"USE YOUR BRAIN": The Skill That Makes AI 100X More Powerful Than ANYONE Thought Possible

The one thing that's going to matter as we get more and more agents is use your damn brain. Yeah. Wild, right? I don't care how good models get, there's a lot of stuff that you have to do before you even talk to the model in order for you to get the most benefit out of it. Let's talk. So, this is a daily upload channel. I do newsy stuff in AI strategy and let's get into the first thing. So, yeah, as I said, do this. But why? Because as the models are getting better, as we have all these agents that are coming onto the scene, like literally like I yapped about it this morning in the last 24 hours in our Discord. I just posted this stuff. I spent about an hour every morning kind of filtering through the noise. Be very keen on who you listen to. Is the person that you're listening to a builder or a marketer? Big question. I build software. So, in here we are seeing slight interruption there, but yes, we've had a lot of stuff that came out in the trailing 24 hours. And I'm not even sure that I'm going to continue being like this news guy cuz it's just like what's not going to change is there's a bunch of skills that you need to get good at in order to leverage them. So, I'll probably just talk about that stuff. But, it was announced that they're coming out with Codeex. You get access to it if you pay 200 bucks. This is their competitor to Claude Code. Claude Code's 100 bucks for max, which makes me so angry because I spent $100 for an afternoon with Claude. A day on the town with Claude Code cost me $100 just lighting it on fire. Cloud Code's building a VS Code extension. Grock iOS app is getting video as an input for voice mode. And then honestly, I'm just going to refer you to this Discord channel, not even mine that's part of our community. This other one because mine's paid. The other one is called, yeah, dev mode. And there's a bunch of bros in here that have built some really good scrapers to detect the web changes as well as app updates as well as new models as well as the specifics of Gemini, Copilot, Cloud Web, GPT web, all that stuff. My community is more so like the whole mission is, hey, let's figure out how to leverage AI in coding and in content to push what's possible. And yeah, there's a lot of smart people in there, way smarter than me. We got Microsoft people. We're getting Googers. It's dope. But yeah, so that's some of the news that's come out. I'm assuming that next week's going to go bananas when Gemini releases a bunch of new stuff. I'm all in on Google. Might get tattooed on my forehead. Just super bullish on Google. Oh man. But this goes into the idea of, okay, now what? Like you just yapped. You just ran the motor. You hit us with the fire hydrant of information. Now what? There's some stuff that you suck at probably because you watch a video and this guy's I will show you how to oneshot this SAS app. Or maybe you're somebody that is already good at coding but doesn't know how to leverage it. Or maybe you're somebody that is coming from Bolt or any of these training wheel apps and you're a vibe coder. You need to get better at product strategy, systems thinking and system patterns, a little bit of architecture, bunch of this stuff. Highly recommend checking out Klein. Their Twitter is just full of bangers. But they basically said that the biggest lever for AI generated coding is the planning process. And planning is actually a role. It's called product management. That's what you figure out the what and the why for what you're about to build. That combines research, figuring out what works, what doesn't work, why should I build this. Always starting with the problem. And highly recommend going and checking this out. Why does planning work so well? It ensures a shared understanding. So AI truly grasps what you're trying to achieve and its constraints. Also, if you don't come in with a dope prompt, then the degradation goes haywire. Meaning like the quality of stuff is just going to be awful. With client, you can at least manage the context window. With cursor, it's hey, start a new chat. With winter surf, I have no idea cuz I don't use it. But what I've been doing is make your own tools. Weird, right? Like I'm the same guy that made a Taskmaster video. Yeah, I use Taskmaster when I'm making a form or like a static website and that's fine. It's fantastic for that. But as soon as you have something complex beyond like basic CRUD, then Taskmaster is going to degrade the quality of the tasks that you're going to create. And so I'll show you something that was from our talk today. We did a workshop and we like yapped for a half hour afterwards, which is awesome. If I didn't know where Figma is, I'm in Figma. Wow, embarrassing. But I basically talk about how don't skip over the thinking. Garbage in, garbage out. AI is not your idea architect. It can pretend like it is, but it's not going to be as good as you actually using your noggin. And then how multi-turn degradation is a thing. And I covered basically like these are the skills that you'll get by getting better, which these have an ROI as models get better. You can't get leverage on like basically if AI is a multiplier, what's zero times zero, right? Like you want leverage. So if you're 10x dev and the model got better, then your multiplier is better. But if you're mid or you're a junior, you want to get better. So it's it's a call out to all the vibe coders that you still need to get good. If you bought protein from someone that's jacked, you're not going to be jacked. Yeah. Planning is at a premium. And then I talked about basically how you chain all these prompts together and then you go into agent mode but you're in discovery. So on the lefth hand side you have ideation, right hand side you have implementation and then I walk through a framework that I've been using and then what it looks like in a chat. So left hand side I was making memory banks across the entire codebase the entire codebase. But I was making memory banks that would help me scale up on my codebase. So what that looked like was having AI docs and then having rules associated with said AI docs. So if I go into here, I had memory bank and then I'd have a memory bank for the core API which is fast API. I'd have a memory bank for the client or the web which in my case is tan stack start and then I'd have one for the infra it's okay if you're using GCP or Terraform or superbase or whatever for databases and deploy CI/CD anything cloud related and that helped me basically pull out like the Master Chief sniper rifle from Halo 3 and be like this is it like only listen to this context and It didn't require any tooling outside of the markdown files that you can yank from client and then using agent mode and ask mode. So when you're early on making a feature and using your brain, then you'll be using ask a bajillion times and then you can move on. And I structured all this for the guys in the group, but here's like the high level. It's basically make rules, do some research, and then have the memory bank. What else? I also modified memory bank so that I have file trees. So I originally built this in flask and I was like, "Oh no, I don't have real time. I don't have like background tasks or maybe I do, but it's better in fast API." So I went and did the research, found the guy who invented fast API and then went to his GitHub, checked out his stars, saw who he is following, and then you go through this like spiderweb effect, which you can't really emulate by going through perplexity. perplexity. All these LM are doing is they're going and finding the things that have popularity that hit on high SEO things. So that's not always the dub. You'll find someone that's got like three views on a blog post who invented the thing. And it's a point that one of our members made today which really resonated with me, which is like find the teacher of the teachers and then you have to validate are they doing the thing. So all that said, I'm excited about all this. I'm going to be putting together some sort of framework that like ties all these loose ends together. If you learn anything in this video, make sure that you subscribe so you can keep up to date. We recently had I made a folder actually of nice these are called nice things people said and it's like went to a cursor meet up yesterday for power users and although it was exciting basically nothing was a surprise and the guides and the cutting edge stuff you teach us is ahead of everyone there. So, that was sick. And then John said, "Thanks. I'm really impressed by your knowledge." Thanks, John. You the man. YTM. Aiden said something nice. I don't know. There's a lot of nice people, but there's a lot of smarter people than me. So, if you want to hang out with people that are better at like architecture, I don't know. That's kind of the beauty of it. There's a lot of ways to make money. I'm not trying to get rich off a community, but I'm just shephering everybody together. And then eventually we'll probably break it out into pods where there'll be like dedicated marketing path, dedicated directory because there's a lot of people in there that don't even know what others do right now. But by going and being active in the Discord, everyone's learning. So I'm just trying to make it easier. Check that out. Links in the description. Make sure you like the video if you like pancakes or like any sort of food. Not even if you like the video, just like the video. And I'll I'll see you in the next one. Bye.