V3 substrate convergence: probes P0–P5 prove one palette across distance/edge/causality/discovery/reasoning#624
Merged
Conversation
The local 8-channel Bach-counterpoint cascade accumulator in layered.rs was named CausalEdge64, colliding with the canonical causal_edge::CausalEdge64 that the same file imports (aliased SpoEdge). domino.rs re-exported the local shadow. Rename the local type to CascadeChannels8 (8 packed u8 channels); the to_spo / from_spo collapse still targets the canonical edge unchanged. After this, the identifier CausalEdge64 in thinking-engine refers ONLY to the canonical type. Pure rename, zero behavior change. cargo check -p thinking-engine green. First code delta of the OSINT-substrate convergence roadmap (OGAR docs/OSINT-SUBSTRATE-REUSE-MAP.md P0); precondition for the P1 distance-identity probe. Board: EPIPHANIES E-CE64-NAME-COLLISION-DEDUP (same commit, hygiene rule). Co-Authored-By: Claude <noreply@anthropic.com>
…scovery The V3 'one causal-distance format' claim, pinned to shipped code. One f32 source (deepnsm subspace_distance_table) → a fixed quantizer → the u16 palette, read byte-identically by both integer consumers: SpoDistances::s_dist (planner nars_engine) and MatrixDistance::distance (arm-discovery oracle). For 4096 deterministic index pairs: nars u16 == arm u32 == quantized deepnsm entry, plus symmetry, plus causal_distance(mask=0b111) == plane sum. Integer-exact; the only float is the deepnsm source, collapsed before either consumer sees it. deepnsm already ships the 6×256 CAM-PQ shape, so no re-bake was needed — the roadmap's 'fix deepnsm→6×(8:8)' hedge is already satisfied upstream. Adds lance-graph-arm-discovery as an osint dev-dep (path, the third source). Test: crates/lance-graph-osint/tests/p1_distance_identity.rs (2 passed). Board: EPIPHANIES E-P1-DISTANCE-IDENTITY-GREEN (same commit, hygiene rule). Co-Authored-By: Claude <noreply@anthropic.com>
P2: OSINT palette indices → CausalEdge64::pack_v2 → s_idx/p_idx/o_idx/causal_mask/ frequency/confidence all round-trip; causal_distance of two edges' heads == the per-plane palette sum (edge and distance engine index the same palette). P3: causal_edge::CausalMask and SpoDistances::causal_distance's mask byte are the SAME 3-bit convention (S=0b100,P=0b010,O=0b001) — each mask keeps exactly its planes (S→s, PO→p+o, SPO→s+p+o, None→0); Level-2 Intervention (PO) is strictly less than Level-3 Counterfactual (SPO) when the Subject term is non-zero, i.e. Pearl's do-calculus ladder is a mask over palette planes. Integer-exact, deterministic, no new deps. Test: p2_p3_edge_pearl.rs (3 passed). Board: EPIPHANIES E-P2-P3-EDGE-PEARL-GREEN (same commit, hygiene rule). Co-Authored-By: Claude <noreply@anthropic.com>
… wire
OSINT dual-use fixture (militaryUse ⟹ impact, perfect correlation over 1000
rows) mined by arm-discovery's AerialProposer over a MatrixDistance oracle:
(1) exact integer evidence — support_ppm=500000, confidence_ppm=1000000;
(2) determinism — two mines byte-identical, no seed;
(3) edge join — arm_to_truth_u8 → TruthU8{255,254} → CausalEdge64::pack_v2
round-trips freq/conf. Discovery output IS the edge wire.
Live-SpoStore promotion stays gated on D-ARM-7 (Jirak floor). In-memory
mine→truth→edge path only. Test: p4_discovery_edge.rs (3 passed).
Board: EPIPHANIES E-P4-DISCOVERY-EDGE-GREEN (same commit).
