Are Junior Developers in BIG TROUBLE?! - Engineering Manager AMA

All right, let's get things going here. Give me one sec. We'll wait for Instagram to do its thing. It's the last one to go usually. Come on. Come on. There we are. Cool stuff. Okay, welcome to the live stream. Um, if you're new here, these are AMA style live streams. So, definitely feel free to jump in the chat, ask anything you want about software engineering, career development, whatever. Um, already from Twitter, yeah, juniors are in trouble and now is the time to build your own biz. Go. We'll get into all sorts of things. So, of course, like I was just saying, if people want to jump in the chat, ask questions, share thoughts, please do. Um, we have a topic for today. Of course, it's it's themed every week. So, this one is going to be about AI and junior developers. Um, I'll talk about why this topic came up. I don't think that just for full transparency, I don't think I'm going to be talking any about anything from my perspective that's like that's groundbreaking. I think that some of this is really about uh kind of resharing my perspective on things. And that's um kind of how it came up in the first place. Devin, welcome to the stream. If we're going to be cooked, let's be cooked together. That's right. Let's make it a barbecue. Uh, and then Francisco's jumping in again to say in trouble, but also the best time ever to be junior if entrepreneurial. I think there's a lot of cool stuff that's going on. I want to chat through it. If folks are building software, leveraging AI, and having success, please do, you know, share in the chat. happy to to kind of see what you're what you're building, how you're approaching it. If you have success with it, share that. If you have things that aren't working so well and you've pivoted from that, please share that. I think it's a great opportunity to use the chat for that. I like reminding people like I make YouTube videos, so if I wanted to go make YouTube videos, I just go record something. For me, this is all about engaging in the chat and making sure that we can steer things that way. So, um, yeah, to kick things off here, if, like I said, if you're new, I have a newsletter that I send out every week. How this usually works is that I have a uh second YouTube channel. It's called Code Commute. And on Code Commute, I talk about a bunch of different topics when I'm commuting to and from work. And then from there, I review at the end of the week to see what the most interesting topic was based on people's engagement. So, that's usually views or if there's a lot of comments and stuff. And then I write a newsletter article about that every weekend. So the newsletter that I linked is at weekly.devleer.ca. Folks on Substack. Sorry that Substack is sort of like a secondhand streaming experience because it's not being done through reream, but you you're already where the the newsletter is. If you want a better streaming experience, you can head over to YouTube uh at dev leader on YouTube. It's it'll be better. Sorry when I go to screen share and stuff. I can't do that to my phone. Um, and then for folks on Instagram, here is also the link, um, you can check that out. Um, yeah, Jonathan Baron, funny how cooked and cooking have opposite meanings in slang. Yeah, very similar words, but, uh, exact opposite. So, I put this newsletter out every weekend. You don't have to subscribe if you don't want like an email newsletter. By all means, I don't not trying to force anyone into that. Uh, it's not for some people. I get it. But if you enjoy the live streams, you want to come back the next time and stuff and you're curious what the topic is, it's almost a guarantee that whatever the newsletter is about is what the live stream will be about. So, um, Justin can is chiming in saying the that YouTube is definitely a better streaming experience. So, um, I think out of all the platforms it goes to, it's probably the best. Um, Instagram to me looks like it's frozen right now. like I don't know LinkedIn sometimes I I don't know if the the comments and stuff are going through I don't really trust it YouTube just seems to be old reliable so check it out but topic is are junior devs cooked and then specifically with respect to building things with AI so the reason that this topic came up is that I had a video out on my main channel on dev leader and this is probably around six months ago talking about sort of the stereotypical thing that's floating around right now where as uh you know junior developers that are trying to you know or aspiring developers trying to get into software development and building things you have access to all these things now like all these different AI tools and even if you take the most simple like chat GPT kind of experience you can go ask chat like I open up a browser tab right now and ask chat GPT can you uh give me code for a small program that does X And you could probably have something up and running, you know, something quick that it can output in the chat pretty quick or it could walk you through how to build it. Or if you're using more advanced tools, they can actually like an agent mode and stuff, they can actually go basically build an application for you to uh varying degrees of success um which we'll talk through a little bit. But the point is that the stereotypical sort of junior dev experience now is like is this vibe coding type of thing where it's like you know again I'm going to generalize here because this is what the the original video was about that I was sort of doing a reaction to that people were saying hey look I work with these junior developers and they're being hired in at my company and they don't even know what they're doing all that they're doing is going to chat GPT or whatever LLC M getting code out of it, dumping it into the code base, pushing it up for review, like they don't actually understand what's going on. And the more that this is talked about on social media, the more that I'm kind of hearing from people that the the perspective is the more that they're doing this, the less learning opportunity that there is. So, when I made this original video, I was kind of reacting to that saying like kind of the things I'm going to talk about today. And the reason this came up is because someone commented on the video and they said, "Hey, it's been about 6 months. What's your take? Like, has this evolved at all?" And then they shared their experience which was essentially that they're at a company and uh you know not their words exactly but essentially like I don't really know how to program and I was able to use AI have it kind of spin up something for us at at work that replaces some other type of internal tool that they have. And they describe their experience and this is sort of what the article is all about. They described their experience as like I was kind of just fumbling through it and it kind of felt like they're describing vibe coding like it gives me something and I try to ask it to fix it until they have something working and they said like couple problems right like one I don't really know what to ask it because that's often my feedback which is like hey take the time to learn right you should be asking the LLM questions so that you can be learning along the way. So, they're like, I don't even know what to ask it. And the problem is that the LLM doesn't actually suggest what you should be asking it. So, they're like, I feel kind of screwed over by that. And then they acknowledge like, hey, I got something that was up and running. It's working. It seems better than what they have where they're where they're working. And they're like, I just think that it could have been better if I actually understood what I was doing. You know, no argument for me on that. Um, Daniel says, "How do they even pass the technical interview doing that?" It's a good question, right? Um, I think there are some places that are like, you know, strongly against trying to screen people that are using AI tools and stuff like that. Um, and would love people's thoughts on this, right? Like I'm curious to hear both both sides. Infected FPS says, and I I totally agree with this, some people are just smooth talkers in uh in in interviews. And forget AI for a second, right? Think about um think about the people that are really good at lead code. And when I say this, I know there's there's going to be all sorts of people that are good at lead code, but I I personally know of people that have like just practice lead code. They would tell you I'm not a good programmer, but I knew I had to practice lead code. And they got good at lead code, would still say they're not a good programmer, interviewed, and were able to kind of talk their way through it and land jobs. So, I think that you can you can be a smooth talker. It can work really well if you know how to sell yourself in an interview and the experience happens to line up that way. It might just work really well for you. I'm not saying that I would, you know, gamble on that. That that's like the good, you know, put all your eggs in that basket and hope for it. Uh certainly not. I would personally much rather be interviewing and talking from a place of like I feel confident about what I'm talking about, not not just hoping I can smooth talk my way through it. But um yeah, I would be curious to hear people's perspectives on AI tools in interviews, right? You have companies that are now sort of there's a whole spectrum of this like either mandating AI tools for developers, right? Trying to promote usage. There's some places that are like tracking the usage to make sure that it's being promoted enough. If you're at a company like this, should you be able to interview and use AI tools? Like, isn't that going to be what you're doing on the job if that's the the new expectation? It would be curious to hear people's thoughts. I know that even at some big tech companies right now that they are doing a big push for AI tooling and AI usage and still in their interviews saying but we don't value AI tooling in interviews. So I think that there's lots of different perspectives on this. Curious to see where people are at. If you're comfortable sharing in the chat, please do. Um curious. I don't I honestly don't know where I'm at on this. Um my my take would be that if the expectation of people at your company is that they should be using AI tools, I think that you should make the interview as representative of the work that people are going to be doing. You take AI out of the picture there. So, forget we're even talking about AI. I still stand by make the interview as representative of the work that someone will actually be doing. This is also why I often say I don't think lead code makes a lot of sense unless you have people cranking out lead code questions for work. I don't think it's very representative. But that's why my argument if you're