Design decisions to date

Status: S1 (concept and pillars) is in progress. “Defaulted” means no steering answer arrived and the default was taken.

#	Question	Chosen (defaulted)	Rejected for now	Metric to confirm	Revisit when
1	How is geometry stored?	Per-cell solid/open on index-keyed Voronoi sites, softened by a smooth density field	Height map; strict voxel grid (kept as a cost contrast)	Cells rebuilt per chunk with zero seam mismatches	Surface mesher is picked
2	How does the world sprawl without stitching?	Keyed jitter plus cluster dropout with a block guard	Global relaxed Voronoi (needs neighbours)	Area CV ~0.5 at jitter 1.0 + dropout (measured 0.503)	Edit cost on dense areas is measured
3	How is structure handled?	Support propagated over the cell-adjacency graph	Rigid-body physics; Minecraft-style no-collapse	Dig-a-pocket test: unsupported count rises predictably	Player builds are added to the graph
4	What is the setting?	Open: candidates in research	Plain survival	A pillar set the setting delivers via mechanics (MDA check)	Pillars are written

Notes

• Decisions 1-3 are backed by figures in procgen. Decision 4 is not yet made.
• The first three come with caveats: the support rule is graph-density sensitive, and the volume render is a ray-march sample, not a mesher.
• The authoritative decision records live in the Lodestar repo (docs/design/decisions.md, DR-nnn); this table is a readable mirror.