Co-Authored-By: Claude <noreply@anthropic.com>
is_a(A,B) ∧ is_a(B,C) ⊢ is_a(A,C) via CausalEdge64::syllogize: Figure::Chain, middle term consumed, deduction truth composes exact (f=255, c=163), mantissa +1, Pearl mask SPO&SPO=SPO. Negative cases: no shared term ⇒ None; identical (S,O) ⇒ None (revision, not syllogism); syllogize is pure/deterministic. The part_of/is_a ontology chaining OSINT needs, entirely on the palette. Test: p5_syllogize.rs (4 passed). Board: EPIPHANIES E-P5-SYLLOGIZE-GREEN. Co-Authored-By: Claude <noreply@anthropic.com>
Records the operator's 1-9 RungLevel reminder as an invariant: the contract's RungLevel (0 Surface=observation .. 6 Counterfactual .. 9 Transcendent) is a dependency stack, not a menu — ShaderDispatch.rung 'elevates on sustained BLOCK' bottom-up, so counterfactual rests on grounded observation+hypothesis rungs. Maps the P1-P5 convergence probes onto the rungs they ground (P1=observation, P3/P4=hypothesis, P3-SPO/P5=counterfactual), with the D-ARM-7 Jirak floor as the stack guard. Reframes P6 (awareness rollover) as THE rung-elevation mechanism. Flags RungLevel duplication (thinking-engine vs contract) as TD-RUNGLEVEL-DUP. EPIPHANIES E-RUNG-LADDER-IS-A-DEPENDENCY-STACK; TECH_DEBT TD-RUNGLEVEL-DUP. Co-Authored-By: Claude <noreply@anthropic.com>
…erently all_projections reads 3 per-plane palette cells for one (a,b) pair and produces all 8 Pearl projections by masked summation — 3 cache reads -> 8 causal questions (the Morton-tile 2^3 all-at-once amortization). Proven un-redundant: (1) the 8-vector is a pure function of the 3 scalar plane distances; (2) monotone ladder, SPO upper-bounds every sub-question; (3) the SAME two candidates rank OPPOSITELY under Association (SO) vs Intervention (PO) — 8 real retrievals, not one distance in 8 hats. Test: p3b_spo_questions_amortized.rs (3 passed). Capstone: records the assembled V3 stack (AriGraph/CausalEdge/thinking-style/rung/ OGAR-AST-transpile ClassView+bitmask-as-focus+askama-ERB) and the two closed rabbit holes (SoA-first migration; rs-graph-llm replayable orchestration on OGAR). Board: EPIPHANIES E-SPO-2CUBE-GIVES-QUESTIONS-AND-CANDIDATES. Co-Authored-By: Claude <noreply@anthropic.com>
…denda A (correction, [G]): rig = the RAG + LLM-API template-discovery membrane on the cold/external side (with spider + rs-graph-llm); templates flow cold->hot as compiled deterministic reflexes, LLM never on the hot path. Completes the V3-stack capstone which had omitted rig. B (conjecture, [S]): 6x2x8bit 'exact' centroid = 6x palette256^2 (12B, 65536 positions/subspace) vs CAM-PQ 6x256 (6B, quantized). The OSINT 6x(8:8) identity tenant ALREADY has this shape, so the semantic axes double as the exact-centroid distance code (no separate CAM-PQ code); preserves the one-distance-format (Pearl-maskable palette sums). Marked CONJECTURE, needs PROBE-6x2x8-vs-CAMPQ (fidelity + storage/cache) before any adoption over CAM-PQ. Board: EPIPHANIES E-RIG-DISCOVERY-MEMBRANE + E-6x2x8-EXACT-CENTROID. Co-Authored-By: Claude <noreply@anthropic.com>
…ested feedback loop Grounded in on-disk maturity (not doctrine alone): graph-flow-action-ogar is shipped (rs-graph-llm <-> OGAR Do arm, contract-only); AriGraph V3 is a commit contract with an untested round-trip; rig is unwired; surrealdb kv-lance is a scaffold (8 TODOs). Decision: it's a layering not an OR — hot reasoning path = rs-graph-llm + rig-cold-membrane + AriGraph-V3-tissue (Option 2, doctrine-aligned); cold projection = AriGraph V3 -> Lance SPO-G -> lancedb now / surrealdb-kv-lance later (Option 1, beside not between). Untested = the feedback round-trip (PROBE-RIG-ARIGRAPH-FEEDBACK, stubbed LLM, no network), blocked P9-class on AriGraph write-back needing lance-graph core/protoc. Testing only the decision half would be a manufactured green. Board: EPIPHANIES E-RIG-ARIGRAPH-FEEDBACK-LAYERING. Co-Authored-By: Claude <noreply@anthropic.com>
