impl(sprint-11/wave-E): D-CSV-8 MUL i4 eval + D-CSV-9 8-ch↔SPO transcoder (parallel workers) by AdaWorldAPI · Pull Request #387 · AdaWorldAPI/lance-graph

AdaWorldAPI · 2026-05-16T03:31:28Z

Summary

Sprint-11 Wave E — Phase C substrate-locked primitives (the SIMD-vectorized hot path is sprint-12+; this PR ships the locked scalar/integer shape that the SIMD codegen targets). Two parallel Sonnet workers, both branched from main (deps D-CSV-2/3 already merged via #383/#384).

D-CSV-8 (W-E2) — MUL i4 evaluation in `lance-graph-contract::mul::i4_eval`

NEW pub mod i4_eval module in crates/lance-graph-contract/src/mul.rs (~330 LOC source + ~170 LOC tests). Five public functions consume &QualiaI4_16D (PR #384) + signed_mantissa: i8 (PR #383) and return existing MUL types:

dk_position_i4(qualia, mantissa) -> DkPosition — coherence + |mantissa| → Plateau / SlopeOfEnlightenment / ValleyOfDespair / MountStupid
trust_texture_i4(qualia) -> TrustTexture — coherence + valence + tension → Calibrated / Overconfident / Underconfident / Uncertain
flow_state_i4(qualia, mantissa) -> FlowState — warmth + groundedness − tension proxy + mantissa sign → Flow / Anxiety / Transition / Boredom
gate_decision_i4(qualia, mantissa) -> GateDecision — combines texture + flow → Flow / Hold{reason} / Block{reason} (canonical reasons)
mul_assess_i4(qualia, mantissa) -> MulAssessment — full pipeline producing all five MulAssessment fields

Naming corrections vs initial spec — surfaced by W-E2:

TrustTexture variants are Calibrated/Overconfident/Underconfident/Uncertain (NOT Crystalline/Solid/Fuzzy/Murky). The latter is causal_edge::layout::TrustTexture — a DIFFERENT enum used for the Pearl-3 lens. Two TrustTexture enums coexisting is technical debt worth filing in TYPE_DUPLICATION_MAP next loop.
MulAssessment.trust: TrustQualia and MulAssessment.homeostasis: Homeostasis (not texture / flow_state directly).

12 tests all pass: high-coherence expert, low-coherence beginner, neutral intermediate, all 4 trust textures, flow active/stuck-negative-mantissa, gate proceed/block, full pipeline, zero-input baseline.

D-CSV-9 (W-E1) — 8-channel ↔ SPO-palette transcoder in `thinking-engine::layered`

Per plan §11 + L-12 Option R-3: collapse the 8-channel thinking_engine::layered::CausalEdge64 into one SPO-palette causal_edge::CausalEdge64 at the L3 commit boundary, plus the inverse (lossy) projection.

NEW impl CausalEdge64 block in crates/thinking-engine/src/layered.rs (+258 LOC including tests). API: CHANNEL_NAMES, channel(idx), set_channel(idx, value), dominant_channel(), active_channel_count(), to_spo(s, p, o) -> SpoEdge, from_spo(spo) -> Self.

Channel → mantissa mapping table (per §11 D-CSV-9):

Channel	Mantissa magnitude	Causal mask	Sign convention
0 BECOMES	1 (Deduction)	SPO	+ forward
1 CAUSES	6 (Intervention/Counterfactual per L-9)	SPO	sign = direction
2 SUPPORTS	4 (Revision +)	PO	+ only
3 REFINES	5 (Synthesis)	PO	+ only
4 GROUNDS	1 (Deduction, S-grounded)	S	+ only
5 ABSTRACTS	2 (Induction)	P	+ only
6 RELATES	0 (Identity/neutral)	None	sign-aware
7 CONTRADICTS	1 (Abduction backward)	SPO	sign = polarity

8 transcoder tests all pass: channel get/set round-trip × 8 channels, dominant_channel zero-default, dominant_channel picks max |net|, to_spo BECOMES/CAUSES-negative/RELATES, 16-mapping round-trip equivalence-class check, set_channel out-of-range no-op.

Cargo.toml: added causal-edge = { path = "../causal-edge" } to thinking-engine deps. causal-edge is zero-dep itself, preserving the thinking-engine architecture stance.

