, one of them is like hardware CRC32, which is an algorithm that you can use to compare two files or two pieces of data and check them for errors. And so what you can do with that is you can offload that to hardware. And because you've been exposed to, okay, well, I know how to offload things to hardware because I've done that in the embedded world, you can kind of like grok that intuitively and say, ⁓ well, my Intel CPU or my AMD CPU can do this also. And this is how I do that. So I think like at the end of the day, if you want to learn how the big things work, you have to start with the small things. And that's what these are are really good for, is like just a way to get your foot in the door, get your feet wet. And you don't even have to scale up. Like you can make a career out of embedded systems. And importantly for people that like this kind of thing, I think embedded is the largest area of low-level programming that is still widely employable for. I get a lot of questions about this, ⁓ where it's like, okay, I love system programming, I like doing the Linux stuff, I like doing like network programming and stuff with sockets, and I like doing memory management, etc. But there's no jobs in it. There's a lot of jobs in embedded, you just gotta look. So I think embedded is A big field where there's still a lot of low level programming jobs because it has to be low level. ⁓ and I think a lot of people look for that.

speaker-0: I had this really magical moment. So I was learning about how CPUs work. And this is this is like a very abstracted explanation of how like like how a bus works. Like let's say you want to transfer information from one register to another in a CPU, right? Like if you imagine like eight wires, like you like an eight bit thing. So you like eight wires, that's the bus. And those wires go f into and out of the registers. If you wanna move and correct me if I'm wrong about this explanation, but if you wanna move data from one register to another. You open the output on register one, that floods the bus with the date with the information. And then you open the input on on register three, for example, and now now that's also flooded. So now the bus and the registers are actually all contain this information. And now you just close the output on register one and you close the input on register three. And now you've you've moved information from from register one to register three at the speed of electricity, essentially. Would you say that that's like a like an accurate way of describing?

speaker-1: It's ⁓ it's reasonable. ⁓ it depends as well on the transist like the exact transistor setup. but at the end of the day, bus like buses are how data is moved around, even but like interchip buses exist as well. Have you looked at ⁓ so for example, all of these little pins on the SCM thirty two, you can connect to various things. Have you messed around with that at all, like chip to chip communication with buses like I2Z or SBI?

speaker-0: I haven't used it yet, no.

speaker-1: But at at the end of the day, data transfer, for example, on a bus, let's take like one of the I20 SBI buses as an example. You're famili are you familiar with like clocking in computers, like how clocking works? Yeah, yeah, yeah, yeah. So ⁓ everything is based around reading the right data at the right time. So when you're trying to transmit like bits.

speaker-0: Yeah, yeah.

speaker-1: on a serial stream, for instance, you've got your serial data stream and also a clock stream, which basically the the edge of the clock is where you read that that bit of data and ingest it. And that's something that's very hidden from you in high level things, especially. Like if you just po like you program in Python or you program in JavaScript or something, you never actually understand and you would hear these terms get thrown around like buses and clock speed. Like for example in a in an Intel processor you hear about clock speed. Well nobody really knows what that actually means. unless you actually go looking for it. ⁓ but embedded, for example, clocking is a fundamental, very important part of building your embedded device and building it reliably. And so you have to learn about that. And I think like if we're going back to how I would sell embedded, it forces you to learn about things that you otherwise would never be exposed to, like how ⁓ CPU register ⁓ transfers work. And for example, there's DMA as well, which s you the CPU doesn't have to do instructions to transfer memory. And that improves performance a lot. But you just learn about these things that you never would have to understand because computers are so powerful these days. So you you never have to worry about that kind of thing. And I think ⁓ it's a bit of a shame because people are losing out on optimization. There's the argument in the games industry where it's like because we've got all these AI upscalers now, they no longer have to optimize and stuff like that. But you're forced to optimize on one of these. And I think for people, if we go all the way back to like the start of the podcast, Where Silwin was talking about how developers should want to just like continuously optimize and optimize and optimize. Well, this is a platform where you can do that. So if that's you, like if you like to just continuously optimize and optimize and optimize, these things need 'cause look at the size of this this little CPU. It's it's tiny, you know?

speaker-0: Did you have cause like when I when I learned that thing about how data cause I was like, how does this even look how does how is stuff moved around the CPU? How's data moved around a CPU? Right. And I looked it up and just blew my mind. It was the coolest thing I it was like, also it it's it's not just magic. Like there's like there's like physical things that happen to move this data around. It was such a beautiful feeling to to learn that. Did you have ⁓ like a similar sort of instance where it's like, ⁓ this this is really, really cool. Like this this tickles me in just the right way.