…; probes = de-blackboxing Records the operator's strategic frame: the substrate accumulated as a black-box wishlist that buried the two real layers (deepnsm = Semantic Transformer, OGAR = transpiler). The P0-P5 convergence probes ARE the cure — each converts a wishlist claim into an integer-exact green assert (P1 de-blackboxed the Semantic Transformer itself; P2-P5 proved everything rides the palette). Names what's still black box (the untested rig->AriGraph runtime loop) and the standing rule: no new substrate claim enters canon without a green probe, else entropy re-accumulates. Also documents what the LangGraph work gave rs-graph-llm: graph-flow (stateful execution + interrupt/resume) + graph-flow-kanban (Rubicon envelope carrying thinking-style + Rung depth, advancing on Flow/Block/Hold) + graph-flow-action(-ogar). Board: EPIPHANIES E-SUBSTRATE-IS-DEEPNSM-PLUS-OGAR. Co-Authored-By: Claude <noreply@anthropic.com>
|
Warning Review limit reached
Next review available in: 1 minute Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (5)
📝 WalkthroughWalkthroughDocumentation entries were added/updated in EPIPHANIES.md and TECH_DEBT.md describing probe (P0-P5) validation results and a duplicated-enum debt ticket. Five new integration test files were added to lance-graph-osint validating distance, edge, projection, discovery, and syllogism invariants. In thinking-engine, the local ChangesBoard documentation updates
OSINT probe test suite
CausalEdge64 to CascadeChannels8 rename
Estimated code review effort: 2 (Simple) | ~15 minutes Possibly related issues
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 20b5d0d20f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Comment on lines
+55
to
+59
| let path = std::env::temp_dir().join(format!( | ||
| "p1_osint_codebook_{}_{}.bin", | ||
| std::process::id(), | ||
| bytes.len() | ||
| )); |
There was a problem hiding this comment.
When p1_distance_identity runs under Rust's default parallel test harness, both tests call synth_codebook() in the same process and derive the same temp filename from only process::id() and bytes.len(). One test can remove or truncate this file while the other is still writing/loading it, making Codebook::load_binary intermittently fail or read a partial file; include a per-call unique component or use a tempfile to isolate concurrent tests.
Useful? React with 👍 / 👎.
Both tests in p1_distance_identity run concurrently under the parallel harness and each removes its codebook file after loading; the pid+len-only temp path collided, so one test could truncate the other's load mid-read (intermittent load_binary failure). Add a per-call AtomicU64 sequence to the filename so each call gets a unique path. Deterministic content unchanged; 2 tests still green. Addresses codex review comment on PR #624. Co-Authored-By: Claude <noreply@anthropic.com>
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
crates/lance-graph-osint/tests/p2_p3_edge_pearl.rs (1)
26-32: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winDuplicate
splitmix64helper.Identical to the one in
p1_distance_identity.rs. As more P-series probes are added (P4/P5 already exist), consider hoisting shared test PRNG/helpers into atests/common/mod.rsto avoid this spreading further.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@crates/lance-graph-osint/tests/p2_p3_edge_pearl.rs` around lines 26 - 32, The splitmix64 helper is duplicated here and matches the implementation in p1_distance_identity.rs, so shared test PRNG code should be centralized. Move splitmix64 into a common test utility such as tests/common/mod.rs and update p2_p3_edge_pearl.rs to use that shared helper instead of keeping a local copy, so future P-series tests can reuse the same function.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@crates/lance-graph-osint/tests/p1_distance_identity.rs`:
- Around line 43-64: The synthetic temp file path in synth_codebook is not
unique across parallel test runs, causing write/load/remove races between the
tests that call it. Update synth_codebook to generate a per-call unique path
instead of using only process::id and byte length, while keeping the
deterministic bytes unchanged. Use the existing synth_codebook and
Codebook::load_binary flow, and ensure the temporary file name cannot collide
when tests execute concurrently.
