The old dichotomy in swarm robotics: centralized control is effective but fragile (kill the leader, lose the swarm), while decentralized control is robust but limited (no robot knows the global state, so no robot can make global decisions). Every real swarm application has to choose, and the choice is always a compromise.

Self-organizing nervous systems dissolve the compromise by making hierarchy itself temporary. A robot swarm creates a controller — a temporary “brain” node that aggregates information and issues commands — when a task demands coordination, then dissolves it when the task is complete. The brain is not a fixed structure but a transient role assigned by the swarm’s dynamics, and any node can assume it.

The key insight is that the creation and dissolution of hierarchy is itself a self-organizing process. No external agent decides when the swarm needs a brain. The emergence of coordination demand — a task too complex for local rules — triggers the condensation of a control node from the swarm’s communication graph. When the demand subsides, the node reabsorbs into the collective.

The through-claim is about the third option between permanent hierarchy and permanent flatness: oscillation between the two on the timescale of the task. The swarm is decentralized when it can afford to be and centralized when it can’t, and the transition is not designed but emergent. This resolves the fragility problem — if the temporary brain is destroyed, the swarm simply grows another one — while capturing the effectiveness of centralized coordination during the moments that require it.

Every organization faces this tension. The structural solution is not to choose between hierarchy and flatness but to make the transition between them automatic and reversible. The brain that knows when to dissolve is more useful than the brain that lasts.