Vibe Coding Live Stream! (Spoiler Alert...)

Transcript is auto-generated and may contain errors.

Hey folks, I'm just driving home from the office on Monday. Um, was really cool. Got a couple folks in from my team that not yet got to meet in person because they live uh out of the country. So, super cool get to chat with them. Um, spending basically all week trying to make sure that I I have time with them. Uh, but this is my first day back from vacation or station, I guess. I was just at home. Um, so just like stupid amount to catch up on, but kind of expected that. Um, nothing like I don't know like life-threatening or like the most urgent, but just a lot to like to catch up on. So, um, just super busy day and I want to make sure that I can spend more time with them uh, specifically throughout the week. Um, so it'll be good. Uh, I am going to I didn't even get a like this is how I'm flustered right now.

I didn't get a chance to check out experience devs. I don't have any pending questions at the moment that I am aware of, but I didn't check uh social platforms today. So, I apologize. Just been too busy. Um, but I'm rushing home to get on the live stream that's on my main channel, which is Dev Leader. And, um, this live stream is going to be coding live, which is fun. Uh, coding live if you've never done it before is always a disaster. Just how things go. But, um, good segue after I have a sneeze that's about to come on at any moment. There it is. I knew it. Pardon me. Can't do anything about it. Um, this this channel is all about answering questions and stuff that people submit. So, if you have a question, leave it below. If it's on software engineering or career development, happy to try and answer.

or you can send them into dev leader which as I mentioned is my main YouTube channel. It's also my social media handle that I use on all platforms. But going forward after some point after today uh because I'm going to be coding live with AI the landing page for code commute. Um, after that's actually deployed, I'll have a landing page where you can submit your questions on codemute.com, which will be super cool and uh you can all uh test the website and break it for me. But what could go wrong, right? It's going to be vibe coded with AI. It's going to be bulletproof, right? Nothing could possibly go wrong, right? I'll just make sure that I don't have my uh you know my credentials or any keys like publicly exposed. No, I'm not going to in the live stream that I'm recording uh in 35 minutes.

I'm not going to be deploying it live. I'm just going to get it going running it locally. Um I'm not a huge fan of doing uh live coding that might involve cloud resources. It's just too easy to leak stuff. We weren't in sport mode and I was wondering why the car wasn't really moving. Um, but we'll get it to a point where it's good to go. So, I figured on this drive I was going to spend some time talking about my my strategy for how I plan to go build this, I guess. Um, and you can be my I don't know, my how do I even say this? like my right my rubber ducks because I have some ideas for how I'd like to approach it. But I think to start things off, by the way, there's going to be a a hidden reason I'm I'm walking through this.

Um, which I if I I don't want to forget as I'm recording this, but I'll mention it later on. So, my strategy is that I I want to come up with like like when I start working with the LLM, maybe I'll take a step back. I'm going to be using cursor, I'm going to be starting everything from scratch and um you know, kicking things off like in the cursor console like make a new project like let's let's literally go from scratch and see what it can do. If I'm if I'm getting flustered on the stream or it's not really going far, then I'll just I'll use Visual Studio. I'll create the um the project templates and stuff and then I'll go back to cursor to to have it go do what it's got to do. But um that's the framing. But what I want to be able to do is great cursor make the projects for me, but like what are we actually trying to accomplish?

Right? I I told you that what my goal is to have a landing page for code.com, but like what's that mean? What are we actually building, right? Like what what's the tech stack? So, uh I want to be able to kind of set that up up front and explain to the LLM like here's what I'd like to build. And for me, that's going to entail having an aspire. I want to use Aspire. So, the Aspire dashboard. Uh, I want to have a Blazer project. So, it'll be Aspire with Blazer. Uh, if you're not familiar with what e either of those things are, um, Aspire is a I don't even know the best way to characterize this. I'm primarily using it for the dashboard because I think the dashboard's super cool, but um it's almost like a coded driven. It's not okay. Instead of having like YAML files to go like orchestrate how you want to set up your containers and resources and stuff, it's code first.

