Stop Using Claude Code The Wrong Way

0:00If you're using Claude Code, you're probably using it wrong, and that is okay because for the first few months I had been using it wrong, too. In this video, I'm going to help with that. Most people open Claude Code and just start typing away. They treat it like Chat GPT with a code editor attached, and that's the different concept. It has so much more power than this. They get frustrated when Claude doesn't understand their project, when it writes code and it doesn't fit the whole idea of what you're trying to build. It keeps forgetting what you're trying to build, and then it goes into this constant loop of shipping a feature, debugging, and then you spiral and give up. There's a pattern that's taking over in 2026. It's called the Claude Markdown file. This is a huge Markdown file that's absolutely

0:46necessary to direct the vision of the project you're trying to build. So, when Leon Vansil made a video called stop using Claude Code the normal way and showing this technique, it got 76,000 views because the people finally understood what they were missing. So, I'm going to break down why the Claude Markdown file is becoming the standard for setting up your project and how I use it every single day inside of my mission control dashboard in future projects I build out. And what the real meta looks like is that when you combine it with skills, agents, sub-agents, and your product requirements document, you will fully understand the difference between just starting Claude Code out of the box and using these Markdown files as actual tools to progress your

1:31project. The problem starts the moment you open Claude Code. You get excited. You're ready to go to work. You're ready to start shipping features, but Claude has no idea what you're trying to build. You have this idea in your head, and maybe you want to build a dashboard, add a new feature, change the landing page, fix a bug, whatever. And you start describing what you want, and you feel like you're making progress. It's okay. I've been there before. But what's really happening behind the scenes is Claude has no context of your project. It doesn't know your tech stack. It doesn't know your coding conventions. It doesn't know your entire file structure. It basically doesn't know you and your business logic. So, it makes assumptions, and those assumptions are usually wrong for your specific project. I'm sure you've opened Claude Code and you started working on your landing page, and it spits out a very generic

2:17vibe-coded landing page, and we've all seen it before. This is because Claude has no understanding of your your tech stack. You want to use Tailwind. You want to use a nice front-end design skill. It has none of that. before we keep going, if you're watching this and you want to actually build with some of these tools, not just watch videos about them, then you're going to want to check out our community down below, Shipping School. We have a full Claude Code course, a full Open Claude course, and four live bootcamps every single week where we actually help you get set up from scratch. Like actually set this thing up, not just watch a tutorial and figure it out by yourself. And we also provide one-on-one coaching, so you could book a call with me. We could share screens, and I can help you get

3:02Claude Code or Open Claude running on your machine. That's it. No fluff. I built this community because watching YouTube's only gets you so far. We launched it just 3 days ago, and we have over 55 members. You need people around you who are actually building, people who hold you accountable, and coaches who can help you when you get stuck. I'll put the link in the description down below. Get in now before the price goes up. So, I see this all the time in Shipping School and a lot of the questions of like, how can I get going? And people just think they could just jump in the terminal and start working. But that's why we have that course built out inside Shipping School that breaks down the exact process from the product requirements document to the plot prod Markdown file to the agents mark Markdown file and as well as getting your sub-agents spin up. So, you know,

3:48people will spend an hour trying to explain their project to Claude, and then Claude writes the code that uses different frameworks or puts files in completely wrong places, or basically just uses a completely different naming convention than the rest of their code base. Files get lost. You get frustrated and give up. So, it's basically like hiring a developer and not giving them any documentation about your project. They might be brilliant, and they're going to waste a ton of time figuring out what you actually want. That's where the Claude Markdown file comes in. Think of it as like giving Claude a brain transplant for your specific code base. Claude Markdown is basically a file called a Markdown file that lives inside of the root of your project. It's basically a comprehensive guide that teaches Claude everything about your

4:33project, your tech stack. So, if you're using Neon for database, Polar for payments, or maybe you're using Supabase, you're using Stripe, it needs to know these frameworks to be able to build accordingly. It needs to know your file structure, your business logic, everything, and all that can be included inside of this Markdown file. So, when you start a new chat inside of Claude Code, the first thing you tell Claude is to read that Markdown file, and suddenly Claude starts understanding your project, and it's been amazing. It gets full understanding of what's going. So, Leon Vansil demonstrates this perfectly in his viral video the last few months. He shows basically how most people just open Claude Code and start rambling about what they want versus the pro approach where you create that Markdown

5:18file and serves as the single source of truth for your entire project. And the results are night and day, and trust me, I've been there before. I've started projects within Bolt and brought them over into Claude. Started shipping features, things were breaking, and wondering why. We don't want to start that way. We want to start the correct way with correct Markdown files, like I said, product requirements document, the Claude Markdown, agent Markdown. All these are necessary before you start working in your project. So, instead of asking Claude, you know, clarifying questions for 20 minutes and making sure it gets the whole understanding, it just gets to work when you put the work in on these Markdown files. So, instead of writing code that doesn't fit your project or has nothing to do with your

6:03project, it instead writes code that looks like you wrote yourself. Now, I want to share with you how I actually use this in my daily workflow because I think my setup is a bit different than what most people are doing. Every morning we have the 6:00 a.m. Rizza agent fire off on a cron job. It does a lot of the content and copy writing for me. You know, the outlines get saved to my mission control dashboard behind me. So, when I'm ready to review and edit those scripts and outlines, I open Claude Code. But before I start editing, I make sure Claude has read my Claude Markdown file for Mission Control. And this file contains everything about my project. You know, the fact that I'm using a Next.js 14 app, the fact that I'm using Tailwind for the styling, the fact that I'm using Neon Postgres. You

