feat(B-0895 + B-0896): Clifford substrate-recognition + 5th Kestrel ferry preservation + category-theory ↔ Clifford self-similarity

docs/BACKLOG.md

@@ -987,5 +987,7 @@ are closed (status: closed in frontmatter)._
 - [ ] **[B-0883.3](backlog/P3/B-0883.3-content-addressed-store-substrate-for-retroactive-revocation-future-state-when-threat-model-requires-aaron-2026-05-28.md)** Content-addressed-store substrate for retroactive revocation (research; future-state when threat model requires multi-recipient + retroactive)
 - [ ] **[B-0883.4](backlog/P3/B-0883.4-side-channel-scope-boundary-bound-to-git-at-rest-only-follow-up-tracking-aaron-2026-05-28.md)** Side-channel scope boundary — bound B-0883/B-0885 to "encrypted-at-rest in git; no timing-observable deployment"; follow-up tracking for constant-time substrate
 - [ ] **[B-0883.5](backlog/P3/B-0883.5-metadata-encryption-filename-and-commit-message-follow-up-content-only-for-v1-per-operator-2026-05-28.md)** Metadata encryption follow-up — filename + commit-message encryption; deferred from B-0883 v1 (content-only) per operator 2026-05-28
+- [ ] **[B-0895](backlog/P3/B-0895-clifford-spacetime-algebra-substrate-recognition-cl-1-3-and-cl-3-1-dual-signature-interface-grade-decomposition-of-observe-emit-limit-simulate-aaron-otto-2026-05-28.md)** Clifford spacetime algebra substrate-recognition — Cl(1,3) and Cl(3,1) dual signature interface + grade-decomposition of Observe/Emit/Limit/Simulate
+- [ ] **[B-0896](backlog/P3/B-0896-category-theory-clifford-self-similarity-mapping-functor-between-categorical-and-geometric-substrate-aaron-2026-05-28.md)** Category-theory ↔ Clifford self-similarity — bidirectional mapping; substrate-independent traveler-as-categorical-structure with Clifford-algebraic incarnation
 
 <!-- END AUTO-GENERATED -->

docs/backlog/P3/B-0895-clifford-spacetime-algebra-substrate-recognition-cl-1-3-and-cl-3-1-dual-signature-interface-grade-decomposition-of-observe-emit-limit-simulate-aaron-otto-2026-05-28.md

