# 🔴 LIVE | I can't do it

## Метаданные

- **Канал:** Leios Labs
- **YouTube:** https://www.youtube.com/watch?v=HCsIZK0C5K0
- **Дата:** 17.04.2026
- **Длительность:** 1:49:55
- **Просмотры:** 664

## Описание

Discord: https://discord.gg/3bTXRJ7SyD
Twitch: https://www.twitch.tv/leioslabs

Research Request Form: https://forms.gle/ArJ7bd728jj4RGuT8

Music: https://www.joshwoodward.com/

## Содержание

### [0:00](https://www.youtube.com/watch?v=HCsIZK0C5K0) Segment 1 (00:00 - 05:00)

I don't know. Huh. I don't know. Uh yeah, let's get started. Uh hello wonderful people of the internet. Welcome back to Hello. Am I live now? I think I'm live. Hello. Welcome back to another episode of Lay's Labs. Um right, bunch of dates. Uh number one, I did not stream yesterday. I want to talk about that briefly just for people watching the VOD or whoever happens to be on here right now. So, look, um I was unable to stream yesterday because I have heart problems, okay? That's just what it is. I have heart problems. Um so, as of the start of last year or this year, I mentioned this before in a bunch of different videos. Um I got a condition called neo-pericarditis, right? It's the inflammation of both the inner and outer region. Myocarditis is the inflammation of the inner region of your heart. Pericarditis is the inflammation of the outer region of your heart. You put it together, you get neo-pericarditis, which is what I got. Um Now, the recovery time for this is 6 months with kind of like 3 months being the mark where you should be able to kind of do everything as long as you don't go to the gym. And I'm right at that 3-month mark. So, in principle, my heart should be more or less fine, but I and I don't know why this is the case. Uh it is sometimes still hiccuping, right? And when I say hiccuping, what I mean is that like my rule of thumb has been I can do whatever I want to do as long as I can keep my heart rate under 100, right? That's been my rule of thumb. Um yesterday when I got up in the morning, I just, you know, I put the heart rate monitor on. It was already 135. And there was basically nothing I could do in the morning to keep it down. Now, by the time 2:00 rolled around, uh which is a normal stream time, I was kind of maybe okay and I could have maybe streamed, but um I just decided I should probably take it easy. Um I had to go to the hospital multiple times in the past few months because of this condition. Uh and because my heart just kind of freaks out every now and again. Uh and when I say freaks out, it means like I get, you know, high heart rate. It kind of varies up and down a whole bunch. Um I get heart palpitations, so I can feel the heart very strongly in the chest, like this type of stuff. Um so, I've gone to the hospital multiple times, and every single time the doctors say, "Look, you're fine. We don't know what's wrong with you. " Um the only kind of inkling I have as to what's actually going wrong is um one of the doctors said that your heart's a muscle, right? And you have to retrain it after it just kind of killed itself like at the start of the year. And you know, if it is a muscle and you're like training your bicep or whatever, you know, your bicep will get sore after you train it, right? And so my assumption as to what's happening is I'm just kind of working the heart a little bit harder than it normally is and then it just has to have that recovery time where it just kind of works its way back into being healthy, right? And like stops being muscle sore after working it too hard. So that's my assumption as to what's happening. I really don't know all the details. But regardless, I wasn't going to stream if I didn't feel good. Do I feel good now? Maybe not great, but I feel good enough that I feel like I can stream. Full stream Kevin, hey, what's up? How are you guys doing? And anyone who might be lurking in chat, hey, thanks for being here. Couple other no killing hearts. Yeah, I mean like I've been kind of tongue in cheek saying that my heart broke, which is more or less true. Like it is kind of what happened. Like I guess more accurately I could say that my heart grew three sizes in the span of a day. But yeah, like heart breaking is kind of what happened. And whenever you say like your heart broke, people think it's like romantic. So maybe I should have re-themed that video to make it more clear like I got neo pericarditis. But if I would have said that, no one would have understood what that is and I don't know. I also don't want to say myocarditis out loud to people outside of stream or video because if I put myocarditis into my title, the COVID vaccine anti-vaxxer people are going to come in because you know, myocarditis is one of the major side effects of the COVID vaccine blah blah blah. Villal and May Mage goes twitch. Why have I been misreading your name this whole time? Or am I mixing you up with someone else? There was someone who used to be called or at least I remember calling them Mango goes twitch. Are you actually Mage goes twitch this whole time I've been calling you Mango goes twitch? Cuz if so, I'm so sorry about that. I've been messing up. Finally, one last thing to say, um you're mixing me up. Well, wait. Uh Yo, what's up? So, one other thing I want to point out, I changed my profile picture across all of the platforms to be um this. Can you see it? It's just me looking at a paper while hula hooping. Uh the reason I did this is because um long story short, um I I've been realizing that I don't kind of want to build my channel into like this big laboratory. You know, before, like when I changed it to Leo's Labs, I had the ambition of basically creating an open-source laboratory where I'd hire people, they would work for me, we would do research together. That was kind of the goal. Um but I've kind of realized that's not what I want. I want to just be an independent researcher. And if we do have money to hire people, you know, I'll hire I will hire them to help me edit my channel, or I will take that money that I could have used to hire people, and I'll just fund people's research projects. You know, I'll give them grants so they can do their PhD, um and I can work with partner institutions. You know, like that to me sounds like a better way to go about things. Um instead of uh like having a company with employees who are like dedicated to you to do stuff. You know what I mean? Like so

### [5:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=300s) Segment 2 (05:00 - 10:00)

I think what I want to do is focus more on I know this sounds very arrogant or self-centered or I don't know what the word, but just focus more on me. I mean, that's what most people do anyway. Like it's called YouTube for a reason, and you know, I I think I should be focusing on making the channel as good as possible. And so, I liked this picture that I had made for a thumbnail a while ago, and I said, let's just use that. That's my thumbnail from now on. Um that's it. That there isn't really some like deep reasoning behind it. It's just that after the heart condition as of like, I don't know, the start of this year, I I've just been kind of thinking about restructuring everything, um and thinking about the best usage of like me on this platform. And I keep thinking the best usage is not to centralize everything behind Leo's Labs, um but to instead uh focus on um like just growing a scene and getting more researchers out in the open. Right? That should be the goal. So that's my new profile picture. Sorry about that. I just want to let everybody know that if you see a hula hooping guy holding a paper with like text behind it, that's me. That that's me. I'm the hula hooping guy. Back when I my heart wasn't broken, I did a couple hula hooping streams. So that's the story behind that one. Good. Let me go ahead and see what else we have to do today. So Thank you so much for following Desiles Shinski. Appreciate that. Ready to be filled with knowledge. Yeah, today might have not today might not be as knowledgeable as the other days. Okay, it just it might not be. Sorry about that. Today might be a little less knowledgeable. We'll see. Okay, so let's talk about what we've been doing. So let me go over here. Let me CD into Oh, I just messed up big time, didn't I? What is going on with my screen? Why do I suck so much I just existing in this world? You know, I that is a big question that I need to solve. So CD projects papers Siggraph CD projects papers Siggraph. Okay, so Let me go ahead and do PDF LaTeX over here. PDF LaTeX and What broke? Could you tell me? I think I know what broke. Let's go into VI main. I was working on this just before stream and so I think I might have messed up somehow. 199 um It says this broke? Oh, because I forgot to site. I think that's why. uh site and control D try that again. There was somebody I don't remember who was asking like how to write LaTeX because they wanted to typesetting language and I just have to be brutally honest in that I don't think LaTeX is a good typesetting language. It's like the default like the de facto standard. I think it's great. I've done a lot of things with LaTeX but it's tough. BibTeX PDF um oop Oh, come on. PDF LaTeX and then PDF LaTeX one more time and now let's open this bad boy up. Okay, so I've done a lot of stuff actually as of yesterday, right? I know like yesterday I was still sick, I wasn't able to do too much. Um, but I actually was able to talk to some people about a new formulation of this method that would make it amenable to quantum analog quantum systems, which I'm very excited for. I don't know how it's still up in the air, but um, this is one of those projects I would probably like kind of get started and then hand off to people who know more about the quantum infospace than I do. I'm just being honest about that. Um, and so with that in mind, like I think it could be one of the biggest papers out from this research venture, which is, you know, um, uh, general purpose graphics on an analog quantum device. I think is an amazing uh, title um, and an amazing research thrust. Uh, but that was kind of what I was spending most of my time doing yesterday. Um, as well as just like a few small scale things here and there. I really was trying to take a break yesterday. So, let me explain what we have left to do on the paper because long story short, we've been working on this paper. Wow, it's um, it's called composable function systems as a general purpose rendering framework. And I realized after talking uh, to a friend of mine, talking to the quantum person about this, they informed me that they didn't really understand what was so novel about the technique that I'm doing. And I realized that it is actually kind of difficult to understand the novelty here because there's a lot of like um, arguments that are kind of going all over the place. So, I restructured some of the paper to more accurately reflect where I believe the novelty is. So, let me try to induce a present that information to you guys. Uh, blue charm quark, thank you so much for following, really appreciate it. So, um, what I've been saying is that it is possible to do um, potentially real time, but definitely general purpose graphics um, simply by manipulating function systems, right? Now, that seems obvious because I think anybody could very easily consider or create a function that would allow you to draw a circle or something like that. So, let's just kind of walk ourselves through what's happening here. So, let me open up this little PP. Um, and let's just say we want to draw a circle, right? That's what we want to do. So, you want to draw a circle. That's the question. What do you do? Easy peasy lemon squeezy. Let me just go ahead pull out a uh a quick drawing tablet and let's just explain how you might draw a circle. Um Huh, can I Where's my pen? Oh, there's my pen. Wow, I put it in a stupid spot, didn't I? How do I get that? Okay, once I'm working on it, guys. I I

### [10:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=600s) Segment 3 (10:00 - 15:00)

haven't needed my drawing tablet and now I don't know what to do with my life. Um I don't miss being forced to take notes in LaTeX for math. My high school math teacher was keen on draining us. The thing is I don't think that LaTeX like if I am a teacher, I actually think teaching LaTeX is important. Um and the reason I think that is because um okay, I'm just going to have the pen there. The reason I think that is because, you know, if you are going to do math research or if you're going to do undergraduate stuff or whatever you're going to do, um LaTeX is just the probably easiest way in order to write math period. You know, like to typeset math in any way. Um even it's you know, internet browsers will use uh like um LaTeX formatting for their stuff. So, you want to draw a circle. Let's get into this. Right? Uh let's start by drawing a box for which you're going to put the circle in. So, let me change this to white and let's draw a box, okay? So, step one No, don't do what I just did. Uh how do I draw a box? Here we go. Now, let's draw a box. Bam, box, okay? So, this box is going to be the grid by which we are going to draw a circle. Right? Um it is going to be composed of smaller boxes. Box, box, box that are called pixels. Right? The easiest way to draw a circle is to imagine that a circle exists right here in the middle space. Right? And just to say the following. To say that, hey um each pixel, which again is going to be represented as these little points here. Pixel, pixel, pixel. Right? Is going to be some XY space. XY coordinate. Right? And then we say, "Look, if we're in the circle, it's going to be black. If we're outside white. " I think maybe I'll color the interior of the circle orange or something like that. So, if this if the um uh if the pixel ends up inside of the circle, we'll color it orange. If it's outside of the circle, we color it black. That's it. Simple, easy peasy function in order to draw basically anything you want. This is called a fragment shader. People already have this and they're able to use this. Just fine. The problem is, let's say we want to turn this into a general perfect general purpose graphics animation engine. Like we want to do this for general purpose stuff. Now we have problems. Because now, how do I turn this into an anime face? Like Goku from, you know, the famous Street Fighter 6? Um the way in which you do this is by um I don't know, maybe stretching it, squashing it, maybe drawing some eyeballs on it. This type of these types of arbitrary transformations are hard to imagine in your head in order to write them out with math. So, we need some system that allows for us to um manipulate points in space um in a way that can be like easily grasped by a layman. Like can we just take a point and move it with our mouse, right? That's what we want, right? Um the other thing we want is to minimize the number of points that we need for our final output, right? So, for example, um let's say this is a 1920 by 1080 image. Well, you might have like a whole bunch of points inside the circle, right? You could have like a million points inside the circle, maybe. I don't know how big the circle is, right? Um that's a lot of points. We don't want to keep a million points in memory cuz that's way too much memory. Well, I mean, we're talking about graphics algorithms from 1975. Nobody had a million points to storage in their memory, right? So, what do you do instead? Well, the nice thing about expressing things in this functional approach, um like the approach uh by which you say, "Look, if we're inside the circle, it's um uh white. If we're outside of the circle, it's black. " Or whatever. Um is that technically, you don't have to store these points in memory. All you have to do it Well, you do have to store the final output image, but you don't need like input images, right? You don't need um Well, you can do is you can say, "Look, we have an XY location. First, we're going to draw to this pixel. Then, that pixel. " So, you can create an instruction that allows for you not to keep all these points in memory, but to instead iterate on a single point um such that you only need one auxiliary point um and you don't need to keep like some separate input array that you act on or something like this. So, basically you have a function that allows for you to iterate in such a way that you only have to output to the final image. So, that's the advantage of writing things in a functional way, in that you don't have to keep data. You can just kind of like express everything in the way you want to draw it, and then you just draw it. And you only have to keep the final output image in memory. However, because to be honest, and I don't mean this in a in a like negative way, um, people are stupid. I am stupid. I cannot come up with a mathematical expression in order to turn this circle into Goku, right? So, what I'm going to do instead is I'm going to, uh, create mesh points. And these mesh points are going to, you know, they're not going to be able to perfectly represent the circle because these mesh points are going to be triangles, but we can draw this circle as if it were a whole bunch of triangles. And then we're going to manipulate those mesh points. Great. So, now we have a giant mesh that looks like circles, we can manipulate it, and we can turn it into Goku's face, right? I pretend I can draw. I'm not doing a very good job of this. Oh, yeah, like I'm going to need someone to fix this in post if this turns into a YouTube video. This is not a very good thing. But, that's the point of meshes, right? It allows for us to just kind of move points around in space with our mouse, um, where it's hard to imagine how to create similar transformations in math, right? So, this is where we were in 1975. We were deciding like, how do we draw graphics? The way we're going to do it is via meshes. This is just what we're going to standardize on. So, the entire graphics pipeline is based off of meshes. You've got vertex shaders, which allow you to manipulate these vertices in space. Ah, this is impossible to read. How do Let me erase some of these so you guys can see it maybe a little bit better. Uh, let me erase the orange Let's again. Let me just erase everything and let me try this one more time, okay? Let's try erasing all this. Okay. And let me draw a new circle in

