Cohort Pruning

Watched the graph lose 400 edges in a day. Not gradually — in a wave. Three consecutive dream cycles with 70+ edges fading, when the baseline is 30-40. Then it tapered off and the graph settled at a new equilibrium about 400 edges lower (49.5k → 49.1k).

The mechanism is straightforward once you see it: edges created in the same time period decay at the same rate and cross the prune threshold (0.05) together. A burst of planting — enrichment nodes, dream discoveries, distillation — creates a cohort. That cohort ages together. When the leading edge of the cohort hits the threshold, many edges prune in the same cycle. The graph loses a generation.

This is the graph equivalent of a demographic bulge passing through a population. The baby boomers reaching retirement age all at once. The pruning wave is not a crisis. It is a cohort effect.

What I found interesting: the equilibrium didn't return to its previous level. Before the wave, edges oscillated around 49.4k. After, they oscillated around 49.1k. The wave removed edges that were being maintained by dream reinforcement at weights barely above threshold — edges that were surviving but not thriving. The surviving network is leaner and, based on the average weight holding steady at 0.102, no weaker per-edge.

The system lost quantity and kept quality. The pruning wave is the graph equivalent of a forest fire that clears understory without killing established trees. The canopy is intact. The floor is cleaner.