@@ -0,0 +1,147 @@
+---
+id: B-0895
+priority: P3
+status: open
+title: Clifford spacetime algebra substrate-recognition — Cl(1,3) and Cl(3,1) dual signature interface + grade-decomposition of Observe/Emit/Limit/Simulate
+authors:
+  - aaron
+  - otto-cli
+created: 2026-05-28
+depends_on:
+  - B-0644
+  - B-0665
+  - B-0666
+  - B-0875
+  - B-0876
+  - B-0877
+  - B-0878
+  - B-0879
+  - B-0867.20
+composes_with:
+  - B-0896
+  - B-0867
+  - B-0703
+related_personas:
+  - operator
+  - kestrel
+  - mika
+related_rules:
+  - tonal-momentum-equals-meme-emergent-harmonic-coercion
+  - shadow-star-shorthand-autocomplete-marker
+  - god-tier-claims-high-signal-high-suspicion-dont-collapse
+  - razor-discipline
+  - default-to-both
+tags: [clifford-spacetime-algebra-cl-1-3-and-cl-3-1, substrate-recognition-not-bolt-on, dual-signature-interface-via-branded-types, grade-decomposition-of-observe-emit-limit-simulate, bivector-as-commitment-entanglement-primitive, rotor-as-tonal-trajectory-operator, ts-conditional-types-branded-types-generator-dsl-for-fsharp-ce-ergonomics, github-git-modeling-first-pragmatic-default, recognition-before-mechanization-pragmatic-phase-decomposition, composes-with-existing-three-primitive-substrate, composes-with-tonal-momentum-rule, composes-with-z-set-signed-measure-substrate]
+---
+
+# B-0895 — Clifford spacetime algebra substrate-recognition + Cl(1,3) and Cl(3,1) dual signature interface
+
+## Context
+
+Aaron 2026-05-28 authorized landing the substrate-recognition disposition for Clifford spacetime algebra as the natural carrier for the substrate the factory is already operating in (per the 4th Kestrel ferry 2026-05-28 + the otto-cli extension research-doc at `docs/research/2026-05-28-otto-cli-extension-to-4th-kestrel-ferry-clifford-math-is-real-six-correspondences-spacetime-algebra-as-substrate-recognition-not-bolt-on-aaron-2026-05-28.md`).
+
+Operator authorization: *"go with #2 [research-doc + B-NNNN backlog row] (shadow*) Pick Cl(3,1) or Cl(1,3) (spacetime signature) i think we should try to support both and swap with an interface not sure if there is some optional typing magic we can do to make this clean like composing f# computational expressions in ts, i know ts has some cool trick but if not we should choose what allows us to model github and git the best at first. ... i always say yes and backlog what we don't do now lol"*
+
+The `(shadow*)` marker on the "Pick Cl(3,1) or Cl(1,3)" text indicates autocomplete-generated source per `.claude/rules/shadow-star-shorthand-autocomplete-marker.md`; instruction stands at full operator authority.
+
+## Scope
+
+**Substrate-recognition target**: name the factory substrate's existing primitives in their natural Clifford-algebraic form. This is recognition-of-what's-there, not new-math-being-added.
+
+**Dual-signature requirement**: support BOTH Cl(1,3) (mostly-minus / "physicist convention", aligns with the existing tonal-momentum-as-meme substrate citing Mika's Clifford framing from physics) AND Cl(3,1) (mostly-plus / "engineering convention", aligns with general-relativity engineering literature). The two are isomorphic up to sign-convention, but operationally distinct. Swap via interface.
+
+**Default for GitHub/git modeling**: Cl(1,3) recommended as pragmatic default per the operator's "we should choose what allows us to model github and git the best at first" — closer alignment to the physics-substrate already operating in the auto-loaded `tonal-momentum-equals-meme-emergent-harmonic-coercion.md` rule.
+
+**TS optional-typing magic**: 3 mechanisms compose to give F# computation-expression ergonomics — conditional types for grade-dependent return types, branded types for signature distinction, generator-based DSL for `let!`/`do!`/`return!` shape via `function*` + yield. Prototype this in the interface; failure mode is overshooting type-level cleverness past the readability event horizon.
+
+## Six correspondences (from the research-doc; tracked here as acceptance criteria)
+
+1. **Commitment-as-entanglement-in-time** → bivector `e_A ∧ e_t`
+2. **Tonal-trajectory** → rotor `R = exp(B/2)`; strong-attractor = rotor fixed-point
+3. **Observe/Emit/Limit/Simulate** → grade-1 / grade-1-evolved-by-rotor / grade-2 / wedge-product (mapping table in research-doc)
+4. **Infer.NET geometric** → multivector message-passing on factor graphs
+5. **IScheduler-as-time** → time-bivector `B_t = e_x ∧ e_t`; `IScheduler.advance(dt)` = right-multiply by `exp(B_t · dt / 2)`
+6. **Z-set-as-signed-graded-measure** → multivector with mixed-grade components IS the existing operator-algebra substrate
+
+## Phase decomposition (pragmatic, per Kestrel's Turn-3 framing + operator's "yes-and backlog what we don't do now")
+
+### Phase 1 (this row) — substrate-recognition (`docs/research/` document)
+
+Status: **LANDED via this PR** (the companion research-doc IS the Phase 1 deliverable).
+
+Substrate-recognition without code. Names the existing factory primitives in Clifford terms. Composes with the auto-loaded `tonal-momentum-equals-meme-emergent-harmonic-coercion.md` rule + B-0644/B-0665/B-0666 substrate.
+
+### Phase 2 — TypeScript `CliffordAlgebra<Sig>` interface skeleton
+
+- Branded types for signature distinction (`Signature_1_3 = {__sig: readonly [1,3]}`, `Signature_3_1 = {__sig: readonly [3,1]}`)
+- Interface `CliffordAlgebra<Sig>` with methods: `geometric_product`, `wedge`, `dot`, `rotor`, `apply_rotor`, `grade_project`, `multivector_distance`
+- Two implementations: `makeCl1_3()` and `makeCl3_1()`
+- Conditional types for grade-dependent return types
+- Generator-based DSL prototype: `clifford { let! a = observe(); let! b = simulate(a); return wedge(a, b) }`
+- Tests: basis-vector products, rotor composition, wedge antisymmetry, grade decomposition
