Paco Cantero, our Co-Founder, runs four businesses in parallel. He's turning 50 this week. And he just built his own AI-powered personal knowledge management system from scratch.
No terminal. No VS Code. No coding background required. Just Claude Code's chat interface, a local folder, and a clear structure for his life.
Here's how he did it, and why it matters for anyone who feels like their PKM system is broken.
The Problem: Scattered Knowledge, Zero Context
If you use productivity tools, you probably recognize this pattern. Your journal lives in Day One. Your meeting notes sit in Tana. Your saved articles pile up in Reader. Your tasks are in Todoist. Your goals are... somewhere.
Each tool works fine on its own. But none of them talk to each other. And when you ask AI for help, it starts from zero every single time because it has no idea who you are or what you care about.
Paco had the same frustration. "I was struggling about really implementing AI in a natural way," he says in the video. "The AI inside Tana wasn't aware of my context."
That last part is the real issue. Context is everything.
What He Built: "Mindset"
Paco calls his system "mindset." It runs entirely on his Mac using three components:
A SQLite database stores all structured data: journal entries, contacts, interactions, goals, habits, projects. Think of it like a personal Notion database, but one that AI can read and write to directly. Paco has over 2,880 journal entries indexed in there.
Markdown files hold the skills (basically SOPs for AI), the CLAUDE.md context file, and readable content like articles and coaching outputs. These are plain text files that live in a regular folder on his computer.
Claude Code ties it all together. Not the regular Claude chat, not Claude in a browser. The Claude Code interface, which connects directly to a local folder and can read, write, and interact with everything inside it.
The key insight: you don't need to know what SQLite is. When Paco told Claude Code he wanted to build a personal system, it recommended SQLite on its own. It set up the tables, the relationships, the whole structure. Paco just described what he needed in plain English.
How It Actually Works Day to Day
Paco's daily workflow is simple. He journals. That's it. Two or three lines of natural language about what happened, what he's thinking, what he worked on.
From those few lines, his system generates:
AI coaching on every entry. Each journal triggers a "skill" (a markdown file that acts as an SOP for Claude). The coaching skill analyzes the entry, pulls relevant context from his 2,880+ journals, and delivers key insights and reflection questions. No scheduling a "review session." No weekly recap. Coaching happens every single time he journals.
Automatic CRM updates. When Paco processes a meeting through Tana, the system detects the people involved, logs the interaction, and connects it to his contact records. He never fills in a CRM field manually.
Content creation pipeline. When he writes articles, a skill generates the full article, thumbnail sentences, LinkedIn intro, tweets, quotes, and deep-thought discussion points. One effort, multiple deliverables.
Life dashboard with visual alerts. Red dots show up when he's neglecting a key area of his life. If he hasn't journaled about French practice in a while, the dot turns red. No manual tracking. The system calculates it from journal frequency.
The "Read Only" Principle
Here's something Paco was firm about: the interface is read-only. He never clicks "edit" or "create new entry" in the dashboard. His only input method is chatting with Claude Code in natural language.
"I forced myself not to use an edit or create new entry here," he explains. "My only entry point is Claude Code."
This sounds limiting, but it's the whole point. You describe what happened. AI handles the categorization, the database updates, the connections between entries, the coaching output. You just read the results and react.
Thomas confirmed it during the interview: "Everything that you show is read only. It's all automatically done by AI, by just engaging with AI."
The ICOR "My Life" Structure Behind It
This system isn't random. It follows ICOR's My Life methodology, which gives your entire life a clear hierarchy:
Dimensions are the top-level areas (think: Growth, Business, Family, Health). Paco has six.
Key Elements sit inside each dimension. Under Growth, for example, he has Chess, Coding, French, and Piano.
Goals connect to Key Elements. Each goal needs at least one project or habit linked to it. If a Key Element has no goal, the system flags it with a red dot.
Habits and Projects are how goals get achieved. Period. That's the ICOR principle: a goal can only be reached through a habit or a project. Nothing else counts.
Journals feed everything. Every entry gets tagged to the right Key Element, which rolls up to the Dimension, which connects to Goals. After 2,880 entries, the AI has deep context about every area of Paco's life.
The structure matters because without it, AI can't give you useful insights. As Paco puts it: "If you are not able to explain clearly to AI how your life is structured, it's impossible to really receive those amazing insights."
What You Need to Build This
The barrier is lower than you'd expect:
- Claude Code (part of the Claude subscription, about $100/month)
- A local folder on your Mac or Windows machine
- TablePlus (free) to view your SQLite database when you want to peek at the raw data
- MCP connections to your existing tools (Tana, Reader, Todoist, Day One, or whatever you use)
You don't need to open a terminal. You don't need VS Code. You don't need to understand databases. Claude Code handles the technical setup. You just need to know what you want your system to do.
Paco built the whole thing in a few weeks of part-time work. It didn't affect his four businesses at all.
Why This Matters
The real shift here isn't technical. It's about how we interact with our tools.
Paco stopped clicking through fields and forms. He stopped scheduling review sessions. He stopped trying to maintain a complex system manually. Instead, he journals naturally and lets AI do the organizing, the connecting, and the coaching.
"I have increased my time to think," he says toward the end of the video. That's the payoff. Not more data, not more dashboards. More time to actually reflect. The operational overhead of productivity systems, the custom fields, the mandatory dropdowns, the weekly reviews, all of that disappears when AI handles it from natural language input.
For busy professionals running multiple responsibilities, that's not a nice-to-have. It's the difference between a system that creates more work and one that actually works for you.
Watch the Full Video and Try It Yourself
The full 54-minute deep dive shows Paco's screen as he walks through every feature of his system. If you're a myICOR member, you can download the scaffold Paco built and start implementing this in your own Claude Code setup.
And if you want to see how Thomas is building an AI team using the same ICOR methodology for business knowledge management, check out the video linked at the end.