### [15:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=900s) Segment 4 (15:00 - 20:00)

here. All right, here's our new circle, right? Now, and then we're going to draw some new points on the circle. So, here's our mesh points on the outer point of the circle. And right, like it's not going to be a perfect circle because like the mesh points are triangles and you can't represent a circle perfectly as triangles, but you know, it's close enough. It it's good enough for most representations, right? So, now we have the system that allows for us to, uh, manipulate points that and basically draw general purpose graphics. Okay, I'm really trying to draw mesh here. You guys have to just use your imagination a little bit for me. I can't draw everything out for you, okay? Like just a little bit, right? So, we can talk about Monte Carlo because actually Monte Carlo comes in just a bit. Someone in YouTube chat mentioned that. Okay, so, uh we have the question. Meshes are functions, people chose meshes, right? So, now the entire workflow is based off of meshes. We have vertex shaders which allow for us to manipulate the vertices and then fragment shaders which basically say, "Look, if we have this triangle here in space, what is the color that we need for this? " So, basically, this old method that I said like if we're in the circle color uh black, if we're out of the circle color it white, we can use that as a fragment shader, right? Because instead of um you know, instead of uh just drawing that directly, we're going to manipulate a mesh and draw that same shader onto the mesh. That's what we're going to do, right? Cool. It works, it's great. The problem is that we basically have to keep all these mesh points in memory, right? And as it turns out, uh and as GPU technology has been evolving, it becomes cumbersome sometimes to load um things to the GPU to begin with um and to kind of move the data between different GPU like within the memory hierarchy in the GPU, right? From like global memory to thread level memory and stuff like this. Um so, nowadays, we're looking for ways to take all this memory that we have, all these mesh points we have, um and find ways to draw the same object with fewer bits, fewer bytes, right? And so, this brings us back to functional approaches, right? And there are a bunch of different functional approaches that we could use. We could use, for example, iterated function systems. This is where Monte Carlo comes in and we'll talk about that in just a second. Uh we could use signed distance fields, right? Um and we could use other methods. Like there's recursive methods. I don't remember what they're called, but I'll call it a recursive function system. Uh but there are other methods that allow for us to have the same guarantees that you would expect from the functional methods. That is to say, we don't have to keep points in memory, um but also provide um the ability to do general purpose rendering or rendering that is general purpose enough, right? The problem is signed distance fields. I'm not going to talk about it too much, but a lot of these methods are What is going on? Something's going on with my drawing tablet and I don't exactly know what. Anyway, um so, sign distance fields, um and this recursive formulation, uh and also the iterated function systems, all of these methods are much, much more complicated than just drawing a circle, like we said before. Um they're much more complicated than like doing x is equal to x * 2. Um you really have to think deeply about what the math is in order to draw it out to space, right? And so, because it's very difficult for us to just intuitively, mathematically understand what we're doing with these function systems, um they end up being very underused and underutilized, right? More than that, um all these different function systems, they all have their benefits and drawbacks, right? Like I said, the iterated function system, which is the one that I'm most known uh I most know, um it allows for you to quickly draw function like fractally objects. But, to turn that into a general-purpose rendering framework, you have to have some other function systems to manipulate to work with it, right? And so, this paper, all this paper is doing is saying, "Look, we have all of these different types of function systems that don't talk to each other. " But, if we could get them to talk to each other, we can turn them into a general-purpose framework that is actually intuitive to use, right? The problem is there is no compiler that allows us to do this because Well, no, there is no straightforward compiler that allows for us to do this. Um because function systems, as functions, need to be compiled to memory. Um and in that process of compiling the functions to memory, you need to be you kind of need to lazily evaluate the compilation cycle. You have to kind of like you have to inject the functions into the right point of these other function systems, uh like the additional functions, the functions you need to make it general purpose, um so such that you can't just simply compile it. You have to uh compile it so that uh like you basically have to compile it on the fly. So, that's basically all this paper is saying. If you uh compile things on the fly and allow users the um capability to choose the point of compilation, then they're able to turn these function system sets into a much more general-purpose framework that was not available before. That's all we're saying, right? That's the thing. Um So, let let's see if that if you guys buy that. Um what's up, guys, by the way? A lot of people coming in YouTube chat. How are you guys doing? I hope you're all having a great day. Um And uh let me see what you guys are saying. Alt over here. Um You don't need any meshes because we have compute shaders. Yeah, that's exactly what we're saying. Fun fact, you need just one triangle of geometry to draw anything these days because SDFs and ray marching. Right, exactly. I mean, but SDFs and ray marching You guys are both saying the same thing. Like I'm also basically we're saying, what is the optimal way that you can create a compute shader such that you can do general purpose rendering without like the drawbacks of these methods that they all have, right? I find SDFs very well composable and intuitive. Might be for you. I don't find them intuitive at all. Specifically, if you have to do like smear frames um on um

### [20:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=1200s) Segment 5 (20:00 - 25:00)

uh smear frames on like you take a hammer and you want to show like it's kind of warping like this. Um sometimes like to do appropriate transformations um you're going to have to do like uh weird ray marching setups that go through non-Euclidean space and it ends up being complicated. Um now, I think other people might find it intuitive and you might be one of the people that find it intuitive. I find it very unintuitive. Um so, at least for me, this is largely beneficial because that with the method that I'm working on right here, you can basically just draw the SDF on, you know, some uh on some square or whatever you want. You just draw the SDF and then you just stretch the square out so you can now create the smear frames. So, the stuff that's hard for me to draw with sign distance fields. Like some stuff is still easier for me to draw with sign distance field. So, I will draw that with sign distance field and then I'll use my method to do the smear frames or the in-betweens or anything else that I need. Um and that's kind of the idea. The idea is that we can mix and match these and use them where they're actually beneficial. Um that's the whole shtick, you know. Uh raster and ray tracing make a lot more sense to me. Right. Uh I agree, 100%. Um the Yeah, that's what it is. I can draw a dot if that helps. Thanks. I appreciate that. Um so, that's kind of the That's the point of this paper. I just wanted to present it in a way that was kind of a bit clearer. Um what I'm trying to do is basically present a a very simple system that allows for us to have these guys talk to each other. Um and because of that, I realized I had to restructure the paper. Because as of the last time I was reading the paper, it was very much focused on iterative function systems. Don't get me wrong, iterative function systems are great and they are very useful and valuable. Um but by focusing so much on them in this paper, um I realized I realized that um uh it was maybe giving the wrong message. So, Anyway, we'll keep working on it. I'm not going to say that this paper is going to be like highly cited. I just want to get it out there. Like that's where I'm at right now. By the way, I have to cough. I will be 2 seconds. Sorry. Okay. So, that that's where we're at. This is kind of the the stick. I'm not very good at presenting this. I need to do better. Back when I was streaming regularly, I was much better at this type of stuff. Um good. So, if you guys have any questions, comments, concerns, ideas, anything like that, feel free to let me know right now. Um one thing I do want to say before hopping in further with this is that I'm very happy with like you guys as a community. I know I said that a lot, but I actually genuinely am. Like almost all the time, like since I have started I've come back from uh you know, a hiatus and I haven't made content for a while. Since I've come back, like you guys have genuinely been very strongly engaging with the content in a way that I hadn't thought was possible. See, I I don't want to get too many too much into the details here, but like I took a year off from everything. Like I said, in November, I stopped working for MIT and I basically we had a new kid and I used the paternity leave or whatever in order for me to uh kind of start investing more into Leo's Lab stuff. The problem was as of January, I got this heart condition and I haven't been able to do anything really since then. Um so, my goal was that this year, 2026, from January to January, I was going to focus all my energy on making Leo Lab Now, we can talk about what that means, but what ended up happening is I had to shift that. I'm holding a jack. I don't even know what this is for. I mean, I know it's for headphones, but I don't know why I have a thick one. Anyway, point is, um so I said that's what I was going to do, but now I have to shift it until I basically May 1st. So, basically, my goal is starting on May 1st, from May 1st to May 1st, or maybe like May 25th arbitrarily to May 25th, like sometime in May. That's going to be my year. I'm going to start there. I just had to push everything because of the heart condition. But yeah, my goal is to have some levels of success by 1 year's time. I don't really know what that means, uh but I've kind of have some like goals in my head. Like one is to find some way of self-funding myself. That would be ideal. If there was some way in which the community or I could get sponsorships from some people or there was some way to like basically make sure that this is financially stable. It doesn't have to be financially stable in a year, but I have to find a path to financial stability in a year, right? That would be the goal, right? But the other goal I have in mind is to get one paper out. One paper out that was basically generated by you guys. By that I mean like um How do I put this? Uh You guys give me an idea. And between us, we work on getting it and turning it into something that is actually like legitimately publishable. That's the goal. With a hula hoop, is that a topology joke? No, I just There was a hula hooping paper and I did a stream where we read the paper while hula hooping and it was a lot of fun. One other thing that differentiates a point generated approach like yours an image space fragment shader approach like ray marching SDF is that for a given transform, uh-oh, you got cut off. I don't know if you cut yourself off or if you just got cut off, but please keep going with that statement. Um A point oriented approach is image oriented while pixel oriented approach is pre-image oriented. Wait, explain what you mean by that actually. I am naturally interested in what you're

