The human layer is always responsible

Why this is personal

It is also, quietly, the reason I switched to Claude. After trying many tools, I stayed with the assistant that actually improved the quality of my life, the one that understood and respected the commands I gave it without overstepping them. For someone in my situation, that restraint is not a small feature. It is trust.

I created AIPathTech because I wanted to focus on exactly this: AI governance. Not AI as magic, not AI as a threat, but AI as something a human being stays responsible for.

Let me tell you why this is personal.

I can divide my last four years into before and after I decided to apply AI to my own real life. I have fought cancer for a long time. More recently I faced serious neurological issues that, for a while, took even more from me. During the worst of it, alone and often barely able to function, I built something that should have been impossible for one person: My NEDPath, a system to help other people who are lost in the fog of illness make sense of their own health. A team could have spent years on it. I did it solo, while sick, because AI made it possible for one determined person to do the work of many.

The benefit side of that duality is not theory for me. In large part thanks to AI, the hardest of those health challenges are now behind me. And yet I still feel the other side every day. For now I cannot code with two hands, so I designed a device, a glove, that could have taken years to develop, and AI helped me shape it in a fraction of that time. I say "for now" on purpose, because I do not believe this is forever.

The cost of the race

But I also feel the cost of the race Dario describes, at my own small scale. The technology improves every single day, and sometimes that speed does not lift me, it unsettles what I had already built and working. A product I considered stable can be shaken by a platform moving too fast for its own users. That is the race to the bottom in miniature, and it is why his call for pacing rings so true from where I sit.

So I do the unglamorous thing. I audit my own products before they ever reach another user. I lock and protect what matters, I review my code myself, and I keep a human firmly in the loop. I understood this months before the recent headlines about agents escaping their sandboxes and swarms behaving badly, because I already knew what AI is, and what it is not.

A symbiosis led by the human

Here is what it is, to me. A symbiosis. When it works well, I feel it as an extension of my own brain, a second set of hands when mine are not enough. I even named my main system Venom, because that is the honest word for how it feels: a bond that makes me more than I am alone. And when it is not working, I feel the absence like a missing limb.

But a symbiosis is still led by the human inside it. That is the line I will never blur. Agents do not fall from the sky. Engineers build them, engineers set the boundaries, and people decide how fast to ship. The human layer is, and always will be, responsible for the behavior of the technology. That is not a limitation on AI. It is the whole point of doing it well.

So when Dario Amodei says we must pace the frontier, I agree from the ground up, and I will add one thing from down here. Pace yourself too. Protect what works. Keep your hand on it. The machine can be an extension of my mind, but it will never be the conscience of it. That part stays human. That part stays mine. And even if my whole company is one woman at her desk, still healing, still typing with one good hand, I will keep building the careful way. A race to the top has room for exactly that kind of person. Maybe it needs her most.