WEBVTT 0:00:15.707 --> 0:00:22.913 All right, maybe before we jump 0:00:26.254 --> 0:00:33.460 Straight into business, Patrick? How are you doing now that you got your audio issues sorted out? 0:00:36.008 --> 0:00:36.919 I'm doing okay. Yeah. 0:00:39.079 --> 0:00:50.580 Have you thought about, reflected on this year when it comes to programming work? You know, stuff you've been studying? What's your overall feel of 2023? 0:00:52.292 --> 0:01:07.024 I mean, I feel like 2023 was the year everyone decided AI was going to happen, but I'm a little depressed. I'm a little depressed to find out the new Gemini demo videos were all fake. So that makes me sad in my 0:01:07.581 --> 0:01:18.179 heart. Yeah, well they got in trouble last time because they were not fake when they released Bard, and it was really embarrassing. So this time they totally faked it. You really can't win if you're Google. 0:01:21.098 --> 0:01:48.840 But yeah, no. I think it's been a good year. It's been a good year. I feel like we're stabilizing after the whole, you know, 2020 stuff. Just say that we'll leave it there. Um. Yeah. So yeah, so I think it's good, but I'm curious what's going to happen with the job market? You know, it's everyone's a little nervous. So hopefully things are up. Crypto's ripping again. So I'm starting to hear people get excited, which is exciting to see people excited about something. So AI and Crypto being exciting again could be cool. 0:01:48.840 --> 0:02:00.653 So yeah, I mean this time I actually have crypto for the first time. So you know, we've missed every other crypto wave. Maybe we'll actually catch this one if it goes to the 0:02:02.324 --> 0:02:20.481 moon. I saw that there are upcoming two SpaceX launches scheduled for this evening as we're recording this podcast, which is December 11th. So I think that's pretty cool from like two launch pads very near to each other in Cape Canaveral. So yeah, a space space is becoming big again. That's cool. 0:02:20.869 --> 0:02:25.088 Cool. So now where you live, do you see the SpaceX launches like 0:02:25.290 --> 0:02:27.771 In the sky? Go look outside. Yes. Okay. 0:02:28.277 --> 0:02:29.205 All right. Very 0:02:30.555 --> 0:02:34.909 Cool, and it's not too cloudy. I'm far enough away where yeah, it's often obscured. 0:02:35.432 --> 0:02:36.720 By clouds? Okay, got it. 0:02:38.824 --> 0:03:32.339 The thing about the job market is really interesting. I really don't know what's going to happen there. I mean, one thing that I guess really scared me about this year was just a huge RTO push, and did you feel at any point? So, just like a bit of context: Patrick and I both moved out of the Bay Area, you know, since sometime around, you know, sometime during the catastrophe. And, you know, yeah. So hearing stories on LinkedIn of folks saying, 'Oh, you know, I have to pack up and move to Seattle or something like that,' that was definitely a scary one of the scariest things of the year, but it seems like it's kind of stabilized at this point. You know, there's two different camps; neither one seems to really be budging. 0:03:34.292 --> 0:05:26.410 Yeah, I don't know. I listened to actually some economists talk about it, which was different than hearing software engineers talk about it because software engineers are obviously colored by personal vested interest, let's say, because it's highly personal. So, I heard some economists talk about it, and they were sort of saying that if you look at the rate that it was before and the rate it's going to be now and like where it's settling out, and sort of like barring some sort of drastic change, that it is just going to be a much much larger portion of people spend more time working from home. Whether that's fully from home or part of the week at home, they're like the sort of cats out of the bag—everyone already figured out that it is okay and it does work. So there will probably be, you know, it's a gradient or probably a hybrid work for the foreseeable future, but they were kind of indicating that they believe it's here to stay. If nothing else, it's competitive advantage for companies that don't do a fully in-office thing to recruit sort of like key talent. And so, hearing someone outside the field say that and they were also talking about sort of implications for like just how much gas is consumed by people, right? Or not office space. And we forget, you know, it's the classic traffic waves thing. If you've ever seen that, it actually doesn't take a lot of change to sometimes swing wildly supply chains or bandwidth or, you know, insert your favorite example here. And so, sometimes actually only a 10 decrease in the demand for like commercial real estate, as an example, leads to mass vacancies just because you'll find new and creative ways or do something different or go with the barely manageable office. And so, I think we've not fully seen the impacts of those things yet because those leases are often long. So, we'll see. Yeah, you know. 0:05:27.360 --> 0:07:01.100 I'm usually a little bit bearish on VR. I mean, I have an Oculus. I actually used it yesterday. I use it, you know, several times a week for exercise—do boxing and other things on the Oculus—and so I'm using it regularly. But, you know, I never really saw the sort of commercial appeal, or it hasn't been that in my opinion. There hasn't really been that killer app. But I wonder if this is really, you know, that the opportunity it's going to reveal itself, right? Because what you'll hear folks say is, 'You miss out on the water cooler chat,' or 'You can't go on the whiteboard and whiteboard with other people.' But, you know, you're not doing that all day. I mean, there were definitely days when I was going in the office where, you know, even when everyone was in an office, they weren't in my office. You know, there was a ton of a whole urban sprawl of buildings in our campus, and then there was still Seattle and other places. And so there were definitely days where I basically was in a conference room almost the entire day, just talking to people from all over. And so those days, you know, you people wouldn't even know if you were at home or not. And then the actual days where you want to get around a whiteboard, you know, if you were to schedule those, they're not really—they don't really need to be very spontaneous, even though they were—but you could schedule those, and then everyone could put on their VR headset and sit around a VR whiteboard and brainstorm. You know, they could do that once a week. 0:07:01.100 --> 0:08:48.674 Yeah, I actually VR is a good one. I think we haven't—we're talking about predictions. I think I think 2023 VR wasn't a thing for me yet, but I have hopes for 2024. There are some key things: the new Oculus coming late in the year. And okay, I guess I'll save it for 2024 predictions. But yeah, that's a great point. What the I guess the thing I hear about remote work is that I think resonates is depending on the kind of personality you have, depending on where you are in your career, and depending on like the nature of your tasking makes a very large impact. And the thing is those things are not constant over your career, and they're not constant even throughout the year. And so, I think that people say this, you know, 'Oh, it's very difficult to innovate in far-flung places.' But I don't know. I think it requires, like you said, careful planning, and I'm not clear that everyone thrived under spontaneous interruption. Let's sit down and talk about this. And we have this come up on my team recently. We were trying to—let's call it innovate, you know, not just turn the crank on some stuff. We were trying to like figure out some new stuff, and we made a point to say like, yeah, we're going to do this on Thursday. I think it was on Monday, and then, you know, Tuesday and Wednesday when we briefly met together, we sort of added things for people to kind of think about and sort of like not go prepare slides for, but just when we came together on Thursday or whatever be ready to sort of think these are the sets of things. And it was actually really interesting because rather than making it halfway through the brainstorming session and someone going, 'Oh, what about this?' We kind of had a notional like set of things we needed to make sure we handled and discussed. Now, wasn't necessarily completely effective, but at least you know I found it—yes, it took a little bit more thought, but it was okay. 0:08:50.969 --> 0:08:57.415 Yep, yep, that makes sense. Yeah, during the pandemic, we also had some different brainstorming sessions, and... 0:08:59.052 --> 0:09:37.600 We used some software—forgot what it was called—but it was really quite good. Man, I wish I could remember what it's called, but it had mapping. Yeah, that's right. Yeah, it had a bunch of these different kind of icebreakers and different corporate kind of tools, different mind-mapping tools all in one. And you kind of share with other people; it's like a Google Docs kind of environment where we're all working on interactively. And so yeah, I think there'll be just more of that as companies realize that there's really no going back, as you said. 0:09:37.600 --> 0:09:47.635 Did you get on the portable gaming PC rage this year? I feel like—I feel like that was another thing this year. Did you get any? 0:09:47.635 --> 0:09:59.549 I was so tempted, but I didn't. Yeah, I am gonna get a Steam Deck, I think, but I'm going to wait a little bit. Yeah, I just can't pull the trigger yet, but it's been looking really nice. I will. 0:10:00.072 --> 0:10:27.139 Say for the couple of long-haul flights I took this year, if you're willing to pack the Steam Deck, oh, it's so nice. It's just like because I don't get a ton of time to just sit, and then I had three, four, five, six hours to just sit and play video games, you know, as long as if you know you didn't kill your battery or you had a charging strategy. But that was just such a game changer for me. Yes, first-world problems that you have to sit for six hours with nothing to do, but it was really. 0:10:28.405 --> 0:10:36.184 Awesome. Yeah, that's great. Um yeah, I think the Steam Deck looks super nice. I definitely want to get my hands on one. 0:10:37.247 --> 0:11:14.963 But there's other alternatives too now coming out depending on like what side of. Yeah, I saw that the ROG Ally or something like that. Yeah, and there's some sort of like other second-tier companies making them too. There's a bunch, but yeah, the Steam Deck obviously, I think has like—most people have, well, I shouldn't say many people have a Steam library that's just like a huge backlog, and there's like integration with that. I don't know, it's like the ease of it, right? Just sitting down and playing it, and the user base is there. And for games that have bad key mapping, you know, there's always like a user-submitted key mapping ranked by popularity, and you can just switch over to someone else's. And yeah, that's very nice. Yeah, actually speaking of 0:11:14.963 --> 0:11:23.687 games the uh did you see this fortnight thing where you you now have lego fortnight there's all these other games inside of fortnight that that came out i 0:11:24.025 --> 0:12:14.785 Will I played for once. I'm on the ball this time. I saw it and I was like, oh, this is really cool. So I have kids that are, you know, I guess upper elementary, early middle school. I told them, and they're like, 'No, that's dumb.' They're like we just want to play real Fortnite. And I was like, no, no, no, let's play like this is gonna be great. So we started up, and my daughter and I did play a bunch of the LEGO Fortnite. I will say for being just a blatant ripoff of Minecraft, but like in Fortnite world, I really enjoyed it. It was like a nice fresh take on, you know, not exactly the voxel, not exactly Minecraft, but in my opinion—maybe I'm biased—that's clearly where they're trying to go. Minecraft is such a juggernaut, and I think that, you know, they want to piece that action, but I really enjoyed it. The game was—it had that cycle where you just always like ready to go play, do the next thing, like get one more little step up. They had it very well polished. 