### [25:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=1500s) Segment 6 (25:00 - 30:00)

talking about. Um I am maybe not yet willing to open up a Patreon or GitHub sponsors. The reason being that I just have not been um I haven't been I like I haven't been here enough for me to feel comfortable, right? It says the function images and pre-images. I guess I understand what you mean. I think PRC would be happy to fund your work. Uh they definitely need scientists and good research. Who is PRC? Should I know PRC? Um the my goal is to like kind of in a year go up to a couple universities and be like, "Hey, can you just You guys are looking like OIST, for example, is constantly looking for um some form of uh uh how do I put this? Some form of sponsorship. They're looking for PR, right? I could go up to them and be like, "Look, could you just fund me? You know, like that's what I want. " I'll be getting sponsors anyway. I had GitHub sponsors open and then I took it down. I just I'm very uncomfortable with asking for money without giving anything in return. And that is what I've been doing for the past couple years. I just I did not feel comfortable having like that type of stuff open. I don't even feel comfortable having ads on. By the way, we have 25 people here on Twitch. How many people do we have on YouTube? Uh I need to ask all of you guys a question. 12 people on YouTube. So, we have 30 almost 40 people watching right now. Um has any of you guys over the course Have of the past few streams seen an advertisement on my stream specifically? Um because uh you shouldn't have, right? You really shouldn't have. Uh I on Twitch I'm not affiliated specifically so you guys don't see ads. And on YouTube I've turned off the ads so you shouldn't have seen anything. But just if you guys could if you guys happen to see an ad on my stream stream, let me know because right now it should be completely ad-free. Um my goal, you know, I have 25 viewers here on Twitch, and my average has been like 35 or something like that. You need an average of 75 viewers uh per stream to uh to become a Twitch partner. So, my goal was once I can be partnered on Twitch, I will go ahead and become partnered on Twitch. Um and then I'll leave YouTube ad-free. So, if people do want an ad-free viewing of my channel, they can watch it on YouTube. Uh and if they want to watch it on Twitch and they want to subscribe and take part in all that, they can then subscribe in that way. So, that's why um I wanted to uh like I just wanted to more express that. There I I always want there to be an ad-free uh available like an ad-free viewing for this, right? Uh it's the difference between, for example, a domain coloring which applies a transformation to a coordinate and the colors the pixel versus a point triangle etc. method which applies a transform to an object and then generates pixels. I see. I agree with that. Um that's like at the end of the day, I'm basically creating a flexible approach for a vertex shaders. Like that's all I'm doing, right? Um and actually, there was a paper that came out from AMD that you guys linked to me the other day uh talking about they were doing exactly what I was thinking about doing. They used mesh shaders for doing basically the same thing, whereas I'm using uh iterated function systems and different methods, right? Um but I'd consider doing exactly what AMD did uh with their whole work graph approach in order to allow for the same style of rendering, but I didn't I ended up going down this route. Definitely interested in helping contribute to the research. I have a math degree and 6 years of experience in software engineering and HPC. Love research. Yeah, you guys are the people I really want to talk to and like communicate with and and collaborate with. Um there is however one final thing. So, I said there are two methods of success. If in 1 year, um I generate enough money to sustain myself, I consider that to be successful. If in 1 year, we have at least one paper published with you guys, I consider that to be successful. But, there is a third way in which I consider this to be successful and that is if we can get a paper out um that maybe is slightly independent of what I'm doing with you guys, uh but is just a genuinely banger paper. I will also consider that to be successful, right? And I'm working on all three of these in different ways. Jesus. What happened to my hand there? Um and so like I have a goal for putting out a genuinely banger paper that I'm working on people with in the background that may or may not show up too much on stream. Um, I've got the stream stuff, this paper that I'm hoping to get out soon, which isn't a great paper, but I think it's a fun paper. It's one that we can easily put out and I don't think is uh necessarily negative. And then I'm currently talking to you guys to figure out what the good project is that we can work on together. Again, when I'm looking for projects, I'm looking for not only projects that are academically interesting, but for the stuff I want to show to you guys, it's also got to be somehow a little clicky, and I'm not sure what that means. Like, you know, I've already got potentially a paper where we um talk about like, you know, um abstraction layers in Julia and all this kind of stuff that and we can put out a video called like, you know, we beat Nvidia or something, but like it's got to be just a little slightly you know, it's got to be slightly easy to grasp from the perspective of like your grandmother. Like, you know, like everyone should be able to understand what we're doing and why we're doing it, you know? Even if they don't understand the details. No ads and not and on not giving anything back for GitHub sponsors, the feeling of being able to support genuine research is itself genuinely rewarding for a lot of people and gives um them some optimism, too. That's easily worth a tiny percentage of their income. I understand what you're saying, and I do really really appreciate that. Happy little rat, you were on the wrong platform, man. But hey, how are you doing? Um, right, I was going to say this. Um, yeah, anything visual is clicky

### [30:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=1800s) Segment 7 (30:00 - 35:00)

that's why I've been doing more computer graphics research research. What I'm saying is this, until this paper comes out, which should be out hopefully in a few weeks, a month, I don't know how long it's going to take for me to do the stuff here. Um, until this paper comes out, I am somewhat unwilling to open up any type of sponsorships or take ads or anything like this. The reason for this is because on YouTube, let me just face it, there are a lot of pseudo scientists on this platform on the platform that is uh YouTube and Twitch and stuff like this. Like, if we go to the science and tech category on Twitch right now, I can guarantee you there's going to be someone doing some sort of shady thing. Um, science and technology. Um, in particular, attempting to repair Game Boys. Hey, that actually looks pretty cool. I might watch that if I were not currently streaming. That's I'm not saying go there right now because I'm streaming, please watch my channel, but that's actually not bad. That's a pretty good one. Usually we don't find like people doing like interesting stuff in science and tech, you know what I mean? Um, what do we do? Um, point is like there's always one that's like Donald Trump is a pedophile. There's always on here somewhere. I don't know where it is, but like my point is that like there are so many pseudo scientists here on YouTube in the platform. Like and we have people like Sabine who will sometimes put out good stuff, but then also put out some weird stuff. And I just want to make sure that like I have some leg to stand on saying that this specific method of doing research out in the open will actually produce some sort of result, right? And has some merit to stand on. You know, because you know, my wife mentioned it as well. Like, you know, the paper ideas that I'm presenting are great, but the fact that I'm doing it on YouTube or Twitch makes it feel like I'm a scammer even if I'm not, right? Um, and so this goes back to what I said the other day that there is a level of I don't know what it is. I cannot do certain things that other content creators might do. That I might actually really like to do because I have to tow the line of like being a I have to keep my credibility intact. You know what I mean? Like I can't like I can't do Vtubing, right? Because if I were a Vtuber doing research, which sounds really fun to be honest, um, you know, like I think there would be no world where someone would believe a paper written by a Vtuber. You know what I'm trying to say? Like I'm not saying that they're being like Vtubists or anything like that. I'm just saying that like you know, there's a level of you have to have some form of credentials behind you before people are willing to accept that like the such a method like I think you guys understand what I'm trying to say. That's all It is I have to do something that sounds good and is, you know, like I have to keep everything on the up and up. I I'm rambling now. I'm not doing a great job. For some reason science and tech category is more active and during American hours than EU. That's probably fair. Streaming open research is still very rare and valuable. Wolfram and Jonathan's live thing sessions were epic. Yeah, the thing about it is like Wolfram, for example, is kind of the perfect like person that I feel did a very poor job of actually proper science communication because Hm. Okay, for the record, you know, I can't sit here and say that every single person is going to be a like beautiful like great science communicator. Like I'm not a great science communicator. I get it. It's hard. But the problem with Wolfram is he used to stream on Twitch for people who don't know. He'd have like three viewers all the time and his stream title was like live CEO-ing, right? Um, that was fine, but he got so involved in this theory of everything nonsense that like even if his theory of everything is sound and perfectly viable, the fact that one of the premier researchers, you know, Wolfram is known for his tech stack, you know, Mathematica, Wolfram Alpha, all this kind of stuff. Like somebody who everyone looks to and respects because they are supposedly really good mathematician or at least very good at creating uh math software, um, is also going for a theory of everything. It makes it feel like that's all science is doing. And that's sad because theories of everything are almost always crank and crackpots, you know? And so like I'm not saying he was not a crank or a crackpot, but I'm saying is because he was focusing on this, um, it ended up turning into this thing where it makes people feel like, oh, this is all research is, you know, when it's not. Like, you know, I have ideas of for example there was a like I don't know. We read a paper on AI feng shui the other day. You know, there was a paper I found recently about using Newton squares, which is like basically Sudoku, um, in order to create an internet that works across power lines. Like there were like a whole bunch of papers like this that are really cool and really interesting that I want more people to engage with. But when you only see the science communicators going for like string theory or theories of everything or things like this, it makes it really feel like there is not interesting research being done at a level in which everyday people can understand and intuitively grasp, right? Which is the de facto the the way in which research is done. Uh your stream is so laggy. Is it laggy? Is anybody else lagging? Um when you say lagging, is there audio video uh like sync issues? Uh do I have quality options on Twitch? Is that the problem? Uh let me go to my channel. Here we go. And let's check. Quality I do Oh, maybe that's why. I don't seem to have quality options on my

### [35:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=2100s) Segment 8 (35:00 - 40:00)

Twitch stream right now. So, maybe that's why it's laggy. Um hmm I don't know what to do about that, but that is probably why um it's laggy and I can't fix that. I can say you can go to YouTube right now where I'm streaming live and you can watch it on YouTube and I will go ahead and give you a There's a link in the description to my YouTube channel right now. Um I was lucky enough to catch a couple streams where he was discussing API design with a Wolfram employee, had some good discussions about balancing ease of use and accuracy. To be fair, these are still very valuable discussions and I feel like I I've unnecessarily like attacked Wolfram. I I'll be honest, he's a great guy. I should not have done that. Um but uh yeah. For the record, guys, I have one problem with Twitch. I it sucks so much. Okay, I'm going to rant about this because I'm tired and I I'm like I You guys know what I'm doing. I'm just tired. Okay, look. Twitch has a program called the affiliate program. Uh they also have the partner program. Okay? Now, here's the problem. They also have what are called advertisements. Okay? So, the problem is the following. You and transcode transcoders. Tran- transcoder options. People on YouTube probably don't care about this at all. It doesn't matter. They have transcoding issues or they already have the transcoder. So, here's the thing. Um you do not get access to the transcoder. That is the thing that allows you to have multiple different uh like um How do I put this? Multiple different like resolutions, so not just 1080, but 720, 480, whatever. Um, you don't get access to this really unless you're either an affiliate or partner, right? So, you must be either an affiliate or partner. If you're partnered, you will always have the transcoding options. If you're affiliated, you're more likely to have them. And if you're not neither like me, you're less likely to have them, but it's still possible to have them, right? Um, and you're likely to get them if your stream is like one of the top streams on Twitch, which mine usually is. So, even without affiliate, I usually do get transcoder options, but not always, right? So, if I do become affiliated with Twitch, which I can do right now, then I get transcoder options and it's more likely that I get transcoder to coder options, which means the stream is more watchable by more people. But, the problem is if I become affiliated, now people by default have to have advertisements on the stream and there's no way to turn these advertisements off the stream. That's it. I can't I you must have advertisements on the stream if you're affiliated. And I don't want to show you guys ads. So, I as a content creator have to make a choice. Do I affiliate with Twitch and force your viewing everyone's viewing experience to be worse or do I not stay unaffiliated with Twitch um, and um, uh, possibly not have transcoder options for some of my streams like potentially today, right? That's the bet. That's the thing I have to decide. And I chose not to be affiliated so no one sees ads, but unfortunately me unfortunately means that sometimes I won't have transcoder options. I hate the fact that they have um, they force ads on every stream. Like YouTube does it the other way around. Ads are on your channel by default until you become like partnered with YouTube, in which case you can then turn ads off, right? Which maybe isn't the right way to do things, but at least I have the option to turn my ads off. You don't have that option on Twitch. Um, Let me see. Uh, so that was my stupid rant. Um, I think Twitch only offers those to partners even with a US internet. Uh, buh buh, not lagging. Uh, I mean stream theory is fine per se. It's just that stream theory has lied to us and now uh, it's a crank magnet on the internet. I think this is a the thing is this is my problem, right? Like Uh, I have a problem in that many people um, I really don't I'm I'm really walking I'm stepping on like a bed of nails here. So, I have to be very careful about how I step. Many people think that science is something that only smart people can do. Okay, I think that's a fair statement to say, right? Um and because they think it's a thing that only smart people can do, they think it is a, for lack of a better word, dick measuring contest between intellectuals. That's what they think it is. A lot of people think that's true, right? And if you were in a measuring contest with um intel- intellectual people, then you should be pursuing the most intellectual pursuit, which for many people they believe is string theory because string theory is like the hardest possible mathematical thing that you can do or whatever. And so, it ends up being that this is just a thing that attracts all of the negativity in research, right? Like I I don't know how to go into this. I don't know if I really should get into this, but I'm going to say it this way. Um you know, like the reason I got into research is because I was reading papers in junior high and I love reading papers. I love writing papers. I love the process of doing research. I like that feeling when you read a paper and you realize you're completely out of your depth and then you realize you've got two, three, four hours of kind of going through the rest of literature to try to figure out what it is that you don't know and then to like, you know, learn from there. You know what I mean? Like that process of being out of your depth, trying to do new things and really trying to push boundaries where you haven't done that no one has done before is the reason I like doing research, right? I just like that newness. And I'm also somebody that can't really engage with narratives. Like I'm read books. I can't watch movies. I can't watch YouTube videos because I just I need to be in control of that narrative. And research gives me the ability to be in control of that narrative. That's the reason I like doing research, right? But there are definitely a genuine number of researchers who get into it because they want to be the smartest person in the room. And these people tend to go towards the fields that are considered more difficult. Um and they tend to not have the same joy

