What works today
The corpus-ingest pipeline, the deterministic compute chassis, the boundary contract,
and every persistence store are real, tested, and exercised end to end by
cargo test --locked and npm test.
- PCGen
.pcc/.lstparsing into canonical source-IR - Pilot compute + boundary contract โ deterministic, fail-honest
- Per-domain engines: spellbook, skill allocation, feat prerequisites, equipment effects, level-up
- Desktop workbench: create, load, clone, export, recompute
- Campaign manager (local) with on-disk persistence and conflict detection
- Self-update chain + browser-handoff feedback to GitHub issues
What it isn't (yet)
Codex today is a developer proof harness plus a buildable desktop workbench surface, not a finished end-user product. The honest single-class path is Fighter, levels 1-3, for any race โ every other class/level combination returns real claim-blocking diagnostics rather than fabricated output. The desktop UI surface is bounded and demo-oriented, not a general character builder.
We say so up front rather than pretending. That posture is the project's central engineering commitment.