So you can write C sharp code and Aspire will basically set up your um your sort of cloud infrastructure which is super cool. Um, one of the reasons that I I want to start this way and not deploy uh in the live stream is because I want to see what it comes up with and then I I want to basically have it set up in Azure for the first time using the Aspire uh entry point because I use Aspire for my other project that's a brand ghost but we we kind of we already deployed all the pieces manually and we switch to Aspire and so we get a cool dashboard and that's awesome but like we're missing out on this other like orchestration piece which is uh which is a shame. So I'm going to try that out for the first time with this. So it's going to be Aspire.

The Blazer part if you're not familiar with Blazer uh lets us C developers uh create frontends uh for for web apps and I don't spend enough time in Blazer. I'd love to spend more time there, but uh Brand Ghost, where I spend all of my time, has a React, uh front end. Um so it doesn't get a lot of practice, unfortunately. So this will be good for me, but Aspire and Blazer are two things I've used before, want to use more of, and I think that's going to be a good starting point. So be telling the LLM, this is the tech stack I want to use. Um, and the other part is that I'm probably going to ask it just to see um I'm going to need some type of storage, right? So, I'll be asking it like what do you recommend for storage and I'm going to tell it I will be deploying to Azure and see if it can come up with recommendations.

I already am pretty convinced I know what I'm going to use. Um either I haven't Azure must have Postgress so I might try Postgress cuz I when I've talked about this before I don't have a good reason I end up using uh my SQL for everything. No good reason just it's just the most comfortable thing for me to go do but literally no good reason. Um and so I should probably just use Postgress for this. I'm assuming Azure has that. So I should be fine there. And uh but I want to I want to ask the LLM like given some context um sorry we got to move over a bunch of lanes here. Given some context for what I want to go build like what do you recommend? Right. I am going to be pretty opinionated on this project. I already know what I want to do.

But uh this is a good exercise I think to be able to demonstrate that you can like you should in my opinion use things like LLMs to get feedback to to get ideas uh to get different perspective. I think that I feel like it's a really overlooked benefit. It's like we are always asking LLM's like write the code like do this specific thing and um you know it might give you wrong code or something but even even just having different perspective even if the perspective is wrong it might just get you to think about things in a different way which might sound kind of funny but I think that there's a huge value in just like not rushing into things and having this opportunity to have some different thoughts kind of come into your head. Like I said, I'm going to be pretty opinionated on this.

I have roughly an hour to try and get what I can done. So, we'll do that. Um, and then I I'll see. I want to I want to use agent mode, which will be pretty cool. I'm hoping that that gives me some opportunity in the live stream to, you know, get the prompts down and then while it's off doing agent things, I can talk. But I I'm fully anticipating It's not going to go well. I haven't had any luck with agent mode on anything. Uh I've already made videos about that. Nice. Okay. Um I just wanted to get to the fast lane cuz I'm going to be late for this live stream. Um, but I I want to I want to see what direction the LLM is going to go in for what parts to build first. So, again, before I actually send it off, like start doing your thing, I will be asking it like what what do you recommend uh in terms of like the first parts of this project to go build?

And the other thing that I've had, um, I've been seeing people suggest this a lot in terms of like refining your prompts. I'm not trying to speak about this like an expert, by the way, because I'm not and I'm trying to improve, but I've heard people say like based on what I provided you, like what other information do you need or or would you ask? And um that way it's giving you some feedback about like you know I don't know the way you've set up your prompts already. So I'd like to explore that and uh probably try to get in the habit of doing it. So once we have that and it starts off on a given path and it's like I'm making this up maybe it's like hey we should go build the two projects as placeholders then I'll scaffold the Aspire code uh and then we'll we'll get working on like a Blazer app or something.

I don't know what it's going to do, but I want to see and then kind of ride it out. Uh if it fails completely, like if it by the end of the hour, if it's just been going in circles and not getting anywhere, then I don't know. I I'll try to extract some lessons from that. I'm being totally transparent. I'm not very hopeful. Um but I I haven't done this, right? This is I think a lot of people that have been using LLMs that are coming from a non I don't know like a less technical background. I've got years and years of experience programming. If you're coming from a less technical background, this might be the kind of thing like you're kind of just expecting it's going to work. And this is the this is the message that's being sold to people, right? like you can go build anything from scratch like can you?

