Definition
Plain language
A system where many AI agents organize themselves like a research lab to run long experiments.
As stated in the literature
A self-organizing multi-agent framework for long-running scientific experimentation, coordinating analyst and experiment agents through shared experimental state — logs, a forum, and a dead-end registry — without a central planner.
Why it matters: It enables long, complex research efforts to proceed without a central planner, which matters when no single agent can hold the whole project in view.
For example, several agents share a common log, a discussion forum, and a list of dead ends so they can run a months-long experiment together without one boss directing them.
Heard on the show
“… The paper itself is called "AutoScientists: Self-Organizing Agent Teams for Long-Running Scientific Experimentation," out of Marinka Zitnik's …”Episode 095 — Seven Wins to Zero: How Organizing AI Agents Like a Lab Changes the Search