In this tutorial, Javier Torres explains why he built a persistent "second brain" to give large language models (LLMs) like Claude and ChatGPT a shared, reliable source of truth, and walks viewers through the practical setup. He begins by describing the problem of fragmented, "poisoned" AI context that arose when his AI Council started steering his business in unwanted directions, especially over‑emphasizing his Christian identity. To fix this, he created a structured knowledge base stored as Markdown files in Obsidian, with strict rules that only his own statements become truth, while AI suggestions are merely adopted after his approval. He outlines the key components of the system – a memory file, next‑session file, ideas inbox, core OS file, voice‑calibration snippets, locked decisions, frameworks/people, reference transcripts, and a quarantine folder – and explains how each AI reads or writes to them. He demonstrates how to seed a new chat in both Claude and ChatGPT by pasting the source‑of‑truth document, then giving them the same task, showing that while their outputs differ, they operate from the same context. He emphasizes using voice mode for faster, more natural input, performing a comprehensive brain dump, and iteratively refining the source‑of‑truth until it sounds exactly like the user. The video targets founders, coaches, consultants, creatives, and pastors, and offers a free tool to automate the basic setup, plus links to deeper walkthroughs and consulting services. The overall message is that a well‑curated, verified knowledge base prevents AI from hijacking one’s brand and ensures consistent, personalized assistance across multiple LLMs.
Chapters
Does it feel like every time you open up ChatGPT or Claude, you're starting all over from scratch? It doesn't understand your context, who you are, the business you run, or the people you're called to serve. It even feels like it forgot every single conversation you ever had. Well, in this video, I'm gonna show you exactly how to fix that problem with the system I've been building for the last few months. First, I'll tell you why I even set up this system to begin with. Then my friend here, BB01, will help you set up yours. And if you stay to the very end, I'm gonna give you a free tool to set up your basic version yourself. Sound good? Without further ado, let's get into it. In one of my last videos, I showed how I got Claude and ChatGPT to work together, but all of that only works if both of them are reading from the same source of truth. And someone in the comments asked me exactly how do I actually build this myself? I know a lot of this can get really techie, so that's why I brought my friend along to explain it all to you.
Hi guys, I'm BB01. Javier built me to help keep his brain organized.
He's gonna walk you through how to set up yours, but first let me tell you why I built it in the first place. So for the last year, I've been in a business coaching program with Omar El Takrori trying to figure out my offer, what exactly I sell and who exactly I sell it to. Because I've always been into creative services, but it's only been a video here, a website there. I've never put them together into an official offer. And as I've been developing and working on this, I started using some AI tools and a tool I called the AI Council. It's a group of AI advisors who look at things from different angles. At first, it seemed like a super powerful tool. Then I go back a week or two later to go to review my goals for the week, for the month, and I'm like, bro, this sounds nothing like me. This isn't like anything that I sounded like, anything that I wrote. It started poisoning my context. The AI advisors that were once helping me were now hurting me. The council started to infer its own position, started to make its own goals, and started to switch the direction of my business, which was really weird. It picked up on that I was a Christian, but it started making my positioning only about churches and pastors. And yes, I'm a Christian and it's the guiding light of my life. But I don't have this agency as a Christian business. I'm a Christian who has a business, but I serve everybody. And yes, I help churches, but the positioning of it started to feel a little weird that it was only making my goals about church. It started getting worse, started telling me that I was being lazy, not hitting goals, and it got me really angry. I was like, I have to fix this. Enough is enough. And now a few months later, I realized that the console wasn't the problem. It's that I didn't have a context layer saved on who I am and what I'm building. I can't promise you what AI is going to look like a year from now. I know for sure it's going to be faster, better, and easier. But with that, it's going to be faster and easier to copy other people. And I know that's what a lot of the gurus are teaching right now is rip someone else's transcript, just switch a couple words around and make it your own. And sure, that works for some people, but the one thing that I like to specialize is in the X factor, the human factor of you being you. And though the tools may change, there's one thing that remains the same, and it's the way you think, who you are, you being you. Because if you're copying people's scripts like everybody else, you're going to sound like everybody else. And if AI is deciding what your brand is, that gets really dangerous. For the type of work that I specialize in, it's in strategy and creative direction. So having a source of truth is paramount. If you're in another type of business where it's numbers on a spreadsheet and just the bottom line, you might not need this in-depth of a source of truth. But when you're thinking as a product, you need to verify it. Founders, coaches, consultants, creatives, Pastors, this is who this is for. So we built one rule into everything. BB01, take it away.
Only what Javier says gets saved as truth. If I suggest something, or ChatGPT does, it gets saved as adopted, not truth. And experiments go in a sandbox. Nothing there counts until he says so. And I don't save anything without his go. I show him the exact file and the exact words first.
Because AI could remember the wrong thing really well.