0:12:15.679 --> 0:12:30.985 Yeah, so you know I had the opposite situation with respect to kids where my older son was the one who told me about it. He's like, 'Dad, we should try this. Kids at school are playing it.' And so, and so we tried it, and yeah, I agree. I think it was pretty good. I like how, you know, 0:12:32.757 --> 0:13:21.440 I guess for me the days of being creative are just kind of over on video games, you know? I just—I don't have it in me to like spend a bunch of time to meticulously create my base and everything. And so, you know, I like how they had the prefabs where you could just say, 'I want a house,' and then it would just show you all the different pieces, and you just had to collect all the resources. So, you know, my son built his own kind of custom house. My four-year-old through tantrums until we put him in his own world where there were no enemies. So he wanted to play too. He's like, 'Oh, I want to play.' You know, everyone else is playing. And so we're like, yeah, come on into. So I made him an account and everything. He joined our server, and then he was like, you know, bitten to death by a spider or something, and threw a tantrum. 0:13:23.837 --> 0:14:59.780 Uh yeah, I mean another game that I got on just recently that I've been enjoying is—the pickup put down is Dave the Diver. Have you heard about Dave the Diver? Oh man, I've heard about that. I heard it was really good. Okay, it's such a goofy like mechanic, but it's just sort of like a job simulator kind of game, but you have like one—you rotate through various, you know, sort of like parts of your day in the life of a slightly overweight diver by day and working at a sushi bar by night. So during the day, you need to dive into the ocean and capture fish, and then at night, you work with your sushi chef to make the proper dishes from the food that you caught and serve patrons that come in. And so it's basically a series of mini-games, right? So diving is its own mini-game with a set of trade-offs, and then the sushi at night—there's you need to bus, you know, the sushi chef makes the dish; you got to take it to the person who wants the dish, and then you got to like clean the table, grind up the wasabi, serve tea. You know, there's like various little things you need to do. So yeah, but it has that really nice like repetition rate where you don't need to sit down and play for, you know, eight hours or two hours or one hour. You know, you can play for 10 minutes, 15 minutes and do one dive or do one night at the sushi bar, and it's not super pressing that like you're gonna be hungry if you don't go get apples or you're gonna—like as far as I can tell, there's very low repercussions for screwing up, which allows you to sort of like experiment and try and do different things. So I've really been enjoying Dave the Diver on my Steam Deck. 0:14:59.780 --> 0:16:49.425 Actually, yeah, actually related to that, I got a game called Cuba Factorium, which is like a—yeah, it's a weird name, but it's basically like a watered-down or simplified Factorio. Okay? Um but there's a couple of mechanics they did really well. Number one, similar to what you said, there's no pressure. Like the people—it's not it's more like a RimWorld or a Dwarf Fortress actually than a Factorio. So you have a, you know, a group of people and you kind of give them tasks by like drawing blueprints of buildings and all of that. Um but there's there's no time pressure. So there's the people don't starve or anything like that. You could just let it idle, and um—it's like playing Factorio with no monsters. You know, if you idle, you'll slowly accumulate some resources, but you know, with you working on the efficiencies, that's the way to push the game forward. Um the other thing they did really well is, you know, with Factorio or any of these games, the first time you play it, you don't really know what you're doing, and so you make a bunch of bad design decisions, and then you feel really compelled to start over, right? Um what this game does is—is you basically hop from island to island, and so they did a really good job with the pacing. Like when I unlocked, uh, when I unlocked conveyor belts, and like I'd up until then been doing the whole thing which is people carrying goods back and forth, of course, like it massively changes the way you want to design the base. And so that's just the time where they're like, 'Okay, you know, you unlock conveyor belts, you won this level, go to the next level,' and start from scratch, but you have all the research from the first level and a bunch of resources and all. 0:16:50.438 --> 0:17:58.174 That oh, that's awesome. I saw for Dyson Sphere Program which I feel had a pretty good cut at that mechanic as well. Uh adding sort of enemies and you know you have to kind of start your world over, but I also learned that apparently—which I don't know, I guess it goes into the economics of game design—but they have a thing where when you're building up your world, they do a sort of weighted average but recording of your peak science output in a given save. So if you're making—I don't even remember the blue cubes at a certain rate, not the total you produce, you can't just let the game sit there and run, but like how many cubes per second you reach at your peak gives you these like metadata points, and then you can use those points in the future as like expenditure, almost like a roguelike to sort of skip ahead in your research in your new setting or they're going to add—I guess permadeath or death, and so you can spend these metadata credits to like restore it. So it's this really interesting dynamic where you should continue building up your worlds to like have this really massive focus on certain science production because then later you'll be able to use those as like a sort of shortcut for future gaming. So I thought that was really interesting. 0:17:59.372 --> 0:18:35.920 Very cool. You'll have to check that out. So okay, let's jump into questions, but before that maybe we'll—we'll do the typical like a recap and prediction. So we did talk about 2023, my predictions for 2024. Um and we called out one of them which is I do think there'll be some type of VR plus work. You know, I just think the audience for that is massive, and there's a huge schism right now that that could bridge over. Um so, yeah, I do think that that's something that will come around. 0:18:37.999 --> 0:19:11.108 You know, I feel like the cost of doing all this AI work is going to become a bigger and bigger problem. You know, I think a lot of these—a lot of these ML training machine learning training jobs are very inefficient because they're very complex. You have to, you know, do all this data transformation, and then you have to train the model, and the data transformation is embarrassingly parallel, but the training the model is not. 0:19:12.947 --> 0:19:58.840 Um, and so the whole thing is extremely inefficient. And, you know, I don't know if it's really—it's gotten a little bit better, but it hasn't gotten better. It hasn't kept up with the price of GPUs per hour and all that, which is skyrocketing. So I would say next year I definitely expect to see more around lowering the cost of doing all this. There's also some environmental aspects. Um you said I think it's like every Bitcoin transaction uses a gallon of water or something. I don't know, there's something wild like that, but um I think efficiency is going to be more and more important. And uh I guess maybe at one final prediction. 0:20:00.872 --> 0:20:22.034 Um, I do think that that—I think that this RISC-V is going to is going to really take off, and we're you know we're going to see like a whole bunch of cheaper hardware come from that. Can you tell me what that is? Oh yeah, so or is it V or five? I—I've 0:20:22.742 --> 0:20:24.580 always said V, but I don't know if that's correct. 0:20:24.580 --> 0:21:23.155 Okay, so it's basically a simpler architecture. And Patrick, you probably know a lot more than I do, but you know you have x86 and you have ARM, you know different processors. So for example, if you build some software on your PC that's probably using x86, um, and so you can't just run that on your phone. So like, you know, different processors need different, you know, uh compilers and all of that. And so this RISC-V or Risc V is a completely open source specification. So, so you know anyone can go and make a processor, and um, and you code that you compiled on one RISC-V machine or cross-compiled to a RISC-V architecture will run on another one. And so, you know, it kind of blows up the duopoly on architectures. Did I get that right? You probably know way more than I do about this. Yeah. 0:21:23.492 --> 0:23:23.360 yeah no no i think you you have it right i guess like to put terms on it so risk v is a isa an instruction set architecture so you're right the two big ones are arm and x86 but x86 is sort of notoriously like very complicated uh and really implemented by intel and amd are the only two people who really make x86 chips x86 being that that instruction set architecture the byte code that you know maps to specific operations uh and so then there's arm and many people license arm to uh have their own implementations though uh so you arm different than x86 has a lot of or at least in my opinion has like a lot of uh modularity to it so you can add you know extra things like if you're building a digital signal processor a dsp you may add a lot of multiply accumulates which are useful for like ir fir filters um versus you know if you're not you don't include that one so it is more you know modular but you pay licensing fees to arm for you know basically the instructions on like how to do all that like what you know buying existing blocks to drop into an fpga or to uh your design for an asic and so risk v is supposed to be yeah more collaborative and unencumbered by the licensing you need to do and so lots of people hopefully will build these modules out and then sort of manufacturers who are trying to compete very very very cheaply uh won't won't have to sort of invent their own instruction set which like you said allows for better compiler optimization and targeting it and some portability in in many cases and so you know you get sort of economies of scale you wouldn't otherwise but you know a lot of these uh chip manufacturers are trying to start up and and kind of gain momentum uh by being in this ecosystem so for things like a raspberry pi or an arduino or whatever they could use this and not have to pay uh you know licensing fees to which would be part even if you buy a chip off the shelf that already 0:23:23.360 --> 0:23:44.972 paid like part of the price you pay for it would be the licensing fee to arm in a lot of cases if it's arm based which is very common for embedded chips and so yeah so i interesting prediction that that'll take off it's been a slow burn growth and we have started to see some uh some more production outputs from folks for for those chips so that's a that's a good prediction yeah 0:23:46.052 --> 0:23:47.100 What are your predictions? 0:23:47.100 --> 0:24:06.083 So I well, I was gonna press you on yours first. Hang on. Oh, sure, I was gonna make you get a little bit more, a little bit more bold here. So you said VR for work that was one of your predictions. So I was gonna push you on say what do you think that's going to look like? That was that was a little hand wavy. So, so okay, I'm like gotta hold you to it. 0:24:06.539 --> 0:24:16.208 Next year, I think that the right play here is is, you know, the idea of you wearing a VR headset all day. I just don't see that happening. 