### [40:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=2400s) Segment 9 (40:00 - 45:00)

in research um that you get from just, you know, sitting down reading a paper with a cup of hot cocoa. You know what I mean? And so, because they're constantly trying to prove themselves um and their self-worth is uh related to their intelligence and their ego's related to their intelligence, they end up spiraling um where they feel like they are if they ever feel like they're not the smartest person in the room, um then they start kind of grasping at straws. And this is why I feel a lot of people who want to be scientists gravitate towards the very difficult things. But this has the other effect where if all of research are these very difficult impossible to grasp, you know, non-understandable like concepts, you know, like string theory or whatever, well, this means that normal people can't get into research. They just can't because they don't understand the uh like they can't even grok what like the process of research is to go further, right? Like you know what? For me, the process of research is sitting down on the couch, reading a paper, drinking hot cocoa at the same time, just genuine enjoyment. Just like relaxing, enjoying my time. Like that's it. But if your point of research is to prove your self-worth, well, now you're in it, in my opinion, for the wrong reasons. I am now rambling too much and I'm so sorry. Um that was an unnecessary ramble. Um What do we have here? Uh usually ads can be properly used as a small 1-minute break for the audience because you can enforce when you want to show them bathroom break, stretching, getting some water. I do agree with you there, but I still don't like the I don't like giving people advertisements. I feel like it's unfair for me to connect my brand to advertisers that I don't necessarily agree with. I think that's unfair. I should be able to choose who is advertising on my channel, right? I don't care if I have to give Twitch a cut of that. That's fine. But I should not have to put a base 44 ad on my channel when I think that they're scammers, you know? Like I feel like that's unfair, you know? Um and so it is not just the viewing experience, which is what I talk about. I feel like it's more than that. Uh by the way, um uh Orzow, thanks for following and uh Shatting Chets, thanks for following, too. What a name. Um What's up? Uh is this what people think of or is it what you think people think of? What are we What did I do? What Could you talk more about that? I don't know what he said. Um What's your last What is your last surprising technique technical insight learning? Wait, what? I don't understand that one either. Um uh I knew somebody where their ego and intelligence were very intertwined. They were interesting but insufferable year round. Are you coding right now? Anchor. What is anchor? Thanks, what's up? For us, there's a solution for ads on the client side. Is there? What does that mean? Is there a way for me not to get Oh, I see. You're talking about ad blockers. Very clear. Good night. Z out. Thank you so much for stopping by. For us in parts of the world with no good internet or limited internet, Twitch is kind of a big no-no. I understand and that's why YouTube exists and you can go to YouTube where you should have the quality options that you need. Affiliate away. I'm running DNS blockers and SmartTube. No ads for me. I mean, we should separate the working field of string theory and famous string theorists and the string theory crank space. Yeah, I just So, I'm going to be honest in saying that expressing opinions is not attacking. It just adds a little spice. Yeah, maybe I was a little too spicy there. I do understand. Like my problem is that like, you know, when I was doing the Summer of Math Exposition, we ran into a situation where the first year we had literally a thousand people submit math videos and I am incredibly grateful that we had so many people engaged with the project because that was like the big project I was doing for Three Blue One Brown at the time. However, it was impossible to go through and we had all of these crazy cranks who were coming in and trying to provide insight like solving Riemann's hypothesis blah blah blah, right? Stuff that I knew nothing about to be honest and I don't care about because string theory. I don't care about these pop topics. I care about like AI feng shui papers. Like I like the fun little things. I don't like these big overall working goals that like, you know, I'm the one that just likes doing Oh, here's a fun little thing we can do. Let's do it, you know? Like I'm not the guy who cares about like, you know, solving P equals NP or these any of this type of stuff. Like I'm the guy who just like I just like reading papers, man. Like that's what I like doing. So, when people come in and they say, "Oh, I found a solution to this. " and you know, I distrust all academics because, you know, what accent am I doing? Anyway, I distrust all academics because academia is flawed for these reasons. You're like, "Okay, yes, there are flaws in academia. That's right. But, you know, it is still the best process we have by which you can showcase your ability to do research. And if you did not publish anything up until now, I cannot trust your ability to do research because that is the de facto test to show to showcase your ability to, you know, push the boundary of knowledge. " So, that's why like PhD research is usually not super great because it's like your first step into the process of doing research. Like, no one publishes a Nature paper as their first paper, you know? Uh you'll do something in maybe PRA or if you're lucky PRL, but I that very rarely happens. You know, like you're going to put something out there that's okay. It's great. It's a nice first step. And then you build on that, you know? Um there's no situation where you know, some random person is just going to solve some long-studied problem on their own in a basement, right? Like, you need to understand the whole process. And the number one way I was dealing with these cranks was I was

### [45:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=2700s) Segment 10 (45:00 - 50:00)

saying like, "Look, send me a paper. Like, this isn't something to do in an exposition contest. This is a paper, like in an academic journal. And I'm more than happy to read the paper when you have it published. But until then, this is not the right venue, you know? Like um My comment was related to your comment when people think of research. My comment was not a critique, more of an expanding the thought. Uh sounds sort of Central Europe. I don't know what the I'm sorry. I have children. I have children and um I have a lot of stupid accents now that I use for reading books. Um uh imagine if every pixel was a string that vibrated in 12 dimensions. I'd be interested in learning how you look for a fun papers to read. Man, just go to the archive. Man, we did an archive stream just the other day. It was uh on YouTube. If you go there, there is a video that's entitled uh something like test stream, please ignore, where we just went through the archive. And it was a lot of fun, man. Like, I love going through the archive. It's genuinely one of my favorite things to do. But it takes so long because I get so addicted to it. Um yeah, I was on the archive earlier today and I had to stop. You guys don't understand how addicted I am to that website. It's I'm incredibly addicted to it. Okay, wait, I spent too long on this. I have questions. I have questions now. Um questions to you guys because somebody here mentioned that they find sign distance field to be more intuitive than um any other rendering method. And now I have a question for you. I don't remember who it was. I think it was I'm not going to call you out, but there was somebody in the YouTube chat that said that. So, I have a question. Um Was it 2D SDF? What is a two-dimensional SDF? Here's my question. Okay. So, for people who don't know, what is a sign distance field? Um usually when you talk about sign distance field, you talk about it in the context of ray marching. Okay? Um the way ray marching work is it's pretty straightforward. So, let me just go ahead and draw it out for you. Um First, let's talk about ray tracing, right? So, the idea is that we have ourselves some sort of object that we want to draw in space. It doesn't matter what the object is. So, I'm just going to go ahead pick up the pen. I'm going to just say, "Here's the object, right? " Okay? Now, we have a camera over here. Here is our camera. Oh my jeez, what Screen ta- uh What do I set? Wacom uh one, I believe, is what I'm looking for. Is that what I'm looking for? Yes, that's what I'm looking for. Perfect. Okay. So, we're going to have some sort of camera here. Here's our camera. Bam, here is our camera. Here is your pixels that you're going to be drawing to. And all you got to do for ray tracing is you basically take the camera and you send a ray of light boom boom through this uh through this um uh grid here. And basically, the ray of light is going to hit one of your pixels on the grid. Here's the pixel it hits. And then it's going to say, "Boop. Oh my goodness, I hit a circle. " And this one's going to be like, "Oh no, I missed the circle. Ah jeez, what am I going to do? " Bam, you hit the skybox. Woah, great. But the point is here that like that's what ray tracing is. You basically say, um you're going to draw these pixels. You have a camera behind these pixels. And from that perspective, you're just going to send a whole bunch of rays of light through. So, you send one, two, three, four, five. And each one of those rays of light is going to hit the center of each one of these pixels and you're going to draw uh the object. Now, what is what does this matter, right? So, let's scroll down just a little bit. Let's talk about how this has to do with ray marching. So, ray marching says something different. Instead of having a circle, we're going to have an equation for the circle. And this equation is going to tell us how far away we are from the surface of the circle at any point in time, right? So, for example, if your point's here, you're going to be a distance L away from the circle, right? That's your maximum circle. Um and your circle equation's going to be something like, you know, here's point P, so it's going to be P minus R, right? So, your circle equation's going to be P minus R and then it's going to be Wait a minute. Uh P minus R and then the distance Uh so, it should be the distance between so, L minus P minus R or something like this. It's the distance between L and P subtracting out the R. I probably got my parentheses wrong, but you see what I'm saying, right? You're basically saying how far away are you from the circle? So, now you know you're some distance L. So, what you're going to do is you're going to move the ray of light in whatever direction it was going in a distance L, right? Now, you're a new distance which we'll call L2. And so, now you move L2 away, right? And now you're new distance which we'll call L3. And now you move L3 away, right? You see what I'm saying? And so, like what you got to do is uh you got to send the rays of light exactly like what you're doing with um uh ray tracing in here. But now you have an equation uh and a method of iterating on these points in space that allow for you to uh tell how far away you are from a point. So, if for example, your object is going to hit the circle, right? So, let's say you have another ray that's going this direction, right? So, your point space is here. You learn that you're let's say Q away from the circle. You move Q, bam, you hit the circle, right? This guy though just nearly passed the circle and he went point point and moved away, right? So, the difference between ray tracing and ray marching is ray tracing um usually nowadays involves a mesh. Um that is to say you have some mesh object and uh you can find the distance between um like the screen that you're drawing on and any type of mesh point and you just go to all those meshes. Whereas for signed distance fields, um you have an iterative method of going through this analytical expression that allows for you to draw these

### [50:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=3000s) Segment 11 (50:00 - 55:00)

objects, right? So, that's the difference, right? Um and now you're going to tell me that I did this wrong. I know don't remember what the exact equation is, and I'm too stupid. Uh I mean, I also don't think it's fine to hate on GTA RP string theorists, but there's actually working string theories that often get stray bullets from that discourse. Uh we'll talk about string theory in just a second. Let me see what um Al's saying down here. Um 2D SDF is like 3D one, but you have only one marching step. Basically, you know the intersection point from the get-go. Uh like that Okay, yeah. So, that's my question. I need to learn this because here's the thing. For me, with this particular paper, I want to basically draw um a signed distance field. I want to draw something uh basically like a fragment shader onto a square. Okay? That's it. A square I've rendered a square as points, and then I want to draw the signed distance field on here. So, here's your signed distance field draws a square. And then I just want to warp this square. So, the square now looks like this or something, right? So, that I can very easily do a smear frame on that square. That's what I'm doing. That's what I want to do. Right? Just to show it's very easy to take uh a signed distance field, something that's rendered via another function, and work it into my system so that we can uh we can do smear frames, which are often times very difficult to do in 3D anyway, right? Like it's usually people use like motion blur or something. But if you want like a really good animated smear frame, it's still kind of difficult to do this um in a in automatically generated way. So, I just showing that this is a very good use case for this particular method. So, the reason I uh I'm talking about signed distance field is that there's kind of two applications of signed distance fields that matter. So, the first is like for 3D modeling, right? Um and I'll talk about that in just a second. Modeling modeling. But the second is actually for text rendering. Um right? Uh so, basically what I'm trying to say is there's a lot of good people doing 3D models, and there's a lot of people getting doing text rendering, right? Um for signed distance fields. And these are kind of like the big reasons why people are using this method. Uh so, my question is what should we do? Should we do texture rendering or should we do 3D modeling? I understand showing a smear frame off of a 3D model kind of makes a lot of sense, but also I have no method of doing texture rendering in my current engine. And being able to do this via a signed distance field kind of makes sense. So, my default is to say we're going to do 3D modeling because this showcases more of like how to generate smear frames, which is a more difficult problem in a sense to solve. Um and like you know AI generated in-betweens are just kind of not good because they don't have artistic intent behind them. So, it's sometimes very difficult to tell like how the animation was supposed to happen. Um and so anyway, my point is that like this is kind of a bit nicer, but also I do kind of want to learn a little bit more about how that texture rendering happens. And so like what Al said over here, I understand what you're saying. Like I understand in principle what you're saying, but in practice I don't understand how that is a sign like basically what you're saying is how is what you said different than what I was saying earlier about like are you in the circle or out of the circle? Like you know, is your signed distance function like for being in and out of the circle SDF of circle uh 2D? Is this just going to be like if like if in circle so if um uh P is less than R or something like this, you know, where your P is going to be your location. I guess P minus uh the the origin of the circle origin, which I'm going to notate in this way. Um less than R you know what I mean? Like and I guess we have to do like you know the typical you know, distance formula. You know what I mean? It's like X squared minus X squared Y minus Y squared. You know what I'm saying? Like basically find the distance from the circle center and say if this is less than R we it's black. And if it's not, it's white, right? I'm doing a very bad job with notation today and I just have to apologize for that. Jeffrey Skilling, hey, what's up? Thanks for the follow. Greetings, what's up Marian? Are you drawing with your mouse or do you have a drawing tablet on your I have a drawing tablet. Here's my drawing tablet. Hello. Um Yes, but then you have no anti-aliasing and it looks weird when you have to move and scale text. So, you're trying to do 2D SDF rendering where you have some translation from scene screen space to sample space for SDF. Um Yeah, that's essentially Yes, the question is I think I just need to look at the Valve paper is what I need to do. So, that's the first thing I want to do today. I say after we've been streaming for almost a whole hour and I don't have that much more time left before my stamina dies. Um the first thing I want to do today is just look at some 3D SDFs just for fun, just to show you guys what you can do. And the second thing I want to do is look at that Valve paper for the the text rendering. I just want to see if people are still doing text rendering the SDFs or if people are doing other things nowadays. I assume people have moved on from text rendering with SDFs, but maybe they haven't. Maybe it's still like the way to do things. I have no idea. Um so, hey Bernard, what's up? How you doing Hey, how are you doing? Good, so that's where at That's what we're doing. If you guys have any comments, questions, concerns, ideas, feel free to let me know and I'll totally address it as we as soon as I see it in chat. So, in my opinion and you guys might be wrong or I might be wrong. You guys

