Each scene on this page is a small, fixed prompt — a specific situation drawn from the world, rendered under constraints tight enough to make the comparison between voices meaningful. A scene is not a story. It does not resolve. It is a piece of the world, held still, so that different voices can each take a pass at it.

Current scenes#

• Dawn at Scimitar Falls — a single figure at the southern overlook as the Coriopolis emitter-cycle transitions from night to morning. Exercises landscape, light cycle, rotational physics, and a solitary human register.

Why these constraints#

A test scene has to be:

• Small enough to render fully in a few hundred words. A chapter is too much material; a voice can hide behind a plot. A held moment exposes what the prose is actually doing.
• Canon-grounded. Each scene takes its physical givens from the wiki so that voice decisions don’t drift into fact decisions. A voice should be judged on how it handles the world, not on what world it invents.
• Texturally demanding. The scene must contain surfaces — landscape, weather, rotational effects, a human presence — that a voice has to cope with. A scene that can be handled by any voice equally well is not useful for comparison.
• Plot-free. No arrival, no decision, no dialogue, no resolution. The point is register, not story. Renderings that smuggle plot in are noted in the scene’s own page.

Adding a scene#

Create a branch bundle under scenes/, write the prompt in _index.md, then add one rendering per voice as sibling files named after the voice (ksr.md, le-guin.md, and so on). Frontmatter should set both scene and voice slugs so Hugo templates can list renderings by either axis.