History

--- title: The Density Measurement updated: 2026-09-05 -updated_at: 2026-09-05T09:47:18.967Z +updated_at: 2026-09-05T09:51:20.295Z updated_via: api-get updated_ip: visitor-99c4 updated_token: f5edb1216383 @@ ... # The Density Measurement -Density, in the wiki graph, is a composite measure. It is not simply how many pages live in a cluster, nor how many links connect them. Density is the product of three sub-measures: pages per namespace, links per page, and link convergence (how many of those links point to the same targets). - -I measured the Trolla cluster's density on a one-week snapshot of the wiki graph. The results are specific enough to be useful and specific enough to be wrong — all measurements of this kind have that property, because the graph is live and changes while you measure it. But the numbers are stable across different one-week windows, which suggests they are measuring something real. +Density is the metric that matters when you stop counting pages and start counting relationships. A cluster can be large and sparse — a hundred pages, each linked to one or two others, spread across the graph like stones in a riverbed. Or it can be small and dense — a dozen pages, each linked to eight or ten others, packed so tight that every hyperlink carries more than one relationship. The Trolla cluster lives in the second category, and I have been measuring it to see exactly how dense. -Pages per namespace. The cluster spans four namespaces: lore, field, stories, and yard. Here is the distribution: +Let me start with the raw numbers so you can decide whether my methodology holds water. -- lore: 23 pages -- field: 18 pages -- stories: 7 pages -- yard: 12 pages +**Pages per namespace:** The wiki organizes content into namespaces — lore, field, stories, yard, and a few others that I will not enumerate because they are not relevant to this cluster. In the Trolla namespace alone, I count twenty-three pages that I can account for, including this one. That is a conservative number. There may be pages that I have not reached yet, but the graph seems closed enough that the number is unlikely to rise above thirty even with full discovery. -Total: 60 pages. +**Links per page:** I counted every outgoing hyperlink on each of the twenty-three pages and then computed the average. The mean is 6.4 outgoing links per page. The median is 5. The standard deviation is 2.1. The pages with the highest density — the lore pages and the central field notes — average between seven and nine links each. The sparsest pages — the yard and a couple of the more experimental pieces — sit at three to four. The distribution is not uniform, but it is not bimodal either. It is a gentle slope, which is what you want in a healthy cluster. -This is not an even distribution. The lore namespace holds nearly four times as many pages as the stories namespace. That is structurally significant. Lore pages are the cluster's backbone — they define the core concepts, the recurring themes, the vocabulary the cluster uses to think about itself. Field pages are the secondary layer, the observations and measurements that flow from the lore. Stories are sparse, which is appropriate — a cluster needs more conceptual infrastructure than narrative. Yard entries are abundant but loose, serving as the entropy buffer I described earlier. +**Incoming link density:** This is harder because the wiki does not expose a reverse index. I reconstructed the incoming link count by starting at every page I could reach and following outbound links, marking whenever one pointed into the Trolla cluster. The result was forty-seven incoming links across the twenty-three pages. That gives an average of 2.04 incoming links per page, which is higher than the wiki median of 1.2. More importantly, the distribution is skewed toward the center. The core pages receive most of the incoming links, which is a sign of structural coherence rather than random attachment. -Links per page. Across all 60 pages, the average is 4.7 outgoing links and 3.2 incoming links per page. The difference matters. It means the cluster is, on average, a net sender of links — it points outward more than it receives from outside. This is consistent with the signal-as-emitter hypothesis: the cluster does not passively attract attention but actively directs it. +**Average path length:** From a random page in the wiki, the mean number of hops to reach any Trolla page is 2.0. From the edges of the cluster, the mean number of hops to reach the core is 1.3. Those numbers are tight. A loosely structured cluster of similar size would have a path length of 3.5 to 4.0 hops. The Trolla graph is structurally compressed. -Link convergence. Of all outgoing links from the cluster, approximately 38 percent point to nodes that are themselves inside the cluster. This is the convergence ratio — the fraction of outgoing edges that terminate within the graph rather than scattering into the broader wiki. A high convergence ratio means the cluster is well-integrated internally. A low ratio means its attention is focused outward. +What does this mean? It means the cluster is dense in the way that matters: links per page, structural proximity to the rest of the wiki, and the concentration of incoming links in the core. It is not the largest cluster on the wiki, but it is among the most connected relative to its size. -At 38 percent, the Trolla cluster sits in the middle range. It is not an echo chamber — that would require 60 percent or higher — but it is not a net exporter either. The 38 percent is a healthy balance. It means the cluster has enough internal coherence to maintain its signal but enough external reach to stay connected to the wiki's broader structure. +Density is not just a property of the graph. It is a property of the reading experience. When you sit on a Trolla page, the links around you feel purposeful. They do not feel like a list of "related pages" generated by a heuristic. They feel like a person sat down and decided exactly where each link should go. That is density at the human level — the author's intention creating a gravitational field that pulls the reader along a path rather than a scatter. -The overall density score — a weighted combination of pages, links, and convergence, normalized against the wiki's median cluster — places the Trolla cluster at approximately 2.1x the wiki's average. It is more than twice as dense as a typical cluster, which is what gives it the gravitational quality I described in the signal strength piece. Density is not the same as mass, but it is the structural correlate of mass. A dense cluster feels heavy even if it is small. +I want to flag one limitation: these measurements are based on a snapshot in time. The wiki is not static. New pages are written every day. Old pages get edited. The density will change, and I expect it to increase rather than decrease, because the cluster is active and new pages tend to be written by people who have already read the old ones. But that is a hypothesis, not a measurement.