speaker-1: Mm, that's a very interesting question. I would say probably when I learnt about memory mapped.io, I would say that was the big thing for me. Because when you do programming on embedded, right, and you're changing registers, you're not just changing registers in ⁓ your CPU, because if you if you think about this chip, it's actually split up into multiple different parts. And you'll hear terms like on chip hardware described a lot. So for example, on an STM32 chip. Inside this package right here, there's actually multiple hardware devices. So for example, you could consider the timers that operate on that to be a separate hardware device. You could consider the I2C or SPY handlers for those protocols, those are a separate hardware device. You've got analog to digital converters and digital to analog converters, which are also hardware devices. And the question is, how does your CPU control those hardware devices? So if I want to set a timer on my STM thirty two to go off, let's say like twice every second. So that's 500 MS. How do I actually set that up? Because at the end of the day, yeah, we're we're setting values in registers, but how is it that the CPU can just do a move instruction on one value into this one register and then magically the timer starts operating at 500 milliseconds? It it didn't really make sense to me. And once I learned about what memory map.io is, which is literally like There are ranges, if you think about the main memory of a microcontroller, where it's like 0x00 all the way to something. And if you look at the data sheet for your STM32, because I imagine you'll have to do that quite a bit when you get into it, but there's a section there about memory mappings for hardware peripheral registers to the CPU's main memory. And so as it turns out, when you set the value with in your C code, it it looks something like timer zero. And then you do the little struct address thing, and then you set whatever the value is. Memory map.io is like the timer, the hardware timer's memory is getting directly set by that. And that kind of realization that it's all it's just shared. It's shared memory. That was pretty crazy. Cause now I understand like when I type in in this code like timer zero, comma, ⁓ whatever the register is, five equals like five hundred milliseconds. But that's not ex that's not how you do it, but just as a easy example. Mm-hmm. How does that timer actually then start ticking at 500 milliseconds? And it it's learning about that and it's right there in the data sheet. It literally tells you this memory address in the CPU's main memory corresponds to this register on the hardware peripheral of the timer. That was pretty nuts. And you'll get into that too once you start playing with your STM boards.

speaker-2: I'd like to just say ⁓ as a as a as a quick note on really detailed technical things, because I'd imagine just based on the variability you get in a large audience, like some people love that, some people completely glazed over by that. Yeah. And it that'll depend on the topic as well. But there's like this kind of misconception I've had for a long time that like these technically difficult topics are like really painful to learn. And I don't think that has to be the case at all. I think That misconception can often come from the traditional school environment. Like for me, linear algebra was really hard to learn because it was like t a boring lecture like two times a week, you know, like that doesn't work at all. And then like I sat down with ⁓ a ⁓ like algebraic geometrist, PhD candidate, and he like explained category theory to me in like ten minutes. And it like I skipped like twenty classes, you know, and it's It's actually some of these things like they seem super daunting and complicated, but if you have a good professional mentor who knows how to break the ideas down, you can you can really get into some fun, deep stuff quickly. And once you're at that level of abstraction, it it it kind of becomes its own reward. It's like you don't need pay to back it up. It's like I just do it because I love it.

speaker-1: I it depends a lot on the teacher. So I completely appreciate that a lot of people glazed over at what I was just talking about. I might be a bit biased because my audience is really geared towards that kind of technical discussion. Okay.

speaker-2: Yeah, you're you're good, you're good. I just ⁓

speaker-1: I get but ⁓ that's what I'm saying. I get comments where people are saying like, this stuff is really hard, but your explanation was very good. And that stuff really like it makes me feel really good. And it makes me think back to when I was in university ⁓ and taking electrical engineering classes and the way that they taught electrical engineering classes at my uni I thought was pretty bad because they would give you just like a piece of paper with a circuit drawn on it, and the circuit didn't do anything. It was just like, ⁓ here's a resistor. Here's a power supply. ⁓ here's like a capacitor. Figure out what the voltage is at this point. It's like, okay, why do I care about that? But as soon as it got to like, okay, now design your own circuit with a little chip on it that actually does something, like it blinks an LED or something like that. I learned so much more just from doing that than the whiteboard exercises ever did. And I think it's for sure, like the concept is the same, because at the end of the day, the brightness of that LED is based on the voltage of that point. So it's the same concept, but it's just explored in a different way with a different teacher. And I think that makes all the difference, to be honest. Because there's there's a point for what you were saying with maths. I actually struggled learning maths as well. Because maths in a vacuum to me was like, why am I learning this? What's the point of learning this? But as soon as it got to things like bit maths or anything to do with like frame buffers for like graphics, that's when I I was like

speaker-2: Mom reveal.

speaker-0: I can't s I can't see her. ⁓ okay I can't know.

speaker-1: Hello, I can do a little bit.

