Note: This article was originally published on LinkedIn.
In my previous blog, I shared why I started building a Second Brain. It was a quest for peace amidst digital chaos and a way to not just consume, but to tame the exponential power of AI. Today, we’re diving under the hood. How do you build a robust, flexible foundation in a Windows environment without getting locked into the whims of one specific AI provider? Welcome to the blueprint of my digital nervous system.
The Philosophy: Freedom over Convenience
As a Windows user, I have a clear philosophy for my tech stack: it must be robust, but also extremely flexible. We live in an era where the AI market moves faster than we can keep up with. Today’s ‘SOTA’ (State of the Art) is tomorrow’s ‘Legacy.’
My biggest fear? Vendor lock-in. You don’t want to build your entire knowledge system around a single interface, only to discover later that another model is twice as smart for half the price. The engine of your Second Brain must be independent of the interface you use to interact with it.
The Gemini Pro “Opportunity”
My journey began with a stroke of luck: six months of free access to Gemini Pro with the purchase of my new phone. It was a great opportunity, but it triggered a critical question: was I really going to dump all my thoughts into Google’s browser environment?
That felt like “scattershotting.” Working in a web interface is fleeting; you are a guest on someone else’s platform. I wanted to be the owner of my own data. What if I want to switch to Anthropic’s Claude in a year? Or what if local LLMs (Large Language Models) running on my own hardware suddenly take the lead? I needed a foundation that was model-agnostic.
The Core Requirements
Before I started, I established a list of non-negotiable requirements:
- Windows-native: No hassle with complex emulations; it needs to run smoothly in my daily work environment.
- Future-proof: The ability to switch AI providers (OpenAI, Google, Anthropic, or local via Ollama) at any time without the architecture collapsing.
- GUI over CLI: While I’m not afraid of a terminal, I want a professional graphical interface for my daily workflow. Stability comes from clarity.
- Local First: Processing can happen in the cloud, but the storage and control must remain local.
Discovering Goose
After intensive sparring with both Copilot and Gemini, one name kept rising to the top: Goose.
Developed by the team at Block (formerly Square) and now part of the Agentic AI Foundation (AAIF), Goose turned out to be the missing link. It’s not just a simple chatbot; it’s an agentic assistant that nests within your own file structure.
Why Goose is the ‘Middleware’ King:
- MCP (Model Control Protocol): This is the holy grail. Goose works with MCP and local extensions. This means I speak a “language” that is independent of the AI itself. If I want to switch models tomorrow, Goose handles the translation; my folder structure remains identical.
- Mature GUI: No cryptic configuration files, but a clean wizard setup for API keys and model selection.
- Contextual Awareness: Because Goose runs locally, it can truly “read” my files. It understands folders, sees connections, and can execute tasks within my Windows environment. This is crucial: a Second Brain must be able to save, process, and update its own instructions.
Initialization: The “Second Brain” Interview
Once the technical connection was established, the real work began: laying the foundation. I used the starter prompt by Brian Madden for this. The process works like an in-depth interview; the AI asks questions to clarify your intentions and goals.
Tip: Be as detailed as possible during this interview. It’s better to give long answers than short ones. This forces the AI to fully grasp the context and build a folder system that truly works for you. Don’t worry about micromanaging the folder structure, the AI will find the data when it needs it.
After the interview, the system builds a folder structure, including a System folder. This contains the instructions for your Second Brain in Markdown (.md) files. This is essentially a manual the AI writes for itself. The beauty? You can read these files yourself and manually nudge the system if needed. But my advice would be to let the 2nd Brain change it’s own instructions by telling it why it need to change and how that would work for you. The Markdown files can be read by tools like Obsidian.
My interview resulted in a clear vision: I want to grow both personally and professionally, building the “Patrick Kaak” brand as a trusted advisor and subject matter expert. I instructed my Second Brain to challenge me looking from different views and act as a critical sparring partner.
The Golden Rule: “No file, no memory”
During the build, I hit a snag. Goose is a powerful agent with its own extensions (like an internal to-do list). This is handy until you open another tool and realize that data is nowhere to be found.
My mantra is now: No file, no memory. Everything important must land as a physical file in my folder structure. I regularly perform a “sanity check” by asking: “What are my current priorities, and in which file did you record them?” If it answers from its temporary cache without updating the file, it gets a (figurative) slap on the wrist.
Training the Brain: My Voice, My Style
Standard AI output often sounds like an overeager, bland intern. To make my Second Brain a true extension of myself, I ‘fed’ it all my old blogs and texts. I had saved a lot of them during all the years in an archive, so I could import those quickly. I also exported my full LinkedIn profile including every answer I ever wrote in a discussion. This taught the model my tone of voice, sentence structure, and specific sense of humor and also gave it insights in what I find interesting. The result? The output feels like a mirrored version of my own thinking, saving me 80% of the time in post-editing.
The Fuel: Journaling as a System
A Second Brain without input is an empty library. Inspired by Jim Kwik (Limitless) and Tiago Forte (Building a Second Brain), journaling has become the heart of my system, using the book 101 Days Journaling as a guide.
- Morning Routine: I ‘program’ my day. We discuss intentions and affirmations. Here, I combine this with the “baby steps” philosophy of Michael Pilarczyk: what small step will I take today to reach my larger goals?
- Evening Routine: This is where the magic happens. I process the inbox, celebrate successes, and let go of the day’s baggage. The AI checks if I’ve met my personal goals (like moving more) and business targets.
The best part? The AI acts as an emotional filter. Negative thoughts are reframed into learning points. Furthermore, the system creates new notes from the reflections, builts up a todo list, and draws connections between my daily reflections and old notes I had long forgotten.
The Tweakers Feed: Intelligent Filtration
We are drowning in information but starving for knowledge. Since I spend a lot of time on Tweakers.net, I set up a workflow where Goose monitors the RSS feed.
- Selection: It presents a shortlist based on my expertise (hacks, data leaks, security trends).
- Synthesis: It creates a 5-point summary for the approved articles.
- The “Why”: Applying the Kwik/Forte tactic, I force myself to note why a specific piece is relevant to my work. This builds a living archive with context for future blogs or advice. And better, the AI also creates a ‘why’ and checks this with my ‘why’ to discuss differences in thoughts.
Security and Portability: The OneDrive Migration
A brain tied to a single hard drive is a risk. A computer crash would essentially be a digital lobotomy. That’s why my entire folder structure lives in OneDrive. This provides backup, version control, and above all: omnipresence. Whether I’m on my desktop, my laptop, or in a cloud environment, my brain travels with me.
What’s Next?
The infrastructure is ready. The foundation is poured, and the walls are up. But a house only becomes a home when it’s lived in.
In Part 3, we’ll look at the ‘Capture’ phase: how do I ensure that every thought, link, and idea I encounter during the day lands seamlessly into this system via an omnichannel capture workflow?
Until the next update. Keep your brain sharp, but your system sharper!
Did you find this interesting? What tool do you currently use to manage your digital notes, and does it feel like a system or a graveyard of unread text? Let me know in the comments!