The constraint
The first useful decision was to stop adding meanings.
It would have been easy to create separate labels for “extends”, “supports”, “compares”, “explains”, “uses”, and “evidence for”. That would make every new article a taxonomy exercise and turn the map into a dense assertion diagram.
Mannvit instead keeps four authored relationships and derives inverse links and guide membership at runtime. The author writes the forward claim once, with a note; the system resolves the reader-facing reverse context.
Builds on: a prerequisite concept or method.
Applies: a real build, case, or use of another method.
Challenges: a genuine later revision from evidence.
Related: a deliberate lateral connection, kept intentionally small.
Editorial test
Treat every visible edge as a reader promise.
The graph audit checks that targets resolve, notes exist, relationship types do not overlap, and hubs are not used as generic semantic targets. It cannot decide whether an idea is useful; that remains editorial judgement.
01
State the transition
Write what the reader gains by following the relationship, not just the target title.
02
Choose the narrowest true meaning
Use a prerequisite or application relationship when it is real; do not default to related.
03
Derive structure
Let the silo determine guide membership so authors do not manually repeat hub links.
04
Remove weak edges
Keep contextual links in the body when useful, but remove map lines that do not represent a next step in thought.
Evidence boundary
The graph can prove that Mannvit authored a navigable pathway. It cannot yet prove which pathway readers choose.
That needs post-launch navigation data. Until then, sparse and legible is a better default than a complete-looking mesh of weak connections.
Bring the content library that has become difficult to navigate.
We can separate structural membership from the few relationships that should actually guide a reader.
Bring the page, report, or workflow as it is now.
We reply with the clearest next step, or an honest no.