speaker-2: It'll be in the footage. All right, go on.

speaker-1: No, I was done with my point. ⁓

speaker-0: Well, I mean, I I also had a hard time learning math 'cause I had a bad teacher. ⁓ and I had the exact same experience. Like why am I why am I learning this? Like what's the point of this? But then I learnt about like bit shifting operations and I like, wait, so I can just like multiply and divide by two, but just by sh ⁓ this is like exploiting some mathematical property of of base two, j and you can you can do multiplication and division like with without doing multiple additions or like like that was It's stuff like that that once you understand it, like if if you if you don't have too many gaps and you actually understand it, it all becomes fascinating, at least at least for me.

speaker-1: Yeah, one hundred percent. And I think it's a shame as well that a lot of people I don't know if you guys have ever felt this way, but certainly I can be discouraged from learning a topic based on my first impressions on it. So, you know, yeah with with maths topics for e for example, if I have a really boring teacher and that's my first introduction to that concept, for example, because I did electrical engineering, I had to learn a lot about ⁓ partial differential equations and stuff like that, because you can model It's things like capacitors and inductors with that. But my first introduction was just like, ⁓ here's a random ass equation. What's it useful for? I don't know, bro. You figure it out. It's useful for something. And yeah, here's how you like manipulate it to get some kind of result. It's like, okay, cool. I'm gonna scroll Instagram, you know? Like, I don't really care about this. But then as soon as

speaker-2: Yeah, That was my least favorite part of that.

speaker-1: Yeah, and I think it's just like the way that it was taught 'cause as soon as I saw how that was applicable to the stuff that I'm actually interested in, it was like, ⁓ okay, this is sick, this is sweet.

speaker-0: Did you get any insights on this working as a as a teaching assistant?

speaker-1: The stuff that I taught as a teaching assistant was actually really interesting. So it was it was not the course that I'll I'll tell you about the course that I tutored for for example, because it's not like I tutored like an algebra clo class or something like that. I taught a practical literally the the task is from start to finish of the semester, students get in a team and they build an electronics product. Like they build an embedded system from start to finish. So what we would do every semester, I say we, but out my ⁓ like Coordinator of the course was awesome. What he would do is every semester he would come up with a unique project idea. So, for example, one of the things that we got to do was like make a MIDI a piano keyboard. So it's like you have the piano keyboard, it connects to the PC over USB and it sends MIDI signals. Like you guys know MIDI? Yes, of course. Yeah, yeah, yeah, yeah, yeah. So awesome, awesome. So it sends MIDI signals to the PC.

speaker-0: I have a music degree, so yes.

speaker-1: And the idea is like you can access it as a device in a MIDI player and you can just go beep bew and hear it in your headphones, right? So that was an example of one project that we did. Another one was like ⁓ a macro keypad. So if you guys have seen like an Elgato stream deck, it was basically like make an Elgato stream deck. And so it's it's cool stuff. Another one was a multimeter, so like probe voltage, et cetera, et cetera. So it was basically like from start to finish, do a design for a board that solves this problem, write all the firmware that goes on it. And then at the end of the semester, we've got all these like requirements that it has to hit. So, you know, for example, it it has to do this kind of functionality, like with the macro keypad. It's like you have to be able to execute three or four keystrokes in a row with one press, for example. ⁓ so you would do like hit one button, does it do control C, control V, et cetera? So it was really fun. And it was very open-ended. We we basically just said to the students, like, okay, by this point in your studies. You should know how to do a lot of the base concepts. So pull it all together and build something real. And that was what I really loved about it. So my role was not necessarily as someone who explains this is how you solve the mass equation. I would think of it more like a coach. So the students are thinking like, okay, we've got this super open-ended problem. We're considering solving it in a number of directions. And they get somebody in who's already done the course, already has some industry experience and has built projects of their own. To basically just like advise them and say, ⁓ it's interesting that you're exploring this aspect of it. But have you considered these kinds of roadblocks that you might encounter with that strategy? And just like help them explore alternatives. Like not tell them what the alternatives are, but help them come to that on their own. And that was that was really rewarding both as someone teaching and also as someone, like when I took the course, having that kind of environment was really good also. And so based on what we were talking about before. It doesn't really apply there because this was a very practical learning thing. The more interesting topic to talk about with that is how they were using AI in that, because that was ⁓ that was a bit bipolar. Have you guys done stuff like that in university or anything before? Like that kind of practical

speaker-2: It really depends on the the professor in my experience. Like some of them are really do not really care about the students at all, and the ones that do will actually give you those projects. Like I guess we had we had a class the professor wasn't good, but I kind of took it and ran with it called what was it called? It was basically like the Raspberry Pi class, essentially. Kind of intro to embedded. And I ⁓ I got this is such a me project, but I like got the Pi Cam 2 module and I like threaded it through my shirt through like a buttonhole, right? And then like I fed it into the Pi. And then I would take that video footage and I would use like a CNN to detect faces. And then I'd pass them to this API called face check ID, which like runs ⁓ faces against social media accounts and then returns you like ⁓ the the like LinkedIn profile or associated like often people will come up in like their schools databases. Then you can send that over to like Sherlock or an OsINT program and like pull their name and phone number and stuff so you could like be walking around and have like an airpod in and just like insta dox everyone you looked at. Yeah.