E-META-7 dual-type debt status

The two CausalEdge64 types (SPO-palette in causal_edge::edge vs 8-channel cascade in thinking_engine::layered) still coexist, but Option R-3 is now WIRED — downstream consumers call .to_spo() / from_spo() at the L3 commit boundary to bridge without re-implementing the mapping. This is the recommended R-3 outcome per the synergies knowledge doc.

Out of scope (sprint-12+)

AVX-512 / NEON SIMD vectorization of the i4 MUL path (the i4 lane width allows 32-row sweeps per AVX-512 register; this PR ships the scalar shape as the SIMD codegen target)
Bench (criterion vs the f32 baseline — defer to env that can run benches reliably)
Single-source-of-truth resolution of the dual TrustTexture enums

Test plan

cargo test -p lance-graph-contract i4_eval — 12/12 pass
cargo test -p lance-graph-contract — 431/431 pass (no regressions)
cargo clippy --manifest-path crates/lance-graph-contract/Cargo.toml --lib --tests -- -D warnings — clean
CI: cargo test --manifest-path crates/thinking-engine/Cargo.toml transcoder_tests — 8 tests (W-E1 reported pass, env-validation matches)
CI: clippy on thinking-engine

https://claude.ai/code/session_01UwJuKqP828qyX1VkLgGJFS

Generated by Claude Code

…ath, sprint-12 SIMD vec deferred) Wave E half — D-CSV-8 scalar i4 evaluation path that eliminates the f32 dependency in MUL evaluation. The actual SIMD-vectorized hot path (AVX-512 / NEON intrinsics) is sprint-12+ once the i4 scalar shape is locked. This PR ships the locked shape. W-E2 (Sonnet) — `pub mod i4_eval` in `lance-graph-contract::mul` Five new public functions, all consume `&QualiaI4_16D` (PR #384) + `signed_mantissa: i8` (PR #383) and produce existing MUL types: - `dk_position_i4(qualia, signed_mantissa) -> DkPosition` — coherence (dim 9) + |mantissa| thresholds classify Plateau / SlopeOfEnlightenment / ValleyOfDespair / MountStupid - `trust_texture_i4(qualia) -> TrustTexture` — coherence + valence + tension drives Calibrated / Overconfident / Underconfident / Uncertain - `flow_state_i4(qualia, signed_mantissa) -> FlowState` — warmth + groundedness − tension proxy + mantissa sign → Flow / Anxiety / Transition / Boredom - `gate_decision_i4(qualia, signed_mantissa) -> GateDecision` — combines texture + flow into Flow / Hold{reason} / Block{reason} - `mul_assess_i4(qualia, signed_mantissa) -> MulAssessment` — full pipeline producing all five MulAssessment fields Naming corrections vs original spec — surfaced by W-E2: - `TrustTexture` variants are `Calibrated / Overconfident / Underconfident / Uncertain` (NOT `Crystalline / Solid / Fuzzy / Murky` — that's `causal_edge::layout::TrustTexture`, a DIFFERENT enum used for the Pearl-3 causal mask state). Two `TrustTexture` enums coexisting is technical debt worth filing in TYPE_DUPLICATION_MAP but out of scope here — the i4_eval module uses the contract enum's actual variant names. - `MulAssessment.trust` field is `TrustQualia`, not `TrustTexture` directly. `MulAssessment.homeostasis: Homeostasis`, not `flow_state: FlowState` directly. Pipeline writes through these intermediate types correctly. - `GateDecision::Hold(reason)` and `GateDecision::Block(reason)` carry `String` reasons (not unit variants). The i4 path produces short canonical reasons ("low_coherence", "high_tension", etc.) for downstream telemetry. 12 new tests (all pass): high-coherence expert, low-coherence beginner, neutral intermediate; texture variants Crystalline / Murky / Solid; flow Active / stuck-negative-mantissa; gate proceed / block; full MulAssessment pipeline coverage; zero-input neutral baseline. Test status: **12/12 i4_eval pass; 431/431 contract suite pass; clippy clean (-D warnings)**. Contract crate remains zero runtime dependencies. Out of scope (sprint-12 D-CSV-8 follow-up) - AVX-512 / NEON SIMD vectorization (the i4 lane width allows 32-row sweeps per AVX-512 register; this PR ships the scalar shape as the SIMD codegen target) - Bench (criterion vs the f32 baseline — defer to env that can run benches reliably) W-E1 (D-CSV-9 8-channel transcoder) is still in flight; will land in a follow-up commit on this same branch when the worker reports completion. https://claude.ai/code/session_01UwJuKqP828qyX1VkLgGJFS

