I’ve been trying to write this without turning it into something it’s not.

It’s not a startup, at least not in the way that word works right now. It’s also not fully defined. But it’s already in motion.

I’m putting together a small, in-person environment in NYC. We will not be remote, we will not be distributed. We will actually be in the same space.

This sits at the point where something is forming but not fully clear yet—right before it gets turned into a direction. Most things get decided there just to move forward.

This doesn’t rush that move.

AI is part of this.

But not to generate outputs or decide what to build. Rather, to stay with what’s forming without turning it into something easier just to move with.

What I’m working in is not ambiguity. It’s gestation.

Things form on their own timeline. And the only real work is not interrupting that process by turning them into something too early.

To start, this is probably 2–4 people.

At least 1–2 engineers who can build agentic systems—not just pipelines or tools, but things that can hold state, contradiction, partial formation without needing to resolve it immediately.

People who don’t need everything clearly defined before they begin.

This is in-person in NYC for a reason.

A lot of what this depends on doesn’t show up remotely—how quickly something gets resolved, what gets forced into clarity, what gets flattened without anyone noticing.

This only works if that can be seen in real time.

There’s no fixed roadmap.