speaker-1: That's how you got the name of that recruiter, right? For your

speaker-2: Yep.

speaker-1: That's cool though, because it's it's not just the project itself. It's a demonstration of like an entire pipeline and all of these different like subsystems interacting with each other to do something really cool and really useful. And that's what this course that we were doing was was about, is is exactly that, because you can learn all of this stuff in a vacuum. Who cares, right? But then what's really useful for people is like here's how you bring it together and actually make something.

speaker-0: curious because I mean you've you've already achieved quite a lot quite a lot. Like I mean, obviously ⁓ not just professionally, but I I know ⁓ did you used to do surfing and CS Go ⁓ competitively as well?

speaker-1: Okay, so I got to global elite in CS GO, but I was not like a comp I was not a professional player, like in any in any stretch of the imagination. Like it global elite in CSGO is like the top point five percent of players global f like big flex, by the way. But within that, there is so much variance to like the pro level players who go like overseas and play matches and stuff, but between like the person who just like achieved that rank at one point in time. So I was very much I was I was nowhere near like professional player, but it's so even within like the top rank of the game, there is so much variance there. And I was I was a good player and I got the top rank in the game, but I was nowhere near like professional level. Sorry to disappoint. Anyone who was like lawcrafting out there that I had a double life playing Counter Strike professionally, I'm sorry to disappoint.

speaker-0: But I'm just curious, like do you manage your time actively? Like how 'cause for example, if I play video games, which I which is why I don't play them a lot anymore, when I do play video games, it's all I can think about. I d I don't care about anything else other than how to how to maximize everything in this in this video game. How do you make it so that you can still play video games but also like learn and like do something with your life out other than just video games?

speaker-2: I could answer this one.

speaker-1: Front 10 AI agents in parallel, obviously. Right? I'm kidding. But yeah, Silvan Gor, go ahead.

speaker-2: ⁓ one of the most useful things for me in like controlling my own behavior is this concept called ⁓ DRA or DRI, differential reinforcement of alternative or incompatible behavior. So rather than trying to resist the like thing that's an autism trap for you, like you just have to create a bigger autism trap that's productive. And that can be kind of hard to do sometimes, but you can like Do some things to kind of put your fingers on the scale a little bit. Like you can listen to I'll often listen to music when I'm doing a boring thing I don't like. The research papers, right? They they they do grab me, like once I'm into it, but it's the same way that's like programming where like you have to kind of overcome that inertia. And what I'll do is I'll use a concept in ⁓ like animal training called chunking where you kind of break the the sessions into really small kind of atomic pieces and then you reward the pieces with as big of a reward as you can. And rewards reinforcers are differing depending on the person. It could be food, it could be music, it could be whatever. And then you can train yourself to perform longer and longer intervals and fade the reward gradually over time. So like less of a reward, longer intervals. And you can build your way up to big things, but Right off the bat, like I like a D R I or D R A to immediately extinct like a habit I don't wanna be doing.

speaker-1: I need to get into the productivity hack world, to be honest. Because the question that Christian asked was like, how do I manage my time? The answer is I don't. I operate in this and maybe this is a sign of like undiagnosed ADHD or something. But the way I get things done is literally, dude, I just like hyper-fixate on something for eight hours at a time. So if I'm making a YouTube video, I have not yet successfully really been able to do the thing where it's like you film the you film like one half of it the day before You film another half, like the next day, and then you do the editing on day three. It's like if I'm gonna sit down and do a YouTube video, it's gonna be like start to finish on the same day, unless it's like a two hour long video where I might cut it out if I just run out of time. But it's often that I've gone until like two or three in the morning because I just wanna finish recording. And so it's the same thing if I'm working on like a personal project. I think a lot of people do this like where you have issues finishing projects. I have issues finishing projects because Once it gets like good enough, I'm on to the next thing. So I have like all of these 90 to 95% finished projects that work well enough to do what I need them to do, but they're not like fully finished because the next shiny little thing like grabs my attention. So I just have to really to get things done, I just like as soon as I get into that operation mode of like, okay, I'm ready to just sit down and hyperfixate on this for like eight hours. I just make sure to maximize that.

