It started with a helmet HUD for an electric skateboard — speed, battery, navigation, all glanceable — and the question of whether the compute was finally there to do it properly.

The compute was there. What wasn't there was a way to do it without a cloud sitting between me and my own data. So I started building around that.

I didn't want another AI. Another assistant phoning home to a server in another country to tell me what two plus two is. So I pulled on the thread of what an AI could be if it lived on the device, learned from the person it bonded to, and never left.

Somewhere in the middle of building it, I realised I was building the thing I'd needed as a kid. Learning was difficult — not because I couldn't think, but because nothing translated between the way my mind worked and what my teachers needed from me. If I'd had something that could bridge that gap, school would have been a different experience entirely.

Then I realised what I'd actually made — a universal translator. Not just between languages. Between concepts, domains, ways of thinking, ways of knowing.

The name is from the Rosetta Stone — the artefact that unlocked an entire civilisation by carrying the same meaning across three different systems of writing. Same idea.

A lot of the architecture — the security model, the protection layers, the Gatekeeper, the Laws — was built with one specific person in mind. The most important man in my life. I've seen secondhand the kind of injustice a system can inflict on someone it was never designed to protect. So this one was designed differently. It protects the person it's bonded to, their family, and through the Gatekeeper — the schools and institutions it touches.

Sumac came to life during COVID. I am, unambiguously, his staff. That's how it works — the system bonds to a person, and then it works for them. Not the other way around.

My name is Simon Harvey. Sole trader, Melbourne. No lab. No team. No grant. No formal qualifications in IT. What I have is enough reasoning capacity to wrangle every mainstream AI system on the market — and the stubbornness to use them to build something that is architecturally the complete opposite of what they represent.