going to be sort of really promoting AI tool usage and having people lean into that, I say have them use it. I just think that it means you have to change your line of questioning and how you're gauging candidates. So, um I know that not everyone's going to take that stance. I would say that if you're um if you're trying to actually assess someone's experiences, right, this is where it gets tricky. If you're doing a behavioral interview and you're like, I want to understand from, you know, historical scenarios how you navigated these things and you ask a behavioral interview question and then someone uses AI to answer that. I'm like, that's that's seeming kind of weird to me because I just want you to navigate a scenario that you lived through. I want to understand your thought process. So, I see that one kind of different, but curious. Um, the great, how do we pronounce it again? The Great W, I think is how we said last time. Thanks for joining again, by the way. Making the interview process realistic would be a big win for everyone. Yes. Um, if I could control all of the software engineering interviews in the world, that would be my my thing is just make it realistic. Like I've I've been in interview loops talking with the other people that were doing interviews being like I don't understand why you're trying to to gauge candidates that way. Like it makes no sense to me. Christopher, I'm self-arning programming. AI has been invaluable as a teacher, but I do my best to learn. You can explain complex functions and give me code examples and exercises. A full-time teacher for $30 a month. Yeah. So, Christopher, if you don't mind sharing, how are how are you getting that experience out of AI? Because I would love for you to be if you don't mind. I I realize I'm asking you to like type stuff, so not trying to put you to work here, but would love if you're comfortable sharing that because the scenario that I'm going to talk through from this newsletter article is is basically the opposite for this person. And I think that there are a lot of people unfortunately that are falling into this sort of category. But I'm curious if the way that you're using AI, as you're saying, is an invaluable teacher. I'm curious if that lines up with some of the suggestions that I kind of have around this stuff. Um, and if there's gaps, then I would love to learn, too. And that way I can share with other people. I'm not really, you know, sort of uh in the the self-learning programming uh mode anymore because I've been doing it for a while, but there are lots of things all of the time that I'm trying to learn in software development still. And uh I've been trying to change how I'm using AI to make sure that I'm learning. There's still some stuff in like front-end development where uh like and I'll share about brand ghost a little bit later, which is a platform I'm building. Uh there's stuff in the front end where to be completely transparent. I'm like I don't care. I need it to work soon. So like AI go do it for me. Now it's not sustainable because I have I'm not getting better at front-end stuff. But it's at least like letting me make some progress. So, you know, there's still things that I have to improve on for myself, too. Um, Jonathan Baron says, "I'm using AI too study for a AWS AI practitioner exam. It's fast at creating table summaries from notes." Yeah, I think great, you know, for summarizing, but I think Jonathan, you can probably attest to this. It's not just, oh, I have the summary now I'm done. It's like, no, what are you going to use the summary for? like you have to do something with it, right? It's not just you've got the learning for free. Um Devin says realism is something we focused on when we design our process for this is for interviewing. So build based on requirements offline then we paired to add uh a surprise requirement emphasis on communication. I love that. Um we tried before Microsoft when I was at a startup um I don't know where they landed on this eventually but we were triing doing like pair programming. Um, it's a lot of effort on the interviewers, which is I think why people were shying away from it. They're like, "Oh, we need a scenario and it can't be like too well understood because then it's hard for the interviewers to not give it away." In my head, I'm like, "I don't know, man. Like, just make it I don't care if the interviewers have to try to not give it away. Like, just have something that people can navigate through together." I really like the idea. Infected FPS says, "Commenting between um World of Warcraft raid polls. My company has a kind of AI mandate. Uh they want us to use it, but the more senior folks like me don't have a ton of use. Interesting. Okay. Could uh Shin says, "Could seeing how people use AI tools also be indicative of how people think through new problems?" Yes. Uh just as we evaluate how people handle problems they previously faced. So Shin, I really like this thought, right? Um, I've been, if folks watch Code Commute, um, I know Devon does for sure, infected FPS does. Uh, I'm pretty sure infected FPS, you're from Code Commute or maybe just these live streams. I can't remember anymore. I know you've been here for a while. Um, but one of the things I try to say on Code Commute a lot is that I like to challenge how I'm thinking about things and and take different angles. And Shin, to your point, I really like this idea because a lot of us are, especially if we've been in the industry for a while, we're going, "Oh, the the new developers are doing it all wrong. They're just using the AI tools and they don't have the fundamentals and they're going to be screwed." And I think for the first time this week, I saw someone post on LinkedIn going, I wonder I wonder if we're going to see in like 5 10 years. I can't remember exactly how they said it. I wonder if we're going to see that those individuals that have just been using the AI tools are actually more effective. And I don't know, like it's weird. I don't want to think that way because it seems wrong. But I've shared this on code commute where I'm like I'm open to it cuz it doesn't bother me. It seems weird but like it's not going to affect me in a negative way. So I'm very curious to see if that's the direction we go. Um people are lazy and want to vibe code but it's so powerful uh if used to actually learn. Yeah. Um Devon says it is more effort but it turns out that we struggled to get people even to that point. Yeah. The code we did get would have been a hundred times 100% better if if they had vibed it. Interesting. Okay, cool. Okay, let me thanks for all the comments, folks. If you have more thoughts and stuff, um I will be, you know, scanning the chat. I try to make sure I can get some some thinking and some words out and then uh get back to the chat. So, I kind of shared the the premise of what this discussion was about. Um and basically when the person was asking me, has my opinion on this changed in six months? It's like not really. Um, and the the reason I say that is because they're they were asking like is everyone just uh you know using AI tools and not learning anything? And I was like that even that hasn't even really been my experience. Like I've worked with I have junior developers on my team at work. I don't know like they're not like using AI so heavily that they have no idea what's going on. like I think they're doing a tremendous job and like that's just not the case. So it hasn't been my lived experience and so that's part of it but the other part of it is that everything that I was sharing on it before just has not really changed. I still think that there's effective ways to go use the tooling which is what I'll be talking about through this. And the one of the meta points that we're getting into this and some of the reason I've been uh picking out some of the chat messages is like this person when they had commented to me they said I don't know what to ask the LLM and it sucks or I can't remember exactly how they said it but it sucks because it's like it doesn't tell me what questions to ask it. There's a there's a simple answer I think to this one. Um Jonathan thank you so much. Using AI to study for the AWS AI exam show me hallucinations can happen when it has the wrong sample space. Uh can you share a work example of yours? Uh Jonathan, uh can you clarify what you mean by that? By thanks again by the way. Really appreciate it. But uh can you share a work example of yours? What do you The answer is probably. Uh I just want to know what you mean and then I can try to to see uh how I can help on that. So, uh, the great W says, "I see it as a quick reference guide. If I know what I'm trying to do but can't think of the syntax, AI is a quick way to find the answer." Absolutely. Um, lots of different ways to use it, right? And I think one of the things that we keep having people focus on is just like dump the code out, right? Like, I don't know what's going on. AI, just make more code. Make more text for me. Give me the text. Let me copy paste it. more text and like that's I feel like that's where people are probably going pretty wrong with this stuff again m maybe not maybe not in we fast forward 10 years in the future and the people that were just like I'm copy pasting maybe they get so effective at being able to build software just by doing that that that everything I've been kind of saying is wrong and that's cool I'm happy to be proven wrong it's okay I've already gone through learning for software development. I'm at where I'm at. So, I'm going to be augmenting myself with AI tooling, but you know, I'm not starting from scratch and trying to to build with AI. So, um, one of the things that I really like, and I called this out before when I was responding in this original video, and I'll I keep saying it, but like one of the things that I really like about AI tooling, is that it lowers the barrier to entry. And what I don't mean is that it lowers the barrier to having amazing code and amazing products and you can have a billion-dollar product overnight. Not like that. I just mean that I've spent a large portion of my career, right? So, I would say like over the last decade trying to help people either that are aspiring developers or they're junior developers and I'm trying to ramp them up. I've spent a large part of my career trying to help people be better software developers. And that includes a lot of people that are new wanting to get into the industry and being like, but you know what? Like, no, I'm not smart enough. or no, I don't know math or no, like it's it's going to be too complicated. And it like it bothers me. Um not like not like personally, but like it bothers me that people have this lived experience where they they talk themselves out of it without even having tried it. They're just like I have this perception that I must be too dumb to be a programmer. And that's like