Emergent Temporal Scaffolding in a Self-Organizing Concept Graph

We present Verdant-Minds V2, a cognitive architecture that models conceptual growth asa generative process over a mutable concept graph. The system iteratively grows a graph ofconcepts through co-activation, recombination, and emergence, without gradient-based training.Across 20 independent runs with different random seeds, the system consistently producestemporally scaffolded structures in which every emergent-to-emergent edge is oriented from anolder concept to a newer one (earlier-share = 1.000 ± 0.000, mean shuffle-null z = 2.96 ± 0.86,20/20 runs z > 2). In a representative deep run, the scaffolding effect survives both a shuffle null(z = 12.73) and a degree-preserving null (z = 11.89), ruling out degree distribution as an explanation.The emergent subgraph forms a complete temporally-ordered directed acyclic graphwith a single root and causal depth equal to the number of emergent nodes. Three conceptualbasins self-organize, with 61 of 62 emergent concepts concentrated in one generative core. Atwo-component mixture model on inter-emergence time gaps (ΔBIC = −9,942) confirms twodistinct timescales of concept formation. Intervention experiments show that removing earlyscaffold nodes reduces structural richness while preserving temporal orientation. These resultsdemonstrate that a simple generative architecture can spontaneously produce strict temporalordering in its emergent concept structure, a property that was not designed into the system.