Loading
Every figure on this page is computed from the 205 article files at build time. Nothing is entered by hand, which means the numbers move whenever the corpus does — including the unflattering ones.
The wiki is deliberately narrow and deep: 205 long articles rather than thousands of stubs. Mean article length sits at 1,389 words and the median at 1,358, so the average is not being dragged around by a handful of outliers.
Four cuts through the same 205 articles. Portals overlap — an article can sit in up to three, which is why the portal column sums to 352 rather than 205. Status, type, and horizon are exclusive: every article has exactly one of each.
Shares are of the article total, so they add to more than 100%.
Type governs what an article must contain. A technology carries a readiness level; a person and an event do not.
Horizon records when a technology could plausibly matter at scale, as claimed by the people working on it. It is an attribution, not a forecast by this site.
84 of 205 articles carry a technology readiness level; the rest are people, events, or concepts where the scale means nothing. The mean across those 84 is 6.0, and the largest cluster falls in the translation band — TRL 5–7, 34 articles. TRL is a coarse instrument and it is applied conservatively here: a level is assigned to the specific claim an article makes, not to the field the article belongs to.
Edges are counted as undirected pairs: if two articles both link to each other, that is one edge, marked reciprocal. Reciprocity is the strongest relatedness signal the site has, because it means the connection was judged worth making from both ends. Density is the share of all possible article pairs that are actually joined. Most pairs have no reason to be connected, so the figure stays small however hard the corpus is cross-linked; the question it answers is how far the average reader is from any other subject.
Inbound links are a decent proxy for which ideas the rest of the field leans on. These ten are the load-bearing concepts.
Articles no other article points at. They stay reachable through search and the index, but never by reading — each one is a missing link somewhere else.
Every article has at least one inbound link.
8 links point at 8 topics that have no page yet. The ones requested most often are the ones worth writing first.
Length is bounded on both ends by editorial policy rather than by how much there is to say. The shortest article runs 913 words and the longest 2,043, a ratio of 2.2 to one. A reference work with a flat length profile is easier to read across than one where entries range from a sentence to an essay.
Buckets are 200 words wide, chosen from the observed spread rather than fixed in advance.
Word counts exclude frontmatter, code blocks, table markup, and headings, so they measure prose rather than file size. Reading time assumes 225 words per minute.