0:24:16.208 --> 0:24:20.258 Monitors, not like having nine monitors spread out like a, you know, stock trader. 0:24:21.102 --> 0:25:17.734 Yeah, yeah. And you know even like if you forget about the weight and all the encumbrance of wearing the headset and everything, um, you know people kind of like I think people are underestimating just like the proprioception part of it, like just the fact that like you know you look in the real world and you touch things. Like you put your hand on your desk, and your desk is real and it's right there. You know, I think um I mean maybe there's all sorts of tricks they could do with, you know, like Neural Radiance Fields and all these things like actually recreate your desk and have it be sort of like so accurate that when you put your hand on it in VR, you feel the same way. But um but yeah, I'm a little bit, you know, skeptical you're just gonna wear it all day. I think what you'll what you'll want is, um, you know everyone on the team gets a headset. It's it's, you know, it's probably cheaper than monitors at this point. Right? It's only like 300 bucks. Like some monitors are three or four hundred dollars. So 0:25:17.734 --> 0:25:22.746 On which headset you thought? But okay, so you're oh, like you're talking about Oculus headset? Yeah? 0:25:23.050 --> 0:25:35.031 I mean it seems to me like Oculus has just dominated. I to be honest, I haven't even heard of any of the other headsets in a long time. I know there's the HTC Vive, there's one from Steam, yep. 0:25:35.790 --> 0:25:44.970 So I was gonna say next year there's rumors that there will be a new Valve Index. There's the Apple Vision Pro rumored to come or oh, that's right. 0:25:44.970 --> 0:25:46.185 So I didn't. 0:25:46.304 --> 0:25:48.885 I know where you were going with this, but okay, you. 0:25:48.902 --> 0:25:52.885 Know I've—I didn't even think about the Apple one. That one, it's not out yet. 0:25:53.324 --> 0:25:54.218 Right. No, next. 0:25:54.454 --> 0:26:52.977 Year. Okay. Maybe yeah, but so I don't know as much about the hardware, but I think that there'll be some type of app and maybe it'll be, you know, heterogeneous where some people could be on Apple, some people could be on Oculus, but it'll be a way where, you know, when you want to do something creative—you maybe want to record a song or go on the whiteboard or something like that. You and your co-workers can do this on VR, and then it's sort of like all integrated back into the workplace where, you know, all your notes are in Slack, everything everybody said was like, you know, there's dictation, and so you get this like this report afterwards along with the whiteboard. And it's animated, so you could kind of see how the whiteboard evolved over time. So yeah, I think there's going to be some kind of cool VR work meeting. 0:26:54.968 --> 0:27:13.160 Okay. And then you talked about lowering AI training costs, improved efficiency—so not just adding more and more parameters to your models but focusing on getting more bang for your buck and being able to train quicker. Is that the call there? 0:27:13.160 --> 0:27:49.744 Yeah, that's right. Yep. I think the cost is getting totally out of control. The models keep getting bigger, although that's starting to plateau, but you know now you're seeing these ensemble models like Mistral put out one where they created an ensemble of eight different other large language models. And so it's like a mixture of experts type approach. So you have—it's faster than running each of those eight models because you get some synergy there, but. 0:27:51.364 --> 0:28:04.780 Yeah, I think the inference, you know, folks try to optimize the inference. It's really difficult because the inference—you know so many people have spent so much time on that. So, you know once the model is trained, you. 0:28:06.872 --> 0:29:16.740 Know go ahead and going ahead and evaluating it. You know there's a bunch of folks working on that and they've made a lot of really good progress, but I think the actual training part of it is something that has lagged behind. And there are talks of AMD kind of upping their software game and creating a real rival for Nvidia. I mean, that could drive some of the cost, some of the price down, but still I think that most people I talk to they're wasting the GPU hours. So you know you get these GPUs from AWS or Amazon Web Services or Google Cloud Platform, and they charge you per hour. And you know if you turn on the GPU machine and you don't use it, you're still paying per hour for the whole thing. And in many of these cases, you know the program is designed so that loading the data is so slow that you're only using like five percent of the GPU, but you're paying for all of it. So yeah, this is something that we just—we're gonna have to fix as an industry. 0:29:16.740 --> 0:29:24.025 All right. And then you had RISC-V, but I don't know how it took it to make a bold claim there. 0:29:24.464 --> 0:29:50.535 Well, you know I've been trying to get my hands on basically like a Raspberry Pi equivalent but in RISC-V, and they're still really expensive. So I just think that with all the geopolitics stuff around, you know, the intellectual property and everything, I just feel like something's going to open up there, but I have to confess I know very little about that side of it. It just feels like there's a lot of energy. 0:29:52.189 --> 0:29:58.340 there we have a benefit of live chat we're getting feedback that it's risk five and wikipedia agrees it's risk five so 0:30:02.027 --> 0:30:04.170 Okay, wait. Where are you seeing the live? 0:30:04.964 --> 0:30:11.039 Chat? Oh, in the general room. Uh, so okay, we'll use AI to go back and auto-edit all of this, so we're not 0:30:13.739 --> 0:32:08.040 Yeah, all right. So, I have—I guess like first, I'll make some AI ones. I think next year we're going to see someone claim to have gotten AGI, and it's going to unleash like a little bit of a debate about what does that actually mean? Like what is... what is? I think some people feel that means like sentience, like oh, it has some ongoing consciousness and like continuity to itself. Where you know right now we have like a look back where—and you can correct me if I'm wrong; I might be misunderstanding—but today, like you talked to ChatGPT, it can only remember a certain number of tokens before basically, you know, tokens start falling off the back. So AGI at some level will be able to recall indefinitely and, you know, sort of shape itself going forward over time and sort of build on it. But I—I don't know. I think someone's going to claim that it just has to be expert enough across enough domains, and so it's oh, it's creative and it's doing this, you know, we've reached AGI. So someone's going to claim it because I feel there's a lot of political zeitgeist. And yeah, you know, just saying it, and we're gonna have the debate. We saw Google put out some guidelines on what they think qualifies as AGI or not, as a—as a, you know, trying to get in and shape and define it. But I think that'll be sort of one thing. I think another thing we'll see is a lot of people focus on this like code writing, code suggestion. I'm actually—I don't know. Whatever. Maybe I'm short-sighted. I think it will come eventually. I actually think what will be more impactful and I'll make a bet that someone will come out with this next year is you see a lot in summarization. And so I think having something like when you do a, you know, a GitHub, you know, pull request or whatever, that having something that comes in and does, you know, basically like an expert-level suggester peer reviewer. So we always talk on this, you know, program or on this podcast—a difficult responsibility of everyone to, you know, doing a not super high latency peer review. But what if your 0:32:08.040 --> 0:33:30.636 first peer review was sort of just like from an AI suggesting, you know, oh, that's genius? Not not just in your review but actually across your entire code base. I've looked at it, done some sort of summarization and said, you know, hey, it looks like you're duplicating a lot of code from this other module over here. Have you, you know, looked at just using that one instead or, you know, making it more extensible or flex? I think that'd be super cool. I think someone's gonna—someone's working on it; they're just not talking about it yet. Oh, to be clear, that's AI. I don't actually know that. No, I think you're right. I mean, it's a great idea. And then my final sort of AI one is I think we're gonna see video generation but for memes. So everyone sort of focused on this like deep fake, and you know, I think that's happening in the high production value, but I think someone's gonna find the right, you know, TikTok-able, okay? Whatever. I don't know how you say that, like the right, you know, sort of easy everyone can just like go make the goofy videos. They were arranged for a while—I think it was like Korea or something where they would do the like, you know, news reports with the little like avatars, you know, marching, right? But like imagine that but like you just type in the scene, you know, whatever, and you know, I want to make a joke like this and someone's delivering the punchline for you as like a cartoon character something like that. Something that is within striking distance of today, but it hits like a nerve of like it's funny. It's not like creepy, you know? It's kind of obviously fake, but but it's still, you know, pretty 0:33:31.581 --> 0:34:47.029 Funny. Yeah, have you seen the uh the fake Pixar-like movie movies where it shows like the yeah, it has like um trying to think of an example, but it looks like the cover of a Pixar movie and um um, you know, the characters look like Pixar characters, but they're—they're fake? Like people generated it with the Bing Image Generator. And um yeah, there was one. It was like it was like Texans and it was a bunch of Pixar children like holding holding guns and stuff. Yeah, I mean, I'm sure you can find like the level of tastefulness, you know, you can get whatever you want there. But but yeah, and so they actually looked really, really good. I mean, you know, I knew that it was fake just from the content, but but the quality looked real. Um like it looked like a, you know, professional person artist had made this parody. Nice. So yeah, I mean, but video is a whole other level. Maybe I'll push you a little bit on that, like uh, you know, like what do you think? It's going to be video or images, or where are we at now with video generation? So from what 0:34:47.029 --> 0:36:12.602 I kind of like gather I've not gone and tried them yet. I think there's a bunch that are fairly close, but I've seen some really, really awesome ones, but they require you to still like start with a good template, like a Blender, you know, sort of render, and it doesn't have to be super precise or sort of drawing and getting there. So I think my understanding is part of it is the sort of continuity. So like if you wanted to make a scene right from like a movie or even for memes, like you need characters that you can sort of cut away and cut back to, and like they're not just some new complete random, right? So how do you sort of like give that guidance and and have the AI sort of keep—we talked about that—that sort of look back. It needs the continuity across the entire, you know, clip that it sort of knows all the moving pieces whether they're on screen or off. And so I think images are kind of are—I don't want to say like already there, but you can go in and you can get like if you see the pictures coming out now of people, it's always that I still see this: oh, you can tell it's AI. You can tell it's not AI because the person has five fingers on each hand, and I'm like, 'Oh, I feel like—oh, they fix that.' Yeah. Like all that stuff is so fixed it's really ambiguous at this point when you get a picture especially because it's blended. If you look at like Photoshop in the public app now has you, you draw a little lasso and type in what you want, and it'll put it in there. So it becomes this very difficult like, uh, you know, what's real versus fake. I think have you 0:36:13.058 --> 0:36:16.939 ever been to This Person Does Not Exist.com a long time ago? 0:36:16.939 --> 0:36:21.580 Is it is it updated? Do they keep adding new versions, like I mean? 0:36:21.580 --> 0:36:42.454 It these people look like they exist. I mean, it's just solved. Um I mean, I'm not an expert, but like yeah, I mean all of these people look totally real, you know? What actually what gives it away is the background, but the person's face if you had if you had just cropped out just silhouetted out the background, the person's face looks totally real there. 