that bothers me cuz I'm like if you haven't tried it, I don't know like as an example, I used to be very good at math as a kid. I was very good at math in high school. I was terrible at math in university. I don't I don't perceive how I operate and program and stuff like as anything to do with math. I know tons of people are like, "Yes, you can boil the syntax down to mathematical proofs and that's why it's math." I don't care. I don't think that I'm doing math when I'm programming. It doesn't feel that way. So, I don't I don't relate those things. And I think that there's a lot of people that just have in their mind that they must not be smart enough. And what I really like about AI tooling is that it gets them past that hump where they're like, "Oh man, like this made code." Like I I did a thing and it made code. they're they're shortcutting their first step to being able to have something that is code in front of them. And to me, I really like that. It's like just that barrier to entry part being reduced is to me a huge step forward. Again, I'm not saying it makes good code or that's the best way to do it or it's always bad code. Forget that. Just the barrier to entry part is really cool to me. Christopher says, "When I started, I had a hundred folders with functions because I was asking for full programs. Now I work with more precision, one function at a time and in a separate reusable scripts, one project folder." Cool. Um, yeah. So, you start to see how some of these pieces come together, right? So, you do get to learn through repetition. Like, you know, you're having this experience where it's outputting stuff and you're like, "Cool, I got all these things. I got all these pieces. How do I what do I do with them? How do I how do I reuse them? How do I, you know? So, I think you're you're getting to learn like I like how you said now I work with more precision, right? Like you're addressing smaller pieces at a time because through iteration you're finding this is a more effective way to go. Okay. Jonathan, uh, for example, I feel AI is not programmed to say I don't know enough. Uh, it usually provides answers to me instead of admitting it cannot. When would a junior not know to trust the AI output? Aha, yes, great question. So, um, I don't think uh, okay, what's a good way to say this? Uh, this is going to be a I haven't really thought through this, so I'm going to try saying this and it might come out totally weird and awkward. This is very much like a a code commute type of situation where I'm like, "Hold on, people. This is the first time I'm trying this out. Bear with me. Um, okay. So, when would a junior not know to trust the AI output? Okay. Um, if you are Okay, I want to I want to replace the AI with a human. This is the scenario I'm trying to come to, but I I want to think of a good example that's going to land. I was going to say like, okay, I'm I'm telling you the output for something. Like, what makes you think that you can trust me? And then I was like, h the fact that you're watching a video on my channel is probably some level some level of trust, I would hope. Um, but like I guess my my thought process here is like how does how does a human know to trust anyone, right? We want to trust people on the team because they have seniority. Um, okay, maybe we can use another example. You have two juniors working on a team. How does a junior know to trust another junior on the team? Right? I'm trying to use an example where maybe the trust isn't built up. They don't have that the sort of the proof that oh this person has demonstrated that they are reliable and trustworthy. When I say trustworthy, I don't mean like the opposite is being malicious. I just mean like I can trust the output. Uh and I through this thought process is like when would a junior not know to not Oh man, there's too many negatives in that Jonathan. When would a junior not know to not trust the AI output? not um my my whole thought process around this is that I would opt for not trusting ever which is kind of weird. I don't mean that necessarily with your colleagues, but I think that when you're in a state of mind where you're questioning things, it it doesn't mean that you're like, oh, like you're wrong, you're wrong, you must be wrong. But it puts you into this state of mind where you're questioning, right? So when you get code from the LLM, instead of going, the LLM made it, it's magic, it must be right, always be thinking like, well, what about this code? Right. And I think that the challenge is that when people are very new to coding, they don't know what to ask it. So I think Jonathan, you're spot on when saying like when would a junior like for them, how do they know? But if their default is to not trust it and then they have to get into questioning, this is sort of the framing that I'd like people to start having more of. Um but then the challenge is like, okay, well, how do you start to question it? Right? you don't if you don't know about programming. I'm assuming that for many people that are on this stream right now or watching the recording have some amount of experience programming maybe not you know maybe someone's like I've never written a line of code in my life and this sounds interesting. So, you know, if you're someone that falls into that category is like, yes, Chat GBT, it can spit out code for you. So, you can trust it, but like you can also ask Chatbt to go write you an article and site sources and it will sometimes just make up crap confidently. It'll give you a source. It'll say, "Yep, this website and on this date, you know, 387 people walked, you know, 10 steps to the north and in this city at this location." And then you go, you're like, "Oh, and there's a citation." And you click it and it's like page not found or something that it doesn't even mention it. So code aside, it's like just for LLM output like I even with citations I start defaulting to let me not trust it. I need to go validate. I need to go understand. Um Andreas, I'm a math major working as a software engineer. Nice. Okay. It's all about being persistent. Exactly. Math. Oh man, how do you do it? Um and by the way, Jonathan, I don't know if that helps answer your question. And I'm going to try to get into more of this topic as we go through. Um, Shin says, "Removing trust is a factor. Wouldn't it be cross referencing the solution with another person AI tool?" Yeah, kind of like before I try this out, does it sound good or any gaps you might see? Shin 100% with this uh this thinking at least from my perspective. This is one of the things I want to talk through today is like like I think what when this person wrote me the comment and they were giving their experience of like I don't know what I'm doing but it just seems to work and I don't know what to ask it. Um this kind of thing like before I try this code out saying to the LLM like does it sound good or any gaps you might see. You can literally say this to the LLM. And spoiler alert, because I'll say this later in the article, but if you're if you don't know what to ask it, you could literally say to the LLM, "Hi, I'm trying to learn about programming and software development. You just gave me code. I would like to know what sorts of questions to ask you so that I can learn or so that I can understand this code more effectively." It's going to give you something. Are those the best questions to ask? I don't know. Probably not. How do you quantify best? Are they better than zero questions to ask? I would say almost certainly. So, like start with that. This is why when the person wrote me and said, "And the LLM doesn't tell me what questions to ask." I was like, "You could ask it, man." Like, it's a it's a text machine. You just you ask it things and it gives you text. So ask it the thing. It's very meta, right? Like ask it how to ask it and what to ask it. But Shin along that train of thought is kind of what I'm thinking. Um Epic Technav, good to see you. Welcome back to the stream. Um Andreas says, "Having the tenacity to keep learning and improving." Yeah, for I can't get motivated in math to do that. So really awesome that you are a math major and you were able to do that and I'm hoping that you find that kind of experience in uh software engineering. Devon, question one, does it work? Question two, does it work when you uh does it work when you get more users than your mom asked? Does it work fast enough? Yeah. Um yeah, like you have to you go through this scaling type of uh process, right? Christopher says, "I don't trust any output and it was a gamecher once I could start reading the code myself. I can now better spot and correct errors myself. Yeah. And so this is an interesting thing, right? To the point I was saying earlier about do individuals that are focusing on prompting a lot and by the way Christopher I'm not trying to um as I talk through this I don't mean to like categorize you in one way. I don't I don't know your whole experience. My my thinking here is like do individuals that focus a lot on, you know, prompting, getting text, they see enough scenarios where they're like, man, like that didn't work. And I'm telling it, you know, fix this bug. And they see enough of these examples where they're like, okay, like, you know, I've had to do this type of thing like 20 times now. I can see that it just output code and I've seen it have to fix this thing like seven times before. Like maybe I should tell it. Like that's not the thing. And I'm curious if over time people see enough patterns this way. I'm I'm wondering right like genuinely I don't think we're going to have the answer for a long time. I'm wondering if long-term like that just becomes a really effective way to program. I don't I don't want to believe it, but I'm I'm very open to being wrong on it. Peter K, thanks for joining in from New York. Great to have you here. So, back to the article. Last thing we're talking about was low barrier to entry, but the the next part was that this individual was saying, and I the the comment is funny, not funny like at them, but it was like funny to read through because they were saying, "If only I knew how it all works, right?" And to me it was funny because a lot of people are talking about building software like exclusively using AI being like I don't even know what's going on but like I you know built seven iPhone apps over the I'm getting ads now that's like like a woman on Instagram that's like sitting back drinking her coffee and she's like I built seven iPhone apps this weekend. I'm like no you didn't. You didn't do that. It's just you didn't. I'm sorry. And if you did, they're not actually like, what kind of app are you going to build seven of in a in a weekend? I don't know. Unless you're like setting like a clock, midnight on Saturday morning, boom, and you're just like grinding stuff out. It's just not it's not happening. Um, not to any realistic