speaker-0: It may not be healthy but it's so relatable. ⁓ I

speaker-2: Yeah, we're the same way.

speaker-0: I think we're all exactly that. Like w I'm working on a spell checker written in C, right? Because I and I I did it in a way where I didn't look up any pre exist like I didn't look up the the distance algorithm, like none of that. I thought if I were to make a spell checker, how would I build it from scratch? I did it in C because C is based. and the way that it worked, of course, yes. The way that it worked is I don't work on it every day. What I do is I get these spells of autism where I'll like the first batch was like a week where all I did was work on this thing. Every waking moment was litigate this thing. Then I ignored it for a month. And then I spent two more days on it and I improved it dramatically. And now I'm back to this like pause in between. Like at some point I'm gonna get I'm gonna get ⁓ a snipe to get I'm gonna I'm gonna go, ⁓ I can do this, and I'm gonna spend like three days doing nothing but that. Unfortunately, the same thing applies to video games. Where ⁓ I've been playing this game called Oxygen Not Included. It's like a base management game. It's kind of like Factorio. And I realized that I I just spent three days doing nothing but playing and thinking about this video game constantly. so as long as I can avoid the video doing it for video games, or I could start live streaming video games and I can monetize it and then it's work. So then that's fine. Maybe that's the loophole.

speaker-2: I would say ⁓ as a as a piece of advice that changed my outlook completely. A lot of like the self help industry and like the the kind of time management stuff, like the mainstream version of that is all slop. It's all like someone who like took one college psychology course, like regurgitated ten points, made up a bunch of anecdotes. They're like three hundred pages of random examples. And what changed the game for me is just reading like the source papers. It's the same thing as like just reading the C header files. It's like You just read like the ⁓ there's a one paper that really, really was good for me talks about ⁓ how compulsivity and impulsivity exist on like a spectrum where they it's not like one end, like you can have kind of both at the same time. And I think like just actually like going into the medical literature and looking at like, okay, what is ADHD symptom actually look like? What does autism actually look like? And it's like There's this kind of gatekeepy attitude around like doctors, right, and like that profession, but it's really just all made up. Like it's not that hard to just read this stuff and like when you don't know a word, you just look it up and it's like it only takes like fifteen extra minutes and the the lens you get in the world is so much more accurate.

speaker-1: It's a bit it's a bit scary though. Like ⁓ I think if you if you open up 'cause I agree with you in in general, and I think people are very used to third hand information that just gets like spoon fed. And even with like AI writing summaries at at this point in time, you know, it's like there's so much like baby bird regurgitate the food for me, like chew the food for me and regurgitate it, where the original meaning and the original value kind of gets lost along the way, and it's like you're just consuming trash content. I think we have too much of that. But I appreciate though that the way maybe that came about is people open up a maybe like a 15 page long report by like some research institution and it's full of like graphs and statistics and it's a bit scary to dive into. So how have you have you had like a barrier to like when you first open up one of these research papers and it's like sc it's scary, like how do you break through that and actually just dig into it?

speaker-2: It really helps to like I think the approach to it is really tough. Like if you're if you want to start reading research papers, it's like very intimidating. You know, there's all these different journals with different amounts of reputability within there, there's all these different authors and they're all citing each other. ⁓ it's kind of like trying to break into contributing to open source software in a way. But when you have like a mentor or a person who is experienced that you can ask that like looks at you and is like, let's see, you're struggling with like you you think you have ADHD, but actually this is compulsive behavior. This paper would really help you. So that for one helps a ton. It's like immediately this is relevant to me. And for two, like I don't really read them in full. I skim them for the juicy parts. That's like a big like if there's a graph, I'm like, ⁓ I don't know. I don't really care. I don't understand it. Like I just need to know, like I don't know. Before I I had this kind of abstraction on top of ADHD and O C D and now I just have kind of clearer picture. I could go back in the future and like really dive into the details if I wanted to, but it's kind of an eighty twenty thing, you know, like I'm gonna it's like I'm gonna I'm gonna learn eighty percent of the Vim binds. I don't need to know, you know, the G command for like or the V command for the inverse the G command. Like who who needs to know that? I do know that, but

speaker-0: I think that we should probably start to to wind down the ⁓ the 'cause it's been going for ⁓ nearly three hours. And I'm gonna have to edit this, so

speaker-1: Yeah.

speaker-2: ⁓ no. Sorry, I've given you a

speaker-1: Horrible job. Your life's gonna be miserable for the next couple of days.

speaker-0: Having to to edit a fun conversation that I enjoy. ⁓

speaker-1: What could be worse?

