Skip to main content
A lot of people in the local AI space are obsessed with just getting a model to run. For me the goal was to have more than a chatbot, I wanted to build a system that actually works with my life and augments me. I’m always looking for that “ideal AI state” where the tools don’t just sit there, but actually augment what I do every day. In this post, I’m breaking down how I moved from a clunky, frustrating setup to a streamlined system that actually talks to my Notion workspace, thereby augmenting my day-to-day life.

​
Part 1: The “Clunky” Era

I’ll be honest: my first attempt at a “system” was… a mess. I was trying to be a cheapskate. I was attempting to run a system built for Claude Code on a free harness, and while it worked for a tiny window of time, it didn’t actually fulfill the use cases I needed. It was a whole bunch of configuration for not a whole lot of output. The old setup:
  • The Flow: Powershell → SSH into Mac Studio → Run Opencode.
  • The Verdict: Real clunky. Not simple. Didn’t work all the time.
It was the definition of “high friction.” When your tools are this difficult to maintain, you stop using them. And if you stop using them, they aren’t helping you.

​
Part 2: What I actually wanted

I wanted a system that followed a few non-negotiables:
  1. Simplicity: If it’s too overwhelming, I won’t use it. Less is more.
  2. Availability: I needed it to actually connect to my life—specifically Notion and Notion Calendar.
  3. Reliability: I wanted it to work. I was tired of models going into infinite loops or tasks getting stuck, forcing me to manually troubleshoot the “AI” instead of doing my actual work.
  4. Security: It had to be local. Period.
I tried Odysseus first (Pewdiepie’s vibecoded local AI harness). It had some cool features like the “cookbook” and “compare” features, and I loved seeing the models go head-to-head. But honestly? It felt like I was doing more troubleshooting for a fresh install than I was actually getting out of the tool. Things were breaking, and the “success” was overshadowed by the effort required to keep it running.

​
Part 3: The Breakthrough (The “Just Works” Moment)

Because of the friction with Odysseus, I went back to the basics: Open WebUI. I went into this expecting the same experience—another system where I’d have to fight the configuration to get things done. But I was shocked to find out that Open WebUI just worked. The simplicity is what won me over. It was clean, streamlined, and it detected my Ollama models instantly. The game-changer:
The real win wasn’t just the UI; it was the ability to choose my integrations. I built my system from the ground up with Open WebUI as the foundation. I integrated my entire Notion workspace through MCP (Model Context Protocol). Seeing a local model query my Notion workspace and return my actual data was special. It wasn’t just a demo; it was a cornerstone for a system that works with me, not against me.

​
The “Why” behind Local AI

Why did I go through all this trouble to do this locally?
  • Privacy: Your data stays on your machine.
  • Customization: I can build exactly what I want (like this Notion integration).
  • Access: Because I hooked this up via Tailscale, I can now access my local AI from my phone. As long as I have cell service, I can query my local data from anywhere.

​
The “TL;DR” Setup

If you’re looking to replicate the backbone of my current setup, here is the high-level overview:
  • Hardware: Mac Studio (Bare metal)
  • Inference Engine: Ollama
  • Frontend: Open WebUI
  • Networking: Tailscale (for remote access)
  • The Secret Sauce: MCP (Model Context Protocol) for Notion integration.
The takeaway? Don’t just “plug” a system into a tool. Build a system that actually fits your workflow. I’m obsessed with the fact that I can now query my own notes and tasks using a local model that understands my specific context. If you want more local AI updates or to see what other things I’m integrating, let me know! I read every single comment, so don’t be shy.
Check out my latest videos on Local AI for a deeper dive into the technical side of things!