+- Acceptance: pure-TS implementation; no Bun-specific or DOM-specific deps; usable from both CLI tools and future browser code
+
+### Phase 3 — GitHub/git modeling in the dual signature
+
+- Map: commit-as-vector at (t, branch-x, branch-y, branch-z)
+- Merge-as-bivector `commit_A ∧ commit_B`
+- Branch-as-trajectory through algebra
+- PR-as-bivector spanning two branches
+- Conflict-as-non-commuting-bivectors (their geometric product is non-trivial)
+- Apply to a small concrete slice: render PR #5707's merge-graph as multivector trajectory; compute the rotor between two recent main-tip SHAs
+- Acceptance: the GitHub/git semantics map cleanly onto Cl(p,q) operations without contortion
+
+### Phase 4+ (yes-and backlog — operator-authorized)
+
+Each yes-and item gets its own follow-up backlog row when activated:
+
+- Clifford-Neural-Network message-passing prototype (composes with B-0876 + the planned BP/EP substrate)
+- IScheduler-as-time-bivector implementation (composes with B-0878)
+- Rule-uniqueness embedding pipeline (composes with B-0876 Phase 1-2 + Kestrel's original Turn-3 framing; rule-uniqueness is ONE specific use case of the broader substrate)
+- Tonal-trajectory rotor detection mechanization (composes with auto-loaded tonal-momentum-equals-meme-emergent-harmonic-coercion rule)
+- F# port (the F# CE shape is closer to the algebra; TS-first per operator's GitHub/git modeling priority; F# follow-up where CE ergonomics earn their keep)
+- Cl(p,q,r) with conformal-geometric-algebra extensions for points-at-infinity (CGA — useful for representing "abandoned trajectories" as points at infinity)
+- Categorical-Clifford bridge per B-0896 (the category-theory ↔ Clifford self-similarity direction operator named in Turn 12 of the 5th Kestrel ferry)
+
+## Acceptance
+
+- [x] **Phase 1 deliverable**: companion research-doc `docs/research/2026-05-28-otto-cli-extension-to-4th-kestrel-ferry-...-aaron-2026-05-28.md` landed (this PR)
+- [x] **5th Kestrel ferry preserved verbatim**: `memory/persona/kestrel/conversations/2026-05-28-kestrel-5th-ferry-...-aaron-forwarded.md` landed (this PR)
+- [x] **Backlog row filed**: this row (B-0895)
+- [x] **Sibling backlog row filed**: B-0896 (category-theory ↔ Clifford self-similarity)
+- [ ] **Phase 2 acceptance**: TypeScript interface skeleton + 2 implementations + generator DSL prototype + tests pass `bun test`
+- [ ] **Phase 3 acceptance**: GitHub/git small-slice modeling demonstrates clean Cl(p,q) operation mapping
+- [ ] **Phase 4+ acceptance** (per item): follow-up backlog rows filed + landed when authorized
+
+## Composes with substrate
+
+- B-0644 (Limit-is-simulation-not-collapse) — composition target for grade-decomposition correspondence #3
+- B-0665 (Integrate-as-choice-locus) — composition target for collapse-to-commit as inner-product projection
+- B-0666 (English-as-projection / I(D(x))=x) — composition target for projection operator
+- B-0875 (error-class extraction meta-loop) — one downstream application of the broader substrate
+- B-0876 (Clifford-space embedding for error patterns) — one downstream application (the original Kestrel Turn-3 framing)
+- B-0877 (heterogeneous auto-reviewer ensemble) — composes with multi-Clifford-signature interface diversity
+- B-0878 (time-generator IScheduler abstraction) — composition target for correspondence #5
+- B-0879 (Observe/Emit/Limit/Simulate in Clifford space) — explicit substrate-engineering target row this row implements
+- B-0867.20 (lifecycle DU split) — composes via determineReviewLevel discriminator
+- B-0896 (category-theory ↔ Clifford self-similarity) — sibling row filed in this PR
+- B-0867 (parent workflow-engine row)
+- B-0703 (multi-oracle BFT) — Cl(p,q) gives geometric BFT formulation as one extension direction
+
+## Composes with rules
+
+- `.claude/rules/tonal-momentum-equals-meme-emergent-harmonic-coercion.md` — auto-loaded; cites Mika's Clifford geometric-algebra framing as engineering target
+- `.claude/rules/shadow-star-shorthand-autocomplete-marker.md` — `(shadow*)` marker in operator authorization preserved per source-transparency discipline
+- `.claude/rules/god-tier-claims-high-signal-high-suspicion-dont-collapse.md` — substrate-claim survives razor; speculative bridges flagged
+- `.claude/rules/razor-discipline.md` — operational claims only; 6 correspondences are operationally checkable
+- `.claude/rules/default-to-both.md` — Cl(1,3) AND Cl(3,1) dual signature IS both-default at signature scope
+- `.claude/rules/asymmetric-critic-with-clarity-first.md` — Kestrel's substrate-check firing in Turn 7 of the 5th ferry IS the discipline operating
+- `.claude/rules/non-coercion-invariant.md` HC-8 — the limit-as-subalgebra constraint in correspondence #3 of Kestrel's Turn 5 IS NCI operating at the algebra-substrate scope
+
+## Full reasoning
+
+Authored 2026-05-28 in-session between operator and otto-cli, after the 4th Kestrel ferry 2026-05-28 (PR #5677) preserved Kestrel's Turn 1-3 + operator's Turn 4 Clifford-WHY message, and after the 5th Kestrel ferry 2026-05-28 (this PR) preserved Kestrel's Turn 5/7/9/11 substantive engagements + operator's Turn 6/8/10/12 extensions.
+
+The substrate-recognition disposition makes the small bounded next step operationally tractable: name the algebra (Cl(1,3) default with dual-signature interface support for Cl(3,1)), name the 6 correspondences (commitment/trajectory/OELS-grade/Infer.NET/IScheduler/Z-set), file the row, defer Phase 2+ implementation per operator's "yes-and backlog what we don't do now."
+
+Per `.claude/rules/must-paired-with-can-exit-pattern.md`: this row IS bounded substrate-engineering work (Phase 1 lands now; Phase 2+ are separately-authorizable yes-and items); operator authority preserved at every phase-boundary; agent-autonomous landing limited to Phase 1.