MagVin Desk v0: The Problem

The Problem

I had fifteen years of documents sitting idle on my drives, and I finally wanted to do something with them.

My financial records stretched back a decade: receipts, subscriptions, and spending logs for a four-person household. I'd been tracking expenses for years, yet I'd never actually analysed them. The data existed, but the insights didn't.

Professional materials had accumulated across twenty years of international training and curriculum development, including course designs, assessment frameworks, workshop resources, and anonymised participant feedback. I'd trained professionals across the Middle East, Southeast Asia, and China, and I had more than enough material to build a serious portfolio if I could ever find a way to make sense of it all.

Medical records for my entire family from clinics across five countries also sat idle on my hard drive, and when you move internationally, no one carries your history forward. So I wanted to build our own longitudinal health archive, something comprehensive enough that I could hand it to a new doctor anywhere in the world and say, "Here's what you need to know."

In short, my problem was that I had the data, but I couldn't do anything useful with it.

What I Tried

I started on macOS with an M2 MacBook Air maxed out at 48GB RAM and 2TB storage, alongside a 2019 iMac I'd upgraded over the years. I’ve been running an Apple-focused household for a decade, and I genuinely love the ecosystem.

I tried Automator workflows for file organisation and experimented with Shortcuts, building them with ChatGPT's guidance. I set up cloud sync chains to move voice notes from my phone through transcription into Markdown, and I tested AI-assisted journaling by feeding daily notes into ChatGPT and copying the summaries into Day One.

Some of it worked, and the workflows handled simple tasks reliably. But nothing I built could scale to what I actually needed: running local language models, processing documents in bulk, or doing serious AI inference. That kind of work requires GPU compute, and even on maxed hardware, my Macs weren’t really designed for it.

The Ceiling

I still love working in macOS. My M2 and iMac sit beside my PC right now, and I use them daily alongside my iPad and iPhone. The Apple ecosystem handles communication, writing, and creative work beautifully, and I have no intention of leaving it behind.

Unfortunately, it doesn’t yet handle local AI very well. The work I wanted to do required a powerful GPU for offline document intelligence, vector embeddings, and language-model inference. Apple doesn't ship that capability in a consumer machine, so I realised that I needed different hardware to move forward.

The Decision

In July 2025, I bought an Alienware Aurora R16: Core i7-14700F, RTX 4070 Super, and upgraded it immediately to 64GB RAM. It was my first PC in over ten years, and the decision came down to practicality.

I wanted a desktop with serious compute power that I could plug in and start working with immediately. The Alienware was available locally, reasonably priced, and powerful enough to run everything I had in mind. The purchase made sense.

I wasn't yet sure what I was building, but I knew I finally had the hardware to find out.

What Came Next

The PC arrived July 11, and by mid-July, I had my first working prototype: a tangle of scripts, folders, and ambition that would eventually become MagVin Desk v1.

That's where the real story of architecture begins.