rhi

Appearance

Open Threads ​

A registry of open work that is not expected to land soon — large or long arcs, paused or abandoned sessions, parked design questions, anything left open without a clear path to imminent completion.

The discriminator is liveness, not scope. A thread belongs here if it is still genuinely open and parked for the foreseeable future, regardless of whether it touches one repo or many. A single-repo arc someone walked away from mid-flight belongs here just as much as a cross-project convention question.

File here:

  • Large or long-running arcs that won't complete soon.
  • Paused or abandoned sessions — real work someone left mid-flight.
  • Parked questions: a design or convention decision left open, with no owner driving it to a near-term answer.

File in the project's TODO.md instead when the item is actively owned inside one repo and being driven toward completion there. Live, progressing backlog is TODO work, not an open thread — even if it's large. The registry is for what's parked, not for everything that's unfinished.

When a thread resolves, move it to closed.md with a one-line reason rather than deleting it, so the archive stays auditable.

Threads ​

Cross-cutting design questions ​

These are parked convention/architecture questions whose answer would propagate across repos. None has an owner driving it to a near-term decision.

  • harness-orchestrator-fit — Does the Claude Code harness assume an orchestrator-class top-level model? Affects model defaults in every repo's CLAUDE.md. Untested.
  • domain-generator-corpus — Would a generator-corpus (worked examples + counterexamples) outperform a rule-corpus (CLAUDE.md) for the failing constraint classes?
  • design-decisions-convention — What's the standard filename + pointer pattern for prior design decisions, and where's the ephemeral/load-bearing cutoff? Partial, inconsistent adoption across repos.
  • context-md-adoption — Should CONTEXT.md be a mandatory ecosystem-wide artifact or stay opt-in? Partial adoption, no decision recorded.
  • out-of-scope-stance — Should the ecosystem CLAUDE.md forbid "out of scope" as a deferral pattern, and how does that square with legitimate project-boundary discipline?
  • claude-md-saturation-curve — Is CLAUDE.md churn saturating (converging on the implicit-constraint set) or still mining? Measurable but unmeasured.
  • compaction-loss-rate — What's the actual rate at which compaction silently strips load-bearing prior agreements? Measurement-blocked.
  • specs-as-software — Is @spec: / @protocol: a software-taxonomy-only namespace, or a shared ecosystem concept concord/paraphase/rescribe can reference?

Substrate & scope questions ​

  • worldbuilding-namespace — How should the worldbuilding lens namespace be structured in software-taxonomy, and how should that convention propagate if paragarden universes export into the corpus?
  • lua-on-mobile — Is a mobile (Android / Termux) target for crescent/moonlet in scope? JIT is disallowed on Android; interpreter fallback works but the in-scope decision is unresolved.
  • Deterministic simulation testing — What does an open, instrumentation-optional testing substrate look like across levels (in-process, wasm, emulator, hypervisor)? Unresolved: the oracle problem, the managed-runtime frontier, whether exploration and reproduction are better as separate tools.

Paused arcs ​

  • existence-persistent-awareness — The fluffy-rolling-meadow "Persistent Awareness Display" plan (time/money degradation HUD) was written, presented for sign-off, and never approved or discarded. Plan file untouched since 2026-04-10; not tracked in existence's TODO.md.
  • scribble-awaiting-reincarnate-stability — Scribble (sketch-level creative environment) was scaffolded in early 2026 and has been dormant since mid-March. The platform-layer dependency on reincarnate is ready; the blocker is reincarnate's overall toolchain being broken/in flux.