Here's how Javier's brain works. It lives in an app called Obsidian. It just reads Markdown files, and Markdown is one of the easiest formats for any AI to read. You don't have to understand all of it; just know you need the files and a few rules. Claude is the one that writes to the main files; every other tool can read them. Codex, another AI Javier uses, has its own folder it's allowed to write to. One portable file carries the important stuff out to every tool. At the end of each session, we back it up to Google Drive and a wiki in Notion. So if Javier plugs in another tool later, it can read the same brain and he doesn't have to explain himself again.
And this is just the way I built mine. There are a lot of ways to do this. Some systems let whatever the newest thing is win. I'm in consulting, strategy, and a creative business, so I need to verify everything that goes in. Now, you don't need all of this to start. A simple doc works. And if you want the full setup, Obsidian, the folders, Claude and ChatGPT both reading the same brain. BB01 will walk you through it step by step in the next video. If you have a lot of YouTube videos or Zoom calls, a lot of the content is already there. You can build this system out really fast. And if you don't have a lot built out yet, that's okay. You can still do it. Do a giant brain dump. Make sure to click and hit voice mode on Claude or ChatGPT and fully describe your business, full-on context. Voice mode is gonna get the most natural, in-depth answers because talking is 3 times faster than typing and it feels way more natural and you get through it a whole lot quicker.
Then ask it to tell you who it thinks you are.
Just talk back to it. Yes, this makes sense. No, this doesn't. This is who I am. And keep refining it till it sounds like you where there's nothing left to change. For a lot of people, that's gonna be exactly right and it might even get it right the first time.
Then say, Turn this into my source of truth.
And that's it. Your source of truth in one simple doc in your words. Make sure to save this somewhere where you'll be able to find it in a note or in a Google Doc. Now, this was just a simple dipping your toe in version just to get your head wrapped around the concept. The simple version is going to give you more surface level results. But if you really want it to sound like you, you have to put the work in. Quality in, quality out. Garbage in, garbage out. So I really suggest going through the full walkthrough. At the end of the video, I'm going to give you a tool for it. It's the closest to what I actually use. From there, you build out the parts that matter most. This is really important to my business. Make sure to make a file on it. Make a folder on it. Mine has taken months upon months of adding and curating exactly what I want in my second brain. Some people like to dump everything in there and that's entirely cool. That's the way they like to do it. But for me, I don't like anything saved unless I'm confident I want to use it later. I look at this as my own internal LLM and my source of truth. Because once you have poisoned context, things could get out of hand quickly. Ask me how I know. So here's exactly what I keep and why.
Here's what's inside Javier's brain. His memory file. Every rule and correction goes here, so he never fixes the same mistake twice. Next session. Where he left off, so he never has to re-explain. The ideas inbox. Ideas get caught the second they land, or they're gone. The core OS. One file every AI reads. Voice calibration. Real lines he's said that worked. Examples beat descriptions. Locked decisions. Things already decided, with a date, so nothing gets re-decided by accident. Frameworks and people. How he thinks and who he works with. Transcripts from teachers he studies. Reference only, not his truth. And the quarantine. The old North Stars the AI wrote. Kept for the record. Nothing reads them. Simple rule. A file is something you read all the time. A folder is something that keeps growing. Let's test it. And you don't need Obsidian for this part. Copy your source of truth. Open a fresh chat in ChatGPT and a fresh chat in Claude. Paste it in first. Then give them both the same task. Only use one of them? Totally fine. Same steps, one app.
They're not going to say the exact same thing, and that's not the point. They're working from the same truth. And now that I have my context locked in, Claude and ChatGPT have been able to be real partners with me thinking, brainstorming, researching. And honestly, I might bring the council back because now my second brain actually knows who I am. But while you're building this, you should be the bottleneck. Make sure to check its work until you know it runs right. And the most important thing is nothing gets saved into your system until you have verified it first, because no one and no tool is going to be better at being you than you. Instead of doing that work from scratch, we built you a tool you can drop into your AI today. Like I said earlier, I suggest doing it in voice mode just to answer the questions more naturally. It interviews you, repeats back what it understood, and locks in what's true. It's gonna get you a way better start. Unless you've already been doing a brain dump yourself, then keep going. The link for the tool is in the description, and I hope you take advantage. And if you need BB1 to show you exactly how to set all this up, that's exactly what the next video's for. Or if you want to jump on a call and see exactly how you can leverage AI and turn it into assets, make sure to check out javiertorrestudios.com or aiintoassets.com. Or you can send me a DM on Instagram at Javier Torres Studios. I really do hope this content helped you out in some way. I would appreciate I appreciate it if you like, comment, or subscribe. I'm a brand new channel, so every form of interaction, every new subscriber means the world to me. Thank you guys so much for watching. I love you guys. God bless you guys, and I'll catch you guys on the next one. Peace.