---
Nitpick comments:
In `@crates/lance-graph-osint/tests/p2_p3_edge_pearl.rs`:
- Around line 26-32: The splitmix64 helper is duplicated here and matches the
implementation in p1_distance_identity.rs, so shared test PRNG code should be
centralized. Move splitmix64 into a common test utility such as
tests/common/mod.rs and update p2_p3_edge_pearl.rs to use that shared helper
instead of keeping a local copy, so future P-series tests can reuse the same
function.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 1b37e27a-0867-425b-8312-32bfc588807e
⛔ Files ignored due to path filters (1)
crates/lance-graph-osint/Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (12)
.claude/board/EPIPHANIES.md.claude/board/TECH_DEBT.mdcrates/lance-graph-osint/Cargo.tomlcrates/lance-graph-osint/tests/p1_distance_identity.rscrates/lance-graph-osint/tests/p2_p3_edge_pearl.rscrates/lance-graph-osint/tests/p3b_spo_questions_amortized.rscrates/lance-graph-osint/tests/p4_discovery_edge.rscrates/lance-graph-osint/tests/p5_syllogize.rscrates/thinking-engine/src/domino.rscrates/thinking-engine/src/engine.rscrates/thinking-engine/src/layered.rscrates/thinking-engine/src/qualia.rs
…oss-repo arc lance-graph-ogar tracks OGAR main by branch and carries a compile-time COUNT_FUSE (mirror::CODEBOOK.len() == ogar_vocab::class_ids::ALL.len()). OGAR #145's osint mint (+2 to ALL) will break the fuse in every consumer the instant it merges, because the contract mirror (65 rows) has no osint concept rows yet. Can't pre-add them to #624 (would break its own CI vs unmerged OGAR main). Tracked with the coordinated-arc resolution + a merge-ordering decision flagged to the operator. Co-Authored-By: Claude <noreply@anthropic.com>
…ds off) Operator correction: classid = (V3-substrate 0x1000)(domain 0x07)(app_prefix 01); 0x1000 is the V3 marker (not q2's app prefix), and 0x0701 is domain:app (not a flat concept) — the 64k dynamic-class space (256 domains x 256 app-classes -> ClassView x bitmask). The human-readable form 0x07:01::1000 is a DELIBERATE post-V3 reorder, postponed to protect the little-endian byte contracts P1-P5 are built on. Hands off: do not reorder classid bytes or rewrite codebook rows toward the human-readable form; that's a later, deliberately-triggered arc with its own LE round-trip probe. Orthogonal to ISS-OGAR-OSINT-MIRROR-PENDING (row count). Board: EPIPHANIES E-CLASSID-HUMANREADABLE-REORDER-DEFERRED. Co-Authored-By: Claude <noreply@anthropic.com>
… directive) When the substrate impl starts, route all classid field extraction through ONE split-order abstraction ((before)::Domain:appid::(after)) so the deferred human-readable reorder (0x1000_0701 <-> 0x07:01::1000) is a one-place flag flip with a single round-trip probe, not an N-site migration that can desync (I-LEGACY-API-FEATURE-GATED trap). Flag defaults to the current stored order. Board: EPIPHANIES E-CLASSID-SPLIT-ORDER-IS-A-FLIP. Co-Authored-By: Claude <noreply@anthropic.com>
…y contract (keep it) Operator ruling: the two repos are one coupled pair moved together every session; the COUNT_FUSE is the INTENTIONAL dependency contract enforcing that parallel movement, not a break to engineer away. Corrects my earlier rev-pin recommendation (option 1 REJECTED). OSINT mirror resolution: keep the fuse, land the 2 mirror rows + domains_agree arm in parallel with OGAR #145 (option 2, brief transient red is fine). Target design: same-binary compile -> lance-graph plug-and-play; when the contract clicks, OGAR emits the vocab (build-time codegen) -> hand-mirror + fuse retire. Ready patch documented in ISS-OGAR-OSINT-MIRROR-PENDING. Board: EPIPHANIES E-OGAR-LANCEGRAPH-MOVE-IN-PARALLEL; ISSUES status updated. Co-Authored-By: Claude <noreply@anthropic.com>
AdaWorldAPI
pushed a commit
that referenced
this pull request
Jul 1, 2026
