How to build your OINC* SuperBrain in four weeks
Obsidian + Instinct + Notion + Claude. A system to retrieve what you save, connect it with what you do, and build with it.

How to build your OINC* SuperBrain in four weeks
I find an article I want to read. A tool I want to try. An idea that might work for a project.
I open a tab.
I send myself the link on WhatsApp.
I save it to come back later.
And later I never come back.
A lot of those things ended up in a tab I closed unread, or in a chat with myself I'd never open again.
I started building a system to recover those ideas, connect them with what I'm doing and see what could come out of it. Saving was only the beginning. The interesting part was being able to use what I'd saved.
I call it the OINC* SuperBrain: Obsidian, Instinct, Notion and Claude.
In that order it sounds like oink. It made me smile and it stuck. 🐷
In this article I'll tell you how I've set it up, how you can set up yours, what still doesn't work and what came out of it without me looking for it.
The problem
Saving is the easy part. Anyone can have a bookmarks folder full of links they'll never open again. The hard part is deciding what deserves to stay, finding it when you need it and turning something you saved into something of your own.
That's what I ask of the system: that a reference can connect to an idea, a person or a project. And that when I come back to it, I can check where it comes from and what I'm relying on.
That's where the OINC* combination works for me. Connectivity helps me find relationships between what I learn and what I'm doing. Verification lets me go back to the sources and tell a fact apart from an interpretation or an idea yet to be checked.
That way, the projects and agents I work with can lean on a shared source of truth, instead of starting each time from loose information. For me, that's the difference: being able to retrieve what I know, cross-check it and turn it into work of my own.
What is OINC*
Four tools with different roles, connected in a single flow.
Obsidian stores and connects. It's the long-term memory. At least once a week I import the content from Notion into the vault. Imported notes stay read-only: corrections are made in Notion. In Obsidian I can see the graph and explore the relationships between what I've been saving.
Instinct captures. It's the agent I talk to on WhatsApp. I pass it a link, a screenshot or a voice note and it turns it into material I can review: a Spanish summary, the main ideas and a critical read. If the content is in English, also a full translation.
Notion organizes. It's the operational inbox. Every capture lives there with its properties. That's where I review and correct it.
Claude reviews. It works with the vault's content, helps me find connections and spot what doesn't add up. I use it to cross-check what I'm building, not to take it as valid without review.
The name reads OINC*: Obsidian, Instinct, Notion and Claude. Captures usually enter through Instinct, get organized in Notion and move on to Obsidian as long-term memory. But to work with them I don't always have to walk that path: I can ask Instinct about a tool or a GitHub repository, explore an idea with Claude, or go straight to Notion and Obsidian. I enter wherever it suits me. The rule I keep is that nothing goes back from Obsidian to Notion unless I ask for it.
How to build it in four weeks
This is how I'd do it if I started today. One week per piece, and no moving to the next until the previous one works on its own.
Week 1: Capture
A single input channel. Mine is WhatsApp, because it's what I have open all day.
And two doors:
What others write goes into a Notion database I call Superbrain. Articles, posts, books, podcasts, tools.
What I write goes into another base: Content Archive. You just start the message with «Mine:». It's transcribed verbatim. No summary, no translation, no critical read.
That separation is the most important decision in the whole system. If you mix what's yours with everyone else's, in three months you won't know what you thought and what you read on X.
Every capture always keeps the same things: original link, author, date, type, language, summary and extracted ideas.
Week 2 · Organize
In Notion, few properties and well defined.
The one that matters most is Type: Article, Post, Book, Podcast, Tool... A capture without a type doesn't appear in any catalogue.
And the rule that took me longest to learn: each database carries its contract in the description. What each property means, which options exist and which ones are retired.
It looks like bureaucracy. It's what allows several agents to write in the same place without breaking it.
In the first weeks I ended up with four places for the same thing: Tool, Repository, Resource and a Tools folder in Obsidian. Some of those options were created by Instinct on its own. Every new option split a concept in two. Since the contract is written, a tool is always a capture with Type: Tool. There is nowhere else.
On top of everything, an entry page: «System map · read this first». Any agent or person who arrives knows what's there, how it connects and what they can't touch.
Week 3: Connect
This is where Obsidian comes in.
The official import plugin brings the Notion databases into the vault. I run it myself, by hand, once a week, or sometimes earlier because I can't wait (the visual system can be addictive). Each note arrives with its Notion identifier, and the relations come down as links.
Two rules:
Imported material is not edited in Obsidian. If something's wrong, it's corrected in Notion and comes down in the next import. A single truth.
Connection before classification. Folders and tags tidy the library. What adds value is the link: which capture feeds which piece, which piece derives from which.