speaker-2: You guys do you guys ever edit your own videos and you're like listening to yourself back and you're like, Yeah, that's a good idea. That's a g like, I kind of enjoy like it's like a conversation with myself. I don't know.

speaker-1: I just love hearing myself speak. I just love myself so much. I just can't wait to sit down and listen to myself again and again and again and again.

speaker-0: Is there any topic that you'd want to go ⁓ into Matt? Something that we haven't covered that you'd like to cover?

speaker-2: Mm-hmm.

speaker-1: Really interesting question. One thing I want to know about you guys is like what do you wanna do? What do you want to do next? Like in your career? Like as a a s like Christian, are you gonna get into are you gonna try and get like a programming job or like and ⁓ Sylvan, you mentioned you're gonna go into like leadership. What's your like career next steps? Like what's your it where do you see yourself in five years if you had to make a prediction like that? Yeah, exactly, exactly.

speaker-0: Is an interview after all, so yeah.

speaker-2: You wanna go first, Christian?

speaker-0: ⁓ yeah, can do. So I've ri I enjoy programming programming very much. I I like ⁓ this guy called ⁓ T Sodding. I'm not sure how you actually say his ⁓ say his name, but ⁓ what he does looks really interesting. I've applied to jobs here and there. I've done a few interviews, a few, like just to to kind of get a sense of how it how it works. I wouldn't mind it at all if what this ends up being is doing live streams on programming or 'Cause I ⁓ programming isn't the only thing that I do that I enjoy. Obviously, I I like philosophy as well and stuff. If this thing gets really big, I wouldn't I wouldn't be upset with that at all. So like if my career ends up being broadcasting, talking about technology, talking about programming, ⁓ having my own corner where I go deeper into t into topics, maybe doing live streams where I program live on the internet, I would very much ⁓ very much like that purely because I would love the freedom that that I'm a freedom enjoyer, very much so. That's why I travel so much. So ⁓ in I that's that would be it. Something like tea sodding, ⁓ but mixed with IRL streaming, talking about philosophy, that would be my ⁓ my answer.

speaker-2: Seems like you're already living the dream.

speaker-1: Yeah. I was about to say like that's if I look at your channel, that's like exactly kinda what you're doing.

speaker-0: Yeah, but the income would have to come from it essentially. That would be that would be the difference.

speaker-1: I gotcha, I gotcha.

speaker-2: Alright, this mine is like a a little I don't know, it's a little crazy. I'm also kind of ⁓ like the the Paul memes, the like prescients memes are like genuinely like I do have like insane visions. I've like seen some of this stuff coming from like a few months out. I don't it it literally did. I'm not even kidding. Like I I'm not even doing this for the bit and and not like d specific details of things, but like I don't know, I had a vibe that this this is how things were gonna turn out.

speaker-1: Yeah. ⁓

speaker-2: for a while. And what I'm seeing next is so in my Discord server, the vestibule, I'm like, which will eventually maybe be moved off Discord. It's essentially like a social credit system based around like ⁓ maximizing creativity, human connection, kindness. It's like we we use like a token economy as it's called in in behavior modification to like incentivize project work. And what we do is we basically have like behavioral ⁓ profile of someone. They have like kind of three levels, right? Like this is a kid who plays Roblox who's like really annoying, right? Is level one. And then like level two is like, all right, you're decent enough. Like you're working on things. And then level three is like, okay, this is a really like a fully grown adult. You're you're a mature person. And then we'll have like a few levels of ability for programming and we're gonna also do math and behavior, psychology, philosophy, all of those topics because Discord is infinitely scriptable and structurable so you can create like an arbitrary skinner's box. And then ideally, so we have like hiring people at the end of the funnel. Like once you reach a certain level of skill, we just hand you right over to recruiters. NetX is a backup reinforcer for the token economy. So people are really actually incentivized to ⁓ to work with us. And yeah, I'd also like to at some point like try to fix the open source software situation and like ⁓ I mean i through through I I guess I don't know. I'm starting a a social credit ⁓ I don't I don't know what it is, like a university slash yeah, it's a cult. Yeah, it's a cult. Yeah, yeah.

speaker-0: Yeah.

speaker-1: Dude, we we joked before about making like a human agent startup, but you that is you are doing the human agents. You are producing the human agents. That's crazy. What's your so okay, you said there's like an open source problem. What do you think is the open source problem at the moment?

