Roadmap: improve distance LOD visual quality

[MichaelFisher1997]

Goal

Improve distance LOD visual quality while keeping the system engine-native, incremental, and safe to ship.

Scope

This roadmap covers the rendering, data, meshing, streaming, persistence, and quality-control work needed to make distant terrain look more stable, detailed, and configurable.

Phase Issues

• Phase 1: polish LOD seams, fades, and representative sampling #687
• Phase 2: introduce richer LOD source data #688
• Phase 3: add a column/span LOD mesh builder #689
• Phase 4: improve hierarchical LOD fallback #690
• Phase 5: persist generated LOD source data #691
• Phase 6: add LOD quality controls and diagnostics #692

Success Criteria

• Distant terrain transitions are less visible during movement.
• LOD boundaries avoid cracks, holes, and obvious tile borders.
• Far terrain better represents water, cliffs, vegetation, caves, and overhang-prone terrain.
• LOD startup and movement streaming remain stable under normal gameplay.
• Cached LOD source data avoids needless regeneration across sessions.
• Quality can be tuned without code changes.

[github-actions]

This is a Feature Request / Roadmap — a well-structured multi-phase plan for improving distance LOD visual quality across 6 sequential phases.

Validation:

• Parent issue Roadmap: improve distance LOD visual quality #686 correctly references all 6 phase issues (Phase 1: polish LOD seams, fades, and representative sampling #687–Phase 6: add LOD quality controls and diagnostics #692)
• Each phase has clear scope, work items, and acceptance criteria
• Phases are properly ordered with dependencies (Phase 1 must land before Phase 2, etc.)
• No duplicate issues found

Architectural Observations:

1. Sequential dependency chain is critical — Phase 3's column/span mesh builder depends on Phase 2's richer data structure; Phase 5 persistence depends on Phases 2–3
2. The world-lod module is the central integration point — existing LOD manager, seam stitching, and renderer tests in src/tests.zig:67-69 will need updates
3. LODSimplifiedData (current single-height representation) is the baseline that Phases 2–3 extend while maintaining fallback

Suggestion: Consider splitting Phase 1 into two separate issues if the seam-polish work and shader fade work have significantly different review timelines, since they touch different subsystems (meshing vs. rendering).

No critical blockers identified — this is a solid phased approach to a well-scoped feature area.

opencode session  |  github run

[MichaelFisher1997]

fucking this off, gonna make a v2 LOD module