6:49know, all this is really important. The fact that I have specialized agents with specific names like Atlas and the Rizza. It basically contains all my content conventions, and you know, I prefer casual conversational writing. It knows this. Basically, the fact that I avoid certain AI slop phrases. This is huge. In a humanized voice as well as I have fed some of my content from previous videos back into the agents to be able to know that, hey, this is my voice. This is how I sound. This is the common phrases that I use. So, when I paste in a script that the Rizza generated and ask Claude to help me refine it, Claude doesn't just see the script. It sees the script context of my entire content system. It knows my voice. It knows my

7:35conventions, and it knows my goals. And then the edits it suggests are actually useful. You know, with Claude Code out of the box, it does produce very generic generic landing pages, generic setups, and they're not specifically tailored for the thing that you're trying to build. Adding this Markdown file greatly changes everything. So, instead of generic improvements and generic advice, it suggests changes that actually fit my specific style and audience. Basically, the same thing happens when I'm building features for Mission Control. When I want to add a new dashboard page or I want to improve my agent pipeline, I don't have to explain my entire tech stack every single time. Claude already knows that I'm using Next.js. It knows my file structure, and it knows how my

8:20agents are set up. I can just say, "Add a new page to track my video performance." Boom, done. And Claude knows exactly what that means in the context of my project. Here's where it gets more interesting. The meta isn't just Claude Markdown. The real power comes when you combine it with skills, agents, and sub-agents. I'm going to explain what I mean by that cuz there's a lot of confusion when it comes to agents, sub-agents, skills. What are they? Skills are reusable code snippets or functions that Claude can access. Think of them as a toolbox, basically. Instead of Claude rewriting the same database connection code every single time, it just uses the skill you've already defined. Agents, they are autonomous systems that can use these skills to complete tasks. So, for

9:05example, my agent Atlas, the CEO, it can read my calendar, check my email, update the dashboard, and send me daily briefings. It's not just answering questions like a typical chatbot. It's actually doing work. Sub-agents are specialized agents that focus on specific tasks. So, the Rizza, for example, handles the content creation, Inspect Deck handles social media, and also we have the Jizza doing podcast outreach. So, when you combine all this with a Markdown file, you get something really powerful. Claude understands your project enough to just write the code and to understand how the code fits in the larger system of the agents and the automation. So, maybe you're going through a little frustration right now. You got to back up and look at the bigger picture. Do I have my Claude

9:51Markdown file? Do I have my agents Markdown file and my product requirements document? If things are not working for you right now, you need to back up and get working on those files. Have Claude help you. Have Claude help you with your idea and start working with the type of framework and architecture with the tools that you want to use to build. I use Neon for database, you could use Superbase, you could use Convex. What are you using for payments? Polar, uh Superwall, um Lemon Squeezy, or what are you using? PayPal, Stripe? You know, you need to give this context first before you start building. So, when I ask Claude to help me improve one of my agents like the Rizza, you know, it's you know, copywriting, script generation, whatever, it doesn't just suggest better prompts. It suggests improvements that would work better with my approval workflow inside of my

10:38mission control. It understands that scripts need to be stored in my Neon database and that they need to integrate with my thumbnail generation pipeline. Also, that they need to fit my publishing schedule. That's the kind of contextual understanding only happens when Claude has deep knowledge about your entire project. Here's why most people are failing with this approach. They try to build a Claude markdown file all at once. They sit down, they try to document their entire project from scratch, and that can be really overwhelming if you're just getting started in this space. You forget the important details or maybe explain things in a way that makes sense to you but not to Claude. The better approach is to build your Claude markdown file iteratively. You know, start with the basics, your tech stack, your main file

11:23structure, your core conventions. Then, as you work on your project, add to the file whenever you notice Claude making assumptions that are wrong for your project. If Claude keeps putting components in the wrong folder, add a section about your component organization. If Claude keeps using different naming conventions, then add a section about how you name things. If Claude keeps forgetting about specific business rules, document those rules. Over time, your Claude markdown file becomes this comprehensive guide that actually makes every interaction with Claude more productive. The other mistake I see people thinking is like you could just copy someone else's Claude markdown file. No, don't do that. That's not going to work. That doesn't work for every project because it's

12:08different. Everyone's building different things. Your conventions, your business logic, your goals are unique to you and your project. You have to actually build it yourself and build a file that actually reflects your specific needs. I guarantee you once you do this, the productivity gains are incredible. I could jump into Claude Code and start building features immediately without spending like that first 10 to 20 minutes explaining all the context. Claude understands my project well enough to suggest improvements I wouldn't have thought myself. It's like having a senior developer who's been working on the project for years, except it's available 24/7. It never gets tired of your questions. So, if you're opening Claude Code and you're just getting started and you're starting to type and you're missing out on probably the biggest productivity improvement you

12:53could probably make to your whole workflow. Get that file going. If you don't, building is going to be much harder. Create a Claude markdown file. Start simple, but start today. Get it going. Document your tech stack, your conventions, your goals, then build on it as you work. And if you want to learn how to set up systems like this and how to build agents that work with Claude Code, how to create automation pipelines that actually make your life easier, that's exactly why we started a community 6 days ago with over 71 members building and shipping, working together side by side, sharing ideas, four live bootcamps every single week, one-on-one coaching to help you get this thing going. So, we have a full Claude Code course that walks you step by step on how I built multiple SaaS products, how I set up my PRD, how I set up my

13:38landing page, payments, and walking you through the process. So, we do live calls every week. Like I said, it's not just a course, sit there, watch it, goodbye. We actually work with you every single week to help you take the next step wherever you're at. I'll put the link in the description. Check it out. Get in there before the price goes up. If you guys haven't subscribed to the channel already, we're dropping four to six videos every single day, dropping workflows, tips, journey videos, AI news. And if you want to stay in the loop and figure out what's going on, hit subscribe, hit the bell notification button, and we'll see you in just about an hour. Bye-bye.