0:36:43.349 --> 0:38:14.680 Was I'm not—I'm not big on like YouTuber personality people, but I think I—I'm gonna—I'm gonna try. I think it's Peter McKinnon. He had a he's like a you know famous for doing photography stuff. He's anyways, he had a video where the users submitted using kind of this Photoshop tool where they either extended the background or they started with a picture but then they edited it in some way, like inserting a shed, inserting a bison, like whatever. So they would give him a picture and people would send them in. He wouldn't see him before, and he would get I forget some number of seconds, you know, 20 seconds, 30 seconds to look at the picture on his computer and figure out which parts were fake and which parts were real. And of course, you know, he's—he's a like if he zoomed in and really thought about it or he knew the location, he would probably know. But in general, he had a really hard time because, you know, it you just don't know. It was like oh, these entire mountains in this beautiful rocky, you know, landscape, or Swiss Alps? It's just completely fake. It was actually just somebody's backyard. And so it's really difficult at the level these tools are getting when you start to blend between them. And I think for first glance stuff again that sort of Instagram, TikTok, whatever, you know, just passing by how much content you consume—a lot of that's going to become sort of the quote-unquote airbrushed. It's going to become augmented in this way. But I think the next step is where you don't even need like you said the background, right? You don't even need that. It's just going to be completely end-to-end. I feel like that's so at hand it's not a bold bold move to guess images which is why I went video. I think video is still not there yet. It's very difficult to get stable, looks like it was shot in a video camera. 0:38:15.689 --> 0:38:41.220 Yeah, that makes sense. I was only—no, that's fine. Any other predictions for 2024? So let's see. Let me see if I have a recap. So you're saying oh, uh, a um a peer reviewer AI peer reviewer. Yeah, we'll also throw in AI like document writer, you know, like documenting their functions and stuff. I feel like that's—that would kind of come for free if you 0:38:42.891 --> 0:38:51.620 had the other one um and uh oh and um um the video generation like a TikTok type thing with generative AI? Yeah, any do you want to do a third prediction? 0:38:51.620 --> 0:40:50.420 Sure, I'll do. I think I had one more, but I've already forgot it. Uh, I think the third one will make a space. We always like to make a space prediction because I'm bullish. I think there's a big unlock that can happen if we can get to space really cheap, like lots of mass to orbit. And I feel there's a lot of different sort of companies working in this space now. A lot of them have really cool technologies, a lot of the demos of pieces of their systems this year—not just SpaceX with you know their stuff that they're testing out in Texas, but also the what is that SpinLaunch and people who are just building various like Stage One or Stage Two components. I think there's a lot of really cool stuff. So I think next year we're going to see people reaching orbit with a lot of these test platforms and sort of start a new space race because if you can get things up there and you can get fuel up and refuel, all of a sudden—I mean they talk about that like bringing one asteroid back. I don't think we'll see an asteroid come back for the next 10 years, but you start to get to where a lot of interesting things become possible if it's super cheap to get things into orbit and vice versa, get things from orbit down onto Earth. I think a lot of stuff we talk about—one of the problems for electric vehicles is a lot of the materials that go into them are scarce or ecologically impactful to mine. But if it's really easy to get things up into orbit and asteroids into orbit, or you know things out to asteroids to mine them, it could be pretty straightforward. It's probably much more ecological to sort of bring down a cargo box of lithium than it is to mine lithium out of the Earth. And so also you can have extending—I don't know if the lithium isn't someone going to fact-check me. I don't know if that was a good one to pick, but a lot of these metals that are in asteroids and stuff I think become accessible. And so I think that's why I'm 0:40:50.420 --> 0:41:25.229 bullish on it overall. But I think next year we're going to see a lot of the people building these sort of test platforms, and I think there's probably a lot in stealth mode because if we can see all these that are out in public, I think there's a lot more that are keeping their sort of head to the ground and just focused on working. But once they start launching, you can't really keep that secret anymore or else they look like some sort of weapon, so they have to become public. So I think we're going to see a lot more stealth uncoverings again. I don't know any secret details there, but I think getting to orbit for a lot of these startups is going to be a thing for next year. 0:41:26.022 --> 0:42:05.796 Wow, it's a bold prediction. Yeah, we'll go that all right. All right, so before we get into questions, I'll talk about our first winner. Our first winner is Frederick J. in Germany. I won't say your last name, Frederick, but you won a T-shirt. So I'll be reaching out to you. Isn't it? Is it Guten Tag? Doesn't that mean like welcome? I know a little, a very tiny amount of German, but anyways, Frederick, thank you so much for being a patron, and I will be in contact with you to get a T-shirt to give. 0:42:08.024 --> 0:43:07.480 T-shirt. All right, and with that said, our first question is from Don on Discord. He says, 'I'm an early career dev, been working for about three years, I'm finding my current role.' And he put in parentheses Fang, so for people who don't know it, it's Facebook, Amazon, Apple, Netflix, Google—one of those companies. I'm part of a huge machine. All the work I do is liaising with other teams rather than shipping code. I have an offer from a startup where I'd be shipping a lot more stuff.' You know, he's worried that at his current position he's missing out on a lot of lessons just because it's so much discussion, not that much working on your craft and writing code. But he's also undervaluing the soft skills that he's doing right now. How do you think about the trade-off between coding and soft skills, Patrick? How do you feel about that trade-off? 0:43:07.480 --> 0:43:13.617 Man, I'm taking off guard because that's not where I thought the question was going. So I was prepping myself. Okay, well maybe I'll. 0:43:15.962 --> 0:43:19.017 take a crack at it. I mean, I think you. 0:43:20.569 --> 0:45:19.217 know, man, so I don't know if this is going to be a distraction or diversion or not, but at some point in my career, I realized—I realized that like not having very good soft skills and not being really aware of what was going on in the company, in the news, you know, in like the programming news, but that was really hurting me. You know, at some point I realized that like I was doing a lot of stuff that could have been done a lot easier if I had just been more aware. So for example, I mean this isn't exactly what happened, but imagine you know you spent all this time trying to figure out a way to really quickly get data from a CSV file and so like you're writing all this crazy C++ code or something to like quickly seek through this CSV file and pull rows. And then someone comes to you and says, 'Oh, there's this thing called a database.' And you know, you just put it in the database, and they've done it all for you. It's that kind of thing where you know I realized at some point in my life—I realized that like yeah, I should actually read more and talk to more people and spend like a little bit less time just like coding absolutely everything from scratch.' So I do think that you know personally I kind of was too far on one end of it. Now on the other end, you do spend a lot of time in big companies, you know, kind of like doing a lot of social work with a lot of different teams and coordinating and all of that, and that does take an extraordinary amount of time. I mean there's a lot more training time at bigger companies, you know, and these are sort of like these soft skills training courses like effective communication and all these things that that you might have. 0:45:20.213 --> 0:46:27.207 to take. And actually the bigger issue is that the code you probably do spend a decent amount of time coding even at a bigger company, but that coding work going to be more like integration work and you're not going to be building as much from scratch. So I do think there's a real trade-off there. What I would say is it's probably below the noise floor. In other words, when I think about reasons to change jobs, I think you know what is your immediate team and manager like? You know, what is the actual thing you're building? Do you feel like when you as a team are done, do you want to look back and say, 'Oh, really that's really cool that we built that?' Or is it just turning a dial or turning a cog in this machine that you're not totally disconnected from? So I think those are the big things. And yeah, I'd be less concerned about whether there's more or less soft skills versus coding, but Patrick, what's your thoughts all right? 0:46:28.557 --> 0:46:32.691 Thank you for buying me time. No, that was a good answer, I guess. 