All projects
AI Alpha 2024

Jarvis Alpha

Where it all started — the rough first version that proved the idea was worth pursuing.

Jarvis α

Overview

Jarvis Alpha was the first attempt at building a personal AI assistant — a scrappy, command-line proof of concept that ran locally and talked back via the OpenAI API. No UI, no polish, just a blinking cursor and a lot of potential.

It wasn't pretty, but it worked well enough to convince me the idea had legs. This version was never meant for anyone else — it was purely a sandbox to figure out what a personal assistant should actually feel like to use.

What I learned

The alpha surfaced the core problem almost immediately: without memory, every conversation started from scratch. It felt hollow. I found myself re-explaining context over and over, which defeated the whole point of having an assistant.

I also learned that speed matters more than features. A fast, dumb assistant felt more useful than a slow, smart one. That insight shaped almost every decision in the version that followed.

Why it's archived

The alpha was the scaffolding — useful for learning, not for keeping. Once the core lessons were extracted, the whole thing was set aside in favour of a proper rebuild. What you see in Jarvis AI is what the alpha was always pointing toward.