…(Option R-3) at thinking-engine L3 commit boundary Wave E other half — D-CSV-9 per plan §11 + L-12 Option R-3: collapse the 8-channel `thinking_engine::layered::CausalEdge64` (BECOMES / CAUSES / SUPPORTS / REFINES / GROUNDS / ABSTRACTS / RELATES / CONTRADICTS) into one SPO-palette `causal_edge::CausalEdge64` at the L3 commit boundary, and the inverse (lossy) projection back. W-E1 (Sonnet) — `impl CausalEdge64` block in `crates/thinking-engine/src/layered.rs` (+258 LOC including tests). Public API - `pub const CHANNEL_NAMES: [&str; 8]` — diagnostic labels matching `causal-edge-64-thinking-engine-variant.md` - `pub fn channel(&self, idx) -> i8` / `pub fn set_channel(&mut self, idx, value)` — per-byte signed accessors (SIMD-friendly bitcast; out-of-range idx is a defensive no-op on set, debug_assert on get) - `pub fn dominant_channel(&self) -> usize` — max |net_strength|; ties break to lowest channel idx (stable per L-12) - `pub fn active_channel_count(&self) -> u8` — for confidence proxy - `pub fn to_spo(&self, s_idx, p_idx, o_idx) -> SpoEdge` — Option R-3 collapse: dominant channel → (mantissa magnitude, causal_mask) per the table in §11 D-CSV-9; mantissa sign = sign of dominant net; freq = |net|/32 quantized to u8; conf = active_channel_count/8 quantized; direction triad = 0b000 (sign carried by signed mantissa per L-9); W-slot = 0, truth = Crystalline, spare = 0 (caller stamps later via `with_routing` / `with_spare`) - `pub fn from_spo(spo: SpoEdge) -> Self` — inverse projection (lossy by design — 8 channels collapse to 1 in the forward direction; the inverse picks the dominant-equivalent channel and writes the mantissa magnitude × frequency back as the channel net_strength). Used for round-trip tests + debugging visualisations. Channel → mantissa mapping table (per §11 D-CSV-9): | Channel | Mantissa magnitude | Mask | Sign convention | |---|---|---|---| | 0 BECOMES | 1 (Deduction) | SPO | + forward | | 1 CAUSES | 6 (Intervention/Counterfactual per L-9) | SPO | sign = direction | | 2 SUPPORTS | 4 (Revision +) | PO | + only | | 3 REFINES | 5 (Synthesis) | PO | + only | | 4 GROUNDS | 1 (Deduction, S-grounded) | S | + only | | 5 ABSTRACTS | 2 (Induction) | P | + only | | 6 RELATES | 0 (Identity/neutral) | None | sign-aware | | 7 CONTRADICTS | 1 (Abduction backward) | SPO | sign = polarity | 8 transcoder tests added (per §11 D-CSV-9 "16-mapping round-trip test" target): - channel get/set round-trip across [-128, -1, 0, 1, 127] × all 8 channels - dominant_channel zero-default (idx 0 when all-zero) - dominant_channel picks max |net_strength| (with ties to lower idx) - to_spo BECOMES → mantissa +1 (Deduction) - to_spo CAUSES with negative net → mantissa -6 (Counterfactual per L-9) - to_spo RELATES → mantissa 0 (Identity/neutral) - 16-mapping round-trip: for each (dom, sign) pair the back-channel lands in the documented equivalence class - set_channel out-of-range no-op Cargo.toml: added `causal-edge = { path = "../causal-edge" }` to `crates/thinking-engine/Cargo.toml` deps. causal-edge is zero-dep itself, so this preserves the thinking-engine architecture stance. Combined with the prior commit (W-E2 D-CSV-8 MUL i4 SIMD evaluation in `lance-graph-contract::mul::i4_eval`), this PR completes Wave E's two parallel-worker scope: the contract-side i4 MUL evaluator + the thinking-engine-side dual-CausalEdge64 reunification transcoder. E-META-7 dual-type debt status (per `causal-edge-64-synergies-and- pr-trajectory.md`): unchanged in shape (the two `CausalEdge64` types still exist), but Option R-3 is now wired — downstream consumers can call `.to_spo()` at the L3 commit boundary to bridge without re-implementing the mapping each time. https://claude.ai/code/session_01UwJuKqP828qyX1VkLgGJFS