For that there are two relations between the databases. A piece of mine knows which captures it was born from. A capture knows whether I've already used it or it's still waiting.
It's the week when the graph stops being a ball of hair and starts drawing something.
Week 4: Review
Claude Code works inside the vault with a rules file, a CLAUDE.md, that it reads before doing anything.
The most useful thing in that file isn't the instructions.
It's the source hierarchy:
What I've published is a primary source.
What I have in draft is what I think, not what I've published.
What someone else writes is worth what its original link is worth.
An AI-made summary is never evidence.
Without that hierarchy, the system ends up citing the summary of a summary as truth.
And one safety rule: Claude doesn't move, rename or delete anything without showing me first a table with each change, its reason and whether it's reversible. I approve. Then it executes.
The last piece is a cloud agent that runs every Sunday. It reviews the brands I've worked with and their founders, looks for new news and job changes, and leaves me a report. It doesn't change anything on its own. It proposes.
Instinct reviews too.
And it also has to be reviewed
Instinct doesn't just store. Every capture comes out with a critical read: what it claims, what it doesn't prove, where it falls short.
Over time I started using it for something more. Before sending an important proposal, I pass it to it with one instruction: find where this breaks, not where it's fine.
It doesn't get tired, isn't afraid to disagree with me and doesn't need to look good to me.
But the review goes both ways.
In a mid-course audit I found captures stuck in the inbox, empty properties the contract marked as required, and missing translations. Nothing serious. All invisible if you don't look.
An agent that writes in your system needs the same three things as a new person on the team: written rules, someone who checks, and corrections in the right place. Mine are made in Notion and flow down to Obsidian on their own.
What I didn't expect: Caset*
Caset* started with a request to Instinct. One day I asked it to turn one of the articles I'd saved into a narration, like a podcast, and pass it to me on a page where I could listen to it.
A while later I had the article narrated, with its author's name and the original link.
I'd asked for it for myself. To listen to what I wasn't going to read on screen. Then I started sharing those first audio articles. They were well received and over the weekend I set out to shape the brand and the product.
That's where @caset_es came from: an app that turns what you save to read into cassettes you can listen to. It already has a website, a waiting list and its own brand system.
It wasn't in these four weeks' plan. The need to listen to the articles gave me the idea. In OINC* I had tools, references and ideas I'd been saving and could pull back to give it shape.
The key was being able to bring those pieces together when the need appeared. What I'd saved started turning into something of my own.
The rules that hold it together
Tools change. Rules keep this from becoming another junk drawer.
Other people's and mine, separate. Always.
Every note keeps its origin: the source and its author.
A synthesis can't rest on another synthesis alone. You have to go back to the sources so the system doesn't end up repeating its own conclusions.
Not one invented fact. What can't be verified stays as pending, not as a fact.
Nobody creates new properties without asking. Not the agents, not me in a hurry.
The numbers
As of October 27:
• 261 captures in the Superbrain.
• 112 tools and 103 articles.
• 26 captures of GitHub repositories.
• 92 entries in the Content Archive. 62 marked as published.
• 54 cassettes in the Caset* catalogue.
These numbers say what's in the system. They don't explain on their own what I've done with it.
I still don't have a reliable measure of how many captures have ended up in work of my own. One reference can feed several pieces. And a product like Caset* can come from connecting many of them.
Capturing gives me material. Deciding what's worthwhile and what I do with it is still on me.
There are two other figures I keep an eye on: 79% of captures have English as their original language and 39% link to X as their main source. That doesn't measure where I discovered each thing, but it does show a concentration. If I always look in the same places, I also limit the ideas I can connect with.
What I'd do differently
I'd stop fiddling with the structure much earlier. In the first two weeks I spent more time designing properties than writing. Since October 7 the structure has been frozen until November 2: no new databases or plugins. Only broken things get fixed.
I'd configure attachments from day one. The importer pulled every image and every audio into the vault. I ended up with 327 MB of mp3s doing nothing there.
I'd check links before changing a title. In my Obsidian import, the title becomes the file name. Changing it can break the links that depend on it.
I'd write the rules before letting the first agent in. What it can change, what it must preserve and when it has to ask. That contract should have been there from the start.
What I've learned
Week 1: Saving gives me freedom to think. I no longer have to remember everything.
Week 2: I don't want a content repository. I want a system that helps me make sense of what's in my head.
Week 3: Capturing isn't thinking. Saving lets me come back. What matters is what I do when I come back.
Week 4: The value isn't in having four tools. It's in what I can do by connecting them.
I started trying not to lose what I found. Now I'm interested in what I can build with it.
OINC* is still under construction.
If you're a founder or work at a brand and want to know more about OINC*, or you're looking for resources to get started, drop a comment or message me privately. I'll share what's working for me.