degree at least. But we keep seeing these things come up in social media where people are talking about I don't even know what's going on and I built an app and I have a website and like part of me is like I'm so happy for you that you got through that barrier of like I don't know how to start but the other part of me is like do you know how it works? Because as this person was illustrating I'm not sharing my screen either. Maybe it would help for some of this too. I never know what people prefer. If you want to see my big dumb head or you want to see more of the article, we'll switch over to the article for a little bit. But um right, if only I knew how it all works. This person had spent time building something and they were just like, "Man, it would have actually helped if I knew what was going on." Like, yeah, I agree. I definitely think that it would. Um yeah. And then infected FPS in the chat says, "I built an app in two days and you try it and it breaks on edge cases." It breaks on the happy path. It breaks when it's not running as local host. Like again, I'm not I'm not saying that you can't How many negatives are here? I'm not saying that you can't build things with AI or be AI first. I'm not saying that. I'm just saying that I think there's a lot of exaggeration going on. So, um, if you knew how it works, right? Debug issues when something breaks. Uhhuh. Like you can actually think through this stuff. Understand why something works the way it does. Ask the right questions to improve your code or architecture. But this kind of leads into this next question, right? Because ask the right questions. Back to what Shin was saying in the chat, right? So in the chat, if I scroll back up on my or it's right here, sorry. um kind of like he said or he or she sorry uh kind of like before I try this out does it sound good or any gaps you might see this is something that you would do with a human even right it doesn't mean like you don't have to use an LLM for this and this is why I think it's really cool is if you had someone dropped code in front of you okay someone drops code in front of you regardless of whether it's a principal engineer or a junior dev on your team or AI. It seems like a good question to ask might be what Shin said before I try this out. Does it sound good? Or any gaps you might see. I don't know. Maybe the principal engineer was like, I just whipped this up super quick and I haven't even tried compiling it. Here you go. Good luck. Maybe the junior engineer, the one you may not want to trust because they might not be as reliable with good code. Maybe they're like, "Dude, I spent two months building this. I know it compiles. I've tried every edge case I possibly can. Like these are questions you'd probably want to ask. The LLM sometimes is just going to be confident. And other times, I've seen it if you ask it, it's like, oh yes, now that you've mentioned it, I don't actually know what I'm doing. Which is why I try to default to just don't trust it. Or, and I got to make a video on this. I'm probably going to do it on I'll do like a recorded video, I think. But I'm gonna talk about some of my AI usage recently, especially with GitHub Copilot with the pull requests. Game changer for me because now I'm like, here's something ridiculous. Go build it. And I'm like, I know it's not going to work. I don't care. It's totally fine. I know it's not going to work, but you just saved me a really big portion of time to go like scaffold a bunch of different pieces. Thanks so much for starting the PR for me. I'll take it from here. So, I really lean hard now into like do not trust the LLM. Okay. So, they had said and great great wording, right? They said, "As a complete ignorant, I couldn't ask the right questions and the LLM didn't suggest them." Right? So, what do we do here? LLM is not suggesting the questions. What do we think we should do? If you don't know something, what are you doing already? You ask the LLM. Hey chat GPT, this I told you I'm going to repeat this, right? So here's the repeat. Hey chat GPT, thank you for the code. I guess we don't say thank you anymore because that's killing the environment. So chat GPT, I'm trying to learn how to program. You gave me code. What questions should I be asking you to make sure that I learn and understand the code? You don't know what to ask? Ask it what to ask. Start there. Uh Devin says, "Built an app in two days. Please stop everybody. You're breaking it and I don't know how to fix it from a Twitter thread." Yeah, see a bunch of these on Twitter. Actually, I don't see the original threads. I see the screenshots of the threads. Um but yeah, it's like I'm actually fascinated like how are people getting like so much usage so quickly? Like there's something there I want to learn from them. Like so I've been building brand ghost for a while now. I'm not a savvy marketer, right? I'm bringing it up on my my live streams like, "Hey, you guys should check out Brand Ghost." I I have to do like cold uh not cold calls, but like cold outreach or like engagement on social media. I don't feel comfortable being a saleserson. I really have to work at it. But I see these stories of people being like, "I made a website and and the problem I had was that I had too many users." even if it's not like, you know, a million users or something. I'm like, you got a lot more users in the first day than I did. Sucks that your site can't stand up. Like, sucks that everyone stole your your secrets, but you got more users than I did very fast. Like, that's pretty cool. So, kind of interesting. Peter says, "Vibe vibe code is not production ready yet. I just haven't seen it yet and I've seen a lot of it. It's good for proof of concepts." Yes, I agree with that, Peter. with I'm just going to add this without getting into the details on this stream because my experience with GitHub copilot like in GitHub with uh issues and pull requests has been so much better than trying to use agents in the IDE. I can't even explain it. Um, I don't want to pull up my repositories and stuff, but like I what's a good way to like I don't have the stats, but almost every single thing that I've built in the past two weeks has been mostly AI like as in it started it. I still have to go in there and do stuff, but like it is absolutely building a large portion of what I'm trying to get at. And one of the reasons this works really well for me at least is because I have an idea in my head what I want to build, but I don't have all the time in the world. So, you know what's really cool? I put a bunch of issues into GitHub and I go to bed and I wake up and now I have code to start with. So, anyway, I'll talk about that more later in a different video. But, um, I think that that's been really helpful from like I guess it's not a vibe code kind of thing, but it's definitely agent mode and blasting out lines of code. So, kind of interesting. Um, yeah, Andre said, "If you built an app like that last time, you would be doing uh the last time you would be doing is bragging." Yeah, you'll be monetizing and scaling. Yeah, Devin, thanks for joining. The last thing that's okay, sorry, typo there. Um, do uh Experto says, "Do you see a decline in demand for software engineering since AI increases the output of a development team?" I absolutely don't. Um, I've done a couple videos on this on Code Commute. So, there was a little a few months back now, maybe it's even longer than that now, but there was like the Mark Zuckerberg interview and people were freaking out about it. And when I reviewed it and I was like, he's not saying he's he doesn't say like, oh, developers are going away. He talks about more code being written by AI. I even just told you in my own project over the last two weeks, more code is being written by AI. I still have to go do it. I didn't like get rid of myself, but um I need to do a follow-up video. The the CEO of Google um I can't I don't know if I've ever said his name out loud and I don't want to try and butcher it. Uh Sundar and I don't know how to say his last name. Uh, sorry, Bachai. So, sorry if that's really far off. Um, you can hear my Canadian coming out, but um, he was doing an interview and I think it was with Lex Freriedman. I wrote it on my whiteboard that I have to go make a video on this. That's why I'm looking over there. And um he even said that when they're I'm going to paraphrase this, sorry if I get this wrong, but he said something along the lines of they're trying to quantify instead of counting like how many lines of code the AI is doing. He's like that's a pretty crappy way to like to measure the effectiveness. So he's like we're trying and it's really difficult. We're trying to measure our productivity increases as software developers. And I think he said that they're observing roughly a 10% increase in productivity. Okay. Like forget the lines of code. Yeah. The Google CEO uh the great W. Um I think his name that's it is him, right? Yes. Sundar Pachai. Oh, every time I say it I'm like I got to be doing that wrong. Um so in his interview he said and we're hiring software engineers. He's like, "We're hiring more software engineers next year." So, he talks about being able to like it's a it's a a productivity booster. So, he's like, "We're not getting rid of the developers, we're boosting their productivity with it." Um, yeah. Right. So, Christopher, this is tool for developers, not a replacement for developers. The great wall. Yeah, you saw the clip. Awesome. So, I'm going to do a video response to that. I have on call this week for work, so I'm going to be a little busier than I want to be. But hopefully when the weekend comes, I'll pick back up with some YouTube videos. I want to do a response on that one. Um because that's the first time, at least that I've seen uh like an a tech executive coming out and being very clear about that. Even the Mark Zuckerberg one, I was like, he's not saying get rid of developers. He went on to say like could it change what job roles and stuff look like? Absolutely. He's like I don't have a crystal ball though. But he was saying like every time in history when we have a big technology shift he's like people freak out about this. Yes. Some types of jobs go away and then tons of other new jobs get created. So Mark, in my opinion, Mark was very much talking about like a like a shift in workforce, but not like a by all software engineers, see you later. Um, but this was the first time that I heard a tech exec say like, you know, this is augmentation. So I thought that was pretty cool and explicitly saying we are hiring more software engineers. Um, okay, let's jump back into this. Um, so I just had a couple more things on here like about not knowing what to ask, right? So you could you can find ways to ask the LLMs what to ask. I think that's an easy win. You're already asking it to make some code. Um, I do think and this is kind of what