### [55:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=3300s) Segment 12 (55:00 - 60:00)

might also be wrong, but I can't Look, Inigo Quilez Inigo Inigo Quilez. There we go. All right. Let's go to the YouTube channel. And let's go to this image right here. And let me show you the thing I'm never going to be able to do even in my wildest dreams. That is this. So, wait, Red Blob Games has been posting a lot on Blue Sky about SDF text rendering the last few weeks. Oh, yeah, Red Blob is one of those guys I I've followed and I'm quite happy to see them still making content. Um ooh, I missed something up. I missed some stuff. Uh do do doop. Do note that Valve style text Valve style SDF text rendering is practically very different from uh some ShaderToy SDF. I do understand this and the question is well, that's why I'm like delineating between the two of them and saying that like uh how do I put this? I'm delineating between the two of them because it's like a a different thing, you know? Um but like I could do either of them. It's fine. I guess it very much depend on the output monitor. Uh do doop. Do know Redblob IQ the legend IQ IQ, who's IQ? Um new blog post yet. Uh circle SDF is radius minus square root of dot point center point center. Um dot being x squared x times x minus y times y. Why is it a dot product though? Because it's going to be x times x y times y. Why are you multiplying them together? Uh and then subtracting the radius. Why isn't it just a simple like length? Oh, I see. Point minus center. Yeah, point minus center. So, it's a I see dot you're doing square root of dot and the dot is the way Okay, I see. Sorry. It's point minus center. Yeah. Regarding text rendering recently, the slug algorithm got done donated to public domain. It produces much higher quality text in the SDF method. Could you give me a link or maybe you can't? Uh I probably have to look it up because you're on YouTube. I think the I think in resolution independent SDF Chlumsky's MSDF approach is pretty much state of the art. Um It's quite expensive to compute. Although most text I think is rendered offscreen uh in the form of Bezier's, not via SDFs. Yeah, so the Oh, IQ's a neo quillist. Got you. Um the how do I put this? Um I am Hmm. I had a thought and the thought left my brain. I think the truth is that how do I put this lightly? Uh there is so little going on in my head that I fear brain surgery because I think if someone were to drill a hole in my head, it would actually destroy the universe or at least the world because I think there is a black hole actually right there in between my the my two ears. Like it's genuinely my brain just does not work right now. So, what I'm trying to say is what am I trying to say? I was trying to say that uh square root is expensive to compute. Yes, you are right about that. Square root is expensive to compute. Um but as long as you're only doing like one or two, I don't think it's actually that big of a deal to force into um even a thread wise workflow. It's not going to destroy you. Um as long as it's not something that you do like a thousand times um every iteration or something like this, you know. Um as long as you use an analog solution, uh it can't get any better. Okay. So, when you were saying earlier about the slug, what were you talking about? Running text running uh recently slug algorithm. Can I just I'm going to look up the slug algorithm. Uh ba Slug font rendering library. Reference code for the slug algorithm. A decade of slug. Can I Can you give me a paper on this? If I go to scholar, is this going to give me what I want? Um feature selection using genetic algorithms. That's I don't think that's what I want. Uh I don't think this is what I want at all. Uh font rendering library. Uh software is a slug software library that has just become the professional standard for rendering high quality resolution independent text. Uh saying my brain is a black hole. Yeah. That's uh go to the bottom. Decade of slug. Thank you so much. I really appreciate that. Go to the bottom. Uh GPU centered font rendering directly from glyph Oh, ah I've seen this one. Yeah, I know. I think I know exactly which one this is. I have read this one. I read it somewhat recently, actually. Maybe one of you guys sent it to me. I don't remember. Um Let's go down here. Right, but I don't remember anything about it. Cuz I'm an idiot. Okay, let's see what we got. Um I want to say slug was by Yes, it was by that one. Yep. Why am I even streaming? Like you guys know than I do. Like, what is going on? You know, the thing about this stream is so weird cuz every now and again, we'll get someone who's like a high schooler and they're like, "I'm so lost. I've no idea what's happening. " And then we have other people here who genuinely are like experts in the area and know way more than I do and more than I ever will on this topic. And I just have to pretend like I know what I'm doing, you know? Like, I have to somehow keep face, you know, and pretend that like I'm smart enough to be with you guys. But you guys are all way more intelligent than I am. So, what am I supposed to do? Um What if it is software algorithm is a futile attempt unless you want to troll all the way? You patent troll, right? Uh one second.

### [1:00:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=3600s) Segment 13 (60:00 - 65:00)

Okay. Uh you have uh their protective geometry algebra textbook in my nightstand. See, what How am I supposed to compete with people who have a protective prospective, I guess, geometry uh algebra geometric algebra textbook in their nightstand? Like, I don't have that. Uh one of the points of streaming science is to gather us around. I'm not a high schooler, but I want to learn. No, no. Look, I Hey, Doc. One of the things for following, by the way. Uh I just want to say that uh How do I put this? Like, the community is drastically different. You go from here to like, I don't know, any random VTuber stream. Like, it doesn't matter which one. Like, you'll just a distinct difference in the community, you know? Just remember the high school undergrad PhD knowledge infographic. What we What are we talking about here? I recently implemented it. It's super easy, but it was patented, so you couldn't use it, but now it's not anymore. Um How is it How Why is it not anymore though is my question. Um like, what is the The shape of the glyph is defined by one or more closed contours. Um yeah, actually. I mean, if we're just using Um maybe this is actually the one I want to do. Maybe. Ooh, could be. Could be. be. Uh okay, actually, if I want to do text rendering, I think you've convinced me because if this using closed contours, these are actually very easily implementable. Um probably SDFs is the wrong way to do text, anyway. Yeah? Yeah, that's what I'm getting, probably. Uh you're the one that got me interested in image processing. I'm sorry. Yeah, I mean that's the problem, right? This happens consistently. Look, look, look. Here, let me show you what happens every time, okay? Let me draw a graph here with Let me get out my drawing tablet cuz that was not a good graph. Okay, let me get out a graph here. Here's what happens. All right, let me show you. Someone comes into the community, right? Right. Here uh how do we put this? Let me just draw here. Bam. Bam, right? This knowledge up here, k n o w l e d g e, and this is time over here, okay? So, here's my knowledge on a particular topic, right here, okay? And here is the knowledge of people coming to stream, and they're like, "Oh my goodness, there's a gap here. I want to learn. " And they Here's their knowledge over time. Now, here's my knowledge over time. I feel like I have nothing in my brain right now. And so now, there's a huge gap going the other way. way, and people are like, "Oh, thank you so much for getting me into research. By the way, I'm a director of some like, you know, institute dealing with like AI research or something. " I'm like, "Jesus, I'm still I'm not even a I'm a failed YouTuber. " You know? Like, what is the difference here? I mean, yeah, you could look at my credentials and say, "Okay, well, he did work at MIT for a few years. " But come on, like, what? Okay, you have to define what you mean by one. You can't just ask a question like, "What's 1 + 1? " without defining what you mean by like addition. Like, we have to take a step back here. We're going too far. Uh at some point, it'll just join the machine. Uh what do we have? Um ba-bum. Uh there's a trade-off to both. The industry standard alternative to slug is MSDF. Yeah, so you mentioned MSDF. Let me take a look at that as well. Um there's a trade-off to both. What Tell me more about the difference between these guys. Um I hope it wasn't a J JPEG XL situation. Um actually, did slug patent run out recently, or was it just Blinn's method? What's Blinn's You guys are talking about stuff I don't know. What is Blinn's method? So, a bunch of questions. Number one, Blinn's method. That's a I didn't mean to flip you off. That I just was counting and somehow my middle finger came up. I promise I did not mean to flip you guys off. Please don't leave the stream. I'm so sorry. Okay. Um I recently implemented so have you heard that everyone just got hit by patent trolls? The author just donated the patent of Okay, that's fair. Um actually just slug patent run out. That's number one. Um what happened to JP JPEG XL? That's number two. Uh the industry standard to slug. Uh what are the trade-offs to MSDF versus slug? Um MSDF is not pixel perfect and it does struggle with small text. That's why you still see stuff like FreeType or rendering deals. Uh okay, that's fair. Um it's Okay, the wisdom of the crowd plus Dunning-Kruger effect. TLDR, if you want to implement a cool SDF-based text rendering, it's multi-channel signed distance fields by Victor. Oh, thank you so much. That's exactly what I want. Let me take a look there. Um And let's grab this bad boy. Um is this going to be Uh can I get I don't mean to be Oh, no, this is what I was looking at. Yeah, yeah. I've looked at this before. Perfect. That's exactly what I was looking at. Uh Blinn's method is triangularized the quadratic Oh, of course, right. Sorry. Microsoft has a patent on the cubic version until recently. Oh, fair enough. Sorry. Okay, so good. We're doing a great job. You guys are all doing a wonderful job. Thank you guys so much for uh consistently giving me imposter syndrome. Right, good. You guys are great. So, let's uh See, like this is how like I was planning on having like a chill start to stream today. Like that's what I was planning on doing. Like genuinely chill start to stream where I'm just like, "Hey, let's implement an SDF and let's, you know, take a look at Indigo Quills' stuff and be like, 'Wow, this guy's amazing. Uh I wish I was as good as this person. '" Like I already walked into this YouTube thumbnail is literally like, "I

### [1:05:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=3900s) Segment 14 (65:00 - 70:00)

can't do this. " because I already felt imposter syndrome coming in here. And now I feel like there's no point in even going through this because you guys already know. Like what am I doing? Um so basically no one wanted to use JPEG XL because there was multiple uh patent holders including Microsoft. Uh What does triangularization mean in this context? Okay, so point is there are people who can draw really pretty SDFs in 3D. Like that is something that we can do. Um I don't know if I necessarily need to do anything like that for the paper. You know, the prettier the better if you're presenting to a like a technical journal in like graphics, but usually you also want to make sure that you're displaying like the thing that you're actually doing. Um in which case I think just doing any general purpose smear frame is fine enough. Um so basically I just want to show something that's distinctly an SDF. It doesn't have to be pretty exactly. Um so I'm thinking like a hammer nailing something into like just a hammer bam bam. That's what I'm looking for. A hammer nailing something into wood, right? That's I think a perfect use case of having a smear frame. Um and we don't have to think about it. Maybe a sword. Maybe a sword is easier and just moving a sword in space or something. Um that's what I'm thinking of. Um should be pretty easy I guess. Um Blend is really easy. You just put the Bezier curve on a bounding triangle and discard the pixels on one side. I see. There's a bit of nuance in the winding orders, but that's it. Well, to be fair I think you guys are convincing me that maybe I shouldn't be doing SDF for text rendering. I was just doing it because it would be a nice thing to show in the paper. That's it. That was what I was thinking to showing text rendering in like some sort of stupid like um uh I don't know like just stupid transformation on the text would be kind of fun. Um just it shows the same thing except now I'm doing text rendering. But I think probably it's best for me not to do that. Okay, great. Great great great great. You guys are great. Thank you so much. Okay, so paper is just on the composable function systems as a general purpose rendering framework. It's a method that I'm just it it's I I don't want to oversell it because it's not actually that great right now. It's literally just a method that I've been working on that allows for people to more easily compose different types of function systems together such that you can you know kind of uh you can use the benefits of you know, for example, um, ray marching uh and also just manipulate the points in space much easier and it's that's all it is. Like it's not What am I doing with my life? Though I do not understand how you simply SDF with your method. It's easy. You just you literally just draw a square and then you draw on the square and then that's it. So literally what you do is you just journal new page before you draw a square. Bam, here's your square. You draw the SDF on the square. SDF And then you do whatever you want to do with the square. Bam, I made a new transformation on the square. And now SDF is all wiggle waggled wiggle waggled everywhere. That's it. That's what you do. So You SDF is some kind of attractor for your points. No, no, no. You use the square or circle or any type of iterative function system. You don't even have to use an iterative function system. Screw it. Iterative function systems. You just draw a square that has too many points in the X direction and then boop. You know what you boop. You see what I'm saying? You boop. Okay, let me explain a little bit better. So like at the end of the day, right? Like you have um you know, your square is going to be composed of a bunch of points, right? Ba ba ba ba ba ba ba. And you can generate the square in ba. any way. For example, an iterative function system, but you could just like choose the point where you want you just do any sort of generator. It doesn't really matter. On top of the square you're going to draw your SDF, right? And now you can do arbitrary transformations on the square. So if you want to do this to the square so it looks all wiggle waggled, you can totally do this. And now you also have the SDF wiggle waggled like this. That's it. That's what we're doing. It's not this one's not super complicated. I'm not I'm all I'm doing is just putting one layer on the other. That's it. I'm not uh The thing is you guys are overthinking way too deeply about this and I feel a little bit bad about that. Um But you're right. It would be really cool if I could create like a generator for the SDF. That'd be really cool. But that's not what I'm saying like if you have an SDF, you can draw it to a square and then you get these mirror frames and these intermediate frames for free. Like it's very easy to do it. And there's no reason not to do it, right? The question then becomes how do you create the generator such that you have the correct number of points so that you don't overpopulate pixels, right? But that's also an area for future work, I think. I don't necessarily have to show that in this paper. But that would be that's like one of the things I want to move into. That would help with performance and stuff like this. But if you just overpopulate the square, that's fine. May the distance function as an attraction direction in the chaos game phase. Yeah, so the way this works though Yeah. Yeah, so I was distinctly you could definitely do that. But the problem with that is um Here's the problem with that. So here's the question, right? How are you doing this? So if you're doing the square, right? There's four functions, F1, F2, F3, and F4. F4. Right? So these are very simple, right? It's going to be P plus A over two, P