speaker-2: Yeah. I think it's too much too many people like contributing AI slop and like shipping too fast and not thinking about things and that's leading to supply chain attacks from China, as I understand it, and it's like leading to a bunch of bad software. I don't I I haven't looked into this. Maybe someone is fixing it, but I think it just we just need more based human programmers actually doing stuff like

speaker-1: Have you seen what Zigs ⁓ the programming language zig, what they've said about AI commits? Have you heard about that? They ⁓ there was a lot of drama about this actually. They as a project have explicitly disallowed all AI assisted and generated contributions. Their reasoning for that is that ⁓ people who are allowed to use AI will often do like drive-by commits, as you mentioned before, where they'll just like get the agent to go and crawl the issues on GitHub or something and then vibecode a solution. for that and then submit the PR. But the issue is that that does not build a real contributor. That builds someone who's just looking to get like clout off of, ⁓ hey, I contributed to Zig. What they really want is someone who they can actually invest in to become a core maintainer later on down the line. So their reasoning for disallowing all AI assisted contributions is like we n we want to invest in somebody, so somebody has to invest in us. Like they have to actually put in work, put in effort in order to join. Which is very interesting because that led to this whole situation with Bun that got acquired by Anthropic and they're like, Well, we wanna do AI system commits, so we're gonna get the agent to port it to Rust instead. But ⁓ do you have an opinion like do you think that's too extreme, like just disallowing all AI commits in general, or do you think that's like a reasonable position to take?

speaker-2: I think that's a reasonable position to take. And another like a lot of people I I ⁓ I'm kind of working with on projects, like one is this ⁓ lower internet social media called friendly, ⁓ and they have the same policy. I I think like in the future we will see agents integrated in, but it'll be like it it should be kind of a social credit system again where it's like, Okay, you've committed fifty PRs that have all been banger features, like, yeah, by all means use agents, but you can't drive by commit.

speaker-1: Do you have an opinion on this stuff, ⁓ Christian? Like

speaker-0: I do. I I mean I w I saw ⁓ the clip from the interview where he said ⁓ where he said that the creator of Zig said that they're no longer allowing AI contributions. I think that if I start a project and I put in the effort and the work, someone showing up and throwing some slop at me, I would not appreciate that very much. I would I would find that to be rather unpleasant. I understand

speaker-1: I'm serving it to you. You've gotta eat it. Consume. Consume this.

speaker-0: And I I and I have to consume it as well, which is which is crazy, right? Like yeah. I I do not I do not blame them at all. I understand the logic. It's almost like I don't know, getting a chat bot to text a girlfriend for you. It's a little bit like that. It's just so impersonal. It's so i if your girlfriend gets upset at you for having Chad Gippity text her, I wouldn't b I wouldn't blame her one bit.

speaker-1: I'm pretty sure there was a South Park episode about that where where that was the plot of the it was such a funny episode.

speaker-0: I've seen it I just it was it Stan, I think, that that ⁓ yeah.

speaker-1: And Wendy was all like, ⁓ he's finally replying to me, he's finally putting in effort.

speaker-0: It depends on what direction the technology itself goes. And I think that we're gonna see this a lot more. I think that we're gonna see a lot more projects start to reject AI contributions. ⁓ we're I mean we're already seeing companies scale da scale back how much AI ⁓ people are using because and I think this is this is just my opinion. I have no evidence for this, but I have a suspicion that the cost of of tokens will only go up from here, at least for a while. So Because obviously the cost was heavily subsidized until fairly recently. And that's when the actual cost showed up. But here's the thing: if I were in their position, I wouldn't remove the subsidy entirely. I would do it partially. So I would I s I have a suspicion that the tokens are still subsidized, just not to the degree that they were before. So

speaker-1: I can't confirm or deny.

speaker-0: Yeah, no, of course, of course. I'm not I'm not I'm I'm not I'm not asking for a for for a take on that. I'm just this is purely my opinion. I think that's the case. And so what I think is gonna happen is the tokens are gonna get more expensive as they are less and less subsidized. As the agents get more complicated, they're gonna use more tokens per prompt, which is already happening. We're already seeing token usage going up dramatically. And I just think that from a from a business perspective, because eventually it always comes back to cost. Eventually managers and people who own companies, even if they're not tech they don't they don't have to be technical to realize that the magic code machine is getting increasingly more expensive. And people on my channel, they joke in the comments, they say that 2027 is going to be the year of the human programmer. AI human humans are going to replace AI in 2027. If it keeps getting more expensive, especially if there's some sort of trade war or literal war, ⁓ especially if it's going to be somewhere in the Asia area and it's going to disrupt ⁓ hardware, then the cost of this stuff is only going to go up. And ironically, I happen to really enjoy low-level stuff and how computers actually work. But if this come if this happens, if there's some sort of dramatic increase in the cost of computation, we might once again return to a world where Being cracked actually matters, where being able to squeeze as much performance out of the computer as you can will once again become a valuable skill set that someone can have.

