Why I’m Building My Rogue AI Second Brain
I made this AI after watching a NetworkChuck video about building a local AI and thought — why can’t I take this further and make it into a second brain for myself? My goal wasn’t just to run AI locally, but to have it pull my tasks, keep me accountable, and prevent me from falling into the feedback loop of second-guessing myself with ChatGPT.
I don’t believe in paying higher prices just because of inflation when most of this can be built with free, open tools. For me, the best way to truly own my privacy and data was to build the concept myself — and while doing so, I’d also get hands-on experience with Python, Streamlit, DevOps workflows, and Security+ study material. Each phase has become a way to measure and benchmark my growing toolkit.
My Roadmap So Far
To stay focused, I broke the project into phases. Each step adds a layer of capability to the dashboard while keeping everything local.
✅ Finished
- Phase 1: Vector Memory (Chat Log Recall) – completed July 13, 2025. The system now recalls past notes and documents on demand.
- Phase 2: Calendar Integration (Google Calendar) – currently syncing when I open the Memory Search tab.
- Phase 3: Notion To-Do Integration – pulls in my Notion tasks so I can see tasks and events in one place.
🔄 In Progress
- Phase 1a: Zep + AutoGen Agent – adding human-like memory recall on top of my vector memory.
⏳ Next Up
- Phase 4: Energy-Level Logic – connect Apple Health, Google Fit, and weather data to adapt tasks based on my energy and environment.
- Phase 5: Unified Prompt Chain – merge memory, personas, and energy awareness into one workflow.
- Phase 6: Docker + GitHub Actions – containerize the system and automate deployment with DevOps tooling.
Future Plans
Once the basics are solid, I want to “DevOpsify” this project — containerize it, version control it properly, and maybe even make parts of it shareable without giving away personal data. I also see opportunities to link what I’m learning for Security+ into the build itself — things like access control, logging, and identity management.
Takeaway
This project isn’t just about passing Security+ or replacing ChatGPT — it’s about proving that anything can be built locally if you’re willing to put in the time. You don’t need enterprise budgets or endless subscriptions. And if you fall on your face along the way? That’s part of the process. Every mistake is just another layer of learning in the second brain I’m building for myself.