0:46:34.952 --> 0:46:56.114 Like I not to cop out, I don't think there's a singular correct answer. I mentioned this before, but I think same thing applies here. This isn't something that is you set a dial between the two—soft skills and programming skills—and you leave it there. That's definitely the wrong answer regardless of where you set the dial. And I think depending on 0:46:58.220 --> 0:48:55.900 where you are in your career, what situation calls for. But okay, specifically at a Fang company, I will say—I'll give that caveat first—having been at two of the letters in the Fang company, um Jason can one-up me there, but that's fine. I will say that soft skills are often in short supply, and so having them, yeah, that's true, a critical critical team member. But when it comes time for promotion, you're being able to point to specific innovative pieces of code or integrations that you did—they're much more defensible. It is very, very hard for your manager, your director, to go to another director or manager and say, 'This person is such a such a collaborator; they are such a ability to just find the right problem.' Those things are all true, and they're all necessary, and they help you get the results and the leverage. So you really have to be clutch at them. But if they come at the cost of being able to point to like you're helping everyone else do their amazing coding, you're very high risk of having all those people you help get promoted before you do. So in a purely selfish not best thing for the project—now helping the project succeed is ultimately good for you—but I think I've suffered from this before doing exactly these things, running around helping people, doing integrations, making sure and then myself not being able to point to specific code contributions has hurt me at Fang companies. So that probably has different dynamics on different teams and maybe even different of the letters in the company. But at a startup, but in those companies I will say you have to be cognizant that you have to get the right work at the right time, which is a soft skill, but being able to execute and write that code that everyone agrees is—is you know chef's kiss, right? That it doesn't have to be a lot of code. It has to be you have to 0:48:55.900 --> 0:49:23.483 point to specific code that no one else was writing, that you were writing. That's the way to sort of light the engine on the rocket ship. And so playing that balance—the fiddle of moving between soft skill, hard skill, soft skill, hard skill, you know filling your pipe—I think that's a that's a skill in itself, and it is really hard because you never know the right answer, and not even in retrospect do you really learn whether or not you had the dial set correctly. 0:49:25.272 --> 0:50:59.637 Yep, yep. I agree with all of that. That makes a ton of sense. The other thing is, you know, the big difference in my opinion between a startup and a FAANG company is that in the case of the startup, everyone's taking this gigantic collective risk, and either everyone wins or everyone loses at the company. And so that creates, you know, a sort of sense of camaraderie. You know, that naturally creates this—this like sense of, of, you know, you're all winning or losing together. But in the FAANG companies, that's not really true. Most people are being evaluated, and they're being incentivized as an individual. So, you know, you could get promoted on a project that completely fails, and as long as you aren't laid off—as long as you're just put onto some other project in that FAANG company, then you did great versus like the startup. If the startup fails, everyone's stock goes to zero. All right. So we have a three-part question. Actually, these are three very distinct, really great questions, so we'll just go through them one by one. The first one is: How and when do you think the current hype for AI will end? It's a good. 0:51:01.172 --> 0:51:07.635 question. I mean, it does feel like 2012 all over again with AlexNet and all of that. I think. 0:51:15.398 --> 0:51:20.494 That here's what I think, and this is pretty speculative. I think the. 0:51:21.996 --> 0:53:20.680 Q* thing from OpenAI is not going to work. I'm pretty skeptical. You know, in general, you know, I have this theory that things that are secretive usually don't work when it comes to research. You know, the idea that like there's going to be some big reveal. So, you know, okay, look at ChatGPT for example. So, you know, Facebook had something that predated ChatGPT called Galactica where it would write research papers, and it worked pretty well, and it would write, you know, it would generate research papers based on a topic. And so ChatGPT is built on the shoulders of all of these other bottles, right? The thing that OpenAI did really well is connect—connect that technology to everyday people, market it really well, you know, handle the fallout when it does hallucinate and generate fake things, you know, handle that fallout appropriately, and and all that. But it was, you know, all these things are incremental. In general, everything is much more incremental than people would like to have you believe. And so this idea of merging, you know, kind of policy optimization, you know, making decisions whether it's in a board game or decisions in the real world—if you if you have a way to do that, connecting that with language understanding, you know, I feel like the first approach to really do that. I mean, we have Cicero and some of these approaches for some board games, but all of these are, you know, very specific to board games. And as we know from AlphaGo, you know, Google wasn't able to take. 0:53:20.680 --> 0:54:39.484 the computer—the the the code and all that behind AlphaGo and use it to like trade stocks or something, or I don't know, use it to do anything else. So, so you know, I'm skeptical that like even though the people behind Cicero went to OpenAI, I'm skeptical that this Q* thing is going to really be a huge revolution. It could be, but I'm skeptical. And so I think when the when people realize that actually using this new wave of AI to do things in the real world is a lot harder than everyone thinks, I think at that point the hype will end. And you know, it'll be, you know, part of another cycle. And eventually, we will figure out an elegant powerful way to do policy optimization. You know, imagine like how would you combine these large language models with robotics, for example? You know, like someone's going to crack that in an elegant way, and then that'll be the next type curve. But I think this hype curve will die when people realize that it's—it's pretty good for speaking and pulling data and engaging. It's good for search and relevance type stuff, but it's difficult for actually doing things. What about you, Patrick? Yeah, so. 0:54:40.615 --> 0:56:38.014 I mean, I guess it's an easy question, which is like maybe it already has—or next oh. And I think there's a couple of like human factors things that just go into that, which is in a relative performance, you know, metric that each new OpenAI release isn't can be the same sort of like absolute improvement, but on a relative basis that goes down, right? And so the leap from GPT-2 to GPT-3—whatever. It feels so much more powerful because it just seems so much cooler, but over time, each one of those increments you have to actually give the same percentage increase. Your absolute increase has to be significantly more. And they've already said, 'Hey, listen, like we've reached the end of parameter scaling.' We know who knows how much truth or, you know, PR there is to that, but I think we're seeing that, which is we're not seeing the same leap forward. Already doesn't feel as big. That said, I think normalizing this out, you kind of mentioned for search purposes—Jason, like being able to even going from pre-I'll call it pre-Google search engines where they literally were doing like fuzzy string matching. I mean, I know there were others to being able to give just something vaguely in line with your question and get back an answer at least, you know, the next step to writing it—even worse. I was asking, you know, OpenAI ChatGPT the question about some computer science problem, and I didn't know what it was called. I didn't know the right search terms of print in Google, so I just like explained the nature of the problem. It failed me, but it did give me a bunch of relevant things that matched what I was saying whereas I wouldn't have been able to put like 'I have a graph and I'm trying to count the number of, you know, neighbors of this or that, you know, in this particular way.' Google would have just matched like 'count number of neighbors in a graph' or something right and give me some algorithm for adjacency matrix, you know, counting anyways. So, I think. 0:56:38.419 --> 0:57:17.012 it's already here, but I think the hype is already like people—people have the attention span of a gnat. I don't know. Like I feel like that is already dead. So now it's just a bunch of people are going to get to work integrating the current capabilities. The question, like you said, if you go back to previous AI booms, they never really manifested and being used in everyday life in a super super meaningful way. The question is whether the current batch given 12, 18, 24 months are they going to be like a new industrial revolution or not, right? And sure maybe they'll continue to be hype cycles after that, but will they actually affect meaningful change? I think it's yet to be seen. 0:57:18.801 --> 0:57:20.640 Yep, yep. That makes sense. All right. 0:57:20.640 --> 0:57:22.058 That was one. What do? 0:57:23.054 --> 0:57:28.400 you think about the current shape of the show? Is there anything you'd like to change next year? 0:57:32.892 --> 0:58:04.988 Patrick. AI, Patrick. Oh man, you know that's a great idea. We should have some type of AI in the show. I mean, it really behooves us as technologists and as AI people to put some type of AI, you know. One thing I've always wanted to do is is the speech-to-speech translation. I think Facebook has—oh geez, I forgot what it's called. Facebook has this model where they do speech-to-speech translation. So, in other words, you can take embedding. 0:58:05.039 --> 0:58:07.560 and then they pull it back out again. 0:58:07.560 --> 0:58:23.129 Right, but in a different language so you can actually take one of our episodes and you know you can convert it so that we're speaking, you know, Portuguese or something, but it has our intonations and everything. Oh. 0:58:24.833 --> 0:59:00.692 Okay, well that I was partly joking. I think like we actually did change the show this year, so I think trying you know Jason and I are pretty busy, so I think we've gone to a little bit of a different format. And yeah, I actually think this is—I'll say more sustainable long term for us. So yeah, I'm happy, happy with this. And it's been going a long time. I listen to other podcasts and they say, 'Oh, you know, this is our, you know, eighth year or ninth year.' And I'm like, 'Wow, they've been going a long time now? Wait a minute. Hey, wait, like we got longer than that!' But I kind of don't like it. It's not real to me. I don't know. It's very—it's very strange have a very strange relationship with this podcast, I guess. 0:59:02.042 --> 1:00:21.779 Yeah, totally. I mean, I think you know one thing we did this year, as Patrick said alluded to, is like we just made it a little bit more effortless. We just, you know, we get on, we talk—I think what we're doing right now really casually. Maybe next year I'd like to somehow get more of the community involved, but I don't—I don't have any clear plans. But I'm happy with what we're doing right now where we're keeping it, you know, pretty casual. I think you know when you have a guest on the show, you know that guest is on your show to either promote something or to advance a cause, or and you know they have a goal. And anytime anyone has a goal, there's a chance that they don't meet their goal, right? And so—and so it just, you know, no one's really, you know, come back to us and said, 'Oh, I hated that show. I hated being on your show.' Like nothing like that has happened. Every guest has been—yeah, exactly. Exactly every guest has been nothing but polite to us. But it does—it does kind of put a bit of a weight. It's much, much, much nicer if Patrick and I can just, you know, roll out of bed on Monday and say, 'Hey, you know, we should do a show this week.' So, um, we're gonna—we're gonna stick with this format at least for the early part of 2024 and see how it goes. Um, last question was anything exciting in tech coming up? I think we covered that in the beginning. But thank you so much, Vitkos, off for your for your questions. 1:00:24.797 --> 1:00:25.894 Are we doing? 