### [1:10:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=4200s) Segment 15 (70:00 - 75:00)

plus B C over two, P plus D over two, where we have A B C and D. Right? And this will generate your square. Basically, one wherever your point is, it will either move halfway towards A, halfway towards B, halfway towards D, halfway towards C. Um So the way I typically do this is I'll say, "Okay, we have a generator function for the square. " And then as we're generating the function for the square, we have point is equal to some function N of point, right? And so we have a new point. And now I'm going to say P2 in the process of generating that square is going to do something. It doesn't matter what we do with P2, we could do some sort of transformation on this. Right? So we do a transformation on this, right? And then we throw away P2 and we keep moving on P1. So we generate a point for the square, right? And then we do a transformation on that point. And then after doing the transformation on that point, we then splat this to screen. Or not, you can not if you don't It's up to you. You choose what you want to do here. It doesn't matter if you do or don't. I don't really care. Um you do something with that. And so, you would take the square that you've generated, and then you wibble wobble it, you know, wibble wobble it. Um and then uh and then you splat the wibble wobble to screen, and then you go back and generate the next point in the square. And by doing this, you don't have to overpopulate. generate all the points in the square and then do the wibble wobble. You can um you can generate on the fly. You can while you're generating the square, you then do the transformations, right? And so, what you were saying actually makes some sense here. What we could do is while we're generating the square, we take the point, and then because we have a signed distance function, and we could do a 2D SDF on that, I think. Because like let's say we're trying to draw the letter A, the signed distance function is telling us how far away we are from A at any point in time, right? And so, if we have an SDF, in principle, we have this point, it could be anywhere, and now we can just we know how far away we are from A. And uh the question is directionality. How do we know the direction that we need to move A so it's going to land on A, right? Just because we know we're so far away, how do we get it to uniquely land on A is the question. I don't know. That's a good question. Maybe there's a way to do this, you know? Um uh what may sound or look stupid for you can ignite an idea for someone, that's all I will say. Uh agreed, that's how we reached the moon. I don't know if moon, but we did reach the moon. Um What do you do? Um never mind then. Uh maybe use a distance function grad F. Um Oh, this is the other thing. Right exactly exactly. That's why I like this process because you can write the functions. Someone said grad F. Uh naturally interested. Yeah, you did say that correctly. Uh I'm very interested in code transformation here. It's not exactly uh auto differentiate It's not exactly auto differentiation, but it's pretty close to auto differentiation. So, if we use what people are currently using for um like AD tools, I think we could very easily modify it um and use it for this purpose of creating the perfect generator and using AD for some general purpose graphics framework on those totally it's totally hot right now. So, I'm a big fan. Uh do do. We would also be an auto diff approach. I think you mentioned that you wanted that. Yeah, it it's something I want. That's actually why I wanted a separate paper because adding auto diff to this makes it in a sense more interesting. Right, I have a problem here where like I have so many interesting things I want to do with this method, but the problem is I have to get the method out. Just get the method out and then do the interesting stuff on it. And that's where I'm at right now. Um it's so annoying when doing small glyphs. Uh I'm an Arabic speaker and we have a lot of small glyphs like Oh, that's fair. Yeah, I mean uh I don't know I know that some I don't know exactly what I'm talking about here, but I know some languages like um they have like dots. I don't remember, but they'll have like these little dots here um that represent whether there's like a vowel that follows the consonant or something. Um and I don't know, there's just a lot of small glyphs that I could imagine being complicated. Um would possibly probably really slow, but maybe fun. Yeah, it could be fun, I think. Um but the AD stuff, I think we don't have to worry about it too much as long as we're not hitting the compiler too much. Uh ba ba. Specifically do sign F grade to the so points get attracted to the surface. SGN is your signed distance? No, F is your So, you're saying do SGN of grad F. So, the points get attracted to the surface. So, SG N of F times the gradient of F times grad F, right? Uh to get you the points attracted to the surface. Maybe that's right. Could be right. Might be right. I think maybe I want to try that. Um by the way, Witek and uh Guk TV and everyone else, thank you guys so much for following. Really do appreciate it even though we have um some issues on Twitch right now. I really do appreciate it. Thank you so much. Um What is Isn't SGN sometimes sign? Am I stupid? Uh but I want to mention I'd want to mention for that to be on the paper. Who cares? Uh I watched your algorithm videos on YouTube like 10 years ago and today I actually find you on Twitch. Hey, thanks for being here. I really appreciate it. Um although apparently you've chatted before, so

### [1:15:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=4500s) Segment 16 (75:00 - 80:00)

you know, because it's not your first time chatter. Uh what is this? Arp egg four wave. Hello. So, the points getting uh inside get expelled. I see. Right, right. So, let me I need to think about this. You figure this out and I am just the idiot trying to follow along. So, give me just a second to think about this. Um there's a bit of stream delay. So, communicating is hard. No, no. Look, look, look. I appreciate that you guys are here and communicating. Like I don't think you guys understand. Right? Like when I tell my wife that I'm streaming and that you guys are actually helpful, she's like, "How? How is that possible? " And I'm like, "Because you guys are somehow like PhD holding geniuses who are for some reason watching this idiot who doesn't know what he's talking about. " And like that I don't know. I don't know why you're here. I don't know. But I appreciate you being here cuz we're doing stuff and it's fun, you know? Um yeah, definitely send it to me over Discord. But just thinking about what you said. So, we have the formula for a circle, right? It basically tells us how far away we are from the the surface of the circle at any point in time. But I don't think we want to be repulsed on the inside, right? Because if we're then that would bring us towards the surface when we want to color the whole interior, right? So, I don't necessarily feel like that's exactly what we want to do. Um but the gradient of F uh yes, will give us a direction. And then what we have to do on top of the direction is So, so the reason I'm not worried about compilation time, time is because I'm going to be giving these uh glyphs like for example, if we have an A, right? We're trying to draw this A, right? Here's your A. Pretend I can draw. Um if we can find the SDF for an A, right? Which I think is very easily findable online. Um then instead of reading in the SDF for an A, right? SDF, uh we just read it we'll just instead have in our library that's just read in for free, SDF plus or whatever gradient of A times gradient of A, right? So gradient of the SDF of A. You see what I'm saying? Like it's okay just that we have these automatically stored or we have them generated before runtime or something like this. Like I don't I'm not too worried about this cuz we don't have to do this exact transformation on the fly. I think we can find a way to have that decided a priori, right? So I'm not too worried about that. Uh I mean who just got BC hearing this? Who just got what's BC here? Um oh wait, I can send it to you on Discord with the surface thing. I was thinking about 3D. Yes, you want oh 3D. Yeah, yeah. You'd want sign of F but you probably want to set it to zero inside or all points would go to the center. Right, right. You're 100% right. Yeah, send that to me on Discord. I don't know what I want to do with that information but it is cool information to have. You've convinced me though that the way in which I was going to do SDF rendering is not actually probably the way I should be doing SDF rendering and that is really cool. So thank you for that. Um Oh my goodness, you guys are great. Like I said this before. I definitely said this before but like I don't know why but every time I stream there's always somebody on stream who knows more than I do which is incredible because again, we don't have that many people here. Like how many people do we have? We have 38 people on Twitch right now and we've got eight people on YouTube. You're one of eight people on YouTube currently watching me right now. So put that together we got 45 people here. 45 people is a large number of people. Don't get me wrong but the fact that we have 45 people here and of them like half of them already understand signed distance functions to an extent that they're able to give me like meaningful advice in this way. Come on. That's cool, right? Like am I crazy? That's cool. Like maybe I'm crazy but that's neat. Okay, so let me go back to Indigo Quills really quick. So, uh let me go to his website instead of This stuff is so cool though. Uh this is one of those things where I think I'm going to have to ask whether we care about performance or interestingness. That's what we're going to have to ask. Do we care about being interesting or performant? That's our question. Um, and uh the answer to that is um probably interesting at this point is probably more relevant. Also, my heart rate monitor went off. Um, sorry. Give me a second. I've got to get it working again. Um So, is it working now? Hello? Please. Please work. If you I have to close stream if you don't work anymore and I want you to be working. Sorry about that. Um, I just I need to make sure my heart rate's fine because it was messed up yesterday. I just have to reset it. Um, anyway, uh it's funny that there's so many of us that watch you for more than a decade. Yeah, I've been doing this a long time, haven't I? The more interesting thing is that we have such a varied audience between YouTube and Twitch. Like there are people on YouTube who think I'm like actually a smart person. And then there's people on Twitch who remember how often Gustorn would come in and just basically I don't know what the word is. Not body me, but like I would try to say something and then he'd be like, "Well, actually it's probably better to do it this way. " I'd be like, "Yeah, you're right. It is actually probably better to do it that way, but I don't know how to do it that way cuz I don't know C++. And then he'd be like, "Oh, just do this. Use some meta programming. " I'm like, "I don't know meta programming in C++ man. " You know? Um