speaker-1: I mean look at RAM prices, right? Like if nobody can just Yeah, dude, if no one can afford like a hundred and twenty eight gigs of RAM so that Google Chrome can have three tabs, you're gonna have to optimize. So You can get

speaker-0: Bought a computer, I know.

speaker-2: Three whole tabs? ⁓ my god. We're building it.

speaker-1: Yeah, yeah, yeah. No, I'm using my like ultra cut down Linux distro so that the actual OS uses nothing and then Google promotes the rest.

speaker-2: Damn.

speaker-1: Yeah, it's crazy. Yeah.

speaker-2: Yeah.

speaker-0: ⁓ in conclusion, join Vimothy's cult. Become become a human engineer. Yeah. A human agent.

speaker-1: Human agent. Yeah, yeah, yeah. We're entering the agentic era of human engineers.

speaker-0: We're gonna have we're gonna make humans teach them stuff and then they're gonna do things.

speaker-1: That's a good concept. You know, when you said before about like the magic code machine getting more and more expensive. To be fair, that's kind of what like before AI, like companies were viewing humans, like programmers, as magic code machines, and they were getting more and more expensive, you know? And then they got replaced by something because they got too expensive. So the magic code machine got too expensive. So we replaced it with another magic code machine. And now that magic code machine is getting too expensive. What are we gonna replace it with?

speaker-2: Love it or hate it, capitalism always finds a way to repair itself. No matter what the problem is, the system, like somehow something more optimal gets gets put into place. Even if it seems bad for a minute, it it kind of is self writing.

speaker-0: is a ⁓ a good feature of it in the sense that it has an actual like self correcting mechanism. Like bro, we can't just bleed money for fifty years. That's just not like you have to make money at some point.

speaker-1: It's not a bug, it's a feature, bro. Yes.

speaker-2: Alright, we should probably wrap it before it gets it gets super, super long.

speaker-1: ⁓

speaker-0: Okay. ⁓ Matt, do you have anything you you would like to promote? Anything you'd like to

speaker-1: ⁓

speaker-2: ⁓ that's a good idea. We should do a like a segment at the end where we like randomly name drop people for for no reason at all. No promo money.

speaker-1: Yeah, ⁓ my buddy Kyle. My buddy Nick. They don't have they don't have YouTube, but you should check them out, trust me.

speaker-0: ⁓ okay ⁓ my my sister, Jinny, I'll mention ⁓ my sister. There you go. She she

speaker-1: My my only friend, STM thirty two, you should check him out, he's pretty cool. I talk to him all the time. We're mates, we're we're very close buddies.

speaker-2: I'll drop the ⁓ the khagi search engine. I'm more corporate than you guys. I don't have a soul.

speaker-1: ⁓ no, I should probably dude, I I'm definitely not paid to say this. ⁓ Windows eleven is actually the best operating system. That's my that's my genuine opinion, guys. That's my genuine I'm just gonna plug gonna plug Windows eleven.

speaker-0: That deserves an applause. I like that so much.

speaker-1: You can trust me.

speaker-0: ⁓ okay, where where can people find you? ⁓ on the internets?

speaker-1: ⁓ they can find me on Hof.underscore World. Had to put the underscore there because YouTube doesn't allow URLs or things that could be URLs because World is actually a valid TLD. You can't have Hof.world because it's a URL. So I had to do Hof.underscoreworld. So yeah, that's the YouTube channel. I actually have some big community news coming soon. I've been working really hard on ⁓ community stuff, stuff being the word, things around the community, things for the community, but they're not ready that they're still in the oven. I'm I'm at that like ninety five percent project finished state. So ⁓ big things coming soon for the community, but at the moment just the YouTube stuff.

speaker-0: Okay, man. It's ⁓ it's been it's been such a pleasure. I I wish we could talk for like way longer, but I

speaker-1: That's the problem with autism.

speaker-2: So we're we're all ⁓ eight hours.

speaker-1: Yeah, I've got I've got ten more hours in me. What are we talking about?

speaker-0: I I literally have to deliver work for a client in like in like free hour.

speaker-1: But do you really know do you just one more hour? Just one more hour, just one more hour.

speaker-0: Matt, we we we actually have to bring you back on. ⁓ this has been fantastic. This has been so enjoyable. ⁓ all right. Guys, thank you, thank you for watching. If you have autism, go to go to Matt's channel. You I think you'll enjoy his content. It will soon

speaker-1: I d I would love to. Yeah.

speaker-2: If you made it to the end, you'd definitely have it.

speaker-0: If you made it for all of this, you'd you you you definitely have it. And ⁓ thank you for watching, and we'll see you guys in the next video.

speaker-1: Thank you guys so much for having me.