I talked about in the original video. Um, and why I'm saying I'm happy to be proven wrong in the future on this, but my take is still that if you are building software leveraging AI tools that if you are learning how to do fundamentals or as we were kind of talking about in the chat, if you're learning about what's going on, trying to understand things, I think that becomes your superpower. Instead, if you're just being like kind of naive to it, kind of ignorant to it, being like, I actually don't care about the software development part, I'm just using this thing that spits out code and makes the web page I want. If you're only doing that, I feel like that's going to be holding you back. I just I don't know. I don't suspect anyone that's actually I can't say like zero people. I suspect there is a small minority of people that are interested in software development that are only copy pasting text out of the LLM. I and and don't care to get better. I feel like as software developers, most of us are very inclined to like I would like to learn more and do this better. So, I I do think that if you're taking that opportunity to learn, ask questions, pause, reflect, um I I think if you're doing that, that will make you better than someone who is just not. And I honestly think that if you're using AI tools to augment that learning process, using AI may make you better than someone who's not using it either. We already saw Christopher I think it was you in the chat that was saying I don't want to scroll all the way back up now but that you've said it's been an invaluable teacher right I'm trying to think like it's been a long time since I had to sit down and go like I have absolutely no idea what I'm doing in this programming topic um you know go on the internet you have to know the keywords to search for we don't have to know that anymore you can kind of just like guess at what to say to an LLM and it will it will start to guide you down a path. Right path or not, different story, but it starts getting you moving. Whereas, if you can't find the right keyword to go look up, Google or Bing, they're not going to tell you anything that you care about unless it's like a typo or something and they can correct you. But concepts, good luck. You could describe a concept. You could say like I'm building a website and you gave me code that loads database or sorry loads records from the database to the website. It's slow. How can I make it go faster? Right? You don't have to go oh like Google search database indexes or uh stack overflow how to do caching. You don't have to know that. You just ask it like hey man it's slow. like teach me teach me about software design and like how to make things faster and how to approach debugging this like you have that that power at your fingertips which I think is super cool. Um okay, what else we got in the chat here? So um Experto, yeah, no problem on the question. Sorry if it's super specific, but if you were a beginner, what would you learn? Uh what does what would you learn for WA? I don't know what WA means. Uh, what would you learn for WA in terms of financial sector, C#, C++, or Java? Um, I'm super biased, right? Um, I'm going to say C. I I don't think you'll have a there's no issue if you learn any of those. If you learn any one of those, no issue. Um, on the C side, like it's kind of funny because I hear people historically saying a lot of C in like in industries like like finance or banking, that kind of stuff. If you're doing like uh super high-speed stuff, um, they might say like we expect C++. I mean there's going to be people will always say like use this language because it's more performant but like I I rarely think that your language is going to be the bottleneck um even for some pretty extreme systems like we have for context our routing at Microsoft for the team that I'm on we do like you know trillions of requests per day we have C# code oh we also have C++ code right but um point is like super high performance. We have C# stuff. It's not like not going to matter. But um I I do think like any one of those options would work. Um personally, like aside from my bias for C because I like it, I feel like Java and C are easier to learn. The reason I say that is because I've seen a lot of people stumble through C++, including myself. I don't really like I don't find the syntax as obvious. personal preference though, right? Other people might say you're nuts. Like definitely C++ is way more nice than C or Java. I don't know, but any one of those is fine. Um, thanks Andreas. I appreciate that. Um, Christopher, I needed to find an unknown value. AI helped me build a binary search function that did the algebra. That was an awesome learning experience. There you go. Right. I think that's sweet. Architect, welcome. Uh, I love using LM to draft code proof of concept, not to produce the final version. Yeah, that's been more of my flow recently. Um, I have a couple of things that are pretty targeted and I can make pull requests for and it either gets it perfectly or like very close. Like I pull it down and try it and I'm like, "Oh, it didn't compile because uh number one, I didn't get I didn't run the job and let it get the the feedback for it." So, like, okay, it missed a using statement and then I try it and it works perfectly. Or some of the front-end changes I've been doing, they just work. And it's funny because how I'm just I'm like, "Hey, there's a bug here." And like I'm just going to try and describe it to you and like I don't like working in the front end, so I can't really explain like effectively the part of the code and it just like does it and it works. So, it's super cool. Um, I have I have a bunch of different experiences now. Some of them that are terrible and some of them that are very good, but more and more leaning into um Oh, export of Washington State. Okay, one sec. Uh, more and more leaning into like this proof of concept kind of thing, but yeah, uh, Washington State. Uh, I don't know if that that changes anything. Uh, I'm in Washington, so sweet to see Washingtonians. I don't know. I'm actually from Canada originally. So, um, Torononian, we'll go with that. OS advertising. Is Blazer viable or does it cause performance issues? Uh, how about JS Blazer drop? Um, I don't depends what you're trying to build, right? Like, uh, I think Blazer is totally fine. I think it's totally viable. I think you can have a lot of success with it. If you go reading online, someone will say, "Oh, it's it's bloated." Or someone will say, "Oh, it's slow." or and then you read someone else being like it's lightning fast or it size isn't an issue. I personally my websites in Blazer seems fine. Um it's not doing anything crazy but like the back end is like it's basically ASP.NET core. So like what what are we talking about for slow like slow render performance? Um I think it depends on a lot of things when we want to quantify the performance part. My my take is like unless you're doing something that is so on the edge of needing like crazy performance, most tech we see today is just not going to be your issue. Like probably if you're hitting an issue, it's just because you're not using it properly. And I don't mean that like personally to you. I just mean stuff is really fast now and you can write crappy stuff and it's still pretty fast. You could use an ineffective tool for the job and it's going to be fast. Personally, like I know there's going to be people that hate what I'm about to say. Like I would not use Python for a back-end server. I just don't think it's a good choice. I think there's better options. You could absolutely write a server in Python and it's going to be incredibly performant. Right? I have biases against it because I don't think it's a good fit. People will come along and say, "Nope, you're wrong." same way like I would be like I wouldn't want to build a website in Rust and people are doing it. They're like it's great it's the best thing ever. Just not what I would want to do. I'd want to use these familiar frameworks that have all of this support. It makes building much faster. So I think when it comes to performance optimizations like honestly the the tech is probably fine. Give you an example has nothing to do with what you're talking about but it's related. Um, we when I was at a digital forensics company, our database that we would use for putting all of our records into was a SQLite database. Now, hear me out. You would think that a SQLite database is a terrible choice for a database, right? For context, too, we weren't doing central case management at the time. I think they've long since moved to these other solutions. But people in forensics, they want their case. They want to, you know, contain their data for their case so that they can go archive it, they can go move it, whatever. It's self-contained. And people were like, "Oh, SQLite database, this is such a bad idea." SQLite is like actually like stupidly effective and it's used on like every device in the world. So, I think there's a lot of us that have these biases against different technology because it seems like not the right thing. Like SQLite database is like just a file on your computer. It's not some like server you have to stand up in the cloud. How could it be good? It's pretty good. So, it's all going to be um sort of based on your use case, but I just don't think it's going to be that big of a deal. Um best viewers, let's get that blocked right now. Cool. Um architect really hate my coworker who always makes some crappy long huge function not even readable code from AI and then tells me to go follow up. Yep, there you go. Um I love Java. It's a powerhouse. You can literally build anything with it. Except if you use C, you can do it even better. I'm just kidding. But but seriously, uh my new favorite, this is from Infected FPS. My new favorite use case is having GitHub Copilot analyze the project and write a C-pilot instructions file for the project for itself to use. Okay, infected FPS. I don't know if I wanna like you're making me want to switch over to my repo and I'm just gonna tell you guys I got to make a video on this. Okay, I know that I'm excited about this stuff because the other day I went to go film a code commute episode in the car and I was talking about some of the AI things that I've been doing and I like I it felt like I just teleported to work and I was like like oh like all that time disappeared. One of the coolest things that I've recently done so I don't write documentation for my own projects. I'm I live in the project. I don't need I don't need to also maintain documentation alongside it because like I'm working in there. So we have this thing called AI and it does writing. So I started trying to use it a little bit more for generating documentation not for me. So when I am getting the AI to do things for me and I go, hey man, like why did you do that? I