…s the reserved root slot Surfaced when merged OGAR #145 + lance-graph #624 met: OGAR shipped osint_system at 0x0700 (CC==0x00, the reserved domain-root slot per canon), which the mirror's in-tree zero-slot test forbids — so the parallel-mirror is blocked on a remap decision (Option A: move to 0x0702 + q2 update; Option B recommended: 0x0700 is the OSINT domain-root/default class, drop from the concept CODEBOOK, mirror only osint_person 0x0701). lance-graph main CI is green (fuse is in the excluded lance-graph-ogar); consumers vendoring lance-graph-ogar are the exposure. Board: ISSUES ISS-OSINT-SYSTEM-ROOT-SLOT-VIOLATION (blocks ISS-OGAR-OSINT-MIRROR-PENDING). Co-Authored-By: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this is
The static half of the OSINT-substrate convergence roadmap, proven against shipped code as a ladder of small, integer-exact
#[test]probes. Each probe converts a previously-black-box "we think this converges" claim into a green assert — de-blackboxing the accumulated-wishlist substrate back into its two real layers: deepnsm (the Semantic Transformer) + OGAR (the transpiler).Zero production-dep changes. One
arm-discoverydev-dep onlance-graph-osint, one local type rename, five test files, board entries. Nothing on any hot path moved.The probe ladder (all green)
CausalEdge64shadowed the canonicalcausal_edge::CausalEdge64→ renamedCascadeChannels8; canonical is now the soleCausalEdge64cargo check -p thinking-enginesubspace_distance_table→ quantize → one palette →SpoDistances::s_dist(planner) ≡MatrixDistance::distance(arm-discovery), byte-exact / 4096 pairsCausalEdge64::pack_v2round-trips s/p/o + mask + freq/conf;causal_distanceof two edges = plane sumcausal_edge::CausalMask≡causal_distancemask byte (S=100,P=010,O=001);PO < SPO= do-calculus confounder projectionall_projections= 8 Pearl questions from 3 cached reads (Morton-tile 2³ amortization); Association vs Intervention rank the same candidates oppositely — questions and candidates inherentlymilitaryUse⟹impactwith exact ppm (500k/1M), deterministic no-seed,arm_to_truth_u8 → CausalEdge64is_a(A,B)∧is_a(B,C) ⊢ is_a(A,C)viasyllogize— deduction truth exact (f=255,c=163), mantissa +1Proven claim: the V3 256×256 palette is one integer-exact metric across five vertices — distance sources, edge carrier, causal-mask semantics, ARM discovery, and multi-hop NAL reasoning. The only float is the deepnsm source, collapsed before any consumer sees it (the "NARS edge").
Architecture recordings (board, append-only)
E-RUNG-LADDER-IS-A-DEPENDENCY-STACK— the 0–9RungLevelladder gates higher reasoning on grounded lower rungs (rungelevates on sustained BLOCK, bottom-up); P1–P5 are the per-rung capabilities; D-ARM-7 Jirak floor is the stack guard.E-SPO-2CUBE-GIVES-QUESTIONS-AND-CANDIDATES— V3-stack capstone; the 2³ projection is the atom every layer shares.E-RIG-DISCOVERY-MEMBRANE+E-6x2x8-EXACT-CENTROID— rig as the cold RAG/LLM template-discovery membrane; the 6×(8:8) tenant can double as an exact-centroid distance code ([S]conjecture, probe named).E-RIG-ARIGRAPH-FEEDBACK-LAYERING— the rig/rs-graph-llm/surrealdb/AriGraph-V3 stack is a hot-loop + cold-projection split; the feedback round-trip is UNTESTED and honestly blocked (P9-class, AriGraph write-back needs protoc).E-SUBSTRATE-IS-DEEPNSM-PLUS-OGAR— the strategic frame: probes = de-blackboxing; standing rule "no new substrate claim without a green probe".TD-RUNGLEVEL-DUP—RungLevelduplicated in thinking-engine vs the canonical contract (deferred dedup).What's NOT in scope (honest edge)
The runtime loop (rig → CausalEdge → AriGraph V3 → next cycle) is asserted, not proven — gated on wiring AriGraph write-back through the protoc boundary (probes P6–P9). This PR is the static substrate only.
Verification
Each probe runs via
cargo test --manifest-path crates/lance-graph-osint/Cargo.toml --test <name>(osint is a standalone/excluded crate). P0 viacargo check -p thinking-engine. All green on branch HEAD.🤖 Generated with Claude Code
Generated by Claude Code
Summary by CodeRabbit
New Features
Bug Fixes
Refactor