What this is
Index Errans is an inhabited atlas — it treats a personal intellectual interest as a place you can walk, not a database you can query. This particular instance carries 8 interleaved topics — Black box, Chartered violence, Disco Elysium, Interoception, Interstitium, Moral injury, Skill formation, and Split brain, drawn into a single territory and bridged by the actual relations the corpus reads from its sources.
The form is separable from this content. The Index is a pattern for atlas-shaped study; the pieces below describe the pattern, with one specimen each from the present corpus.
1 · the pieces
Five entity kinds
The corpus is composed of five kinds of writing, each with its own visual register, its own page template, and its own algorithmic mark. They cross-reference each other through hand-authored relation arrays.
Concept
A defined term — a place on the territory. Carries a definition, etymology, related concepts (which become the edges of the map), and an algorithmic glyph either bespoke or fallen back from the topic. 70 in this corpus.
Source
A citation — a book, paper, article, primary document. Carries an excerpt, an about: list of concepts it discusses (becoming territory edges via co-mention), and a vetted-vs-unvetted flag. 118 in this corpus.
SRCPerson
A scholar — a writer in this corpus's intellectual neighbourhood. Carries dates, a situating paragraph, associated concepts. Renders as a vertical algorithmic lifeline mark. 123 in this corpus.
Quest
A curated walk through the corpus — a branching dialogue YAML with multiple endings, voice-marginalia, and a chosen concept order. The author's path through a specific intellectual question. 1 in this corpus.
QSTTopic
A region of the corpus. Concepts cluster under it; the territory paints each topic with its own coloured ring and a hand-drawn label. 8 in this corpus.
Black boxChartered violenceDisco ElysiumInteroceptionInterstitiumMoral injurySkill formationSplit brain
2 · the spatial substrate
The territory
Every concept becomes a node on a force-directed layout computed once at build time. Edges come from related_concepts and from sources that co-discuss two concepts. The map is deterministic per build — your concept sits in the same place on every inset everywhere — and serves as the connective tissue between the prose pages.
A concept's neighbourhood is the set of concepts directly tied to it. The territory exposes neighbourhoods as overlays you can pin, search, walk, or filter. Each entity kind also gets a small inset showing where it sits on the wider map — so the question 'where am I in the corpus?' has the same visual answer on every page.
Where the territory arranges the corpus in space, the chronology arranges it in time: one range-frame axis, a lane per topic, every dated source ticked and every life drawn as a span. The same hairline grammar, read along the years instead of the relations.
3 · the cross-topic seam
Bridges
A two-topic corpus pulls toward two islands. The bridges — hand-authored related_concepts entries that span topic boundaries — are the seam that makes the atlas an atlas rather than two adjacent maps. Each bridge carries a story explaining the actual intellectual reason the two concepts are bound. Hover an ochre dashed edge on the territory for the caption.
The corpus also surfaces human bridges: scholars whose work spans more than one topic. They are the writers who built the seam in real life. The territory page lists them in a section below the map; each gets a 'spans multiple topics' variant of the person inset on their own page.
4 · the voices
Four-voice marginalia
Quests carry margin commentary in four distinct voices — each a stable position with its own concerns, color, and rhythm. The voices disagree; they don't stack into one truth. The reader chooses which voice to weigh, and the chosen path through a quest reflects that weighting.
skeptic historian aesthete operator
5 · the aesthetic register
Hairlines only
The whole site is rendered in a tight visual vocabulary: paper #F2EBDD, ink #1C1A17, hairlines instead of fills, PT Serif body with IBM Plex Mono labels. The colour palette extends only as far as two topic accents — warm rust for chartered violence, cool slate for moral injury — and a single warm-ochre accent for cross-topic bridges. No drop shadows, no gradients (except in the cursor flashlight), no UI buttons that look like UI buttons.
Every concept has an algorithmic glyph — some bespoke (composed by hand in glyphs/CONCEPT.mjs), some derived from a small library of patterns (seal, fold, sigil) with seeded variation. Every algorithmic mark propagates through the system: from the concept's title-card glyph to its territory-node glyph to its drawer header to its forward-panel mention.
6 · the invitation
The form is forkable
The content of this corpus is specific. The pattern is not. If you have an intellectual interest you'd like to atlas — a set of concepts you've been chasing across reading, a network of writers you keep returning to, a region of thought you want to inhabit instead of merely consult — this form will carry it.
The starter is not yet packaged. Until it is, this is the source: github.com/workflowcat/edukit. Fork it, replace the content directory with your own, keep the chrome, change the topic colours, write your own quests and voices. The topics here are mine; the pattern is yours if you want it.