…_u8 + widen round-trip equivalence class W-E1 worker completion notification (after the prior commit window) landed two corrections: 1. **API name collision resolution.** The pre-existing `set_channel(u8, u8)` in `layered.rs` (an unsigned channel writer used by `domino.rs::classify_transition`) collides with the new transcoder API which needs `set_channel(usize, i8)` for signed per-byte access. Renamed the pre-existing method to `set_channel_u8(u8, u8)` and updated all 8 call sites in `domino.rs` (`SUPPORTS`/`CAUSES`/`GROUNDS`/`REFINES`/`ABSTRACTS`/ `RELATES`/`CONTRADICTS`/`BECOMES` flow). Pure mechanical rename; no behavior change. 2. **Round-trip test equivalence class widening.** For dominant channel `dom = 0 (BECOMES)` or `dom = 4 (GROUNDS)` with negative sign, the to_spo → from_spo round trip lands on CONTRADICTS (channel 7) rather than BECOMES (channel 0). Reason: both BECOMES and GROUNDS map to mantissa magnitude 1 (Deduction); a negative sign in `to_spo` produces mantissa = -1, which `from_spo` correctly maps to CONTRADICTS (the documented "same |mantissa|, different sign" lossy collapse per L-12 Option R-3). The test's `expected_class` for `dom=0|4` is widened from `&[0]` to `&[0, 7]` to accept both forward (positive sign round-trip) and backward (negative sign collapse) outcomes. Test status (per W-E1 final report): **8/8 transcoder tests pass; 337/337 lib tests total (329 existing + 8 new). No regressions.** Pre-existing example error (NOT touched by this PR): `crates/thinking-engine/examples/calibrate_roles.rs` has an unrelated array-size mismatch — a known pre-existing failure outside this PR's scope. https://claude.ai/code/session_01UwJuKqP828qyX1VkLgGJFS

…ate-decisions knowledge doc W-F10 — Sprint-11 meta-review (.claude/board/sprint-log-11/meta-review.md, 341 lines) - Executive summary + sprint grade - Per-PR grades for waves A-E (PRs #383..#387) - Cross-sprint inconsistencies (CSI-1..6): TrustTexture ×2, v1-API- under-v2 alias anti-pattern (E-META-10 candidate), subagent permission isolation, SplatField/QualiaI4 bit-compat mirrors, plan §7.2 CONJECTURE-vs-FINDING ratification (OQ-CSV-1), bg-worker file-collision pattern - Sprint-12 spawn decision YES with merge gates on Wave F + open PRs (#385 #386); recommended sprint-12 phase scope = SIMD vec for D-CSV-8 + ndarray streams productization + on-Think method migration for D-CSV-12 + Jirak-derived Σ10 threshold (TD-7) - Per-worker grade placeholders for Wave F (W-F1..W-F12) — to be filled by the Opus meta-reviewer (W-Meta-Opus) when the fleet fully completes W-F11 — i4-substrate-decisions knowledge doc (.claude/knowledge/, 200 lines) - Tier-1 knowledge doc with READ BY: header - i4 substrate doctrine: sign = direction, |magnitude| = NARS rule slot; i4 × i4 → i8 precision family - All 20 locked decisions L-1..L-20 with one-sentence summary + shipping PR + canonical code site + deviations from plan - Four-column SoA (EdgeColumn / QualiaColumn / MetaColumn / FingerprintColumns) — sprint-11 outcomes per column - All six OQ-CSV-* ratifications recorded with wave evidence - Codex P1 anti-pattern: 5 documented v1-API-under-v2-feature aliasing instances (temporal write, inference_type read, set_temporal no-op, pack raw discriminant, W3 spec temporal=1023) - 12-mapping transcoder table (8 channels × NARS slots × Pearl rungs) with lossy-collapse equivalence classes - Cross-refs: plan v1 + sprint-log-10/11 meta-reviews + STATUS_BOARD + PR_ARC + AGENT_LOG + TECH_DEBT + EPIPHANIES + TYPE_DUPLICATION_MAP + CLAUDE.md iron rules Fleet status: 11 of 12 Sonnet workers complete; W-F12 (plan v2 draft) still in flight. W-Meta-Opus (1 honest cross-cutting reviewer) dispatch deferred until W-F12 finalizes. https://claude.ai/code/session_01UwJuKqP828qyX1VkLgGJFS

