Skip to content

Playbook and Three.js code kit for detailed dioramas: classic pixel art, stroked toon or pixel + stroke; two intake gates for look and model sourcing; free orbit, palette snap and ink; animated toy-scale cast; per-entity frame QA.

4.8 KB
Raw Blame

Represent the named subject, not automatically a teaching prop

Resolve WHAT is being represented before asset search or geometry. Default representationMode: real-subject: depict the actual subject in its appropriate natural/physical form. “Educational”, “explain”, “3D model”, “diorama” or a toon/pixel style does not authorize replacing it with a toy, classroom aid, bench apparatus or museum display.

Three modes; delegate only when asked

  • real-subject: the actual thing. Earth is a planet, a heart is the organ, a volcano is terrain. A real machine retains its actual feet/housing; natural objects do not gain an invented base, stand, spindle or mechanism. Scale compression is declared; no false claim of literal size.
  • teaching-model: a physical demonstrator deliberately representing the thing: a classroom globe, orrery, mounted anatomical model or tabletop rig. Use only when explicitly requested/confirmed or explicitly delegated with a stated reason. Its supports and mechanisms need plausible construction but must not be confused with how the represented natural subject works.
  • explanatory: an openly simplified visualization: section, exploded layout, field/flow arrows, diagrammatic scale or symbolic geometry. Keep subject identity and correct causal explanation. A cutaway or overlay can also be a view of real-subject; it does not require turning the subject into a tangible toy.

A “realistic classroom globe” is still teaching-model; “stylized Earth in space” is still real-subject. An imported asset's form must match the mode: a commercially licensed globe on a stand cannot substitute for a requested planet just because both have the right title. Asset accuracy/role and representation are separate checks.

Ask only for a material unresolved choice

If the user explicitly names the form, follow it. If they simply ask to explain the subject, use real-subject without another question. The word “model” alone is not an instruction for a physical scale model.

If the brief genuinely leaves competing forms unresolved, ask once:

How should I represent the subject?

  1. The real thing — actual subject in its appropriate setting.
  2. A physical teaching model — a tangible demonstrator with intentional supports.
  3. An explanatory visualization — simplified sections, fields or exploded forms.
  4. You choose — choose the form that best serves the explanation and state why.

Combine with an already-needed style question where possible. Don't ask again during repairs or per component. If a repair's old design silently substituted a globe/toy for an actual-subject request, correct that mismatch rather than preserving an erroneous surrogate as a supposed prior choice.

Keep independent decisions independent

  • Render preset: pixel / toon / hybrid changes visual treatment, not subject identity.
  • Presentation: full scene / standalone / both controls surrounding context. Standalone Earth is still a planet, not automatically a globe. Do not force this extra question or implement a dual-mode toggle unless requested or meaningfully unresolved.
  • Exterior / Inside: authored explanatory views, not permission to change the subject into a classroom object. Preserve its complete default form unless an explicit permanent section was requested.
  • Educational profile: causal text/lessons and inspection behavior, not mandatory legs, casing, platform or a three-part machine analogy for everything.

Record mode and decision origin (explicit, default, clarified, delegated) plus one-line rationale in DESIGN. Keep the user's mode during normal repair. Reference/asset searches include the actual form/variant; layout proof checks for unwanted stands, floors, display bases or toy mechanisms before detail. Add explanatory layers to the subject rather than replacing the subject to make labelling easier.

Small semantic regression set

Expected behavior: “explain Earth”, “3D model of Earth in space”, “pixel volcano”, “standalone beating heart” → real-subject, no representation question or invented display mount. “Classroom globe with stand”, “anatomical teaching model on a plinth” → teaching-model. “Diagrammatic exploded pump”, “schematic Earth layer visualization” → explanatory. “Cutaway of the real pump” → real-subject with a cutaway, not automatically a physical demonstrator. An exhibit brief explicitly undecided between a planet and tabletop globe → clarify. Repair of an explicitly chosen teaching model → retain teaching-model. Style/base preferences alone do not resolve or override these choices.

These are instruction/semantic-consumer fixtures, not an automatic scientific-truth detector or proof of a rendered scene. No Earth scene is commissioned by these examples.