Planning My AI Workflow for Managing MonoRepo Chaos

Rude boy. Okay. Yeah. Hey, I'm Parker Rex. Today we're talking about AI stuff. We're talking about three tiered architectures, talking about some AI news. Yeah. This is a daily channel. I led tech for a startup that sold for 23 million bucks. We scaled a bunch of tech. Yeah. Yeah. Yeah. Yeah. Yeah. Let's go. So, first of all, we got some dope new members. Oh my gosh, this thing that I'm doing is working. I really wanted to have a group of great nerds together and it's so great because I am not your guru. I'm just someone that's shephering it, right? And I share what I know, but I have gaps. There's so many better devs in our group now that it's just absolutely bananas. Like we had this guy join yesterday and his name he was actually kind of a kind of ripping on me on the on the comments a little bit but he introduced me to this pieces thing and it's basically I think the way he explained it was it's like malware plus LLM but I wanted to talk about it just for a sec. I hate that they have scrolljacking but they had a good podcast or master class whatever you want to call it explaining what it does but it's kind of like uh I don't know especially remember stuff that you're doing. It's kind of like recall, but I don't know if any of you guys have used it, but it has nine months worth of context, and it's supposed to basically just follow along what you're doing while you're coding. So, if you get lost, you won't run in circles. And I'll show you an example from one of the things. Whoa, that's huge. Where is this? Is it my DMs? Let's see. Yeah, that's not it. This. So, it's basically just like remembers all the stuff that you're doing. And I guess that's pretty helpful. But I do think this will be something that's built into the OS at some point, but I just want to call that out and see if anyone's using it. I'm not, but it sounds like an interesting tool. And then Indy Dev Dan, I watched this guy's videos for a long time and if you haven't checked out his channel, he's got a bunch of good stuff on it. I stopped watching him for a little while just because I don't know, some of the stuff was overkill for my use case. doesn't mean he's bad. I think he's probably a way better engineer than me. But I'm all about like the practicality of stuff and just flying around. So, I want to see like real use cases. And anyways, his most recent video, he basically lays out these three folders that you should have in all your projects. And it's really like covering for context and AI docs. And then I think the last one's I should probably know this, right? Let's just look. That's not it. Yeah. So, the cloud commands is the last one. And I think this is close to what I want, which is I don't really want to have any tools besides just cursor and shell scripts. And what you need is basically a set of different things that you're doing over and over and over again. So in my case, what I'm doing, and I'll show you this in practice in a sec, is basically like you'd have a shell script plus the files that you want. The shell script would include the like file tree. So the structure, what that looks like in practice is something like this. Make it bigger. Everyone's going to complain. Oh my gosh, light mode, dude. I live in Palm Beach, Florida. It's beautiful out. Sorry, you're in a dungeon, but it looks like a practice is not there. That's a front end one I need to do. But you end up with something like this where it's like you can see the actual like file tree both on the existing ver oops the existing version of what you have and then what you want. So if you're doing a refactor, I did a refactor from flask to fast API and this is all agentic. So I didn't actually write much of this code but I'm reading it right. So you have that something that's always dealing with your context. And then AI docs for me it's a couple things that I do where it's I do a pitch is the first thing. The pitch comes before the PRD. Then you have the PRD and then you have examples and this is a mixture of NDV Dan stuff and like Klein you have an overview and then what that looks like in practice is oops you have examples. So examples of how we're going to use these in the project the PRD so that like wraps it all together all the requirements and poking holes in the idea and this is a product of I don't know like 50 chats just like going through and kind of working through it because I trust myself more than I trust the AI to do it. So we work together the file structure comparison for a refactor. If you're starting net new, you can have it actually give you what you want, but I always say start from scratch. If you've never done it, like I started from scratch and did the flask one and then I was like, "Oh, this isn't as good." But that's learning. And then the implementation tasks. So that's what I'd include in the AI docs. So this has more to it than taskmaster, right? because it's like already has dependent structure built in and then already has more because it has snippets and relative paths and then the overview which is this is how we're going to do it. I think we're getting closer where you wouldn't need all these tools. The only tool that I I really am considering actually I'll call it out is augment but I'm not sure. I'm still indifferent about it. I haven't talked about it much just because I don't want to bash on stuff that I haven't made a clear decision on. So now let's talk about the like use of this. So I think like the title of this video like the main point is basically like how to use AI and bigger projects and this will be a high level of it because this is a daily upload but the more in depth like this is like layer one depth ADHD content and then we'll go to like layer two depth will be on the main channel and then layer three depth is just like raw recording of three hours of me actually working and that's in the school. Yeah, it's paid. So three- tiered architecture. I want to talk about that and what I'm doing with my product Echo. And so if you didn't know what Echo is, Echo basically solves for Echo. The problem that it solves is YouTube videos are annoying to produce. Not to actually record. This part's fun. All the stuff around it sucks. So that's the why. And you always start with a Y. And then the what is the UI I can upload to and get all metadata, titles, thumbnails. In a later version, it'll recursively check the performance of the video against others like it or in its sector and then autooptimize. So you wouldn't have to go and like rehash or whatever, which would be cool. And then the how we're getting into now. So that's the actual like technicals. So the evolution from prototype to where I am now was initially, if you watch the channel, cool. Mounted a bucket. Originally I was like, whoa, cool. I can mount a bucket. and I'll just drop stuff in here and then it'll do it. And I did that. And I made a few videos where they were auto uploading. First versions of it like broke YouTube because