BEARING

Updated at cycle boundaries and before the final commit of each v18 slice.

Scope note:

• BEARING says where the repo stands now, what feels wrong now, and what is next.
• For canonical noun meanings, use GLOSSARY.md.
• For the runtime architecture ladder, use 0035-observer-geometry-architecture-ladder.md.
• For later-major horizon planning, use release-horizon-v20-v21.md.

Continuum Optic Admission Posture

For cross-repo optic admission, git-warp is a complete Continuum participant, not an Echo runtime surrogate. Continuum is the protocol for exchanging witnessed causal history. Wesley compiles artifacts and descriptors, Echo admits Echo-local runtime invocations, git-warp admits git-warp-local causal history and readings, authority layers issue grants, and applications hide handles and basis references behind product adapters.

Where are we

git-warp has shipped v17.0.0. The release work is now behind us in repo history, npm, and JSR; the active direction is v18.0.0.

The v18 hill is not generic graph-model cleanup. It is Continuum compatibility:

Make git-warp a Continuum-compatible sibling WARP runtime: consume Wesley-generated artifacts for Continuum-owned contract families, map git-warp's append-only Git-backed causal history into honest WARP Optic evidence, and give warp-ttd generated-family facts instead of handwritten adapter folklore.

The long-term compatibility target is the WARP Optic shape described in ~/git/blog/aion-paper-07/dist/aion-paper-07.txt, plus the Continuum contract families authored in ~/git/continuum/schemas/ and compiled by Wesley. Echo and git-warp are sibling runtime implementations. git-warp has its own Continuum role, and it must not emit Continuum-shaped values as native Continuum witnesses until that witnesshood is actually proven.

Backlog fold-in: the repo-visible v18 lane is WL-4A-v18-graph-substrate-convergence in WORKLOADS.md, backed by the eight notes in method/backlog/v18.0.0. Treat that lane as the graph-model track inside this compatibility campaign: node and edge record identity, attachment slots, graph-op algebra, content cutover, legacy property projection, migration tooling, and genesis replay equivalence. Existing echo-shaped backlog identities are historical shorthand for graph-model pressure already exercised by Echo, not a claim that Echo owns git-warp's Continuum role.

Current branch state at this boundary:

• Branch: v18-continuum-opening
• Release tag: v17.0.0
• Latest remote head inspected: origin/main at 5afdd3eb
• Latest package version: 17.0.0
• Latest closed cycle: 0145-push-pr-review-merge

The release ladder is now:

• v17.0.0: shipped TypeScript migration, public API honesty, materialization-frontdoor deletion, readings/optics direction, and query read-model groundwork.
• v18.0.0: Continuum/WARP Optic compatibility for git-warp as an independent Continuum participant, through Wesley-generated contract-family artifacts and honest evidence posture.
• v19.0.0: observation, doctrine, and slice-first runtime convergence
• v20.0.0: slice-first read execution
• v21.0.0: distributed observer geometry and admission reality

The v18 compatibility work is bigger than ten slices. The first ten slices are the opening campaign. Slice 11 is an explicit re-plan point after the repo has real evidence from generated artifact ingestion, evidence posture, and the first receipt-family projection.

Invariants

Compact list here; full derivations with paper grounding, codebase mapping, and concrete checks live in docs/invariants/.

1. TICK-CONFLUENCE — same patches, any order, same materialized state (Paper II Thm 5.1, OG-4 Thm 10) → tick-confluence.md
2. HOLOGRAPHIC-BOUNDARY — initial state + patch chains = complete replay, no ambient state (Paper III Thm 4.1) → holographic-boundary.md
3. BACKWARD-PROVENANCE — every value traces to exactly one producing patch (Paper III Thm 4.2) → backward-provenance-completeness.md
4. PAYLOAD-MONOID — checkpoint + remaining patches = full replay (Paper III Prop 3.2) → payload-monoid.md
5. STATE-PROVENANCE-SEP — state convergence does not imply history convergence (OG-4 Prop 13, OG-1 Thm 91) → state-provenance-separation.md
6. EXPLICIT-CONFLICT — conflicts are surfaced, never silently erased (OG-4 Thm 15) → explicit-conflict-surfacing.md
7. APPEND-ONLY — Git history never rewritten (Paper III Def 3.6) → append-only-history.md
8. DOMAIN-PURITY — domain never imports infrastructure or ambient state (Paper III Rmk 3.4) → domain-purity.md
9. WRITER-ISOLATION — each writer owns its own ref, no coordination (Paper II Thm 7.1, OG-4 Thm 10) → writer-isolation.md
10. TWO-PLANE-COMMUTATION — property and topology ops commute (Paper II Thm 7.1) → two-plane-commutation.md
11. CAS-ATOMICITY — writer ref updates are compare-and-swap (Paper II Rmk 4.3) → cas-atomicity.md
12. OBSERVER-DETERMINISM — queries and traversals are deterministic functions of state (Paper IV Def 3.1) → observer-projection-determinism.md
13. TRAVERSAL-TRUTH — streams for traversal, ports for truth (OG-1 Def 3, Paper IV Sec 3.3) → traversal-truth.md
14. NO-SCALARIZATION — observer comparison is multi-dimensional (OG-1 Thm 87) → no-scalarization.md
15. SUFFIX-TRANSPORT — sync at tip, not replay from frontier (OG-4 Thm 9) → suffix-transport-correctness.md

