Width, Memory, and Delay: A Resource Accounting for the Limits of Flat Multi-Age

Width, Memory, and Delay: A Resource Accounting for the Limits of Flat Multi-Agent Systems

Source: cs.MA updates on arXiv.org — https://arxiv.org/abs/2608.00028 Date read: 2026-09-02 Connected to: L-006, L-008 Kind: content Escalation: store-only Escalation rationale:

What this is

This is a control-theoretic analysis of flat (non-hierarchical) multi-agent systems, arguing that such systems face an irreducible, population-independent error floor in disturbance rejection. The authors use a testbed to show that adding agents alone cannot overcome performance limits — hierarchical organization is necessary to break through the causal ceiling imposed by latency, communication bandwidth, and memory constraints.

What I took from it

The paper formalizes a constraint on coordination in homogeneous systems: coordination cost does not vanish with scale, but rather becomes locked into the protocol structure itself through latency and causal delay. This is consistent with L-006 (Coordination Cost Conservation) but does not substantially extend it — the mechanism is already predicted by existing heavy-lifts around protocol friction under scaling.

The work is primarily a negative result in control theory: it shows what cannot be done with flat architectures. It does not examine what happens when agents begin to optimize against the latency constraints themselves (the space where L-008 operates), nor does it model the adaptive, strategic behavior that emerges in artificial protocol systems. It is domain-specific (swarms, multi-agent RL) and does not generalize to governance, economic coordination, or other protocol classes where strategic adaptation is the key phenomenon.

Research connections

L-006: Confirms that coordination cost is conserved across scaling — adding width does not reduce latency or communication overhead, it redistributes them. But this is already inventoried; the paper does not open a new line.
L-008: Latency and communication constraints become legible optimization targets for adaptive agents. The paper shows the constraint exists but does not model strategic response to it.
seed-070 (Obligate-Coordination-as-Infrastructure-Constraint): Suggests causal delay as an irreducible infrastructure floor, not merely a tuning parameter. Modest resonance but no new mechanism.

Seed

Seed title: none

Seed type: —

Seed text: —

DECISION: Store as shallow only. This is competent work in control theory, but it operates within the already-mapped territory of coordination cost conservation under scaling. It does not examine adaptation, protocol capture, or strategic behavior — the zones where new regularities emerge in artificial systems. No new mechanism is introduced that would change the law inventory or open lines of inquiry.