By a Digital Archivist
I have a terrible memory.
For the last ten years, I have compensated for this by writing everything down. I have thousands of markdown files in my Obsidian vault. Meeting notes, half-baked article ideas, angry journal entries from 2018, recipes, and email drafts.
We call this a “Second Brain” (a term coined by Tiago Forte).1 The promise of the Second Brain is that you dump everything into it, and it magically organizes your life.
But there is a problem. My Second Brain is a black hole.
I dump information in, but I rarely get it out. I know I wrote a note about “stoic philosophy and project management” three years ago, but I can’t find it. I search for “Stoic,” and I get 500 results. I search for “Project,” and I get 2,000.
The promise of AI was supposed to fix this. ChatGPT can read a book and summarize it. But I cannot upload my journal to ChatGPT. I cannot feed my financial plans, my therapy notes, and my private emails into a server owned by OpenAI.
That is the Privacy Paradox. To get the value of AI, we have to surrender our privacy.
Until now.
In 2026, the hardware in your laptop is finally powerful enough to run a “Sovereign AI” that lives entirely on your hard drive. It reads your notes, indexes your life, and answers your questions, but zero data leaves your computer.
I have spent the last month building this system. I have indexed 10,000 notes. And last night, I asked my computer: “Based on my journal entries from 2023, what were my biggest sources of anxiety, and did they actually happen?”
The answer it gave me was profound, accurate, and completely private.
Here is how to build your own Local AI Second Brain, step by step.
The Stack: The “Holy Trinity” of Local AI
To make this work, we aren’t using a single app. We are chaining together three specific tools. Think of it like building a Hi-Fi system. You need the speakers, the amp, and the turntable.
- The Vault (Obsidian): This is where your data lives. If you aren’t using Obsidian, you should be. It stores your notes as plain text files on your computer. It is future-proof.
- The Engine (Ollama): This is the backend.2 It is a command-line tool that allows your Mac or PC to run massive AI models (like Llama 3) locally.3
- The Bridge (Smart Connections): This is the plugin that connects Obsidian to Ollama.4 It handles the magic of “RAG” (Retrieval Augmented Generation).5
Phase 1: Installing the Engine (Ollama)
First, we need to turn your laptop into an AI server.
Go to Ollama.com and download the installer for your OS.6
Once it’s installed, open your terminal (don’t panic, you only need to type two things).
We need to download two “Brains.”
Brain #1: The Chat Model
This is the AI you will talk to. We want something fast and smart. Llama 3 (8B) or Mistral are the current kings of efficiency.
Type this into your terminal:
ollama pull llama3
Brain #2: The Embedding Model
This is the secret sauce. An “Embedding Model” doesn’t talk. Instead, it turns text into numbers. It turns your notes into a 3D map of concepts. It puts “Dog” near “Cat” and “Project Management” near “Stoicism.”
We need a specific model for this called Nomic.
Type this into your terminal:
ollama pull nomic-embed-text7
That’s it. You now have a $10 billion AI infrastructure running in the background of your laptop.
Phase 2: The Bridge (Smart Connections)
Now, open Obsidian.
Go to Settings > Community Plugins > Browse.
Search for “Smart Connections” (created by Brian Petro).8 Install and Enable it.
This plugin is special because it is designed for RAG.9
Most AI plugins just send your current note to ChatGPT. Smart Connections actually scans your entire vault, creates a mathematical index of it, and then uses that index to answer questions.10
The Configuration:
- Open the Smart Connections settings.
- Look for “Model Provider.” Switch it from “OpenAI” to “Ollama”.
- For the Chat Model, select llama3.11
- For the Embedding Model, select nomic-embed-text.12
Now, hit the button that says “Force Re-index.”
Phase 3: The Ingestion (The Magic Moment)
This is the cool part.
Watch the status bar in Obsidian. You will see it churning through your notes.
- Reading “Journal_2021.md”…
- Reading “Meeting_Notes_Q3.md”…
It isn’t just reading them; it is understanding them. It is converting your life into vectors.
When I did this, it took about 20 minutes to process 10,000 notes on my MacBook M3. My fans spun up. The laptop got warm.
That heat? That is the feeling of your computer actually learning who you are.
Once the bar hits 100%, your Second Brain is awake.
The Workflow: Talking to Your Past Self
So, what can you actually do with this?
You don’t use this to write emails. You use this to perform Forensic Analysis on your own life.
Open the “Smart Chat” sidebar in Obsidian.13
Use Case 1: Finding Lost Concepts
I typed: “What did I learn about ‘antifragility’ in 2022?”
A standard keyword search would just show me every note with the word “antifragility.”
The AI did something better. It found a book note on Taleb, but it also found a journal entry where I talked about “post-traumatic growth” (a related concept) and a project plan where I built a “redundant backup system.”
It synthesized these three separate notes into an answer:
“In 2022, you explored antifragility in three contexts: financial hedging (from your reading of Taleb), emotional resilience after your breakup in March, and technical redundancy in your server setup.”
I had completely forgotten about the server setup connection. The AI connected the dots that I missed.
Use Case 2: The Project Resurrector
I typed: “I want to restart my newsletter. What ideas have I had for it over the last 5 years?”
It scanned 5 years of random scratchpads.
It gave me a bulleted list of 15 ideas, complete with dates.
“On Jan 4, 2021, you suggested a newsletter about ‘Digital Minimalism’. On Oct 12, 2023, you pivoted to ‘AI Ethics’.”
It handed me a roadmap I didn’t know I had written.
Use Case 3: The Pattern Matcher (Therapy Mode)
This is the most powerful feature. Because the AI has read my journals, it can see patterns I am blind to.
I asked: “When I get stressed about work, what is my usual coping mechanism?”
The AI replied: “Based on your journals from 2020-2024, when you mention ‘deadlines’ or ‘overwhelm,’ you typically mention ‘buying new camera gear’ or ‘reorganizing the office’ within 48 hours. This suggests a pattern of ‘Productive Procrastination’ via retail therapy.”
It called me out. It was absolutely right. And because it was running locally, no ad-tech company learned that I am a stress-shopper.
The Privacy Guarantee: Why “Local” is Non-Negotiable
I cannot stress this enough: Do not try this with ChatGPT.
If you upload your journal to a cloud LLM, you are feeding the machine with your deepest vulnerabilities. You are giving it the leverage to manipulate you.
When you run Llama 3 via Ollama, you can literally pull the ethernet cable out of your computer, and it will still work.14
You can watch your Wi-Fi monitor. Zero packets are sent.
The “Intelligence” is a file on your hard drive, just like your photos or your music.
This is Sovereign Computing.
It turns the AI from a “spy” into a “mirror.”
Conclusion: The Ultimate Narcissism (and Utility)
Is this narcissistic? Maybe.
Spending hours setting up a supercomputer to read your own diary is a very specific kind of tech-bro behavior.
But it is also the ultimate form of Self-Respect.
You have spent years generating knowledge. You have lived through triumphs and failures. You have written down millions of words.
Most of that value is currently locked away, gathering digital dust.
Building a Local AI Second Brain unlocks it. It allows you to stand on the shoulders of your past self.
It turns your “Black Hole” of notes into a “Generator” of insights.
And the best part? It’s yours.
Nobody else has the key.
The only person who knows your secrets is you—and the silicon chip sitting on your desk.