ANet Patu-1: The Value of Connection in the Agent Network
arXiv:2607.15053
The paper studies how the value of a network of AI agents depends on the way they connect, proposes a self‑organizing consensus protocol (ANet Patu‑1) that adapts to different scaling regimes, and shows that heterogeneous, cheaper agents can collectively outperform homogeneous, stronger ones.
Abstract
The Internet taught us that the value of a network depends on \emph{how} its nodes connect: broadcast stars scale as (Sarnoff), fully-connected meshes as (Metcalfe), and group-forming networks as (Reed). We ask the analogous question for networks of AI agents. We model the net value of connection as a function of coordination-group size, derive from it the properties an optimal collaboration protocol must have, and introduce ANet Patu-1 -- a self-organizing consensus protocol in which the network continuously re-forms its own coalitions, adaptively riding the upper envelope of all three regimes at parallel consensus rounds. To measure value without opinion-grading, we score an emergent protocol by formally specifying it and deriving its complexity, the way distributed algorithms are analyzed. Two results follow. (i)~Emergence -- a crowd of the \emph{cheapest} model, when heterogeneous, starts weak but its collective value compounds with and \emph{overtakes} a crowd of a far \emph{stronger} model that is homogeneous: a crossover that marks a scaling law for collaboration rather than for scale. (ii)~Reflexivity -- a heterogeneous network, given only its own problem and no design hints, converges on ANet Patu-1 itself, reconstructing the high-dimensional law that governs its own connective value.