1:00:26.316 --> 1:00:52.202 Another giveaway? Let's do it. So the next person to get a T-shirt is Rob G. Um I don't have Rob's address, but I do have your email address. And so Rob, I will be in contact with you to get a T-shirt. So congratulations, Rob, and thank you so much for being a patron. Rob's been a patron for over a year. So thank you, Rob. Nice. 1:00:55.122 --> 1:01:20.900 I have a small segue before we go into another question—something that we haven't talked about but has been really interesting is 3D printers. So we've talked briefly about 3D printers in the past. I think 3D printers are also sort of like coming into their own this year. So we didn't make—I didn't have a bold prediction about it, but I'm excited to see what's going to happen with the quality and amount. Do you have a 3D printer? 1:01:20.900 --> 1:01:31.589 Jason? Yeah, I do. I actually last Black Friday I replaced my 3D printer, and now I have an Ender S1 Pro, which I really enjoy. Nice. 1:01:32.027 --> 1:02:34.634 Nice. I think we save it up for another episode to talk about, but recently—I think I'll call it a Gen 3. So I had a Gen—not a Gen Zero, like a RepRap, like a you know do everything yourself from scratch, but I got on at Gen One, I'll call it, which was like you procured 2020 aluminum extrusion and you like screwed stuff together, and you know it's just like a kit. Um I had one of those personal—I call that a 10 one. Then Gen Two was like the Ender. I had an Ender. Yeah, it was great, great. I'll—I'll say I think it's a clear Gen Three. I got a Bambu X1C, and okay. Uh yeah, we'll talk about in a future episode, but it is fundamentally a different interaction than Gen One and two. And I printed so many things. I've had exactly one issue once, and I haven't like touched it. Like, wow, it's—it's crazy. So that's amazing. This is the unlock, and so it's just gonna ripple out. I think people have been a little burned, so it's gonna take a little time, but I think—yeah, I think there's a revolution. 1:02:35.039 --> 1:02:44.540 Brewing? Have you been following the nerfs and the Gaussian Splatting and all of this tech, or basically you could take a couple of pictures of something and get a 3D model? 1:02:44.540 --> 1:03:37.982 I am so intrigued. I see. I don't go on X—yeah, I don't go on Twitter that much. When I do, I have some people who have been doing this, and it blows my mind. I don't—I literally don't understand it at all except I know how like traditional photogrammetry works in theory, right? Like taking pictures, matching features. Jason and I have talked about this before. We used to do some work in this field prior, so like I understand like matching features from thing to thing, understanding parallax, like estimating pose, so like a traditional pipeline for building that stuff. I understand. When I watch the nerfs and Gaussian Splatting, I—I don't understand. Like I just see the things like thin wires and strings that traditionally would have been impossible to do via, you know, other mechanisms; they just would have screwed up. And I see them work, and it's just like they blow my mind away, but I—I haven't done the deep dive to figure out what they're actually doing. 1:03:38.522 --> 1:04:05.539 Yeah, we should do a show on it. Um I know a decent amount—I know a decent amount about it. But yeah, that's really cool. Um Yeah, I feel like that's all kind of coming to a head, isn't it? Like imagine if you break something, it's not completely shattered; it's just—it's just broken in two. You could like, you know, put the two pieces next to each other, you know, get a 3D model out and then 3D print or replacement something that'd be freaking awesome. Oh, see? 1:04:07.463 --> 1:04:14.787 We already—but we already passed our prediction thing. It has to be like you can't no one do this this year. We gotta bring them for 2025 if you. 1:04:15.023 --> 1:04:36.720 Do it. We have to be advisors on my company or something. All right, so Rick Vegas asks what are the essential things you think every programmer should know? For example, commands, types, for loops, dictionaries, arrays—what are the important things that cross all languages? 1:04:38.462 --> 1:04:47.963 Machine opcodes. You should have memorized assembly set for, for you know, yeah, at least three or four different for risk. 1:04:47.963 --> 1:06:04.420 Five. Yeah, I mean it's a good question. I would say—I mean there's definitely, you know, definitely data structures. You know, actually, I would say university does a pretty good job of covering the abstractions in a pretty general way, right? So data structures, you know, they use sorting as a way to teach recursion and also teach, you know, algorithms and complexity. So, you know, a lot of people get turned off when we talk about all the different sorting stuff in undergrad. They say themselves a lot like, 'Why would I learn this when it's like dot sort in almost every language?' But it's really just a case study in, you know, different ways of solving the same problem and how to relate them to each other and everything. So, I would say—you know, it's hard to give that answer in five minutes, but I think the undergrad curriculum is pretty good at giving kind of the overall basics, and you don't mean you have to get an undergrad, but it's a really good launch point. 1:06:05.980 --> 1:07:28.731 Yeah. If I had to add one thing, I agree with you. A lot of people would say like Hash Map or something, which is definitely also like a good one when you start growing up in size, but if I had to add something to the data structures and algorithms and a sort of way of thinking and decomposing things, I would add string manipulation. It doesn't have to be like Regexes, but just like being able to—and maybe it comes from doing programming competition, I don't know—but so many times I see folks struggle like, 'I have a string; I don't know how to get the thing I want out of the string.' And it's just like no be comfortable. It could be crappy, it could be non-efficient; it doesn't matter. Just know how to—like, 'I have a string; the string isn't what I want; I want some piece of it or some recombination of it.' And that requires, of course, like you know arrays because mostly we treat strings as, you know, arrays of characters. You know, it requires looping. It requires, you know, arrays of arrays and being comfortable in that, but like just really be comfortable like taking a string, chopping it up, and getting the pieces you want out of it. They don't really teach that in data structures. Now they teach things you would need for doing it, but that specific additional one—just very, very common—oh, and debugging! Oh man. 1:07:29.963 --> 1:07:51.546 That's a great one. Yeah, actually, that is one thing that's not taught very well in school because it's so language specific. But just getting whichever language, you know, when you start your first job, chances are you probably won't have any control over the language. They'll say, 'Hey, you know, you're writing in C++ or Python or whatever.' I thought— 1:07:51.546 --> 1:07:53.183 You were gonna name something esoteric. 1:07:54.179 --> 1:08:19.187 Oh yeah. It's your first day on the job and you have to write in what was it? PL/I or something that one where it's all images, but yeah, knowing how the debugger works in that language and how to profile and find out what—how where your code is fast and slow is really important. 1:08:19.913 --> 1:08:28.907 Oh, I wasn't thinking that would be. But yes, you mentioned that was good performance analysis as well. Yeah, yeah, yeah. Okay, yeah. Both all the things you need. 1:08:29.633 --> 1:09:25.034 To know. Yeah, debugging and perf analysis super super important. All right. So Hoover—Hoover Shoes—asks, 'Oh, it's gonna be a tough one. I'm a Computer Science student who's still deciding how I want to organize my life.' Like for example, what purpose? And he gave some examples: raise a family, effective altruism, doing cool engineering and science—what purpose his career should serve? What question is were there profound experiences you had that made you want to pursue a particular life?' And I think he means outside of work or, you know, sir, you know, a life that your career affords you. So yeah, I think man, I have to think about this for a minute. I don't know if you— 1:09:25.034 --> 1:11:21.820 Have a quicker answer, Patrick. I mean, I guess so part of this is very person-to-person dependent. So right? No, if you're a religious person, that could give you—I'll say like externally like a way of grounding yourself. And so for me, that's definitely like a part of it. But just in general, I would say, leaving that aside, that you kind of have to pick in my opinion and decide what is work going to define? Like who you are? Or are you going to have like a definition of who you are and like you work to kind of achieve it? I guess like hand wave without getting into, you know, having thought too right about it. And I think that both kinds are okay. And I'll give an example which is I routinely have career conversations with people. You know, I've talked to people over the years—managers. There are some people who will say, 'I never want to work on an ad system. Ads are bad, and I don't want to work on anything that has to do with serving ads, building ads.' Building ads like you know just, and you can insert any topic here. Other people might say, 'You know, I don't have a specific moral like, you know, problem with ads so long as if the technology I'm working on and I'm learning something and I'm engaged and I have a good team.' You know, these things are more important. I think this is a slice of that same problem which is do you go to work and you spend a lot of time with the people there and on those topics? So it is perfectly fine to want those to align very closely with your internal beliefs, but I think it is also okay to say you have a very strong external—several people might call that work-life balance. You might call it whatever, but you have very strong external—you have a family, you know, whether that's your own, you know, spouse and children or whether it's your parents or your brothers, like it doesn't matter—or friend group, and you're really focused on a hobby outside of work and work as a way to make money and provide good insurance. 1:11:21.820 --> 1:12:59.447 Support your life. I think both are okay, and being understanding that there are people who turn that dial between those two. And again, all right, the hobby horse—I don't think it's static over time. I think various times people can have that knob turn back and forth. So I don't know that there's advice you can give you. I think it's a decision you should make and be thoughtful about like what is the reason that you work? And why do you work? And why do you work where you work? And you should reevaluate it from time to time, and maybe come to the realization that it should switch. And for some people, I will say, it becomes a bit of a stumbling block if you allow it always to be turned to 'work defines me.' Then it becomes very difficult to switch jobs because you're almost switching identities. And I will just say in—I guess I'll say at least in Western society in our current capitalistic whatever you think about that, allowing work to have that control over you is something that doesn't lead you to a game theoretic optimum, right? Like you are going to end up being probably underpaid, under thing if you allow work to have a sort of hold over you. That you have to do that maybe if you go run a startup and your own founder, and that—that is I'll leave that aside. I don't I can't sort of speak to that; I've never done it. But that may be a way out. But otherwise, if you're working at sort of like a medium or big company and you know that's a dangerous place in the dial to be for too long, which is 'I allow this work to be who I am.' So that would be my only caution there, but otherwise, I think it's something that individuals all will and you will work on a team where everyone on that team defines that differently. 