Skip to content

Display Is Not Probe

Observation card

Field Value
Status partial (doctrine; UX overlays still evolving)
Verified against 5ce15c13 (docs ontology; sealed path base 6e69d792)
Claim boundary Measurement ≠ visualization mapping

The split in one breath

Side What it is Authority
Probe / instrument Transport computation + sealed observation channels Scientific
Display Coloring samples or host buffers for human eyes Presentation only

If a pixel is magenta, brown, or white on the plate, that is a mapping result unless a caption names a sealed channel and a Complete snapshot backs the class.

World viewport vs Observation Plate

Surface Public name What it shows
Large 3D Godot view World viewport Scene meshes, lighting, locomotion camera
Rectangular film inset Observation Plate Mapped transport samples / host film buffer

The world viewport is not the Observation Plate. Portal lettering can warp in the plate while the mesh looks “normal” in the viewport—or the reverse presentation artifacts can dominate. Scientific reading prefers the plate plus Inspector counts, not the free camera beauty view alone.

Internal pair: Observation Plate (FilmView / film buffer); Transport Lens (GrinFilmCamera implementation).

Measurement vs mapping

Instrument  →  Observation channel (sealed)  →  Display Mode (mapping)  →  Observation Plate
  • Measurement: e.g. cathedral.probe.outcome stores ProbeOutcomeCode per sample.
  • Mapping: e.g. NormalRGB takes a normal vector and paints \(n\cdot 0.5+0.5\).
  • Plate: shows the mapped colors (and may use legacy fallback fills when data is missing).

Changing Display Mode must not change the outcome histogram. If it appears to, you are not reading a sealed plane—or the frame was re-generated.

Semantic plane vs visual color

Layer Example Use for
Semantic plane HitGeometry, MaxStepsExhausted Region Probe seeds, histograms, evidence
Visual color Magenta block, brown NormalRGB face Teaching, reconnaissance, never sole class proof

RGB (or NormalRGB) cannot identify scientific regions. Unresolved Regions are connected components on the outcome plane (default: MaxStepsExhausted), carried by cathedral.probe.region_label—not flood-fills of pink pixels.

Why RGB fails as a region classifier

  1. Fallback fills reuse a fixed presentation color for many host failure/missing paths.
  2. Upscale blocks (80×45 → plate) create rectangular magenta that looks “semantic” but follows the sample grid.
  3. BackgroundResolved and HitGeometry can both shade non-magenta or both fall into fallback depending on path—color does not encode the enum.
  4. Opacity (e.g. 40%) blends plate with the viewport and further confuses class reading.

Compact rules

  1. Viewport ≠ Observation Plate.
  2. Display Mode ≠ Instrument.
  3. Plate color ≠ outcome code.
  4. Region Probe outlines come from sealed labels, not lasso-on-magenta.
  5. Complete frame ≠ successful geometry resolution.
  6. Deeper Region Refinement ≠ greater truth—only more Transport Effort.

See also