I Was Starting To Forget...

So I recently went through what I think a lot of developers are going through now in this AI era. And even though you might be creating the architecture, reading code, running tests, managing context, a lot of us are actually losing some of our skills when it comes to writing code and some of the lower level aspects of software development. Now you do have muscle memory. So if it if it hasn't been too long, once you get back into it, it's easy to get back. But what about after a year or 2 years or 5 years without writing any code? You're going to lose that part of your skill set. So let's talk about that. Hey guys, real quick. So for the past year or so, I've been hinting at a new interactive platform that my team and I have been working on. And you can finally go to start.dev and you can join the weight list. We have early access coming up and we also have a Discord server and an X account where we post regular updates and screenshots and videos and all kinds of stuff. So if you're interested in that, I'll have the links in the description below. Okay, so everybody knows that the way that we write code is changing. In fact, there's a lot of people that aren't writing code at all anymore. And that includes top level seniors at giant companies. And I hear from a lot of different people. I like to ask questions about their situations and experiences because I I want to know how this change is panning out within the industry and not from Twitter or or YouTube, but from real people. And I just talked to a guy that took my AI course and he interviewed for the role of AI native software engineer and he got it. He got the role. I'm actually going to make a video on his experience because it really surprised me that in the interview they had no type of actual coding syntax testing which is very different from what I'm used to. They let him use clawed code during the interview. Now what I think a lot of people get wrong is that you can't use AI in a way to where the output where your code is is at the same level or even better than without it. I just think we're moving towards higher level concepts being more important than syntax typing. But with that said, I still think it's important to know your syntax when it comes to, you know, data structures. You should be able to to deal with simple algorithms, patterns in React if that's what you're using. Of course, creating components, fetching data, all that stuff to me is is mandatory to know, and you should be able to do that with or without the help of AI. I'm not saying you should always do it without AI, but you should know how. I like to create a lot of stuff on my downtime. I and in the mornings, I like building little tools that can just kind of make my life easier. I log a lot of the stuff at bradtravery.dev. And I keep devlogs for certain projects, both open source and private. And what I've been doing is every morning I think of some kind of small project to build that will help me somehow. And when I create my projects, I'm using AI tooling, but I'm also, you know, in control of the architecture, always looking at the code and optimizing. But I had realized that I hadn't actually written any code uh from scratch outside of courses, you know, creating uh course projects and so on. But I have I've been working on this this uh SAS called Vidpipe.ai AI that turns YouTube videos into a content pipeline of articles and newsletters, social posts, things like that. And there's a part of that source code I call the worker and it uses something called bullmq and it uses reddus and on and upstach to cue and process the the video to article jobs. basically the the background engine that does all the heavy lifting and it watches a channel, grabs new videos, pulls transcripts, uh runs it through the AI, and generates everything on its own. And and this is a a a good example of a situation where I couldn't just tell AI to do the work. You know, I I tried that, but it failed miserably. And this is the the meat of this project, right? Uh, and I I needed the worker to handle generation at scale, to retry jobs that fail, uh, deal with rate limits and and make sure that the same video never gets processed twice and a whole bunch of other stuff. So, obviously, I had to handle this manually. And this is where I realized that I I was struggling. And I mean, I wouldn't call this project easy by any means, but I had a much harder time with it than I would have, you know, a year or two ago. um literally for two weeks I worked on this part of the project and I I need that part of my brain to work and not go away. So I can't outsource that to AI. So what I decided to do from here on out is always have some kind of side project where I use no AI at all outside of maybe asking some questions and and prototyping. No claud code, no 100line autocompletes, so shut off co-pilot. um just me code and logic. And uh I'll still use AI tooling for all my other stuff. That's just the way it is now. But I'll always have at least one project that's just straight handcoded. And it'll take a lot more time, but the purpose isn't really the project itself and getting getting whatever it is out there. It's the whole process and wanting to ret retain my skill set and make my brain, you know, keep working in certain ways. And I've been really getting into, you know, home lab stuff lately, networking hardware, creating servers. So, my first idea was to create a home lab inventory tracker. I plan on adding a social aspect as well. And I just started it, but I will be building in public. The only thing that I used AI for was prototyping, just the overall Tailwind theme and UI ideas. The rest of it I'll be building by hand. And I'll be honest with you, it's it's slow. um stuff that AI would have knocked out in a few minutes. It's taking me some real time, but I can already feel that part of my brain switching back on and I'm actually thinking through the problems myself instead of just reviewing what you know what the AI wrote and that's the whole point. So that's kind of what I suggest just everyone do. Uh any software developer, no matter where you are in your career, no cloud code, no codecs, and I'm not saying don't use those ever. You can use them with every single other other project you have, but pick one real project that you're excited about. Not a to-do list, not something like that, something you would genuinely use and build the whole thing by hand. I would say the most you could do is ask chat GPT questions if you're stuck and just prototyping cuz who wants to type out HTML and CSS? This should be able to keep you on your toes so that you don't forget everything about writing actual code. And it's not just syntax. There's a lot of low-level um skills and and just problem solving that you do lose with AI. But let me know if you guys have this problem as well. Those of you that are really leaning into AI, have have you tried to really sit down without it and build something? And what's your experience been? And if you do take this route and you do build something using no AI at all, tag me on X so I can check it out. and I'll see you guys