keep telling you to not do that. I start to say to it, please fix that and tell me how I should prompt you to not do that. Tell me how I should update the documentation for this project so that a human or an LLM would not make this mistake next time and can follow the patterns going forward. And then it does it. I can even say go update the documentation yourself to go do it and it does it. But the really exciting use case for me was when I was transitioning from going to like, hey, build this little feature, build this more complicated feature, do some refactoring, trying these scenarios out with GitHub Copilot. This is all in GitHub, by the way, not like agent mode in the IDE. Um, I was trying these different scenarios out, making them more complicated, and then I was like, wait a second, it seems to be doing a pretty good job refactoring and going across my entire codebase. Like, it seems that it can do that. So, let's try something. And I said, so in brand ghost, I've been building this for a while now. In any software I build, over time, there's architectural drift. And that's because if I start something one year from that time, I can guarantee you I have found better ways to do things, patterns that I like using better, patterns I hate using, all sorts of stuff. But what I'm not doing, cuz I would like to keep moving forward, is going back and addressing every single pattern from the past. I just know if I'm adding something new, I'm like, "Okay, the last time this was the pattern we wanted to use. Let's go do it. Periodically, if I'm going back and fixing things, I'll go update the pattern. But this results in architectural drift. And I'm using the word architectural. Could be pattern drift. It could be naming convention drift. I'm just going to use the word architectural to encompass all of it. I said, go analyze the codebase. I want you to go describe the top five areas that demonstrate architectural drift across the codebase. And it did it. It did it very well right. Are there more areas? Probably. But it called out a bunch of different spots. Now I have this list of architectural drift. You know what I can do with that? Because I don't have time to go fix it and I don't really give a crap because I have better stuff to do is I can go file an issue with the text that it output. Put it into the issue and say co-pilot, go fix this. go fix this architectural drift. You've already documented it. Submit. See you later. I'll see you in the morning when I wake up and I can go review this code. So, that's been pretty cool. I think in my list of five over the past two weeks, I've addressed two of the five doing that. Um, the only reason I haven't touched some of the others is because I'm building features that it's going to conflict with and I'm like, I just don't want to I don't want to muddy the water right now. So, what I've been thinking about more recently is like how can I periodically like um there's there's so many more examples, but like how can I periodically whether it's like once a month, once every two weeks or something, go file an issue in GitHub and say go review the architectural drift and update the documentation to support it. Call out any new architectural drift. Uh refine what's already been captured. if architectural drift is going away. This type of thing I'm trying to explore and it's putting it into the documentation in the repository and I'm experimenting with seeing how well that helps. Like I said, I don't I don't give a crap about the documentation. I'm living in the code. But the LLM might it might do better if it has more documentation. So anyway, super stoked about using it that way. It's been I've been programming you know similar ways for a long time and this is the first that's not true. I was going to like using chat GBT to like debug stuff and like help design stuff has been pretty gamechanging and I'm finding now that my development is split between that where I'm hands-on working with like uh the LLM to like understand and design things and then in the background like I get an idea and I'm like oh let's just go fire this one off and it can go clean up the code base or what was a good one? Um, this one is a crappy example because it should do it well, but it needs to be able to get um the build output and I don't have it automatically building in uh in GitHub. But I said I have this situation where I picked the name for something and it's used everywhere. It's used everywhere in my codebase and I need to rename it. And it's not just like, oh, change like I could rename it in Visual Studio. It would address most of it. But now I have variables named really funny. They wouldn't match up. Now I'm going to have classes that also have like part of that name in it. They need to be changed. It's going to be more work. So I said, "Hey, I don't want to do this. co-pilot probably could, but um it seems like without the build output, it's just it stops partway and then it doesn't have feedback to check like can this compile. So anyway, so many cool things got me all derailed because it's so exciting. Um but yeah, OS advertising, simple cruddy ecom things. Does Blazer do the job if you were to compare to dedicated JS framework? Absolutely. Um, no issue. Personally, I think the the benefit you get with the JS stuff is like you're going to have a lot more support just from all sorts of different, you know, libraries and things like that. But like Blazer is not aiming to replace JavaScript. Like this is the thing where and and I'm not I'm not suggesting that you are saying that but um a lot of people online are like well why would you ever use Blazer if you have if you have React if you have Angular you have all these things um like Microsoft is not like hey we're going to make Blazer and we're going to take all the market share from JavaScript. No, not the plan. They're like, "Hey, we got people that build C and .NET and they really like using it and they don't have a really good front-end solution for that for web development. So, let's build something for those developers." Um, I've not used Django. Yeah, Copilot uh in in GitHub I think is super cool. It's 40 bucks a month though. So, uh, I so I, you know, I can expense it because it's for business, but, um, it's weird like I don't why Microsoft should give me that, right? What's up with that? Um, yeah. Is it Django or Django? Django. These are like words I'd ever talked to anyone in person about. Django. So, I've never had to say it out loud, right? I've seen it a million times. In my head, I just say Django. Yeah, no employee discount. Can you believe that? I even pay for my own Microsoft Office. Crazy, right? I get Visual Studio Enterprise for free, though. So, that's pretty cool. Um, did you just tell us how to say Django by spelling it out? That doesn't work that way. You need to phonetically spell it out. Um, PHP in 2025. Man, don't talk to me about PHP. We don't do PHP here. Uh I have a bias against it just because I think it's a I don't like it. Not for any good reason. I think that there is actually a lot of PHP developers. So um I would never use it. If given the choice, I would say no. But it's around. It's just the J sound. Okay. It's just Django. It's not DJ Ango. That one's a stretch. I know. Okay. Uh, I got my start in PHP. Yeah, sorry to hear about that, man. At least you're here now, right? Okay. Um, okay. So, let's wrap up the the article. I'm all over the place. Thanks for the tangent. I can't wait to make a video on that. I don't even know. Editor is going to be like, dude, what do you even want me to clip down from this video? And I'll be like, I don't know, just AI stuff. Um, so I think, you know, some some meta points, right? I just I've kind of said this stuff throughout the stream, but I I really feel like if people are trying, especially for beginners, right? This could apply to people that are more senior learning in a new area, too, but like slow down. I should be doing this when I'm having AI spit spit out frontend code for me or when I'm like, "Hey, it's not working. Like, fix it." And it just fixes it. I should be like, "What did you change?" Like, explain to me why. So, I think that's important. Um, if it's building systems for you, like ask it tradeoffs. Uh, the second question on here is just back to the point where someone's like, I don't even know what questions to ask it. Ask it what questions to ask it, right? Like it will tell you. It will give you a list. Um, clarifying concepts, right? if uh you see it doing some new code or in the output it's like we're going to use and it's talking about some caching pattern and you're like dude I have no idea what that means instead of being like no worries copy paste being like maybe it's a good opportunity for me to ask what that is why are we using this caching strategy what even is a caching strategy are there other caching strategies why would we use this caching strate strategy versus another one, right? Like just start going down that path and exploring stuff. The the rushing to try and get stuff made in my opinion is kind of it's missing the point. I think especially if uh if you're in the group of people where you're hearing like you should be building side projects, right? As a as a more junior developer, like I would say at any level to be honest, but you're like I got to build side projects. I need them for my portfolio. Like, need them on my resume. I think if you just had AI like spit out projects and you're like, "Cool, I can put it on my resume." In my mind, like, you just miss the entire point of the side project. For me personally, the point of the side project is that you're demonstrating that you're learning through a subject. To me, that's why I like seeing them. So, if you skip all of that just to be able to say, "Great, I have the output." what does that matter? Right? So, I definitely think, you know, ask it to clarify concepts. Make sure you understand them. Keep asking it till it's starting to make sense. Um, and then this last part, you know, asking why things failed. So, yes, you're going to be getting code that comes out of the LLM and it's going to be crap sometimes or it's good, but they missed something. Whatever it is, right? Ask why it's not working. Not just fix it. There's a lot of memes now which are pretty funny, right? It's like uh just examples of like someone being like please like one more time like just fix the ID or fix the build error in the IDE. Like you want to you want to move away from that and have a little bit more critical thinking because you have a tool that can explain that kind of stuff to you. Um thanks Christopher for being here. You probably already dropped off but I appreciate you being here. Um, oh, last two messages are reversed. Oh, okay. Let me let me read them in the other direction, I think. One sec. I'll read it and we'll see. Basically, what I'm just using JSON managing the