What the view claims
The original full-article MDS field has pairwise-distance correlation 0.369 , so even the control is a lossy overview rather than literal geometry. Every selector arm retains its unrelaxed construction separately from a measured minimum-clearance display transform. The latter changes legibility, never ego identity or the relation graph used to construct the field.
There are no cluster containers, persistent root boundaries, global edges, implied plane faces, or canvas labels here. The perimeter does not introduce a second clustering: it reuses the canonical fingerprint model and its stable axis order. The nearest-six baseline remains a bounded navigation identity, not an assertion that every shown relation is strong. The absolute lens uses a fixed cosine threshold of 0.60 ; it leaves 43 concepts with no direct qualifying relation on this snapshot, while degree ranges up to 46 . Origin-affinity and path-cohesion retain bounded nearest-neighbor reach but replace rank-driven intensity with fixed-scale similarity evidence. Multi-node selection and the corner map always use the active connectivity lens; strongest is a union using the best reach, weakest is a strict intersection using the weakest reach, and average counts missing reach as zero.
A single MDS coordinate system places every approved concept on one shared semantic surface; communities do not bound or partition it.
The world view and the local view carry different evidence: global position summarizes all pairwise distances, while each recursive ego fingerprint preserves the concept's exact nearest-neighbor signature.
Minimum-clearance relaxation is a legibility transform only. Its displacement and correlation are recorded so it cannot silently become part of the semantic claim.
R57's five graph-layout arms are fitted into the same world rectangle and can now be switched without rebuilding a glyph. Every nested world stencil and overview follows the active coordinate array, so the miniature remains an exact copy of the surrounding world.
Adaptive LOD makes recursion conditional on available screen area rather than requiring one impossible density for every zoom scale.
Each ego now carries its own origin-anchored local MDS. Member size is an absolute function of similarity to that ego's focus; overlap is preserved and resolved through canonical-identity hover propagation.
The selectable world-stencil glyph repeats the complete global map inside every root. Its internal fill moves forward through two neighbor generations at rest and three on inspection.
The surrounding world asks the transpose question: it highlights every root whose own forward diffusion includes the selected concept. Inner propagation and outer inclusion therefore no longer duplicate one view.
Depth occupies sharply separated intensity bands, while maximum-product path similarity controls the gradient within each band. Third-degree marks and outer traces are deliberately near the visibility threshold.
The default world-stencil view is theme-aware monochrome. Root fills and persistent container boundaries are removed so the repeated point field defines each glyph's own visual gravity.
Branch color remains available only as a comparison mode and only on relationship strokes. Neutral fills prevent shared hues from implying connections that do not exist.
The label-free inset uses a square world-aspect detail field at low zoom and becomes a global overview with a viewport rectangle at high zoom.
Every canonical mark inside a root now uses the outer-world circle radius under the exact same similarity transform as its position. Zoom therefore reveals the same map rather than a second map with inflated circles.
The internal coordinate field is two percent smaller than R48, leaving a measured boundary margin. Because positions and radii shrink together, the readable display's non-overlap is preserved inside every root. Fill while the outer inverse hairline uses its own much faster hop fade.
The rank baseline is retained because six stable neighbors make navigation and recursive identity bounded, but its weakest edge varies by origin.
The absolute-threshold lens is retained because variable degree and zero degree are the missing evidence in a forced-k graph.
The origin-affinity lens is retained because it preserves the bounded stencil while exposing whether second- and third-hop nodes remain directly similar to the origin.
Path cohesion remains explicitly experimental: it can distinguish a cohesive route from a bridge, but averages can conceal one weak edge.