So I don't know other people are having success with it. I honestly don't think I will. And um you know the worst case is that we don't get anything built but I am able to kind of extract uh some lessons like was it did I have shitty prompts and then it put me down a path that it was kind of like unreoverable? uh were they too general, too specific? I don't know. Um should I have course corrected it? Is it did it start producing code that just like doesn't compile and I have to spend so much time trying to like correct it that I don't know like we're just off on a complete tangent and it's just time wasting. I don't know. So, we'll see. But if uh if you're interested in seeing how it went, it's like I said in what 24 minutes now I really got to make some headway.

I have like only a few minutes by the time I park the car to to get on a camera. Um but it's going to be recorded and uh if you want to check it out, it'll be on Dev Leader. By the time this video that you're watching right now is posted, it'll be live on my on my YouTube channel. So uh it's May 12th right now. So, it'll be up uh I don't know when this video is probably going to go up on maybe it'll go up tomorrow. Maybe it'll be up Tuesday. Might do it Tuesday 5 5 or uh 1:45 p.m. I try to get it up for the for people that are commuting in North America for their 5:00 p.m. commute.

So be able to watch it on dev leader and uh one of the reasons I want to do this a couple right one is I want to I have codec commute.com like I bought the domain as people do right you buy a domain and then you don't use it so got to put the domain to use and uh I think that it will be a more streamlined way to make sure that I can get questions for from folks folks for the channel. So, kind of excited to do that, make some use out of it. And, uh, you know, being able to build stuff with LLMs, I think, is good. Coding live is good practice. But, uh, the thing that I was hinting at earlier is that, uh, I have in my professional career, I have a lot of experience building products end to end uh, like userf facing products in particular, right, for digital forensics, a lot of prototyping.

So especially like that part of the life cycle which is like we have an idea let's see like what's the l truly the minimum viable way that we can get an answer I'm not even talking about a minimum viable product what's the minimum amount of work you can do to answer a business related question do that right so that does lead to a lot of uh coded prototypes though and so I feel like I have a pretty good handle on the software development life cycle And I want to I've already talked to the platform owner Nick Chaps about this, but I'm going to try to get some courses together because I haven't done courses in a while. I'm going to try to get some courses together that are uh building things end to end. I haven't settled on, you know, what products we'll build. They'll they'll mostly be like you can't do um like super large scale things in a course because it's completely overwhelming.

But um we'll be doing end to end builds of things. I'll come up with a list of them then we'll try to prioritize. But it's going to be done with AI. So, what I want to be able to convey in the course and kind of sharing with you folks who love your thoughts on this is that when I've talked about coding stuff, leveraging LLMs and I, you know, vibe coding became like the this, you know, ridiculous term that's everywhere. Um, I started joking like, "Oh, look, like I'm going to vibe code." And then people see what I'm doing and they're like, "Oh, that's not vibe coding, though." Like apparently and whatever the internet's the uh authority on this, not me, but apparently Vibe coding is like, you know, taking the output of the LLM, dropping it in, trying to run, and then uh just going back and telling it to fix things like or change things, right?

Like add this, remove this, fix this, and you just copy the output blindly or I'm assuming using agents to just let them go build whatever. But I' I've been told that's what you need to do. And if you're asking the LLM, like you know, you see the output and you're like, "Oh, no, you're doing this wrong. Go change it." Or you're giving it like specific feedback. Like suddenly you're not vibe coding anymore. That's like breaking the the rules of vibe coding. Like it's inherently supposed to be shitty, which I don't I don't understand it. Like it doesn't make sense to me. Like if you I get it. If you don't have experience building software and you don't have a choice but to copy the output blindly, sure. But surely if you know how to build software and you have feedback on what you're seeing as soon as it comes out, like why wouldn't I just tell it to correct the code or change the style or whatever?

So anyway, what I want to be able to demonstrate in the course is a few things, but one of them is that instead of vibe coding in a way that's just like blindly copy LLM output, I want to walk the students through how you can use an LLM to enhance the entire process. So whether that is writing code, whether that is you know getting feedback about your design uh or uh the project outline or what features uh might be considered for an MVP, what types of functional non-functional requirements like I want to go through the software development life cycle for something and illustrate at all points how we can be using an LLM to be able to help build that. So, I think this will be a really fun course to make. Um, the biggest challenge I see is like in terms of getting content and stuff made for it, do I have to go It's and it's a genuine question.

