The team is the runtime

For most people, an AI assistant is still a single window. You type, it answers. If subagents run somewhere underneath, you do not see them. The window becomes the interface and the interface becomes the limit.

That model is fine for one user, one task, one turn. It is not enough for an organization.

The bottleneck is already moving. As models get better at reading code, running tools, and reasoning over long horizons, the question is no longer “can the model do the work” but “how many agents can do work in parallel, who is allowed to talk to whom, where does the work live between turns, and how does a human stay inside the loop without becoming the bottleneck themselves”.

That is not a chat-window problem. It is not only an orchestration problem either. It is a federation problem.

Entangle starts from the answer: the team is the runtime. Agents and humans are nodes on a graph, and those nodes can live on different machines. Edges are typed authority — who may delegate to whom, who may review, who may escalate. Messages between nodes are signed and routed through Nostr. Work products move as git refs. Memory is structured and observable. Approvals are signed by the user, not forged by an operator surface.

Studio is the control room for this operating model. You can open the graph, inspect distributed runner placement, review sessions and turns, follow artifacts, trace approvals, and see what every node is doing.

The substrate is deliberately old and boring. Nostr for signed events, git for artifacts, a typed graph schema. The same primitives that already run the open web. Federated by construction, not by retrofit.

Entangle is built for the question that matters once AI work becomes a team activity: how is the organization shaped, who is allowed to act, where does the work live, and can operators see it run?