Skip to content

Chapter 3 — Hermetic Closure

Act II — Trusting · No hard prerequisites · Validation chapter

Core Question

When is a render silently wrong — and what does the failure look like from the outside?

Hermetic Closure Hero — budget=32 vs budget=700 side by side at 640×360

Same scene. Same camera. Opposite truth. Left: budget=32, closure 0.0% — every pixel is unresolved budget noise that looks like a render. Right: budget=700, closure 100.0% — every pixel is a real transport result. The renders are visually indistinguishable. Source: output/hermetic_hit_closure/20260604T023019Z/

Hermetic Closure — 3-panel with failure map

3-panel extended view: silent failure · budget exhaustion heatmap · full closure. The middle panel shows where the budget is exhausted — failures cluster in the high-curvature annular zone, not randomly.

What to look at

Inspect: The failure storyboard image — look at where the failure pixels cluster. They are not random; they trace the annular high-curvature zone from Chapter 1's heat map.

Contradiction: The budget=32 render looks reasonable. No obvious holes, no glitch artifacts. The Validation HUD is the only thing that reveals 0.0% closure — 100% of pixels are unresolved budget noise.

What would make it stronger: A side-by-side render with both HUDs visible simultaneously — budget=32 (left, HUD: 0%) and budget=700 (right, HUD: 100%). This is the chapter's highest-priority pending screenshot. See capture recipe below.

What the Visitor Sees

The hermetic contract: Every ray entering a hermetic scene volume must reach a definitive classification (geometry hit, escaped, portal event) before the integrator exits. A render that satisfies the contract is closed — every pixel is a real transport result. One that violates it is open — some pixels are unresolved budget exhaustion, displayed as noise that looks like a legitimate render.

Budget = 32 (0% closure)

The image looks reasonable. Geometry is present, shading is plausible, there are no obvious holes or glitches. A developer looking at this image without instrumentation would not know anything was wrong.

The Validation HUD tells a different story:

Closure:        0.0%
Phase:          budget_saturated
Rays resolved:  0 / 200
Budget used:    32 / 32 (100% — every ray exhausted all steps)

Every pixel in this render is noise. No ray reached a classification endpoint before running out of steps.

Budget = 700 (100% closure)

The image looks essentially similar. The HUD:

Closure:       100.0%
Phase:         plateau
Rays resolved: 200 / 200
Budget used:   far below 700 (plateau — additional budget adds zero cost)

Every pixel is a real transport result.

The cliff. There is a sharp transition around budget ≈ 300 where closure jumps from near-zero to near-complete. Below the cliff the renderer fails silently. Above it, every pixel is resolved.

Hermetic hit closure recovery heatmap

Adaptive recovery heatmap. Recovery is high in the annular curvature band and low at corners. The pattern matches the transport field's curvature structure: adaptive scaling helps most where the field is strongest. Source: output/hermetic_hit_closure/20260514T040157Z/

The failure storyboard shows that failure pixels are not randomly distributed. They cluster in the high-curvature annular region — the same zone the Chapter 1 curvature heat map identified as optically expensive. Integrating through high-curvature regions requires more steps to find the ray's endpoint.

The recovery heatmap shows that adaptive budget scaling recovers most failures in the annular band but cannot recover corners, which have a different failure mechanism.

Artifacts

Artifact File Notes
Failure storyboard misterylabs_artifacts/visuals/hermetic-hit-closure-storyboard.png Failure pixel distribution
Recovery heatmap misterylabs_artifacts/visuals/hermetic-hit-closure-recovery.png Adaptive recovery map
Validation summary misterylabs_artifacts/validation/hermetic-hit-closure.md Cell table, closure %, phase
Card misterylabs_artifacts/cards/hermetic-hit-closure.md Ready for MisterY Labs

Missing: A side-by-side render showing budget=32 (left) and budget=700 (right) simultaneously with both Validation HUDs visible — the clearest single-image demonstration of the cliff. This is the chapter's highest-priority pending screenshot.

Sample World

hermetic_closure_world — design proposal

Scene: test-hermetic-curved-room.tscn

Three budget presets:

Preset Budget Expected Closure Label
failure 32 0.0% "Silent failure"
transition 300 ~50% "Cliff edge"
success 700 100.0% "Full closure"

The hermetic_closure overlay colors pixels green (hit), blue (escaped), red (budget exhausted). At budget=32 the film is entirely red. The Validation HUD is mandatory in this world — it is the only mechanism for detecting the failure.

Build priority: 3.

Validation Question

At budget=700, step=0.015, row traversal: what is the closure percentage?

Expected: 100.0% (plateau phase). Below 95% indicates regression.

At budget=32, same settings: Expected 0.0% (budget_saturated). Above 5% suggests the scene is less expensive than the artifact baseline — worth investigating.

Falsification of the core claim: If closure at budget=32 is above 10%, either the scene or the integration budget calculation has changed. The budget=32 / 0% result should be stable across runs.

Key Insight

A renderer can fail completely and look fine. The only way to know is to measure closure. "Plausible" is not "correct."

The Validation HUD Is Not Optional

Silent failure

At integration budget=32, the hermetic curved room produces an image that passes casual visual inspection but contains zero correctly classified pixels.

The hermetic closure overlay and Validation HUD are the only mechanisms for detecting this. If you run this scene without instrumentation, you cannot distinguish a correct render from a silent failure.

This is why the hermetic fixture contract exists: transport correctness is not gradual — it falls off a cliff. The right response is to measure closure on every render, not to assume correctness from visual inspection.

Next Chapter

Chapter 4 — Coherence Basin →: Chapter 3 showed that insufficient budget causes transport failure. Chapter 4 asks whether there are regions where no amount of budget or precision can achieve convergence.

Bridge: "We know budget determines closure. But what if the problem isn't budget at all — what if some regions of the field are fundamentally unstable?"

Capture Recipe — Side-by-Side HUD

Target: A single composite image showing budget=32 (left half, HUD visible) and budget=700 (right half, HUD visible) rendered from the same camera pose on test-hermetic-curved-room.tscn.

Steps:

  1. Open test-hermetic-curved-room.tscn in the Godot editor.
  2. Set RayBeamRenderer.MaxSteps = 32. Run and screenshot the full viewport including the Validation HUD overlay. Save as budget_32_hud.png.
  3. Set RayBeamRenderer.MaxSteps = 700. Run and screenshot. Save as budget_700_hud.png.
  4. Composite side-by-side (e.g., ImageMagick: convert budget_32_hud.png budget_700_hud.png +append chapter_03_sbs_hud.png).
  5. Copy result to Docs/assets/observatory/hermetic-closure-sbs-hud.png and update the figure in this chapter.

Expected HUD values:

Panel MaxSteps Closure Phase
Left 32 0.0% budget_saturated
Right 700 100.0% plateau

Promotion path: Copy to misterylabs_artifacts/visuals/hermetic-closure-sbs-hud.png and update manifest.json artifact a04.