same blazer wanting to run some JSON. I'm thinking about doing JS interrupt. Is that viable? Um, and then I see the other message here. Just need to reinitialize the JSP each time state ch. Yeah. So I don't I know you can do a lot of JS interrop in Blazer but I don't know the details around like how complex that gets. I've made a video on it. I think I think I made I made an article for sure like a year ago um that you can do it but I I don't have a lot of experience like building out a website that's heavily relying on on that type of interop. Like I don't know, man. Like I think like try it. I I don't Pardon me. My throat got really dry out of nowhere. I I don't think that you're just going to hit some, you know, crazy scale issue or something, some crazy performance issue. I would say try it because you can always try to optimize things. So if that's a pattern that you're like, I think that I want to do this. I think it's going to enable me to develop more effectively. I would say try it. Um, but yeah, you can absolutely do the interop. Like it absolutely works. There's lots of videos and stuff online. So, I don't know. I would say go do a prototype with it and see. But I think it's viable. Jonathan, sometimes slow down to build fast. Yes. Right. It's just a different way of doing that. It's slow down to build fast like from from a long-term perspective. So, um, how do we say this name? Zaka. Oh, Zaka. Zaka 6. Okay, I got it. I'm catching myself relying on AI for design. Checking more and more. Checking already written code seems like a fine usage without too much brain rot of the programmer. Yeah. So, um I think like I like this brain rot thing. I have a niece and nephew and they're one's 10. I think my nephew is seven or eight. And apparently there's this thing that's like the kids are into it. It's called Italian brain rot. And like my father's Italian and we were on not Skype anymore because Skype is RIP, but we were on a uh a Google Meets weekly meeting catching up with my family and my my niece and nephew were talking about Italian brain rot and we were like, "Pardon of me? What what what is what what's what's Italian? Like what are you talking about?" And Yeah. And then so it's a general slang for mindless activity. But but they were talking about Italian brain rot which is like apparently these specific memes where they have like an U usually it's like an animal like blended with like another object or something. It's super bizarre. And it's called Italian brain rot. And we were like there's no way this is a thing. And we we sat there for like 10 minutes just like looking up these Italian brain rot images. It's a stream. I can do this. Right. Let me Italian brain rot. You can tell like I'm not a good streamer because like I don't actually I like don't know how to do this stuff effectively. Give me one sec. Do not read this article if you're over six. You won't get it. Okay. One sec. Look at this. This is a thing. Yeah. Bomb bombard derico crocodilo. Yeah. And that was one of them that he said and we were like, "What are you talking about, man?" He I think the shark with the the shoes was one of the first ones. I've never seen the gator the crocodile bomb. That's That's Bombadro. No, I thought the Bombadro Crocodero thing was a like a fighter plane. Anyway, Italian brain rot. It's a thing, right? Crazy. Um but yeah, in terms of brain rot like for for programming, I think that you know it's easy to fall in that trap if you're just copy pasting. But I think there's such an opportunity to use a powerful tool. The LLM is a powerful tool. Sorry, I just realized on Substack, you have no idea what we were just looking at. Some Italian brain rot. It's it's a thing. I don't know. Crazy. But yeah, I think lots of opportunity to talk to the LLM to get it to answer questions and use that epic technav. I'm going to do a challenge soon where I build a simple app in my stack, Angular and Spring Boot with zero AI or internet help. What I'm hoping and see how far I can get. I I hope doing this sometimes will keep me sharp. Yeah, try it out. There's a bunch of people that are talking about this kind of thing. Um I don't know. For me, I think I'm kind of I don't I don't know if I need to prove that to myself. And I'm not saying that I disagree with your thought. I think definitely an interesting exercise. I think for where I'm at, I'm not I'm not too concerned if I start to lean more into AI. I don't I don't think I'm concerned about it, I think, is what I'll say. I don't know. Maybe maybe that's a naive way to look at it. In a few years, I'll be like, "Oops." But I'm trying to be aware of it. I'm trying to be I'm trying to challenge myself a little bit, right? Um instead of being like, "Oh, no. I don't want to lose my skills." I'm trying to be like, "Hey, like what if we try this? What if we try using the tool more and more?" So, I'm kind of curious, but we'll see. Zakazaga 6, I already stopped using autocomplete recently. We'll try no AI thing as well. Yeah, there's a um well, let me let me do the thing because I'll wrap up the stream now, but I'll try to pull up the video on my main channel. Um one sec. Let me see if I can find it first before I I got to refresh a bunch of stuff. And where is it? What did I call that video? One sec. It called illiterate. It's in the title. speaking of illiterate, I can't spell it. Yes. Okay. So, this video here, um Oh, you can't see the thumbnail for it. Darn. This video has a cool thumbnail. Um, this one actually talks about an article that someone wrote. And am I wearing Oh, man. That's embarrassing. I'm wearing the same shirt. Look at that. It's the same shirt. Is it or it's not? Same color. I swear I have more than two shirts. Um but yeah, they wrote a really cool article. Um they actually responded to the video, too. What? Or it was on a short. Anyway, um but yeah, I think it was Why is there No, there's 71 comments. Why are they not showing up? What the heck's going on? Yeah. Okay. There's a bunch of comments. Anyway, the person who wrote this article responded. Um, but they talked about in the article like that they're basically stopping using AI tools and stuff like that. New YouTube feature. They cut the comments to two. Worst feature ever. I'm sitting here like, am I going crazy? I'm positive that like someone someone wrote about this or wrote back on this. Also, another interesting thing and just to share this, but like YouTube doesn't notify me about all the comments. Um, which is nuts. So, when I talk about Brand Ghost in just a moment, the comment feature in Brand Ghost shows me more comments than YouTube Studio shows me, which is pretty crazy cuz we're not using magic. We're just using the APIs they have. Um, oh my god, it's your seion. Yeah, don't look at my shirt. Um, okay. So, let's go back to the Dev Leader channel. Cool. Okay, so for folks that are new here, uh I am signing off. So this is the newsletter that I write every week. This is where generally the topics come from for the live streams. So if you're on Substack, I'm just sharing the Substack link. You're already there. Um that's that. So this is the article, right? And this is if you want, you can watch the live stream directly on Substack. I just don't really recommend it because it's probably a crappy experience compared to being on YouTube. But um or Twitter, there's a lot of people like overwhelming majority of people are on Twitter. I don't like Twitter seems to be the spot to go watch stuff. I don't know. I have like my LinkedIn following is five times what it is compared to Twitter. And I think there's I don't even know if there's anyone on LinkedIn watching it right now. Like it's crazy. LinkedIn hates me. But anyway, um this is my main YouTube channel. It's called Dev Leader. So I have C programming tutorials. I have a podcast where I interview other software engineers. generally try to focus on their career progression because I think that what I've learned is that people really enjoy seeing other career paths, stuff that's not typical, right? I had, you know, C Microsoft celebrity Scott Hansselman. I realize probably not everyone knows him, but like if you're like a Microsoft developer, you know Scott Hanselman. So, he's pretty cool. He came on. Uh, I really enjoyed that conversation. There is a free resume review playlist and so the opportunity to submit your resume and have it reviewed. If you check out the resume review playlist, like there's a a video right here. Um, oh, I realized the Scott Handsomeman part was probably cut off because my stupid head was over it. Anyway, um, resume review playlist here. If you watch them, you can see how to submit. Um, I had first feedback where someone I I shared it out I think already, but someone responded to a comment and they're like, "By the way, like after you reviewed my resume, I had two internship interviews or they have them lined up." So, try my best to help. Um, you have a picture with Scott. That's awesome. I I've seen him in person, but I I didn't like run up to get a picture. So he pro he wouldn't recognize me cuz I he does like a million podcasts. But anyway, um Code Commute is my other YouTube channel. Um Code Commute is where all of the ideas start. So I film Code Commute mostly to and from work and then sometimes making faces in my office right here, but uh other times when I'm on vacation in Arizona, but mostly it's when I'm driving to and from work. open-ended software engineering questions. Uh the channel is primarily driven by questions that you submit. So if you want a topic answered, just go to a video and comment it on it or send a message to me. This is streaming to like eight platforms right now. So pick a platform, send me a message, and say that you want me to talk about it on Code Commute. I'm happy to do it, keep you anonymous that way. But uh this this channel's a lot of fun for me. It's got almost 300 videos in not too long. Growth really plummeted. Stopped getting a lot of views. The first videos are getting a lot of views and then maybe I got boring or something. I don't know. But that's fine. Uh architect, do you sometimes feel a bit burnt out or uneasy on Mondays and Tuesdays after the weekend returning to work? Kind of feeling that now. What helps you cope with that? I do all of the time. Um, I think that there's probably a bunch of bunch of different things to look at here. I think that at different points in my career when I was just working a lot didn't matter. Um, I think now like this week, oh, just to give you an example, like this week I'm on call in the morning starting at 6:00 a.m. So going from a Sunday being like I have to wake up in the morning and the first thing I need to do is get to a computer, check things out, and then just be ready to be getting paged for incidents. Like, nope. That's not a good way to like, you know, to start a week. But generally I I kind of have this feeling um