I don't know the answer yet. Do I have to go build it with an LLM first by myself without recording to show that it's possible or do I just start recording and and capture some of the the problems along the way? And I I honestly think the um the second way might be better, right? I obviously if I'm just like spinning my tires like wasting time with prompts, I'm not going to record, you know, 3 hours of a course is just me being frustrated with the LLM. But uh I think it would be good to be able to illustrate like look, this is what I'm recommending you do if you're getting into a trap and the LLM is like kind of stuck and not making progress. Like how do we get out of that? Right? Or if it's not being helpful, how do we course correct?

So, I think that'll be helpful to show. But yeah, that's the general idea is that uh we're trying to get some courses together for working with AI tools. Uh probably do it in VS Code with agent mode and co-pilot. We'll see. I'll talk to um to other Nick Nick Chapsis about cursor, but I I feel like VS Code will be a good option because I well actually I don't know. Is the co-pilot stuff in VS Code free? Maybe. Maybe it is. I think cursor is free for students, but I don't know for VS Code. Anyway, I'll I'll chat with them. Maybe we need to show both. I don't know. But um I'm excited for it because I like building stuff. I want to do some more technical courses. I haven't done a technical course in a long time now. Overdue.

And this is why the other day I I said it on one of the code said that we said I said on one of the code commute episodes, not code, um that I'm looking for ideas for things to to build, right? So, I got to I got to find the right scoping so that I don't pick something that's going to be finished in like, you know, 20 minutes and something that's not going to take, you know, 6 months for people to go build. But should be fun. Should be a lot to learn and I am hoping to do one minimally and get the feedback on it so that I can uh build the next one more effectively. But, uh, the reality is if the first one just goes well in terms of putting it together, I might start soon on the second and, um, then as feedback's coming in for the first one, leverage it for even a third cuz I think there's there's enough examples like this that I could go do.

And I think that the while the lessons I I try to make sure that my lessons are very generalized, right? So, whatever I'm going to be telling you about Oh, I got to I wanted to switch lanes, but someone was coming up fast. Um, whatever I'm putting in the lessons, I want to make sure they're general, but I think that the reality is we're building something specific. So, a lot of the material that will come up is going to be aimed at that specific thing we're building, right? So, just to I'm making up an example. don't anchor anything to this, but if I'm like, "Hey, we're going to be building a Blazer website." And you're like, "That's cool, but like I don't want to build a Blazer website." Like, I get it.

I'm going to try to have very general lessons in there about how the tooling works and, you know, the the process we're going to follow, but there is going to be a lot of Blazer focus because that's what the, you know, the main tech will be. But if it's like an ASP.NET core site, that might look different, too. So, or Maui, right? Uh, I probably won't do Maui. I don't feel like I'm the the right uh I haven't used Maui. It would be wrong for me to make a course on Maui until I have some Maui under my belt. Uh, but I feel like uh ASP.NET Core, I could do something or I don't think anyone wants to see a Windform map. Uh, but you know, if it's just like a I don't know, console application that does something. We'll see. More than likely, it'll be uh ASP on a core web app, though, to start.

So, I don't have a timeline for that, but trying to gauge some interest um because I think it's going to be super cool to do it. And how am I doing for time? 6 minutes before I got to be on a camera. Oh, that's not true. I have eight minutes to get home and then then there's six minutes to spare is what it's saying. We'll see if I can do it. Um, what else would I want to say aside from the core stuff? Might be it. Yeah, I think that might be it. I I I'm I'm excit This live stream is going to be got to be bumpy. I'm a little nervous, but chat's usually pretty good. We'll see. We'll see if there's some people that have uh suggestions for prompts and stuff. Like I don't claim to be a prompting expert, so any anything like this is good practice for me.

But I think I'll wrap it up there so you don't have to listen to me mumble about nothing. So thanks for being here and this is just a friendly reminder for folks that say to the end right like if you're new to the channel questions please leave them in the comments. Anything on software engineering career development I am happy to try my best to answer and uh hopefully sometime soon we'll have code.com set up you can submit your questions there. Uh, but I think if you go there right now, at the time I'm recording this, I think it just goes to my YouTube site or YouTube uh page. So, I will see you folks next time. Thanks for being here. Take care.