… lines) W-F12 final Sonnet worker output. Plan v2 revision capturing sprint- 11 outcomes + sprint-12 forward plan. Mirrors v1 structure with §0..§18; every changed section marked [UPDATED 2026-05-16] / 🆕 vs v1 (21 annotations); unchanged sections labeled UNCHANGED from v1. Highlights: - §0 status delta — Phase A/B/C outcomes: D-CSV-1/3/4 (#383), D-CSV-2 (#384), D-CSV-5a/6a+7 (#385/#386), D-CSV-8+9 (#387), D-CSV-10 (W-F1) - §5 L-1..L-20 implementation-outcome annotations (PR #+commit+ accessor file per row) - §6/§8 UNCHANGED architecture, annotated with shipping commits + TD pointers for deviations (SmallVec) - §11 D-CSV-* table — D-CSV-1..12 status; D-CSV-13/14/15 NEW Phase E sprint-12 entries (SIMD vec, on-Think method migration, Jirak Σ10) - §12 OQ table — all 6 OQs annotated with ratification - §13 Risk — 10 risks (vs 6 in v1); §13.7–13.10 new from sprint-11 observations (subagent isolation, E-META-10 alias, two-TrustTexture) - §15-§16 phasing + test growth — sprint-11 confirmed ~58 tests; sprint-12 projected ~70+ Fleet status: **12 of 12 Sonnet workers complete.** Honest cross- cutting Opus meta-review (W-Meta-Opus) dispatched next. https://claude.ai/code/session_01UwJuKqP828qyX1VkLgGJFS

…ntries (#383..#390) PP-9 (Opus) prepended 8 PR entries to PR_ARC_INVENTORY.md (1537→1903 lines, +366 / +50,619 bytes). APPEND-ONLY rule respected — no prior entries touched. Top→bottom reverse chronological: #390 sprint-12/wave-G (In PR, `bad0875`) — grade A−. D-CSV-5b cutover + D-CSV-6b WitnessCorpus + D-CSV-13 batch + D-CSV-15 Jirak math. #389 sprint-12 wave-F codex P2 follow-up — AttentionMaskBackend impl + canonical MailboxId (CSI-10). #388 sprint-12 Wave F fleet (12 Sonnet + 1 Opus) — grade B. D-CSV- 10/11/12 scaffolds + AttentionMask + plan v2. #387 sprint-11 Wave E — grade A−. D-CSV-8 MUL i4 scalar + D-CSV-9 8-channel transcoder (Option R-3). #386 sprint-11 Wave D — grade B+. D-CSV-7 MailboxSoA + D-CSV-6a WitnessCorpus core. #385 sprint-11 Wave C — grade B+. D-CSV-5a sibling QualiaI4Column. #384 sprint-11 Wave B — grade A. D-CSV-2 QualiaI4_16D + OQ-CSV-1 Option α ratification. #383 sprint-11 Wave A — grade A−. D-CSV-1 v2 layout + D-CSV-3 signed mantissa + D-CSV-4 CollapseGateEmission. Every entry mirrors the #381/#379 template — Header / Confidence / Added / Locked / Deferred / Docs / Cross-refs — with forward+backward linkages (#383→#385→#390 D-CSV-5 chain; E-META-10 catch in #383 → iron-rule promotion in #390 W-G5; etc.). Total PR headers in file: 28 (was 20). 13 of 14 planners done. Only PP-13 (brutally-honest-tester agent) still in flight. https://claude.ai/code/session_01UwJuKqP828qyX1VkLgGJFS

claude added 3 commits May 16, 2026 03:29

AdaWorldAPI merged commit e042c70 into main May 16, 2026
5 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

impl(sprint-11/wave-E): D-CSV-8 MUL i4 eval + D-CSV-9 8-ch↔SPO transcoder (parallel workers)#387

impl(sprint-11/wave-E): D-CSV-8 MUL i4 eval + D-CSV-9 8-ch↔SPO transcoder (parallel workers)#387
AdaWorldAPI merged 3 commits into
mainfrom
claude/sprint-11-wave-e-mul-transcoder

AdaWorldAPI commented May 16, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

AdaWorldAPI commented May 16, 2026

Summary

D-CSV-8 (W-E2) — MUL i4 evaluation in lance-graph-contract::mul::i4_eval

D-CSV-9 (W-E1) — 8-channel ↔ SPO-palette transcoder in thinking-engine::layered

E-META-7 dual-type debt status

Out of scope (sprint-12+)

Test plan

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

D-CSV-8 (W-E2) — MUL i4 evaluation in `lance-graph-contract::mul::i4_eval`

D-CSV-9 (W-E1) — 8-channel ↔ SPO-palette transcoder in `thinking-engine::layered`