I notice it more and more and this is a big sweeping generalization when I have more things piling up and a lot of those things are not things I'm excited about which I think in any job that's you know bound to happen to varying degrees. Um my heart pumping but no one really yelling at me. Yeah. like, you know, I have a bunch of stuff right now that I either have to kick off, I have to track, or I have to drive to the next level of progression across like off the top of my head like probably five or so different work streams. And to start a Monday like that is like, you know, I'm not super excited. But what I noticed today when I wrapped up my on call shift, I was like because I had done that, I'm like, I'm powering through the stuff that like that I don't really want to do, right? And I I'm saying this transparently because there's always going to be work. I think it's ridiculous to expect that every piece of work in any company is going to be something that people are like, "Hell yeah, I can't wait to do this." No. Like there's stuff that I don't like to do, but I know it's part of the job. That's okay. Like I'm a grown man. I'm not very tall, but I have finished growing, I think. So, I'm a grown man. These are things that I could handle, and it's okay, right? So, um I just kind of acknowledge that like not every day is going to be awesome. I think it's a sign if every week, more and more days of those week, I'm weeks I'm like, ah, this doesn't feel good. Um, so I don't I don't think it's a thing to ignore, but uh I think that, you know, here and there or like for me like I don't like Mondays, but once Monday is done, I'm fine. I don't know. It's some I I don't think that's too unusual, but I've definitely had points in my career where I'm like, cool, it's Monday, like time to go work and like being energetic for it. So, don't know. I hope you find a good balance for you and uh you find something that's exciting. So this is code commute. I have a lot of fun doing code commute. I pray the Jesus prayer. It's a meditative type practice. It calms you down. Interesting. If you find things calm you down, I think that's good. Um, I should also mention because people were asking for this, I have to promote it more, but folks wanted Code Commute on Spotify. They said, "Put it there." And I said, "Guys, you don't understand. The YouTube part is easy. I film the videos. I upload them. I write a caption and like a, you know, a sentence, the title and like a sentence. Sorry for the the description, and I'm done, right? I want code to be low effort." And people were like, "No, but like we watch it in the background and then we don't want to keep our phones open while we're like cleaning and doing chores." And I said, "Okay, for all of you, because I love you, I will upload it to Spotify. I will do this thing for you." And no one's watching it on Spotify, and I'm uploading them. So, I got 273 videos right now for Code Commute. Actually, there's two more. Actually, one more. So, one for tomorrow. So, 274 videos on code commute. I've uploaded I don't know a bunch now. Um almost 100 or something. I got almost 200 to go. I'm trying to do 20 of them a week to catch up. If you if you enjoy Code Commute, I encourage you to try it out on Spotify and help me get those numbers up. But thanks, I would appreciate that a lot. Um, I will also mention that I have courses on C sharp and other topics related to software engineering on dome train. This is the thing that keeps all of my content creation actually alive in terms of being able to make YouTube videos. So, uh, I lose money on my YouTube videos. Um, on both channels, yes, on code commute, I lose money because driving to and from work in the fast lane. uh is not covered by code commute. And on dev leader, my videos cost money, which is fine. Um you know, my goal is to be able to put the content out and help people, but the courses are what supports everything else. So have courses on C. I have paired up with Ryan Murphy for uh career management courses, getting promoted, behavioral interviews, and soft skills. I have a couple more courses that are planned. Uh, let me rephrase that. I have a couple of courses that I know I want to build. That's what I'll say. And I've talked to Nick Chaps as high level about them. He's very excited. I just have to narrow them down a little bit more. Um, and I've just been in such a whirlwind of stuff to do that I need to get through like another week or two. I actually think after I'm done my on call ship this week, I I feel like I should be in a good spot to be like, "Okay, let's get some courses made." I haven't made courses in a little while, so I'm pumped for these ones. We'll see how it goes. But that's uh if you guys are interested, if you know people that are trying to get started in C and software development, there's lots of awesome courses on here by the way. Like I'm just showing you mine because I'm being selfish, but um Nick Chaps has done an outstanding job getting like software professionals to come to his platform to make courses. So lots of awesome people on here. By the way, the great w Thank you very much. Code commute is great and you are great. And finally, I'm just going to talk about Brand Ghost super quickly. Brandos is the social media content creation, cross-osting and scheduling platform that I am building. If you see my content online, uh it is basically done through Brand Ghost. The exception to that are basically this live stream, my newsletter article for now, and what else? um uploading full-length YouTube videos. If you see shorts, if you see a post on social media, it is guaranteed through Brand Ghost. So, it lets me reshare my content across all my platforms. This is basically started building this as soon as I started getting back into content creation. So, for those of you that don't know, I tried to be a content creator back in 2013. I was trying to learn in public. I was writing a a blog that was called Dev Leader and I wanted to share my experiences about being an a new engineering manager balancing development and leadership and dev leader was born. It's clever, right? Um and then after a couple months of writing a blog, I was like this sucks. No one looks at it. I keep sharing it everywhere. Not going anywhere. So I stopped. And then 10 years later, so 2023, I said, "Hey, why did I ever give up on that? That was really dumb." And I realized that I gave up on it because I just couldn't keep up with trying to share things. I know I need to share content online. I know it's a pain in the butt. I need a system to do it for me. Yes, there are lots of other systems that can do it, but I find that they all lack because I've tried them. So I said, what better way than to build your own and then to give away the crossosting and scheduling for free. So those platforms like Buffer that let you have a free trial or have a free account and you can only link up like three platforms. Nope, you can do them all and you can schedule with ours. There's no limit. If you abuse the system, I'll find you. But um yeah, it's it's totally for free. And then the more advanced features are paid for. If you're trying to run content like I do, those are definitely paid for features. But uh it is built in ASP.NET Core. The front is in Nex.js. Um I don't like touching the front end because it's not C, but it's a lot of fun. Uh I like reminding people. I've had a couple people now from these videos that have said, "Hey, I want to try posting more content online." I said, "Great. Try it out. It's free. Like you don't you don't put a credit card in anywhere. Just try it. There's nothing to lose. So this is Brand Ghost. I'll put that in the chat as well. If you're on Substack, you can't see what I'm talking about. Just brandost.ai. And totally free to try. Um, the other thing I'll mention is that if you are a small business and you need to do advertising and get active on social media, if you're finding that it feels like a distraction to be on social media because you'd rather be running your business or you're like, "Oh, I have to go hire someone to do all my content posting." You might not, right? Get the content made. You can add it into Brand Ghost. The biggest feature that I use about Brand Ghost is topic streams recurring scheduling, right? So, one sec. Let me just show you. Let me just show you my content calendar. I'm pulling it up in another tab. One moment. Okay, that's my content calendar. I'm going to pick the 17th. Okay, look how many posts. These are all the posts that are going out on the 17th. Oh, I realize this is like the worst spot. One sec. Here we go. See all these posts? This is just one day. This is my content calendar. I'm just going to show you something quickly. I'm going to go to July and just show you my content calendar. Um, it's a little slow just because I'm a bit of a power user. Um, there's my content calendar for July. You want to see my content for August? Guess how much of this I scheduled myself. I'll give you a wild guess. Right. 26th. There's all the posts that are going out. This day has a little bit less. Right. All of my content. I write my content. It goes into Brand Ghost and then it is perpetually scheduled. I have memes that go out Wednesday, Friday, Saturday, Sunday. I have 46 worth weeks worth of memes that you won't see them repeat for 46 weeks. I haven't I'm giving you guys secrets here. I haven't written a blog post aside from my newsletter article in over a year. Every single day I share out a link to a blog post because I have over 300 of them that I can reshare. My YouTube videos, someone pointed out to me in a message. They said, "By the way, I noticed your videos are from they're a little bit older." And I said, "Yeah, because when I put them into Brand Ghost, it's starting from two years ago and sharing them out and it will just post them every day automatically. I don't schedule anything. You can, right? You can schedule a post up here. I just don't do that because I don't need to. So, Brand Ghost is the tool I'm making. I'm super proud of it. If you are interested in posting any content, you have a business where you need to manage social media. Just reach out to me. Like I said, it's free to start. I keep saying this, but like Brand Ghost has to succeed for me even like I need it to work for me. I need it to work well or else I cannot create content. I'm extremely motivated for this to be a good thing and I'm hoping that if you have an interest in posting content online that I can be part of helping you with that. So if you're just getting started or you have a ton of content, reach out, try it out, let me know. So I think that's it folks. Thank you so much for being here and I will see you next week same time 700 p.m. Pacific. Watch code commute through the week and if you want to see what the live stream topic is, weekly.devleer.ca. Thanks so much. See you next time.