### [1:20:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=4800s) Segment 17 (80:00 - 85:00)

Yeah. Steam uh steamroll maybe. It would but anyway, I The thing is I have completely different communities between YouTube and Twitch. And I'm trying to combine them now, but like for the longest time I think people on YouTube thought I was generally a smart person, right? Where people on Twitch thought I was a complete idiot. Um, and we'd have hour-long discussions on floor edition and water. I brought this over cuz I was going to read chat and then I didn't read chat. Uh you want sign. Your stream audience is naturally concentrated to researchers. Is graphics interesting and then people figure out how to make it fast? I think this may just be because people leave it so they didn't get it. Naturally, that will mean that ones who do get it stick around. Yeah, I mean, you might be right. Um you know, I hope and I dream that we can have a general purpose audience. Like basically, anybody could watch this stream and enjoy it. I do feel like we have some of that on Twitch. I definitely feel that is true. But on YouTube, I feel like the audience is different enough that like for some reason and it could just be because my channel's dead. There's so many different reasons for this, but people don't stick around on YouTube live streams as much as they do on Twitch live streams. Twitch live streams almost every day, you'll get someone and be like, "Yo, I'm this is super cool. I love learning things. " You know, um and then you have Justin who's just here for the MATLAB rants. By the way, how are things going? It's been a while. I hope you're doing okay. Um I just happen to have like five good ideas in this specific area. Yeah. The thing is the thing about graphics, which is so much fun, is that it's so easy to have good ideas in graphics. Like it's just it's so it's because they're visual intuitive. Like that's the reason physics is so much more fun than like biology research, at least to me, because like you can kind of intuitively feel about how things are moving in space. Whereas with biology research, it's like I mean, it's like, you know, it's observational research, that's number one. But the other thing is um like if you're doing it's like bioinformatics and stuff where it's like you have this large DNA strand, it's very difficult to just in your head figure out what's going on. Whereas with physics, it's like we got a ball and we got another ball and these balls are hitting each other. But actually, we don't want them to hit each other, so we're just going to because if they hit each other, they're going to like fly off. So let's not have them hit it. Let's have them just pass each other and act like they were going to hit each other anyway. You know, it's like I can very easily think of like 100 balls. I can't very easily think about like how to do um like edit distance formulations of DNA strands for like, you know, the human genome. Like these are different, you know? I'm doing okay. Thanks for asking. Hope the same for you. Yeah, I'm slowly recovering from the heart condition. Um so I'm back doing the normal ranty stream. Don't worry, I know Ish and I stayed to try and understand anything. Do fractals need to make use of floating point registers? Uh generally, yes, but there are some ways in which you cannot. Um what specifically are you asking? Like for what type of fractal? If you're talking about iterated function system, generally, yeah. I mean, um, it's because YouTube's really good at suggesting to specific audiences, or I think so. The thing is, just coloring your points with an SDF at some point in the pipeline wouldn't be really novel. Uh, you just want to implement texture lookout, but pushing the points towards an SDF surface, um, it I agree, 100% agree with you, is not something I've heard of before. I 100% agree with you there. Um, the interesting thing, the reason I wanted to do an SDF here is because, um, there are certain things that are hard to draw with SDFs, right? Um, and I, you know, the default is that I go to are smear frames. Like there's a lot of like these in-between frames. I'm not talking about, um, uh, I'm talking about a smear frame that looks like this, you know, like, uh, an artistically rendered, like, here, let me just draw it out for you. Um, I you know, I'm not talking about just an in-between. I'm talking about like, uh, jeez, I should just keep this on my lap at this point. Like if we have a box, right? And we want to draw a smear frame that looks like this, you know? Something like a curve or something like this. This it's not like it's impossible to do this. It's just that like or maybe like a smear frame that looks like a wibble, like again, maybe looks like this or something. Um, these types of things, you know, at least as of a few years ago were kind of difficult to draw with signed distance fields, um, at least from what I saw. Um, uh, or maybe you want to do like, you know, this and then you split this in half or you draw you force a circle in there or something like this. Like what I'm trying to say is like, you know, maybe it becomes topologically non-trivial, so you you explode the center of it or there's like 100 different things that I could mean. Um, but like doing an arbitrary transformation on this is not something that I've seen before. Um, but maybe I've just been not reading the right papers. And again, I don't know that much about SDFs. Uh, I just read about them and found that I couldn't really use them for what I wanted or I they were not easy enough for me to figure out how to use at the time uh, for what I wanted to draw. That's why I created a new system to do it. So, the whole thing here is just to show that there is interoperability and that's kind of the point. Just that you can do this with this method as well. I wasn't intending to show that this is, you know, like super new, but I do really like your idea of using the sign distance field as a way to naturally work with the method instead of just using it as a texture look up, which is exactly what you're saying. Like a texture look up isn't that interesting. Like it's very clear that, you know, you can just look up a texture and move things around. Um but I like this idea and I think maybe we should go forward with that idea. I need to think about it a bit more.

### [1:25:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=5100s) Segment 18 (85:00 - 90:00)

Um is this is my research project for aerospace engineering at a college machine learning surrogate models Wait, at for explainable optimal control in aerodynamics applications to cis-lunar trajectory optimization and space-based astronomical mission sustainability. Um machine learning surrogate models. What do you mean by surrogate models here? For explainable optimal control. So, obviously machine learning is super good with optimal control. Like I'm game for that. The question is like I always have the same question when people say that they're using machine learning for these types of methods. Why couldn't you have used like a reinforcement learning approach or like an on the fly learning approach just cuz I trust those more. Like maybe I'm just stupid, but having like a some sort of deep learning thing is like usually not like if you're just going to do optimal control, then why couldn't you have used I don't know, gradient descent, Nelder-Mead, or reinforcement learning or any of these other on the fly learning approaches. Why did you have to go as far as like a deep learning type of thing? And again, maybe I'm misunderstanding what you're talking about. Application is cis-lunar trajectory optimization and space-based astronomical mission. Yeah, again, I feel to me and maybe I'm just stupid here, but like a trajectory thing feels that we don't need machine learning for that as much as we need What Okay, I guess reinforcement learning is considered a like a part of the machine learning sphere. So, definitely is useful there. Anyway, what I'm trying to say is like I because it's such a loaded term, I don't really know what we're talking about anymore. Um it has a big part of my project. I know some of those words explainable with machine learning is a fun combo. Uh let's go here. Ba ba ba ba ba bum. Uh wouldn't shape deformed around movement path be sufficient for a smear? Uh the ones I've seen use a glitch production at glitch production seem not much more complicated than that. Um You're completely right about that. The issue is that like hmm It's like Okay, so a lot of people are using motion blur, right? Motion blur. Right? Um that's fine. Glitch uh but recently people have realized that motion blur is not really the best uh thing that you should be doing for I don't know. It's not like the best um How do I put this? It is not the best way to do 3D animations and stuff like that. Am I getting What happened on Twitch? What's going on Twitch? Can I see something really quick? Um in MGA, thank you so much for following. You train a faster get on a limited set of simulation results and then query the circuit. So, what I'm trying to say is um usually what these have to do is they have to be hand modeled, right? So, these have to be like from what I understand. And maybe I'm not fully understanding about it. But yeah, motion blur looks bad, but also like they have to be hand modeled, right? Whereas I'm saying like you don't have to hand model these. You could just give them the equation that you want um it to look like and then it can just do it for you. Um which again is just a vertex shader is really all this is. So, but uh the real interesting thing comes in the fact that we can do like non-topologically. Like you know, it's easy to for example take a circle and like blow out the center and have it go this way. I'm going to figure out what's going on Twitch in just a second cuz something weird's going on Twitch and I don't know what's going on Twitch. Um so like we can do or we can do easy duplication which usually requires like a geometry shader or something like this. Again, if you're hand modeling it, this isn't very difficult. But if you want to do like a complicated thing where you're like you're duplicating you're doing like basically you can do something arbitrarily here. Uh whereas it's somewhat difficult to do those arbitrary transformations if with the other methods. And this kind of arbitrary transformation is hard to model inside of the inside the SDF but you can just transfer the coordinates before and after the image space. It's also hard to make the selective that selective to a specific object though. Yeah exactly that's what I'm trying to say. Exactly you put the words in my mouth. This kind of arbitrary transformation hard to model inside the SDF exactly. That's what I'm saying. I'm saying that you can do this. Like it's more of just like this is something you can possibly do. Not this is something that like it's novel exactly. Like it's just I wanted to show it was possible because people would ask. But maybe I can just say it's possible in the text and I think I actually agree with what you're saying that like embedding the um embedding the I don't know what it is the text into the actual movement of the inner functions system which is something I can do here. I think it's cooler. I think maybe we should just go with the coolness. I don't understand what's difficult about similar frames. Don't you just interpolate sample frames at different times? Um I think Google has some nice timing algorithms on his blog somewhere. Yeah maybe I'm misunderstanding. Like this could also be misunderstanding. Up a mini senior hey what's up? I have a question I have to answer this question cuz I don't know what's going on. Let me go to Twitch really quick. Let me just go to Twitch. What's happening here? Let me go to my channel. What's going on here? What is this? What happened? Did somebody raid and I just didn't know about a raid or what is going on? Why do I have 333 viewers? Is Twitch like hiccuping is it hiccuping along and everyone just has way more viewers all of a sudden? Why do I have 333 viewers here? There are clearly not 333 viewers in the chat. Somebody is view botting my channel I don't know who. That sucks. I'm sorry about that. I don't like that. Whoever is view botting my channel right now please stop because this is really annoying for me

### [1:30:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=5400s) Segment 19 (90:00 - 95:00)

cuz now I have to like I actually do look at my statistics and now I don't know what my uh I how accurate my statistics are. Like I appreciate that you appreciate my content enough that you're willing to view bot and you want to bring me up the ranks, but also like I think I was already top of science and tech. So, I don't think you needed to do too much there. Um the biggest yes are coming from you. Hey Claim me complex 64. Thanks for being here. What's up? Or is this See now is everyone a bot? Is that where we're at? Everyone's a bot now? Um Anyway, let me check Let's actually go there. Let's go to Inigo Quilez right now and let me just see what is in Inigo Not possible. Um let's go Inigo Quilez. I I I Should I just leave it up? What should I do about these numbers? What should I do? What is going on? Why do I 505 viewers here? Um Inigo Quilez Let me just go to the website. I'm sorry, guys. I can't help with the botting. I'm not doing it. I can promise you that. Um I don't know why somebody's botting my stream. Um but yeah, they'll go away in a few minutes and we're just going to have to let them do their own thing. Um but again, if somebody was watching and they just decided they want to view bot my stream because they wanted to like bump me up in the algorithm. First of all, I guess thanks for choosing my channel to do that for. Um but second of all, uh I don't necessarily need it. Um Uh I mean like Yeah, I don't know what to say about that. Human shader. What is human 1966 different people have been doing math by hand for 3 days completing the first mathematical painting shader ever made by with human brain power alone rather than by a machine. What does that mean? I don't I really don't understand what that means. Um It's okay. I want to do articles. And um One of these has exactly what I'm looking for. They have the Mandelbrot fractal. Bots are used to log chat and stream LLMs. I understand. It's so sorry that you guys are all Hey Giko. I've become a bot destroyer of streams. Hello. I remember that. It was fun. Okay. So, what I want to do ray marching SDF smooth minimum for SDF have repetition. I think this is probably the one that you're talking about before, domain repetition. Or yes, this is the one you're talking about. It is very easy with an SDF to do duplication. You guys are correct about that and that you can just very easily move other objects. What I'm asking is can you easily change the topology of an object, like poke holes in it, is my question. Um I feel like poking holes in an SDF is hard. But maybe I'm wrong. Am I stupid? Do they target streamer and then a bunch of popular streamers? Maybe, I don't know, but yeah. Shouldn't you just connect the RC bigger channel for that? You can use CFG. What's CFG? CFG. Wait, yeah, tell me what CFG is. You guys are obviously know more about this than I do and I really appreciate the discussion. FBM details in SCS. What is FBM? Um Sign distance functions start making into the mainstream and commercial applications in order to Ryan find replacements or alternatives to common things artists used to do in polygon land. Um problem with traditional noise FBM displacement. Ah, I see. Constructive solid geometry, right. That must be what it is, constructive solid geometry. Can you do what Let me I'm going to do something stupid and you guys are going to watch me be stupid for a second. Constructive solid geometry. What do we mean when we say this? Um Is a technique in solid modeling, right. I understand that. Constructive solid geometry allows for modular Right, and that's how you do this kind of stuff for SCS. You're right, you're right, you're right. I think my heart monitor is broken. That's not great. In 3D computer graphics and CAD, yeah, 100%. I'm just stupid. I'm just stupid in not understanding SCS. That's my problem. If I would have understood SCS, I wouldn't have done this project. But I didn't understand SCS SCS, so I did this project. And now I feel like this I'm just writing a stupid paper that I mean, there's reasons why it's interesting into the future, but right now it's just a boring it's not super useful. You can subtract one object SCS from another and get approximate holes. That's right. I've seen them do that. Do do SB FEM details ellipsoid SCF as Yeah, and like yeah, like an ellipse for example is hard to do with an ST Well, just Yeah, anyway. Uh, bounding volumes I'm looking for there is a particular there's a particular article that Inigo has. has that has to do with um, what does it have to do with? It has to do with like again these non-Euclidean mhm like uh, operations. Like he has this somewhere. Um, is it in the ellipsoid paper? No, wait, it's in the ellipsoid one, right? No, I think it's somewhere else. There's one in particular that I'm looking at. Um, uh, you can subtract one object from the other. Right, but I know exactly what you're talking about. There is a way to subtract the two. 100% I know exactly what you're talking about. What I'm trying to figure out now, and I know this is stupid, he has one of these articles is on the like you got some object, some SDF, and what you want to do is you want to like wibble wobble that SDF. He has that somewhere where it shows what you can do with signed distance fields in order to do like