1:13:00.915 --> 1:14:09.023 Yeah, fantastic answer. I mean, similar to Patrick, I think for me, you know, I have a lot of things outside of work—things relate to kind of church service and some of these other things that when I look back on are a big part of my life, I think. I mean, I saw this thing that like I don't really get to. I am the kind of person that cries in movies. I am guilty of that. You know, there's something about like when the music plays a certain way, like they've figured out, and you know, I realized at some point I came into a movie about halfway through and I started getting real emotional, and I realized like I don't even know what's going on in this movie. It's just when they have the music a certain way and the people look a certain way and the angles a certain way, the camera—like it just triggers something in me. So I'm kind of a sap in that way, but I saw, but I don't really feel that way about things I see on the internet or quotes or anything like that. But I found this one quote and it said it. 1:14:10.643 --> 1:14:28.665 said something like there's a younger version of you who would be like so proud of you, you know, like who you are right now, right? And that really hit me. It's like man, you kind of forget about what you were like, you know, like so many years ago. And the whole journey, right? So I think for me, 1:14:30.218 --> 1:14:35.095 I try to stay pretty—I can't. I want to use 1:14:36.681 --> 1:15:44.960 the word 'pure.' It's not really the right word, but like I want to keep pretty, like, not compromised, you know? Like if—I mean it doesn't mean like I just have no filter and I just tell everyone like, oh, you know, you're fat or something like no. It's not. But it's like... but it's like, you know, in my head I have kind of like an unadulterated view of myself and, you know, my place I'm in and the world around me and everything. And so I can kind of make predictions that come true at least, you know, locally. I'm not predicting like, you know, anything like the lottery or anything, but yeah. But like locally in my life, I can kind of make predictions that more or less come true because I have a pretty clear view of what's going on. I think that's something that I think is really important. And, you know, if you have that, then you can kind of say, like where do I want to be and, like, you know, what is the path that I could take to get there? And just doing that one step at a time. 1:15:47.219 --> 1:17:15.360 And so I think as Patrick said, there's times where, you know, I'm really passionate about some work thing, and so, you know, life is kind of getting in the way of solving this problem or getting some work thing done. And there are other times where, you know, I really have a lot going on outside of work, and, you know, work is really there to pay the bills or to be something that I do during the day before I do the thing I really want to do. So it's, you know, you're constantly going to keep moving through that space. I don't think that you'll ever be just focused on one or the other. But yeah, I do think it's important to really take inventory of yourself. You could look back really far, like as that slogan was suggesting, or you could even look back in the past week and say, 'You know what did I do this week? Did I kind of end this week kind of better than I started it? Did I waste a lot of time this week? And do I have a whole bunch of regrets? Is there things I can fix there, or am I overall pretty happy?' I think there are so many people that, you know, on social media and other places, that are trying to tell you how to be happy or give you purpose in your life. And I don't know, I'm a little skeptical. I haven't listened to any of them, and I look back and I feel like it's probably a good decision. 1:17:15.360 --> 1:19:04.386 Yeah, I mean, you made some good points there. I think for me, I guess if I iterate on what you're saying too, is I think people get a little hung up about setting specific goal posts that they want to—it's like very, very. Like if you think about all the design space of everywhere you can be: family, marriage, whatever. They set very specific goals, like 'I have to be married.' 'I want to be married by this age.' And people set very specific things and work-related as well. For me, it's about, and I think you're sort of alluding to that Jason is, like each day, each month, each year, trying to just have that compounding growth. So for where I am right now, given the extent of my abilities, being honest about that, and saying if I want to grow, it's not just like go take a supplement or go take a course. It's like I need to start spending a little bit more time each week and not trying to have these step-function changes but just slowly move the heading of my vector through the space and realize that further away items are going to have a bigger cone of uncertainty, and what I want today for me in five years is very different than what I will probably end up being in five years. But I want to look back like that, like look back and say, 'Oh, I'm still happy that this is a good spot to have been in.' And be realizing your circumstances and what is and isn't in your control. But personally, I've never resonated with this—like you have to have a five-year goal and be pushing with it, or where are you going to end up? Maybe maybe it's a strict necessity for some people, but I've found for myself it's like I've never really had a goal other than to be somewhat better than I am today—like being more and having grown across all the things in five years without specifically needing to have done this bucket list or checklist in five years or I will be sort of like have missed my goal. 1:19:05.466 --> 1:20:02.369 Yeah, I mean looking back one thing that I feel is pretty universal that I'm really kind of happy with is that I didn't—I didn't really have any fear or shame. You know, like putting myself out there. I'll try. Like The Show is a great example. Every episode we talk about a programming language—spoiler alert: Patrick and I are not experts on Unity, right? Yeah. Like every now and then we'll get an email from someone who's really an expert on a language telling us how like 'we don't know what we're talking about.' And that's fine. I mean it's one email in the span of many, many, many positive emails so it doesn't bother me at all. But I mean we do get that, and I think that person's probably right. But you know Patrick and I kind of put ourselves out there to talk about as many programming languages as possible, and we can. 1:20:03.940 --> 1:22:03.940 You know, we can do our best to talk about the language, to explain it to folks as best we can, and we're not really deterred by the fact that we're not experts in every language. And so I think that kind of percolates through everything. I had this philosophy like this is so many years ago in college. I would always go on a date—on one date with anybody, any girl who if we seemed like we got along maybe in a class or something, I'd always try and ask to on a date, or if a girl asked me to go on a date, I'd go on one date with anybody. And so when you have that philosophy, you get rejected all the time. People have boyfriends, or they are just not into you, or whatever. But I think all of those experiences were really useful, and eventually, you know, I went on a date with a girl who then we got married. And I'm really proud to be married to her today, right? So so I think that, you know, don't really be afraid. It doesn't mean you do dumb stuff, right? It doesn't mean like if you're thinking about skydiving without a parachute—yeah, be afraid, right? But like but like don't be afraid to do social things, you know, really kind of push yourself, and you can learn lessons from that. And there are definitely things I regret, you know, arguments I got into at work, the times where I raised a big stink out of something about something that really wasn't that big a deal. Everyone has these kind of—they're not skeletons because everyone knows about it, but what's the word? But things that you regret that kind of weigh on you, right? But there's that also needs 1:22:03.940 --> 1:22:47.912 to be in the context of all the things you did, all the things you said that had a huge impact, a positive impact on people. So so yeah, I think the biggest thing that I could try and tell people is to just take chances and explore and learn as much as you can about yourself, about people around you, and about programming languages that got really deep. All right. So next question from Champion Two-Timer? Oh, two-timer champion! That's cool. Okay. I just figured it out as I said the name, and now 1:22:47.963 --> 1:22:49.400 we want to know what their champion does. 1:22:49.853 --> 1:23:10.170 Yeah, you should tell us, Champion. If you're in the Discord right now. Actually, I'm gonna ask Champion: Tell us what he's a champion of. But in the meantime, what kind of programmers were you in college and what was your first job when you got out? Patrick and I had the same job, so this is going to be interesting. Patrick, why don't you go first? 1:23:10.626 --> 1:24:41.009 Was a terrible programmer. No. And in college—I mean, I used it to try a lot of different things, so I guess I don't exactly know how to answer what kind of programmer I was in college, but I definitely wasn't super sure what I was going to do when I got out. My first job I took was at a defense contractor, but I did it as a part of a program where I would get to try a bunch of different projects there over the first few years that I was going to work there. And so that really appealed to me, and also I was going to get my Master's degree at the same time. So for me coming out of college was like about continuing my education but also about getting to try a bunch of different things because I feel like it paid off for me. But I've always been in the philosophy that the earlier you in your career, you really want to sample as many things as you can—like become good in things at some point. Though it's like, you know, pyramid, you want to have a good foundation, but as you go up, you want to narrow in your focus until you become better and better and better at certain things, but that you have a good foundation that you've built up on. So yeah, I guess coming out of college first was at a defense contractor. I enjoyed my time there. I learned a lot. I wasn't ready to go straight into—I mentioned earlier being at a couple thing companies, but like I wasn't ready for that right out of college. I needed to mature a little, grow a little before getting sort of thrown into that. So I appreciated that first job, but there was a time when it was no longer the right job for me, and so I was thankful that I was able to 1:24:43.067 --> 1:26:21.145 rotate out of that. Yeah. In college, you know, I was really competitive. I actually won the high school programming contest, and my high school—and so because of that, I ended up doing a lot of low-level programming just because most of these competitions, you use C++—you could use Java, but there were some things where Java was not fast enough, and so your answer would be correct if it was in C++. And so I had a brief period on the competitive side where I was using Java and I switched back to C++, and that gave me like a very low-level mindset where you know the joke about the CSV file and not using a database—I mean, that's not far from the truth when I was just starting. And so that's kind of how I was in college. It's just like, 'You know, any problem? Let's just throw C++ at it.' Not no third-party libraries, no nothing. Just write everything in C++. I went to the same job as Patrick; that's how we met was at our first job. And I guess the thing that makes that kind of interesting is—you know, it was just a job in the area. I think that so many people might feel like their first job really defines who they are. I mean, 1:26:22.799 --> 1:26:57.291 Actually, my first tech job I had while I was still in college, I worked part-time at the State of Florida doing some tech stuff for them. But yeah, I think you know I wouldn't really stress too much about your first job. I think whatever you get your hands on is going to be great experience. And you know, we were there. Patrick, how long were you at our first job? I was there about five years. I think I was there exactly five years. Yeah, I think. 