Back to all projects

Project

Orbital Lab

A hands-on experiment in watching several AI agents make, critique, and improve something together without hiding how the work happened.

multi-agent systemsworkbenchartifactsdecision gateslocal models

Primary theme Agents, Simulations & Play

Also in Sensemaking & Decisions

Year 2026

Status Prototype

Why I made it

Multi-agent systems are hard to trust when the orchestration is invisible and outputs have no clear lineage. I wanted to see who did what, why an artifact changed, and where a human decision still mattered.

What came of it

The vertical slice runs a five-agent work loop, stores every run separately, versions and evaluates artifacts, and presents the state as an interactive spatial workbench.

A little more

Orbital Lab explores an interface for agent collaboration that feels alive without hiding the underlying execution record.

Represent orchestration as typed operations over visible project state, make artifacts first-class, and keep a human decision gate in the loop.

Keep exploring

You might also like

See all projects