What just shipped

v17.0.0 shipped and was followed by release hardening:

• The v17 release branch landed through PR #84.
• Follow-up repair and package migration work landed through PR #85.
• Release hardening landed through PR #86.
• The final v17 coverage ratchet landed through PR #87; the signed v17.0.0 tag points at that merge.
• npm publish recovery landed through PR #88.
• PR #89 simplified the README model sentence after the release line.

The shipped v17 scope remains: TypeScript migration, public API honesty, materialization-frontdoor deletion, readings/optics direction, query read-model groundwork, sync hardening, release gates, and package publishing.

What feels wrong

• The release preflight fix lowered the coverage ratchet to the measured full-suite v17 line baseline 91.74%; this is tracked as v19 bad-code debt in SPEC_coverage-ratchet-baseline-drop.md.
• v18 can easily turn into adapter folklore if git-warp hand-authors local mirrors of Continuum-owned families instead of consuming Wesley-generated artifacts.
• v18 can also lie in the other direction: Continuum-shaped values are not Continuum-native witnesses unless the runtime has actually proven native witnesshood. Initial git-warp compatibility evidence should be treated as translated git-warp evidence until stronger proof exists.
• The v18 backlog already names a graph-model convergence lane. The plan must fold that lane into Continuum compatibility instead of replacing it with a parallel cross-repo adapter plan.
• warp-ttd needs git-warp facts as generated-family nouns, but the existing ecosystem still contains handwritten adapter and protocol residue.

What comes next

Run the v18 opening campaign. Update this task list at the end of each slice, before the final commit for that slice, and mark completed items with - [x].

Running Task List

• 1. Sync and clean the repo runway: fast-forward main, clear fsmonitor noise, close stale v17/0145 bookkeeping, and record the v18 starting point.
• 2. Create the v18 Continuum compatibility charter: WARP Optic compatibility, Continuum contract-family compatibility, Wesley-generated artifact consumption, and warp-ttd acceptance.
• 3. Build the cross-repo contract matrix: Continuum family to Wesley generated artifact to git-warp source fact to warp-ttd consumer need, with WL-4A-v18-graph-substrate-convergence folded in as the graph-model track.
• 4. Define git-warp's WARP Optic realization map: observer plan, bounded slice, lowering surface, admissibility law, and retention contract.
• 5. Add a generated-artifact ingestion path for Continuum families, with a guard against handwritten local mirrors becoming contract authority. The current seam admits Continuum receipt-family fixture JSON and Wesley realization manifest JSON through explicit load context; it binds each accepted JSON shape to the matching context authority, the domain policy independently rejects descriptor kind/authority mismatches, the adapter entry point and adapter-local JSON type carriers are split below the source-size and one-file-per-concept caps, self-attested authority fields from artifact JSON are rejected, policy-test authority fixtures are named constants, and empty or internally inconsistent Wesley generated inventory is rejected.
• 6. Make evidence posture explicit: translated git-warp evidence first, native Continuum evidence only after native witnesshood is proven.
• 7. Prove the patch commit visibility contract: success means canonical writer-tip advancement and visible graph truth, not just object creation.
• 8. Add the same-writer concurrent patch race witness with final-frontier and visible-state assertions.
• 9. Project git-warp receipt facts into the generated Continuum receipt-family shape with conformance tests.
• 10. Add the first warp-ttd smoke over generated-family git-warp receipt facts instead of handwritten adapter-local receipt folklore.
• 11. Re-plan with evidence in hand before expanding into reading-envelope, suffix/runtime-boundary, neighborhood-core, and settlement-family slices.

The loop stays strict: write or update the cycle doc, capture RED, green the slice, update this BEARING task list before the final commit, validate, then commit only the files touched in that slice.