1:26:57.426 --> 1:27:00.767 It was just shy of five. I think it was like four, four and- 1:27:01.695 --> 1:27:07.298 Some change. Yeah, apparently at five you got, I think a clock or something, but I left too quickly. Yeah, I know. 1:27:08.901 --> 1:27:42.280 I literally hit the fifth year, but I think I left so quickly after that I didn't get my clock. But you know it was five amazing years. And you know, it was basically from my point of view for Patrick might be a little different. My point of view is basically just throwing a dart at a dartboard. I just knew about this company, interviewed, and then just worked there for five years, and it was extremely rewarding. So if just a random draw gave a really rewarding job like that, then the takeaway for me is that it really doesn't matter what your first job is. You're gonna—you're gonna learn a lot. Yeah, I. 1:27:43.275 --> 1:28:47.755 I mean, I'll drop right here and then we'll probably have to go into lightning round for the rest of these questions. But also, I don't think Jason—no offense to him or myself—but neither of us came from sort of target schools for big tech companies. So like, right that they don't recruit there, but like we didn't come from schools that you know traditionally you see as people going there. And so I'll say the same thing. People get really worried about exactly what college they go to, exactly what first job they have. I think it's just about building up your character and your skills and your abilities. And you know, it might take time, and it can be harder or easier based on certain decisions. But like, I guess like Jason said, it's—I think it's a fairly robust kind of thing for the most part, you know. And so but again, it's it's about being that like realistic about where you are and where you want to be. And do you need to grow? And so some people have a sort of outsized view. You got to be humble but practical. Like don't—don't err too far on either side. Don't underestimate your skills, and don't overestimate them either. You have to be realistic about where you are and where you want. 1:28:49.054 --> 1:29:11.852 to be yep yeah totally well so we'll do the final shirt the final shirt is going to a gentleman name and i hope i don't butcher names a reeb a reeb k um a reeb has been a patron for four years thank you so much a reeb for your patronage and uh and we will uh get a t-shirt over to 1:29:12.679 --> 1:29:42.320 You. All right. I'll read the next question here. Jason, all right, which is—a lot of people already specialize? Oh, I think I already answered this. People already specialize, at least knowing what path they want to be on, whether it's AI, security, DevOps, data engineering, web dev. I'm an early career back-end software engineer, and I get to play a little bit with almost everything, and that's what I like about it. But from a career perspective, should I be thinking more about specializing? Is specialize even real, or is it just like a title? This is from Cup of Geo. 1:29:42.320 --> 1:30:26.811 Yeah, it's a great question. I think that—I think you should, you know, within the context of engineering, you should do what you enjoy, you know, what gives you the most strength. So, you know, again within the context of your job. So if at your job, you know, you do find yourself doing a lot of different things, you're hopefully on a team and you know maybe you can do some horse trading and say, 'Hey, I'll do the—you know, this part of the work.' You know, I'll do the refactoring if you take this other item, or what have you? And you can kind of find the thing that you really enjoy and double down on that. Um, and so you 1:30:28.583 --> 1:31:45.800 know in that sense like it's going to kind of guide you to be more and more specialized over time. But at the same time, like—you know, you have to kind of keep your eye on the kind of overall market and kind of where the puck is going. So, you know, I think if like if you're at a company that uses what's it called, ASP.NET or something, right, which like every year you know loses market share, well then you take a little bit of time to learn to learn—you know, Node.js or something just so that so that you're kind of skating to where the puck is going there. Um even if you're not using it at work. So, so you know it's a balance to strike, I would say. Really, it depends on what gives you kind of strength and energy as an engineer. And different people have different things. Some get bored really easily if they start specializing. Other people—you end up in the sort of analysis paralysis when they have too many options and they really just need to pick something and start specializing. So it's a very personal question, but I think the biggest thing is just monitor yourself and adjust accordingly. 1:31:45.800 --> 1:33:44.924 Yeah, I—I think for a slightly different tech, but I already you know kind of mentioned it. I actually don't think it's super important to worry about what you want to be when you grow up. You know, I was having this conversation with my kids and I don't really know what I want to be when I grow up. I mean, my current specialty as a software engineer is like—I guess what I've been working on the last many years, it's completely something I didn't really necessarily knew existed. It's not something I would have been working on. It's in a field supporting things that I didn't like, probably weren't even an option when I was in college. And so I think that like it's very hard to predict some of those things. And I started out doing—you know, we talked about a defense contract doing like low-level embedded programming, doing multi-year government maintenance contracts, like really weird stuff. And then I, you know, went to a FAANG company and did, you know, giant large-scale distributed MapReduce projects. And you know, I've done everything in between, and I've sort of taken the tack that like I want to always be learning new things and figuring out how stuff works. And there are—I think places in the space that you can learn stuff that's just too niche or like it's sort of the expected value of that sort of niche skill is is low, so it may have really high payout but very low probability of that payout. Um and you can you can gamble on those things. If you look back, you know, there are things I—I don't want to. Oh, I don't want to attack anyone. There are technologies we've even named so far which could be considered like big gambles with potentially big payouts, but they're just that. And you may like that; that may be exciting, or you may say which has kind of been more my attack is take things which are interesting to me but are somewhat stable or adjacent or you know skills that that will sort of go along with me and that has served me well. So I feel that there's a lot of can be a lot of peer pressure, and a lot of people just know or think that they know, and they might 1:33:46.139 --> 1:35:13.259 be wrong. But a lot of people who think that they know and just don't—don't feel that pressure like do stuff that you like. And you know, for me it's about, and Jason kind of mentioned those things about regrets. Don't I try to look at the upside. There's always been something to learn from all the things I've done. Some stuff has been truly terrible and maybe less learned, but I'm still thankful that I did those things and make sure that I know how to like not do them again or how could I maybe have avoided that? Right? There's always sort of lessons to be learned. So I—I mean, I think first 10 years, I think you don't really have to pick a specialist. I don't know if you ever do. But yeah, I would also say there is a danger and sort of on a given project like, you know, saying that I just will do anything. I don't want to do any one thing, and you know every job I take must be different, and sort of being a renaissance person can be a little dangerous as you want to get promoted further and further. So I think there's about—you know it there's different ways of playing that game and sort of having enough attention spans that you really learn something, but then realizing that you can have that time slicing over time is probably better and more stable than sort of going into a project and saying I want to do a little bit of everything here. Um and sort of forcing yourself to be super spread thin doesn't allow you to learn in depth any one thing. And so I would say learn things deeply sequentially rather than—you know, spread yourself too thin. 1:35:13.259 --> 1:36:10.977 Yep, that makes sense. All right, we're up to our final question. Thank you so much for the questions. Really interesting stuff. We're going to end on a really light note here. Do you guys have a favorite? This is from Sky2C. Do you guys have a favorite esoteric language? Um I mentioned it a little bit earlier, although I think I got the name wrong, but Piet, which is the one where it's a picture—the source code is literally a picture and the compiler reads that picture and turns it into some algorithm. I just think that's really cool. Um you know, I could imagine some kind of like it'd be cool to get a to get a Piet program that sorts onto a canvas and put it up in my house somewhere that would be pretty neat. Maybe I'll work on that for 2024, but—but yeah, I just thought that you know crossing media there, I think is really interesting. Maybe we can have ChatGPT create a Piet program for me or something. 1:36:12.529 --> 1:36:17.119 I was going to go Egyptian hieroglyphics before I realized that it meant a programming language, not a uh. 1:36:18.959 --> 1:36:56.505 I don't really, you know, we joke a lot about, but I've never really sat down and tried to program in many esoteric languages. Probably some assembly language, you know. I it's interesting the difference between CISC and RISC architectures and assembly. So I did some programming in Motorola 68K assembly, which is CISC, so complex instruction sets, and I actually kind of like that. I guess I don't know if that's considered esoteric. I might offend some people by saying it, but I feel like that was a pretty cool opportunity, so you know for a non-mainstream language. For today, I think you know that would be my. 1:36:58.007 --> 1:37:27.032 Pick. That makes sense. All right, let me just do a quick check of the other channels and the live show chat. I think we're good. All right, thank you everybody for a bunch of really interesting questions. It's been an amazing year. Thanks Patrick for taking so much time to be a huge part of the show. And we will catch you all next year, Patrick. Any final? 1:37:27.032 --> 1:37:53.037 Words. Thank you Jason and thank you for the compliments. Thank you for all the listeners out there as we beam this across. If you're an alien listening, send us your cobalt in an asteroid. All right. No, no, it's been great. It's another good year, and so yeah, I'm happy. I'm thankful, thankful for lots of things, but thank you Jason. Thank you for sticking with this, and yeah, all the work you do. 1:37:53.644 --> 1:38:32.740 As well. Yeah, totally. All right, everyone, we'll catch you later. Have a good year. Music by Eric Barndoller. Programming Throwdown is distributed under a Creative Commons Attribution-ShareAlike 2.0 license. You're free to share, copy, distribute, transmit the work, to remix, adapt the work, but you must provide attribution to Patrick and I, and ShareAlike, and kind you.