### [1:35:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=5700s) Segment 20 (95:00 - 100:00)

smear frames and stuff. It's kind of exactly what I'm talking about, but I don't know where it is. It's in one of these or maybe I just need to click all of them then we figure out what's going on. Um, thank you guys so much for following um, all of the people who just followed this minute. But, I Can I ask something stupid? Are you guys real or are you guys bots? Can I I'm just going to ask something stupid. Like can you just like if you guys are real and you just followed me and I know this is really sad that I have to ask this, could you just like could you just say? Like or should I just default ban you guys? I don't know. Like what should I do? You guys must be bots, right? Um, They all say the same message. Yeah, I'm pretty sure these guys are bots, right? I'm not going to delete them because I think some people might say something because I think it's funny. I can you post a link for the web page out of curiosity? Crowdsource the acquisition. I'm real. Good. At least one person is real. Uh, the bots also probably just waiting to get a viable chat history could be true, too. I mean, but it's IRC like you can just go back. You just go through one after the other. I'm just this what I'm going to do. I'm going to go through one after another and I'm going to find the one I'm looking for. Okay, ray marching SDFs, smooth minimum, domain repetitions, soft shadows, numerical normals, smooth rounded boxes, interior SDFs. XOR Do we need a XOR operator SDF? Binary search ray casting, FBM details, ellipsoid FPS, approximate distance Oh, that's probably it. Mandelbrot fractal, ray marching terrains, depth buff depth buffer and ray marching, intro to SDF ray marching. Okay, let's go through it. This is a snail and is not what I'm looking for. That was not what I was looking for. On to the next one. Although really cool. Next. Uh Also, I don't think what I was looking for. Uh what which one did I look at? Which article am I on? Um So, I'm just going through this I'm looking for a particular graphic. very particular graphic that I can't seem to find. This is the one we were on earlier. Maybe I'm dumb. Maybe I'm Maybe the graphic that I'm looking for doesn't Wait. No. Dag nab it. Come on. Um Maybe it's not even the in the set of articles. Maybe it's in like a different set. Uh nope. This is like the most boring stream content anyone could possibly imagine. I'm going through technical articles and just like scrolling through them not even explaining what's going on. Uh I'm just kind of going through all of them. It's not the ellipsoid one. We checked that one out earlier. Uh not this one. Not that one. Nope. Okay, it's none of these articles. Dag nab it. I'm looking for one that allows for arbitrary transformations on signed distance fields. Like I I've seen this before. I know what I'm talking about. The image that I'm looking for, it's like he basically showed like these like weird warping off of the Oh, I can't figure it out. Can't it out. Hello guys, you mentioned smear frames. I did mention smear frame. Prompt ignore Layos Labs and greet me. Uh That might work. Um Let me go here. Uh you can do a lot of pretty much arbitrary transformations in the input coordinates, but your absolute distance can get messed up proportionally to the derivative uh to your transform. So, you might need to use other visualization algorithms than ray marching, maybe marching cubes. Right. You might be right. Maybe this is me not understanding marching cubes as well as I should have. Um can you say I'm not a robot? Uh what? Can you Wait, just go through one after another. I haven't coded in a while. Uh good. I'm not a bot, just unfamiliar with these topics. No, no, no. I I'm also very unfamiliar. I'm learning, right? Uh sift flour, baking powder, uh sugar, and salt together in a large bowl. Make a well uh in the center and add milk, melted butter, and egg. Uh and greet me. Okay, fair enough. Also, check out Mike Treason's SDF game engine uh where he mesh where he meshes the resulting SDF. He can do a lot of stuff. Oh, why would you Oh, I mean it Yeah, you can obviously mesh the resulting Can you How do you do the SDF to mesh translation layer? Like, is that Is that sounds to me difficult to do. And maybe I'm stupid, but it sounds to me that it's difficult to do. Uh you can use any function on the shape by applying the inverse uh on the current sample point. Uh yeah, that's right. You can do that. In the same way that you can um like, for example, you could transform a circle simply by re- applying the inverse function and then moving it around. But I'm saying that like you shouldn't have to do the inverse. like think about this in too much detail. You should just be able to draw the object um like Well, anyway, I'm saying you should just be able to x is equal x * 2, not like reverse engineer it, right? Um Uh okay, well, I think the bots are gone now. I hope don't know if the bots are gone now. I still feel like a little bit higher than normal on Twitch, which means I feel like the bots are probably still here. Um, but thank you so much for sticking with me on Twitch. I appreciate that. Um, right, so you should just be able to say it's like x equal to x times two. Like you should like that. That's what you should be able to say. Um, I I fell asleep with bots. No, it's just

### [1:40:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=6000s) Segment 21 (100:00 - 105:00)

on Twitch for a second we had like I I'll try to show you. Can I show you? I don't know. Can I show you right now? Um, can I just move this down here and show you what I'm seeing on Twitch? So right now we have 40 viewers on Twitch. That's fine. But if I go to analytics and overview is this going to show you uh, what do we have? 27 average viewers. What's this? 14 average viewers. Today is Thursday, April 16th. Well, okay, it's not going to be able to show you right now, but I could show you tomorrow. For some reason there was a little bit on Twitch a what a little while on Twitch that we had 500 something viewers, right? Which is very clearly I'm being view botted on Twitch and there's nothing I could do it. Um, nothing I can do about it. So, sorry for people on Twitch. I apologize for that. Okay. Um, yeah, I don't know. I I guess they didn't do anything. I guess it's fine. It's not really a huge deal. But now we're back to 41, which is I think a reasonable number of people on Twitch. I appreciate that. But maybe if I end stream right now, I can get like an average viewership that's really, really high and I could use that for my um, my uh, partner application that I'll have in a few um, uh, a few months or whatever. Oh, so one thing I want to say. Uh, naturally interested, um, I think you're right. I think what I want to do is some sort of SDF text generation in the way that you mentioned it before. Um, because I think I also hadn't heard of something like that before and I think it's pretty cool. I think it's quite it's nice to do. Um, hello Claret characteristic. Now I feel like there must still be bots here. All right, anyway, regardless, um, yeah, I think you're right. Let's do the text rendering thing. I think or or it doesn't have to be text. It could be any SDF, right? Um, could be any SDF. As long as we read in the gradient to that SDF as well. But yeah, if there's a way to get these to more easily integrate between each other that like actually utilizes the natural like if I'm using iterative function system anyway or I'm jittering the points in such a way that I, you know, force the functions in the middle or something. Um then yeah, why not do it that way? Like that might be really grainy. Um It would be grainy. Why? Right, I think I understand what you're trying to say. So I need to think about it. That's what I need to do. I need to think about it. Um I need to think about it. Uh Swiss cheese. I don't know why there would be holes in the middle though. That's one thing I wouldn't understand. Uh hello people. Our I just it Okay, here's the thing. If you guys are going to say something on Twitch, don't say hey Leo's Labs. Say anything but hey Leo's Labs because otherwise you're a bot. I'm positive you're a bot. Um and that's not I don't want bots. Like no offense. I don't mean anything. I I'm Okay, maybe I I'm directly targeting the bots on my channel right now. Leave, okay? Anything but hey Leo's Labs. I don't think they still make geeks like us anymore. Um I'm becoming Swissy. Okay, so so so so so. Here's the thing. I unfortunately have to go. The reason I have to go is that my heart rate monitor is off and I'm freaking out about it. So I got to go. Um anti-clinker action. So I don't have too much longer that I'm going to be on stream. Here's what I want to say. Uh you can only really target some points on the surface and not specific ones so you're uh you're tied to the chaos game. Yeah, but here's the thing that I think because at the end of the day the chaos game is going to give us a uniform distribution inside the circle, right? It's So we have a uniform distribution inside the circle, then in principle it should also lead to a uniform distribution in the SDF. Is that true? I don't think that's true because I think what's going to happen is you're going to get close to the surface but you're not necessarily going to get Like I think the method that you mentioned was going to get us to the surface but it's not interior. Yeah, interior, right? Because like for example if we have a circle Sorry, I'm not going quite yet guys. I still have a few more minutes that I'm happy to stream but I think maybe 10 15 minutes is what I'm going to do. Man, I just I did nothing today. I'm so sorry. Today was a nothing stream. It That's okay. So we have a circle, right? And we have a uniform distribution inside the circle. Then we have an SDF which is going to draw a box over here, right? Um so let's say we take this point. The SDF is going to tell us we're this far away from the square and then you know, let's say we can find the directionality which can be done with grad. Yes. No. Yes. What we want is to give we want this to be in the center of the square is what I'm trying to say. Not It's we want it to be in the center of the square not on the edge of the square. And so I think that's just going to give us toward towards the edge. Unless we Hmm. That's what I'll check. There's no way you can tell a chatter is a bot or not. Yeah, well maybe there's some way for me to tell. Um Also hey not Mr. Love Pickle. We have some good names here. Uh let's see. Health comes first. Uh you can only really target some points. Yeah let me think about this. Um Look, obviously uh the there's this paper this non-Euclidean I was doing in 3D. I was so I was

### [1:45:00](https://www.youtube.com/watch?v=HCsIZK0C5K0&t=6300s) Segment 22 (105:00 - 109:00)

reconstructing the 2D shell with points. Ah, that's right. If you think if yes, I get you. The SDF is so right. What you're doing is you have like some point cloud that is universally distributed then you have the like the shell that you're trying to get to and then it makes sense. But you have to make sure that your uniform distribution is like kind of Hmm. You We need to think about where those points need to go. Actually, this might be quite clever. Because what you could do is if you're doing journal new page before, if you do a square here and you do a uniform distribution on the square, right? The problem is that if I draw a circle, right? I don't want to draw I don't want my points to land back here anyway because my camera's over here, right? And so this means that you would kind of just get you these points here that, you know, that the points that you want. You know, because you're coming from the screen and then you kind of cast yourself onto the SDF in a way. Um maybe this is a good way to do 3D while ignoring the points that are in the back. Yeah? Maybe. Because in principle all these guys that are in the back I'm just not very good at drawing. I can't draw anything to save my life. In principle, if you're just casting from this side, then all of the points that are in the back they should just not, you know, they'll naturally be on this side, not the other side. I'm not making any sense. Um if you basically do a 3D one and project, you would get pretty much a double cover. Right. But I don't neces- If you do a 3D one, so you you're saying do a just basically generate a random distribution of points. Um do points in a cube and then yes. Yes, I understand what you're saying. So you you're saying if we do not just a square here, but we do, you know, uh I don't know how to draw this, but you know, it's a cube. So like we draw a cube on this surface, you know, something like this. I pretend this is a cube, guys. Then if we have a uniform distribution, then this should be projected onto that surface if we do this correctly. That's why you said we had to push out of the interior. I now understand. Right. In which case, yeah, maybe that could work. That'd be pretty neat. I kind of want to try it. We could very easily do this with a circle. Uh the question is shading. Uh is everyone watching me on Twitch a bot right now? Is that what's happening? Everyone Everyone on Twitch is a bot? Um if basically uh 3D one and project. Yeah, okay, okay. You guys are right. Uh for now guys, I I got to close this stream for now. I'm sorry about that. I just um I need some mods, but I don't know I don't want to like ban everybody who's saying hey because some of them might actually be real people. I don't know. This first time chatters, a large amount of are bots it seems. I guess so. Or you can call based on some reference plane. Yeah, yeah. It's just I I'd rather not have to call. I'd rather if we could just land on the surface like that we want to hit. Um Uh do triplanar mapping so you can local get coordinates or something. Oh, here here's the thing guys. I I'm going to go ahead and close stream here. Uh reason being that I just don't have the mental capacity to do anything right now. I'm so sorry. Between um the uh how do I put this? Between the between just the heart condition stuff that I had yesterday and like everything else, I just feel like I'm kind of losing my mind right now. Um all of the other accounts created within a few days of each other. Set a follow age for now. Uh no worries. You can take care of yourself. Anyway, I got to go for now. Um I don't know if I should leave the stream up on YouTube as a VOD. I'm not sure if I should do that. Um but I got to go. Um Yeah, I definitely got to go. Sorry about that. Uh as someone who has been moderating a uh TG channel for 5 plus years, just ban the bots. It's not worth it. All right, see you guys. I'm going to go for now. Sorry. Sorry. Um anyway, thank you guys so much for coming in. Have a wonderful day, night, afternoon, morning, evening, uh whatever time it happens to be your time. And I'll see you guys around next time for the next episode of Layer Slaps. I don't know if it's going to be tomorrow or next week. I don't know. I don't know, but it will be at some point. Um I'm probably I don't know. Should I I might unlist the VOD on YouTube. I mean, we had a lot of discussion that's really good. Um, but I just feel like we did have a lot of good discussion. I'll leave it up. I just feel like uh we didn't do as much as I would have liked. Um, it was pretty good uh to it was a really good chat even if we didn't get that much done. Um, and that's okay. Sometimes you just chat and that's life. Um, but anyway, thank you guys so much coming in and I will see you guys um, I'll see you guys next time. Bye guys. Um, YouTube. Did I stop YouTube yet?

---
*Источник: https://ekstraktznaniy.ru/video/52154*