I accidentally sent 75 versions of the same video. But it proved the the thing would be helpful because it could do anything. Like I started with one thing. Oh, it does subtitles and chapters. And then I just started layering more. And I was like, whoa, I could do a lot of stuff with this. And I started talking about on the internet. Then I started talking about it to friends and realized, oh, companies want this. So then I'm like, okay, I got to go to the next level and also kind of need a front end because I can't really see what's going on. And like staring at Docker logs or G GCS, not GCS, but like a staring at like logs basically to know if it's working isn't as nice as having a front end. So then that brought up like okay I need to think about this and there was a learning curve involved for it. There was stuff I didn't need I didn't necessarily know. So out of 10 this is what I do whenever I'm like starting something new because I've learned so many different things and you just have to be honest which is like zero through 10 where are you on the tech that you think you need for this? That's a big question. So I knew I needed I didn't think I needed I knew I needed Docker and I was like a three out of 10. And then I knew I needed Python because there's just gonna be like a lot of agentic stuff and I want to use Google and Python's the best way to do that. It's the only way to do it. So I was like two out of 10 on this. And then I knew I didn't want to use next because I've had so many issues with it because I am doing a separate backend. So I wanted to try tanstack start which I've used tan stack query but that's not routing. So it's like I'm like a one. And so I'm looking at this and I'm like oh gosh there's a lot of stuff. And then there's also GCP and I was probably like a three. So what I did was I kind of like looked at this and I'm like, okay, like knowing this, there's going to be this idiot tax that I'm going to pay down with time. Am I willing to do that? And yeah, easy for me. I am. So what that meant was just going and reading docs. The Docker docs are awesome. Python, there's so much different documentation out there. It's easy to learn. It's really like simple syntax. Tanstack was probably the toughest just because it's at first like I was just reading the tan stack start docs but then you realize that it's just tan stack router with some goodies on top of it. So there's plenty of examples there and then GCP there's a fantastic number of examples. So that's when I said okay I'm gonna just go for it. And then that's when this three- tiered architecture came into place. So I'm looking at it now and I'm like, okay, what do I need to do? I have this new mixture of client and like cursor and the ndv dan thing kind of with like scripts. How am I going to do this? So the next step for me and what I'll be recording on the main channel is basically pulling this all together in a three- tiered architecture where I'll have let me create DB So everything will be relying on front end DB back end. These credentials, this was for when I was just doing it manually. So these are service accounts. I'm not going to click on them, but there are service accounts that allow me to do this stuff with admin perms, which like no user on the planet wants to deal with. So because we have a front end now, that won't even need to be there. And it'll just make this like a lot more fun to work on. I think that's an issue too with people is when they AI code stuff, they're giving it the wheel too much and you need to take the wheel. Like even if you're a newbie, like newbies overindex the power, seniors under index the power. There needs to be somewhere in between. Whereas if you're a newbie, good, glad, welcome to the game. You've gotten more dopamine than every other programmer on the planet by them not just getting hello world, you're getting like an app, tic-tac-toe or whatever. But now you need to skill up. And so then to that point like they'll just let stuff fly and they hate to work on it because they're confused. So you need to like as you're going, you know, you kind of clean the kitchen as you're cooking. It's kind of the same with cleaning your code base as you're coding so that it's just easier to work on. It's more maintainable. So in my case, I'm going to end up cleaning all this so that I don't have like a bunch of crap. And then it'll just be at the top level. It'll be like backend front end DB docs and then docs I'm going to move everything out of these different tiers. So it'll be like backend docs and then front end docs and then at the core it'll be docs and that way I'll have the ability to manipulate all the areas in the specifics. If there's a road map for the back end, an implementation plan and those tasks, then those are separate and the front end will be separate and the database will be separate and then there'll be one that's above that and the parent which is just docs which is something that touches all of them. So it's like the connection between the three for instance if I want to do like a docker swarm or whatever or frankly just like gluing them together. So that's what I'm excited about. That's going to be recording in general. the community is growing and I'm I'm like really pretty excited about it honestly like this open telemetry thing I talked about yesterday but I think this will be I'm not going to jump ship on any of my projects to go and implement this but it's something that's on my horizon I know that this is so sick like I can just tell because we manually did this at delivery dudes where it's like you have these things and then you react to them but you want agents to react to And so with open telemetry, you can kind of like stream all your stuff under the hood down. But that's exciting. And then we have a master class on Friday that our boy Hari is going to be doing. So he's going to be talking about all like the Nad Vibe marketing stuff because he's in the field doing it. He's got clients and he's producing a bunch of content for them. So that'll be cool. So if you're interested in that, we do the marketing ones on Fridays, every other Friday. And then it's basically like we flip-flop between coding ones and marketing ones because everyone needs to know both. I don't care what you think. That's what I think. So yeah, and it's half off right now until I launch I vibe with AI which I'll actually be building it public on here too. Then once I get Echo up and running, then I'll go and switch into the I Vibe with AI stuff and all the content will be around that. And make sure you like the video because you know that's a good thing to do. I also need to build this some Discordy stuff. If anyone's built with Discord, please let me know what's the best practice for verification of members. That's like an easy one, but I want to have like content feeds in there with Vertex. So, if you're interested in that and you want to build something together, hit me up, subscribe, and I'll see you tomorrow. Paste.